Browse the corpus

Introduction Infertility is conceptualized as a major crisis in life. A crisis evokes emotional reactions that are classified into four main phases: the initial phase (shock, surprise, denial); the reactive phase (frustration, anger, anxiety, guilt, grief, depression, isolation); the adaptive phase (acceptance) and a resolution phase (planning for future solutions) (1,2). Reactions during a crisis are determined by factors such as: influences of the event itself, pre-existing personality, cultural factors and social support from close family and friends (3). The conceptual framework of crisis theory (2,4) is modified for the purpose of the infertility crisis (1). Although the crisis theory may initially be used, this theory is considered less useful for the adaptive phase as infertility may consist of recurring stressful events (5). Women who have undergone unsuccessful tubal surgery for female infertility factor mostly remain in the reactive phase for 2 years after surgery (6). Life crises are considered time-limited with duration of 6 weeks or less for the reactive phase (1). The infertility crisis may be ongoing or chronic after recurrent treatment failure.

e undergone unsuccessful tubal surgery for female infertility factor mostly remain in the reactive phase for 2 years after surgery (6). Life crises are considered time-limited with duration of 6 weeks or less for the reactive phase (1). The infertility crisis may be ongoing or chronic after recurrent treatment failure. Infertility represents many losses, such as, the loss of fertility and reproductive ability and the loss of a child and biological offspring (1). Thus, grief is a normal reaction to a distressing situation, such as a loss (7). However, when the loss is of a potential, not an actual loss, the couple may not realize they are allowed to grieve (1). Grief reactions are often self-limited and with preserved self-esteem (8). The duration of normal grief reactions depends on the grieving process, which is considered important for successful adjustment (1,7). However, the grieving process may be hampered or prolonged, thus causing complicated grief. Complicated grief occurs when grief reactions persist more than 2 months after a loss and is consistent with the definition of major depression (9). Therefore, distinguishing between grief reactions, complicated grief and depression is important but can be difficult (8,10). Symptoms that are not characteristic for normal grief reactions are excessive guilt, suicidal ideation and feelings of worthlessness (9).

consistent with the definition of major depression (9). Therefore, distinguishing between grief reactions, complicated grief and depression is important but can be difficult (8,10). Symptoms that are not characteristic for normal grief reactions are excessive guilt, suicidal ideation and feelings of worthlessness (9). Approximately 30–40% of couples undergoing IVF treatment will remain childless after treatment (11,12). Grief is one of the main experiences of being childless in women 2 years after ending unsuccessful IVF (13). Previous qualitative studies exploring the experience of childlessness after undergoing unsuccessful IVF have focused on women (13–15) or interviews have been conducted with the couple together, not individually (16,17). Hence, there is a need to obtain a more in-depth understanding about both men and women's experiences of remaining childless after undergoing unsuccessful IVF treatment. The objective of the current qualitative-approach study was to explore the experience of undergoing unsuccessful IVF treatment and of remaining childless 3 years after IVF in a sample of women and men in Sweden.

Approximately 30–40% of couples undergoing IVF treatment will remain childless after treatment (11,12). Grief is one of the main experiences of being childless in women 2 years after ending unsuccessful IVF (13). Previous qualitative studies exploring the experience of childlessness after undergoing unsuccessful IVF have focused on women (13–15) or interviews have been conducted with the couple together, not individually (16,17). Hence, there is a need to obtain a more in-depth understanding about both men and women's experiences of remaining childless after undergoing unsuccessful IVF treatment. The objective of the current qualitative-approach study was to explore the experience of undergoing unsuccessful IVF treatment and of remaining childless 3 years after IVF in a sample of women and men in Sweden. Material and methods Men and women, who had undergone IVF or ICSI (intra-cytoplasmic sperm injection) treatment at the Centre of Reproduction, University Hospital, Uppsala, Sweden, participated. In Sweden, assisted reproductive technology (ART) is offered by both public and private clinics. The Centre of Reproduction is a public clinic and infertile couples are offered three subsidized IVF treatments with a waiting list of 3 months at the time of the study. Subsidized counseling was offered to all couples at their first visit to the public clinic.

ology (ART) is offered by both public and private clinics. The Centre of Reproduction is a public clinic and infertile couples are offered three subsidized IVF treatments with a waiting list of 3 months at the time of the study. Subsidized counseling was offered to all couples at their first visit to the public clinic. The methodological approach was qualitative with semi-structured individual interviews (18) and qualitative content analysis (19). Participants were identified from the clinic's database and recruitment was in two steps. First, an invitation letter including a written consent to participate was sent to women (n = 49) who had undergone IVF and had no further IVF treatments at the public clinic after a minimum of 3 years. Secondly, men and women were contacted individually by phone, by the first author (HV) about a week after the letter was received; this provided an opportunity for the potential participants to ask questions, to check for exclusion criteria unavailable in the database and to book a place and time for an interview. Reasons for declining participation in the study were not assessed. Exclusion criteria were: Swedish language difficulties, having a biological child from previous relationship and having already adopted a child.

to check for exclusion criteria unavailable in the database and to book a place and time for an interview. Reasons for declining participation in the study were not assessed. Exclusion criteria were: Swedish language difficulties, having a biological child from previous relationship and having already adopted a child. All interviews were conducted in a non-clinical room at the hospital in 2006 by the first author. Written informed consent was obtained and socio-demographic and fertility data were collected via a questionnaire. Fertility history, such as, on a previous pregnancy loss, was obtained from the medical records. The interviews started after informing the participant that she or he could withdraw participation at any time. A pre-tested and revised interview-guide covering the following topics was used: life situation as involuntary childless; mental health during and 3 years after IVF; partner relationship and social network and support; and decision-making when to end IVF treatment. All interviews were tape-recorded, lasted about 20–60 minutes (mean 40 minutes) and were then transcribed verbatim by the first author. Field notes on the interaction during the interview were made and taken into consideration during the analysis. The interviewer was known to some participants due to working as a midwife at the Centre of Reproduction.

, lasted about 20–60 minutes (mean 40 minutes) and were then transcribed verbatim by the first author. Field notes on the interaction during the interview were made and taken into consideration during the analysis. The interviewer was known to some participants due to working as a midwife at the Centre of Reproduction. Data analyses All interviews were analyzed using qualitative content analysis (19) by the first author and checked against the co-authors. After careful reading of the transcribed interview, the text was divided into meaning units. A meaning unit is a piece of text with a specific content that relates to the aim of the study. All meaning units were then condensed; a process of shortening the text while still preserving the core content, into condensed meaning units. The condensed meaning units were further shortened into codes; a labeling that allows the data to be understood in relation to the context. Codes were then grouped into categories depending on similarities and differences in content (19). The study was approved by the Ethics Committee, Uppsala University, Uppsala, Sweden.

Data analyses All interviews were analyzed using qualitative content analysis (19) by the first author and checked against the co-authors. After careful reading of the transcribed interview, the text was divided into meaning units. A meaning unit is a piece of text with a specific content that relates to the aim of the study. All meaning units were then condensed; a process of shortening the text while still preserving the core content, into condensed meaning units. The condensed meaning units were further shortened into codes; a labeling that allows the data to be understood in relation to the context. Codes were then grouped into categories depending on similarities and differences in content (19). The study was approved by the Ethics Committee, Uppsala University, Uppsala, Sweden. Results The study group consisted of 19 participants: 10 women and 9 men. Seven were couples and two male and three female participants did not bring their partners to the study. Of the latter participants, two of the men and two of the women had a biological child with partners from a previous relationship and these partners were excluded. One female participant was divorced. Three of the male and two of the female participants were planning for adoption at the time of the interview. Demographic and fertility information are presented in Table 1. The participants had their last IVF treatment at the public clinic at a mean of 38 months prior to the data collection. The analyses resulted in two main categories and seven sub-categories describing the experiences of the participants (see Table 2). Quotations from the interviews are presented below and labeled by M (male) or F (female), code number and age.

ent at the public clinic at a mean of 38 months prior to the data collection. The analyses resulted in two main categories and seven sub-categories describing the experiences of the participants (see Table 2). Quotations from the interviews are presented below and labeled by M (male) or F (female), code number and age. Table 1 Demographic and fertility data given by women and men. Women (n = 10) Men (n = 9) Age (years) 39.6 (35–43) 38.8 (36–46) University/college (n) 3 4 High school 7 5 Employee (n) 9 7 Unemployed 1 0 Student 0 2 Duration of infertility (years)a 7.0 (2–10) 4.7 (1–7) Infertility factorsb (n) Female 2 4 Male 4 1 Unexplained/not known 4 4 Previous pregnancy loss (n, %) No 4 (40%) Yes 6 (60%) Numbers of IVF/ICSIb (n) Public 2.3 (1–3) 2.1 (1–3) Numbers of IVF/ICSIb (n) 3 2 Private 3.0 (2–4) 2.5 (2–3) Counseling (n) During IVF 5 1 After IVF 0 0 a Time before last IVF. b Data from data base. Table 2 Men's and women's experiences after undergoing unsuccessful IVF treatment

Women (n = 10) Men (n = 9) Age (years) 39.6 (35–43) 38.8 (36–46) University/college (n) 3 4 High school 7 5 Employee (n) 9 7 Unemployed 1 0 Student 0 2 Duration of infertility (years)a 7.0 (2–10) 4.7 (1–7) Infertility factorsb (n) Female 2 4 Male 4 1 Unexplained/not known 4 4 Previous pregnancy loss (n, %) No 4 (40%) Yes 6 (60%) Numbers of IVF/ICSIb (n) Public 2.3 (1–3) 2.1 (1–3) Numbers of IVF/ICSIb (n) 3 2 Private 3.0 (2–4) 2.5 (2–3) Counseling (n) During IVF 5 1 After IVF 0 0 a Time before last IVF. b Data from data base. Table 2 Men's and women's experiences after undergoing unsuccessful IVF treatment Experiences in relation to unsuccessful IVF treatment Putting up a shield Late realization of the need of professional support Affected partner relationship Frustrated at the ending of IVF Experiences of remaining childless after IVF failure Unanswered questions after ending IVF Feeling excluded and lacking understanding Loss of future life goals Putting up a shield Unsuccessful IVF affected mental health negatively. One male reflection was that not everyone undergoing IVF was mentally strong enough to deal with failure after failure. Women explained how they had been putting up a shield of optimism prior to treatment and not showing any reactions after failure. Undergoing unsuccessful IVF treatment was described by women in terms of grief. The experience after treatment failure was described as losing someone close, and suicidal thoughts were revealed. Men had no knowledge on possible emotional reactions after unsuccessful treatment, and the reactions of their wives were unexpected. Not allowing oneself to grieve, to ignore grief and just continue with the next treatment were described by both men and women.

losing someone close, and suicidal thoughts were revealed. Men had no knowledge on possible emotional reactions after unsuccessful treatment, and the reactions of their wives were unexpected. Not allowing oneself to grieve, to ignore grief and just continue with the next treatment were described by both men and women. during these five years, after each try it was just… new attempt … we never allowed ourselves to grieve (…) we ignored it and went on to the next attempt … and that was also the same with those around us … (F0537). Late realization of the need of professional support A lack of professional support for handling the grief reactions after IVF failure was described by both men and women. They did not request support during treatment and not until IVF treatment ended did they realize that professional support would have been beneficial. Men felt it was expected of them to take the supportive role toward their partner and show no sadness or grief when treatment failed. Handling their own and their partners’ grief reactions after treatment failure were considered difficult. In retrospect, a lack of knowledge about grief might have hindered their partners from proper grieving, according to some men.

upportive role toward their partner and show no sadness or grief when treatment failed. Handling their own and their partners’ grief reactions after treatment failure were considered difficult. In retrospect, a lack of knowledge about grief might have hindered their partners from proper grieving, according to some men. it was more a support role I took on myself as I thought this was required (…) but to allow grief… allowed us both to be sad … instead of … yes, now she is sad so I have to comfort her and keep myself together a bit longer … then I can't be sad at the same time … so you kept yourself together and swept it under the mat (…) at the same time I hinder her from grieving properly (M0436). Individual support was suggested by men as this would enable them to better understand their spouses' unexpected grief reactions. Mandatory counseling after IVF failure, to able to process grief in between the treatments, was suggested by both men and women. Men expressed that their partner would have needed professional support both during and after treatment. Affected partner relationship Partner relations were affected both positively and negatively by unsuccessful IVF. Some men and women described how the relationship with their partner had been strengthened in a critical situation. Other men and women experienced strain in the relationship and temporary separations after ending IVF. Difficulty in communicating was one reason mentioned by women who experienced strain in the relationship. Another reason was a lack of self-esteem, expressed as worthlessness, upon not being a proper woman.

situation. Other men and women experienced strain in the relationship and temporary separations after ending IVF. Difficulty in communicating was one reason mentioned by women who experienced strain in the relationship. Another reason was a lack of self-esteem, expressed as worthlessness, upon not being a proper woman. when I felt so bad … there was something wrong with me I wasn't a proper woman, I didn't understand why my husband wanted to stay with me … because I couldn't give him what he wanted … I felt worthless … I thought I couldn't do anything … (F0741). The subject of sexuality was spontaneously mentioned during the interview by some women but not by men. Sexual problems were a reason for temporary separation after IVF. The lust and enjoyment of sexual life disappeared during IVF treatment and had not returned.

when I felt so bad … there was something wrong with me I wasn't a proper woman, I didn't understand why my husband wanted to stay with me … because I couldn't give him what he wanted … I felt worthless … I thought I couldn't do anything … (F0741). The subject of sexuality was spontaneously mentioned during the interview by some women but not by men. Sexual problems were a reason for temporary separation after IVF. The lust and enjoyment of sexual life disappeared during IVF treatment and had not returned. Frustrated at the ending of IVF During the treatment neither men nor women were given advice about whether to end or continue further IVF treatments. Men considered receiving information about the prognosis of undergoing further IVF would have been helpful in their decision-making. The reasons for discontinuing IVF treatment included the women's emotional reactions, medical and economical factors. Men explained that it was their wife's decision not to continue further IVF treatment and they did not always agree with their spouse's decision, and frustration over the decision was expressed. The women themselves stated that it had been their decision to discontinue IVF. After the last IVF treatment, a final consultation with a health professional was lacking according to both men and women. Treatments were described as being too forced and the ending often being abrupt. Both men and women were frustrated at being left on their own when treatment failed.

ir decision to discontinue IVF. After the last IVF treatment, a final consultation with a health professional was lacking according to both men and women. Treatments were described as being too forced and the ending often being abrupt. Both men and women were frustrated at being left on their own when treatment failed. … finished … so quick … that was the end, it was finished … there was also this … someone to talk to afterwards is needed … so it doesn't become as detached as it feels today … it went so fast … (M0337). Unanswered questions after ending IVF Three years after IVF there were still unanswered questions. Women were concerned that they had not been thoroughly evaluated, not being fully examined or not having the fertility factor finally explained were the reasons given by both men and women as to why it was difficult to process childlessness. Both men and women described that an unexplained fertility factor was difficult to accept and it was difficult to explain the cause of infertility to others. Men considered an explained infertility factor was easier to deal with as it could be taken care of. A female factor for infertility was described by women as hard to process, as the husband then could have a child with another woman. However, an explained factor was more difficult for the one carrying the factor according to both men and women.

infertility factor was easier to deal with as it could be taken care of. A female factor for infertility was described by women as hard to process, as the husband then could have a child with another woman. However, an explained factor was more difficult for the one carrying the factor according to both men and women. yes, I think that means a lot … I think that is so for the person who has something wrong with them … that person … I have been there … I think that he can have children … with another woman … you don't just go out and have a child with whoever … I think that it is a grief for those that can't … (F0835). Feeling excluded and lacking understanding For men, childlessness meant being left out in social situations when friends and colleagues were talking about their children; men felt excluded. … to be involuntarily … childless … means some form of exclusion … unless you have … I experience that it is something you can't share with many others … on a different level … it can be ordinary things … such as if you have friends that talk about their children and so on … so I would probably see that as some kind of exclusion … (M0937).

… childless … means some form of exclusion … unless you have … I experience that it is something you can't share with many others … on a different level … it can be ordinary things … such as if you have friends that talk about their children and so on … so I would probably see that as some kind of exclusion … (M0937). Women described childlessness as not being like others, with inability to have children and create a family. Both men and women had experienced upsetting questions and a lack of understanding from their family members, friends and colleagues. Men expressed that being childless affected contact with their own parents, as there were no grandchildren connecting them. The lack of understanding was explained as lack of knowledge about infertility and its treatment, and a disinterest in what it meant to be childless. Women expressed feelings of frustration and anger on the lack of understanding. Men and women communicated differently about their fertility problems. Men described how they seldom informed their closest family, friends and colleagues about the fertility problems, and as a consequence their social network avoided confronting them about their childlessness. Women described more often how they communicated with the family and friends, who consequently tried to support them, but women experienced loss of privacy by telling colleagues at work.

olleagues about the fertility problems, and as a consequence their social network avoided confronting them about their childlessness. Women described more often how they communicated with the family and friends, who consequently tried to support them, but women experienced loss of privacy by telling colleagues at work. Loss of future life goals Three years after ending IVF mental health was described as relatively stable among men. Women revealed being both more mentally stable than during IVF but some were still suffering from grief; it was difficult not knowing how to handle grief. Pain and the absence of a child were the experiences of childlessness described among men. Women described the experience in terms of grief, emptiness, and meaninglessness. … grief … it is a deep grief … yes, that is perhaps the strongest word that comes directly … emptiness … yes, the meaningfulness … there is something missing in your life … in any case when you reach this age … you feel … what is life about … it maybe that … the child part so to say, … it feels as if you have been cheated … (F0939).

s a deep grief … yes, that is perhaps the strongest word that comes directly … emptiness … yes, the meaningfulness … there is something missing in your life … in any case when you reach this age … you feel … what is life about … it maybe that … the child part so to say, … it feels as if you have been cheated … (F0939). Most men and women had not reached an adaptation to childlessness. Men described how they tried to learn to live with the pain by denying it. Men were uncertain whether their spouse had accepted childlessness as the fertility problem was not discussed at home. Women described how they were physically and rationally close to accepting childlessness; however, there would never be an emotional acceptance. Both men and women described that having children was something they had considered self-evident. Blaming oneself for being childless and feelings of guilt were expressed by women, but not by men. The women felt a loss of control over the situation because they were childless. The loss of parenthood, of not having a family including a child, was described as a different life situation; but at the same time, a life situation with more freedom. Some women still had hopes for spontaneous pregnancy but also expressed how the hope for a child had taken a great deal of time in their life. This was one reason for ending the process, but they were still grieving about being childless. The future, as described by men and women, was being alone in old age. Women explained that there would be no one to succeed them and thought about what to do with their inheritance. Both men and women described disappointment at not being able to give their parents the joy of grandchildren and not being able to have their own grandchildren.

d by men and women, was being alone in old age. Women explained that there would be no one to succeed them and thought about what to do with their inheritance. Both men and women described disappointment at not being able to give their parents the joy of grandchildren and not being able to have their own grandchildren. Discussion Three years after ending IVF treatment most men and women were still processing and had not adapted to remaining childless, indicating the grieving process was unresolved. Unsuccessful treatment was experienced by women in terms of grief, whereas men took upon themselves a supportive role and did not express grief reactions, but expressed a need for professional counseling in how to handle grief.

and had not adapted to remaining childless, indicating the grieving process was unresolved. Unsuccessful treatment was experienced by women in terms of grief, whereas men took upon themselves a supportive role and did not express grief reactions, but expressed a need for professional counseling in how to handle grief. The results from the current study support the statement that the crisis theory is insufficient for infertile couples (5), as men and women were still processing and had not adapted to childlessness. Men and women tend to describe emotional reactions according to typical gender roles (16). This was corroborated by the differences in experiences between men and women, such as men having the supportive role, and was in accordance with a study of women and men reacting differently 2 years after unsuccessful tubal surgery (6). Women in the current study described the experience of undergoing unsuccessful treatment in terms of grief, in accordance with previous studies (13–15). Furthermore, women in the current study also described symptoms of depression after IVF failure: as a lack of self-esteem, expressed as worthlessness. A loss of control and blaming oneself for being childless and feelings of guilt were also expressed by women, but not by men. Depression includes feelings of worthlessness, such as loss of self-esteem and is contrary to normal grief reactions where self-esteem is not affected (8). Guilt and loss of control, such as sense of control, are considered as vulnerability factors for distress after IVF (20). There are two aspects of loss of control: of bodily functions and of future life goals (21). Following multiple pregnancy losses and failed treatments, women may be particularly vulnerable to chronic grief (22). Previous pregnancy loss was an independent risk factor for major depression in women in our previous study (23) and had occurred among more than half of the women in the current study. Furthermore, pre-existing personality is considered a determinant factor (3) and women with neurotic personality may be more vulnerable to develop depression after treatment failure (24). Relational strain among men and women has also been shown to be a significant determinant for developing depressive symptoms after failed treatment (25). For some couples, this was one of the reasons for another loss, which included temporary separations or divorce.

e to develop depression after treatment failure (24). Relational strain among men and women has also been shown to be a significant determinant for developing depressive symptoms after failed treatment (25). For some couples, this was one of the reasons for another loss, which included temporary separations or divorce. More losses due to infertility and fewer gains of childlessness, were reported in this study than in previous studies reporting more gains (13,17). A lack of professional support and how men took the supportive role after treatment failure were also described and indicated that men were in need of their own individual support and were in need of their own individual support while undergoing IVF treatment. Although assisted reproduction has been developed during the last two decades, little seems to have improved in terms of support and counseling for couples undergoing IVF, a situation described by Schmidt a decade ago (26). Suggestions for improvement for those providing IVF were given by men in this study which included individual and mandatory support, including information about possible grief reactions and how to handle them, as there was no prior awareness of the need for professional support after failed treatment. These are clinical implications that are possible for a specialized health professional in the IVF team, such as the midwife, to assess for all couples undergoing IVF. Such improvements are needed to facilitate the grieving process after undergoing unsuccessful IVF treatment.

rofessional support after failed treatment. These are clinical implications that are possible for a specialized health professional in the IVF team, such as the midwife, to assess for all couples undergoing IVF. Such improvements are needed to facilitate the grieving process after undergoing unsuccessful IVF treatment. One consequence of grief and depressive symptoms after IVF failure was to discontinue treatment. The decision to end treatment was described by men as being the woman's decision, one reason being grief reactions after treatment failure. The most common reason for discontinuing treatment is emotional burden (27,28). Thus, to limit dropout from further treatment, more effective support may allow the couple an opportunity to reflect upon their options undergoing treatment. Furthermore, to have follow-up appointments, already booked prior to IVF, could be another option for counseling and support when treatment has failed. If depressive symptoms are identified and adequately treated, the chance may be that more couples continue and optimize the chance of successful outcome after treatment (28).

rmore, to have follow-up appointments, already booked prior to IVF, could be another option for counseling and support when treatment has failed. If depressive symptoms are identified and adequately treated, the chance may be that more couples continue and optimize the chance of successful outcome after treatment (28). Lack of a structured end to the IVF treatment may be a factor hampering the processing of childlessness. However, there are no rituals, when ending IVF, for the experience of ongoing childlessness to facilitate an adaptation (1,5). To have some kind of structured summary after IVF, as a final consultation, is important to process the experience of involuntary childlessness. Improved consultation is also suggested to facilitate the decision-making process for the couple on when to end IVF (29). However, ending treatment most often means that the possibilities for having a biological child are over and includes giving up hope of pregnancy. This was expressed by women in the current study by being rationally close to accepting childlessness but not emotionally. Therefore, cessation of treatment does not end infertility for women, as there is no emotional acceptance, indicating increased vulnerability to develop depression (30).

iving up hope of pregnancy. This was expressed by women in the current study by being rationally close to accepting childlessness but not emotionally. Therefore, cessation of treatment does not end infertility for women, as there is no emotional acceptance, indicating increased vulnerability to develop depression (30). Unanswered questions 3 years after IVF were considered as interfering with the processing of childlessness and previous studies indicate that an unexplained infertility factor can be a distress factor hampering the processing of childlessness (14,31) and may lead to depression in men (23). Explanation of the infertility diagnosis, at a final consultation may increase the sense of control (6), facilitate the processing of childlessness and decrease the risk of developing depression. Undoubtedly, a final consultation would be beneficial for couples in resolving remaining issues. With a shorter waiting time for infertility evaluation and treatment and fewer visits to the clinic than a decade ago (26), there is less time and opportunity for explaining the cause of infertility either before or during IVF. Therefore, the result of the evaluation, whether an explained or unexplained factor has been found, needs to be thoroughly clarified to the couple not only during IVF but also at a final consultation. Information about other treatment options or alternatives available or whether to end treatment needs to be discussed. Counseling may be helpful for the couple in their decision-making process within this often ambiguous situation.

thoroughly clarified to the couple not only during IVF but also at a final consultation. Information about other treatment options or alternatives available or whether to end treatment needs to be discussed. Counseling may be helpful for the couple in their decision-making process within this often ambiguous situation. Social isolation, a feeling of being excluded with no understanding from their social network, was experienced by both men and women after IVF, and communication about fertility problems with close family members and friends differed between men and women, resulting in less social support among men. A lack of social support or discontent with support given leads to more grief and depression in women (32). For those who actively seek social support and express negative feelings, a healthier mental health outcome is described than for those who deny negative feelings and do not seek or accept support from their social network (3).

l support or discontent with support given leads to more grief and depression in women (32). For those who actively seek social support and express negative feelings, a healthier mental health outcome is described than for those who deny negative feelings and do not seek or accept support from their social network (3). A loss of future life goals was described among men and women in the current study who had not adapted to remaining childless 3 years after undergoing IVF. Another study revealed that among women who consider themselves as definitely childless, substantial proportion have high scores of complicated grief and depressive symptoms (32). It is important for health professionals to distinguish between normal grief and complicated grief (10) and identify and follow the individuals at risk of developing depression after IVF failure and to offer evidence-based treatment. However, among some women there was still a hope for spontaneous pregnancy and among other participants there were early plans for adoption, indicating that they were not planning to remain childless.

fy and follow the individuals at risk of developing depression after IVF failure and to offer evidence-based treatment. However, among some women there was still a hope for spontaneous pregnancy and among other participants there were early plans for adoption, indicating that they were not planning to remain childless. Qualitative research aims to attain meaning and understanding of individual's experience (19,33). The qualitative design of this study enabled assessment of the experiences of infertile individuals in more detail and depth compared to a quantitative approach using questionnaires. One concern with assessment by questionnaires is that self-report measures may be susceptible to social desirability bias. Infertile couples may want to present as well-adjusted emotional state especially prior to treatment (34) and may respond more positively in questionnaires (29). Qualitative research interviews with its interactive characteristics have a potential to diminish this tendency. The interviews were all conducted, transcribed and analyzed by the same person, who had been working as a midwife at the clinic where the participants previously had undergone IVF. This provided an initial insight and knowledge in the research field, a pre-understanding considered important in qualitative research. However, when interpreting the results, on the other hand, it is important to clarify the pre-understanding (19,33). Therefore, to problematize the interpretation, the co-author who had not worked in an IVF-setting discussed alternative interpretations if considered necessary. However, transferability of these results has to be made with caution. The extent of the transferability of the result of a qualitative study to other settings and populations depends on cultural and traditional similarities or differences (18).

IVF-setting discussed alternative interpretations if considered necessary. However, transferability of these results has to be made with caution. The extent of the transferability of the result of a qualitative study to other settings and populations depends on cultural and traditional similarities or differences (18). A limitation of the present study is that men with male infertility factor and women with university level education were few compared to those attending the clinic where recruitment was done (35). Furthermore, individuals with other ethnic backgrounds than Swedish are not included. Another limitation was that sexual problems, spontaneously mentioned by women, were not a topic guiding the interviews and therefore the male participants' experiences of sexuality was not assessed. In conclusion, unresolved grief was the main experience of remaining childless after undergoing unsuccessful IVF. The recommendations for reproductive health professionals are to provide additional individual support, as men and women were experiencing involuntary childlessness differently. Information and counseling in an early phase of IVF, concerning grief reactions following IVF failure is recommended. A structured final consultation after IVF, with time to discuss the result of the infertility evaluation and if other treatment alternatives are available, may facilitate the grieving process after undergoing unsuccessful IVF treatment. This study was supported by research grants from the Family Planning Foundation and the Foundation of Caring Sciences, Uppsala University.

In conclusion, unresolved grief was the main experience of remaining childless after undergoing unsuccessful IVF. The recommendations for reproductive health professionals are to provide additional individual support, as men and women were experiencing involuntary childlessness differently. Information and counseling in an early phase of IVF, concerning grief reactions following IVF failure is recommended. A structured final consultation after IVF, with time to discuss the result of the infertility evaluation and if other treatment alternatives are available, may facilitate the grieving process after undergoing unsuccessful IVF treatment. This study was supported by research grants from the Family Planning Foundation and the Foundation of Caring Sciences, Uppsala University. Declaration of interest The authors report no conflicts of interest. The authors are responsible for the content and writing of the paper.

Introduction Modern healthcare is expensive, especially surgery performed in a regular operating theatre. General anesthesia involves life-threatening risks and undesirable side-effects. Local anesthesia minimizes these drawbacks and can be administered by the surgeon. This promotes simplification. Female sterilization in an office situation using laparoscopy with a low-pressure pneumoperitoneum administered under local anesthesia has been performed for more than 30 years (1), but has not gained widespread popularity. A survey of the American Association of Gynecologic Laparoscopists (2) showed that only 5.1% of its members performed office-based laparoscopy under local anesthesia. This is a low incidence for doctors with a special interest in laparoscopy. One explanation can be that active insufflation of gas gives too much pain and necessitates heavy conscious sedation and surveillance by specially trained personnel (3). Even a low-pressure pneumoperitoneum is painful. By contrast, a zero-pressure (gasless) pneumoperitoneum is painless. Office-based laparoscopy and gasless laparoscopy are not new procedures, but gasless laparoscopy under local anesthesia is a new procedure. Two articles have been published on gasless laparoscopic female sterilization but these operations were done under general anesthesia (4,5).

(gasless) pneumoperitoneum is painless. Office-based laparoscopy and gasless laparoscopy are not new procedures, but gasless laparoscopy under local anesthesia is a new procedure. Two articles have been published on gasless laparoscopic female sterilization but these operations were done under general anesthesia (4,5). Carbon dioxide gas is cold, very dry, dissolves in water and provokes hypothermia, desiccation, tissue irritation, and acid-base and blood gas changes. By comparison, room air is warmer, humid, insoluble, not irritating and available everywhere. Gasless laparoscopy depends on mechanical lifting of the abdominal wall with a passive inflow of room air. An open access technique for trocar insertion is safer than a closed access technique and is therefore a better choice for office laparoscopy (1). The open access technique results in a 1.5–2 cm wide hole in the abdominal wall, so using a micro-laparoscope is meaningless. An ordinary laparoscope has advantages in focal distance and field of vision.

rocar insertion is safer than a closed access technique and is therefore a better choice for office laparoscopy (1). The open access technique results in a 1.5–2 cm wide hole in the abdominal wall, so using a micro-laparoscope is meaningless. An ordinary laparoscope has advantages in focal distance and field of vision. Material and methods This was a prospective pilot study to evaluate a mechanical approach to lifting, used for office-based gasless laparoscopic sterilization. Women with a body mass index (BMI) < 30 kg/m2, no serious illnesses and no known abdominal adhesions were informed about general and local anesthesia during presterilization counseling sessions. All women without contraindications for office surgery chose local anesthesia and were given written information about the procedure. Between September 2003 and December 2005, 35 women were sterilized in a low resource setting situated one floor below a regular operating theatre. These women had a mean age of 39 years (range 27–48), BMI 24.1 kg/m2 (19–30). Fourteen women had had a previous abdominal operation. All women but one had a normal-sized uterus. The procedure room had resuscitation equipment including oxygen and suction, an emergency tray with diazepam (5 mg/ml), atropine (1 mg/ml), catastrophic adrenaline (0.1 mg/ml) and a narcotic antidote, electrocardiography equipment, pulse oximeter and an alarm button.

ration. All women but one had a normal-sized uterus. The procedure room had resuscitation equipment including oxygen and suction, an emergency tray with diazepam (5 mg/ml), atropine (1 mg/ml), catastrophic adrenaline (0.1 mg/ml) and a narcotic antidote, electrocardiography equipment, pulse oximeter and an alarm button. Women had fasted for at least 6 hours and were premedicated orally with 200 mg ibuprofen and 1 g paracetamol/30 mg codeine phosphate about 1 hour before the start of surgery. After voiding, each patient walked to the procedure room where she was given intravenous fluid. Personnel included the surgeon, an assistant to arrange the instruments and a midwife to monitor the patient's blood pressure, pulse rate, respiratory rate, blood oxygen saturation, electrocardiogram, level of sedation and to give medications on request from the surgeon. Only the surgeon was dressed in sterile clothing. The patient was cleaned by the assistant and draped by the surgeon and mildly sedated with 5 mg diazepam and 25 mg meperidine: doses that could be repeated once. The muscle relaxant effect of diazepam was of value in the mechanical stretching of the abdominal wall. A mixture of 40 ml 1% lidocaine hydrochloride/adrenaline and 60–80 ml 0.9% sterile NaCl was used for local anesthesia. The anterior region of the cervical portio was anesthetized and a Hulka forceps attached. The abdominal wall just beneath the umbilicus was anesthetized and a >12.5-mm trocar was inserted into the abdominal wall using an open access technique. The lifting technique used the camera trocar as an anchoring device in the abdominal wall. The open trocar gas inlet allowed a free inflow of room air. An operative (0°, 10 mm) laparoscope with a 6-mm working channel was used. The trocar/abdominal wall were lifted with a loop of polydioxanone suture (PDS #1) snared around the shaft of the trocar with a hang knot and needle-driven through the fascia, cutaneous tissue and skin in the lower end of the abdominal wall incision. The loop suture was attached to a horizontal metal arm mounted on the operating table and placed above the woman.

loop of polydioxanone suture (PDS #1) snared around the shaft of the trocar with a hang knot and needle-driven through the fascia, cutaneous tissue and skin in the lower end of the abdominal wall incision. The loop suture was attached to a horizontal metal arm mounted on the operating table and placed above the woman. Padded shoulder supports were an important prerequisite, because if a woman were to slide downwards on the table she would be likely to get scared and tense her abdominal muscles to stay put. This would tend to press the intestines into the pelvis. However, despite the shoulder supports it was necessary for the women to lie with their pelvic region 10 cm outside the table. This position gave necessary sliding distance to prevent the Hulka forceps from being blocked by the table when it was placed in steep Trendelenburg position.

ss the intestines into the pelvis. However, despite the shoulder supports it was necessary for the women to lie with their pelvic region 10 cm outside the table. This position gave necessary sliding distance to prevent the Hulka forceps from being blocked by the table when it was placed in steep Trendelenburg position. Mechanical lifting of the abdominal wall with the camera trocar as an anchoring device and with the trocar gas inlet open, led to passive filling of the abdomen with air. Lifting the skin and subcutaneous tissue 6–8 cm above the symphysis pubis in the same way using a towel clamp allowed sufficient air into the abdomen for laparoscopic sterilization. The combination of mechanical lifting, the Trendelenburg position and the forward rotation of the uterus created adequate intra-peritoneal space to identify, anesthetize, coagulate and cut the tubes. If the mechanical lifting procedure did not create sufficient space to identify the tubes, a small amount of additional room air (1) was insufflated actively using a rubber bulb. A towel clamp closing the upper end of the abdominal wall incision then secured air tightness. The mechanical lifting supported most of the weight of the abdominal wall and only a small volume of actively insufflated air and a low intra-abdominal pressure increase were needed to create additional space. After the tubes had been anesthetized, coagulated, divided with hook-scissors and the intra-peritoneal air been reduced to a minimum, the abdominal wall opening was closed. The women were able to sit up for a minute and then walk back to the recovery room.

minal pressure increase were needed to create additional space. After the tubes had been anesthetized, coagulated, divided with hook-scissors and the intra-peritoneal air been reduced to a minimum, the abdominal wall opening was closed. The women were able to sit up for a minute and then walk back to the recovery room. All women received postoperative information from the surgeon or the midwife. They were given oral medication for 2 days (200 mg ibuprofen × 3 and 1 g paracetamol/30 mg codeine phosphate × 3) and a questionnaire to be sent back in 1 week (pain score, worst painful moment, satisfaction rate, complications, validity of the presterilization counseling session).

tion from the surgeon or the midwife. They were given oral medication for 2 days (200 mg ibuprofen × 3 and 1 g paracetamol/30 mg codeine phosphate × 3) and a questionnaire to be sent back in 1 week (pain score, worst painful moment, satisfaction rate, complications, validity of the presterilization counseling session). Results All women had a successful tubal ligation. The overall satisfaction rate was 97%. Two of the first 10 women in the study had a small amount of filtered room air insufflated actively, but when the camera trocar lifting procedure was complemented by supra-pubic lifting, there was no need for active air filling because the intra-abdominal laparoscopic view was always good. No woman reported the mechanical lifting moment to be painful: one stating, ‘It was like being lifted in the pants'. Thirty-four of 35 women were satisfied in that they answered ‘yes’ to the question of whether they would recommend the same operation to their best friend. One woman was not satisfied; the reason was much pain when one of the tubes was electro-coagulated. This was the only woman to report that the interprocedural pain was worse than was expected from the information given in the presterilization counseling. Per/postoperative pain was measured using a visual 0–10 analog scale (VAS). The 34 satisfied women reported an average score of 2.6 (range 0–7.5) and three expressed no pain at all. The potentially painful moments during the operation were the different needle pinpricks, the Hulka forceps manipulations, unintentional rough touching of pelvic organs and anesthetic failure. One woman reported VAS 7.5 for the steep Trendelenburg position and one woman VAS 7.0 for the Hulka forceps application. Minor sedation was sufficient. All women walked to and from the operating room. No operation was converted to general anesthesia and there were no complications except for one wound infection. All women left for home after 1–5 hours and all submitted a completed outcome questionnaire.

for the Hulka forceps application. Minor sedation was sufficient. All women walked to and from the operating room. No operation was converted to general anesthesia and there were no complications except for one wound infection. All women left for home after 1–5 hours and all submitted a completed outcome questionnaire. Discussion In the first 10 operations, a 1.2-mm puncture needle was used for tubal anesthesia. This needle had a tendency to push the tube in front of itself rather than to penetrate. Later, a 0.4-mm needle was used. When using a 10-mm laparoscope, a >12.5 mm camera trocar with an open high-flow gas inlet is necessary for the free passive inflow of room air. A smaller trocar can reduce the inflow of air, and result in a negative intra-abdominal pressure, that limits the space for inspection and instrumentation and causes pain. The used mechanical lifting procedure has a short setup time (< 1–2 min), introduces no extra devices into the peritoneal cavity, causes no trauma to the peritoneal surface and does not interfere with surgical movements.

intra-abdominal pressure, that limits the space for inspection and instrumentation and causes pain. The used mechanical lifting procedure has a short setup time (< 1–2 min), introduces no extra devices into the peritoneal cavity, causes no trauma to the peritoneal surface and does not interfere with surgical movements. To further simplify the process, the surgeon can use an amnioscope (20 × 200 × 25 mm), which permits direct visual inspection and instrumentation of the tubal areas (Figure 1). Short instruments are then used for an optimal visual distance. Since 1993, the author has performed more than 200 female sterilizations using an amnioscope. The only drawback compared to using a laparoscope are a less comfortable working position for the surgeon and the fact that the personnel and patient cannot watch the procedure on a monitor. An amnioscope with or without an operative laparoscope is a good choice, that supports a high inflow of room air and to insert a second working instrument beside the laparoscope to displace obscuring loops of distended bowel, if any. Figure 1 Mechanical lifting

To further simplify the process, the surgeon can use an amnioscope (20 × 200 × 25 mm), which permits direct visual inspection and instrumentation of the tubal areas (Figure 1). Short instruments are then used for an optimal visual distance. Since 1993, the author has performed more than 200 female sterilizations using an amnioscope. The only drawback compared to using a laparoscope are a less comfortable working position for the surgeon and the fact that the personnel and patient cannot watch the procedure on a monitor. An amnioscope with or without an operative laparoscope is a good choice, that supports a high inflow of room air and to insert a second working instrument beside the laparoscope to displace obscuring loops of distended bowel, if any. Figure 1 Mechanical lifting Using an amnioscope and operate under local anesthesia and mild sedation in an office setting conforms well to the statement of the WHO Task Force on Female Sterilization (6). ‘The ideal female sterilization would involve a simple, easily learned, onetime procedure that could be accomplished under local anesthesia and involve a tubal occlusion technique that caused minimum damage. The procedure would be safe, have high efficacy, be readily accessible, and be personally and culturally acceptable. The cost for each procedure would be low and there would be minimal costs for the maintenance of equipment'.

hed under local anesthesia and involve a tubal occlusion technique that caused minimum damage. The procedure would be safe, have high efficacy, be readily accessible, and be personally and culturally acceptable. The cost for each procedure would be low and there would be minimal costs for the maintenance of equipment'. In developing countries, where minilaparotomy is a common approach, women would probably benefit from this amnioscope lifting procedure. Compared with minilaparotomy, it gives a much better view of the pelvic organs, an aesthetically more acceptable scar, a shorter recovery period and less complications/ complaints (6). Using an ordinary anesthesia frame in the lifting process, with the horizontal arm draped in a sterile sleeve, makes the procedure progress even more smoothly. In developing countries, where resources are limited for the purchase and maintenance of more sophisticated laparoscopic equipment, this amnioscope lifting procedure is truly a cheaper and safer option for female sterilization than the traditional laparoscopic procedure presently used in the developed world. Office laparoscopic sterilization requires, in contrast to office hysteroscopic sterilization, no scheduling of surgery according to the woman's menstrual cycle and has an almost 100% first-attempt success rate, an immediate effect on fertility, no need for tubal patency control, no material costs, a reversal success rate of 55–75% (7), an unchanged possibility of IVF and can be performed within 48 hours of delivery.

heduling of surgery according to the woman's menstrual cycle and has an almost 100% first-attempt success rate, an immediate effect on fertility, no need for tubal patency control, no material costs, a reversal success rate of 55–75% (7), an unchanged possibility of IVF and can be performed within 48 hours of delivery. Office laparoscopy is claimed to cost much less than traditional laparoscopy. In one study, there was an almost 80% reduction in costs (8), which is in agreement with the present study. Our total costs in 2006 were calculated to be NOK 2,895 (US $446), which represents a 75% reduction. Gasless laparoscopic sterilization and hysteroscopic sterilization (Essure® device) can both be done in an office-based setting with the same OR-team and operation time. The Essure procedure is however approximately US $1,575 more expensive due to additional costs for the Essure® devices (ESS305, Conceptus, Inc., USA, Retail Price: $1,299) and for tubal occlusion control (HSG US $275) and is contraindicated within 6 weeks after delivery and in women with hypersensitivity to nickel or allergy to contrast media. A comparative study/literature review of hysteroscopic sterilization versus laparoscopic tubal sterilization (9) showed overall standard complication rates for laparoscopic sterilization of 0.8–0.9% (6,9) and a major complication rate for hysteroscopic sterilization of 3.2% (9). Moreover, correct use of local anesthesia removes the single greatest source of risk in conventional laparoscopic sterilization procedures, general anesthesia.

overall standard complication rates for laparoscopic sterilization of 0.8–0.9% (6,9) and a major complication rate for hysteroscopic sterilization of 3.2% (9). Moreover, correct use of local anesthesia removes the single greatest source of risk in conventional laparoscopic sterilization procedures, general anesthesia. A gastight mechanical lifting procedure, that creates adequate intra-peritoneal space for female sterilization under local anesthesia, is also useful in surgery under general anesthesia. Therefore, in conventional gas laparoscopy, the author also uses the described lifting technique. Such lift-assisted laparoscopy makes surgery in low gas pressure (1–6 mm Hg) possible with sustained optimal or adequate view (10) and, in case of need, to take temporary measures in a ‘gasless’ condition. An immediate shift between low, standard and zero gas pressure is possible. With no gas pressure, conventional open surgery instruments such as clamps, scissors and powerful suction devices can be used and with standard gas pressure, the complementary abdominal wall lifting means a ‘double’ outcome in forming the intra-abdominal space. A special slit-trocar facilitates the shifting maneuver between the gas-based and gasless technique (Figure 2). The finding that the centrally positioned camera trocar, except for its normal function as a gastight sleeve for the laparoscope, is also a perfect anchoring device for mechanical lifting, is an enhancement for laparoscopic surgery. ‘Yesterdays’ gasless laparoscopy must not be confused with lift-assisted laparoscopy. The European Association for Endoscopic Surgery states that ‘gasless laparoscopy has no clinically relevant advantages compared to low-pressure (5–7 mm Hg) pneumoperitoneum’ (11). There may be two exceptions to this: laparoscopy under local anesthesia and lift-assisted laparoscopy under general anesthesia if it is required to use a conventional open surgery instrument. All patients benefit from low-pressure gas laparoscopy and especially high-risk patients, including pregnant women.

oneum’ (11). There may be two exceptions to this: laparoscopy under local anesthesia and lift-assisted laparoscopy under general anesthesia if it is required to use a conventional open surgery instrument. All patients benefit from low-pressure gas laparoscopy and especially high-risk patients, including pregnant women. Figure 2 Slit-trocar for gas/gasless laparoscopy The evaluated mechanical lifting technique can be applied effectively in office laparoscopic sterilization, even to overweight women. Risks of general anesthesia and active gas insufflation are eliminated and room air, two strings and a needle, replace a CO2 gas filling machine and narcosis Declaration of interest The author has invented the slit-trocar.

Key Message Women with intellectual disability have a higher risk of adverse pregnancy outcome and should be considered a risk group. Caregivers need to be aware of this and provide better tailored pre- and intrapartum care and support. Introduction Studies describing pregnancy and childbearing in women with intellectual disability (ID) are few and knowledge is limited. However, Australian studies show that the health of mothers with intellectual limitations is worse than other women's health (1) and one-third of women interviewed reported moderate to severe depression, anxiety, and stress during pregnancy (2). Other studies revealed a higher incidence of preeclampsia (3) and preterm births (28%) (4). Internationally, according to the World Health Organization, ID is defined as a person with an intelligence quotient below 70, a derogation of adaptive capacity, and a debut before 18 years of age (5). During most of the 20th century, women with ID were institutionalized and/or sterilized to prevent pregnancy. In Sweden, involuntary sterilizations occurred between 1934 and 1975 (6). Currently, there is no legislation prohibiting childbearing in this group; however, negative attitudes towards pregnancies and parenthood exist (7). Few previous Scandinavian studies have investigated pregnancy and childbirth in women with ID. One Swedish study described involuntary sterilizations of women with ID after childbirth or abortion (8).

ation prohibiting childbearing in this group; however, negative attitudes towards pregnancies and parenthood exist (7). Few previous Scandinavian studies have investigated pregnancy and childbirth in women with ID. One Swedish study described involuntary sterilizations of women with ID after childbirth or abortion (8). The aim of the present study was therefore to investigate antenatal health and demographic factors as well as pregnancy and delivery outcomes (mode of delivery, pain relief during labor, preterm birth and discharge from hospital) in women with ID in Sweden compared to women without ID or any psychiatric diagnosis. Material and methods The data were collected from two registers and the use of personal identification numbers allows information between different registers to be linked (9–11). The dataset comprised linkage of the two health registers: the National Patient Register (NPR) and the Medical Birth Register (MBR) (12,13). The NPR has a nationwide coverage of facility care from 1987 and out-patient care was added from 2001. The MBR, established in 1973, contains demographic and medical information from all hospital births and the few home births in Sweden. It covers 98–99% of all births and compiles information related to antenatal and perinatal factors that could be of importance for the health of the infant. The study was approved by the Regional Ethical Committe in Uppsala (327/2007).

and medical information from all hospital births and the few home births in Sweden. It covers 98–99% of all births and compiles information related to antenatal and perinatal factors that could be of importance for the health of the infant. The study was approved by the Regional Ethical Committe in Uppsala (327/2007). All women with an ID diagnosis of International Classification of Diseases (ICD) 8 to 10, chapter V (ICD-8 codes 311–315, ICD-9 codes 317–319, ICD-10 codes F70-F79) were identified from the NPR. This group was linked to MBR, which resulted in a sample including all pregnancies and births to ID women from 1973 to 2006 in Sweden (n = 2332 births). To enable a more stringent comparison, only single pregnancies and primiparas were selected for analysis. Data from 1999–2007 were included because, during this period, all pregnancy and delivery variables of interest for the study were represented. To allow comparison, all women without ID or any other psychiatric diagnoses (n = 340 624) who had their first child (only singletons) during the same period were identified from the Swedish Medical Birth Register and served as a control group. This resulted in 326 women with ID and 340 624 women without ID or any other psychiatric diagnosis.

women without ID or any other psychiatric diagnoses (n = 340 624) who had their first child (only singletons) during the same period were identified from the Swedish Medical Birth Register and served as a control group. This resulted in 326 women with ID and 340 624 women without ID or any other psychiatric diagnosis. For comparison we selected the majority of variables on demographic factors, maternal health, pregnancy and delivery outcomes that were available in MBR during the entire time-period of investigation. Some of these variables had also been used in other studies, which enabled comparisons (1–3,7). We omitted a small number of variables that were not well registered (many missing values) such as smoking in the third trimester. The outcome variables were maternal health and socio-demographic circumstances at registration for antenatal care, mode of delivery, pain relief during labor, preterm birth and discharge from hospital. The proportion of missing values in all variables was similar in the two groups except for ‘working at registration’ where data were missing for 24.5% of women with ID and 14.2% of women without ID. Missing values were excluded from the analysis.

ery, pain relief during labor, preterm birth and discharge from hospital. The proportion of missing values in all variables was similar in the two groups except for ‘working at registration’ where data were missing for 24.5% of women with ID and 14.2% of women without ID. Missing values were excluded from the analysis. Maternal age was defined as completed years at the time of delivery and cohabiting if the woman lived with the child's father at the time she registered for antenatal care. Body mass index (BMI) ≤ 24.9 was categorized as lean and normal, overweight as BMI 25.0–29.9 and obese as BMI ≥ 30. Gestational age was based on the results of ultrasonography performed early in the second trimester. Preterm birth is defined by the World Health Organization as the delivery of a child before 37 completed weeks of gestation. Discharge to a place other than home means that the woman was not discharged directly to home from hospital but to another care facility. ID was here defined as non-syndromic ID (14).

trimester. Preterm birth is defined by the World Health Organization as the delivery of a child before 37 completed weeks of gestation. Discharge to a place other than home means that the woman was not discharged directly to home from hospital but to another care facility. ID was here defined as non-syndromic ID (14). Statistical analysis Statistical analyses were with the SPSS 15.0 software program for Windows (IBM Statistical Package for the Social Sciences). Descriptive statistics were used to describe the frequencies of all the variables presented. The differences between women with ID and women without ID were analyzed by chi-squared test or risk ratios (RR) with 95% confidence intervals (Mantel–Haenszel method). Interval scaled variables were analyzed by comparing means with the independent-samples t-test. Variables that differed in the univariate analysis were inserted into a binary logistic regression analysis. To reveal associations between preterm birth, cesarean section (CS), nitrous oxide, discharge to place other than home and ID we used unadjusted binary logistic regression and binary logistic regression adjusted for maternal age, BMI, cohabitation, working, smoking and epilepsy (age was used as a continuous variable).

To reveal associations between preterm birth, cesarean section (CS), nitrous oxide, discharge to place other than home and ID we used unadjusted binary logistic regression and binary logistic regression adjusted for maternal age, BMI, cohabitation, working, smoking and epilepsy (age was used as a continuous variable). Results At the first antenatal visit more women with ID did not cohabit (36.6%) with the father of the expected child than did women without ID (6.2%) Table 2. More than half of the women with ID did not work (54.1%), compared with 14.3% of the women without ID. In the comparative analysis, the mean age of the women with ID was 24.2 years (range 16–46) and the mean age of the women without ID was 28.3 years (range 11–55) (Table 1). The proportion of teenage births (11–19 years) was higher in women with ID (18.4%) than in women without ID (3.3%). It was more common for women with ID to give birth before the age of 24 years (58.6%) than for women without ID (22.4%). Few women in both groups gave birth at an age of 40 years or older. Most women with ID (88.7%) had attended antenatal care services. The diagnosis preeclampsia was registered in 1.1% of the cases. Diabetes was registered in 0.8% of the cases. Table 1 Characteristics of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID

Results At the first antenatal visit more women with ID did not cohabit (36.6%) with the father of the expected child than did women without ID (6.2%) Table 2. More than half of the women with ID did not work (54.1%), compared with 14.3% of the women without ID. In the comparative analysis, the mean age of the women with ID was 24.2 years (range 16–46) and the mean age of the women without ID was 28.3 years (range 11–55) (Table 1). The proportion of teenage births (11–19 years) was higher in women with ID (18.4%) than in women without ID (3.3%). It was more common for women with ID to give birth before the age of 24 years (58.6%) than for women without ID (22.4%). Few women in both groups gave birth at an age of 40 years or older. Most women with ID (88.7%) had attended antenatal care services. The diagnosis preeclampsia was registered in 1.1% of the cases. Diabetes was registered in 0.8% of the cases. Table 1 Characteristics of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID Women without ID, mean n = 340 624 Women with ID, mean n = 326 p-value Missing data Total % Age (year) 28.28 24.20 <0.001* 0.3 (range) (11–55) (16–46) Weight at registration (means in kilogram) 66.67 68.47 0.015* 11.8 (range) (32–188) (38–128) Body mass index 24.02 25.42 <0.001* 13.6 (range) (15.01–63.17) (15.23–45.91) * Student's t-test. Table 2 Sociodemographic and maternal health data of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID Women without ID (ref) Women with ID

Women without ID, mean n = 340 624 Women with ID, mean n = 326 p-value Missing data Total % Age (year) 28.28 24.20 <0.001* 0.3 (range) (11–55) (16–46) Weight at registration (means in kilogram) 66.67 68.47 0.015* 11.8 (range) (32–188) (38–128) Body mass index 24.02 25.42 <0.001* 13.6 (range) (15.01–63.17) (15.23–45.91) * Student's t-test. Table 2 Sociodemographic and maternal health data of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID Women without ID (ref) Women with ID n % n % Relative risk 95%CI Missing data Total% Not cohabiting with the father at registration 19,826 6.2 108 36.6 5.9 4.9–7.1 6.5 Not working at registration 41,708 14.3 133 54.1 3.8 3.2–4.5 14.2 Smoking at registration 24,949 7.9 84 27.9 3.5 2.9–4.4 7.2 Epilepsy 1371 0.4 18 5.5 13.7 8.6–21.8 0.0 Preterm birth (weeks 22–29) 1712 0.5 5 1.5 3.0 1.3–7.3 Preterm birth (weeks 30–36) 19,055 5.6 35 10.7 1.9 1.4–2.7 Prolonged gestation (weeks 43–45) 2483 0.7 2 0.6 0.9 0.8–1.1 CI, confidence interval. The mean weight at registration for antenatal care was about 2 kg higher in women with ID (68.5 kg) than in women without ID (66.7 kg). The mean BMI at registration was 25.4 and 24.0, respectively. The occurrence of obesity in women with ID was higher (20.1%) than in women without ID (8.6%). The mean weight gain during pregnancy was 13.1 kg (weight at delivery – weight at registration).

her in women with ID (68.5 kg) than in women without ID (66.7 kg). The mean BMI at registration was 25.4 and 24.0, respectively. The occurrence of obesity in women with ID was higher (20.1%) than in women without ID (8.6%). The mean weight gain during pregnancy was 13.1 kg (weight at delivery – weight at registration). More than one-quarter (27.9%) of the women with ID smoked at the first antenatal visit compared with 7.9% of women without ID. Smoking decreased over time in the ID group from 33.8% (1999–2001 and 33.0% (2002–2004) to 19.5% (2005–2007). A medical history of epilepsy was registered 14 times more often for women with ID (5.5%) than for women without ID (0.4%) at the first antenatal visit. For women with ID, 12.2% of pregnancies ended with a preterm birth (<37 weeks of gestation), compared with 6.1% of the women without ID. Over time, preterm births occurred in 23.3% (1999–2001), 19.2% (2002–2004) and 17.3% (2005–2007) of women with ID.

5.5%) than for women without ID (0.4%) at the first antenatal visit. For women with ID, 12.2% of pregnancies ended with a preterm birth (<37 weeks of gestation), compared with 6.1% of the women without ID. Over time, preterm births occurred in 23.3% (1999–2001), 19.2% (2002–2004) and 17.3% (2005–2007) of women with ID. Spontaneous onset of labor occurred in 74.5% of women with ID and 81.5% women without ID (Table 3). Vaginal birth was equally common in both groups. Instrumental delivery by vacuum extraction occurred in 8.9% women with ID and 13.8% in women without ID. CS was more common in women with ID (24.5%) than in women without ID (17.7%). Epidural blockade was administered equally to both groups. However, there was a difference in the use of nitrous oxide: 59.5% in women with ID vs. 75.8% in women without ID. The same difference in the use of nitrous oxide was found in the subgroup of women who had a normal vaginal birth (71.9 vs. 84.1%). The use of nitrous oxide among women with ID increased over time: 51.1% (1999–2001), 56.9% (2002–2004) and 67.7% (2005–2007). The women with ID were more likely to be discharged from the maternity ward directly to a place other than their homes (6.5%) than were women without ID (2.4%) were. The multivariate analysis included statistically significant variables from the univariate analysis (Table 4). An ID diagnosis was associated with all the outcomes investigated. Table 3 Delivery data of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID Women without ID (ref) Women with ID

Spontaneous onset of labor occurred in 74.5% of women with ID and 81.5% women without ID (Table 3). Vaginal birth was equally common in both groups. Instrumental delivery by vacuum extraction occurred in 8.9% women with ID and 13.8% in women without ID. CS was more common in women with ID (24.5%) than in women without ID (17.7%). Epidural blockade was administered equally to both groups. However, there was a difference in the use of nitrous oxide: 59.5% in women with ID vs. 75.8% in women without ID. The same difference in the use of nitrous oxide was found in the subgroup of women who had a normal vaginal birth (71.9 vs. 84.1%). The use of nitrous oxide among women with ID increased over time: 51.1% (1999–2001), 56.9% (2002–2004) and 67.7% (2005–2007). The women with ID were more likely to be discharged from the maternity ward directly to a place other than their homes (6.5%) than were women without ID (2.4%) were. The multivariate analysis included statistically significant variables from the univariate analysis (Table 4). An ID diagnosis was associated with all the outcomes investigated. Table 3 Delivery data of women without intellectual disability (ID) or any psychiatric diagnosis, and women with ID Women without ID (ref) Women with ID n % n % Relative risk 95% CI Missing data Total% Cesarean section 60,130 17.7 80 24.5 1.4 1.1–1.7 0.0 Vacuum extraction 47,134 13.8 29 8.9 0.6 0.5–0.9 0.0 Epidural anesthesia 142,664 41.9 123 37.7 1.1 0.9–1.2 0.0 Nitrous oxide 258,265 75.8 194 59.5 0.8 0.7–0.9 0.0 No pharmacological pain relief 17,812 5.2 23 7.1 1.4 0.9–2.0 0.0 No pain relief at all 4644 1.4 4 1.2 0.9 0.3–2.4 0.0 Discharge to place other than home 7995 2.4 21 6.5 2.8 1.8–4.2 1.1 CI, confidence interval.

Epidural anesthesia 142,664 41.9 123 37.7 1.1 0.9–1.2 0.0 Nitrous oxide 258,265 75.8 194 59.5 0.8 0.7–0.9 0.0 No pharmacological pain relief 17,812 5.2 23 7.1 1.4 0.9–2.0 0.0 No pain relief at all 4644 1.4 4 1.2 0.9 0.3–2.4 0.0 Discharge to place other than home 7995 2.4 21 6.5 2.8 1.8–4.2 1.1 CI, confidence interval. Table 4 Unadjusted odds ratios (95% confidence interval) and adjusted odds ratios (95% confidence interval) for preterm birth, cesarean section, non-use of nitrous oxide and discharge to place other than home in women with ID vs. women without ID Women with ID vs. women without ID Crude OR (95%CI) Adjusted OR (95%CI) Preterm birth 2.15 (1.55–3.00) 1.68 (1.06–2.67) Cesarean section 1.52 (1.18–1.95) 1.55 (1.11–2.17) Non-use of nitrous oxide 2.13 (1.71–2.66) 1.89 (1.43–2.50) Discharge to other place than home 2.88 (1.85–4.49) 2.24 (1.22–4.17) Adjusted for maternal age, obesity, cohabitation, working, smoking and epilepsy. CI, confidence interval. Discussion This study shows several differences between women with ID and women without ID during pregnancy and childbirth, e.g. a higher proportion of teenagers, obesity and single status among women with ID. Women with ID more often had preterm births and CS. Women with ID used less nitrous oxide as pain relief during delivery than women without ID did, and were more often discharged from the maternity ward to a place other than home. Moreover, there was an overrepresentation of smokers among women with ID at the first antenatal visit.

more often had preterm births and CS. Women with ID used less nitrous oxide as pain relief during delivery than women without ID did, and were more often discharged from the maternity ward to a place other than home. Moreover, there was an overrepresentation of smokers among women with ID at the first antenatal visit. This study was prospective, based on a standardized collection of data gathered at maternal care units and hospitals; therefore, recall bias was avoided. Maternal care is free of charge and home deliveries are rare in Sweden, thus, selection bias is unlikely. A weakness with the Swedish MBR is the lack of systematic documentation of some variables (11). Sources of errors include incorrect documentation in medical records, changes in medical records, missing records, and deficient transmission into the Swedish MBR. Medical records on pregnancy and childbirth do not include education level and housing condition variables and these data are therefore lacking in MBR. If these variables could be included in medical records, the MBR data would be strengthened. Some characteristics and socio-demographic variables had more missing values than others had, but this was balanced by the large amount of data.

level and housing condition variables and these data are therefore lacking in MBR. If these variables could be included in medical records, the MBR data would be strengthened. Some characteristics and socio-demographic variables had more missing values than others had, but this was balanced by the large amount of data. The NPR includes women with the diagnosis ID. The ID diagnosis includes the classifications: mild, moderate, severe, profound, other and non-specific ID. In this study, information about the etiological factor and exclusion of the psychotic group or the severity in women with ID were not known. In the group of women without ID, all psychiatric diagnoses were excluded to eliminate incorrect adjacent classifications; this could mean this group was healthier than the general population. The results in this study should be compared with similar studies and may not be generalizable to all populations. However, in populations with a similar socio-demographic structure, such as in Sweden, similar results would be expected.

ssifications; this could mean this group was healthier than the general population. The results in this study should be compared with similar studies and may not be generalizable to all populations. However, in populations with a similar socio-demographic structure, such as in Sweden, similar results would be expected. More women with ID (36.6%) did not cohabit with the father at the first antenatal visit compared with women without ID (6.2%). One reason could be more pregnancies in women with ID were unplanned or the pregnancies occurred before cohabitation. Women with ID (54.1%) worked less than women without ID (14.3%). Even so, there is a possibility this proportion was overestimated, as data for one-quarter of the women were missing, possibly indicating lack of work in these women. Unpaid work and daily occupational training based on governmental funding can be registered as ‘work’. Generally, in Sweden, people with ID are less often gainfully employed (27.6%) than the general population (75.1%) (15,16). A medical history of epilepsy was more common in women with ID (5.5%) than in women without ID (0.4%), and a higher prevalence of epilepsy is more associated with ID than in the general population (17,18).

, in Sweden, people with ID are less often gainfully employed (27.6%) than the general population (75.1%) (15,16). A medical history of epilepsy was more common in women with ID (5.5%) than in women without ID (0.4%), and a higher prevalence of epilepsy is more associated with ID than in the general population (17,18). During pregnancy, women with ID were on average four years younger than women without ID were. One reason could be that women with ID did not have the level of education or professional work that often competes with childbearing. Alternatively, they may have wished to become a mother early in life. However, there could be other reasons for not using contraceptives, e.g. low compliance with contraceptive methods, side effects, inconvenience, no steady partner for the moment, and poor economy. In the UK, women with ID have limited knowledge about sexual health issues and approximately half of the women lack basic knowledge about reproduction (19): the prevalence of non-contraceptive use is high (40.8%) (20). More women with ID were obese than women without ID. An explanation may be that women with ID registered later in the pregnancy with antenatal care than women without ID, and had already gained some weight due to the pregnancy. However, the average weight gain between registration with antenatal services and delivery (13.1 kg) indicated a substantial period must have elapsed between the two registered weights. Obesity is a key risk factor for ill-health among women with ID (21,22).

ID, and had already gained some weight due to the pregnancy. However, the average weight gain between registration with antenatal services and delivery (13.1 kg) indicated a substantial period must have elapsed between the two registered weights. Obesity is a key risk factor for ill-health among women with ID (21,22). The overrepresentation of smokers among women with ID could be explained partly by their younger age at childbirth, as younger women in Sweden smoke more than older women do (23). However, it is also possible the women with ID did not receive ‘tailored’ information about smoking, did not understand the smoking information, did not want to stop smoking during pregnancy, or did not receive adequate support to be able to avoid smoking. In Finland, the prevalence of smoking in early pregnancy (25.7%) is higher among young, primiparas, unmarried women with preterm birth (24). As tobacco is the only substance reported in the MBR, the use of alcohol or other drugs among these women could not be assessed. However, it is assumed women with ID live in more deprived settings where alcohol and substance abuse are more prevalent (25).

mong young, primiparas, unmarried women with preterm birth (24). As tobacco is the only substance reported in the MBR, the use of alcohol or other drugs among these women could not be assessed. However, it is assumed women with ID live in more deprived settings where alcohol and substance abuse are more prevalent (25). A higher proportion of women with ID (12.2%) had a gestation length of less than 37 full weeks, compared with 6.1% in the women without ID, and the proportion of preterm deliveries was constant over time. In Australia, the proportion of preterm births among women with ID is 28% (4), although the causes of preterm births are unknown. Reasons for shorter gestation length could be that women with ID have more difficulty in understanding and interpreting signs and symptoms of pregnancy complications such as premature contractions. Other medical reasons include preeclampsia, which is more prevalent during pregnancy among women with ID than among women without ID (OR = 2.85) (3). However, in our data retrieved from MBR a diagnosis of preeclampsia was only reported in 1.1% of women with ID, possibly due to underreporting. A shorter gestational length may pose different problems for women with ID. If labor starts unexpectedly, they may not have received sufficient or appropriate information and are thus unprepared, which can cause more anxiety and stress during delivery, and lead to more CSs being performed on medical as well as humanitarian indications.

onal length may pose different problems for women with ID. If labor starts unexpectedly, they may not have received sufficient or appropriate information and are thus unprepared, which can cause more anxiety and stress during delivery, and lead to more CSs being performed on medical as well as humanitarian indications. During labor, nitrous oxide is an effective and safe analgesic and is the most common pain relief in Sweden (26,27). Totally, women with ID used this method (59.5%) less than women without ID did (75.8%) and the same difference was found in the subgroup of women who had a normal vaginal birth (71.9 vs. 84.1%). This could reflect less need for pain relief or the lack of ability to express the need. Alternatively, the midwife might not have interpreted the women's difficulty in expressing their need and misjudged or underestimated the women's need for pain relief. Another possible explanation is that these women had less knowledge about pain relief during delivery and consequently did not express any wishes. Epidural anesthesia is the most potent analgesia during delivery and the use of epidural analgesia did not differ between the two groups. Severe pain requiring epidural analgesia may be more obvious and easier for women to express and for midwives to understand and handle.

y and consequently did not express any wishes. Epidural anesthesia is the most potent analgesia during delivery and the use of epidural analgesia did not differ between the two groups. Severe pain requiring epidural analgesia may be more obvious and easier for women to express and for midwives to understand and handle. Spontaneous onset of labor was more common in women without ID (81.5%) than in women with ID (74.5%); however, delivery ended more often with a CS in women with ID (24.5%) than in women without ID (17.7%). The indications could be either medical or psychosocial for both elective and acute CS. Severe preeclampsia in primiparas is associated with the child being small for gestational age (28) and this condition may lead to CS. Further, moderate to severe depression, substance abuse, anxiety and stress during pregnancy (2), and specific types of anxiety, e.g. psychosocial stress, family functioning and fear of childbirth, may have associations with CS (29). Another reason for preterm by itself can include presentations other than vertex presentations that exclude vaginal delivery. Vacuum extraction demands more participation from the woman, which may be difficult to obtain from women with ID, or professionals may consider this collaboration is difficult to establish.

29). Another reason for preterm by itself can include presentations other than vertex presentations that exclude vaginal delivery. Vacuum extraction demands more participation from the woman, which may be difficult to obtain from women with ID, or professionals may consider this collaboration is difficult to establish. Women with ID were more often discharged from the maternity ward to place other than home than were women without ID. Children born to women with ID were more frequently admitted to a neonatal intensive care unit (OR = 2.51) (3) and, as gestation length was shorter in women with ID, the children might have needed more neonatal care. Another reason could be that women with ID needed medical care or referral to an institution for planning and assessment of bonding with their babies and training in parenting skills.

sive care unit (OR = 2.51) (3) and, as gestation length was shorter in women with ID, the children might have needed more neonatal care. Another reason could be that women with ID needed medical care or referral to an institution for planning and assessment of bonding with their babies and training in parenting skills. Women with ID are a neglected group in society, yet they still have to elucidate their feelings and needs, as every woman does. They want to be treated with respect and consideration, to have the opportunity to express their wishes and to control their lives from their own point of view. The focus has been on others’ reactions to the pregnancy rather than on the pregnant women themselves (7). To understand the needs and experiences of women with ID in childbearing, it is necessary to recognize them as women instead of medical problems and to acknowledge their human rights (30): they have a right to be heard. This study confirmed women with ID are vulnerable during pregnancy and delivery. Further studies should investigate if their newborn children have a similar increased vulnerability.

ecessary to recognize them as women instead of medical problems and to acknowledge their human rights (30): they have a right to be heard. This study confirmed women with ID are vulnerable during pregnancy and delivery. Further studies should investigate if their newborn children have a similar increased vulnerability. Conclusion Women with ID had more preterm births, used less nitrous oxide as pain relief in labor, and the delivery ended more often with CS. A higher proportion of women with ID smoked at the first antenatal visit. These findings imply that women with ID should be regarded as a risk group. We suggest the knowledge and skills of prepartum and intrapartum caregivers should be improved to enhance more accessible, interactive and better-tailored information, support and care for women with ID. Statistical advice from Katarina Selling and Lars Berglund, Uppsala, Sweden, is gratefully acknowledged. Funding The study was funded by Sävstaholm Foundation, Uppsala University, Uppsala County Council and The Family Planning Fund at Uppsala University Hospital, Sweden. Abbreviations BMIbody mass index CScesarean section ICDInternational Classification of Diseases IDintellectual disability MBRMedical Birth Register NPRNational Patient Register ORodds ratio RRrelative risk

Key Message Newborns of mothers with intellectual disability are a risk group. They have a higher risk of being small for gestational age and of stillbirth and perinatal death compared with newborns of mothers without intellectual disability or any other psychiatric diagnosis. Introduction Few studies have described newborn health in children born to mothers with intellectual disability (ID), and knowledge regarding this topic is limited. According to the World Health Organization, ID is defined as an intelligence quotient below 70 and a reduced adaptive capacity that appears before 18 years of age (1). Previous Australian studies found that children born to mothers with ID were more likely than other children to be born preterm (28%), to have low birthweight (22%) and to be admitted to neonatal intensive care (2,3); however, there was no difference in Apgar scores at one minute between children born to mothers with ID and children born to mothers without ID (3). The incidence of preterm birth among all children born in Sweden (1995–2009) was 5% (4) and the incidence of low birthweight (1998–2007) was 6% (5). The aim of this study was to describe mode of birth, preterm birth rates, Apgar scores, the incidence of being small for gestational age (SGA), stillbirth and overall perinatal death in children born to mothers with ID in comparison to children born to mothers without ID or any other psychiatric diagnosis.

Introduction Few studies have described newborn health in children born to mothers with intellectual disability (ID), and knowledge regarding this topic is limited. According to the World Health Organization, ID is defined as an intelligence quotient below 70 and a reduced adaptive capacity that appears before 18 years of age (1). Previous Australian studies found that children born to mothers with ID were more likely than other children to be born preterm (28%), to have low birthweight (22%) and to be admitted to neonatal intensive care (2,3); however, there was no difference in Apgar scores at one minute between children born to mothers with ID and children born to mothers without ID (3). The incidence of preterm birth among all children born in Sweden (1995–2009) was 5% (4) and the incidence of low birthweight (1998–2007) was 6% (5). The aim of this study was to describe mode of birth, preterm birth rates, Apgar scores, the incidence of being small for gestational age (SGA), stillbirth and overall perinatal death in children born to mothers with ID in comparison to children born to mothers without ID or any other psychiatric diagnosis. Material and methods A total of 326 children born to mothers with ID were compared with 340 624 children born to mothers without ID or any other psychiatric diagnosis between 1999 and 2007 in Sweden. Data were obtained from two of the Swedish National Board of Health and Welfare healthcare registers, namely the National Patient Register (http://www.socialstyrelsen.se/register/halsodataregister/patientregistret/inenglish) and the Medical Birth Register (MBR) (6), which covers 98–99% of all births in Sweden. The variables for comparison were selected on the following grounds: data of good quality (i.e. data with few missing cases) that were found to be important in previous studies (i.e. preterm birth and birthweight; 2,3). The Regional Research Ethics Committee in Uppsala, Sweden, approved this study (327/2007).

n Sweden. The variables for comparison were selected on the following grounds: data of good quality (i.e. data with few missing cases) that were found to be important in previous studies (i.e. preterm birth and birthweight; 2,3). The Regional Research Ethics Committee in Uppsala, Sweden, approved this study (327/2007). A Swedish population-based sample of all women with ID, as defined in International Classification of Diseases (ICD) 8–10, chapter V (ICD-8 codes 311–315, ICD-9 codes 317–319 and ICD-10 codes F70–F79), was identified from the National Patient Register (no women were excluded). The group of women with ID was limited to women with an ID diagnosis and did not include women with syndromes, neuropsychiatric diseases or other psychiatric diagnoses. This sample was linked to the MBR at the National Board of Health and Welfare to identify women with ID who gave birth between 1999 and 2007. A comparison group included all other women who had given birth during the same period, excluding those with any psychiatric diagnosis. Women with ID were younger (mean age 24.2 vs. 28.3 years), with a higher proportion of teenagers (58.6 vs. 22.4%), were more obese (20.1 vs. 8.6%) and smoked more during early pregnancy (27.9 vs. 7.9%) than those without ID. The prevalence of smoking decreased over time in the ID group, from 33.8 (1999–2001) to 33.0 (2002–2004) and then to 19.5% (2005–2007). Other variables included in the analysis were not cohabiting with the child's father at the time of registration (36.6% of mothers with ID vs. 6.2% of those without ID), not working at the time of registration (54.1 vs. 14.3%) and epilepsy (5.5 vs. 0.4%).

8 (1999–2001) to 33.0 (2002–2004) and then to 19.5% (2005–2007). Other variables included in the analysis were not cohabiting with the child's father at the time of registration (36.6% of mothers with ID vs. 6.2% of those without ID), not working at the time of registration (54.1 vs. 14.3%) and epilepsy (5.5 vs. 0.4%). Information on single-born children to mothers with ID and to mothers without ID or any other psychiatric diagnosis was obtained from the MBR. The data set extended from 1999 to 2007 and included 505 children born to mothers with ID and 525 664 children born to mothers without ID. To allow a more stringent comparison, only first-born children (those whose mothers were primiparous) were selected, resulting in 326 children born to mothers with ID and 340 624 children born to mothers without ID. The newborn health outcome pattern was the same for both the total sample (505/525 664 and the selected study sample (326/340 624), except for congenital malformations (p = 0.038), for which the difference between groups did not remain significant when the sample was reduced to first-borns of primiparous mothers. We chose, however, to analyse the more stringent sample to avoid the influence of parity on the main outcome variables. Preterm birth was defined as a birth before 37 completed weeks of gestation. Small for gestational age was defined as a birthweight and/or birth length of at least two standard deviations below the mean for the infant's gestational age. Stillbirth, according to the Swedish Act before 2008, was defined as an infant born after 28 completed weeks of gestation without signs of life at and after birth. Perinatal death was defined, according to the Swedish convention at the time, as death near the time of birth, from 28 weeks of gestation to one week postnatally. Neonatal death was defined as a child dying in the first 28 days of life. Body mass index (BMI; in kilograms per square meter) was based on height and weight. For the non-pregnant population, a BMI ≤ 24.9 kg/m2 was categorized as lean and normal, a BMI of 25.0–29.9 kg/m2 overweight and a BMI ≥ 30 kg/m2 obese.

tnatally. Neonatal death was defined as a child dying in the first 28 days of life. Body mass index (BMI; in kilograms per square meter) was based on height and weight. For the non-pregnant population, a BMI ≤ 24.9 kg/m2 was categorized as lean and normal, a BMI of 25.0–29.9 kg/m2 overweight and a BMI ≥ 30 kg/m2 obese. Statistical analysis Statistical analyses were conducted with the Statistical Package for the Social Sciences (SPSS) 15.0 software program for Windows. Descriptive statistics were used to describe the frequencies of all variables presented. The differences between the case cohort and the control cohort were analysed using the chi-squared test. A significance level of p ≤ 0.05 was chosen, and risk ratios with 95% confidence intervals were estimated (Mantel–Haenszel method). Variables that differed significantly between groups in a univariate analysis were inserted into a binary logistic regression analysis. To reveal associations with Apgar score <7 points at five minutes, prematurity, SGA, stillbirth and perinatal death, we used unadjusted binary logistic regression and binary logistic regression adjusted for the effects of maternal characteristics and birth outcome, such as age, BMI, cohabitation with the child's father, working, smoking, epilepsy and cesarean section (CS). Age was used as a continuous variable.

h and perinatal death, we used unadjusted binary logistic regression and binary logistic regression adjusted for the effects of maternal characteristics and birth outcome, such as age, BMI, cohabitation with the child's father, working, smoking, epilepsy and cesarean section (CS). Age was used as a continuous variable. Results A greater proportion of children born to mothers with ID than children born to mothers without ID were born by CS (24.5 vs. 17.7%; Table 1). Apgar scores <7 at one and five minutes were more prevalent among children born to mothers with ID than those born to mothers without ID. More children of women with ID compared with women without ID were born preterm (12.2 vs. 6.1%) and were SGA (8.4 vs. 3.1%). Stillbirth was almost four times more prevalent among the children born to mothers with ID than among those born to mothers without ID. Stillbirths were proportionally distributed over time as follows: 2.2% in 1999–2001, 0.9% in 2002–2004 and 0.8% in 2005–2007. Perinatal death was more than four times more common among children born to mothers with ID (1.8%) than among those born to mothers without ID (0.4%). Table 1 Newborn health in children born to mothers without intellectual disability (ID) compared with children born to mothers with ID Children of mothers without ID [n = 340 624 (%)] Children of mothers with ID [n = 326 (%)]

Results A greater proportion of children born to mothers with ID than children born to mothers without ID were born by CS (24.5 vs. 17.7%; Table 1). Apgar scores <7 at one and five minutes were more prevalent among children born to mothers with ID than those born to mothers without ID. More children of women with ID compared with women without ID were born preterm (12.2 vs. 6.1%) and were SGA (8.4 vs. 3.1%). Stillbirth was almost four times more prevalent among the children born to mothers with ID than among those born to mothers without ID. Stillbirths were proportionally distributed over time as follows: 2.2% in 1999–2001, 0.9% in 2002–2004 and 0.8% in 2005–2007. Perinatal death was more than four times more common among children born to mothers with ID (1.8%) than among those born to mothers without ID (0.4%). Table 1 Newborn health in children born to mothers without intellectual disability (ID) compared with children born to mothers with ID Children of mothers without ID [n = 340 624 (%)] Children of mothers with ID [n = 326 (%)] Characteristic n Percentage n Percentage Relative risk 95% Confidence interval Missing data (percentage of total data) Live born 339 551 (99.7) 322 (98.8) 1.0 0.9–1.1 0.0 Cesarean section 60 130 (17.7) 80 (24.5) 1.4 1.1–1.7 0.0 Apgar score One minute <7 points 21 634 (6.4) 32 (9.8) 1.5 1.1–2.2 0.7 Five minutes <7 points 5064 (1.5) 12 (3.7) 2.5 1.4–4.4 0.8 Ten minutes <7 points 1967 (0.6) 3 (0.9) 1.6 0.5–4.9 3.7 Born preterm 0.1 Weeks 22–29 1712 (0.5) 5 (1.5) 3.0 1.3–7.3 Weeks 30–36 19 055 (5.6) 35 (10.7) 1.9 1.4–2.7 Weeks 37–42 317 022 (93.1) 284 (87.1) 0.9 0.8–1.1 Weeks 43–45 2483 (0.7) 2 (0.6) 0.8 0.2–3.4 Small for gestational age 10 466 (3.1) 27 (8.4) 2.7 1.9–3.9 0.7 Large for gestational age 7203 (2.1) 7 (2.2) 1.0 0.5–2.1 0.7 Congenital malformation 12 878 (3.8) 18 (5.5) 1.5 0.9–2.3 0.0 Stillbirth 1073 (0.3) 4 (1.2) 3.9 1.5–10.4 0.0 Perinatal death 1455 (0.4) 6 (1.8) 4.3 1.9–9.6 0.0 Neonatal death 525 (0.2) 2 (0.6) 4.0 1.0–16.0 0.0 The multivariate analysis included the outcome variables that differed between the groups in the univariate analysis (Table 2). In the logistic regression analysis, an ID diagnosis was associated with Apgar score <7 at five minutes, preterm birth, SGA, stillbirth and perinatal death (crude odds ratio). When the model was adjusted for effects of maternal characteristics and mode of birth, the ID diagnosis was associated with SGA, stillbirth and perinatal death (adjusted odds ratio). In a subanalysis, we compared smoking mothers with ID vs. non-smoking mothers with ID, and the result did not differ between the groups with respect to Apgar score <7 points at five minutes, prematurity, SGA and perinatal death.

ID diagnosis was associated with SGA, stillbirth and perinatal death (adjusted odds ratio). In a subanalysis, we compared smoking mothers with ID vs. non-smoking mothers with ID, and the result did not differ between the groups with respect to Apgar score <7 points at five minutes, prematurity, SGA and perinatal death. Table 2 Crude odds ratio (95% confidence interval) and adjusted odds ratio (95% confidence interval) for preterm birth, Apgar score <7 points at five minutes, small for gestational age, perinatal death and stillbirth in women with intellectual disability (ID) compared with women without ID Women with ID vs. women without ID Characteristic Crude odds ratio (95% confidence interval) Adjusted odds ratio (95% confidence interval) Preterm birth 2.15 (1.55–3.00) 1.58 (0.99–2.51) Apgar <7 points at five minutes 2.59 (1.45–4.61) 1.83 (0.81–4.15) Small for gestational age 2.86 (1.93–4.24) 2.25 (1.37–3.69) Perinatal death 4.37 (1.95–9.19) 4.25 (1.72–10.50) Stillbirth 3.93 (1.46–10.56) 4.53 (1.65–12.44) Odds ratios were adjusted for maternal age, obesity, cohabitation with the child's father, working, smoking, epilepsy and cesarean section. Discussion The main findings of this study included an increased risk for mothers with ID to have children being SGA, and an overrepresentation of stillbirths and perinatal deaths among children born to mothers with ID. A higher proportion of children of mothers with ID were born preterm and by CS.

Characteristic Crude odds ratio (95% confidence interval) Adjusted odds ratio (95% confidence interval) Preterm birth 2.15 (1.55–3.00) 1.58 (0.99–2.51) Apgar <7 points at five minutes 2.59 (1.45–4.61) 1.83 (0.81–4.15) Small for gestational age 2.86 (1.93–4.24) 2.25 (1.37–3.69) Perinatal death 4.37 (1.95–9.19) 4.25 (1.72–10.50) Stillbirth 3.93 (1.46–10.56) 4.53 (1.65–12.44) Odds ratios were adjusted for maternal age, obesity, cohabitation with the child's father, working, smoking, epilepsy and cesarean section. Discussion The main findings of this study included an increased risk for mothers with ID to have children being SGA, and an overrepresentation of stillbirths and perinatal deaths among children born to mothers with ID. A higher proportion of children of mothers with ID were born preterm and by CS. An Australian study revealed a higher incidence of pre-eclampsia in women with ID (3), and pre-eclampsia is associated with SGA (7). We did not have information about pre-eclampsia among the mothers of children in this study, but it cannot be exluded that this higher incidence of pre-eclampsia also exists in Sweden. Stillbirths and perinatal death among children born to intellectually disabled mothers were four times higher than the proportion among children born to mothers without ID; 1.2 vs. 0.3 and 1.8 vs. 0.4%, respectively. Increasing maternal age, obesity, unemployment and smoking were also associated with stillbirth, peri- and neonatal death, which is supported by several studies (8–15). The relative reduction in smoking over time among mothers with ID did, however, not significantly change the proportion of stillbirths in our study. Decreased fetal movements commonly precede stillbirth, but we still lack proper evidence on how to use fetal movement assessments in clinical practice to prevent stillbirths (16). Mothers with ID may have greater difficulties than mothers without ID in understanding and reacting to decreases in fetal movements.

study. Decreased fetal movements commonly precede stillbirth, but we still lack proper evidence on how to use fetal movement assessments in clinical practice to prevent stillbirths (16). Mothers with ID may have greater difficulties than mothers without ID in understanding and reacting to decreases in fetal movements. A decrease in fetal movements is likewise a well-known risk factor for SGA, and in our study children born to mothers with ID were more often SGA (8.4%) than those born to mothers without an ID (3.1%). We also found an association between SGA and older maternal age, not living with the child's father, unemployment, CS and smoking. Prenatal smoking has previously been reported to increase the risk of SGA (odds ratio 3.08) (17). Maternal smoking at the first antenatal visit has decreased over time in Sweden from 31 (1983) to 7% (2008) (18). This development probably reflects an increased understanding among the general public of the health risks associated with smoking during pregnancy. Despite the decrease in smoking over time, there was an overrepresentation of smoking among mothers with ID compared with mothers without ID. This could partly be explained by their younger mean age at childbirth. However, other reasons (i.e. inadequate information about smoking, difficulty in understanding the information, a desire to keep smoking or inadequate support for smoking cessation) cannot be excluded.

hers with ID compared with mothers without ID. This could partly be explained by their younger mean age at childbirth. However, other reasons (i.e. inadequate information about smoking, difficulty in understanding the information, a desire to keep smoking or inadequate support for smoking cessation) cannot be excluded. In the present study, there were no data about alcohol or substance abuse other than tobacco among the mothers. It is likely that women with ID live in more deprived settings, which could lead to alcohol and/or substance misuse. Moderate to severe predelivery depression, anxiety and stress among mothers with ID were inferred in an Australian study (19), and maternal substance use was a risk factor for SGA (odds ratio 2.40) (20). Such information was not available from the register data that we used, but it can be assumed that if mothers with ID more often use benzodiazepines during pregnancy, the use of such drugs could contribute to the greater incidence of preterm births, SGA and lower Apgar scores in their children compared with the children of mothers without ID. The maternal use of benzodiazepines during pregnancy is believed to increase the risk of preterm delivery (adjusted odds ratio 6.79), low birthweight, low Apgar score, neonatal intensive care, respiratory distress syndrome (21) and CS (22,23). There are conflicting results regarding selective serotonin reuptake inhibitor treatments during pregnancy. A Canadian study affirmed that maternal use of selective serotonin reuptake inhibitor antidepressants is associated with preterm birth, low birthweight and respiratory distress (24). Furthermore, a recent study has documented an increased risk of persistent pulmonary hypertension of the newborn (25). However, a recent review article claims that the suspected increase in risk of preterm birth, low birthweight or SGA has not been confirmed (26).

m birth, low birthweight and respiratory distress (24). Furthermore, a recent study has documented an increased risk of persistent pulmonary hypertension of the newborn (25). However, a recent review article claims that the suspected increase in risk of preterm birth, low birthweight or SGA has not been confirmed (26). Apgar score at one minute was not shown to differ between children born to women with ID compared with those born to women without ID (3). Our study suggested a lower Apgar score at five minutes in women with ID compared with women without ID, but the lower Apgar score was probably due to maternal characteristics and mode of delivery. In the multivariate analysis, after adjusting for these variables we were unable to verify an association between Apgar score <7 at five minutes and ID. A higher proportion (12.2%) of children born to mothers with ID were born preterm, compared with children born to mothers without ID (6.1%). Preterm birth was also associated with maternal obesity, smoking and CS. Prenatal smoking has in previous studies demonstrated an increased risk of preterm delivery (27), as well as extreme prematurity (odds ratio 7.25) (17). The increased preterm birth rate in women with ID could be compared with results from Australian studies with 28% preterm births (2) and an odds ratio of 1.76 (99% confidence interval 0.59–5.28) compared with a reference group of women without ID (3). The increased risk of preterm birth in women with ID is most probably due to CS and/or other maternal characteristics than to their ID diagnosis, because no association was found in the logistic regression analysis after adjusting for these variables.

0.59–5.28) compared with a reference group of women without ID (3). The increased risk of preterm birth in women with ID is most probably due to CS and/or other maternal characteristics than to their ID diagnosis, because no association was found in the logistic regression analysis after adjusting for these variables. The maternal characteristics and the mode of birth in children born to women with ID increase the risk for low Apgar score and preterm birth. Together with an increased risk of low birthweight, it is more likely that there will be complications for the infant and more frequent admissions to a neonatal intensive care unit with such issues as dyspnea, temperature drops, infections, cramps, and difficulty with breastfeeding and bonding with the mother. A recent study from the UK (28) stated that it is more common for children born to mothers with mild to moderate ID to be admitted to a special neonatal care unit and not to be breastfed at discharge.

ssues as dyspnea, temperature drops, infections, cramps, and difficulty with breastfeeding and bonding with the mother. A recent study from the UK (28) stated that it is more common for children born to mothers with mild to moderate ID to be admitted to a special neonatal care unit and not to be breastfed at discharge. In general, antenatal care attendance in Sweden is very high. Roughly 90% of pregnant women have made an initial antenatal visit within the first 12 full pregnancy weeks (6). A strength of the present study is the prospective nature; recall bias was avoided by using data gathered at maternal care units and hospitals. In Sweden, maternal care is free of charge and home deliveries are rare; thus, selection bias is unlikely. One weakness of the MBR is its lack of certain documentation, such as missing records, incorrect documentation, changes in medical records, and incorrect transmission into the MBR (6). It is a shortcoming that medical records do not include data on education level. If this variable were available, it could serve as a proxy for socioeconomic status in the MBR. The clinical significance of Apgar score at one minute is weak, and Apgar score registration at 10 minutes is often missing because midwives commonly do not register the Apgar score at 10 minutes if it has reached 10 points at one and/or five minutes. We therefore concentrated the analysis on Apgar score at five minutes. The place to which the children were discharged (i.e. directly to home or to another place) could have been a useful variable for analysis in this study; however, there were too many missing values for child discharge. Data on discharge are registered in separate medical records, one for the mother and one for the child, and the transfer of data to MBR from these records may occur while the child is still in the neonatal unit. The congenital malformation variable included various deformations and chromosomal abnormalities, but an analysis of the subgroups within this variable was not possible.

one for the mother and one for the child, and the transfer of data to MBR from these records may occur while the child is still in the neonatal unit. The congenital malformation variable included various deformations and chromosomal abnormalities, but an analysis of the subgroups within this variable was not possible. The women with ID were identified in the National Patient Register, which means that they have a registered facility care or out-patient care on some occasion. They are therefore a clearly defined sample, but there may be other mothers with ID not registered in the National Patient Register, and we cannot claim that our sample is representative for all women with ID in Sweden, because it probably consists predominantly of women with mild ID. We had no data about the underlying causes of ID or its degree of severity. It has, however, been reported that women with a mild ID more often have children than women with a moderate or severe ID (29). We have no reasons to believe that this sample represents the most severe cases of ID and that differences between groups are overestimated. We had no data about lifestyle factors (except smoking) and use of medical products or the mothers’ socioeconomic situation, all of which may have an impact on pregnancy outcomes. The results of this study cannot be generalized to other populations, although it is reasonable to believe that the same findings would occur in populations with sociodemographic conditions similar to those of the Swedish population.

thers’ socioeconomic situation, all of which may have an impact on pregnancy outcomes. The results of this study cannot be generalized to other populations, although it is reasonable to believe that the same findings would occur in populations with sociodemographic conditions similar to those of the Swedish population. In conclusion, unborn and newborn children of mothers with ID have a higher likelihood of perinatal death or being SGA. The children are more often born preterm and by CS. These pregnant women and their newborns should therefore be considered a risk group. Furthermore, we believe that there is a need to increase the knowledge and skill of health professionals caring for mothers with ID to enable healthy choices during pregnancy for these women, to enhance their preparedness for birth and their ability to care for the newborn child. The authors acknowledge valuable statistical advice from Katarina Selling (Statisticon, Uppsala) and Lars Berglund (Uppsala Clinical Research Center). Funding The study was funded by Sävstaholm Foundation, Uppsala University, Uppsala County Council and the Family Planning Fund at Uppsala University Hospital. BMIbody mass index CScesarean section ICDInternational Classification of Diseases IDintellectual disability MBRMedical Birth Register SGAsmall for gestational age

Key Message Accurate surveillance of maternal mortality, even in countries with few cases like Sweden, requires better tools. More complete reporting can be achieved by using information available from national registers and death certificates. Introduction Although 99% of maternal deaths occur in low-income countries, healthy young women still die in high-income countries due to complications of pregnancy and childbirth. Maternal mortality decreased in Europe until the 1980s 1–3 but recent data from several European countries have indicated increasing maternal mortality rates 4–7. In part, these changes are due to improved data assessment, but demographic changes such as increased maternal age and migration may be contributing to this rise 4,5,8–10. Underestimation of maternal mortality appears to be substantial, even in high-income countries 11–13, including Sweden 15,16. Inadequate statistics not only impede the possibility of determining trends but also affect inter-country comparisons of maternal mortality rates 11,17. Although the absolute number is small, maternal mortality is tragic, largely because it is often preventable. Accurate surveillance of maternal deaths may therefore lead to changes in patient care.

impede the possibility of determining trends but also affect inter-country comparisons of maternal mortality rates 11,17. Although the absolute number is small, maternal mortality is tragic, largely because it is often preventable. Accurate surveillance of maternal deaths may therefore lead to changes in patient care. The World Health Organization (WHO) defines maternal death in International Classification of Diseases (ICD)-9 and ICD-10 2004 as the death of a woman while pregnant or within 42 days (that is, days 0–41) of termination of pregnancy – irrespective of the duration and site of the pregnancy – from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes. This definition allows the identification of maternal deaths, based on their causes, as either direct or indirect. Direct deaths are those resulting from complications to pregnancy, delivery, and postpartum; from interventions, omissions or incorrect treatment; or from a chain of events resulting from any of the above. Indirect deaths are those resulting from preexisting disease or diseases that developed during pregnancy but were not due to direct obstetric causes, although they may have been aggravated by physiological effects of pregnancy. Deaths due to hemorrhage, puerperal sepsis or complications of anesthesia are examples of direct maternal deaths; deaths from epilepsy or aggravation of an existing cardiac disease are classified as indirect.

t due to direct obstetric causes, although they may have been aggravated by physiological effects of pregnancy. Deaths due to hemorrhage, puerperal sepsis or complications of anesthesia are examples of direct maternal deaths; deaths from epilepsy or aggravation of an existing cardiac disease are classified as indirect. ICD-10 has been in use in Sweden since 1997, preceded from 1987 to 1996 by ICD-9. Pregnancy-related deaths are all those deaths occurring during pregnancy or within 42 days after the end of pregnancy, irrespective of cause of death. This definition was introduced in ICD-10 to facilitate the identification of maternal deaths under circumstances in which cause of death attribution is inadequate. Fortuitous deaths are defined as deaths from unrelated causes which happened to occur during pregnancy or the puerperium. Along with the reporting method of the British Centre for Maternal and Child Enquiries, we prefer to use the term “coincidental” instead of “fortuitous,” since coincidental is a more accurate description and the term fortuitous could imply “fortunate” 19. Suicides are not included in the current ICD definition of maternal death, in contrast to both the Centre for Maternal and Child Enquiries reports and a proposal for classification by the WHO Working Group on Maternal Mortality and Morbidity Classifications 20. Cancers, accidents, and homicides were classified as coincidental deaths, although the British reports classify hormone-dependent cancers as indirect maternal deaths, and homicides as direct, indirect or coincidental, depending on the individual circumstances b19.

on Maternal Mortality and Morbidity Classifications 20. Cancers, accidents, and homicides were classified as coincidental deaths, although the British reports classify hormone-dependent cancers as indirect maternal deaths, and homicides as direct, indirect or coincidental, depending on the individual circumstances b19. The concept of late maternal death was included in ICD-10 in order to capture deaths from pregnancy-related events that occur between 6 weeks and 1 year postpartum 21. A complication occurring during pregnancy can lead to death more than 42 days later, and increasingly available modern life-sustaining procedures and technologies enable more women to survive beyond this period. For the purpose of international reporting of maternal mortality, only those maternal deaths occurring before the end of the 42-day period are included. However, the recording of late deaths is useful for national analytical purposes.

staining procedures and technologies enable more women to survive beyond this period. For the purpose of international reporting of maternal mortality, only those maternal deaths occurring before the end of the 42-day period are included. However, the recording of late deaths is useful for national analytical purposes. Different methods have been used to address the problem of underreporting maternal deaths. In a recent WHO publication, an adjustment factor of 1.5 is applied to account for misclassification of maternal deaths in countries whose civil registration is otherwise characterized as complete, that is, with good attribution of cause of death 22. This adjustment factor is the median of underreporting of maternal deaths in civil registration based on available studies. Early pregnancy deaths, later deaths in the postpartum period, deaths at extremes of maternal age (youngest and oldest), and indirect deaths caused by cerebro- and cardiovascular diseases are the most common cases not reported as instances of maternal death 11,23. With the publication of ICD-10, WHO recommended the inclusion on death certificates of questions regarding current pregnancy and pregnancy within 1 year preceding death 18 but this has not been implemented in Sweden, although shown to be useful 24. The benefits of routine linkage of birth and death registers have been shown in several studies 12,13,16,25. In Denmark, in contrast to the other Nordic countries, the number of maternal deaths reported to WHO is based on linkage between the civil registration system and the national patient register 25. Confidential enquiries, which first began in England and Wales more than 50 years ago 19, have identified more maternal deaths compared with the civil registration systems in several countries 9,11,14,25. In Sweden the maternal mortality group of the Swedish Society of Obstetrics and Gynecology was formed in 2007 and has so far assessed five to seven cases yearly, reported on a voluntary basis from Swedish hospitals, but a routine linkage system of registers is not yet in place.

systems in several countries 9,11,14,25. In Sweden the maternal mortality group of the Swedish Society of Obstetrics and Gynecology was formed in 2007 and has so far assessed five to seven cases yearly, reported on a voluntary basis from Swedish hospitals, but a routine linkage system of registers is not yet in place. Sweden's official statistics on maternal mortality, reported to the WHO by the National Board of Health and Welfare, is based on the underlying cause of death, defined as the disease or injury that initiated the pathological chain resulting in death; or the circumstances surrounding the accident or act of violence that caused a lethal injury. Thus, only deaths with an underlying cause of death identified in ICD-9 chapter XI (630–676) or ICD-10 chapter XV (O00–O95, O98–O99) or A34 (obstetrical tetanus) are reported as maternal deaths. The thought behind this is to capture deaths directly associated with a pregnancy, whereas death from an aggravated preexisting condition will be assigned the same code as the primary disease. For example, when a woman dies from a myocardial infarction during labor, the underlying cause of death will most likely be coded as cardiac disease (ICD chapter IX), and the pregnancy as a contributing cause of death; she will not be reported as a maternal death, although the definition of a maternal death clearly embraces this indirect death. ICD provides coding rules for assigning an underlying cause of death 18, but the knowledge of clinicians on how to complete death certificates has been shown to be deficient 26. The objective of our study was to use the existing information in Swedish national registers and cause of death certificates to determine the number of maternal deaths that in a way would conform to a greater degree with the ICD-9 and -10 definitions of a maternal death than the method currently used. The “new” number acquired was used to recalculate maternal and pregnancy-related mortality ratios from 1988 to 2007.

use of death certificates to determine the number of maternal deaths that in a way would conform to a greater degree with the ICD-9 and -10 definitions of a maternal death than the method currently used. The “new” number acquired was used to recalculate maternal and pregnancy-related mortality ratios from 1988 to 2007. Material and methods We used three Swedish national registers: the Cause of Death Register (CDR), the Medical Birth Register (MBR), and the National Patient Register (NPR). The study population comprised 27 957 recorded deaths of women of reproductive age (defined by the WHO as 15–49 years old) between 1988 and 2007. These women were identified through the CDR, a record that includes all residents, whether that person was a citizen or was present in Sweden at the time of death. However, undocumented migrants and those who died while seeking asylum or visiting Sweden are not included. The register is based on death certificates issued by an attending physician or a physician conducting an autopsy. Underlying and contributory causes of death were coded in the CDR according to ICD-9 from 1988 to 1996, and ICD-10 since 1997. The MBR contains information on women who give birth to a living child, or to a stillbirth with a gestational age ≥28 weeks (since 2008: ≥22 weeks) or a birth weight ≥500 g. The NPR has registered data on hospital discharges since 1964 but it is first considered complete for all public hospitals in Sweden after 1987. Beginning in 2001 the NPR also includes a register of out-patient care. All three records are maintained by the Swedish National Board of Health and Welfare.

th weight ≥500 g. The NPR has registered data on hospital discharges since 1964 but it is first considered complete for all public hospitals in Sweden after 1987. Beginning in 2001 the NPR also includes a register of out-patient care. All three records are maintained by the Swedish National Board of Health and Welfare. Through the CDR we obtained information on deceased women who had an underlying cause of death or a contributory cause of death listed in the ICD chapter “Pregnancy, childbirth and the puerperium” in ICD-9 (630–676) and ICD-10 (O00–O99); obstetric tetanus (A34 in ICD-10); or choriocarcinoma (181 in ICD-9, C58 in ICD-10). By linking CDR to MBR through each resident's personal identification number, we acquired information on women who had given birth within 1 year of their death. When linking CDR to NPR we obtained information on women who had been admitted to hospital or had an out-patient specialist appointment designated with an ICD code that appears in the chapters mentioned above. Using this approach, all women having a diagnosis related to pregnancy in at least one of the three registers within 1 year before they died were identified. Copies of the cause of death certificates of these women were obtained from the National Board of Health and Welfare. The certificates were independently scrutinized by two obstetricians (A.E. and U.H.) and a decision on classifying each as a direct or indirect maternal death, or a coincidental death, was made after discussion. Pregnancy-related mortality was calculated, as well as maternal mortality ratios, both excluding and including suicides, and compared with the 75 cases with an underlying cause of death included in the ICD chapter for Pregnancy, Childbirth, and the Puerperium, that is, maternal deaths reported to WHO by the National Board of Health and Welfare.

y was calculated, as well as maternal mortality ratios, both excluding and including suicides, and compared with the 75 cases with an underlying cause of death included in the ICD chapter for Pregnancy, Childbirth, and the Puerperium, that is, maternal deaths reported to WHO by the National Board of Health and Welfare. SAS version 9.2 and SAS ENTERPRISE GUIDE 4.2 software package (SAS Institute, Cary, NC) were used for the record linkage. Ethics approval for this study was not needed according to Swedish laws on ethical review, since all women were deceased. The Regional Ethics Committee in Uppsala, Sweden, confirmed that the study did not fall into the category of research requiring ethical clearance (2008/381, 14 January 2009). Results In our study population, we identified 491 women in the CDR or NPR with at least one diagnosis from the ICD chapter cited earlier, or a delivery registered in the MBR within 1 year prior to the woman's death. In 164 cases a woman died within 42 days after giving birth or having a pregnancy-related diagnosis in the NPR (Figure 1); 327 women died on day 42 to 364 (Table 1). Figure 1 Maternal deaths in Sweden (excluding and including suicides), 1988–2007, identified through Cause of Death Register, Medical Birth Register, and National Patient Register. Table 1 Underlying causes of death in women who died on day 42 to 364 inclusive after termination of pregnancy in Sweden 1988–2007

Results In our study population, we identified 491 women in the CDR or NPR with at least one diagnosis from the ICD chapter cited earlier, or a delivery registered in the MBR within 1 year prior to the woman's death. In 164 cases a woman died within 42 days after giving birth or having a pregnancy-related diagnosis in the NPR (Figure 1); 327 women died on day 42 to 364 (Table 1). Figure 1 Maternal deaths in Sweden (excluding and including suicides), 1988–2007, identified through Cause of Death Register, Medical Birth Register, and National Patient Register. Table 1 Underlying causes of death in women who died on day 42 to 364 inclusive after termination of pregnancy in Sweden 1988–2007 Cause of death Number of deaths Percentage Malignancy of breast or genital organs 22 6.7 Other malignancy 78 23.9 Disease of the circulatory system 39 11.9 Disease of the respiratory system 11 3.4 Endocrine and metabolic disease 7 2.1 Disease of the nervous system 6 1.8 Sepsis 4 1.2 Other diseases 13 4.0 Pregnancy, childbirth, and the puerperium 2 0.6 Suicide 64 19.6 Traffic accident 31 9.5 Homicidea 15 4.6 Other external causeb 30 9.2 Unclear cause of death 5 1.5 Total 327 100 a Relationship between victim and perpetrator not coded in the International Classification of Diseases. b Other accidents, substance abuse, and deaths from external causes by undetermined event.

Cause of death Number of deaths Percentage Malignancy of breast or genital organs 22 6.7 Other malignancy 78 23.9 Disease of the circulatory system 39 11.9 Disease of the respiratory system 11 3.4 Endocrine and metabolic disease 7 2.1 Disease of the nervous system 6 1.8 Sepsis 4 1.2 Other diseases 13 4.0 Pregnancy, childbirth, and the puerperium 2 0.6 Suicide 64 19.6 Traffic accident 31 9.5 Homicidea 15 4.6 Other external causeb 30 9.2 Unclear cause of death 5 1.5 Total 327 100 a Relationship between victim and perpetrator not coded in the International Classification of Diseases. b Other accidents, substance abuse, and deaths from external causes by undetermined event. Of the 75 women with an underlying cause of death connected to pregnancy, one proved to be a late maternal death and one was assessed as a false register entry (discussed below). One maternal death could not be classified as direct or indirect with the information available. When we reviewed cases in the CDR with a contributory causes of death related to pregnancy, another 21 maternal deaths (five direct and 16 indirect) were identified (Figure 1). In these five direct deaths, pregnancy was not noted as the condition precipitating the chain of events that led to the death recorded on the certificate. However, these deaths were correctly coded according to ICD rules. Our assessment was that a complication of the pregnancy was the underlying cause of death. Through the MBR we found 53 deaths listed in the CDR that occurred within 42 days after delivery that were not registered with a diagnosis related to pregnancy. Of those, 27 were assessed as maternal deaths (five direct and 22 indirect). Finally, through the NPR we identified 10 women who died within 42 days after they had been registered in the NPR with a diagnosis related to pregnancy, but without a birth being registered in the MBR or a pregnancy-related diagnosis in the CDR. Two of these were assessed as maternal deaths, one direct (sepsis) and one indirect (epilepsy). In summary, after step-wise register linkage and review of death certificates, we identified 50 “new” highly probable maternal deaths and 11 suicides during pregnancy or the puerperium. There were 2 060 678 live births in Sweden from 1988 to 2007, which gives a maternal mortality ratio of 3.6 if one uses maternal deaths as defined by the underlying cause of death. When we count all 123 maternal deaths, the ratio rises to 6.0, and if we include suicides, the ratio becomes 6.5. The majority of the women in our study population (n = 100) were born in Sweden (59 “known” and 41 “new” cases) and 23 were born abroad (14 “known” and 9 “new” cases).

ined by the underlying cause of death. When we count all 123 maternal deaths, the ratio rises to 6.0, and if we include suicides, the ratio becomes 6.5. The majority of the women in our study population (n = 100) were born in Sweden (59 “known” and 41 “new” cases) and 23 were born abroad (14 “known” and 9 “new” cases). Among the deaths within 42 days after the termination of pregnancy, three cases were excluded as false entries. According to the registers these women had given birth after their death, and no cause of death certificates could be found. After reviewing register data and death certificates, we also excluded 10 women we believe were miscoded and had not been pregnant shortly before their death. The most common coding mistake in this group was a diagnosis of postpartum deep venous thrombosis given to a person who died from a malignancy, with no registered birth within a year before her death. Thus, a total of 151 pregnancy-related deaths were identified, giving a pregnancy-related mortality ratio of 7.3/100 000 live births.

n coding mistake in this group was a diagnosis of postpartum deep venous thrombosis given to a person who died from a malignancy, with no registered birth within a year before her death. Thus, a total of 151 pregnancy-related deaths were identified, giving a pregnancy-related mortality ratio of 7.3/100 000 live births. Table 1 lists the 327 women who died between day 42 and 364 after delivery or after having a diagnosis related to pregnancy registered in the NPR. Only two women could clearly be classified as late maternal deaths: one peripartum cardiomyopathy and one choriocarcinoma. Both late maternal deaths were identified through the CDR. The cardiomyopathy was coded as the underlying cause of death connected to pregnancy, and thus incorrectly included in the official statistics. In total, four cases of choriocarcinoma were found. Three of those women died after having the disease more than 1 year. No case of obstetric tetanus was found. The ICD-10 diagnoses O96 (death from any obstetric cause occurring more than 42 days but less than 1 year after delivery) and O97 (death from sequelae of direct obstetric causes occurring 1 year or more after delivery) were not assigned to any woman during the survey period.

ear. No case of obstetric tetanus was found. The ICD-10 diagnoses O96 (death from any obstetric cause occurring more than 42 days but less than 1 year after delivery) and O97 (death from sequelae of direct obstetric causes occurring 1 year or more after delivery) were not assigned to any woman during the survey period. Discussion The major finding of this study is the occurrence of 64% more maternal deaths than those identified through underlying cause of death only, giving a maternal mortality ratio of 6.0 instead of 3.6, which is equivalent to the adjustment factor of 1.5 used by WHO. Nearly all “new” maternal deaths (48 of 50) were found by searching among the contributory causes of death in the CDR and by linking CDR to MBR; only two were identified through linkage with the NPR. Among cases included in the official statistics with an underlying cause of death related to pregnancy complications (that could, therefore, primarily be expected to be a direct maternal death) we identified both direct (58/73) and indirect deaths (14/73). Although assumed to be exclusively indirect deaths, we also found both direct (11/50) and indirect (39/50) maternal deaths among the “new” cases. The same proportion of foreign-born women was found among the cases reported to WHO (14/73) as among the “new” cases (9/50), indicating that immigrants were missed to the same extent as Swedish-born women.

ely indirect deaths, we also found both direct (11/50) and indirect (39/50) maternal deaths among the “new” cases. The same proportion of foreign-born women was found among the cases reported to WHO (14/73) as among the “new” cases (9/50), indicating that immigrants were missed to the same extent as Swedish-born women. Among the 29 maternal deaths identified through the MFR and NPR (having no pregnancy diagnosis registered in CDR), we assessed 23 cases as maternal deaths (three direct and 20 indirect), based on the combined information from registers and death certificates – despite the fact that there was no information about pregnancy on the certificates. Most probably the doctor completing the death certificate was either unaware of the recent pregnancy, or failed to recognize the importance of recording associations between a death and a pregnancy. Information about the pregnancy was recorded on the death certificate in six cases (three direct and three indirect maternal deaths) but this was not taken into account as an underlying or a contributory cause of death when the death was coded in the CDR. These cases illustrate substandard registration of the cause of death by the presiding doctor, as well as deficient secondary assessment of the death certificate at official level.

maternal deaths) but this was not taken into account as an underlying or a contributory cause of death when the death was coded in the CDR. These cases illustrate substandard registration of the cause of death by the presiding doctor, as well as deficient secondary assessment of the death certificate at official level. We calculated maternal mortality ratios both excluding and including 11 suicides during pregnancy and within 42 days after termination of pregnancy. The Centre for Maternal and Child Enquiries classifies suicides as indirect maternal deaths 19, in contrast to studies strictly applying the ICD-rules 2,25 that classify suicides as coincidental deaths. In the Netherlands, suicides occurring during pregnancy and the puerperium are classified as a maternal deaths only if a clear correlation with the pregnancy can be found; otherwise they are considered coincidental deaths 27. The WHO Working Group on the Classification of Maternal Deaths and Severe Maternal Morbidities has decided to characterize suicides in pregnancy, as well as deaths from puerperal psychosis and postpartum depression, as direct maternal deaths in the coming WHO classification 20. The prevalence of suicides during gestation and the year following the pregnancy has been shown by several studies to be lower than that of suicides of women who were not pregnant 28,29. This protective effect seems to be stronger during pregnancy than postpartum. However, the risk of suicide has been shown to be higher in the year after a miscarriage or an induced abortion than after a term pregnancy 29. Three recent British reports found suicide more common than previously thought and considered it the leading overall cause of maternal death in the United Kingdom 19. Associations between pregnancy and suicide are deserving of further study in Sweden.

a miscarriage or an induced abortion than after a term pregnancy 29. Three recent British reports found suicide more common than previously thought and considered it the leading overall cause of maternal death in the United Kingdom 19. Associations between pregnancy and suicide are deserving of further study in Sweden. The most common underlying causes of death among women dying on day 42 to 364 after their pregnancy were malignancies, suicides, and vascular diseases, just as in the overall population of women of reproductive age 30. Although late indirect maternal deaths may be among these, this could not be verified with the data available. A pregnancy-related psychiatric disorder was mentioned on the death certificates of five women who committed suicide.

, and vascular diseases, just as in the overall population of women of reproductive age 30. Although late indirect maternal deaths may be among these, this could not be verified with the data available. A pregnancy-related psychiatric disorder was mentioned on the death certificates of five women who committed suicide. By means of record linkage, we have been able to identify deaths that occurred within 1 year of pregnancies recorded in the national registers. Deaths during pregnancy or after a spontaneous abortion in women who did not receive specialist care and, therefore, were not registered in the NPR would not be identifiable unless the pregnancy was noted on the death certificate. A death due to complications after an induced abortion is reported in the CDR, but induced abortions are not registered in the NPR according to Swedish law. Therefore, other possible deaths within a year after induced abortions (irrespective of cause) could not be identified through the record linkage in this study unless the abortion was noted on the death certificate. No direct deaths due to abortion were registered during this study period or from 1980 to 1988 2. It is unlikely that direct deaths in early pregnancy or after spontaneous or induced abortions would be missed. Women without a complete personal identification number (asylum seekers, undocumented migrants, and visitors) are not included in the CDR, and possible maternal deaths in this group were not identified. Maternal deaths at ages below 15 or above 49 are theoretically possible but were not identified since we started with a population of women who died at a reproductive age according to the WHO definition. Basing our classification on register data and death certificates only, an overestimation of maternal deaths through misclassification could not be ruled out. A careful review of individual medical records might reveal a few coincidental deaths among cases we assessed as indirect maternal deaths, depending on the timing of the death and the medical history of the woman. However, the assessment of medical records was beyond the scope of this study.

ion could not be ruled out. A careful review of individual medical records might reveal a few coincidental deaths among cases we assessed as indirect maternal deaths, depending on the timing of the death and the medical history of the woman. However, the assessment of medical records was beyond the scope of this study. Although a maternal death is a rare event in Sweden, surveillance of maternal mortality is a necessity. We need the appropriate tools to detect trends. The analysis of every death by the Swedish Maternal Mortality Group provides a deeper understanding of this unfortunate phenomenon and extracts vital information by scrutinizing the chain of events leading to a maternal death. A Nordic collaboration was formed in 2011 that gives us the opportunity to learn from cases occurring in our five relatively small populations (Denmark, Finland, Iceland, Norway and Sweden) and to return recommendations to health professionals.

nformation by scrutinizing the chain of events leading to a maternal death. A Nordic collaboration was formed in 2011 that gives us the opportunity to learn from cases occurring in our five relatively small populations (Denmark, Finland, Iceland, Norway and Sweden) and to return recommendations to health professionals. The current system for identifying maternal deaths in Sweden according to the ICD definition is inadequate: physicians fail to complete death certificates correctly, information about pregnancy is not always taken into account when certificates are coded, direct maternal deaths are not always coded with an underlying cause of death connected to pregnancy, and indirect maternal deaths are not captured when the underlying cause of death alone is considered. By using information that is already available in national registers and on death certificates, the accuracy of statistical records could be improved substantially. Since maternal death occurs so infrequently, Swedish clinicians may be unaware of the importance of searching for associations between pregnancy and death. The introduction of check boxes on Swedish death certificates for current pregnancy and pregnancy within a year preceding death may increase this awareness. In combination with routine record linkage, it would most probably improve the completeness of maternal death data. International praxis for handling inconsistencies between the ICD definition of maternal death and the ICD rules for coding causes of death also need to be improved to capture maternal deaths better and to increase the comparability of maternal mortality statistics.

ably improve the completeness of maternal death data. International praxis for handling inconsistencies between the ICD definition of maternal death and the ICD rules for coding causes of death also need to be improved to capture maternal deaths better and to increase the comparability of maternal mortality statistics. Funding This work was supported by a grant from the Swedish Council for Working Life and Social Research (FAS 2007–2026) and by the Medical Faculty, Uppsala University. CDRCause of Death Register ICDInternational Classification of Diseases MBRMedical Birth Register NPRNational Patient Register WHOWorld Health Organization

Introduction Follicle flushing has been shown to be ineffective in increasing the number of aspirated oocytes. The last Cochrane analysis concluded that “there is no evidence that follicular aspiration and flushing is associated with … an increase in oocyte yield”. Follicular flushing even seemed to be disadvantageous, as “the operative time is significantly longer and more opiate analgesia is required for pain relief during oocyte retrieval” 1. This statement was based on different studies in which follicle aspiration was performed in polyfollicular in vitro fertilization (IVF) (normal responders) and oligofollicular IVF (low responders) following controlled ovarian hyperstimulation. However, many IVF centers have introduced monofollicular IVF, as this technique promises to have some advantages in certain groups of patients 2. Therefore it remains to be evaluated if this statement also applies to monofollicular IVF.

oligofollicular IVF (low responders) following controlled ovarian hyperstimulation. However, many IVF centers have introduced monofollicular IVF, as this technique promises to have some advantages in certain groups of patients 2. Therefore it remains to be evaluated if this statement also applies to monofollicular IVF. Material and methods In all, 164 aspirations were performed in monofollicular IVF-cycles in 2011, following written consent. Monofollicular IVF was defined as natural cycle IVF (n = 68) (including 5000 units human chorionic gonadotropin, hCG), IVF with low doses of clomiphene citrate (n = 77) (including clomiphene citrate 25 mg/day from day 6 until induction of ovulation plus 5000 units hCG) and IVF with low doses of human menopausal gonadotropin (n = 15) (including 75 units human menopausal gonadotropin from day 6 until induction of ovulation plus 5000 units hCG). hCG was given 36 h before aspiration. Where luteinizing hormone was already increased at the time of follicle monitoring, follicle aspiration was performed without hCG. Only those cycles with one follicle were analyzed.

75 units human menopausal gonadotropin from day 6 until induction of ovulation plus 5000 units hCG). hCG was given 36 h before aspiration. Where luteinizing hormone was already increased at the time of follicle monitoring, follicle aspiration was performed without hCG. Only those cycles with one follicle were analyzed. Follicles were aspirated without anesthesia and analgesia, using gauge 19 single lumen needles (250 mm Hg) as described elsewhere 3. After initial aspiration, follicles were flushed and aspirated three times each with 2 mL flushing medium with heparin (SynVitro® Flush, Origio, Berlin, Germany). Fertilization was achieved by standard intracytoplasmic sperm injection. The oocyte yield, the proportion of mature oocytes, the fertilization rate (pronuclear stage) and the transfer rate were analyzed for each aspiration step. Embryos were transferred two to three days after aspiration. Fisher's exact test was used to analyze the difference in oocyte numbers and transfer rates. Results Average patient age was 37.0 years (SD ± 3.8, range 28–45) and basal levels of follicle stimulating hormone were between 2.0 and 47.4 (average 9.1 ± 8.2). Overall cancellation rate due to premature ovulation was 20/115 (17.4%). Follicle size at the time of aspiration was 19.0 ± 2.1 mm. Follicle aspiration took around 30 sec. Flushing extended the aspiration time by less than 1 min.

asal levels of follicle stimulating hormone were between 2.0 and 47.4 (average 9.1 ± 8.2). Overall cancellation rate due to premature ovulation was 20/115 (17.4%). Follicle size at the time of aspiration was 19.0 ± 2.1 mm. Follicle aspiration took around 30 sec. Flushing extended the aspiration time by less than 1 min. Total oocyte yield/aspiration was 44.5% in the aspirate, 20.7% in the 1st flush, 10.4% in the 2nd flush and 4.3% in the 3rd flush (Table 1). In each aspiration step, the oocyte yield was reduced by around 50%. By flushing, the total oocyte yield increased significantly (p < 0.01) by 80.9%, from 44.5 to 80.5%. Table 1 Total oocyte yield, proportion of mature oocytes and fertilization rate in follicular aspirates and in different flushings in monofollicular in vitro fertilization (IVF)

Total oocyte yield/aspiration was 44.5% in the aspirate, 20.7% in the 1st flush, 10.4% in the 2nd flush and 4.3% in the 3rd flush (Table 1). In each aspiration step, the oocyte yield was reduced by around 50%. By flushing, the total oocyte yield increased significantly (p < 0.01) by 80.9%, from 44.5 to 80.5%. Table 1 Total oocyte yield, proportion of mature oocytes and fertilization rate in follicular aspirates and in different flushings in monofollicular in vitro fertilization (IVF) Monofollicular IVF Aspirations, n 164 Mean age, years ± SD 37.0 ± 3.8 (28–45) Mean follicular diameter, mm ± SD 19.0 ± 2.1 Total oocytes/aspirations, n 132/164 Total oocytes/aspirations, % 80.5 Aspirate 1st flush 2nd flush 3rd flush Oocytes/group/aspirations, n 73/164 34/1645 17/164 7/164 Oocytes/group/aspirations, % 44.5 20.7 10.4 4.3 MII oocytes/aspirated oocytes, n 67/73 31/34 16/17 7/7 MII oocytes/aspirated oocytes, % 91.8 91.2 94.1 100 Fertilized oocytes/MII oocytes, n 33/67 16/31 10/16 4/7 Fertilized oocytes/MII oocytes, % 49.3 51.6 62.5 57.1 Fertilized oocytes/aspirated oocytes, n 33/73 16/34 10/17 4/7 Fertilized oocytes/aspirated oocytes, % 45.2 47.1 58.8 57.1 Transfer rate/group/aspiration, n 33/164 16/164 10/164 4/164 Transfer rate/group/aspiration, % 20.1 9.8 6.1 2.4 Total transfer rate/aspiration, n 63/164 Total transfer rate/aspiration, % 38.4 The proportions of metaphase II oocytes/aspirated oocytes were similar in all four groups (91.8, 91.2, 94.1 and 100%). The proportion of fertilized (pronuclear) oocytes/ metaphase II oocytes were also similar in all four groups (49.3, 51.6, 62.5 and 57.1%) as well as the proportion of fertilized (pronuclear)/aspirated oocytes (45.2, 47.1, 58.8 and 57.1%).

es/aspirated oocytes were similar in all four groups (91.8, 91.2, 94.1 and 100%). The proportion of fertilized (pronuclear) oocytes/ metaphase II oocytes were also similar in all four groups (49.3, 51.6, 62.5 and 57.1%) as well as the proportion of fertilized (pronuclear)/aspirated oocytes (45.2, 47.1, 58.8 and 57.1%). Transfer rates in each group/total aspirations were 20.1, 9.8, 6.1 and 2.4%. By flushing, the total transfer rate increased significantly (p < 0.01) by 91.0%, from 20.1 to 38.4%. As the transfer rate dropped to 2.4% in the last flushing step, we defined a maximum of three flushing steps to be useful. Clinically relevant bleeding, injury to pelvic organs or peritoneal infection was never observed. Discussion The history of IVF shows a progression from the original monofollicular IVF to polyfollicular IVF, sometimes with a very high number of follicles. However, the number of follicles has once again declined in many centers, partly because of a reduction in the stimulation dose and partly because of the increasing average age of the patients. Specialists in reproductive medicine are therefore increasingly being confronted with an oligofollicular reaction. Irrespective of this, monofollicular natural cycle IVF is being increasingly propagated as a treatment method. With the increase in oligo- and monofollicular IVF techniques, a re-evaluation of the aspiration techniques is necessary.

medicine are therefore increasingly being confronted with an oligofollicular reaction. Irrespective of this, monofollicular natural cycle IVF is being increasingly propagated as a treatment method. With the increase in oligo- and monofollicular IVF techniques, a re-evaluation of the aspiration techniques is necessary. In this study, we could show that follicle flushing is beneficial, at least in monofollicular IVF. Further studies will show whether our observations can be extended to oligofollicular IVF. When interpreting our results, it should be considered that there is a certain possibility that the oocytes remain in the aspiration needle during the aspiration and are therefore washed back into the follicle during flushing. The needle and aspiration tubing had a capacity of 0.7 mL in our system. This is equivalent to about 15% of the mean follicle volume of approximately 4 mL, which is aspirated from a follicle size of 20 mm. The probability of the oocytes remaining in the needle and tubing is therefore about 15%. The oocyte retrieval rate from the aspiration may thus be calculated to be 6.7% higher than the 44.5% stated in our study when the needle is removed and completely evacuated after the aspiration. Nevertheless, even with this calculation, washing the follicle is clearly superior to simple aspiration.

herefore about 15%. The oocyte retrieval rate from the aspiration may thus be calculated to be 6.7% higher than the 44.5% stated in our study when the needle is removed and completely evacuated after the aspiration. Nevertheless, even with this calculation, washing the follicle is clearly superior to simple aspiration. In the Cochrane analysis mentioned above 1, it was shown that flushing would lead to a longer duration of anesthesia, which can be disadvantageous for the oocytes as well as for the patient. As aspiration of a single follicle is unproblematic without anesthesia and possible without analgesia 3, this argument does not apply to monofollicular IVF. The three flushings prolonged the aspiration time by less than 1 min in our study. The flushing itself was not painful. Our aspiration technique was stated by 85% of surveyed women to be as or less painful than having a blood sample taken 3. The effect of follicle flushing has also been described in a French publication 4. Mendez Lozano et al. performed an aspiration without flushing in 79 women and with triple flushing in 47 women with a total of 10 mL flushing medium. The percentage of patients with a good embryo was 28.8% in the group without flushing and 37.8% in the group with flushing; however, the difference was not significant. The effect of the individual flushes was not analyzed. These studies basically showed similar results to ours. However, no significance could be assigned and additional detailed data from the individual flushes were not ascertained.

hing and 37.8% in the group with flushing; however, the difference was not significant. The effect of the individual flushes was not analyzed. These studies basically showed similar results to ours. However, no significance could be assigned and additional detailed data from the individual flushes were not ascertained. In conclusion, three flushings almost doubled not only the number of aspirated oocytes but also the transfer rate in monofollicular IVF. These results indicate, first, that the oocyte yield can be increased by flushing and, secondly, that the oocytes, collected by flushing, are as mature and fertilizable as those aspirated without flushing. Funding No special funding. hCGhuman chorionic gonadotropin IVFin vitro fertilization

Key Message Enabling all healthcare providers who are expected to provide skilled birth attendance to provide the signal functions of emergency obstetric care is urgently needed. Introduction At the turn of the century, the international community agreed upon eight Millennium Development Goals (MDG), two of which are specifically related to maternal and newborn health (MDG 4 and 5). The maternal mortality ratio and the proportion of births attended by a skilled birth attendant (SBA) are two indicators that are used to monitor progress towards the attainment of MDG 5 (1). A maternal death is a relatively rare event and trends in maternal mortality over time are difficult to measure. The “proportion of births attended by skilled health personnel” is therefore often used as a more appropriate proxy indicator to track progress towards MDG 5. Skilled attendance at birth requires two key components: an SBA and an enabling environment that includes drugs and equipment, a functional referral system and enabling policies (2–4). Studies have demonstrated a positive correlation between the proportion of deliveries taking place with an SBA and a reduction in maternal deaths (5,6). The aim is to have a skilled attendant present at 90% of all births by 2015 (7). However, the World Health Organization estimates of coverage show that in 2011, only 18% of births in Bangladesh were attended by skilled personnel, in Nepal this was 19%, Pakistan 39% and India 47% (8).

reduction in maternal deaths (5,6). The aim is to have a skilled attendant present at 90% of all births by 2015 (7). However, the World Health Organization estimates of coverage show that in 2011, only 18% of births in Bangladesh were attended by skilled personnel, in Nepal this was 19%, Pakistan 39% and India 47% (8). An SBA is defined as “an accredited health professional such as a midwife, doctor or nurse who has been educated and trained to proficiency in the skills needed to manage normal (uncomplicated) pregnancies, childbirth and the immediate postnatal period, and in the identification, management and referral of complications in women and newborns” (3). Potentially life-threatening complications require immediate recognition and management, which is included in the package of care described internationally as emergency (or essential) obstetric care (EOC). Currently, EOC consists of nine interventions (or signal functions). Seven of these signal functions constitute what is referred to internationally as basic EOC. These include parenteral administration of antibiotics, oxytocics and anticonvulsants, manual removal of a retained placenta, removal of retained products of conception, assisted vaginal delivery (ventouse delivery) and newborn resuscitation using a bag and mask. Two additional signal functions – cesarean section and blood transfusion – constitute comprehensive EOC (9).

cs, oxytocics and anticonvulsants, manual removal of a retained placenta, removal of retained products of conception, assisted vaginal delivery (ventouse delivery) and newborn resuscitation using a bag and mask. Two additional signal functions – cesarean section and blood transfusion – constitute comprehensive EOC (9). Currently, a multitude of different cadres of healthcare workers provide care during pregnancy and childbirth but it is not always clear if these providers are considered to be an SBA according to the existing international definition or which of the signal functions of EOC these healthcare providers can and are expected to perform (9,10). The objectives of this study were to explore which of the existing cadres of healthcare providers are considered SBAs in four countries in Asia, the length of their training, which signal functions of EOC each cadre of provider is expected to perform and whether training and legislation are in place to support this.

). The objectives of this study were to explore which of the existing cadres of healthcare providers are considered SBAs in four countries in Asia, the length of their training, which signal functions of EOC each cadre of provider is expected to perform and whether training and legislation are in place to support this. Material and methods Between November 2011 and March 2012, we carried out a cross-sectional, descriptive study in four South Asian countries: Bangladesh, India, Nepal and Pakistan. The countries were chosen purposively because of their involvement in the Making it Happen programme (11) which aims to build capacity for healthcare providers to provide skilled birth attendance and emergency obstetric and early newborn care. Data were collected using a structured questionnaire in English. Key informants were purposively selected according to their involvement in training, regulation, recruitment and deployment of SBAs in these countries. Key informants included personnel working in the maternal and newborn or reproductive and child health units of ministries of health, key staff of regulatory bodies such as nursing and medical councils or professional associations and academic staff of teaching institutions (medical and nursing schools) involved in training and deployment.

ed personnel working in the maternal and newborn or reproductive and child health units of ministries of health, key staff of regulatory bodies such as nursing and medical councils or professional associations and academic staff of teaching institutions (medical and nursing schools) involved in training and deployment. Using a semi-structured questionnaire and telephone or face-to-face in-depth interviews, we obtained information for each country regarding: the cadres of staff providing maternity services, whether these cadres conducted deliveries, whether they were considered to be SBAs, which of the EOC signal functions they were expected to provide, or were trained and/or legislated to provide. Informed consent was obtained in writing from all participants before completing the questionnaire or answering any study-related questions. Key informants were free to decline participation or withdraw at any time without given reason. All participants were asked to complete the questionnaire by email, which was supplemented by teleconferences or face-to-face interviews in-country. In the case of conflicting answers from key informants in the same country, all key informants were contacted again to clarify diverging information until consensus was reached. In cases where there was no consensus, the majority view is presented in this paper. Data were entered and analyzed using Microsoft Excel 2007.

n the case of conflicting answers from key informants in the same country, all key informants were contacted again to clarify diverging information until consensus was reached. In cases where there was no consensus, the majority view is presented in this paper. Data were entered and analyzed using Microsoft Excel 2007. Results A total of 33 questionnaires were completed. The overall response rate for the initial round of questionnaires sent out by email was 40.7% (33/81): 33.3% in India (7/21) and Nepal (5/15), 35% in Bangladesh (7/20) and 56% in Pakistan (14/25). Of the 33 questionnaires, 40.6% were completed by informants from regulatory bodies or professional associations (Nursing/Medical Boards), 40.6% by academic staff of teaching institutions and 18.8% by informants from Ministries of Health. The profile and distribution on non-respondents was similar to that of respondents.

the 33 questionnaires, 40.6% were completed by informants from regulatory bodies or professional associations (Nursing/Medical Boards), 40.6% by academic staff of teaching institutions and 18.8% by informants from Ministries of Health. The profile and distribution on non-respondents was similar to that of respondents. The cadres of healthcare providers considered to be an SBA in each country are presented in Table 1. The number of cadres ranged from seven in Bangladesh and Nepal, eight in India to nine in Pakistan. In total across these countries, there were 13 cadres of providers considered to be SBA, of which six could be categorized as “medical” and seven as “nursing” cadres. Medical cadres such as obstetrician–gynecologists and medical officers with additional training in obstetrics as well as nursing cadres such as staff nurses and senior staff nurses are considered SBAs in all the countries. Job titles differed between countries and not all cadres were considered an SBA in all countries. The length of training for obstetrician–gynecologists and medical doctors/officers, staff nurses and senior staff nurses is comparable across the four countries. Major differences in length of training exist for three cadres: medical assistants, SBAs working at community level and lady health visitors. In India and Nepal, there is no cadre called community SBA; instead assistant nurse midwives are employed at community level and are trained for 18 months in both countries. Lady health visitors, a recognized cadre in both India and Pakistan, are considered SBA but the length of their training differs. Regulatory bodies (nursing or medical councils) exist for all cadres considered to be SBAs and formal registration is mandatory and in place for these cadres. No regulatory bodies were reported to exist for maternal and child health workers or for traditional birth attendants.

but the length of their training differs. Regulatory bodies (nursing or medical councils) exist for all cadres considered to be SBAs and formal registration is mandatory and in place for these cadres. No regulatory bodies were reported to exist for maternal and child health workers or for traditional birth attendants. Cadres of healthcare providers, whether considered a skilled birth attendant (SBA) or not, and training duration

but the length of their training differs. Regulatory bodies (nursing or medical councils) exist for all cadres considered to be SBAs and formal registration is mandatory and in place for these cadres. No regulatory bodies were reported to exist for maternal and child health workers or for traditional birth attendants. Cadres of healthcare providers, whether considered a skilled birth attendant (SBA) or not, and training duration Cadre Bangladesh India Nepal Pakistan Obstetrician 5 years (MBBS) + 1 year for Diploma (DGO), 3 years MD to 4 years FCPS 5 years (MBBS) + 4 years specialist training 5 years (MBBS) + 3 years 5 years (MBBS) + 4 years specialist training Medical Officer with additional training in O&G 5 years (MBBS) + 6 months or 1 year EOC training 5 years (MBBS) + 16 weeks EMOC or 18 weeks for anesthesia training 5 years (MBBS) + 6 months O&G training 5 years (MBBS) + 1 year Medical Officer without additional training in O&G 5 years (MBBS) 5 years (MBBS) 5 years (MBBS) 5 years (MBBS) MDGP NA NA 5 + 3 years (4–6 months O&G training) NA Ayurvedic Doctor NA 5.5 years1 NA NA Medical Assistant 3 years2 NA 3 years 1 year Sub-Assistant Community Medical Officer 3 years2 NA NA NA Senior Staff Nurse 4 years previously; now 3 years 4 years 4 years 4 years Staff Nurse 3 years 3 years 3 years 3 years Assistant Staff Nurse 3 years2 NA NA NA Midwife No separate midwifery cadre yet; training to start 2012 No separate training, nurses with midwifery training are considered “midwives” (SSN) No separate training, nurses with midwifery training are considered “midwives” (SSN) 18 months Direct entry (community midwife) or via nursing (RNM) Auxiliary Nurse Midwife NA 18 months (from 2012: 2 years) 18 months NA Family Welfare Visitor 18 months (not including Midwifery), some FWVs have additional 6 months SBA training NA NA NA Lady Health Visitor NA ANM training + additional 6 months training NA 18 months Family Health Worker NA NA NA 12 months Maternal and Child Health Worker NA NA 6 months2 NA Community SBA 6 months NA NA 18 months Traditional Birth Attendant None None None None ANM, Auxiliary Nurse Midwife; DGO, Diploma Gynecology & Obstetrics; FCPS, Fellow of College of Physicians and Surgeons; MBBS, Bachelor of Medicine; MDGP, Medical Doctor with General Practice training; O&G, Obstetrics & Gynecology; RNM, Registered Nurse Midwife; SSN, Senior Staff Nurse; NA, cadre not present;

None None ANM, Auxiliary Nurse Midwife; DGO, Diploma Gynecology & Obstetrics; FCPS, Fellow of College of Physicians and Surgeons; MBBS, Bachelor of Medicine; MDGP, Medical Doctor with General Practice training; O&G, Obstetrics & Gynecology; RNM, Registered Nurse Midwife; SSN, Senior Staff Nurse; NA, cadre not present; 1 Considered as SBA only if trained (21-day SBA training). 2 Not conducting deliveries.

None None ANM, Auxiliary Nurse Midwife; DGO, Diploma Gynecology & Obstetrics; FCPS, Fellow of College of Physicians and Surgeons; MBBS, Bachelor of Medicine; MDGP, Medical Doctor with General Practice training; O&G, Obstetrics & Gynecology; RNM, Registered Nurse Midwife; SSN, Senior Staff Nurse; NA, cadre not present; 1 Considered as SBA only if trained (21-day SBA training). 2 Not conducting deliveries. Key informants were asked to clarify which cadres of healthcare providers are expected to perform each of the signal functions of EOC and conformity between performance (P), training (T) and authorization (A) to perform each EOC signal function was examined (Table 2). Across all four countries the healthcare providers reported to be SBAs perform parenteral administration of antibiotics and oxytocics as well as newborn resuscitation with a bag and mask. Anticonvulsants are administered by 80–100% of SBA cadres in Pakistan, India and Nepal. In Bangladesh, anticonvulsants were given by only 43% of cadres expected to function as an SBA. Manual removal of a retained placenta, removal of retained products of conception and assisted vaginal delivery were mainly performed by specialist and general medical doctors; whereas, the majority of nursing cadres did not conduct these procedures. In some settings, healthcare providers are trained but in practice do not perform certain signal functions, such as in Nepal five of seven cadres are trained to perform manual removal of placenta, but only three are authorized to perform this. Conversely, some healthcare providers perform signal functions but this is not included in their training currently, such as in Bangladesh all seven cadres perform newborn resuscitation but this is included in training for only four.

to perform manual removal of placenta, but only three are authorized to perform this. Conversely, some healthcare providers perform signal functions but this is not included in their training currently, such as in Bangladesh all seven cadres perform newborn resuscitation but this is included in training for only four. Table 2 Performance (P), authorization (A) and training (T) of emergency obstetric care signal functions by cadres of healthcare providers considered to be skilled birth attendants (SBA) Figure 1 shows the percentage of healthcare providers considered to be SBAs who perform, are trained and authorized to perform the signal functions of EOC for each country. Figure 1 Proportion of healthcare provider cadres considered to be a skilled birth attendant reported to perform, trained and authorised to perform signal functions of emergency obstetric care in Bangladesh (a), India (b), Nepal (c) and Pakistan (d).

Figure 1 shows the percentage of healthcare providers considered to be SBAs who perform, are trained and authorized to perform the signal functions of EOC for each country. Figure 1 Proportion of healthcare provider cadres considered to be a skilled birth attendant reported to perform, trained and authorised to perform signal functions of emergency obstetric care in Bangladesh (a), India (b), Nepal (c) and Pakistan (d). Discussion Skilled birth attendance is essential to save maternal and newborn lives and key to attaining MDG 4 and 5. Ensuring that there are sufficient competent healthcare providers who can be considered SBAs is therefore crucial. This study mapped out the different cadres of health workers providing maternity services in these four countries in Asia and reports on whether these cadres are expected to provide skilled birth attendance, what they are trained and legislated to do and whether this includes the provision of EOC. This study shows that not all healthcare providers who currently conduct deliveries in South Asia are considered to be SBAs by key informants from ministries of health, regulatory bodies such as nursing and medical councils or professional associations.

d legislated to do and whether this includes the provision of EOC. This study shows that not all healthcare providers who currently conduct deliveries in South Asia are considered to be SBAs by key informants from ministries of health, regulatory bodies such as nursing and medical councils or professional associations. For any healthcare provider who is expected to provide skilled birth attendance, training and legislation needs to be in place (12,13). This includes training and legislation regarding the signal functions of EOC. It is important that it is clear to the healthcare providers themselves, as well as to the staff supervising and managing the healthcare provider, which of the signal functions of EOC can be provided. This study shows that in a number of settings in South Asia this is currently not clear. Our findings are in line with a previous study from sub-Saharan Africa that also reported a lack of clarity and consensus with regard to which signal functions of EOC could be provided by different cadres of SBAs (10). Both this study and the previous one from sub-Saharan Africa report that key signal functions of EOC (manual removal of a retained placenta, removal of retained products of conception and assisted vaginal delivery) are currently mainly provided by medical staff.

f EOC could be provided by different cadres of SBAs (10). Both this study and the previous one from sub-Saharan Africa report that key signal functions of EOC (manual removal of a retained placenta, removal of retained products of conception and assisted vaginal delivery) are currently mainly provided by medical staff. Because of the global scarcity of healthcare providers, healthcare facilities often have no allocation of medical doctors and nursing or midwifery staff are expected to provide the majority of maternity care. It is, therefore, important that existing nursing and midwifery cadres are trained in the immediate recognition and management of obstetric complications and legislated to perform these. “Task shifting” is the term used in the literature to describe the additional allocation of tasks previously only provided by medical staff to healthcare providers who are nurses or midwives. This has been an effective strategy in other settings (14,15). For maternity care, task shifting means that providers who were previously not able to provide components of EOC will be enabled to do so. Providing additional “skills and drills” practical training has proved to be effective in increasing the knowledge and skills of healthcare providers (16). However, training on its own does not guarantee that healthcare providers apply the new knowledge and skills they have acquired. In India, medical officers received additional training in obstetrics. However, it was observed that only those doctors with an interest in obstetrics who were posted to facilities with a sufficiently high caseload were able to put acquired knowledge and skills into practice (17). This illustrates that even if training and legislation are in place, motivation and an enabling environment need to be present (12,13,18).

only those doctors with an interest in obstetrics who were posted to facilities with a sufficiently high caseload were able to put acquired knowledge and skills into practice (17). This illustrates that even if training and legislation are in place, motivation and an enabling environment need to be present (12,13,18). Similarly, research has shown that in some settings, trained and competent healthcare providers are in place but the staff lack the authorization from a professional body to provide components of EOC (2,3,10), constituting a poor use of existing staff (18). In the four South Asian countries included in this study, the different cadres of healthcare providers expected to provide skilled birth attendance were authorized (legislated) to provide most of the interventions they performed and it was rather lack of training that was highlighted as a reason why not all signal functions of EOC were provided. Review of the pre-service curricula of healthcare provider cadres expected to function as SBAs will clarify gaps in existing training (19). Recently the International Confederation of Midwives has provided guidelines to inform pre-service training curricula for cadres of staff providing skilled birth attendance (20). Assessment of healthcare providers’ competence in the work place will also help to determine if additional training is required.

aining (19). Recently the International Confederation of Midwives has provided guidelines to inform pre-service training curricula for cadres of staff providing skilled birth attendance (20). Assessment of healthcare providers’ competence in the work place will also help to determine if additional training is required. Our mapping exercise revealed that obtaining consistent information from a country can be challenging. There is often a lack of clarity and agreement between different institutions and organizations responsible for training, legislation and employment of healthcare providers. Representatives of ministries of health, regulatory bodies and training institutions are often not aware of discrepancies that exist between performance, training and regulation. Therefore, there is a need to establish clear guidelines and disseminate any already existing guidelines describing the role and responsibility of each cadre of staff providing maternity services. This must include dissemination to all involved in monitoring, supervision, training and accreditation of healthcare providers.

efore, there is a need to establish clear guidelines and disseminate any already existing guidelines describing the role and responsibility of each cadre of staff providing maternity services. This must include dissemination to all involved in monitoring, supervision, training and accreditation of healthcare providers. A limitation of this study is that the information was provided by key informants working at the government, senior management and training institutions in each country who may not be aware of the exact reality on the ground. Although consensus among key informants was obtained, information provided was not verified by comparison with training curricula or direct observation of healthcare provider practice. There is a need for further research that would clarify what healthcare providers working in different types of health facilities are able to do. This would include direct observation and assessment of practice to establish which signal functions of EOC are performed (or not) and to identify the various barriers and enablers for this. Conclusion Strategies encouraging delivery to take place with an SBA can only be successful if health workers have the necessary knowledge and skills to provide signal functions of EOC. The results of this study show that currently only a low percentage of healthcare providers who work as SBAs perform these key life-saving interventions in South Asia. We wish to thank the key informants in the four countries who agreed to participate and kindly contributed to the compilation of the data.

Conclusion Strategies encouraging delivery to take place with an SBA can only be successful if health workers have the necessary knowledge and skills to provide signal functions of EOC. The results of this study show that currently only a low percentage of healthcare providers who work as SBAs perform these key life-saving interventions in South Asia. We wish to thank the key informants in the four countries who agreed to participate and kindly contributed to the compilation of the data. Conflict of interest All authors declare they have no conflicts of interest in connection with this article. Funding The study was funded by DFID/UKAid. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. EOCemergency obstetric care MDGMillennium Development Goals SBAskilled birth attendant

Key Message The vast majority of Swedish gamete donors are satisfied with their contribution after donation. The oocyte and sperm donors who expressed ambivalence before the donation were less satisfied 2 months after donation. Introduction Third-party conception would not be possible without a gamete supply from oocyte and semen donors. Previous studies have shown gender distinctions between oocyte and sperm donors 1–5. Women have appeared to be more involved in the process and outcome of the donation than their male counterparts 6. Oocyte donors seem more often found to be motivated by empathy towards the infertile couple, whereas sperm donors are more likely to be curious about their own fertility 4. In addition, results from earlier international studies on oocyte donors have shown that the higher the level of pre-donation altruistic motivation, the higher the post-donation satisfaction 7,8. In contrast, higher levels of pre-donation procedure-related ambivalence among oocyte donors have been associated with lower levels of post-donation satisfaction 8. In addition, women who believed that others would support them were more likely to report previous intention to donate than women who felt they would not be supported 9–13.

er levels of pre-donation procedure-related ambivalence among oocyte donors have been associated with lower levels of post-donation satisfaction 8. In addition, women who believed that others would support them were more likely to report previous intention to donate than women who felt they would not be supported 9–13. Presentation of information in terms of gains and losses can be powerful and could potentially influence an individual's preferences and decision-making process 14. Oocyte donation entails greater personal costs and medical risks than semen donation because the oocyte donors are exposed to possible physical risks as well as emotional and psychological burdens. In addition, the level of donor satisfaction is dependent on multiple factors, including time required, level of personal inconvenience, actual donation experience, and follow-up care 3,8,15,16. It has also been 17 reported that oocyte donors' attitudes towards various clinical scenarios change following donation, reflecting an overall expression of having greater reservations following the donation process. Other studies suggest that oocyte donors might not be aware of or consider in depth the broader implications of being a donor 2 and that counseling increases the donor's insight into the potential impact of gamete donation 10,15,18.

ng an overall expression of having greater reservations following the donation process. Other studies suggest that oocyte donors might not be aware of or consider in depth the broader implications of being a donor 2 and that counseling increases the donor's insight into the potential impact of gamete donation 10,15,18. Anecdotal reports had suggested that those who choose to be donor candidates had experienced transient depression or anxiety to a greater degree than those who did not want to be donors 19 and had negative perceptions about the techniques involved 10,20. However, previous research has shown that oocyte donors tend to be positive about their donation experience, are psychologically well adjusted, that levels of satisfaction are high and that many would donate again 7,12,20–26. There is clearly inadequate empirical information available to understand donors' emotional and psychological adjustment after donation. In a follow-up study, it has been reported that there was a lack of flexibility regarding anonymity and information about the outcome of donations 21. The author stated that improved donor satisfaction was likely to improve future donor recruitment and retention. To improve the readiness of the general population to act as gamete donors and given the relative importance and long-term impact of oocyte donation, more research on the psychosocial consequences for donors is needed.

stated that improved donor satisfaction was likely to improve future donor recruitment and retention. To improve the readiness of the general population to act as gamete donors and given the relative importance and long-term impact of oocyte donation, more research on the psychosocial consequences for donors is needed. Our aim for this study was twofold: first, to study differences in ambivalence before donation and symptoms of anxiety and depression in relation to satisfaction after donation between oocyte and sperm donors', and second, to study experiences of donor treatment and perception of donation. Material and methods The Swedish multicenter study on gamete donation is a prospective longitudinal study of donors and recipients of donated gametes, including two comparison groups, carried out by collecting data from all fertility clinics performing gamete donation in Sweden, i.e. at the university hospitals in Stockholm, Gothenburg, Uppsala, Umeå, Linkoping, Örebro, and Malmö. During 2005–2008 gamete donors were approached regarding participation.

onated gametes, including two comparison groups, carried out by collecting data from all fertility clinics performing gamete donation in Sweden, i.e. at the university hospitals in Stockholm, Gothenburg, Uppsala, Umeå, Linkoping, Örebro, and Malmö. During 2005–2008 gamete donors were approached regarding participation. Donors were recruited through advertisements in local newspapers, blood donation programs, family/friends, and former oocyte donors. Potential donors spoke with the donor nurse coordinator on the telephone, who gave an overview of the principles of donation including the medical procedure and the social consequences. At the clinic visit the potential donors completed a medical questionnaire to provide information about surgical, medical, gynecologic/obstetric, and social history. The coordinator then conducted the assessment of the donor motivation and qualifications for admittance to the program. Potential donors received written information about the donation process and then had an interview with the medical doctor who reviewed medication, time commitment, legal and ethical issues. Donors were queried as to whether they would or would not participate in the donor program. Donor candidates were then assessed and evaluated by the clinic psychologists. If approved, potential donors were registered.

interview with the medical doctor who reviewed medication, time commitment, legal and ethical issues. Donors were queried as to whether they would or would not participate in the donor program. Donor candidates were then assessed and evaluated by the clinic psychologists. If approved, potential donors were registered. All women and men accepted as donors of oocytes/sperm were approached and asked if they would be willing to participate in the study. The only exclusion was persons who did not speak and/or read Swedish. Participants completed two questionnaires: on acceptance and 2 months after donation. Participation was rewarded with gift vouchers (worth approximately 12€). Of 251 eligible oocyte donors, 41 withdrew from the treatment procedure and 29 declined participation, whereas 181 (86%) agreed and completed the first questionnaire. Of these, 165 (91%) also completed the second questionnaire (Figure 1). Of 173 eligible sperm donors, 54 withdrew from the medical procedure. Of 156 eligible sperm donors approached, 119 (76%) accepted and completed the first questionnaire (Figure 1). Eighty-nine (76%) completed the second questionnaire. Demographic data collected included age, highest stage of education, civil state, number of children, type of donation (anonymous/known to the recipient couples), and number of donations/treatments. Figure 1 Flow diagram of participation of eligible oocyte donors.

Of 251 eligible oocyte donors, 41 withdrew from the treatment procedure and 29 declined participation, whereas 181 (86%) agreed and completed the first questionnaire. Of these, 165 (91%) also completed the second questionnaire (Figure 1). Of 173 eligible sperm donors, 54 withdrew from the medical procedure. Of 156 eligible sperm donors approached, 119 (76%) accepted and completed the first questionnaire (Figure 1). Eighty-nine (76%) completed the second questionnaire. Demographic data collected included age, highest stage of education, civil state, number of children, type of donation (anonymous/known to the recipient couples), and number of donations/treatments. Figure 1 Flow diagram of participation of eligible oocyte donors. Ambivalence was measured with a modified Swedish version of an adapted Donor Ambivalence Scale for oocyte donors by Klock et al. 8. Permission was received from the author for using the scale in this study. The scale consists of seven multiple-choice items that measure mixed feelings about the donation. Responses are combined into a summary score between 0 and 7, with higher scores indicating greater ambivalence. In this study we considered a score ≥4 to indicate high ambivalence. The Hospital Anxiety and Depression Scale (before and after donation), which measures anxiety and depressive symptoms, comprises 14 items (seven anxiety and seven depressive). Each scale was dichotomized into two levels, no depressive/anxiety symptoms (scores <8) and depressive/anxiety symptoms present (scores ≥8) 27.

Ambivalence was measured with a modified Swedish version of an adapted Donor Ambivalence Scale for oocyte donors by Klock et al. 8. Permission was received from the author for using the scale in this study. The scale consists of seven multiple-choice items that measure mixed feelings about the donation. Responses are combined into a summary score between 0 and 7, with higher scores indicating greater ambivalence. In this study we considered a score ≥4 to indicate high ambivalence. The Hospital Anxiety and Depression Scale (before and after donation), which measures anxiety and depressive symptoms, comprises 14 items (seven anxiety and seven depressive). Each scale was dichotomized into two levels, no depressive/anxiety symptoms (scores <8) and depressive/anxiety symptoms present (scores ≥8) 27. Experience of donor treatment after donation was assessed using five study-specific questions about the care. The five response alternatives were collapsed into two categories “Not good at all/Not enough” to “Very good/Enough”. Satisfaction with donation was measured after the donation with seven items developed on the basis of earlier research (3,8,20,25). In addition one single item assesses overall experience. The items were translated to Swedish, tested and found to be accurate in a pilot study. Each question was dichotomized into satisfied/not satisfied and a summary score computed. A score >6 was considered to mean individual satisfaction.

lier research (3,8,20,25). In addition one single item assesses overall experience. The items were translated to Swedish, tested and found to be accurate in a pilot study. Each question was dichotomized into satisfied/not satisfied and a summary score computed. A score >6 was considered to mean individual satisfaction. Perceptions of the donation were assessed using four study-specific questions. The response categories were formulated to indicate levels of agreement (“agree”, “neutral”, “disagree”, and “cannot form an opinion”). The outcome of the donation, one single item was formulated on donation outcome: “Do you know the pregnancy result of the donation treatment?” The three response alternatives were “I do not know”, “the recipient was pregnant”, and “the recipient was not pregnant”. Statistical analysis This was performed using IBM SPSS v.17 (Armink, NY, USA). Chi-squared testing was used to detect bivariate differences on the measurements of anxiety, depressive symptoms, experience, ambivalence, perception, and satisfaction in comparing oocyte and sperm donors. Anxiety and depressive symptoms were both dichotomized into whether anxiety/depressive symptoms were present or not, using the cut-off value of 8 (≥8 indicating anxiety/depressive symptoms).

ents of anxiety, depressive symptoms, experience, ambivalence, perception, and satisfaction in comparing oocyte and sperm donors. Anxiety and depressive symptoms were both dichotomized into whether anxiety/depressive symptoms were present or not, using the cut-off value of 8 (≥8 indicating anxiety/depressive symptoms). Ambivalence was similarly defined into two levels indicating ambivalence or no ambivalence, as was perception and satisfaction. Binary logistic regression was used to evaluate the outcome satisfaction using ambivalence, anxiety, and depressive symptoms before and after treatment as predictors in three separate models. All models included background data such as age, education, and previous biological children. The study was approved by the Regional Ethics Review Board in Linköping (dnr M129-05-050223, T113-07 080122). Results Sociodemographic data are presented in Table 1. The mean age was 30.4 years (standard deviation 4.5) for oocyte donors and 33.9 years (standard deviation 7.6) for sperm donors. Twenty-three (14%) oocyte donors and five (3%) sperm donors were so-called known donors, meaning that the recipient couple and the donors were known to each other. Of the donors, 156 (86%) knew persons that had infertility experience (data not shown). Table 1 Sociodemographics of 165 oocyte donors and 89 sperm donors

Results Sociodemographic data are presented in Table 1. The mean age was 30.4 years (standard deviation 4.5) for oocyte donors and 33.9 years (standard deviation 7.6) for sperm donors. Twenty-three (14%) oocyte donors and five (3%) sperm donors were so-called known donors, meaning that the recipient couple and the donors were known to each other. Of the donors, 156 (86%) knew persons that had infertility experience (data not shown). Table 1 Sociodemographics of 165 oocyte donors and 89 sperm donors Oocyte donors Sperm donors n % n % pValue Age < 0.001 Mean, standard deviation 30.39 4.53 33.91 7.60 0.753 n≤30 years 69 41.8 35 39.8 n>30 years 96 58.2 53 60.2 Marital status Single 39 23.8 35 38.9 0.003 In a relationship 18 11.0 16 17.8 Cohabitation/Married 107 65.2 38 42.2 Education Elementary school 7 4.3 0 0.0 0.002 High school 81 49.4 29 32.2 University 76 46.3 61 67.8 Biological children No 110 67.5 31 35.2 < 0.001 Yes 53 32.5 57 64.8 Donor known to the recipient couple 0.001 No 141 86.0 114 96.7 0.007 Yes 23 14.0 5 3.3 The Hospital Anxiety and Depression Scale assessment of the oocyte donors before and 2 months after donation showed that 13/161 (8%) had anxiety before donation and 24/163 (15%) had anxiety after donation. For depressive symptoms there were 3/161 (2%) and 7/193 (4%) scoring above the cut-off level, respectively. Very few sperm donors showed depressive symptoms after the donation and no differences could be seen between oocyte and sperm donors (Table 2). The donors' anxiety and depressive levels did not show any differences in relation to negative or positive perceptions of satisfaction after the donation. The same results were found regarding satisfaction in relation to anxiety and depression levels after donation (data not shown).

n oocyte and sperm donors (Table 2). The donors' anxiety and depressive levels did not show any differences in relation to negative or positive perceptions of satisfaction after the donation. The same results were found regarding satisfaction in relation to anxiety and depression levels after donation (data not shown). Table 2 Oocyte donors' (n = 165) and sperm donors' (n = 89) symptoms of anxiety and depression before and after donation Oocyte donors Sperm donors n % n % pValue Anxiety symptoms before donation No 148 91.9 83 93.3 0.703 Yes 13 8.1 6 6.7 Anxiety symptoms after donation No 139 85.3 81 92.0 0.120 Yes 24 14.7 7 8.0 Depression symptoms before donation No 158 98.1 88 98.9 0.655 Yes 3 1.9 1 1.1 Depression symptoms after donation No 156 95.7 87 98.9 0.174 Yes 7 4.3 1 1.1 Some donors thought that they did not receive sufficient information about practical issues and future consequences, but almost all (98%) were satisfied with their overall experience compared with oocyte donors (86%, p < 0.003) (Table 3). Table 3 Oocyte donors' (n=165) and sperm donors'(n=89) experience of donor treatment

Oocyte donors Sperm donors n % n % pValue Anxiety symptoms before donation No 148 91.9 83 93.3 0.703 Yes 13 8.1 6 6.7 Anxiety symptoms after donation No 139 85.3 81 92.0 0.120 Yes 24 14.7 7 8.0 Depression symptoms before donation No 158 98.1 88 98.9 0.655 Yes 3 1.9 1 1.1 Depression symptoms after donation No 156 95.7 87 98.9 0.174 Yes 7 4.3 1 1.1 Some donors thought that they did not receive sufficient information about practical issues and future consequences, but almost all (98%) were satisfied with their overall experience compared with oocyte donors (86%, p < 0.003) (Table 3). Table 3 Oocyte donors' (n=165) and sperm donors'(n=89) experience of donor treatment Oocyte donors Sperm donors n % n % pValue How did you experience during the first contact with the clinic at the time when you wanted to donate? Bad 14 8.5 7 7.9 0.853 Good 150 91.5 82 82.1 What was your experience of meeting the staff at the clinic before the donation? Bad 5 3.0 8 9.0 0.041 Good 159 97.0 81 91.0 Did you get enough information about practical issues regarding donation? Not enough 15 9.1 12 13.5 0.286 Enough 149 90.9 77 86.5 Did you get enough information about future consequences regarding the donation? Not enough 21 12.8, 3 3.4 0.014 Enough 143 87.2 86 96.6 How do you view your overall experience of the donation? Bad 23 14.1 2 2.2 0.003 Good 140 85.9 87 97.8 Twenty-five (15%) oocyte donors reported a bad experience with being given hormones and 30 (18%) considered the oocyte retrieval painful. In general, most oocyte donors had a positive experience of donation (n = 130, 79%).

o you view your overall experience of the donation? Bad 23 14.1 2 2.2 0.003 Good 140 85.9 87 97.8 Twenty-five (15%) oocyte donors reported a bad experience with being given hormones and 30 (18%) considered the oocyte retrieval painful. In general, most oocyte donors had a positive experience of donation (n = 130, 79%). The donation had a positive impact on the lives of the donors (Table 4). There were no measured satisfaction differences between oocyte and sperm donors except that a higher percentage of oocyte donors regarded donation as a major event in their lives. A majority of donors (58%) did not know about the outcome of their donation; 93% of sperm donors vs. 44% of oocyte donors. Table 4 Oocyte donors' (n = 165) and sperm donors' (n = 89) satisfaction with the donation

The donation had a positive impact on the lives of the donors (Table 4). There were no measured satisfaction differences between oocyte and sperm donors except that a higher percentage of oocyte donors regarded donation as a major event in their lives. A majority of donors (58%) did not know about the outcome of their donation; 93% of sperm donors vs. 44% of oocyte donors. Table 4 Oocyte donors' (n = 165) and sperm donors' (n = 89) satisfaction with the donation Oocyte donors Sperm donors n % n % pValue I am happy to help couples unable to have children by other means Agree 162 99.4 88 97.8 0.366 Neutral 0 0.0 1 1.1 Disagree 1 0.6 1 1.1 I feel as though I have made a contribution to my fellow human beings Agree 160 97.6 82 92.1 0.109 Neutral 3 1.8 4 4,5 Disagree 1 0.6 3 3.4 My life is more content Agree 79 48.2 49 55.7 0.473 Neutral 69 42.1 33 37.5 Disagree 16 9.8 6 6.8 I feel that I gave something away without receiving anything back Agree 12 7.5 6 6.7 0.105 Neutral 14 8.8 16 18.0 Disagree 133 83.6 67 75.3 This is the highlight (a major event) in my life Agree 60 37.0 25 27.8 0.021 Neutral 64 39.5 29 32.2 Disagree 38 23.5 36 40.0 I think I will brood about it for the rest of my life Agree 5 3.1 4 4.6 0.835 Neutral 13 8.1 7 8.0 Disagree 143 88.8 76 87.4 The perspectives of the oocyte donors as reported after donating differed from those of sperm donors (Table 5). Oocyte donors had a feeling of more support from family and friends compared with sperm donors. Despite this, a larger proportion of oocyte donors reported the donation to be completed.

agree 143 88.8 76 87.4 The perspectives of the oocyte donors as reported after donating differed from those of sperm donors (Table 5). Oocyte donors had a feeling of more support from family and friends compared with sperm donors. Despite this, a larger proportion of oocyte donors reported the donation to be completed. Table 5 Oocyte donors' (n = 165) and sperm donors' (n = 89) perceptions post-donation Oocyte donors Sperm donors n % n % pValue I am concerned over my future fertility Agree 13 7.9 5 5.6 0.796 Neutral 14 8.5 8 9.0 Disagree 138 83.6 76 85.4 I feel that my family and friends are proud of my donor contribution Agree 117 74.1 17 23.3 <0.001 Neutral 25 15.8 28 38.4 Disagree 16 10.1 28 38.4 It is hard for family and friends to understand all the aspects of my donation Agree 22 14.3 10 15.6 0.002 Neutral 25 16.2 24 37.5 Disagree 107 69.5 30 46.9 The donation is for me totally completed/finished after the donation procedure Agree 73 46.5 19 22.4 0.001 Neutral 23 14.6 21 24.7 Disagree 61 38.9 45 52.9 Responders aged 30 or more were approximately 60% less satisfied than those aged below 30 (Table 6). After adjustments of sociodemographic background variables, those who did not indicate ambivalence were almost five times more satisfied compared with those who did indicate ambivalence. Table 6 Logistic regressions predicting satisfaction for 165 oocyte donors and 89 sperm donors

Oocyte donors Sperm donors n % n % pValue I am concerned over my future fertility Agree 13 7.9 5 5.6 0.796 Neutral 14 8.5 8 9.0 Disagree 138 83.6 76 85.4 I feel that my family and friends are proud of my donor contribution Agree 117 74.1 17 23.3 <0.001 Neutral 25 15.8 28 38.4 Disagree 16 10.1 28 38.4 It is hard for family and friends to understand all the aspects of my donation Agree 22 14.3 10 15.6 0.002 Neutral 25 16.2 24 37.5 Disagree 107 69.5 30 46.9 The donation is for me totally completed/finished after the donation procedure Agree 73 46.5 19 22.4 0.001 Neutral 23 14.6 21 24.7 Disagree 61 38.9 45 52.9 Responders aged 30 or more were approximately 60% less satisfied than those aged below 30 (Table 6). After adjustments of sociodemographic background variables, those who did not indicate ambivalence were almost five times more satisfied compared with those who did indicate ambivalence. Table 6 Logistic regressions predicting satisfaction for 165 oocyte donors and 89 sperm donors Model 1 pre-anxiety/-depression Model 2 post-anxiety/-depression Model 3 ambivalence OR 95% CI OR 95% CI OR 95% CI Sperm donor Reference level Reference level Reference level Oocyte donor 1.18 0.52–2.69 1.27 0.56–2.91 1.13 0.45–2.81 University Reference level Reference level Reference level Elementary school 1.34 0.13–13.87 0.98 0.10–10.11 0.85 0.08–8.66 High school 1.36 0.65–2.85 1.34 0.64–2.78 1.17 0.54–2.55 Age, >30 years Reference level Reference level Reference level Age, ≤ 30 years 0.43 0.18–1.01 0.41 0.18–0.97 0.46 0.19–1.12 Single Reference level Reference level Reference level In a relationship 2.53 0.68–9.45 2.10 0.58–7.65 2.72 0.71–10.46 Cohabitation, married 2.74 0.97–7.72 2.36 0.87–6.37 2.43 0.85–6.98 Biological children, yes Reference level Reference level Reference level Biological children, no 1.27 0.49–3.32 1.24 0.48–3.22 1.17 0.43–3.20 Pre-anxiety symptoms, yes Reference level – – – – Pre-anxiety symptoms, no 1.09 0.22–5.26 – – – – Pre-depression symptoms, yes Reference level – – – – Pre-depression symptoms, no – – – – – – Post-anxiety symptoms, yes – – Reference level – – Post-anxiety symptoms, no – – 1.95 0.62–6.11 – – Post-depression symptoms, yes – – Reference level – – Post-depression symptoms, no – – – – – – Ambivalence, yes – – – – Reference level Ambivalence, no – – – – 4.71 1.34–16.51 Discussion In this national study of identifiable donors the vast majority were satisfied with their contribution after the donation. Those oocyte and sperm donors who expressed ambivalence before the donation reported less satisfaction after donation. This is in line with a previous study 8 showing a negative correlation between pre-donation ambivalence and post-donation satisfaction. In a recent Swedish study, sperm donors (39%) were found to be more ambivalent compared with oocyte donors (21%) before donation using the same ambivalence scale as the present study 4. In contrast, we found no gender differences in donor satisfaction post-donation in the present study.

and post-donation satisfaction. In a recent Swedish study, sperm donors (39%) were found to be more ambivalent compared with oocyte donors (21%) before donation using the same ambivalence scale as the present study 4. In contrast, we found no gender differences in donor satisfaction post-donation in the present study. Both oocyte and sperm donors were rather well educated. The sperm donors were more likely to have a university degree compared with oocyte donors. The demographic variables, emotional stress pre-donation and post-donation were not associated with the gamete donors' level of satisfaction with the donation. The level of anxiety and depressive symptoms of the oocyte and sperm donors were stable over time in the present study. This is in line with earlier studies of oocyte donors' characteristics 3,28. There were positive attitudes to the overall experience of the donation. These results are in line with earlier studies reported in two reviews 12,26. Our previous studies on personality characteristics among both oocyte and sperm donors showed that they are autonomous, stable, and well adjusted 28,29. The results indicated that oocyte and sperm donors in general felt less worried, and suffered less from uncertainty and shyness. Purewal and van den Akker 30 have found that lower scores of perceived behavioral control in gamete donors predicted more willingness to donate. High levels of perceived behavior control predicted possible donation group and low levels of perceived behavior control predicted non-donation group in Purewal and van den Akker 30

l and van den Akker 30 have found that lower scores of perceived behavioral control in gamete donors predicted more willingness to donate. High levels of perceived behavior control predicted possible donation group and low levels of perceived behavior control predicted non-donation group in Purewal and van den Akker 30 The majority of the donors in the present study did not know the outcome of the donation; however, there were gender differences between the donor groups. Very few sperm donors knew the outcome of the donation. Kalfoglou and colleagues 18,21 found that 76% of anonymous oocyte donors were not told about the outcome but 75% of those donors wanted to know the outcome. Also, in two other studies 3,31 from the USA, the corresponding figures of the wish to know were the same. Our study of potential oocyte donors from the general population in Sweden 32 found that almost 50% said that they would not want information about the outcome of the donation. The level of interest of knowing the outcome of the donation could also reflect the motivation and satisfaction. It may be hard for family and friends to fully understand and give support to donors. In our study the donation seems to be acknowledged more readily by the families and friends of oocyte donors than of semen donors, perhaps because for oocyte donors the “medical” intervention involved is less embarrassing to discuss with family and friends than the very much more private “masturbation” involved in sperm donation.

y the donation seems to be acknowledged more readily by the families and friends of oocyte donors than of semen donors, perhaps because for oocyte donors the “medical” intervention involved is less embarrassing to discuss with family and friends than the very much more private “masturbation” involved in sperm donation. The seven clinics' practices concerning how the information is provided and choosing the potential donors are not identical but do follow the recommendations from the National Board of Health and Welfare. Policies concerning information on the results of the donation, i.e. pregnancies or number of children born, may differ in some respects between clinics. When a program views a donor as a patient, rather than simply as a donor, the attitudes and experiences of each as a donor becomes a necessary component of care. This is probably more pronounced among oocyte donors because they go through medical interventions that place more demands on the oocyte donors than on sperm donors. It may seem simpler for sperm donors to make their donations, but even sperm donation is not a single event. The donors have to go through medical psychological screening, blood tests, and repeated semen donation. The time commitment and sometimes also the absence of financial incentives can make it very difficult to recruit gamete donors, although this is not always reported 1.

but even sperm donation is not a single event. The donors have to go through medical psychological screening, blood tests, and repeated semen donation. The time commitment and sometimes also the absence of financial incentives can make it very difficult to recruit gamete donors, although this is not always reported 1. The main strength of the present study is the large sample size and that it is a prospective study including all fertility clinics performing gamete donation in Sweden. Distinct inclusion criteria and high response rates contribute to the external validity. However, no information is available about the donors who were not accepted or who chose not to participate in the present study, and it is possible that they have a different view of the questions asked. In conclusion, the vast majority of donors expressed 2 months after donation that they were satisfied with their contribution. Oocyte and sperm donors who expressed ambivalence before the donation were less satisfied after the donation. There is a need for a longer follow-up to catch donor reflections about any offspring which might have future implications for them. We thank the statistician Marie Bladh. Funding Merck Serono provided support during the implementation of the study. Grants were received from Medical Research Council of Southeast Sweden (FORSS), Medical Research Council (RFR) of Uppsala and Örebro, Medical Faculty at Uppsala University and Grants for Caring Sciences at Uppsala University.

We thank the statistician Marie Bladh. Funding Merck Serono provided support during the implementation of the study. Grants were received from Medical Research Council of Southeast Sweden (FORSS), Medical Research Council (RFR) of Uppsala and Örebro, Medical Faculty at Uppsala University and Grants for Caring Sciences at Uppsala University. Conflict of interest The authors have stated explicitly that there are no conflicts of interest in connection with this article.

Abbreviations ACRalbumin to creatinine ratio BMIbody mass index BPblood pressure CIconfidence interval MAPmean arterial blood pressure ORodds ratio RRrisk ratio Key Message Albumin to creatinine ratio is an independent prognostic factor for maternal and neonatal adverse outcomes in suspected preeclampsia, though the prognostic value appears larger for maternal outcomes. Therefore albumin to creatinine ratio could play an important role in healthcare research and clinical practice in the future. Introduction Preeclampsia is defined as the presence of raised blood pressure (BP; ≥140/90 mmHg) after 20 weeks’ gestation, in a previously normotensive non‐proteinuric patient with one or more of the following: significant proteinuria (≥0.3 g/24 h), maternal organ dysfunction or uteroplacental dysfunction 1, 2. Suspected preeclampsia is the most frequent clinical presentation to obstetric units. Preeclampsia is associated with severe complications such as seizures, stroke, multiple organ failure and perinatal mortality if not recognized and managed properly.

4 h), maternal organ dysfunction or uteroplacental dysfunction 1, 2. Suspected preeclampsia is the most frequent clinical presentation to obstetric units. Preeclampsia is associated with severe complications such as seizures, stroke, multiple organ failure and perinatal mortality if not recognized and managed properly. The spot urinary protein to creatinine ratio and the albumin to creatinine ratio (ACR) have been studied in patients with renal disease, diabetes and preeclampsia to assess proteinuria. Albumin excretion is considered to reflect glomerular damage more accurately than total protein excretion, and albuminuria may be a marker of systemic endothelial cell dysfunction 3. The majority of international organizations now recommend spot proteinuria tests in the assessment of suspected preeclampsia. ACR has been shown to be an accurate indicator of proteinuria in women with preeclampsia 4, 5, 6. Despite this evidence, the obstetric community has not widely adopted the use of ACR as yet, and protein to creatinine ratio or 24‐h urine collection are more commonly employed.

in the assessment of suspected preeclampsia. ACR has been shown to be an accurate indicator of proteinuria in women with preeclampsia 4, 5, 6. Despite this evidence, the obstetric community has not widely adopted the use of ACR as yet, and protein to creatinine ratio or 24‐h urine collection are more commonly employed. As well as being useful in the diagnosis of preeclampsia 4, 6, ACR has potential to be useful in predicting adverse pregnancy outcomes 7, 8. New prognostic factors are needed in this area 9, 10, 11, 12. Prognostic factors can guide clinical decision‐making and patient counseling, and inform the design and analysis of new trials 10, 11, 12. They can also improve prognostic models, which produce absolute risk predictions for women based on a set of individual characteristics 9. Before including a new factor in a prognostic model, it is important to quantify its independent prognostic value over and above existing prognostic factors. Factors that add additional (independent) prognostic information are difficult to find, but are necessary to improve the discrimination performance of prognostic models 12.

luding a new factor in a prognostic model, it is important to quantify its independent prognostic value over and above existing prognostic factors. Factors that add additional (independent) prognostic information are difficult to find, but are necessary to improve the discrimination performance of prognostic models 12. The aim of this study was to examine the prognostic value of baseline ACR (ACR at first presentation) to predict maternal and neonatal adverse outcomes in women referred with suspected preeclampsia. There were two objectives: (i) to examine whether ACR is prognostic for adverse maternal and neonatal outcomes when no other factor is considered (unadjusted prognostic effect) and (ii) to evaluate whether ACR is a prognostic factor for such outcomes after adjusting for existing prognostic factors (independent prognostic effect). Material and methods Study design We performed a retrospective cohort study of pregnant women undergoing ACR test in the obstetric Day Assessment Units of two hospitals in National Health Service Lothian trust between December 2009 and February 2012. The Simpson Centre for Reproductive Health is a tertiary referral center with more than 6500 deliveries per annum. St John's Hospital is a district general hospital with approximately 2600 deliveries per annum. Women were excluded if they had not delivered by the end of February 2012.

n December 2009 and February 2012. The Simpson Centre for Reproductive Health is a tertiary referral center with more than 6500 deliveries per annum. St John's Hospital is a district general hospital with approximately 2600 deliveries per annum. Women were excluded if they had not delivered by the end of February 2012. All pregnant women with urinary ACR results were identified from a biochemistry database (APEX, ApexHealthware). Women were included if they had booked for their pregnancy prior to 14 weeks and if they were referred from primary care to the hospital Day Assessment Unit with suspected preeclampsia [suspected hypertension (generally ≥140/90 mmHg) and at least 1+ proteinuria on dipstick testing]. Women were excluded if they had multiple pregnancy, proteinuric renal disease or proven urinary tract infection, or if the ACR was measured for another indication (for example diabetes). Women who had their first ACR sent prior to 20 weeks’ gestation were also excluded, as this suggests a chronic hypertensive or proteinuric disorder or underlying renal pathology.

nancy, proteinuric renal disease or proven urinary tract infection, or if the ACR was measured for another indication (for example diabetes). Women who had their first ACR sent prior to 20 weeks’ gestation were also excluded, as this suggests a chronic hypertensive or proteinuric disorder or underlying renal pathology. We performed systematic review of medical records collecting predefined characteristics (demographic and clinical) to maximize accuracy and minimize missing data. We used multiple data sources to collect neonatal outcome data in order to increase confidence that no cases of perinatal mortality or significant morbidity were missed. Data were acquired from the maternity electronic patient records database TRAK (supplied by Intersystems) and the neonatal unit electronic patient records database BadgerNet (supplied by Clevermed) systems. Demographic features were recorded at booking visit, clinical and laboratory data at the time of first ACR measurement and subsequent antenatal visits and at delivery, and the outcome of mothers and babies was collected for every pregnancy.

electronic patient records database BadgerNet (supplied by Clevermed) systems. Demographic features were recorded at booking visit, clinical and laboratory data at the time of first ACR measurement and subsequent antenatal visits and at delivery, and the outcome of mothers and babies was collected for every pregnancy. ACR measurement taken on first hospital assessment for suspected preeclampsia was used in the analysis (i.e. follow‐up measurements were not included). ACR was calculated from urine samples in the biochemistry labs of The Royal Infirmary of Edinburgh. Immunoassays (Abbott Architect), turbidmetric and kinetic alkaline picrate (Jaffe) were used to calculate the concentrations of albumin and creatinine, respectively, in the urine sample. From this, the albumin (mg/L)/urine creatinine (mmol/L) was calculated.

in the biochemistry labs of The Royal Infirmary of Edinburgh. Immunoassays (Abbott Architect), turbidmetric and kinetic alkaline picrate (Jaffe) were used to calculate the concentrations of albumin and creatinine, respectively, in the urine sample. From this, the albumin (mg/L)/urine creatinine (mmol/L) was calculated. Existing prognostic factors were: gestational age at ACR measurement, essential hypertension, preexisting diabetes, gestational diabetes, social deprivation index, body mass index (BMI), mean arterial BP, current smoking status, parity and maternal age recorded from the clinical record at booking (<14 weeks). Deprivation was recorded as social multiple index of deprivation [a postcode‐based Scottish Index of multiple deprivation from 2012 – five groups ranging from most deprived index 1 to least deprived index 5] 13. BMI was recorded as <18.5, 18.5–24.99, 25.0–29.99, 30.0–34.9, 35.0–39.9 and >40. Mean arterial BP(MAP) was recorded as diastolic BP+ 1/3 (systolic BP‐diastolic BP). MAP was used in place of systolic or diastolic BP because previous evidence suggests it is a better prognostic factor for preeclampsia than BP measured during the first or second trimester of pregnancy 14. Data on development of gestational diabetes (to allow exclusion and diagnosed using Scottish Intercollegiate Guidelines network guideline) 15 and gestation at ACR (days) were also recorded.

suggests it is a better prognostic factor for preeclampsia than BP measured during the first or second trimester of pregnancy 14. Data on development of gestational diabetes (to allow exclusion and diagnosed using Scottish Intercollegiate Guidelines network guideline) 15 and gestation at ACR (days) were also recorded. The primary maternal outcome was a composite adverse maternal outcome, defined as one or more of: use of intravenous magnesium sulfate for seizure prophylaxis, use of intravenous antihypertensives, admission to intensive care unit/or high dependency unit for hypertension, placental abruption, eclampsia or HELLP (haemolysis, elevated liver enzymes, low platelets). The primary neonatal outcome was a composite adverse neonatal outcome, defined as one or more of: iatrogenic preterm delivery <34 weeks, birthweight <5th centile (calculated from sex‐specific birthweight centile charts) 16, abnormal umbilical artery Doppler [absent or reversed end‐diastolic (ARED) flow], arterial cord pH <7.1, need for ventilation, neonatal or intrauterine death. Secondary outcome was gestation at delivery (weeks). No formal power calculation was performed, and we included all data available over a three‐year time period to maximize sample size. In prognosis research, a typical rule of thumb is that at least 10 events (cases with the outcome of interest) are required to evaluate every one candidate prognostic variable 17. In our study, over 200 women had a maternal composite adverse outcome, thus the sample size was considered adequate for the analysis performed.

prognosis research, a typical rule of thumb is that at least 10 events (cases with the outcome of interest) are required to evaluate every one candidate prognostic variable 17. In our study, over 200 women had a maternal composite adverse outcome, thus the sample size was considered adequate for the analysis performed. In all, 3.9% of women had one or more missing values for data on existing prognostic factors. Due to the small proportion missing we considered a complete case multivariable analysis sufficient 18. Thus, only a complete case analysis was performed, and the relatively few women with missing data were excluded from the multivariable analysis but included in the ACR‐only analysis. Primary analyses The baseline characteristics of the sample were summarized by primary outcome status with differences between groups assessed using unpaired t‐tests or Mann–Whitney U‐tests for continuous and chi‐square tests for binary data. Univariable and multivariable logistic regression models were used to examine the unadjusted and the adjusted (independent) prognostic association of ACR with each binary primary outcome. The multivariable analysis was adjusted for a predefined set of factors that we considered to be prognostic factors, as described above.

ariable and multivariable logistic regression models were used to examine the unadjusted and the adjusted (independent) prognostic association of ACR with each binary primary outcome. The multivariable analysis was adjusted for a predefined set of factors that we considered to be prognostic factors, as described above. For the continuous variable “ACR” the assumption of linearity of the prognostic effect on the log‐odds scale was examined using fractional polynomials. Fractional polynomials of degree two were used to obtain an appropriate transformation for ACR, for which the linearity assumption did not hold 19. This suggested that a logarithmic transformation was needed for ACR. Thus, the logistic models estimated the prognostic value of ACR as summarized by an (adjusted) odds ratio (OR), giving the (adjusted) relative odds of the outcome for two individuals that differ in log‐ACR by 1 unit. To avoid deletion of patients with undefined log‐transformed ACR values [log (0)], 0.01 was added across all the entries of ACR following transformation of the data. Similarly, univariable and multivariable models were fitted for the secondary outcome, gestation weeks at delivery using linear rather than logistic regression.

For the continuous variable “ACR” the assumption of linearity of the prognostic effect on the log‐odds scale was examined using fractional polynomials. Fractional polynomials of degree two were used to obtain an appropriate transformation for ACR, for which the linearity assumption did not hold 19. This suggested that a logarithmic transformation was needed for ACR. Thus, the logistic models estimated the prognostic value of ACR as summarized by an (adjusted) odds ratio (OR), giving the (adjusted) relative odds of the outcome for two individuals that differ in log‐ACR by 1 unit. To avoid deletion of patients with undefined log‐transformed ACR values [log (0)], 0.01 was added across all the entries of ACR following transformation of the data. Similarly, univariable and multivariable models were fitted for the secondary outcome, gestation weeks at delivery using linear rather than logistic regression. For the neonatal composite outcome, gestational age at ACR measurement was adjusted for as a binary outcome after categorizing to age <34 weeks and age ≥34 weeks. This categorization was enforced by the clinical team in advance of the analysis as follows: Women who had the first ACR test before 34 weeks represented a group with suspected preterm preeclampsia vs. women with suspected later onset preeclampsia. Preterm preeclampsia is a more severe clinical condition and is associated more often with neonatal adverse outcome including premature delivery. Part of the composite adverse neonatal outcome is iatrogenic preterm delivery prior to 34 weeks.

For the neonatal composite outcome, gestational age at ACR measurement was adjusted for as a binary outcome after categorizing to age <34 weeks and age ≥34 weeks. This categorization was enforced by the clinical team in advance of the analysis as follows: Women who had the first ACR test before 34 weeks represented a group with suspected preterm preeclampsia vs. women with suspected later onset preeclampsia. Preterm preeclampsia is a more severe clinical condition and is associated more often with neonatal adverse outcome including premature delivery. Part of the composite adverse neonatal outcome is iatrogenic preterm delivery prior to 34 weeks. The rationale for the above was based on the existing literature 20, 21, 22, 23. Secondary analysis The discrimination performance of the entire multivariable model was summarized to ascertain its potential as a prognostic model, using the apparent C statistic [area under the receiver operating characteristic (ROC) curve] where 0.5 indicates no discrimination (between those with and those without the outcome) beyond chance and 1 indicates perfect discrimination. The C‐statistic is equivalently defined as the probability that the predicted risk for a randomly selected individual with the outcome is higher than that for a randomly selected individual without the outcome 24.

n (between those with and those without the outcome) beyond chance and 1 indicates perfect discrimination. The C‐statistic is equivalently defined as the probability that the predicted risk for a randomly selected individual with the outcome is higher than that for a randomly selected individual without the outcome 24. Sensitivity analysis Alongside the univariable and multivariable logistic regression analyses to obtain ORs, Poisson regression with robust standard errors was used to obtain (adjusted) risk ratios (RRs). The dataset included extreme values (two entries ACR = 2000 and one entry where ACR = 0). Therefore a sensitivity analysis was run to examine the effect of excluding these values. All analyses were performed in STATA version 12 (StataCorp, College Station, TX, USA) and the regression models fitted using maximum likelihood estimation. This was a retrospective study on samples already obtained and the study was approved through the University of Edinburgh and registered with the University of Edinburgh and NHS Lothian on 29/2/2012. No external ethics committee was required. An agreement with the data holder was in place to use the data, for the purposes of this study, which were anonymous and unlinked.

ained and the study was approved through the University of Edinburgh and registered with the University of Edinburgh and NHS Lothian on 29/2/2012. No external ethics committee was required. An agreement with the data holder was in place to use the data, for the purposes of this study, which were anonymous and unlinked. Results In all, 941 pregnant women had an ACR performed during the study period. A total of 224 records were excluded due to predefined exclusion criteria, leaving a cohort of 717 women. Complete data (on ACR and existing prognostic factors for the multivariable analysis) was available for 689 women. Women's characteristics are detailed in Table 1. The majority of first ACR measurements were performed between 35 and 40 weeks’ gestation (interquartile range 35–40 weeks, median 37 weeks and standard deviation 4 weeks). Table 1 Baseline maternal characteristics (values are numbers and percentages of the presence of a given characteristic)

Results In all, 941 pregnant women had an ACR performed during the study period. A total of 224 records were excluded due to predefined exclusion criteria, leaving a cohort of 717 women. Complete data (on ACR and existing prognostic factors for the multivariable analysis) was available for 689 women. Women's characteristics are detailed in Table 1. The majority of first ACR measurements were performed between 35 and 40 weeks’ gestation (interquartile range 35–40 weeks, median 37 weeks and standard deviation 4 weeks). Table 1 Baseline maternal characteristics (values are numbers and percentages of the presence of a given characteristic) Characteristic Participants (n = 717) Maternal age at delivery (years), mean (SD) 29.93 (6.06) Booking characteristics Nulliparity 57.18% Essential hypertension 9.34% Preexisting diabetes 2.79% Current smoker 15.85% Scottish index of multiple deprivation 1 (most deprived) 21.51% 2 22.63% 3 20.39% 4 15.39% 5 (least deprived) 20.11% Body mass index <18.5 2.32% 18.5–24.99 33.48% 25.0–29.99 28.55% 30.0–34.9 20.14% 35.0–39.9 9.71% >40 5.80% Booking systolic BP, mean (SD) 115.26 (12.48) Booking diastolic BP, mean (SD) 69.78 (9.81) Booking mean arterial BP, mean (SD) 84.94 (9.95) Development of gestational diabetes 3.35% Gestational age at ACR test (weeks), median (IQR) 37.43 (35.0–39.14) ACR result (mg/mmol), median (IQR) 4.40 (1.40–23.60) Gestational age at delivery (weeks), median (IQR) 39.43 (38.00–40.43) BP, blood pressure; HDU, high dependency unit; HELLP, hemolysis elevated liver enzymes low platelet count syndrome; ICU, intensive care unit; IQR, interquartile range; SD, standard deviation.

9.14) ACR result (mg/mmol), median (IQR) 4.40 (1.40–23.60) Gestational age at delivery (weeks), median (IQR) 39.43 (38.00–40.43) BP, blood pressure; HDU, high dependency unit; HELLP, hemolysis elevated liver enzymes low platelet count syndrome; ICU, intensive care unit; IQR, interquartile range; SD, standard deviation. John Wiley & Sons, LtdAdverse maternal outcomes Of the 717 women included, 204 experienced a composite adverse maternal outcome (28.5%) (Table 2). Thirty women had more than one adverse event (n = 174 one event, n = 26 two events, n = 4 three events), leading to a total of 238 adverse outcomes. Supporting Information Table S1 shows the maternal characteristics for the women with and without composite adverse maternal outcomes. MAP and maternal age at booking were comparable between the two groups. There was no significant difference between the two groups regarding essential hypertension, gestational diabetes, and smoking or social deprivation index. Univariable analysis showed that mean ACR, median gestational age at ACR measurement, mean maternal age, preexisting diabetes and BMI differed between the two outcome groups (Table S1). Table 2 Number of maternal and neonatal adverse outcomes

John Wiley & Sons, LtdAdverse maternal outcomes Of the 717 women included, 204 experienced a composite adverse maternal outcome (28.5%) (Table 2). Thirty women had more than one adverse event (n = 174 one event, n = 26 two events, n = 4 three events), leading to a total of 238 adverse outcomes. Supporting Information Table S1 shows the maternal characteristics for the women with and without composite adverse maternal outcomes. MAP and maternal age at booking were comparable between the two groups. There was no significant difference between the two groups regarding essential hypertension, gestational diabetes, and smoking or social deprivation index. Univariable analysis showed that mean ACR, median gestational age at ACR measurement, mean maternal age, preexisting diabetes and BMI differed between the two outcome groups (Table S1). Table 2 Number of maternal and neonatal adverse outcomes Values are numbers Maternal adverse outcomes (Total n = 238) Use of magnesium sulfate 12 Use of intravenous antihypertensives 15 Admission to HDU or ICU for hypertension 196 Abruption 7 Eclampsia 0 HELLP 8 Neonatal adverse outcomes (Total n = 192) Iatrogenic preterm delivery <34 weeks 33 Birthweight <5th centile 98 Abnormal Dopplers (AEDF or REDF) 11 Arterial cord pH <7.1 12 Need for ventilation 32 Intrauterine death 5 Neonatal death 1 AEDF, absent end‐diastolic flow; HDU, high dependency unit; HELLP, hemolysis elevated liver enzymes low platelet count syndrome; ICU, intensive care unit; REDF reversed end‐diastolic flow.

ntile 98 Abnormal Dopplers (AEDF or REDF) 11 Arterial cord pH <7.1 12 Need for ventilation 32 Intrauterine death 5 Neonatal death 1 AEDF, absent end‐diastolic flow; HDU, high dependency unit; HELLP, hemolysis elevated liver enzymes low platelet count syndrome; ICU, intensive care unit; REDF reversed end‐diastolic flow. John Wiley & Sons, LtdAdverse neonatal outcomes Of 717 neonates, 146 experienced a composite adverse neonatal outcome (20.4%) (Table 2). Twenty‐eight neonates had more than one adverse event (n = 118 one event, n = 15 two events, n = 8 three events and n = 5 four events), leading to a total of 192 adverse outcomes. Maternal age was comparable between the two groups. There were differences in median gestational age at ACR measurement, mean ACR, smoking, BMI and MAP between the groups (see Supporting Information Table S2). Unadjusted and adjusted prognostic value of ACR for maternal and neonatal adverse outcomes Univariable logistic regression analysis of all 717 women (Table 3) showed that log ACR is prognostic for both maternal [OR 1.52, 95% confidence interval (CI) 1.38–1.684] and neonatal (OR 1.13, 95% CI 1.02–1.25) composite adverse outcome. These unadjusted estimates imply that a unit increase in log‐transformed ACR increases the odds of maternal and neonatal adverse outcomes by 52% and 13%, respectively. Table 3 Logistic regression results for unadjusted and adjusted models for the primary outcomes: composite maternal and composite neonatal outcomes

Unadjusted and adjusted prognostic value of ACR for maternal and neonatal adverse outcomes Univariable logistic regression analysis of all 717 women (Table 3) showed that log ACR is prognostic for both maternal [OR 1.52, 95% confidence interval (CI) 1.38–1.684] and neonatal (OR 1.13, 95% CI 1.02–1.25) composite adverse outcome. These unadjusted estimates imply that a unit increase in log‐transformed ACR increases the odds of maternal and neonatal adverse outcomes by 52% and 13%, respectively. Table 3 Logistic regression results for unadjusted and adjusted models for the primary outcomes: composite maternal and composite neonatal outcomes Model Variable Composite maternal adverse outcome Composite neonatal adverse outcome OR (95% CI) p‐value ROCa OR (95% CI) p‐value ROCa

Unadjusted and adjusted prognostic value of ACR for maternal and neonatal adverse outcomes Univariable logistic regression analysis of all 717 women (Table 3) showed that log ACR is prognostic for both maternal [OR 1.52, 95% confidence interval (CI) 1.38–1.684] and neonatal (OR 1.13, 95% CI 1.02–1.25) composite adverse outcome. These unadjusted estimates imply that a unit increase in log‐transformed ACR increases the odds of maternal and neonatal adverse outcomes by 52% and 13%, respectively. Table 3 Logistic regression results for unadjusted and adjusted models for the primary outcomes: composite maternal and composite neonatal outcomes Model Variable Composite maternal adverse outcome Composite neonatal adverse outcome OR (95% CI) p‐value ROCa OR (95% CI) p‐value ROCa Unadjusted ACRb 1.52 (1.38–1.68) <0.001 0.70 (0.66–0.74) 1.13 (1.02–1.25) 0.022 0.557 (0.504–0.610) Adjusted ACRb 1.60 (1.42–1.80) <0.001 0.76 (0.72–0.80) 1.15 (1.02–1.29) 0.025 0.718 (0.668–0.760) Gestational age at ACR 0.88 (0.83–0.92) <0.001 0.25 (0.16–1.29) <0.001 Maternal age 1.04 (1.08–1.08) 0.019 0.99 (0.95–1.02) 0.505 Essential hypertension 0.78 (0.38–1.60) 0.505 1.62 (0.79–3.33) 0.19 Preexisting diabetes 0.68 (0.12–3.72) 0.655 1.77 (0.40–7.88) 0.452 Gestational diabetes 1.02 (0.38–2.77) 0.964 0.73 (0.19–2.77) 0.64 Smoking 0.85 (0.50–1.45) 0.55 1.94 (1.16–3.26) 0.012 Nulliparity 0.96 (0.66–1.40) 0.826 1.16 (0.77–1.75) 0.471 Social deprivation index 1 1 1 2 0.80 (0.46–1.41) 0.451 0.95 (0.53–1.72) 0.876 3 1.10 (0.63–1.92) 0.73 0.78 (0.42–1.45) 0.437 4 0.623 (0.33–1.19) 0.152 0.78 (0.39–1.55) 0.473 5 0.424 (0.26–0.80) 0.008 0.84 (0.43–1.62) 0.603 Body mass index <18.5 1 1 18.5–24.99 1.06 (0.32–3.50) 0.93 0.21 (0.07–0.64) 0.006 25.0–29.99 1.47 (0.44–4.93) 0.535 0.12 (0.04–0.38) <0.001 30.0–34.9 0.70 (0.20–2.48) 0.581 0.11 (0.03–0.37) <0.001 35.0–39.9 0.50 (0.13–1.98) 0.321 0.16 (0.04–0.57) 0.005 >40.0 0.56 (0.13–2.39) 0.434 0.09 (0.02–0.41) 0.002 Mean arterial blood pressure 1.02 (1.00–1.05) 0.041 0.99 (0.97–1.01) 0.381 ROC, receiver operating characteristic.

0.44–4.93) 0.535 0.12 (0.04–0.38) <0.001 30.0–34.9 0.70 (0.20–2.48) 0.581 0.11 (0.03–0.37) <0.001 35.0–39.9 0.50 (0.13–1.98) 0.321 0.16 (0.04–0.57) 0.005 >40.0 0.56 (0.13–2.39) 0.434 0.09 (0.02–0.41) 0.002 Mean arterial blood pressure 1.02 (1.00–1.05) 0.041 0.99 (0.97–1.01) 0.381 ROC, receiver operating characteristic. a C statistic. b Log‐transformed ACR (albumin creatinine ratio). John Wiley & Sons, LtdMultivariable analysis (based on the 689 women with complete data, Table 3) also showed that log ACR is an independent prognostic factor for maternal composite adverse outcome (OR 1.60, 95% CI 1.43–1.80) and neonatal composite adverse outcome (OR 1.15, 95% CI 1.02–1.29). This implies that a unit increase in log‐transformed ACR, after adjusting for other factors, increases the odds of adverse maternal composite outcome by 60% and of adverse neonatal outcome by 15%. Unadjusted and adjusted prognostic value of ACR for gestation at delivery Univariable (coefficient −0.38, 95% CI −0.48 to −0.27, p < 0.001) and multivariable linear regression (coefficient −0.46, 95% CI −0.54 to −0.38, p < 0.001) shows a prognostic effect of log ACR for gestational age at delivery (Supporting Information Table S3). The adjusted estimate implies that for every unit increase in log‐transformed ACR, the average gestational age at delivery is decreased by about 0.5 weeks.

ession (coefficient −0.46, 95% CI −0.54 to −0.38, p < 0.001) shows a prognostic effect of log ACR for gestational age at delivery (Supporting Information Table S3). The adjusted estimate implies that for every unit increase in log‐transformed ACR, the average gestational age at delivery is decreased by about 0.5 weeks. Discrimination performance of the multivariate models The apparent C‐statistic for the multivariable models was 0.76 (95% CI 0.72–0.80) for composite maternal adverse outcome and 0.72 (95% CI 0.67–0.77) for composite neonatal adverse outcome (Table 3). If ACR is removed, then the C‐statistic of the multivariable models is reduced considerably to 0.67 (95% CI 0.64–0.72) for maternal composite outcome; however, for the neonatal outcome the C‐statistic and its 95% CI barely change. This suggests that ACR is more important in terms of providing additional discrimination as to outcome risk predictions, for the maternal outcome. Sensitivity analysis Results from the Poisson model with robust standard errors were consistent with those of logistic regression analysis. In both the univariable and multivariable analysis ACR still had significant prognostic ability for maternal (unadjusted RR 1.31, 95% CI 1.24–1.39; adjusted RR 1.32 95% CI 1.25–1.41) and neonatal outcomes (RR 1.10, 95% CI 1.01–1.19; adjusted RR 1.10, 95% CI 1.02–1.19) (Supporting Information Table S4). This implies that, after adjusting for other factors, a unit increase in log‐transformed ACR increases the risk of adverse maternal outcome by 32% and of fetal adverse outcome by 10%.

% CI 1.25–1.41) and neonatal outcomes (RR 1.10, 95% CI 1.01–1.19; adjusted RR 1.10, 95% CI 1.02–1.19) (Supporting Information Table S4). This implies that, after adjusting for other factors, a unit increase in log‐transformed ACR increases the risk of adverse maternal outcome by 32% and of fetal adverse outcome by 10%. The sensitivity analysis, excluding the extreme values (ACR = 2000 and ACR = 0), did not alter any conclusions for either primary and secondary outcomes (Supporting Information Tables S5 and S6). Supporting Information Figures S1 and S2 show the predicted probability of maternal adverse composite outcomes for ACR (Figure S1) and log ACR (Figure S2) based on the univariable and multivariable models excluding extreme values (ACR = 2000 and ACR = 0). To illustrate the appropriate fit of a linear relation between log ACR and the log‐odds of a maternal composite outcome, Figure 1 shows the unadjusted linear relationship alongside the observed risk. Figure 1 Graph of the predicted probability of maternal composite adverse outcome against albumin (mg/L) creatinine (mmol/L) ratio (ACR). The adjusted (red) and unadjusted (blue) models were fitted using log e (ln)‐transformed ACR and the logit was obtained using the coefficients from the fitted model multiplied by the means/medians of all other continuous adjustment factors, the most common category of the categorical adjustment factors and the values of log ACR. [Color figure can be viewed at wileyonlinelibrary.com]

log e (ln)‐transformed ACR and the logit was obtained using the coefficients from the fitted model multiplied by the means/medians of all other continuous adjustment factors, the most common category of the categorical adjustment factors and the values of log ACR. [Color figure can be viewed at wileyonlinelibrary.com] Discussion Based on this retrospective cohort study, we show that log ACR is an independent prognostic factor for composite adverse maternal and neonatal outcomes. We suggest that a unit increase in log‐transformed ACR is associated with a 30% increased risk of maternal adverse composite outcome and a 10% risk of neonatal adverse composite outcome (corresponding to increased odds of 60 and 15%, respectively). We also demonstrated that in this population a 1‐unit increase in log ACR was associated with a decrease in gestation at delivery by approximately 0.5 weeks (approximately 3 days). Based on the secondary analyses we showed that although ACR adds prognostic value, the overall discrimination performance of the multivariable models was only moderate. Thus, additional prognostic factors are required to improve performance further, for a clinically useful model to identify those most likely to have an adverse outcome. In terms of improving discrimination performance (as measured by the C‐statistic), ACR appears to be more important for maternal outcomes than for fetal outcomes.

ional prognostic factors are required to improve performance further, for a clinically useful model to identify those most likely to have an adverse outcome. In terms of improving discrimination performance (as measured by the C‐statistic), ACR appears to be more important for maternal outcomes than for fetal outcomes. A systematic review 25 and study that used ORs and appropriate tests on two ACR thresholds 26 have already indicated a prognostic ability of ACR for adverse outcomes associated with preeclampsia. Nonetheless, three of five of the studies included in the systematic review 25 were conducted 30 years ago with ACR tests that had different thresholds and that were performed in heterogeneous populations 7. Previous work is also limited by the use of thresholds to categorize (or dichotomize) ACR values 26. Other studies have found that the degree of proteinuria does not correlate with adverse outcome 6, 27. A major strength and uniqueness of our study was that ACR was analyzed as a continuous variable 28. Categorization of continuous predictors leads to loss of information, and hence loss of power, as well as poor predictive performance, and hence poor clinical usefulness 29, 30, 31. It also leads to data dredging (to find the “best” threshold) and does not reflect the underlying prognostic trend.

ontinuous variable 28. Categorization of continuous predictors leads to loss of information, and hence loss of power, as well as poor predictive performance, and hence poor clinical usefulness 29, 30, 31. It also leads to data dredging (to find the “best” threshold) and does not reflect the underlying prognostic trend. A log transformation was identified as the most appropriate scale on which to incorporate ACR in the model, suggesting that the effect of a 1‐unit increase in ACR depends on the actual value of ACR itself. Other strengths include the use of stored samples to measure ACR using standardized measurement methods, the collection of ACR values blind to the outcome status, the reasonably large cohort itself, and the very small amount of missing data.

effect of a 1‐unit increase in ACR depends on the actual value of ACR itself. Other strengths include the use of stored samples to measure ACR using standardized measurement methods, the collection of ACR values blind to the outcome status, the reasonably large cohort itself, and the very small amount of missing data. This study had some limitations. The primary outcomes were “composite” to increase the power to detect the prognostic ability of ACR. Moreover, the outcomes are objective, and clinical severity is similar within each group. However, it is difficult to examine the effect size of the prognostic factor of interest for each outcome separately 32. It is instead presumed that the effect size is related to all the components of the composite outcome. It is recommended that components of composite outcomes be considered secondary outcomes and that the related results are provided alongside primary analysis. This was not possible in this study due to the small number of events in most of the components of the composite outcome. However, these components were carefully selected to ensure that they were comparable in magnitude of severity and direction of effect.

t the related results are provided alongside primary analysis. This was not possible in this study due to the small number of events in most of the components of the composite outcome. However, these components were carefully selected to ensure that they were comparable in magnitude of severity and direction of effect. A further potential limitation results from the retrospective design of our study, as it is difficult to exclude the possibility of intervention bias in observational studies of this type. ACR results were available to clinicians, and may have influenced management decisions and thereby affected maternal and neonatal outcomes. However, these effects are likely to be small, as decision‐making in women with preeclampsia is based on the whole clinical presentation, not just the amount of proteinuria. We have shown that in women with suspected preeclampsia the ACR at presentation is an independent predictor of adverse outcome. As an indication of the potential usefulness of ACR in practice, Figure 1 shows how the value of ACR would change the predicted probability of an adverse outcome for a woman who otherwise would have median values of other covariates included in our model. However, clinical management of women with preeclampsia is directed by multiple factors, for example BP control, hematological and biochemical parameters, symptomatology and fetal considerations, including gestation. Thus, no single factor determines management or, in particular, intervention via delivery. Our data suggest that ACR should be considered within this clinical assessment.

by multiple factors, for example BP control, hematological and biochemical parameters, symptomatology and fetal considerations, including gestation. Thus, no single factor determines management or, in particular, intervention via delivery. Our data suggest that ACR should be considered within this clinical assessment. A recent series on prognosis research 9, 10, 11, 12 discusses how a single prognostic factor (such as ACR) rarely predicts individual outcome risk accurately, and usually does not suitably discriminate between high‐risk and low‐risk individuals. This is why prognostic models are needed, as they utilize multiple prognostic factors in combination to improve individual risk prediction accuracy and to discriminate better the underlying risk across individuals 33. Future work should focus on identifying further independent prognostic factors for adverse outcomes in order to further improve the discrimination performance of prognostic models. This may include the examination of the prognostic value of multiple measurements of ACR over time. In due course, a prognostic model could be developed incorporating a large set of prognostic factors (including ACR), followed by internal and external validation to ensure reliability of the model predictions. At that stage, its use in clinical decision‐making could be evaluated, for example based on values of high predicted risk that warrant clinical action.

developed incorporating a large set of prognostic factors (including ACR), followed by internal and external validation to ensure reliability of the model predictions. At that stage, its use in clinical decision‐making could be evaluated, for example based on values of high predicted risk that warrant clinical action. Funding There was no funding for this study. S.J.S. and F.D. are supported by Tommy's (registered charity nos 1060508 and SCO39280), who contribute to research infrastructure costs. Supporting information Table S1. Maternal characteristics for women who experienced maternal adverse composite outcome; values are numbers and percentages unless otherwise stated. Click here for additional data file. Table S2. Maternal characteristics for neonatals who experienced adverse composite outcomes; values are numbers and percentages unless otherwise stated. Click here for additional data file. Table S3. Linear regression results for the unadjusted and adjusted model for the secondary outcome; gestational age at delivery. Click here for additional data file. Table S4. Poisson regression with robust SE results for ACR (log‐transformed) for unadjusted, adjusted models, where the response is composite maternal/neonatal adverse outcome. Click here for additional data file. Table S5. Logistic regression results with extreme ACR values removed for log‐transformed ACR for unadjusted and adjusted models for the primary outcomes; composite maternal adverse outcome and composite neonatal outcome. Click here for additional data file.

Table S4. Poisson regression with robust SE results for ACR (log‐transformed) for unadjusted, adjusted models, where the response is composite maternal/neonatal adverse outcome. Click here for additional data file. Table S5. Logistic regression results with extreme ACR values removed for log‐transformed ACR for unadjusted and adjusted models for the primary outcomes; composite maternal adverse outcome and composite neonatal outcome. Click here for additional data file. Table S6. Linear regression results with extreme ACR values removed for log‐transformed ACR for the unadjusted and adjusted model for the secondary outcome; gestational age at delivery. Click here for additional data file. Figure S1. Graph of the predicted probability of maternal composite adverse outcome (AO) against albumin (mg/L) creatinine (mmol/L) ratio (ACR). The adjusted (red) and unadjusted (blue) models were fitted using log‐transformed ACR and the logit was obtained using the coefficients from the fitted model multiplied by the means/medians of all other continuous adjustment factors, the most common category of the categorical adjustment factors and the values of log ACR. Click here for additional data file. Figure S2. Graph of the predicted probability of maternal composite adverse outcome (AO) against the log‐transformed albumin (mg/L) creatinine (mmol/L) ratio (ACR). The adjusted (red) and unadjusted (blue) models were fitted using log‐transformed ACR and the logit was obtained using the coefficient of log ACR from the fitted model multiplied by the values of log ACR. Click here for additional data file.

Abbreviations c‐junV‐jun avian sarcoma virus 17 oncogene homolog DRGdorsal root ganglia EP2prostaglandin E2 receptor EPHectEndometriosis Phenome and Biobanking Hormonization Project ERestrogen receptor GWASgenome‐wide association studies HEECshuman endometrial endothelial cells hTERThuman telemorase reverse transcriptase HUVECshuman umbilical vein endothelial cells K‐rasV‐Ki‐ras2 Kirsten rat sarcoma viral oncogene homolog L‐THPlevo‐tetrahydropalmatine NKnatural killer PrkdcSCIloss‐of‐function mutation in the mouse homologue of the human PRKDC (severe immunodeficient mice) TGF‐β1transforming growth factor β1 TIMP‐1tissue inhibitor of matrix metalloproteinase 1 WERFWorld Endometriosis Research Foundation Key Message Due to its complex etiology and pathophysiology, endometriosis research requires careful selection of appropriate in vitro and in vivo models that, in combination with the use of well‐characterized human tissue can enhance the identification of anxiously awaited new therapies. Introduction Endometriosis is characterized by the growth of endometrium‐like tissue outside the uterus, most commonly on the pelvic peritoneum and ovaries 1. The presence of ectopic endometrial deposits (lesions) in the peritoneal cavity are thought to cause the defining symptoms of endometriosis, which are debilitating pelvic pain and infertility. Hence, these multicellular tissue deposits impact at least two biological systems; the nervous system and the reproductive system (Figure 1).

f ectopic endometrial deposits (lesions) in the peritoneal cavity are thought to cause the defining symptoms of endometriosis, which are debilitating pelvic pain and infertility. Hence, these multicellular tissue deposits impact at least two biological systems; the nervous system and the reproductive system (Figure 1). Figure 1 The presence of endometriosis lesions impacts the reproductive system and the nervous system. Endometriosis lesions present within the pelvic cavity cause the defining symptoms of the condition, which are infertility and chronic debilitating pelvic pain. Lesions generate an inflammatory environment, which may have a negative impact on developing oocytes and implanting blastocysts. This inflammation is also thought to activate nerve fibers that innervate lesions. In addition to disease‐specific cellular and molecular changes, extensive lesions and adhesions can cause distortion of the pelvic organs and nerve compression that may also contribute to fertility problems and pain. Animal models of endometriosis are required to dissect out specific disease mechanisms that impact on the reproductive and nervous system. [Color figure can be viewed at wileyonlinelibrary.com]

sions can cause distortion of the pelvic organs and nerve compression that may also contribute to fertility problems and pain. Animal models of endometriosis are required to dissect out specific disease mechanisms that impact on the reproductive and nervous system. [Color figure can be viewed at wileyonlinelibrary.com] To identify new means of treating the symptoms of endometriosis, it is critical to understand the complex pathophysiology of the condition. However, not only are there multiple theories on the origin of endometriosis (retrograde menstruation/transplantation, metaplasia), there are many different proposed theories on its pathogenesis (such as genetics, immune response, environment). Hence, there are multiple considerations and end points to consider during the design of experiments aimed at identifying potential therapeutic targets for endometriosis. Tools for endometriosis research fall into two broad categories: patient‐derived tissue and fluid and cells isolated from these sources, or models (in vitro and in vivo; Figure 2). Each have pros and cons and have differing utilities depending on the research question. In the current review, we discuss relevant human tissue resources and laboratory models for use in endometriosis research and we highlight the strengths and weaknesses of each.

e sources, or models (in vitro and in vivo; Figure 2). Each have pros and cons and have differing utilities depending on the research question. In the current review, we discuss relevant human tissue resources and laboratory models for use in endometriosis research and we highlight the strengths and weaknesses of each. Figure 2 Research question specific models for the study of endometriosis. Many theories on the pathophysiology and etiology of endometriosis exist. Hypothesis‐driven research will be facilitated with careful experimental design and selection of appropriate human tissue from patients with and without endometriosis and appropriate combinations of in vitro and in vivo laboratory models. [Color figure can be viewed at wileyonlinelibrary.com]

gy of endometriosis exist. Hypothesis‐driven research will be facilitated with careful experimental design and selection of appropriate human tissue from patients with and without endometriosis and appropriate combinations of in vitro and in vivo laboratory models. [Color figure can be viewed at wileyonlinelibrary.com] Material and methods We conducted a primary computerized literature search for relevant publications in “PubMed” that related to laboratory models for identifying therapeutic targets for endometriosis. We searched using the following key words: endometriosis AND human tissue OR mouse model OR rat model OR animal model OR in vitro. One author (EG) selected relevant abstracts and the full texts were obtained. EG reviewed studies that were published from 2000 to 2016 and although those studies formed the basis of our review, some older publications were included as deemed appropriate for historical but pivotal models of endometriosis. Reference lists from other publications were also examined for any relevant studies that were not extracted from the initial literature search. Articles were included if they (i) described the use of human tissue for the identification of novel genetic traits or pathways that could represent therapeutic targets for endometriosis (studies using human tissue were included only if patients and controls were confirmed via laparoscopy to have or not have endometriosis), (ii) described a unique in vivo rodent model of endometriosis, or (iii) described relevant in vitro techniques for the study of endometriosis. This is not a systematic review.

iosis (studies using human tissue were included only if patients and controls were confirmed via laparoscopy to have or not have endometriosis), (ii) described a unique in vivo rodent model of endometriosis, or (iii) described relevant in vitro techniques for the study of endometriosis. This is not a systematic review. Patient‐derived tissue, fluid and cells The appropriate methodology for the collection of patient‐derived material The analysis of human samples has provided valuable advances in our understanding of biological changes associated with endometriosis in both the peritoneum and the endometrium. As yet there are no validated clinical biomarkers of endometriosis 2, so the reference standard for diagnosis of the condition is the macroscopic visualization of lesions during laparoscopy. The procedure provides clinicians with an opportunity for the collection of tissue and fluid biospecimens for use in endometriosis research.

e are no validated clinical biomarkers of endometriosis 2, so the reference standard for diagnosis of the condition is the macroscopic visualization of lesions during laparoscopy. The procedure provides clinicians with an opportunity for the collection of tissue and fluid biospecimens for use in endometriosis research. Many centers worldwide have been collecting tissue and fluids from patients with and without endometriosis undergoing laparoscopy, as well as surgical and clinical phenotype data for a range of research purposes. However, huge variations in the collection of such data and specimens exist that could limit comparisons in data and reproducibility of results from different studies. The World Endometriosis Research Foundation (WERF) Endometriosis Phenome and Biobanking Harmonization Project (EPHect) spearheaded a mission to promote the adoption of internationally agreed standard operating procedures for tissue and fluid sample collection, processing and storage as well as surgical and clinical phenotype data collection. The WERF EPHect working group developed surgical and clinical questionnaires and evidence‐based standard operating procedures that were published in four concurrent articles 3, 4, 5, 6. This standardization will optimize sample quality, reduce variability and enable large‐scale cross‐center, epidemiologically robust endometriosis research. At the time of writing this review there are 13 registered centers using the WERF EPHect tools (http://endometriosisfoundation.org/ephect/centres-using-werf-ephect-tools/). A list of human biological specimens being used in endometriosis research is provided in Table 1.

epidemiologically robust endometriosis research. At the time of writing this review there are 13 registered centers using the WERF EPHect tools (http://endometriosisfoundation.org/ephect/centres-using-werf-ephect-tools/). A list of human biological specimens being used in endometriosis research is provided in Table 1. Table 1 Human tissues and fluid biospecimens used in endometriosis research Tissue Fluid Eutopic endometrium Blood Ectopic endometrium (lesion) Urine Unaffected peritoneum (adjacent and distal or prone and distal) Saliva Myometrium Peritoneal fluid Subcutaneous abdominal fat Endometrial fluid Omental/visceral fat Menstrual effluent John Wiley & Sons, LtdLesions vary in their location and invasiveness, biopsies can therefore be from the ovaries (endometriomas), the peritoneal lining, or deep infiltrating nodules. This heterogeneous nature of lesion biopsies, presentation of the condition, disease classification, variation in sample metadata, and the lack of correlation between perceived severity of disease and symptomatology can make data generated in endometriosis studies difficult to interpret. Standardization of clinical phenotype data collection and biopsy recovery through WERF EPHect will enable the generation of more reliable and comparable data. Samples should also be explicitly characterized and comprehensive details of sample metadata (cycle stage, disease severity, pain scores, history of subfertility) should be published alongside any results using human tissue to ensure transparency of any potential confounding factors.

more reliable and comparable data. Samples should also be explicitly characterized and comprehensive details of sample metadata (cycle stage, disease severity, pain scores, history of subfertility) should be published alongside any results using human tissue to ensure transparency of any potential confounding factors. Appropriate use of experimental controls is critical for human tissue data to be informative. Peritoneum and endometrium control samples should be included in studies that analyze gene expression in endometriosis lesions (particularly homogenized samples), because lesion biopsies are often contaminated with surrounding peritoneal tissue and in these cases it should be included as a control. It is not sufficient to compare gene expression in endometriosis lesions only to the endometrium or peritoneum alone. It is not entirely necessary to include peritoneal controls in immunohistochemical analysis of particular cell types, for example, because it is easy to distinguish the lesion boundary from surrounding peritoneum. Biopsies of peritoneum from sites adjacent and distal to lesions in patients with endometriosis and from sites prone to endometriosis in patients without the condition (Figure 3) also provides useful biological information when analyzed as additional controls within an experiment. Menstrual cycle phase is also known to have obvious impacts on peritoneal fluid concentration 7 and composition, and a profound impact on endometrial 8, and likely also endometriosis lesion, gene expression. Exogenous hormones can also modulate gene expression in these samples. For example, oral contraceptive use increases the expression of cyclooxygenase‐2 in the eutopic and ectopic endometrium of women with endometriosis 9. Stromal cells isolated from ovarian endometriomas exhibit increased expression of aromatase, estradiol 17β‐dehydrogenase 1, steroid sulfatase and estrogen sulfotransferase 10. Whether these gene expression changes are maintained when cells are isolated and cultured is uncertain because the majority of published studies use cells derived from patients that have not had exogenous hormone treatment 3–6 months before laparoscopy. Access to true control samples collected from fertile patients with no pain and no endometriosis is often limited due to the fact that few women undergo laparoscopy (or endometrial biopsy) without symptoms (decline in laparoscopic sterilization).

t have not had exogenous hormone treatment 3–6 months before laparoscopy. Access to true control samples collected from fertile patients with no pain and no endometriosis is often limited due to the fact that few women undergo laparoscopy (or endometrial biopsy) without symptoms (decline in laparoscopic sterilization). As women may exhibit no symptoms even though they have lesions, consistent with endometriosis detected at laparoscopy, caution is necessary in defining “control” samples. Figure 3 Peritoneal and endometriosis lesions collected at time of surgery. Taking biopsies of peritoneum from sites adjacent and distal to lesions in patients with endometriosis, and from sites prone to endometriosis in patients without the condition provides useful biological information when analyzed as additional controls within an experiment. [Color figure can be viewed at wileyonlinelibrary.com] Below we discuss a limited number of key studies that have significantly enhanced our understanding of endometriosis and also provide examples of identification of genes or pathways that could be targets for therapeutic intervention.

Figure 3 Peritoneal and endometriosis lesions collected at time of surgery. Taking biopsies of peritoneum from sites adjacent and distal to lesions in patients with endometriosis, and from sites prone to endometriosis in patients without the condition provides useful biological information when analyzed as additional controls within an experiment. [Color figure can be viewed at wileyonlinelibrary.com] Below we discuss a limited number of key studies that have significantly enhanced our understanding of endometriosis and also provide examples of identification of genes or pathways that could be targets for therapeutic intervention. Identification of genes associated with endometriosis, gene expression trends, and novel pathways Microarray transcriptomic studies in combination with pathway analysis have yielded informative results on biological changes taking place in the eutopic endometrium of women with endometriosis; Burney et al. demonstrated a dysregulation of the proliferative to secretory transition in women with endometriosis. In secretory phase endometrium of women with endometriosis, the authors identified persistent expression of genes associated with DNA synthesis and cellular mitosis and decreased expression of progesterone‐regulated genes, suggestive of a “progesterone‐resistant” phenotype 11.

ry transition in women with endometriosis. In secretory phase endometrium of women with endometriosis, the authors identified persistent expression of genes associated with DNA synthesis and cellular mitosis and decreased expression of progesterone‐regulated genes, suggestive of a “progesterone‐resistant” phenotype 11. Genome‐wide association studies (GWAS) performed on DNA extracted from blood or saliva, in combination with replication studies, have been hugely informative in the identification of genetic loci associated with endometriosis risk and associations with other genetic traits 12, 13. GWAS studies provide an opportunity for identifying new drug targets for endometriosis; genes discovered using this approach can be investigated for their ability to be targeted by small molecule inhibitors and therapeutic antibodies or protein therapeutics. For example, 155 of 991 genes implicated in disease from GWAS (15.6%) have an associated drug project in development 12. The implementation of WERF EPHect standard operating procedures for the collection, processing, and storage of tissue and biofluid specimens will enhance such large‐scale cross‐center collaboration, such as GWAS. Validation of results generated in GWAS studies remains a challenge, as does exploring the mechanistic relevance of identified genes.

EPHect standard operating procedures for the collection, processing, and storage of tissue and biofluid specimens will enhance such large‐scale cross‐center collaboration, such as GWAS. Validation of results generated in GWAS studies remains a challenge, as does exploring the mechanistic relevance of identified genes. Analysis of cytokine profiles of peritoneal fluid samples identified a subset of patients with a shared consensus signature of elevated cytokines associated with macrophage infiltration and activation. Bioinformatics analysis identified an enrichment of V‐jun avian sarcoma virus 17 oncogene homolog (c‐jun), FBJ murine osteosarcoma viral oncogene homolog (c‐fos) and activator protein 1 (AP‐1) transcription factor binding sites among the measured consensus cytokine signature. Subsequent inhibition of upstream kinases of c‐jun resulted in an attenuation of cytokine expression by macrophages isolated from peritoneal fluid from women with endometriosis 14, suggestive of a novel therapeutic strategy for limiting inflammatory mechanisms that drive endometriosis.

red consensus cytokine signature. Subsequent inhibition of upstream kinases of c‐jun resulted in an attenuation of cytokine expression by macrophages isolated from peritoneal fluid from women with endometriosis 14, suggestive of a novel therapeutic strategy for limiting inflammatory mechanisms that drive endometriosis. How studies on cells derived from patient tissues/fluids have informed our understanding of the pathophysiology of the disorder In a study from our group, gene expression analysis performed on human endometriosis lesions compared with endometrium and peritoneum revealed an upregulation of the axonal guidance molecule SLIT3 in lesions 15. In this same study immunofluorescence performed on human endometriosis lesions indicated that endothelial cells lining the blood vessels of lesions express estrogen receptor β (ERβ) and not ERα, and that blood vessels and nerves are found in close proximity in lesions. These findings informed in vitro and in vivo studies exploring ER regulation of Slits in blood vessel–nerve crosstalk in endometriosis 15. In another study from our team, analysis of peritoneal fluid revealed increased levels of transforming growth factor β1 (TGF‐β1) and lactate in women with endometriosis, and an increase in glycolysis‐related genes 16. This informed investigation of TGF‐β1 regulation of glycolysis genes and lactate levels in peritoneal mesothelial cells from women with and without endometriosis and led to the hypothesis that the “Warburg effect” (a high rate of glycolysis and lactic acid fermentation) seen in tumorigenesis is a key contributor to the pathophysiology of endometriosis and may be modulated by TGF‐β 16. These two studies emphasize the power of integrated studies that use a number of different models to explore hypothesis‐driven endometriosis research.

igh rate of glycolysis and lactic acid fermentation) seen in tumorigenesis is a key contributor to the pathophysiology of endometriosis and may be modulated by TGF‐β 16. These two studies emphasize the power of integrated studies that use a number of different models to explore hypothesis‐driven endometriosis research. Endometriosis lesions are complex multicellular tissue deposits (Figure 4). The hallmarks of an endometriosis lesion are the presence of endometrioid epithelial and stromal cells that resemble the cellular organization of the eutopic endometrium. These lesions are highly vascularized and we now know that they are innervated, enabling a dialogue between the lesion microenvironment and the nervous system. Immune cells including macrophages, mast cells, T and B lymphocytes and natural killer (NK) cells are also present within lesions and contribute to the inflammatory microenvironment of the lesion. The pelvic peritoneum may also play an important role in the establishment and progression of endometriosis by providing a surface for the attachment of endometrial fragments. Hence, many different cellular interactions can take place within a lesion and appropriate cells are required for the study of these interactions. To complement studies on intact tissue biopsies a number of studies have also focused on analysis of different cell types isolated from patient biopsies.

ometrial fragments. Hence, many different cellular interactions can take place within a lesion and appropriate cells are required for the study of these interactions. To complement studies on intact tissue biopsies a number of studies have also focused on analysis of different cell types isolated from patient biopsies. Figure 4 Endometriosis lesions are multicellular tissue deposits. Typical endometriosis lesions contain endometrioid glandular structures (made up of epithelial cells) and stromal cells, similar to the eutopic endometrium. Lesions become vascularized, innervated and infiltrated by immune cells such as macrophages, mast cells and natural killer cells. Isolation of specific cell types from lesions is difficult due to the limited amount of tissue available from biopsies. In most cases the exploration of cellular interactions must be recreated using cell models and disease‐specific effects on physiology must be modeled in vivo. [Color figure can be viewed at wileyonlinelibrary.com]

specific cell types from lesions is difficult due to the limited amount of tissue available from biopsies. In most cases the exploration of cellular interactions must be recreated using cell models and disease‐specific effects on physiology must be modeled in vivo. [Color figure can be viewed at wileyonlinelibrary.com] Eutopic endometrium Non‐cancerous epithelial cells are difficult to propagate from primary tissue due to their short lifespan and usually enter senescence after 2 weeks in culture. For this reason, only a limited number of studies use epithelial cells from patients with endometriosis. Those that have, indicate altered gene expression and enhanced migratory abilities in epithelial cells isolated from patients with endometriosis 17. Stromal cells are much easier to isolate and propagate from primary tissue and many studies have used stromal cells isolated from the eutopic endometrium from patients with and without endometriosis undergoing laparoscopy. Endometrial stromal cells may be induced to decidualize in vitro using progesterone and cAMP. This approach has been used as a model for studying the potential effects of endometriosis on endometrial receptivity and differences in gene expression in cells from women with and without endometriosis. For example, decreased Notch signaling and connexin 43 expression identified in the endometrium of women with endometriosis have been implicated in impaired decidualization using eutopic stromal cells 18, 19. Immune cells resident in the eutopic endometrium of patients may also be isolated with flow cytometry. For example, uterine NK cells have been analyzed using flow cytometry and found to exhibit decreased levels of killer cell inhibitory receptors 20. Increased levels of uterine NK progenitor cells have also been identified with flow cytometry in the eutopic endometrium and are thought to contribute to endometriosis‐associated infertility 21.

K cells have been analyzed using flow cytometry and found to exhibit decreased levels of killer cell inhibitory receptors 20. Increased levels of uterine NK progenitor cells have also been identified with flow cytometry in the eutopic endometrium and are thought to contribute to endometriosis‐associated infertility 21. Ectopic endometrium Epithelial cells have been isolated from lesion biopsies and used to explore the effect of inhibition of the Wnt/β‐catenin pathway on gene expression and the functional end points proliferation, migration and invasion 22. The very limited amount of epithelial cells that can be recovered from ectopic endometrium means multiple experiments cannot be performed from cells isolated from one patient and large patient numbers are required for a study. Primary stromal cells isolated from endometriosis lesions have also been used in a number of studies to demonstrate for example, that these cells also have an increased migration and invasion ability 23.

riments cannot be performed from cells isolated from one patient and large patient numbers are required for a study. Primary stromal cells isolated from endometriosis lesions have also been used in a number of studies to demonstrate for example, that these cells also have an increased migration and invasion ability 23. Peritoneum In women with endometriosis the phenotype of the mesothelial cells that line the peritoneal cavity has been reported to be altered, such that ectopic endometrial cells are more likely to attach and invade underlying structures 24. Primary human peritoneal mesothelial cells can be isolated during laparoscopic surgery by gentle brushing of the pelvic mesothelium using a specialized brush, the cells can then be dislodged and established in culture 16. These cell types have been used in studies that shed new light on the role of the mesothelial cell in endometriosis pathophysiology including a Warburg‐like metabolic reprogramming 16, 25, 26, 27. Peritoneal mesothelial cells have also been used in a co‐culture model to simulate interactions between endometrial stromal cells and human peritoneal mesothelial cells in normal and pathophysiological states 28.

lial cell in endometriosis pathophysiology including a Warburg‐like metabolic reprogramming 16, 25, 26, 27. Peritoneal mesothelial cells have also been used in a co‐culture model to simulate interactions between endometrial stromal cells and human peritoneal mesothelial cells in normal and pathophysiological states 28. Immune cells derived from peritoneal fluid or peripheral blood The peritoneal fluid is an incredibly useful resource for the study of the role played by immune cells in the etiology of endometriosis. T and B lymphocytes, NK cells and macrophages 14 isolated from the peritoneal fluid of women with and without endometriosis have been analyzed in a number of studies 29, 30, 31. The presence of endometriosis has also been hypothesized to effect immune profiles systemically; T, B and NK cells isolated from the peripheral blood from women with endometriosis have also been analyzed 29, 30, 31. In vitro and in vivo model systems In vitro models The number of cells that can be isolated from lesions is limiting for many studies. Additionally, nerves for example, cannot be isolated from biopsies and for these reasons cell models are required. Primary patient‐derived cells or cell models provide a limited “snap shot” of gene expression or cell function and at best can be used to provide some information on cell–cell interaction when used in co‐culture or three‐dimensional culture systems.

isolated from biopsies and for these reasons cell models are required. Primary patient‐derived cells or cell models provide a limited “snap shot” of gene expression or cell function and at best can be used to provide some information on cell–cell interaction when used in co‐culture or three‐dimensional culture systems. Epithelial and stromal cells The limited number of cells that can be isolated from endometriosis lesions is often the reason that many researchers decide to use cell lines. Normal human endometrial epithelial cells have been immortalized using human papillomavirus and human telemorase reverse transcriptase (hTERT) 32 and used as a control cell in a number of endometriosis studies. Purified epithelial cells isolated from ovarian endometriomas (E'mosis1 and 2) have also been immortalized by combined transfection of human cyclinD1, cyclin‐dependent kinase 4 (cdk4) and human telomerase reverse transcriptase (hTERT) 33. Another widely used endometriotic epithelial cell line (12Z) was first established from an active peritoneal lesion 34. As each of these cell lines is generated from individual patients, it cannot be assumed that results obtained from their interrogation are a true representation of endometriotic epithelial cells. However, they are useful for the analysis of signaling pathways or functional studies investigating invasion, migration and proliferation that may need more cells than can be isolated from primary tissue. The ectopic stromal cell line 22B (derived from an active red peritoneal lesion) has been used in a number of studies 34, 35 but poses the same limitation as the ectopic epithelial cell line discussed above.

igating invasion, migration and proliferation that may need more cells than can be isolated from primary tissue. The ectopic stromal cell line 22B (derived from an active red peritoneal lesion) has been used in a number of studies 34, 35 but poses the same limitation as the ectopic epithelial cell line discussed above. Mesothelial cells The humen pleural cavity mesothelial cell line MeT‐5A has also been used in endometriosis studies and is thought to mirror the phenotype of primary human peritoneal mesothelial cells. A peritoneal mesothelial cell line has been established through transfection with simian vacuolating virus 40 TAg (SV40 T) antigen 36. Peritoneal mesothelial cell line (LP9) is a commercially available peritoneal mesothelial cell line that has been used in a limited number of endometriosis studies 37. Primary mesothelial cells derived from different locations and cell lines are thought to have a similar phenotype to peritoneal mesothelial cells. However, as with most cell lines they are not useful for exploring disease‐specific phenotype.

cell line that has been used in a limited number of endometriosis studies 37. Primary mesothelial cells derived from different locations and cell lines are thought to have a similar phenotype to peritoneal mesothelial cells. However, as with most cell lines they are not useful for exploring disease‐specific phenotype. Immune cells Although peritoneal fluid is readily accessible during laparoscopy, analysis of its immune cells may not represent immune cell phenotypes present within endometriosis lesions. For example, tissue‐resident macrophages are known to have a different phenotype to peritoneal fluid macrophages. Peripheral blood monocyte‐derived macrophages may be plated and activated using different cytokines and estradiol to generate a phenotype similar to an endometriosis macrophage 38; this model is useful when large numbers of cells are required for gene expression and functional studies. Due to the small size of lesion biopsy material, flow cytometry for the isolation of lesion‐resident immune cells is extremely difficult and limits any manipulation of these cells ex vivo.

ometriosis macrophage 38; this model is useful when large numbers of cells are required for gene expression and functional studies. Due to the small size of lesion biopsy material, flow cytometry for the isolation of lesion‐resident immune cells is extremely difficult and limits any manipulation of these cells ex vivo. Nerve fibers Mechsners’ group analyzed a chicken dorsal root ganglia (DRG) explant model to explore the effects of peritoneal fluid on neurite outgrowth 39. DRG are thought to contain predominantly sensory neurons, and this group were also able to recover sympathetic ganglia to aid their studies showing an imbalance between sensory and sympathetic innervation in lesions of women with endometriosis; incubation of chicken DRG with peritoneal fluid from women with endometriosis resulted in increased neurite outgrowth from sensory ganglia and a decrease in neurite outgrowth from sympathetic ganglia 40. Rat DRG explants have also been used to explore both endothelial cell–nerve and macrophage–nerve crosstalk to explore the mechanisms underlying endometriosis‐associated pain 15, 38, whereas dissociated rat DRG neurons have been used in gene expression studies 15, 38. Human embryonic stem cells can be rapidly converted into sensory neurons with a nociceptor‐like phenotype using small molecule inhibitors 41, this technique holds great potential as a platform for exploring peripheral nerve mechanisms in endometriosis. Recently, this model has been used to explore ER‐mediated expression of nociceptive ion channels 42.

rapidly converted into sensory neurons with a nociceptor‐like phenotype using small molecule inhibitors 41, this technique holds great potential as a platform for exploring peripheral nerve mechanisms in endometriosis. Recently, this model has been used to explore ER‐mediated expression of nociceptive ion channels 42. Endothelial cells Ectopic endometrial fragments must acquire a vasculature to survive and form a lesion; moreover, invading immune cells extravasate from the bloodstream into vascularized lesions to contribute to the inflammatory environment of the lesion, hence neovascularization/angiogenesis are an important process in the etiology of endometriosis. Anti‐angiogenic drugs are being explored as a potential therapy for endometriosis and endothelial cells are used as a model in endometriosis research. To our knowledge there are no studies documenting the isolation and propagation of endothelial cells from endometriotic lesions. Endothelial cells can be isolated from the human endometrium 43 and have been immortalized and shown to retain their phenotype in culture 44. These human endometrial endothelial cells (HEECs) have been used in endometriosis research exploring the crosstalk between endothelial cells and nerve fibers; conditioned media from rat DRG can enhance network formation by HEECs 15. Human umbilical vein endothelial cells (HUVECs) are also often used as an endothelial cell model for angiogenesis research. However, endothelial cells are known to have vascular‐bed‐specific responses and HUVECs are isolated from macrovessels compared with microvessels like the endometrial endothelial cells. To this end, HUVECs express several extracellular matrix proteins not expressed in HEECs 43.

ial cell model for angiogenesis research. However, endothelial cells are known to have vascular‐bed‐specific responses and HUVECs are isolated from macrovessels compared with microvessels like the endometrial endothelial cells. To this end, HUVECs express several extracellular matrix proteins not expressed in HEECs 43. Three‐dimensional in vitro models in endometriosis research Co‐culture models can be used to assess the effect of one cell type on another 45. Three‐dimensional (3D) cell culture models are reported to mirror phenotype and gene expression in vivo more closely than 2D culture systems, moreover 3D culture offers an opportunity to explore the interactions between different cells relevant to endometriosis research. Ex vivo culture of human endometrial tissue in a 3D fibrin matrix has been shown to mimic the early stages of endometriosis invasion, gland and stroma formation and sprouting of new vessels 46. Recent advances in microfabrication technologies have enabled researchers to more closely mirror physiological interactions and microenvironment of human tissue. “Organs‐on‐Chips” 47 use a microfluidic system to reconstitute the organ architecture on the chip. The major cellular constituents of the endometrium are isolated and then reassembled in microfluidic devices that are fabricated with biocompatible materials and hydrogels. These devices individually compartmentalize each cell type in independent chambers and allow imaging. Furthermore, “organ” to “organ” communication can be assessed by connecting two “organ‐on‐chips”; in the review the authors suggest that placing a “liver‐on‐chip” upstream of an “endometrium‐on‐chip” has the potential to assess the effect of metabolites on endometrial function 48. This approach could potentially revolutionize our in vitro studies and allow identification of novel targets and testing of potential therapeutics in endometriosis without the need for animal models of disease.

of an “endometrium‐on‐chip” has the potential to assess the effect of metabolites on endometrial function 48. This approach could potentially revolutionize our in vitro studies and allow identification of novel targets and testing of potential therapeutics in endometriosis without the need for animal models of disease. Animal models In vivo models can be used to investigate the effect of the presence of endometriosis lesions on functional outcomes such as fertility or pain that requires behavior analysis. Moreover, in vivo models allow the preclinical testing of potential therapies that are required for us to move towards anxiously awaited new therapeutic options. Spontaneous endometriosis occurs only in humans and some primates such as rhesus monkeys and baboons 49, that have menstrual cycles. Non‐human primates offer the most physiologically relevant animal model of endometriosis in terms of phylogeny and reproductive anatomy but their uses are limited due to cost and ethical concerns. The study of endometriosis in the Baboon is an attractive option for a number of reasons, including noninvasive cycle monitoring, continuous breeding, collection of naturally occurring peritoneal fluid, spontaneous retrograde menstruation and human‐like minimal to severe endometriosis. Research in baboons has been pivotal in our understanding of some key pathophysiological features of endometriosis, which have been reviewed elsewhere 50. Rodent models of endometriosis have been developed and as they offer a more tangible laboratory model of the disease they will be the focus of this section of the review.

rch in baboons has been pivotal in our understanding of some key pathophysiological features of endometriosis, which have been reviewed elsewhere 50. Rodent models of endometriosis have been developed and as they offer a more tangible laboratory model of the disease they will be the focus of this section of the review. Rat The rat autologous model as developed by Vernon and Wilson in 1985 51 uses intact animals allowing the exploration of the effects of endometriosis on fertility and fecundity and has been used historically to demonstrate that ectopic endometrium is associated with an increased frequency of luteinized unruptured follicles and altered follicular development 52. This model was modified slightly by Berkley et al. 53, who sutured small pieces of uterus not only to the mesenteric cascade but also onto the abdomen and ovary to more closely mirror the distribution of lesions in women. Berkley's group was the first to use the rat model to explore the association between endometriosis and increased pelvic nociception and demonstrated that the rats had vaginal hyperalgesia in‐line with viscera–visceral referred hyperalgesia that could be suggestive of altered pain responses at the central nervous system level 53. This study preceded a number of pivotal publications that broke new ground on our understanding of pain mechanisms in endometriosis; the group were the first to demonstrate that lesions are actively innervated 54, they demonstrated that endometriosis influences pain behaviors induced by ureteral calculosis 55 again exploring the phenomenon of “viscera–visceral” hyperalgesia, and that estradiol levels modulate endometriosis‐induced vaginal nociception 56, 57, innervation, vascularization and growth factor content of lesions 58. Much of our understanding of pain mechanisms in endometriosis has stemmed from these publications.

again exploring the phenomenon of “viscera–visceral” hyperalgesia, and that estradiol levels modulate endometriosis‐induced vaginal nociception 56, 57, innervation, vascularization and growth factor content of lesions 58. Much of our understanding of pain mechanisms in endometriosis has stemmed from these publications. In a refinement of the autotransplant model, MRI was used to serially and noninvasively monitor lesion growth and to obtain lesion volumes allowing for a reduction in animal usage 59. Subsequently, the rat model has been used in many studies as a test‐bed for potential therapeutics including melatonin 60, doxycycline 61, Etanercept (tumor necrosis factor α antibody) 62, and gene therapy delivered via polymeric micelles 63. The natural, mixed dopamine receptor antagonist Levo‐tetrahydropalmatine (L‐THP) was shown to reduce lesion growth and to alleviate generalized hyperalgesia 64. This list is certainly not exhaustive and we apologize to authors of the studies we cannot cite due to reference limitations. Zhao et al. also demonstrated an up‐regulation of gene expression in dorsal root ganglia that was associated with increased nociception and this was attenuated by L‐THP treatment 64. The rat model has also been used to demonstrate that tissue inhibitor of matrix metalloproteinase 1 (TIMP1) secreted from ectopic uterine tissue explants resulted in poorer embryo quality and deleterious effects on ovulation 65 and the authors suggested that novel TIMP‐1‐modulating therapies may be developed to alleviate infertility in women with endometriosis.

rate that tissue inhibitor of matrix metalloproteinase 1 (TIMP1) secreted from ectopic uterine tissue explants resulted in poorer embryo quality and deleterious effects on ovulation 65 and the authors suggested that novel TIMP‐1‐modulating therapies may be developed to alleviate infertility in women with endometriosis. Recently, the rat has been used to explore pain mechanisms using innovative techniques. Dmitrieva et al. used telemetric assessment to record visceromotor responses induced by vaginal distention, allowing observer‐independent recording 66. In a unique take on the rat model, Alvarez et al. grafted autologous uterine tissue onto the gastrocnemius muscle allowing for in vivo electrophysiological recordings from sensory neurons innervating the graft and the exploration of agents injected directly into the graft 67. Uterine fragments have also been grafted onto the sciatic nerve to mimic neuropathic pain that might arise in endometriosis 68. Although these studies allow in‐depth mechanistic studies on innervating peripheral nerves, this model lacks authentic interactions between peritoneum and ectopic endometrium and does not allow physiologically relevant exploration of spinal cord engagement and vicera–visceral hyperalegsia resulting from convergent neurons. For example, using the Vernon and Wilson model, Chen et al. elegantly showed pelvic organ cross‐sensitization mediated via p38 in the rostral ventromedial medulla of the spinal cord 69.

low physiologically relevant exploration of spinal cord engagement and vicera–visceral hyperalegsia resulting from convergent neurons. For example, using the Vernon and Wilson model, Chen et al. elegantly showed pelvic organ cross‐sensitization mediated via p38 in the rostral ventromedial medulla of the spinal cord 69. A number of studies evaluating the validity of the rat as a model of endometriosis have analyzed the gene expression profiles of ectopic tissue deposits in the rat with the aim of comparing them to endometriosis lesions in women. Many common pathways were identified including the inflammatory response, angiogenesis, extracellular matrix re‐modeling and wound healing 70, 71, 72. Although the rat model replicates certain aspects of the disease, all of the modifications discussed above rely on the suturing of uterine fragments (endometrium plus myometrium) onto different sites and do not truly simulate dissemination of tissue into the peritoneum and formation of lesions from shed endometrial tissue. One study explored the use of fibrin glue and found less local inflammation and fewer adhesions at the site of the graft 73 but again this study used full thickness uterine fragments. An obvious limitation of using the rat model is the low number of commercially available animals engineered to have genetic manipulations. In this respect mouse models clearly have an advantage over the rat model.

nd fewer adhesions at the site of the graft 73 but again this study used full thickness uterine fragments. An obvious limitation of using the rat model is the low number of commercially available animals engineered to have genetic manipulations. In this respect mouse models clearly have an advantage over the rat model. Mouse Since the publication of the initial autologous mouse model of endometriosis 74, which was based on the rat model described by Vernon and Wilson, many adaptations and refinements have been published (summarized in Table 2). Using mice to model endometriosis has the added benefit of the wide availability of genetically engineered animals that express ubiquitous or cell‐specific fluorescent proteins to track cells or monitor explants as well as conditional or constitutive gene deletions to aid in studies investigating gene function. Table 2 Comparison of different endometriosis mouse models

Mouse Since the publication of the initial autologous mouse model of endometriosis 74, which was based on the rat model described by Vernon and Wilson, many adaptations and refinements have been published (summarized in Table 2). Using mice to model endometriosis has the added benefit of the wide availability of genetically engineered animals that express ubiquitous or cell‐specific fluorescent proteins to track cells or monitor explants as well as conditional or constitutive gene deletions to aid in studies investigating gene function. Table 2 Comparison of different endometriosis mouse models Model Benefits Limitations Heterologous (intact tissue) Humanized mouse model of endometriosis. Human tissue can be manipulated before xenografting Uses immunodeficient mice – cannot analyze full immune cell complement. Usually grafted subcutaneously – does not mirror authentic endometrium–peritoneum interactions Heterologous (human endometriosis cell lines) Bypasses problems in accessing tissue and variability of patient material. Used for pain studies Immunodeficient mice. Not a true recapitulation of an endometriosis lesion Autologous No rejection response and good for analysis of immune cell contribution Does not allow analysis of host/donor contribution Syngeneic (suturing tissue to peritoneal lining) Easy to localize lesions and measure regression in drug treatment studies. Induced in intact mice – useful for studies on fertility. Genetic manipulation of donor or host Suturing induces an inflammatory response Syngeneic (injection, whole uterine fragments) Immunocompetent recipient mice. Genetic manipulation of donor or host Uses ovariectomy and supraphysiological levels of estradiol. Difficult to localize all lesions unless reporter mice used as donors or labeling of tissue. Both myometrium and endometrium injected Syngeneic (injection of ‘menstrual’ material) Lesions phenocopy those recovered from women. Mirrors process of retrograde menstruation. Mice exhibit changes in sensory behavior and molecular changes in nervous system Uses ovariectomy and supraphysiological levels of estradiol. Difficult to localize all lesions unless reporter mice used as donors or labeling of tissue John Wiley & Sons, LtdHuman tissue xenografts The availability of mice homozygous for the Prkdc SCID (severe combined immunodeficient mice) or mice with a mutation in the forkhead box protein N1 (FOXN1) gene (Nude mice; athymic with greatly reduced number of T cells) allows human endometrium to be xenografted either subcutaneously 75 or intraperitoneally 76 to generate humanized models of endometriosis because the mice do not mount a rejection response.

e) or mice with a mutation in the forkhead box protein N1 (FOXN1) gene (Nude mice; athymic with greatly reduced number of T cells) allows human endometrium to be xenografted either subcutaneously 75 or intraperitoneally 76 to generate humanized models of endometriosis because the mice do not mount a rejection response. This approach is very useful for determining in vivo how a “human” lesion would respond to a treatment and for elegantly determining cells originating from endometrial tissue vs. peritoneum/mesothelium 77. Because these animals are immunocompromised, injection of endometrial tissue into the peritoneum results in the formation of lesions and the tissue does not require suturing. Human endometrial tissue can also be manipulated in vitro before xenografting into the host 78. An alternative heterologous model uses human immortalized endometriosis epithelial and stromal cells that are re‐suspended in matrigel and xenografted intraperitoneally into ovariectomized, estradiol‐supplemented nude mice 79. However, lesions that form using this model cannot truly recapitulate the histological appearance of typical endometriosis lesions because they are not intact tissue fragments. These heterologous models limit the analysis of the host immune response to ectopic tissue; the inflammatory response being a key area for exploration of the pathophysiology of endometriosis.

ot truly recapitulate the histological appearance of typical endometriosis lesions because they are not intact tissue fragments. These heterologous models limit the analysis of the host immune response to ectopic tissue; the inflammatory response being a key area for exploration of the pathophysiology of endometriosis. Syngeneic mouse models Syngeneic models of endometriosis bypass the limitation of the heterologous models and allow full analysis of the immune system. It is still unknown if the immunological perturbation observed in women with endometriosis is a fundamental defect of the immune system or a consequence of endometrial tissue at ectopic sites. Using a syngeneic model, induction of endometriosis was shown to inhibit spleen leucocyte function 80, which has clear implications for immunological therapies.

logical perturbation observed in women with endometriosis is a fundamental defect of the immune system or a consequence of endometrial tissue at ectopic sites. Using a syngeneic model, induction of endometriosis was shown to inhibit spleen leucocyte function 80, which has clear implications for immunological therapies. A number of variations of the syngeneic model exist; donor uterine tissue (full thickness or endometrial fragments only) are sutured onto the peritoneal lining or injected into the peritoneal space using a syringe or small laparotomic incision 81. Donor material is usually collected at estrous or following pregnant mare serum gonadotropin 82 or estradiol priming 83. A recently published study from our group used mouse “menstrual” donor endometrial tissue to inoculate syngeneic recipients 84. “Menstrual” endometrium was generated by hormonal manipulation using subcutaneous estradiol injections and a progesterone implant plus a decidualization stimulus to mimic the human menstrual cycle 85. Following progesterone withdrawal (4 h) decidualized bleeding endometrium was recovered and used to inoculate ovariectomized and estradiol‐supplemented recipients. Lesions were recovered after 21 days that mirrored human lesions in histological appearance, vascularization, innervation and influx of macrophages 15, 38, 84. More recently we have used our model to demonstrate that mice with endometriosis lesions exhibit robust changes in sensory behavior (both spontaneous and evoked) and elevated cyclooxygenase‐2 expression in the spinal cord and brain indicative of central sensitization (Greaves et al., 2017, accepted). The model can be used for preclinical testing of potential therapies for targeting endometriosis‐associated pain; we have demonstrated that a highly selective prostaglandin E2 receptor (EP2) antagonist could significantly attenuate both abdominal and secondary hyperalgesia.

nsitization (Greaves et al., 2017, accepted). The model can be used for preclinical testing of potential therapies for targeting endometriosis‐associated pain; we have demonstrated that a highly selective prostaglandin E2 receptor (EP2) antagonist could significantly attenuate both abdominal and secondary hyperalgesia. Interestingly, the spiny mouse (Acomys cahirinus) was recently found to undergo spontaneous decidualization, endometrial shedding and bleeding, demonstrating for the first time menstruation in a rodent 86. Whether this mouse spontaneously develops endometriosis and its use in endometriosis research is still to be determined. A number of studies describe an improvement on versions of the mouse model of endometriosis to allow non‐invasive monitoring of lesions. In one such study, human endometrial fragments were transduced with green fluorescent protein, transplanted into nude mice and imaged through the skin. However, because the adenoviral expression vector is transient the fluorescence faded by the third week so longitudinal studies are limited 87. Another study used donor mice engineered to ubiquitously express luciferase and uterine fragments were sutured to the peritoneal wall of non‐luminescent recipient mice. Anesthetized mice were injected with luciferin (intravenously or intraperitoneally) and bioluminescence was imaged. This approach was used successfully to monitor the efficacy of antiangiogenic therapy 88. The ability for non‐invasive monitoring of lesions is very desirable for preclinical models of endometriosis.

ice. Anesthetized mice were injected with luciferin (intravenously or intraperitoneally) and bioluminescence was imaged. This approach was used successfully to monitor the efficacy of antiangiogenic therapy 88. The ability for non‐invasive monitoring of lesions is very desirable for preclinical models of endometriosis. Spontaneous model of endometriosis A spontaneous model of ovarian endometriosis has also been described with induced expression of oncogenic K‐ras (V‐Ki‐ras2 Kirsten rat sarcoma viral oncogene homolog) in ovarian surface epithelium resulting in benign epithelial lesions on the ovary that exhibit a simple endometrioid glandular structure; however, no associated stroma was observed 89. This activated K‐ras model was modified by xenografting “menstrual” endometrium expressing activated K‐ras into a subcutaneous pocket on the ventral abdomen in intact immunocompetent recipients 90.

esions on the ovary that exhibit a simple endometrioid glandular structure; however, no associated stroma was observed 89. This activated K‐ras model was modified by xenografting “menstrual” endometrium expressing activated K‐ras into a subcutaneous pocket on the ventral abdomen in intact immunocompetent recipients 90. Many of the published mouse models use ovariectomized mice supplemented with estradiol as this avoids the natural variations in estradiol in mice during the estrous cycle, and uniform availability of estradiol (usually at higher concentrations than in intact mice) is believed to promote lesion establishment and growth. This does, however, prevent the analysis of how endometriosis affects fertility. Endometriosis can be induced in intact mice, but this model uses full thickness uterine fragments sutured to the peritoneal wall. The intact model has been used to demonstrate that endometriosis is associated with decreased oocyte and embryo quality 91 and is associated with reduced pregnancy rate 92. These studies merit further exploration of the link between fertility and ectopic tissue.

ickness uterine fragments sutured to the peritoneal wall. The intact model has been used to demonstrate that endometriosis is associated with decreased oocyte and embryo quality 91 and is associated with reduced pregnancy rate 92. These studies merit further exploration of the link between fertility and ectopic tissue. Summary The etiology of endometriosis is complex. The availability of human tissue for endometriosis research has provided informative descriptive data on the disorder, and the adoption of internationally agreed‐on standard operating procedures for tissue and fluid sample collection (the EPHect initiative) will significantly advance this research approach. An additional layer of complexity is disease‐associated epigenetic regulation and studies should be undertaken on the epigenetic component of endometriosis and chronic pain susceptibility. Genomic approaches and next‐generation sequencing are now more affordable than ever, making these large‐scale studies more achievable. Model systems are required to explore disease‐specific mechanisms and also to validate genomic pathways discovered using GWAS and other genomic approaches. The advent and availability of different models for use in endometriosis research has considerably improved in recent years; however it is important that the research question be carefully considered before selection of the most relevant models. The physiologically relevant menstruating mouse model of endometriosis also exhibits robust changes in sensory behavior and molecular alterations in the central nervous system highlighting the model as particularly promising for the study of pain mechanisms in endometriosis, which still remain largely unknown. Reversal of the endometriosis‐associated pain state has also been demonstrated with a highly specific EP2 receptor antagonist indicating that this model is an ideal preclinical platform for testing potential therapeutics for endometriosis‐associated pain. Syngeneic mouse models of endometriosis can be used to easily explore cellular contributions of donor/host tissue and genetic manipulation can be used to enhance this. The syngeneic model also has the added benefit of a full immune system, so they can be used to explore this important aspect of endometriosis pathophysiology. Many of the published mouse models of endometriosis can be further improved by adapting them to allow non‐invasive in vivo monitoring of lesion size.

hance this. The syngeneic model also has the added benefit of a full immune system, so they can be used to explore this important aspect of endometriosis pathophysiology. Many of the published mouse models of endometriosis can be further improved by adapting them to allow non‐invasive in vivo monitoring of lesion size. This review has highlighted the use of many different cell types and models in use for endometriosis research. Wherever possible, primary cells should be isolated from patient samples and sample sizes should be selected to achieve appropriate power. When this is not possible, models should be chosen with care. Given the heterogeneity of patient endometriosis samples, data generated using cell lines derived from endometriosis tissue must be interpreted with caution (generated from one patient). New microfluidic techniques for 3D cell culture hold particular promise for enhancing in vitro studies. The real power of these models can be seen when ex vivo human tissue, in vitro and in vivo techniques are used in combination to produce high‐quality, clinically relevant data that advances our understanding and identifies possible future therapeutic targets.

Abbreviations IUCintrauterine contraception LNG‐IUSlevonorgestrel‐releasing intrauterine system LPlidocaine‐prilocaine NSAIDnonsteroidal anti‐inflammatory drug RCTrandomized clinical trial VASvisual analogue scale Key message Analgesia and cervical priming can be effective in reducing intrauterine contraception placement‐related pain when compared with placebo, but routine use remains a subject for debate. Targeted use of effective strategies in women identified as being at greater risk of experiencing pain may be a useful approach. 1 INTRODUCTION Intrauterine contraceptives (IUC) are highly effective and safe methods of pregnancy prevention.1 Greater awareness and use could help to reduce the incidence of unplanned or mistimed pregnancy.2, 3 The US‐based Contraceptive CHOICE Project reported that provision of accurate, unbiased counseling on long‐acting reversible contraceptive methods, and removal of cost barriers, led to the majority of the women choosing IUC.2 A subsequent review of the impact of the CHOICE project showed that use of long‐acting reversible contraceptive methods led to higher rates of continued contraceptive use and a reduction in the average annual rates of pregnancy, birth and abortion among teenage participants.3

to the majority of the women choosing IUC.2 A subsequent review of the impact of the CHOICE project showed that use of long‐acting reversible contraceptive methods led to higher rates of continued contraceptive use and a reduction in the average annual rates of pregnancy, birth and abortion among teenage participants.3 For most women, pain experienced with IUC placement is mild to moderate and less than anticipated.4, 5 However, some women remain anxious about the possibility of pain or are more likely to be affected by factors such as nulliparity, or a long time period since delivery. In addition, anatomical, cultural or psychological elements can contribute to a more painful experience.6, 7 Fear of pain at the time of placement can therefore be a barrier to choosing IUC.6, 8, 9 In a survey of pain and discomfort, both at the time of IUC placement and as a recollection, experienced by parous and nulliparous women participating in a UK‐based contraceptive service, Murty9 found that women anticipating pain were more anxious and more likely to take analgesia before placement. Although their pain scores during the procedure were similar to those of women who had not taken analgesia, their recollection of the pain experienced when asked 6 months later was greater than they reported immediately post‐placement. Healthcare professional concerns about difficult and/or painful placement may also discourage discussion of IUC as a contraceptive option and lead to the counseling of women on other, less effective methods.10

n of the pain experienced when asked 6 months later was greater than they reported immediately post‐placement. Healthcare professional concerns about difficult and/or painful placement may also discourage discussion of IUC as a contraceptive option and lead to the counseling of women on other, less effective methods.10 A literature review to evaluate the evidence for strategies to minimize pain experienced during IUC placement, carried out in 2012, led to a consensus that no prophylactic pharmacological intervention had been adequately studied to support its routine use.11 Furthermore, in a Cochrane Review of interventions to minimize pain associated with IUC placement, the authors concluded that some oral analgesics and lidocaine formulations are effective in reducing placement‐related pain in specific groups but that most of the evidence came from single trials and were of moderate quality.8 We undertook an updated review to determine whether there was any new evidence for pharmacological interventions to minimize pain associated with IUC placement, to identify whether nonpharmacological or procedure‐related interventions could prove helpful. We also set out to identify any factors that may assist in predicting the women most likely to experience pain.

ther there was any new evidence for pharmacological interventions to minimize pain associated with IUC placement, to identify whether nonpharmacological or procedure‐related interventions could prove helpful. We also set out to identify any factors that may assist in predicting the women most likely to experience pain. 2 MATERIAL AND METHODS The broader objectives of this review, together with the diversity of populations and interventions and lack of consistent and validated assessment of pain experience, limits the possibility of a systematic review or meta‐analysis.12 Consequently, we opted to undertake a narrative review, an approach that is based on systematic methodologies, employs a bibliographic research strategy,13, 14 and looks at the evidence contributing to the clinical concept of pain relief during IUC placement.

s the possibility of a systematic review or meta‐analysis.12 Consequently, we opted to undertake a narrative review, an approach that is based on systematic methodologies, employs a bibliographic research strategy,13, 14 and looks at the evidence contributing to the clinical concept of pain relief during IUC placement. We searched the PubMed and Cochrane databases to update and extend the findings of our original review.11 We conducted a search for publications in any language that reported pharmacological interventions to reduce pain with IUC placement using the terms “intrauterine contraception” AND “insertion” AND “pain” published from December 2012 to September 2018. We then conducted a search of publications from January 1980 to September 2018, the timeframe of the original and our updated review combined, to identify other approaches that may influence pain experience or predict the experience of pain. We used a combination of text and MeSH terms: “intrauterine contraception” OR “levonorgestrel‐releasing intrauterine system” OR “LNG‐IUS” OR “IUD” AND “pain” OR “anxiety” OR “fear” OR “counseling” OR “insertion” OR “placement” OR “initiation” OR “cervical priming” OR “cervical ripening.” The search was not limited to randomized controlled trials and results were cross‐referenced and duplicate publications were removed.

g intrauterine system” OR “LNG‐IUS” OR “IUD” AND “pain” OR “anxiety” OR “fear” OR “counseling” OR “insertion” OR “placement” OR “initiation” OR “cervical priming” OR “cervical ripening.” The search was not limited to randomized controlled trials and results were cross‐referenced and duplicate publications were removed. The search identified 550 publications (Figure 1). Those not relevant to pain management either before or after IUC placement; those included in the previously published review11; and those reporting the findings of reviews were excluded. The final number of publications included in our review was 43. We assessed these publications for information relating to the effect of pharmacological interventions (pre‐insertion oral or local analgesia, cervical priming, post‐insertion analgesia), nonpharmacological strategies or procedure‐related factors on the pain experienced with IUC placement. We also assessed publications for factors that may help to predict the likelihood of experiencing pain with placement. Assessment of risk of bias of R randomized clinical trials (RCTs) included in the review was carried out according to the Cochrane Collaboration Handbook for Systematic Reviews of Interventions.15 As the intention was to conduct a narrative review rather than a systematic review the assessment of bias was limited to RCTs. No non‐English publications were identified for inclusion. Figure 1 Literature selection process for review

The search identified 550 publications (Figure 1). Those not relevant to pain management either before or after IUC placement; those included in the previously published review11; and those reporting the findings of reviews were excluded. The final number of publications included in our review was 43. We assessed these publications for information relating to the effect of pharmacological interventions (pre‐insertion oral or local analgesia, cervical priming, post‐insertion analgesia), nonpharmacological strategies or procedure‐related factors on the pain experienced with IUC placement. We also assessed publications for factors that may help to predict the likelihood of experiencing pain with placement. Assessment of risk of bias of R randomized clinical trials (RCTs) included in the review was carried out according to the Cochrane Collaboration Handbook for Systematic Reviews of Interventions.15 As the intention was to conduct a narrative review rather than a systematic review the assessment of bias was limited to RCTs. No non‐English publications were identified for inclusion. Figure 1 Literature selection process for review 3 RESULTS Information regarding study participants, intervention, method of pain evaluation, comparison of pain scores during IUC placement between the intervention and placebo/control groups, and level of evidence from each study reporting pain‐relieving interventions is summarized in Table 1. The clinical relevance and statistical significance of differences in pain scores between the intervention and placebo/control groups are also noted. There is a degree of overlap across the study descriptions as some studies described multiple interventions relevant to more than one category.

is summarized in Table 1. The clinical relevance and statistical significance of differences in pain scores between the intervention and placebo/control groups are also noted. There is a degree of overlap across the study descriptions as some studies described multiple interventions relevant to more than one category. Table 1 Studies of pre‐placement pharmacological interventions for reduction of pain associated with IUC placement Author(s) N Population Type of study Interventions Method of evaluation of paina Effect of intervention vs placebo/control on pain score during IUC placement procedure Level of evidencec Pre‐insertion oral analgesia Crawford et al16 72 Mainly parous RCT 20 mg ketorolac vs placebo Multiple time points; 0‐10 cm VAS Reduction in mean pain scores vs placebo at IUC placement (4.2 vs 5.7 cm, P = 0.031). Reduction in overall pain (3.6 vs 4.9 cm, P = 0.047) and pain 10 min post‐procedure (1.1 vs 2.5, P = 0.007) 2 Ngo et al17 118 Mainly nulliparous RCT 550 mg naproxen sodium vs placebo Multiple time points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (69 vs 66 mm, P = 0.89). Reduction in median pain scores at 5 min (17 vs 26.0 mm, P = 0.01) and at 15 min (13 vs 24.0 mm, P = 0.01) post‐placement.

Reduction in overall pain (3.6 vs 4.9 cm, P = 0.047) and pain 10 min post‐procedure (1.1 vs 2.5, P = 0.007) 2 Ngo et al17 118 Mainly nulliparous RCT 550 mg naproxen sodium vs placebo Multiple time points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (69 vs 66 mm, P = 0.89). Reduction in median pain scores at 5 min (17 vs 26.0 mm, P = 0.01) and at 15 min (13 vs 24.0 mm, P = 0.01) post‐placement. 2 Singh et al18 80 Nulliparous RCT N2O/O2 vs no pain relief Measured at time of placement on 0‐100 mm VAS No difference in mean pain score vs control at IUC placement (54.3 ± 24.8 mm vs 55.3 ± 20.9 mm, P = 0.86) 2 Bednarek et al19 202 Nulliparous and parous RCT Ibuprofen vs no pain relief Multiple time points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (38.0 mm vs 41.5 mm, P = 0.5) 2 Castro et al20 100 Only cesarean delivery and nulligravidas RCT NSAID (400 mg ibuprofen) vs 2% lidocaine intracervical injection Multiple time points; 0‐100 mm VAS No difference in mean pain score vs lidocaine injection at IUC placement (48.1 ± 27.5 mm vs 44 ± 24.5 mm, P = 0.4) 2 Pre‐insertion cervical priming Maged et al21 120 Cesarean delivery RCT 600 μg misoprostol vs placebo Measured at time of placement on 0‐10 cm VAS Reduction in mean pain score vs placebo at IUC placement (5.7 ± 1.4 cm vs 6.5 ± 0.9 cm) 2 Espey et al23 82 Nulliparous RCT 400 μg misoprostol vs placebo Measured at time of placement on 0‐10 cm VAS No difference in mean pain score vs placebo at IUC placement (5.8 ± 2.0 cm vs 5.9 ± 2.0 cm, P = 0.94) 2 Abdellah et al24 140 Cesarean delivery RCT 800 μg misoprostol vs placebo Measured at time of placement on 0‐10 cm VAS Reduction in mean pain score vs placebo at IUC placement (2.7 ± 0.6 cm vs 4.3 ± 0.8 cm, P = 0.001) 2 Scavuzzi et al25 179 Nulliparous RCT 400 μg misoprostol vs placebo Multiple time‐points; VAS 0‐10 cm (analyzed as absent or mild [0‐5] vs moderate or severe [6‐10] Reduction in moderate or severe pain score vs placebo at IUC placement (OR 0.30, 95% CI 0.16‐0.55) 2 Lathrop et al26 73 Nulliparous RCT 400 μg misoprostol vs placebo Multiple time points; 0‐100 mm VAS Increase in mean pain score vs placebo immediately prior (10.84 vs 2.11 mm; P = 0.005) and post‐placement (46.5 vs 35.4 mm, P = 0.04) 2 Pre‐insertion local anesthesia Mody et al27 64 Nulliparous RCT 20‐mL 1% buffered lidocaine paracervical block vs placebo Multiple time‐points; 0‐100 mm VAS Reduction in median pain score vs placebo at IUC placement (33 vs 54 mm, P = 0.002)b.

(10.84 vs 2.11 mm; P = 0.005) and post‐placement (46.5 vs 35.4 mm, P = 0.04) 2 Pre‐insertion local anesthesia Mody et al27 64 Nulliparous RCT 20‐mL 1% buffered lidocaine paracervical block vs placebo Multiple time‐points; 0‐100 mm VAS Reduction in median pain score vs placebo at IUC placement (33 vs 54 mm, P = 0.002)b. Reduction in median pain scores at other time‐points: uterine sounding (30 vs 47 mm, P = 0.005); 5 min after placement (12 vs 27 mm, P = 0.005); overall pain perception (30 vs 51 mm, P = 0.015). Increased pain score with block administration vs placebo (30 vs 8 mm, P = 0.003) Conti et al28 220 Nulliparous and parous RCT 20 mL of 2% lidocaine gel self‐administered ≤ 15 min prior to IUC placement vs placebo gel Multiple time‐points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (65 [1‐99] mm vs 59 [5‐100] mm, P = 0.09). Reduction in median pain scores at speculum placement vs placebo (7 (0‐81) mm vs 11 (0‐80) mm, P = 0.046) Abd Ellah et al29 48 Nulliparous and parous RCT Self‐inserted novel lidocaine dual‐response in situ gel vs placebo Multiple time‐points; 0‐10 cm VAS Reduction in mean pain scores vs placebo at IUC placement (median [IQR]:3[2‐3.75] cm vs 6[5.5‐7] cm, P = 0.0001)b. Reduction in median pain scores at other time‐points: tenaculum placement (median [IQR]: 2[1‐2] cm vs 4[3‐4] cm, P = 0.0001); uterine sound insertion (median[IQR]: 3[2‐3] cm vs 5[4‐6] cm, P = 0.0001)

0, 31, 32, 34, 35, 36, 37 The majority of these interventions involve pre‐insertion local anesthesia with lidocaine preparations. There is a need, however, for standardization of doses of these local analgesic agents, not only to optimize pain relief but also to maximize safety by preventing spread to vaginal tissue.31 The studies described reinforce the message that pain with IUC placement is not confined to the insertion of the device; use of the tenaculum and the uterine sound can also contribute to an uncomfortable experience. Assessment of post‐placement pain was limited in the studies we reviewed yet this may be important to consider for two reasons. First, although lack of time may limit the achievement of optimal effect of oral or local analgesia for IUC placement, these agents may help in the relief of post‐placement pain. Second, post‐procedural pain may contribute to the memory of a painful experience and lead to the sharing of negative views when talking to other women about IUC.

Abd Ellah et al29 48 Nulliparous and parous RCT Self‐inserted novel lidocaine dual‐response in situ gel vs placebo Multiple time‐points; 0‐10 cm VAS Reduction in mean pain scores vs placebo at IUC placement (median [IQR]:3[2‐3.75] cm vs 6[5.5‐7] cm, P = 0.0001)b. Reduction in median pain scores at other time‐points: tenaculum placement (median [IQR]: 2[1‐2] cm vs 4[3‐4] cm, P = 0.0001); uterine sound insertion (median[IQR]: 3[2‐3] cm vs 5[4‐6] cm, P = 0.0001) 2 Akers et al30 95 Nulliparous RCT 1% lidocaine paracervical block vs a sham control (1 cm depression of the vaginal epithelium at paracervical block sites with a wooden cotton‐tipped applicator) Multiple time‐points; 0‐100 mm VAS Reduction in mean pain score vs sham block at IUC placement (30.0 [95% CI 20‐58] mm vs 71.5 [95% CI 66.0‐82.0] mm, P = 0.006)b. Analysis of the secondary outcomes found that the scores across all six VAS assessments (baseline, speculum, tenaculum, block, uterine sounding, IUC placement, and speculum removal were lower in the lidocaine block group compared with the sham block group (27.7 [95% CI 16.0‐40.2] vs 53.9 [95% CI 44.0‐57.8], P < 0.001) 2 Karasu et al31 151 Parous RCT Lidocaine spray (10%, net 40 mg) vs 2 g lidocaine cream vs 10 mL lidocaine injection (20 mg/mL) Tenaculum and IUC placement; 0‐10 cm VAS (categorized as none [0], mild [1‐3], moderate [4‐6], severe [7‐10]) Reduction in mean pain score vs LP cream and control (no anesthesia) at IUC placement (2.85 ± 2.5 cm vs 4.0 ± 1.7 cm vs 4.25 ± 1.9, respectively).

vs 2 g lidocaine cream vs 10 mL lidocaine injection (20 mg/mL) Tenaculum and IUC placement; 0‐10 cm VAS (categorized as none [0], mild [1‐3], moderate [4‐6], severe [7‐10]) Reduction in mean pain score vs LP cream and control (no anesthesia) at IUC placement (2.85 ± 2.5 cm vs 4.0 ± 1.7 cm vs 4.25 ± 1.9, respectively). Greater proportion of women experiencing no pain at IUC placement with spray vs cream, injection and control groups (25.5% vs 1.9% vs 2.1% vs 2.0%, respectively, P = 0.01) and at tenaculum placement (39.2% vs 7.5% vs 6.4% vs 4.1%, respectively, P = 0.01) 2 Abbas et al32 120 Parous RCT Lidocaine‐prilocaine cream (LP) vs placebo Multiple time‐points; 0‐10 cm VAS Reduction in median pain score vs placebo at IUC placement (3.0 vs 6.5 cm, P = 0.0001)b 2 Torky et al41 420 Parous Non‐randomized comparator study Lidocaine gel or lidocaine spray vs no anesthesia Multiple time‐points; 0‐10 cm VAS No difference in mean pain score at IUC placement: lidocaine gel vs lidocaine spray vs no anesthesia (4.9 ± 1.9 cm vs 4.6 ± 1.9 cm vs 5.9 ± 1.5 cm respectively, P = 0.059). Reduction in mean pain score during cervical traction with use of local anesthesia (4.5 ± 2.1 cm for lidocaine gel; 4.0 ± 2.2 cm for lidocaine spray vs 6.0 ± 1.5 cm for no anesthetic, P = 0.003)

2 Abbas et al32 120 Parous RCT Lidocaine‐prilocaine cream (LP) vs placebo Multiple time‐points; 0‐10 cm VAS Reduction in median pain score vs placebo at IUC placement (3.0 vs 6.5 cm, P = 0.0001)b 2 Torky et al41 420 Parous Non‐randomized comparator study Lidocaine gel or lidocaine spray vs no anesthesia Multiple time‐points; 0‐10 cm VAS No difference in mean pain score at IUC placement: lidocaine gel vs lidocaine spray vs no anesthesia (4.9 ± 1.9 cm vs 4.6 ± 1.9 cm vs 5.9 ± 1.5 cm respectively, P = 0.059). Reduction in mean pain score during cervical traction with use of local anesthesia (4.5 ± 2.1 cm for lidocaine gel; 4.0 ± 2.2 cm for lidocaine spray vs 6.0 ± 1.5 cm for no anesthetic, P = 0.003) 2 Elkhouly and Maherl21 200 Mostly nulliparous RCT 1% lidocaine intracervical block, misoprostol (400 μg), oral naproxen (500 mg) vs placebo Multiple time‐points; 0‐10 cm VAS No difference in mean pain score vs placebo at IUC placement with lidocaine block vs misoprostol, oral naproxen and placebo (5.3 ± 1.61 cm, 5.02 ± 1.12 cm, 4.9 ± 1.24 vs 5.16 ± 1.21 cm, respectively, P = 0.460) 2 Rapkin et al33 59 Nulliparous RCT Self‐administered 2% lidocaine gel vs placebo Multiple time‐points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (61 mm [IQR 53‐71]) vs (69 mm [IQR 63‐80], P = 0.06). Reduction vs placebo at tenaculum placement (32 [IQR 18‐54] mm vs 56 [IQR 26‐75] mm, P = 0.02)

2 Elkhouly and Maherl21 200 Mostly nulliparous RCT 1% lidocaine intracervical block, misoprostol (400 μg), oral naproxen (500 mg) vs placebo Multiple time‐points; 0‐10 cm VAS No difference in mean pain score vs placebo at IUC placement with lidocaine block vs misoprostol, oral naproxen and placebo (5.3 ± 1.61 cm, 5.02 ± 1.12 cm, 4.9 ± 1.24 vs 5.16 ± 1.21 cm, respectively, P = 0.460) 2 Rapkin et al33 59 Nulliparous RCT Self‐administered 2% lidocaine gel vs placebo Multiple time‐points; 0‐100 mm VAS No difference in median pain score vs placebo at IUC placement (61 mm [IQR 53‐71]) vs (69 mm [IQR 63‐80], P = 0.06). Reduction vs placebo at tenaculum placement (32 [IQR 18‐54] mm vs 56 [IQR 26‐75] mm, P = 0.02) 2 Fouda et al34 90 Parous RCT Diclofenac potassium (2 × 50 mg 1 h prior) plus application of 3 mL 2% lidocaine gel on the cervical lip and cotton swab soaked in 2% lidocaine gel in the cervical canal (3 min prior) vs placebo Multiple time‐points; 0‐10 cm VAS Slight reduction in mean pain score vs placebo (3.14 ± 0.92 cm vs 3.94 ± 1.3 cm, P = 0.001) 2 Aksoy et al35 200 Parous RCT Lidocaine spray (10%, net 40 mg) vs placebo Measured at time of placement on 0‐10 cm VAS Reduction in mean pain score vs placebo at IUC placement (1.01 ± 1.20 cm vs 3.23 ± 1.60 cm, P < 0.001)b. Decrease in the proportion of women scoring ≥ 4 on the pain‐VAS vs placebo (6% vs 41%, P < 0.001)

2 Fouda et al34 90 Parous RCT Diclofenac potassium (2 × 50 mg 1 h prior) plus application of 3 mL 2% lidocaine gel on the cervical lip and cotton swab soaked in 2% lidocaine gel in the cervical canal (3 min prior) vs placebo Multiple time‐points; 0‐10 cm VAS Slight reduction in mean pain score vs placebo (3.14 ± 0.92 cm vs 3.94 ± 1.3 cm, P = 0.001) 2 Aksoy et al35 200 Parous RCT Lidocaine spray (10%, net 40 mg) vs placebo Measured at time of placement on 0‐10 cm VAS Reduction in mean pain score vs placebo at IUC placement (1.01 ± 1.20 cm vs 3.23 ± 1.60 cm, P < 0.001)b. Decrease in the proportion of women scoring ≥ 4 on the pain‐VAS vs placebo (6% vs 41%, P < 0.001) 2 Tavakolian et al36 92 Parous RCT Lidocaine‐prilocaine cream (LP) vs placebo Multiple time‐points; 0‐10 cm VAS (categorized as none [0 points], mild [1‐3], average [4‐6] and severe [9], extremely severe [10] Reduction in mean pain score vs placebo at IUC placement (2.65 ± 2.53 cm vs 4.61 ± 2.55 cm, P < 0.001)b 2 Tornblom‐Paulander et al37 218 Nulliparous RCT Lidocaine (8.5 ml) of a novel formulation (SHACT) vs placebo Multiple time‐points; 0‐100 mm VAS (starting at 10 min post‐placement) Reduction in maximum pain score at IUC placement vs placebo at 10 min (28.3 ± 24.6 mm vs 44.2 ± 26.0 mm, P < 0.0001) 2 Micks et al38 24 Nulliparous RCT 0.5 mg nitroglycerin gel (1 mL) vs placebo gel Multiple time‐points; 0‐100 mm VAS No difference in mean pain scores vs placebo (55.0 ± 29.7 mm vs 57.4 ± 22.1 mm, P = 0.82) 2 Allen et al39 145 Mostly parous RCT 2% lidocaine gel vs placebo Measured at time of placement on 0‐10 cm VAS No difference in mean pain scores vs placebo (35.2 ± 27.7 mm vs 36.7 ± 30.0 mm, P = 0.8) 2 Nelson, Fong40 40 Nulliparous and parous RCT 1.2% lidocaine infusion (via endometrial aspirator) vs saline Multiple time‐points; 0‐9 scale (where 0 = no pain and 9 = worst pain imaginable) No difference in mean pain scores vs placebo (2.95 vs 3.75, P = 0.37) 2 VAS, visual analogue scale; CI, confidence interval; OR, odds ratio.

ong40 40 Nulliparous and parous RCT 1.2% lidocaine infusion (via endometrial aspirator) vs saline Multiple time‐points; 0‐9 scale (where 0 = no pain and 9 = worst pain imaginable) No difference in mean pain scores vs placebo (2.95 vs 3.75, P = 0.37) 2 VAS, visual analogue scale; CI, confidence interval; OR, odds ratio. a For studies in which pain was evaluated at different time‐points, the time‐points included one or more of the following: at speculum insertion, at tenaculum placement, during device placement, shortly after device placement (10 min) and longer intervals after placement (1 h, 2 h, 1 d, 2 d, and 3 d). b Difference in pain scores between intervention and control/placebo regarded as clinically relevant according to Olsen et al.61 c Oxford Center for Evidence‐Based Medicine60 levels of evidence: Level 1, systematic review of randomized trials; Level 2, randomized trial or observational study with dramatic effect; Level 3, non‐randomized controlled cohort/follow‐up study; Level 4, case series, case‐control studies or historically controlled studies; Level 5, mechanism‐based reasoning.

cine60 levels of evidence: Level 1, systematic review of randomized trials; Level 2, randomized trial or observational study with dramatic effect; Level 3, non‐randomized controlled cohort/follow‐up study; Level 4, case series, case‐control studies or historically controlled studies; Level 5, mechanism‐based reasoning. John Wiley & Sons, Ltd3.1 Characteristics and methods of included studies Among the reviewed publications, 26 described pre‐placement pharmacological interventions to minimize pain associated with IUC placement.16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 Of these, 25 were RCTs16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 and one was a nonrandomized comparator study.41 Eight publications—of which six described RCTs,42, 43, 44, 46, 47, 48 one was a pilot feasibility study45 and one was a pooled analysis49described nonpharmacological interventions. A further nine publications described factors related to the experience of pain with IUC placement.50, 51, 52, 53, 54, 55, 56, 57, 58 Of these, one was an RCT,50 three were non‐RCTs,51, 52, 53 three were prospective cohort studies,54, 55, 56 one was a case‐control study,57 and one was a secondary analysis of the US‐based Contraceptive CHOICE Project.58

bed factors related to the experience of pain with IUC placement.50, 51, 52, 53, 54, 55, 56, 57, 58 Of these, one was an RCT,50 three were non‐RCTs,51, 52, 53 three were prospective cohort studies,54, 55, 56 one was a case‐control study,57 and one was a secondary analysis of the US‐based Contraceptive CHOICE Project.58 All studies evaluating pharmacological interventions except for Nelson and Fong40 reported the use of a 10‐cm or 100‐mm visual analogue scale (VAS), where 0 is equivalent to “no pain” and 10 is equivalent to “worst pain ever”, by study participants to indicate the severity of pain experienced. Nevertheless, there were wide variations in the assessment of the experience of pain in terms of timing (IUC placement only); IUC placement plus other time‐points (speculum insertion, tenaculum placement, uterine sounding); post‐placement assessment (multiple time intervals); overall perception of pain.

in experienced. Nevertheless, there were wide variations in the assessment of the experience of pain in terms of timing (IUC placement only); IUC placement plus other time‐points (speculum insertion, tenaculum placement, uterine sounding); post‐placement assessment (multiple time intervals); overall perception of pain. 3.2 Pre‐insertion pharmacological therapy: Oral analgesia Five RCTs evaluated the effectiveness of oral analgesia on the experience of pain with IUC placement16, 17, 18, 19, 20 and one RCT compared multiple analgesic agents.21 One RCT, reported a 15‐mm reduction in mean pain score with oral ketorolac (20 mg) given 40‐60 minutes before IUC placement when compared with placebo.16 Although statistically significant, the authors suggested that the time required for the maximum analgesic effect of ketorolac (1‐2 hours after administration) may have affected the outcome. The effect of the intervention may have been more accurate, although less practical, if measured at a later time‐point.16 Four studies, evaluating 550 mg naproxen sodium,17 N2O/O2 18 and ibuprofen compared with placebo,19, 20 found no difference in mean pain scores at the time of IUC placement between the treatment and control groups; however, a reduction in median pain score was observed at 5 minutes (9 mm) and 15 minutes (11.2 mm) after IUC placement in the naproxen sodium group when compared with placebo.17

n compared with placebo,19, 20 found no difference in mean pain scores at the time of IUC placement between the treatment and control groups; however, a reduction in median pain score was observed at 5 minutes (9 mm) and 15 minutes (11.2 mm) after IUC placement in the naproxen sodium group when compared with placebo.17 3.3 Pre‐insertion pharmacological therapy: Cervical priming Three RCTs found pain scores with IUC placement to be lower following vaginal administration of misoprostol when compared with placebo (Table 2).22, 24, 25 Using cut‐offs of “absent or mild (VAS score of 0‐5 cm)” and “moderate to severe (VAS score of 6‐10 cm)”, Scavuzzi et al25 reported a reduction in moderate‐to‐severe pain with 400 μg misoprostol administered 4 hours before IUC placement in nulliparous women (risk ratio 0.56 [32/86 vs 62/93]; 95% confidence interval (CI) 0.41‐0.75; P = 0.00008). Two further studies found misoprostol (600 μg 6 hours prior and 800 μg 3 hours prior, respectively) reduced mean pain scores at placement by 0.8‐1.6 cm compared with placebo in women with a history of only elective cesarean delivery.22, 24 Although the 1.6‐cm reduction in pain score with misoprostol compared with placebo was statistically significant, Abdellah et al24 highlighted multiple factors limiting the validity of their study: pain minimization by women after a successful procedure, time taken for misoprostol to take effect, the lack of availability of the 52 mg levonorgestrel‐releasing intrauterine system (LNG‐IUS), and a high insertion failure rate.24 Furthermore, the frequency of abdominal cramping was increased in the misoprostol group in all three studies.22, 24, 25 Espey et al23 evaluated 400 μg of misoprostol administered 2‐8 hours before IUC placement in nulliparous women and found there to be no difference in pain scores at placement between the treatment and placebo groups. Lathrop et al26 found there to be an 11‐mm increase in pain experienced with misoprostol compared with placebo.

al23 evaluated 400 μg of misoprostol administered 2‐8 hours before IUC placement in nulliparous women and found there to be no difference in pain scores at placement between the treatment and placebo groups. Lathrop et al26 found there to be an 11‐mm increase in pain experienced with misoprostol compared with placebo. Table 2 Predictive factors for increased potential for painful IUC placement52, 53, 54, 56, 58, 62

al23 evaluated 400 μg of misoprostol administered 2‐8 hours before IUC placement in nulliparous women and found there to be no difference in pain scores at placement between the treatment and placebo groups. Lathrop et al26 found there to be an 11‐mm increase in pain experienced with misoprostol compared with placebo. Table 2 Predictive factors for increased potential for painful IUC placement52, 53, 54, 56, 58, 62 Physical factors detected during routine history or examination Psychological and sociocultural factors Low parity (1‐2 live births) Number of years in education (≥7) Longer interval between last birth and placement (>13 months) Presence of mood disorders Nonbreastfeeding at time of placement (irrespective of time since last birth) History of sexual trauma Presence of cervical resistance and pain Previous negative reaction to vaginal examination Uterine length Previous placement reported as painful Dysmenorrhea Awareness of the potential for pain from a friend or family member Multiple cesarean deliveries Anticipation or expectation of pain Menstruation (for nulligravidas) Age (adolescence) Difficulty or pain when using uterine sound Size of inserter John Wiley & Sons, Ltd3.4 Pre‐insertion local anesthesia A total of 16 studies, of which 14 were RCTs, evaluated the effect of pre‐insertion local anesthesia on pain experienced with IUC placement.27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 Interventions included lidocaine gel, lidocaine‐prilocaine (LP) cream, lidocaine intracervical block, lidocaine infusion, and nitroglycerin gel. Nine RCTs reported lower VAS pain scores in the treatment group compared with the control/placebo group.27, 29, 30, 31, 32, 34, 35, 36, 37, 41

, 34, 35, 36, 37, 38, 39, 40, 41 Interventions included lidocaine gel, lidocaine‐prilocaine (LP) cream, lidocaine intracervical block, lidocaine infusion, and nitroglycerin gel. Nine RCTs reported lower VAS pain scores in the treatment group compared with the control/placebo group.27, 29, 30, 31, 32, 34, 35, 36, 37, 41 3.4.1 Lidocaine spray Two RCTs reported lower placement‐related pain scores with lidocaine spray (net 40 mg) when compared with 2 g lidocaine cream or 10 mL lidocaine (20 mg/mL) injection31 and placebo.35 In the study by Karasu et al,31 a greater number of women in the lidocaine spray group reported no pain with placement compared with the injection and cream group (25.5% vs 1.9% vs 2.1%; P < 0.001). In the study by Aksoy et al,35 use of lidocaine spray reduced mean pain scores by 2.2 cm when compared with placebo (1.01 ± 1.20 vs 3.23 ± 1.60, P > 0.001). The authors also found a 35% reduction in the number of women scoring ≥ 4 on the 10‐cm VAS (6% vs 41%, P < 0.001).35 Neither study involved nulliparous women, nor did they assess pain at time‐points other than IUC placement, an observation noted by Aksoy et al35 as being useful in evaluating the delayed prostaglandin‐related cramping response that many women can experience. Karasu et al31 discussed the variability in dosing options for all three lidocaine preparations used in the study and the risks involved in increasing doses over and above those routinely used in gynecological practice. A nonrandomized study41 compared the effect of cervical lidocaine spray, cervical lidocaine gel and no topical anesthesia in 420 women: mean pain scores were 1‐1.5 cm lower with the use of lidocaine preparations compared with no anesthetic.41

of time may limit the achievement of optimal effect of oral or local analgesia for IUC placement, these agents may help in the relief of post‐placement pain. Second, post‐procedural pain may contribute to the memory of a painful experience and lead to the sharing of negative views when talking to other women about IUC. Multiple factors limit the ability to compare outcomes in the studies described: study populations, randomization protocols, concealment procedures, types of IUC placed, timing of pain assessment, and dose, formulation, and timing of analgesia all varied and limited the external validity of the findings. Although the instrument to measure pain (10‐cm or 100‐mm VAS) was used consistently across the majority of studies, the different assessment points for pain (both during the placement procedure and immediately following), use of inconsistent cut‐offs, and the potential for different anchor points limit inter‐study comparison. Additionally, most of the studies evaluated pain in all women who are looking for IUC placement and it is well known that many women experience no pain or minimal pain at IUC placement.4, 5, 6

creasing doses over and above those routinely used in gynecological practice. A nonrandomized study41 compared the effect of cervical lidocaine spray, cervical lidocaine gel and no topical anesthesia in 420 women: mean pain scores were 1‐1.5 cm lower with the use of lidocaine preparations compared with no anesthetic.41 3.4.2 Lidocaine paracervical block Three RCTs evaluated the effect of lidocaine paracervical block on pain experienced with IUC placement.27, 30, 31 Akers et al30 found there to be a 41.5 mm reduction in mean pain score when comparing injection of 1% lidocaine with a sham control in nulliparous women (P = 0.006). The authors highlighted limitations relating to the proportion of highly educated and insured women included in the study; the use of small IUC devices and the lack of blinding of study coordinators responsible for collecting participant data.30 Mody et al27 reported a 21 mm reduction in median pain score with 1% lidocaine block compared with placebo in nulliparous women (P = 0.002). However, there was an increase in pain associated with the lidocaine injection (30 vs 8 mm, P = 0.003).27 The study by Karasu et al31 described above found that 10 mL 2% lidocaine injection, a dose higher than that used in most clinics, led to a lower mean pain score at placement when compared with LP cream and control (no anesthesia) (2.9 ± 1.4 vs 4.0 ± 1.7 vs 4.25 ± 1.9) but no difference in the proportion of women experiencing no pain at placement (2.1% vs 1.9% vs 2.0%). Two RCTs comparing lidocaine paracervical block with nonsteroidal anti‐inflammatory drugs (NSAIDs) and placebo (400 mg ibuprofen and 500 mg oral naproxen/placebo, respectively) found no differences in mean pain scores during placement.20, 21

proportion of women experiencing no pain at placement (2.1% vs 1.9% vs 2.0%). Two RCTs comparing lidocaine paracervical block with nonsteroidal anti‐inflammatory drugs (NSAIDs) and placebo (400 mg ibuprofen and 500 mg oral naproxen/placebo, respectively) found no differences in mean pain scores during placement.20, 21 3.4.3 Lidocaine‐prilocaine cream Two studies31, 32 reported 1.9‐3.5 cm reductions on a 10‐cm VAS in mean pain scores at multiple time‐points during placement in parous women following application of LP cream when compared with the control group. In the study by Karasu et al,31 use of LP cream did not lead to lower pain scores at either tenaculum or IUC placement when compared with controls (no anesthesia). 3.4.4 Lidocaine gel Authors of two studies29, 37 reported lower placement‐related pain scores with novel formulations of lidocaine gel when compared with placebo. A single, 8.5‐mL dose of a formulation designed to minimize leakage and prolong presence in the target tissues (SHACT) reduced maximum pain by 16.1 mm when compared with placebo in >200 nulliparous women.37 Abd Ellah et al29 reported the findings of a small study (n = 48), involving self‐administration of a dual‐response, in situ lidocaine gel formulation in both parous and nulliparous women. Median pain scores were 2‐3 cm lower at all steps of IUC placement (including tenaculum placement and uterine sounding) with the lidocaine gel compared with placebo. The authors attributed the lower pain scores to the controlled release behavior of the gel demonstrated in rheological studies.29

s and nulliparous women. Median pain scores were 2‐3 cm lower at all steps of IUC placement (including tenaculum placement and uterine sounding) with the lidocaine gel compared with placebo. The authors attributed the lower pain scores to the controlled release behavior of the gel demonstrated in rheological studies.29 Three RCTs28, 33, 39 and one non‐RCT41 reported no difference in median pain scores with IUC placement with 2% lidocaine gel compared with placebo. 3.4.5 Other local anesthesia approaches Two RCTs, one comparing the effect of infusion of 1.2 mL of 2% lidocaine solution, administered by endometrial aspirator with placebo40 and one comparing the effect of 0.5 mg nitroglycerin gel (1 mL) with placebo,38 found no differences in mean pain scores between the treatment and control groups. 3.5 Post‐insertion pharmacological therapy No new or additional studies were found in the updated literature search.

3.4.5 Other local anesthesia approaches Two RCTs, one comparing the effect of infusion of 1.2 mL of 2% lidocaine solution, administered by endometrial aspirator with placebo40 and one comparing the effect of 0.5 mg nitroglycerin gel (1 mL) with placebo,38 found no differences in mean pain scores between the treatment and control groups. 3.5 Post‐insertion pharmacological therapy No new or additional studies were found in the updated literature search. 3.5.1 Non‐pharmacological pain management The broader search of the literature yielded four studies evaluating nonpharmacological strategies that may indirectly help to reduce pain experienced with IUC placement.42, 44, 45, 46 A successful reduction in anxiety (but not pain) was reported with use of inhaled lavender compared with placebo.42 Also, Arsenijevic et al44 reported a reduction in uterine and cervical injury during priming, which may impact on pain experience, with the use of a continuous, controllable balloon dilator compared with the Hegar dilator in a three‐arm study involving 120 women. Authors of a pilot feasibility study of a novel suction cervical retractor reported a 15‐mm reduction in mean pain score compared with placebo at IUC placement.45 Use of ultrasound‐guided IUC placement led to a 2.6‐cm reduction in mean pain score compared with the traditional, non‐guided technique (2.4 vs 5.0, P < 0.001).46 However, the authors acknowledged that presence of a full bladder could add to the pain experienced with speculum insertion. It was also noted that the need for availability of equipment and sonographic knowledge when placing the probe may not be feasible in all clinics.

non‐guided technique (2.4 vs 5.0, P < 0.001).46 However, the authors acknowledged that presence of a full bladder could add to the pain experienced with speculum insertion. It was also noted that the need for availability of equipment and sonographic knowledge when placing the probe may not be feasible in all clinics. A pooled analysis from three phase II studies involving the use of a modified placement device (EvoInserter®, insertion tube diameter of 3.8 mm) for the low‐dose 13.5 mg LNG‐IUS suggested that the reduced diameter contributed to ease of placement and manageable pain.49 Additional post‐hoc analyses showed a significant association (P = 0.0001) between the women's evaluation of pain on placement that was maintained following adjustment for age and parity. However, all data were retrospective and comparative data were not available across all three studies. As options for pain management and cervical dilation were left at the discretion of the investigators, conclusions regarding their impact on ease of placement and experience of pain are not possible.49 One of the studies that contributed to this pooled analysis reported a greater proportion of women experiencing “no pain” or “only mild pain” during placement in a randomized comparison of the EvoInserter® with the standard 52 mg LNG‐IUS with a larger (4.75 mm) insertion tube (72.3% vs 57.9%,).59

in are not possible.49 One of the studies that contributed to this pooled analysis reported a greater proportion of women experiencing “no pain” or “only mild pain” during placement in a randomized comparison of the EvoInserter® with the standard 52 mg LNG‐IUS with a larger (4.75 mm) insertion tube (72.3% vs 57.9%,).59 Two further RCTs found no difference in mean pain scores between vulsellum and a single‐tooth tenaculum43 or between tenaculum and Littlewood forceps47 when stabilizing the cervix for IUC placement. A comparison of delayed vs immediate emptying of a pre‐filled bladder also failed to show any significant difference in mean pain scores.48 3.6 Predictors of the pain experience Eight publications described factors that may help to predict women more likely to experience pain with IUC placement.50, 51, 52, 53, 54, 55, 56, 57, 58 The identified factors are summarized in Table 2. Chi et al57 suggested that higher education (>7 years), low parity (1‐2 live births), a longer interval between birth and placement (>13 months), and nonbreastfeeding at time of insertion were all associated with severe pain on IUC insertion. Goldstuck and Matthews51 reported an increase in pain experienced in women with greater cervical resistance when measured using a Salter Abbey electronic force gauge with a modified recording probe into which the IUC was inserted. The authors also reported that expected pain was higher than immediate pain experienced by study subjects (P = 0.001).

rted an increase in pain experienced in women with greater cervical resistance when measured using a Salter Abbey electronic force gauge with a modified recording probe into which the IUC was inserted. The authors also reported that expected pain was higher than immediate pain experienced by study subjects (P = 0.001). Newer studies identified several additional factors that may exert a role in the experience of pain: timing of insertion relative to menstruation, previous cesarean delivery, history of dysmenorrhea, expected pain, baseline anxiety, and size of insertion tube.50, 52, 53, 54 One RCT50 compared pain with IUC placement in both nulliparous and parous women within days 1‐7 of menstruation and at any day without vaginal blood loss and found no difference in pain experienced. A non‐randomized study investigating menstrual and gynecologic history as predictors of difficult or painful IUC placement reported severe dysmenorrhea as the only predictor of placement‐related pain.52 Santos et al53 compared pain experienced during IUC placement in three group of women; nulligravidas, parous with vaginal delivery and parous only with cesarean delivery. Women who had undergone cesarean delivery were at greater risk of pain (moderate to severe [VAS score 4‐10 cm] experienced by 88%) than women in the other two groups, potentially due to an anatomically distorted uterus caused by hypertrophic scarring.53

rous with vaginal delivery and parous only with cesarean delivery. Women who had undergone cesarean delivery were at greater risk of pain (moderate to severe [VAS score 4‐10 cm] experienced by 88%) than women in the other two groups, potentially due to an anatomically distorted uterus caused by hypertrophic scarring.53 Additionally, a linear regression analysis of the placebo‐controlled study evaluating the effect of high‐dose lidocaine gel on pain with IUC placement found nulliparity, interval IUC placement and a history of dysmenorrhea to be predictive of pain.39 In a multivariable analysis of the results of a prospective cohort study comparing pain with IUC placement in women with and without the experience of vaginal delivery by the same authors,54 a history of vaginal delivery was associated with a 15.5‐point reduction in mean pain score when compared with no previous vaginal delivery (P = 0.009). Other predictors of pain were “expected pain” and “placement difficulty”.54

acement in women with and without the experience of vaginal delivery by the same authors,54 a history of vaginal delivery was associated with a 15.5‐point reduction in mean pain score when compared with no previous vaginal delivery (P = 0.009). Other predictors of pain were “expected pain” and “placement difficulty”.54 A prospective cohort study56 found that the mean post‐placement pain experienced by nulliparous adolescents was higher than that among parous adult women on each day of the 2‐week study (P < 0.05) and the greatest mean difference occurred in the first 4 days. The authors suggested that pain scores may have been affected by the detailed pre‐ and post‐placement counseling of adolescent women provided by a trained Pediatric and Adolescent Gynecologist, which is uncommon in many settings and, therefore, different when compared with routine practice.56 There was also a recommendation to use ibuprofen by one professional responsible for this arm of the study and eight adolescents had the placement procedure under sedation.56

rained Pediatric and Adolescent Gynecologist, which is uncommon in many settings and, therefore, different when compared with routine practice.56 There was also a recommendation to use ibuprofen by one professional responsible for this arm of the study and eight adolescents had the placement procedure under sedation.56 A secondary analysis of 1149 participants in the US Contraceptive CHOICE Project looked at whether anticipated pain affected actual pain experienced during IUC placement.58 After controlling for parity, history of dysmenorrhea and type of IUC, higher anticipated pain was associated with an increase in experienced pain (adjusted risk ratio for 1 unit increase in anticipated pain, 1.19; 95% CI 1.14‐1.25).58 Nulliparity, history of dysmenorrhea, and placement of an LNG‐IUS (with a 4.8‐mm inserter) were all associated with an increase in mean pain score with IUC placement. As CHOICE was a prospective cohort study, real‐time collection of anticipated and actual pain data limits recall bias and strengthens the study.58 However, lack of information regarding the use of pain‐relieving interventions, other than the routine offering of premedication with NSAIDs, may be a limiting factor.

ment. As CHOICE was a prospective cohort study, real‐time collection of anticipated and actual pain data limits recall bias and strengthens the study.58 However, lack of information regarding the use of pain‐relieving interventions, other than the routine offering of premedication with NSAIDs, may be a limiting factor. A prospective cohort study by Narayan et al55 also looked at anticipated and actual pain related to IUC compared with implant placement. Although women choosing IUC had been told to expect the procedure to be painful, anticipated pain with both methods was similar. When asked about actual pain experienced via the post‐visit survey, women choosing IUC (n = 50) reported greater levels of pain than expected (7.0 vs 6.0, P = 0.004); whereas women choosing an implant reported less actual pain (2.0 vs 5.0, P = 0.001).55

ure to be painful, anticipated pain with both methods was similar. When asked about actual pain experienced via the post‐visit survey, women choosing IUC (n = 50) reported greater levels of pain than expected (7.0 vs 6.0, P = 0.004); whereas women choosing an implant reported less actual pain (2.0 vs 5.0, P = 0.001).55 3.7 Assessment of risk of bias and quality of included studies Each RCT included in the review was assessed for selection, performance, detection, attrition, and reporting bias by three authors (two assessors and one moderator) and the results are shown in Figure 2. The majority of the remaining non‐RCTs identified during our review described factors that may help identify those women at increased risk of pain with IUC placement. We did not assess these publications for risk of bias as their purpose was to provide insights into the potential for greater experience of pain than healthcare professionals may consider when counseling women regarding IUC. Given the nature of the review, ie narrative rather than systematic, assessment of quality of the RCTs did not extend beyond categorization according to the guidance of the Oxford Center for Evidence‐Based Medicine.60 Figure 2 Assessment of risk of bias of RCTs [Color figure can be viewed at http://www.wileyonlinelibrary.com]

3.7 Assessment of risk of bias and quality of included studies Each RCT included in the review was assessed for selection, performance, detection, attrition, and reporting bias by three authors (two assessors and one moderator) and the results are shown in Figure 2. The majority of the remaining non‐RCTs identified during our review described factors that may help identify those women at increased risk of pain with IUC placement. We did not assess these publications for risk of bias as their purpose was to provide insights into the potential for greater experience of pain than healthcare professionals may consider when counseling women regarding IUC. Given the nature of the review, ie narrative rather than systematic, assessment of quality of the RCTs did not extend beyond categorization according to the guidance of the Oxford Center for Evidence‐Based Medicine.60 Figure 2 Assessment of risk of bias of RCTs [Color figure can be viewed at http://www.wileyonlinelibrary.com] Although 12 RCTs were identified as being at low risk of bias within these categories,16, 19, 23, 24, 25, 32, 35, 36, 37, 38, 39 when looking at the potential for “other sources” of bias, all were seen as being at high risk due to a number of factors: they offered other analgesics besides the study drugs, recruited women who were looking for IUC placement and, in some cases, there was overlap in the VAS pain scores between treatment and placebo groups or the VAS pain scores in both groups were low (<4).

all were seen as being at high risk due to a number of factors: they offered other analgesics besides the study drugs, recruited women who were looking for IUC placement and, in some cases, there was overlap in the VAS pain scores between treatment and placebo groups or the VAS pain scores in both groups were low (<4). 4 DISCUSSION Our updated review set out to identify whether new RCTs contributed to further knowledge regarding effective pain strategies for IUC placement and identified 13 RCTs describing pharmacological interventions that result in lower pain scores when compared with placebo or control groups.16, 22, 24, 25, 27, 29, 30, 31, 32, 34, 35, 36, 37 The majority of these interventions involve pre‐insertion local anesthesia with lidocaine preparations. There is a need, however, for standardization of doses of these local analgesic agents, not only to optimize pain relief but also to maximize safety by preventing spread to vaginal tissue.31

mediately following), use of inconsistent cut‐offs, and the potential for different anchor points limit inter‐study comparison. Additionally, most of the studies evaluated pain in all women who are looking for IUC placement and it is well known that many women experience no pain or minimal pain at IUC placement.4, 5, 6 Authors of nine of the RCTs reviewed here described the differences in pain scores between the intervention (oral ketorolac, vaginal misoprostol, lidocaine [spray, gel or paracervical block] and LP cream) and the placebo/control as being clinically relevant.16, 24, 27, 29, 30, 32, 35, 36, 37, 38 The differences in pain scores within these studies ranged from 15 to 41 mm. A systematic review by Olsen et al61 suggests 17 mm as the minimum median effect size on the VAS to be considered clinically relevant but note that it can vary according to whether there are single or multiple measurements. Using a median effect size of ≥17 mm as a benchmark, use of lidocaine paracervical block, novel lidocaine gels, LP cream, and lidocaine spray, were shown to achieve clinically relevant reductions in pain at the time of IUC placement when compared with placebo/control.27, 29, 30, 32, 35, 36 There is a need for more studies that emphasize the importance of clinically relevant changes in the VAS pain evaluation or focus on and recruit only women with a history of pain or difficulty with IUC placement. Grouping the pain responses into mild (0‐3), moderate (4‐6) and severe (7‐10) could provide more useful and relevant comparisons of different interventions.

he importance of clinically relevant changes in the VAS pain evaluation or focus on and recruit only women with a history of pain or difficulty with IUC placement. Grouping the pain responses into mild (0‐3), moderate (4‐6) and severe (7‐10) could provide more useful and relevant comparisons of different interventions. Our review summarizes the findings of multiple studies looking at factors that increase the risk of pain with IUC placement. The effect of existing anxiety and beliefs about expected pain on actual pain during placement should not be under‐estimated. Women with mood disorders, a history of sexual trauma, a previous negative reaction to vaginal examination, a previous IUC placement that was painful, or awareness of the potential for pain from a friend or relatives may be more anxious about placement. When looking at whether anticipated pain vs actual pain could be a barrier to use of IUC, Narayan et al55 suggested that women choosing IUC may expect placement to be more painful and experience greater levels of pain than those choosing other contraceptive methods. However, it did not appear to affect their willingness to recommend IUC to friends either immediately following placement or 6 months later. Nevertheless, identifying women who anticipate pain and using evidence‐based pain management strategies may help to improve their experience. New studies, which evaluate these strategies in different groups of women, could help to change the perception of IUC placement and encourage more women to select IUC as a method.

evertheless, identifying women who anticipate pain and using evidence‐based pain management strategies may help to improve their experience. New studies, which evaluate these strategies in different groups of women, could help to change the perception of IUC placement and encourage more women to select IUC as a method. 4.1 Areas of potential future research In addition to the ongoing need to evaluate pain‐relieving strategies in a systematic, validated way, there are several potential areas for future research, such as the impact of pain‐relieving strategies on the severity and duration of post‐placement pain and timing of administration of pre‐placement analgesia. For example, optimum administration time needed to achieve peak plasma concentrations of ketorolac (2 hours ahead of placement) was described as impractical for IUC placement in a busy outpatient setting by the authors.16

erity and duration of post‐placement pain and timing of administration of pre‐placement analgesia. For example, optimum administration time needed to achieve peak plasma concentrations of ketorolac (2 hours ahead of placement) was described as impractical for IUC placement in a busy outpatient setting by the authors.16 A number of studies described clinically relevant effects in placement‐related pain when using lidocaine gel or spray, or LP cream,29, 31, 32, 35, 36, 37 including novel formulations designed to maximize retention in the cervical canal. However, the practical benefit of local application of lidocaine may be limited by the length of time required for lidocaine to take effect in a busy clinic (3 minutes for spray; 7 minutes with the speculum in place for LP cream)35, 36 and contribute to existing anxiety around the procedure. Although only a small study, the significant reduction in placement‐related pain with a self‐administered, dual‐response lidocaine gel, compared with placebo, described by Abd Ellah et al29 showed the potential for the adaptation of existing agents to overcome limitations with administration.

iety around the procedure. Although only a small study, the significant reduction in placement‐related pain with a self‐administered, dual‐response lidocaine gel, compared with placebo, described by Abd Ellah et al29 showed the potential for the adaptation of existing agents to overcome limitations with administration. Given that size and flexibility of the IUC can affect pain during placement, using smaller inserters and devices is likely to improve the experience for many women. The studies describing cervical priming showed that a moderate reduction in pain with IUC placement can be achieved with the use of vaginal misoprostol when compared with placebo.22, 24, 25 These studies also showed that misoprostol increased the likelihood of successful IUC placement and eased the procedure from a healthcare professional perspective in women who had a history of cesarean delivery or were nulligravidas. A study by Bahamondes et al62 also showed that misoprostol can increase the likelihood of successful IUC placement after previous insertion failure when compared with placebo: the risk ratios of successful placements in the evaluable population and the intent‐to‐treat population (95% CI) were 1.41 (8.2‐43.0) and 1.32 (0.3‐36.9), respectively. However, the side effect of pain caused by misoprostol‐induced uterine contractions remains an important consideration and may require adjunctive treatment with NSAIDs.

f successful placements in the evaluable population and the intent‐to‐treat population (95% CI) were 1.41 (8.2‐43.0) and 1.32 (0.3‐36.9), respectively. However, the side effect of pain caused by misoprostol‐induced uterine contractions remains an important consideration and may require adjunctive treatment with NSAIDs. The previous literature review highlighted a lack of studies investigating the potential for nonpharmacological strategies, including cognitive treatment approaches, to minimize the pain and anxiety associated with IUC placement.11 Our updated review found no studies evaluating the use of cognitive treatment in this area despite growing recognition of the links between psychological factors and pain experience.63 Although we found evidence of a significant and clinically relevant reduction in placement‐related pain with the use of ultrasound to guide placement when compared with control (non‐guided technique),46 pain experience appears to be unaffected by the type of forceps used.3, 43, 47 Research into this area, therefore, remains important.

d evidence of a significant and clinically relevant reduction in placement‐related pain with the use of ultrasound to guide placement when compared with control (non‐guided technique),46 pain experience appears to be unaffected by the type of forceps used.3, 43, 47 Research into this area, therefore, remains important. In conclusion, the number of publications found in the updated literature review indicates that the desire to identify ways to minimize pain with IUC placement remains an important goal. Evidence of clinically relevant effects on the experience of pain with pre‐placement analgesia when compared with placebo or controls is reported in studies evaluating different lidocaine preparations but routine use remains a subject for debate. Novel formulations of lidocaine gel, designed to ease administration, minimize leakage or prolong retention, appear particularly promising and further evaluation of these preparations in women who experience difficulty with placement would be a useful next step in achieving a more positive interpretation of the findings. Although placement‐related pain, insertion difficulties, and failures are uncommon, some women may be anxious or be at greater risk of painful placement and would, therefore, benefit from pain relief. Using an individual approach, guided by factors predictive of an increased risk of experiencing pain with IUC placement, could help to improve the experience for women. It may also contribute to the identification of effective, tailored strategies for routine use. In the modern era of medicine, our inability to recommend any positive treatment for pain relief with IUC placement creates professional discomfort. Further studies that use consistent approaches to timing of pain assessment could help to identify strategies to minimize the experience of pain and change the perception of IUC placement. In doing so, this may encourage more women to choose IUC as their contraceptive method.

placement creates professional discomfort. Further studies that use consistent approaches to timing of pain assessment could help to identify strategies to minimize the experience of pain and change the perception of IUC placement. In doing so, this may encourage more women to choose IUC as their contraceptive method. CONFLICT OF INTEREST This publication and its content are solely the responsibility of the authors. KGD has participated in advisory board meetings on contraception and been the PI of clinical trials on IUS and other contraceptives, received honorarium for presentations on contraception for Bayer AG, MSD/Merck, HRA‐Pharma, Exeltis, Natural Cycles, and Mithra. These potential conflicts of interest have been reviewed and managed by KI. AdSS has participated as a member of Advisory Boards for Bayer AG. JTJ has received payments for consulting and research support from Bayer AG, Abbvie, Merck, HRA Pharma, Sebela, and the Population Council; consulting only from Cooper Surgical; and research support only from Daré, Mithra, and Medicines360. These companies and organizations may have a commercial or financial interest in the results of this research and technology. These potential conflicts of interest have been reviewed and managed by OHSU. TP has participated in Advisory Boards for Bayer AG and Shionogi Ltd. She has given presentations on Intrauterine Contraception and counseling and is a member of the INTRA Group. LB participated in advisory board meetings on contraception for Bayer AG and for MSD/Merck. IM and MR have nothing to disclose.

Abbreviations ECSemergency cesarean section RCTrandomized controlled trial Key message The majority of obstetric outcomes occur with a very low incidence. Our sample size calculations showed that when using rare obstetric outcomes large sample sizes were required. Multicenter studies, international collaborations or alternative study designs to randomized controlled trials could be considered. 1 INTRODUCTION In high‐income countries the majority of pregnancies have a good outcome, and many adverse obstetric outcomes rarely occur. This makes demonstrating clinically relevant and statistically significant effects of new interventions a challenge. Randomized controlled trials (RCTs) are considered the gold standard for establishing causal inference in healthcare interventions, and are therefore frequently applied as study designs.1, 2 If RCTs have adequate statistical power, the expectation is that significant differences between groups will be a result of the intervention.1

ntrolled trials (RCTs) are considered the gold standard for establishing causal inference in healthcare interventions, and are therefore frequently applied as study designs.1, 2 If RCTs have adequate statistical power, the expectation is that significant differences between groups will be a result of the intervention.1 To ensure the quality of scientific work, calculating and reporting a study's sample size is fundamental.2 However, sample size calculations are sparsely reported in scientific papers, and many trials do not achieve the target sample size stated before starting the trial.3 When a study is underpowered, there is a risk of not finding the true difference between the groups, affecting the quality of the study.4 Two obstetric papers from 1997‐2000 on the introduction of continuous electronic fetal monitoring in obstetric care and on the potential bias when comparing small and large maternity institutions when studying stillbirth rates, respectively, discuss the implications of study design, rare outcome measures and large sample sizes.5, 6 The two studies concluded that, for rare outcomes, very large sample sizes were needed to detect statistically significant differences between study groups. Even though these two studies emphasize the implications of sample size calculation in obstetric outcomes, many researchers still include rare outcomes in their study design without having sufficient power to do so.

ery large sample sizes were needed to detect statistically significant differences between study groups. Even though these two studies emphasize the implications of sample size calculation in obstetric outcomes, many researchers still include rare outcomes in their study design without having sufficient power to do so. Our study had three objectives: first, to report incidences of 16 obstetric outcomes; secondly, to calculate sample sizes for tentative studies using three selected outcomes: neonatal mortality, Apgar score <7 at 5 minutes and emergency cesarean section (ECS); and thirdly, to discuss the implications for study design in obstetrics when choosing outcome measures. 2 MATERIAL AND METHODS 2.1 Population and study design The study was a registry‐based study. Data were retrieved from the Danish Medical Birth Registry and included all deliveries in Denmark from 2008 to 2015. The Danish Medical Birth Registry contains information on all deliveries in Denmark, thus providing data on the mother, the child, the pregnancy and the delivery. The annual number of deliveries in Denmark is approximately 60 000, with 96‐98% of deliveries in public hospitals and 2‐3% at home, mostly attended by midwives from public maternity departments.7 Currently there are 23 public maternity departments in Denmark, all with access to specialists in obstetrics and anesthesiology. Midwives attend all deliveries and an obstetrician is only involved in the event of complications.

hospitals and 2‐3% at home, mostly attended by midwives from public maternity departments.7 Currently there are 23 public maternity departments in Denmark, all with access to specialists in obstetrics and anesthesiology. Midwives attend all deliveries and an obstetrician is only involved in the event of complications. In this study we operated with two populations: the total population and the study population. We used the former, which included all deliveries in Denmark in the study period, to report incidences of important obstetric outcomes. For calculating sample sizes for tentative studies, we used a study population that included all intended vaginal deliveries with a term (gestational age ≥37 weeks) singleton in cephalic presentation. Stillbirths and homebirths were excluded. Incidences of the obstetric outcomes were also reported for the study population. 2.2 Outcome measures It is the scientific question raised that defines whether an obstetric event is an intervention, an outcome or even a population, not the event itself. In our study we have defined 16 relevant and used obstetric outcomes in the literature and are aware that these outcomes in other studies could be defined as interventions or constitute a study population.

that defines whether an obstetric event is an intervention, an outcome or even a population, not the event itself. In our study we have defined 16 relevant and used obstetric outcomes in the literature and are aware that these outcomes in other studies could be defined as interventions or constitute a study population. We chose 16 obstetric outcomes for reporting incidences of the total population: preeclampsia, hemolysis, elevated liver enzymes, and low platelets (HELLP) syndrome, eclampsia, induction of labor, oxytocin augmentation, umbilical cord prolapse, shoulder dystocia, vacuum extraction, ECS, postpartum hemorrhage ≥1000 mL, manual exploration of the uterus, stillbirth, Apgar score <7 at 5 minutes, preterm delivery <37 weeks of gestation, low birthweight <2500 g and neonatal mortality. We chose these outcomes considering the relevance for the patients and the severity of the outcomes. For the study population we report incidences of 14 obstetric outcomes. For sample size calculation in tentative studies, we chose three outcomes from the core outcome set for key stakeholders in maternity care: neonatal mortality, Apgar score <7 at 5 minutes and ECS.8 The chosen outcomes reflect different incidences: one extremely rare, one rare and one more common.

We chose 16 obstetric outcomes for reporting incidences of the total population: preeclampsia, hemolysis, elevated liver enzymes, and low platelets (HELLP) syndrome, eclampsia, induction of labor, oxytocin augmentation, umbilical cord prolapse, shoulder dystocia, vacuum extraction, ECS, postpartum hemorrhage ≥1000 mL, manual exploration of the uterus, stillbirth, Apgar score <7 at 5 minutes, preterm delivery <37 weeks of gestation, low birthweight <2500 g and neonatal mortality. We chose these outcomes considering the relevance for the patients and the severity of the outcomes. For the study population we report incidences of 14 obstetric outcomes. For sample size calculation in tentative studies, we chose three outcomes from the core outcome set for key stakeholders in maternity care: neonatal mortality, Apgar score <7 at 5 minutes and ECS.8 The chosen outcomes reflect different incidences: one extremely rare, one rare and one more common. Neonatal mortality is defined as death before the age of 28 completed days after live birth.9 Apgar score is used to assess the condition of the newborn at 1 and 5 minutes after birth and it is a validated predictor of neonatal survival.10 The Apgar score at 5 minutes is the best predictor of neonatal survival.10 Cesarean section is linked to a wide range of complications, such as uterine rupture and abnormal invasive placenta, which leads to higher risk of maternal and neonatal morbidity and mortality.11

is a validated predictor of neonatal survival.10 The Apgar score at 5 minutes is the best predictor of neonatal survival.10 Cesarean section is linked to a wide range of complications, such as uterine rupture and abnormal invasive placenta, which leads to higher risk of maternal and neonatal morbidity and mortality.11 2.3 Statistical analyses The data have been used as part of another study.12 Before analysis, the dataset was checked for logical errors. We recoded missing data for maternal weight and height with unrealistic values and checked whether there was consistency between the diagnosis of the delivery and the surgical intervention or procedure coded. The selected outcomes are reported as incidences, both for the total population and for the study population. The incidences of neonatal mortality, Apgar score <7 at 5 minutes and ECS in the study population formed the basis for the tentative sample size calculations. We calculated the sample sizes for the comparison of two proportions, which necessitates a proportion of the outcome, and the researcher to consider the intervention effect and the desired maximum risk of statistical errors. The statistical tool is provided in Supporting Information Appendix S1.

mple size calculations. We calculated the sample sizes for the comparison of two proportions, which necessitates a proportion of the outcome, and the researcher to consider the intervention effect and the desired maximum risk of statistical errors. The statistical tool is provided in Supporting Information Appendix S1. We calculated the sample size necessary for tentative RCTs and cohort studies to be able to detect risk reductions of 25 and 50% at the 5% level with a power of 80 and 90%, respectively. For the RCTs we calculated the sample size required for comparing two proportions with equal‐sized groups (i.e. 1:1 ratio), whereas for the cohort study we calculated the sample size also required for two proportions but with unequal sized groups. We calculated sample sizes for proportion of exposed women of 5, 10 and 25%.13 Incidences were computed using IBM SPSS® version 24 (IBM Corp., Armonk, NY, USA) and sample size calculations were made using SAS® software package version 9.4 (SAS Institute, Cary, NC, USA). As missing data were rare, imputation was not applied. 2.4 Ethical approval Approval was obtained from the Danish Data Protection Agency (file no.: 2012‐58‐0004). As this was a registry‐based study, ethical approval was not required according to the Danish Research Ethics Committee Law.14

Incidences were computed using IBM SPSS® version 24 (IBM Corp., Armonk, NY, USA) and sample size calculations were made using SAS® software package version 9.4 (SAS Institute, Cary, NC, USA). As missing data were rare, imputation was not applied. 2.4 Ethical approval Approval was obtained from the Danish Data Protection Agency (file no.: 2012‐58‐0004). As this was a registry‐based study, ethical approval was not required according to the Danish Research Ethics Committee Law.14 3 RESULTS From 2008 to 2015, there were 465 919 deliveries in Denmark. The study population, including intended vaginal deliveries with term singletons in cephalic presentation, consisted of 381 567 deliveries. There were missing data for Apgar score <7 at 5 minutes in 1260 deliveries (0.3%). There was no missing data for the variables neonatal mortality or ECS. Table 1 shows the sociodemographic characteristics of the total population and the study population. In general, Danish women were most likely to deliver at term, to be 25‐34 years of age, to be non‐smokers, and to have a normal body mass index (i.e. 18.5‐24.9 kg/m2). Table 1 Sociodemographic characteristics of the total population and the study population in Denmark from 2008 to 2015 Characteristics Total populationa Study populationb

Table 1 shows the sociodemographic characteristics of the total population and the study population. In general, Danish women were most likely to deliver at term, to be 25‐34 years of age, to be non‐smokers, and to have a normal body mass index (i.e. 18.5‐24.9 kg/m2). Table 1 Sociodemographic characteristics of the total population and the study population in Denmark from 2008 to 2015 Characteristics Total populationa Study populationb n (%) n (%) 465 919 (100) 381 567 (100) Singleton deliveries 456 014 (97.9) 381 567 (100) Twin deliveries 9794 (2.1) — Triplet/quadruplet deliveries 111 (0.0) — Breech deliveries 19 244 (4.1) — Singleton vaginal breech deliveries 1987 (0.5) — Planned cesarean section 43 407 (9.3) — Gestational age <37 weeks 30 544 (6.6) — 37+0 to 39+6 weeks 206 986 (44.4) 159 973 (41.9) ≥40 weeks 228 193 (49.0) 221 594 (58.1) Missing data 196 (0.0) — Maternal age (years) <25 57 907 (12.4) 49 985 (13.1) 25‐34 310 442 (66.6) 257 365 (67.4) 35‐39 81 335 (17.5) 62 532 (16.4) ≥40 16 235 (3.5) 11 685 (3.1) Parity Nulliparous 212 445 (45.6) 177 674 (46.6) Multiparous 248 976 (53.4) 203 893 (52.5) Missing data 4498 (1.0) 3605 (0.9) Smoking during pregnancy No 402 816 (86.5) 330 631 (86.7) Smoking cessation during pregnancy 13 676 (2.9) 11 588 (3.0) 1‐20 cigarettes per day 41 061 (8.8) 33 215 (8.7) >20 cigarettes per day 1417 (0.3) 1095 (0.3) Missing data 6949 (1.5) 5038 (1.3) Body mass index (kg/m2) <18.5 18 321 (3.9) 15 054 (3.9) 18.5‐24.9 276 318 (59.3) 229 585 (60.2) 25‐29.9 95 519 (20.5) 77 202 (20.2) 30‐34.9 37 128 (8.0) 29 342 (7.7) ≥35 20 302 (4.4) 15 618 (4.1) Missing data 18 331 (3.9) 14 766 (3.9) a The total population included all deliveries with gestational age 20+0 to 45+0.

.3) Body mass index (kg/m2) <18.5 18 321 (3.9) 15 054 (3.9) 18.5‐24.9 276 318 (59.3) 229 585 (60.2) 25‐29.9 95 519 (20.5) 77 202 (20.2) 30‐34.9 37 128 (8.0) 29 342 (7.7) ≥35 20 302 (4.4) 15 618 (4.1) Missing data 18 331 (3.9) 14 766 (3.9) a The total population included all deliveries with gestational age 20+0 to 45+0. b The study population included all term singleton (≥37 weeks of gestational age) with intended vaginal cephalic delivery. John Wiley & Sons, LtdTable 2 reports incidences of the 16 obstetric outcomes. Most outcomes occurred at a low incidence. The only outcomes with an incidence >10% were induction, oxytocin augmentation of labor and ECS. In the total population, the incidence of neonatal mortality, Apgar score <7 at 5 minutes and ECS was 0.4, 0.9 and 12.2%, respectively. In the study population, the incidence of neonatal mortality, Apgar score <7 at 5 minutes and ECS was 0.05% (95% confidence interval (CI) 0.04‐0.06), 0.58% (95% CI 0.55‐0.60) and 10.5% (95% CI; 10.4‐10.6), respectively. Table 2 Incidences of obstetric outcomes in Denmark from 2008 to 2015 Outcomea Total populationb Study populationc

John Wiley & Sons, LtdTable 2 reports incidences of the 16 obstetric outcomes. Most outcomes occurred at a low incidence. The only outcomes with an incidence >10% were induction, oxytocin augmentation of labor and ECS. In the total population, the incidence of neonatal mortality, Apgar score <7 at 5 minutes and ECS was 0.4, 0.9 and 12.2%, respectively. In the study population, the incidence of neonatal mortality, Apgar score <7 at 5 minutes and ECS was 0.05% (95% confidence interval (CI) 0.04‐0.06), 0.58% (95% CI 0.55‐0.60) and 10.5% (95% CI; 10.4‐10.6), respectively. Table 2 Incidences of obstetric outcomes in Denmark from 2008 to 2015 Outcomea Total populationb Study populationc n (%) n (%) 465 919 (100) 381 567 (100) Pregnancy outcomes Preeclampsia 13 874 (3.0) 9836 (2.6) HELLP 1177 (0.3) 472 (0.1) Eclampsia 249 (0.05) 147 (0.04) Induction of labor 102 499 (22.0) 93 174 (24.4) Delivery outcomes Oxytocin augmentation 100 791 (21.6) 92 975 (24.4) Umbilical cord prolapse 514 (0.1) 275 (0.1) Shoulder dystocia 4449 (1.0) 4344 (1.1) Vacuum extraction 32 816 (7.0) 30 943 (8.1) Emergency cesarean section 56 619 (12.2) 40 416 (10.6) Postpartum hemorrhage ≥1000 mL 10 627 (6.4) 8393 (6.2) Manual exploration of the uterus 6634 (1.4) 4858 (1.3) Neonatal outcomes Stillbirth 1721 (0.4) — Apgar score <7 at 5 minutes 4034 (0.9) 2201 (0.58) Preterm delivery <37 weeks’ gestation 30 544 (6.6) — Low birthweight <2500 g 21 648 (4.6) 4949 (1.3) Neonatal mortality 1310 (0.3) 184 (0.05) Abbreviation: HELLP, hemolysis, elevated liver enzymes, and low platelets.

4858 (1.3) Neonatal outcomes Stillbirth 1721 (0.4) — Apgar score <7 at 5 minutes 4034 (0.9) 2201 (0.58) Preterm delivery <37 weeks’ gestation 30 544 (6.6) — Low birthweight <2500 g 21 648 (4.6) 4949 (1.3) Neonatal mortality 1310 (0.3) 184 (0.05) Abbreviation: HELLP, hemolysis, elevated liver enzymes, and low platelets. a One delivery can be represented more than once. b The total population included all deliveries in Denmark from 2008 to 2015 with gestational age 20+0 to 45+0 weeks. In the event of multiple fetuses in one pregnancy, an outcome among one or more of the children counts. c The study population included all term (≥37 weeks’ gestational age), singleton, intended vaginal cephalic delivery in Denmark from 2008 to 2015. John Wiley & Sons, Ltd Table S1 reports incidences of the obstetric outcomes stratified by year.

b The total population included all deliveries in Denmark from 2008 to 2015 with gestational age 20+0 to 45+0 weeks. In the event of multiple fetuses in one pregnancy, an outcome among one or more of the children counts. c The study population included all term (≥37 weeks’ gestational age), singleton, intended vaginal cephalic delivery in Denmark from 2008 to 2015. John Wiley & Sons, Ltd Table S1 reports incidences of the obstetric outcomes stratified by year. Figure 1 and Table 3 report the sample sizes calculated for tentative RCTs and cohort studies. As shown, the incidence of the outcome measure affected the sample size. When using neonatal mortality with an incidence of 0.05% as the outcome in a tentative RCT with an expected risk reduction of 50% and power of 80%, our sample size calculation showed required sample size of 195 036 deliveries. Using Apgar score <7 at 5 minutes, with an incidence of 0.58%, as the outcome in a tentative RCT, with the same risk reduction and same power, 16 254 deliveries were required. For ECS with an incidence of 10.5%, 818 deliveries were required for a tentative RCT. Figure 1 and Table 3 also report the sample sizes required for studies with a power of 90% and for studies with a risk reduction of 25%.

in a tentative RCT, with the same risk reduction and same power, 16 254 deliveries were required. For ECS with an incidence of 10.5%, 818 deliveries were required for a tentative RCT. Figure 1 and Table 3 also report the sample sizes required for studies with a power of 90% and for studies with a risk reduction of 25%. Figure 1 The total sample size required for tentative randomized controlled trials (1:1) and cohort studies (1:3, 1:9, 1:19) with neonatal mortality, Apgar score <7 at 5 minutes (<7/5) and emergency cesarean section as the outcome. Changes of a 25 and 50% reduction in outcomes are plotted against incidences, using an 80 and 90% power and a 5% significance level Table 3 Sample sizes for tentative randomized controlled trials and cohort studiesa Outcome (incidence) Power Risk reduction 50% 25% Proportion of exposed Proportion of exposed 5% 10% 25% 50% 5% 10% 25% 50% Neonatal mortality (0.05%) 80% 884 820 476 510 241 584 195 036 4 576 400 2 431 690 1 190 132 916 518 90% 1 259 520 673 250 334 172 261 096 6 280 940 3 327 940 1 615 172 1 226 960 Apgar score <7 at 5 minutes (0.58%) 80% 73 680 39 680 20 112 16 254 377 260 200 460 98 112 75 786 90% 104 840 56 040 27 816 21 758 517 760 274 340 133 148 101 454 Emergency cesarean section (10.6%) 80% 3740 2010 1016 818 18 820 10 000 4884 3764 90% 5300 2830 1400 1092 25 760 13 650 6620 5038 a Total sample sizes required for tentative classical randomized controlled trials with an allocation of 1:1 ratio (i.e. 50% exposed) and cohort studies with a proportion of exposed of 5, 10 and 25%.

cesarean section (10.6%) 80% 3740 2010 1016 818 18 820 10 000 4884 3764 90% 5300 2830 1400 1092 25 760 13 650 6620 5038 a Total sample sizes required for tentative classical randomized controlled trials with an allocation of 1:1 ratio (i.e. 50% exposed) and cohort studies with a proportion of exposed of 5, 10 and 25%. John Wiley & Sons, LtdOur results illustrate that an expected lower risk reduction increased the sample sizes. Using neonatal mortality as the outcome in a tentative RCT with 80% power and changing the risk reduction from 50% to 25% resulted in a fourfold increase in the required sample size to 916 518 deliveries. Using Apgar score <7 at 5 minutes or ECS as outcomes, the same fourfold increase in the required sample size was seen when the expected risk reduction changed from 50% to 25% (Figure 1, Table 3). Our results furthermore illustrate that changing the power from 90% to 80% had a small impact on the required sample sizes. Using Apgar score <7 at 5 minutes as the outcome in a tentative RCT with 50% risk reduction and 80% power instead of 90%, the required sample size decreased from 21 758 to 16 254 deliveries. Using ECS as the outcome in an RCT with 50% risk reduction and 80% power instead of 90%, the required sample size decreased from 1092 to 818 deliveries (Figure 1, Table 3).

tcome in a tentative RCT with 50% risk reduction and 80% power instead of 90%, the required sample size decreased from 21 758 to 16 254 deliveries. Using ECS as the outcome in an RCT with 50% risk reduction and 80% power instead of 90%, the required sample size decreased from 1092 to 818 deliveries (Figure 1, Table 3). The study design furthermore affected the required sample sizes. When changing the proportion of exposed women from 50% to a smaller proportion, larger sample sizes were required. If Apgar score <7 at 5 minutes was used as the outcome in a tentative cohort study with 25% exposed instead of an RCT (50% exposed) with a power of 80% and risk reduction of 50%, the required sample size increased from 16 254 deliveries to 20 112 deliveries (Table 3). The same tendency was seen when the proportion of exposed women was even smaller. If the proportion of exposed was 10%, the required sample size was 39 680, and if the proportion of exposed was 5%, the sample size was further increased to 73 680 deliveries (Table 3). Table S2 reports the required sample sizes for all 14 outcomes from the study population. 4 DISCUSSION We found that the majority of obstetric outcomes occurred at a very low incidence. Our sample size calculations showed that the choice of study design, the outcome incidence and the change from 50% to 25% in the risk reduction all contributed to the required sample size of the tentative studies. Changing the power from 90% to 80% did not have large impact on the required sample size.

4 DISCUSSION We found that the majority of obstetric outcomes occurred at a very low incidence. Our sample size calculations showed that the choice of study design, the outcome incidence and the change from 50% to 25% in the risk reduction all contributed to the required sample size of the tentative studies. Changing the power from 90% to 80% did not have large impact on the required sample size. A strength of our study is that the reported incidences are based on a large data source of 456 014 deliveries and the sample size calculations on a data source of 381 567 deliveries. The deliveries represent the period from 2008 to 2015, making the data fairly current. Registry‐based research always involves the uncertainty associated with inaccurate reporting. Several studies show, however, that data from the Danish Medical Birth Register are valid in terms of diagnosis on most well‐defined outcomes, such as preeclampsia, birthweight, oxytocin augmentation of labor, vacuum extraction and cesarean section.15, 16

s involves the uncertainty associated with inaccurate reporting. Several studies show, however, that data from the Danish Medical Birth Register are valid in terms of diagnosis on most well‐defined outcomes, such as preeclampsia, birthweight, oxytocin augmentation of labor, vacuum extraction and cesarean section.15, 16 The study demonstrated that low incidence of the outcome affected the sample size. In Danish settings, if an RCT with 90% power was required to show a significant reduction of 25% of Apgar score <7 at 5 minutes, the study would take more than 2 years and require the inclusion of all deliveries. However, it might not be possible to include all eligible patients and some do not want to participate in the study. Thus, the time it takes to recruit patients will be longer than anticipated. This entails the researcher in the planning phase of an RCT to be realistic about recruitment and retention of participants in the study. This could be done through a feasibility study.

atients and some do not want to participate in the study. Thus, the time it takes to recruit patients will be longer than anticipated. This entails the researcher in the planning phase of an RCT to be realistic about recruitment and retention of participants in the study. This could be done through a feasibility study. Our sample size calculations showed that a major contributor to the required sample size was the change from 50% to 25% in the risk reduction. Applying a risk reduction of 50% instead of 25% to the sample size calculation in a tentative RCT with a rare outcome such as Apgar score <7 at 5 minutes would still require a large‐scale multicenter study. Multicenter studies have the advantage of including more participants in shorter time. However, multicenter studies are considerably more complex to run than single‐site studies. Furthermore, the sample size calculations depend upon the assumption that the differences between the compared interventions in the centers are unbiased estimates of the same quantity. Based on previous studies, a reduction in risk of 50% or 25% in Apgar score <7 at 5 minutes is probably unrealistic.17, 18 A realistic and still clinically important reduction in Apgar score <7 at 5 minutes might be 10%, which would require an even larger sample size.

enters are unbiased estimates of the same quantity. Based on previous studies, a reduction in risk of 50% or 25% in Apgar score <7 at 5 minutes is probably unrealistic.17, 18 A realistic and still clinically important reduction in Apgar score <7 at 5 minutes might be 10%, which would require an even larger sample size. With a more common outcome such as ECS, conducting RCTs is more feasible because of the requirement of smaller sample sizes to achieve the adequate power. This might explain why ECS is often seen as an outcome in obstetric studies.17, 19, 20, 21 ECS may be a relevant outcome, but it is also easier to obtain power to show statistically significant results compared with a more rare outcome. Furthermore, in many studies an effect on the more common outcomes is often found and the interpretation is that a given intervention has only affected these common outcomes. The intervention, however, could potentially also have affected the rare outcomes, but the study might be underpowered to show this effect. Meta‐analyses of RCTs are a way of increasing the power of the estimated intervention effect. However, meta‐analyses are, like single‐site studies, prone to risk of systematic and random error.22, 23

With a more common outcome such as ECS, conducting RCTs is more feasible because of the requirement of smaller sample sizes to achieve the adequate power. This might explain why ECS is often seen as an outcome in obstetric studies.17, 19, 20, 21 ECS may be a relevant outcome, but it is also easier to obtain power to show statistically significant results compared with a more rare outcome. Furthermore, in many studies an effect on the more common outcomes is often found and the interpretation is that a given intervention has only affected these common outcomes. The intervention, however, could potentially also have affected the rare outcomes, but the study might be underpowered to show this effect. Meta‐analyses of RCTs are a way of increasing the power of the estimated intervention effect. However, meta‐analyses are, like single‐site studies, prone to risk of systematic and random error.22, 23 Sometimes used in studies with rare outcomes,24, 25, 26 composite outcomes combine several variables, which are considered to be equivalent, into one outcome to increase the total incidence of these outcomes. Composite outcomes enable the study to be performed with a smaller sample and/or in less time. However, composite outcomes often provide an unclear reflection of the effect because the outcomes are not necessarily equivalent in terms of severity or measurements, and it is possible that the exposure increases the risk of one complication and decreases the risk of another. In the latter situation, the possible effect of the exposure may be camouflaged.27

an unclear reflection of the effect because the outcomes are not necessarily equivalent in terms of severity or measurements, and it is possible that the exposure increases the risk of one complication and decreases the risk of another. In the latter situation, the possible effect of the exposure may be camouflaged.27 RCTs are considered the gold standard for establishing causality between exposure and outcome in healthcare interventions.1, 2 RCTs are usually expensive and time‐consuming, and not all research questions can be answered by interventional studies due to ethical considerations. A possible alternative study design to RCTs is observational studies, such as cohort or case–control studies. These designs facilitates studies using rare outcomes as they allow for inclusion of large populations due to no intervention and no treatment, just as the exposure does not have to be administered to the participants.28 Thus, even though our sample size calculations revealed larger sample sizes for cohort studies of unequal groups, also when studying rare outcomes such as Apgar score <7 at 5 minutes, obtaining the required sample size in a cohort study would be easier than in an RCT. This design would even allow studying rare outcomes of low exposure and also even with a relatively low reduction in risk. There is a limitation when studying extremely rare outcomes because the required sample sizes will be extremely large, making it impossible to recruit participants within a reasonable time frame. Cohort studies, however, have the limitations that they are prone to bias, the data may be inaccurate and misclassified, and deducing causal conclusions is not possible, weakening the study's internal validity. The International Network of Obstetric Survey Systems (INOSS), a multi‐country collaboration, facilitates studies of rare and severe outcomes in pregnancy and childbirth through international cooperation.29 International collaborative work on registry‐based data might be a good approach to obtain sufficient sample size when studying rare outcomes. Sometimes RCTs are not the most feasible study design, and many historical cases exist in which treatments with a convincing change are based on observational studies.30

tion.29 International collaborative work on registry‐based data might be a good approach to obtain sufficient sample size when studying rare outcomes. Sometimes RCTs are not the most feasible study design, and many historical cases exist in which treatments with a convincing change are based on observational studies.30 The attention given to relevant outcome measures, as in the CROWN initiative31 is very important to obtain high quality evidence, but it is also essential to consider whether an RCT with rare outcomes is actually feasible, or if an alternative study design must be chosen. To our knowledge, the medical literature on sample size calculation and power regarding obstetric outcomes and choice of study design is sparse. Our results support the findings from two other studies investigating this issue. Mongelli et al demonstrated that when introducing electronic fetal monitoring, it was much easier to detect an increase in the incidences of cesarean section than a reduction in morbidity because of the different sample sizes and time needed to detect a significant change in the two outcomes.5 Moster et al demonstrated that large sample sizes were needed when comparisons of safety between different sizes of delivery units were made for low‐risk pregnancies, including stillbirth as the outcome measure.6 Our findings furthermore provide insights into sample sizes in relation to study design in both rare and more common obstetric outcomes.

t large sample sizes were needed when comparisons of safety between different sizes of delivery units were made for low‐risk pregnancies, including stillbirth as the outcome measure.6 Our findings furthermore provide insights into sample sizes in relation to study design in both rare and more common obstetric outcomes. 5 CONCLUSION Based on Danish national data from an 8‐year period, we found that several obstetric outcomes occur rarely. Consequently, very large sample sizes are required to achieve adequate statistical power in tentative RCTs. This necessity entails a risk of studies being underpowered or only showing an effect on common outcomes when an effect on rare outcomes might also exist. Focusing on international multicenter collaboration and prioritizing a feasible study design can provide high quality evidence when investigating rare outcomes. CONFLICT OF INTEREST The authors have stated explicitly that there are no conflicts of interest in connection with this article. Supporting information Click here for additional data file. Click here for additional data file. Click here for additional data file. ACKNOWLEDGMENT We would like to express our deepest gratitude to senior advisor Steen Rasmussen, MSc (econ), MPH, Rigshospitalet, for managing the national registry data.

Abbreviations GnRHagonadotropin‐releasing hormone agonist UPAulipristal acetate Key message Non‐inferiority of ulipristal acetate in terms of intraoperative blood loss could not be established. Pretreatment with gonadotropin‐releasing hormone agonist seems to be more favorable than ulipristal acetate for several operative outcomes and subjective surgical parameters. 1 INTRODUCTION Laparoscopic myomectomy seems to have several advantages over the laparotomic approach. Smaller incisions result in less postoperative pain, shorter hospital stay and faster recovery.1, 2, 3 However, laparoscopic myomectomy can be difficult when fibroids are large and numerous. This may result in extensive intraoperative bleeding and the need for a conversion to a laparotomy. Medical pretreatment prior to surgery might reduce these risks by decreasing fibroid size and vascularization of the fibroid. Only two medications are registered for the pretreatment of fibroids. Gonadotropin‐releasing hormone agonists (GnRHa) are considered the gold standard. Pretreatment with GnRHa improves pre‐ and postoperative hemoglobin level and reduces uterine and fibroid volume.4 Ulipristal acetate (UPA), a selective progesterone receptor modulator, has recently been approved for preoperative treatment of uterine fibroids. UPA has pro‐apoptotic and anti‐proliferative effects on the fibroid and normal myometrial tissue remains unaffected.

an be changed into defects with contact with the cavity. Healing is a long‐term, ongoing process of tissue remodeling, peaking in intensity within 6 months after the primary insult.15, 16 On the basis our data we assume that CS scar healing should be completed after 6 months. This is consistent with previous studies.14 The most important decision is whether to use an SLT or DLT to improve the scar quality and decrease the risk of uterine rupture and dehiscence in the subsequent pregnancy. Data from Swedish registers demonstrate no significant difference in the rate of uterine rupture.17 An earlier meta‐analysis including retrospective and prospective studies reported that locked SLT, compared with DLT, is associated with a fourfold increase in the risk of uterine rupture.18 We do not have any clinical data from subsequent pregnancies in our patients and are thus unable to confirm this finding. We have shown that DLT is associated with smaller defects and with thicker RMT. This is in contrast to evidence based on randomized trials which does not support a specific type of uterine closure. SLT and locked first layer are possibly associated with thinner residual RMT.19

Only two medications are registered for the pretreatment of fibroids. Gonadotropin‐releasing hormone agonists (GnRHa) are considered the gold standard. Pretreatment with GnRHa improves pre‐ and postoperative hemoglobin level and reduces uterine and fibroid volume.4 Ulipristal acetate (UPA), a selective progesterone receptor modulator, has recently been approved for preoperative treatment of uterine fibroids. UPA has pro‐apoptotic and anti‐proliferative effects on the fibroid and normal myometrial tissue remains unaffected. No randomized trials are available reporting on surgical outcomes comparing pretreatment with GnRHa or UPA. In this double‐blind randomized controlled trial, we evaluate whether pretreatment with UPA was non‐inferior to pretreatment with GnRHa (11.25 mg) on intraoperative and postoperative outcomes of laparoscopic myomectomy. 2 MATERIAL AND METHODS 2.1 Study design We performed a double‐blind randomized controlled trial in nine hospitals in the Netherlands comparing UPA and GnRHa prior to laparoscopic myomectomy. Participating hospitals were selected on extensive experience (>150 per year) with level 3 and 4 gynecological laparoscopic surgery as defined by Royal College of Obstetricians and Gynaecologists.

randomized controlled trial in nine hospitals in the Netherlands comparing UPA and GnRHa prior to laparoscopic myomectomy. Participating hospitals were selected on extensive experience (>150 per year) with level 3 and 4 gynecological laparoscopic surgery as defined by Royal College of Obstetricians and Gynaecologists. 2.2 Study population Premenopausal women for a laparoscopic resection of a maximum of two FIGO (PALM‐COEIN classification) type 3, 4, 5, 6, or two to five uterine fibroids with a diameter of 5‐12 cm were eligible for participation in this study. Exclusion criteria were age below 18 years, pregnancy, suspicion of malignancy, use of hormonal agents, chronic use of anticoagulants, coagulopathy, contraindication to laparoscopic procedures or allergy to leuprolide acetate or UPA.

ine fibroids with a diameter of 5‐12 cm were eligible for participation in this study. Exclusion criteria were age below 18 years, pregnancy, suspicion of malignancy, use of hormonal agents, chronic use of anticoagulants, coagulopathy, contraindication to laparoscopic procedures or allergy to leuprolide acetate or UPA. 2.3 Randomization and treatment Eligible women visiting the outpatient clinic of one of the participating centers were informed about the study by their gynecologist. After written informed consent was given, participating women were randomly allocated in a one‐to‐one ratio to receive either GnRHa or UPA. Randomization was performed using a computer‐generated randomization system and stratified by center. Women in the GnRHa group received a single intramuscular injection of leuprolide acetate (11.25 mg in 1 mL) and daily placebo tablets for 12 consecutive weeks. Participants in the UPA group received daily oral UPA 5 mg for 12 consecutive weeks and a one‐time placebo injection containing 1 mL saline. Study materials and medication packaging were identical for both groups. Treatment was preferably started in the first week of the menstruation period. Surgery was performed within a month after the last tablet. Participants and gynecologists were blinded to treatment allocation during the entire study period. Statistics were performed by an independent statistician blinded to the allocated study groups.

preferably started in the first week of the menstruation period. Surgery was performed within a month after the last tablet. Participants and gynecologists were blinded to treatment allocation during the entire study period. Statistics were performed by an independent statistician blinded to the allocated study groups. 2.4 Outcome measures The primary outcome was intraoperative blood loss. Secondary outcomes were time of surgery, time of enucleation, time of suturing, surgical ease and reduction in fibroid volume. For a careful explanation of all outcome measures, see Appendix S1. 2.5 Laparoscopic myomectomy Surgery was performed by experienced surgeons. The procedure was performed under general anesthesia after administration of prophylactic broad‐spectrum antibiotics. An expert meeting of participants was held on 2 June 2014 to reach consensus on the surgical technique. Relevant surgical characteristics were divided in: “standard use”, “never use” or “optional use”, defined as: Standard use—use of barbed sutures, use of (any) uterine manipulator, use of blue dye in uterine cavity in order to diagnose whether the cavity was opened or not. For fibroids >8 cm it was allowed to apply bulldogs on the uterine artery and infundibulopelvic ligament. Never use—vasoconstrictive medication such as glypressin or use of bulldogs for fibroids <8 cm. Optional use—single administration of 1000 mg of tranexamic acid or use of hemostatic or anti‐adhesive products on uterine incision were allowed only after operative blood loss has been calculated.

nfundibulopelvic ligament. Never use—vasoconstrictive medication such as glypressin or use of bulldogs for fibroids <8 cm. Optional use—single administration of 1000 mg of tranexamic acid or use of hemostatic or anti‐adhesive products on uterine incision were allowed only after operative blood loss has been calculated. 2.6 Statistical analyses The trial was a non‐inferiority trial with the following null hypothesis: UPA is non‐inferior to GnRHa in terms of blood loss during surgery with a maximum difference of 150 mL considered acceptable based on previous studies on this subject.5 The assumed standard deviation was 250 mL for intraoperative blood loss based on a survey in three hospitals in the Netherlands. Based on a two‐group t test of equivalence in means, using an one‐sided significance level of 2.5% (one‐sided), and a Type II error of 20% (80% power) this yields a sample size of 90 women (45 in each study arm).

deviation was 250 mL for intraoperative blood loss based on a survey in three hospitals in the Netherlands. Based on a two‐group t test of equivalence in means, using an one‐sided significance level of 2.5% (one‐sided), and a Type II error of 20% (80% power) this yields a sample size of 90 women (45 in each study arm). The analyses of the primary outcome intraoperative blood loss were performed both according to the per protocol principle and intention‐to‐treat principle. All other analyses were performed according to the intention‐to‐treat principle. Normality of the data was assessed visually by means of QQ plots. Because the primary outcome itself did not appear to be normally distributed, but became normally distributed after a log‐transformation, we used the following procedure for testing non‐inferiority. To take into account the non‐inferiority margin of 150 mL defined on the original scale, we added 150 mL to the observed blood losses for GnRHa but left observed blood losses for UPA unchanged. This was done to create a setting in which the difference of the log‐transformation of the adapted blood loss was 0 exactly if the blood loss in UPA pretreated women is 150 mL higher than pretreatment with GnRHa (the null hypothesis of the non‐inferiority test). Non‐inferiority was concluded if the confidence interval for the differences in means for the log‐transformation of these adapted outcomes lay entirely below 0 mL.

oss was 0 exactly if the blood loss in UPA pretreated women is 150 mL higher than pretreatment with GnRHa (the null hypothesis of the non‐inferiority test). Non‐inferiority was concluded if the confidence interval for the differences in means for the log‐transformation of these adapted outcomes lay entirely below 0 mL. Normally distributed data were summarized as mean ± standard deviation and were compared with the independent t test. For continuous outcomes that were not normally distributed we present median and interquartile range. Depending on the exact distribution, we used an independent sample t test on the log‐transformed outcome or the Mann‐Whitney U test when comparing the outcomes between the groups. The Mann‐Whitney test was also used for comparison of ordinal variables between the groups. Dichotomous and categorical outcomes were summarized by frequencies and percentage. Chi‐square test and Fisher's exact test were used to compare the distribution of these outcomes between groups. To assess differences between baseline and after 3 months within a group, we used the paired t test for normally distributed variables, the Wilcoxon signed rank test for non‐normally distributed data and the McNemar test for dichotomous variables.

Normally distributed data were summarized as mean ± standard deviation and were compared with the independent t test. For continuous outcomes that were not normally distributed we present median and interquartile range. Depending on the exact distribution, we used an independent sample t test on the log‐transformed outcome or the Mann‐Whitney U test when comparing the outcomes between the groups. The Mann‐Whitney test was also used for comparison of ordinal variables between the groups. Dichotomous and categorical outcomes were summarized by frequencies and percentage. Chi‐square test and Fisher's exact test were used to compare the distribution of these outcomes between groups. To assess differences between baseline and after 3 months within a group, we used the paired t test for normally distributed variables, the Wilcoxon signed rank test for non‐normally distributed data and the McNemar test for dichotomous variables. Linear regression analyses were performed to correct the analyses for comparison of mean blood loss and total surgery time between pretreatment groups for potential confounders. A variable was considered a confounder when the regression coefficient for groups changed by >10% when the confounder was added to the model. Potential confounders with a skewed distribution were log‐transformed to decrease of the impact of outliers. All analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, USA). Non‐inferiority of blood loss was tested at a one‐sided significance level of 2.5%. A two‐sided significance level of 5% was used for all other analyses.

Linear regression analyses were performed to correct the analyses for comparison of mean blood loss and total surgery time between pretreatment groups for potential confounders. A variable was considered a confounder when the regression coefficient for groups changed by >10% when the confounder was added to the model. Potential confounders with a skewed distribution were log‐transformed to decrease of the impact of outliers. All analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, USA). Non‐inferiority of blood loss was tested at a one‐sided significance level of 2.5%. A two‐sided significance level of 5% was used for all other analyses. 2.7 Ethical approval This study was approved by the National Central Committee on Research Involving Human Subjects (CCMO ‐ NL49916.029.14), by the ethics committee of VU Medical Center Amsterdam (Ref. No. 2014/421, date 17‐12‐2014) and by the boards of all participating hospitals. The trial protocol has been registered on ClinicalTrials.gov (NCT02288130).

proved by the National Central Committee on Research Involving Human Subjects (CCMO ‐ NL49916.029.14), by the ethics committee of VU Medical Center Amsterdam (Ref. No. 2014/421, date 17‐12‐2014) and by the boards of all participating hospitals. The trial protocol has been registered on ClinicalTrials.gov (NCT02288130). 3 RESULTS 3.1 Women Women were enrolled between May 2015 and July 2017. Due to disappointing inclusion rates in most participating centers and expiration of study medication, the intended number of inclusions was not met. Six of nine participating hospitals included women (Table S1). Of 68 eligible women, 55 were randomized: 30 allocated to UPA and 25 to leuprolide acetate (Figure 1). One woman randomized to UPA in retrospect did not meet the inclusion criteria (fibroid >12 cm) and was excluded from analysis. In the UPA group, two women dropped out before the end of pretreatment; one woman withdrew informed consent directly after allocation, so no follow up occurred and one woman underwent abdominal hysterectomy 6 weeks after start of medication due to persistent severe abdominal pains and fibroid growth. In the GnRHa group, all women completed pretreatment. A total of three women did not undergo a laparoscopic myomectomy. Two women (one in each group) refused surgery because they preferred homeopathic therapy. For one woman allocated GnRHa, surgery was cancelled due to major decrease in fibroid volume from 80 cm3 to 1 cm3. Figure 1 CONSORT flow diagram

3 RESULTS 3.1 Women Women were enrolled between May 2015 and July 2017. Due to disappointing inclusion rates in most participating centers and expiration of study medication, the intended number of inclusions was not met. Six of nine participating hospitals included women (Table S1). Of 68 eligible women, 55 were randomized: 30 allocated to UPA and 25 to leuprolide acetate (Figure 1). One woman randomized to UPA in retrospect did not meet the inclusion criteria (fibroid >12 cm) and was excluded from analysis. In the UPA group, two women dropped out before the end of pretreatment; one woman withdrew informed consent directly after allocation, so no follow up occurred and one woman underwent abdominal hysterectomy 6 weeks after start of medication due to persistent severe abdominal pains and fibroid growth. In the GnRHa group, all women completed pretreatment. A total of three women did not undergo a laparoscopic myomectomy. Two women (one in each group) refused surgery because they preferred homeopathic therapy. For one woman allocated GnRHa, surgery was cancelled due to major decrease in fibroid volume from 80 cm3 to 1 cm3. Figure 1 CONSORT flow diagram The two treatment groups were similar in demographic characteristics and hemoglobin level before pretreatment (Table 1). However, treatment groups differed in fibroid characteristics, due to lack of stratification for fibroid size at baseline. The mean diameter of the largest fibroid was significantly higher in women allocated UPA than women allocated GnRHa (8.5 ± 1.9 vs 7.4 ± 1.6 cm; P = 0.035; 95% CI .1‐2.0). Other fibroid characteristics at baseline (ie, type, uterine volume and total fibroid volume planned for resection) did not show any significant difference.

n diameter of the largest fibroid was significantly higher in women allocated UPA than women allocated GnRHa (8.5 ± 1.9 vs 7.4 ± 1.6 cm; P = 0.035; 95% CI .1‐2.0). Other fibroid characteristics at baseline (ie, type, uterine volume and total fibroid volume planned for resection) did not show any significant difference. Table 1 Baseline characteristics of the study population Ulipristal acetate (n = 29) leuprolide acetate (n = 25) P value Age (years; mean ± SD) 38.4 ± 5.8 41.4 ± 5.6 0.058 Body mass index (kg/m2; median, range) 24.5 (18.8‐38.2) 25.9 (18.0‐42.5) 0.466 Parity (median, range) 0 (0‐2) 1 (0‐4) 0.112 Race (n; %) Caucasian 17 (59) 11 (44) 0.357 Black 3 (10) 6 (24) Other 9 (31) 8 (32) Medical history of abdominal surgery (n, %) 14 (48) 8 (32) 0.225 Hemoglobin (mmol/L; mean ± SD) 7.8 ± 1.0 7.7 ± 1.1 0.742 Indication for surgery (n)a Heavy menstrual bleeding 11 13 NA Abdominal pain 9 8 Subfertility 8 2 Mechanical complaintsb 15 7 Other 2 1 Type of largest fibroidc (n, %) 3 — 3 (12) 0.093 4 1 (3) 2 (8) 5 4 (14) 4 (16) 6 9 (31) 3 (12) 2‐5 15 (52) 13 (52) Mean diameter of largest fibroid (cm; mean ± SD) 8.5 ± 1.9 7.4 ± 1.6 0.035 Total fibroid volume planned for resection (cm3; median, IQR) 316.3 (184.7‐462.6) 246.0 (130.2‐344.3) 0.094 Uterine volume (cm3; median, IQR) 530.5 (392.8‐774.5) 421.2 (327.5‐819.8) 0.450 Abbreviations: IQR, interquartile range; SD, standard deviation. Bold indicates statistically significant values. a Patients could have more than one indication for surgery. b Eg, constipation, urinary retention. c Type of fibroid following the FIGO (PALM‐COEIN) classification.

Total fibroid volume planned for resection (cm3; median, IQR) 316.3 (184.7‐462.6) 246.0 (130.2‐344.3) 0.094 Uterine volume (cm3; median, IQR) 530.5 (392.8‐774.5) 421.2 (327.5‐819.8) 0.450 Abbreviations: IQR, interquartile range; SD, standard deviation. Bold indicates statistically significant values. a Patients could have more than one indication for surgery. b Eg, constipation, urinary retention. c Type of fibroid following the FIGO (PALM‐COEIN) classification. John Wiley & Sons, LtdCompliance to study medication was high in both groups. Only two women (one in each group) reported that they forgot to take their oral study medication daily (UPA group 50 tablets remaining; GnRHa group 8 tablets remaining). For the per protocol analyses of the primary outcome, the woman who did not comply with UPA was excluded from analyses.

ation was high in both groups. Only two women (one in each group) reported that they forgot to take their oral study medication daily (UPA group 50 tablets remaining; GnRHa group 8 tablets remaining). For the per protocol analyses of the primary outcome, the woman who did not comply with UPA was excluded from analyses. 3.2 Intraoperative blood loss As intraoperative blood loss was not normally distributed, non‐inferiority could not be assessed in the standard manner using the 95% CI for the differences in mean blood loss. The alternative approach using a log‐transformation of the adapted outcomes as described in the statistical analysis yielded a 95% CI for which the upper bound exceeded 0 (95% CI for difference −0.06 to 0.30 for both per protocol and intention to treat analyses) and hence this trial does not support the alternative hypothesis that UPA is non‐inferior to GnRHa. After correction for confounder mean diameter of the largest fibroid, the 95% confidence interval for the mean difference of the logs still contained 0 (per protocol analyses: 95% CI for difference −0.20 to 0.13; intention‐to‐treat analyses: 95% CI for difference −0.20 to 0.14), which did not alter the conclusion.

all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.017, test significance of variable Closure technique is 0.065 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.493 Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Presence and defect contact with the uterine cavity Yes No Normal scar Yes No Normal scar 6 weeks 75 (50.3) 48 (32.2) 26 (17.4) 87 (49.7) 53 (30.3) 35 (20.0) 6 months 91 (61.1) 27 (18.1) 31 (20.8) 93 (53.1) 36 (20.6) 46 (26.3) 12 months 84 (56.4) 40 (26.8) 25 (16.8) 92 (52.6) 35 (20.0) 48 (27.4) John Wiley & Sons, Ltd4 DISCUSSION The prevalence of scar defects on transvaginal ultrasound in our study was 83.2% in the SLT and 72.6% in the DLT group at the 12‐month follow up. We observed a higher prevalence of defects compared with other groups after first CS (37‐61%).3, 9, 10, 12 The discrepancy is most likely to be explained not only by the different definitions of defect that have been used in different studies, but also by including defects that were not in contact with the uterine cavity. The 12‐month prevalence of more serious defects (RMT <2.5 mm)9, 13, 14 in contact with the cavity was 12.2% in the SLT and 6.8% in the DLT group.

RHa. After correction for confounder mean diameter of the largest fibroid, the 95% confidence interval for the mean difference of the logs still contained 0 (per protocol analyses: 95% CI for difference −0.20 to 0.13; intention‐to‐treat analyses: 95% CI for difference −0.20 to 0.14), which did not alter the conclusion. Pretreatment with UPA results in significant higher median intraoperative blood loss compared with pretreatment with GnRHa (525 mL [interquartile range 348‐1025; range 100‐2275] vs 280 mL] 100‐500; range 40‐2200], P = 0.002) (Table 2). Number of fibroids removed, fibroid type and mean diameter of largest fibroid at baseline were tested for potential confounding using linear regression analysis. Only mean diameter of the largest fibroid at baseline appeared to be a confounder. Correction for this confounder resulted in a P value of 0.011. Table 2 Results on intraoperative outcomes Ulipristal acetate (n = 26) leuprolide acetate (n = 23) P value Intraoperative blood loss (mL; median, IQR) 525 (348‐1025) 280 (100‐500) 0.002 a , b Total surgery time (min; median, IQR) 188 (132‐231) 125 (100‐175) 0.023 a , c Time of enucleation (min; median, IQR) 51 (31‐86) 35 (26‐66) 0.203 Fibroid 1 49 (31‐86) 33 (25‐60) 0.191 Fibroid 2 10 (5‐15) 8 (5‐25) 0.916 Suturing time (min; median, IQR) 42 (29‐51) 25 (14‐40) 0.009 Fibroid 1 40 (28‐48) 22 (14‐33) 0.003

Ulipristal acetate (n = 26) leuprolide acetate (n = 23) P value Intraoperative blood loss (mL; median, IQR) 525 (348‐1025) 280 (100‐500) 0.002 a , b Total surgery time (min; median, IQR) 188 (132‐231) 125 (100‐175) 0.023 a , c Time of enucleation (min; median, IQR) 51 (31‐86) 35 (26‐66) 0.203 Fibroid 1 49 (31‐86) 33 (25‐60) 0.191 Fibroid 2 10 (5‐15) 8 (5‐25) 0.916 Suturing time (min; median, IQR) 42 (29‐51) 25 (14‐40) 0.009 Fibroid 1 40 (28‐48) 22 (14‐33) 0.003 Fibroid 2 5 (3‐40) 18 (7‐30) 0.325 Time of morcellation (min; median, IQR) 13 (5‐25) 7 (4‐14) 0.085 Number of fibroids removed (n; %) 1 21 (81) 19 (83) 0.868 ≥2 5 (19) 4 (17) Weight of fibroids removed (g; median, IQR) 349 (185‐561) 140 (61‐272) 0.001 Hemoglobin drop (mmol/L; mean ± SD) 1.8 ± 1.3 1.0 ± 0.8 0.012 Bulldogs on uterine artery (n, %) 4 (15) 2 (9) 0.671 Bulldogs on ovarian vessels (n, %) 10 (39) 3 (13) 0.044 Opening of cavum (n, %) 4 (15) 5 (22) 0.716 Conversion rate (n, %) 3 (12) 0 (0) 0.237 Complication rate (n, %)d 4 (15) 2 (9) 0.671 Abbreviations: IQR, interquartile range; SD, standard deviation. Bold indicates statistically significant values. a Unpaired t test performed on log‐transformed variable intraoperative blood loss and total surgery time. b After correction for confounder ‘mean diameter of largest fibroid’: P = 0.011. c After correction for confounder ‘mean diameter of largest fibroid’: P = 0.053. d All intraoperative complications in both groups were blood loss >1 L.

a Unpaired t test performed on log‐transformed variable intraoperative blood loss and total surgery time. b After correction for confounder ‘mean diameter of largest fibroid’: P = 0.011. c After correction for confounder ‘mean diameter of largest fibroid’: P = 0.053. d All intraoperative complications in both groups were blood loss >1 L. John Wiley & Sons, LtdHemoglobin level within 48 hours of surgery compared with hemoglobin level before surgery was statistically significantly lower in the UPA group compared with GnRHa (−1.8 ± 1.3 vs −1.0 ± 0.8 mmol/L; 95% CI for difference 0.2‐1.4; P = 0.012). 3.3 Secondary outcomes 3.3.1 Differences from baseline to 3 months Table 3 shows changes in fibroid characteristics and hemoglobin levels between baseline and after 3 months of pretreatment. Change in mean diameter of the largest fibroid was found to differ between pretreatment with UPA and pretreatment with GnRHa (−3.6% [−15.5 to 10.4] vs −14.6% [−40.7 to −5.6]; P = 0.003). Total reduction in volume of the fibroids that were planned for resection was statistically significantly less after UPA than after GnRHa pretreatment (−7.2% [−35.5 to 54.1] vs −38.4% [−71.5 to −19.3]; P = 0.001). Reduction in uterine volume after 3 months was also statistically significantly smaller in the UPA group than with GnRHa (−6.4% [−24.3 to 51.0] vs −26.2% [−63.4 to 4.2]; P = 0.020). Hemoglobin levels increased significantly in both pretreatment groups during pretreatment (Table 2). Table 3 Changes in characteristics of fibroids and hemoglobin levels after 3 months’ pretreatment

3.3 Secondary outcomes 3.3.1 Differences from baseline to 3 months Table 3 shows changes in fibroid characteristics and hemoglobin levels between baseline and after 3 months of pretreatment. Change in mean diameter of the largest fibroid was found to differ between pretreatment with UPA and pretreatment with GnRHa (−3.6% [−15.5 to 10.4] vs −14.6% [−40.7 to −5.6]; P = 0.003). Total reduction in volume of the fibroids that were planned for resection was statistically significantly less after UPA than after GnRHa pretreatment (−7.2% [−35.5 to 54.1] vs −38.4% [−71.5 to −19.3]; P = 0.001). Reduction in uterine volume after 3 months was also statistically significantly smaller in the UPA group than with GnRHa (−6.4% [−24.3 to 51.0] vs −26.2% [−63.4 to 4.2]; P = 0.020). Hemoglobin levels increased significantly in both pretreatment groups during pretreatment (Table 2). Table 3 Changes in characteristics of fibroids and hemoglobin levels after 3 months’ pretreatment Ulipristal acetate (n = 29) leuprolide acetate (n = 25) P value Mean diameter of largest fibroid (cm; mean ± SD) Baseline 8.5 ± 1.9 7.4 ± 1.6 3 months 8.4 ± 2.3 5.8 ± 2.1a Change from baseline to 3 months in cm −0.1 ± 1.7 −1.6 ± 1.5 0.002 Change from baseline to 3 months in % (median, IQR) −3.6% (−15.5 to 10.4) −14.6% (−40.7 to −5.6) 0.003 Total fibroid volume planned for resection (cm3; median, IQR) Baseline 316.3 (184.7‐462.6) 246.0 (130.2‐344.3) 3 months 319.4 (163.0‐506.4) 105.9 (54.4‐195.9)b Change from baseline to 3 months in cm3 −33.0 (−109.5 to 95.8) −80.2 (−183.6 to −33.2) 0.012

Change from baseline to 3 months in % (median, IQR) −3.6% (−15.5 to 10.4) −14.6% (−40.7 to −5.6) 0.003 Total fibroid volume planned for resection (cm3; median, IQR) Baseline 316.3 (184.7‐462.6) 246.0 (130.2‐344.3) 3 months 319.4 (163.0‐506.4) 105.9 (54.4‐195.9)b Change from baseline to 3 months in cm3 −33.0 (−109.5 to 95.8) −80.2 (−183.6 to −33.2) 0.012 Change from baseline to 3 months in % −7.2% (−35.5 to 54.1) −38.4% (−71.5 to −19.3) 0.001 Uterine volume (cm3; median, IQR) Baseline 530.5 (392.8‐774.5) 421.2 (327.5‐819.8) 3 months 598.2 (284.6‐830.6) 272.1 (180.4‐508.8)c Change from baseline to 3 months in cm3 −28.1 (−161.2 to 107.3) −151.5 (−256.6 to 9.4) 0.081 Change from baseline to 3 months in % −6.4% (−24.3 to 51) −26.1% (−63.4 to 4.2) 0.020 Hemoglobin (mmol/L, mean ± SD) Baseline 7.8 ± 1.0 7.7 ± 1.1 3 months 8.4 ± .8d 8.2 ± 0.9d Change from baseline to 3 months 0.5 ± 0.9 0.5 ± 0.9 0.859 Abbreviations: IQR, interquartile range; SD, standard deviation. Bold indicates statistically significant values. a Statistically significant compared with baseline (P < 0.001). b Statistically significant compared with baseline (P < 0.001). c Statistically significant compared with baseline (P = 0.015). d Statistically significant compared with baseline (ulipristal acetate P = 0.012; GnRHa P = 0.013).

Bold indicates statistically significant values. a Statistically significant compared with baseline (P < 0.001). b Statistically significant compared with baseline (P < 0.001). c Statistically significant compared with baseline (P = 0.015). d Statistically significant compared with baseline (ulipristal acetate P = 0.012; GnRHa P = 0.013). John Wiley & Sons, Ltd3.3.2 Other intraoperative outcomes In 81% of women pretreated with UPA, only one fibroid was removed, compared with 83% of women pretreated with GnRHa (P = 0.868). Unadjusted analysis of total surgery time showed a longer surgery time in the UPA group compared with GnRHa (188 minutes [132‐231] vs 125 minutes [100‐175]; P = 0.023) (Table 2). A potential confounding effect on surgery time of number of fibroids removed and the mean diameter of the largest fibroid at baseline was tested. Only mean diameter of the largest fibroid at baseline appeared to be a confounder. After correction for this confounder, the difference in surgery time between both groups did not reach statistical significance (P = 0.053). No difference in time of enucleation and time of morcellation was found between both groups. Suturing time for the largest fibroid was longer in the UPA group (40 minutes [28‐48] vs 22 minutes [14‐33]; P = 0.003). The weight of the fibroids removed was significantly higher in women pretreated with UPA (349 g [185‐561] vs 140 g [61‐272]; P = 0.001).

eation and time of morcellation was found between both groups. Suturing time for the largest fibroid was longer in the UPA group (40 minutes [28‐48] vs 22 minutes [14‐33]; P = 0.003). The weight of the fibroids removed was significantly higher in women pretreated with UPA (349 g [185‐561] vs 140 g [61‐272]; P = 0.001). In women pretreated with UPA, the ovarian vessels were clamped more frequently than in the GnRHa group (10 times vs 3 times; P = 0.044) because in 52% of the women in the UPA group, the mean diameter of the largest fibroid remained >8 cm despite pretreatment. In the GnRHa group, this was 16% (P = 0.004). No difference was found in frequency of clamping of the uterine artery, opening of the uterine cavum or conversion rate. Six intraoperative complications were reported, four in the UPA group and two in the GnRHa group (all intraoperative blood loss of >1 L).

e pretreatment. In the GnRHa group, this was 16% (P = 0.004). No difference was found in frequency of clamping of the uterine artery, opening of the uterine cavum or conversion rate. Six intraoperative complications were reported, four in the UPA group and two in the GnRHa group (all intraoperative blood loss of >1 L). 3.3.3 Surgical ease Most items of the surgical assessment tool show a significant difference between both treatment groups (Table 4). Procedures of women pretreated with UPA were found to be more difficult than procedures of women pretreated with GnRHa (4 [3.0‐5.0] vs 3 [2.0‐4.0]; P = 0.011). Surgeons in the UPA group were less satisfied compared with those in the GnRHa group (2 [2.0‐3.3] vs 2 [1.0‐2.0] P = 0.027). Surgeons found it more difficult to identify cleavage planes in women pretreated with UPA than with GnRHa (4 [2.7‐4.0] vs 3 [1.0‐4.0]; P = 0.035). They also reported more difficulties with morcellation of fibroids pretreated with UPA compared with GnRHa (3 [2.0‐3.0] vs 2 [2.0‐3.0]; P = 0.011). In women pretreated with UPA, surgeons reported softer fibroids than in women pretreated with GnRHa (2 [1.0‐3.0] vs 3 [2.0‐4.0]; P = 0.017). This resulted in a poorer grip on the fibroid during surgery in the UPA group (2 [1.0‐3.0] vs 1 [1.0‐2.0]; P = 0.001). Stitching the myometrium was assessed as more difficult in women who received UPA than GnRHa (3 [3.0‐3.0] vs 3 [2.0‐3.0]; P = 0.011). The subjectively evaluated bleeding tendency of the tissue at surgery was reported to be higher in the UPA group than the GnRHa group (3 [3.0‐4.0] vs 3 [2.0‐3.0]; P = 0.004). The grip of barbed sutures in the myometrium and the anatomical result did not differ between the treatment groups.

] vs 3 [2.0‐3.0]; P = 0.011). The subjectively evaluated bleeding tendency of the tissue at surgery was reported to be higher in the UPA group than the GnRHa group (3 [3.0‐4.0] vs 3 [2.0‐3.0]; P = 0.004). The grip of barbed sutures in the myometrium and the anatomical result did not differ between the treatment groups. Table 4 Results on surgical ease 1 2 3 4 5 P value Difficulty of entire procedure (median, IQR) a 1 = very easy; 2 = easy; 3 = moderate; 4 = difficult; 5 = very difficult UPAb 0.011 GnRHac Satisfaction with entire procedure (median, IQR) 1 = very satisfied; 2 = satisfied; 3 = moderate; 4 = unsatisfied; 5 = very unsatisfied UPA 0.027 GnRHa Difficulty finding cleavage planes fibroid/capsula (median, IQR) 1 = very easy; 2 = easy; 3 = moderate; 4 = difficult; 5 = very difficult UPA 0.035 GnRHa Difficulty of morcellation (median, IQR) 1 = very easy; 2 = easy; 3 = moderate; 4 = difficult; 5 = very difficult UPA 0.011 GnRHa Consistency of fibroid (median, IQR) 1 = very soft; 2 = soft; 3 = normal; 4 = firm; 5 = very firm UPA 0.017 GnRHa Grip on fibroid (median, IQR) 1 = good; 2 = moderate; 3 = bad UPA 0.001 GnRHa Grip of barbed sutures in myometrium (median, IQR) 1 = good; 2 = moderate; 3 = bad UPA 0.151 GnRHa Ease of stitching (median, IQR) 1 = very easy; 2 = easy; 3 = moderate; 4 = difficult; 5 = very difficult UPA 0.011 GnRHa Bleeding tendency of tissue at surgery (median, IQR) 1 = almost none; 2 = little; 3 = normal; 4 = more than average; 5 = very bloody UPA 0.004 GnRHa Result anatomically (median, IQR) 1 = very satisfied; 2 = satisfied; 3 = moderate; 4 = unsatisfied; 5 = very unsatisfied UPA 0.054 GnRHa a IQR = interquartile range.

1 = very easy; 2 = easy; 3 = moderate; 4 = difficult; 5 = very difficult UPA 0.011 GnRHa Bleeding tendency of tissue at surgery (median, IQR) 1 = almost none; 2 = little; 3 = normal; 4 = more than average; 5 = very bloody UPA 0.004 GnRHa Result anatomically (median, IQR) 1 = very satisfied; 2 = satisfied; 3 = moderate; 4 = unsatisfied; 5 = very unsatisfied UPA 0.054 GnRHa a IQR = interquartile range. b UPA = ulipristal acetate (). c GnRHa = Gonadotropin Releasing‐hormone Agonist (). John Wiley & Sons, Ltd3.3.4 Side‐effects, postoperative complications and serious adverse events To assess the prevalence of the most common side‐effects, headaches and hot flushes, the Menopause questionnaire by Oldenhave et al6 was used. The number of women reporting moderate to severe headaches after 3 months was not significantly higher compared with baseline in both groups (UPA: 36 vs 32%; GnRHa: 16 vs 35%). The number of women experiencing moderate to severe hot flushes was significantly higher after 3 months for both groups (UPA: 7 vs 43%, P = 0.006; GnRHa 12 vs 65%, P = 0.004). The frequency of hot flushes after 3 months did not differ between groups (P = 0.111). A total of six postoperative complications were reported, four in the UPA group and two in the GnRHa group. For detailed information on these complications see Appendix S2.

John Wiley & Sons, Ltd3.3.4 Side‐effects, postoperative complications and serious adverse events To assess the prevalence of the most common side‐effects, headaches and hot flushes, the Menopause questionnaire by Oldenhave et al6 was used. The number of women reporting moderate to severe headaches after 3 months was not significantly higher compared with baseline in both groups (UPA: 36 vs 32%; GnRHa: 16 vs 35%). The number of women experiencing moderate to severe hot flushes was significantly higher after 3 months for both groups (UPA: 7 vs 43%, P = 0.006; GnRHa 12 vs 65%, P = 0.004). The frequency of hot flushes after 3 months did not differ between groups (P = 0.111). A total of six postoperative complications were reported, four in the UPA group and two in the GnRHa group. For detailed information on these complications see Appendix S2. 4 DISCUSSION In this double‐blinded randomized controlled trial, non‐inferiority of UPA to GnRHa in terms of intraoperative blood loss could not be established (using a predefined non‐inferiority margin of 150 mL). This can be explained by the limited sample size; however, it can not be excluded that UPA is inferior to GnRHa as a pretreatment for laparoscopic myomectomy. Median intraoperative blood loss was 245 mL higher in the group pretreated with UPA than GnRHa. This was in line with the higher hemoglobin drop postoperatively in women pretreated with UPA. Mediation analysis showed that the difference in intraoperative blood loss between both groups can partly be explained by less heavy fibroids after pretreatment with GnRHa and lower fibroid weight, in turn associated with lower intraoperative blood loss (Appendix S3). Suturing time of the first fibroid is significantly longer in women who received UPA as pretreatment compared with GnRHa. This could explain the higher intraoperative blood loss in the UPA group or the larger fibroid size at the time of surgical removal. Within the confines of this trial, pretreatment with UPA results in a significantly smaller decrease in fibroid volume and uterine volume compared with GnRHa. Also, laparoscopic myomectomies in women pretreated with UPA were subjectively evaluated as more difficult than in women pretreated with GnRHa.

surgical removal. Within the confines of this trial, pretreatment with UPA results in a significantly smaller decrease in fibroid volume and uterine volume compared with GnRHa. Also, laparoscopic myomectomies in women pretreated with UPA were subjectively evaluated as more difficult than in women pretreated with GnRHa. No previous studies have been published comparing UPA with GnRHa prior to laparoscopic myomectomy. Previous randomized trials compared GnRHa with placebo or immediate surgery before laparoscopic myomectomy.7, 8, 9 No randomized trials have been performed comparing UPA with placebo before laparoscopic myomectomy. Assumptions on intraoperative blood loss made in designing this study are based on a systematic review and meta‐analysis by Chen et al5 comparing pretreatment with GnRHa with no pretreatment. Mean intraoperative blood loss in the included studies (n = 3) for women pretreated with GnRHa varied from 172 to 199 mL. This is lower than the median intraoperative blood loss found in our study of 280 mL in women pretreated with GnRHa. This can be explained by smaller fibroid diameter or fibroid volume of the previous studies (ie, volume of largest fibroid is ±65 cm3 in these studies, whereas median volume of fibroid planned for resection in our study is 246 cm3). The predefined non‐inferiority margin of 150 mL may have been too small, considering the larger fibroids included in our study. This could be why non‐inferiority was not established.

, volume of largest fibroid is ±65 cm3 in these studies, whereas median volume of fibroid planned for resection in our study is 246 cm3). The predefined non‐inferiority margin of 150 mL may have been too small, considering the larger fibroids included in our study. This could be why non‐inferiority was not established. A randomized trial by Donnez et al10 comparing UPA with GnRHa, did not show significant differences in reduction of fibroid volume after 3 months of pretreatment between the two groups. The difference with our study may be explained by the size of the fibroids included in these studies. In the study by Donnez et al, the median cumulative volume of the three largest fibroids at baseline is 79.6 cm3 in the UPA group and 59.2 cm3 in the leuprolide acetate group. These fibroids are much smaller than the total fibroid volume planned for resection in our study (ie, 316.3 cm3 and 246.0 cm3, respectively, at baseline). Well conducted studies on surgical ease in pretreated women undergoing laparoscopic myomectomy are very limited. Only two retrospective studies have been published on this subject comparing surgical experience in women pretreated with UPA, with no hormonal pretreatment prior to myomectomy11, 12 showing overall no difference in surgical experience.

gical ease in pretreated women undergoing laparoscopic myomectomy are very limited. Only two retrospective studies have been published on this subject comparing surgical experience in women pretreated with UPA, with no hormonal pretreatment prior to myomectomy11, 12 showing overall no difference in surgical experience. This is the first randomized controlled trial performed on intraoperative outcomes comparing UPA with GnRHa prior to laparoscopic myomectomy. This trial was performed double‐blind, resulting in women and surgeons who were unaware of the pretreatment received. This is particularly important for subjective outcomes such as surgical ease. An important limitation of this study is that the anticipated total of 90 women could not be recruited due to expiration of study medication and disappointing inclusion rates. These can be explained by physician preferences for one of the treatments, an overestimation of most participating centers of the number of laparoscopic myomectomies they perform on a yearly basis, and the fact that many eligible women were already (pre)treated with UPA or GnRHa in another hospital before they were referred to one of the participating centers. Despite this, several outcomes reached statistical significance. It is not possible to conclude whether some of these significant differences were caused by chance (type I error). Another limitation is the fact that we did not stratify for fibroid size at baseline, resulting in an unbalanced distribution of the fibroid size in both groups. A regression analysis was performed to correct for this confounder; however, it cannot be excluded that this difference at baseline had a subsequent effect on many of the other endpoints such as blood loss and weight of fibroids. Additionally, we present the stratified results of fibroids ≤8 cm or >8 cm (Table S2). The direction and trend of the differences between UPA and GnRHa remain the same.

t cannot be excluded that this difference at baseline had a subsequent effect on many of the other endpoints such as blood loss and weight of fibroids. Additionally, we present the stratified results of fibroids ≤8 cm or >8 cm (Table S2). The direction and trend of the differences between UPA and GnRHa remain the same. The majority of women were included in one center. Sensitivity analyses did not show differences in intraoperative blood loss for this center compared with other centers (Table S1). An additional limitation of the study may be that the questionnaire to assess surgical ease is non‐validated due to very limited studies on this subject, so a total score could not be given. Surgeons were blinded to the pretreatment received in both treatment arms and each question should be interpreted as an individual item without calculation of a total score. Also, since the majority of women were included in one center, surgical ease was determined by a limited number of surgeons and should be interpreted as such.

e blinded to the pretreatment received in both treatment arms and each question should be interpreted as an individual item without calculation of a total score. Also, since the majority of women were included in one center, surgical ease was determined by a limited number of surgeons and should be interpreted as such. 5 CONCLUSION Our study did not demonstrate non‐inferiority of UPA as a pretreatment compared with GnRHa. We had an underpowered study with a relatively small number of women. Confirmation of our findings is needed to make any final conclusions and based on our data we advise that larger studies, potentially of a superior study design, are carried out. Furthermore, fibroids in our study were large and these large fibroids in particular may benefit from volume reduction to facilitate a successful laparoscopic approach. From that perspective, volume reduction is important, since volume seems to be related to surgical ease and surgery time. Future studies should aim to confirm this. Supporting information Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file.

Abbreviations AOPangle of progression AORadjusted odds ratio AUCarea under curve BMIbody mass index BSBishop score CScesarean section IOLinduction of labor PCAposterior cervical angle ROCreceiver‐operating characteristics SWEshear wave elastography Key message Shear wave elastography is a useful predictor of cesarean section. The combination of sonographic cervical length and elastography is superior to the Bishop score in the prediction of the outcome of labor induction.

IOLinduction of labor PCAposterior cervical angle ROCreceiver‐operating characteristics SWEshear wave elastography Key message Shear wave elastography is a useful predictor of cesarean section. The combination of sonographic cervical length and elastography is superior to the Bishop score in the prediction of the outcome of labor induction. 1 INTRODUCTION Approximately one in five inductions of labor (IOL) results in an emergency cesarean section (CS) due to failure to reach active phase or labor, failure to progress beyond the active phase or fetal distress.1 The expected outcome and management of IOL has traditionally been based on vaginal digital assessment of the cervix to assess the Bishop Score (BS).2 However, the BS has been shown to be subjective3 and has relatively low predictive performance.4 Recent research has therefore focused on the use of ultrasound for more objective assessment of individual components of BS. Besides cervical length, posterior cervical angle (PCA)5 and angle of progression (AOP)6 can be measured sonographically to reflect the cervical position and fetal head descent, respectively. Although some studies have shown that these ultrasonic measurements are superior to digital assessment, their predictive values remain suboptimal for clinical use.7, 8 One main reason for this is the inability to measure cervical consistency, a major component of BS, using conventional ultrasound technology. Hence strain‐based sonoelastography has also been investigated for measuring uterine cervical stiffness. However, this requires human movements on the ultrasound probe to generate a ‘stress’ on the target tissue. It is limited to measuring relative strain of the target tissue in comparison with its adjacent tissues, but the cervix lacks good surrounding tissues to act as reference.9 Several small‐cohort studies have conflicting results regarding the predictive value of strain‐based elastography.10, 11, 12, 13 In contrast, shear wave elastrography (SWE) uses ultrasound pulses to generate shear waves across a target tissue, and the shear wave velocity (‘v’) correlates to the tissue stiffness. Young's modulus (E) in kPa can be estimated using the formula E ≅ 3ρv2, where ρ is the density of the tissue (kg/m3), which is assumed to be constant.14 SWE assessment of the cervix has recently been reported to have good intra‐ and interobserver reproducibility.15 Cervical stiffness using SWE was also shown to decrease with gestational age.15 Yet its value in predicting IOL outcome is unknown.

ρ is the density of the tissue (kg/m3), which is assumed to be constant.14 SWE assessment of the cervix has recently been reported to have good intra‐ and interobserver reproducibility.15 Cervical stiffness using SWE was also shown to decrease with gestational age.15 Yet its value in predicting IOL outcome is unknown. Therefore the aim of the present study is to evaluate whether cervical SWE can improve the predictive performance of the outcome of IOL when combined with other ultrasound‐based assessments as compared with BS alone. 2 MATERIAL AND METHODS 2.1 Study design This was a prospective observational study conducted between September 2015 and November 2017. Women admitted for IOL were invited to participate. The inclusion criteria were: (1) Chinese women carrying singleton pregnancies; (2) ≥37 gestational weeks; (3) vertex presentation; (4) normal fetal well‐being on cardiotocography. Women who had a history of cervical surgery or any contraindication for vaginal delivery were excluded.

ined not only by the different definitions of defect that have been used in different studies, but also by including defects that were not in contact with the uterine cavity. The 12‐month prevalence of more serious defects (RMT <2.5 mm)9, 13, 14 in contact with the cavity was 12.2% in the SLT and 6.8% in the DLT group. We show that the presence of a defect and the scar position are relatively stable during the first postpartum year but that their appearance changes. Defects with or without contact with the uterine cavity changed statistically between controls. Both types of defects occurred more frequently in the 6‐week follow up and were more common in the SLT group. The most notable example was the defect without contact with the uterine cavity, which represented 30% of scar defects at 6 weeks and 20% at 6 months. During healing, this structure can disappear and can be changed into defects with contact with the cavity. Healing is a long‐term, ongoing process of tissue remodeling, peaking in intensity within 6 months after the primary insult.15, 16 On the basis our data we assume that CS scar healing should be completed after 6 months. This is consistent with previous studies.14

for IOL were invited to participate. The inclusion criteria were: (1) Chinese women carrying singleton pregnancies; (2) ≥37 gestational weeks; (3) vertex presentation; (4) normal fetal well‐being on cardiotocography. Women who had a history of cervical surgery or any contraindication for vaginal delivery were excluded. 2.2 Measurement of cervical SWE Participants were assessed on admission for IOL. Before the digital assessment for the BS, women were asked to empty the bladder and in modified lithotomy, a transvaginal scan of the cervix was performed using an SE 12‐3 probe (3‐12 MHz) of the SuperSonic Imagine ultrasound system (Aixplorer supersonic imagine, Aix‐en‐Province, France). The probe was inserted gently without any pressure being exerted on the cervix, and the mid‐sagittal view of the cervix was identified by clear visualization of internal os, canal and external os. The cervical image was magnified to occupy at least 75% of the screen. Once an optimal image of the cervix was obtained, a sampling box was put over the anterior lip and then the posterior lip of the cervix. To optimize the quality of the elastogram color image, each time the size of the sampling box was adjusted to just fit either the anterior or the posterior cervix. Each of the cervical lips was divided into three equal parts along its longitudinal axis: the inner part (proximal one‐third), the middle part, and the outer part (distal one‐third). Then the SWE value of each region of interest (ROI) was measured with a 5‐mm‐diameter circle placed at the center of each ROI. The SWE value was automatically displayed in pressure units (kPa) on the screen (Figure 1). The sampling method was repeated two more times from each cervical lip so as to obtain three independent SWE measurements of each ROI. The average of these three measurements was used for analysis.

ced at the center of each ROI. The SWE value was automatically displayed in pressure units (kPa) on the screen (Figure 1). The sampling method was repeated two more times from each cervical lip so as to obtain three independent SWE measurements of each ROI. The average of these three measurements was used for analysis. Figure 1 The shear wave elastic measurements. The shear wave elastic measurements were made on the inner, middle and outer parts of the anterior (A) and posterior (B) cervical lip The scan were performed by trained ultrasonographers. Their intra‐ and interobserver reproducibility was assessed among the first consecutive 30 women, who were assessed twice by the same operator, and then reassessed by the second operator. All the measurements were made on the independent images selected from the saved images. Both operators were blinded to the measurements made by each other.

server reproducibility was assessed among the first consecutive 30 women, who were assessed twice by the same operator, and then reassessed by the second operator. All the measurements were made on the independent images selected from the saved images. Both operators were blinded to the measurements made by each other. 2.3 Other ultrasonic assessment The cervical length was measured as the linear distance between the internal os and the external os.16 The PCA was defined as the inferior angle between the line joining the internal os and external os, and the line across the lower segment of the posterior uterine wall (Figure 2).5 A transperineal ultrasound was then performed using a curved XC6‐1 probe (1‐6 MHz) and the AOP was measured on the sagittal view between a line crossing the longitudinal axis of pubic symphysis intersecting a line through its inferior point tangential to the outer edge of the fetal skull (Figure 3).6 The pulsatility indices (PI) of the fetal middle cerebral artery (MCA) and the umbilical artery (UA) were also measured and the cerebroplacental ratio (CPR)17 was calculated. Estimated fetal weight (EFW) was calculated from the measurements of fetal biometry18 and expressed as percentiles.19 Figure 2 The measurement of the posterior cervical angle, which is the inferior angle between the line joining the internal os and external os, and the line across the lower segment of the posterior uterine wall

2.3 Other ultrasonic assessment The cervical length was measured as the linear distance between the internal os and the external os.16 The PCA was defined as the inferior angle between the line joining the internal os and external os, and the line across the lower segment of the posterior uterine wall (Figure 2).5 A transperineal ultrasound was then performed using a curved XC6‐1 probe (1‐6 MHz) and the AOP was measured on the sagittal view between a line crossing the longitudinal axis of pubic symphysis intersecting a line through its inferior point tangential to the outer edge of the fetal skull (Figure 3).6 The pulsatility indices (PI) of the fetal middle cerebral artery (MCA) and the umbilical artery (UA) were also measured and the cerebroplacental ratio (CPR)17 was calculated. Estimated fetal weight (EFW) was calculated from the measurements of fetal biometry18 and expressed as percentiles.19 Figure 2 The measurement of the posterior cervical angle, which is the inferior angle between the line joining the internal os and external os, and the line across the lower segment of the posterior uterine wall Figure 3 The measurement of the angle of progression, which is the angle between a line crossing the longitudinal axis of pubic symphysis intersecting a line through its inferior point tangential to the outer edge of the fetal skull [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2 The measurement of the posterior cervical angle, which is the inferior angle between the line joining the internal os and external os, and the line across the lower segment of the posterior uterine wall Figure 3 The measurement of the angle of progression, which is the angle between a line crossing the longitudinal axis of pubic symphysis intersecting a line through its inferior point tangential to the outer edge of the fetal skull [Color figure can be viewed at http://wileyonlinelibrary.com] 2.4 Management of IOL Subsequently, an independent obstetrician, blinded to the ultrasound findings, performed the per‐vaginal digital examination to determine the BS. The decision on the method of IOL was based on the BS. The standard practice of the studying unit was that, when the BS was ≥6, the cervix was regarded as ripened or favorable, and the IOL proceeded with amniotomy and/or syntocinon infusion; when the BS was <6, vaginal prostaglandin E2 (PGE2) gel or dinoprostone pessary was used. All clinical staff were blinded to the ultrasound findings. Failure to enter active phase was defined as failure of the cervix to efface and dilate to 3 cm in 12 hours after amniotomy or initiation of syntocinon infusion, or remaining unfavorable (BS <6) in 24 hours after a single pessary of 10 mg dinoprostone or three doses of 3 mg PGE2 gel. Failure to progress in the active phase was defined as cervical dilation slower than 1 cm/h for 4 hours during the active phase of labor. Fetal distress was defined as the presence of pathological cardiotocography which required immediate delivery.20

single pessary of 10 mg dinoprostone or three doses of 3 mg PGE2 gel. Failure to progress in the active phase was defined as cervical dilation slower than 1 cm/h for 4 hours during the active phase of labor. Fetal distress was defined as the presence of pathological cardiotocography which required immediate delivery.20 2.5 Sample size Our previous model to predict outcome of induction based on BS alone gave an AUC of 0.65.7 To detect a change of 0.1 in AUC with a new prediction model would require a minimum sample size of 425 for a type 1 error of 0.05, 80% power and assuming a CS for failed induction of 20% and a correlation between AUC of 0.5. Planned sample size was increased by a further 10% to 468 to allow for up to 10% failure rate to measure one or more ultrasound markers.

w prediction model would require a minimum sample size of 425 for a type 1 error of 0.05, 80% power and assuming a CS for failed induction of 20% and a correlation between AUC of 0.5. Planned sample size was increased by a further 10% to 468 to allow for up to 10% failure rate to measure one or more ultrasound markers. 2.6 Statistical analyses The primary outcome of IOL was successful vaginal delivery vs CS. Secondary comparisons were made between the group of vaginal delivery and the subgroups of: (1) CS for failure to enter active phase, (2) CS for failure to progress and (3) CS for fetal distress. The maternal characteristics, fetal and cervical ultrasound parameters were compared using the vaginal delivery group as the reference. Normality of variables was tested using the Kolmogorov–Smirnov test. Normally distributed continuous variables were compared using the Student t test, and non‐normal distributed parameters were compared with the Mann‐Whitney U test. Categorical variables were compared using the chi‐square test or Fisher exact test as appropriate. The intra‐ and interobserver reproducibility was assessed by the intraclass correlation coefficient (ICC) and Bland‐Altman graphs. SWE values between different cervical regions were compared with paired Wilcoxon signed rank test. A backward stepwise conditional elimination method was used to generate the regression model and to determine the independent predictors for all CS and for CS due to failure to enter active labor.21 Receiver‐operating characteristics (ROC) curves were then constructed for the regression models to determine their discriminative ability. The optimal cutoff was determined by the Youden index.22 The area under curve (AUC) was compared using the DeLong test. Statistical Package for Social Science (SPSS) version 20.0. (IBM Corp., NY, USA) and medcalc statistical Software version 18 (MedCalc Software bvba, Ostend, Belgium) were used for the statistical analysis. A two‐tailed P value <0.05 was considered statistically significant.

rve (AUC) was compared using the DeLong test. Statistical Package for Social Science (SPSS) version 20.0. (IBM Corp., NY, USA) and medcalc statistical Software version 18 (MedCalc Software bvba, Ostend, Belgium) were used for the statistical analysis. A two‐tailed P value <0.05 was considered statistically significant. 2.7 Ethical approval This study was approved by the Joint Chinese University of Hong Kong‐New Territories East Cluster Clinical Research Ethics Committee on 21 April 2015 (CRE: 2015.141). Written informed consent was obtained from all participants. 3 RESULTS A total of 500 pregnant women were recruited, of which 25 were excluded because 12 cases had early sign of spontaneous onset of labor, 4 cases had CS due to suspected macrosomia and 9 cases declined IOL, leaving 475 cases for IOL. The demographic characteristics of the studied population are shown in Table 1. The primary indications for IOL included post‐term pregnancy in 234 (49.3%), gestational diabetes mellitus in 75 (15.8%), suspected macrosomia in 46 (9.7%), oligohydramnios in 28 (5.9%), intrauterine growth restriction in 27 (5.7%), hypertension in 22 (4.6%), history of precipitate labor in 19 (4%), advanced maternal age in 16 (3.4%) and other reasons in 8 (1.7%). Half of the cases (49.9%) needed cervical ripening due to unfavorable cervix. Following IOL, 393 (82.7%) resulted in vaginal delivery and 82 (17.3%) had emergency CS, of which 40 were due to failure to enter active phase, 20 due to fetal distress, 19 due to failure to progress, and three due to other reasons.

8 (1.7%). Half of the cases (49.9%) needed cervical ripening due to unfavorable cervix. Following IOL, 393 (82.7%) resulted in vaginal delivery and 82 (17.3%) had emergency CS, of which 40 were due to failure to enter active phase, 20 due to fetal distress, 19 due to failure to progress, and three due to other reasons. Table 1 The demographic characteristics of the 475 women who underwent induction of labor Characteristics Value Maternal age (y) 32 (19‐45) Maternal height (cm) 158 (144‐177) BMI at delivery (kg/m2) 27.34 (19.07‐42.83) Nulliparous 274 (57.7%) Gestational age (wk) 40.1 (37‐42) Bishop score ≥6 238 (50.1%) Birthweight (g) 3427 (1966‐4195) Note Data are given as median (range) or n (%). Abbreviation: BMI, body mass index. John Wiley & Sons, LtdThe ICCs of intra‐ and interobserver reproducibility were >0.85 in each ROI (Table S1, Figures S1 and S2). Table 2 gives the SWE values at each ROI and shows that an elastic gradient exists along the longitudinal axis, with the inner part being significantly stiffer than the middle part, and the middle part being significantly stiffer than the outer part, along both the anterior (5.4 kPa vs 4.8 kPa vs 3.8 kPa; all P < 0.001) and posterior lips (5.0 kPa vs 4.7 kPa vs 3.9 kPa; all P < 0.001). The SWE values at different ROI are also significantly intercorrelated with each other (Spearman coefficients are shown in Table 3) (all P < 0.001). Hence, for subsequent comparison we used the inner cervical SWE (the mean SWE of the inner anterior and inner posterior cervix), the stiffest region.

kPa; all P < 0.001). The SWE values at different ROI are also significantly intercorrelated with each other (Spearman coefficients are shown in Table 3) (all P < 0.001). Hence, for subsequent comparison we used the inner cervical SWE (the mean SWE of the inner anterior and inner posterior cervix), the stiffest region. Table 2 Shear wave elastic values at different cervical regions (kPa) Region Inner part Middle part Outer part Inner vs Middle Middle vs Outer Anterior cervical lip 5.4 (4.3‐6.5) 4.8 (3.8‐5.6) 3.8 (3.1‐4.7) P < 0.001 P < 0.001 Posterior cervical lip 5.0 (4.0‐6.0) 4.7 (3.8‐5.7) 3.9 (3.1‐4.7) P < 0.001 P < 0.001 Note Data are given as median (interquartile). The data were compared with paired Wilcoxon signed rank test. John Wiley & Sons, LtdTable 3 The Spearman coefficients between shear wave elastic values at different regions Region Anterior cervical lip Posterior cervical lip Middle Outer Inner Middle Outer Anterior cervical lip Inner 0.77 0.566 0.633 0.509 0.41 Middle — 0.7 0.548 0.539 0.419 Outer — — 0.544 0.536 0.57 Posterior cervical lip Inner — — — 0.781 0.618 Middle — — — — 0.649 Note The Spearman correlation was performed. All P < 0.001.

n Anterior cervical lip Posterior cervical lip Middle Outer Inner Middle Outer Anterior cervical lip Inner 0.77 0.566 0.633 0.509 0.41 Middle — 0.7 0.548 0.539 0.419 Outer — — 0.544 0.536 0.57 Posterior cervical lip Inner — — — 0.781 0.618 Middle — — — — 0.649 Note The Spearman correlation was performed. All P < 0.001. John Wiley & Sons, LtdComparison of maternal characteristics, the BS, fetal and cervical sonographic measurements between the vaginal delivery group and the whole group of CS is illustrated in Table 4. Table 5 shows the odds ratio (OR) of variables in the univariate analysis and adjusted odds ratio (AOR) in the multivariate analysis for the prediction of CS. The results indicated that the significant independent predictors, in order of strength, multiparity (AOR 0.102, 95% confidence interval [CI] 0.048‐0.22), cervical length (AOR 1.717, 95% CI 1.077‐1.663), inner cervical elasticity (AOR 1.338, 95% CI 1.001‐1.598) and maternal height (AOR 0.894, 95% CI 0.845‐0.946). Combining these four factors, the AUC for the prediction of CS was 0.815 (95% CI 0.777‐0.85) (Figure 4). Body mass index (BMI) ≥0 kg/m2, Bishop score and AOP were not independent predictors. Table 4 The comparison of maternal characteristics, Bishop score, fetal and cervical sonographic measurements between the vaginal delivery group and cesarean group Factors Vaginal delivery (n = 393) cesarean group (n = 82) P

John Wiley & Sons, LtdComparison of maternal characteristics, the BS, fetal and cervical sonographic measurements between the vaginal delivery group and the whole group of CS is illustrated in Table 4. Table 5 shows the odds ratio (OR) of variables in the univariate analysis and adjusted odds ratio (AOR) in the multivariate analysis for the prediction of CS. The results indicated that the significant independent predictors, in order of strength, multiparity (AOR 0.102, 95% confidence interval [CI] 0.048‐0.22), cervical length (AOR 1.717, 95% CI 1.077‐1.663), inner cervical elasticity (AOR 1.338, 95% CI 1.001‐1.598) and maternal height (AOR 0.894, 95% CI 0.845‐0.946). Combining these four factors, the AUC for the prediction of CS was 0.815 (95% CI 0.777‐0.85) (Figure 4). Body mass index (BMI) ≥0 kg/m2, Bishop score and AOP were not independent predictors. Table 4 The comparison of maternal characteristics, Bishop score, fetal and cervical sonographic measurements between the vaginal delivery group and cesarean group Factors Vaginal delivery (n = 393) cesarean group (n = 82) P Maternal age (≥35 y)a 139 (35.4%) 40 (48.8%) 0.023 Maternal height (cm)b 158 (155‐162) 155 (152‐159) <0.001 BMI at delivery (≥30 kg/m2)a 80 (20.4%) 30 (36.6%) 0.002 Multiparousa 192 (48.9%) 9 (11%) <0.001 Bishop scoreb 6 (4‐6) 3 (3‐4.5) <0.001 EFW (g)b 3325 (3073‐3566) 3556 (3095‐3783) 0.013 EFW <10th centilea 28 (7.1%) 8 (9.8%) 0.395 UA PIc 0.77 (.67‐.87) 0.76 (.67‐.85) 0.662 MCA PIc 1.35 (1.1‐1.59) 1.35 (1.08‐1.51) 0.681 CPRc 1.75 (1.45‐2.09) 1.73 (1.41‐2.14) 0.946 Cervical length (cm)c 2.4 (1.6‐3.0) 2.9 (2.2‐3.5) <0.001 Posterior cervical angle (°)c 112 (98‐125) 109 (91‐124) 0.04 Angle of progression (°)c 87 (80‐96) 84 (75‐88) <0.001 Inner cervical SWE (kPa)c 5.1 (4.2‐6.0) 5.8 (4.9‐7.0) <0.001 Data are given as median (interquartile range) or n (%).

‐2.09) 1.73 (1.41‐2.14) 0.946 Cervical length (cm)c 2.4 (1.6‐3.0) 2.9 (2.2‐3.5) <0.001 Posterior cervical angle (°)c 112 (98‐125) 109 (91‐124) 0.04 Angle of progression (°)c 87 (80‐96) 84 (75‐88) <0.001 Inner cervical SWE (kPa)c 5.1 (4.2‐6.0) 5.8 (4.9‐7.0) <0.001 Data are given as median (interquartile range) or n (%). Abbreviations: BMI, body mass index; CPR, cerebroplacental ratio; EFW, estimated fetal weight; inner cervical SWE, mean of shear wave elasticity of anterior and posterior inner cervix; MCA, middle cerebral artery; PI, pulsatility index; UA, umbilical artery. a Chi‐square test or Fisher exact test as appropriate. b Mann‐Whitney U test. c Student t test. John Wiley & Sons, LtdTable 5 Univariate analysis and multivariate analysis for prediction of cesarean delivery Variables Univariate analysis Multivariate analysis Odds ratio (95% CI) P Adjusted odds ratio (95% CI) P Maternal height 0.907 (.864‐.952) <0.001 0.894 (0.845‐0.946) <0.001 BMI ≥30 kg/m2 2.257 (1.353‐3.767) 0.002 — — Multiparous 0.129 (.063‐.265) <0.001 0.102 (0.048‐0.22) <0.001 EFW 1.001 (1.000‐1.001) 0.023 — — Bishop score 0.605 (.515‐.71) <0.001 — — Cervical length 1.916 (1.451‐2.530) <0.001 1.717 (1.183‐2.492) 0.004 Angle of progression 0.953 (.931‐.974) <0.001 — — Inner cervical SWE 1.43 (1.214‐1.684) <0.001 1.338 (1.077‐1.663) 0.009 Binary logistic regression was performed. Equation 1: Loge (odds) = 13.686 – 2.279*parity (0 for nulliparous, 1 for multiparous)‐ 0.112*mat height + 0.541*cervical length + 0.291* inner cervical SWE.

Maternal height 0.907 (.864‐.952) <0.001 0.894 (0.845‐0.946) <0.001 BMI ≥30 kg/m2 2.257 (1.353‐3.767) 0.002 — — Multiparous 0.129 (.063‐.265) <0.001 0.102 (0.048‐0.22) <0.001 EFW 1.001 (1.000‐1.001) 0.023 — — Bishop score 0.605 (.515‐.71) <0.001 — — Cervical length 1.916 (1.451‐2.530) <0.001 1.717 (1.183‐2.492) 0.004 Angle of progression 0.953 (.931‐.974) <0.001 — — Inner cervical SWE 1.43 (1.214‐1.684) <0.001 1.338 (1.077‐1.663) 0.009 Binary logistic regression was performed. Equation 1: Loge (odds) = 13.686 – 2.279*parity (0 for nulliparous, 1 for multiparous)‐ 0.112*mat height + 0.541*cervical length + 0.291* inner cervical SWE. Abbreviations: BMI, body mass index; EFW, estimated fetal weight; inner cervical SWE, mean of shear wave elasticity of anterior and posterior inner cervix. John Wiley & Sons, LtdFigure 4 The prediction of all cesarean deliveries. ROC curve for the prediction of all cesarean deliveries after induction of labor with AUC of 0.815 (95% CI 0.777‐0.85)

Abbreviations: BMI, body mass index; EFW, estimated fetal weight; inner cervical SWE, mean of shear wave elasticity of anterior and posterior inner cervix. John Wiley & Sons, LtdFigure 4 The prediction of all cesarean deliveries. ROC curve for the prediction of all cesarean deliveries after induction of labor with AUC of 0.815 (95% CI 0.777‐0.85) Table 6 shows the comparison of maternal characteristics, and fetal ultrasound parameters and cervical measurements between the group of vaginal delivery and the three subgroups of CS for different indications. Compared with the group of vaginal delivery, women who had CS for failure to enter active phase had a significantly higher inner cervical SWE value (median 6.9 kPa vs 5.1 kPa; P < 0.001), as well as a longer cervix, smaller AOP and PCA. However, none of these sonographic parameters was different when comparing the vaginal delivery group with either the subgroup of CS for fetal distress (5.2 kPa) or the subgroup of CS for failure to progress in the active phase (5.0 kPa). Women who failed to enter active phase were also significantly shorter, more obese, having a lower prevalence of multiparity, lower Bishop score, a higher EFW by univariate analysis (Table 6). After multivariate analysis, only parity, cervical length and inner cervical SWE were independent predictors for failure to enter active phase (Table 7), with an AUC of 0.888 (95% CI 0.853‐0.916; Figure 5). If the two cervical ultrasound measurements were replaced by the Bishop score, the AUC significantly dropped to 0.819 (95% CI 0.778‐0.855). The difference between two AUCs was 0.0687 with a 95% CI of 0.0175‐0.12 (DeLong test: z = 2.631, P = 0.009). As the multi‐parity is the most significant predictor for success of IOL, we further focused on the nulliparous subgroup, and found that sonographic prediction was even stronger than BS among nulliparous women (AUC 0.816, 95% CI 0.759‐0.864 vs 0.68, 95% CI 0.615‐0.74; P = 0.0054) (Figure 6). The sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio and negative likelihood ratio of the regression models in all women and in the nulliparous subgroup are shown in Table 8.

women (AUC 0.816, 95% CI 0.759‐0.864 vs 0.68, 95% CI 0.615‐0.74; P = 0.0054) (Figure 6). The sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio and negative likelihood ratio of the regression models in all women and in the nulliparous subgroup are shown in Table 8. Table 6 The comparison of maternal characteristics, Bishop score, fetal and cervical ultrasonic measurements between the vaginal delivery group and different groups of cesarean section Factors Vaginal delivery (reference) (n = 393) cesarean for failure to enter active phase cesarean for failure to progress in active phase cesarean for fetal distress (n = 40) P (n = 19) P (n = 20) P

Table 6 The comparison of maternal characteristics, Bishop score, fetal and cervical ultrasonic measurements between the vaginal delivery group and different groups of cesarean section Factors Vaginal delivery (reference) (n = 393) cesarean for failure to enter active phase cesarean for failure to progress in active phase cesarean for fetal distress (n = 40) P (n = 19) P (n = 20) P Maternal age (≥35 y)a 139 (35.4%) 18 (45%) 0.227 9 (47.4%) 0.287 11 (55%) 0.075 Maternal height (cm)b 158 (155‐162) 156 (154‐160) 0.062 154 (152‐158) 0.002 153 (150‐159) <0.001 BMI at IOL (≥30 kg/m2)a 80 (20.4%) 16 (40.0%) 0.004 6 (31.6%) 0.25 6 (30.0%) 0.394 Multiparousa 192 (48.9%) 2 (5%) <0.001 3 (15.8%) 0.005 3 (15%) 0.003 Bishop scoreb 6 (4‐6) 3 (2‐4) <0.001 4 (3‐5) 0.003 4 (3‐6) 0.028 EFW (g)b 3325 (3073‐3566) 3527 (3276‐3770) 0.008 3641 (3511‐3826) <0.001 3082 (2871‐3498) 0.111 EFW <10th centilea 26 (7.0%) 1 (2.5%) 0.498 0 0.617 7 (36.8%) <0.001 UA PIc 0.77 (0.67‐0.87) 0.75 (0.66‐0.85) 0.521 0.79 (0.70‐0.85) 0.75 0.76 (0.67‐0.80) 0.558 MCA PIc 1.35 (1.10‐1.59) 1.39 (1.22‐1.66) 0.087 1.16 (1.03‐1.47) 0.084 1.17 (.99‐1.40) 0.02 CPRc 1.76 (1.45‐2.09) 1.89 (1.56‐2.21) 0.051 1.47 (1.27‐1.93) 0.093 1.64 (1.29‐2.09) 0.19 Cervical length (cm)c 2.4 (1.6‐3.0) 3.3 (2.7‐3.7) <0.001 2.5 (1.6‐3.0) 0.795 2.4 (2.1‐3.5) 0.122 Posterior cervical angle (°)c 112 (98‐125) 102 (84‐121) 0.018 121 (106‐128) 0.6 110 (95‐126) 0.786 Angle of progression (°)c 87 (80‐96) 81 (73‐88) <0.001 86 (82‐88) 0.081 85 (79‐86) 0.101 Inner cervical SWEc 5.1 (4.2‐6.0) 6.9 (5.5‐7.6) <0.001 5.0 (4.6‐5.7) 0.937 5.2 (4.6‐5.9) 0.793 All the comparisons were made with in the vaginal delivery group. Data are given as median (interquartile range) or n (%).

95‐126) 0.786 Angle of progression (°)c 87 (80‐96) 81 (73‐88) <0.001 86 (82‐88) 0.081 85 (79‐86) 0.101 Inner cervical SWEc 5.1 (4.2‐6.0) 6.9 (5.5‐7.6) <0.001 5.0 (4.6‐5.7) 0.937 5.2 (4.6‐5.9) 0.793 All the comparisons were made with in the vaginal delivery group. Data are given as median (interquartile range) or n (%). Abbreviations: BMI, body mass index; CPR, cerebroplacental ratio; EFW, estimated fetal weight; MCA, middle cerebral artery; PI, pulsatility index; UA, umbilical artery. a Chi‐square test or Fisher exact test as appropriate. b Mann‐Whitney U test. c Student t test. John Wiley & Sons, LtdTable 7 Univariate analysis and multivariate analysis for prediction of cesarean delivery for failure to enter active phase Variables Univariate analysis Multivariate analysis Odds ratio (95% CI) P Adjusted odds ratio (95% CI) P Multiparous 0.055 (0.013‐0.232) <0.001 0.029 (0.006‐0.142) <0.001 Bishop score 0.52 (0.411‐0.658) <0.001 — — Cervical length 3.019 (1.981‐4.602) <0.001 2.556 (1.462‐4.467) 0.001 Angle of progression 0.942 (0.914‐0.971) <0.001 — — Inner cervical SWE 1.825 (1.448‐2.299) <0.001 1.689 (1.234‐2.311) 0.001 Binary logistic regression was performed. Equation 2: Loge (odds) = −7.228 – 3.533*parity (0 for nulliparous, 1 for multiparous) + 0938*cervical length + 0.524* inner cervical SWE. Abbreviation: SWE, shear wave elastography.

Multiparous 0.055 (0.013‐0.232) <0.001 0.029 (0.006‐0.142) <0.001 Bishop score 0.52 (0.411‐0.658) <0.001 — — Cervical length 3.019 (1.981‐4.602) <0.001 2.556 (1.462‐4.467) 0.001 Angle of progression 0.942 (0.914‐0.971) <0.001 — — Inner cervical SWE 1.825 (1.448‐2.299) <0.001 1.689 (1.234‐2.311) 0.001 Binary logistic regression was performed. Equation 2: Loge (odds) = −7.228 – 3.533*parity (0 for nulliparous, 1 for multiparous) + 0938*cervical length + 0.524* inner cervical SWE. Abbreviation: SWE, shear wave elastography. John Wiley & Sons, LtdFigure 5 The prediction of cesarean section for failure to enter active phase. ROC curves compare the predictive ability of parity with ultrasonographic measurement (cervical length with inner cervical SWE, black line: AUC 0.888 (95% C: 0.853‐0.916) and parity with Bishop score (dashed line: AUC 0.819 (95% CI 0.778‐0.855) (P = 0.009). The diagnostic odds ratio is 17.41 (sensitivity of 82.5% and specificity of 78.7%) and 9.65 (sensitivity of 80% and specificity of 70.7%), respectively [Color figure can be viewed at http://wileyonlinelibrary.com]

95% C: 0.853‐0.916) and parity with Bishop score (dashed line: AUC 0.819 (95% CI 0.778‐0.855) (P = 0.009). The diagnostic odds ratio is 17.41 (sensitivity of 82.5% and specificity of 78.7%) and 9.65 (sensitivity of 80% and specificity of 70.7%), respectively [Color figure can be viewed at http://wileyonlinelibrary.com] Figure 6 The prediction of cesarean section for failure to enter active phase among nulliparous women. ROC curves compare the predictive ability of sonographic measurement (cervical length with inner cervical SWE (black line: AUC 0.816, 95% CI 0.759‐0.864) and Bishop score (dashed line: AUC 0.68, 95% CI 0.615‐0.74) (P = 0.0054). The diagnostic odds ratio is 12.34 (sensitivity of 70.0% and specificity of 84.1%) and 3.80 (sensitivity of 65.0% and specificity of 67.2%), respectively [Color figure can be viewed at http://wileyonlinelibrary.com] Table 8 The screening performance of different predictors of failure to enter active labor Predictor AUC (95% CI) Cutoff Sensitivity Specificity PPV NPV +LR −LR Parity + CL + inner cervical SWE 0.888 (0.853‐.916) >0.1031 82.5% 78.7% 29.2% 97.7% 3.88 0.22 Parity + Bishop score 0.819 (0.778‐.855) >0.0989 80.0% 70.7% 21.8% 97.2% 2.73 0.28 Nulliparity CL + inner cervical SWE 0.816 (0.759‐.864) >0.2247 7.0% 84.1% 38.4% 92.3% 4.4 0.36 Bishop score 0.680 (0.615‐.740) >0.1601 65.0% 67.2% 28.3% 9.6% 1.98 0.52 Receiver‐operating characteristics curves were constructed and the Youden index was used to determine the optimal cutoff.

0% 70.7% 21.8% 97.2% 2.73 0.28 Nulliparity CL + inner cervical SWE 0.816 (0.759‐.864) >0.2247 7.0% 84.1% 38.4% 92.3% 4.4 0.36 Bishop score 0.680 (0.615‐.740) >0.1601 65.0% 67.2% 28.3% 9.6% 1.98 0.52 Receiver‐operating characteristics curves were constructed and the Youden index was used to determine the optimal cutoff. Abbreviations: AUC, area under curve; CL, cervical length; inner cervical SWE, mean of shear wave elasticity of anterior and posterior inner cervix; LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value. John Wiley & Sons, LtdIn the subgroup of CS indicated for failure to progress in the active phase, EFW was significantly higher and the Bishop score lower than in the vaginal delivery group. The BMI was also higher and maternal height less, and there were fewer multiparous women. In the subgroup of CS for fetal distress, the MCA PI were significantly lower, the proportion with EFW below 10th percentile was higher, the mothers were shorter and there were fewer multiparous women; the BS was also lower, all based on univariate analysis (Table 6).

eight less, and there were fewer multiparous women. In the subgroup of CS for fetal distress, the MCA PI were significantly lower, the proportion with EFW below 10th percentile was higher, the mothers were shorter and there were fewer multiparous women; the BS was also lower, all based on univariate analysis (Table 6). 4 DISCUSSION This is the first study using SWE to predict the outcome of IOL, and it demonstrated that (1) the stiffness of the cervix decreases towards the outer cervix; (2) the inner cervical SWE and cervical length are independent predictors for overall CS, as well as for the subgroup of CS indicated for failure to enter active phase, but AOP, PCA and the BS are not; (3) a model using the combination of cervical length, inner cervical SWE, parity and maternal height, can achieve an AUC of .815 in the prediction of overall CS after IOL; and, with the former three factors, an AUC of .888 for CS for failure to enter active phase.

to enter active phase, but AOP, PCA and the BS are not; (3) a model using the combination of cervical length, inner cervical SWE, parity and maternal height, can achieve an AUC of .815 in the prediction of overall CS after IOL; and, with the former three factors, an AUC of .888 for CS for failure to enter active phase. Our finding of decreasing stiffness from the inner to the outer part of the cervix is concordant with several studies that have shown the spatial heterogeneity in the stiffness within the cervix using SWE.15 This has been hypothesized to be attributable to the cervical collagen fiber orientation.23 The collagen cross‐link around the internal os is significantly more heterogeneous than that around the external os, and therefore the stroma around the internal os functions distinctively from the external os.23 Hernandez‐Andrade et al found that the stiffness of the inner cervix is more predictive of spontaneous preterm delivery.24 They showed that a hard internal os at 16‐24 weeks is 80% less likely to have spontaneous preterm delivery compared with a soft internal os.24 In a small cohort study, a hard internal os was associated with the failure of IOL.10 Therefore, the inner cervical SWE was selected in the regression analysis. Besides the objective measurement of the cervical stiffness, SWE also has a potential advantage over manual examination, as the latter cannot easily access the innermost part of the cervix.

d internal os was associated with the failure of IOL.10 Therefore, the inner cervical SWE was selected in the regression analysis. Besides the objective measurement of the cervical stiffness, SWE also has a potential advantage over manual examination, as the latter cannot easily access the innermost part of the cervix. In contrast to previous studies, our study showed that SWE is useful in the prediction of overall CS following IOL. So far, only a few small‐scale studies have attempted to evaluate elastography in predicting the outcomes of IOL, and their results are controversial.10, 11, 12, 13 Pereira et al11 concluded that elastography is not useful in predicting the IOL outcome. However, our study differs from Pereira's in many significant ways. First of all, they used a semiquantitative strain‐based elastography, which relies on the internal organ movement, whereas the shear wave elastography we used has the advantage of quantifying the cervical stiffness independent of adjacent tissues and operators’ movements or internal organ movement. Secondly, they focused on a small spot at the internal os of the canal, which is the gland or sometimes the mucus, but we surveyed the cervical elasticity on the stroma, which contributes to the mechanical strength23 and then selected the stiffest inner cervix as the reference.25 Thirdly, they recruited only 99 pregnant women, whereas the sample size of our cohort is five times larger. A recent meta‐analysis26 combined the findings of four small‐cohort studies of a total of 323 subjects,12, 13, 27, 28 and suggested that strain‐based elastography might be predictive of successful IOL. However, the reported AUC of cervical elastography was only 0.55, which was no better than that of the BS (0.51) and much poorer than that of cervical length (0.70). Our findings are superior to the meta‐analysis in several ways. First, the heterogeneity among the reviewed studies could impact the power of the meta‐analysis. Secondly, we were able to show that both cervical elastography and cervical length are independent factors, whereas BS is not. Finally, we achieved a relatively high AUC of 0.815 and 0.888, respectively, for the prediction of overall CS after IOL and that of the subgroup requiring CS for failure to enter active phase of labor.

, we were able to show that both cervical elastography and cervical length are independent factors, whereas BS is not. Finally, we achieved a relatively high AUC of 0.815 and 0.888, respectively, for the prediction of overall CS after IOL and that of the subgroup requiring CS for failure to enter active phase of labor. Even after excluding multi‐parity, which is the strongest predictor of IOL outcome, and focus on nulliparous women, the combination of cervical elastography and cervical length was even better than BS with the difference of 0.136 between two AUCs We also found that PCA and AOP, which are respectively the proxies of cervical position and fetal head station in the BS, are no longer independent predictors when SWE is included. This result provides further evidence of intercorrelation between the different components of the BS.29 As shown in our comparison of the regression models (Figure 5, Table 8), sonographic measurement of cervical length and SWE is superior to manual assessment of the BS in predicting failure to enter active phase. Transvaginal ultrasonic examination also causes less pain than digital examination.30 However, SWE is not yet readily available in routine ultrasound machine, and it is also expensive to purchase such a machine with cutting‐edge technology.

ior to manual assessment of the BS in predicting failure to enter active phase. Transvaginal ultrasonic examination also causes less pain than digital examination.30 However, SWE is not yet readily available in routine ultrasound machine, and it is also expensive to purchase such a machine with cutting‐edge technology. Our subgroup analysis showed that the SWE and other sonographic parameters are not useful in predicting CS indicated due to fetal distress or failure to progress. This is biologically understandable, as fetal distress is unrelated to cervical favorability but rather restricted fetal growth and fetal compromise, whereas failure to progress is related more to large fetal size, as reflected from our results (Table 6). Our findings also indicate that it is not straightforward to create a prediction model for all CS. Whereas a large fetal weight increases the chance of CS for slow progress, a small fetus is associated with CS for fetal distress. The effect of fetal weight may be masked when overall CS is the primary outcome.

Table 6). Our findings also indicate that it is not straightforward to create a prediction model for all CS. Whereas a large fetal weight increases the chance of CS for slow progress, a small fetus is associated with CS for fetal distress. The effect of fetal weight may be masked when overall CS is the primary outcome. The major strengths of our study are that, by measuring elasticity in different regions of the cervix, we demonstrated that the inner part of the cervix is the most useful predictor of different regions of the cervix. The large sample size from a homogeneous ethnic group is another advantage of our study. However, the overall number of CS of 80 can only allow a maximum of eight variables for multivariate analysis. Therefore we could only select the eight strongest variables based on univariate analysis.21 Nonetheless, our study has tested multiple clinical and ultrasonic variables, of which the combination has significantly improved the prediction compared with using clinical variables alone.31 The choice of the method of IOL was based on the BS alone. It is worth investigating in future research whether SWE may provide a better guide of IOL method and improve the chances of success.

c variables, of which the combination has significantly improved the prediction compared with using clinical variables alone.31 The choice of the method of IOL was based on the BS alone. It is worth investigating in future research whether SWE may provide a better guide of IOL method and improve the chances of success. 5 CONCLUSION Shear wave elastography is a useful tool in pre‐IOL assessment of the stiffness of the cervix, which is an independent predictor of overall CS, and specifically CS indicated for the failure to enter active phase. PCA, AOP and the Bishop score were not independent predictors of CS. The combination of sonographic cervical length and shear‐wave elastography is superior to the Bishop score in predicting failure of IOL. CONFLICT OF INTEREST The authors have stated explicitly that there are no conflicts of interest in connection with this article. Supporting information Click here for additional data file. Click here for additional data file. Click here for additional data file. ACKNOWLEDGMENTS We thank Ms WM Yuan and Ms HW Cheung for their contribution in the case recruitment.

Abbreviations CScesarean section DLTdouble‐layer technique RMTresidual myometrial thickness SLTsingle‐layer technique Key message Deficient uterine scar healing represents a side effect with potential negative long‐term consequences. This study demonstrates that a double‐layer technique with the first continuous nonlocking suture followed by a second continuous nonlocking suture is associated with better uterine suture healing. 1 INTRODUCTION There are multiple factors driving the increase in cesarean section (CS) rates internationally. Demographic factors in the economically developed world partly explain the rise.1, 2 Deficient uterine scar healing represents a side effect with negative consequences. Serious obstetric complications may occur in the subsequent pregnancy such as uterine scar dehiscence (0.6‐3.8%), uterine scar rupture (0.2‐3.8%) and cesarean scar pregnancy, which may be associated with morbidly adherent placenta.3, 4 In the long‐term, women with a scar defect may also suffer from gynecological problems.5, 6, 7 Transvaginal ultrasound is validated tool to evaluate uterine scar defects commonly referred to as “niche”.8, 9 Healing and scar maturation are influenced by the suture technique, number of previous CS deliveries and patient's medical and obstetric history.10 There are many techniques for the closure of the uterine incision. In the short‐term, no clear benefit of any of the randomized comparisons has been shown. The objective of this study was to compare the effect of single‐ vs double‐layer closure technique on uterine scar morphology in primiparas after CS.

ric history.10 There are many techniques for the closure of the uterine incision. In the short‐term, no clear benefit of any of the randomized comparisons has been shown. The objective of this study was to compare the effect of single‐ vs double‐layer closure technique on uterine scar morphology in primiparas after CS. 2 MATERIAL AND METHODS A prospective randomized study was conducted between November 2011 and September 2014 in the Institute for the Care of Mother and Child in Prague (tertiary perinatological center). The study included a cohort of nulliparous women with a singleton pregnancy who underwent first delivery by any type of CS. Women were invited for three consecutive control visits at 6 weeks, 6 months and 12 months postpartum. Women who did not complete all postoperative visits were excluded from the analysis.

center). The study included a cohort of nulliparous women with a singleton pregnancy who underwent first delivery by any type of CS. Women were invited for three consecutive control visits at 6 weeks, 6 months and 12 months postpartum. Women who did not complete all postoperative visits were excluded from the analysis. The goal of the proposed study is to test the null hypothesis that the mean thickness of the uterine scar in the two groups is equal. Women with a uterine scar defect had thinner full lower uterine segment and thinner myometrial layer. The optimal cut‐off value varied from 2.0 to 3.5 mm for full lower uterine segment thickness and from 1.4 to 2.0 mm for myometrial layer.11 We calculated the sample size to observe a difference of 1.0 mm of myometrial thickness between uterine closures. For the comparison of the influence of both surgical techniques on muscle layer thickness we used a two‐tailed test, which means that an effect in either direction will be interpreted. The significance level for a test of the null hypothesis was set at 0.050 (alpha). With the proposed sample size of 103 each for the two groups, the study would have a power of 90.1% to yield a statistically significant result. This computation assumes that the mean difference for myometrial layer thickness is 1.0 mm and the common within‐group standard deviation is 2.2 mm (on the basis of our feasibility studies). This effect was selected as the smallest effect that would be important to detect, in the sense that any smaller effect would not be of clinical or substantive significance. It was also assumed that this effect size was reasonable, in the sense that an effect of this magnitude could be anticipated in this field of research. Women indicated for CS were randomly allocated after opening a consecutively numbered envelope containing the uterine closure technique group, either single‐layer technique (SLT) or double‐layer technique (DLT) closure of the hysterotomy. This study represents a subanalysis from a large prospective cohort of healthy women with a first singleton pregnancy who delivered at or beyond 37 weeks. The SLT involved a single continuous nonlocking suture including decidua. In DLT the first layer comprised a continuous nonlocking suture including decidua, followed by a second continuous, nonlocking suture (Figure 1). In both techniques, we used 0/0 polyglactin 910 suturing material with a blunt needle (Vicryl, Ethicon®, Diegem, Belgium).

gle continuous nonlocking suture including decidua. In DLT the first layer comprised a continuous nonlocking suture including decidua, followed by a second continuous, nonlocking suture (Figure 1). In both techniques, we used 0/0 polyglactin 910 suturing material with a blunt needle (Vicryl, Ethicon®, Diegem, Belgium). Additional hemostasis sutures could be laid regardless of the closure method. All obstetricians who performed the surgery had similar experience, and those with less training were always supervised by a senior attending physician. Figure 1 Uterine suture technique. (A) Unlocked single‐layer closure. Decidua was incorporated in the suture. The uterine serosa is not included in the suture. (B) Double‐layer closure. First layer unlocked suture including decidua. Second layer unlocked suture taking superficial part of myometrium. Uterine serosa is not included in the suture [Color figure can be viewed at http://wileyonlinelibrary.com]

porated in the suture. The uterine serosa is not included in the suture. (B) Double‐layer closure. First layer unlocked suture including decidua. Second layer unlocked suture taking superficial part of myometrium. Uterine serosa is not included in the suture [Color figure can be viewed at http://wileyonlinelibrary.com] During the follow‐up control visits at 6 weeks, 6 and 12 months, all patients underwent a three‐dimensional transvaginal ultrasound of the uterus using a GE Voluson E8 Expert ultrasound system (General Electric, Zipf, Austria) equipped with a 2.8‐10 MHz transvaginal probe in the lithotomy position with an empty urinary bladder. Imaging was performed in the mid‐sagittal plane with the angle of acquisition set at 120°. Two volume datasets in longitudinal and transverse sections were acquired and stored for later analysis using the software 4d view (General Electric Medical Kretz technik, Zipf, Austria). Data analysis were undertaken by two of the authors (J.H. and L.K.), blinded to clinical data. The following sonographic features were assessed: position of the uterus (anteflexion or retroflexion), visibility of the CS scar, presence of a scar defect (yes or no); any visible defect in the scar was classified as a defect. The following scar measurements were taken: total myometrial thickness, residual myometrial thickness (RMT), scar width, the distance between the scar and the external cervical os. We also noted myometrial defects with or without contact with the uterine cavity (scar defect character) (Figure 2).

lassified as a defect. The following scar measurements were taken: total myometrial thickness, residual myometrial thickness (RMT), scar width, the distance between the scar and the external cervical os. We also noted myometrial defects with or without contact with the uterine cavity (scar defect character) (Figure 2). Figure 2 (A) Transvaginal ultrasound demonstrating measurement of total myometrial thickness (1) and residual myometrial thickness (2). (B) Schematic diagram showing CS scar placement and dimensions measurement: total myometrial thickness (1), residual myometrial thickness (2), width of the scar defect (3), distance between the scar and the external cervical ostium (4), external cervical ostium (5), myometrial defects without contact with the uterine cavity (6) [Color figure can be viewed at http://wileyonlinelibrary.com] Primary outcome was the mean RMT related to the SLT or DLT. Secondary outcomes included: frequency of uterine defects, the position of the uterus, frequency of uterine defects with respect to the position of the uterus, position of the scar and frequency of defects with respect to the stage of labor in which the CS was performed. We recorded the incidence of selected postoperative complications at 6 weeks postpartum.

f uterine defects, the position of the uterus, frequency of uterine defects with respect to the position of the uterus, position of the scar and frequency of defects with respect to the stage of labor in which the CS was performed. We recorded the incidence of selected postoperative complications at 6 weeks postpartum. Statistical analyses were performed using SPSS software version 13.0 (IBM Corp., Armonk, NY, USA). Participant baseline and ultrasound characteristics were presented. The SLT and DLT group were compared. For continuous, normally distributed variables, we used Student's t test. The nonparametric tests (Wilcoxon‐Mann‐Whitney test) were used for continuous, non‐normally distributed variables. To test symmetry in the contingency table with dichotomous variables we used Fisher's exact test. A P value <0.05 was considered to be significant. Apart from simple tests, linear model (two‐way ANOVA) was carried out to examine the effect of vaginal findings during indication for CS on the scar. To test the development of categorized variables (including dichotomous variables) over time and dependence on suturing techniques, the generalized linear mixed model was used. 2.1 Ethical approval The design of this study was approved by the institutional ethics committee (ethics committee number 3/2010).

Statistical analyses were performed using SPSS software version 13.0 (IBM Corp., Armonk, NY, USA). Participant baseline and ultrasound characteristics were presented. The SLT and DLT group were compared. For continuous, normally distributed variables, we used Student's t test. The nonparametric tests (Wilcoxon‐Mann‐Whitney test) were used for continuous, non‐normally distributed variables. To test symmetry in the contingency table with dichotomous variables we used Fisher's exact test. A P value <0.05 was considered to be significant. Apart from simple tests, linear model (two‐way ANOVA) was carried out to examine the effect of vaginal findings during indication for CS on the scar. To test the development of categorized variables (including dichotomous variables) over time and dependence on suturing techniques, the generalized linear mixed model was used. 2.1 Ethical approval The design of this study was approved by the institutional ethics committee (ethics committee number 3/2010). 3 RESULTS A total of 540 pregnant women (270 in the SLT, 270 in the DLT) were included in the study. Drop‐out rate was 216 cases (40%) (Figure 3). Table 1 summarizes the basic demographics and obstetric data of the 12‐month visit cohort and those who dropped out. Six weeks postpartum, there were no differences in uterine position, presence or localization of the defect and scar sonomorphology. In all, 324 women attended the 12‐month visit, 149 of whom underwent SLT and 175 DLT. Their demographic and obstetric data did not differ, nor did the incidence of selected complications (Table 2). Repeated observational data on uterine position, presence and type of defects with missing cases are presented in Tables 3, 4, 5, 6. The CS scar measurements at 12 months postpartum are listed in Table 7. Defects in the SLT group were significantly (0.002) wider (4.8 vs 4.0 mm). The RMT in the SLT group were significantly (0.019) thinner (4.6 vs 5.2 mm). At 12 months there were no significant difference in the total myometrial thickness (0.777). One year postpartum, defects in the DLT group were significantly (0.002) closer to the external cervical os (30.0 vs 33.0 mm). Women who underwent CS at the stage of full cervical dilation had scars that were closer to the external cervical os (0.000), RMT was thinner (0.010) and width of the defect smaller (0.001). Other ultrasound scar measurements were not influenced by the vaginal finding at the time of the CS indication.

30.0 vs 33.0 mm). Women who underwent CS at the stage of full cervical dilation had scars that were closer to the external cervical os (0.000), RMT was thinner (0.010) and width of the defect smaller (0.001). Other ultrasound scar measurements were not influenced by the vaginal finding at the time of the CS indication. Figure 3 Flow chart summarizing selection of participants who underwent single‐ or double‐layer uterine suture technique. We observed five protocol violations Table 1 Demographics and obstetric data comparison of study subjects who attended the 12‐month visit and women who dropped out. Ultrasound outcomes comparison of study subjects attending the 12‐month visit and women who dropped out after the 6‐week check‐up Parameter Follow up P 12‐month visit (n = 324) Drop out (n = 216) Demographic and obstetrics details Maternal age (y) 31.7 ± 3.8 31.1 ± 4.2 0.139b BMI (kg/m2) 22.9 (20.2‐24.8) 23.1 (20.5‐27.8) 0.925c Gestational age at delivery (gestational weeks) 40.2 (39‐41) 40.2 (39‐41) 0.928c Previous surgery on the uterus 43 (13.3) 23 (15.0) 0.612d Assisted reproduction: IVF/ICSI 21 (6.5) 9 (5.9) 0.489d Gestational diabetes 12 (3.7) 7 (4.6) 0.409d Hypertensive disorders 11 (3.4) 4 (2.8) 0.443d Type of cesarean section Acute in pregnancy 15 (4.6) 2 (0.9) 0.062d Acute during labor 132 (40.7) 92 (42.6) Planned in pregnancy 39 (12.0) 29 (13.4) Planned during labor 138 (42.6) 93 (43.1) Hysterotomy closure Single‐layer 149 (46.0) 100 (46.3) 0.167d Double‐layer 175 (54.0) 116 (53.7) Cervical dilation No dilation 91 (28.1) 54 (25.0) 0.360d

Type of cesarean section Acute in pregnancy 15 (4.6) 2 (0.9) 0.062d Acute during labor 132 (40.7) 92 (42.6) Planned in pregnancy 39 (12.0) 29 (13.4) Planned during labor 138 (42.6) 93 (43.1) Hysterotomy closure Single‐layer 149 (46.0) 100 (46.3) 0.167d Double‐layer 175 (54.0) 116 (53.7) Cervical dilation No dilation 91 (28.1) 54 (25.0) 0.360d Partial dilation 166 (51.2) 125 (57.9) Full dilation 67 (20.7) 37 (17.1) 6‐week control (n = 153)a Uterine position Anteflexion 176 (54.3) 88 (56.1) 0.570d Retroflexion 148 (45.7) 65 (42.5) CS scar description Intact scar 62 (19.1) 38 (24.8) 0.199d Presence of scar defect 262 (80.9) 115 (75.2) Scar measurements and localization Total myometrial thickness (mm) 11.5 (9.7‐13.0) 10.5 (9.2‐12.1) 0.115c Residual myometrial thickness (mm) 5.8 ± 2.2 5.6 ± 1.9 0.340b Width of the defect (mm) 4.3 (2.9‐5.9) 4.2 (2.8‐5.8) 0.180c Scar‐external cervical os distance (mm) 31.0 (27.0‐35.0) 33.0 (28.0‐35.4) 0.211c Abbreviations: CS, cesarean section; IVF/ICSI, in vitro fertilization/intracytoplasmatic sperm injection. a The 63 women who did not attend the 6‐week check‐up were not included. Characteristics are presented as mean ± SD for normally distributed variables, median and interquartile range for non‐normally distributed variables. Categorical variables were presented as total number (percentage in group). b Student′s t test. c Wilcoxon‐Mann‐Whitney test. d Pearson Chi‐square test.

a The 63 women who did not attend the 6‐week check‐up were not included. Characteristics are presented as mean ± SD for normally distributed variables, median and interquartile range for non‐normally distributed variables. Categorical variables were presented as total number (percentage in group). b Student′s t test. c Wilcoxon‐Mann‐Whitney test. d Pearson Chi‐square test. John Wiley & Sons, LtdTable 2 Demographics, obstetric data and selected complications at 6 week postpartum of study subjects who finished the 12‐month follow up. Characteristics are presented as median and interquartile range for non‐normally distributed variables. Categorical variables are presented as total number (percentage in group) Parameter Closure technique P Single‐layer (n = 149) Double‐layer (n = 175) Demographic and obstetrics details Maternal age (y) 31.0 (29.0‐34.0) 32.0 (29.0‐34.0) 0.392a BMI (kg/m2) 22.4 (20.4‐25.3) 22.3 (20.1‐24.2) 0.602a Gestational age at delivery (gestational weeks) 40.0 (39.0‐41.0) 40.0 (40.0‐41.0) 0.446a Type of cesarean section Acute in pregnancy 8 (5.4) 8 (4.6) 0.850b Acute during labor 62 (41.6) 70 (40.0) Planned in pregnancy 16 (10.7) 22 (12.6) Planned during labor 63 (42.3) 75 (42.9) Cervical dilation No dilation 42 (28.2) 49 (28.0) 0.245b Partial dilation 82 (55.0) 84 (48.0) Full dilation 25 (16.8) 42 (24.0) 6‐week complications None 135 (90.6) 159 (90.9) 0.263b Maternal infectious morbidity 7 (4.7) 7 (4.0) Operative procedures on wound 4 (2.7) 1 (0.6) Other (placental remnants, transfusion) 3 (2.0) 8 (4.6) a Wilcoxon‐Mann‐Whitney test. b Pearson Chi‐square test.

Partial dilation 82 (55.0) 84 (48.0) Full dilation 25 (16.8) 42 (24.0) 6‐week complications None 135 (90.6) 159 (90.9) 0.263b Maternal infectious morbidity 7 (4.7) 7 (4.0) Operative procedures on wound 4 (2.7) 1 (0.6) Other (placental remnants, transfusion) 3 (2.0) 8 (4.6) a Wilcoxon‐Mann‐Whitney test. b Pearson Chi‐square test. John Wiley & Sons, LtdTable 3 Repeated observation data on presence of scar defects with missing cases. Categorical variables are presented as total number (percentage in group) Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Presence of scar defect Yes No Yes No 6 weeksa 187 (81.0) 44 (19.0) 191 (77.6) 55 (22.4) 6 monthsb 148 (78.7) 40 (21.3) 155 (76.4) 48 (23.6) 12 monthsc 124 (83.2) 25 (16.8) 127 (72.6) 48 (27.4) a Total number of women who attended the 6‐week follow up = 477. b Total number of women who attended the 6‐month follow up = 391. c Total number of women who attended the 12‐month follow up = 324. John Wiley & Sons, LtdTable 4 Repeated observation data on uterine position with missing cases. Categorical variables are presented as total number (percentage in group) Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Uterine position Anteflexion Retroflexion Anteflexion Retroflexion 6 weeksa 122 (52.8) 109 (47.2) 141 (57.3) 105 (42.7) 6 monthsb 123 (65.4) 65 (34.6) 151 (74.4) 52 (25.6) 12 monthsc 90 (60.4) 59 (39.6) 125 (71.4) 50 (28.6) a Total number of women who attended the 6‐week follow up = 477. b Total number of women who attended the 6‐month follow up = 391.

Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Uterine position Anteflexion Retroflexion Anteflexion Retroflexion 6 weeksa 122 (52.8) 109 (47.2) 141 (57.3) 105 (42.7) 6 monthsb 123 (65.4) 65 (34.6) 151 (74.4) 52 (25.6) 12 monthsc 90 (60.4) 59 (39.6) 125 (71.4) 50 (28.6) a Total number of women who attended the 6‐week follow up = 477. b Total number of women who attended the 6‐month follow up = 391. c Total number of women who attended the 12‐month follow up = 324. John Wiley & Sons, LtdTable 5 Repeated observation data on scar defects in anteflexion/retroflexion (AVF/RVF) uterus with missing cases. Categorical variables are presented as total number (percentage in group) Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Scar defect in AVF uterus Yes No Yes No 6 weeksa 105 (86.1) 17 (13.9) 116 (82.3) 25 (17.7) 6 monthsb 102 (82.9) 21 (17.1) 117 (77.5) 34 (22.5) 12 monthsc 81 (90.0) 9 (10.0) 90 (72.0) 35 (28.2) Follow up Scar defect in RVF uterus Yes No Yes No 6 weeksa 82 (75.2) 27 (24.8) 75 (71.4) 30 (28.6) 6 monthsb 46 (70.8) 19 (29.2) 38 (73.1) 14 (26.9) 12 monthsc 43 (72.9) 16 (27.1) 37 (74.0) 13 (26.0) a Total number of women who attended the 6‐week follow up = 477. b Total number of women who attended the 6‐month follow up = 391. c Total number of women who attended the 12‐month follow up = 324. John Wiley & Sons, LtdTable 6 Repeated observation data on scar defect presence and defect contact with the uterine cavity with missing cases. Categorical variables are presented as total number (percentage in group)

b Total number of women who attended the 6‐month follow up = 391. c Total number of women who attended the 12‐month follow up = 324. John Wiley & Sons, LtdTable 6 Repeated observation data on scar defect presence and defect contact with the uterine cavity with missing cases. Categorical variables are presented as total number (percentage in group) Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Presence and defect contact with the uterine cavity Yes No Normal scar Yes No Normal scar 6 weeksa 111 (48.1) 76 (32.9) 44 (19.0) 118 (48.0) 73 (29.7) 55 (22.4) 6 monthsb 112 (59.6) 36 (19.1) 40 (21.3) 114 (56.2) 41 (20.2) 48 (23.6) 12 monthsc 84 (56.4) 40 (26.8) 25 (16.8) 92 (52.6) 35 (20.0) 48 (27.4) Notes Yes: scar defect is present and is in contact with the cavity; No: scar defect is present and is not in contact with the cavity, normal scar defect is not present. a Total number of women who attended the 6‐week follow up = 477. b Total number of women who attended the 6‐month follow up = 391. c Total number of women who attended the 12‐month follow up = 324. John Wiley & Sons, LtdTable 7 Sonographic cesarean section (CS) scar measurement and localization at 12‐month follow up and relation to cervical dilation at time of delivery. Characteristics are presented as mean ± SD for normally distributed variables, median and interquartile range for non‐normally distributed variables Parameter Closure technique P Single‐layer (n = 149) Double‐layer (n = 175) CS scar description Scar measurements and localization Total myometrial thickness (mm) 10.0 (8.5‐11.5) 10.0 (8.6‐11.7) 0.777b

John Wiley & Sons, LtdTable 7 Sonographic cesarean section (CS) scar measurement and localization at 12‐month follow up and relation to cervical dilation at time of delivery. Characteristics are presented as mean ± SD for normally distributed variables, median and interquartile range for non‐normally distributed variables Parameter Closure technique P Single‐layer (n = 149) Double‐layer (n = 175) CS scar description Scar measurements and localization Total myometrial thickness (mm) 10.0 (8.5‐11.5) 10.0 (8.6‐11.7) 0.777b Residual myometrial thickness (mm) 4.6 (±1.9) 5.2 (±2.2) 0.019a Width of the defect (mm) 4.8 (3.2‐6.6) 4.0 (3.0‐5.4) 0.002b Scar‐external cervical os distance (mm) 33.0 (29.0‐37.0) 30.0 (25.7‐34.7) 0.002b Residual myometrial thicknes <2.5 mm 18 (12.2) 12 (6.8) 0.019a CS scar description and relation to vaginal finding at CS indication Total myometrial thickness No cervical dilation 10.5 (±2.6) 10.1 (±2.0) 0.533c Partial cervical dilation 10.2 (±2.1) 10.2 (±1.8) Full cervical dilation 8.8 (±2.4) 9.9 (±2.4) Residual myometrial thickness No cervical dilation 4.8 (±1.7) 5.7 (±2.1) 0.010c Partial cervical dilation 4.7 (±2.0) 5.0 (±1.9) Full cervical dilation 4.3 (±2.2) 5.0 (±2.5) Width of the defect No cervical dilation 5.6 (±2.8) 4.3 (±2.1) 0.001c Partial cervical dilation 4.7 (3.2‐6.0) 4.0 (3.2‐5.5) Full cervical dilation 4.0 (2.9‐7.3) 4.0 (3.1‐5.3) Scar‐external cervical os distance No cervical dilation 35.5 (29.3‐37.0) 33.0 (29.0‐39.0) 0.000c Partial cervical dilation 33.6 (±6.1) 30.1 (±5.5) Full cervical dilation 28.9 (±9.0) 27.0 (±5.8) Student′s t test. Wilcoxon‐Mann‐Whitney test.

Partial cervical dilation 4.7 (3.2‐6.0) 4.0 (3.2‐5.5) Full cervical dilation 4.0 (2.9‐7.3) 4.0 (3.1‐5.3) Scar‐external cervical os distance No cervical dilation 35.5 (29.3‐37.0) 33.0 (29.0‐39.0) 0.000c Partial cervical dilation 33.6 (±6.1) 30.1 (±5.5) Full cervical dilation 28.9 (±9.0) 27.0 (±5.8) Student′s t test. Wilcoxon‐Mann‐Whitney test. Two‐way ANOVA (with interaction).

Partial cervical dilation 4.7 (3.2‐6.0) 4.0 (3.2‐5.5) Full cervical dilation 4.0 (2.9‐7.3) 4.0 (3.1‐5.3) Scar‐external cervical os distance No cervical dilation 35.5 (29.3‐37.0) 33.0 (29.0‐39.0) 0.000c Partial cervical dilation 33.6 (±6.1) 30.1 (±5.5) Full cervical dilation 28.9 (±9.0) 27.0 (±5.8) Student′s t test. Wilcoxon‐Mann‐Whitney test. Two‐way ANOVA (with interaction). John Wiley & Sons, LtdLongitudinal observational data on uterine position and presence and type of defects in this cohort are presented in Tables 8, 9, 10, 11. The incidence of scar defects was not statistically significant between controls. The difference between groups with different suturing methods was statistically significant (0.036). A higher proportion of the defects were seen in the SLT group.The position of the uterus varied greatly between controls (0.000), especially between 6‐week and 6‐month controls. The difference between the 6‐ and 12‐month controls was already statistically insignificant. The difference between groups with different suturing methods was not statistically significant. The combination of uterine position and scar defect presence changed significantly between controls (0.001) and it varied significantly depending on the suturing method (0.003). A higher incidence of uterine position in retroflexion and defects was seen in the 6‐week control in the SLT group. Defects with or without contact with the uterine cavity changed statistically between controls (0.017). Both types of defects occurred more frequently in the 6‐week follow up than in stage II and III controls. The difference based on suturing method is at the limit of statistical significance (0.065). Both types of defects are more common in the SLT group. For any of the variables uterine position, scar defect, combination of both and scar placement, the difference between groups according to the suturing method did not change statistically significantly over time (interaction of time and suture type is not significant).

types of defects are more common in the SLT group. For any of the variables uterine position, scar defect, combination of both and scar placement, the difference between groups according to the suturing method did not change statistically significantly over time (interaction of time and suture type is not significant). Table 8 Longitudinal observation data on Presence of scar defect in women with all three follow up controls (total number of women who attended the 12‐month follow up, with all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.247, test significance of variable Closure technique is 0.036 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.370 Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Presence of scar defect Yes No Yes No 6 weeks 123 (82.6) 26 (17.4) 140 (80.0) 35 (20.0) 6 months 118 (79.2) 31 (20.8) 129 (73.7) 46 (26.3) 12 months 124 (83.2) 25 (16.8) 127 (72.6) 48 (27.4) John Wiley & Sons, LtdTable 9 Longitudinal observation data on Uterine position in women with all three follow‐up controls (total number of women who attended 12‐month follow up, with all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.000, test significance of variable Closure technique is 0.132 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.270

all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.000, test significance of variable Closure technique is 0.132 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.270 Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Uterine position Anteflexion Retroflexion Anteflexion Retroflexion 6 weeks 78 (52.3) 71 (47.7) 97 (55.4) 78 (46.6) 6 months 97 (65.1) 52 (34.9) 128 (73.1) 47 (26.9) 12 months 90 (60.4) 59 (39.6) 125 (71.4) 50 (28.6) John Wiley & Sons, LtdTable 10 Longitudinal observation data on Scar defect in anteflexion/retroflexion (AVF/RVF) uterus in women with all three follow‐up controls (total number of women who attended 12‐months follow up, with all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.001, test significance of variable Closure technique is 0.003 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.821

all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.001, test significance of variable Closure technique is 0.003 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.821 Follow up Closure technique Single‐layer (n = 149) Double‐layer (n = 175) Scar defect in AVF uterus Yes No Yes No 6 weeks 70 (89.7) 8 (10.3) 80 (82.5) 17 (17.5) 6 months 81 (83.5) 16 (16.5) 94 (73.4) 34 (26.6) 12 months 81 (90.0) 9 (10.0) 90 (72.0) 35 (28.2) Follow up Scar defect in RVF uterus Yes No Yes No 6 weeks 53 (74.6) 18 (25.4) 60 (76.9) 18 (23.1) 6 months 37 (71.2) 15 (28.8) 35 (74.4) 12 (25.5) 12 months 43 (72.9) 16 (27.1) 37 (74.0) 13 (26.0) John Wiley & Sons, LtdTable 11 Longitudinal observation data on Scar defect presence and defect contact with the uterine cavity in women with all three follow‐up controls (total number of women who attended 12‐month follow up, with all three controls = 324). Test of dependence of outcome on time (generalized linear mixed model): test significance of variable Time is 0.017, test significance of variable Closure technique is 0.065 and test significance of interaction Time with Closure technique (change of closure technique outcome in time) is 0.493

The most important decision is whether to use an SLT or DLT to improve the scar quality and decrease the risk of uterine rupture and dehiscence in the subsequent pregnancy. Data from Swedish registers demonstrate no significant difference in the rate of uterine rupture.17 An earlier meta‐analysis including retrospective and prospective studies reported that locked SLT, compared with DLT, is associated with a fourfold increase in the risk of uterine rupture.18 We do not have any clinical data from subsequent pregnancies in our patients and are thus unable to confirm this finding. We have shown that DLT is associated with smaller defects and with thicker RMT. This is in contrast to evidence based on randomized trials which does not support a specific type of uterine closure. SLT and locked first layer are possibly associated with thinner residual RMT.19 Prospective studies using transvaginal ultrasound of the scar, favor an unlocked suture with exclusion of the decidua to optimize the placement of the muscle layers and their regeneration.19, 20 Roberge et al demonstrate in a randomized controlled trial that DLT with a first unlocked layer excluding the decidua, compared with locked SLT including the decidua, is associated with a greater RMT and healing ratio.21 But the lack of statistical power did not allow the authors to draw a definitive conclusion. It is possible that excluding the decidua from the first suture induces a better adaptation of myometrium. We have included decidua in the first suture and thus are unable to confirm this suggestion.

er RMT and healing ratio.21 But the lack of statistical power did not allow the authors to draw a definitive conclusion. It is possible that excluding the decidua from the first suture induces a better adaptation of myometrium. We have included decidua in the first suture and thus are unable to confirm this suggestion. The reason for the more distal scars and defects in women with CS at full dilation is clear. At stage II of labor, the cervix creates a continuous birth canal with the uterine cavity and is pulled up. Therefore, the incision is finally located caudal to the external cervical ostium. In contrast to other studies, we were not able to prove that scars with larger defects reside more caudally than intact scars or scars with smaller defects.22 But those findings were influenced by cases with more than one CS. We have shown a change in uterine position between 6 weeks and 1 year from retroflexion to anteflexion, with no relation to closure technique. We do not have data on its position before the pregnancy and thus were unable to demonstrate that the CS and the closure technique affected it. The reason for change in uterine position is unknown: it could be due to tissue healing and scar remodeling, but it may also interfere with healing and tissue extension. Vikhareva Osser et al observed more scar defects in women with a uterus in retroflexion.9 A higher incidence of uterine position in retroflexion and defects was only seen in our study in the 6‐week control in the SLT group.

tissue healing and scar remodeling, but it may also interfere with healing and tissue extension. Vikhareva Osser et al observed more scar defects in women with a uterus in retroflexion.9 A higher incidence of uterine position in retroflexion and defects was only seen in our study in the 6‐week control in the SLT group. Our study is not limited only to CS performed in women before or in early labor. Women in advanced labor were also included. The women in both groups were selected from the same caucasian community and the comparability between the two groups was high. The scar was longitudinally evaluated by two independent observers blinded to the treatment allocation in a population having a primary CS. Another strength of our study was the uniform use of a specific suture method for both closures. On the other hand, cervical dilation, duration of labor or oxytocin augmentation are factors that increase the risk of a large scar defect in non‐pregnant women.23

e treatment allocation in a population having a primary CS. Another strength of our study was the uniform use of a specific suture method for both closures. On the other hand, cervical dilation, duration of labor or oxytocin augmentation are factors that increase the risk of a large scar defect in non‐pregnant women.23 The main limitation of the study is the drop‐out rate. We did not examine further why women discontinued the follow up. Another limitation is that RMT represents an indirect evaluation of scar healing and demonstrates a surrogate outcome for the prediction of negative consequences. On the other hand, there are enough data showing that the presence of scar defects and RMT value are correlated with such adverse events.10, 23 Also, we did not perform an ultrasonographic examination according to menstrual cycle, as recommended by another group.7 Synchronization was not possible because many women were breastfeeding and had secondary amenorrhea. This study was primarily an urogynecological research targeted at the influence of the first pregnancy and delivery on female pelvic floor; therefore this subanalysis, which is focused on scar assessment after CS, was not registered as a randomized control trial. 5 CONCLUSION Our data demonstrate the benefit of DLT. Defects in SLT group were more common, wider and had thinner RMT. Most changes in the scar area occurred during the first 6 months. Although recent discussion has focused mainly on the number of suture layers, the current knowledge highlights the importance of decidua suture exclusion.

a demonstrate the benefit of DLT. Defects in SLT group were more common, wider and had thinner RMT. Most changes in the scar area occurred during the first 6 months. Although recent discussion has focused mainly on the number of suture layers, the current knowledge highlights the importance of decidua suture exclusion. CONFLICT OF INTERESTS The authors have stated explicitly that there are no conflicts of interest in connection with this article. This study was supported by PROGRES Q 34, Charles University project, Prague, Czech Republic.

Abbreviations CIconfidence interval DNdelivered neonates ICHintracranial hemorrhage INDintrapartum/neonatal death IRDSinfant respiratory distress syndrome LGAlarge for gestational age NECnecrotizing enterocolitis PERINEDNetherlands’ perinatal registry PNDperinatal death PPHNpersistent pulmonary hypertension of the newborn RRrelative risk SGAsmall for gestational age Key message Small‐for‐gestational‐age males have increased risks of antepartum death; all males born after 32+0 weeks have increased risks of neonatal morbidity compared with females. We found no increased risk of any mortality in normal weight or large‐for‐gestational‐age males after 28+0 weeks.

SGAsmall for gestational age Key message Small‐for‐gestational‐age males have increased risks of antepartum death; all males born after 32+0 weeks have increased risks of neonatal morbidity compared with females. We found no increased risk of any mortality in normal weight or large‐for‐gestational‐age males after 28+0 weeks. 1 INTRODUCTION In pregnancy, fetal sex is known to affect placentation,1 intrauterine growth,2 preterm birth3, 4, 5, 6, 7, 8, 9 and perinatal outcome.10, 11, 12, 13 Previous studies have suggested a male predominance in miscarriage,10 antepartum death,11, 12, 13 perinatal death, fetal distress, respiratory distress syndrome and low Apgar scores.14 Although these studies provide valuable information about gender differences, outcomes should be interpreted with care because of methodological weaknesses. First, in some studies, outcomes were adjusted for absolute birthweight and gestational age at delivery while not taking into account that healthy males are on average heavier than healthy females at any given gestational age.2 This may lead to comparison of small‐for‐gestation (SGA) males with normal weight females and falsely suggest a higher risk of adverse outcome among males. In other studies, outcomes were based on calculated male to female ratio for adverse outcome, expressed as the number of adverse outcomes among males divided by the number of adverse outcomes among females. Since, historically, more males are born, this comparison may also falsely suggest a higher risk of adverse outcome among males. Secondly, previous studies did not investigate whether there was an association between gestational age at delivery, fetal growth and perinatal outcome.

ber of adverse outcomes among females. Since, historically, more males are born, this comparison may also falsely suggest a higher risk of adverse outcome among males. Secondly, previous studies did not investigate whether there was an association between gestational age at delivery, fetal growth and perinatal outcome. These methodological limitations can be overcome when birthweight percentiles are used to express growth instead of absolute birthweight, stratification for SGA, normal weight and large‐for‐gestational‐age (LGA) males, females and gestational age at delivery is performed, and the fetus at risk approach is used to rule out bias through unequal numbers of male and female infants. The objective of this hypothesis‐generating study was to evaluate the association between fetal sex and adverse pregnancy outcome, including antepartum death, intrapartum/neonatal death and neonatal morbidity, whilst correcting for fetal growth and gestational age at delivery.

These methodological limitations can be overcome when birthweight percentiles are used to express growth instead of absolute birthweight, stratification for SGA, normal weight and large‐for‐gestational‐age (LGA) males, females and gestational age at delivery is performed, and the fetus at risk approach is used to rule out bias through unequal numbers of male and female infants. The objective of this hypothesis‐generating study was to evaluate the association between fetal sex and adverse pregnancy outcome, including antepartum death, intrapartum/neonatal death and neonatal morbidity, whilst correcting for fetal growth and gestational age at delivery. 2 MATERIAL AND METHODS This study was performed in a nationwide cohort with the use of the Netherlands Perinatal Registry (PERINED). The PERINED consists of population‐based data that contain information on pregnancies, deliveries and re‐admissions until 28 days after birth. The PERINED database is obtained by a validated linkage of three different registries: the midwifery registry, the obstetrics registry and the neonatology registry of hospital admissions.15, 16 Records are entered in the PERINED registry at the child's level. The coverage of the PERINED registry is approximately 96% of all deliveries in the Netherlands. It contains pregnancies of ≥22 weeks, gestational age and is used primarily for annual assessment of the quality indicators of obstetric care.

missions.15, 16 Records are entered in the PERINED registry at the child's level. The coverage of the PERINED registry is approximately 96% of all deliveries in the Netherlands. It contains pregnancies of ≥22 weeks, gestational age and is used primarily for annual assessment of the quality indicators of obstetric care. We included all white European women who delivered a singleton baby between 25+0 and 42+6 weeks of gestation in The Netherlands between 1 January 1999, and 31 December 2010. We excluded all women who delivered an infant with congenital anomalies.17 Women of other ethnicities were excluded to avoid bias through differences in optimal weight for gestation in other groups not taken into account in the Dutch birthweight reference curves and to avoid bias through the presence of different risk profiles for adverse pregnancy outcomes among non‐white European women. Gestational age at delivery was defined as extremely preterm (25+0‐27+6 weeks), very preterm (28+0‐31+6 weeks), moderate to late preterm (32+0‐36+6 weeks) and term (37+0‐42+6 weeks). The Dutch reference curves for birthweight by gestational age stratified for parity and sex were used.18 SGA was defined as a birthweight below the 10th percentile for gestation (<p10), appropriately grown for gestation (normal weight) as a birthweight between the 10th and 90th percentile (p10‐90) for gestation, and LGA as birthweight above the 90th percentile (>p90) for gestation.

ratified for parity and sex were used.18 SGA was defined as a birthweight below the 10th percentile for gestation (<p10), appropriately grown for gestation (normal weight) as a birthweight between the 10th and 90th percentile (p10‐90) for gestation, and LGA as birthweight above the 90th percentile (>p90) for gestation. Our outcome measures were antepartum death, intrapartum/neonatal death, perinatal death, a composite of neonatal morbidity, and a composite of adverse perinatal outcome. Antepartum death was defined as stillbirth that occurred between 25+0 weeks gestational age and the onset of labor. Intrapartum and neonatal death were defined as death between the onset of labor and 28 days after birth. Perinatal death was defined as antepartum death, intrapartum death and neonatal death with definitions as described. The composite measure of neonatal morbidity consisted of: infant respiratory distress syndrome (IRDS), neonatal sepsis, necrotizing enterocolitis (NEC), meconium aspiration, persistent pulmonary hypertension of the newborn (PPHN), periventricular leukomalacia, Apgar score <7 at 5 minutes and intracranial hemorrhage (ICH), all as judged by the clinician within the first admission after birth. If a neonate suffered from neonatal morbidity and died within 28 days after birth, it was considered to have suffered intrapartum/neonatal death and not morbidity. The composite measure of adverse perinatal outcome consisted of the composite measure of neonatal morbidity and perinatal death.

irst admission after birth. If a neonate suffered from neonatal morbidity and died within 28 days after birth, it was considered to have suffered intrapartum/neonatal death and not morbidity. The composite measure of adverse perinatal outcome consisted of the composite measure of neonatal morbidity and perinatal death. PERINED registered demographic and obstetric characteristics including maternal age, parity, hypertensive complications of pregnancy (chronic hypertension, pregnancy‐induced hypertension, preeclampsia)19 and socioeconomic status. Parity was categorized into nulliparous and multiparous women. The socioeconomic status score is based on mean income level, the percentage of households with a low income, the percentage of inhabitants without a paid job and the percentage of households with on average a low education in a postal code area.20 The continuous socioeconomic status score was categorized into a low, middle and high group, based on percentile ranges (<25th percentile, 25‐75th percentile, >75th percentile).

me, the percentage of inhabitants without a paid job and the percentage of households with on average a low education in a postal code area.20 The continuous socioeconomic status score was categorized into a low, middle and high group, based on percentile ranges (<25th percentile, 25‐75th percentile, >75th percentile). Demographic and obstetric baseline characteristics were compared between males and females using Student's t test and Chi‐square test as appropriate. We tested for interaction between fetal sex and gestational age at delivery and between fetal sex and birthweight percentile. These tests were performed separately for all outcome measures. If interaction was found to be present (defined as a P < 0.001), analyses were performed, stratified for gestational age at delivery (25+0‐27+6 weeks, 28+0‐31+6 weeks, 32+0‐36+6 weeks, 37+0‐42+6 weeks) and birthweight percentile group (SGA, normal weight and LGA). The fetus at risk approach was used to determine the association between fetal sex and pregnancy outcome, expressed as relative risk (RR) with 95% confidence intervals (CI) within the strata for birthweight percentile and gestational age at delivery. 2.1 Antepartum death For a fetus to be at risk of antepartum death (AD) at 36+0 weeks it is necessary to be alive at 36+0 weeks. Consequently, the risk of antepartum death was calculated as a proportion of the ongoing pregnancies (OP) at a particular gestation. Risk of antepartum death at week n = ADn/OPn*100.

The fetus at risk approach was used to determine the association between fetal sex and pregnancy outcome, expressed as relative risk (RR) with 95% confidence intervals (CI) within the strata for birthweight percentile and gestational age at delivery. 2.1 Antepartum death For a fetus to be at risk of antepartum death (AD) at 36+0 weeks it is necessary to be alive at 36+0 weeks. Consequently, the risk of antepartum death was calculated as a proportion of the ongoing pregnancies (OP) at a particular gestation. Risk of antepartum death at week n = ADn/OPn*100. 2.2 Intrapartum/neonatal death In concurrence, the risk of intrapartum/neonatal death (IND) at any gestational age is obtained by dividing the number of intrapartum and neonatal deaths at that gestation by the number of neonates at risk of intrapartum/neonatal death at that gestation. The neonates at risk of intrapartum/neonatal death at a certain gestational age include all delivered neonates (DN) that did not die antepartum, meaning all pregnancies with onset of labor. Risk of intrapartum/neonatal death at week n = INDn/DNn*100. 2.3 Perinatal death The risk of perinatal death (PND) at any gestational age is obtained by dividing the number of neonates with perinatal death (antepartum or intrapartum/neonatal death) in each stratum by the total number of deliveries in each birthweight category. Risk of perinatal death at week n = PNDn/DNn*100. 2.3.1 Composite neonatal morbidity The denominator used for the risk of neonatal morbidity (NM) is all alive fetuses (AF) born at that gestation, since only alive fetuses can suffer morbidity.

2.3 Perinatal death The risk of perinatal death (PND) at any gestational age is obtained by dividing the number of neonates with perinatal death (antepartum or intrapartum/neonatal death) in each stratum by the total number of deliveries in each birthweight category. Risk of perinatal death at week n = PNDn/DNn*100. 2.3.1 Composite neonatal morbidity The denominator used for the risk of neonatal morbidity (NM) is all alive fetuses (AF) born at that gestation, since only alive fetuses can suffer morbidity. Risk of neonatal morbidity at week n = NMn/AFn*100. 2.3.2 Composite adverse perinatal outcome The numerator is all neonates with perinatal death (PND) or morbidity (NM) in each stratum and the denominator all deliveries in each birthweight category (DN). Risk of adverse perinatal outcome at week n = (PNDn + NMn)/DNn*100. A post‐hoc analysis for all separate components of morbidity was performed to differentiate further which outcome contributes most to the found difference between male and female neonates. Data were analyzed with the SAS statistical software package version 9.2 (SAS Institute Inc., Cary, NC, USA). 2.4 Ethical approval The data in the perinatal registry are anonymous and exempt from ethical approval. The Netherlands Perinatal Registry gave approval for the use of the data for this study (approval number 13.73).

Data were analyzed with the SAS statistical software package version 9.2 (SAS Institute Inc., Cary, NC, USA). 2.4 Ethical approval The data in the perinatal registry are anonymous and exempt from ethical approval. The Netherlands Perinatal Registry gave approval for the use of the data for this study (approval number 13.73). 3 RESULTS From 1 January 1999 until 31 December 2010 a total of 2 078 327 singleton pregnancies between 25 and 42 weeks without congenital anomalies were identified in the PERINED database. After exclusion of non‐white European women (n = 335 426 [16%]) and infants with an unknown gender (n = 70 [0.004%]), our study population consisted of 1 742 831 pregnancies.

ber 2010 a total of 2 078 327 singleton pregnancies between 25 and 42 weeks without congenital anomalies were identified in the PERINED database. After exclusion of non‐white European women (n = 335 426 [16%]) and infants with an unknown gender (n = 70 [0.004%]), our study population consisted of 1 742 831 pregnancies. Baseline characteristics of the cohort are presented in Table 1. There were more male (n = 895 272 [51.4%]) than female infants (n = 847 559 [48.6%]) in the cohort. There were no statistically significant differences in maternal baseline characteristics between the two groups. However, women with a male fetus were more likely to have a hypertensive complication of pregnancy compared with women with a female fetus (9.3 vs 9.0%, P < 0.001). Women with a male fetus were more likely (all P < 0.001) than women with a female fetus to undergo an emergency cesarean section (8.5 vs 7.1%) or vaginal instrumental delivery (12.2% vs 10.0%). The rate of preterm delivery (<37 weeks of gestation) was higher among males than among females (6.4 vs 5.4%). Average birthweight in males was 124 g (95% CI 122‐126) higher than in females (3525 vs 3401 g; P < 0.001). Finally, male and female fetuses were equally likely to be SGA (8.9%) or LGA (11%). Table 1 Characteristics of the 1 742 831 singleton white European pregnancies without congenital anomalies in the Netherlands, 1999‐2010

Baseline characteristics of the cohort are presented in Table 1. There were more male (n = 895 272 [51.4%]) than female infants (n = 847 559 [48.6%]) in the cohort. There were no statistically significant differences in maternal baseline characteristics between the two groups. However, women with a male fetus were more likely to have a hypertensive complication of pregnancy compared with women with a female fetus (9.3 vs 9.0%, P < 0.001). Women with a male fetus were more likely (all P < 0.001) than women with a female fetus to undergo an emergency cesarean section (8.5 vs 7.1%) or vaginal instrumental delivery (12.2% vs 10.0%). The rate of preterm delivery (<37 weeks of gestation) was higher among males than among females (6.4 vs 5.4%). Average birthweight in males was 124 g (95% CI 122‐126) higher than in females (3525 vs 3401 g; P < 0.001). Finally, male and female fetuses were equally likely to be SGA (8.9%) or LGA (11%). Table 1 Characteristics of the 1 742 831 singleton white European pregnancies without congenital anomalies in the Netherlands, 1999‐2010 Male infants Female infants P value (n = 895 272) (n = 847 559) Maternal characteristics Maternal age, years (mean), (SD) 30.7 (4.6) 30.7 (4.7) 0.24 Nulliparous, n (%) 426 995 (47.7) 403 106 (47.6) 0.08 Low socioeconomic status, n (%) 172 737 (19.3) 163 006 (19.2) 0.30 Pregnancy and delivery characteristics Conception IVF/ICSI, n (%) 24 856 (2.8) 23 375 (2.8) 0.46 Hypertensive disordersa, n (%) 83 170 (9.3) 76 100 (9.0) <0.0001 Induction of labor, n (%) 191 330 (21.4) 182 001 (21.5) 0.10 Mode of delivery Spontaneous vaginal delivery, n (%) 654 129 (73.1) 647 524 (76.4) <0.0001 Instrumental vaginal delivery, n (%) 109 448 (12.2) 84 552 (10.0) Elective cesarean, n (%) 55 269 (6.2) 55 556 (6.6) Emergency cesarean, n (%) 76 426 (8.5) 59 957 (7.1) Neonatal characteristics Gestational age at delivery (wk), median (IQR) 39 (38‐40) 40 (39‐40) <0.0001 Delivery <32 wk GA, (%) 8110 (0.9) 6438 (0.8) <0.0001 Delivery <37 wk GA, (%) 57 353 (6.4) 45 980 (5.4) <0.0001 Birthweight (g), mean (SD) 3525 (597) 3401 (564) <0.0001 Birthweight <p10 (SGA) 79 442 (8.9) 75 691 (8.9) Birthweight p10‐p90 (normal weight) 719 424 (80.4) 677 826 (80.0) Birthweight ≥p90 (LGA) 96 406 (10.8) 94 042 (11.1) Abbreviation: SD, standard deviation.

1 Delivery <37 wk GA, (%) 57 353 (6.4) 45 980 (5.4) <0.0001 Birthweight (g), mean (SD) 3525 (597) 3401 (564) <0.0001 Birthweight <p10 (SGA) 79 442 (8.9) 75 691 (8.9) Birthweight p10‐p90 (normal weight) 719 424 (80.4) 677 826 (80.0) Birthweight ≥p90 (LGA) 96 406 (10.8) 94 042 (11.1) Abbreviation: SD, standard deviation. a Preexisting hypertension, pregnancy‐induced hypertension, preeclampsia. John Wiley & Sons, LtdInteraction between fetal sex and gestational age at delivery was statistically significant for antepartum death (P < 0.001), intrapartum/neonatal death (P < 0.001) and neonatal morbidity (P < 0.001). Interaction between fetal sex and birthweight percentile was also significant for all outcome measures (all P < 0.001). Therefore, outcomes are presented stratified for four strata of gestational age at delivery (25+0‐27+6 weeks, 28+0‐31+6 weeks, 32+0‐36+6 weeks, 37+0‐42+6 weeks) and three strata of birthweight percentile (SGA, normal weight and LGA). 3.1 Antepartum death Table 2 shows the relative risk for antepartum death separately for four strata of gestational age at delivery (25+0‐27+6 weeks, 28+0‐31+6 weeks, 32+0‐36+6 weeks, 36+6‐42+6 weeks) and three strata of birthweight percentile (SGA, normal weight and LGA). Table 2 Antepartum death rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 397/79 442 0.50 371/75 691 0.49 1.02 (0.89‐1.17) 28‐31 wk GA 303/78 863 0.38 206/75 223 0.27 1.40 (1.17‐1.67)

3.1 Antepartum death Table 2 shows the relative risk for antepartum death separately for four strata of gestational age at delivery (25+0‐27+6 weeks, 28+0‐31+6 weeks, 32+0‐36+6 weeks, 36+6‐42+6 weeks) and three strata of birthweight percentile (SGA, normal weight and LGA). Table 2 Antepartum death rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 397/79 442 0.50 371/75 691 0.49 1.02 (0.89‐1.17) 28‐31 wk GA 303/78 863 0.38 206/75 223 0.27 1.40 (1.17‐1.67) 32‐36 wk GA 370/78 017 0.47 264/74 710 0.35 1.34 (1.15‐1.57) 37‐42 wk GA 503/73 106 0.69 420/70 774 0.59 1.16 (1.02‐1.32) Birthweight p10‐p90 (normal weight) 25‐27 wk GA 277/719 424 0.04 304/677 826 0.04 0.86 (0.73‐1.01) 28‐31 wk GA 404/718 042 0.06 422/676 696 0.06 0.90 (0.79‐1.03) 32‐36 wk GA 597/713 388 0.08 563/673 002 0.08 1.00 (.89‐1.12) 37‐42 wk GA 937/671 723 0.14 978/640 051 0.15 0.91 (0.83‐1.00) Birthweight ≥p90 (LGA) 25‐27 wk GA 31/96 406 0.03 29/94 042 0.03 1.04 (0.63‐1.73) 28‐31 wk GA 46/96 243 0.05 44/93 850 0.05 1.02 (0.67‐1.54) 32‐36 wk GA 36/95 757 0.04 24/93 409 0.03 1.46 (0.87‐2.45) 37‐42 wk GA 131/93 090 0.14 122/90 754 0.13 1.05 (0.82‐1.34) All birthweights 25‐27 wk GA 705/895 272 0.08 704/847 559 0.08 0.95 (0.85‐1.05) 28‐31 wk GA 753/893 148 0.08 672/845 769 0.08 1.06 (0.96‐1.18) 32‐36 wk GA 1003/887 162 0.11 851/841 121 0.10 1.12 (1.02‐1.22)

Birthweight p10‐p90 (normal weight) 25‐27 wk GA 277/719 424 0.04 304/677 826 0.04 0.86 (0.73‐1.01) 28‐31 wk GA 404/718 042 0.06 422/676 696 0.06 0.90 (0.79‐1.03) 32‐36 wk GA 597/713 388 0.08 563/673 002 0.08 1.00 (.89‐1.12) 37‐42 wk GA 937/671 723 0.14 978/640 051 0.15 0.91 (0.83‐1.00) Birthweight ≥p90 (LGA) 25‐27 wk GA 31/96 406 0.03 29/94 042 0.03 1.04 (0.63‐1.73) 28‐31 wk GA 46/96 243 0.05 44/93 850 0.05 1.02 (0.67‐1.54) 32‐36 wk GA 36/95 757 0.04 24/93 409 0.03 1.46 (0.87‐2.45) 37‐42 wk GA 131/93 090 0.14 122/90 754 0.13 1.05 (0.82‐1.34) All birthweights 25‐27 wk GA 705/895 272 0.08 704/847 559 0.08 0.95 (0.85‐1.05) 28‐31 wk GA 753/893 148 0.08 672/845 769 0.08 1.06 (0.96‐1.18) 32‐36 wk GA 1003/887 162 0.11 851/841 121 0.10 1.12 (1.02‐1.22) 37‐42 wk GA 1571/837 919 0.19 1520/801 579 0.19 0.99 (0.92‐1.06) Abbreviations: CI, confidence interval; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated with fetus at risk approach, meaning the numerator is all neonates with antepartum death in each stratum and the denominator all women remaining pregnant at the beginning of the gestational age stratum in each birthweight category. John Wiley & Sons, LtdAntepartum death occurs more often in SGA males than SGA females after 28+0 weeks (from 28+0 to 31+6 weeks, RR 1.40, 95% CI 1.17‐1.67; from 32+0 to 36+6 weeks, RR 1.34, 95% CI 1.15‐1.57; from 37+0 to 42+6 weeks, RR 1.16, 95% CI 1.02‐1.32).

a RR calculated with fetus at risk approach, meaning the numerator is all neonates with antepartum death in each stratum and the denominator all women remaining pregnant at the beginning of the gestational age stratum in each birthweight category. John Wiley & Sons, LtdAntepartum death occurs more often in SGA males than SGA females after 28+0 weeks (from 28+0 to 31+6 weeks, RR 1.40, 95% CI 1.17‐1.67; from 32+0 to 36+6 weeks, RR 1.34, 95% CI 1.15‐1.57; from 37+0 to 42+6 weeks, RR 1.16, 95% CI 1.02‐1.32). For normal weight and LGA infants, there was no statistically significant difference between males and females. The analysis for all infants (not stratified into SGA, normal weight and LGA) shows that antepartum death occurs more often in males than in females born between 32+0 and 36+6 weeks (RR 1.12, 95% CI 1.02‐1.22). 3.2 Intrapartum and neonatal death Table 3 shows the rates of intrapartum/neonatal death. In most groups, there were no significant differences between males and females. However, normal weight males born between 25+0 and 27+6 weeks GA had an increased risk of intrapartum/neonatal death compared with females (RR 1.20, 95% CI 1.02‐1.40). Table 3 Intrapartum and neonatal death rate in males and females by birthweight percentiles and gestational age category

3.2 Intrapartum and neonatal death Table 3 shows the rates of intrapartum/neonatal death. In most groups, there were no significant differences between males and females. However, normal weight males born between 25+0 and 27+6 weeks GA had an increased risk of intrapartum/neonatal death compared with females (RR 1.20, 95% CI 1.02‐1.40). Table 3 Intrapartum and neonatal death rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 61/182 34 41/97 42 .79 (0.58‐1.08) 28‐31 wk GA 68/543 13 30/307 9.8 1.28 (0.85‐1.92) 32‐36 wk GA 110/4541 2.4 103/3672 2.8 0.86 (0.66‐1.13) 37‐42 wk GA 263/72 603 0.36 232/70 354 0.33 1.10 (0.92‐1.31) Birthweight p10‐p90 (normal weight) 25‐27 wk GA 298/1105 27 186/826 23 1.20 (1.02‐1.40) 28‐31 wk GA 203/4250 4.8 173/3272 5.3 0.90 (0.74‐1.10) 32‐36 wk GA 303/41 068 0.73 202/323 88 0.62 1.18 (0.99‐1.41) 37‐42 wk GA 549/670 786 0.08 470/639 073 0.07 1.11 (0.98‐1.26) Birthweight ≥p90 (LGA) 25‐27 wk GA 21/132 16 36/163 22 0.72 (0.44‐1.17) 28‐31 wk GA 33/440 7.5 31/397 7.8 0.96 (0.60‐1.54) 32‐36 wk GA 24/2631 0.91 27/2631 1.0 0.89 (0.51‐1.54) 37‐42 wk GA 64/92 959 0.07 53/90 632 0.06 1.18 (0.82‐1.69) All birthweights 25‐27 wk GA 380/1419 27 263/1086 24 1.08 (0.94‐1.25) 28‐31 wk GA 304/5233 5.8 234/3976 5.9 0.99 (0.84‐1.16) 32‐36 wk GA 437/48 240 0.91 332/38 691 0.86 1.06 (0.92‐1.22) 37‐42 wk GA 876/836 348 0.10 755/800 059 0.09 1.11 (1.01‐1.22)

28‐31 wk GA 203/4250 4.8 173/3272 5.3 0.90 (0.74‐1.10) 32‐36 wk GA 303/41 068 0.73 202/323 88 0.62 1.18 (0.99‐1.41) 37‐42 wk GA 549/670 786 0.08 470/639 073 0.07 1.11 (0.98‐1.26) Birthweight ≥p90 (LGA) 25‐27 wk GA 21/132 16 36/163 22 0.72 (0.44‐1.17) 28‐31 wk GA 33/440 7.5 31/397 7.8 0.96 (0.60‐1.54) 32‐36 wk GA 24/2631 0.91 27/2631 1.0 0.89 (0.51‐1.54) 37‐42 wk GA 64/92 959 0.07 53/90 632 0.06 1.18 (0.82‐1.69) All birthweights 25‐27 wk GA 380/1419 27 263/1086 24 1.08 (0.94‐1.25) 28‐31 wk GA 304/5233 5.8 234/3976 5.9 0.99 (0.84‐1.16) 32‐36 wk GA 437/48 240 0.91 332/38 691 0.86 1.06 (0.92‐1.22) 37‐42 wk GA 876/836 348 0.10 755/800 059 0.09 1.11 (1.01‐1.22) Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated as: the numerator is all neonates with intrapartum or neonatal death in each stratum and the denominator all deliveries without antepartum death in each birthweight category. John Wiley & Sons, LtdThe analysis for all infants (not stratified into SGA, normal weight and LGA) shows that intrapartum/neonatal death occurs more often in males than in females born between 37+0 and 42+6 weeks (RR 1.11, 95% CI 1.01‐1.22). 3.3 Perinatal death Table 4 shows the rates of perinatal death. In most groups, there were no significant differences between males and females. However, SGA males born between 37+0 and 42+6 weeks GA had an increased risk of perinatal death compared with females (RR 1.06, 95% CI 1.01‐1.12).

John Wiley & Sons, LtdThe analysis for all infants (not stratified into SGA, normal weight and LGA) shows that intrapartum/neonatal death occurs more often in males than in females born between 37+0 and 42+6 weeks (RR 1.11, 95% CI 1.01‐1.22). 3.3 Perinatal death Table 4 shows the rates of perinatal death. In most groups, there were no significant differences between males and females. However, SGA males born between 37+0 and 42+6 weeks GA had an increased risk of perinatal death compared with females (RR 1.06, 95% CI 1.01‐1.12). Table 4 Perinatal death rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 458/579 79 412/468 88 0.77 (0.37‐0.73) 28‐31 wk GA 371/846 44 236/513 46 0.96 (0.89‐1.05) 32‐36 wk GA 480/4911 9.8 367/3936 9.3 1.02 (0.96‐1.09) 37‐42 wk GA 766/73 106 1.0 652/70 774 0.92 1.06 (1.01‐1.12) Birthweight p10‐p90 (normal weight) 25‐27 wk GA 575/1382 42 490/1130 43 0.97 (.90‐1.04) 28‐31 wk GA 607/4654 13 595/3694 16 0.89 (0.84‐0.95) 32‐36 wk GA 900/41 665 2.2 765/32 951 2.3 0.97 (0.92‐1.01) 37‐42 wk GA 1486/671 723 0.22 1448/640 051 0.23 0.99 (0.95‐1.03) Birthweight ≥p90 (LGA) 25‐27 wk GA 52/163 32 65/192 34 0.95 (0.73‐1.22) 28‐31 wk GA 79/486 16 75/441 17 0.97 (0.81‐1.15) 32‐36 wk GA 60/2667 2.2 51/2655 1.9 1.08 (0.89‐1.27) 37‐42 wk GA 195/93 090 0.21 175/90 754 0.19 1.04 (0.94‐1.14) All birthweights 25‐27 wk GA 1085/2124 51 967/1790 54 0.95 (0.89‐1.01) 28‐31 wk GA 1057/5986 18 906/4648 19 0.95 (0.90‐0.99)

Birthweight p10‐p90 (normal weight) 25‐27 wk GA 575/1382 42 490/1130 43 0.97 (.90‐1.04) 28‐31 wk GA 607/4654 13 595/3694 16 0.89 (0.84‐0.95) 32‐36 wk GA 900/41 665 2.2 765/32 951 2.3 0.97 (0.92‐1.01) 37‐42 wk GA 1486/671 723 0.22 1448/640 051 0.23 0.99 (0.95‐1.03) Birthweight ≥p90 (LGA) 25‐27 wk GA 52/163 32 65/192 34 0.95 (0.73‐1.22) 28‐31 wk GA 79/486 16 75/441 17 0.97 (0.81‐1.15) 32‐36 wk GA 60/2667 2.2 51/2655 1.9 1.08 (0.89‐1.27) 37‐42 wk GA 195/93 090 0.21 175/90 754 0.19 1.04 (0.94‐1.14) All birthweights 25‐27 wk GA 1085/2124 51 967/1790 54 0.95 (0.89‐1.01) 28‐31 wk GA 1057/5986 18 906/4648 19 0.95 (0.90‐0.99) 32‐36 wk GA 1440/49 243 2.9 1183/39 542 3.0 0.99 (0.95‐1.03) 37‐42 wk GA 2447/837 919 0.29 2275/801 579 0.28 1.01 (0.99‐1.04) Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated as: the numerator are all neonates with perinatal death (antepartum or intrapartum/neonatal death) in each stratum and the denominator all deliveries in each birthweight category. John Wiley & Sons, LtdIn contrast, the analysis for all infants (not stratified into SGA, normal weight and LGA) shows that perinatal death occurs less often in males than in females born between 28+0 and 31+6 weeks (RR 0.95, 95% CI 0.90‐0.99). 3.4 Composite morbidity The composite morbidity of males and females is shown in Table 5. Table 5 Composite neonatal morbidity rate in males and females by birthweight percentiles and gestational age category

John Wiley & Sons, LtdIn contrast, the analysis for all infants (not stratified into SGA, normal weight and LGA) shows that perinatal death occurs less often in males than in females born between 28+0 and 31+6 weeks (RR 0.95, 95% CI 0.90‐0.99). 3.4 Composite morbidity The composite morbidity of males and females is shown in Table 5. Table 5 Composite neonatal morbidity rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 94/121 78 48/56 86 0.91 (0.78‐1.05) 28‐31 wk GA 324/475 68 190/277 69 0.99 (0.90‐1.10) 32‐36 wk GA 535/4431 12 366/3569 10 1.18 (1.04‐1.33) 37‐42 wk GA 1723/72 340 2.4 1320/70 122 1.9 1.27 (1.18‐1.36) Birthweight p10‐p90 (normal weight) 25‐27 wk GA 674/807 84 527/807 82 1.01 (0.97‐1.06) 28‐31 wk GA 2602/4047 64 1861/3099 60 1.07 (1.03‐1.11) 32‐36 wk GA 3754/40 765 9.2 2550/32 186 7.9 1.16 (1.11‐1.22) 37‐42 wk GA 9292/670 237 1.4 6533/638 603 1.0 1.36 (1.31‐1.40) Birthweight ≥p90 (LGA) 25‐27 wk GA 87/111 78 105/127 83 0.95 (0.84‐1.08) 28‐31 wk GA 248/407 60 197/366 54 1.13 (1.00‐1.28) 32‐36 wk GA 274/2607 11 213/2604 8.2 1.28 (1.08‐1.52) 37‐42 wk GA 1698/92 895 1.8 1233/90 579 1.4 1.34 (1.25‐1.44) All birthweights 25‐27 wk GA 855/1039 82 680/823 83 0.99 (0.95‐1.04) 28‐31 wk GA 3174/4929 64 2248/3742 60 1.07 (1.04‐1.11) 32‐36 wk GA 4563/47 803 9.5 3129/38 359 8.2 1.17 (1.12‐1.22) 37‐42 wk GA 12 713/835 472 1.5 9086/799 304 1.1 1.34 (1.30‐1.37) Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age.

All birthweights 25‐27 wk GA 855/1039 82 680/823 83 0.99 (0.95‐1.04) 28‐31 wk GA 3174/4929 64 2248/3742 60 1.07 (1.04‐1.11) 32‐36 wk GA 4563/47 803 9.5 3129/38 359 8.2 1.17 (1.12‐1.22) 37‐42 wk GA 12 713/835 472 1.5 9086/799 304 1.1 1.34 (1.30‐1.37) Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated as: the numerator is all alive neonates with morbidity in each stratum and the denominator all alive neonates in each birthweight category. John Wiley & Sons, LtdThe risk of composite morbidity was significantly increased in males compared with females in most (8/12) strata, with relative risks ranging from 1.07 to 1.36. All males born after 32+0 weeks of gestation have an increased risk of neonatal morbidity compared with females. The analysis for all infants (not stratified into SGA, normal weight and LGA) shows that composite morbidity was significantly increased in males compared with females born after 28+0 weeks, with relative risks ranging from 1.07 to 1.34. 3.5 Composite adverse perinatal outcome The composite adverse neonatal outcome of males and females is shown in Table 6. Table 6 Composite adverse perinatal outcome rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 552/579 95 460/468 98 0.97 (0.95‐0.99) 28‐31 wk GA 695/846 82 426/513 83 0.99 (0.94‐1.04) 32‐36 wk GA 1015/4911 21 733/3936 19 1.11 (1.02‐1.21) 37‐42 wk GA 2489/73 106 3.4 1972/70 774 2.8 1.22

Table 6 Composite adverse perinatal outcome rate in males and females by birthweight percentiles and gestational age category Male infants Female infants RRa (95% CI) (n = 895 272) % (n = 847 559) % Birthweight <p10 (SGA) 25‐27 wk GA 552/579 95 460/468 98 0.97 (0.95‐0.99) 28‐31 wk GA 695/846 82 426/513 83 0.99 (0.94‐1.04) 32‐36 wk GA 1015/4911 21 733/3936 19 1.11 (1.02‐1.21) 37‐42 wk GA 2489/73 106 3.4 1972/70 774 2.8 1.22 (1.15‐1.30) Birthweight p10‐p90 (normal weight) 25‐27 wk GA 1249/1382 90 1017/1130 90 1.00 (0.98‐1.03) 28‐31 wk GA 3209/4654 69 2456/3694 66 1.04 (1.01‐1.07) 32‐36 wk GA 4654/41 665 11 3315/32 951 10 1.11 (1.06‐1.16) 37‐42 wk GA 10 778/671 723 1.6 7981/640 051 1.2 1.29 (1.25‐1.32) Birthweight ≥p90 (LGA) 25‐27 wk GA 139/163 85 170/192 89 0.96 (0.89‐1.05) 28‐31 wk GA 327/486 67 272/441 62 1.09 (0.99‐1.21) 32‐36 wk GA 334/2667 13 264/2655 10 1.26 (1.08‐1.47) 37‐42 wk GA 1893/93 090 2.0 1408/90 754 1.6 1.31 (1.22‐1.40) All birthweights 25‐27 wk GA 1940/2124 91 1647/1790 92 0.99 (0.97‐1.01) 28‐31 wk GA 4231/5986 71 3154/4648 68 1.04 (1.02‐1.07) 32‐36 wk GA 6003/49 243 12 4312/39 542 11 1.12 (1.08‐1.16) 37‐42 wk GA 15 160/837 919 1.8 11 361/801 579 1.4 1.28 (1.25‐1.31) Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated as: the numerator are all neonates with perinatal death (antepartum or intrapartum/neonatal death) or morbidity in each stratum and the denominator all deliveries in each birthweight category.

Abbreviations: CI, confidence interval; GA, gestational age; LGA, large‐for‐gestational age; RR, relative risk ratio; SGA, small‐for‐gestational age. a RR calculated as: the numerator are all neonates with perinatal death (antepartum or intrapartum/neonatal death) or morbidity in each stratum and the denominator all deliveries in each birthweight category. John Wiley & Sons, LtdThe risk of composite adverse neonatal outcome was significantly increased in males compared with females in most (7/12) strata, with relative risks ranging from 1.04 to 1.31. All males born after 32+0 weeks of gestation have an increased risk of adverse neonatal outcome compared with females. The analysis for all infants (not stratified into SGA, normal weight and LGA) shows that composite morbidity was significantly increased in males compared with females born after 28+0 weeks, with relative risks ranging from 1.04 to 1.28.

John Wiley & Sons, LtdThe risk of composite adverse neonatal outcome was significantly increased in males compared with females in most (7/12) strata, with relative risks ranging from 1.04 to 1.31. All males born after 32+0 weeks of gestation have an increased risk of adverse neonatal outcome compared with females. The analysis for all infants (not stratified into SGA, normal weight and LGA) shows that composite morbidity was significantly increased in males compared with females born after 28+0 weeks, with relative risks ranging from 1.04 to 1.28. The individual components of the composite neonatal morbidity (IRDS, sepsis, NEC, meconium aspiration, PPHN, periventricular leukomalacia, Apgar score <7 at 5 minutes and ICH) were also analyzed separately (Tables S1‐S8). IRDS was increased in males compared with females at most gestational ages, regardless of birthweight percentile (Table S1). Sepsis was increased in all term males compared with females regardless of birthweight. Between 32+0 and 36+6 weeks the risk of sepsis was only increased in normal weight males (Table S2). PPHN was only increased in normal weight males at term compared with females (Table S5). For periventricular leukomalacia, although in only one stratum, a decreased risk was found for SGA males between 25+0 and 27+0 weeks of gestation (Table S6). The risk of ICH was increased in all normal weight males compared with females regardless of gestational age at delivery and in LGA males born between 28+0 and 31+6 weeks of gestation (RR 1.52, 95% CI 1.08‐2.13) (Table S6). The incidence of an Apgar score <7 at 5 minutes was increased in all term males compared with females regardless of birthweight. Furthermore, SGA males between 25+0 and 27+0 and 32+0 and 36+6 weeks of gestation were at increased risk (Table S7). The incidence of NEC (Table S3) and meconium aspiration (Table S4) was not significantly different between males and females.

increased in all term males compared with females regardless of birthweight. Furthermore, SGA males between 25+0 and 27+0 and 32+0 and 36+6 weeks of gestation were at increased risk (Table S7). The incidence of NEC (Table S3) and meconium aspiration (Table S4) was not significantly different between males and females. 4 DISCUSSION In our study, we analyzed 1 742 831 singleton deliveries and assessed differences in pregnancy outcomes between males and females after adjustments for differences in fetal growth. SGA males have an increased risk of antepartum death and all males born after 32+0 weeks of gestation have an increased risk of neonatal morbidity compared with females. Differences in neonatal morbidity are mainly caused by increased risks of sepsis, IRDS and ICH in males compared with females. Our study has some limitations. These limitations are mainly related to the fact that we performed a database study and consequently had to rely on information that was recorded in the registry. To assess differences in the incidence of adverse outcome, large numbers are needed. In our opinion, it would not have been feasible to collect enough data using a different study design.

elated to the fact that we performed a database study and consequently had to rely on information that was recorded in the registry. To assess differences in the incidence of adverse outcome, large numbers are needed. In our opinion, it would not have been feasible to collect enough data using a different study design. Secondly, the Dutch Perinatal registry did not contain reliable data on the use of betamethasone, maternal smoking, diabetes or maternal body mass index because these data were not mandatory fields in the perinatal registration during the study period, and as a result not reliably registered. All these factors are associated with adverse pregnancy outcome, but it is unlikely that there is a fetal sex‐based bias. Betamethasone administration is based on gestational age and is not sex‐dependent, and no publications about a relation between maternal smoking, diabetes or BMI and fetal sex could be retrieved. Adverse neonatal outcome could not be more clearly defined than that judged by the clinician within the first admission after birth, therefore we do not expect a fetal sex‐based bias. In addition, there are probably unknown factors causally related to perinatal outcome, such as maternal smoking and hypertensive disorders. By adjusting for birthweight percentile it is possible that our results are partly distorted by a biasing path through the unmeasured factors.21

not expect a fetal sex‐based bias. In addition, there are probably unknown factors causally related to perinatal outcome, such as maternal smoking and hypertensive disorders. By adjusting for birthweight percentile it is possible that our results are partly distorted by a biasing path through the unmeasured factors.21 Women with a male fetus were more likely to have a hypertensive complication of pregnancy compared with women with a female fetus (9.3 vs 9.0%, P < 0.001). This is in accordance with findings in previous research.22 The association might be term‐dependent, but this is outside the scope of this study.

not expect a fetal sex‐based bias. In addition, there are probably unknown factors causally related to perinatal outcome, such as maternal smoking and hypertensive disorders. By adjusting for birthweight percentile it is possible that our results are partly distorted by a biasing path through the unmeasured factors.21 Women with a male fetus were more likely to have a hypertensive complication of pregnancy compared with women with a female fetus (9.3 vs 9.0%, P < 0.001). This is in accordance with findings in previous research.22 The association might be term‐dependent, but this is outside the scope of this study. The PERINED database does not contain quantitative data on the method of pregnancy dating. According to the ruling guideline during the study period, pregnancy dating was predominantly performed by first trimester ultrasound measurements (crown‐rump length). If no first trimester dating was performed, dating was based on head‐circumference measurement or last menstrual period. In the rare cases when dating was done in the second trimester, a sex‐dependent difference in estimation, since boys are already somewhat larger than girls on average at that time23 may disturb these analyses. Because the majority of ultrasound dating was done in the first trimester we do not expect a substantial influence on our results. Also, in case of antepartum death—especially in the preterm period—the gestational age of delivery is not the same as the moment of fetal demise. Although this might cause a structural underestimation of birthweight percentile, it is unlikely that this time‐effect is different between males and females. Also, the difference between fetal demise and delivery and thus possible overestimation of SGA in case of antepartum death is likely small at term, because—according to the Dutch protocol—all pregnant women in the term period undergo weekly checkups including Doppler auscultation of the fetal heart rate.

s and females. Also, the difference between fetal demise and delivery and thus possible overestimation of SGA in case of antepartum death is likely small at term, because—according to the Dutch protocol—all pregnant women in the term period undergo weekly checkups including Doppler auscultation of the fetal heart rate. We used population‐based birthweight percentiles.18 Individual growth potential and placental characteristics might have enabled more accurate prediction of growth restriction and adverse outcome.24, 25 We were not able to correct for this because maternal length and weight, and placental weight and pathology are not registered in the Dutch Perinatal Registry. We do not expect a systematical bias. Furthermore, ultrasound growth charts used to detect SGA fetuses are not gender‐specific. Therefore it is possible that girls (who are by definition smaller than male fetuses) are more often SGA compared with boys, perhaps leading to earlier intervention (induction) and thereby preventing antepartum death. This possible bias cannot be ruled out and should be investigated in further studies.

gender‐specific. Therefore it is possible that girls (who are by definition smaller than male fetuses) are more often SGA compared with boys, perhaps leading to earlier intervention (induction) and thereby preventing antepartum death. This possible bias cannot be ruled out and should be investigated in further studies. Our data did not allow us to make statements about the influence of mode of delivery on neonatal death and morbidity. Table 1 shows differences in induction of labor and instrumental deliveries between males and females. These are often a result of suspected compromised fetal condition rather than a confounder of adverse outcome. We can only hypothesize that the higher rate of vaginal instrumental delivery among males might be associated with the increased risk of ICH among males; another explanation could be that the ICH occurred antenatally and caused fetal distress. This is not within the scope of this paper and should be investigated in future studies. Finally, there might be an association between the higher incidence of cesarean sections among males and the risk of IRDS, although information on timing of the cesarean section and administration of betamethasone is lacking.

ress. This is not within the scope of this paper and should be investigated in future studies. Finally, there might be an association between the higher incidence of cesarean sections among males and the risk of IRDS, although information on timing of the cesarean section and administration of betamethasone is lacking. Finally, some causes of neonatal morbidity are predominantly a problem in the extreme preterm period, whereas others occur at all gestational ages or mainly at term. We decided to use a composite outcome containing important causes of neonatal morbidity available in the PERINED registry to provide useful information for clinicians. Stratification into four categories of gestational ages at delivery ensures that an automatic distinction is made between causes of morbidity that predominantly occur among extremely preterm infants and causes of morbidity that occur at all gestational ages. Separate tables for all causes of morbidity are provided as Tables S1‐S8. The main strength of this study is the size (1 742 831 pregnancies) and composition of the cohort. Data are derived from a large, well‐maintained population‐based national perinatal registry (1999‐2010). Cohorts of all previous studies were smaller (549 up to 469 152 pregnancies)11, 25 and often included anomalous fetuses and/or twins. The proportion of male infants, the incidence of antepartum deaths, intrapartum/neonatal deaths and composite neonatal morbidity that we found in this study are in accordance with previous research.7, 14, 26, 27, 28, 29, 30

smaller (549 up to 469 152 pregnancies)11, 25 and often included anomalous fetuses and/or twins. The proportion of male infants, the incidence of antepartum deaths, intrapartum/neonatal deaths and composite neonatal morbidity that we found in this study are in accordance with previous research.7, 14, 26, 27, 28, 29, 30 Secondly, as discussed in the introduction, previous studies falsely suggest a higher risk of adverse outcome among males because the absolute number of adverse outcomes among males was divided by the number of adverse outcomes among females. Using the fetus at risk approach avoids bias through unequal numbers of male and female fetuses at a certain gestational age. Also—in contrast to previous studies—outcomes were stratified for birthweight percentile and gestational age at delivery. This enables comparison of male and female infants within the same birthweight percentile category and gestational age group, and rules out bias through the association between male sex and premature delivery. As a result, the outcomes most closely represent the influence of fetal sex on the outcomes of interest and allowed us to check for associations between fetal growth, gender and adverse outcome. To our knowledge this is the first study that tested for interaction and consequently performed analyses stratified for gestational age at delivery and birthweight percentile.

Secondly, as discussed in the introduction, previous studies falsely suggest a higher risk of adverse outcome among males because the absolute number of adverse outcomes among males was divided by the number of adverse outcomes among females. Using the fetus at risk approach avoids bias through unequal numbers of male and female fetuses at a certain gestational age. Also—in contrast to previous studies—outcomes were stratified for birthweight percentile and gestational age at delivery. This enables comparison of male and female infants within the same birthweight percentile category and gestational age group, and rules out bias through the association between male sex and premature delivery. As a result, the outcomes most closely represent the influence of fetal sex on the outcomes of interest and allowed us to check for associations between fetal growth, gender and adverse outcome. To our knowledge this is the first study that tested for interaction and consequently performed analyses stratified for gestational age at delivery and birthweight percentile. This was a hypothesis‐generating study. Unknown or unmeasured factors may have caused these associations. After appropriate correction for birthweight percentile and gestational age, the results of this study put previous studies in perspective that showed increased antepartum death, intrapartum/neonatal death and neonatal morbidity in males compared with females.10, 11, 12, 13, 14 These higher adverse outcome rates might not be caused by actual increased chances of adverse outcome among males but by the increased risk of preterm delivery in males,3, 4, 5, 6, 7, 8, 9 by adjustment for absolute birthweight instead of birthweight percentile, and calculation of male/female ratios of adverse outcome that was not adjusted for differences in the total number of male and female infants born. In contrast to previous studies, we found no increased risk of antepartum, intrapartum/neonatal and perinatal death in normal weight and LGA males born after 28+0 weeks. We found an increased risk of antepartum death (after 28+0 weeks) and neonatal morbidity (after 32+0 weeks) among SGA males, and only increased risk in perinatal death among SGA males born between 37+0 and 42+6 weeks.

partum, intrapartum/neonatal and perinatal death in normal weight and LGA males born after 28+0 weeks. We found an increased risk of antepartum death (after 28+0 weeks) and neonatal morbidity (after 32+0 weeks) among SGA males, and only increased risk in perinatal death among SGA males born between 37+0 and 42+6 weeks. This study shows increased composite morbidity and composite adverse perinatal outcome among all males born after 32+0 weeks of gestation. The higher neonatal morbidity rate in males seems to be mainly caused by higher rates of IRDS, sepsis and ICH in males. The effect of higher neonatal morbidity on long‐term health outcomes in males in our cohort is unsure. Previous research showed higher rates of bronchopulmonary dysplasia and increased mortality in very preterm SGA infants and very low birthweight infants.31, 32 We hypothesize that genetic differences between males and females might underlie the fetal sex‐related differences. This study gives no clues on how to decrease neonatal morbidity in male infants. The results of this study matter because they correct the erroneous idea that all male infants suffer more antepartum, intrapartum/neonatal and perinatal death than female infants. The main implication of this study is awareness of differences between male and female perinatal outcome. Our study has further focused on which domain these differences lie.

cause they correct the erroneous idea that all male infants suffer more antepartum, intrapartum/neonatal and perinatal death than female infants. The main implication of this study is awareness of differences between male and female perinatal outcome. Our study has further focused on which domain these differences lie. Further intervention studies should be aware of sex differences in perinatal outcome and should perform prespecified analyses separately for male and female infants. The importance of this advice has already been shown in a recent intervention study that found differences in effectiveness of allopurinol between male and female fetuses.33 This could lead to a more tailored approach fit for the individual neonate to improve outcomes. In addition, future research could be aimed at unraveling mechanisms that might play a role in the increased neonatal morbidity in males. 5 CONCLUSION SGA males have an increased risk of antepartum death and all males born after 32+0 weeks of gestation have an increased risk of neonatal morbidity compared with females. Differences in neonatal morbidity are mainly caused by increased risks of IRDS, sepsis, PPHN, Apgar score <7 at 5 minutes and ICH in males compared with females. In contrast to findings in previous studies, we found no increased risk of antepartum, intrapartum/neonatal and perinatal death in normal weight and LGA males born after 28+0 weeks. CONFLICT OF INTEREST B.W.M. reports consultancy for ObsEva, Merck and Guerbet. B.W.M. is supported by a NHMRC Practitioner Fellowship (GNT1082548). Supporting information

5 CONCLUSION SGA males have an increased risk of antepartum death and all males born after 32+0 weeks of gestation have an increased risk of neonatal morbidity compared with females. Differences in neonatal morbidity are mainly caused by increased risks of IRDS, sepsis, PPHN, Apgar score <7 at 5 minutes and ICH in males compared with females. In contrast to findings in previous studies, we found no increased risk of antepartum, intrapartum/neonatal and perinatal death in normal weight and LGA males born after 28+0 weeks. CONFLICT OF INTEREST B.W.M. reports consultancy for ObsEva, Merck and Guerbet. B.W.M. is supported by a NHMRC Practitioner Fellowship (GNT1082548). Supporting information Click here for additional data file. ACKNOWLEDGMENTS We thank all Dutch midwives, obstetricians, neonatologists and other perinatal healthcare providers for the registration of perinatal information and the Foundation of the Netherlands Perinatal Registry (http://www.perinatreg.nl) for permission (13.73) to use the registry data.

Cluster trials are used increasingly to test interventions that need to be implemented at a group level, or that require testing on large populations, but they are notoriously susceptible to selection bias. The problem also affects individually randomized trials, especially open ones, where participants may exclude themselves after they discover their treatment allocation, or if treatment takes time to organize, such that one group has less opportunity to receive it. In this type of trial, the problem is well recognized, and the solution is clear, namely to enter participants in the trial before randomization and to include all participants in their allocated group, whatever treatment they received, so called “analysis by intention‐to‐treat”. For cluster trials, “analysis by intention‐to‐treat” requires first that all participants be recruited to the cluster before randomization or, if this is practically impossible, that recruitment post‐randomization is sufficiently automatic to rule out selective exclusion. Then second, the analysis must be conducted on the whole sample. Unfortunately, even major funders and high impact journals continue to miss this source of bias.

ter before randomization or, if this is practically impossible, that recruitment post‐randomization is sufficiently automatic to rule out selective exclusion. Then second, the analysis must be conducted on the whole sample. Unfortunately, even major funders and high impact journals continue to miss this source of bias. The stepped wedge PARROT trial reported in the Lancet in March1 is a recent example. Stepped wedge trials are a subtype of cluster trials, in which intervention centers act as their own controls, the control period precedes the intervention, and the intervention typically continues after the trial ends.2 The PARROT researchers were testing the effect of introducing a new test for preeclampsia, placental growth factor (PlGF).3 They randomized whole maternity units, each caring for many thousands of women over the study period, but analyzed only the minority who had suspected preeclampsia. On the face of it, maternal severe adverse outcomes were reduced (24/447 women in the concealed testing group vs 22/573 women in the revealed group, odds ratio [OR] 0.32, 95% CI 0.11‐0.96; P = 0.043), with no significant effect on perinatal adverse outcomes (86 revealed vs 63 concealed, OR 1.45, 95% CI 0.73‐2.90). The authors recommended adoption of PlGF testing.

tcomes were reduced (24/447 women in the concealed testing group vs 22/573 women in the revealed group, odds ratio [OR] 0.32, 95% CI 0.11‐0.96; P = 0.043), with no significant effect on perinatal adverse outcomes (86 revealed vs 63 concealed, OR 1.45, 95% CI 0.73‐2.90). The authors recommended adoption of PlGF testing. However, the number of women identified with suspected preeclampsia had increased by 28%, from 447 to 573, during the period when the new PlGF test results were revealed and acted upon. As the recruitment periods were identical and birth rates had not altered, this is unlikely to have occurred by chance, and was probably due to inclusion of additional women with milder disease. No‐one can now identify the corresponding women during the control periods, but it is likely that they also had mild disease and good outcomes. If so, we can crudely model what the results would have been, had there been an additional 126 women and babies with good outcomes in the control period. The rate of severe adverse maternal outcomes would have become about 4% in both groups and the trend towards more adverse baby outcomes would have increased (15% vs 11%), favoring controls. Without access to the full data set we cannot estimate the statistical significance, but the conclusion of the trial might well have been the opposite!

verse maternal outcomes would have become about 4% in both groups and the trend towards more adverse baby outcomes would have increased (15% vs 11%), favoring controls. Without access to the full data set we cannot estimate the statistical significance, but the conclusion of the trial might well have been the opposite! The authors of the recent AFFIRM trial,4 also in the Lancet, avoided the problem. Hospitals were allocated at random to a package of care that encouraged mothers to be aware of, and report, alterations in fetal movements, and encouraged staff to respond appropriately. The authors reported all perinatal deaths occurring in each hospital over the relevant time periods, not just those in women who had experienced altered movements. The result was clear. The package was ineffective with an adjusted OR 0.98 (95% CI 0.83‐1.17) for perinatal mortality. Similar problems have affected, and been avoided, in cluster trials of school‐based sex education interventions. “Baby think it over”, a school‐based pregnancy prevention program in which teenage girls cared for a simulated infant, was evaluated in a cluster trial published in the Lancet in 2016.5 A higher proportion of the intervention group went on to have at least one birth as teenagers, 97/1267 (8%) vs 67/1567 (4%) control (RR 1.36, 95% CI 1.10‐1.67, P = 0.003) or at least one termination of pregnancy as the first pregnancy event (9% vs 6%). The headline results were that use of the infant simulator was harmful.

gher proportion of the intervention group went on to have at least one birth as teenagers, 97/1267 (8%) vs 67/1567 (4%) control (RR 1.36, 95% CI 1.10‐1.67, P = 0.003) or at least one termination of pregnancy as the first pregnancy event (9% vs 6%). The headline results were that use of the infant simulator was harmful. Unfortunately, only about half the girls in the intervention schools could be recruited because of the availability of school health nurses and infant simulators. This gave an opportunity for selection bias. Bolzern et al6 tested baseline factors for nominal statistical significance, and showed that some differences could not have occurred by chance; the intervention group was more socioeconomically disadvantaged (P = 0.000000000019) and had lower educational attainment (P = 0.0000000015). Teachers were probably recruiting girls who they thought were at higher risk, to the intervention groups. Analyzing pregnancies and abortions among all the girls in the intervention and control clusters, which would have avoided the problem, was not done.

000019) and had lower educational attainment (P = 0.0000000015). Teachers were probably recruiting girls who they thought were at higher risk, to the intervention groups. Analyzing pregnancies and abortions among all the girls in the intervention and control clusters, which would have avoided the problem, was not done. In contrast, the investigators of SHARE, a cluster trial of school‐based peer‐led sex education published in the BMJ,7 did exactly that. Whole schools were allocated to intervention or control and every female member of the relevant class was followed up, whether or not they actually participated. There were no significant differences between the groups in registered conceptions per 1000 pupils (300 SHARE vs 274 control; difference 26, 95% CI −33 to 86), or in terminations per 1000 pupils (127 vs 112; difference 15, 95% CI −13 to 42) between ages 16 and 20 years. The results were disappointing for supporters of the intervention, but secure. The present authors are involved in two ongoing cluster trials in obstetrics. JGT and KW with the GBS‐3 trial in the UK,8 and HF with the CDC4G trial in Sweden.9 GBS‐3 is a cluster trial testing the effect of adding routine Group B streptococcus screening at either 36 weeks or intrapartum, to the UK's present risk‐based screening policy. Eighty maternity units will be randomized. The primary end point, all‐cause early neonatal sepsis, will be measured from routine data for every birth within the clusters, whether or not the mother underwent screening.

tococcus screening at either 36 weeks or intrapartum, to the UK's present risk‐based screening policy. Eighty maternity units will be randomized. The primary end point, all‐cause early neonatal sepsis, will be measured from routine data for every birth within the clusters, whether or not the mother underwent screening. CDC4G is a cluster trial testing the effect of lowering the threshold for diagnosing gestational diabetes. The primary outcome is the rate of fetal macrosomia. If this were measured among pregnancies where diabetes has been diagnosed, the result would be biased because the new threshold will almost certainly diagnose more women with diabetes. Instead it will be measured among all women delivering within the participating hospitals whatever their diabetes results. So long as both trials analyze everyone in the randomized units, they should avoid the problems of PARROT and “Baby Think it Over”, and produce reliable results like AFFIRM and SHARE.

Abbreviations EFWestimated fetal weight FGRfetal growth restriction NDIneurodevelopmental impairment RCTrandomized controlled trial Key message The data of 25 included studies, reporting 2895 pregnancies complicated by fetal growth restriction, diagnosed before 32 weeks of gestation indicate that overall survival was 81%. In 12% of the surviving children cognitive impairment and/or cerebral palsy was diagnosed. 1 INTRODUCTION Severe early‐onset fetal growth restriction (FGR) with placental insufficiency as its mechanism1 is an obstetric condition that is mostly managed in tertiary‐care hospitals. By consensus, FGR is defined as onset before 32 weeks of gestation, a fetal abdominal circumference or estimated fetal weight (EFW) below the 3rd centile or absent end‐diastolic flow in the umbilical artery, or abdominal circumference or EFW below the 10th centile combined with a pulsatility index of the uterine artery above the 95th centile and/or pulsatility index of the umbilical artery above the 95th centile.2 This patient group needs high amounts of care and has a high likelihood of iatrogenic premature delivery, both for fetal and for secondary maternal indications, such as the development of the maternal syndrome of preeclampsia.3 As these FGR children are usually born very preterm, the condition carries significant risks of neonatal mortality, major and minor morbidity, and long‐term health sequelae.4, 5 These risks are not only strongly related to gestational age, but also to the extent of growth restriction. Reported survival rates vary.3

a.3 As these FGR children are usually born very preterm, the condition carries significant risks of neonatal mortality, major and minor morbidity, and long‐term health sequelae.4, 5 These risks are not only strongly related to gestational age, but also to the extent of growth restriction. Reported survival rates vary.3 Counseling patients with severe early‐onset FGR about perinatal prognosis is difficult because of the uncertain influence of different prognostic variables of the condition. Furthermore, the widespread variability of existing data on survival and long‐term prognosis of the fetus makes decision‐making in this patient group even more difficult. Overview of total mortality is often lacking in literature on this subject. For example, many studies describe the prognosis of live‐born neonates after FGR and do not take antenatal death into account. From an obstetric perspective, long‐term outcomes can only be interpreted optimally if they are presented together with the proportions of antenatal and neonatal death.6 The aim of this systematic review is to describe the chances of overall (antenatal and neonatal) survival, and long‐term morbidity and neurodevelopment based on the total number of fetuses at first FGR diagnosis to inform patients and obstetricians in their counseling and decision‐making.

antenatal and neonatal death.6 The aim of this systematic review is to describe the chances of overall (antenatal and neonatal) survival, and long‐term morbidity and neurodevelopment based on the total number of fetuses at first FGR diagnosis to inform patients and obstetricians in their counseling and decision‐making. 2 MATERIAL AND METHODS 2.1 Data sources An information specialist (JL) performed a broad search in OVID MEDLINE from 2000 to 27 April 2019. The search consisted of controlled terms, including MeSH terms, and text words for FGR and antenatal/perinatal mortality or neurodevelopment in infants with demonstrated FGR, combined with search filters to retrieve primary and secondary studies (the latter only as a check). We searched from 2000 onwards because neonatal health care has changed fundamentally in the current millennium. No further restrictions were applied. The complete search strategy is shown in the Supplementary material (Table S1). The retrieved records were imported and de‐duplicated in endnote X7. The included studies were screened for additional relevant cited or citing references. 2.2 Main outcomes measures Six important research questions were identified: What is, in severe early‐onset FGR, the chance of intrauterine death? What is, in live‐born neonates after severe early‐onset FGR, the chance of neonatal death? What is, in surviving children after severe early‐onset FGR, the chance of neurodevelopmental impairment (NDI) at or before 5 years of age in long‐term follow up?

2.2 Main outcomes measures Six important research questions were identified: What is, in severe early‐onset FGR, the chance of intrauterine death? What is, in live‐born neonates after severe early‐onset FGR, the chance of neonatal death? What is, in surviving children after severe early‐onset FGR, the chance of neurodevelopmental impairment (NDI) at or before 5 years of age in long‐term follow up? What is, in surviving children after severe early‐onset FGR, the mean cognitive score at or before 5 years of age? What is, in surviving children after severe early‐onset FGR, the mean motor score at or before 5 years of age? What is, in surviving children after severe early‐onset FGR, the chance of cerebral palsy at or before 5 years of age? 2.3 Eligibility criteria Records covering singleton pregnancies diagnosed with FGR, as defined by trialists, diagnosed before 32 weeks of gestation, were included when the antenatal and perinatal data on mortality were reported. If a study included patients diagnosed with FGR before and after 32 weeks of gestation (for example between 24 and 38 weeks of gestation) the study was only included if data on the subgroup below 32 weeks of gestation was reported separately in the publication. Because of the progress of quality of obstetric and neonatal care, only patient groups (partially) born in or after the year 2000 were included. Furthermore, only records published in English and with an available full text were included.

subgroup below 32 weeks of gestation was reported separately in the publication. Because of the progress of quality of obstetric and neonatal care, only patient groups (partially) born in or after the year 2000 were included. Furthermore, only records published in English and with an available full text were included. Records were excluded if they only described neonates born after FGR, evaluating the postnatal data, without describing the antenatal and perinatal mortality. 2.4 Data collection Titles and abstracts of all search results were independently screened by 2 researchers (AP and IMB). Discrepancies were resolved by discussion with a third researcher (WG). The full text of potentially eligible studies was assessed. Relevant data were extracted from the full text by 2 researchers independently (AP and IMB) and compared for purpose of completeness and correctness. The quality of the evidence was rated using the GRADE instrument.7 3 RESULTS The literature search identified 2602 unique records, and 2 additional records were identified through reference and citation checks. After title and abstract screening, 269 full‐text records were assessed for eligibility; 25 studies comprising 2895 patients were included in the systematic review (Figure 1). Figure 1 Flowchart article selection. FGR, fetal growth restriction; GA, gestational age [Color figure can be viewed at http://wileyonlinelibrary.com]

3 RESULTS The literature search identified 2602 unique records, and 2 additional records were identified through reference and citation checks. After title and abstract screening, 269 full‐text records were assessed for eligibility; 25 studies comprising 2895 patients were included in the systematic review (Figure 1). Figure 1 Flowchart article selection. FGR, fetal growth restriction; GA, gestational age [Color figure can be viewed at http://wileyonlinelibrary.com] 3.1 General characteristics of the studies Table 1 summarizes the characteristics of the included studies. The number of included pregnancies varied from 8 to 503. FGR was defined differently among the included studies; some studies focused on the EFW or abdominal circumference only, whereas other studies included Doppler measurements as well (Table 1). Table 1 Characteristics of included studies Study design Number of patients Definition of FGR Gestational age at diagnosis FGR (wk + d), (mean ± SD or median [IQR]) EFW at diagnosis FGR (g) (mean ± SD or median [IQR]) Proportion of patients with preeclampsia or HELLP at diagnosis FGR Ali et al9 Clinically retrospectively registered, open, parallel, randomized controlled trial 80 AC <10th centile with increased HC:AC ratio Group 1 (n = 34) mean 30 ± 0.5 Group 2 (n = 34) mean 30 ± 0.3 Group 1 (n = 34) mean 1202 ± 72 Group 2 (n = 34) mean 1209 ± 48 0% Ali et al8 Clinically registered, open, parallel, randomized clinical trial 60 AC or birthweight <10th centile Group 1 (n = 30) mean 30 ± 0.5 Group 2 (n = 30) mean 30 ± 0.4 Group 1 (n = 30) mean 1193 ± 51 Group 2 (n = 30) mean 1216 ± 63

Study design Number of patients Definition of FGR Gestational age at diagnosis FGR (wk + d), (mean ± SD or median [IQR]) EFW at diagnosis FGR (g) (mean ± SD or median [IQR]) Proportion of patients with preeclampsia or HELLP at diagnosis FGR Ali et al9 Clinically retrospectively registered, open, parallel, randomized controlled trial 80 AC <10th centile with increased HC:AC ratio Group 1 (n = 34) mean 30 ± 0.5 Group 2 (n = 34) mean 30 ± 0.3 Group 1 (n = 34) mean 1202 ± 72 Group 2 (n = 34) mean 1209 ± 48 0% Ali et al8 Clinically registered, open, parallel, randomized clinical trial 60 AC or birthweight <10th centile Group 1 (n = 30) mean 30 ± 0.5 Group 2 (n = 30) mean 30 ± 0.4 Group 1 (n = 30) mean 1193 ± 51 Group 2 (n = 30) mean 1216 ± 63 0% Aoki et al14 Retrospective cohort study 17 <5th centile (not defined what needs to be <5th centile) Median 25.4 (22.6‐27.7) Median 513 (260‐741) 17/17 = 100% Baschat et al22 Prospective cohort study 44 AC <5th centile and umbilical artery Doppler PI more than 2 SD above the gestational mean by local reference values Median 25+1 (range 16+4 to 31+6) Not described Not described Belghiti et al23 Retrospective cohort study 10 FGR patients with reported outcomes <5th centile (not defined what needs to be <5th centile) 25+0 to 25+6 Not described for subgroup FGR 10/10 = 100% Fox et al 817 Retrospective case‐control study 252 EFW <25th centile 21.0 ± 1.0 Not described Not described Fujisaki et al24 Prospective, 1‐arm, interventional pilot study 14 EFW ≤5th centile Median 25+3 (22+6 to 25+5) Mean 418 ± 160 0% Groom et al11 Triple‐blind, placebo‐controlled, parallel, phase II‐III trial randomized at the participant level 122 At 22+0 to 27+6 wk of gestation: AC ≤3rd centile

t described Not described Fujisaki et al24 Prospective, 1‐arm, interventional pilot study 14 EFW ≤5th centile Median 25+3 (22+6 to 25+5) Mean 418 ± 160 0% Groom et al11 Triple‐blind, placebo‐controlled, parallel, phase II‐III trial randomized at the participant level 122 At 22+0 to 27+6 wk of gestation: AC ≤3rd centile At 28+0 to 29+6 wk of gestation: EFW <700 g Group 1 (n = 63): Mean 24.5 ± 1.7 Group 2 (n = 59): Mean 24.8 ± 1.7 Group 1 (n = 63): Mean 479.3 ± 148.1 Group 2 (n = 59): Mean 495.7 ± 170.2 16/122 = 13.1% Hasegawa et al25 Retrospective cohort study 26 <5th centile (not defined what needs to be <5th centile) Group 1 (n = 17) median 25.3 (21.4‐29.9) Group 2 (n = 9) median 25.3 (20.4‐28.1) Not described Not described Herraiz et al26 Observational prospective cohort study 74 EFW <3rd centile or EFW <10th centile + abnormal fetal Doppler Group 1 (n = 37): 27.0 ± 2.8 Group 2 (n = 36): 27.9 ± 2.0

16/122 = 13.1% Hasegawa et al25 Retrospective cohort study 26 <5th centile (not defined what needs to be <5th centile) Group 1 (n = 17) median 25.3 (21.4‐29.9) Group 2 (n = 9) median 25.3 (20.4‐28.1) Not described Not described Herraiz et al26 Observational prospective cohort study 74 EFW <3rd centile or EFW <10th centile + abnormal fetal Doppler Group 1 (n = 37): 27.0 ± 2.8 Group 2 (n = 36): 27.9 ± 2.0 Not described 36/74 = 48.6% Kubo et al27 Open label, phase 1 clinical trial 8 (<32 wk) EFW ≤ to −1.5 SD on ultrasonography from the Japanese standard table Median 28+4 (26+0 to 30+5) Median 967 (708‐1164) Not described Lawin‐O'Brien et al28 Multicenter retrospective study of databases 245 AC ≤3rd centile for gestational age, AC calculated according to UK recommended standard and Altman and Chitty chart Median 23+4 wk (range 22+0 to 25+6) Median 353 g (range 166‐677) 81/245 = 33% Lees et al10 Prospective multicenter non‐blinded management trial 503 AC below 10th centile according to local standards and abnormal umbilical artery Doppler PI above 95th centile based on local standards, irrespective of the presence of absent or reversed EDF Mean 29+0 ± 11 Mean 881 ± 217 g 195/503 = 38.8% Maged et al29 Prospective non‐randomized study 50 EFW <10th centile or AC <10th centile with abnormal umbilical artery Doppler indices Group 1 (n = 25): Mean 27.4 ± 1.6 Group 2 (n = 25): Mean 28.1 ± 1.5

Not described 36/74 = 48.6% Kubo et al27 Open label, phase 1 clinical trial 8 (<32 wk) EFW ≤ to −1.5 SD on ultrasonography from the Japanese standard table Median 28+4 (26+0 to 30+5) Median 967 (708‐1164) Not described Lawin‐O'Brien et al28 Multicenter retrospective study of databases 245 AC ≤3rd centile for gestational age, AC calculated according to UK recommended standard and Altman and Chitty chart Median 23+4 wk (range 22+0 to 25+6) Median 353 g (range 166‐677) 81/245 = 33% Lees et al10 Prospective multicenter non‐blinded management trial 503 AC below 10th centile according to local standards and abnormal umbilical artery Doppler PI above 95th centile based on local standards, irrespective of the presence of absent or reversed EDF Mean 29+0 ± 11 Mean 881 ± 217 g 195/503 = 38.8% Maged et al29 Prospective non‐randomized study 50 EFW <10th centile or AC <10th centile with abnormal umbilical artery Doppler indices Group 1 (n = 25): Mean 27.4 ± 1.6 Group 2 (n = 25): Mean 28.1 ± 1.5 Not described Not described Petersen et al30 Retrospective cohort study 33 patients, with 36 pregnancies EFW <10th centile for GA and at least 2 of the following: normal karyotype, notched uterine artery Doppler waveforms in the second trimester, placental histology changes consistent with uteroplacental insufficiency Median 24 (range 18‐29) Median 364 g (range 167‐496) Not described Rizzo et al31 Cohort study (unclear whether prospective or retrospective) 31 EFW <10th centile for population standard confirmed at birth Median 26.1 wk (range 22.6‐29.1) Not described 0/31 = 0% Savchev et al32 Retrospective analysis of a prospective cohort 211 subgroup <32 wk EFW <10th centile Mean 28.1 ± 4.0 wk Mean 1061 ± 494 g 74/211 = 35.1% Sharp et al12 Randomized placebo‐controlled trial 135 AC or EFW <10th centile and absent or reversed EDF in the umbilical artery Group 1 (n = 70): Median 25.1 (24.0‐27.5)

chev et al32 Retrospective analysis of a prospective cohort 211 subgroup <32 wk EFW <10th centile Mean 28.1 ± 4.0 wk Mean 1061 ± 494 g 74/211 = 35.1% Sharp et al12 Randomized placebo‐controlled trial 135 AC or EFW <10th centile and absent or reversed EDF in the umbilical artery Group 1 (n = 70): Median 25.1 (24.0‐27.5) Group 2 (n = 65): Median 25.6 (24.1‐27.4) Group 1 (n = 70): Median 451 (352‐613) Group 2 (n = 65): Median 436 (326‐594) 24/135 = 17.8% Simonazzi et al13 Retrospective cohort study 16 EFW and/or AC <5th centile Median 22+3 (range 20+0 to 23+3) Median 324 g (range 248‐509) Not described Story et al33 Retrospective cohort study 20 EFW <3rd centile Median 21+4 (range 18+2 to 24+0) Not described Not described Takahashi et al34 Prospective cohort study 18 <1.5 SD Japanese standard Median 23.0 (range 18‐25) Not described 0/18 = 0% Temming et al35 Retrospective cohort study 355 EFW <10th centile using Warsof growth curves before 20+0 wk of gestation and Hadlock growth curves from 20+0 wk of gestation onward Mean 19.5 ± 0.9 Not described Not described Von Dadelszen et al36 Case‐control study 27 AC <5th centile Group 1 (n = 17) median 21+1 (19+5 to 23+2) Group 2 (n = 10) median 22+4 (21+1 to 23+4) Not described Not described Yildirim et al37 Retrospective cohort study 300 EFW <10th centile Group 1 (n = 137) median 30.8 (CI 30.3‐31.3) wk Group 2 (n = 163) 30.1 (CI 29.6‐30.6) wk Not described 184/300 = 61.3% Zhang‐Rutledge et al38 Retrospective cohort study 254 EFW ≤10th centile Group 1 (n = 91): Average 21+5 Group 2 (n = 163): Average 21+3

Not described Not described Yildirim et al37 Retrospective cohort study 300 EFW <10th centile Group 1 (n = 137) median 30.8 (CI 30.3‐31.3) wk Group 2 (n = 163) 30.1 (CI 29.6‐30.6) wk Not described 184/300 = 61.3% Zhang‐Rutledge et al38 Retrospective cohort study 254 EFW ≤10th centile Group 1 (n = 91): Average 21+5 Group 2 (n = 163): Average 21+3 Not described Not described Abbreviations: AC, abdominal circumference; EDF, end‐diastolic flow; EFW, estimated fetal weight; FGR, fetal growth restriction; GA, gestational age; HC, head circumference; IQR, interquartile range; PI, pulsatility index; SD, standard deviation. John Wiley & Sons, LtdTable S2A,B (see Supplementary material) shows the judgment of risk of bias of the individual studies. Two8, 9 of the 5 included randomized controlled trials (RCT)8, 9, 10, 11, 12 were judged as “unknown” risk of bias. This judgment was mostly based on the fact that these studies were retrospectively registered and not blinded, and that some of the baseline criteria and outcomes were not reported for pregnancies that involved neonatal death. The other RCTs and the observational studies included were generally judged as “low” risk of bias. 3.2 Synthesis of the results The results on mortality are summarized in Table 2. When combining all data on mortality, of 2895 pregnancies, 355 (12.3%; range 0%‐53%) ended in an antenatal death. Of 2540 live‐born children, 1 child was lost to follow up. In all, 192 (7.6% of 2539 and 6.6% of the total of 2895 pregnancies; range 0%‐71%) neonatal deaths occurred, and 2347 (81%; range 14%‐100%) of pregnancies survived.

l data on mortality, of 2895 pregnancies, 355 (12.3%; range 0%‐53%) ended in an antenatal death. Of 2540 live‐born children, 1 child was lost to follow up. In all, 192 (7.6% of 2539 and 6.6% of the total of 2895 pregnancies; range 0%‐71%) neonatal deaths occurred, and 2347 (81%; range 14%‐100%) of pregnancies survived. Table 2 Outcome data on mortality Number of patients in final analysis GA at delivery (wk + d) (mean ± SD or median [IQR]) Birthweight (g) (mean ± SD or median [IQR]) Antenatal death Live born Neonatal death Survival at discharge Ali et al9 73 7/80 lost to follow up Group 1 (n = 34) mean 36 ± 0.9 Group 2 (n = 34) mean 36 ± 0.7 Group 1 (n = 34) mean 2022 ± 25 Group 2 (n = 34) mean 2324 ± 19 0/73 = 0% 73/73 = 100% 5/73 = 6.8% 68/73 = 93.2% Ali et al8 55 5/60 lost to follow up Group 1 (n = 25) mean 36.8 ± 0.8 Group 2 (n = 20) mean 34.8 ± 0.6. (among surviving babies) Group 1 (n = 25) mean 1854 ± 262 Group 2 (n = 20) mean 1694 ± 169 (among surviving babies) 0/55 = 0% 55/55 = 100% 10/55 = 18.2% 45/55 = 81.8% Aoki et al14 17 Median 27.3 (23.7‐29.3) wk Median 568 g (300‐764) 1/17 = 5.9% 16/17 = 94.1% 2/16 = 12.5% 14/17 = 82.4% Baschat et al22 44 Median 29+6 (range 26+4 to 37+6) for live birth Median 26+6 (range 25+1 to 28) for stillbirth

Group 2 (n = 20) mean 1694 ± 169 (among surviving babies) 0/55 = 0% 55/55 = 100% 10/55 = 18.2% 45/55 = 81.8% Aoki et al14 17 Median 27.3 (23.7‐29.3) wk Median 568 g (300‐764) 1/17 = 5.9% 16/17 = 94.1% 2/16 = 12.5% 14/17 = 82.4% Baschat et al22 44 Median 29+6 (range 26+4 to 37+6) for live birth Median 26+6 (range 25+1 to 28) for stillbirth Median 725 (range 420‐2260) 10/44 = 22.7% 34/44 = 77.3% 1/34 = 2.9% 33/44 = 75.0% Belghiti et al23 10 with reported outcome Median 26+2 (25+6 to 26+6) Median 507 (429‐553) 1/10 = 10.0% 9/10 = 90.0% 6/9 = 66.7%, of which 4 intrapartum death 3/10 = 30.0% Fox et al17 252 Mean 39.2 ± 2.4 Mean 2999 ± 682 g 4/252 (1.6%) 248/252 (98.4%) 2/248 = 0.8% 246/252 = 97.6% Fujisaki et al24 14 Median 29+0 (26+6 to 35+3) Median 604 (437‐1340) 2/14 = 14.3% 12/14 = 85.7% 1/12 = 8.3% 11/14 = 78.6% Groom et al11 122 Group 1 (n = 63): Mean 31+5 ± 4+4 Group 2 (n = 59): Mean 31+2 ± 4+4 Group 1 (n = 63): 1233 ± 774 Group 2 (n = 59): 1184 ± 823 19/122 = 15.6% 103/122 = 84.4% 9/103 = 8.7% 94/122 = 77.0% Hasegawa et al25 26 Group 1 (n = 17) median 28.7 (24.7‐31.7) Group 2 (n = 9) median 28.5 (26.1‐32.4) Group 1 (n = 18) median 695 (424‐1016) g Group 2 (n = 8) median 568 (426‐654) g 1/26 = 3.8% 25/26 = 96.2% 1/25 = 4% 24/26 = 92.3% Herraiz et al26 73 1/74 lost to follow up Group 1 (FGR) mean 30.1 ± 3.2 Group 2 (FGR+PE) mean 29.4 ± 2.5 Group 1 (FGR): mean 994 ± 419 Group 2 (FGR+PE): 925 ± 308 4/73 = 5.5% (1 TOP and 3 IUFD) 69/73 = 94.5% 6/69 = 8.7% 63/73 = 86.3% Kubo et al27 8 Median 37+0 (35+2 to 37+0) Median 2157 (1553‐2281) 0/8 = 0% 8/8 = 100% 0/8 = 0% 8/8 = 100% Lawin‐O'Brien et al28 245

1/74 lost to follow up Group 1 (FGR) mean 30.1 ± 3.2 Group 2 (FGR+PE) mean 29.4 ± 2.5 Group 1 (FGR): mean 994 ± 419 Group 2 (FGR+PE): 925 ± 308 4/73 = 5.5% (1 TOP and 3 IUFD) 69/73 = 94.5% 6/69 = 8.7% 63/73 = 86.3% Kubo et al27 8 Median 37+0 (35+2 to 37+0) Median 2157 (1553‐2281) 0/8 = 0% 8/8 = 100% 0/8 = 0% 8/8 = 100% Lawin‐O'Brien et al28 245 79/324 lost to follow up 128 (52.2%) <28+0 53 (21.6%) 28+1 to 32+0 24 (9.8%) 32+1 to 36+0 36 (14.7%) >36+0 Survived: Median 1020 g (range 435‐3420) Neonatal death: median 560 (range 313‐2550) Fetal death median 422 (range 155‐2570) Feticide/TOP median 345 (range 220‐512) 89 fetal death and 33 feticide/TOP = 122/245 = 49.8% 123/245 = 50.2% 22/123 = 17.9% 101/245 = 41.2% Lees et al10 503 9/511 lost to follow up Mean 30+5 ± 16 Mean 1013 ± 321 g 12/503 = 2.4% 491/503 = 97.6% 27/490 = 5.5% (1 live‐born lost to follow up) 463/503 = 92.0% Maged et al29 50 Group 1 (n = 25): Mean 35.3 ± 1.8 Group 2 (n = 25): Mean 34.8 ± 1.9 Group 1 (n = 25): Mean 2067 ± 352 Group 2 (n = 25): 1733 ± 361 4/50 = 8.0% 46/50 = 92.0% 4/46 = 8.7% 42/50 = 84.0% Petersen et al30 33 patients, with 36 pregnancies IUFD: Median 25 wk (range 21‐27) Live birth: Median 27 wk (range 24‐31) IUFD: Median 308 (range 170‐480) Live birth: Median 486 (320‐553) 19/36 = 52.8% 17/36 = 47.2% 12/17 = 70.6% 5/36 = 13.9% Rizzo et al31 31 Median 28.3 wk (range 23.6‐30.4) Median 590 g (range 312‐915) 7/31 = 22.6% 24/31 = 77.4% 3/24 = 12.5% 21/31 = 67.7% Savchev et al32 211 Mean 34.6 ± 8.0 wk Mean 1647 ± 765 g 9/211 = 4.3% 202/211 = 95.7% 6/202 = 3.0% 196/92.9% Sharp et al12 135 Group 1 (n = 70): Median 28.1 (26.7‐29.7)

70.6% 5/36 = 13.9% Rizzo et al31 31 Median 28.3 wk (range 23.6‐30.4) Median 590 g (range 312‐915) 7/31 = 22.6% 24/31 = 77.4% 3/24 = 12.5% 21/31 = 67.7% Savchev et al32 211 Mean 34.6 ± 8.0 wk Mean 1647 ± 765 g 9/211 = 4.3% 202/211 = 95.7% 6/202 = 3.0% 196/92.9% Sharp et al12 135 Group 1 (n = 70): Median 28.1 (26.7‐29.7) Group 2 (n = 65): Median 28.4 (27.3‐30.1) Group 1 (n = 70): Median 604 (496‐766) Group 2 (n = 65): Median 590 (430‐842) 43/135 = 31.9% 92/135 = 68.1% 17/92 = 18.5% 75/135 = 55.6% Simonazzi et al13 16 Group 1 (n = 4) median 34 wk (30‐36) Group 2 (n = 11) median 28 wk (24‐30) Group 1 (n = 4) median 1598 g (1100‐1750) Group 2 (n = 11) median 630 g (408‐951) 1/16 TOP = 6.25% 15/16 = 93.8% 3/15 = 20% 12/16 = 75.0% Story et al33 20 Survived (n = 12): median 32+0 (range 27+1 to 39+0) Neonatal death (n = 2): 26 wk and 31+5 IUFD (n = 6): median 26 (range 24+2 to 27+2) Survived (n = 12): median 980 (range 720‐2090) Neonatal death (n = 2): 620 and 1050 Fetal death (n = 6): median 450 (range 424‐530) 6/20 = 30% 14/20 = 70% 2/14 = 14.3% 12/20 = 60.0% Takahashi et al34 18 Median 28.5 wk (26.0‐30.3) Median 625 g (3630‐850) 5/18 = 27.8% 13/18 = 72.2% 2/13 = 15.4% 11/18 = 61.1% Temming et al35 355 Mean 37.2 ± 3.4 Mean 2725 ± 763 g 9/355 (2.5%) 346/355 (97.5%) 5/346 (1.4%) 341/355 = 96.1% Von Dadelszen et al36 27 Group 1 (n = 17): median 25+6 (23+5 to 29+5) Group 2 (n = 10): median 27+1 (25+4 to 32+6) Not described 14/27 = 51.9% 13/27 = 48.1% 2/13 = 15.4% 11/27 = 40.7% Yildirim et al37 300 Group 1 (n = 137): median 32.8 (95% CI 32.3‐33.3) wks Group 2 (n = 163): median 31.3 (95% CI 30.84‐31.7) wk

6/20 = 30% 14/20 = 70% 2/14 = 14.3% 12/20 = 60.0% Takahashi et al34 18 Median 28.5 wk (26.0‐30.3) Median 625 g (3630‐850) 5/18 = 27.8% 13/18 = 72.2% 2/13 = 15.4% 11/18 = 61.1% Temming et al35 355 Mean 37.2 ± 3.4 Mean 2725 ± 763 g 9/355 (2.5%) 346/355 (97.5%) 5/346 (1.4%) 341/355 = 96.1% Von Dadelszen et al36 27 Group 1 (n = 17): median 25+6 (23+5 to 29+5) Group 2 (n = 10): median 27+1 (25+4 to 32+6) Not described 14/27 = 51.9% 13/27 = 48.1% 2/13 = 15.4% 11/27 = 40.7% Yildirim et al37 300 Group 1 (n = 137): median 32.8 (95% CI 32.3‐33.3) wks Group 2 (n = 163): median 31.3 (95% CI 30.84‐31.7) wk Group 1 (n = 137) median 1390 (95% CI 1308‐1473) g Group 2 (n = 163) median 1071 (95% CI 1011‐1131) g 58/300 = 19.3% (perinatal death) 242/300 = 80.7% 29/242 = 12.0% 9 deaths >28 d (9/242 = 3.7%) 204/300 = 68.0% Zhang‐Rutledge et al38 254 Group 1 (n = 91): 37.1 Group 2 (n = 163): 38.3 (not described whether means or medians are presented) Group 1 (n = 91): Average 2605 Group 2 (n = 163): Average 2936 4/254 = 1.6% 250/254 = 98.4% 6/250 = 2.4% 244/254 = 96.1% Abbreviations: FGR, fetal growth restriction; GA, gestational age; IQR, interquartile range; IUFD, intrauterine fetal death; PE, preeclampsia; SD, standard deviation; TOP, termination of pregnancy.

Group 2 (n = 163): 38.3 (not described whether means or medians are presented) Group 1 (n = 91): Average 2605 Group 2 (n = 163): Average 2936 4/254 = 1.6% 250/254 = 98.4% 6/250 = 2.4% 244/254 = 96.1% Abbreviations: FGR, fetal growth restriction; GA, gestational age; IQR, interquartile range; IUFD, intrauterine fetal death; PE, preeclampsia; SD, standard deviation; TOP, termination of pregnancy. John Wiley & Sons, LtdA subset of the studies report neonatal morbidity (see Supplementary material, Table S3). When combining the data, 34% of the live‐born neonates experienced respiratory distress syndrome (2 studies, range 34%‐36%), 9.1% had bronchopulmonary dysplasia (4 studies, range 4%‐19%), 4.3% had intraventricular hemorrhage (10 studies, range 0%‐25%), 5.6% had necrotizing enterocolitis (9 studies, range 0%‐22%), 2.6% had persistent pulmonary hypertension of the newborn (2 studies, range 1.9%‐9.1%), 12.5% had retinopathy of prematurity (4 studies, range 2%‐29%) and 30% had sepsis (4 studies, range 25%‐64%). One study used a composite outcome for severe neonatal morbidity13 and 1 study used a composite for respiratory distress syndrome and chronic lung disease.14

tension of the newborn (2 studies, range 1.9%‐9.1%), 12.5% had retinopathy of prematurity (4 studies, range 2%‐29%) and 30% had sepsis (4 studies, range 25%‐64%). One study used a composite outcome for severe neonatal morbidity13 and 1 study used a composite for respiratory distress syndrome and chronic lung disease.14 The ages at which the neurodevelopmental outcome was assessed, the types of tests used for the assessment and the definition of NDI differed between studies. Therefore, not all studies could be included in the evidence table. From the 476 children (402 from 1 larger study, the remainder from 6 small studies) who underwent neurodevelopmental assessment (Table 3), 58 children (12%; 0%‐27%) suffered from cognitive impairment and/or cerebral palsy. Overall, cerebral palsy rates in the 7 studies were low: varying from 1% to 10%. NDI was diagnosed in 50 children (11% of surviving children assessed). Eight percent of 629 pregnancies resulted in a surviving infant with NDI. Only Lees et al,10 reporting 10% NDI among the assessed children, included all important domains in the definition of NDI (Bayley III score, cerebral palsy, hearing loss and visual loss). Table 3 Outcome data on long‐term follow up Number of surviving children assessed in follow up Age at assessment Definition of NDI Neurodevelopmental test used Proportion of children with NDI Aoki et al14 12/14 = 85.7% Not described “Handicapped” (not further explained) Not described 3/12 = 25.0% of surviving children 3/17 = 17.6% of all pregnancies

Table 3 Outcome data on long‐term follow up Number of surviving children assessed in follow up Age at assessment Definition of NDI Neurodevelopmental test used Proportion of children with NDI Aoki et al14 12/14 = 85.7% Not described “Handicapped” (not further explained) Not described 3/12 = 25.0% of surviving children 3/17 = 17.6% of all pregnancies Fujisaki et al24 11/11 = 100% 18 mo “Mental retardation” was defined as developmental quotient of <70 Kyoto Scale of psychological Development 2001 3/11 = 27.3% of surviving children 3/14 = 21.4% of all pregnancies Hasegawa et al25 23/23 = 100% 2 y (corrected) Neurological complications were defined as cerebral palsy or mental retardation diagnosed by independent pediatric neurologists at corrected age of 2 y Not described 5/23 = 21.7% of surviving children (1 cerebral palsy, 4 mental retardation) 5/26 = 19.2% of all pregnancies Lees et al20 443/461 = 88% 2 y, corrected for prematurity A cognitive Bayley III score or corrected Bayley II mental development index score of less than 85 or an estimated cognitive delay of more than 3 mo, cerebral palsy, with a GMFCS of more than 1, hearing loss needing hearing aids, or severe visual loss (legally certifiable as blind or partially sighted) Bayley III Scales of Infant and Toddler Development or corrected Bayley II 39/402 = 9.7% of surviving children (443 with known outcome, but 402 neurodevelopmental assessed) 39/502 = 7.8% of all pregnancies Cerebral palsy 6/402 = 1.5% of surviving children Cerebral palsy 6/502 = 1.2% of all pregnancies

Lees et al20 443/461 = 88% 2 y, corrected for prematurity A cognitive Bayley III score or corrected Bayley II mental development index score of less than 85 or an estimated cognitive delay of more than 3 mo, cerebral palsy, with a GMFCS of more than 1, hearing loss needing hearing aids, or severe visual loss (legally certifiable as blind or partially sighted) Bayley III Scales of Infant and Toddler Development or corrected Bayley II 39/402 = 9.7% of surviving children (443 with known outcome, but 402 neurodevelopmental assessed) 39/502 = 7.8% of all pregnancies Cerebral palsy 6/402 = 1.5% of surviving children Cerebral palsy 6/502 = 1.2% of all pregnancies Petersen et al30 5/5 = 100% 2 y of age (corrected) Developmental delay was defined as >1 SD below the mean Griffiths Mental Developmental Scales 1/5 = 20.0% of surviving children 1/36 = 2.8% of all pregnancies Simonazzi et al13 12/12 = 100% Median 30 mo (24‐58 mo) Not described Not described 0/12 = 0% Cerebral palsy 1/12 = 8.3% of surviving children Cerebral palsy 1/16 = 6.3% of all pregnancies Takahashi et al34 11/11 = 100% Between 2 and 13 y median 6 y Developmental quotient <70 Kyoto scale of development and the Wechsler Intelligence Scale for Children III 0/11 = 0% Cerebral palsy 1/11 = 9.1% of surviving children Cerebral palsy 1/18 = 5.6% of all pregnancies Abbreviations: GMFCS, gross motor function classification system; NDI, neurodevelopmental impairment; SD, standard deviation.

Takahashi et al34 11/11 = 100% Between 2 and 13 y median 6 y Developmental quotient <70 Kyoto scale of development and the Wechsler Intelligence Scale for Children III 0/11 = 0% Cerebral palsy 1/11 = 9.1% of surviving children Cerebral palsy 1/18 = 5.6% of all pregnancies Abbreviations: GMFCS, gross motor function classification system; NDI, neurodevelopmental impairment; SD, standard deviation. John Wiley & Sons, LtdTables 4 and 5 present the quality of evidence for our research questions on the mortality and long‐term neurodevelopment, respectively. Our fourth and fifth research questions were not addressed in any of the included studies. Table 4 Evidence table on mortality outcomes No. of studies Certainty assessment Effect Certainty Importance Study design Risk of bias Inconsistency Indirectness Imprecision Other considerations No. of events No. of individuals Antenatal death—RCT 5 Randomized trials Seriousa Seriousb Not serious Not serious None 74 888 LOW CRUCIAL Antenatal death—observational studies 20 Observational studies Not serious Very seriousc Serious Not serious None 281 2007 VERY LOW CRUCIAL Neonatal death—RCT 5 Randomized trials Seriousa Seriousd Not serious Not serious None 68 813 LOW CRUCIAL Neonatal death—observational studies 20 Observational studies Not serious Very seriouse Seriousf Not serious None 123 1726 VERY LOW CRUCIAL Abbreviation: FGR, fetal growth restriction; GA, gestational age; RCT, randomized controlled trial. a Two out of 5 randomized controlled trials were rated as unknown risk of bias. b Rate of antenatal death varies between 0% and 31.9%.

CRUCIAL Neonatal death—observational studies 20 Observational studies Not serious Very seriouse Seriousf Not serious None 123 1726 VERY LOW CRUCIAL Abbreviation: FGR, fetal growth restriction; GA, gestational age; RCT, randomized controlled trial. a Two out of 5 randomized controlled trials were rated as unknown risk of bias. b Rate of antenatal death varies between 0% and 31.9%. c Rate of antenatal death varies between 0% and 52.8%. d Rate of neonatal death varies between 5.5 and 18.5%. e Rate of neonatal death varies between 0% and 70.6%. f Definitions of FGR and GA at inclusion differ. John Wiley & Sons, LtdTable 5 Evidence table on neurodevelopmental outcomes No. of studies Certainty assessment Effect Certainty Importance Study design Risk of bias Inconsistency Indirectness Imprecision Other considerations No. of events No. of individuals Neurodevelopmental impairment at or before 5 y of age in long‐term follow up—RCT 1 Randomized trials Not serious Not serious Not serious Not serious None 39 402 HIGH CRUCIAL Neurodevelopmental impairment at or before 5 y of age in long‐term follow up—observational studies 2 Observational studies Not serious Not serious Seriousa Seriousb None 6 28 LOW CRUCIAL Cerebral palsy at or before 5 y of age—RCT 1 Randomized trials Not serious Not serious Not serious Not serious None 6 402 HIGH IMPORTANT Cerebral palsy at or before 5 y of age—observational studies 2 Observational studies Not serious Not serious Serious Seriousb None 1 28 LOW IMPORTANT Abbreviation: FGR, fetal growth restriction, GA, gestational age; RCT, randomized controlled trial.

CRUCIAL Cerebral palsy at or before 5 y of age—RCT 1 Randomized trials Not serious Not serious Not serious Not serious None 6 402 HIGH IMPORTANT Cerebral palsy at or before 5 y of age—observational studies 2 Observational studies Not serious Not serious Serious Seriousb None 1 28 LOW IMPORTANT Abbreviation: FGR, fetal growth restriction, GA, gestational age; RCT, randomized controlled trial. a Definitions of FGR and GA at inclusion differ. b Small sample size of 1 study (5 children). John Wiley & Sons, Ltd4 DISCUSSION The aim of this systematic review was to collate evidence on the perinatal mortality, morbidity and long‐term (neuro‐)development of pregnancies complicated by early‐onset FGR. Particularly in pregnancies with fetal compromise around the limits of viability, information on fetal and neonatal prognosis could offer a guide in decision‐making for parents and obstetricians. We found that antenatal mortality was about twice as high as neonatal mortality. Only a few studies reported on the number of children diagnosed with relevant neonatal morbidity, such as respiratory distress syndrome, bronchopulmonary dysplasia, persistent pulmonary hypertension of the newborn and retinopathy of prematurity. Also, a minority of the studies reported outcomes of long‐term follow up. Moreover, neurodevelopmental assessments were performed at different ages and different neurodevelopmental measures were used.

syndrome, bronchopulmonary dysplasia, persistent pulmonary hypertension of the newborn and retinopathy of prematurity. Also, a minority of the studies reported outcomes of long‐term follow up. Moreover, neurodevelopmental assessments were performed at different ages and different neurodevelopmental measures were used. The strength of this systematic review is the broad literature search and the strict inclusion criteria. We excluded studies that included all their patients before 2000, as the level of (neonatal) health care was essentially different in that period. Many studies that reported long‐term follow up did not include the antenatal and/or neonatal mortality of the sample studied,5, 15 which could create selection bias and may lead to numbers on healthy survival of early‐onset severe FGR to be too optimistic. Therefore, we also predefined to exclude studies that used live birth or survival as starting criteria, as we consider it crucial to include data on all‐type mortality to allow proper conclusions about prognosis from the obstetric perspective. Severity of brain damage is not only associated with FGR, but also with perinatal/neonatal management, and survival bias was therefore taken into account.

val as starting criteria, as we consider it crucial to include data on all‐type mortality to allow proper conclusions about prognosis from the obstetric perspective. Severity of brain damage is not only associated with FGR, but also with perinatal/neonatal management, and survival bias was therefore taken into account. One weakness of this systematic review is the lack of consistency in the definition of FGR in the included studies. As is highlighted in Table 1, only the minority of the included studies report in detail on the definition of FGR that was used. Studies basing the diagnosis of FGR only on growth parameters are especially at risk of having included small‐for‐gestational‐age pregnancies as well, even though the risk of including small‐for‐gestational‐age pregnancies without placental insufficiency is higher above 32 weeks of gestation compared with pregnancies below 32 weeks of gestation.16 In particular, the study of Fox et al17 included a wide range of pregnancies based on the EFW <25th centile. Due to the fact that these pregnancies were antenatally diagnosed as being complicated by FGR, despite the wide definition used to diagnose the FGR and the possible bias that this could cause, we decided to include the study in the systematic review. Exclusion of this study led to an increase in the overall mortality from 18.9% to 20.4%: in total, 351 pregnancies ended in antenatal death and 190 in neonatal death, out of 2643 pregnancies (mortalities of 13.2% and 7.2%, respectively). The gestational age and EFW at diagnosis of FGR varied between the included studies and within some of the individual studies (with wide ranges or SD), possibly representing pregnancies with variable prognosis. The variety of definitions of FGR used and the range of gestational age and/or EFW of the included pregnancies are 2 of the reasons why the quality of evidence for most outcomes was rated very low, low or moderate, because the quality of evidence was downgraded due to serious indirectness18 based on differences in study populations.

definitions of FGR used and the range of gestational age and/or EFW of the included pregnancies are 2 of the reasons why the quality of evidence for most outcomes was rated very low, low or moderate, because the quality of evidence was downgraded due to serious indirectness18 based on differences in study populations. Another weakness is the lack of consistent information about hypertensive disorders of pregnancy as they share pathophysiology and often coincide. Interventions in the management of this syndrome may have caused bias in an unknown direction.19 One large well‐designed RCT20 provides high‐quality evidence on the mortality and morbidity outcomes and neurodevelopmental outcomes at 2 years of age.10 Limitations of this study are that it is a trial on patient management and some pregnancies were excluded because of fetal distress. However, the advantage of this RCT was the strict inclusion criteria of FGR and the relatively well‐organized follow up with high attrition rate.

velopmental outcomes at 2 years of age.10 Limitations of this study are that it is a trial on patient management and some pregnancies were excluded because of fetal distress. However, the advantage of this RCT was the strict inclusion criteria of FGR and the relatively well‐organized follow up with high attrition rate. Currently, there are no specific evidence‐based therapies for early‐onset severe FGR. In the absence of therapeutic interventions, standard management consists of intensive maternal and fetal monitoring and counseling with timed delivery. Increased fetal surveillance is performed in the period of fetal viability, so that decisions around management and timing of delivery, usually by cesarean section, can be made.3 Informed choices depend on data on fetal and neonatal survival and morbidity. Because of the higher antenatal mortality, we hypothesize that changing thresholds for intervention to decrease antenatal mortality may result in increased postnatal mortality or increased rates of NDIs. The aim for joint obstetric and neonatal care is to improve overall survival without impairments.

val and morbidity. Because of the higher antenatal mortality, we hypothesize that changing thresholds for intervention to decrease antenatal mortality may result in increased postnatal mortality or increased rates of NDIs. The aim for joint obstetric and neonatal care is to improve overall survival without impairments. Regarding the variability of prognostic profiles between patients, a systematic review of individual patient data would be useful, to be able to individualize prognostic counseling as much as possible. We excluded studies reporting on wider ranges of gestational age. This included 2 well‐designed studies investigating long‐term neurodevelopment.6, 21 In these studies, 10 out of 34 (29%) and 14 out of 149 (10%) children, respectively, had an abnormal IQ score, of which the latter percentage is in line with the findings of this systematic review. Together with the studies included in our analyses that reported on long‐term neurodevelopment, it illustrates the need for more prospective studies starting at diagnosis of FGR and extending to early school age development of the surviving children.

er percentage is in line with the findings of this systematic review. Together with the studies included in our analyses that reported on long‐term neurodevelopment, it illustrates the need for more prospective studies starting at diagnosis of FGR and extending to early school age development of the surviving children. 5 CONCLUSION In this systematic review based on 25 studies comprising 2895 pregnancies complicated by severe early‐onset FGR, we found that the overall rates of antenatal and neonatal death were 12.3% and 6.6%, respectively. Of the 476 children included in the long‐term follow up, 12.2% of the survivors (7.9% of all pregnancies) were affected by NDI and/or cerebral palsy. Data on neurodevelopment were much less reported and mostly during toddler years, and not school age. Conclusions at an individual level are hampered by the differences in study quality and prognostic characteristics. A future analysis with individual patient data might further improve individual patient counseling. Longer follow up in prospective FGR cohorts is needed to provide data on the balance between mortality and NDI. CONFLICT OF INTEREST None. Supporting information Click here for additional data file.

Abbreviations CIconfidence interval EWSearly warning system ICUintensive care unit MEOWSmodified early obstetric warning score Key message Current MEOWS trigger thresholds vary, based arbitrarily on clinical consensus or expert opinion. Limited evidence links implementation of MEOWS with improved outcomes in the general pregnant population and further research is required to develop predictive models and validate evidence‐based MEOWS. 1 INTRODUCTION Maternal deaths in the UK are rare (4.65 per 100 000 live births in 2012‐2014) compared with the global burden (216 per 1 00 000 live births in 2015).1, 2 99% of maternal mortalities occur in low‐income and middle‐income countries. Many women die from preventable causes. Despite a 44% reduction in maternal deaths globally between 1990 and 2015, approximately 830 women still die every day. There has been no significant change in the maternal death rates in the UK between 2009‐2011 and 2012‐2014, although there is a downward trend. The women whose deaths were reported in the last UK triennial report left behind 358 children.

aths globally between 1990 and 2015, approximately 830 women still die every day. There has been no significant change in the maternal death rates in the UK between 2009‐2011 and 2012‐2014, although there is a downward trend. The women whose deaths were reported in the last UK triennial report left behind 358 children. Delay is a major contributor to maternal mortality. The effective management of an acutely unwell pregnant woman can be challenging and requires an appropriate and timely response. An approach to tackling this clinical challenge is the use of early warning systems (EWS). EWS involve the routine monitoring and recording of vital signs or clinical observations on specifically designed charts with linked escalation protocols. They list criteria for abnormal physiological parameters that trigger a color‐coded or weighted scoring system aimed to guide the frequency of monitoring, need for, and urgency of clinical review. Color‐coded systems trigger a clinical response when one or more abnormal observation is recorded in the red zone or two or more mildly abnormal parameters in the amber zone. In this commentary we will examine the evidence base behind modified early obstetric warning scores (MEOWS) and discuss the issues surrounding current practice.

tems trigger a clinical response when one or more abnormal observation is recorded in the red zone or two or more mildly abnormal parameters in the amber zone. In this commentary we will examine the evidence base behind modified early obstetric warning scores (MEOWS) and discuss the issues surrounding current practice. The UK Confidential Enquiry into Maternal Deaths published in 2007 recommended the introduction of MEOWS for all acute obstetric admissions, including women in early pregnancy.3 In 2012 the Royal College of Physicians introduced national early warning scores for use in the general adult population.4 Thereafter, there has been a growing trend towards the use of similar charts in the pregnant population. The MEOWS included in the 2007 UK Confidential Enquiry consisted of scores of respiratory rate, oxygen saturation, temperature, heart rate, blood pressure, assessment of urine, including for proteinuria, color of amniotic fluid, neurological response, pain score, assessment of lochia, and an overall assessment of whether the woman appears well.3 Subsequent reports have recommended their use but have also acknowledged their shortcomings and the lack of an evidence‐based standardized approach in maternity populations. The latest report published in 2016 raises the issue of ensuring that MEOWS remain fit for purpose in view of developments in modern information technology, near patients’ testing, and sepsis red flags.1

so acknowledged their shortcomings and the lack of an evidence‐based standardized approach in maternity populations. The latest report published in 2016 raises the issue of ensuring that MEOWS remain fit for purpose in view of developments in modern information technology, near patients’ testing, and sepsis red flags.1 A principal component of an EWS relies on a phase of decompensation in the physiological parameters measured; signaling an opportunity to intervene and prevent further deterioration. Concerns have been raised in translating this principle directly to the pregnant population with its unique physiological adaptations, which may allow for a longer period of apparent compensation but one that can be followed by an abrupt deterioration. For this reason the level of consciousness as a parameter in an early warning score has been criticized. A reduced or altered level of consciousness is not an early warning sign but a red flag indicating established illness.1 Further, there is a paucity of data from the low‐risk pregnant population needed to establish the normal distributions of parameters for gestation‐specific vital signs. The Oxford‐based Pregnancy Physiology Pattern Prediction study, currently in progress, aims to collect these data and will provide a valuable basis for the development of a national obstetric EWS.5 It may assist in addressing the question of whether gestation‐specific MEOWS are necessary. However, it may not provide an answer as to how feasible this system may be or which vital signs or combinations are most predictive of maternal deterioration.

e a valuable basis for the development of a national obstetric EWS.5 It may assist in addressing the question of whether gestation‐specific MEOWS are necessary. However, it may not provide an answer as to how feasible this system may be or which vital signs or combinations are most predictive of maternal deterioration. 2 WHAT IS THE CURRENT EVIDENCE? A recent analysis of charts of vital signs from consultant‐led maternity units across the UK demonstrated considerable variation in EWS and escalation protocols.6 Smith et al extracted data from 120 such units and found they included a wide range of supposedly normal vital signs. In some units these parameters overlapped with those used to identify sepsis, a leading cause of maternal morbidity and mortality, as well as mild diastolic and severe systolic hypertension, as defined by the National Institute for Health and Care Excellence. A variation in escalation protocols was also noted in the use of a color‐coded trigger system, an aggregate weighted scoring system or a combination of both. Bick et al performed a cross‐sectional survey across the maternity services of the UK National Health Service.7 All except one of the 108 organizations that responded reported using EWS. Of these, 99% used them antenatally, 76% for women in labor, and 100% for postnatal. The challenges reported included overlapping EWS with the use of a partogram in labor, staff shortages, and delays in obtaining clinical reviews when escalation was required.

of the 108 organizations that responded reported using EWS. Of these, 99% used them antenatally, 76% for women in labor, and 100% for postnatal. The challenges reported included overlapping EWS with the use of a partogram in labor, staff shortages, and delays in obtaining clinical reviews when escalation was required. A systematic review conducted to aid in the development of an Irish national guideline looked at 33 studies separated into four categories—descriptive studies, clinical practice guidelines, effectiveness studies, and development and validation studies.8 They found little high‐quality evidence that MEOWS were being developed and tested for their predictive ability. The evidence currently available focuses on selected high‐risk obstetric populations. This review found that compliance with recording observations varied and was related to training issues. It improved after audit and diminished with the increasing length of inpatient stay.

ped and tested for their predictive ability. The evidence currently available focuses on selected high‐risk obstetric populations. This review found that compliance with recording observations varied and was related to training issues. It improved after audit and diminished with the increasing length of inpatient stay. An ethnographic study conducted in two UK hospitals over a 7‐month period also found that charts were used at the health provider’s discretion.9 The study, included in the systematic review, reported that MEOWS charts were perceived by some staff members as a challenge to their clinical judgement. Staff members, particularly in postnatal wards, raised concerns over what they perceived as the medicalization of healthy women, and the impact that the universal use of MEOWS would have on care, such as support for breastfeeding. There is a particular need to demonstrate their benefit with appropriate professional education in low‐risk women. Charts may have their highest impact in a population where clinical concerns are least, such as on the postnatal ward. It was reported that the positive aspects of MEOWS charts include their use as a visual aid to recognize negative trends especially where the decline is gradual. Defined thresholds enable staff to identify deviations from them and legitimize summoning help across hierarchical boundaries. Accordingly, these charts were seen in some cases as useful in escalating care quickly. They may also increase team awareness of critical situations.

ds especially where the decline is gradual. Defined thresholds enable staff to identify deviations from them and legitimize summoning help across hierarchical boundaries. Accordingly, these charts were seen in some cases as useful in escalating care quickly. They may also increase team awareness of critical situations. A validation study looking prospectively at 676 consecutive maternity inpatient admissions reviewed MEOWS triggers and morbidity.10 The MEOWS chart used was adapted from the 7th CEMACE report. Outcomes at 30 days were retrieved from hospital records and patients’ notes including morbidity, death, admission to ICU, and discharge. The data analyzed were tested for associations between abnormal triggers and the presence of morbidity by calculating the relative risk for individual parameters. They report 89% sensitivity (95% CI 81%‐95%) and 79% specificity (95% CI 76%‐82%) and concluded that MEOWS charts are a useful bedside test for predicting morbidity. A low positive predictive value of 39% indicated a low prevalence of maternal morbidity. The authors of this study conclude that this supports the widespread use of MEOWS for all obstetric patients. Although MEOW’s ability to identify morbidity is promising, further evidence is needed to demonstrate the benefit of introducing this to the pregnant population as a whole. This would include its benefit over established clinical practice and health economic benefits.

widespread use of MEOWS for all obstetric patients. Although MEOW’s ability to identify morbidity is promising, further evidence is needed to demonstrate the benefit of introducing this to the pregnant population as a whole. This would include its benefit over established clinical practice and health economic benefits. A retrospective case‐controlled study carried out across 2 tertiary obstetric units in Canada extracted MEOWS variables in the 24 h prior to women’s admission to the ICU.11 This validation study found MEOWS to be very sensitive in predicting ICU admission and reported it had a high negative predictive value. However, MEOWS was not specific, and many controls who had no complications also triggered on thresholds used. Four variables (maximum temperature, heart rate, systolic blood pressure, and respiratory rate) were significantly associated with admission to the ICU. Secondary analyses and the use of the four variable model had high sensitivity, specificity, negative predictive value, and area under the receiver operating characteristic curve for admission to the ICU.

heart rate, systolic blood pressure, and respiratory rate) were significantly associated with admission to the ICU. Secondary analyses and the use of the four variable model had high sensitivity, specificity, negative predictive value, and area under the receiver operating characteristic curve for admission to the ICU. Shields et al conducted a pilot study over 6 sites in the USA following the implementation of a maternity early warning trigger tool.12 They aimed to evaluate prospectively the use of a pathway‐specific early warning trigger tool and determine whether it was associated with reduced maternal morbidity compared with controls in non‐pilot sites. The tool addressed four main causes of morbidity—sepsis, cardiovascular dysfunction, severe preeclampsia or hypertension, and severe hemorrhage, and included recommendations for patients’ assessment and treatment. The tool had two levels of activation—non‐severe, which required two triggers, or severe, requiring a single abnormal trigger. They reported a significant reduction in Centers for Disease Control‐defined severe maternal morbidity (P = 0.01) and composite maternal morbidity (P = 0.01) comparing baseline and after the implementation of the maternity early warning trigger tool. The authors suggest that areas to focus on in the development of an EWS are ease of use and an alert frequency low enough to prevent alarm fatigue but high enough for clinicians to associate value with the trigger.

y (P = 0.01) comparing baseline and after the implementation of the maternity early warning trigger tool. The authors suggest that areas to focus on in the development of an EWS are ease of use and an alert frequency low enough to prevent alarm fatigue but high enough for clinicians to associate value with the trigger. Carle et al performed a retrospective secondary analysis on physiological data collected for a routine audit from a critical care obstetric population.13 They set out to develop (n = 2240) and internally validate (n = 2200) an obstetric early warning system using separate samples of women. The most abnormal measurements within the first 24 h following admission to critical care were analyzed. Significant variables (P < 0.05) were weighted following multiple logistic regressions and used to create a statistical obstetric trigger system. The primary outcome was survival. They reported an area under the receiver operator curve of 0.995 (95% CI 0.992‐0.998) for the resulting statistical score. Additional variables were then added to create a clinical score. This included high systolic and diastolic blood pressure and temperature >38°C, although these variable were not demonstrated to be statistically significant. Their addition to this score was felt necessary to make the system clinically relevant and acceptable and did not result in a significant decrease in the score’s discriminatory ability. A sensitivity of 97% and a specificity of 87% were quoted with a score of 12. This may indicate it is able to discriminate between survivors and non‐survivors but it does not validate a system for the early detection of deterioration. Although this article reports on a large data set, it cannot be directly translated to the ward setting and further external validation is required. Furthermore, a retrospective secondary analysis of data has potentially missed clinically relevant data. The non‐ventilated respiratory rate was absent from 96% of the cases that resulted in death in the validation set of data.

nnot be directly translated to the ward setting and further external validation is required. Furthermore, a retrospective secondary analysis of data has potentially missed clinically relevant data. The non‐ventilated respiratory rate was absent from 96% of the cases that resulted in death in the validation set of data. 3 DISCUSSION The principle of a maternity‐specific EWS to structure surveillance for hospitalized women is intuitively sound. Their current widespread use and policy support, including recommendations following confidential enquiries and from the National Health Service Litigation Authority, are not, however, currently backed by a strong evidence base. There is limited evidence linking the implementation of MEOWS charts to improved outcomes in the general pregnant population. Deaths in the UK obstetric population are rare and the challenge is to equate the impact of MEOWS with meaningful outcomes. There appears to be room for improvement as substandard care continues to be highlighted in recurrent confidential enquiries.1 Future research aiming to identify the components of MEOWS that are most strongly related to adverse outcomes and their relative interdependence on accompanying implementation outcomes may help to simplify management strategies and escalation protocols. The UK Royal College of Anaesthetists has recommended only one intermediate step prior to review by a senior clinician.14

OWS that are most strongly related to adverse outcomes and their relative interdependence on accompanying implementation outcomes may help to simplify management strategies and escalation protocols. The UK Royal College of Anaesthetists has recommended only one intermediate step prior to review by a senior clinician.14 The wide variation that exists in chart layout and implementation undermines confidence in the validity of MEOWS.6 Current trigger thresholds vary and are arbitrarily based on clinical consensus or expert opinion. Additional research is required to develop predictive models and to validate MEOWS in a general obstetric population, preferably with well‐designed controlled studies and appropriate health economics. Future studies could have a global focus, including the collection of clinical variables in low‐resource settings where intervention bias is less likely and adverse outcomes are more common. The issues of compliance in monitoring and recording may be addressed with the introduction of automated monitoring and trigger devices. The benefits of an automated EWS include improving the audit trail, the environmental benefits of a paperless system, time efficiency, and improved consistency. However, MEOWS should not be considered a substitute for clinical evaluation and assessment. CONFLICT OF INTEREST AS has developed an automated vital signs monitoring device (CRADLE Vital Signs Alert) with Microlife, incorporating a traffic light triage system. AS is principle investigator on the ongoing CRADLE Projects. JS is co‐investigator on the CRADLE 3 trial.

Abbreviations CCICanadian Classification of Health Interventions DCS, Diabetes Complications Severity Index GDMgestational diabetes ICD‐10‐CAInternational Classification of Disease Version 10 Canadian Modification Key message Iatrogenic delivery is associated with increased risk of neonatal morbidity/mortality for women with diabetes in the late preterm/early term period. However, iatrogenic delivery at 38, 39 or 40 weeks’ gestation was associated with lower rates of neonatal morbidity/mortality for women with gestational diabetes. 1 INTRODUCTION Diabetes in pregnancy has been shown to increase the risk of perinatal mortality, morbidity and congenital anomalies.1, 2 Risks of adverse outcomes are not constant across all types of diabetes; women with type 1 diabetes typically have the highest rate of adverse outcomes, followed by women with type 2 diabetes and women with gestational diabetes (GDM).3 Fetuses exposed to diabetes in utero appear to have a different growth trajectory, with excess weight in the trunk and shoulders, compared with fetuses of women without diabetes;4, 5 this places them at an increased risk for macrosomia and birth injuries such as shoulder dystocia and brachial plexus injuries, even after controlling for fetal size.6, 7 These risks can be reduced with good glycemic control pre‐conceptionally and throughout pregnancy.1, 8 The Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) study shows that there is a continuous linear association between maternal glucose levels and birthweight.9

injuries, even after controlling for fetal size.6, 7 These risks can be reduced with good glycemic control pre‐conceptionally and throughout pregnancy.1, 8 The Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) study shows that there is a continuous linear association between maternal glucose levels and birthweight.9 It has been hypothesized that iatrogenic delivery at earlier term gestation reduces the risk of neonatal complications (as opposed to expectant management).6 The clinical decision about when to deliver women with diabetes depends on many maternal and fetal factors and the potential risk of adverse outcomes.6, 10, 11 A 2018 Cochrane systematic review attempted to determine the optimal timing of delivery for women with preexisting diabetes but concluded that there was insufficient evidence to answer this question as no trials had been published in this population.12 A similar Cochrane systematic review, focusing on women with GDM, also concluded that there was limited data to support clinical decision making.13 To date, only three small randomized controlled trials of 200 women,14 425 women,15 and 100 women,16 respectively, have been published on timing of delivery among women with diabetes. Among women with insulin‐requiring diabetes in pregnancy, Kjos et al14 found that infants born following expectant management were significantly larger than infants born following induction of labor. The GINEXMAL trial found no significant differences in cesarean delivery rates or severe maternal/neonatal morbidity between women with labor GDM randomized to induction of labor or expectant management between 38+0 and 39+0 weeks of gestation.15 However, infants born following induction of labor were significantly more likely to have hyperbilirubinemia.15 Finally, the third small trial by Worda et al16 also did not observe significant differences in large‐for‐gestational age or cesarean delivery following induction of labor at 38 vs 40 weeks of gestation among women with insulin‐controlled GDM.

uction of labor were significantly more likely to have hyperbilirubinemia.15 Finally, the third small trial by Worda et al16 also did not observe significant differences in large‐for‐gestational age or cesarean delivery following induction of labor at 38 vs 40 weeks of gestation among women with insulin‐controlled GDM. To our knowledge, no study has examined the impact of gestational age at delivery on maternal and neonatal morbidity in women with diabetes, and how this varies by type of diabetes. As such, this study aimed to quantify the week‐specific risks of maternal and neonatal morbidity and mortality in women who delivered iatrogenically following induction of labor or pre‐labor cesarean section compared with expectant management for women with type 1 diabetes, type 2 diabetes and GDM.

by type of diabetes. As such, this study aimed to quantify the week‐specific risks of maternal and neonatal morbidity and mortality in women who delivered iatrogenically following induction of labor or pre‐labor cesarean section compared with expectant management for women with type 1 diabetes, type 2 diabetes and GDM. 2 MATERIAL AND METHODS Our study population was drawn from all non‐anomalous singleton hospital births in Canada (except Quebec) at or beyond 36 weeks’ gestation from 2004 to 2014. Individual level de‐identified delivery data were obtained from the Canadian Institute of Health Information Discharge Abstract Database. Data were entered into the Discharge Abstract Database by trained health coders and this database has been validated for use in perinatal epidemiologic studies.17 Delivery was identified using International Classification of Disease Version 10 Canadian Modification (ICD‐10‐CA) codes for live births and stillbirths. Mother‐infant dyads were excluded if delivery occurred at <36 weeks’ or >42 weeks’ gestation (as delivery prior to this stage would not be clinically recommended in the absence of other issues impacting maternal or fetal health), gestational age data were missing or the infant had a congenital anomaly.

and stillbirths. Mother‐infant dyads were excluded if delivery occurred at <36 weeks’ or >42 weeks’ gestation (as delivery prior to this stage would not be clinically recommended in the absence of other issues impacting maternal or fetal health), gestational age data were missing or the infant had a congenital anomaly. The following hierarchical algorithm was used to classify women with diabetes: codes for Type 1 diabetes (Mother: ICD‐10‐CA E10.x, O24.5) superseded all other diabetes codes, and codes for Type 2 diabetes (Mother: ICD‐10‐CA E11.x, O24.6) superseded GDM codes. Women with codes for GDM (Mother: ICD‐10‐CA O24.8; Infant: ICD‐10‐CA P70.0) who did not have other codes for preexisting diabetes were classified as having GDM. The comparison group consisted of women with none of the above‐mentioned codes for diabetes. Women with impaired glucose tolerance but who had not been diagnosed with type 1, type 2 or GDM, were therefore included in the comparison group. For each woman with type 1 or type 2 diabetes, a Diabetes Complications Severity Index (DCSI) score was calculated. The DCSI is an ICD‐10 based scoring system that assigns zero (no complications), one (no severe complications, but at least one mild complication) or two (one or more severe complications) points based on the presence of any of seven diabetic complications, namely, cardiovascular, cerebrovascular or metabolic complications or nephropathy, neuropathy, peripheral vascular disease or retinopathy.18

), one (no severe complications, but at least one mild complication) or two (one or more severe complications) points based on the presence of any of seven diabetic complications, namely, cardiovascular, cerebrovascular or metabolic complications or nephropathy, neuropathy, peripheral vascular disease or retinopathy.18 Women were categorized as undergoing iatrogenic delivery if codes were present for labor induction or pre‐labor cesarean delivery. Labor induction was identified based on a Canadian Classification of Health Interventions (CCI) code. Currently, no CCI or ICD‐10‐CA codes exist to identify pre‐labor cesarean delivery;19 however, a validated algorithm has been developed and used in multiple studies to identify labor using ICD‐9‐CM codes.20, 21, 22 Women were classified as having a pre‐labor cesarean section if they had a CCI code for cesarean delivery and no ICD‐10‐CA codes were present indicating that labor occurred. Women who did not deliver in a specific week of gestation were categorized as undergoing expectant management at that week of gestation and served as the comparison group, regardless of whether they later delivered following spontaneous onset of labor or iatrogenic delivery. Women who experienced spontaneous labor onset and subsequent delivery at a given week of gestation were excluded from the comparison (expectant management) group for that particular week of gestation (Figure 1; Figure S1). Figure 1 Study flow diagram

Women were categorized as undergoing iatrogenic delivery if codes were present for labor induction or pre‐labor cesarean delivery. Labor induction was identified based on a Canadian Classification of Health Interventions (CCI) code. Currently, no CCI or ICD‐10‐CA codes exist to identify pre‐labor cesarean delivery;19 however, a validated algorithm has been developed and used in multiple studies to identify labor using ICD‐9‐CM codes.20, 21, 22 Women were classified as having a pre‐labor cesarean section if they had a CCI code for cesarean delivery and no ICD‐10‐CA codes were present indicating that labor occurred. Women who did not deliver in a specific week of gestation were categorized as undergoing expectant management at that week of gestation and served as the comparison group, regardless of whether they later delivered following spontaneous onset of labor or iatrogenic delivery. Women who experienced spontaneous labor onset and subsequent delivery at a given week of gestation were excluded from the comparison (expectant management) group for that particular week of gestation (Figure 1; Figure S1). Figure 1 Study flow diagram The primary outcomes were a composite of maternal mortality or severe maternal morbidity, and severe neonatal morbidity or mortality. Maternal death or severe maternal morbidity was defined as the occurrence of one or more of the following conditions/procedures in the immediate perinatal period: maternal death prior to discharge, obstetric embolism, obstetric shock, postpartum hemorrhage with hysterectomy or other procedures to control bleeding, sepsis, thromboembolism or uterine rupture. The specific ICD‐10‐CA and CCI codes used can be found in Table S2. The above‐listed outcomes were selected to reduce the probability of confounding by indication, as they are unlikely to be an indication for labor induction or pre‐labor cesarean delivery but may be an unintended consequence of these procedures.19 Severe neonatal morbidity or mortality was defined as: birth asphyxia, fetal asphyxia, intraventricular hemorrhage (grade 3 or 4), neonatal convulsions, other disturbances of cerebral status of newborn, respiratory distress syndrome, birth injury to central nervous system, birth injury to peripheral nervous system, birth injury to skeleton, fetal fracture of humerus or clavicle to facilitate delivery, shoulder dystocia, stillbirth or neonatal death. The specific ICD‐10‐CA and CCI codes used can be found in Table S2.

, respiratory distress syndrome, birth injury to central nervous system, birth injury to peripheral nervous system, birth injury to skeleton, fetal fracture of humerus or clavicle to facilitate delivery, shoulder dystocia, stillbirth or neonatal death. The specific ICD‐10‐CA and CCI codes used can be found in Table S2. 2.1 Statistical analyses At each gestational age, the observed rate of each outcome was calculated per 100 deliveries following iatrogenic delivery and per 100 ongoing pregnancies following expectant management. These separate denominators were chosen as they mimic the real‐life decisions that clinicians make (ie, iatrogenic delivery or expectant management). Of note, a woman who was expectantly managed at 37 weeks of gestation could later be delivered iatrogenically at 38 weeks if her clinical situation changed. These observed rates were used to generate unadjusted risk differences and risk ratios. Logistic regression models were derived to calculate adjusted risk ratios, risk differences, and absolute predicted risks for each outcome at each week of gestation (ie, at each week of gestation, the risk of adverse outcomes following iatrogenic delivery was compared with the same risk among ongoing pregnancies). All models were adjusted for year, parity and the obstetric comorbidity score (a validated composite risk score that included preexisting chronic disease, pregnancy‐associated disease and maternal age23, 24). Although preexisting diabetes is a component of the obstetric comorbidity score, it was not included in the index for this study, as the type of diabetes was the primary exposure variable. Models examining neonatal outcomes were further adjusted for infant sex. As there is an increased risk of recurrence for many adverse obstetric events, a sensitivity analysis was conducted, restricted to the first birth during the study period for all women. All analyses was conducted using STATA SE Version 14 (StataCorp.).

s examining neonatal outcomes were further adjusted for infant sex. As there is an increased risk of recurrence for many adverse obstetric events, a sensitivity analysis was conducted, restricted to the first birth during the study period for all women. All analyses was conducted using STATA SE Version 14 (StataCorp.). 2.2 Ethical approval Ethics approval for this study was obtained from the Conjoint Health Research Ethics Board at the University of Calgary on 3 September 2014 (REB14‐1314). Patients were not directly involved in this study.

s examining neonatal outcomes were further adjusted for infant sex. As there is an increased risk of recurrence for many adverse obstetric events, a sensitivity analysis was conducted, restricted to the first birth during the study period for all women. All analyses was conducted using STATA SE Version 14 (StataCorp.). 2.2 Ethical approval Ethics approval for this study was obtained from the Conjoint Health Research Ethics Board at the University of Calgary on 3 September 2014 (REB14‐1314). Patients were not directly involved in this study. 3 RESULTS Overall, 3 029 523 women delivered during the study period and their records could be linked to their infants. Women were excluded due to missing information on gestational age (n = 3857), delivery at <35 weeks of gestation (n=142 398) or >42 weeks of gestation (n= 353), having an infant with a congenital anomaly (n=141 883). Overall, 2 707 471 women were included in this study; including 5889 women with type 1 diabetes, 9422 women with type 2 diabetes and 138 917 women with GDM. The prevalence of diabetes‐related complications was low—97.0% of women with type 1 diabetes and 99.1% of women with type 2 diabetes had a DCSI of zero, indicating no diabetes‐related complications. Women with any form of diabetes were approximately twice as likely to have an iatrogenic delivery compared with women without diabetes and were also more likely to experience an adverse maternal or neonatal outcome (Table 1). Similar characteristics were observed when the analysis was restricted to first births during the study period (Table S3). The incidence of iatrogenic delivery by gestational age differed by type of diabetes, and elective cesarean section was used more frequently in women with any type of diabetes than in women without diabetes (Figure 2).

cs were observed when the analysis was restricted to first births during the study period (Table S3). The incidence of iatrogenic delivery by gestational age differed by type of diabetes, and elective cesarean section was used more frequently in women with any type of diabetes than in women without diabetes (Figure 2). Table 1 Characteristics of women delivering singleton non‐anomalous infants ≥36 wk of gestation in Canadian hospitals (excluding Quebec) between 2004 and 2014 Characteristic Women without diabetes n = 2 553 243 Women with type 1 diabetes n = 5889 Women with type 2 diabetes n = 9422 Women with gestational diabetes n = 138 917 Maternal age (y), mean (SD) 29.3 (5.6) 29.8 (5.3) P < 0.001 32.6 (5.5) P < 0.001 32.1 (5.3) P < 0.001 Gestational age at delivery (wk), mean (SD) 39.2 (1.2) 37.6 (1.0) P < 0.001 37.9 (1.1) P < 0.001 38.6 (1.2) P < 0.001 Iatrogenic delivery*, % (95% CI) 32.1 (32.1‐32.2) 68.7 (67.5‐69.9) P < 0.001 67.9 (66.9‐68.8) P < 0.001 52.2 (51.9‐52.5) P < 0.001 Mode of delivery, % (95% CI) Spontaneous vaginal 63.7 (63.7‐63.8) 30.7 (29.5‐31.9) 42.3 (41.3‐43.3) 53.0 (52.7‐53.2) Operative vaginal 10.6 (10.6‐10.7) 9.9 (9.2‐10.7) 7.3 (6.8‐7.9) 10.0 (9.8‐10.1) Cesarean section 25.6 (25.6‐25.7) 59.4 (58.1‐60.6) P < 0.001 50.4 (49.4‐51.4) P < 0.001 37.0 (36.8‐37.3) P < 0.001 Hypertensive disorders of pregnancy, % (95% CI) 4.8 (4.8‐4.8) 17.7 (16.7‐18.8) P < 0.001 15.3 (14.6‐16.1) P < 0.001 9.3 (9.1‐9.4) P < 0.001

Mode of delivery, % (95% CI) Spontaneous vaginal 63.7 (63.7‐63.8) 30.7 (29.5‐31.9) 42.3 (41.3‐43.3) 53.0 (52.7‐53.2) Operative vaginal 10.6 (10.6‐10.7) 9.9 (9.2‐10.7) 7.3 (6.8‐7.9) 10.0 (9.8‐10.1) Cesarean section 25.6 (25.6‐25.7) 59.4 (58.1‐60.6) P < 0.001 50.4 (49.4‐51.4) P < 0.001 37.0 (36.8‐37.3) P < 0.001 Hypertensive disorders of pregnancy, % (95% CI) 4.8 (4.8‐4.8) 17.7 (16.7‐18.8) P < 0.001 15.3 (14.6‐16.1) P < 0.001 9.3 (9.1‐9.4) P < 0.001 Parity, % (95% CI) Nulliparous 42.0 (41.9‐42.1) 44.5 (43.0‐45.9) 28.8 (27.8‐29.8) 35.2 (34.9‐35.5) Multiparous, no history of prior cesarean section 45.4 (26.3‐28.9) 27.6 (26.3‐28.9) 45.4 (44.3‐46.5) 43.9 (43.6‐44.3) Multiparous, prior cesarean section 12.6 (12.6‐12.6) 28.0 (26.7‐29.3) P < 0.001 25.8 (24.8‐26.8) P < 0.001 20.9 (20.6‐21.1) P < 0.001 Diabetes complications severity index score, mean (SD) — 3.5 (.2) 1.4 (.2) — Male infant sex, % (95% CI) 50.6 (50.6‐50.7) 49.9 (48.6‐51.2) P < 0.25 50.2 (49.2‐51.3) P < 0.33 51.5 (51.2‐51.7) P < 0.001 Infant birthweight (g), mean (SD) 3450.3 (481.1) 3710 (615.8) P < 0.001 3562.7 (623.4) P < 0.001 3446.4 (526.6) P < 0.04 Large for gestational age infant >97th percentile, %, 95% CI 9.2 (9.2‐9.3) 44.6 (43.3‐48.9) P < 0.001 31.3 (30.3‐32.2) P < 0.001 15.5 (15.3‐15.7) P < 0.001 Severe maternal morbidity/mortality, % (95% CI) 0.4 (0.4‐0.4) 0.6 (0.4‐0.8) P < 0.02 0.8 (0.6‐1.0) P < 0.001 0.5 (0.5‐0.6) P < 0.001 Severe neonatal morbidity/mortality, % (95% CI) 7.7 (7.7‐7.8) 20.8 (19.7‐21.8) P < 0.001 15.6 (14.9‐16.4) P < 0.001 9.8 (9.7‐10.0) P < 0.001

Large for gestational age infant >97th percentile, %, 95% CI 9.2 (9.2‐9.3) 44.6 (43.3‐48.9) P < 0.001 31.3 (30.3‐32.2) P < 0.001 15.5 (15.3‐15.7) P < 0.001 Severe maternal morbidity/mortality, % (95% CI) 0.4 (0.4‐0.4) 0.6 (0.4‐0.8) P < 0.02 0.8 (0.6‐1.0) P < 0.001 0.5 (0.5‐0.6) P < 0.001 Severe neonatal morbidity/mortality, % (95% CI) 7.7 (7.7‐7.8) 20.8 (19.7‐21.8) P < 0.001 15.6 (14.9‐16.4) P < 0.001 9.8 (9.7‐10.0) P < 0.001 Chi‐square tests were used to assess differences between groups for categorical variables, and ANOVA and t tests were used to assess differences between groups for continuous variables. P values compare women with diabetes with the general population. * Defined as a delivery following induction of labor or pre‐labor cesarean delivery. John Wiley & Sons, LtdFigure 2 Iatrogenic delivery by week of gestation. (A) Incidence of iatrogenic delivery by gestational age and type of diabetes. (B) Method of iatrogenic delivery by gestational age and type of diabetes

Chi‐square tests were used to assess differences between groups for categorical variables, and ANOVA and t tests were used to assess differences between groups for continuous variables. P values compare women with diabetes with the general population. * Defined as a delivery following induction of labor or pre‐labor cesarean delivery. John Wiley & Sons, LtdFigure 2 Iatrogenic delivery by week of gestation. (A) Incidence of iatrogenic delivery by gestational age and type of diabetes. (B) Method of iatrogenic delivery by gestational age and type of diabetes For all women, the absolute risk of severe maternal morbidity/mortality was very low, typically impacting less than 1% of women regardless of whether they were iatrogenically delivered or expectantly managed (Figure 3; Table S4). The absolute risk of severe maternal morbidity/mortality generally displayed a U‐shaped relation with gestational age. Among women without diabetes, iatrogenic delivery was associated with a significantly higher adjusted risk of severe maternal morbidity/mortality at 36, 37, 38, and 39 weeks of gestation compared with expectant management, and a decreased risk of severe maternal morbidity/mortality at 41 weeks of gestation. For example, at 36 weeks, there were 1.6 excess cases of severe maternal morbidity/mortality per 1000 deliveries following iatrogenic delivery compared with expectant management, whereas at 39 weeks, there were .5 excess cases of maternal morbidity/mortality per 1000 deliveries. No significant differences were observed between iatrogenic delivery and expectant management in gestational age‐specific rates of severe maternal morbidity/mortality among women with any form of diabetes. Results were similar when restricted to first births during the study period (Table S5).

lity per 1000 deliveries. No significant differences were observed between iatrogenic delivery and expectant management in gestational age‐specific rates of severe maternal morbidity/mortality among women with any form of diabetes. Results were similar when restricted to first births during the study period (Table S5). Figure 3 Predicted adjusted risk of maternal morbidity/mortality per 100 deliveries for (A) women without diabetes, (B) women with type 1 diabetes, (C) women with type 2 diabetes and (D) women with gestational diabetes. Estimates are adjusted for year, obstetric comorbidity score and parity. *Indicates statistically significant results. Maternal death or severe maternal morbidity was defined as the occurrence of one or more of the following conditions/procedures in the immediate postpartum period: maternal death prior to discharge, obstetric embolism, obstetric shock, postpartum hemorrhage with hysterectomy or other procedures to control bleeding, sepsis, uterine rupture, thromboembolism

bidity was defined as the occurrence of one or more of the following conditions/procedures in the immediate postpartum period: maternal death prior to discharge, obstetric embolism, obstetric shock, postpartum hemorrhage with hysterectomy or other procedures to control bleeding, sepsis, uterine rupture, thromboembolism The absolute risk of severe neonatal morbidity/mortality was much higher than the risk of maternal morbidity/mortality and also displayed a U‐shaped relation with gestational age (Figure 4; Table S6). Among women without diabetes and women with GDM, iatrogenic delivery was associated with an increased adjusted risk of neonatal morbidity/mortality at 36 and 37 weeks of gestation compared with expectant management. However, such women had a lower adjusted risk of neonatal morbidity/mortality at 38, 39 and 40 weeks of gestation following iatrogenic delivery. For example, at 36 weeks, there were 81.3 and 76.7 excess cases of neonatal morbidity/mortality per 1000 deliveries in women without diabetes and women with GDM, respectively, following iatrogenic delivery compared with expectant management, whereas at 39 weeks, there were 25.1 and 27.3 fewer cases of neonatal morbidity/mortality per 1000 deliveries in women without diabetes and with GDM, respectively, following iatrogenic delivery compared with expectant management. Increased risks of severe neonatal morbidity/mortality following iatrogenic delivery were also observed for women with type 1 diabetes at 36 (98.3 excess cases per 1000 deliveries) and 37 weeks of gestation (44.5 excess cases per 1000 deliveries) and for women with type 2 diabetes at 36 weeks of gestation (77.9 excess cases per 1000 deliveries) compared with expectant management. Results were similar when restricted to first births during the study period (Table S7).

ss cases per 1000 deliveries) and 37 weeks of gestation (44.5 excess cases per 1000 deliveries) and for women with type 2 diabetes at 36 weeks of gestation (77.9 excess cases per 1000 deliveries) compared with expectant management. Results were similar when restricted to first births during the study period (Table S7). Figure 4 Predicted adjusted risk of neonatal morbidity/mortality per 100 deliveries for infants born to (A) women without diabetes, (B) women with type 1 diabetes, (C) women with type 2 diabetes and (D) women with gestational diabetes. Estimates are adjusted for year, obstetric comorbidity score, parity and infant sex. *Indicates statistically significant results. Severe neonatal morbidity or mortality was defined as the occurrence of one or more of the following conditions/procedures during the birth hospitalization: birth asphyxia, fetal asphyxia, intraventricular hemorrhage (grade 3 or 4), neonatal convulsions, other disturbances of cerebral status of newborn, respiratory distress syndrome, birth injury to central nervous system, birth injury to peripheral nervous system, birth injury to skeleton, fetal fracture of humerus or clavicle to facilitate delivery, shoulder dystocia, stillbirth, neonatal death

neonatal convulsions, other disturbances of cerebral status of newborn, respiratory distress syndrome, birth injury to central nervous system, birth injury to peripheral nervous system, birth injury to skeleton, fetal fracture of humerus or clavicle to facilitate delivery, shoulder dystocia, stillbirth, neonatal death 4 DISCUSSION Our study suggests that there may be some benefit to the infant if delivery of women with diabetes occurred at 38, 39 or 40 weeks’ gestation. Timing of delivery for a woman with diabetes is a complex issue for patients and care providers and depends on several individual and contextual factors. The results of this study may be useful in supporting discussions on the risks and benefits of early delivery. However, one must recognize that even with the use of modern diabetes management tools, ideal glycemic control in women with type 1 diabetes is elusive, neonatal complications rates remain high25 and stillbirth is associated with poor glycemic control in women with type 1 or type 2 diabetes.11

nefits of early delivery. However, one must recognize that even with the use of modern diabetes management tools, ideal glycemic control in women with type 1 diabetes is elusive, neonatal complications rates remain high25 and stillbirth is associated with poor glycemic control in women with type 1 or type 2 diabetes.11 Currently no consensus exists in the literature or in clinical guidelines as to the optimal timing of delivery for women with diabetes.2, 6 Clinical practice patterns vary across provider types and settings. A survey of clinicians participating in an international conference on the management of patients with GDM found that for patients with diet‐controlled GDM, 10% delivered their patients at 38 weeks of gestation, 14% at 39 weeks, 57% at 40 weeks and 16% at 41 weeks.26 For patients with medication‐controlled GDM, 28% of respondents delivered their patients at 38 weeks of gestation, 46% at 39 weeks and 19% at 40 weeks.26

h GDM found that for patients with diet‐controlled GDM, 10% delivered their patients at 38 weeks of gestation, 14% at 39 weeks, 57% at 40 weeks and 16% at 41 weeks.26 For patients with medication‐controlled GDM, 28% of respondents delivered their patients at 38 weeks of gestation, 46% at 39 weeks and 19% at 40 weeks.26 The bulk of the literature on the timing of delivery in women with diabetes comes from retrospective observational studies and suffers from methodological limitations such as a failure to adjust for important confounders, an inability to distinguish diabetes sub‐types and a lack of information on glycemic control.27 A study using California vital statistics data from 1997 to 2006 found that the risk of stillbirth increased for all women with increasing gestational age, and that women with GDM had an increased risk of stillbirth compared with women without GDM at 36‐41 weeks of gestation.28 Even though the relative risk of stillbirth is increased for women with GDM, the absolute risk of stillbirth remains low—this translates to large numbers of women needing to be delivered at earlier gestational ages to prevent a single case of stillbirth (4435 at 38 weeks of gestation, 1518 at 39 weeks, and 1311 at 40 weeks).28 Results from the present study examining morbidity patterns corroborate these findings, and show a decreased risk of neonatal morbidity for women with GDM who delivered at 38, 39 or 40 weeks of gestation. The present study also shows the impact that late preterm and early term birth has on child health: infants born following iatrogenic delivery at 36 and 37 weeks of gestation for women with GDM and type 1 diabetes had an increased risk of severe neonatal morbidity/mortality.

elivered at 38, 39 or 40 weeks of gestation. The present study also shows the impact that late preterm and early term birth has on child health: infants born following iatrogenic delivery at 36 and 37 weeks of gestation for women with GDM and type 1 diabetes had an increased risk of severe neonatal morbidity/mortality. Our study, and the bulk of the literature in this area, supports delivery at 38, 39 or 40 weeks of gestation for women with GDM. However, there is insufficient evidence on timing of delivery for women with type 1 and type 2 diabetes. Our findings suggest no maternal benefit and little or no additional neonatal benefit of iatrogenic delivery at 39 rather than 38 weeks of gestation for women with type 1 or type 2 diabetes. However, these women have a much greater risk of stillbirth compared with women with GDM, and ideal glycemic control is challenging to achieve in women with type 1 diabetes. For this reason, there is little justification for delaying delivery of women with preexisting diabetes beyond 38 weeks of gestation, particularly in the presence of poor glycemic control, which is an independent risk factor for stillbirth.11