Browse the corpus

Use of each category of contraceptive method was indicated in four dichotomous variables whereby usual use of the method of interest was coded ‘1’, else ‘0’. These dichotomous variables were the dependent variables in binary logistic regression models. Initially each binary logistic regression model predicted the crude (univariable) odds of use of the method of interest by relationship status and duration, followed by mutual adjustment for both partnership characteristic variables. Statistical analyses were performed using Stata/SE 15 (StataCorp LP) taking into account the survey’s sample design, using the svy function. Results Contraceptive method use among all women, by age Table 1 shows usual contraceptive method, sexual activity and pregnancy status among all women aged 16–49 years, stratified by age group. Only 6.2% had not had sex with a man in the year prior to interview, and 15.0% were either currently pregnant, trying to conceive or menopausal, totalling 21.2% unlikely to have been at risk of unintended pregnancy. The methods most commonly used by those at risk were oral/injectable hormonal methods (26.0%), barrier methods (16.4%), and unreliable methods or none (14.8%). Use of hormonal methods was highest in the youngest age group, decreasing with successive age group. The reverse was true for unreliable and no method use, which increased with age group. Reliance on sterilisation also increased with age group and was reported by almost one in five 35–49-year-olds, but by virtually no 16–24-year-olds.

Key messages Use of barrier methods was higher in short-term relationships among younger participants, but this was not seen among the oldest respondents. Relationship duration was more strongly associated with the contraceptive method used than was self-defined status of relationship, and this pattern was more obvious among younger participants. Asking about partnership characteristics may help providers support women in their use of contraception by considering their priorities and preferences at different life stages. Background An estimated 16% of pregnancies in Britain are unplanned.1 Since contraception is available free of charge under National Health Service provision, and the prevalence of contraceptive use is high, efforts to reduce this number need to extend beyond addressing non-use. The risk of unplanned pregnancy resulting from method discontinuation has been shown to be almost as high as that resulting from no method use.2 Consequently, an understanding of influences on method choice is essential to contraceptive provision compatible with users’ lifestyles, and so less likely to be discontinued.

The risk of unplanned pregnancy resulting from method discontinuation has been shown to be almost as high as that resulting from no method use.2 Consequently, an understanding of influences on method choice is essential to contraceptive provision compatible with users’ lifestyles, and so less likely to be discontinued. Studies in the United States (US) have shown both age3 4 and relationship characteristics to be more strongly associated with contraceptive method than other demographic factors.5 Contraceptive use has been associated with the number of sexual partners in Australia and Europe,6–9 the quality and level of trust in relationships,5 10 and type of relationship in Britain and the US.11–16 Other US studies and one in Britain have shown that condom use decreases with the relationship duration,11–14 17 18 while hormonal method use increases.13–16 19 Other studies have shown use of any method increases with relationship duration.10 15 18

methods was highest in the youngest age group, decreasing with successive age group. The reverse was true for unreliable and no method use, which increased with age group. Reliance on sterilisation also increased with age group and was reported by almost one in five 35–49-year-olds, but by virtually no 16–24-year-olds. Table 1 ‘Usual’ contraceptive method use among women aged 16–49 years, stratified by age (column percentages)

elationships,5 10 and type of relationship in Britain and the US.11–16 Other US studies and one in Britain have shown that condom use decreases with the relationship duration,11–14 17 18 while hormonal method use increases.13–16 19 Other studies have shown use of any method increases with relationship duration.10 15 18 To date, no studies using large, representative samples of women have examined how associations between method use and relationship characteristics differ by age group, or have distinguished between relationship duration and type. Further, few have taken account of fertility intentions.13 20 21 The aim of this study was twofold: first, to describe contraceptive method use among women in Britain in 2010–2012 by age group, and second, to examine associations in different age groups between type of partnership, duration of relationship, and contraceptive method use by women (reporting in 2010–2012) at risk of unintended pregnancy. The study uses data from the third National Survey of Sexual Attitudes and Lifestyles (Natsal-3). Methods Study design and setting Data were analysed from Natsal-3, a cross-sectional survey of 15 162 16–74-year-old men and women living in Britain, interviewed between September 2010 and August 2012 (response rate 57.7%). The survey used a multistage, clustered and stratified probability sample. Natsal-3 data were weighted to adjust for the unequal probabilities of selection and a non-response post-stratification weight was applied.22 Natsal-3 was approved by the Oxfordshire Research Ethics Committee A (reference: 09/H0604/27).

Methods Study design and setting Data were analysed from Natsal-3, a cross-sectional survey of 15 162 16–74-year-old men and women living in Britain, interviewed between September 2010 and August 2012 (response rate 57.7%). The survey used a multistage, clustered and stratified probability sample. Natsal-3 data were weighted to adjust for the unequal probabilities of selection and a non-response post-stratification weight was applied.22 Natsal-3 was approved by the Oxfordshire Research Ethics Committee A (reference: 09/H0604/27). Participants Female Natsal-3 participants of ‘reproductive age’ (aged 16 to 49 years),23 were included in initial analyses of contraceptive method use (and stratified by age group) for whom usual contraceptive method was reported (n=5857). To explore patterns of method use by relationship type and duration by age group, we excluded those currently pregnant, trying to conceive or menopausal (n=802), those who had not had heterosexual sex in the recent time period (n=89), and participants who relied on sterilisation (n=379) (as few such women existed in the two younger age groups). We also excluded women who did not report year of both first and most recent occasions of sex with their most recent partner (n=129), or relationship status with their most recent partner (n=2). After applying these criteria, the analysis included 4456 women (online supplementary files 1 and 2).

wo younger age groups). We also excluded women who did not report year of both first and most recent occasions of sex with their most recent partner (n=129), or relationship status with their most recent partner (n=2). After applying these criteria, the analysis included 4456 women (online supplementary files 1 and 2). Measures Outcomes The primary outcome was the contraceptive method usually used. Responses to the question ‘Which would you say is your most usual (contraceptive) method these days?’ were grouped into four categories (indicating type of method, as opposed to grouping based on effectiveness): unreliable or no usual method; barrier methods; oral/injectable hormonal; and long-acting reversible contraceptives (LARC) (figure 1). Participants were allowed to select up to three usual methods. For the 261 women who indicated use of more than one method, the most effective method reported was coded as their usual method. Figure 1 Contraceptive methods grouped by type. Exposure variables Relationship duration was calculated from the dates of first and most recent occasion of sex with the most recent sexual partner and grouped into mutually exclusive categories: 1 day (where the dates of first and last sex were the same); more than 1 day to 6 months; 6 months to 1 year; 1 to 3 years; 3 to 5 years; and 5 years or more.

ration was calculated from the dates of first and most recent occasion of sex with the most recent sexual partner and grouped into mutually exclusive categories: 1 day (where the dates of first and last sex were the same); more than 1 day to 6 months; 6 months to 1 year; 1 to 3 years; 3 to 5 years; and 5 years or more. Relationship status was derived from the question ‘Which one of these descriptions applies best to you and [that person/name] at the time you most recently had sex?’ Responses were grouped into four mutually exclusive categories: recently met, not steady, steady but non-cohabiting, married/cohabiting. Statistical analyses Initial analyses described the association between contraceptive method group usually used and age group (16–24, 25–34 and 35–49 years), including those with limited recent heterosexual sexual experience and those currently pregnant. From this point, further analyses were restricted to women at risk of unplanned pregnancy (those not answering ‘I would definitely like (more) children and am currently trying’ to a fertility intention question, those pre-menopausal, and those with recent heterosexual activity). A cross-tabulation of contraceptive method type usually used and relationship status and duration showed trends in method use by relationship characteristics in different age groups.

children and am currently trying’ to a fertility intention question, those pre-menopausal, and those with recent heterosexual activity). A cross-tabulation of contraceptive method type usually used and relationship status and duration showed trends in method use by relationship characteristics in different age groups. Use of each category of contraceptive method was indicated in four dichotomous variables whereby usual use of the method of interest was coded ‘1’, else ‘0’. These dichotomous variables were the dependent variables in binary logistic regression models. Initially each binary logistic regression model predicted the crude (univariable) odds of use of the method of interest by relationship status and duration, followed by mutual adjustment for both partnership characteristic variables. Statistical analyses were performed using Stata/SE 15 (StataCorp LP) taking into account the survey’s sample design, using the svy function. Results Contraceptive method use among all women, by age

methods was highest in the youngest age group, decreasing with successive age group. The reverse was true for unreliable and no method use, which increased with age group. Reliance on sterilisation also increased with age group and was reported by almost one in five 35–49-year-olds, but by virtually no 16–24-year-olds. Table 1 ‘Usual’ contraceptive method use among women aged 16–49 years, stratified by age (column percentages) Total Age range (years) 16–24 25–34 35–49 Unweighted Weighted n % (95% CI) n % (95% CI) n % (95% CI) n n % (95% CI) Unreliable method or none 126 7.6 (6.3 to 9.2) 295 12.3 (10.9 to 13.9) 352 19.4 (17.4 to 21.5) 773 655 14.8 (13.6 to 16.0) Barrier methods 357 22.1 (19.9 to 24.4) 400 18.0 (16.2 to 20.0) 234 12.9 (11.4 to 14.6) 991 727 16.4 (15.3 to 17.5) Oral/injectable methods 768 45.5 (42.7 to 48.2) 832 33.2 (31.2 to 35.2) 244 13.2 (11.5 to 15.0) 1844 1154 26.0 (24.8 to 27.3) LARC 220 12.1 (10.4 to 13.9) 287 11.4 (10.0 to 13.0) 193 11.3 (9.8 to 13.1) 700 511 11.5 (10.5 to 12.6) Male or female sterilisation 3 0.2 (0.1 to 0.6) 87 3.6 (2.8 to 4.5) 280 18.5 (16.4 to 20.7) 370 449 10.1 (9.1 to 11.3) Not sexually active† 36 2.5 (1.7 to 3.5) 121 4.2 (3.5 to 5.0) 159 6.8 (5.7 to 8.1) 316 226 5.1 (4.5 to 5.8) Only same-sex experience‡ 14 1.0 (0.6 to 1.6) 24 1.0 (0.6 to 1.5) 23 1.3 (0.8 to 1.9) 61 49 1.1 (0.8 to 1.5) Currently pregnant§ or menopausal 160 9.2 (7.8 to 10.8) 359 16.3 (14.7 to 18.1) 283 16.7 (14.8 to 18.8) 802 666 15.0 (13.9 to 16.2) Total Unweighted 1684 28.8 2405 41.1 1768 30.2 5857 100 Weighted 939 21.2 (20.1 to 22.2) 1331 30.0 (28.8 to 31.3) 2167 48.8 (47.3 to 50.4) 4437 100 Proportions are weighted to account for the survey sampling design and non-response.

.8 to 10.8) 359 16.3 (14.7 to 18.1) 283 16.7 (14.8 to 18.8) 802 666 15.0 (13.9 to 16.2) Total Unweighted 1684 28.8 2405 41.1 1768 30.2 5857 100 Weighted 939 21.2 (20.1 to 22.2) 1331 30.0 (28.8 to 31.3) 2167 48.8 (47.3 to 50.4) 4437 100 Proportions are weighted to account for the survey sampling design and non-response. Total N includes women aged 16–49 years, who reported their usual method of contraception. One woman reported her usual method of contraception as an emergency method and was excluded from these results. †Not sexually active in the last year. ‡Only same-sex sexual experience in the last year. §Includes women currently pregnant or trying to conceive. CI, confidence interval; LARC, long-acting reversible contraception. Contraceptive method use by age and relationship characteristics among women at risk of unintended pregnancy Online supplementary file 3 gives the distribution of women by partnership characteristics. Table 2 shows contraceptive method group use by partnership characteristics by age group (online supplementary file 4 provides the crude odds ratios and online supplementary file 5 gives additional descriptive analyses of condom use by partnership characteristics). 10.1136/bmjsrh-2017-200037.supp3Supplementary file 3 10.1136/bmjsrh-2017-200037.supp4Supplementary file 4 10.1136/bmjsrh-2017-200037.supp5Supplementary file 5 Table 2 Distribution of ‘usual’ contraceptive method use by relationship duration and relationship status, stratified by age

Contraceptive method use by age and relationship characteristics among women at risk of unintended pregnancy Online supplementary file 3 gives the distribution of women by partnership characteristics. Table 2 shows contraceptive method group use by partnership characteristics by age group (online supplementary file 4 provides the crude odds ratios and online supplementary file 5 gives additional descriptive analyses of condom use by partnership characteristics). 10.1136/bmjsrh-2017-200037.supp3Supplementary file 3 10.1136/bmjsrh-2017-200037.supp4Supplementary file 4 10.1136/bmjsrh-2017-200037.supp5Supplementary file 5 Table 2 Distribution of ‘usual’ contraceptive method use by relationship duration and relationship status, stratified by age Unreliable or no method Barrier methods Oral and injectable hormonal methods LARC Total Weighted column % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) Unweighted n Weighted n Total row % Age: 16–24 years

Table 2 Distribution of ‘usual’ contraceptive method use by relationship duration and relationship status, stratified by age Unreliable or no method Barrier methods Oral and injectable hormonal methods LARC Total Weighted column % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) Unweighted n Weighted n Total row % Age: 16–24 years Relationship duration 1 day 16.3 (11.5 to 22.5) 33.6 (27.2 to 40.5) 39.1 (32.5 to 46.2) 11.0 (7.5 to 16.0) 272 155 100 18.8 (16.6 to 21.3) >1 day<6 months 5.2 (3.0 to 8.8) 39.2 (32.3 to 46.6) 46.2 (39.4 to 53.1) 9.4 (6.0 to 14.6) 233 126 100 15.3 (13.4 to 17.5) ≥6 months<1 year 10.6 (6.7 to 16.4) 18.9 (13.6 to 25.8) 58.4 (50.8 to 65.6) 12.1 (7.9 to 18.0) 212 117 100 14.3 (12.4 to 16.4) ≥1 year<3 years 6.4 (4.3 to 9.6) 16.4 (12.7 to 21.0) 60.0 (54.6 to 65.2) 17.1 (13.4 to 21.7) 367 204 100 24.8 (22.5 to 27.3) ≥3 years<5 years 9.2 (5.9 to 13.9) 24.0 (18.3 to 30.8) 54.5 (47.2 to 61.6) 12.3 (8.3 to 17.8) 224 127 100 15.4 (13.4 to 17.6) ≥5 years 8.1 (4.4 to 14.4) 19.3 (13.5 to 26.8) 53.0 (44.7 to 61.1) 19.6 (13.8 to 27.3) 167 93 100 11.4 (9.6 to 13.3) Relationship status Recently met 13.0 (6.1 to 25.5) 44.8 (32.5 to 57.7) 31.0 (20.7 to 43.5) 11.2 (5.5 to 21.5) 70 42 100 5.1 (4.0 to 6.5) Not steady 11.8 (8.5 to 16.3) 32.4 (27.0 to 38.2) 44.4 (38.2 to 50.8) 11.4 (8.0 to 15.9) 332 180 100 21.9 (19.5 to 24.5) Steady* 6.8 (5.0 to 9.1) 22.2 (18.9 to 26.0) 57.7 (53.5 to 61.9) 13.3 (10.7 to 16.4) 748 398 100 48.4 (45.4 to 51.3) Married/cohabiting 11.3 (7.7 to 16.2) 19.8 (15.5 to 25.0) 52.1 (46.3 to 57.9) 16.8 (12.8 to 21.7) 325 203 100 24.7 (22.3 to 27.2) Total n n Unweighted (n) 133 353 771 218 1475 100 100 Weighted (n) 77 (9.3%; 7.8 to 11.1) 206 (25.0%; 22.6 to 27.6) 429 (52.1%; 49.1 to 55.0) 112 (13.6%; 11.7 to 15.8) 823 100 100

4 to 51.3) Married/cohabiting 11.3 (7.7 to 16.2) 19.8 (15.5 to 25.0) 52.1 (46.3 to 57.9) 16.8 (12.8 to 21.7) 325 203 100 24.7 (22.3 to 27.2) Total n n Unweighted (n) 133 353 771 218 1475 100 100 Weighted (n) 77 (9.3%; 7.8 to 11.1) 206 (25.0%; 22.6 to 27.6) 429 (52.1%; 49.1 to 55.0) 112 (13.6%; 11.7 to 15.8) 823 100 100 Age: 25–34 years

4 to 51.3) Married/cohabiting 11.3 (7.7 to 16.2) 19.8 (15.5 to 25.0) 52.1 (46.3 to 57.9) 16.8 (12.8 to 21.7) 325 203 100 24.7 (22.3 to 27.2) Total n n Unweighted (n) 133 353 771 218 1475 100 100 Weighted (n) 77 (9.3%; 7.8 to 11.1) 206 (25.0%; 22.6 to 27.6) 429 (52.1%; 49.1 to 55.0) 112 (13.6%; 11.7 to 15.8) 823 100 100 Age: 25–34 years Relationship duration 1 day 29.9 (23.2 to 37.7) 27.5 (21.1 to 35.5) 29.1 (22.8 to 36.3) 13.2 (8.8 to 19.3) 204 105 100 10.3 (9.0 to 11.9) >1 day<6 months 15.1 (10.1 to 22.0) 28.3 (19.3 to 39.5) 42.8 (33.9 to 52.2) 13.7 (8.8 to 20.7) 165 78 100 7.7 (6.5 to 9.2) ≥6 months<1 year 19.4 (12.0 to 29.9) 22.5 (13.8 to 34.5) 42.2 (32.6 to 52.3) 15.9 (9.9 to 24.5) 121 63 100 6.2 (5.0 to 7.6) ≥1 year<3 years 18.0 (13.0 to 24.4) 23.9 (17.3 to 32.1) 47.1 (39.2 to 55.2) 10.9 (6.9 to 16.8) 206 113 100 11.1 (9.5 to 12.9) ≥3 years<5 years 20.3 (15.4 to 26.3) 17.8 (13.2 to 23.5) 50.0 (43.1 to 57.0) 11.9 (8.2 to 17.1) 253 134 100 13.3 (11.7 to 15.0) ≥5 years 16.9 (14.4 to 19.7) 22.8 (19.8 to 26.0) 43.5 (40.2 to 46.9) 16.8 (14.2 to 19.8) 914 521 100 51.4 (48.6 to 54.1) Relationship status Recently met 17.3 (10.0 to 28.3) 30.4 (19.8 to 43.6) 41.9 (30.4 to 54.4) 10.4 (5.3 to 19.3) 82 42 100 4.1 (3.2 to 5.2) Not steady 22.7 (17.3 to 29.1) 23.6 (17.3 to 31.3) 40.7 (33.9 to 47.9) 13.0 (9.1 to 18.1) 271 128 100 12.6 (10.9 to 14.5) Steady* 18.7 (15.0 to 23.0) 18.7 (14.7 to 23.6) 47.1 (42.1 to 52.3) 15.5 (12.0 to 19.7) 439 196 100 19.3 (17.5 to 21.3) Married/cohabiting 18.2 (15.9 to 20.9) 24.0 (21.2 to 26.9) 42.5 (39.4 to 45.7) 15.3 (13.1 to 17.9) 1071 649 100 64.0 (61.6 to 66.3) Total n n Unweighted (n) 357 395 825 286 1863 100 100 Weighted (n) 191 (18.8%; 17.0 to 20.9) 235 (23.2%; 20.9 to 25.6) 438 (43.2%; 40.7 to 45.7) 151 (14.9%; 13.1 to 16.8) 1014 100 100

1.3) Married/cohabiting 18.2 (15.9 to 20.9) 24.0 (21.2 to 26.9) 42.5 (39.4 to 45.7) 15.3 (13.1 to 17.9) 1071 649 100 64.0 (61.6 to 66.3) Total n n Unweighted (n) 357 395 825 286 1863 100 100 Weighted (n) 191 (18.8%; 17.0 to 20.9) 235 (23.2%; 20.9 to 25.6) 438 (43.2%; 40.7 to 45.7) 151 (14.9%; 13.1 to 16.8) 1014 100 100 Age: 35–49 years

1.3) Married/cohabiting 18.2 (15.9 to 20.9) 24.0 (21.2 to 26.9) 42.5 (39.4 to 45.7) 15.3 (13.1 to 17.9) 1071 649 100 64.0 (61.6 to 66.3) Total n n Unweighted (n) 357 395 825 286 1863 100 100 Weighted (n) 191 (18.8%; 17.0 to 20.9) 235 (23.2%; 20.9 to 25.6) 438 (43.2%; 40.7 to 45.7) 151 (14.9%; 13.1 to 16.8) 1014 100 100 Age: 35–49 years Relationship duration 1 day 37.4 (26.6 to 49.5) 17.2 (9.8 to 28.5) 24.0 (15.7 to 34.7) 21.4 (13.2 to 32.9) 91 93 100 7.2 (5.7 to 9.0) >1 day<6 months 49.5 (35.6 to 63.5) 20.4 (11.3 to 33.9) 11.8 (5.5 to 23.7) 18.3 (9.1 to 33.4) 52 42 100 3.2 (2.4 to 4.3) ≥6 months<1 year 43.5 (30.7 to 57.3) 21.0 (11.9 to 34.5) 20.3 (11.6 to 33.1) 15.1 (7.6 to 27.9) 58 45 100 3.4 (2.6 to 4.6) ≥1 year<3 years 42.7 (31.5 to 54.8) 22.4 (13.6 to 34.7) 17.9 (10.1 to 30.0) 16.9 (9.1 to 29.2) 77 67 100 5.1 (4.1 to 6.5) ≥3 years<5 years 53.8 (41.6 to 65.7) 16.4 (9.7 to 26.3) 17.7 (10.1 to 29.3) 12.1 (6.0 to 22.7) 79 86 100 6.6 (5.2 to 8.3) ≥5 years 35.7 (32.0 to 39.5) 22.3 (19.3 to 25.5) 22.8 (19.6 to 26.2) 19.3 (16.4 to 22.5) 761 974 100 74.5 (71.8 to 77.1) 37.4 (26.6 to 49.5) 17.2 (9.8 to 28.5) 24.0 (15.7 to 34.7) 21.4 (13.2 to 32.9) 91 93 100 7.2 (5.7 to 9.0) Relationship status Recently met 17.0 (7.3 to 34.9) 25.4 (11.8 to 46.3) 15.2 (6.5 to 31.7) 42.3 (24.9 to 61.9) 29 25 100 1.9 (1.3 to 2.9) Not steady 52.5 (43.3 to 61.6) 17.3 (11.9 to 24.4) 16.3 (10.9 to 23.7) 13.9 (8.2 to 22.6) 148 125 100 9.5 (8.0 to 11.3) Steady* 42.8 (35.4 to 50.5) 21.8 (15.7 to 29.5) 19.6 (14.1 to 26.6) 15.8 (10.9 to 22.4) 198 170 100 13.0 (11.1 to 15.2) Married/cohabiting 36.0 (32.3 to 39.7) 21.8 (18.9 to 25.0) 23.1 (19.9 to 26.6) 19.2 (16.5 to 22.3) 743 987 100 75.5 (72.9 to 78.0) Total n n Unweighted (n) 449 231 246 192 1118 100 100 Weighted (n) 497 (38.1%; 34.9 to 41.3) 280 (21.4%; 18.9 to 24.1) 285 (21.8%; 19.2 to 24.7) 244 (18.7%; 16.3 to 21.3) 1307 100 100 Proportions are weighted to account for the survey sampling design and non-response. Total N includes women aged 16–49 years, with at least one previous opposite- sex partner and whose most recent sexual experience was with an opposite- sex partner, who are not currently trying to conceive or are pregnant. A total of 275 women indicated more than one usual method of contraception: 232 women selected an oral and injectable method or LARC in addition to a barrier method, and the more effective method was selected as their usual method.

with an opposite- sex partner, who are not currently trying to conceive or are pregnant. A total of 275 women indicated more than one usual method of contraception: 232 women selected an oral and injectable method or LARC in addition to a barrier method, and the more effective method was selected as their usual method. *Relationship status steady, but not cohabiting. LARC, long-acting reversible contraception. 16–24-year-olds Unreliable and no method(s) Less than 10% of 16–24-year-olds at risk of unintended pregnancy were using an unreliable or no contraceptive method. This proportion did not vary consistently by relationship status but decreased significantly with duration. After adjustment for relationship status and duration, the association between use of an unreliable method or none and relationship duration remained significant, but the association with relationship status was weakened (figure 2A and online supplementary file 6). 10.1136/bmjsrh-2017-200037.supp6Supplementary file 6 Figure 2 Forest plots of adjusted odds ratios for use of each group of contraceptive methods, by relationship duration and relationship status. A: 16–24 years; B: 25–34 years; C: 35–49 years. Barrier methods Barrier method use was reported by one in four women at risk of unintended pregnancy, but decreased significantly as relationships increased in length and became more established. After adjustment, the odds of barrier method use were only significantly lower for women in relationships lasting 1 to 3 years' duration, and among women in steady or married/cohabiting relationships.

of unintended pregnancy, but decreased significantly as relationships increased in length and became more established. After adjustment, the odds of barrier method use were only significantly lower for women in relationships lasting 1 to 3 years' duration, and among women in steady or married/cohabiting relationships. Oral and injectable methods Hormonal method use accounted for over half of usual use by women at risk of unintended pregnancy. Use was highest among women in relationships of 1 to 3 years' duration and those in ‘steady’ relationships. Adjusted odds showed significant increased use of oral/injectable methods in relationships lasting between 6 months and 3 years, as well as among steady, non-cohabiting couples. LARC The proportion of women using LARC did not vary with relationship duration nor status in this age group, and this was confirmed in the adjusted logistic regression results. 25–34-year-olds Unreliable and no method(s) Almost one-fifth of women in this age group used an unreliable or no method. This proportion increased with relationship duration but not status. Adjusted results showed a significant decrease in use of unreliable methods or none for all relationships lasting longer than 1 day (with the exception of relationships of 6 months to 1 year) (figure 2B).

his age group used an unreliable or no method. This proportion increased with relationship duration but not status. Adjusted results showed a significant decrease in use of unreliable methods or none for all relationships lasting longer than 1 day (with the exception of relationships of 6 months to 1 year) (figure 2B). Barrier methods Almost one-quarter of women at risk of unintended pregnancy usually used a barrier method. This proportion did not vary significantly by relationship characteristics. After adjustment, only women whose most recent relationship lasted between 3 and 5 years were less likely to report use of barrier methods. Oral and injectable methods The proportion of women reporting oral/injectable method use increased with relationship duration but did not vary significantly by status. After adjustment, use of these methods significantly increased with relationship duration, but notably, decreased with relationship stability. Women whose most recent relationship had lasted only 1 day were less likely than other women to use hormonal methods, but women who said they had recently met their partner were more likely than married/cohabiting women to have used such methods. LARC There were no significant associations between use of LARC and relationship duration or status.

Oral and injectable methods The proportion of women reporting oral/injectable method use increased with relationship duration but did not vary significantly by status. After adjustment, use of these methods significantly increased with relationship duration, but notably, decreased with relationship stability. Women whose most recent relationship had lasted only 1 day were less likely than other women to use hormonal methods, but women who said they had recently met their partner were more likely than married/cohabiting women to have used such methods. LARC There were no significant associations between use of LARC and relationship duration or status. 35–49-year-olds Unreliable and no method(s) Some 38.1% of women in this age group reported usual use of an unreliable or no method. This proportion varied by relationship status but not duration. Adjusted odds ratios showed women who had recently met their partner were less likely to use an unreliable or no method (figure 2C). Barrier methods Use of barrier methods varied with neither relationship status nor duration. Oral and injectable methods The adjusted odds of using an oral or injectable method were not significantly associated with relationship status nor duration. LARC Use of LARC varied significantly by relationship status, but not duration. After adjustment, women who had recently met their partner were significantly more likely to use LARC than women in other types of relationship.

Oral and injectable methods The adjusted odds of using an oral or injectable method were not significantly associated with relationship status nor duration. LARC Use of LARC varied significantly by relationship status, but not duration. After adjustment, women who had recently met their partner were significantly more likely to use LARC than women in other types of relationship. Discussion Principal findings These data from a nationally-representative survey of sexually-active women in Britain illustrate differences in method use by the nature and duration of the relationship with the most recent sexual partner and how the extent of these differences varies with age. Use of barrier methods was appreciably higher in short-term than longer-term relationships among younger participants, but this pattern was not seen among respondents in the oldest age group. While there was considerable correspondence between the duration and type of relationships and their association with contraceptive method use, some differences were observed. Relationship duration was marginally more strongly associated with the contraceptive method used than was relationship type, and this pattern was more obvious among younger participants. Moreover, relationship type and duration operated differentially in their influence on hormonal method use among 25–34-year-old women.

elationship duration was marginally more strongly associated with the contraceptive method used than was relationship type, and this pattern was more obvious among younger participants. Moreover, relationship type and duration operated differentially in their influence on hormonal method use among 25–34-year-old women. Comparison of our findings to those of other studies is difficult because of differences in methodology, in the definition of relationship types and in the categorisation of contraceptive methods. Typically, studies that have compared contraceptive method use in different age groups,4 24 have not taken relationship characteristics into account and, conversely, studies that compared contraceptive method use by relationship characteristics have not done so in different age groups.6–10 13 14 16 25 These differences in study design are likely to partially explain the equivocal nature of findings of many previous studies14 and also apparent anomalies between other studies and our own. At the aggregate level, we did not find higher levels of LARC use among younger participants than older ones, as others have.26 27 Nevertheless, their use increased markedly among 16–24-year-old women whose relationship had lasted 5 or more years; equivalent to the prevalence among 35–49-year-olds.

studies and our own. At the aggregate level, we did not find higher levels of LARC use among younger participants than older ones, as others have.26 27 Nevertheless, their use increased markedly among 16–24-year-old women whose relationship had lasted 5 or more years; equivalent to the prevalence among 35–49-year-olds. Study strengths and weaknesses Our large population-based dataset enabled us to estimate the prevalence of contraceptive use in different age groups by relationship type and duration, and to exclude women not apparently at risk of unplanned pregnancy. The study also has limitations. For multi-partnered participants, cross-sectional data are unsuited to tracing pathways of use and consequently we have made several assumptions in this article. First, we have assumed that respondents used their ‘usual’ method of contraception with their most recent sexual partner. When respondents reported several usual methods, the most effective may not have been the method used for the most recent sex. Second, the most recent sexual partner may not have been the main sexual partner during the previous year. Relationship duration was calculated objectively as the time between first and most recent sex with the respondent’s most recent sexual partner. This measure might not truly reflect relationship duration as it would be defined by the participant. We restricted our analyses to women, as misreporting of usual contraceptive method may be more common by men, since the majority of the methods are operationalised by women. We were limited to use of data generated by the questions asked in Natsal-3, which did not probe affective attributes of the relationship which others have linked to contraceptive method use.5 10 Finally, while we have no reason to believe contraceptive practices might have changed in the time since interview, these analyses are based on data reported by participants up to 8 years ago.

atsal-3, which did not probe affective attributes of the relationship which others have linked to contraceptive method use.5 10 Finally, while we have no reason to believe contraceptive practices might have changed in the time since interview, these analyses are based on data reported by participants up to 8 years ago. Implications for policy and practice Our data should be interpreted in the context of women’s lives, and their evolving priorities and preoccupations through the life course. Younger women, and those in more transient relationships, may be more motivated to avoid pregnancy than those who are older and in more stable relationships. Thus, the fact that one in six 16–24-year-olds uses an unreliable/no method of contraception with a partner they have just met is of concern. So, also, is the relatively low prevalence of reliable method use in shorter-term relationships among older participants. Alternatively, it is likely that although not explicitly trying to conceive, for some women a pregnancy would not be unwelcome, resulting in some ambivalence in relation to contraceptive use.28 An acknowledgement of pregnancy intention as a continuum is important to further understand women’s contraceptive needs. To date, sexual health policy remains focused on younger adults, but evidence that the majority of unplanned pregnancies occur among women aged 25 years and over,1 has prompted pleas for age-specific health promotion strategies.29 30

tention as a continuum is important to further understand women’s contraceptive needs. To date, sexual health policy remains focused on younger adults, but evidence that the majority of unplanned pregnancies occur among women aged 25 years and over,1 has prompted pleas for age-specific health promotion strategies.29 30 The independent associations of relationship type and duration with method use, and the suggestion that duration might be the stronger influence, has implications for policy and practice. Conventionally in clinical record-taking, data are collected on relationship type. Our findings suggest that there are benefits to be gained from asking about relationship duration in helping support women in their use of contraception, to help them consider their priorities and preferences at different life stages. Our findings highlight the complex nature of associations between age, partnership characteristics and method use, and signify that every woman needs to be assessed individually. Future research Future research should recognise that the associations between relationship characteristics and contraceptive method use are more complex than might be suggested from aggregate data often used to assess the need for health promotional messages and public health interventions. Studies need to take account of both age and the nature of the relationships in which women use contraception. 10.1136/bmjsrh-2017-200037.supp1Supplementary file 1 10.1136/bmjsrh-2017-200037.supp2Supplementary file 2

Future research Future research should recognise that the associations between relationship characteristics and contraceptive method use are more complex than might be suggested from aggregate data often used to assess the need for health promotional messages and public health interventions. Studies need to take account of both age and the nature of the relationships in which women use contraception. 10.1136/bmjsrh-2017-200037.supp1Supplementary file 1 10.1136/bmjsrh-2017-200037.supp2Supplementary file 2 Natsal-3 was a collaboration between University College London (London, UK), the London School of Hygiene and Tropical Medicine (London, UK), NatCen Social Research, Public Health England (formerly the Health Protection Agency), and the University of Manchester (Manchester, UK). The authors thank the study participants; the team of interviewers from NatCen Social Research; Heather Wardle, Vicki Hawkins, Cathy Coshall, and operations and computing staff from NatCen Social Research; and Professor Carol Dezateux at UCL GOS ICH for writing assistance. Contributors: NF analysed and interpreted the data. NF, MJP, IMT and KW were major contributors in writing the manuscript. All authors read and approved the final manuscript.

Natsal-3 was a collaboration between University College London (London, UK), the London School of Hygiene and Tropical Medicine (London, UK), NatCen Social Research, Public Health England (formerly the Health Protection Agency), and the University of Manchester (Manchester, UK). The authors thank the study participants; the team of interviewers from NatCen Social Research; Heather Wardle, Vicki Hawkins, Cathy Coshall, and operations and computing staff from NatCen Social Research; and Professor Carol Dezateux at UCL GOS ICH for writing assistance. Contributors: NF analysed and interpreted the data. NF, MJP, IMT and KW were major contributors in writing the manuscript. All authors read and approved the final manuscript. Funding: Natsal-3 was supported by grants from the Medical Research Council (G0701757) and the Wellcome Trust (084840), with contributions from the Economic and Social Research Council and Department of Health. NF was supported by a UK/EU Fees Studentship Award for her MSc in Demography & Health at LSHTM. All research at Great Ormond Street Hospital NHS Foundation Trust and UCL Great Ormond Street Institute of Child Health is made possible by the NIHR Great Ormond Street Hospital Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: None declared. Patient consent: Not required. Ethics approval: The Natsal-3 study was approved by the Oxfordshire Research Ethics Committee A (reference: 09/H0604/27). Participants provided oral informed consent for interviews.

Funding: Natsal-3 was supported by grants from the Medical Research Council (G0701757) and the Wellcome Trust (084840), with contributions from the Economic and Social Research Council and Department of Health. NF was supported by a UK/EU Fees Studentship Award for her MSc in Demography & Health at LSHTM. All research at Great Ormond Street Hospital NHS Foundation Trust and UCL Great Ormond Street Institute of Child Health is made possible by the NIHR Great Ormond Street Hospital Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: None declared. Patient consent: Not required. Ethics approval: The Natsal-3 study was approved by the Oxfordshire Research Ethics Committee A (reference: 09/H0604/27). Participants provided oral informed consent for interviews. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: The data that support the findings of this study are available from the Natsal research team but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the Natsal research team.

Key messages Formative evaluation of online information to support abortion access in the Republic of Ireland, Northern Ireland, and England shows ‘useful’ information is extremely limited. Information retrieved by users is not always accurate within the jurisdiction where the search took place. Even ‘good’ web pages are flawed. The user needs to be central to the design of web pages. Introduction A recognition of three issues informs this paper: (1) the use of the internet as a source of information about accessing services; (2) the impact of jurisdiction on how services are accessed; (3) the lack of guidance on producing useful internet-based information on how to access abortion services. The internet is a key source of information related to stigmatised conditions or treatments.1 However, web-based access information has been queried and concerns raised— globally and in the UK.2–4 In the context of abortion care, usefulness for access is a central question in these discussions. Research in abortion care has highlighted that abortion care seekers using internet searches have difficulties finding information useful for accessing services.5–7 WHO argues that poor information is a significant barrier to abortion care and emphasises the need to produce and disseminate clear, accurate information about how and where to access abortion services.8 This is rooted in a concern that poor information can delay care, leading to patients accessing abortion at a later stage and at greater risk.

mation is a significant barrier to abortion care and emphasises the need to produce and disseminate clear, accurate information about how and where to access abortion services.8 This is rooted in a concern that poor information can delay care, leading to patients accessing abortion at a later stage and at greater risk. Abortion providers have tried to address these concerns through investing in website design and how service information is presented. However, at the time of writing there is no framework for or guidance on what providers should consider when trying to design web pages that facilitate access effectively. In the absence of such guidance, designers need to rely on existing quality assessment mechanisms such as WebMedQual9 or DISCERN.10 This reliance is problematic for two reasons. First, these frameworks are quality assessment tools. Their concerns are predominantly issues of scientific accuracy, authorship and rigour. Second, where ‘usefulness for access’ is discussed in these tools, they do not appreciate the peculiarities of access to abortion care. Of particular concern to abortion care access is the accuracy of information within or relevance of information to the location where information searches take place.8 Pathways to accessing abortion care are frequently contingent on specific user-related factors (for instance, whether a pregnancy presents a risk to health or whether the pregnancy is the result of rape or incest). Furthermore, pathways can also vary significantly between jurisdictions. In the UK, for example, the policy and legal context of users in Northern Ireland is significantly different to users in England. Whether an information source facilitates access to care is very much contingent on the jurisdiction in which the user is based.

, pathways can also vary significantly between jurisdictions. In the UK, for example, the policy and legal context of users in Northern Ireland is significantly different to users in England. Whether an information source facilitates access to care is very much contingent on the jurisdiction in which the user is based. This article highlights that, despite significant investment in designing technically accurate and clearly expressed information resources online, those seeking access to abortion services still face difficulties in accessing relevant, useful information. It illustrates this through the application of a specially-designed formative evaluative instrument to information retrieved through searches in three jurisdictions: the Republic of Ireland, Northern Ireland, and England. In each of these areas there are specific controls on how, when and where an abortion can be accessed and penalties for procuring or providing an abortion outside of these access pathways. Given these differences, information on how to access abortion services may be useful in one jurisdiction but not another. The findings presented in this article provide insight into the challenges faced by abortion services information seekers relying on the internet and draw attention to areas that abortion and health services should be aware of when producing useful web-based information.

y be useful in one jurisdiction but not another. The findings presented in this article provide insight into the challenges faced by abortion services information seekers relying on the internet and draw attention to areas that abortion and health services should be aware of when producing useful web-based information. Since the research was conducted (in 2016), there have been significant changes in how abortion can be accessed in each country included in the research. In the Republic of Ireland the constitutional ban on abortion was removed by referendum on 26 May 2018. There has also been a campaign to decriminalise abortion in England and Northern Ireland. Women living in Northern Ireland since 2017 no longer have to pay for abortion services in England as private patients. That said, legislation for the procurement and provision of abortion after these changes is yet to be enacted in any of the countries and at the time of publication women still continue to have restricted access to abortion. Furthermore, it is not yet clear what the effect, if any, of these changes will have on abortion information on the internet. The fact that abortion access is currently a core debate in two of the countries included in the research adds a particular timeliness to the paper.

ntinue to have restricted access to abortion. Furthermore, it is not yet clear what the effect, if any, of these changes will have on abortion information on the internet. The fact that abortion access is currently a core debate in two of the countries included in the research adds a particular timeliness to the paper. Methodology Objectives The research aimed to interrogate the usefulness of web-based information on accessing abortion services retrieved through internet searches in different jurisdictions from the perspective of service users. This interrogation would provide a formative evaluative mechanism for enabling user-focused design of abortion access information web pages.

rogate the usefulness of web-based information on accessing abortion services retrieved through internet searches in different jurisdictions from the perspective of service users. This interrogation would provide a formative evaluative mechanism for enabling user-focused design of abortion access information web pages. Research design The research team evaluated information relevant to accessing abortion services and abortion care on the internet. This includes clinic-based care and self-administration/telemedicine services. (At the time of the research, self-administration/telemedicine services were only available in the two Irish jurisdictions. These services were provided through an activist organisation registered in the Netherlands offering misoprostol to those living in jurisdictions where abortion access is heavily restricted.) The research adopted a formative evaluative approach.11 Formative evaluation uses appraisal to identify areas requiring further development within objects of evaluation. In the context of questions of service ‘utility’, its objective is not to provide definitive assessments of usefulness but to enable the effective implementation and design of useful services. As such, our research objectives were to (1) assess the merits of existing sources from a user-centred perspective, and (2) suggest how the utility of online information could be improved.

ve is not to provide definitive assessments of usefulness but to enable the effective implementation and design of useful services. As such, our research objectives were to (1) assess the merits of existing sources from a user-centred perspective, and (2) suggest how the utility of online information could be improved. Formative evaluation is useful not just in the design of services but also as a means of opening up debates on approaches and practices within services. In the context of abortion service information, it draws attention to the problems of placing information on different pages within the website. This is arguably a strategy of organisations in jurisdictions where abortion is criminalised. At the same time, given that research suggests information seekers do not always look beyond the first page of websites, such strategic placement can be problematic. Through formative evaluation, organisations can see the limitations of their approaches to website design and begin to consider how they can facilitate access as much as possible.

hat research suggests information seekers do not always look beyond the first page of websites, such strategic placement can be problematic. Through formative evaluation, organisations can see the limitations of their approaches to website design and begin to consider how they can facilitate access as much as possible. Data collection The researchers sought to replicate a search pattern similar to that of an individual wishing to access abortion services (therefore an individual who has made the decision to access abortion) using an established methodological approach.12 13 Data collection reflected our user-centred approach. We assessed individual web pages retrieved rather than the complete website, an approach resonant with evidence14 that health information seekers commonly decide to use a particular service or source based on the first web page accessed.

methodological approach.12 13 Data collection reflected our user-centred approach. We assessed individual web pages retrieved rather than the complete website, an approach resonant with evidence14 that health information seekers commonly decide to use a particular service or source based on the first web page accessed. The three countries—England, Northern Ireland, and the Republic of Ireland—were selected for three reasons. First, when the research was conducted, there was specific legislation regarding the procurement of abortion services in each jurisdiction (this was highly restrictive in Northern Ireland and the Republic of Ireland). Despite ongoing changes in abortion law and policy at the time of publication, it is likely that access to abortion will be contingent on user-related factors with penalties for methods of access outside established pathways. Second, both jurisdictions on the island of Ireland have historic problems with both the dissemination of accurate abortion information and abortion stigma. Third, while England has historically a more liberal approach to abortion access than Irish jurisdictions, abortion access is closely regulated there and the types of abortion service (and user) are very similar. The fact that it is not entirely liberal in its approach to abortion provision and access makes it a good comparator for the Irish jurisdictions.

a more liberal approach to abortion access than Irish jurisdictions, abortion access is closely regulated there and the types of abortion service (and user) are very similar. The fact that it is not entirely liberal in its approach to abortion provision and access makes it a good comparator for the Irish jurisdictions. We used two popular search engines: www.google.co.uk and www.bing.co.uk (.ie in the Republic of Ireland). Searches focused on the relevance to the search query. This meant that web pages retrieved were deemed (by the search engines) to relate to accessing abortion services. These web pages were not always the first page of the individual website. Searches were carried out by the researchers in Northern Ireland (27 July 2016), the Republic of Ireland (24 August 2016) and England (15 September 2016). We used results from the first two pages of search hits as previous studies2 14 have found individuals are unlikely to search past even the first page of results.

e. Searches were carried out by the researchers in Northern Ireland (27 July 2016), the Republic of Ireland (24 August 2016) and England (15 September 2016). We used results from the first two pages of search hits as previous studies2 14 have found individuals are unlikely to search past even the first page of results. The researchers spoke to professionals in three organisations—Brook Advisory Clinic Liverpool, the Family Planning Association Belfast, and the Irish Family Planning Association Dublin—to identify appropriate search terms. We chose these organisations based on their extensive knowledge and experience in working with women seeking abortion. We provided an extensive list of possible searches to the organisations and asked each to choose the top three based on their own experience or to suggest alternatives. All organisations agreed that the term ‘termination’ would not be used by an abortion seeker, nor would terms such as ‘crisis pregnancy’. All were in agreement that the following four searches would be the most frequently heard in their professional experience:How to get an abortion Where can I get an abortion? Getting an abortion I need an abortion Sample and inclusion Grounds for inclusion in the study were that identified web pages offered advice or information on accessing abortion services. We applied one screening question: “Is the information about accessing services and/or service-user support?”. Web pages that recorded ‘yes’ were then evaluated. News or commercial web pages and campaigning sites which were not intended to assist access to services were excluded.

rmation on accessing abortion services. We applied one screening question: “Is the information about accessing services and/or service-user support?”. Web pages that recorded ‘yes’ were then evaluated. News or commercial web pages and campaigning sites which were not intended to assist access to services were excluded. Measurement tool development In the absence of an existing user-focused evaluative framework for abortion information online, the researchers designed and applied a user-focused five-item tool—the Abortion Service Information Assessment Tool (ASIAT). ASIAT was designed with the user-perspective in mind. Our assessment criteria are resonant with the arguments of Zhang,15 who proposed that frameworks for assessing online information should not just include source indicators—such as those identified by WebMedQual9 and DISCERN10—but also user- and situation-related indicators. This includes the location of the user, the level of user health literacy, age, gender, and requirements (ie, is the search for access to services or for support or information about a specific condition?).

indicators—such as those identified by WebMedQual9 and DISCERN10—but also user- and situation-related indicators. This includes the location of the user, the level of user health literacy, age, gender, and requirements (ie, is the search for access to services or for support or information about a specific condition?). ASIAT combined two issues: (1) whether information seekers would be inclined to use the web page; and (2) whether it would be useful to them. In relation to the first issues, criteria considered in the appraisal included the use of jargon, presence of links to online and offline services, references to existing research, and indications that the web page is monitored by and compliant with professional standards. These criteria reflect the fact that the inclination of information-seekers to use particular online health information is based on heuristics (the design and ‘feel’ of a web page when first encountered) as well as authorship, evidence or the presence of markers of quality compliance.9 14 In relation to the second issue, ASIAT assessed jurisdictional relevancy, reflecting the fact that how abortion services are accessed differs across jurisdictions. Descriptions of pathways to services may be useful to information seekers in one jurisdiction but not another.

resence of markers of quality compliance.9 14 In relation to the second issue, ASIAT assessed jurisdictional relevancy, reflecting the fact that how abortion services are accessed differs across jurisdictions. Descriptions of pathways to services may be useful to information seekers in one jurisdiction but not another. ASIAT ASIAT uses a five-point Likert scale with scores ranging from 5 (lowest) to 25 (highest). Scoring bands were as follows, 5–12 (poor), 13–19 (average) and 20–25 (good). Two researchers conducted preliminary searches and assessments of five web pages independently. They then discussed their assessments and used these to refine the tool. Having refined the tool and updated the scores for the first five results, they then assessed the remaining websites. The final questions used to assess information quality and the criteria informing scoring are outlined in table 1. Table 1 Abortion service information appraisal tool questions and criteria Question Criteria Value range (1–5) Is the language used accessible? Jargon and technical terms, clarity of expression Minimum=Jargon heavy, unclear Maximum=No jargon, clearly expressed Are links provided to regulated online and offline abortion services? Obviousness of links, frequency of links Minimum=No links Maximum=Links provided clearly and obviously and multiple times Are links provided to regulated offline care providers including support groups and information providers? Obviousness of links, frequency of links Minimum=No links Maximum=Links provided clearly and obviously and multiple times

Are links provided to regulated online and offline abortion services? Obviousness of links, frequency of links Minimum=No links Maximum=Links provided clearly and obviously and multiple times Are links provided to regulated offline care providers including support groups and information providers? Obviousness of links, frequency of links Minimum=No links Maximum=Links provided clearly and obviously and multiple times Is information relevant to the jurisdiction presented clearly? References to jurisdictional difference, need for prior knowledge Minimum=No reference to or recognition of jurisdictional difference Maximum=Clear and obvious references to jurisdictional difference; no prior knowledge required Is information quality assured? References to/citations from current research, signs that web page is subject to professional standards and compliance Minimum=No references, citations or signs of professional compliance Maximum=Multiple references to range of evidence/research; indicators that web page is regulated by and compliant with professional standards Statistical analysis Non-parametric tests were used when investigating differences using ordinal scaled data and differences are presented using the median and IQR. All data were analysed using the Statistical Package for Social Sciences (SPSS) version 22.0 (SPSS Inc, Chicago, USA). Means (95% CI) and SD are given where data are normally distributed. Patient and public involvement No patients were involved in this study.

Is information quality assured? References to/citations from current research, signs that web page is subject to professional standards and compliance Minimum=No references, citations or signs of professional compliance Maximum=Multiple references to range of evidence/research; indicators that web page is regulated by and compliant with professional standards Statistical analysis Non-parametric tests were used when investigating differences using ordinal scaled data and differences are presented using the median and IQR. All data were analysed using the Statistical Package for Social Sciences (SPSS) version 22.0 (SPSS Inc, Chicago, USA). Means (95% CI) and SD are given where data are normally distributed. Patient and public involvement No patients were involved in this study. Results Overall, 619 web pages were reviewed with 83 (13.4%) meeting the criteria for phase 2 screening. Phase 2 web pages included: 22 (26.5%) from the Republic of Ireland; 31 (37.3%) from Northern Ireland; and 30 (36.1%) from England. Google retrieved 73 (87.9%) of the 83 web pages that met initial screening criteria, whereas Bing retrieved 46 (55.4%).

reviewed with 83 (13.4%) meeting the criteria for phase 2 screening. Phase 2 web pages included: 22 (26.5%) from the Republic of Ireland; 31 (37.3%) from Northern Ireland; and 30 (36.1%) from England. Google retrieved 73 (87.9%) of the 83 web pages that met initial screening criteria, whereas Bing retrieved 46 (55.4%). We identified seven categories of ownership: abortion provider (including abortion pill providers); statutory health provider; campaigning or activist organisation; non-statutory health provider (including sexual and reproductive health information centres); interactive information forums; personal blogs or magazines; and other. The vast majority of web pages (n=52) were authored by abortion and/or health organisations, followed by magazines (n=12) and interactive information forums (n=11). Six pages were owned by activist or campaigning organisations (see online supplementary table 1). 10.1136/bmjsrh-2018-200086.supp1Supplementary file 1 Of the indicators identified in other measures—language, links with regulated services, content, and authorship—quantitative data revealed a somewhat mixed picture (see online supplementary tables 2–4). While the majority of web pages retrieved used accessible and jargon-free language (n=69), a significant portion (n=23) had no links with regulated services and the content of 32 web pages was either not authored by specialists, unreferenced, or lacked any sign of compliance with professional standards.

ry tables 2–4). While the majority of web pages retrieved used accessible and jargon-free language (n=69), a significant portion (n=23) had no links with regulated services and the content of 32 web pages was either not authored by specialists, unreferenced, or lacked any sign of compliance with professional standards. The relevancy of information to the jurisdiction where the search took place was problematic in all three jurisdictions (see table 2). Northern Ireland searches returned the highest number of web pages with clearly irrelevant information (n=15/31, 48.4%); this was followed by England (n=14/30, 46.7%) and the Republic of Ireland (n=10/22, 45.5%). Table 2 Information relevant to the jurisdiction

The relevancy of information to the jurisdiction where the search took place was problematic in all three jurisdictions (see table 2). Northern Ireland searches returned the highest number of web pages with clearly irrelevant information (n=15/31, 48.4%); this was followed by England (n=14/30, 46.7%) and the Republic of Ireland (n=10/22, 45.5%). Table 2 Information relevant to the jurisdiction Is information relevant to the jurisdiction presented clearly? Number of pages (%) No 39 (47.0) Yes, but difficult to find and requires prior knowledge 7 (8.4) Yes, but requires prior knowledge 5 (6.0) Infrequently 3 (3.6) Multiple times, requires no prior knowledge 29 (34.9) In terms of performance against the ASIAT tool, of the 83 web pages assessed, 22 (26.5%) were considered poor, 35 (42.1%) as average, and 26 (31.3%) as good; 6.9% were classified as excellent. Overall web pages rated as good or excellent in terms of usefulness accounted for 4.2% of the entire sample retrieved (n=619) from searches. Mean ASIAT score across all 83 web pages were 16.33 (SD 4.5). No significant differences were found between ASIAT score and jurisdiction (χ2=1.573, P=0.455), although web pages retrieved within the Republic of Ireland had the highest overall score (17.3, SD 4.5) with Northern Ireland web pages returning the lowest (15.7, SD 4.5) (see table 3). Table 3 Mean ASIAT score by jurisdiction Jurisdiction ASIAT score England 16.4 (4.9*) Republic of Ireland 17.3 (4.5*) Northern Ireland 15.7 (4.5*) Total 16.33 (4.5*) *SD. ASIAT, Abortion Service Information Assessment Tool.

Is information relevant to the jurisdiction presented clearly? Number of pages (%) No 39 (47.0) Yes, but difficult to find and requires prior knowledge 7 (8.4) Yes, but requires prior knowledge 5 (6.0) Infrequently 3 (3.6) Multiple times, requires no prior knowledge 29 (34.9) In terms of performance against the ASIAT tool, of the 83 web pages assessed, 22 (26.5%) were considered poor, 35 (42.1%) as average, and 26 (31.3%) as good; 6.9% were classified as excellent. Overall web pages rated as good or excellent in terms of usefulness accounted for 4.2% of the entire sample retrieved (n=619) from searches. Mean ASIAT score across all 83 web pages were 16.33 (SD 4.5). No significant differences were found between ASIAT score and jurisdiction (χ2=1.573, P=0.455), although web pages retrieved within the Republic of Ireland had the highest overall score (17.3, SD 4.5) with Northern Ireland web pages returning the lowest (15.7, SD 4.5) (see table 3). Table 3 Mean ASIAT score by jurisdiction Jurisdiction ASIAT score England 16.4 (4.9*) Republic of Ireland 17.3 (4.5*) Northern Ireland 15.7 (4.5*) Total 16.33 (4.5*) *SD. ASIAT, Abortion Service Information Assessment Tool. A Kruskal-Wallis test was carried out to examine the difference in ASIAT scores between the seven categories of web page (table 4). This revealed a statistically significant difference between the quality of websites among the seven categories of ownership (χ2=25.975, P=0.01). Statutory health provider web pages had the highest median score (19.5, IQR 9.0), with interactive information forums returning the lowest median score (11.0, IQR 4.0).

e 4). This revealed a statistically significant difference between the quality of websites among the seven categories of ownership (χ2=25.975, P=0.01). Statutory health provider web pages had the highest median score (19.5, IQR 9.0), with interactive information forums returning the lowest median score (11.0, IQR 4.0). Table 4 ASIAT scores by resource type Type of resource ASIAT score* Abortion provider 17 (7.0) Statutory health provider 19.5 (9.0) Campaigning or activist organisation 17 (3.0) Non-statutory health provider 18 (8.0) Interactive information forums 11 (4.0) Personal blogs or magazines 14.5 (5.0) Other 18 (4.0) *Median and IQRs presented. ASIAT, Abortion Service Information Assessment Tool. Five web pages received the maximum ASIAT score of 25; these included Brook, Positive Options, NHS and NHS direct. Web pages with the lowest ASIAT scores were Quora (7), WikiHow (9), Jezebel, Netmums and Cosmopolitan (10).

Type of resource ASIAT score* Abortion provider 17 (7.0) Statutory health provider 19.5 (9.0) Campaigning or activist organisation 17 (3.0) Non-statutory health provider 18 (8.0) Interactive information forums 11 (4.0) Personal blogs or magazines 14.5 (5.0) Other 18 (4.0) *Median and IQRs presented. ASIAT, Abortion Service Information Assessment Tool. Five web pages received the maximum ASIAT score of 25; these included Brook, Positive Options, NHS and NHS direct. Web pages with the lowest ASIAT scores were Quora (7), WikiHow (9), Jezebel, Netmums and Cosmopolitan (10). As an example of an excellent web page, the Brook website (www.brook.org.uk) scored highly in all categories. Information was written in an accessible language, was provided by reproductive health specialists, multiple links were provided to offline services, information relevant to the jurisdiction was clear and required no prior knowledge, and the site was quality assured with references to standardisation bodies. Interestingly, the NHS website scored lower as it included information irrelevant to either of the Irish legal jurisdictions, limiting its usefulness to users based there. A poor website, for example, Quora, achieved low scores as information was not provided by professionals, it was not relevant to the jurisdiction, no sources or citations were included, and there were no links to regulated offline services.

of the Irish legal jurisdictions, limiting its usefulness to users based there. A poor website, for example, Quora, achieved low scores as information was not provided by professionals, it was not relevant to the jurisdiction, no sources or citations were included, and there were no links to regulated offline services. Discussion This research offers insight into what happens when user-based indicators are included in evaluations of abortion service information online. Our research demonstrates that when the relevance and/or accuracy of information about accessing services in the jurisdiction where the search took place is included, the number of useful sources available to information seekers declines. Notably there is a limited difference between jurisdictions with different abortion law and policy, suggesting that a lack of sensitivity to the location of the user is a general problem rather than one restricted to specific jurisdictions. Furthermore, echoing Kunst et al’s4 observations, the inclusion of jurisdiction as an evaluative criterion indicates that even web pages which have a highly specialist authorship may not be useful to users.

ensitivity to the location of the user is a general problem rather than one restricted to specific jurisdictions. Furthermore, echoing Kunst et al’s4 observations, the inclusion of jurisdiction as an evaluative criterion indicates that even web pages which have a highly specialist authorship may not be useful to users. More broadly, ASIAT highlights the challenges facing those seeking information on accessing abortion services. While there is substantial information available through web sources, not all of this information meets the needs of or is useful to abortion seekers. This is a substantial problem in the context of abortion. The limited availability of useful web-based information on services presents a further challenge for those8 16 aiming to increase the safety of abortion. Without having accurate information on how to access care, users may delay treatment and have abortions at later stages at greater risk.

in the context of abortion. The limited availability of useful web-based information on services presents a further challenge for those8 16 aiming to increase the safety of abortion. Without having accurate information on how to access care, users may delay treatment and have abortions at later stages at greater risk. There are limitations to this study. ASIAT has not been subjected to peer review in its own right and, as such, the findings can only be interpreted as indicating areas for future investigation. However, the majority of items in ASIAT are informed by questions used in other instruments. Additionally, as a formative evaluative mechanism, ASIAT is intended to highlight areas for development rather than provide conclusive results of quality. A further potential limitation is the fact that the evaluation focuses on the first web page. The usefulness of information on the website as a whole is not included in the assessment. That said, this approach is based on existing research suggesting that internet information seekers frequently only read the first page of websites. Though ASIAT is not comprehensive, it does mimic the actions of internet users. It is also worth noting that the tool was developed in jurisdictions where the right to information is not limited by law (although in the Republic of Ireland it is closely regulated). This needs to be appreciated when adapting the tool for other jurisdictions where the laws around how and when information about service access is provided are more stringent.

s developed in jurisdictions where the right to information is not limited by law (although in the Republic of Ireland it is closely regulated). This needs to be appreciated when adapting the tool for other jurisdictions where the laws around how and when information about service access is provided are more stringent. Conclusion Regardless of its limitations, this research shows that designers of web-based information need to appreciate user-related and situation-related indicators more actively. Discussions on what counts as good information cannot focus solely on issues of content, credibility and heuristics; the context, particularly the jurisdictional context, in which searches are taking place needs to be considered. ASIAT was designed and applied with two objectives in mind: (1) to focus attention on information seekers in the evaluation of information online; and (2) to offer a mechanism for user-centred formative evaluation of abortion access information online. The tool and the results of the evaluation are noteworthy as they focused explicitly on potential users of abortion services. Both the results of the study and the ASIAT tool are intended to enable web page designers to identify areas for improvement in their own information pages. DND and CP contributed equally.

Conclusion Regardless of its limitations, this research shows that designers of web-based information need to appreciate user-related and situation-related indicators more actively. Discussions on what counts as good information cannot focus solely on issues of content, credibility and heuristics; the context, particularly the jurisdictional context, in which searches are taking place needs to be considered. ASIAT was designed and applied with two objectives in mind: (1) to focus attention on information seekers in the evaluation of information online; and (2) to offer a mechanism for user-centred formative evaluation of abortion access information online. The tool and the results of the evaluation are noteworthy as they focused explicitly on potential users of abortion services. Both the results of the study and the ASIAT tool are intended to enable web page designers to identify areas for improvement in their own information pages. DND and CP contributed equally. Contributors: DND: contributed to the design of the tool, data collection, analysis, and writing up of findings. CP: conceived of the original idea and contributed to the design of the tool, data collection, analysis, and writing up of findings. PB: contributed to the design of the tool, statistical analysis, and writing up of the findings. Funding: The research was produced with the support of Wellcome Trust grant 110469/A/15/Z (‘The Liverpool-Ireland Abortion Corridor: History, Activism and Medical Practice’). Competing interests: None declared. Patient consent: Not required.

Contributors: DND: contributed to the design of the tool, data collection, analysis, and writing up of findings. CP: conceived of the original idea and contributed to the design of the tool, data collection, analysis, and writing up of findings. PB: contributed to the design of the tool, statistical analysis, and writing up of the findings. Funding: The research was produced with the support of Wellcome Trust grant 110469/A/15/Z (‘The Liverpool-Ireland Abortion Corridor: History, Activism and Medical Practice’). Competing interests: None declared. Patient consent: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages A substantial proportion of young people in Britain transition into sexual activity under circumstances that are incompatible with positive sexual health. Adverse circumstances of first sexual intercourse were associated with socio-economic status, educational level, source of sex education, relationship with, and virginity status of, the first partner. An exclusive focus on chronological age neglects the importance of contextual circumstances in defining the nature of first sexual intercourse. Introduction The context in which first sexual intercourse occurs generally receives less empirical attention than chronological age at first sexual intercourse. However, an exclusive focus on age neglects individual differences in physical, social and psychological maturity, as well as the emphasis placed by young people themselves on the circumstances in which first sex occurred in evaluating their experiences.1 As a result, some have argued for a more nuanced concept of readiness and appropriateness of timing of first sexual intercourse.2

s in physical, social and psychological maturity, as well as the emphasis placed by young people themselves on the circumstances in which first sex occurred in evaluating their experiences.1 As a result, some have argued for a more nuanced concept of readiness and appropriateness of timing of first sexual intercourse.2 The concept of ‘sexual competence’ represents an alternative approach to timing of first sexual intercourse, considering the contextual attributes of the event, rather than simply age at occurrence. This departs from the traditional framing of all sexual activity among teenagers as problematic, and recognises that young age alone does not threaten sexual health, any more than older age safeguards it.3 With reference to prior work on the role of ‘interactional competence’ in negotiating sexual behaviour,4 5 ‘sexual competence’ was operationalised specifically in relation to first heterosexual intercourse by Wellings et al 6 using four variables measured in the second National Survey of Sexual Attitudes and Lifestyles (Natsal-2). Participants were classified as ‘sexually competent’ at first intercourse if they reported that the event was characterised by contraceptive protection, autonomy of decision, equal willingness of both partners, and that it had occurred at the ‘right time’.

National Survey of Sexual Attitudes and Lifestyles (Natsal-2). Participants were classified as ‘sexually competent’ at first intercourse if they reported that the event was characterised by contraceptive protection, autonomy of decision, equal willingness of both partners, and that it had occurred at the ‘right time’. The use of these domains in defining sexual competence may be considered compatible with the definition of sexual health endorsed by the WHO,7 highlighting the importance of not only physical health, but also mental and social aspects, referring to a "positive and respectful approach to…sexual relationships" and "safe sexual experiences, free of coercion".

exual competence may be considered compatible with the definition of sexual health endorsed by the WHO,7 highlighting the importance of not only physical health, but also mental and social aspects, referring to a "positive and respectful approach to…sexual relationships" and "safe sexual experiences, free of coercion". In this study, we examined the prevalence of ‘sexual competence’ and its component variables in a representative population-based sample of 17–24-year-olds living in Britain. Further, we examined the antecedent factors that are associated with a lack of ‘sexual competence’ at first sexual intercourse. The public health relevance of this study is two-fold. First, given its compatibility with the WHO definition of sexual health, the concept of sexual competence is likely to represent first sexual intercourse that is consistent with well-being and health. Second, studies have shown that psychosocial factors relating to first intercourse, such as autonomy and the emotional experience, are associated with sexual health outcomes.8–10 We have shown lack of sexual competence at first sex (the term ‘sex’ is used specifically in relation to heterosexual intercourse unless explicitly stated otherwise) to be associated with poor subsequent sexual health, as defined by self-reported sexually transmitted infection (STI) diagnosis, testing positive for human papillomavirus (HPV), lower sexual function, unplanned pregnancy, and experience of non-volitional sex.11

terosexual intercourse unless explicitly stated otherwise) to be associated with poor subsequent sexual health, as defined by self-reported sexually transmitted infection (STI) diagnosis, testing positive for human papillomavirus (HPV), lower sexual function, unplanned pregnancy, and experience of non-volitional sex.11 Methods Participants The Third National Survey of Sexual Attitudes and Lifestyles (Natsal-3) is a stratified probability sample survey of 15 162 men and women aged 16–74 years, resident in Britain, conducted in 2010–2012.12 We restricted analyses to sexually experienced respondents aged 17–24 years (n=2825) to ensure relevancy to the recent cohort becoming sexually active in Britain. In order to examine the relationship between education and sexual competence, participants aged 16 years at interview were excluded because they could not be ascribed an educational level.

lly experienced respondents aged 17–24 years (n=2825) to ensure relevancy to the recent cohort becoming sexually active in Britain. In order to examine the relationship between education and sexual competence, participants aged 16 years at interview were excluded because they could not be ascribed an educational level. Measures Participants were asked about their age at and experience of first heterosexual intercourse in the face-to-face component of the interview. These questions were asked with the use of show cards so that respondents did not have to verbalise any sexually explicit terms, instead quoting the letter that corresponded to their preferred answer option, and to help preserve confidentiality (in case of being overheard by other household members). For participants reporting first sexual intercourse at age 12  years or younger, questions about circumstances were asked about their first experience since turning age 13 years due to ethical concerns relating to probing questions about early non-consensual encounters.12 The questions relating to the experience of first intercourse sought to measure whether partners were both equally willing to engage in sexual intercourse; whether the decision to have sex was autonomous (not due to factors external to the self, such as peer pressure or drunkenness); whether the respondent felt their first experience of sexual intercourse had happened at the ‘right’ time; and whether a reliable method of contraception had been used (contraceptive pill or condom) (survey questions have been previously reported).11 As in the study by Wellings et al,6 the measure of sexual competence constructed using these four variables was as follows: respondents who endorsed all four of these items were categorised as ‘sexually competent’ at first intercourse, and respondents who endorsed fewer than all four were categorised as not ‘sexually competent’ at first intercourse. Respondents reporting that their partner was ‘more willing’ at first sex were filtered to an additional question asking whether they were ‘forced’. Those reporting forced first sex were excluded from analyses (n=22) as it was considered inappropriate to classify these respondents in terms of ‘sexual competence’.

ourse. Respondents reporting that their partner was ‘more willing’ at first sex were filtered to an additional question asking whether they were ‘forced’. Those reporting forced first sex were excluded from analyses (n=22) as it was considered inappropriate to classify these respondents in terms of ‘sexual competence’. Potential explanatory variables were selected with the aim of representing key influences in childhood and adolescence, along with those relating to the more immediate context of first sex. Two indicators relating to socio-economic status were examined: the area-level Index of Multiple Deprivation quintiles,13 and educational level of the participant. The ethnicity of participants, and their family structure (whether they lived with both parents) at age 14 years, provided further contextual information. Two variables relating to learning about sexual matters, which potentially have a more direct influence on sexual behaviour, were examined: one based on a question asking participants to identify the ‘main’ source from which they learnt about sexual matters when growing up; the other asked about the level/difficulty of discussing sex with their parents during their teenage years. Finally, factors relating to the immediate context of first sexual intercourse were explored, including the nature of the relationship, the age of the participant at first sex and how this compared with the age of the partner, and the prior sexual experience of the partner.

ith their parents during their teenage years. Finally, factors relating to the immediate context of first sexual intercourse were explored, including the nature of the relationship, the age of the participant at first sex and how this compared with the age of the partner, and the prior sexual experience of the partner. Statistical analysis We present the prevalence of ‘sexual competence’, and the measure’s component parts by age at first sexual intercourse. Unadjusted odds ratios were calculated to examine variation in the prevalence of ‘sexual competence’ at first sex by potential explanatory factors. Multivariable logistic regression was used to determine which factors were independently associated with ‘sexual competence’ at first sex. Two multivariable regression models are presented. The first model includes the variables relating to socio-demographic background factors, how the respondent learnt about sex, age at first sex. The second model also includes the variables indicative of the relationship context in which first sex occurred. This two-stage approach was employed in order that we could examine the independent associations between sexual competence and the variables conceptualised as more distally related to the outcome, before separately evaluating the variables considered to be more proximal to event, adjusted for those at more distal levels. All analyses were conducted using the Stata (Version 13) survey commands, accounting for the weighting, clustering and stratification of the survey data. Patient and public involvement Patients were not involved in this study.

Multivariable logistic regression was used to determine which factors were independently associated with ‘sexual competence’ at first sex. Two multivariable regression models are presented. The first model includes the variables relating to socio-demographic background factors, how the respondent learnt about sex, age at first sex. The second model also includes the variables indicative of the relationship context in which first sex occurred. This two-stage approach was employed in order that we could examine the independent associations between sexual competence and the variables conceptualised as more distally related to the outcome, before separately evaluating the variables considered to be more proximal to event, adjusted for those at more distal levels. All analyses were conducted using the Stata (Version 13) survey commands, accounting for the weighting, clustering and stratification of the survey data. Patient and public involvement Patients were not involved in this study. Results Prevalence and unadjusted odds ratios

Multivariable logistic regression was used to determine which factors were independently associated with ‘sexual competence’ at first sex. Two multivariable regression models are presented. The first model includes the variables relating to socio-demographic background factors, how the respondent learnt about sex, age at first sex. The second model also includes the variables indicative of the relationship context in which first sex occurred. This two-stage approach was employed in order that we could examine the independent associations between sexual competence and the variables conceptualised as more distally related to the outcome, before separately evaluating the variables considered to be more proximal to event, adjusted for those at more distal levels. All analyses were conducted using the Stata (Version 13) survey commands, accounting for the weighting, clustering and stratification of the survey data. Patient and public involvement Patients were not involved in this study. Results Prevalence and unadjusted odds ratios Table 1 shows the proportion of 17–24-year-old respondents who reported the following conditions of first intercourse: unequal willingness; a non-autonomous decision; that sex had not happened at the ‘right time’; and non-use of contraception, by gender and age at first sexual intercourse. The most commonly reported negative feature of first sex was that it was not felt to have occurred at the ‘right time’ (39.7% of women and 26.5% of men). Approximately 10% of young people did not use a reliable contraceptive method at first sex. Among women, a general pattern was observed whereby those who were younger at first sex more commonly reported adverse contextual factors, with statistically significant trends observed for perceived timing, equal willingness, and non-autonomous decision-making. Among men, such a trend was observed for perceived timing and contraceptive use. Some 77.7% of women and 64.7% of men who reported first intercourse at age 13–14 years were categorised as not ‘sexually competent’, declining to 36.3% and 39.6% among those aged ≥18 years at first intercourse.

-autonomous decision-making. Among men, such a trend was observed for perceived timing and contraceptive use. Some 77.7% of women and 64.7% of men who reported first intercourse at age 13–14 years were categorised as not ‘sexually competent’, declining to 36.3% and 39.6% among those aged ≥18 years at first intercourse. Table 1 Proportion of 17–24-year-olds reporting certain circumstances at first sex by age at first sex Age at first sex (years) Not equally willing Not the ’right time' Non-autonomous reason Did not use reliable contraception Not ’sexually competent' N (unweighted/ weighted)* % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) Women 13–14 28.4 (22.7 to 34.9) 71.6 (65.2 to 77.2) 28.8 (23.2 to 35.0) 17.4 (13.0 to 22.8) 77.7 (71.5 to 82.9) 249/119 15 16.0 (12.4 to 20.4) 53.8 (48.1 to 59.5) 23.6 (19.0 to 28.8) 10.2 (7.4 to 13.9) 61.5 (55.7 to 66.9) 352/185 16 15.7 (12.1 to 20.7) 30.8 (26.2 to 35.8) 13.9 (10.8 to 17.8) 6.8 (4.8 to 9.5) 44.9 (39.8 to 50.0) 473/260 17 17.9 (12.4 to 25.0) 31.4 (25.2 to 38.4) 14.3 (10.2 to 19.6) 10.9 (7.2 to 16.2) 47.6 (40.7 to 54.6) 262/152 18–24 13.4 (9.0 to 19.4) 22.1 (16.3 to 29.3) 7.7 (4.4 to 13.1) 11.3 (7.5 to 16.7) 36.3 (29.3 to 44.0) 227/163 P value† 0.009 <0.001 <0.001 0.234 <0.001

Age at first sex (years) Not equally willing Not the ’right time' Non-autonomous reason Did not use reliable contraception Not ’sexually competent' N (unweighted/ weighted)* % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) Women 13–14 28.4 (22.7 to 34.9) 71.6 (65.2 to 77.2) 28.8 (23.2 to 35.0) 17.4 (13.0 to 22.8) 77.7 (71.5 to 82.9) 249/119 15 16.0 (12.4 to 20.4) 53.8 (48.1 to 59.5) 23.6 (19.0 to 28.8) 10.2 (7.4 to 13.9) 61.5 (55.7 to 66.9) 352/185 16 15.7 (12.1 to 20.7) 30.8 (26.2 to 35.8) 13.9 (10.8 to 17.8) 6.8 (4.8 to 9.5) 44.9 (39.8 to 50.0) 473/260 17 17.9 (12.4 to 25.0) 31.4 (25.2 to 38.4) 14.3 (10.2 to 19.6) 10.9 (7.2 to 16.2) 47.6 (40.7 to 54.6) 262/152 18–24 13.4 (9.0 to 19.4) 22.1 (16.3 to 29.3) 7.7 (4.4 to 13.1) 11.3 (7.5 to 16.7) 36.3 (29.3 to 44.0) 227/163 P value† 0.009 <0.001 <0.001 0.234 <0.001 All 17.4 (15.4 to 19.7) 39.7 (37.0 to 42.4) 17.0 (15.1 to 19.1) 10.5 (8.9 to 12.3) 51.7 (48.9 to 54.5) N (unweighted/ weighted) 1567/880 1562/877 1543/865 1566/880 1563/878 Men 13–14 16.6 (11.1 to 24.1) 49.0 (41.4 to 56.6) 14.6 (9.5 to 21.6) 25.0 (19.2 to 31.9) 64.7 (57.3 to 71.5) 210/148 15 8.0 (5.0 to 12.4) 29.6 (23.8 to 36.0) 15.3 (11.4 to 20.4) 10.9 (7.4 to 15.8) 47.3 (40.7 to 53.9) 262/186 16 6.7 (4.1 to 10.8) 20.1 (15.1 to 26.2) 11.4 (8.0 to 16.1) 5.7 (3.5 to 8.9) 34.3 (28.5 to 40.6) 318/234 17 13.1 (7.6 to 21.6) 15.6 (10.7 to 22.2) 7.9 (4.7 to 12.8) 9.5 (6.1 to 14.5) 38.0 (30.7 to 46.0) 214/160 18–24 7.4 (4.6 to 11.6) 23.0 (16.8 to 30.6) 12.4 (8.3 to 18.1) 12.8 (8.6 to 18.6) 39.6 (32.3 to 47.3) 217/180 P value† 0.133 <0.001 0.161 0.014 <0.001

All 17.4 (15.4 to 19.7) 39.7 (37.0 to 42.4) 17.0 (15.1 to 19.1) 10.5 (8.9 to 12.3) 51.7 (48.9 to 54.5) N (unweighted/ weighted) 1567/880 1562/877 1543/865 1566/880 1563/878 Men 13–14 16.6 (11.1 to 24.1) 49.0 (41.4 to 56.6) 14.6 (9.5 to 21.6) 25.0 (19.2 to 31.9) 64.7 (57.3 to 71.5) 210/148 15 8.0 (5.0 to 12.4) 29.6 (23.8 to 36.0) 15.3 (11.4 to 20.4) 10.9 (7.4 to 15.8) 47.3 (40.7 to 53.9) 262/186 16 6.7 (4.1 to 10.8) 20.1 (15.1 to 26.2) 11.4 (8.0 to 16.1) 5.7 (3.5 to 8.9) 34.3 (28.5 to 40.6) 318/234 17 13.1 (7.6 to 21.6) 15.6 (10.7 to 22.2) 7.9 (4.7 to 12.8) 9.5 (6.1 to 14.5) 38.0 (30.7 to 46.0) 214/160 18–24 7.4 (4.6 to 11.6) 23.0 (16.8 to 30.6) 12.4 (8.3 to 18.1) 12.8 (8.6 to 18.6) 39.6 (32.3 to 47.3) 217/180 P value† 0.133 <0.001 0.161 0.014 <0.001 All 9.8 (7.9 to 12.2) 26.5 (23.8 to 29.5) 12.3 (10.3 to 14.6) 12.0 (10.2 to 14.1) 43.6 (40.4 to 46.9) N (unweighted/ weighted)‡ 1226/912 1218/906 1221/908 1225/911 1221/908 *Values of n displayed are the denominators used for calculating the proportions not ‘sexually competent’. Denominators for other components vary slightly due to non-response to individual items. †P value for test for trend. ‡Denominator for ‘sexual competence’ exceeds that of some components: respondents with missing data on any of the components were classified as not sexually competent if any of the non-missing values indicated unequal willingness, not the right time, non-autonomous reason, or non-use of reliable contraception.

†P value for test for trend. ‡Denominator for ‘sexual competence’ exceeds that of some components: respondents with missing data on any of the components were classified as not sexually competent if any of the non-missing values indicated unequal willingness, not the right time, non-autonomous reason, or non-use of reliable contraception. Table 2 shows the prevalence and unadjusted odds ratios of a lack of sexual competence at first sex according to potential explanatory factors. Among both men and women, lack of sexual competence was associated with: living in a more deprived area; lower education level; not living with both parents at age 14 years; first sex occurring before age 16 years; reporting having not been in a ‘steady’ relationship at first sex; the virginity status of the first sexual partner (with the highest odds of lacking sexual competence among those who were uncertain of their partner’s virginity status); and having an older first sexual partner. Among women only, a lack of sexual competence was also associated with black ethnicity; reporting ‘friends’ or ‘other’ as the main source of learning about sexual matters while growing up; and lack of discussion with parents about sexual matters when growing up. Table 2 Proportions and unadjusted odds ratios of respondents aged 17–24 years at interview who were not sexually competent at first sex, by explanatory variables.

Table 2 shows the prevalence and unadjusted odds ratios of a lack of sexual competence at first sex according to potential explanatory factors. Among both men and women, lack of sexual competence was associated with: living in a more deprived area; lower education level; not living with both parents at age 14 years; first sex occurring before age 16 years; reporting having not been in a ‘steady’ relationship at first sex; the virginity status of the first sexual partner (with the highest odds of lacking sexual competence among those who were uncertain of their partner’s virginity status); and having an older first sexual partner. Among women only, a lack of sexual competence was also associated with black ethnicity; reporting ‘friends’ or ‘other’ as the main source of learning about sexual matters while growing up; and lack of discussion with parents about sexual matters when growing up. Table 2 Proportions and unadjusted odds ratios of respondents aged 17–24 years at interview who were not sexually competent at first sex, by explanatory variables. Explanatory variables Men 95% CI OR 95% CI P value* N (unweighted/ weighted) Women 95% CI OR 95% CI P value* N (unweighted/ weighted) % % IMD quintile 1: Least deprived 27.7 21.5 to 34.8 1 <0.001 216/158 43.9 38.0 to 50.0 1 0.008 263/152 2 38.9 32.3 to 45.9 1.66 1.07 to 2.57 233/177 49.0 42.6 to 55.5 1.23 0.86 to 1.75 280/157 3 40.4 33.7 to 47.6 1.77 1.14 to 2.77 231/164 46.3 39.9 to 52.8 1.10 0.78 to 1.56 283/170 4 52.7 45.2 to 60.0 2.91 1.85 to 4.55 258/213 58.4 52.6 to 64.0 1.79 1.28 to 2.52 339/195 5: Most deprived 53.5 47.0 to 59.9 3.00 1.97 to 4.59 283/197 57.6 51.9 to 63.1 1.73 1.24 to 2.42 398/204 Education level of respondent Left school at 17+ years 37.3 33.3 to 41.5 1 <0.001 718/550 46.3 42.8 to 49.8 1 <0.001 930/550 Left school at 16 years with qualifications 51.2 45.6 to 56.6 1.76 1.33 to 2.33 382/262 59.2 54.1 to 64.1 1.68 1.31 to 2.16 479/245 Left school at 16 years with no qualifications 65.4 53.8 to 75.4 3.17 1.89 to 5.30 89/66 70.3 60.0 to 78.8 2.74 1.71 to 4.40 116/55 Ethnic group White 42.4 39.1 to 45.8 1 0.143 1094/802 50.1 47.3 to 53.0 1 0.020 1396/768 Mixed 64.6 46.7 to 79.2 2.47 1.16 to 5.26 38/27 64.2 50.8 to 75.7 1.79 1.02 to 3.12 65/36 Asian 48.4 32.5 to 64.6 1.27 0.65 to 2.50 43/40 45.7 29.1 to 63.3 0.84 0.40 to 1.74 38/27 Black 43.8 26.0 to 63.2 1.06 0.47 to 2.36 32/27 76.5 59.2 to 87.9 3.24 1.44 to 7.28 42/33 Chinese and ’other' 57.2 29.3 to 81.2 1.81 0.56 to 5.82 14/13 55.2 31.6 to 76.7 1.23 0.46 to 3.27 20/14 Family structure <0.001 Both parents 41.3 37.3 to 45.4 1 0.027 816/643 47.6 44.0 to 51.2 1 928/555 One/neither parent 49.2 43.7 to 54.7 1.38 1.04 to 1.83 405/265 58.6 54.3 to 62.8 1.56 1.25 to 1.96 634/322 Main source of information about sexual matters Lessons at school 40.2 34.6 to 46.1 1 0.502 412/310 46.3 41.8 to 50.8 1 0.023 567/322 Mother or father 47.8 36.5 to 59.2 1.36 0.81 to 2.29 92/72 52.0 45.1 to 58.8 1.26 0.90 to 1.75 236/129 Friends 45.5 39.5 to 51.6 1.24 0.89 to 1.73 314/227 55.9 50.7 to 61.1 1.47 1.11 to 1.95 420/230 Other 44.6 39.4 to 50.0 1.20 0.87 to 1.65 396/295 55.5 49.5 to 61.4 1.45 1.07 to 1.95 335/194 Ease discussing sex with parents at age 14 years Easy with one/both 42.8 36.2 to 49.6 1 0.300 312/241

8 1.26 0.90 to 1.75 236/129 Friends 45.5 39.5 to 51.6 1.24 0.89 to 1.73 314/227 55.9 50.7 to 61.1 1.47 1.11 to 1.95 420/230 Other 44.6 39.4 to 50.0 1.20 0.87 to 1.65 396/295 55.5 49.5 to 61.4 1.45 1.07 to 1.95 335/194 Ease discussing sex with parents at age 14 years Easy with one/both 42.8 36.2 to 49.6 1 0.300 312/241 46.9 42.1 to 51.8 1 0.051 501/284 Difficult 52.0 40.0 to 63.7 1.45 0.83 to 2.53 87/71 55.3 45.4 to 64.9 1.40 0.90 to 2.20 132/78 Didn’t discuss with either 42.6 38.8 to 46.6 0.99 0.72 to 1.36 759/550 53.8 50.0 to 57.5 1.32 1.03 to 1.68 821/465 Varied depending on topic 31.5 16.9 to 50.9 0.61 0.26 to 1.44 29/25 40.9 28.2 to 54.9 0.78 0.43 to 1.43 59/31 Sex before age 16 years No 37.0 32.9 to 41.2 1 <0.001 749/574 43.2 39.6 to 46.8 1 <0.001 962/575 Yes 55.0 49.9 to 60.0 2.08 1.58 to 2.74 472/334 67.8 63.5 to 71.9 2.78 2.17 to 3.55 601/304 Relationship with first sexual partner Steady relationship 34.2 30.0 to 38.7 1 <0.001 574/436 39.8 36.3 to 43.3 1 <0.001 968/545 Just/recently met for first time 61.3 53.5 to 68.5 3.04 2.12 to 4.37 207/152 77.3 67.5 to 84.8 5.15 3.07 to 8.64 129/78 Known each other a while, not in steady relationship 47.8 42.4 to 53.3 1.76 1.32 to 2.36 394/286 73.4 68.2 to 78.1 4.19 3.12 to 5.63 384/207 Used to be in steady relationship 52.2 35.9 to 68.1 2.10 1.06 to 4.18 44/33 64.9 48.8 to 78.2 2.80 1.42 to 5.54 61/34.61 Married/living together – – – – – 2/1 16.2 4.9, 42.5 0.29 0.08 to 1.13 20/13 Partner’s first time too Yes, partner’s first time 37.1 32.9 to 41.5 1 <0.001 577/434 41.6 37.2 to 46.0 1 <0.001 573/318 Think it was first time 52.8 37.1 to 68.0 1.90 0.97 to 3.70 50/43 72.1 55.4 to 84.3 3.63 1.71 to 7.72 43/24 Think it was not first time 59.5 47.8 to 70.2 2.49 1.51 to 4.12 94/69 66.9 55.2 to 76.9 2.85 1.70 to 4.78 82/49 No, not first time 46.0 40.7 to 51.3 1.44 1.09 to 1.91 444/325 55.9 52.1 to 59.7 1.79 1.41 to 2.27 825/466 Age difference between respondent and first sex partner Same age 40.2 36.0 to 44.5 1 <0.001 718/535 45.6 41.0 to 50.2 1 0.005 551/314 Partner younger than respondent 38.6 31.3 to 46.5 0.94 0.64 to 1.36 211/168 56.6 41.6 to 70.4 1.56 0.83 to 2.92 54/31 Respondent younger than partner 56.4 49.8 to 62.8 1.92 1.41 to 2.63 286/203 55.0 51.5 to 58.5 1.46 1.16 to 1.85 956/532 *Wald test.

e 40.2 36.0 to 44.5 1 <0.001 718/535 45.6 41.0 to 50.2 1 0.005 551/314 Partner younger than respondent 38.6 31.3 to 46.5 0.94 0.64 to 1.36 211/168 56.6 41.6 to 70.4 1.56 0.83 to 2.92 54/31 Respondent younger than partner 56.4 49.8 to 62.8 1.92 1.41 to 2.63 286/203 55.0 51.5 to 58.5 1.46 1.16 to 1.85 956/532 *Wald test. CI, confidence interval; IMD, Index of Multiple Deprivation; OR, odds ratio. Multivariable regression analyses The results of multivariable logistic regression analyses are presented in tables 3 and 4. The first model (Model 1) includes the variables relating to socio-demographic background factors, how the respondent learnt about sex, and age at first sex. In these adjusted models, the majority of associations observed in the crude analyses were retained, although somewhat attenuated. (Model 1, tables 3 and 4). Table 3 Logistic regression examining predictors of lack of sexual competence at first sex, results adjusted for all other variables in table column (Men)

Multivariable regression analyses The results of multivariable logistic regression analyses are presented in tables 3 and 4. The first model (Model 1) includes the variables relating to socio-demographic background factors, how the respondent learnt about sex, and age at first sex. In these adjusted models, the majority of associations observed in the crude analyses were retained, although somewhat attenuated. (Model 1, tables 3 and 4). Table 3 Logistic regression examining predictors of lack of sexual competence at first sex, results adjusted for all other variables in table column (Men) Outcome: not sexually competent at first intercourse Model 1 Model 2 AOR 95% CI P value AOR 95% CI P value IMD quintile 1: Least deprived 1 0.003 1 0.006 2 1.58 1.00 to 2.48 1.49 0.95 to 2.34 3 1.64 1.03 to 2.62 1.55 0.97 to 2.47 4 2.43 1.51 to 3.92 2.29 1.44 to 3.63 5: Most deprived 2.21 1.40 to 3.48 2.03 1.28 to 3.22 Education level of respondent Studying for/attained further academic qualifications 1 0.002 1 0.047 No academic qualifications 2.36 1.32 to 4.22 2.27 1.25 to 4.13 Academic qualifications typically gained at age 16 years 1.54 1.12 to 2.12 1.49 1.08 to 2.05 Ethnic group White 1 0.136 1 0.130 Mixed 2.84 1.18 to 6.84 2.88 1.12 to 7.41 Asian 1.49 0.73 to 3.05 1.71 0.82 to 3.58 Black 0.75 0.31 to 1.82 0.75 0.30 to 1.87 Chinese and ’other' 0.84 0.13 to 5.46 0.85 0.10 to 7.24 Family structure Both parents 1 0.957 1 One/neither parent 0.99 0.72 to 1.36 1.01 0.74 to 1.39 0.935 Main source of information about sexual matters Lessons at school 1 0.646 1 0.754 Mother or father 1.29 0.69 to 2.42 1.22 0.67 to 2.22 Friends 1.00 0.70 to 1.43 0.95 0.66 to 1.37 Other 0.88 0.62 to 1.25 0.89 0.62 to 1.27 Ease discussing sex with parents at age 14 years Easy with one/both 1 0.272 1 0.197 Difficult 1.65 0.87 to 3.13 1.66 0.87 to 3.17 Didn’t discuss with either 0.99 0.69 to 1.43 0.96 0.67 to 1.38 Varied depending on topic 0.72 0.31 to 1.70 0.64 0.28 to 1.46 Sex before age 16 No 1 <0.001 1 0.001 Yes 1.72 1.27 to 2.34 1.71 1.24 to 2.36 Relationship with first sexual partner Steady relationship 1 <0.001 Just/recently met for first time 2.42 1.57 to 3.73 Known each other a while to not in steady relationship 1.15 0.81 to 1.62 Used to be in steady relationship 1.56 0.66 to 3.70 Married/living together – – – Partner’s first time too Yes to partner’s first time 1 0.0639 Think it was first time 1.75 0.87 to 3.53 Think it was not first time 2.12 1.15 to 3.91 No to not first time 1.19 0.84 to 1.67 Age difference between respondent and first sex partner Same age 1 0.4316 Partner younger than respondent 0.94 0.63 to 1.41 Respondent younger than partner 1.25 0.86 to 1.80 N unweighted 1092 1092 N weighted 817 817 AOR, adjusted odds ratio; CI, confidence interval; IMD, Index of Multipl

ot first time 1.19 0.84 to 1.67 Age difference between respondent and first sex partner Same age 1 0.4316 Partner younger than respondent 0.94 0.63 to 1.41 Respondent younger than partner 1.25 0.86 to 1.80 N unweighted 1092 1092 N weighted 817 817 AOR, adjusted odds ratio; CI, confidence interval; IMD, Index of Multipl e Deprivation. Table 4 Logistic regression examining predictors of lack of sexual competence at first sex, results adjusted for all other variables in table column (Women)

ot first time 1.19 0.84 to 1.67 Age difference between respondent and first sex partner Same age 1 0.4316 Partner younger than respondent 0.94 0.63 to 1.41 Respondent younger than partner 1.25 0.86 to 1.80 N unweighted 1092 1092 N weighted 817 817 AOR, adjusted odds ratio; CI, confidence interval; IMD, Index of Multipl e Deprivation. Table 4 Logistic regression examining predictors of lack of sexual competence at first sex, results adjusted for all other variables in table column (Women) Outcome: not sexually competent at first intercourse AOR Model 1 P value AOR Model 2 P value 95% CI 95% CI IMD quintile 1: Least deprived 1 0.233 1 0.291 2 1.19 0.81 to 1.74 1.18 0.78 to 1.78 3 0.99 0.68 to 1.43 0.96 0.64 to 1.44 4 1.44 0.99 to 2.07 1.43 0.97 to 2.12 5: Most deprived 1.26 0.87 to 1.82 1.22 0.83 to 1.81 Education level of respondent Studying for/attained further academic qualifications 1 0.020 1 0.067 Academic qualifications typically gained at 16 years 1.34 1.02 to 1.76 1.22 0.91 to 1.63 No academic qualifications 1.81 1.07 to 3.08 1.87 1.06 to 3.32 Ethnic group White 1 0.050 1 0.009 Mixed 1.75 0.87 to 3.52 1.47 0.74 to 2.93 Asian 1.54 0.67 to 3.52 2.85 1.10 to 7.41 Black 3.22 1.26 to 8.22 4.67 1.62 to 13.45 Chinese and ’other' 1.66 0.60 to 4.58 2.97 0.80 to 11.01 Family structure Both parents 1 0.029 1 0.245 One/neither parent 1.34 1.03 to 1.73 1.18 0.89 to 1.56 Main source of information about sexual matters Lessons at school 1 0.071 1 0.536 Mother or father 1.38 0.92 to 2.09 1.33 0.85 to 2.09 Friends 1.48 1.09 to 2.02 1.20 0.86 to 1.67 Other 1.25 0.90 to 1.72 1.15 0.80 to 1.64 Ease discussing sex with parents at age 14 years Easy with one/both 1 0.096 1 0.137 Difficult 1.58 0.93 to 2.67 1.51 0.90 to 2.54 Didn’t discuss with either 1.39 1.02 to 1.90 1.39 0.98 to 1.95 Varied depending on topic 0.94 0.50 to 1.77 0.91 0.47 to 1.76 Sex before age 16 years No 1 <0.001 1 <0.001 Yes 2.60 2.00 to 3.38 2.93 2.20 to 3.90 Relationship with first sexual partner Steady relationship 1 <0.001 Just/recently met for first time 4.84 2.71 to 8.64 Known each other a while to not in steady relationship 3.95 2.81 to 5.56 Used to be in steady relationship 2.95 1.45 to 6.01 Married/living together 0.14 0.02 to 1.21 Partner’s first time too Yes to partner’s first time 1 <0.001 Think it was first time 3.90 1.62 to 9.36 Think it was not first time 2.29 1.21 to 4.34 No to not first time 1.52 1.14 to 2.04 Age difference between respondent and first sex partner Same age 1 0.639 Partner younger than respondent 1.40 0.70 to 2.81 Respondent younger than partner 1.04 0.79 to 1.37 N unweighted 1432 1432 N weighted 805 805 AOR, adjusted odds ratio; CI, confidence interval; IMD,

1 to 4.34 No to not first time 1.52 1.14 to 2.04 Age difference between respondent and first sex partner Same age 1 0.639 Partner younger than respondent 1.40 0.70 to 2.81 Respondent younger than partner 1.04 0.79 to 1.37 N unweighted 1432 1432 N weighted 805 805 AOR, adjusted odds ratio; CI, confidence interval; IMD, Index of Multiple Deprivation. After adjustment for variables relating to the immediate relational context in which first intercourse occurred (Model 2, table 4), source of learning about sexual matters and communication with parents about sexual matters were no longer associated with sexual competence among women, potentially indicating a mediatory role of these more proximal factors. Lower educational level, black ethnicity, and sex before 16 years retained their associations with a lack of sexual competence at first sex among women, although at a borderline level for the former. Among men, the associations between a sexual competence and IMD quintile, educational level, and sex before 16 retained statistical significance, even after adjustment for the variables relating to the immediate relational context of first sex (Model 2, table 3).

ng women, although at a borderline level for the former. Among men, the associations between a sexual competence and IMD quintile, educational level, and sex before 16 retained statistical significance, even after adjustment for the variables relating to the immediate relational context of first sex (Model 2, table 3). After adjustment for all other variables in the model, the status of the relationship with the first sexual partner retained its strong associations with sexual competence (Model 2, tables 2 and 3). Among men, this association seemed primarily driven by the increased odds of a lack of sexual competence among those reporting they ‘had just/recently met’ their partner compared with those in a ‘steady relationship’ at the time. Respondents’ knowledge of their first sexual partner’s virginity status also continued to be associated with sexual competence among women after adjustment. A similar association, of borderline statistical significance, was also evident among men. Finally, having had an older partner at first sex was no longer associated with lacking sexual competence among men or women in these fully adjusted models. Discussion This study describes the circumstances of first sexual intercourse using a representative population-based sample of young people living in Britain, and provides an exploration of the antecedent factors associated with a novel measure of the first sexual intercourse experience: ‘sexual competence’.

After adjustment for all other variables in the model, the status of the relationship with the first sexual partner retained its strong associations with sexual competence (Model 2, tables 2 and 3). Among men, this association seemed primarily driven by the increased odds of a lack of sexual competence among those reporting they ‘had just/recently met’ their partner compared with those in a ‘steady relationship’ at the time. Respondents’ knowledge of their first sexual partner’s virginity status also continued to be associated with sexual competence among women after adjustment. A similar association, of borderline statistical significance, was also evident among men. Finally, having had an older partner at first sex was no longer associated with lacking sexual competence among men or women in these fully adjusted models. Discussion This study describes the circumstances of first sexual intercourse using a representative population-based sample of young people living in Britain, and provides an exploration of the antecedent factors associated with a novel measure of the first sexual intercourse experience: ‘sexual competence’. Adverse circumstances of first sex were reported by a substantial proportion of young people. More than a third of women and a quarter of men did not consider that their experience of first sexual intercourse occurred at the ‘right time’, while almost 1 in 5 women reported that they and their partner were not equally willing to have sex on that first occurrence, and a similar proportion of women reported a non-autonomous reason for first sex. While the majority of young people used a reliable contraceptive method at first sex, 1 in 10 did not.

ht time’, while almost 1 in 5 women reported that they and their partner were not equally willing to have sex on that first occurrence, and a similar proportion of women reported a non-autonomous reason for first sex. While the majority of young people used a reliable contraceptive method at first sex, 1 in 10 did not. In relation to the composite measure of ‘sexual competence’, over half of women and more than a third of men were categorised as not being ‘sexually competent’ at first sex. Although age at first intercourse was associated with sexual competence, it did not explain all of the variability in sexual competence – at no age did the prevalence of sexual competence approach zero or 100%. This finding supports the proposition that chronological age may be an overly simplistic indicator of the nature of first intercourse. Furthermore, the associations between several antecedent factors and sexual competence at first sex were retained when adjusting for age at first sex. This provides further evidence that the measure of sexual competence represents a distinct dimension of the experience of first intercourse, which is not simply a function of age.

e, the associations between several antecedent factors and sexual competence at first sex were retained when adjusting for age at first sex. This provides further evidence that the measure of sexual competence represents a distinct dimension of the experience of first intercourse, which is not simply a function of age. In line with previous research, the stability of the partnership was associated with a more positive first sexual experience.14–17 Uncertainty relating to the virginity status of the partner was associated with a lack of sexual competence, potentially suggesting this variable is acting as a proxy for communication between partners. The association between indicators of socio-economic status and sexual competence is consistent with previous research6 15 18 19 and may be explained by the effect of limited life aspirations on sexual behaviour.18 20 Previous research has found that individuals of lower socio-economic status have lower levels of perceived control,21 22 which could be an important psycho-social determinant of sexual competence. The association between ethnicity and sexual competence is consistent with previous research reporting variations in sexual behaviour across different ethnic groups, as is the finding that the association remains after adjustment for broader risk factors (eg, indicators of socio-economic status).23 Further research is warranted to examine the drivers of ethnic variations in sexual behaviour and health.

esearch reporting variations in sexual behaviour across different ethnic groups, as is the finding that the association remains after adjustment for broader risk factors (eg, indicators of socio-economic status).23 Further research is warranted to examine the drivers of ethnic variations in sexual behaviour and health. That young women who had discussed sexual matters with their parents, and those who reported school to be their main source from which they learnt about sexual matters, were more likely to have been sexually competent at first sex resonates with previous research.6 24 Parental communication, and school-based relationships and sex education, may provide the knowledge and skills required to negotiate a positive and safe sexual experience. However, these associations were not observed among men, even in unadjusted analyses. A possible interpretation is that communication and negotiation skills are less important for men in achieving a first sexual intercourse that they reflect positively on. Prior research reports that men generally give more positive accounts of first intercourse as they are more likely to just be happy that they had sex25 26 and less likely to report experiencing pressure from their partner.25 27

s important for men in achieving a first sexual intercourse that they reflect positively on. Prior research reports that men generally give more positive accounts of first intercourse as they are more likely to just be happy that they had sex25 26 and less likely to report experiencing pressure from their partner.25 27 Limitations The response rate to Natsal-3, at 57.7%,12 potentially limits representativeness of the findings. Our reliance on observational data means that the associations detected may be due to unmeasured and/or unknown confounders. The Natsal survey relies on retrospective self-reports relating to an event that could have occurred up to a decade earlier; therefore, it is important to consider the potential for recall bias when interpreting the results. This could explain the strong association observed between relationship with partner at first sex and sexual competence at first sex; perhaps those who reflect on the first sexual intercourse positively, and therefore will be classified at sexually competent, will also be more likely to recall that they were in a ‘steady’ relationship at the time. Finally, despite Natsal-3’s large sample size, a relatively small proportion of participants were of non-white British ethnicity (reflecting the ethnic composition of Britain), meaning that analyses involving specific ethnic groups were limited by small numbers.

ll that they were in a ‘steady’ relationship at the time. Finally, despite Natsal-3’s large sample size, a relatively small proportion of participants were of non-white British ethnicity (reflecting the ethnic composition of Britain), meaning that analyses involving specific ethnic groups were limited by small numbers. Conclusions and implications A substantial proportion of young people in Britain become sexually active under circumstances that are arguably incompatible with sexual health defined in its broad sense encompassing both physical and psycho-social well-being. The antecedent factors associated with sexual competence are of public health relevance for understanding where interventions to improve the conditions of first sex may be best targeted. Inequalities in sexual health have commonly been described in terms of the unequal distributions of STIs,28 unplanned pregnancies,29 and ‘early’ transitions into sexual activity,30 across socio-economic groups. The current findings suggest that these inequalities are also reflected in the nature of first intercourse, indicating that greater efforts are required to reduce the disparities that exist from the very onset of sexual activity. While the results indicate that communication with parents about sex and school-based sex education may help towards the achievement of sexual competence among young women, the same cannot be said for men, suggesting that greater consideration needs to be given to how men can best be equipped to have a safe and positive transition into sexual activity.

mmunication with parents about sex and school-based sex education may help towards the achievement of sexual competence among young women, the same cannot be said for men, suggesting that greater consideration needs to be given to how men can best be equipped to have a safe and positive transition into sexual activity. Previous research has suggested that the experience of first sexual intercourse can have implications for sexual health status later in life.8–10 Analyses of Natsal-3 data have identified lack of sexual competence at first intercourse to be a risk factor for poor subsequent sexual health among young people, independently of age at first sex.11 Therefore, it is possible that targeted interventions aimed at enabling at-risk young people to have a more positive and healthy first sexual experience may result in improvements in sexual health that continue into adulthood.

or for poor subsequent sexual health among young people, independently of age at first sex.11 Therefore, it is possible that targeted interventions aimed at enabling at-risk young people to have a more positive and healthy first sexual experience may result in improvements in sexual health that continue into adulthood. As a research tool, the measure of sexual competence was constructed rather opportunistically by Natsal-2 researchers combining existing variables considered to be necessary for a healthy first sexual intercourse. The finding that age at first sex does not explain all of the variation observed in sexual competence, and nor does it account for the associations found with other antecedent factors, suggests that this measure is not merely a function of age at first sex, but rather tapping into a distinct dimension of the experience in itself. These conclusions support the future use of this measure in research concerned with sexual behaviour among young people, and may represent an alternative indicator by which the effectiveness of interventions to improve sexual health can be assessed. Natsal-3 is a collaboration between University College London (London, UK), the London School of Hygiene and Tropical Medicine (London, UK), NatCen Social Research, Public Health England (formerly the Health Protection Agency), and the University of Manchester (Manchester, UK). The authors thank the study participants; the team of interviewers from NatCen Social Research; and operations and computing staff from NatCen Social Research.

Medicine (London, UK), NatCen Social Research, Public Health England (formerly the Health Protection Agency), and the University of Manchester (Manchester, UK). The authors thank the study participants; the team of interviewers from NatCen Social Research; and operations and computing staff from NatCen Social Research. Funding: Natsal-3 was supported by grants from the Medical Research Council (www.mrc.ac.uk; G0701757) and the Wellcome Trust (www.wellcome.ac.uk; 084840), with contributions from the Economic and Social Research Council and Department of Health. MJP was funded by an ESRC PhD Studentship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: None declared. Patient consent: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Development of male hormonal contraception has been delayed by multiple factors, but new clinical trials are closer than ever to bringing viable products to market. With new technologies, new potential non-hormonal targets for male contraception have been discovered. Developing a greater range of male methods is vital to providing holistic reproductive healthcare. Introduction The Faculty of Sexual and Reproductive Healthcare vision statement1 focuses on patient experience, specifically that patients should have access to the full range of contraceptive options. For men, these options are currently limited to condoms and vasectomy. Both these methods have their positives—condoms continue to play a vital role in infection prevention and are a widely used, easily accessible method of contraception; likewise, vasectomy is a highly effective permanent method of contraception. However, they are not sufficient for many sexually active men who wish to have control over their fertility, and for their partners who wish to share the burden of contraception. We know that sterilisation and condoms are not sufficient for women to control their fertility, and having a greater breadth of contraceptive tools allows more couples to have better reproductive control. This review will help clinicians to understand the male contraceptives that are on the horizon and may start coming to market in the medium term.

condoms are not sufficient for women to control their fertility, and having a greater breadth of contraceptive tools allows more couples to have better reproductive control. This review will help clinicians to understand the male contraceptives that are on the horizon and may start coming to market in the medium term. The life course of a sperm Spermatogenesis begins in the brain. The hypothalamus secretes gonadotrophin-releasing hormone (GnRH), which in turn stimulates the anterior pituitary to release luteinising hormone (LH) and follicle stimulating hormone (FSH). These hormones then act on the Leydig and Sertoli cells within the testis. Leydig cells produce testosterone, which is secreted into the bloodstream. Serum testosterone supports a wide range of male functions and is a key component of the negative feedback loop that controls GnRH and gonadotrophin production. Testicular levels of testosterone are approximately 40 times higher than in blood2 and support Sertoli cells in their role in spermatogenesis. After release from Sertoli cells (‘spermiation’), sperm progress through the seminiferous tubules and into the epididymis for storage, concentration and maturation to functional sperm. Finally, at the point of ejaculation, sperm are transferred from the epididymis via the vas deferens to the urethra and then out of the body (figure 1).

m Sertoli cells (‘spermiation’), sperm progress through the seminiferous tubules and into the epididymis for storage, concentration and maturation to functional sperm. Finally, at the point of ejaculation, sperm are transferred from the epididymis via the vas deferens to the urethra and then out of the body (figure 1). Figure 1 The hypothalamic-pituitary-testicular axis and its role in spermatogenesis. CCBY 4.0 Licence – John Reynolds-Wright. Image accessible via https://flic.kr/p/2hgvk5V6A. FSH, follicle stimulating hormone; GnRH, gonadotrophin-releasing hormone; LH, luteinising hormone. Potential novel male contraceptives can be divided into those based on a hormonal approach, acting via gonadotrophin suppression, and non-hormonal methods. Hormonal strategies for male contraception involve the hypothalamic-pituitary-testicular (HPT) axis, using exogenous testosterone in combination with progestogen to drive the negative feedback response and suppress GnRH, FSH and LH. Exogenous replacement of testosterone prevents hypogonadal side effects from HPT axis suppression but is not delivered at high enough concentrations to support spermatogenesis directly. Non-hormonal methods can act at any point during spermatogenesis or can target sperm maturation, detachment, motility or transport out of the testis. This includes physical methods of infiltration of the vas with a dissolvable compound to block or damage sperm during their movement along the vas.

rmatogenesis directly. Non-hormonal methods can act at any point during spermatogenesis or can target sperm maturation, detachment, motility or transport out of the testis. This includes physical methods of infiltration of the vas with a dissolvable compound to block or damage sperm during their movement along the vas. A brief history of hormonal male contraception The question asked most frequently about male contraception, other than ‘When will it be ready?’, is ‘what is taking so long?’. The following summary on the history of male contraception will hopefully answer this question.

rmatogenesis directly. Non-hormonal methods can act at any point during spermatogenesis or can target sperm maturation, detachment, motility or transport out of the testis. This includes physical methods of infiltration of the vas with a dissolvable compound to block or damage sperm during their movement along the vas. A brief history of hormonal male contraception The question asked most frequently about male contraception, other than ‘When will it be ready?’, is ‘what is taking so long?’. The following summary on the history of male contraception will hopefully answer this question. The quest to develop new forms of male contraception started at more or less the same time as the development of female hormonal contraceptive methods, but has taken a very different course. The culture shift that took place in the wake of the phenomenon of the hormonal contraceptive pill for women3 in 1965 led to the creation of the WHO Special Programme of Research, Development and Research Training in Human Reproduction (WHO-HRP).3 Within the WHO-HRP a task force for developing methods of male fertility regulation was developed. This identified possible modalities for male contraceptives through basic science research projects and multicentre landmark clinical trials of testosterone alone and in combination with progestogens.3 The task force also focused efforts on the development and promotion of no-scalpel vasectomy, which has become the predominant technique globally, and gossypol, which was planned as a reversible contraceptive but unfortunately had to be abandoned due to issues with toxicity and irreversibility.3

ion with progestogens.3 The task force also focused efforts on the development and promotion of no-scalpel vasectomy, which has become the predominant technique globally, and gossypol, which was planned as a reversible contraceptive but unfortunately had to be abandoned due to issues with toxicity and irreversibility.3 Over approximately a decade from the mid-1980s, the task force developed new esters of testosterone with the hope that they would be developed by industry partners for use in treatment of hypogonadal patients and as male contraceptives. However, this did not come to fruition, for reasons including concerns in industry about security of patents, possibility of litigation, perception of a ‘small market’ and competition with their existing female contraceptives. Nevertheless, landmark contraceptive efficacy studies were conducted at this time using weekly injections of testosterone enanthate. The first of these demonstrated that most men (70%) became azoospermic when administered 200 mg/week of testosterone enanthate, and couples then used this as their exclusive method of contraception for 12 months. This proved to be a highly effective contraceptive, with only 0.8 pregnancies per 100 person-years of exposure.4 A second study extended this to ask ‘how low does a sperm count need to go?’, and used 3 million/mL as the cut-off for entry to the efficacy phase. This increased the response rate (92% of men achieved sperm counts below this threshold and entered the efficacy phase), and again excellent efficacy was demonstrated with a Pearl Index of 1.4 (95% CI 0.4 to 3.7).5 Although a Pearl Index of 8.1 (95% CI 2.2 to 20.7) was achieved even in men who were not azoospermic, it was subsequently decided that <1 million/mL was the appropriate sperm count cut-off for future contraceptive efficacy studies.6

ain excellent efficacy was demonstrated with a Pearl Index of 1.4 (95% CI 0.4 to 3.7).5 Although a Pearl Index of 8.1 (95% CI 2.2 to 20.7) was achieved even in men who were not azoospermic, it was subsequently decided that <1 million/mL was the appropriate sperm count cut-off for future contraceptive efficacy studies.6 While these studies demonstrated that hormonal contraception using testosterone was a real possibility for men, subsequent studies have focused on a combination with progestogens, which are potent gonadotrophin suppressors in men as they are in women. In combination, they allow a much lower dose of testosterone to be used, essentially using the progestogen to do the gonadotrophin suppression, with approximately physiological doses of testosterone providing replacement for the suppressed Leydig cell function. The 1990s saw a lot of activity exploring a range of progestogens, although generally in small studies with spermatogenic suppression as the end point, rather than contraceptive efficacy.7 A key limitation at that time was the absence of a long-acting testosterone preparation, although testosterone pellets were used in studies, predominantly in the UK and Australia. In prototype long-acting reversible contraceptive (LARC) studies, testosterone pellets were combined with etonogestrel implants, showing excellent spermatogenic suppression,8 and in combination with depot medoxyprogesterone acetate (DMPA) were used in an efficacy study, with no pregnancies in 55 couples over 35.5 person-years of exposure.9

ing reversible contraceptive (LARC) studies, testosterone pellets were combined with etonogestrel implants, showing excellent spermatogenic suppression,8 and in combination with depot medoxyprogesterone acetate (DMPA) were used in an efficacy study, with no pregnancies in 55 couples over 35.5 person-years of exposure.9 The development of long-acting testosterone undecanoate (TU) by Schering led to a collaborative study with Organon, which developed specific etonogestrel implants for men for a trial of that combination, confirming effective spermatogenic suppression.10 Unfortunately, this industry involvement was short-lived due to changing priorities in both companies, and there are no commercially funded active studies at present. In China, large studies were also conducted using TU alone.11 The combination approach was developed into an efficacy study by WHO and CONRAD (Contraception Research and Development network), using TU with norethisterone enanthate, both given at 8-week intervals (with the anticipation of a single combination injection being developed). This study recruited over 300 couples internationally, with 96.8% of men achieving sufficient spermatogenic suppression to enter the 1-year efficacy phase. While the trial was stopped early by a WHO review panel due to concern over side effects (despite very few men discontinuing treatment), there were just four pregnancies, giving a contraceptive efficacy of 1.59% (CI 0.6 to 4.2),12 thus matching hormonal female methods and substantially better than condoms, the only current reversible male method.

early by a WHO review panel due to concern over side effects (despite very few men discontinuing treatment), there were just four pregnancies, giving a contraceptive efficacy of 1.59% (CI 0.6 to 4.2),12 thus matching hormonal female methods and substantially better than condoms, the only current reversible male method. The essential role of GnRH in regulating reproductive function makes it an obvious target for study, since the early days of development of GnRH agonists in reproductive medicine. Both agonists and antagonists have the potential to act as contraceptives. GnRH agonists do not result in sufficient gonadotrophin suppression in men, but GnRH antagonists are very effective and there have been several studies that have investigated GnRH antagonists (with testosterone); however, these did not reduce sperm concentrations to a greater degree than testosterone-progestogen combinations.13 GnRH antagonists have only been available in parenteral preparations requiring serial injections or infusions, which have practicality and cost implications. New oral GnRH antagonists are becoming available, developed for the treatment of endometriosis,14 and may have a role to play as part of a patient-controlled contraceptive in the future.

ly been available in parenteral preparations requiring serial injections or infusions, which have practicality and cost implications. New oral GnRH antagonists are becoming available, developed for the treatment of endometriosis,14 and may have a role to play as part of a patient-controlled contraceptive in the future. Current and future hormonal approaches Testosterone and Nestorone® (segesterone acetate) transdermal gel A nestorone-testosterone gel (NES/T) has recently entered an international phase IIb clinical trial to establish efficacy and side effect frequency, supported by the US National Institute of Child Health and Human Development and the Population Council. Nestorone® is a novel potent progestogen with minimal activity at the androgen receptor, also licensed by the Food and Drug Administration as a combined vaginal ring for female contraception, not yet available in the UK. This is a patient-controlled, daily-dose method of combined progestogen-testosterone hormonal contraception. Minimal side effects are anticipated due to the relatively low dose of testosterone and progestogen delivered.15 In the development of this approach, NES/T was shown to suppress FSH and LH to less than 1 IU/L (a level consistent with achieving contraceptive effectiveness in other trials) more consistently and effectively than testosterone gel alone. There was a significant decrease in sperm concentrations in the NES/T group despite the duration of treatment being only 4 weeks, and no differences in psychosexual measures between groups or from baseline.16

ing contraceptive effectiveness in other trials) more consistently and effectively than testosterone gel alone. There was a significant decrease in sperm concentrations in the NES/T group despite the duration of treatment being only 4 weeks, and no differences in psychosexual measures between groups or from baseline.16 Kisspeptin-based HPT suppression Kisspeptin is a compound recently discovered to play a significant role in the regulation of GnRH release by the hypothalamus. The kisspeptin-neurokinin B-dynorphin pathway acts on the hypothalamus to upregulate and downregulate the release of GnRH.17 It has potential to be used both in patients with hypogonadism to increase GnRH and sex steroid production, and conversely to suppress GnRH production and thus the HPT axis as a potential contraceptive,18 but we are not aware of any current trials related to male contraception.

e and downregulate the release of GnRH.17 It has potential to be used both in patients with hypogonadism to increase GnRH and sex steroid production, and conversely to suppress GnRH production and thus the HPT axis as a potential contraceptive,18 but we are not aware of any current trials related to male contraception. Synthetic androgen-progestogen compounds and selective androgen receptor modulators Several chemical compounds have been developed that exhibit both androgenic and progestogenic effects and represent potential male contraceptives, combining both aspects of what is currently the leading hormonal approach and might also be useful for hypogonadism. There are various ongoing studies to develop these drugs, both as implantable and oral contraceptives. Recently, phase I trials of two oral preparations, 11-beta-methyl-19-nortestosterone dodecylcarbonate19 and dimethandrolone undecanoate,20 have been conducted showing gonadotrophin suppression and therefore promise as potential male contraceptive pills, although at a very early stage. 7alpha-methyl-19-nortestosterone (MENT) is a potent androgen which does not undergo 5-alpha reduction but is aromatised; this combination may even allow protective effects against prostate disease, while supporting sexual function and bone density.21 When administered as an implant (because of rapid clearance), MENT can provide good gonadotrophin and spermatogenic suppression.22

which does not undergo 5-alpha reduction but is aromatised; this combination may even allow protective effects against prostate disease, while supporting sexual function and bone density.21 When administered as an implant (because of rapid clearance), MENT can provide good gonadotrophin and spermatogenic suppression.22 Non-hormonal approaches acting on spermatogenesis Bisdichloroacetyldiamines Conversion of vitamin A to retinoic acid in the testis is essential for spermatogenesis, as retinoic acid is essential for initiation of meiosis. Meiosis is a potentially attractive target for contraception, as its only function is the production of haploid gametes; thus, sufficient specificity should reduce the likelihood of off-target effects and not interfere with the endocrine function of the testis. Conversion from vitamin A is mediated by alcohol and aldehyde dehydrogenases, which can be reversibly inhibited by administration of bisdichloroacetyldiamines. To date this has been successfully used to induce azoospermia in rabbits, which reversed following withdrawal of the compound, but unfortunately in humans the drug induces an ‘antabuse’-type reaction. Further studies, including the development of agents without this side effect, are warranted as the method may induce azoospermia more rapidly than hormonal methods.23

e azoospermia in rabbits, which reversed following withdrawal of the compound, but unfortunately in humans the drug induces an ‘antabuse’-type reaction. Further studies, including the development of agents without this side effect, are warranted as the method may induce azoospermia more rapidly than hormonal methods.23 Bromodomain testis-specific protein inhibitors Bromodomain testis-specific proteins (BRDTs) are expressed during the development of maturing spermatocytes and are important in the remodelling of cellular chromatin. A BRDT inhibitor has been shown to induce reversible infertility in mice without affecting serum testosterone levels or copulatory behaviour.24 Further development from this important proof of concept may be facilitated by the potential of related drugs in oncology. Lonidamine derivatives Lonidamine was discovered in the 1970s and is a non-hormonal drug that is known to have antispermatogenic characteristics; however, it also had significant side effects, including testicular pain and liver dysfunction. Derivatives of lonidamine, including adjudin and gamendazole, have fewer side effects and act by causing premature detachment of spermatids from the Sertoli cells and subsequent infertility.25 Unfortunately, in animal studies, doses that could give effective contraception either resulted in liver side effects or irreversibility. However, further studies are in progress to modify these compounds to improve reversibility and reduce adverse effects.26

rmatids from the Sertoli cells and subsequent infertility.25 Unfortunately, in animal studies, doses that could give effective contraception either resulted in liver side effects or irreversibility. However, further studies are in progress to modify these compounds to improve reversibility and reduce adverse effects.26 Thermal treatment Transient mild increases in testicular temperature can result in reduction in sperm count particularly when paired with a hormone administration, but is inferior to combined testosterone and progestogen treatment.27 There are some small communities of men practising ‘do-it-yourself’ thermal treatments, either using a hot bath method or with a modified form of underwear to create ‘artificial cryptorchidism’. There is not much current active research; however, a recent French acceptability study reported that one-third of respondents showed interest in this method as a contraceptive.28

-it-yourself’ thermal treatments, either using a hot bath method or with a modified form of underwear to create ‘artificial cryptorchidism’. There is not much current active research; however, a recent French acceptability study reported that one-third of respondents showed interest in this method as a contraceptive.28 Non-hormonal approaches acting on the epididymis The epididymis is responsible for concentrating sperm within the seminiferous fluid and conditioning the lipids and proteins on the surface of the sperm. This process takes several weeks and is essential for maturation of the sperm to improve their functionality. There are several areas currently being explored as potential sites of intervention for a male contraceptive. There are proteins, for example Eppin,29 expressed in the epididymis (and nowhere else in the body) that could be targeted with drugs that disrupt their functions. As these proteins are only expressed in the reproductive tract, there should theoretically be few side effects to these methods.29

for a male contraceptive. There are proteins, for example Eppin,29 expressed in the epididymis (and nowhere else in the body) that could be targeted with drugs that disrupt their functions. As these proteins are only expressed in the reproductive tract, there should theoretically be few side effects to these methods.29 Non-hormonal approaches acting on the vas Intravasal agents The reversible inhibition of sperm under guidance technique involves the injection of a polymer into the vas, which then acts to destabilise the cell membrane of sperm that pass through it, rendering them non-viable. As there is not a complete barrier to fluid passage along the vas, issues relating to back-pressure are avoided. The method can later be reversed either by mechanically massaging the polymerised section of the vas to dislodge the polymer or by reinjection with a dissolving agent. The technique was originally developed in India,30 and a related compound is currently undergoing animal trials in the USA under the brand name Vasalgel. Following successful trials in non-human primates,31 further preclinical studies are under way.

vas to dislodge the polymer or by reinjection with a dissolving agent. The technique was originally developed in India,30 and a related compound is currently undergoing animal trials in the USA under the brand name Vasalgel. Following successful trials in non-human primates,31 further preclinical studies are under way. Adrenoreceptor antagonists Transport of sperm from the epididymis along the vas is mediated by smooth muscle contractility. Alpha-1-adrenoreceptor antagonists, such as tamsulosin, inhibit this contractility, and patients treated with these drugs for other reasons demonstrate reduced sperm content in semen.32 The contraceptive potential of these drugs has been explored only in small studies, which have shown that anejaculation can be generated at doses of 0.8 mg, but this had side effects including transient discomfort lasting up to 10 hours. At a lower dose of 0.4 mg, side effects were reduced, but higher numbers of functional sperm were found in seminal fluid. There is potential that this or a similar compound could be used as an ‘on-demand’ male contraceptive as the effect of these drugs is transient.

ncluding transient discomfort lasting up to 10 hours. At a lower dose of 0.4 mg, side effects were reduced, but higher numbers of functional sperm were found in seminal fluid. There is potential that this or a similar compound could be used as an ‘on-demand’ male contraceptive as the effect of these drugs is transient. Barrier methods Barrier methods will, for the foreseeable future, have an important role in infection prevention. Novel materials and production methods are increasing the diversity of condoms available on the market with an aim to improve sensation while retaining safety and efficacy. Funding from a charitable organisation has been used to develop a self-lubricating condom that may improve pleasure and reduce discomfort, as well as reduce friction-related damage to the condom.33 Microchip delivery Novel methods of delivery of existing compounds may be possible. Microchip technology to deliver progestogen-only contraception for women is currently in development and has been used for delivery of parathyroid hormone.34 This format would be a truly LARC method and could overcome the difficulties in dosing frequency for testosterone alone or with a progestogen. Will novel male methods meet the general principles of contraceptive development? There are five domains broadly sought after in a novel contraceptive method35: Contraceptives must have high efficacy, both in controlled settings and in real-world use.

Microchip delivery Novel methods of delivery of existing compounds may be possible. Microchip technology to deliver progestogen-only contraception for women is currently in development and has been used for delivery of parathyroid hormone.34 This format would be a truly LARC method and could overcome the difficulties in dosing frequency for testosterone alone or with a progestogen. Will novel male methods meet the general principles of contraceptive development? There are five domains broadly sought after in a novel contraceptive method35: Contraceptives must have high efficacy, both in controlled settings and in real-world use. The contraceptive effect must be reversible: while there may be some utility to alternative methods of permanent contraception, safe effective methods of achieving this already exist and the gap in the market is for true, on-demand reversibility. Speed of action: the time taken from initiating the contraceptive until the method offers protection from pregnancy should be as short as possible to prevent unintended pregnancy. Ease of use: novel contraceptives should be simple to use, whether they are patient-initiated (pills and gels) or clinician-initiated (injections and implants), with clear ‘missed pill’ rules. This also includes issues of accessibility. Safety: side effects should be minimal to reduce discontinuation, particularly with regard to secondary sex characteristics, sexual function and sexual pleasure. Non-contraceptive benefits should be maximised and should not be worse than no contraception.

Ease of use: novel contraceptives should be simple to use, whether they are patient-initiated (pills and gels) or clinician-initiated (injections and implants), with clear ‘missed pill’ rules. This also includes issues of accessibility. Safety: side effects should be minimal to reduce discontinuation, particularly with regard to secondary sex characteristics, sexual function and sexual pleasure. Non-contraceptive benefits should be maximised and should not be worse than no contraception. It is clear that fulfilling all of these domains is closer than ever before for some of the approaches discussed above. Efficacy of currently studied methods looks to be high, with low rates of residual spermatogenesis and low rates of pregnancy. Likewise, reversibility following discontinuation appears complete and treatment modalities are familiar and easy to use (injections, gels and pills). Speed of action is where this approach falls down slightly—hormonal methods currently being investigated require several weeks of use before they can be relied on, although this is similar to vasectomy. Male hormonal methods also appear to be largely safe, being based on steroids that have been used in large numbers of people over many years, without detriment to secondary sex characteristics. Hormonal methods can have a negative effect on cholesterol, particularly decreases in high-density lipoprotein cholesterol, which may play a role in heart disease risk, but this must be interpreted with caution as it may largely reflect historical approaches based on high doses of testosterone, and risk of cardiovascular disease is complex and multifactorial. There may be positive non-contraceptive benefits of these drugs in terms of prostate function, muscle mass and bone density that will not be understood until there are longer-term, population-level data, and the use of novel steroids may increase these benefits.

of cardiovascular disease is complex and multifactorial. There may be positive non-contraceptive benefits of these drugs in terms of prostate function, muscle mass and bone density that will not be understood until there are longer-term, population-level data, and the use of novel steroids may increase these benefits. Social probabilities Ample evidence exists to show that high proportions of men find the idea of a male contraceptive acceptable and that women would rely on their male partners to use a male contraceptive method.36 It is important to remember that many women rely on their male partner for contraception already—approximately 54% of women in the UK use either male sterilisation, male condom or withdrawal method as their main method of contraception.37 Thus, men want to engage in reproductive health, but their current options are limited compared with the range of contraceptive choices available to women. Male hormonal contraceptive methods do not currently exist on the market anywhere globally and so we cannot base our assumptions on future use of these methods on how current methods are used. Men may be attracted to male hormonal contraceptives for non-contraceptive benefits, similar to the familiar non-contraceptive benefits of female hormonal methods, and this is an aspect that can potentially be ‘engineered’ into methods based on novel steroids.

uture use of these methods on how current methods are used. Men may be attracted to male hormonal contraceptives for non-contraceptive benefits, similar to the familiar non-contraceptive benefits of female hormonal methods, and this is an aspect that can potentially be ‘engineered’ into methods based on novel steroids. Challenges for services In the current research paradigm, male contraception development involves frequent semen analysis to confirm azoospermia/oligospermia prior to relying on the method as a contraceptive. During later phase studies and then as male contraceptives enter clinical practice, the focus will need to shift from intensive monitoring of sperm concentrations and more to general rules of use. However, semen analysis remains in routine clinical practice for confirming vasectomy efficacy, and home testing for spermatogenic suppression may also be a possibility with several companies developing such technology in relation to infertility.38

ve monitoring of sperm concentrations and more to general rules of use. However, semen analysis remains in routine clinical practice for confirming vasectomy efficacy, and home testing for spermatogenic suppression may also be a possibility with several companies developing such technology in relation to infertility.38 Summary More diverse methods of male contraceptives are needed to help meet the unmet contraceptive need of men and women everywhere. Research into these methods has been ongoing since the 1970s, but development has been hindered by a wide range of factors: challenges for funding of male methods and lack of industry involvement, disbelief that there is a real need for novel male methods, slow development of novel testosterone preparations, and the underlying physiology whereby men produce millions of sperm per day rather than releasing one oocyte per month. Fortunately, we are closer than ever to new hormonal methods coming to the market, and while non-hormonal methods seem at present more distant they have the potential to make a major contribution in the future. Correction notice: Since this article was first published online the middle intial A has been added to the author Richard Anderson. Contributors: JJR-W and RA contributed equally to the manuscript. Funding: The authors' work in this field is supported by funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and is carried out at the MRC Centre for Reproductive Health, which is funded by MRC Centre (grant MR/N022556/1). Competing interests: None declared.

Contributors: JJR-W and RA contributed equally to the manuscript. Funding: The authors' work in this field is supported by funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and is carried out at the MRC Centre for Reproductive Health, which is funded by MRC Centre (grant MR/N022556/1). Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

The Faculty of Sexual & Reproductive Healthcare (FSRH) Clinical Effectiveness Unit is developing a guideline looking at contraceptive options for women with weight issues. We hope that this guideline will include information for those facing bariatric surgery as almost 80% of women requesting this procedure are in their reproductive years.1 These women are also advised to avoid pregnancy for up to 24 months following surgery, making effective, reversible contraception an ideal choice.2 With this in mind, we therefore asked women aged between 18 and 50 years on a bariatric surgery waiting list to complete a voluntary, anonymous online survey about their sexual and reproductive health. Ethical approval was granted by the National Health Service, University of Sunderland and City Hospitals Sunderland NHS Foundation Trust Research Ethics Committees. There were 42 responders with the majority (38%, n=16) aged between 35 and 44 years old, 92% (n=38) were heterosexual and 71% (n=30) had children. All participants (n=42) described their ethnic origin as white with an average body mass index (BMI) of 42 kg/m2.

ot want to use a condom, in a relationship, who usually makes decisions on condom use; (6) Fertility intention and dual protection: do you want more children in the future; did you or your partner use contraceptive (other than condoms) at last intercourse, if partner is using contraceptive, do they need to use condoms. Bivariate associations were determined using chi-squared tests for categorical variables and t-tests for continuous variables. Variables where the majority (≥95% of total study cohort) were within one of the categories were not maintained in the analysis. For continuous variables, we examined the data using categories based on median values, natural breaks, or common categories from similar studies. For these variables, where categorical evaluation did not reveal any nonlinear associations, the variable remained continuous for the analysis. We conducted separate multivariate analyses for women and men using Poisson regression with robust standard errors to estimate the adjusted prevalence ratios (aPR) and 95% confidence intervals (CIs) for variables associated with our primary outcomes. Variables that were significant at P<0.10 in bivariate analysis were included in each multivariate model. Table 1 Demographic, behavioural characteristics and beliefs among study participants by gender

We conducted separate multivariate analyses for women and men using Poisson regression with robust standard errors to estimate the adjusted prevalence ratios (aPR) and 95% confidence intervals (CIs) for variables associated with our primary outcomes. Variables that were significant at P<0.10 in bivariate analysis were included in each multivariate model. Table 1 Demographic, behavioural characteristics and beliefs among study participants by gender Characteristics Female (n=255) Male (n=220) P value* n (%) n (%) Age (years, mean±SD) 32.1±6.36 37.40±5.19 <0.001 Martial status 0.087 Married/committed relationship 249 (97.6) 219 (99.6) Single/dating 6 (2.4) 1 (0.4) Education 0.011 Some secondary or less 219 (85.9) 169 (76.8) Completed secondary or more 36 (14.1) 51 (23.2) Religion 0.264 Catholic 50 (19.7) 47 (21.5) Protestant 186 (73.2) 152 (69.4) Muslim 18 (7.1) 17 (7.8) Other/no religion 0 (0) 3 (1.4) Years since diagnosis (mean±SD) 4.64+3.45 4.62±4.14 0.949 Use of alcohol or drugs before/during sex in last month <0.001 No 229 (92.3) 161 (73.2) Yes 19 (7.7) 59 (26.8) Most recent sexual partner HIV-positive 0.004 No 47 (18.5) 40 (18.2) Yes 163 (64.2) 164 (74.6) Do not know 44 (17.3) 16 (7.3) Most recent sexual partner has sex with other people <0.001 No 81 (31.8) 141 (64.4) Yes 101 (39.6) 30 (13.7) Do not know 73 (28.6) 48 (21.9) Sexual frequency in last month 0.658 ≤weekly 120 (47.1) 108 (49.1) >1x/week 135 (52.9) 112 (50.9) Sexual partners in past month (n) 0.073 1 248 (97.2) 208 (94.6) 2–4 4 (1.6) 11 (5.0) >5 3 (1.2) 1 (0.4) Children (n) 0.006 0 14 (5.5) 6 (2.7) 1–2 123 (48.2) 81 (36.8) >3 118 (46.3) 133 (60.5) Desire more children in the future 0.010 No 160 (62.8) 161 (73.9) Yes 95 (37.2) 57 (26.1) Contraceptive use at last intercourse (IUD, implant, DMPA, pills, sterilisation, not including condom) 0.202 No 140 (55.6) 135 (61.4) Yes 112 (44.4) 85 (38.6) Currently using ART 0.846 No 29 (11.4) 26 (11.8) Yes, <2 years 86 (33.7) 79 (35.9) Yes, >2 years 140 (54.9) 115 (52.3) Believe HIV medications effect transmission risk to partner 0.577 Increases risk or no change 166 (65.6) 149 (68.0) Decreases risk 87 (34.4) 70 (32.0) Believe condoms are effective at preventing pregnancy 0.487 No 28 (11.0) 20 (9.1) Yes 226 (89.0) 200 (90.0) Believe condoms prevent HIV transmission to partner 0.057 No 41 (16.7) 50 (23.8) Yes 205 (83.3) 160 (76.2) Believe one needs to use a condom if using also birth control method 0.064 No 69 (27.3) 44 (20.0) Yes 184 (72.7) 176 (80.0) Can refuse sex if partner did not want to use a condom <0.001 No 127 (50.6) 64 (29.4) Yes 124 (49.4) 154 (70.6) Decision-making about using condom use 0.11

With this in mind, we therefore asked women aged between 18 and 50 years on a bariatric surgery waiting list to complete a voluntary, anonymous online survey about their sexual and reproductive health. Ethical approval was granted by the National Health Service, University of Sunderland and City Hospitals Sunderland NHS Foundation Trust Research Ethics Committees. There were 42 responders with the majority (38%, n=16) aged between 35 and 44 years old, 92% (n=38) were heterosexual and 71% (n=30) had children. All participants (n=42) described their ethnic origin as white with an average body mass index (BMI) of 42 kg/m2. Menstrual cycles were described as regular in half the participants (n=21), and 21% (n=9) reported the average length of bleeding per cycle to be 3–4 days. Nearly half of the respondents experienced heavy menstrual bleeding (HMB) (n=19), defined as needing frequent changes of sanitary protection (pads or tampons) during the day and/or night. Painful periods (dysmenorrhoea) were reported by 60% (n=25). Medication for dysmenorrhoea was taken by 39% (n=12). Bleeding in between periods was experienced by 12% (n=5). Contraception was used by either the woman or her partner in 76% (n=32) of participants, with 5% (n=2) stating that they were not in a relationship. Over half the participants (55%, n=23) were using contraception, with the hormonal intrauterine system (IUS) (n=8) being the most common method, followed by the progestogen implant (n=5) and progestogen-only pill (n=3) (table 1). Table 1 Pre-surgical contraception participant information

Contraception was used by either the woman or her partner in 76% (n=32) of participants, with 5% (n=2) stating that they were not in a relationship. Over half the participants (55%, n=23) were using contraception, with the hormonal intrauterine system (IUS) (n=8) being the most common method, followed by the progestogen implant (n=5) and progestogen-only pill (n=3) (table 1). Table 1 Pre-surgical contraception participant information Question Response % n Do you use contraception at present? Yes 55 23 No 24 10 In a same sex relationship 2 1 No, partner uses condoms 21 9 Not in a relationship 5 2 Do you suffer from or undertake any of the following that may affect your contraceptive choices? Smoking 0 0 History of deep vein thrombosis 5 2 History of heart disease 5 2 Migraine 10 4 Did not answer 83 35 Type of contraception used Progestogen-only pill 7 3 Progestogen implant (eg, Nexplanon) 12 5 Progestogen injection (eg, Depo-Provera) 5 2 Intrauterine device – copper 5 2 Intrauterine device – hormonal (eg, Mirena or Jaydess) 19 8 Condoms – male Female sterilisation 5 2 Male sterilisation 5 2 None 12 5 Did not answer 5 2 26 11 Are there contraceptive methods you have been told you cannot use? Yes 24 10 No 66 28 Did not answer 9 4 Reasons given for not using contraceptive methods (optional response) Irregular cycles 1 Lack of efficacy 1 Familial history of deep vein thrombosis 3 Overweight 3 High blood pressure 2 Who do you speak to about contraception? General practitioner 42 18 Practice nurse 26 11 Family planning clinic 26 11 Gynaecologist 5 2 Genitourinary medicine clinic 2 1 Pharmacist 0 0 Other 0 0 Did not respond 14 6 Menstrual data from participants using hormonal contraception and intrauterine systems (IUSs) or devices (IUDs) (n=20) showed low levels of HMB and high rates of amenorrhoea (table 2). Polycystic ovary syndrome (PCOS) had been diagnosed in 16% (n=7) of the respondents.

2 1 Pharmacist 0 0 Other 0 0 Did not respond 14 6 Menstrual data from participants using hormonal contraception and intrauterine systems (IUSs) or devices (IUDs) (n=20) showed low levels of HMB and high rates of amenorrhoea (table 2). Polycystic ovary syndrome (PCOS) had been diagnosed in 16% (n=7) of the respondents. Table 2 Menstrual patterns in participants using hormonal contraceptive and intrauterine devices

2 1 Pharmacist 0 0 Other 0 0 Did not respond 14 6 Menstrual data from participants using hormonal contraception and intrauterine systems (IUSs) or devices (IUDs) (n=20) showed low levels of HMB and high rates of amenorrhoea (table 2). Polycystic ovary syndrome (PCOS) had been diagnosed in 16% (n=7) of the respondents. Table 2 Menstrual patterns in participants using hormonal contraceptive and intrauterine devices Contraceptive used (n=20) Description of menstrual cycle Average length of bleed Self-reported heavy menstrual bleeding Intrauterine systems (hormonal) (n=8) Regular 4–5 days No Irregular 1–2 days No Irregular Varies Varies Irregular Amenorrhoeic No Irregular Amenorrhoeic No Irregular Amenorrhoeic No Irregular Amenorrhoeic No Irregular Amenorrhoeic No Intrauterine devices (copper) (n=2) Irregular Amenorrhoeic No Irregular 3–4 days No Progestogen implants (n=5) Irregular Amenorrhoeic No Irregular Amenorrhoeic No Irregular Amenorrhoeic No Irregular 4–5 days No Irregular 4–5 days No Progestogen-only pill (n=3) Regular Amenorrhoeic No Regular Amenorrhoeic No Irregular Amenorrhoeic No Progestogen injections (n=2) Irregular Amenorrhoeic No This is the first British survey looking at menstrual bleeding patterns and contraceptive use in obese women before bariatric surgery. No woman was using a contraceptive method where there were safety issues, although 23% who reported being in a heterosexual relationship were not using contraception. This suggests that the UK Medical Eligibility Criteria for Contraceptive Use guideline was being followed, and just three women would need to choose an alternative non-oral method after bariatric surgery.3 The updated US Medical Eligibility for Contraceptive Use do not recommend oral hormonal methods, citing the potential to decrease contraceptive effectiveness associated with the procedure and postoperative complications such as long-term diarrhoea and/or vomiting.4 The reported incidence of PCOS was similar to data found in other studies investigating overweight and obese women. A survey of 563 Swedish women found that prior to surgery, the most common methods were hormonal IUDs (16%, n=86), followed by progestogen-only pills (15%, n=85), and copper IUDs (14%, n=77), with 32% (n=182) reporting not using any contraception, but this did not specify whether this was inclusive of male or female sterilisation5

563 Swedish women found that prior to surgery, the most common methods were hormonal IUDs (16%, n=86), followed by progestogen-only pills (15%, n=85), and copper IUDs (14%, n=77), with 32% (n=182) reporting not using any contraception, but this did not specify whether this was inclusive of male or female sterilisation5 When looking at the menstrual data for women prior to bariatric surgery our findings showed similar levels of dysmenorrhoea in our pre-bariatric surgical cohort when compared with the general population of a similar age.6 Hormonal IUSsystems offer protection against dysmenorrhoea, endometrial hyperplasia, pelvic pain and HMB, in addition to providing contraceptive protection, which may be attributed to their high use in our cohort. Further research is needed to investigate the menstrual changes that occur after bariatric surgery and the effects of gastric bypass procedures on oral hormonal absorption. Contributors: All authors contributed to the writing of this paper, have seen the final version and approve of it. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: No, there are no competing interests for any author. Patient consent: All data was anonymised. No personal information is given in this paper. Ethics approval: NHS and University of Sunderland. Provenance and peer review: Not commissioned; externally peer reviewed.

Governments, advocates, providers, policymakers and other stakeholders who want to fully support women’s rights to access abortion across the globe must address quality of care, in addition to efforts to change abortion laws, train providers and expand service provision. Documenting and working to improve the quality of abortion care is necessary in order to improve service delivery and health outcomes, expand access to safe abortion especially in legally restricted settings, and to ensure the human right to the highest attainable standard of health, as outlined by the WHO.1 Quality of healthcare services is the degree to which services produce desired health outcomes and rely on best available evidence.2 Domains of quality as defined by the WHO3 and the Institute of Medicine (IOM)2 ask whether healthcare is effective, efficient, accessible, acceptable/patient-centred, equitable and safe.

Quality of healthcare services is the degree to which services produce desired health outcomes and rely on best available evidence.2 Domains of quality as defined by the WHO3 and the Institute of Medicine (IOM)2 ask whether healthcare is effective, efficient, accessible, acceptable/patient-centred, equitable and safe. The safety of abortion is well established in settings where abortion is legal,4 despite claims to the contrary from those who seek to restrict access to abortion. A safe abortion is understood to mean a medical (medication), aspiration, or surgical abortion that conforms to WHO guidelines.5 Well-managed self-use of medical abortion is on the spectrum of safe abortion. In settings with high-quality clinical data and access to legal abortion such as the United States, first-trimester abortions carry an extremely low risk of death (0.3–0.5/100 000 abortions) which has remained stable over time. Risk increases in the second trimester (2.5–6.7 deaths per 100 000 procedures for 14–17 and >18 weeks, respectively).4 Major complications are also rare; ≤0.1% of first-trimester abortions4 and up to 0.23% of abortions when all gestational ages are included.4 Abortion in any trimester is 14-fold safer than childbirth in a high-income setting,6 and safer than many other common healthcare procedures. However, quality abortion care includes, but is not limited to, safety.

≤0.1% of first-trimester abortions4 and up to 0.23% of abortions when all gestational ages are included.4 Abortion in any trimester is 14-fold safer than childbirth in a high-income setting,6 and safer than many other common healthcare procedures. However, quality abortion care includes, but is not limited to, safety. In the quality of care literature, quality is further categorised as interpersonal quality, or the healthcare interaction, and technical quality, whether providers meet normative standards for appropriate care or evidence-based criteria. Patient or client experience can assess both interpersonal and technical quality, measures the client-centeredness of care, and “includes any process observable by patients, including subjective experiences (eg, pain was controlled), objective experiences (eg, waited more than 15 minutes past appointment time), and observations of physician, nurse, or staff behaviour (eg, doctor provided all relevant information)”.7 It is important to note that client experience reports are distinct from ‘satisfaction’ ratings in that they reflect specific components of the care experience in contrast to a global rating.

st appointment time), and observations of physician, nurse, or staff behaviour (eg, doctor provided all relevant information)”.7 It is important to note that client experience reports are distinct from ‘satisfaction’ ratings in that they reflect specific components of the care experience in contrast to a global rating. Patient or client experience has been shown to be linked with other elements of quality of in-hospital care, including clinical processes and structures, effectiveness, efficiency and safety.7 8 Client experience also impacts behaviours such as adherence, follow-up decisions and behaviour change.7 Most literature that seeks to describe the relationship between experience of care, domains of quality of care, and health or behavioural outcomes comes from other fields; it is not known whether these relationships hold in abortion care.

impacts behaviours such as adherence, follow-up decisions and behaviour change.7 Most literature that seeks to describe the relationship between experience of care, domains of quality of care, and health or behavioural outcomes comes from other fields; it is not known whether these relationships hold in abortion care. Evidence exists to guide clinical practice in abortion.9 However, assessment of clinical practice remains unstandardised, and very little evidence exists documenting client perceptions of both technical or interpersonal quality, especially from low- and middle-income country settings. Women who receive needed abortion services are nearly universally satisfied, but this may be due to having received a needed service and tells us little about where quality can be improved.10 The limited evidence available suggests that similar elements of care are as important to abortion clients as to patients seeking other types of healthcare: wait times, pain management and experience of pain, and treatment by staff and providers. More work is needed to advance a set of evidence-based, validated quality metrics that help us identify which aspects of the client experience – both technical and interpersonal – are most important to women seeking care and most likely to impact other domains of quality such as effectiveness and safety.

and providers. More work is needed to advance a set of evidence-based, validated quality metrics that help us identify which aspects of the client experience – both technical and interpersonal – are most important to women seeking care and most likely to impact other domains of quality such as effectiveness and safety. Beyond observed relationships between quality of care and health outcomes, quality care is important because it is a human right. All people have the right to demand quality healthcare services from providers, organisations, and ultimately their governments. The rights-based approach explicitly acknowledges that health services provide an opportunity to address stigma and gender bias, and support people to know and achieve their human rights. A rights-based perspective also incorporates informed choice and extends beyond clients and includes providers, who have the right to have the training, supplies and respect they need to do their jobs well.

vide an opportunity to address stigma and gender bias, and support people to know and achieve their human rights. A rights-based perspective also incorporates informed choice and extends beyond clients and includes providers, who have the right to have the training, supplies and respect they need to do their jobs well. Despite advances in development of, and access to, safe abortion technologies, measuring and assuring access to high-quality care needs more attention. To date, there is no standardised, validated set of quality metrics for abortion. Previous efforts to synthesise current indicators focus on the type of information available;10 the next steps are to align these information types by elements of quality and by interpersonal and technical quality, guided by established frameworks that permit comparisons across settings and clinical areas. A validated and widely used set of metrics that capture all domains of quality, including the client experience both within and outside the formal healthcare system, would enable us to assess the safety of services, identify areas for improvement, compare abortion services with other types of services (which may be found lacking in comparison), and highlight the work of providers, administrators and advocates worldwide. Part of normalising abortion services within healthcare systems is establishing a framework for evaluating quality as we do for other core services such as antenatal care and vaccines. We lack both common terminology and measures to assess abortion services across diverse health system settings, especially in low- and middle-income countries. Such measures would allow us to build evidence about the effectiveness, efficiency, accessibility, patient-centered-ness, equity and safety of abortion services, and ultimately to improve abortion care for women across the globe.

ces across diverse health system settings, especially in low- and middle-income countries. Such measures would allow us to build evidence about the effectiveness, efficiency, accessibility, patient-centered-ness, equity and safety of abortion services, and ultimately to improve abortion care for women across the globe. Contributors: NK, BP and BGD conceived the idea; BGD drafted the manuscript; and NK, BP, KA, DM, NMC, SB, KB and CG provided substantive revisions to the manuscript. Funding: Partial support for this work came from the David and Lucile Packard Foundation and the William and Flora Hewlett Foundation. Competing interests: None declared. Patient consent: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Women face many barriers to safe abortion care, especially later in pregnancy. Research documenting the safety of out-of-clinic models of medication abortion provision before and after 12 weeks gestation is needed. This analysis demonstrates that out-of-clinic models could be a safe alternative for women in need of abortions beyond 12 weeks gestation in legally restrictive contexts. Introduction Even where abortion is legal, structural and institutional barriers can delay women from accessing abortion care.1–3 Research has demonstrated that young women, poor women, unmarried women, and women who live further from abortion providers have heightened risk for experiencing delays in accessing abortion care such as later pregnancy recognition, difficulty deciding, fear of judgement, lack of autonomy, and difficulty paying for and travelling to abortion services.1 2 4–7 In settings where abortion is legally restricted or illegal, such delays are compounded by the lack of legal and/or safe options, often pushing women later into pregnancy before they obtain care. While only a small proportion of all abortions occur in the second trimester,8 access to safe abortion beyond the first trimester is an integral component of comprehensive reproductive healthcare.

are compounded by the lack of legal and/or safe options, often pushing women later into pregnancy before they obtain care. While only a small proportion of all abortions occur in the second trimester,8 access to safe abortion beyond the first trimester is an integral component of comprehensive reproductive healthcare. Globally, the use of medications for abortion outside of formal health systems is on the rise.9 10 Increasingly, data from around the world suggest that women who use misoprostol to self-manage abortions can do so safely and effectively.11–13 In light of this evidence, women’s health advocates have utilised a harm-reduction model to combat mortality and morbidity from unsafe abortion by providing women with counselling and information about early medication abortion (<12 weeks' gestation) through websites, hotlines, and social media platforms.14 15 For women in restrictive settings who need services later in pregnancy, however, equivalent safe options are limited or non-existent; as a result, unsafe abortion past the first trimester disproportionately accounts for the majority of global abortion-related morbidity and mortality.8 Because medications for abortion are safe and effective in the second trimester, and are among WHO recommended methods for later abortion,16 17, some hotlines and organisations in restrictive settings have begun to provide counselling and accompaniment services beyond the first trimester for women who do not have a safe legal option.18 In these contexts, accompaniment services include providing support for women along every step of their abortion process, including but not limited to texts and calls to check in during the process, reminders, emotional support and follow-up.

s beyond the first trimester for women who do not have a safe legal option.18 In these contexts, accompaniment services include providing support for women along every step of their abortion process, including but not limited to texts and calls to check in during the process, reminders, emotional support and follow-up. WHO guidelines aimed at expanding the cadres of abortion providers do not include task-shifting recommendations for medication abortion beyond the first trimester due to lack of evidence.19 Given the barriers that women face in obtaining safe abortions in restrictive settings and the risks that accompany unsafe abortion later in pregnancy, there is a compelling need for research that documents the safety, feasibility and acceptability of existing models that expand the cadres of providers for abortion after 12 weeks gestation, including those providing information and support to women who self-manage second trimester abortions outside of formal clinic settings. This article presents an analysis of electronic client records from a safe abortion hotline in Indonesia that provides information and accompaniment services to women seeking abortion beyond the first trimester.

formation and support to women who self-manage second trimester abortions outside of formal clinic settings. This article presents an analysis of electronic client records from a safe abortion hotline in Indonesia that provides information and accompaniment services to women seeking abortion beyond the first trimester. Methods Data for this analysis come from electronic client records from a safe abortion hotline in Indonesia, where abortion is permitted only when a woman’s life is at risk, or in cases of rape or fetal anomaly. All records were anonymised prior to the hotline granting researchers access to them, and no personal identifying information was contained in any of the records. Ninety-six women contacted the hotline between July 2012 and October 2016 seeking information on medication abortion beyond 12 weeks' gestation and are included in this study. Ethical approval for this study was obtained from Allendale Investigational Review Board.

fying information was contained in any of the records. Ninety-six women contacted the hotline between July 2012 and October 2016 seeking information on medication abortion beyond 12 weeks' gestation and are included in this study. Ethical approval for this study was obtained from Allendale Investigational Review Board. The hotline provides abortion counselling and accompaniment services up until 22 weeks' gestation, and employs current evidence-based guidelines for medication abortion for pregnancies above 12 weeks' gestation: 200 mg mifepristone administered orally, followed after 12 to 48 hours by 400 µg oral misoprostol, followed by 400 µ g sublingual misoprostol every 3 hours up to a maximum of five doses.17 20 In instances where women are unable to procure mifepristone, the hotline also employs misoprostol only-regimens. None of the hotline counsellors are clinically trained, but all have undergone intensive training in medication abortion protocols and abortion counselling.

every 3 hours up to a maximum of five doses.17 20 In instances where women are unable to procure mifepristone, the hotline also employs misoprostol only-regimens. None of the hotline counsellors are clinically trained, but all have undergone intensive training in medication abortion protocols and abortion counselling. All women calling the hotline to request information about medication abortion receive standardised and compassionate pre-abortion counselling about the process and what to expect. Gestational age of the caller is ascertained by last menstrual period (LMP) or independently obtained ultrasound results as reported to the hotline by the caller. To ensure women have access to reliable and effective medications, those who are not already in possession of medications are connected with an in-country organisation that can send a package with the required dosage of mifepristone and misoprostol through the mail. Once women confirm they have reliable medications in hand, they are instructed to call their hotline counsellor to begin an intensive, gestation-specific counselling session in preparation for the abortion. The counselling session covers the following topics: how the drugs function, how to manage pain, what the products of conception will look like, management of products of conception, how to manage retained products of conception, how to recognise complication signs, potential interactions with medical personnel in case of emergency treatment seeking, how to confirm abortion completion, what to expect after the abortion, management of emotions, and prevention of future unwanted pregnancy.

onception, how to manage retained products of conception, how to recognise complication signs, potential interactions with medical personnel in case of emergency treatment seeking, how to confirm abortion completion, what to expect after the abortion, management of emotions, and prevention of future unwanted pregnancy. Once the initial counselling session is complete, women are asked to contact their counsellor by phone or text to record the time of oral administration of mifepristone, and the time of all subsequent sublingual administrations of misoprostol. Women are also asked to record the following information in a written diary: timing and description of any side effects experienced, potential warning signs of complications, timing and reported severity of cramping, timing of bleeding, descriptions of expelled products of conception, and additional medications taken for pain management, nausea or other symptoms. Women are in close touch with their counsellor throughout the abortion process and communicate the contents of the diary with their counsellor during regular phone calls. During each communication, the counsellor inputs relayed information about the abortion process into the hotline’s electronic record system. Counsellors follow-up with women for a period of 3 weeks after the start of their abortion to monitor completion and potential warning signs of complications.

regular phone calls. During each communication, the counsellor inputs relayed information about the abortion process into the hotline’s electronic record system. Counsellors follow-up with women for a period of 3 weeks after the start of their abortion to monitor completion and potential warning signs of complications. Variables were developed in keeping with protocols implemented by the hotline. In figure 1, an initial second-trimester hotline caller is defined as anyone calling the hotline seeking information about accompaniment services for second-trimester abortion. A second-trimester medication abortion client is any woman who contacted the hotline beyond 12 weeks' gestation who participates in the initial counselling session. Medication abortion attempts are classified as 1st medication abortion attempts the first time a client administers abortion medications and  2nd medication abortion attempts if, after the client has followed evidence-based protocols,17 20 the medications did not produce their intended effect and, together with their hotline counsellor, the client decides to administer a second dose of 200 mg oral mifepristone followed by 400 µg oral misoprostol and up to five doses of 400 µg sublingual misoprostol. An abortion is characterised as a failed abortion if the abortion medications did not produce their desired effect (no expulsion of the products of conception within 24 hours after the first misoprostol dose). An abortion is characterised as a complete abortion if both the client and the counsellor reported that the placenta has been expelled, the woman is in a stable condition, and the pregnancy has been successfully terminated using abortion medications. A client is categorised as having sought care if, for any reason during the medication abortion process, she decides to seek medical care in a formal health facility. Any medical care directly related to the medication abortion obtained by a medication abortion client at a formal health facility is categorised as treatment received. In the absence of clinical records to assess abortion-related complications, the hotline relies on women’s self-report of symptoms experienced that require extra monitoring and might indicate the need to seek medical care in a formal health facility.

ent at a formal health facility is categorised as treatment received. In the absence of clinical records to assess abortion-related complications, the hotline relies on women’s self-report of symptoms experienced that require extra monitoring and might indicate the need to seek medical care in a formal health facility. Warning signs of a potential complications include heavy bleeding, defined as soaking more than two sanitary pads per hour for more than 2 hours; high fever, defined as a recorded temperature over 38°C for more than 24 hours; and severe pain, defined as pain that inhibits normal activities and is not alleviated by over-the-counter painkillers. Figure 1 Second-trimester medication abortion (MA) hotline callers, July 2012 – October 2016. Descriptive statistics were calculated using Stata Release 14.21 Patient and public involvement The development of the research question and key outcomes was informed by previous collaborations with our hotline partner and ongoing conversations about the importance of building an evidence base to document the safety of out-of-clinic abortions, with an emphasis on documenting women’s experiences. As this article presents a secondary analysis of retrospective anonymised electronic client records from the safe abortion hotline, safe abortion hotline clients were not involved in the design, recruitment or implementation of the study.

of out-of-clinic abortions, with an emphasis on documenting women’s experiences. As this article presents a secondary analysis of retrospective anonymised electronic client records from the safe abortion hotline, safe abortion hotline clients were not involved in the design, recruitment or implementation of the study. Results Ninety-six women with pregnancies beyond the first trimester contacted the hotline during the study period for information on medication abortion. Callers reported gestational ages ranging from 13 to 22 (mean 15.6) weeks. After receiving pre-abortion counselling from the hotline, five of the 96 initial callers did not contact the hotline again, and hotline staff were neither able to confirm receipt of medications nor provide further support through the medication abortion process (figure 1). These women are considered lost-to-follow-up, and both the status of their medication abortion attempts and the outcomes of their pregnancies are unknown.

hotline again, and hotline staff were neither able to confirm receipt of medications nor provide further support through the medication abortion process (figure 1). These women are considered lost-to-follow-up, and both the status of their medication abortion attempts and the outcomes of their pregnancies are unknown. Ninety-one women received telephone support from trained safe abortion hotline counsellors throughout their abortion process, with gestational ages ranging from 13 to 22 (mean 15.6) weeks; 47 women (50.6%) had gestational ages of 13 or 14 weeks (table 1, online figure 2, supplementary file). Overall, 83 out of the 91 women who managed their own abortions beyond the first trimester with telephone support from trained safe abortion hotline counsellors successfully terminated their pregnancies using medication abortion (91.2%). Six callers sought medical care during their medication abortion process (6.6%); these callers all successfully terminated pregnancies after seeking additional care. 10.1136/bmjsrh-2018-200102.supp1Supplementary file 1 Table 1 Abortion characteristics for hotline clients seeking second-trimester medication abortion (2012–2016) (n=91) Characteristics % (n) Gestational age (weeks) (mean (SD)) 15.6 (2.5) Complete abortion by first MA attempt without medical care 89.0 (81.0) Complete abortion by second MA attempt without medical care 91.2 (83.0) Sought medical care 6.6 (6.0) Terminated pregnancy* 97.8 (89.0) *Two callers were lost to follow-up after a failed abortion on their first attempt; the outcome of their pregnancy is unknown.

tion by first MA attempt without medical care 89.0 (81.0) Complete abortion by second MA attempt without medical care 91.2 (83.0) Sought medical care 6.6 (6.0) Terminated pregnancy* 97.8 (89.0) *Two callers were lost to follow-up after a failed abortion on their first attempt; the outcome of their pregnancy is unknown. MA, medication abortion. Eighty-one women had a complete abortion using medications on their first attempt (89.0%) and did not seek medical care. Three women experienced heavy bleeding after their first attempt, a warning sign of a potential complication, and sought medical care. Two of these women underwent dilation and curettage (D&C), one was observed at the hospital and not treated. For seven women, their first attempt at self-managing a second- trimester medication abortion with support from the hotline resulted in a failed abortion (the medications had no effect). Four of these women pursued a second attempt at self managing a second trimester medication abortion with the support from hotline; two of these women had a complete abortion and did not seek medical care, while two sought medical care for heavy bleeding (figure 1, table 2). Both women who sought care after a second medication abortion attempt underwent D&C when they sought care. Two women with a failed abortion after their first attempt were lost to follow-up, and the outcomes of their pregnancies are unknown. One woman sought medical care and obtained a surgical abortion after her first medication abortion attempt.

fter a second medication abortion attempt underwent D&C when they sought care. Two women with a failed abortion after their first attempt were lost to follow-up, and the outcomes of their pregnancies are unknown. One woman sought medical care and obtained a surgical abortion after her first medication abortion attempt. Table 2 Abortion trajectories of second-trimester medication abortion MA clients with failed abortion after first attempt (2012–2016) (n=7) Client Gestational age (weeks) Medications used Abortion trajectory 1 14 Mifepristone and misoprostol (M&M) Complete abortion on second MA attempt (M&M) 2 13 M&M Complete abortion on second MA attempt (M&M) 3 13 Misoprostol alone Lost to Follow-up 4 14 Misoprostol alone Lost to Follow-up 5 14 Misoprostol alone Surgical Abortion 6 14 Misoprostol alone Received D&C after second MA attempt (misoprostol alone) 7 13 Misoprostol alone Received D&C after second MA attempt (M&M) MA, medication abortion.

on second MA attempt (M&M) 3 13 Misoprostol alone Lost to Follow-up 4 14 Misoprostol alone Lost to Follow-up 5 14 Misoprostol alone Surgical Abortion 6 14 Misoprostol alone Received D&C after second MA attempt (misoprostol alone) 7 13 Misoprostol alone Received D&C after second MA attempt (M&M) MA, medication abortion. Characteristics of clients’ final abortion attempt are presented in table 3. Most clients (82.4%) used mifepristone and misoprostol in combination to terminate their pregnancies; 16 women (17.6%) used misoprostol alone. Complete data on the timing of medication administration is missing for 28 women. Out of the 47 women who used the combined mifepristone and misoprostol regime and reported the timing of medication administration to their counsellor, 45 women adhered to the evidence-based guidelines17 20 and took the first dose of misoprostol within 12 to 48 hours of taking mifepristone. The two clients who reported timing of medication administration outside of the evidence-based guidelines took the first dose of misoprostol 60 hours after taking mifepristone; recorded gestational ages of these clients were 17 and 19 weeks respectively, and both clients successfully completed abortions without seeking medical care. Table 3 Abortion characteristics of final abortion attempt for hotline clients seeking second-trimester medication abortion (2012–2016) (n=91)

Characteristics of clients’ final abortion attempt are presented in table 3. Most clients (82.4%) used mifepristone and misoprostol in combination to terminate their pregnancies; 16 women (17.6%) used misoprostol alone. Complete data on the timing of medication administration is missing for 28 women. Out of the 47 women who used the combined mifepristone and misoprostol regime and reported the timing of medication administration to their counsellor, 45 women adhered to the evidence-based guidelines17 20 and took the first dose of misoprostol within 12 to 48 hours of taking mifepristone. The two clients who reported timing of medication administration outside of the evidence-based guidelines took the first dose of misoprostol 60 hours after taking mifepristone; recorded gestational ages of these clients were 17 and 19 weeks respectively, and both clients successfully completed abortions without seeking medical care. Table 3 Abortion characteristics of final abortion attempt for hotline clients seeking second-trimester medication abortion (2012–2016) (n=91) Characteristics % (n) N Medications used 91 Mifepristone and misoprostol in combination 82.4 (75) Misoprostol alone 17.6 (16) Hours between mifepristone and first dose of misoprostol (mean (range)* 32.9 (12–60) 47 Timing of medication doses adheres to evidence-based protocols* 95.7 (45) 47 Warning signs of potential complications† 90 Experienced heavy bleeding 12.2 (11) Experienced high fever 3.3 (3) *Complete data on medication timing missing for 28 callers.

epristone and first dose of misoprostol (mean (range)* 32.9 (12–60) 47 Timing of medication doses adheres to evidence-based protocols* 95.7 (45) 47 Warning signs of potential complications† 90 Experienced heavy bleeding 12.2 (11) Experienced high fever 3.3 (3) *Complete data on medication timing missing for 28 callers. †Data on heavy bleeding and high fever missing for one caller. Heavy bleeding is defined as soaking more than two sanitary pads per hour for more than 2 hours; high fever is defined as recorded temperature of over 38°C for more than 24 hours. Women experienced a variety of normal side effects during the abortion process including nausea, dizziness, mild fever, heavy bleeding, and cramps (data not shown). In total, 11 women experienced heavy bleeding, and three experienced a fever over 38°C lasting more than 24 hours. Fifty-five women used over-the-counter painkillers to help manage pain. Discussion This is the first known documentation of the process and outcomes of second-trimester medication abortion provision outside of a clinic setting by non-clinically trained abortion counsellors. Despite data limitations, our analysis suggests that when women have access to accurate information alongside empathic and evidence-based counselling, they may be able to safely self-manage medication abortions beyond the first trimester.

ision outside of a clinic setting by non-clinically trained abortion counsellors. Despite data limitations, our analysis suggests that when women have access to accurate information alongside empathic and evidence-based counselling, they may be able to safely self-manage medication abortions beyond the first trimester. Our study has some important limitations: data collected by this hotline are not collected for research purposes, nor are they collected by trained researchers – thus data suffer from missing data on sociodemographic information and secondary outcomes of interest, such as timing of medication doses. Despite data imperfections, this analysis demonstrates a clear and compelling need for abortion services beyond the first trimester, and suggests that further exploration of this model of accompaniment for self-management of later abortion is warranted. In addition, we were unable to reliably determine the incidence of abortion complications among women included in our analysis. An agreed-upon definition of and framework for abortion complications in settings where unsafe abortion persists does not exist, and the lack of such a definition stymies efforts to reliably document the safety of abortions in much of the world.22 Accurate estimates of the incidence and severity of abortion complications both in and out of facility settings are essential to developing targeted and effective programmes, policies and interventions to increase access to safe abortion and to improve women’s health. An additional limitation is the small sample size of this study; though this study includes data on all callers to the safe abortion hotline reporting gestational ages over 12 weeks during the study period, second-trimester medication abortion clients are a low percentage of overall callers. Future studies should aim to document second-trimester medication abortions across multiple contexts.

is study includes data on all callers to the safe abortion hotline reporting gestational ages over 12 weeks during the study period, second-trimester medication abortion clients are a low percentage of overall callers. Future studies should aim to document second-trimester medication abortions across multiple contexts. There has been little documentation of women’s experiences self-managing abortions with the support of hotlines or other non-clinically trained abortion counsellors.11 12 15 18 23 24 Evidence from Europe, Canada and the USA suggests that early in pregnancy, a ‘home-based’ medication abortion (where women take mifepristone in a clinic, and return home to take the remaining doses of misoprostol) is as safe, effective and acceptable as clinic-based medication abortion, and may increase access to care by reducing the number of clinic visits a woman is required to attend.23 Abortion beyond 12 weeks gestation, however, is qualitatively different, and objectively runs a higher risk of mortality and morbidity. While some women seeking abortion later in pregnancy express a preference for medication abortion in legal, clinic-based settings,25 26 no research has documented the preferences, expectations or experiences of women in need of later abortion in legally restricted, non-clinic based settings.

of mortality and morbidity. While some women seeking abortion later in pregnancy express a preference for medication abortion in legal, clinic-based settings,25 26 no research has documented the preferences, expectations or experiences of women in need of later abortion in legally restricted, non-clinic based settings. Evidence has demonstrated that the provision of second-trimester abortions using medications requires training in medication protocols, empathic counselling skills, and well-established emergency plans27 – it does not, however, require advanced clinical training or surgical skills as required for dilation and evacuation (the other WHO recommended method for abortion beyond 12 weeks gestation). Globally, evidence has shown that when women cannot access safe abortion services, they often utilise abortion methods that endanger their health.28 Given the demonstrated need for abortions later in pregnancy regardless of the legal setting, safe-abortion hotlines are providing women with essential, evidence-based information on how to self-manage medication abortions beyond 12 weeks gestation. If this model proves to be a safer alternative than existing options, these services have the potential to dramatically reduce abortion-related harms for women worldwide.

bortion hotlines are providing women with essential, evidence-based information on how to self-manage medication abortions beyond 12 weeks gestation. If this model proves to be a safer alternative than existing options, these services have the potential to dramatically reduce abortion-related harms for women worldwide. Our analysis of electronic client records from this safe-abortion hotline in Indonesia suggests that the model of providing accompaniment services for abortions later in pregnancy could be a safe alternative for women in need of abortions beyond 12 weeks gestation in legally restrictive contexts. Further research is warranted to document the safety, feasibility and acceptability of non-clinic-based second-trimester medication abortion, and accompaniment services provided by non-physician providers as a way of expanding access to safe abortion later in pregnancy. The authors would like to thank the staff at the safe abortion hotline in Indonesia for generously providing them with access to their counselling records. Contributors: CG and IH originated the study question. RTJ conducted the analysis. CG and RTJ wrote the original draft of the manuscript. CG, RTJ, SB and IH all contributed in reviewing and editing the manuscript. All authors read and approved the final manuscript. Funding: This work was supported in part by the David and Lucile Packard Foundation. The funders were not involved in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.

Contributors: CG and IH originated the study question. RTJ conducted the analysis. CG and RTJ wrote the original draft of the manuscript. CG, RTJ, SB and IH all contributed in reviewing and editing the manuscript. All authors read and approved the final manuscript. Funding: This work was supported in part by the David and Lucile Packard Foundation. The funders were not involved in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the article for publication. Competing interests: None declared. Patient consent: Not required. Ethics approval: Ethical approval for this study was obtained from Allendale Investigational Review Board. As the data were collected in the process of routine service provision, and the authors only had access to data that were completely anonymised, informed consent was not collected from callers to the safe abortion hotline. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Data are available upon reasonable request; please contact rjayaweera@ibisreproductivehealth.org.

We read David Horwell’s letter ‘End of the road for Essure®?’ published in the July 2017 issue of this journal with interest.1While Bayer’s decision is understandable, it is also disappointing. Essure’s withdrawal from the market leaves laparoscopic clip sterilisation and open tubal ligation (at caesarean delivery), as the only methods for female sterilisation available in UK. Clinicians will have to wait for future hysteroscopic alternatives to emerge to broaden the choice for women once more. This will be difficult in the light of the lawsuits attracted by Essure and the withdrawal of Adiana in 2012 due to its high failure rate and the manufacturer’s inability to keep up with legal costs over patent infringement litigation.2 Having offered the Essure device to our patients we had the opportunity to appreciate first-hand the advantages of the hysteroscopic approach, especially in those who preferred or needed to avoid abdominal incision or general anaesthesia (GA) and who wanted to use the ambulatory setting to be sterilised. We found that with adequate counselling, the hysteroscopic approach was popular with women, despite our informing them of the gradually emerging adverse evidence, reported adverse events and the Faculty of Sexual & Reproductive Health (FSRH) recommendation to avoid its use in women with nickel allergy, heavy menstrual bleeding, chronic pelvic pain, pelvic infections and autoimmune conditions.3

lar with women, despite our informing them of the gradually emerging adverse evidence, reported adverse events and the Faculty of Sexual & Reproductive Health (FSRH) recommendation to avoid its use in women with nickel allergy, heavy menstrual bleeding, chronic pelvic pain, pelvic infections and autoimmune conditions.3 Between September 2009 and June 2017, 200 Essure procedures were performed in our department instead of laparoscopic sterilisations. To gain confidence the first 10 Essure cases were done under GA. The remaining 190 procedures were done by two operators in our out-patient hysteroscopy suite. Of the 200 women, a majority (175) had no previous abdominal surgery. Twenty-five (12.5%) had had laparotomies (three via midline and 22 via low transverse incisions). Generally, women were happy to use additional contraception until the necessary confirmatory test. A total of 185 women attended follow-up; 120 had a hysterosalpingogram (HSG) and 65 had pelvic ultrasound scans. Of these, 10 needed an HSG as we failed to obtain the ideal views on USS. Two cases after an HSG showed possible unilaterally patency and had a further HSG 3 months later, confirming bilateral occlusion. Fifteen women were lost to follow-up.

ow-up; 120 had a hysterosalpingogram (HSG) and 65 had pelvic ultrasound scans. Of these, 10 needed an HSG as we failed to obtain the ideal views on USS. Two cases after an HSG showed possible unilaterally patency and had a further HSG 3 months later, confirming bilateral occlusion. Fifteen women were lost to follow-up. To date, five of the 200 women (2%) have undergone further surgery – three for pelvic pain, one for menorrhagia and fibromyalgia-like symptoms and one for bilateral perforation. Four had bilateral salpingectomy and one a hysterectomy and bilateral salpingectomy. Two pregnancies have occurred following the procedure, one after a misdiagnosed HSG and the other a luteal phase pregnancy. Essure will be missed, especially as a choice for permanent birth control for high-risk women – the very obese, those with one or more midline laparotomies, those who cannot accept or have contraindications for using hormonal methods or where GA is contraindicated. The ease of recovery in women accepting the hysteroscopic versus the laparoscopic approach is also relevant. We agree that it is unlikely that a new office-based procedure will become available very soon. We hope that a future device will not need post-procedural confirmation and continuing use of contraception, and of course will not have the perceived or real side effects reported by women using Essure and Adiana.

Essure will be missed, especially as a choice for permanent birth control for high-risk women – the very obese, those with one or more midline laparotomies, those who cannot accept or have contraindications for using hormonal methods or where GA is contraindicated. The ease of recovery in women accepting the hysteroscopic versus the laparoscopic approach is also relevant. We agree that it is unlikely that a new office-based procedure will become available very soon. We hope that a future device will not need post-procedural confirmation and continuing use of contraception, and of course will not have the perceived or real side effects reported by women using Essure and Adiana. Contributors: SD wrote the draft of the letter. WY provided the data from our department about our experience with Essure and contributed to the writing of the draft. AG reviewed and edited the letter. All authors approved the final version. Competing interests: None declared. Patient consent: Obtained. Provenance and peer review: Not commissioned; internally peer reviewed.

Key message points General practice continues to be the most commonly used source by women for contraceptive supplies. It is therefore important that relevant practitioner skills and willingness to provide all methods be maintained in this setting. Integrated sexual and reproductive health clinics and wider contraceptive provision in community settings, such as in schools and colleges, have helped reach those who have not accessed more traditional services in the past. A strong case can be made for the provision of high-quality services in different settings and with different styles of service delivery to ensure that women and men have access to contraception that meets their needs. Introduction Contraceptive supplies have been available free of charge to all wishing to prevent pregnancy from the National Health Service (NHS) in Britain since the early 1970s. Open-access contraceptive services are provided in general practice, family planning clinics (FPCs) and young people’s services. The availability of a range of contraceptive sources and services including retail outlets helps ensure that the needs of different users are met.

vice (NHS) in Britain since the early 1970s. Open-access contraceptive services are provided in general practice, family planning clinics (FPCs) and young people’s services. The availability of a range of contraceptive sources and services including retail outlets helps ensure that the needs of different users are met. The first decade of the 21st century saw several changes in the delivery of contraceptive services in Britain, largely in response to national strategies to improve sexual and reproductive health.1–4 Guidance relating to long-acting reversible contraceptives (LARC) from the National Institute of Health and Care Excellence extended their availability to general practice.5 Recommendations have been made for availability of contraception in schools and other educational settings to improve access for young people,6 and the trend towards integration of FPCs and genitourinary medicine (GUM) services has increased the likelihood that those attending for the treatment of sexually transmitted infection (STI) leave the service with contraception.7 8 During the same period, the provision of over-the-counter emergency contraception has increased.9 10

towards integration of FPCs and genitourinary medicine (GUM) services has increased the likelihood that those attending for the treatment of sexually transmitted infection (STI) leave the service with contraception.7 8 During the same period, the provision of over-the-counter emergency contraception has increased.9 10 Of interest is the extent to which these changes may be reflected in patterns of contraceptive service use. In this article, we analyse data from the third National Survey of Sexual Attitudes and Lifestyles (Natsal-3) to estimate the prevalence of use of different sources of contraceptive supplies in Britain and its variation by key demographic and behavioural characteristics. In doing so, we update similar information from the second survey (Natsal-2).11 Methods Data were analysed from Natsal-3, a multistage, clustered and stratified probability sample survey of 15 162 British residents aged 16–74 years interviewed between September 2010 and August 2012. Participants were interviewed in their homes using a combination of face-to-face computer-assisted personal interview (CAPI) and computer-assisted self-interview (CASI) for more sensitive questions. The sampling frame was the Postcode Address File. The response rate was 57.7% and the cooperation rate, for all the eligible addresses contacted, was 65.8%. Full details of the methods used in Natsal-3 and the characteristics of the complete sample are reported elsewhere.12–14

rview (CASI) for more sensitive questions. The sampling frame was the Postcode Address File. The response rate was 57.7% and the cooperation rate, for all the eligible addresses contacted, was 65.8%. Full details of the methods used in Natsal-3 and the characteristics of the complete sample are reported elsewhere.12–14 Measures Participants were asked in the CAPI which methods of contraception they or together with a partner had used in the past year. Those reporting use of any method(s) were asked to identify which sources of supply they had used from a card showing all available options with the following question: “Have you got contraception from any of these sources in the last year?”.15 Sources listed were grouped into four categories: general practice (doctor or nurse), community clinic (sexual health (GUM) clinic, family planning/contraceptive/reproductive health clinic, youth services or school/college/university), retail (pharmacy/chemist, website, petrol station/supermarket/other shop, vending machine or mail order) and other (emergency department at hospital, antenatal clinic, private doctor or other). Sociodemographic characteristics used in this analysis included age, ethnic group, and educational attainment. Area of residence was categorised as rural or urban and area-related socioeconomic status was measured using the Index of Multiple Deprivation (IMD).16 IMD is a multi-dimensional measure of area (neighbourhood)-level deprivation based on the participants’ postcode. Sexual health and lifestyle indicators included current relationship status, parity, STI diagnosis and pregnancy, including planning status (ie, ambivalent/unplanned, planned, trying or not pregnant).

D).16 IMD is a multi-dimensional measure of area (neighbourhood)-level deprivation based on the participants’ postcode. Sexual health and lifestyle indicators included current relationship status, parity, STI diagnosis and pregnancy, including planning status (ie, ambivalent/unplanned, planned, trying or not pregnant). Statistical analyses Women and men aged 16–44 years were included if they reported having (i) had vaginal intercourse in the past year, and (ii) not exclusively relied on either female or male sterilisation (including hysterectomy) for contraception. We used Stata (version 13.1) for complex survey analysis to incorporate weighting, clustering and stratification of the Natsal-3 data.13 We calculated the percentage of participants (with 95% confidence intervals) who had obtained contraception in the past year from each source and used the chi-square statistic to determine whether there were statistically significant variation in the reporting of these sources by sociodemographic characteristics and by lifestyle and sexual health variables. Participants were not asked from which sources they obtain specific contraceptive methods, and so we were unable to link method with source used. However, since use of LARC methods was likely to have been of longer duration, we included sole use of LARC (intrauterine contraceptives and implants) as an independent variable. Injectables were not included in the LARC category due to their lower continuation rates. Proportional Venn diagrams were used to represent the extent of overlap in use of different sources among those who had obtained contraceptive supplies.

f LARC (intrauterine contraceptives and implants) as an independent variable. Injectables were not included in the LARC category due to their lower continuation rates. Proportional Venn diagrams were used to represent the extent of overlap in use of different sources among those who had obtained contraceptive supplies. Results Some 4571 women and 3142 men were eligible for inclusion in the analysis; 87.0% of women and 73.8% of men obtained contraception from at least one source in the past year (table 1). Table 1 Use of sources to obtain contraceptives in the past year among women and men aged 16–44 years who had had vaginal sex in the past year Source of contraceptive supplies Women Men Denominators (unweighted, weighted) 4571, 3088 3142, 3182 % (95% CI) % (95% CI) General practice* 59.1 (57.2 to 61.0) 15.3 (13.6 to 17.0) Community clinic 23.0 (21.6 to 24.5) 21.3 (19.6 to 23.0) GUM clinic 8.5 (7.6 to 9.5) 10.8 (9.6 to 12.1) Family planning clinic 14.4 (13.2 to 15.7) 6.8 (5.8 to 8.1) Youth advisory clinic 1.1 (0.8 to 1.6) 2.1 (1.7 to 2.7) School/college/university 2.4 (1.9 to 3.0) 6.4 (5.4 to 7.5) Retail 28.6 (26.9 to 30.3) 54.6 (52.3 to 56.8) Pharmacy/chemist 21.3 (19.8 to 22.9) 35.0 (32.8 to 37.2) Internet website 0.7 (0.5 to 1.2) 3.1 (2.4 to 4.0) Petrol station/supermarket 10.3 (9.2 to 11.5) 24.8 (23.0 to 26.7) Vending machine 0.7 (0.5 to 1.1) 9.4 (8.2 to 10.7) Mail order 0.2 (0.1 to 0.4) 0.7 (0.4 to 1.2) Other 2.9 (2.4 to 3.6) 2.9 (2.3 to 3.6)

54.6 (52.3 to 56.8) Pharmacy/chemist 21.3 (19.8 to 22.9) 35.0 (32.8 to 37.2) Internet website 0.7 (0.5 to 1.2) 3.1 (2.4 to 4.0) Petrol station/supermarket 10.3 (9.2 to 11.5) 24.8 (23.0 to 26.7) Vending machine 0.7 (0.5 to 1.1) 9.4 (8.2 to 10.7) Mail order 0.2 (0.1 to 0.4) 0.7 (0.4 to 1.2) Other 2.9 (2.4 to 3.6) 2.9 (2.3 to 3.6) Emergency department at hospital 0.30 (0.2 to 0.5) 0.2 (0.1 to 0.5) NHS antenatal clinic 0.90 (0.7 to 1.3) 1.0 (0.7 to 1.5) Private doctor 1.3 (0.9 to 1.7) 1.0 (0.7 to 1.4) Other 0.5 (0.3 to 0.8) 0.8 (0.5 to 1.3) Contraception not obtained in the last year 13.0 (11.6 to 14.4) 26.2 (24.2 to 28.3) Number of sources used† 0 13.0 (11.6 to 14.4) 26.2 (24.2 to 28.3) 1 59.8 (57.9 to 61.6) 43.3 (41.0 to 45.5) 2 21.0 (19.6 to 22.5) 20.6 (18.9 to 22.4) 3+ 6.2 (5.5 to 7.1) 10.0 (8.8 to 11.3) *General practice includes doctor, nurse and general practice surgery. †Any source rather than source category. GUM, genitourinary medicine; NHS, National Health Service. Prevalence of use of different sources of contraception The source of contraceptive supplies used by the largest proportion of women (59.1%) was general practice. 28.6% obtained supplies from retail outlets, the majority being pharmacies. 23% of women obtained supplies from community services, predominantly FPCs (14.4%) but 8.5% obtained supplies from GUM services (table 1). 27.3% of women had used more than one source. Women most commonly reported use of both general practice and retail sources (11.0%) (figure 1).

utlets, the majority being pharmacies. 23% of women obtained supplies from community services, predominantly FPCs (14.4%) but 8.5% obtained supplies from GUM services (table 1). 27.3% of women had used more than one source. Women most commonly reported use of both general practice and retail sources (11.0%) (figure 1). The source of contraceptive supplies used by the largest proportion of men (54.6%) was retail outlets (table 1). Among retail suppliers, pharmacies were most commonly used by more than one-third of men, but petrol stations and supermarkets were used by one in four (24.8%) and vending machines by nearly one in ten (9.4%). Just over one-fifth (21.3%) of men used community clinics, of which GUM services were more commonly used than other community services. 15.3% of men reported contraception used by them being from general practice. 30.6% of men used more than one source of contraception, with the greatest overlap in use being community clinics and retail (11.9%) (figure 1). Figure 1 Extent of overlap in the type(s) of source used to obtain contraceptives in the past year among women and men aged 16–44 years who reported vaginal sex in the past year and obtaining contraceptives. These diagrams apply only to those who obtained contraception in the previous year and so the percentages differ from those presented in the tables, where the denominator includes those who had not.

he past year among women and men aged 16–44 years who reported vaginal sex in the past year and obtaining contraceptives. These diagrams apply only to those who obtained contraception in the previous year and so the percentages differ from those presented in the tables, where the denominator includes those who had not. Use of contraceptive sources by characteristics of participants General practice Marked age differences among women were seen in the use of general practice for contraceptive supplies (table 2). Roughly two-thirds of women in their 20s had obtained contraception from this source in the past year, compared with half of 16–17-year-olds and 35–44-year-olds. Statistically significant differences were also seen by other demographic variables. Women living in rural districts, those in cohabiting or ‘steady’ relationships as opposed to those who were single or married, and childless women compared with mothers, all more commonly used general practice for contraceptive supplies. Half of LARC users obtained supplies from general practice. Women with minimum educational qualifications, or none, more commonly used this source compared with those who were more highly qualified. The most marked difference, however, was seen by ethnic group. Compared with women identifying as ‘white’, considerably lower proportions of women in black and minority ethnic groups obtained contraception from general practice. Variation in use of general practice by health behaviour was not statistically significant among women.

ked difference, however, was seen by ethnic group. Compared with women identifying as ‘white’, considerably lower proportions of women in black and minority ethnic groups obtained contraception from general practice. Variation in use of general practice by health behaviour was not statistically significant among women. Table 2 Prevalence of source(s) used to obtain contraceptive supplies in the past year among women aged 16–44 years who had had vaginal sex in the past year by sociodemographic and behavioural characteristics* Sociodemographic and behavioural characteristics General practice Community clinics Retail More than one None Denominators (unweighted, weighted) % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) All 4571, 3088 59.1 (57.2 to 61.0) 23.0 (21.6 to 24.5) 28.6 (26.9 to 30.3) 27.3 (25.7 to 28.9) 13.0 (11.6 to 14.4) Age group (years) p<0.0001 p<0.0001 p=0.0255 p<0.0001 p<0.0001 16–17 242, 118 50.3 (43.5 to 57.0) 53.7 (46.8 to 60.5) 23.9 (18.4 to 30.4) 43.5 (36.8 to 50.5) 11.5 (7.7 to 16.9) 18–19 345, 182 59.7 (53.9 to 65.3) 45.2 (39.4 to 51.1) 21.4 (17.2 to 26.4) 36.2 (30.8 to 42.0) 7.2 (4.5 to 11.4) 20–24 865, 504 67.3 (63.6 to 70.7) 36.5 (32.8 to 40.3) 29.4 (26.1 to 33.0) 40.9 (37.3 to 44.5) 6.1 (4.5 to 8.2) 25–29 1042, 522 66.0 (62.5 to 69.3) 19.7 (17.3 to 22.3) 30.2 (27.1 to 33.5) 27.7 (24.8 to 30.7) 8.8 (7.0 to 11.0) 30–34 725, 464 60.0 (55.8 to 64.0) 17.1 (14.3 to 20.2) 31.0 (27.3 to 35.0) 24.4 (21.2 to 28.0) 13.4 (10.8 to 16.7) 35–44 629, 758 49.6 (45.5 to 53.8) 9.9 (7.6 to 12.7) 27.8 (24.4 to 31.5) 15.0 (12.4 to 18.0) 21.7 (18.4 to 25.4) Ethnic group† p<0.0001 p=0.0002 p=0.0263 p=0.0687

16–17 242, 118 50.3 (43.5 to 57.0) 53.7 (46.8 to 60.5) 23.9 (18.4 to 30.4) 43.5 (36.8 to 50.5) 11.5 (7.7 to 16.9) 18–19 345, 182 59.7 (53.9 to 65.3) 45.2 (39.4 to 51.1) 21.4 (17.2 to 26.4) 36.2 (30.8 to 42.0) 7.2 (4.5 to 11.4) 20–24 865, 504 67.3 (63.6 to 70.7) 36.5 (32.8 to 40.3) 29.4 (26.1 to 33.0) 40.9 (37.3 to 44.5) 6.1 (4.5 to 8.2) 25–29 1042, 522 66.0 (62.5 to 69.3) 19.7 (17.3 to 22.3) 30.2 (27.1 to 33.5) 27.7 (24.8 to 30.7) 8.8 (7.0 to 11.0) 30–34 725, 464 60.0 (55.8 to 64.0) 17.1 (14.3 to 20.2) 31.0 (27.3 to 35.0) 24.4 (21.2 to 28.0) 13.4 (10.8 to 16.7) 35–44 629, 758 49.6 (45.5 to 53.8) 9.9 (7.6 to 12.7) 27.8 (24.4 to 31.5) 15.0 (12.4 to 18.0) 21.7 (18.4 to 25.4) Ethnic group† p<0.0001 p=0.0002 p=0.0263 p=0.0687 p=0.1049 White 3410, 2223 61.4 (59.3 to 63.4) 22.2 (20.7 to 23.7) 28.3 (26.6 to 30.1) 27.4 (25.8 to 29.1) 12.3 (10.9 to 13.8) Black 120, 92 38.7 (28.9 to 49.4) 40.2 (30.2 to 51.0) 23.2 (16.0 to 32.6) 26.1 (18.5 to 35.5) 16.0 (9.2 to 26.5) Black: African 64, 51 46.7 (32.6 to 61.3) 32.7 (21.0 to 47.0) 21.2 (12.1 to 34.5) 23.9 (14.3 to 37.2) 17.8 (8.5 to 33.7) Black: Caribbean 41, 28 29.2 (16.0 to 47.1) 57.9 (38.8 to 75.0) 30.2 (16.0 to 49.7) 41.8 (25.1 to 60.7) 9.9 (3.3 to 26.5) Asian 160, 126 39.3 (30.9 to 48.3) 18.5 (12.3 to 26.9) 39.3 (31.1 to 48.2) 19.2 (13.4 to 26.9) 16.8 (11.2 to 24.3) Mixed/other 163, 113 49.3 (40.2 to 58.5) 31.5 (23.7 to 40.6) 24.7 (17.7 to 33.3) 30.7 (23.0 to 39.5) 19.3 (12.2 to 29.2) Academic qualifications‡§ p=0.0474 p=0.0251 p<0.0001 p<0.0001 p=0.3574

White 3410, 2223 61.4 (59.3 to 63.4) 22.2 (20.7 to 23.7) 28.3 (26.6 to 30.1) 27.4 (25.8 to 29.1) 12.3 (10.9 to 13.8) Black 120, 92 38.7 (28.9 to 49.4) 40.2 (30.2 to 51.0) 23.2 (16.0 to 32.6) 26.1 (18.5 to 35.5) 16.0 (9.2 to 26.5) Black: African 64, 51 46.7 (32.6 to 61.3) 32.7 (21.0 to 47.0) 21.2 (12.1 to 34.5) 23.9 (14.3 to 37.2) 17.8 (8.5 to 33.7) Black: Caribbean 41, 28 29.2 (16.0 to 47.1) 57.9 (38.8 to 75.0) 30.2 (16.0 to 49.7) 41.8 (25.1 to 60.7) 9.9 (3.3 to 26.5) Asian 160, 126 39.3 (30.9 to 48.3) 18.5 (12.3 to 26.9) 39.3 (31.1 to 48.2) 19.2 (13.4 to 26.9) 16.8 (11.2 to 24.3) Mixed/other 163, 113 49.3 (40.2 to 58.5) 31.5 (23.7 to 40.6) 24.7 (17.7 to 33.3) 30.7 (23.0 to 39.5) 19.3 (12.2 to 29.2) Academic qualifications‡§ p=0.0474 p=0.0251 p<0.0001 p<0.0001 p=0.3574 No academic qualifications 270, 162 63.1 (56.3 to 69.5) 27.2 (21.2 to 34.3) 14.8 (10.6 to 20.3) 21.5 (16.5 to 27.6) 9.8 (6.6 to 14.5) Academic qualifications typically gained at age 16 years 1225, 791 62.7 (59.2 to 66.0) 20.0 (17.6 to 22.6) 24.1 (21.4 to 27.0) 22.8 (20.3 to 25.4) 12.1 (9.9 to 14.6) Studying for/attained further academic qualifications 2150, 1463 57.8 (55.3 to 60.3) 23.7 (21.8 to 25.8) 32.8 (30.5 to 35.3) 30.8 (28.7 to 33.0) 13.2 (11.4 to 15.1) Quintiles of Index of Multiple Deprivation p=0.7373 p<0.0001 p=0.8767 p=0.3140 p=0.1343

No academic qualifications 270, 162 63.1 (56.3 to 69.5) 27.2 (21.2 to 34.3) 14.8 (10.6 to 20.3) 21.5 (16.5 to 27.6) 9.8 (6.6 to 14.5) Academic qualifications typically gained at age 16 years 1225, 791 62.7 (59.2 to 66.0) 20.0 (17.6 to 22.6) 24.1 (21.4 to 27.0) 22.8 (20.3 to 25.4) 12.1 (9.9 to 14.6) Studying for/attained further academic qualifications 2150, 1463 57.8 (55.3 to 60.3) 23.7 (21.8 to 25.8) 32.8 (30.5 to 35.3) 30.8 (28.7 to 33.0) 13.2 (11.4 to 15.1) Quintiles of Index of Multiple Deprivation p=0.7373 p<0.0001 p=0.8767 p=0.3140 p=0.1343 1–2 (lower deprivation) 1354, 950 59.6 (56.5 to 62.6) 18.5 (16.4 to 20.9) 28.9 (26.3 to 31.7) 26.0 (23.6 to 28.5) 14.7 (12.6 to 17.2) 3 780, 528 59.9 (55.8 to 64.0) 24.1 (21.1 to 27.4) 29.0 (25.5 to 32.8) 26.7 (23.3 to 30.5) 11.7 (9.0 to 15.1) 4–5 (higher deprivation) 1714, 1070 58.2 (55.2 to 61.2) 26.5 (24.1 to 29.0) 28.0 (25.5 to 30.7) 28.7 (26.2 to 31.2) 12.0 (10.1 to 14.1) Area type¶ p=0.0010 p<0.0001 p=0.8584 p=0.0386 p=0.3058 Urban 3162, 2082 57.6 (55.5 to 59.8) 24.6 (23.0 to 26.3) 28.5 (26.6 to 30.4) 28.1 (26.3 to 29.9) 13.3 (11.8 to 14.9) Rural 686, 465 65.7 (61.5 to 69.7) 15.8 (13.1 to 19.1) 28.9 (25.2 to 32.9) 23.6 (20.2 to 27.4) 11.4 (8.7 to 14.9) Relationship status p<0.0001 p<0.0001 p=0.4251 p<0.0001 p<0.0001

p<0.0001 p=0.8584 p=0.0386 p=0.3058 Urban 3162, 2082 57.6 (55.5 to 59.8) 24.6 (23.0 to 26.3) 28.5 (26.6 to 30.4) 28.1 (26.3 to 29.9) 13.3 (11.8 to 14.9) Rural 686, 465 65.7 (61.5 to 69.7) 15.8 (13.1 to 19.1) 28.9 (25.2 to 32.9) 23.6 (20.2 to 27.4) 11.4 (8.7 to 14.9) Relationship status p<0.0001 p<0.0001 p=0.4251 p<0.0001 p<0.0001 Married/civil partnership 1044, 924 52.5 (49.1 to 55.8) 11.1 (9.1 to 13.5) 29.8 (26.8 to 33.1) 18.0 (15.5 to 20.8) 18.7 (16.0 to 21.6) Living with partner 899, 621 65.2 (61.5 to 68.7) 21.1 (18.4 to 24.1) 27.3 (24.2 to 30.6) 27.1 (24.2 to 30.3) 10.8 (8.4 to 13.7) In a steady relationship 993, 515 65.5 (62.0 to 68.8) 36.3 (33.0 to 39.8) 29.7 (26.4 to 33.2) 38.8 (35.4 to 42.4) 6.7 (5.0 to 8.9) Single, never married 902, 482 57.1 (53.4 to 60.7) 34.1 (30.5 to 37.9) 26.6 (23.3 to 30.2) 32.9 (29.4 to 36.5) 11.7 (9.4 to 14.4) Opposite-sex partners, past year p=0.4996 p<0.0001 p=0.1243 p<0.0001 p<0.0001 1 2889, 2023 59.0 (56.7 to 61.1) 18.8 (17.3 to 20.4) 28.0 (26.1 to 30.0) 23.7 (22.1 to 25.5) 14.4 (12.9 to 16.2) 2+ 937, 510 60.4 (56.8 to 63.9) 39.7 (36.0 to 43.5) 31.0 (27.6 to 34.7) 41.7 (38.2 to 45.4) 6.8 (5.2 to 8.8) Any STI diagnosis (excluding thrush) (past year) p=0.9270 p<0.0001 p=0.4544 p<0.0001 p=0.0038 No 3729, 2482 59.1 (57.2 to 61.1) 22.2 (20.8 to 23.6) 28.7 (27.0 to 30.5) 26.8 (25.2 to 28.4) 13.2 (11.8 to 14.6) Yes 94, 49 59.7 (46.2 to 71.9) 65.4 (52.7 to 76.2) 24.3 (15.0 to 37.0) 56.8 (46.0 to 67.0) 3.7 (1.5 to 9.0) Pregnancy in past year p=0.4000 p<0.0001 p=0.3186 p=0.0002 p=0.0004 Pregnancy: unplanned/ambivalent 306, 178 59.8 (53.0 to 66.3) 32.3 (26.6 to 38.5) 28.4 (22.5 to 35.1) 32.0 (26.4 to 38.1) 6.1 (3.7 to 9.9) Pregnancy: planned 293, 188 60.5 (54.1 to 66.6) 13.9 (10.2 to 18.8) 24.6 (19.7 to 30.3) 20.5 (16.0 to 25.9) 17.2 (12.6 to 23.1) Trying to get pregnant 176, 125 53.3 (44.3 to 62.1) 16.6 (11.3 to 23.7) 20.0 (14.1 to 27.5) 15.2 (10.5 to 21.6) 20.8 (14.0 to 29.6) No pregnancy 1551, 1100 56.2 (53.2 to 59.2) 17.4 (15.4 to 19.7) 26.3 (23.8 to 28.9) 21.0 (18.8 to 23.3) 16.2 (14.1 to 18.7) Number of children p=0.0081 p<0.0001 p=0.0007 p<0.0001 p<0.0001 0 1936, 1210 62.0 (59.5 to 64.5) 29.8 (27.6 to 32.1) 31.9 (29.5 to 34.4) 35.3 (32.9 to 37.7) 9.3 (7.8 to 11.1) 1 783, 459 58.0 (54.0 to 62.0) 19.6 (16.7 to 22.8) 26.0 (22.6 to 29.7) 22.4 (19.3 to 25.8) 13.4 (10.7 to 16.8) 2+ 1119, 874 55.6 (52.1 to 59.1) 15.3 (13.1 to 17.9) 25.3 (22.4 to 28.3) 18.7 (16.4 to 21.3) 17.8 (15.2 to 20.7) Used LARC only (past year) 509, 374 48.0 (42.6 to 53.4) 23.4 (19.5 to 2

37.7) 9.3 (7.8 to 11.1) 1 783, 459 58.0 (54.0 to 62.0) 19.6 (16.7 to 22.8) 26.0 (22.6 to 29.7) 22.4 (19.3 to 25.8) 13.4 (10.7 to 16.8) 2+ 1119, 874 55.6 (52.1 to 59.1) 15.3 (13.1 to 17.9) 25.3 (22.4 to 28.3) 18.7 (16.4 to 21.3) 17.8 (15.2 to 20.7) Used LARC only (past year) 509, 374 48.0 (42.6 to 53.4) 23.4 (19.5 to 2 7.7) 2.0 (1.2 to 3.5) 7.4 (5.5 to 10.0) 30.8 (26.1 to 36.0) *P-values indicate differences within exposure category (ie, source of contraceptive supplies). Row percentages are presented; the categories are not mutually exclusive and add up to more than 100%. †P-value presented corresponds to the categories: White, Black, Asian, Mixed/Other. ‡Participants aged ≥17 years. §English General Certificate of Secondary Education or equivalent. ¶Urban is defined as areas with a population of >=10 000 people. LARC, long-acting reversible contraception; STI, sexually transmitted infection. Among men, the proportions reporting general practice as a source of contraception in the past year were statistically significantly higher among those living with a partner, fathers and those reporting only one sexual partner during the period (table 3). Table 3 Prevalence of source(s) used to obtain contraceptive supplies in the past year among men aged 16–44 years who had had vaginal sex in the past year by sociodemographic and behavioural characteristics* Sociodemographic and behavioural characteristics General practice Community clinics Retail More than one None Denominators (unweighted, weighted) % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) All 3142, 3182 15.3 (13.6 to 17.0) 21.3 (19.6 to 23.1)

Table 3 Prevalence of source(s) used to obtain contraceptive supplies in the past year among men aged 16–44 years who had had vaginal sex in the past year by sociodemographic and behavioural characteristics* Sociodemographic and behavioural characteristics General practice Community clinics Retail More than one None Denominators (unweighted, weighted) % (95% CI) % (95% CI) % (95% CI) % (95% CI) % (95% CI) All 3142, 3182 15.3 (13.6 to 17.0) 21.3 (19.6 to 23.1) 54.6 (52.3 to 56.8) 30.6 (28.57 to 32.67) 26.2 (24.2 to 28.3) Age group (years) p=0.2423 p<0.0001 p=0.0001 p<0.0001 p<0.0001 16–17 195, 116 14.3 (9.8 to 20.4) 58.9 (51.1 to 66.2) 48.0 (40.4 to 55.8) 41.2 (34.0 to 48.8) 7.6 (4.3 to 13.0) 18–19 287, 194 17.6 (13.2 to 23.0) 53.0 (46.2 to 59.6) 56.9 (49.9 to 63.6) 46.7 (39.8 to 53.7) 10.2 (6.9 to 14.8) 20–24 671, 537 15.7 (12.7 to 19.3) 37.3 33.4 to 41.4) 63.1 (59.0 to 67.0) 46.0 (41.9 to 50.2) 16.0 (13.2 to 19.3) 25–29 650, 514 12.9 (10.4 to 15.9) 18.9 (15.6 to 22.6) 57.1 (52.8 to 61.2) 30.1 (26.1 to 34.4) 25.7 (22.2 to 29.4) 30–34 452, 476 18.2 (14.6 to 22.5) 10.8 (7.9 to 14.6) 55.6 (50.6 to 60.6) 26.5 (22.3 to 31.2) 28.3 (23.9 to 33.1) 35–44 452, 802 14.3 (11.0 to 18.3) 5.2 (3.5 to 7.8) 47.0 (42.1 to 52.0) 17.6 (14.1 to 21.8) 38.6 (33.8 to 43.6) Ethnic group† p=0.5357 p=0.0431 p=0.0658 p=0.2708 p=0.0052

16–17 195, 116 14.3 (9.8 to 20.4) 58.9 (51.1 to 66.2) 48.0 (40.4 to 55.8) 41.2 (34.0 to 48.8) 7.6 (4.3 to 13.0) 18–19 287, 194 17.6 (13.2 to 23.0) 53.0 (46.2 to 59.6) 56.9 (49.9 to 63.6) 46.7 (39.8 to 53.7) 10.2 (6.9 to 14.8) 20–24 671, 537 15.7 (12.7 to 19.3) 37.3 33.4 to 41.4) 63.1 (59.0 to 67.0) 46.0 (41.9 to 50.2) 16.0 (13.2 to 19.3) 25–29 650, 514 12.9 (10.4 to 15.9) 18.9 (15.6 to 22.6) 57.1 (52.8 to 61.2) 30.1 (26.1 to 34.4) 25.7 (22.2 to 29.4) 30–34 452, 476 18.2 (14.6 to 22.5) 10.8 (7.9 to 14.6) 55.6 (50.6 to 60.6) 26.5 (22.3 to 31.2) 28.3 (23.9 to 33.1) 35–44 452, 802 14.3 (11.0 to 18.3) 5.2 (3.5 to 7.8) 47.0 (42.1 to 52.0) 17.6 (14.1 to 21.8) 38.6 (33.8 to 43.6) Ethnic group† p=0.5357 p=0.0431 p=0.0658 p=0.2708 p=0.0052 White 2384, 2256 14.9 (13.3 to 16.7) 21.0 (19.3 to 22.8) 53.4 (51.0 to 55.7) 30.7 (28.6 to 32.9) 27.6 (25.4 to 29.9) Black 89, 97 21.2 (13.1 to 32.5) 28.6 (18.8 to 40.9) 67.0 (54.7 to 77.4) 40.1 (28.7 to 52.8) 12.1 (6.5 to 21.4) Black: African 51, 56 18.2 (8.9 to 33.6) 26.0 (14.1 to 43.0) 69.6 (52.4 to 82.7) 41.5 (26.9 to 57.9) 8.1 (3.2 to 19.3) Black: Caribbean 30, 33 22.4 (9.1 to 45.4) 24.4 (12.0 to 43.3) 56.7 (34.6 to 76.4) 31.3 (15.7 to 52.7) 22.0 (9.3 to 43.7) Asian 119, 170 16.4 (10.1 to 25.6) 15.4 (9.2 to 24.7) 63.4 (52.4 to 73.2) 23.1 (15.6 to 32.7) 16.5 (9.8 to 26.5) Mixed/Other 119, 120 17.2 (10.6 to 26.6) 32.1 (23.6 to 42.1) 58.7 (47.5 to 68.9) 33.9 (25.3 to 43.8) 21.9 (13.2 to 34.2) Academic qualifications ‡§ p=0.1968 p=0.0265 p<0.0001 p=0.0001 p<0.0001

White 2384, 2256 14.9 (13.3 to 16.7) 21.0 (19.3 to 22.8) 53.4 (51.0 to 55.7) 30.7 (28.6 to 32.9) 27.6 (25.4 to 29.9) Black 89, 97 21.2 (13.1 to 32.5) 28.6 (18.8 to 40.9) 67.0 (54.7 to 77.4) 40.1 (28.7 to 52.8) 12.1 (6.5 to 21.4) Black: African 51, 56 18.2 (8.9 to 33.6) 26.0 (14.1 to 43.0) 69.6 (52.4 to 82.7) 41.5 (26.9 to 57.9) 8.1 (3.2 to 19.3) Black: Caribbean 30, 33 22.4 (9.1 to 45.4) 24.4 (12.0 to 43.3) 56.7 (34.6 to 76.4) 31.3 (15.7 to 52.7) 22.0 (9.3 to 43.7) Asian 119, 170 16.4 (10.1 to 25.6) 15.4 (9.2 to 24.7) 63.4 (52.4 to 73.2) 23.1 (15.6 to 32.7) 16.5 (9.8 to 26.5) Mixed/Other 119, 120 17.2 (10.6 to 26.6) 32.1 (23.6 to 42.1) 58.7 (47.5 to 68.9) 33.9 (25.3 to 43.8) 21.9 (13.2 to 34.2) Academic qualifications ‡§ p=0.1968 p=0.0265 p<0.0001 p=0.0001 p<0.0001 No academic qualifications 205, 209 19.5 (13.8 to 26.7) 20.7 (15.1 to 27.7) 40.8 (33.1 to 49.0) 26.9 (20.3 to 34.7) 33.7 (26.0 to 42.4) Academic qualifications typically gained at age 16 years 915, 888 15.6 (13.1 to 18.5) 17.6 (15.1 to 20.5) 48.5 (44.8 to 52.3) 25.2 (22.2 to 28.4) 33.0 (29.4 to 36.9) Studying for/attained further academic qualifications 1432, 1408 14.0 (11.9 to 16.3) 22.8 (20.4 to 25.4) 61.6 (58.6 to 64.5) 34.5 (31.6 to 37.4) 21.3 (18.8 to 24.0) Quintiles of Index of Multiple Deprivation p=0.0461 p=0.0866 p=0.8472 p=0.2125 p=0.1546

No academic qualifications 205, 209 19.5 (13.8 to 26.7) 20.7 (15.1 to 27.7) 40.8 (33.1 to 49.0) 26.9 (20.3 to 34.7) 33.7 (26.0 to 42.4) Academic qualifications typically gained at age 16 years 915, 888 15.6 (13.1 to 18.5) 17.6 (15.1 to 20.5) 48.5 (44.8 to 52.3) 25.2 (22.2 to 28.4) 33.0 (29.4 to 36.9) Studying for/attained further academic qualifications 1432, 1408 14.0 (11.9 to 16.3) 22.8 (20.4 to 25.4) 61.6 (58.6 to 64.5) 34.5 (31.6 to 37.4) 21.3 (18.8 to 24.0) Quintiles of Index of Multiple Deprivation p=0.0461 p=0.0866 p=0.8472 p=0.2125 p=0.1546 1–2 (lower deprivation) 1017, 1005 13.0 (10.7 to 15.8) 19.7 (17.2 to 22.6) 55.4 (51.8 to 59.0) 30.5 (27.4 to 33.8) 28.0 (24.8 to 31.5) 3 529, 535 14.5 (11.5 to 18.1) 19.6 (16.3 to 23.5) 54.2 (49.3 to 59.0) 27.3 (23.2 to 31.9) 27.5 (22.9 to 32.6) 4–5 (higher deprivation) 1161, 1098 17.7 (15.1 to 20.6) 23.5 (20.9 to 26.2) 54.0 (50.4 to 57.5) 32.2 (29.1 to 35.5) 23.8 (20.9 to 27.0) Area type¶ p=0.3805 p=0.8600 p=0.002 p=0.217 p=0.002 Urban 2201, 2166 14.9 (13.1 to 16.9) 21.3 (19.5 to 23.3) 56.1 (53.6 to 58.6) 31.1 (28.8 to 33.5) 24.6 (22.5 to 26.9) Rural 506, 472 16.7 (13.4 to 20.7) 21.0 (17.4 to 25.0) 47.5 (42.7 to 52.3) 28.1 (24.2 to 32.4) 33.2 (28.2 to 38.6) Relationship status p=0.0001 p<0.0001 p<0.0001 p<0.0001 p<0.0001

Area type¶ p=0.3805 p=0.8600 p=0.002 p=0.217 p=0.002 Urban 2201, 2166 14.9 (13.1 to 16.9) 21.3 (19.5 to 23.3) 56.1 (53.6 to 58.6) 31.1 (28.8 to 33.5) 24.6 (22.5 to 26.9) Rural 506, 472 16.7 (13.4 to 20.7) 21.0 (17.4 to 25.0) 47.5 (42.7 to 52.3) 28.1 (24.2 to 32.4) 33.2 (28.2 to 38.6) Relationship status p=0.0001 p<0.0001 p<0.0001 p<0.0001 p<0.0001 Married/civil partnership 617, 887 19.1 (15.9 to 22.8) 7.6 (5.7 to 10.2) 47.0 (42.6 to 51.4) 17.9 (14.8 to 21.5) 34.7 (30.5 to 39.0) Living with partner 570, 604 17.3 (14.1 to 21.0) 14.1 (11.4 to 17.3) 44.0 (39.4 to 48.7) 25.2 (21.5 to 29.3) 37.9 (33.4 to 42.6) In a steady relationship 698, 525 10.9 (8.6 to 13.6) 34.9 (30.8 to 39.1) 61.6 (57.4 to 65.6) 41.7 (37.3 to 46.1) 18.4 (15.3 to 21.9) Single, never married 816, 617 11.6 (9.1 to 14.7) 36.2 (32.5 to 40.0) 70.0 (66.5 to 73.4) 44.6 (40.7 to 48.6) 9.1 (7.0 to 11.6) Opposite-sex partners, past year p=0.0289 p<0.0001 p<0.0001 p<0.0001 p<0.0001 1 1802, 1913 16.3 (14.4 to 18.5) 15.4 (13.7 to 17.3) 50.2 (47.5 to 52.9) 25.6 (23.3 to 28.0) 31.5 (29.0 to 34.2) 2+ 882, 702 12.5 (10.2 to 15.4) 37.3 (33.6 to 41.1) 66.3 (62.7 to 69.8) 44.0 (40.1 to 47.9) 11.9 (9.5 to 14.6) Any STI diagnosis (excluding thrush) (past year) p=0.6319 p<0.0001 p=0.5533 p=0.0342 p=0.1018 No 2617, 2552 15.1 (13.5 to 16.9) 20.6 (18.9 to 22.3) 54.7 (52.4 to 56.9) 30.3 (28.2 to 32.4) 26.5 (24.4 to 28.7) Yes 57, 47 18.0 (8.5 to 34.1) 57.4 (42.6 to 71.1) 50.2 (35.7 to 64.7) 45.2 (31.1 to 60.1) 15.5 (7.5 to 29.5) Number of children p=0.0268 p<0.0001 p<0.0001 p<0.0001 p<0.0001 0 1893, 1614 13.5 (11.7 to 15.6) 27.8 (25.6 to 30.1) 61.3 (58.7 to 63.8) 36.8 (34.3 to 39.4) 19.4 (17.4 to 21.6) 1 382, 416 17.3 (13.3 to 22.1) 11.6 (8.7 to 15.3) 43.2 (37.6 to 48.9) 18.7 (14.9 to 23.2) 35.9 (30.4 to 41.8) 2+ 429, 605 18.6 (14.9 to 22.9) 10.5 (7.8 to 13.9) 44.6 (39.4 to 49.9) 22.3 (18.2 to 27.0) 37.4 (32.3 to 42.8) Partner used LARC only (past year) 177, 193 20.4 (14.1 to 28.5) 7.5 (4.4 to 12.6) 9.1 (5.4 to 15.0) 5.2 (2.6 to 10.0) 64.2 (55.6 to 71.9) *P-values indicate differences within exposure category (ie, source of contraceptive supplies). Row percentages are presented; the categories are not mutually exclusive and add up to more than 100%.

C only (past year) 177, 193 20.4 (14.1 to 28.5) 7.5 (4.4 to 12.6) 9.1 (5.4 to 15.0) 5.2 (2.6 to 10.0) 64.2 (55.6 to 71.9) *P-values indicate differences within exposure category (ie, source of contraceptive supplies). Row percentages are presented; the categories are not mutually exclusive and add up to more than 100%. †P-value presented corresponds to the categories: White, Black, Asian, Mixed/Other. ‡Participants aged ≥17 years. §English General Certificate of Secondary Education or equivalent. ¶Urban is defined as areas with a population of >=10 000 people. LARC, long-acting reversible contraception; STI, sexually transmitted infection.

C only (past year) 177, 193 20.4 (14.1 to 28.5) 7.5 (4.4 to 12.6) 9.1 (5.4 to 15.0) 5.2 (2.6 to 10.0) 64.2 (55.6 to 71.9) *P-values indicate differences within exposure category (ie, source of contraceptive supplies). Row percentages are presented; the categories are not mutually exclusive and add up to more than 100%. †P-value presented corresponds to the categories: White, Black, Asian, Mixed/Other. ‡Participants aged ≥17 years. §English General Certificate of Secondary Education or equivalent. ¶Urban is defined as areas with a population of >=10 000 people. LARC, long-acting reversible contraception; STI, sexually transmitted infection. Community clinics Marked age differences were also seen in the use of community clinics for contraception. There was a strong and consistent age gradient with over half of 16–17-year-old women sourcing contraception from community clinics compared with fewer than 10% of those aged 35–44 years. Other variation by demographic characteristic tended to reverse that found in relation to general practice. Community clinics were more commonly used by women living in urban areas, women in black and ethnic minority groups, those not co-habiting with a partner, and childless women. Variation was also seen by demographic characteristics, which were not significantly associated with use of general practice. Notable among these was area-related deprivation; women living in poorer areas more commonly sought contraceptive supplies from community clinics compared with those living in more affluent areas. Statistically significant associations between sourcing contraception from community clinics and sexual health-related variables were found which were also not seen for general practice. Use of community clinics was statistically higher among women diagnosed with an STI, those with more than one partner, and those who had an unplanned pregnancy in the past year.

en sourcing contraception from community clinics and sexual health-related variables were found which were also not seen for general practice. Use of community clinics was statistically higher among women diagnosed with an STI, those with more than one partner, and those who had an unplanned pregnancy in the past year. Among men, the age gradient in use of community clinics was even more marked than among women; 58.9% of 16–19-year-olds had used community clinics in the past year, compared with 5.2% of 35–44-year-olds. Statistically significant differences in use were seen by all the sexual health variables. Retail outlets Among women, few strong or consistent differences in use of retail outlets were seen by demographic characteristics, and none by sexual health variables. A strong gradient was seen by academic achievement with almost one-third of women studying for or having attained post-16 qualifications reporting use, compared with barely 15% of those with minimum educational qualifications or none. Other significant associations were observed in relation to parity and ethnicity, with women without children (31.9%) and Asian women (39.3%) most commonly reporting use of retail outlets. More variation by demographic characteristic was observed among men. Statistically significant differences were seen by age, educational level, area of residence, relationship status and parity. Use was highest among men aged 20–24 years, those with the highest educational qualifications, those living in urban areas, those who were neither married nor cohabiting, and those without children.

istically significant differences were seen by age, educational level, area of residence, relationship status and parity. Use was highest among men aged 20–24 years, those with the highest educational qualifications, those living in urban areas, those who were neither married nor cohabiting, and those without children. Use of more than one source of contraception Age differences were seen in the use of more than one source of contraceptive supplies. More than one-third of women aged under 25 years had used two or more sources in the past year. Statistically significant associations were also seen between multiple use and higher educational level, being unmarried, having more than one sexual partner in the past year, having been diagnosed with an STI, having experienced unplanned pregnancy, and being childless. A broadly similar pattern was observed among men. Statistically significant associations were seen between use of more than one source and being under 25, being unmarried, having two or more partners in the past year, an STI diagnosis in the past year and not having children.

Use of more than one source of contraception Age differences were seen in the use of more than one source of contraceptive supplies. More than one-third of women aged under 25 years had used two or more sources in the past year. Statistically significant associations were also seen between multiple use and higher educational level, being unmarried, having more than one sexual partner in the past year, having been diagnosed with an STI, having experienced unplanned pregnancy, and being childless. A broadly similar pattern was observed among men. Statistically significant associations were seen between use of more than one source and being under 25, being unmarried, having two or more partners in the past year, an STI diagnosis in the past year and not having children. Use of no source of contraception Some 13% of sexually active women reported having used no source of contraceptive supplies in the past year, although their partner may have done so. Age was associated with non-use, being most common among women in the 35–44 years age group. Not having used a source of contraception was significantly less common among women who were unmarried, childless, who had had an unplanned pregnancy during the past year, who had more than one sexual partner and who had been diagnosed with an STI. The proportion was highest among LARC users, rising to 30.8%.

ge group. Not having used a source of contraception was significantly less common among women who were unmarried, childless, who had had an unplanned pregnancy during the past year, who had more than one sexual partner and who had been diagnosed with an STI. The proportion was highest among LARC users, rising to 30.8%. Among men, prevalence of non-use was also higher in older age groups. Not having used a source of contraceptive supplies was significantly less common among unmarried men, those without children, those living in urban areas, and those identifying as Black Caribbean. It was also lower among men reporting more than one sexual partner and among those diagnosed with an STI in the past year. Discussion These data, from a large population-based national survey, provide information on patterns of service use for contraception among women and men in Britain. They show high proportions accessing a source of contraceptive supplies in the past year, nearly 90% of women and almost three-quarters of men. The majority of women, six in ten, use general practice and roughly the same proportion of men use retail suppliers. Little more than one in five women and men access contraception from community clinics, but those who do are younger and at greater risk of poor sexual health. Similar characteristics are seen for the sizeable minority who used more than one source in the past year.

ghly the same proportion of men use retail suppliers. Little more than one in five women and men access contraception from community clinics, but those who do are younger and at greater risk of poor sexual health. Similar characteristics are seen for the sizeable minority who used more than one source in the past year. A strength of this study is that it provides data from a large representative community sample rather than, as is more often the case, from attenders at health services.17–19 We were able to examine use of contraceptive sources by a wider range of factors than is usual, including not only sociodemographic characteristics but indicators of sexual and reproductive health status. Caution is needed in interpreting some findings due to the small numbers in some categories (eg, ethnic group). We also need to bear in mind that, in identifying the source of contraception used during the past year, we cannot be sure that participants were not referring to access by their partner and this is perhaps more likely to apply in the case of men. Nor were we able to establish which contraceptive methods were obtained from which sources, and participants may have interpreted ‘source’ as their main provider or all services involved in obtaining a contraceptive method. For example, contraceptive pill/oral contraceptives users may get their prescription from a general practitioner (GP) but may have to pick up their supplies from a pharmacy. Finally, we were unable to establish whether user preference for method determines choice of source, or vice versa. While most contraception is used for the prevention of pregnancy, other benefits such as management of acne or STI protection may also affect choice of source to obtain methods. The source from which women and men obtain a contraceptive method will depend on a number of variables, including which service is able to provide it. Condoms are rarely available free of charge from general practice in Britain; hormonal methods (emergency contraception excepted) are not available from retail outlets; and methods such as withdrawal and periodic abstinence require no contact with a provider.

of variables, including which service is able to provide it. Condoms are rarely available free of charge from general practice in Britain; hormonal methods (emergency contraception excepted) are not available from retail outlets; and methods such as withdrawal and periodic abstinence require no contact with a provider. Although direct comparisons with data from Natsal-2, conducted between 1999 and 2001, are not possible because of slight differences in question wording (Natsal-2 asked about use of sources of contraceptive advice or supplies while Natsal-3 asked only about supplies), we can nevertheless identify some shifts in patterns of use between the two surveys.10 A marked increase can be seen in use of any source of contraceptive supplies by the whole population aged 16–44 years, but especially among younger men. Two-thirds of men aged under 20 had accessed no source of contraception (for either advice or supplies) in 2000 but a decade later nine in ten reported accessing a service for supplies. Prevalence of sources of contraception is still ranked in the same order as was seen in Natsal-2. General practice remains the source used by the majority of women, the proportion of women using this source is virtually unchanged between the surveys (59.2% in Natsal-2). By contrast, although still accessed by smaller proportions, use of community clinics by both women and men increased between 2000 and 2010, as did use of retail outlets. In Natsal-2, 15.0% of women and 7.7% of men reported use of community clinics in the last year and 22.5% and 42.7%, respectively, reported use of retail. Commensurate with these changes, reported use of more than one source of contraception was higher in Natsal-3 than in Natsal-2, especially among men.

retail outlets. In Natsal-2, 15.0% of women and 7.7% of men reported use of community clinics in the last year and 22.5% and 42.7%, respectively, reported use of retail. Commensurate with these changes, reported use of more than one source of contraception was higher in Natsal-3 than in Natsal-2, especially among men. If these data represent true trends and not simply changes to question wording, they may reflect recent changes in public health policy and practice. The increase in the use of any source of contraception by men may follow from the now broader range of services available, including school- and college-based services. It may also be associated with a shift in focus from traditionally organised, women-centred family planning services towards integrated sexual and reproductive health services likely to have broader appeal in which staff may be better able to discuss contraception with young men attending for STI-related reasons.7 8 The increase in use of community clinics, which is consistent with increases in attendance observed using routinely collected clinic data,20 is likely to be welcomed by commissioners and providers in Britain as validating the considerable efforts made at the start of the 21st century to reach their target clientele. Their higher prevalence of use among the young, those living in more deprived areas, and those at greater sexual health risk is consistent with the growth in the number of dedicated young people’s services, increased accessibility, extended opening hours and capacity to provide drop-in services. The trend towards greater use of retail services by women may reflect greater availability of over-the-counter emergency contraception. The high proportion of LARC users reporting use of no service, almost a third, is likely to be attributable to the growing number of women using LARC methods which require few visits to services.

owards greater use of retail services by women may reflect greater availability of over-the-counter emergency contraception. The high proportion of LARC users reporting use of no service, almost a third, is likely to be attributable to the growing number of women using LARC methods which require few visits to services. This said, general practice is still the mainstay of contraceptive provision in Britain, at least among women. Healthcare professionals in this setting provide services to more women who are single, multi-partnered and living in more deprived areas with contraception than do those in community clinics. General practice supplies twice as many women with LARC methods than community clinics, which may reflect financial incentives to increase LARC uptake introduced in general practice.21 It follows that staff in general practice are still vitally important as a source of contraceptive supplies and need to maintain their skills and willingness to provide all methods. This is especially so in rural areas where community clinics are less commonly situated, a fact consistent with evidence in our data of lower rates of their use outside cities and towns. However, while there has been some evidence that LARC provision has been increasing in general practice, these methods may not be available in many practices requiring referral to more specialist services.17 22

y situated, a fact consistent with evidence in our data of lower rates of their use outside cities and towns. However, while there has been some evidence that LARC provision has been increasing in general practice, these methods may not be available in many practices requiring referral to more specialist services.17 22 We cannot tell whether the increase in use of multiple sources in general, and in particular among young people, is attributable to users ‘shopping around’ to find out which service best suits their needs, whether the mix of contraceptive methods they use requires use of different services or whether young people are more likely to be referred between different services compared with those in older age groups. These issues may require further investigation. Data from Natsal-3 on actual and preferred use of sources of contraceptive supplies showed that one-third of those aged under 25 years reported not having accessed their preferred source and this discordance was associated with use of less effective contraception and markers of poor sexual and reproductive health.23

n. Data from Natsal-3 on actual and preferred use of sources of contraceptive supplies showed that one-third of those aged under 25 years reported not having accessed their preferred source and this discordance was associated with use of less effective contraception and markers of poor sexual and reproductive health.23 These data on variation in the use of different sources by users’ characteristics will be of interest to those planning, purchasing and providing services. They may also provide evidence supporting the retention and expansion of community-based services. In England, local authorities are responsible for ensuring the availability of "confidential, open access services for STIs and contraception as well as reasonable access to all methods of contraception".24 This includes costs of the contraceptive methods or prescriptions and specialist sexual health services. Additional services provided in general practice, such as contraceptive advice and medical examinations of those seeking advice (excluding the fitting of any contraceptive devices), are commissioned by NHS England under the GP contract. Concerns have been raised that the move to the commissioning of enhanced contraceptive services to local authorities will lead to greater centralisation of services and fewer services providing more specialist care and facilities.25 Although access to contraception has widened over the last decade, there remains yet more potential for expanding services in community pharmacies and via the internet, such as by the provision of oral contraceptives.26–28 The increase in the choice of sources for contraceptive supplies and greater choice of contraceptive methods available from these sources have coincided with a decline in teenage pregnancy.29 30 A strong case can be made for continuing to provide high-quality services in different settings and with different styles of service delivery.

he choice of sources for contraceptive supplies and greater choice of contraceptive methods available from these sources have coincided with a decline in teenage pregnancy.29 30 A strong case can be made for continuing to provide high-quality services in different settings and with different styles of service delivery. Natsal-3 is a collaboration between University College London, the London School of Hygiene and Tropical Medicine, NatCen Social Research, Public Health England, and the University of Manchester. Contributors: RSF led on the writing of the paper. RG and KJ conducted the statistical analysis. All other authors (KW, AG, JD, CHM, MP, AMJ, WM) commented on drafts of the paper and were involved in the planning and conduct of the study. Funding: Natsal-3 was supported by grants from the Medical Research Council (G0701757)and the Wellcome Trust (084840), with contributions from the Economic and Social Research Council and Department of Health. Competing interests: AG receives an honorarium from HRPharma. None of the other authors report conflicting interests. Ethics approval: The Natsal-3 study was approved by the Oxfordshire Research Ethics Committee (Reference: 09/H0604/27). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Contraceptive use among men and women living with HIV in Malawi remains poor. Condom use is influenced more by relationship dynamics, fertility intention and self-efficacy than perception of infection or transmission risk. Contraceptive and antiretroviral use does not deter condom use. Introduction Despite significant progress over the past decade to reduce HIV transmission through increased availability of HIV testing, improved access to and utilisation of antiretroviral therapy (ART) and interventions to prevent maternal-to-child transmission (PMTCT), condoms continue to be under-utilised.1 Correct and consistent condom use can reduce HIV transmission risk by 80%.2 Condoms have been promoted among high-risk and HIV-positive individuals with limited success,3 despite recognition of their efficacy in reducing HIV, and prior to the last decade, their status as the principal biomedical intervention for prevention. Low utilisation of condoms has been associated with cost, religious ideology, alcohol or drug use, younger sexual debut, poor knowledge of HIV/AIDS, beliefs of diminished sexual pleasure and male emotional fulfilment, disbelief in prevention efficacy, distrust in relationships, gender inequality and perceptions of modesty.4

tilisation of condoms has been associated with cost, religious ideology, alcohol or drug use, younger sexual debut, poor knowledge of HIV/AIDS, beliefs of diminished sexual pleasure and male emotional fulfilment, disbelief in prevention efficacy, distrust in relationships, gender inequality and perceptions of modesty.4 People living with HIV can now live longer and healthier lives on ART, and as their life experiences change, their sexual behaviours, fertility desires and perceptions of condom need may also shift. Malawi has one of the highest HIV prevalence rates in the world at 10.6%.5 Malawi has achieved great success in curbing the HIV epidemic, demonstrating a declining incidence of HIV of approximately 35% from 2009 to 2012.6 Although condoms are freely distributed at ART clinics and educational messaging is integrated into clinic visits, consistent condom use has remained poor among individuals with HIV who receive ART, even among discordant couples.7 Understanding the dynamic influences that affect condom use among individuals with HIV in care is a critical step towards tailoring interventions to increase condom use. With the expansion of ART across the region, and increasing interest in the role of ART as a preventive strategy to reduce HIV infectivity, information on sexual practices among HIV-positive men and women on ART is critical. The purpose of this analysis is to identify factors associated with condom use among individuals infected with HIV receiving care in Lilongwe, Malawi.

sing interest in the role of ART as a preventive strategy to reduce HIV infectivity, information on sexual practices among HIV-positive men and women on ART is critical. The purpose of this analysis is to identify factors associated with condom use among individuals infected with HIV receiving care in Lilongwe, Malawi. Methods This is a sub-analysis of data collected in a cross-sectional study evaluating reproductive health knowledge, attitudes and practices among HIV-positive individuals. We received approval to conduct the study from the National Health Services Research Committee in Malawi, the institutional review board at Emory University, and the institutional review board at the University of North Carolina-Chapel Hill. The study population included participants who attended one of the Lighthouse Trust clinics, integrated HIV testing, treatment, and care clinics in Lilongwe, Malawi. Methods for recruitment and enrollment were previously published.8 Briefly, between 26 September 2013 and 20 December 2013, we enrolled HIV-positive individuals between the ages of 18 and 45 years who were sexually active in the last 6 months. A sample size of 600 individuals was selected based on feasibility within the study time frame. Individuals were recruited from the waiting area in the clinic and taken to a private area to confirm study eligibility, review and sign written consent, and complete a questionnaire. For this analysis, we excluded individuals who did not report sexual intercourse in the last month.

ibility within the study time frame. Individuals were recruited from the waiting area in the clinic and taken to a private area to confirm study eligibility, review and sign written consent, and complete a questionnaire. For this analysis, we excluded individuals who did not report sexual intercourse in the last month. Utilising a social-ecological framework, we conducted focus group discussions at the clinic prior to this study to inform the development of the questionnaire. Survey questions were a compilation of original study questions and questions used in the Malawi 2010 Demographic and Health Survey9 (DHS). Questions were pilot tested for clarity and intent prior to study initiation. The questionnaire was administered in Chichewa by a trained interviewer using a paper-based, semi-structured questionnaire with 160 questions for women and 130 questions for men. There were two primary outcomes of interest: self-reported condom use at last coitus (No vs Yes) and consistent condom use (Always vs Sometimes or Never) during intercourse over the past month. Condom use referred to use of either the male or female condom; however, female condom use is uncommon in Malawi.

Utilising a social-ecological framework, we conducted focus group discussions at the clinic prior to this study to inform the development of the questionnaire. Survey questions were a compilation of original study questions and questions used in the Malawi 2010 Demographic and Health Survey9 (DHS). Questions were pilot tested for clarity and intent prior to study initiation. The questionnaire was administered in Chichewa by a trained interviewer using a paper-based, semi-structured questionnaire with 160 questions for women and 130 questions for men. There were two primary outcomes of interest: self-reported condom use at last coitus (No vs Yes) and consistent condom use (Always vs Sometimes or Never) during intercourse over the past month. Condom use referred to use of either the male or female condom; however, female condom use is uncommon in Malawi. Potential correlates of interest fell into six domains: (1) Sociodemographics: age, marital status, education, religion; (2) HIV-related: years since diagnosis, current use of ART, partner aware of HIV status, do HIV medications reduce risk of transmission to partner; (3) Sexual characteristics and risk behaviour: partner’s serostatus, do you think your most recent partner has sex with others, number of partners in last month, sexual frequency, use of alcohol or drugs prior to or during sex in the last month; (4) Perception of condoms: effective at preventing pregnancy, effective at preventing HIV transmission risk; (5) Communication and self-efficacy: would you be able to refuse sex if your partner did not want to use a condom, in a relationship, who usually makes decisions on condom use; (6) Fertility intention and dual protection: do you want more children in the future; did you or your partner use contraceptive (other than condoms) at last intercourse, if partner is using contraceptive, do they need to use condoms.

6.7) 50 (23.8) Yes 205 (83.3) 160 (76.2) Believe one needs to use a condom if using also birth control method 0.064 No 69 (27.3) 44 (20.0) Yes 184 (72.7) 176 (80.0) Can refuse sex if partner did not want to use a condom <0.001 No 127 (50.6) 64 (29.4) Yes 124 (49.4) 154 (70.6) Decision-making about using condom use 0.11 7 I do (did) 131 (54.4) 93 (45.2) My partner does (did) 35 (14.5) 31 (15.0) We both do (did) equally 75 (31.1) 82 (39.8) Consistent condom use in past month 0.004 No 157 (61.8) 107 (48.6) Yes 97 (38.2) 113 (51.4) Condom use at last intercourse 0.002 No 113 (44.7) 67 (30.7) Yes 140 (55.3) 151 (69.3) *P value by t-test for continuous variables and Pearson chi-square or Fisher exact test for categorical variables. ART, antiretroviral therapy; DMPA, depot medroxprogesterone acetate; IUD, intrauterine device. Results We screened 623 individuals (349 women and 274 men) and enrolled a total of 562 study subjects (90.2%; 308 women and 254 men). A total of 475 individuals reported intercourse in the last month and were included in this analysis (255 females and 220 males). The mean age of the participants was 34.6 (range 18–45) years. The majority reported being married or in a committed relationship with one partner (n=468; 98.5%), sex with only one partner in the past month (n=456, 96.0%), and had at least one child (n=455; 95.7%). Eighty-eight (18.6%) had a child born with HIV, and 126 (26.6%) had transmitted HIV to a partner.

range 18–45) years. The majority reported being married or in a committed relationship with one partner (n=468; 98.5%), sex with only one partner in the past month (n=456, 96.0%), and had at least one child (n=455; 95.7%). Eighty-eight (18.6%) had a child born with HIV, and 126 (26.6%) had transmitted HIV to a partner. Characteristics of our cohort by gender are presented in table 1. The female participants were younger, more likely to desire children in the future, use contraception at last coitus and think their partner had sex with others. Males were more likely to use alcohol or drugs before or during coitus, know their partner’s status and be able to refuse sex without a condom. Among the female participants, 38.2% (97/254) and 55.3% (140/253) reported always using a condom in the past month and condom use at last intercourse, respectively. Males were significantly more likely to report consistent condom use (p=0.004) and condom use at last intercourse (p=0.002) compared with female participants, with 51.4% (113/220) and 69.3% (151/218) males reporting always using a condom in the past month and condom use at last intercourse, respectively. Among all participants, a majority believed that either the use of ART increases the risk of transmission to partners (n=212, 45.0%) or does not change the risk (n=102, 21.7%).

h 51.4% (113/220) and 69.3% (151/218) males reporting always using a condom in the past month and condom use at last intercourse, respectively. Among all participants, a majority believed that either the use of ART increases the risk of transmission to partners (n=212, 45.0%) or does not change the risk (n=102, 21.7%). Bivariate associations For the female participants, consistent condom use in the past month was significantly associated with non-Catholic status, knowing one’s partner’s HIV status, lower sexual frequency (once weekly or less), no desire for future children, using ART, recognising a role for dual protection (belief that one needs to use a condom when also using a birth control method), the ability to refuse sex without a condom, and shared decision making on condom use (table 2). These same factors (table 3) except for ART use were associated with condom use at last coitus. Additionally, higher education was also associated with condom use at last coitus. Table 2 Demographic, behavioural characteristics and beliefs of female and male participants by reporting consistent condom use in the last month

Bivariate associations For the female participants, consistent condom use in the past month was significantly associated with non-Catholic status, knowing one’s partner’s HIV status, lower sexual frequency (once weekly or less), no desire for future children, using ART, recognising a role for dual protection (belief that one needs to use a condom when also using a birth control method), the ability to refuse sex without a condom, and shared decision making on condom use (table 2). These same factors (table 3) except for ART use were associated with condom use at last coitus. Additionally, higher education was also associated with condom use at last coitus. Table 2 Demographic, behavioural characteristics and beliefs of female and male participants by reporting consistent condom use in the last month Characteristics Consistent condom use (always using condom in last month) Female (n=254) Male (n=220) No (n=157) (61.8%) Yes (n=97) (38.2%) P value* No (n=107) (48.6%) Yes (n=113) (51.4%) P value* n (%) n (%) n (%) n (%) Age (years, mean±SD) 32.0±6.2 32.2±6.7 0.746 36.8±5.4 38.0±4.9 0.086 Marital status 0.804 0.303 Married or committed relationship 153 (61.7) 95 (38.3) 106 (48.4) 113 (51.6) Single or dating 4 (61.7) 2 (33.3) 1 (100.0) 0 (0) Education 0.173 0.370 Some secondary or less 139 (63.5) 80 (36.5) 85 (50.3) 84 (49.7) Completed secondary or more 18 (51.4) 17 (48.6) 22 (43.1) 29 (56.9) Religion 0.009 0.934 Catholic 39 (78.0) 11 (22.0) 23 (48.9) 24 (51.1) Other 118 (57.8) 86 (42.2) 83 (48.3) 89 (51.7) Years since diagnosis (mean±SD) 4.5±3.6 4.9±3.2 0.418 4.0±3.3 5.2±4.7 0.024 Use of alcohol or drugs before/during sex in last month 0.851 0.011 No 139 (61.0) 89 (39.0) 70 (43.5) 91 (56.5) Yes 12 (63.2) 7 (36.8) 37 (62.7) 22 (37.3) Most recent sexual partner HIV-positive 0.063 0.058 No 27 (57.5) 20 (42.6) 14 (35.0) 26 (65.0) Yes 95 (58.6) 67 (41.4) 82 (50.0) 82 (50.0) Do not know 34 (77.3) 10 (22.7) 11 (68.8) 5 (31.3) Sexual frequency in last month 0.013 0.078 ≤weekly 64 (53.8) 55 (46.2) 46 (42.6) 62 (57.4) >1x/week 93 (68.9) 42 (31.1) 61 (54.5) 51 (45.5) Desire more children in the future 0.001 0.002 No 87 (54.4) 73 (45.6) 68 (42.2) 93 (57.8) Yes 70 (74.5) 24 (25.5) 38 (66.7) 19 (33.3) Contraceptive use at last intercourse (IUD, implant, DMPA, pills, Sterilisations, not including condom) 0.761 0.925 No 87 (62.6) 52 (37.4) 66 (48.9) 69 (51.1) Yes 68 (60.7) 44 (39.3) 41 (48.2) 44 (51.8) Currently using ART 0.083 0.374 No 20 (71.4) 8 (28.6) 16 (61.5) 10 (38.5) Yes, <2 years 59 (68.6) 27 (31.4) 37 (46.8) 42 (53.2) Yes, >2 years 78 (55.7) 62 (44.3) 54 (47.0) 61 (53.0) Believe HIV medications effect transmission risk to partner 0.742 0.602 Increases risk or no change 104 (63.0) 61 (37.0) 71 (47.7) 78 (52.4) Decreases risk 53 (60.9) 34 (39.1) 36 (51.4) 34 (48.6) Believe condoms are effective at preventing pregnancy 0.475 0.898 No 19 (67.9) 9 (32.1) 10 (50.0) 10 (50.0) Yes 137 (60.9) 88 (39.1) 97 (48.5) 103 (51.5) Believe condoms are effective at preventing HIV transm

sk or no change 104 (63.0) 61 (37.0) 71 (47.7) 78 (52.4) Decreases risk 53 (60.9) 34 (39.1) 36 (51.4) 34 (48.6) Believe condoms are effective at preventing pregnancy 0.475 0.898 No 19 (67.9) 9 (32.1) 10 (50.0) 10 (50.0) Yes 137 (60.9) 88 (39.1) 97 (48.5) 103 (51.5) Believe condoms are effective at preventing HIV transm ission to partner 0.139 0.314 No 29 (70.7) 12 (29.3) 20 (40.0) 30 (60.0) Yes 119 (58.3) 85 (41.7) 77 (48.1) 83 (51.9) Believe one needs to use a condom if also using birth control method 0.067 <0.001 No 49 (71.0) 20 (29.0) 33 (75.0) 11 (25.0) Yes 107 (58.5) 76 (41.5) 74 (42.1) 102 (58.0) Can refuse sex if partner did not want to use a condom <0.001 <0.001 No 111 (87.4) 16 (12.6) 52 (81.3) 12 (18.8) Yes 42 (34.2) 81 (65.9) 55 (35.7) 99 (64.3) Decision-making about using condom use <0.001 0.025 I do (did) 93 (71.5) 37 (28.5) 45 (48.4) 48 (51.6) My partner does (did) 22 (62.9) 13 (37.1) 20 (64.5) 11 (35.5) We both do (did) equally 28 (37.3) 47 (62.7) 30 (36.6) 52 (63.4) *P value by t-test for continuous variables and Pearson chi-square or Fisher exact test for categorical variables. ART, antiretroviral therapy; DMPA, depot medroxprogesterone acetate; IUD, intrauterine device. Table 3 Demographic, behavioural characteristics and beliefs of female and male participants by reporting condom use at last coitus

ission to partner 0.139 0.314 No 29 (70.7) 12 (29.3) 20 (40.0) 30 (60.0) Yes 119 (58.3) 85 (41.7) 77 (48.1) 83 (51.9) Believe one needs to use a condom if also using birth control method 0.067 <0.001 No 49 (71.0) 20 (29.0) 33 (75.0) 11 (25.0) Yes 107 (58.5) 76 (41.5) 74 (42.1) 102 (58.0) Can refuse sex if partner did not want to use a condom <0.001 <0.001 No 111 (87.4) 16 (12.6) 52 (81.3) 12 (18.8) Yes 42 (34.2) 81 (65.9) 55 (35.7) 99 (64.3) Decision-making about using condom use <0.001 0.025 I do (did) 93 (71.5) 37 (28.5) 45 (48.4) 48 (51.6) My partner does (did) 22 (62.9) 13 (37.1) 20 (64.5) 11 (35.5) We both do (did) equally 28 (37.3) 47 (62.7) 30 (36.6) 52 (63.4) *P value by t-test for continuous variables and Pearson chi-square or Fisher exact test for categorical variables. ART, antiretroviral therapy; DMPA, depot medroxprogesterone acetate; IUD, intrauterine device. Table 3 Demographic, behavioural characteristics and beliefs of female and male participants by reporting condom use at last coitus Characteristics Condom use at last coitus Female (n=253) Male (n=218) No (n=113) (44.7%) Yes (n=140) (55.3%) P value No (n=67) (30.7%) Yes (n=151) (69.3%) P-value* n (%) n (%) n (%) n (%) Age (years, mean±SD) 32.1±6.0 32.1±6.7 0.924 37.7±5.1 37.2±5.2 0.568 Marital status 0.790 Married or committed relationship 110 (44.5) 137 (55.5) Single or dating 3 (50.0) 3 (50.0) Education 0.066 0.562 Completed primary or less 102 (47.0) 115 (53.0) 53 (31.7) 114 (68.3) Some secondary or more 11 (30.6) 25 (69.4) 14 (27.5) 37 (72.6) Religion 0.034 0.595 Catholic 29 (58.0) 21 (42.0) 16 (34.0) 31 (66.0) Other 84 (41.4) 119 (58.6) 51 (30.0) 119 (70.0) Years since diagnosis (mean±SD) 4.5±3.6 4.8±3.4 0.515 4.1±3.4 4.9±4.4 0.197 Use of alcohol or drugs before/during sex in last month 0.447 0.053 No 99 (43.6) 128 (56.4) 43 (27.0) 116 (73.0) Yes 10 (52.6) 9 (47.4) 24 (40.7) 35 (59.3) Most recent sexual partner HIV positive 0.027 0.053 No 21 (45.7) 25 (54.3) 7 (18.0) 32 (82.1) Yes 65 (39.9) 98 (60.1) 52 (31.9) 111 (68.1) Do not know 27 (62.8) 16 (37.2) 8 (50.0) 8 (50.0) Sexual frequency in last month 0.029 0.004 ≤weekly 45 (37.5) 75 (62.5) 23 (21.5) 84 (78.5) >1x/week 68 (51.1) 65 (48.9) 44 (39.6) 67 (60.4) Desire more children in the future 0.004 0.015 No 60 (37.7) 99 (62.3) 42 (26.4) 117 (73.6) Yes 53 (56.4) 41 (43.6) 25 (43.9) 32 (56.1) Contraceptive use at last intercourse (IUD, implant, DMPA, pills, sterilisation, not including condom) 0.400 0.721 No 65 (46.8) 74 (53.2) 40 (29.9) 94 (70.2) Yes 46 (41.4) 65 (58.6) 27 (32.1) 57 (67.9) Currently using ART 0.169 0.886 No 17 (58.6) 12 (41.4) 9 (34.6) 17 (65.4) Yes, <2 years 40 (47.1) 45 (52.9) 23 (29.5) 55 (70.5) Yes, >2 years 56 (40.3) 83 (59.7) 35 (30.7) 79 (69.3) Believe HIV medications effect transmission risk to partner 0.307 0.168 Increases risk or no change 70 (42.7) 94 (57.3) 41 (27.9) 106 (72.1) Decreases risk 43 (49.4) 44 (50.6) 26 (37.1) 44 (62.9) Believe condoms are effective at preventing pregnancy 0.303 0.147 No 15 (53.6) 13 (46.4) 9 (45.0) 11 (55.0) Yes 97 (43.3) 127 (56.7) 58 (29.3) 140 (70.7) Believe condoms are effective at preventing HIV transmission to partner 0.856 0.536 No 18 (43.9) 23 (56.1) 12 (24.0) 38 (76.0) Ye

(49.4) 44 (50.6) 26 (37.1) 44 (62.9) Believe condoms are effective at preventing pregnancy 0.303 0.147 No 15 (53.6) 13 (46.4) 9 (45.0) 11 (55.0) Yes 97 (43.3) 127 (56.7) 58 (29.3) 140 (70.7) Believe condoms are effective at preventing HIV transmission to partner 0.856 0.536 No 18 (43.9) 23 (56.1) 12 (24.0) 38 (76.0) Ye s 86 (42.4) 117 (57.6) 45 (28.5) 113 (71.5) Believe one needs to use a condom if also using birth control method 0.057 0.004 No 37 (54.4) 31 (45.6) 21 (48.8) 22 (51.2) Yes 75 (41.0) 108 (59.0) 46 (26.3) 129 (73.7) Can refuse sex if partner did not want to use a condom <0.001 <0.001 No 87 (69.0) 39 (31.0) 38 (59.4) 26 (40.6) Yes 23 (18.6) 101 (81.5) 28 (18.4) 124 (81.6) Decision-making about using condom use 0.001 0.041 I do (did) 67 (51.2) 64 (48.9) 23 (24.7) 70 (75.3) My partner does (did) 14 (41.2) 20 (58.8) 14 (45.2) 17 (54.8) We both do (did) equally 19 (25.3) 56 (74.7) 18 (22.2) 63 (77.8) *P value by t-test for continuous variables and Pearson chi-square or Fisher exact test for categorical variables. ART, antiretroviral therapy; DMPA, depot medroxprogesterone acetate; IUD, intrauterine device.

s 86 (42.4) 117 (57.6) 45 (28.5) 113 (71.5) Believe one needs to use a condom if also using birth control method 0.057 0.004 No 37 (54.4) 31 (45.6) 21 (48.8) 22 (51.2) Yes 75 (41.0) 108 (59.0) 46 (26.3) 129 (73.7) Can refuse sex if partner did not want to use a condom <0.001 <0.001 No 87 (69.0) 39 (31.0) 38 (59.4) 26 (40.6) Yes 23 (18.6) 101 (81.5) 28 (18.4) 124 (81.6) Decision-making about using condom use 0.001 0.041 I do (did) 67 (51.2) 64 (48.9) 23 (24.7) 70 (75.3) My partner does (did) 14 (41.2) 20 (58.8) 14 (45.2) 17 (54.8) We both do (did) equally 19 (25.3) 56 (74.7) 18 (22.2) 63 (77.8) *P value by t-test for continuous variables and Pearson chi-square or Fisher exact test for categorical variables. ART, antiretroviral therapy; DMPA, depot medroxprogesterone acetate; IUD, intrauterine device. For the male participants, older age, longer time since HIV diagnosis, lower sexual frequency, no desire for future children, not using alcohol or drugs prior to or during sex, recognising a role for dual protection, the ability to refuse sex without a condom, and either making the decision or shared decision-making about condom use were associated with consistent condom use (table 2). Condom use at last coitus for men increased with lower sexual frequency, having a seronegative partner, no desire for future children, recognising a role for dual protection, the ability to refuse sex without a condom, and either making the decision or shared decision-making about condom use.

tent condom use (table 2). Condom use at last coitus for men increased with lower sexual frequency, having a seronegative partner, no desire for future children, recognising a role for dual protection, the ability to refuse sex without a condom, and either making the decision or shared decision-making about condom use. Among female and male participants there was no significant difference in condom use by perception of ART in reducing HIV transmission to partners (among all participants and among the subset of participants using ART, data not shown), perception of condoms for pregnancy prevention, or use of a modern contraceptive method at last coitus other than condoms. Notably, despite only 47% of the couples believing that their partner was monogamous, this perception did not alter condom use behaviour.

mong the subset of participants using ART, data not shown), perception of condoms for pregnancy prevention, or use of a modern contraceptive method at last coitus other than condoms. Notably, despite only 47% of the couples believing that their partner was monogamous, this perception did not alter condom use behaviour. Multivariate models Among both men and women, being able to refuse sex without a condom was the strongest predictor of condom use in our multivariate model, both for consistent condom use and condom use at last coitus (table 4). Among the female participants, other factors associated with increased consistent condom use included using ART for at least 2 years and joint decision-making about condom use. The desire to have children in the future was negatively associated with consistent condom use. These same factors, except for ART use, were associated with condom use at last coitus among women. For men, in addition to the ability to refuse sex without a condom, joint decision-making about condom use was associated with increased consistent condom use, and increased sexual frequency was associated with decreased consistent condom use. Only the ability to refuse sex without a condom was associated with increased condom use at last intercourse among men. Table 4 Multivariate poisson regression models evaluating factors associated with evaluating consistent condom use and condom use at last coitus for female and male participants

Multivariate models Among both men and women, being able to refuse sex without a condom was the strongest predictor of condom use in our multivariate model, both for consistent condom use and condom use at last coitus (table 4). Among the female participants, other factors associated with increased consistent condom use included using ART for at least 2 years and joint decision-making about condom use. The desire to have children in the future was negatively associated with consistent condom use. These same factors, except for ART use, were associated with condom use at last coitus among women. For men, in addition to the ability to refuse sex without a condom, joint decision-making about condom use was associated with increased consistent condom use, and increased sexual frequency was associated with decreased consistent condom use. Only the ability to refuse sex without a condom was associated with increased condom use at last intercourse among men. Table 4 Multivariate poisson regression models evaluating factors associated with evaluating consistent condom use and condom use at last coitus for female and male participants Consistent condom use (past month) Use at last coitus Female Male Female Male (APRs (95% CI)) (APRs (95% CI)) (APRs (95% CI)) (APRs (95% CI)) Age 1.01 (0.98 to 1.04) Education Some secondary or less 1 Completed secondary or more 1.20 (0.94 to 1.51) Catholic Yes 1 1 No 1.50 (0.89 to 2.52) 1.18 (0.86 to 1.61) Years since diagnosis 1.01 (0.99 to 1.04) Use of alcohol or drugs before/during sex in last month No 1 1 Yes 0.82 (0.60 to 1.11) 0.99 (0.81 to 1.20) Sexual frequency in last month ≤weekly 1 1 1 1 >1 x/week 0.79 (0.62 to 1.00) 0.96 (0.76 to 1.22) 0.87 (0.72 to 1.05) 0.90 (0.76 to 1.05) Currently using ART No 1 <2 years 1.31 (0.85 to 2.02) ≥2 years 1.52 (1.03 to 2.23)* Desire more children in the future No 1 1 1 1 Yes 0.69 (0.51 to 0.93)* 0.84 (0.57 to 1.24) 0.74 (0.59 to 0.93)* 0.95 (0.75 to 1.20) Most recent sexual partner HIV-positive Yes 0.89 (0.70 to 1.13) 0.87 (0.67 to 1.13) 1.27 (0.96 to 1.70) 0.88 (0.75 to 1.03) No 1 1 1 1 I do not know 0.66 (0.42 to 1.03) 0.59 (0.28 to 1.24) 1.24 (0.52 to 1.51) 0.66 (0.41 to 1.08) Believe one needs to use a condom if using also birth control method No 1 1 1 1 Yes 1.09 (0.80 to 1.48) 1.88 (1.16 to 3.05)* 1.08 (0.84 to 1.39) 1.24 (0.96 to 1.60) Can refuse sex if partner did not want to use a condom No 1 1 1 1 Yes 4.33 (2.72 to 6.90)* 2.77 (1.68 to 4.58)* 2.27 (1.74 to 2.95)* 1.70 (1.28 to 2.27)* Decision-making about using condom use I do 1 1 1 1 My partner does 1.34 (0.88 to 2.04) 0.89 (0.56 to 1.41) 1.28 (0.96 to 1.70) 0.81 (0.59 to 1.11) We both do 1.50 (1.15 to 1.97)* 0.98 (0.77 to 1.26) 1.24 (1.02 to 1.51)* 0.93 (0.80 to 1.10) The prevalence ratios presented are adjusting for all variables with results included in the table for the specific gender and outcome.

artner does 1.34 (0.88 to 2.04) 0.89 (0.56 to 1.41) 1.28 (0.96 to 1.70) 0.81 (0.59 to 1.11) We both do 1.50 (1.15 to 1.97)* 0.98 (0.77 to 1.26) 1.24 (1.02 to 1.51)* 0.93 (0.80 to 1.10) The prevalence ratios presented are adjusting for all variables with results included in the table for the specific gender and outcome. *P<0.05. APRs, adjusted prevalence ratios; ART, antiretroviral therapy CI , CI interval. Discussion Although our data demonstrate that condom use continues to be inconsistent among this study population, overall data from the Demographic and Health Surveys (DHS) in sub-Saharan Africa suggests increased condom use over the last one to two decades.10 Most participants recognise that condoms reduce HIV transmission risk; however, safe sexual practices seem to be influenced more by other factors, such as relationship dynamics, fertility intention and self-efficacy than perception of infection or transmission risk.

reased condom use over the last one to two decades.10 Most participants recognise that condoms reduce HIV transmission risk; however, safe sexual practices seem to be influenced more by other factors, such as relationship dynamics, fertility intention and self-efficacy than perception of infection or transmission risk. Similar to prior studies, our data demonstrate that couples that jointly make decisions regarding condom use more consistently use condoms, whereas when individuals independently make these decisions, condoms are less often used.11 Negotiation of condom use can be challenging, particularly for women. Promotion efforts to develop a plan for safe sexual practices should be aimed at the couple. When both men and women are able to refuse sex without a condom,4 12 condom use improves, demonstrating that self-efficacy is a strong predictor of condom utilisation. Focused efforts to develop skills that allow individuals to negotiate for condom use and overcome the specific obstacles within relationships that reduce power to refuse sex without a condom could improve self-efficacy and condom use within these groups. Further, these findings suggest that poor communication within a relationship may inhibit condom use.4 13 14 The development of communication strategies for couples at greatest risk, designed to strengthen dialogue regarding sexual and reproductive health within the relationship and increase condom use, could help address the discrepancies between risk and behaviour.

cation within a relationship may inhibit condom use.4 13 14 The development of communication strategies for couples at greatest risk, designed to strengthen dialogue regarding sexual and reproductive health within the relationship and increase condom use, could help address the discrepancies between risk and behaviour. We found that use of effective contraceptive methods did not alter condom use behaviours, further reinforcing the finding that effective pregnancy prevention method promotion will not detract from sexually transmitted infection/HIV prevention efforts. Rather, for men, believing in the need for ‘dual protection’ increased condom use. Similar to other studies, we did not find that ART use and knowledge that ART use decreases transmission risk alters condom use behaviour among our cohort of primarily ART users.15 Rather, similar to one prior study, we found the opposite trend: increased condom use among women on ART for a longer time.16 These results echo what others are reporting: the promotion of specific HIV prevention strategies, such as ART treatment as prevention, will not reduce other HIV prevention behaviours.17 It is unclear why there is a high proportion of individuals who perceive their risk of transmission to partners increases with the use of ART or does not change. This finding may be specific to this cohort or may reflect a broader misunderstanding of the association between low viral load and decreased transmission potential. Further, it is possible that providers choose to not discuss this association given concern that it may lead to altered risk behaviour. Future efforts need to explore patient perceptions further, and potentially increase awareness of the impact of viral suppression on HIV transmission.

decreased transmission potential. Further, it is possible that providers choose to not discuss this association given concern that it may lead to altered risk behaviour. Future efforts need to explore patient perceptions further, and potentially increase awareness of the impact of viral suppression on HIV transmission. The trend toward reduced condom use with greater sexual frequency noted in our study has been reported in other studies,4 suggesting that user fatigue may play a role in condom non-use. Another finding echoed by others is that alcohol use decreases condom use.18 Whereas other studies have noted demographic differences, such as age and education, to be predictive of condom use, demographic differences in our cohort were not predictive. Thus, efforts to improve condom use need to be applied broadly, irrespective of these factors.

thers is that alcohol use decreases condom use.18 Whereas other studies have noted demographic differences, such as age and education, to be predictive of condom use, demographic differences in our cohort were not predictive. Thus, efforts to improve condom use need to be applied broadly, irrespective of these factors. A strength of this study is that we applied two measures used in prior research to evaluate for condom use, allowing us to evaluate different trends that may be related to consistent use over use at last intercourse. A limitation, however, is that both of these are self-reported measures and reporting bias may influence the accuracy of the responses. Social desirability bias may have led both men and women to over-report condom use or other variables such as decision-making for condom use. Another strength of this study is the parallel evaluation of men and women, allowing us to evaluate trends that may differ by gender. However, we did not evaluate couples, so we cannot compare responses or comment on the dynamics that may be related to sexual practices within a specific relationship. We are limited in our ability to comment on any causal relationship between factors evaluated (such as ART use or contraception use) and condom use; a longitudinal evaluation is recommended to further evaluate this dynamic relationship. As we are conducting this evaluation at two clinics in Lilongwe where study population characteristics, such as high rates of monogamy and ART use, may differ from other settings, the generalisability of our findings may be limited. Lastly, although this is one of the larger evaluations of men and women with HIV receiving care, we may still have limited power in our predictive models to detect some true associations, especially in cases where we have lower prevalence of those covariates (such as non-ART users or single individuals). Non-significant findings may be echoed in other studies and add to our evolving understanding on drivers of condom use.

till have limited power in our predictive models to detect some true associations, especially in cases where we have lower prevalence of those covariates (such as non-ART users or single individuals). Non-significant findings may be echoed in other studies and add to our evolving understanding on drivers of condom use. In conclusion, our study highlights that among high-risk men and women living with HIV, condom use continues to be inconsistent. Previously proven interventions to reduce sexual risk behaviours19 must be personalised to each setting, with efforts to target specific barriers that exist for that community. Among the individuals in our clinical setting, HIV- and ART-related factors and contraceptive use were not central in determining condom use. Condom use was influenced more by relationship dynamics, fertility intention and self-efficacy than perception of infection or transmission risk. To effectively target future efforts at HIV and STI prevention, we must recognise the multiple influences that determine condom use, explore methods and employ tools to facilitate condom negotiation and joint decision-making. The authors would like to thank the study participants, the study research assistants (Justin Milonde and Felix Mtunga), and the staff of the Lighthouse Trust Clinics for their assistance and support of the study. Contributors: LH conceived the study and drafted the survey with JT and HT. JT oversaw its operational aspects with SP, HT, TC, BS and LM. LH conducted data analysis and drafted the initial manuscript. All authors read and approved the final manuscript.

The authors would like to thank the study participants, the study research assistants (Justin Milonde and Felix Mtunga), and the staff of the Lighthouse Trust Clinics for their assistance and support of the study. Contributors: LH conceived the study and drafted the survey with JT and HT. JT oversaw its operational aspects with SP, HT, TC, BS and LM. LH conducted data analysis and drafted the initial manuscript. All authors read and approved the final manuscript. Funding: This study was supported by a grant from the Society in Family Planning (PI: Lisa Haddad). Dr Haddad received support through a career development award from the National Institute of Health (1K23HD078153-01A1). Competing interests: None declared. Patient consent: Obtained. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Menstrual restrictions are common among women in the Terai region of Nepal. When controlling for demographic variables, no form of IPV was associated with high menstrual restrictions. Future research should seek to identify how best to capture menstrual restrictions and stigma, including deviations around such norms. Introduction After a decade of research and advocacy by global health experts, menstrual health and hygiene is now recognised as a public health concern.1 2 In low- and middle-income countries, in particular, recent efforts have focused on dealing with the lack of puberty education and menstrual products and the inadequacy of water and sanitation facilities in schools.1 By some estimates, only 47% of schools in low-income countries have water coverage and only 46% have sanitation infrastructure.3 The failure to provide girls with information and hygienic facilities reinforces the notion that menses are a private matter to be hidden and, in part, explains why many girls miss school while menstruating.4 5 Experiencing shame and fear related to menstruation can diminish girls’ self-confidence6 and has implications for their health and well-being into adulthood.7

hygienic facilities reinforces the notion that menses are a private matter to be hidden and, in part, explains why many girls miss school while menstruating.4 5 Experiencing shame and fear related to menstruation can diminish girls’ self-confidence6 and has implications for their health and well-being into adulthood.7 Emerging evidence has linked women’s sanitation and hygiene experiences with increased vulnerabilities to violence.8 9 Specifically, poor access to public toilet facilities—either they are too far away or too few—or poor design of the facilities, dimly lit or isolated, leaves women at risk of sexual harassment, assault and rape.10 11 According to the UN, lack of access is widespread: one in three women does not have access to safe, adequate toilets.12 A recent population-based study in Kenya directly linked violence against women and inadequate sanitation. Winter and Barchi11 found that Kenyan women who practised open defaecation had greater odds (adjusted OR = 1.387) of reporting non-partner violence than women who had access to improved sanitation.11 Women and girls who are menstruating and require more frequent use of these facilities are at increased risk of abuse.9 Menstruation has been linked to gender-based violence in other ways as well. Girls attending school while menstruating can experience psychological abuse and ridicule from male classmates, particularly if they soil their clothing.13

ting and require more frequent use of these facilities are at increased risk of abuse.9 Menstruation has been linked to gender-based violence in other ways as well. Girls attending school while menstruating can experience psychological abuse and ridicule from male classmates, particularly if they soil their clothing.13 There is some evidence of increasing research and programmes focusing on menstrual health and women and girls' vulnerability to violence outside the home, but little research has specifically investigated links between violence within the home (ie, intimate partner violence (IPV)) and menstrual health. IPV may be linked to menstrual health owing to power differentials that disfavour women as well as societal norms that stigmatise menstruation. For instance, in Nepal and other global settings, the onset of menses signals girls’ sexual maturity, triggering their removal from school and increasing their vulnerability to forced and/or early marriages14; early marriage has been reported to increase a women’s likelihood of experiencing IPV.15

gmatise menstruation. For instance, in Nepal and other global settings, the onset of menses signals girls’ sexual maturity, triggering their removal from school and increasing their vulnerability to forced and/or early marriages14; early marriage has been reported to increase a women’s likelihood of experiencing IPV.15 IPV has also consistently been associated with reduced reproductive autonomy and sexual decision-making among women.16 17 Perpetrators of IPV and reproductive coercion are male partners as well as extended family members, including in-laws.18 Similarly, in Nepal, women’s and girls’ autonomy is often restricted during menstruation by family members, husbands and in-laws. Common practices include forbidding a menstruating woman or girl to pray or enter a place of worship, cook or enter the kitchen, or touch men or children.19 20 In extreme cases, a woman may be banished to small outbuildings while menstruating. This practice, known as Chaupadi, is generally prevalent in certain areas within the midwest and far-western development regions of Nepal. One study reported that 21% of 672 women reported such experiences21; and recent media reports have highlighted the potential lethality of such practices.22 23 IPV prevalence in Nepal mirrors the global average, with one in three women facing this type of abuse.24 25 The extent to which IPV might influence or co-occur with these restrictive menstrual practices is unclear.

experiences21; and recent media reports have highlighted the potential lethality of such practices.22 23 IPV prevalence in Nepal mirrors the global average, with one in three women facing this type of abuse.24 25 The extent to which IPV might influence or co-occur with these restrictive menstrual practices is unclear. Despite the shared controlling nature of IPV and menstrual restriction and the shared health consequences, no studies to date have examined the relationship between these two phenomena. Thus, this study aims to record the prevalence of menstrual restrictions experienced by married women, and examine potential associations between IPV in the past year and menstrual restrictions imposed by husbands and/or in-laws among women in three districts of Nepal: Nawalparasi, Kapilvastu and Chitwan. Methods Our analysis used baseline data from a larger randomised control trial26 aiming to reduce IPV in three districts within the Terai region in Nepal (n=1440 female community respondents, n=360 intensive intervention participants, ages 18–49) to document the prevalence of such restrictive practices. The study received ethical approval from George Mason University (8 02 242–1), Emory University (IRB00091115) and the Nepal Health Research Council (178/2015).

in Nepal (n=1440 female community respondents, n=360 intensive intervention participants, ages 18–49) to document the prevalence of such restrictive practices. The study received ethical approval from George Mason University (8 02 242–1), Emory University (IRB00091115) and the Nepal Health Research Council (178/2015). Measures Emotional IPV was assessed using the WHO Multi-Country Study on Women’s Health and Domestic Violence scale.27 Specifically, each woman was asked if her husband had insulted or made her feel bad, belittled or humiliated her in front of other people, done things to scare or intimidate her and threatened to hurt her or someone she cared about in the past 12 months. Economic IPV was assessed using the United Nations Multi-Country Study on Men and Violence three-item scale, which asks if in the past year the woman’s husband had prevented her from getting a job, going to work, trading or earning money; taken her earning or valuables against her will; and kept money from his earnings for alcohol, tobacco or other things for himself when he knew she having trouble affording the household expenses.28 Using the Standard Outcomes for Assessment of Intimate Partner Violence from the What Works to Prevent Violence Global Programme, we assessed physical and sexual IPV in the past year.29 Each woman reported how often her husband had slapped or thrown something at her; pushed or shoved her; hit her with his fist or something else; kicked, dragged, beaten, choked or burned her; threatened or actually used a gun, knife or other weapon against her; physically forced her to have sex; made her feel so afraid that she had sex when she did not want to; and forced her to do something sexual she did not want to. Dichotomous variables were created for each type of violence. Women who answered in the affirmative for any scale item were considered to have experienced that type of violence. A dichotomous ’any IPV' variable was also created. A woman was coded as having experienced any IPV if she reported experiencing emotional, physical or sexual violence.

e created for each type of violence. Women who answered in the affirmative for any scale item were considered to have experienced that type of violence. A dichotomous ’any IPV' variable was also created. A woman was coded as having experienced any IPV if she reported experiencing emotional, physical or sexual violence. Menstrual restriction was assessed using a six-item scale designed in consultation with local Nepalese non-governmental organisations to capture common experiences. Each woman was asked if her husband or in-laws expected her to do any of the following during menstruation: not sleep with her husband, stay away from everyone, not enter a temple or place of worship, stop praying (outside of the temple), not cook, or other restriction. Women could choose all factors that applied. A dichotomous variable was created: women who chose one or none of the items were coded as ‘low menstrual restriction’ and women who selected two or more items were coded as ‘high menstrual restriction.’ Since this is among the first studies to quantify menstrual restrictions, the variable was coded to capture the quartile of respondents who experienced most restrictions.

e or none of the items were coded as ‘low menstrual restriction’ and women who selected two or more items were coded as ‘high menstrual restriction.’ Since this is among the first studies to quantify menstrual restrictions, the variable was coded to capture the quartile of respondents who experienced most restrictions. The demographic variables were drawn from previous research in Nepal on IPV30 31 and included a continuous measure of age and categorical measures of education and husband’s education (none, primary, secondary and higher). Additionally, categorical variables were created for age at marriage (≤14, 15–17 and ≥18) and number of children (0, 1–3, ≥4). A dichotomous measure was created for household income stress (yes/no). In consultation with in-country partners, caste was coded as a dichotomous variable, comprising advantaged castes (eg, upper caste, advantaged Janajatis) and disadvantaged castes/religious minorities (eg, disadvantaged Janajatis, Dalit, Muslim).

). A dichotomous measure was created for household income stress (yes/no). In consultation with in-country partners, caste was coded as a dichotomous variable, comprising advantaged castes (eg, upper caste, advantaged Janajatis) and disadvantaged castes/religious minorities (eg, disadvantaged Janajatis, Dalit, Muslim). Analysis Descriptive statistics of all demographic, menstrual restriction and violence variables were obtained. Bivariate cross-tabulations using Pearson’s χ2 tests assessed the association between demographics and menstrual restrictions, and between menstrual restriction and IPV. Generalised estimating equation logistic regressions were used to assess the relationship between IPV and menstrual restriction while accounting for the clustering within the data. Unadjusted logistic regressions were conducted with each type of IPV (separately) as the predictor and high menstrual restriction as the outcome. Finally, we used two adjusted logistic regressions. The first included any IPV and menstrual restriction, while controlling for demographic variables. The second adjusted logistic regression included all the (separate) types of IPV, so as to isolate the association of each type with the outcome, menstrual restriction, while also controlling for the demographic variables.

he first included any IPV and menstrual restriction, while controlling for demographic variables. The second adjusted logistic regression included all the (separate) types of IPV, so as to isolate the association of each type with the outcome, menstrual restriction, while also controlling for the demographic variables. Patient statement Participants were not directly involved in the design of the research. However, the research team relied on the local knowledge and expertise of our in-country implementation partners to help devise and carry out a study reflective of the communities’ needs and experiences. After conclusion of the larger study, the results of this paper and others will be disseminated to participants via community meetings. Results As shown in table 1, participants had a mean age of 34.5. Around one in three women (31.2%) had no education and almost one in five (18.5%) had more than a secondary education. Husbands were slightly more educated: 14.1% had no education, while 28.4% had attended higher than secondary school. Nearly half (48.5%) of the sample had been married before the age of 18—35.9% between the ages of 15 and 17 and 12.6% at age ≤14. The majority of women (79.7%) had between 1 and 3 children. Just under half of the participants (44.9%) reported experiencing household income stress and just over half (52.9%) were from a disadvantaged caste. Table 1 Demographic characteristics and associations with menstrual restrictions

Results As shown in table 1, participants had a mean age of 34.5. Around one in three women (31.2%) had no education and almost one in five (18.5%) had more than a secondary education. Husbands were slightly more educated: 14.1% had no education, while 28.4% had attended higher than secondary school. Nearly half (48.5%) of the sample had been married before the age of 18—35.9% between the ages of 15 and 17 and 12.6% at age ≤14. The majority of women (79.7%) had between 1 and 3 children. Just under half of the participants (44.9%) reported experiencing household income stress and just over half (52.9%) were from a disadvantaged caste. Table 1 Demographic characteristics and associations with menstrual restrictions Total (n=1800) Low menstrual restriction (%) High menstrual restriction (%) P values Total (%) 100 27.7 72.3 Demographics Age (years) 34.5 33.9 34.8 0.047 Education None 562 (31.2) 36.3 63.7 <0.001 Primary 444 (24.7) 24.6 75.5 Secondary 461 (25.6) 26.5 73.5 Higher 333 (18.5) 19.2 80.8 Husband’s education None 253 (14.1) 34.0 66.0 <0.001 Primary 397 (22.1) 32.5 67.5 Secondary 636 (35.4) 27.8 72.2 Higher 510 (28.4) 20.4 79.6 Age at marriage (years) ≤14 226 (12.6) 30.5 69.5 0.415 15–17 646 (35.9) 28.5 71.5 ≥18 928 (51.6) 26.5 73.5 Number of children None 77 (4.3) 28.6 71.4 0.955 1–3 1435 (79.7) 27.8 72.2 4+ 288 (16.0) 27.1 72.9 Income stress Yes 806 (44.9) 26.6 73.4 0.269 No 990 (55.1) 28.9 71.1 Caste/ethnicity Disadvantaged 952 (52.9) 40.2 59.8 <0.001 Advantaged 848 (47.1) 13.7 86.3 Experience of menstrual restriction was widespread. Almost three out of four women (72.3%) reported experiencing high menstrual restriction, or two or more types of menstrual restriction. Experience of menstrual restriction varied significantly by age (p=0.047) and by education. Women who were more educated (p<0.001) or who had husbands who were more educated (p<0.001) reported higher levels of menstrual restriction. Women from an advantaged caste also reported higher levels of menstrual restriction (p<0.001).

nce of menstrual restriction varied significantly by age (p=0.047) and by education. Women who were more educated (p<0.001) or who had husbands who were more educated (p<0.001) reported higher levels of menstrual restriction. Women from an advantaged caste also reported higher levels of menstrual restriction (p<0.001). As shown in table 2, in the past year 40.4% of women reported any IPV, 29% reported emotional IPV, 17.5% reported economic IPV, 15.7% reported physical IPV and 18.1% reported sexual violence. Unadjusted odds ratios showed that with the exception of physical violence, all other types of IPV were not significantly associated with high menstrual restrictions. When controlling for demographic variables and IPV, adjusted regressions indicated that no type of IPV was associated with high menstrual restrictions. Table 2 Unadjusted and adjusted logistic regressions for associations between past-year intimate partner violence (IPV) and menstrual restrictions

As shown in table 2, in the past year 40.4% of women reported any IPV, 29% reported emotional IPV, 17.5% reported economic IPV, 15.7% reported physical IPV and 18.1% reported sexual violence. Unadjusted odds ratios showed that with the exception of physical violence, all other types of IPV were not significantly associated with high menstrual restrictions. When controlling for demographic variables and IPV, adjusted regressions indicated that no type of IPV was associated with high menstrual restrictions. Table 2 Unadjusted and adjusted logistic regressions for associations between past-year intimate partner violence (IPV) and menstrual restrictions Totals N (%) Low menstrual restriction (%) High menstrual restriction (%) OR, 95% CI P values AOR, 95% CI P values (n=499) (n=1301) IPV variables Any IPV* 726 (40.4) 41.9 39.8 0.91 (0.74 to 1.13) 0.398 1.05 (0.84 to 1.32) 0.674 Emotional IPV 522 (29.0) 29.1 29.0 1.00 (0.79 to 1.25) 0.996 1.30 (0.95 to 1.76) 0.100 Economic IPV 315 (17.5) 20.2 16.5 0.84 (0.64 to 1.09) 0.526 0.91 (0.68 to 1.22) 0.535 Physical IPV 282 (15.7) 19.4 14.2 0.68 (0.52 to 0.90) 0.006 0.73 (0.51 to 1.05) 0.095 Sexual IPV 325 (18.1) 18.8 17.8 0.95 (0.73 to 1.23) 0.679 1.10 (0.81 to 1.51) 0.543 Demographics Age 1.01 (0.99 to 1.02) 0.365 Education Primary 1.42 (1.05 to 1.92) 0.024 Secondary 1.11 (0.78 to 1.59) 0.554 Higher 1.35 (0.84 to 2.17) 0.210 None (Ref) Husband’s education Primary 0.96 (0.68 to 1.37) 0.839 Secondary 1.08 (0.75 to 1.53) 0.678 Higher 0.97 (0.63 to 1.50) 0.900 None (Ref) Caste Advantaged 3.94 (2.97 to 5.20) <0.001 Disadvantaged (Ref) Age at marriage 15–17 1.12 (0.79 to 1.58) 0.527 18+ 1.00 (0.70 to 1.43) 0.983 ≤14 (Ref) Number of children 1–3 0.72 (0.41 to 1.24) 0.236 4+ 0.85 (0.45 to 1.62) 0.622 None (Ref) Income stress Yes 1.04 (0.83 to 1.32) 0.726 No (Ref) *Any IPV includes emotional, physical or sexual violence. The adjusted logistic regression for any IPV, does not include the other IPV variables, but does include menstrual restriction (the main outcome) and demographic control variables (not reported).

22 None (Ref) Income stress Yes 1.04 (0.83 to 1.32) 0.726 No (Ref) *Any IPV includes emotional, physical or sexual violence. The adjusted logistic regression for any IPV, does not include the other IPV variables, but does include menstrual restriction (the main outcome) and demographic control variables (not reported). Discussion These findings indicate that women’s participation in family and social life during menstruation is widely restricted by both husbands and husbands’ family members in this sample of married women from three districts in the Terai region of Nepal. Our study sought to examine whether IPV correlated with experience of menstrual restriction; no such association was identified. Because this study is among the first to assess both menstrual restrictions and potential associations with IPV experiences, further research is clearly needed. It may be that owing to the prevalence of self-reported menstrual restriction among participants in this sample, significant associations were not found because of limited variation in the outcome of interest. Furthermore, some 7 in 10 Nepalese women reported high levels of menstrual restriction in this sample, thus indicating that such restrictions reflect a socially accepted practice. This study did not examine whether or not women accepted this practice, and neither did it investigate whether women resisted or deviated from this socially acceptable practice. Possibly, women who challenge or defy socially acceptable practices, such as menstrual restrictions, may be more vulnerable to IPV. Similar to IPV and gender norms research, it would be important for future research to identify specific situations that are less socially acceptable than others. For instance, challenging the menstrual restriction of not entering a temple might have elicited more negative responses than not entering the kitchen. One limitation of the study is that measurement of menstrual restriction has not been standardised. This study is a critical first step in this process, but additional research is needed. Future research should also seek to identify how best to capture menstrual stigma, norms, and deviations around such norms to obtain a better understanding of this gendered phenomenon.

enstrual restriction has not been standardised. This study is a critical first step in this process, but additional research is needed. Future research should also seek to identify how best to capture menstrual stigma, norms, and deviations around such norms to obtain a better understanding of this gendered phenomenon. The study of social norms and menstrual restrictions is particularly important in light of recent events in Nepal. In August 2017, Nepal passed a law criminalising Chaupadi. This action was widely received as a positive step toward ensuring the safety and well-being of Nepali women and girls. However, as indicated by our findings and prior research,19–21 numerous types of menstrual restriction are faced in Nepal, and in some areas these practices are pervasive. Similar to other practices rooted in gender norms (eg, IPV, child marriage), ending these practices will require, together with legislation, changing and challenging social norms. Since Nepal’s 2015 earthquake, increased programming and funding has focused on gender equity and women’s empowerment. More research is needed to determine how such initiatives might also deal with social norms, stigma and restrictions related to menstruation.

er with legislation, changing and challenging social norms. Since Nepal’s 2015 earthquake, increased programming and funding has focused on gender equity and women’s empowerment. More research is needed to determine how such initiatives might also deal with social norms, stigma and restrictions related to menstruation. The high frequency of menstrual restrictions reported herein highlights the need for global health and development to address this problem. Although not explicitly stated in the sustainable development goals,32 ensuring menstruation is safe and destigmatised is essential to achieving these targets, particularly those related to health, equitable education, gender equality, available and adequate sanitation and inclusive economic opportunity. Integrating efforts to reduce menstrual stigma within existing water, sanitation and hygiene (WASH) initiatives that focus on menstrual hygiene education and facilities is an initial way to tackle this issue jointly with sustainable development goal priorities. Funding: This manuscript has been funded by a grant (#P06254) from UK government aid, via the What Works to Prevent Violence Against Women and Girls? Global Programme. The funds were managed by the South African Medical Research Council. The views expressed do not necessarily reflect the UK government’s official policies. Competing interests: None declared. Patient consent: Not required. Ethics approval: George Mason University (802242-1), Emory University (IRB00091115), and the Nepal Health Research Council (178/2015).

Funding: This manuscript has been funded by a grant (#P06254) from UK government aid, via the What Works to Prevent Violence Against Women and Girls? Global Programme. The funds were managed by the South African Medical Research Council. The views expressed do not necessarily reflect the UK government’s official policies. Competing interests: None declared. Patient consent: Not required. Ethics approval: George Mason University (802242-1), Emory University (IRB00091115), and the Nepal Health Research Council (178/2015). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages In the majority of countries, medication (or medical) abortion accounts for at least half of all abortions, and this has increased over the past 10 years. In the majority of countries, at least 90% of abortions occur under 12 weeks, and the proportion of abortions obtained earlier than 9 weeks has increased in most countries. Findings suggest improvements in access to timely care and a choice of methods in some countries, although research is needed to understand if the observed distributions are a function of women’s preferences or of barriers to care. Introduction Abortion is a common event in women’s lives: in 2010–2014, an estimated 56 million abortions occurred, and nearly 8 million occurred in high-income countries.1 High-income countries with liberal abortion laws often have accessible and reliable abortion data, resulting in a unique opportunity to compare abortion service provision across countries. Data on the method and gestational age of abortion can provide insight on abortion provision within a country, including the options women have to terminate their pregnancies. Trend data on these characteristics can show whether abortion care within a country has changed over time.

n service provision across countries. Data on the method and gestational age of abortion can provide insight on abortion provision within a country, including the options women have to terminate their pregnancies. Trend data on these characteristics can show whether abortion care within a country has changed over time. This review provides the first cross-national comparison of the method and gestational age of abortion in high-income countries, and of trends in these characteristics over time. Comparing how and when pregnancies are terminated across countries and over time is an important first step toward understanding the factors influencing the findings. These factors might include service regulations, provider practices and preferences, and women’s preferences. Such a review can also be used to identify departures from typical patterns, and explore their causes. Ultimately these data are of use to policymakers, programme planners and service providers to ensure women have access to safe, high-quality and timely abortion care.

ractices and preferences, and women’s preferences. Such a review can also be used to identify departures from typical patterns, and explore their causes. Ultimately these data are of use to policymakers, programme planners and service providers to ensure women have access to safe, high-quality and timely abortion care. Methods Official statistics on the number of legal abortions by gestational age and method of abortion were sought for the 40 high-income countries with liberal abortion laws, using information on countries’ abortion laws compiled by the Guttmacher Institute.2 We identified high-income countries using the World Bank country classifications by income level, for the year 2017.3 In conjunction with data collection on abortion incidence, we collected data from 2012 to 2018. We collected data for the most recent year available after 2010, and, when possible, we also collected data for the 10 years prior. Countries were categorised as having liberal abortion laws if abortion was legal without restriction or was legal on socioeconomic grounds, either with or without gestational limits. We also include New Zealand and Israel, where abortion is allowed to preserve a woman’s mental health, because this ground is interpreted relatively liberally in these two countries.2 Among the 24 high-income countries for which data were available, abortion was available on request in 18 countries, on socioeconomic grounds in four, and for the mental health of the woman in two (table 1).4

serve a woman’s mental health, because this ground is interpreted relatively liberally in these two countries.2 Among the 24 high-income countries for which data were available, abortion was available on request in 18 countries, on socioeconomic grounds in four, and for the mental health of the woman in two (table 1).4 Table 1 Legal status, legal indications and data source of abortion in high-income countries with abortion data, 2017

serve a woman’s mental health, because this ground is interpreted relatively liberally in these two countries.2 Among the 24 high-income countries for which data were available, abortion was available on request in 18 countries, on socioeconomic grounds in four, and for the mental health of the woman in two (table 1).4 Table 1 Legal status, legal indications and data source of abortion in high-income countries with abortion data, 2017 Country Gestational age limit Legal status type code (2017)* Waiting period Country recognised approval Conscientious objection Data source Misoprostol Mifepristone Year of mifepristone registration Belgium Through 14th week On request 6 days, after counselling Yes Yes 1999 Allowed Official statistics Canada Law does not indicate gestational age limit On request No waiting period Yes Yes 2015 Varies by province Official statistics Czech Republic Through 12th week On request No waiting period Yes Yes 2013 Allowed unless life and health of mother is at risk, otherwise referral required Official statistics Denmark Through 12th week On request No waiting period Yes Yes 1999 Allowed Official statistics Estonia Through 12th week On request No waiting period Yes Yes 2003 Allowed Official statistics† Finland Through 12th week Socioeconomic grounds No waiting period Yes Yes 1999 No explicit reference in law Official statistics France Through 14th week On request 7 days unless this takes the pregnancy beyond 12 weeks Yes Yes 1988 Allowed but referral required Official statistics Germany Through 14th week On request 3 days, after counselling Yes Yes 1999 Allowed unless life and health of mother is at risk Official statistics Great Britain Through 12th week Socioeconomic grounds No waiting period Yes Yes 1991 Allowed unless life, and physical or mental health of mother is at risk Official statistics Iceland Through 12th week Socioeconomic grounds No waiting period Yes Yes 1999 No explicit reference in law Official statistics Israel Through 12th week Mental health No waiting period Yes Yes 1999 Allowed Official statistics Italy Through 90 days/3 months On request 7 days after first visit Yes Yes 2009 Allowed Official statistics Japan Through 12th week Socioeconomic grounds No waiting period Yes No NA No explicit reference in law Official statistics† Netherlands No gestational age limit for previability abortion On request 5 days after first visit Yes Yes 1999 Allowed Official statistics New Zealand Through 12th week Mental health No waiting period Yes Yes 2001 Allowed and referral not required Official statistics Norway Through 12th week On request No waiting period Yes Yes

No gestational age limit for previability abortion On request 5 days after first visit Yes Yes 1999 Allowed Official statistics New Zealand Through 12th week Mental health No waiting period Yes Yes 2001 Allowed and referral not required Official statistics Norway Through 12th week On request No waiting period Yes Yes 2000 Allowed Official statistics Portugal Through 10th week On request 3 days Yes Yes 2007 Allowed but referral required Official statistics Singapore Through 24th week On request 48 hours, after counselling unless termination is immediately necessary Yes Yes NA Allowed unless life and health of mother is at risk Official statistics Slovakia Through 12th week On request 48 hours after mandated information given No No NA Allowed unless life and health of mother is at risk Official statistics† Slovenia Through 12th week On request No waiting period Yes Yes 2013 Allowed unless life and health of mother is at risk Official statistics† Spain Through 14th week On request 3 days after counselling Yes Yes 1999 Allowed Official statistics Sweden Through 18th week On request No waiting period Yes Yes 1992 No explicit reference in law Official statistics Switzerland Through 12th week On request No waiting period Yes Yes 1999 No explicit reference in law Official statistics USA No gestational age limit for previability abortion On request Varies by jurisdiction Yes Yes 2000 Varies by state Survey *Categories include On Request/Without Restriction to Reason; Socioeconomic Grounds (also to save the woman’s life and physical/mental health); To Preserve Mental Health; To Preserve Health; To Save the Woman’s Life; Prohibited Altogether.

y abortion On request Varies by jurisdiction Yes Yes 2000 Varies by state Survey *Categories include On Request/Without Restriction to Reason; Socioeconomic Grounds (also to save the woman’s life and physical/mental health); To Preserve Mental Health; To Preserve Health; To Save the Woman’s Life; Prohibited Altogether. †Official statistics obtained by questionnaire administered to country representatives. Sources: Center for Reproductive Rights for legal status type code and gestational age-limit; Global Abortion Policies Database for waiting period, conscientious objection, and recognition of mifepristone and misoprostol; Gynuity for year of mifepristone registration. NA, not available; TBD, to be determined.

†Official statistics obtained by questionnaire administered to country representatives. Sources: Center for Reproductive Rights for legal status type code and gestational age-limit; Global Abortion Policies Database for waiting period, conscientious objection, and recognition of mifepristone and misoprostol; Gynuity for year of mifepristone registration. NA, not available; TBD, to be determined. To collect official abortion statistics, we searched the websites of statistical offices or relevant agencies for the most recent official abortion data available (online supplementary supporting table 1). When the search resulted in reports with data from previous years the data for up to 10 years prior were collated. When data were not available online, we contacted agencies directly and asked officials to complete a standardised questionnaire that gathered information from official records on the incidence of legally induced abortions and characteristics of abortions including gestational age and method. This questionnaire was distributed for a study on abortion incidence that began in 2014 and the questionnaires requested data on abortions up to 2017. For the United States, we utilised results from the Guttmacher Institute Abortion Patient Survey,5 as abortion statistics from the Centers for Disease Control and Prevention do not contain data for all states.6 10.1136/bmjsrh-2018-200149.supp1Supplementary file 1

To collect official abortion statistics, we searched the websites of statistical offices or relevant agencies for the most recent official abortion data available (online supplementary supporting table 1). When the search resulted in reports with data from previous years the data for up to 10 years prior were collated. When data were not available online, we contacted agencies directly and asked officials to complete a standardised questionnaire that gathered information from official records on the incidence of legally induced abortions and characteristics of abortions including gestational age and method. This questionnaire was distributed for a study on abortion incidence that began in 2014 and the questionnaires requested data on abortions up to 2017. For the United States, we utilised results from the Guttmacher Institute Abortion Patient Survey,5 as abortion statistics from the Centers for Disease Control and Prevention do not contain data for all states.6 10.1136/bmjsrh-2018-200149.supp1Supplementary file 1 As the legal status and legal indications of abortion can influence the method and gestational age of abortion, we collected information on the legal status, gestational age limits and exceptions, mandatory waiting periods, and conscientious objection clauses in a country (table 1 and online supplementary supporting table 2). For legal status and gestational age limit, we collected information from the Center for Reproductive Rights,7 and for exceptions to the gestational age limit, waiting periods and conscientious objection clauses, from the Global Abortion Policies Database.8 We also collected information from the Global Abortion Policies Database on whether misoprostol and/or mifepristone are recognised in a country, which was supplemented with information from Gynuity on the year mifepristone was registered.9

nd conscientious objection clauses, from the Global Abortion Policies Database.8 We also collected information from the Global Abortion Policies Database on whether misoprostol and/or mifepristone are recognised in a country, which was supplemented with information from Gynuity on the year mifepristone was registered.9 10.1136/bmjsrh-2018-200149.supp2Supplementary file 2 We categorised method of abortion into three types: surgical, medication (or medical) and other. It was not possible to further refine these categories due to differences in abortion reporting, method availability, and terminology between countries. For this review article, medication abortion refers to primarily the regimen of mifepristone and misoprostol, although if a country reported abortions obtained through misoprostol alone, they were also included in the total of medication abortions. In addition, the Netherlands and Canada report a category of surgical abortion plus medication abortion, and these abortions were classified as surgical abortions. The definition of other abortion varies by country and not all countries explicitly state what falls under this category, but it can include methods such as hysterotomy. We collated gestational age data as follows: fewer than 9 weeks (up to 8 weeks and 6 days), 9–12 weeks, and 13 or more weeks. These categories reflect the most common classification of gestational age in country reports. Countries that use different classification categories are indicated in the footnotes of figure 3. Figure 3 Percent distribution of abortions by gestational age, 2017.

We collated gestational age data as follows: fewer than 9 weeks (up to 8 weeks and 6 days), 9–12 weeks, and 13 or more weeks. These categories reflect the most common classification of gestational age in country reports. Countries that use different classification categories are indicated in the footnotes of figure 3. Figure 3 Percent distribution of abortions by gestational age, 2017. Of the 40 high-income countries with liberal laws in the world as of 2017 (as defined by the World Bank, and for countries with populations greater than one million), data were available on the method of abortion in 16 countries and on the gestational ages at which abortions were performed in 21 countries. In 10 countries, the data on abortions are from 2017; in the remaining countries, the most recent year for which data were available ranged from 2011 to 2016. The year of data available for each country can be found in online supplementary supporting table 3. Trend data were not available for all countries, and we present trends on the method of abortion for 11 countries, and the gestational age for 13 countries. The difference between data availability for gestational age and method is attributed to variations in what characteristics countries collect and report in their official abortions statistics. In Great Britain, Scotland reports its abortion statistics separately from England and Wales, and the categories used for gestational age and method differ slightly. Therefore the data for these geographic areas are presented separately in this analysis, but are counted as one country in the above tabulations. In addition, we refer to Great Britain rather than the United Kingdom due to the different abortion law in Northern Ireland.

d for gestational age and method differ slightly. Therefore the data for these geographic areas are presented separately in this analysis, but are counted as one country in the above tabulations. In addition, we refer to Great Britain rather than the United Kingdom due to the different abortion law in Northern Ireland. 10.1136/bmjsrh-2018-200149.supp3Supplementary file 3 We calculated percentage distributions in order to facilitate comparison between countries. In addition, due to concerns with abortion underreporting, percentage distributions have the advantage of being more reliable if abortion reporting is incomplete, as long as the level of underreporting doesn’t vary by gestational age or method of abortion. We assessed the completeness of data for each country from the data collection methodologies in the country reports or published literature. Statistics that were assessed as capturing at least 90% of all legal abortions within the country were classified as complete and footnotes indicate whether a country’s data are incomplete.

d the completeness of data for each country from the data collection methodologies in the country reports or published literature. Statistics that were assessed as capturing at least 90% of all legal abortions within the country were classified as complete and footnotes indicate whether a country’s data are incomplete. When calculating the percentage distribution of abortions, abortions for which the gestational age or method of abortion were not specified were not included in the numerator or denominator unless non-specified abortions accounted for more than 10% of all abortions reported. That is, when non-specified abortions accounted for fewer than 10% of all abortions, we assumed the non-specified abortions had the same distribution according to gestational age and method. The only exception is for Canada, where 18.3% of all abortions were unspecified. In addition, in Canada, country statistics only report gestational age and method data for abortions performed in hospitals, which, according to country statistics, account for only about a third of all abortions in Canada. Despite this limitation, the data from Canada are included as they provide at least some insight into abortion provision in that country. No ethical approval was required for this study as the data do not constitute human subjects research.

When calculating the percentage distribution of abortions, abortions for which the gestational age or method of abortion were not specified were not included in the numerator or denominator unless non-specified abortions accounted for more than 10% of all abortions reported. That is, when non-specified abortions accounted for fewer than 10% of all abortions, we assumed the non-specified abortions had the same distribution according to gestational age and method. The only exception is for Canada, where 18.3% of all abortions were unspecified. In addition, in Canada, country statistics only report gestational age and method data for abortions performed in hospitals, which, according to country statistics, account for only about a third of all abortions in Canada. Despite this limitation, the data from Canada are included as they provide at least some insight into abortion provision in that country. No ethical approval was required for this study as the data do not constitute human subjects research. Results Method of abortion In the majority of countries, medication abortion accounts for at least half of all abortions, with the highest proportions found in the Nordic countries of Finland (97%), Sweden (93%) and Norway (88%) (figure 1). In almost all Northern Europe countries, medication abortion accounts for at least two-thirds of all abortions. In Western Europe, France and Switzerland have the highest proportion of medication abortions at 68% and 75%, respectively. Figure 1 Percentage distribution of abortions by method of abortion, 2017.

Results Method of abortion In the majority of countries, medication abortion accounts for at least half of all abortions, with the highest proportions found in the Nordic countries of Finland (97%), Sweden (93%) and Norway (88%) (figure 1). In almost all Northern Europe countries, medication abortion accounts for at least two-thirds of all abortions. In Western Europe, France and Switzerland have the highest proportion of medication abortions at 68% and 75%, respectively. Figure 1 Percentage distribution of abortions by method of abortion, 2017. In five countries, of which three are in Western Europe, surgical abortion accounts for over three-quarters of the abortion services. Southern Europe has the greatest subregional variation in abortion method, with surgical abortion accounting for 81% of abortions in Italy, compared with medication abortions accounting for 71% of all abortions in Portugal. Over the past 10 years, all countries have seen an increase in the proportion of abortions that were medication abortions (figure 2). In Denmark, Estonia, England  and Wales, France and Iceland, medication abortion is now the more common method of abortion. The change in Iceland is especially noteworthy as medication abortion only became available in Iceland in 2006. The proportion of abortions that were medication abortions has also increased in Belgium, Italy and Germany, although medication abortion still accounts for a small proportion of all abortions in these countries. Figure 2 Proportion of medication and surgical abortions from 2007 to 2017*

Over the past 10 years, all countries have seen an increase in the proportion of abortions that were medication abortions (figure 2). In Denmark, Estonia, England  and Wales, France and Iceland, medication abortion is now the more common method of abortion. The change in Iceland is especially noteworthy as medication abortion only became available in Iceland in 2006. The proportion of abortions that were medication abortions has also increased in Belgium, Italy and Germany, although medication abortion still accounts for a small proportion of all abortions in these countries. Figure 2 Proportion of medication and surgical abortions from 2007 to 2017* Gestational age In 2017, or the most recent year of available data, for almost all countries, over two-thirds of abortions occurred at less than 9 weeks gestation (figure 3). The percentage of abortions occurring at less than 9 weeks gestation ranged from 39% in Canada to 84% in Sweden. [NB. Abortion data in Canada are for hospitals only, which, according to country statistics, account for only about a third of all abortions in Canada.] New Zealand, Canada and Italy are the only countries where fewer than half of all abortions occurred at less than 9 weeks gestation; however, in all countries most abortions occurred by 10 or 11 weeks gestation. In the majority of countries over 90% of abortions occur before 13 weeks. In 10 countries, 95% of abortions occurred before 13 weeks, with the highest proportion of abortions completed before 13 weeks in Germany (97%). The proportion of abortions at 13 weeks and above was highest in the Netherlands, where 18% of abortions occurred after 13 weeks.

over 90% of abortions occur before 13 weeks. In 10 countries, 95% of abortions occurred before 13 weeks, with the highest proportion of abortions completed before 13 weeks in Germany (97%). The proportion of abortions at 13 weeks and above was highest in the Netherlands, where 18% of abortions occurred after 13 weeks. Over the past 10 years there has generally been an increase in the proportion of abortions that were obtained earlier than 9 weeks gestation (online supplementary supporting figure 1). There is a considerable range among countries: from an increase of two percentage points in Finland, to 25 percentage points in Germany. Although New Zealand currently has the lowest proportion of abortions that occur at less than 9 weeks gestation, this value has increased by 20 percentage points in the last 10 years. The Netherlands experienced small declines in the proportion of abortions that were obtained earlier than 9 weeks gestation. 10.1136/bmjsrh-2018-200149.supp4Supplementary file 4

Over the past 10 years there has generally been an increase in the proportion of abortions that were obtained earlier than 9 weeks gestation (online supplementary supporting figure 1). There is a considerable range among countries: from an increase of two percentage points in Finland, to 25 percentage points in Germany. Although New Zealand currently has the lowest proportion of abortions that occur at less than 9 weeks gestation, this value has increased by 20 percentage points in the last 10 years. The Netherlands experienced small declines in the proportion of abortions that were obtained earlier than 9 weeks gestation. 10.1136/bmjsrh-2018-200149.supp4Supplementary file 4 In Norway, Sweden, and England and Wales, information about method of abortion by gestational age was available (online supplementary supporting table 4). In all three of these countries, medication abortions account for at least 83% of abortions performed at less than 9 weeks gestation. This is also seen in Scotland, where medication abortions account for 90% of all abortions at less than 9 weeks gestation (data not shown). In Sweden, mifepristone is given at the hospital/clinic and then women have the option to either take the misoprostol in the hospital/clinic or at home.10 In Sweden, 64% of all abortions were medication abortions where misoprostol was taken at home, which accounted for 75% of all abortions performed at less than 9 weeks' gestation (data not shown). Between 9 and 12 weeks gestation, only England and Wales reported a higher proportion of surgical abortions versus medication abortions. Medication abortion continued to be the more utilised method in both Norway (68%) and Sweden (78%) for abortions performed between 9 and 12 weeksgestation.

gestation (data not shown). Between 9 and 12 weeks gestation, only England and Wales reported a higher proportion of surgical abortions versus medication abortions. Medication abortion continued to be the more utilised method in both Norway (68%) and Sweden (78%) for abortions performed between 9 and 12 weeksgestation. 10.1136/bmjsrh-2018-200149.supp5Supplementary file 5 Data were available on the percentage distribution of the abortions by gestational age of the pregnancy and age group of the women for Norway, Finland, Sweden, and England and Wales (online supplementary supporting figure 2) for 2017. In all of these countries, over 72% of all abortions occurred at less than 9 weeks gestation for all six age groups. However, women aged under 20 years consistently have a lower proportion of abortions at less than 9 weeks, and a higher proportion at 9–12 weeks compared with older age groups. Generally all age groups experienced abortions at 13 weeks and above at very similar proportions, except in Finland where women aged over 40 years had a slightly higher proportion of abortions at 13–18 weeks.

ortion of abortions at less than 9 weeks, and a higher proportion at 9–12 weeks compared with older age groups. Generally all age groups experienced abortions at 13 weeks and above at very similar proportions, except in Finland where women aged over 40 years had a slightly higher proportion of abortions at 13–18 weeks. Discussion The provision of abortion services has changed in the past decade. Medication abortion accounts for at least half of all abortions in the majority of countries. Also in the majority of countries, over 90% of all abortions were completed before 13 weeks, and more than two-thirds of abortions occurred before the first 9 weeks of gestation. Over the past 10 years there has been an increase in both the proportion of abortions that were medication abortions, and the proportion that were obtained before 9 weeks gestation.

90% of all abortions were completed before 13 weeks, and more than two-thirds of abortions occurred before the first 9 weeks of gestation. Over the past 10 years there has been an increase in both the proportion of abortions that were medication abortions, and the proportion that were obtained before 9 weeks gestation. Over the past 10 years, the dramatic increase in the proportion of abortions that are medication abortions is likely reflective of factors related to both supply of and demand for the procedure. There is substantial variation across countries with respect to the year in which misoprostol and mifepristone was first approved for pregnancy termination, but the length of time that the method has been available is not necessarily correlated with the prevalence of the method. France was the first country, other than China, to license the use of misoprostol and mifepristone for early medical abortion in 1988,11 and since then countries have approved this combination regimen at different points, with Canada being the most recent among this group of countries to legalise the combination of misoprostol and mifepristone in 2015. Japan and Slovakia are the only countries cited in this review where combination mifepristone and misoprostol have not been approved for abortion services, although misoprostol alone is recognised as a method.8

ost recent among this group of countries to legalise the combination of misoprostol and mifepristone in 2015. Japan and Slovakia are the only countries cited in this review where combination mifepristone and misoprostol have not been approved for abortion services, although misoprostol alone is recognised as a method.8 The proportion of abortions that were medication abortions may also depend in part on the level of government and public sector support for this method, and on the adequate training of providers. Although France was one of the first countries to approve the use of misoprostol and mifepristone, in the 1990s medication abortion was very highly regulated, and only in 2001 was a law introduced to allow the provision of medication abortion outside of hospitals,11 and medication abortion has been available outside of hospitals since 2005.12 Portugal, in contrast, has a higher proportion of medication abortions; almost three-quarters of abortions occur in public hospitals, and 98% of abortions performed in hospitals are medication abortions. This may partially be due to the fact that abortion is only legal in Portugal up to 10 weeks gestation. Overall in Western Europe there continues to be a high proportion of surgical abortions. However in Belgium, the latest available data are from 2011 so it not possible to discern if this proportion has changed recently.

s may partially be due to the fact that abortion is only legal in Portugal up to 10 weeks gestation. Overall in Western Europe there continues to be a high proportion of surgical abortions. However in Belgium, the latest available data are from 2011 so it not possible to discern if this proportion has changed recently. Mandatory waiting periods can also influence the proportion of abortions that are medication abortions. Ten of the 24 countries in this review require waiting periods after a woman’s first appointment (table 1). These waiting periods range from 48 hours in Scotland and Singapore, to seven days in France and Italy.8 Longer waiting periods can reduce women’s choice of abortion method because it increases the gestational age at which the abortion is performed. For example, in France, medication abortion can only be performed up to 49 days gestation, yet the mandatory 7-day waiting period may impact whether a women is eligible for a medication abortion at the time of her procedure.11

abortion method because it increases the gestational age at which the abortion is performed. For example, in France, medication abortion can only be performed up to 49 days gestation, yet the mandatory 7-day waiting period may impact whether a women is eligible for a medication abortion at the time of her procedure.11 The proportion of abortions that occur earlier than 9 weeks gestation increased in the majority of countries in this review. The increase in abortions earlier than 9 weeks is likely closely related to the increased availability of medication abortion, since when medication abortion was first approved, the recommended protocol is for a medication abortion to occur under 63 days or 9 weeks gestation.13 This trend might also be due to the rise in technologies for earlier detection of pregnancy, and the widespread availability of pregnancy tests that provide accurate results at early gestations.13 There is a strong connection between these two factors, which when coupled with increased access to abortion services, may result in an increase in the proportion of abortions occurring earlier than 9 weeks gestation.

nd the widespread availability of pregnancy tests that provide accurate results at early gestations.13 There is a strong connection between these two factors, which when coupled with increased access to abortion services, may result in an increase in the proportion of abortions occurring earlier than 9 weeks gestation. The proportion of abortions conducted after 9 weeks remained high in a few countries. In Canada, the available data suggest that a large proportion of abortions are obtained after 13 weeks, but data are only available for abortions performed in hospitals, where abortions after 13 weeks are likely overrepresented. In addition, 18.6% of the reported abortions in Canada are unspecified, which could bias these findings if the unspecified abortions are not evenly distributed across gestational age.

r 13 weeks, but data are only available for abortions performed in hospitals, where abortions after 13 weeks are likely overrepresented. In addition, 18.6% of the reported abortions in Canada are unspecified, which could bias these findings if the unspecified abortions are not evenly distributed across gestational age. A country’s gestational age limits for legal abortion can affect the distribution of gestational age of abortion within a country, and in neighbouring countries, if women have to travel to seek an abortion after 13 weeks. For example, there is no gestational age limit for legal abortion in the Netherlands, and the large proportion of abortions done after 13 weeks in the Netherlands are partly due to abortions obtained by non-residents. In 2015, non-residents accounted for 12.6% of women obtaining abortions in the Netherlands, and the majority of those women sought abortions at 13 weeks and above, compared to only 5.1% of Dutch women who had abortions at 13 weeks and above.14 Other evidence found that in 2006 72% of abortions at 13 weeks and above in the Netherlands were to women of non-Dutch residence and origin.15 Therefore, regardless of how rare abortion at 13 weeks and above is in a given country, the Dutch experience suggests that the demand for abortions at greater than 13 weeks exists to some extent in all countries. Safe and legal abortions at 13 weeks and above should be available to all women who need them.

nd origin.15 Therefore, regardless of how rare abortion at 13 weeks and above is in a given country, the Dutch experience suggests that the demand for abortions at greater than 13 weeks exists to some extent in all countries. Safe and legal abortions at 13 weeks and above should be available to all women who need them. Conscientious objection clauses exist for most countries considered in this review (table 1). With adequate and timely referrals and a sufficient number of providers, conscientious objection can potentially exist without being a barrier to care.16 However, examples in Europe indicate that it can delay or prevent women from accessing abortion care,17 and potentially cause women to seek abortions after 13 weeks gestation. Data on the impact of conscientious objection are limited, but, in Italy, a country with a lower proportion of abortions at less than 9 weeks gestation, the prevalence of conscientious objection has led to a shortage of abortion providers in Italy. It is estimated that 82%–91% of providers in Rome and the surrounding areas are conscientious objectors, and abortion services are only provided in 60% of Italian hospitals.18 The prevalence of conscientious objection likely delays women’s access to services, and as indicated in a statement by the Council of Europe in 2016, causes Italian women to travel abroad, or "resort to ending their pregnancy without the support of competent health authorities".19

rovided in 60% of Italian hospitals.18 The prevalence of conscientious objection likely delays women’s access to services, and as indicated in a statement by the Council of Europe in 2016, causes Italian women to travel abroad, or "resort to ending their pregnancy without the support of competent health authorities".19 This article, the first review of gestational age and method of abortion across high-income countries, has several limitations. While the completeness of data were assessed for each country from official reports, the findings may be influenced if abortion statistics are underreported or incomplete. If countries have incomplete data, these findings rely on the assumption that the level of underreporting of abortion does not vary by gestational age or method of abortion, however this information is not available. Anecdotal evidence suggests that medication abortion is more prone to underreporting than surgical abortion, especially with the increase of access to medication abortion outside of the healthcare system. Regulations and common practices within the healthcare system that we were not able to quantify or cite may also drive some of the findings. Similarly, as most data are not available by country of residence, the legal status and/or gestational age limits of abortion within a country and neighbouring countries may also influence the distribution of gestational age. Finally, the findings from these countries may not be generalisable to countries without official statistics, low-income countries, or countries with restrictive abortion laws, and indicate a need for additional research. In particular, findings from low-income countries with liberal abortion laws could shed light on service provision and barriers to care in those settings.

neralisable to countries without official statistics, low-income countries, or countries with restrictive abortion laws, and indicate a need for additional research. In particular, findings from low-income countries with liberal abortion laws could shed light on service provision and barriers to care in those settings. Options for terminating pregnancies are important as studies have found that when given the choice, many women have strong preferences for one method of abortion over the other.20–23 Reasons for selecting medication abortion include fear of clinical procedures or anaesthesia, and the perception that it is "more natural" and similar to a miscarriage.20 Conversely, women select surgical procedures because they are "faster and easier".21 However the percentage distribution of abortion methods in a country does not necessarily indicate a woman’s ability to choose a method. Health system factors such as method availability, provider preference, waiting periods and/or gestational age limits can all influence the method a woman chooses.

ster and easier".21 However the percentage distribution of abortion methods in a country does not necessarily indicate a woman’s ability to choose a method. Health system factors such as method availability, provider preference, waiting periods and/or gestational age limits can all influence the method a woman chooses. Examining the distribution of abortions by method of abortion and gestational age at time of termination is one step in understanding the factors that drive these distributions, and this review highlights the need to understand if the observed distributions are a function of women’s preferences or of barriers to care. More research is needed on the proportion of women obtaining an abortion who used the method of their choice, to shed light on whether women have the resources and options they need to terminate their pregnancies the way they want to. Ultimately, this evidence should enable policy makers and service providers to consider measures needed to ensure women can access safe services that align with their preferences in a timely manner. 10.1136/bmjsrh-2018-200149.supp6Supplementary file 6 The authors gratefully acknowledge data assistance provided by Lorraine Kwok and Cynthia Beavin. The authors also thank Susheela Singh, Akinrinola Bankole and Evert Ketting for their comments on earlier drafts.

Examining the distribution of abortions by method of abortion and gestational age at time of termination is one step in understanding the factors that drive these distributions, and this review highlights the need to understand if the observed distributions are a function of women’s preferences or of barriers to care. More research is needed on the proportion of women obtaining an abortion who used the method of their choice, to shed light on whether women have the resources and options they need to terminate their pregnancies the way they want to. Ultimately, this evidence should enable policy makers and service providers to consider measures needed to ensure women can access safe services that align with their preferences in a timely manner. 10.1136/bmjsrh-2018-200149.supp6Supplementary file 6 The authors gratefully acknowledge data assistance provided by Lorraine Kwok and Cynthia Beavin. The authors also thank Susheela Singh, Akinrinola Bankole and Evert Ketting for their comments on earlier drafts. Contributors: Both authors contributed to the conceptualisation of the paper. Drawing on methodologies from earlier papers by GS and colleagues, AP conducted data collection and analysis. Both authors developed the outline of the paper, and AP wrote the first draft. Both authors reviewed and substantively edited subsequent drafts. Funding: This work was made possible by UK Aid from the UK Government and the Dutch Ministry of Foreign Affairs. The views expressed are those of the authors and do not necessarily reflect the positions and policies of the donors.

Contributors: Both authors contributed to the conceptualisation of the paper. Drawing on methodologies from earlier papers by GS and colleagues, AP conducted data collection and analysis. Both authors developed the outline of the paper, and AP wrote the first draft. Both authors reviewed and substantively edited subsequent drafts. Funding: This work was made possible by UK Aid from the UK Government and the Dutch Ministry of Foreign Affairs. The views expressed are those of the authors and do not necessarily reflect the positions and policies of the donors. Competing interests: None declared. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not required.

Key messages The number of sexual partners a person has in their lifetime is an important predictor of sexually transmitted infection (STI) and their associated health risks. Until now, no current data are available on correlates of lifetime sexual partners at older ages. We found in a large representative sample of older English adults that a higher lifetime number of sexual partners is associated with a number of sociodemographic and behavioural factors. An understanding of who is more likely to have had more sexual partners may help health practitioners to identify individuals who are at greatest risk of STI and their associated health complications across the life course.

We found in a large representative sample of older English adults that a higher lifetime number of sexual partners is associated with a number of sociodemographic and behavioural factors. An understanding of who is more likely to have had more sexual partners may help health practitioners to identify individuals who are at greatest risk of STI and their associated health complications across the life course. Background The number of sexual partners a person has in their lifetime is an important correlate of sexually transmitted infections (STIs) and their associated health risks.1–5 With the exception of HIV and AIDS, the general population is largely unaware of the substantial impact STIs can have on morbidity and mortality.6 For example, a number of sexually transmitted pathogens are known to cause cancer. Sexually acquired human papillomavirus (HPV) plays a causal role in around 70% of cervical, vaginal and anal cancers, 30%–40% of vulval, penile and oropharyngeal cancers, and has been causally linked to non-melanoma skin cancer and cancer of the conjunctiva.7 Hepatitis B virus causes hepatocellular carcinoma, one of the most common forms of cancer.8 Other STIs associated with cancers include Epstein-Barr virus, linked to nasopharyngeal carcinoma and lymphoma; human herpes virus type 8, linked to Kaposi’s sarcoma9; and human T-cell lymphotrophic virus type I (HTLV-I), linked to adult T-cell leukaemia and lymphoma.10 STIs also increase the risk of infertility11 and are associated with acute complications for pregnant women and their infants, such as miscarriage, prematurity, stillbirth and newborn blindness.12 13 STIs are the leading cause of loss of healthy life years in developing countries,13 and account for a substantial number of adverse health events and deaths globally. For example, in 1998 around 20 million adverse health events and almost 30 000 deaths in the US were directly attributable to STIs.14

orn blindness.12 13 STIs are the leading cause of loss of healthy life years in developing countries,13 and account for a substantial number of adverse health events and deaths globally. For example, in 1998 around 20 million adverse health events and almost 30 000 deaths in the US were directly attributable to STIs.14 The significant health consequences associated with STIs, and the high costs to society in terms of healthcare expenditure,15 underscore the importance of identifying those at risk of STI and implementing effective prevention strategies. There is a relatively large literature base on the correlates of number of sexual partners in adolescents. For example, the Youth Risk Behaviour Surveillance System in the US has observed significant associations with sex and ethnicity, with male students (20.9%) more likely than female students (14.4%), and black students (35.6%) more likely than white and Hispanic students (14.2% and 17.6%, respectively), to have had four or more sexual partners during their lifetime.16 Another US-based study found that common correlates of number of sexual partners among black females include alcohol, tobacco, marijuana use, and dating violence; and white females had similar correlates with the addition of physical fighting.17 Among white males, alcohol, tobacco, marijuana use, physical fighting, carrying weapons, and dating violence were strong correlates of number of sexual partners; and black males had similar correlates with the addition of binge alcohol use.17

hite females had similar correlates with the addition of physical fighting.17 Among white males, alcohol, tobacco, marijuana use, physical fighting, carrying weapons, and dating violence were strong correlates of number of sexual partners; and black males had similar correlates with the addition of binge alcohol use.17 While these findings provide important information for the development of interventions to reduce risky sexual behaviour at younger ages, factors that predict the number of lifetime sexual partners in adolescence may differ from those associated with a higher number of lifetime sexual partners in older adults. It is possible that individuals with a relatively high number of sexual partners in adolescence may have an average or below-average number by the time they reach old age, as others ‘catch up’ over time; for example, by remaining single and continuing to date while others settle down and stay with a single partner. Some literature exists on correlates of lifetime number of sexual partners in general adult populations, for example, in a British sample of 4913 men and 6777 women (aged 16 to 74 years) it was found that in women, but not men, low sexual function was associated with a higher number of sexual partners.18 To our knowledge, no current data are available on correlates of lifetime sexual partners at older ages. This information is needed as older adults are at greatest risk of developing cancer19 and many cancers that are common in older adults have been shown to be associated with STIs (eg, liver, anus, penile, sarcoma, prostate). Moreover, HIV diagnosis in later life is associated with shorter survival periods.20 In addition, a focus on older people, who have had increased opportunity for sexual experience on the basis of having had more time during which they have or could have been sexually active, would offer greater insight into factors predictive of a higher number of sexual partners that may aid in the identification of those at greatest risk of STIs across the life course. The identification of correlates of lifetime number of sexual partners in older adults would also offer insight as to how interventions could best be targeted to educate younger adults about the risks associated with having a high number of sexual partners and the practice of safe sex.

k of STIs across the life course. The identification of correlates of lifetime number of sexual partners in older adults would also offer insight as to how interventions could best be targeted to educate younger adults about the risks associated with having a high number of sexual partners and the practice of safe sex. This study therefore aimed to explore correlates of the lifetime number of sexual partners in a nationally representative sample of older adults (aged ≥50 years). We examined associations between self-reported lifetime number of sexual partners and a range of sociodemographic and behavioural variables.

k of STIs across the life course. The identification of correlates of lifetime number of sexual partners in older adults would also offer insight as to how interventions could best be targeted to educate younger adults about the risks associated with having a high number of sexual partners and the practice of safe sex. This study therefore aimed to explore correlates of the lifetime number of sexual partners in a nationally representative sample of older adults (aged ≥50 years). We examined associations between self-reported lifetime number of sexual partners and a range of sociodemographic and behavioural variables. Method Study population Data were from the English Longitudinal Study of Ageing (ELSA), a population-representative longitudinal panel study of men and women aged ≥50 years living in England.21 The study started in 2002, with participants recruited from an annual cross‐sectional survey of households and followed up every 2 years. Data are collected via computer-assisted personal interview (CAPI) conducted face-to-face in the participant’s home or residence, with additional self-completion questionnaires returned to the research office by post after the CAPI. The Sexual Relationships and Activities Questionnaire (SRA-Q) was administered as a self-completion measure in Wave 6 (2012/13) and was returned by 7079 (67%) participants. Of these, 6921 reported their lifetime number of sexual partners and formed the final analytical sample. All participants gave full informed consent to participate in the study, and ethical approval was obtained from the London Multi‐Centre Research Ethics Committee.

012/13) and was returned by 7079 (67%) participants. Of these, 6921 reported their lifetime number of sexual partners and formed the final analytical sample. All participants gave full informed consent to participate in the study, and ethical approval was obtained from the London Multi‐Centre Research Ethics Committee. Measures Number of sexual partners was assessed as part of the SRA-Q,22 which participants completed in private and returned in a sealed envelope. Participants were asked to indicate the number of sexual partners (vaginal/oral/anal sex) they had had in their lifetime (0, 1, 2–4, 5–9, 10–19, 20+). Due to low numbers of participants reporting have had 0 or in excess of 20 partners, we combined these with proximal categories, leaving four groups for analysis: 0–1, 2–4, 5–9 and ≥10 sexual partners. Demographic information collected included age, sex, ethnicity (white vs. non-white) and partnership status (married/cohabiting, separated/divorced, widowed, or single/never married). Socioeconomic status was based on household non-pension wealth (which has been identified as particularly relevant to health outcomes in this age group23), categorised into quintiles across all Wave 6 ELSA participants.

) and partnership status (married/cohabiting, separated/divorced, widowed, or single/never married). Socioeconomic status was based on household non-pension wealth (which has been identified as particularly relevant to health outcomes in this age group23), categorised into quintiles across all Wave 6 ELSA participants. Sexual orientation was assessed with the question: "Which statement best describes your sexual desires over your lifetime? Please include being interested in sex, fantasising about sex or wanting to have sex". Response options were (1) entirely for women, (2) mostly for women, but some desires for men, (3) equally for women and men, (4) mostly for men, but some desires for women, (5) entirely for men, and (6) no sexual desires in lifetime. We categorised participants with desires entirely for a different sex as heterosexual, entirely for the same sex as homosexual and those endorsing response options 2, 3 or 4 as bisexual. We coded the sexual orientation of those reporting no sexual desires as missing.

rely for men, and (6) no sexual desires in lifetime. We categorised participants with desires entirely for a different sex as heterosexual, entirely for the same sex as homosexual and those endorsing response options 2, 3 or 4 as bisexual. We coded the sexual orientation of those reporting no sexual desires as missing. Health-related variables included self-reported smoking status (current smoker, former smoker or never smoker) and frequency of alcohol intake, categorised as never/rarely (never – once or twice a year), regularly (once every couple of months – twice a week), or frequently (3 days a week – almost every day).22 Physical activity was assessed with three items that asked participants how often they took part in vigorous, moderate and low-intensity activities (more than once a week, once a week, 1–3 times a month, hardly ever/never),24 and further categorised into three groups, as previously described:25 inactive (no moderate/vigorous activity on a weekly basis); moderate activity at least once a week; and vigorous activity at least once a week. Limiting long-standing illness was self-reported in response to two questions: (i) "Do you have any long-standing illness, disability, or infirmity? By long-standing I mean anything that has troubled you over a period of time or that is likely to affect you over a period of time". If yes, (ii) "Does this illness or disability limit your activities in any way?". Declaration of a long-standing illness and any form of limitation classified the participant as having a limiting long-standing illness.

g that has troubled you over a period of time or that is likely to affect you over a period of time". If yes, (ii) "Does this illness or disability limit your activities in any way?". Declaration of a long-standing illness and any form of limitation classified the participant as having a limiting long-standing illness. Patient and public involvement Patients and public were not involved in the design of any aspect of this observational study. Statistical analysis Analyses were performed using IBM SPSS Statistics 22. Data were weighted to correct for sampling probabilities and for differential non-response and to calibrate back to the 2011 National Census population distributions for age and sex. The weights accounted for the differential probability of being included in Wave 6 of ELSA and for non-response to the SRA‐Q. Details can be found at http://doc.ukdataservice.ac.uk/doc/5050/mrdoc/pdf/5050_elsa_w6_technical_report_v1.pdf. Bivariate associations between lifetime number of sexual partners and predictors were assessed using one-way analyses of variance (ANOVAs) for continuous variables and χ2 tests for categorical variables. We then used multivariable multinomial logistic regression to analyse independent associations between lifetime number of sexual partners and predictors, with all variables entered into the same model. Separate analyses were carried out on men and women.

for continuous variables and χ2 tests for categorical variables. We then used multivariable multinomial logistic regression to analyse independent associations between lifetime number of sexual partners and predictors, with all variables entered into the same model. Separate analyses were carried out on men and women. We performed a sensitivity analysis in which multivariable models were repeated excluding participants who reported having had no sexual partners, to assess the extent to which their inclusion in the group with one sexual partner affected the results. Results Among men, 29.8% reported having had 0–1 sexual partners in their lifetime, 30.5% had had between 2 and 4 partners, 19.8% had had between 5 and 9 partners, and 19.9% had had 10 or more partners. Among women, the respective figures were 38.5% (0–1), 37.4% (2–4), 15.6% (5–9) and 8.5% (≥10).

We performed a sensitivity analysis in which multivariable models were repeated excluding participants who reported having had no sexual partners, to assess the extent to which their inclusion in the group with one sexual partner affected the results. Results Among men, 29.8% reported having had 0–1 sexual partners in their lifetime, 30.5% had had between 2 and 4 partners, 19.8% had had between 5 and 9 partners, and 19.9% had had 10 or more partners. Among women, the respective figures were 38.5% (0–1), 37.4% (2–4), 15.6% (5–9) and 8.5% (≥10). Bivariate associations between lifetime number of sexual partners and correlates are summarised in table 1. In both men and women, lifetime number of sexual partners was significantly associated with age, partner status, sexual orientation, wealth, smoking status, alcohol intake and physical activity. Those who had had more sexual partners tended to be younger than those who reported few sexual partners (p<0.001). Those who were separated/divorced or single/never married were more likely to report a higher number of sexual partners than those who were married/cohabiting or widowed (p<0.001). Those who were bisexual or homosexual tended to have a higher number of sexual partners than those who were heterosexual (p<0.001). Those in the lowest and highest quintiles of wealth reported more sexual partners than those in the middle quintiles (p<0.001 in men, p=0.006 in women). Current and former smokers reported a higher number of sexual partners than never smokers (p<0.001), and regular/frequent alcohol drinkers reported a higher number of sexual partners than those who were teetotal or rarely drank alcohol (p<0.001). Regular moderate/vigorous physical activity in women and regular vigorous physical activity in men was associated with a higher number of sexual partners (p<0.001). In women, but not in men, there was also a significant association with ethnicity, with white women reporting a higher number of sexual partners than those from ethnic minority groups (p<0.001). In men, but not in women, there was a significant association with limiting long-standing illness, with men without an illness more likely to report having more than one partner (p=0.037).

ion with ethnicity, with white women reporting a higher number of sexual partners than those from ethnic minority groups (p<0.001). In men, but not in women, there was a significant association with limiting long-standing illness, with men without an illness more likely to report having more than one partner (p=0.037). Table 1 Bivariate associations between lifetime number of sexual partners and predictors in men and women

ion with ethnicity, with white women reporting a higher number of sexual partners than those from ethnic minority groups (p<0.001). In men, but not in women, there was a significant association with limiting long-standing illness, with men without an illness more likely to report having more than one partner (p=0.037). Table 1 Bivariate associations between lifetime number of sexual partners and predictors in men and women Predictor n* Lifetime number of sexual partners in men n* Lifetime number of sexual partners in women 0–1 (n=911) 2–4 (n=932) 5–9 (n=604) ≥10 (n=607) P value 0–1 (n=1488) 2–4 (n=1448) 5–9 (n=604) ≥10 (n=327) P value Age (years) (mean (SD)) 3196 68.46 (9.89) 65.44 (9.43) 62.50 (8.57) 60.72 (8.03) <0.001 3548 69.25 (9.87) 64.36 (8.93) 59.77 (7.41) 65.36 (9.68) <0.001 Partner status Married/cohabiting 2111 32.9 31.1 18.9 17.1 <0.001 1868 44.8 34.5 14.7 6.0 <0.001 Separated/divorced 328 4.3 18.6 29.3 47.9 – 500 10.0 45.8 28.2 16.0 – Widowed 195 38.5 38.5 13.3 9.7 – 600 57.7 32.7 6.7 3.0 – Single/never married 249 17.7 19.3 19.7 43.4 – 189 35.4 24.9 18.0 21.7 – Sexual orientation Heterosexual 2701 28.8 29.9 20.0 21.1 <0.001 2831 40.7 37.1 15.5 6.8 <0.001 Bisexual 110 20.0 25.5 20.9 33.6 – 187 21.9 25.7 23.5 28.9 – Homosexual 44 4.5 6.8 4.5 84.1 – 14 14.3 35.7 35.7 14.3 – Ethnicity White 2695 28.4 29.2 19.6 22.8 0.131 3025 40.4 35.7 15.7 8.1 <0.001 Non-white 188 33.0 28.7 22.3 16.0 – 129 61.2 27.9 8.5 2.3 – Wealth quintile 1 (poorest) 493 21.1 22.5 19.5 36.9 <0.001 642 38.6 35.7 16.0 9.7 0.006 2 539 34.0 28.4 19.1 18.6 – 636 41.5 36.5 14.0 8.0 – 3 556 32.0 35.3 17.6 15.1 – 649 44.4 37.0 12.0 6.6 – 4 639 30.2 32.2 20.8 16.7 – 599 44.4 33.4 16.9 5.3 – 5 (richest) 617 25.8 27.6 22.2 24.5 – 573 38.6 33.5 18.5 9.4 – Smoking status Never smoker 2175 31.9 32.8 18.1 17.2 <0.001 2480 46.1 35.1 13.3 5.4 <0.001 Former smoker 294 19.4 16.7 25.5 38.4 254 21.3 33.1 28.0 17.7 Current smoker 414 18.6 18.8 24.6 37.9 – 421 24.5 38.5 20.4 16.6 – Alcohol intake† Never/rarely 449 36.3 30.3 16.3 17.1 <0.001 946 51.5 33.4 8.8 6.3 <0.001 Regularly 1160 30.0 30.0 19.2 20.8 – 1327 39.0 36.5 17.6 6.9 – Frequently 1176 24.0 27.9 22.2 25.9 – 804 33.0 35.4 20.4 11.2 – Physical activity Inactive 585 32.5 25.8 16.4 25.3 <0.001 844 53.1 29.5 11.3 6.2 <0.001 Moderately active at least once a week 1242 31.4 29.1 19.9 19.6 – 1483 39.9 37.8 14.6 8.3 – Vigorously active at least once a week 1059 23.4 31.1 21.5 24.0 – 828 32.6 37.0 21.4 9.1 – Limiting long-standing illness No 1965 27.4 30.6 20.0 21.9 0.037 2007 39.7 36.3 16.1 7.9 0.114 Yes 918 31.4 26.0 19.4 23.2 – 1146 43.9 33.8 14.3 8.0 – Values are perc

ce a week 1242 31.4 29.1 19.9 19.6 – 1483 39.9 37.8 14.6 8.3 – Vigorously active at least once a week 1059 23.4 31.1 21.5 24.0 – 828 32.6 37.0 21.4 9.1 – Limiting long-standing illness No 1965 27.4 30.6 20.0 21.9 0.037 2007 39.7 36.3 16.1 7.9 0.114 Yes 918 31.4 26.0 19.4 23.2 – 1146 43.9 33.8 14.3 8.0 – Values are perc entages unless otherwise stated. All figures are weighted for sampling probabilities and differential non-response. *All sample sizes are provided unweighted. Note sample numbers for each correlate do not always sum to the total sample number due to missing data. †Never/rarely=never – once or twice a year; regularly=once every 2 months – twice a week; frequently=3 days a week – almost every day. SD, standard deviation.

All figures are weighted for sampling probabilities and differential non-response. *All sample sizes are provided unweighted. Note sample numbers for each correlate do not always sum to the total sample number due to missing data. †Never/rarely=never – once or twice a year; regularly=once every 2 months – twice a week; frequently=3 days a week – almost every day. SD, standard deviation. Multivariable models confirmed that in both men (table 2) and women (table 3), younger age, being separated/divorced or single/never married, being a current (and to a lesser extent former) smoker, and drinking alcohol regularly or frequently were independently associated with a higher lifetime number of sexual partners. Homosexuality in men and bisexuality in women were also independently associated with a higher number of sexual partners. In addition, white ethnicity and regular moderate and vigorous physical activity were independently associated with a higher number of sexual partners in women only, and being in the highest and lowest quintiles of wealth was independently associated with a higher number of sexual partners in men only. After adjustment, the absence of limiting long-standing illness was associated with a higher number of sexual partners in women, but not in men. Table 2 Multivariable models testing independent associations between predictors and lifetime number of sexual partners in men

Multivariable models confirmed that in both men (table 2) and women (table 3), younger age, being separated/divorced or single/never married, being a current (and to a lesser extent former) smoker, and drinking alcohol regularly or frequently were independently associated with a higher lifetime number of sexual partners. Homosexuality in men and bisexuality in women were also independently associated with a higher number of sexual partners. In addition, white ethnicity and regular moderate and vigorous physical activity were independently associated with a higher number of sexual partners in women only, and being in the highest and lowest quintiles of wealth was independently associated with a higher number of sexual partners in men only. After adjustment, the absence of limiting long-standing illness was associated with a higher number of sexual partners in women, but not in men. Table 2 Multivariable models testing independent associations between predictors and lifetime number of sexual partners in men Predictor Lifetime number of sexual partners in men (adjusted OR (95% CI))† 2–4 (n=786)‡ 5–9 (n=491) ≥10 (n=486) Age 0.96 (0.95 to 0.97)*** 0.94 (0.92 to 0.95)*** 0.92 (0.91 to 0.94)*** Partnership status§ Separated/divorced 4.95 (2.65 to 9.24)*** 12.97 (6.98 to 24.08)*** 20.18 (10.94 to 37.21)*** Widowed 1.83 (1.23 to 2.73)** 1.38 (0.80 to 2.39) 1.47 (0.81 to 2.67) Single/never married 1.23 (0.77 to 1.95) 1.69 (1.05 to 2.74)* 2.59 (1.65 to 4.07)*** Sexual orientation¶ Bisexual 0.99 (0.55 to 1.77) 0.88 (0.46 to 1.67) 1.29 (0.71 to 2.36) Homosexual 1.24 (0.21 to 7.46) 0.80 (0.10 to 6.50) 11.55 (2.55 to 52.38)** Non-white ethnicity†† 1.03 (0.67 to 1.59) 1.12 (0.69 to 1.82) 0.85 (0.50 to 1.46) Wealth quintile‡‡ 2 0.68 (0.46 to 0.99)* 0.63 (0.41 to 0.97)* 0.41 (0.27 to 0.62)*** 3 0.94 (0.64 to 1.37) 0.81 (0.52 to 1.26) 0.50 (0.32 to 0.77)** 4 0.96 (0.65 to 1.41) 1.05 (0.68 to 1.62) 0.60 (0.39 to 0.94)* 5 (richest) 0.82 (0.55 to 1.23) 1.21 (0.77 to 1.90) 0.94 (0.60 to 1.45) Smoking status§§ Former smoker 0.54 (0.35 to 0.82)** 1.16 (0.77 to 1.74) 1.48 (0.99 to 2.22) Current smoker 0.87 (0.60 to 1.26) 1.93 (1.33 to 2.80)** 2.13 (1.46 to 3.11)*** Alcohol intake¶¶ Regularly 1.10 (0.81 to 1.48) 1.51 (1.04 to 2.18)* 1.97 (1.33 to 2.92)** Frequently 1.34 (0.98 to 1.83) 2.32 (1.58 to 3.39)*** 3.25 (2.17 to 4.88)*** Physical activity††† Moderately active at least once a week 1.02 (0.76 to 1.37) 1.28 (0.90 to 1.82) 0.74 (0.52 to 1.05) Vigorously active at least once a week 1.30 (0.94 to 1.80) 1.59 (1.08 to 2.35)* 1.14 (0.78 to 1.67) Limiting long-standing illness 0.84 (0.66 to 1.08) 1.16 (0.88 to 1.54) 1.13 (0.84 to 1.51) All figures are weighted for sampling probabilities and differential non-response.

1.37) 1.28 (0.90 to 1.82) 0.74 (0.52 to 1.05) Vigorously active at least once a week 1.30 (0.94 to 1.80) 1.59 (1.08 to 2.35)* 1.14 (0.78 to 1.67) Limiting long-standing illness 0.84 (0.66 to 1.08) 1.16 (0.88 to 1.54) 1.13 (0.84 to 1.51) All figures are weighted for sampling probabilities and differential non-response. *P<0.05, **p<0.01, ***p<0.001. †Multivariable multinomial logistic regression was performed. 0–1 sexual partners was the reference group in all analyses (n=775). ‡Unweighted sample sizes. Models were performed on complete cases and as such the analysed sample number differs from the number included in the bivariate analyses. §Reference category: married/cohabiting. ¶Reference category: heterosexual. ††Reference category: white. ‡‡Reference category: 1 (poorest). §§Reference category: never smoker. ¶¶Reference category: never/rarely. †††Reference category: inactive. Table 3 Multivariable models testing independent associations between predictors and lifetime number of sexual partners in women

§Reference category: married/cohabiting. ¶Reference category: heterosexual. ††Reference category: white. ‡‡Reference category: 1 (poorest). §§Reference category: never smoker. ¶¶Reference category: never/rarely. †††Reference category: inactive. Table 3 Multivariable models testing independent associations between predictors and lifetime number of sexual partners in women Predictor Lifetime number of sexual partners in women (adjusted OR (95% CI))† 2–4 (n=1170)‡ 5–9 (n=486) ≥10 (n=243) Age 0.95 (0.93 to 0.96)*** 0.90 (0.89 to 0.92)*** 0.90 (0.88 to 0.92)*** Partnership status§ Separated/divorced 6.98 (4.84 to 10.07)*** 12.68 (8.39 to 19.16)*** 14.53 (8.99 to 23.46)*** Widowed 1.73 (1.33 to 2.26)*** 1.66 (1.08 to 2.55)* 1.38 (0.72 to 2.65) Single/never married 1.56 (0.96 to 2.54) 2.22 (1.26 to 3.89)** 5.33 (2.93 to 9.67)*** Sexual orientation¶ Bisexual 1.10 (0.70 to 1.74) 2.09 (1.27 to 3.46)** 5.87 (3.50 to 9.85)*** Homosexual 3.98 (0.59 to 26.70) 5.61 (0.68 to 46.49) 5.05 (0.47 to 54.57) Non-white ethnicity†† 0.35 (0.21 to 0.58)*** 0.22 (0.11 to 0.46)*** 0.11 (0.03 to 0.35)*** Wealth quintile‡‡ 2 0.97 (0.71 to 1.31) 0.83 (0.55 to 1.25) 0.90 (0.55 to 1.48) 3 1.05 (0.77 to 1.43) 0.90 (0.59 to 1.37) 1.02 (0.60 to 1.74) 4 0.83 (0.60 to 1.15) 1.08 (0.71 to 1.66) 0.74 (0.42 to 1.32) 5 (richest) 0.88 (0.63 to 1.23) 1.15 (0.74 to 1.77) 1.20 (0.69 to 2.10) Smoking status§§ Former smoker 1.17 (0.78 to 1.73) 1.78 (1.14 to 2.78)* 2.24 (1.31 to 3.83)** Current smoker 1.53 (1.13 to 2.07) 1.95 (1.34 to 2.83)*** 3.18 (2.06 to 4.90)*** Alcohol intake¶¶ Regularly 1.23 (0.98 to 1.54) 2.20 (1.58 to 3.06)*** 1.16 (0.78 to 1.75) Frequently 1.59 (1.23 to 2.07)** 3.16 (2.18 to 4.56)*** 2.17 (1.39 to 3.40)** Physical activity††† Moderately active at least once a week 1.51 (1.18 to 1.92)** 1.26 (0.90 to 1.76) 1.68 (1.08 to 2.60)* Vigorously active at least once a week 1.60 (1.19 to 2.14)** 1.74 (1.19 to 2.55)** 1.73 (1.04 to 2.88)* Limiting long-standing illness 1.12 (0.91 to 1.38) 1.53 (1.16 to 2.02)** 1.58 (1.10 to 2.26)* All figures are weighted for sampling probabilities and differential non-response.

.26 (0.90 to 1.76) 1.68 (1.08 to 2.60)* Vigorously active at least once a week 1.60 (1.19 to 2.14)** 1.74 (1.19 to 2.55)** 1.73 (1.04 to 2.88)* Limiting long-standing illness 1.12 (0.91 to 1.38) 1.53 (1.16 to 2.02)** 1.58 (1.10 to 2.26)* All figures are weighted for sampling probabilities and differential non-response. *P<0.05, **p<0.01, ***p<0.001. †Multivariable multinomial logistic regression was performed. 0–1 sexual partners was the reference group in all analyses (n=1198). ‡Unweighted sample sizes. Models were performed on complete cases and as such the analysed sample number differs from the number included in the bivariate analyses. §Reference category: married/cohabiting. ¶Reference category: heterosexual. ††Reference category: white. ‡‡Reference category: 1 (poorest). §§Reference category: never smoker. ¶¶Reference category: never/rarely. †††Reference category: inactive. There were no notable differences in the results when men (n=38) and women (n=28) who reported no sexual partners were excluded (online supplementary tables 1 and 2, respectively). 10.1136/bmjsrh-2018-200230.supp1Supplementary file 1

§§Reference category: never smoker. ¶¶Reference category: never/rarely. †††Reference category: inactive. There were no notable differences in the results when men (n=38) and women (n=28) who reported no sexual partners were excluded (online supplementary tables 1 and 2, respectively). 10.1136/bmjsrh-2018-200230.supp1Supplementary file 1 Discussion Using data from a large, representative sample of older adults living in England, the present study has identified behavioural and sociodemographic factors that are associated with a greater number of lifetime sexual partners. Men were more likely than women to report a higher number of lifetime sexual partners: 39.7% of men and 24.1% of women reported having had at least five sexual partners in their lifetime, and 19.9% of men and 8.5% of women had had 10 or more partners. Despite differences in the absolute number of partners, there were a number of similarities in the factors that predicted lifetime number of sexual partners in men and women. Being younger, separated/divorced or single/never married, being a current or former smoker, and drinking alcohol regularly or frequently were independently associated with a higher number of lifetime sexual partners in both sexes. Sex-specific predictors were also observed. In men but not women, being gay and being in the highest and lowest quintiles of wealth were associated with a higher number of previous sexual partners. In women but not men, being bisexual, of white ethnicity, participating in regular vigorous physical activity and being free of limiting long-standing illness were associated with a greater number of previous sexual partners.

d lowest quintiles of wealth were associated with a higher number of previous sexual partners. In women but not men, being bisexual, of white ethnicity, participating in regular vigorous physical activity and being free of limiting long-standing illness were associated with a greater number of previous sexual partners. The finding that men have a higher number of lifetime sexual partners than women is consistent with previous research in adolescents.26 This may be driven by men’s higher testosterone levels leading to increased feelings of sexual desire,27 and/or the ‘sexual double standard’ that sees more favourable societal attitudes towards promiscuity in males than females.28 It may also be at least partly attributable to social desirability bias, with men more likely than women to overreport the number of sexual partners they have had.29 The finding that among the over-50s, younger age was associated with a higher number of lifetime sexual partners is interesting, and likely reflects changes in attitudes and opinions towards sexuality and changes in rates of divorce and separation30 between generations, even those relatively close in age. Little has been published on older adults’ attitudes towards sexuality and further work in this area is required. Perhaps unsurprisingly, those participants who were not married (separated/divorced or single/never married) reported a higher number of sexual partners. This is likely owing to greater opportunity to engage in sexual activity with multiple sexual partners throughout life, although the frequency of sexual activity reported by this population compared with those who are married is lower (data not shown).

single/never married) reported a higher number of sexual partners. This is likely owing to greater opportunity to engage in sexual activity with multiple sexual partners throughout life, although the frequency of sexual activity reported by this population compared with those who are married is lower (data not shown). In both sexes, smoking and alcohol use were associated with a higher number of lifetime sexual partners. Alcohol use has previously been shown to be associated with a greater number of lifetime sexual partners and a greater chance of HIV transmission.31 32 Having a greater tendency for risk taking has been shown to be associated with smoking and alcohol consumption,33 34 and it is plausible that those who have a tendency to take risks are also likely to have a higher number of lifetime sexual partners. Moreover, alcohol consumption/being drunk affects decision-making processes and may increase chances of unplanned sexual encounters.

associated with smoking and alcohol consumption,33 34 and it is plausible that those who have a tendency to take risks are also likely to have a higher number of lifetime sexual partners. Moreover, alcohol consumption/being drunk affects decision-making processes and may increase chances of unplanned sexual encounters. Homosexuality predicted a higher number of lifetime sexual partners in men only. The literature is mixed on this topic, with some studies indicating that gay men have a higher number of sexual partners compared with heterosexual men and some studies observing a similar prevalence.35 The inconsistency in the data on the number of sexual partners of gay men probably reflects flaws in the sampling techniques of earlier studies (eg, recruiting subjects in gay bars) and their completion before the HIV epidemic.35 However, the present study does suggest that gay men do have a greater number of sexual partners in their lifetime. This is of clinical importance, given that gay men are a key population with a high prevalence of HIV35 and a greater number of sexual partners is known to increase the risk of HIV transmission. In the present study, being in the highest and lowest quintiles of wealth was associated with a higher number of previous sexual partners in men but not in women. Socioeconomic status is an established predictor of divorce, with higher divorce rates among those from the lowest and highest socioeconomic groups.36 This ties in with the finding that those who are not married have a higher number of sexual partners.

higher number of previous sexual partners in men but not in women. Socioeconomic status is an established predictor of divorce, with higher divorce rates among those from the lowest and highest socioeconomic groups.36 This ties in with the finding that those who are not married have a higher number of sexual partners. Women from white ethnic backgrounds reported a higher number of lifetime sexual partners than those from other ethnicities. A plausible explanation is differences in cultural norms between the ethnicities. For example, a large proportion of Asians and only a small proportion of white individuals residing in the UK follow Islam.37 For those who follow Islam, sexual intercourse between unmarried men and women is forbidden and thus this population is likely to have a low number of lifetime sexual partners. It is not clear why an association with ethnicity was only observed in women; further research is required to explore this in greater detail. Another factor that was found to predict number of sexual partners in women only was vigorous physical activity. Women who reported engaging in moderate and vigorous physical activity at least weekly were more likely to report a higher number of lifetime sexual partners than those who were less active. Vigorous physical activity is usually achieved via participating in sport.38 Sport participation has been shown to track across the lifespan39 and may provide a setting for social interaction, thus increasing the potential to meet a greater number of potential sexual partners. However, further research is required to confirm or refute this hypothesis. Moreover, physical activity has been shown to lower levels of depression and depression has been shown to be associated with higher levels of sexual problems. Absence of limiting long-standing illness was associated with a higher number of sexual partners in women. Women who do not have such illnesses may have had greater opportunity across the lifespan for greater social interaction (eg, via sports participation; those with limiting long-standing illness are less likely to participate in sport) and increasing one’s opportunity to meet a greater number of sexual partners. Finally, women who were bisexual had a higher number of lifetime sexual partners than those who were heterosexual. A plausible explanation for this finding is elusive and further research is needed.

ss are less likely to participate in sport) and increasing one’s opportunity to meet a greater number of sexual partners. Finally, women who were bisexual had a higher number of lifetime sexual partners than those who were heterosexual. A plausible explanation for this finding is elusive and further research is needed. This is the first study to identify sociodemographic and behavioural correlates of the number of lifetime sexual partners in an older sample. Strengths of the study include the large, representative sample and data on a wide range of potential correlates. However, findings from the present study must be interpreted in light of its limitations. All measures were self-reported which may have introduced reporting or recall bias (particularly in relation to lifetime number of sexual partners). However, the item on number of sexual partners was included in a paper-based questionnaire rather than in the face-to-face interview, and it was made clear to participants that survey responses would remain anonymous. The sample was almost exclusively white so findings may not generalise to other ethnic groups in which attitudes towards sex may differ. While the measure of smoking took into account historical behaviour by including a category for ex-smokers, data on alcohol intake and physical activity only reflected participants’ current behaviour. The predictive power of these variables in determining number of sexual partners over the life course may therefore have been over- or underestimated. Longitudinal cohort studies that collect measures of these behaviours throughout the life course could assess prospective relationships by asking participants to report their lifetime number of sexual partners in future waves of data collection. Finally, the current data were not able to ascertain for each participant when during the lifespan the period at which the greatest number of sexual partners were acquired. For some participants it is possible that they only had a very high number of sexual partners when adolescents, others middle age, and for some old age.

ly, the current data were not able to ascertain for each participant when during the lifespan the period at which the greatest number of sexual partners were acquired. For some participants it is possible that they only had a very high number of sexual partners when adolescents, others middle age, and for some old age. In conclusion, the present study has identified a number of sociodemographic and behavioural correlates of lifetime sexual partners. This information may help health practitioners to identify individuals who are at greatest risk of STIs and their associated health complications across the life course. Moreover, findings from the present study could also benefit younger adults through targeted interventions to educate groups at risk of having a high number of sexual partners about the risks associated with doing so, and the practice of safe sex. The authors acknowledge their previous publication that used the same dataset (ELSA): Jackson S, Firth J, Veronese N, Stubbs B, Koyanagi A, Yang L, Smith L. Decline in sexuality and wellbeing in older adults: a population based study. Journal of Affective Disorders 2018;245:912-917. Contributors: SJ and LS conceived the idea. SJ carried out statistical analyses. SJ and LS interpreted the findings. SJ and LS drafted the manuscript. All authors provided extensive comments on the manuscript. All authors approved the final manuscript before submission. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared.

Contributors: SJ and LS conceived the idea. SJ carried out statistical analyses. SJ and LS interpreted the findings. SJ and LS drafted the manuscript. All authors provided extensive comments on the manuscript. All authors approved the final manuscript before submission. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not required.

Key messages Consistent condom use among adolescents is suboptimal in Northern Canada – this is particularly concerning as this region has sexually transmitted infection rates seven-fold the national average. Using alcohol and drugs was associated with a higher likelihood of being sexually active among both adolescent boys and girls, and a lower likelihood of consistent condom use among adolescent girls. Gender differences are important to consider: sexually diverse girls had a lower likelihood of consistent condom use while sexually diverse boys had a higher likelihood.

Using alcohol and drugs was associated with a higher likelihood of being sexually active among both adolescent boys and girls, and a lower likelihood of consistent condom use among adolescent girls. Gender differences are important to consider: sexually diverse girls had a lower likelihood of consistent condom use while sexually diverse boys had a higher likelihood. Introduction Over 4 million persons live in the Arctic region, a vast area spanning eight countries.1 There is a rich diversity of experiences, geographies and cultures across the Arctic, yet there are similarities regarding insufficient health resources, infrastructure and research, and a history of Indigenous Peoples’ colonisation that render knowledge acquired in this context valuable for understanding other Arctic and global regions.2 3 In Canada, youth living in the Northwest Territories (NWT) experience adverse sexual health outcomes, including a higher incidence of unintended pregnancies and sexually transmitted infections (STIs).3 4 In 2010, the pregnancy rate among adolescent girls aged 15–19 years in the NWT was 40.4 per 1000 in comparison to the national average of 28.8 per 1000.5 In 2015, the incidence of STIs in the NWT was seven times higher than the national incidence rate of 3.9 cases per 1000.6 It was reported that 1 in 15 men and 1 in 8 women aged 15–24 years in the NWT had at least one STI during 2016.7 In the same year, the prevalence of gonorrhoea and chlamydia in the NWT was 19 times7 and six times8 higher than the reported national prevalence, respectively.

ational incidence rate of 3.9 cases per 1000.6 It was reported that 1 in 15 men and 1 in 8 women aged 15–24 years in the NWT had at least one STI during 2016.7 In the same year, the prevalence of gonorrhoea and chlamydia in the NWT was 19 times7 and six times8 higher than the reported national prevalence, respectively. While the presence of another STI elevates youth’s vulnerability for HIV acquisition and transmission, documented HIV rates in the NWT are comparable with the rest of Canada.9 However, these prevalence estimates may be under-reported as a result of socio-structural barriers that reduce uptake of HIV and STI testing in Northern and rural regions, including geographic isolation, stigma, and reduced access to sexual health services.4 STI incidence and unintended pregnancies may be substantially reduced through safer sex practices, and condom use is a key practice to prevent both outcomes among adolescents.10 11 Studies have explored factors associated with sexual activity and condom use across diverse adolescent populations.10 12 STI knowledge may be a protective factor associated with increased condom use.11 In contrast, drug and alcohol use may increase adolescents’ vulnerability to STIs through reducing sexual inhibitions and compromising sexual negotiation skills including contraceptive use.4 13 14

cross diverse adolescent populations.10 12 STI knowledge may be a protective factor associated with increased condom use.11 In contrast, drug and alcohol use may increase adolescents’ vulnerability to STIs through reducing sexual inhibitions and compromising sexual negotiation skills including contraceptive use.4 13 14 Youth substance use and its association with sexual health is particularly important to assess in the Arctic, where alcohol and substance use play a significant role in contributing to health disparities generally.2 The NWT reported the highest rates of monthly heavy drinking in Canada in 2014.15 Arctic Canadian regions such as the NWT have a high proportion of Indigenous Peoples (50.7%),16 many of whom experience sexual and mental health disparities as a result of historical and ongoing colonisation, racism and discrimination, and the effects of intergenerational trauma from the forced removal of Indigenous children to attend residential schools.17 18 A recent study with adolescent girls in the NWT revealed that substance use was associated with reporting multiple sex partners.19 Limited research has examined substance use and its association with sexual activity and condom use among adolescents in the NWT.

of Indigenous children to attend residential schools.17 18 A recent study with adolescent girls in the NWT revealed that substance use was associated with reporting multiple sex partners.19 Limited research has examined substance use and its association with sexual activity and condom use among adolescents in the NWT. Social-ecological factors are associated with gender and sexual orientation differences in sexual health.12 20 Adolescent girls experience particular socio-ecological vulnerabilities to STI acquisition, including alcohol use norms in communities that can influence sexual decision-making,21 gender norms that reduce their sexual relationship power, and biological vulnerabilities associated with the cervical opening and cervical mucus development.22 Lesbian, gay, bisexual queer and other sexually diverse (LGBQ+) youth in Canada experience increased sexual health disparities in part due to limited access to tailored sexual health information and resources.23 This is evidenced in higher prevalence of STIs among bisexual vs heterosexual women24 and disproportionate syphilis infections among gay and bisexual men vs heterosexual men in Canada.25 Qualitative research in the NWT suggests that LGBQ + youth experience safer sex barriers due to lack of knowledge and safer sex tool availability, in addition to stigma that produces shame and fear of accessing HIV/STI resources.20 Gender and sexual orientation differences in sexual practices are understudied among Northern youth.

research in the NWT suggests that LGBQ + youth experience safer sex barriers due to lack of knowledge and safer sex tool availability, in addition to stigma that produces shame and fear of accessing HIV/STI resources.20 Gender and sexual orientation differences in sexual practices are understudied among Northern youth. Identifying patterns of sexual activity and condom use can inform sexual health programming to promote uptake of safer sex practices.26 However, knowledge gaps remain regarding sexual and condom use practices among NWT youth. The latest territorial/national data on sexual behaviour is from the 2009/2010 Canadian Community Health Survey, which reported that: 38.6% of individuals aged 15–24 years in the NWT had multiple sexual partners, in comparison to the national estimate of 32.5%27; 78.6% of NWT youth had used a condom during their last sexual encounter, higher than the national estimate of 67.9%27; and sexually active men reported higher condom use (73% vs 63%) than women.27 The percentage reporting more than one sexual partner was higher among those aged 15–17 years than those aged 20–24 years (35% vs 30%). However, with the exception of gender and age, these data did not examine further variables associated with sexual practices. In addition, it did not examine consistent condom use.27 Little is known of factors associated with engaging in sex, and using condoms, among Northern adolescents, particularly early adolescents aged 13–15 years in the NWT.

f gender and age, these data did not examine further variables associated with sexual practices. In addition, it did not examine consistent condom use.27 Little is known of factors associated with engaging in sex, and using condoms, among Northern adolescents, particularly early adolescents aged 13–15 years in the NWT. This study aimed to tackle knowledge gaps regarding sexual practices among NWT youth, specifically to examine (a) the prevalence of sexual activity and consistent condom use among NWT youth aged 13–18 years, and (b) socio-demographic, risk, and protective factors associated with sexual activity and consistent condom use among NWT youth.

o tackle knowledge gaps regarding sexual practices among NWT youth, specifically to examine (a) the prevalence of sexual activity and consistent condom use among NWT youth aged 13–18 years, and (b) socio-demographic, risk, and protective factors associated with sexual activity and consistent condom use among NWT youth. Methods We implemented a cross-sectional survey with Northern and Indigenous youth recruited using purposive, venue-based sampling at 23 schools in 17 NWT communities (Aklavik, Whati, Fort McPherson, N’Dilo, Lutselk’e, Fort Liard, Fort Simpson, Yellowknife, Ulukhaktok, Fort Resolution, Behchoko, Inuvik, Tuktoyaktuk, Hay River, Katlodeeche First Nation, Fort Smith, and Norman Wells) between October 2016 to June 2017. The 17 communities were purposively selected out of 25 communities with junior/secondary schools in the NWT, and only 1/24 schools that were offered the opportunity to participate in the research declined. Each school selected classrooms in grades 7–12 to participate. The paper-based survey questionnaires were self-administered among a cohort of secondary school youth aged 13–18 years, who agreed to participate in a sexual health workshop held by a Northern and Indigenous youth sexual health programme (Fostering Open eXpression Among Youth [FOXY]). The survey results included in this analysis are the pre-test results (conducted before the workshops). Self-administration of surveys helped to reduce social desirability bias. Trained research assistants provided details of the informed consent process, were present to answer any questions, and collected the completed surveys that were entered into a database for analyses. The study protocol was approved by the research ethics boards of the University of Toronto and the Aurora Research Institute. All participants provided informed voluntary written consent prior to engaging in the study.

er any questions, and collected the completed surveys that were entered into a database for analyses. The study protocol was approved by the research ethics boards of the University of Toronto and the Aurora Research Institute. All participants provided informed voluntary written consent prior to engaging in the study. Self-reported information on gender and gender identity, sexual orientation (lesbian, bisexual, gay, queer or other sexual diversity [LGBQ+] vs. heterosexual/straight), ethnicity (Indigenous or non-Indigenous), rural vs. urban residence (living in Yellowknife [urban] vs. outside of Yellowknife [rural]), past 3 month drug/alcohol use (‘In the past 3 months have you drank alcohol or used drugs?’ Yes/No) was collected. We assessed STI knowledge using a scale based on the Sexually Transmitted Disease Knowledge Questionnaire,28 which was previously validated in the NWT with adolescent girls19 (Cronbach’s alpha in current study=0.81). The primary outcomes of interest were (a) engagement in any sexual activity, defined as ever engaging in oral, vaginal or anal sex (‘Have you done any of these in your life? (Yes/No): vaginal sex, anal sex, oral sex’) and (b) for those that were sexually active in the past 3 months, consistent condom use within the past 3 months, defined as individuals indicating that they ‘always’ used condoms during vaginal/anal sex (’If you have had vaginal or anal sex in the past 3 months, how often do you use a condom?' Always/Sometimes/Never).

and (b) for those that were sexually active in the past 3 months, consistent condom use within the past 3 months, defined as individuals indicating that they ‘always’ used condoms during vaginal/anal sex (’If you have had vaginal or anal sex in the past 3 months, how often do you use a condom?' Always/Sometimes/Never). Summary statistics were used to compare socio-demographic differences across the primary outcomes by gender. Due to the small sample size, individuals identifying as transgender or other non-cisgender genders (n=7) were excluded from bivariate and multivariate analysis, however frequencies are reported. Statistical comparisons were made using Student’s t-test for continuous variables, and χ2 test for homogeneity for categorical variables. For categorical comparisons with any cell size less than five, Fisher’s exact tests were used to determine statistical significance. Statistical significance was evaluated using two-sided p values at a 0.05 level of significance. We use crude and adjusted Poisson regression models with a robust sandwich error variance to produce adjusted relative risks estimates of the likelihood of experiencing the primary outcomes.29 As a sensitivity analysis, we conducted a log binomial regression model with GEE adjustment for community to examine if findings clustered by community.

Poisson regression models with a robust sandwich error variance to produce adjusted relative risks estimates of the likelihood of experiencing the primary outcomes.29 As a sensitivity analysis, we conducted a log binomial regression model with GEE adjustment for community to examine if findings clustered by community. Patient and public involvement This research is a collaborative partnership with an Indigenous agency, including manuscript authors (CLL, NM, KM) that guided the study foci, research question development, and provided feedback into the outcome measures. The questions were pilot-tested with Northern and Indigenous adolescent peer leaders, many based in the communities the study was conducted in and supporting the sexual health workshops. Results will be shared in factsheets and briefs developed with a NWT Indigenous knowledge translation agency. Results In total 610 individuals completed the survey, 302 (49.5%) self-reported gender as cisgender girls, 298 (48.9%) as cisgender boys, 7 (1.2%) identified as transgender or gender diverse (TG/GD), and 3 (0.5%) had missing information on gender. Individuals with missing information on gender were excluded from further analysis, and transgender and gender diverse participant responses were described in the frequency table, but excluded from further analysis due to the small number.

ender or gender diverse (TG/GD), and 3 (0.5%) had missing information on gender. Individuals with missing information on gender were excluded from further analysis, and transgender and gender diverse participant responses were described in the frequency table, but excluded from further analysis due to the small number. The mean age was 14.2 years (SD: 1.5) with a age range of 13–18 years. We observed a mean STI knowledge score of 3.46 (SD: 3.21); min/max of 0 and 13 correct answers, respectively. Most participants identified as Indigenous (n=444; 73.1%) and 14.0% (n=85) identified as lesbian, gay, bisexual, queer or other sexually diversity (LGBQ+). Most individuals resided outside of Yellowknife in rural communities (n=430; 70.8%). One-third reported ever using alcohol or drugs (n=211; 34.8%). Among sexually active individuals (n=115; 18.9%) less than half (n=54; 47.0%) reported past 3 month consistent condom use. Statistically significant gender differences were observed in STI knowledge, sexual activity, LGBQ+ identity, and alcohol or drug use for cisgender girls verses boys, respectively (table 1). Statistically significant differences were observed across all categories when stratifying by sexual activity, with higher age, rural community of residence vs. Yellowknife, higher STI knowledge, Indigenous vs. non-Indigenous identity, alcohol/drug use, and LGBQ +vs heterosexual identity for sexually active verses non-sexually active individuals, respectively. Stratifying by consistent condom use yielded no statistically significant differences (table 2).

nity of residence vs. Yellowknife, higher STI knowledge, Indigenous vs. non-Indigenous identity, alcohol/drug use, and LGBQ +vs heterosexual identity for sexually active verses non-sexually active individuals, respectively. Stratifying by consistent condom use yielded no statistically significant differences (table 2). Table 1 Socio-demographic characteristics among a sample of Northern adolescents in the Northwest Territories (NWT), Canada (n=607)† Characteristic Adolescent cisgender girl (n=302) Adolescent cisgender boy (n=298) Transgender youth (n=7) P value* Socio-demographic factors

nity of residence vs. Yellowknife, higher STI knowledge, Indigenous vs. non-Indigenous identity, alcohol/drug use, and LGBQ +vs heterosexual identity for sexually active verses non-sexually active individuals, respectively. Stratifying by consistent condom use yielded no statistically significant differences (table 2). Table 1 Socio-demographic characteristics among a sample of Northern adolescents in the Northwest Territories (NWT), Canada (n=607)† Characteristic Adolescent cisgender girl (n=302) Adolescent cisgender boy (n=298) Transgender youth (n=7) P value* Socio-demographic factors Age in years (mean (SD)) 14.2 (1.5) 14.2 (1.4) 14.0 (1.4) 0.735 STI knowledge (mean (SD)) 3.9 (3.6) 3.0 (2.6) 4.6 (4.0) <0.001 Indigenous identity No 74 (24.5% (19.8% to 29.8%)) 88 (29.5% (24.4% to 35.1%)) 1 (14.3% (0.4% to 57.9%)) 0.166 Yes 228 (75.5% (70.2% to 80.2%)) 210 (70.5% (64.9% to 75.6%)) 6 (85.7% (42.1% to 99.6%)) LGBQ+ identity Missing 6 (2.0% (0.7% to 4.3%)) 2 (0.7% (0.1% to 2.4%)) – <0.001 No 234 (77.5% (72.3% to 82.1%)) 279 (93.6% (90.2% to 96.1%)) 1 (14.3% (0.4% to 57.9%)) Yes 62 (20.5% (16.1% to 25.5%)) 17 (5.7% (3.4% to 9.0%)) 6 (85.7% (42.1% to 99.6%)) Rural residence No 82 (27.2% (22.2% to 32.5%)) 92 (30.9% (25.7% to 36.5%)) 3 (42.9% (9.9% to 81.6%)) 0.315 Yes 220 (72.9% (67.5% to 77.8%)) 206 (69.1% (63.5% to 74.3%)) 4 (57.1% (18.4% to 90.1%)) Alcohol or drug use in past 3 months No 180 (59.6% (53.8% to 65.2%)) 212 (71.1% (65.6% to 76.2%)) 4 (57.1% (18.4% to 90.1%)) 0.003 Yes 122 (40.4% (34.8% to 46.2%)) 86 (28.9% (23.8% to 34.4%)) 3 (42.9% (9.9% to 81.6%)) Sexually active No 233 (77.2% (72.0% to 81.8%)) 252 (84.6% (80.0% to 88.5%)) 7 (100% (59% to 100%)) 0.021 Yes 69 (22.9% (2.0% to 28.0%)) 46 (15.4% (11.5% to 20.0%)) –

2%)) 212 (71.1% (65.6% to 76.2%)) 4 (57.1% (18.4% to 90.1%)) 0.003 Yes 122 (40.4% (34.8% to 46.2%)) 86 (28.9% (23.8% to 34.4%)) 3 (42.9% (9.9% to 81.6%)) Sexually active No 233 (77.2% (72.0% to 81.8%)) 252 (84.6% (80.0% to 88.5%)) 7 (100% (59% to 100%)) 0.021 Yes 69 (22.9% (2.0% to 28.0%)) 46 (15.4% (11.5% to 20.0%)) – Sexual practices, risk and protective factors among sexually active youth Age in years (mean (SD)) 15.3 (2.0) 15.6 (1.4) – 0.366 STI knowledge (mean (SD)) 5.8 (4.0) 3.8 (3.0) – 0.006 Number of sexual partners (mean (SD)) 2.3 (1.5) 2.6 (1.7) – 0.280 Types of sexual act‡ Not stated 1 (1.5% (0.0% to 7.8%)) 2 (4.3% (0.5% to 14.8%)) – – Oral 35 (50.7% (38.4% to 63.0%)) 21 (45.7% (30.9% to 61.0%)) – 0.546 Vaginal 64 (92.8% (83.9% to 97.6%)) 42 (91.3% (79.2% to 97.6%)) – 0.779 Anal 10 (14.5% (7.2% to 25.0%)) 5 (10.9% (3.6% to 23.6%)) – 0.576 Consistent condom use No 39 (56.5% (44.0% to 68.4%)) 22 (47.8% (32.9% to 63.1%)) – 0.360 Yes 30 (43.5% (31.6% to 56.0%)) 24 (52.3% (36.9% to 67.1%)) – Alcohol or drug use in past 3 months No 14 (20.3% (11.6% to 31.7%)) 14 (30.4% (17.7%–45.8%)) – 0.214 Yes 55 (79.7% (68.3% to 88.4%)%)) 32 (69.6% (54.2% to 82.3%)) – Indigenous identity No 9 (13.0% (6.1% to 23.3%)) 8 (17.4% (7.8% to 31.4%)) – 0.520 Yes 60 (87.0% (76.7% to 93.9%)) 38 (82.6% (68.6% to 92.2%)) – Rural residence No 3 (4.4% (0.9% to 12.2%)) 10 (21.7% (10.9% to 36.4%)) – 0.004 Yes 66 (95.6% (87.7% to 99.1%)) 36 (78.3% (63.6% to 89.1%)) – *Significance determined at α=0.05, and p values produced via Student’s t-test, χ2 test for homogeneity, or Fisher’s exact test.

% (76.7% to 93.9%)) 38 (82.6% (68.6% to 92.2%)) – Rural residence No 3 (4.4% (0.9% to 12.2%)) 10 (21.7% (10.9% to 36.4%)) – 0.004 Yes 66 (95.6% (87.7% to 99.1%)) 36 (78.3% (63.6% to 89.1%)) – *Significance determined at α=0.05, and p values produced via Student’s t-test, χ2 test for homogeneity, or Fisher’s exact test. †Types of sexual acts are overlapping categories causing the sum total to exceed the number of sexually active youth. ‡Unless otherwise stated, the number of individuals and column percentages along with their 95% binomial confidence intervals are reported in each cell. Those with missing information on variables or who’s self-reported gender was other than cisgender girl/boy were reported but not used in for statistical comparisons due to small numbers. SD, standard deviation; STI, sexually transmitted infection. Table 2 Sexual practices among a sample of Northern adolescents in the Northwest Territories (NWT), Canada (n=607)

‡Unless otherwise stated, the number of individuals and column percentages along with their 95% binomial confidence intervals are reported in each cell. Those with missing information on variables or who’s self-reported gender was other than cisgender girl/boy were reported but not used in for statistical comparisons due to small numbers. SD, standard deviation; STI, sexually transmitted infection. Table 2 Sexual practices among a sample of Northern adolescents in the Northwest Territories (NWT), Canada (n=607) Sexually active Consistent condom use No (n=495) Yes (n=115) P value* No (n=61) Yes (n=54) P value* Age in years (mean (SD)) 13.9 (1.2) 15.4 (1.8) <0.001 15.4 (2.1) 15.4 (1.4) 0.789 Types of sexual act† Not stated – 1 (1.5%(0.0% to 4.7%)) – – 1 (1.9%(0.0% to 9.9%)) – Oral – 56 (48.7%(39.3% to 58.2%)) – 37 (60.7%(47.3% to 72.9%)) 19 (35.2%(22.7% to 49.4%)) 0.008 Vaginal – 106 (92.2%(85.7% to 96.4%)) – 55 (90.1%(79.8% to 96.3%)) 51 (94.4%(84.6%– to 98.8%)) 0.398 Anal – 15 (13.0%(7.5% to 20.6%)) – 7 (11.5%(4.7% to 22.2%)) 8 (14.8%(6.6% to 27.1%)) 0.599 Number of sexual partners (mean (SD)) – 2.4 (1.6) – 2.1 (1.6) 2.6 (1.5) 0.171 STI knowledge score (mean (SD)) 3.1 (3.0) 5.0 (3.7) <0.001 5.4 (3.8) 4.7 (3.6) 0.343 Gender Adolescent cisgender girls 233 (77.2%(72.0% to 81.8%)) 69 (22.9%(18.2% to 28.0%)) 0.021 39 (56.5%(44.0% to 68.4%)) 30 (43.5%(31.6% to 56.0%)) 0.360 Adolescent cisgender boys 252 (84.6%(80.0% to 88.5%)) 46 (15.4%(11.5% to 20.0%)) 22 (47.8%(32.9% to 63.1%)) 24 (52.2%(36.9% to 67.1%)) Transgender youth 7 (100.0%(59.0% to 100%)) 0 (0.0%(0.0% to 41.0%)) – – Indigenous identity No 146 (89.6%(83.8% to 93.8%)) 17 (10.4%(6.2% to 16.2%)) 0.001 9 (52.9%(27.8% to 77.0%)) 8 (47.1%(23.0% to 72.2%)) 0.993 Yes 349 (78.1%(73.9% to 81.8%)) 98 (21.9%(18.2% to 26.1%)) 52 (53.1%(42.7% to 63.2%)) 46 (46.9%(36.8% to 57.3%)) LGBQ+ identity Missing 8 (88.9%(51.8% to 99.7%)) 1 (11.1%(0.3% to 48.2%)%)) (0.0%(0.0%% to 97.5%)) 1 (100.0%(2.5% to 100%)) No 424 (82.2%(78.6% to 85.4%)) 92 (17.8%(14.6% to 21.4%)) 0.038 46 (50.0%(39.4% to 60.6%)) 46 (50.0%(39.4% to 60.6%)) 0.125 Yes 63 (74.1%(63.5% to 83.0%)) 22 (25.9%(17.0% to 36.5%)) 15 (68.2%(45.1% to 86.1%)) 7 (31.8%(13.9% to 54.9%)) Rural residence (n (row %)) No 164 (92.7%(87.8% to 96.0%)) 13 (7.3%(4.0% to 12.2%)) <0.001 4 (30.8%(9.1% to 61.4%)) 9 (69.2%(38.6% to 90.9%)) 0.087 Yes 331 (76.4%(72.2% to 80.4%)) 102 (23.6%(19.6% to 27.8%)) 57 (55.9%(45.7% to 65.7%)) 45 (44.1%(34.3% to 54.3%)) Alcohol or drug use in past 3 months No 371 (93.0%(90.0% to 95.3%)) 28 (7.0%(4.7% to 10.0%)) <0.001 11 (39.3%(21.5% to 59.4%)) 17 (60.7%(40.6% to 78.5%)) 0.094 Yes 124 (58.8%(51.8% to 65.5%)) 87 (41.2%(34.5% to 48.2%)) 50 (57.5%(46.4% to 68.0%)

3.6%(19.6% to 27.8%)) 57 (55.9%(45.7% to 65.7%)) 45 (44.1%(34.3% to 54.3%)) Alcohol or drug use in past 3 months No 371 (93.0%(90.0% to 95.3%)) 28 (7.0%(4.7% to 10.0%)) <0.001 11 (39.3%(21.5% to 59.4%)) 17 (60.7%(40.6% to 78.5%)) 0.094 Yes 124 (58.8%(51.8% to 65.5%)) 87 (41.2%(34.5% to 48.2%)) 50 (57.5%(46.4% to 68.0%) ) 37 (42.5%(32.0% to 53.6%)) *Significance determined at α=0.05, and p values produced via Student’s t-test, χ2 test for homogeneity, or Fisher’s exact test. Those with missing information on variables or whose self-reported gender was other than girl/boy were not used in the statistical comparisons. †Types of sexual acts are overlapping categories causing the sum total to exceed the number of sexually active youth. Column percentages represent the marginal proportions. SD, standard deviation; STI, sexually transmitted infection. Table 3 and table 4 report socio-demographic factors associated with sexual activity and consistent condom use among cisgender girls and boys. In adjusted analyses including socio-demographic variables, age was associated with increased odds of sexual activity among both boys and girls, and rural vs Yellowknife residence was associated with lower odds of sexual activity among boys. Odds of consistent condom use were lower for rural vs Yellowknife residence among boys and girls, and lower for LGBQ+ girls vs heterosexual girls. Among boys, Indigenous identity and LGBQ+ identity were associated with increased odds of consistent condom use.

as associated with lower odds of sexual activity among boys. Odds of consistent condom use were lower for rural vs Yellowknife residence among boys and girls, and lower for LGBQ+ girls vs heterosexual girls. Among boys, Indigenous identity and LGBQ+ identity were associated with increased odds of consistent condom use. Table 3 Socio-demographic factors associated with the odds of sexual activity and consistent condom use among adolescent cisgender girl (aged 13–18 years) participants in the Northwest Territories (NWT), Canada

as associated with lower odds of sexual activity among boys. Odds of consistent condom use were lower for rural vs Yellowknife residence among boys and girls, and lower for LGBQ+ girls vs heterosexual girls. Among boys, Indigenous identity and LGBQ+ identity were associated with increased odds of consistent condom use. Table 3 Socio-demographic factors associated with the odds of sexual activity and consistent condom use among adolescent cisgender girl (aged 13–18 years) participants in the Northwest Territories (NWT), Canada Socio-demographic factor uRR P value aRR* P value aRR† P value Sexual activity (n=115) Age 1.23 (1.12, 1.34) <0.001 1.17 (1.11, 1.24) <0.001 1.16 (1.08, 1.25) <0.001 Indigenous identity 2.17 (1.13, 4.16) 0.020 1.18 (0.89, 1.56) 0.243 1.12 (0.65, 1.95) 0.678 LGBQ+  identity 1.48 (0.94, 2.33) 0.088 1.3 (0.88, 1.93) 0.183 1.35 (0.91, 1.99) 0.134 Rural residence 8.11 (2.62, 25.09) <0.001 0.73 (0.31, 1.74) 0.482 2.58 (0.82, 8.14) 0.105 Past 3 month alcohol/drug use 6.55 (3.67, 11.70) <0.001 4.32 (2.17, 8.61) <0.001 STI knowledge 1.13 (1.08, 1.19) <0.001 1.07 (1.02, 1.13) 0.006 Consistent condom use (n=54) Age 1.02 (0.90, 1.16) 0.734 0.99 (0.9, 1.1) 0.913 1.02 (0.90, 1.15) 0.762 Indigenous identity 0.74 (0.39, 1.44) 0.382 0.83 (0.68, 1.03) 0.092 0.85 (0.41, 1.73) 0.646 LGBQ+  identity 0.40 (0.16, 1.01) 0.052 0.5 (0.26, 0.94) 0.032 0.43 (0.19, 0.96) 0.040 Rural residence 0.41 (0.30, 0.55) <0.001 0.57 (0.47, 0.67) <0.001 0.65 (0.34, 1.27) 0.213 Past 3 month alcohol/drug use 0.41 (0.27, 0.65) <0.001 0.40 (0.27, 0.58) <0.001 STI knowledge 0.98 (0.91, 1.05) 0.527 0.98 (0.92, 1.04) 0.500 *Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, and rural place of residence.

41 (0.30, 0.55) <0.001 0.57 (0.47, 0.67) <0.001 0.65 (0.34, 1.27) 0.213 Past 3 month alcohol/drug use 0.41 (0.27, 0.65) <0.001 0.40 (0.27, 0.58) <0.001 STI knowledge 0.98 (0.91, 1.05) 0.527 0.98 (0.92, 1.04) 0.500 *Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, and rural place of residence. †Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, rural place of residence, alcohol use in the past 3 months, and STI knowledge. aRR, adjusted relative risk; LGBQ+, lesbian, gay, bisexual, queer and other sexually diverse; STI, sexually transmitted infection; uRR, unadjusted relative risk. Table 4 Socio-demographic factors associated with the odds of sexual activity and consistent condom use among adolescent cisgender boy (aged 13–18 years) participants in the Northwest Territories (NWT), Canada

aRR, adjusted relative risk; LGBQ+, lesbian, gay, bisexual, queer and other sexually diverse; STI, sexually transmitted infection; uRR, unadjusted relative risk. Table 4 Socio-demographic factors associated with the odds of sexual activity and consistent condom use among adolescent cisgender boy (aged 13–18 years) participants in the Northwest Territories (NWT), Canada Socio-demographic factor uRR P value aRR* P value aRR† P value Sexual activity (n=115) Age 1.62 (1.42, 1.84) <0.001 1.50 (1.30, 1.74) <0.001 1.51 (1.27, 1.79) <0.001 Indigenous identity 2.02 (0.98, 4.14) 0.056 1.08 (0.83, 1.41) 0.565 1.07 (0.54, 2.13) 0.851 LGBQ+ identity 1.13 (0.39, 3.27) 0.824 1.2 (0.69, 2.11) 0.520 1.48 (0.52, 4.21) 0.463 Rural residence 1.60 (0.83, 3.09) 0.158 0.34 (0.26, 0.46) <0.001 0.63 (0.34, 1.15) 0.131 Past 3 month alcohol/drug use 5.76 (3.24, 10.22) <0.001 – – 3.90 (2.12, 7.19) <0.001 STI knowledge 1.12 (1.02, 1.22) 0.012 – – 1.07 (0.99, 1.15) 0.097 Consistent condom use (n=54) Age 0.92 (0.75, 1.12) 0.389 0.99 (0.90, 1.10) 0.884 0.95 (0.77, 1.18) 0.660 Indigenous identity 1.47 (0.58, 3.77) 0.418 1.73 (1.16, 2.56) 0.007 1.93 (0.92, 4.02) 0.080 LGBQ+ identity 2.05 (1.51, 2.78) <0.001 2.39 (1.93, 2.95) <0.001 2.37 (1.44, 3.91) <0.001 Rural residence 0.83 (0.46, 1.52) 0.553 0.70 (0.56, 0.86) 0.001 0.70 (0.38, 1.30) 0.257 Past 3 month alcohol/drug use 1.31 (0.67, 2.58) 0.432 – – 1.35 (0.73, 2.50) 0.338 STI knowledge 1.00 (0.91, 1.10) 0.979 – – 1.01 (0.92, 1.10) 0.846 *Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, and rural place of residence.

0.46, 1.52) 0.553 0.70 (0.56, 0.86) 0.001 0.70 (0.38, 1.30) 0.257 Past 3 month alcohol/drug use 1.31 (0.67, 2.58) 0.432 – – 1.35 (0.73, 2.50) 0.338 STI knowledge 1.00 (0.91, 1.10) 0.979 – – 1.01 (0.92, 1.10) 0.846 *Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, and rural place of residence. †Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, rural place of residence, alcohol use in the past 3 months, and STI knowledge. aRR, adjusted relative risk; LGBQ +, lesbian, gay, bisexual, queer and other sexually diverse; STI, sexually transmitted infection; uRR, unadjusted relative risk.

0.46, 1.52) 0.553 0.70 (0.56, 0.86) 0.001 0.70 (0.38, 1.30) 0.257 Past 3 month alcohol/drug use 1.31 (0.67, 2.58) 0.432 – – 1.35 (0.73, 2.50) 0.338 STI knowledge 1.00 (0.91, 1.10) 0.979 – – 1.01 (0.92, 1.10) 0.846 *Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, and rural place of residence. †Multivariable models adjusted for age, Indigenous ancestry, LGBQ+ identity, rural place of residence, alcohol use in the past 3 months, and STI knowledge. aRR, adjusted relative risk; LGBQ +, lesbian, gay, bisexual, queer and other sexually diverse; STI, sexually transmitted infection; uRR, unadjusted relative risk. In multivariable models that included socio-demographic variables and modifiable factors such as past 3 month alcohol and drug use and STI knowledge, an increased likelihood of being sexual active was associated with higher age, alcohol or drug use, and STI knowledge among girls; and higher age, and alcohol or drug use among boys. Decreased likelihood of consistent condom use was associated with LGBQ+ identity and alcohol or drug use among girls, and increased likelihood of consistent condom use was associated with LGBQ+ identity among boys. No significant collinearity was observed between the adjusted included variables; all variance inflation factor values were <1.25. A cross-product interaction term was included for both adjusted models (data not shown) yielding P values <0.001, justifying the choice of stratifying the analysis by gender. The sensitivity analysis for clustering by community did not change the results substantially, suggesting the results are similar across communities in the study.

nteraction term was included for both adjusted models (data not shown) yielding P values <0.001, justifying the choice of stratifying the analysis by gender. The sensitivity analysis for clustering by community did not change the results substantially, suggesting the results are similar across communities in the study. Discussion A minority of participants – 22.9% of cisgender girls and 15.4% of cisgender boys – reported being sexually active. Of these sexually active participants, less than half reported past 3 month consistent condom use. Notably, past 3 month alcohol and drug use were significantly associated with being sexually active irrespective of gender, and with reduced likelihood of condom use among adolescent girls in the NWT. This study revealed important gender differences, with girls reporting higher STI knowledge, increased sexual activity, and higher alcohol and drug use than boys.

were significantly associated with being sexually active irrespective of gender, and with reduced likelihood of condom use among adolescent girls in the NWT. This study revealed important gender differences, with girls reporting higher STI knowledge, increased sexual activity, and higher alcohol and drug use than boys. Similar to prior studies, we found that alcohol and drug use were associated with adolescent sexual risk practices, including being sexually active13 and a lower likelihood of condom use.14 Alcohol and drug use among youth is a public health concern that extends beyond sexual health, as it can contribute to long-term health and substance use issues.30 To our knowledge this is among the first quantitative studies to explore factors associated with sexual activity and condom use among adolescents aged 13 years and older in circumpolar regions. Future sexual health programmes with NWT adolescents can work from a syndemics approach that conceptualises the interconnectedness of social and health disparities, including the complex linkages between substance use and sexual risk practices.31

dom use among adolescents aged 13 years and older in circumpolar regions. Future sexual health programmes with NWT adolescents can work from a syndemics approach that conceptualises the interconnectedness of social and health disparities, including the complex linkages between substance use and sexual risk practices.31 Girls were more likely than boys to identify as sexually diverse, and sexually diverse girls reported lower condom use than sexually diverse boys, in comparison with their heterosexual counterparts. Our findings that sexually diverse girls were 60% less likely to consistently use condoms than heterosexual girls corroborates prior research in the NWT that revealed a lack of LGBQ+ safer sex education in secondary schools21 and a general lack of knowledge of safer sex for sexually diverse women. These findings also align with research that situates the low uptake of safer sex practices, and corresponding STI risks, among sexually diverse women in larger social contexts that construct their STI risks as negligible.25 It is also plausible that sexually diverse girls use condoms less as they may have less male sex partners. This is an area for future research. Interestingly, LGBQ+ boys were 130% more likely to consistently use condoms as compared with heterosexual boys. Due to the high HIV and syphilis prevalence among gay and bisexual men in Canada10 there has been a focus on HIV prevention strategies, including condom use, targeting this group. Qualitative narratives from sexually diverse boys in the NWT reinforced the construction of HIV and STI as an issue impacting gay/bisexual men,22 so it is plausible that sexually diverse boys perceive themselves at higher risk for HIV/STI than heterosexual boys and therefore use condoms more frequently and consistently to manage this risk. Findings have implications for tailoring messages for condom use for both sexually diverse and heterosexual boys and girls to improve HIV and STI literacy and address barriers to consistent condom use. Future research can also consider tailoring interventions for adolescent girls and boys living in more rural locations who reported lower consistent condom use than those in Yellowknife.

both sexually diverse and heterosexual boys and girls to improve HIV and STI literacy and address barriers to consistent condom use. Future research can also consider tailoring interventions for adolescent girls and boys living in more rural locations who reported lower consistent condom use than those in Yellowknife. The study has several limitations. There is a likelihood of underreporting sexual activity due to social desirability bias and fear of judgement. Future studies could use tablet-based vs. paper-based surveys to minimise confidentiality concerns. We did not collect data on other mental health variables, such as depression, noted to be associated with both substance use and sexual risks among NWT youth.21 The few variables associated with consistent condom use in adjusted analyses suggests the need for qualitative research to further explore factors shaping condom use decision-making among adolescents in the NWT. Low numbers, particularly among consistent condom use, may mean that we are underpowered to detect statistically significant associations. However, the magnitude/direction of these associations and overall interpretations are unlikely to change with the achieved statistical power. While schools from 68% of regions with junior/secondary schools participated in the study, the representativeness of the sample is limited by the non-probability sampling of regions, and of classrooms within each participating school. Future research can expand to all NWT communities with junior/secondary schools and systematically select participating classrooms within schools to gather more representative data across NWT youth. Despite these limitations, our study provides unique insight into the ways in which sexual activity varies among this sample of NWT adolescents by gender, age, rurality, and alcohol and drug use.

hools and systematically select participating classrooms within schools to gather more representative data across NWT youth. Despite these limitations, our study provides unique insight into the ways in which sexual activity varies among this sample of NWT adolescents by gender, age, rurality, and alcohol and drug use. While STI knowledge was higher among sexually active vs non-sexually active girls, it was not associated with condom use among sexually active girls or boys. This suggests the need to go beyond providing STI information to promoting sexual health literacy and condom negotiation skills, and integrating a syndemics framework31 to explore the role that alcohol and drug use play in adolescents’ lives and sexual practices. These can inform culturally, gender and contextually tailored strategies to promote sexual health and well-being among adolescents in the NWT. Contributors: CHL conceptualised the paper and led the writing. CLL conceptualised the FOXY/SMASH programme with MacNeill, led research implementation and contributed to manuscript writing and interpretation. CHL and CLL are co-principal investigators of this study and NM a co-investigator. JF contributed substantially to manuscript writing. NM and KM led data collection and contributed to interpretation. AY led data analysis and contributed to manuscript writing.

on and contributed to manuscript writing and interpretation. CHL and CLL are co-principal investigators of this study and NM a co-investigator. JF contributed substantially to manuscript writing. NM and KM led data collection and contributed to interpretation. AY led data analysis and contributed to manuscript writing. Funding: The study was supported by funding from the Canadian Institutes of Health Research (CIHR) and the Social Sciences & Humanities Research Council of Canada (SSHRC). Funders played no role in study design or interpretation of findings. CHL was also supported by funding from the Ontario Ministry of Research & Innovation Early Researcher Award, Canada Research Chairs Program and the Canada Foundation for Innovation. Competing interests: None declared. Patient consent: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: The data are not available to be shared without obtaining ethical approval for changes in the data sharing process.

Key messages The need for second-trimester abortion services does not diminish, even with expansion of first-trimester abortion services. Poorer, younger women are more at risk for second-trimester abortion (versus first-trimester). States with health and fetal anomaly legal exceptions do not have higher utilisation of second-trimester abortion services, indicating under-utilisation of these legal exceptions. Introduction Overall, 10%–15% of abortions worldwide occur in the second trimester, yet they account for a disproportionate number of maternal deaths when provided under unsafe conditions.1–3 Reasons women need second-trimester abortion services vary, but include late diagnosis of fetal anomalies, delays in recognition of pregnancy, newly diagnosed or worsening maternal health conditions, and delays associated with access to care.2–4 Access to safe second-trimester abortion is an essential part of global efforts to reduce maternal morbidity and mortality, yet access is restricted in many parts of the world.2 Two-thirds of countries with laws regulating second-trimester abortion allow it only to save the woman’s life.3 While many countries do have exceptions permitting second-trimester abortion, such as the woman’s physical or mental health, rape, fetal anomalies, or socioeconomic reasons, access to services under these exceptions varies by country.5–7 The unmet need for second-trimester abortion remains unknown since many national vital statistics systems do not specifically capture incidence of second-trimester abortions.2

an’s physical or mental health, rape, fetal anomalies, or socioeconomic reasons, access to services under these exceptions varies by country.5–7 The unmet need for second-trimester abortion remains unknown since many national vital statistics systems do not specifically capture incidence of second-trimester abortions.2 In Mexico, abortion laws are determined at the state level. In Mexico City, one of Mexico’s 32 states, women have had access to first-trimester abortion in the private and public sectors since 2007.8 In Mexico’s 31 other states, abortion at all gestational ages remains restricted.7 Gestational age (GA) measures the age of a pregnancy and is reported in weeks. Rape is the only circumstance for which abortion is legally permitted at the federal level.9 Other grounds for legal access vary by state. For example, 14 states permit abortion when the health of a woman is at risk, while 16 permit abortion in cases of fetal anomalies.10 Previous research on abortion in Mexico has estimated incidence using indirect methods11–13 or focused on Mexico City’s pubic abortion programme.8 14 15 Data on in-facility second-trimester abortion is lacking; yet, it is critical to measure documented service provision in order to understand demand for abortion services and population health outcomes. We used Mexico’s hospital discharge register to describe utilisation of second-trimester spontaneous, induced, and post-abortion services, over time (2007–2015) and by state.

; yet, it is critical to measure documented service provision in order to understand demand for abortion services and population health outcomes. We used Mexico’s hospital discharge register to describe utilisation of second-trimester spontaneous, induced, and post-abortion services, over time (2007–2015) and by state. Methods Data and variables We used 2007–2015 data from Mexico’s hospital discharge system, SAEH (Subsistema Automatizado de Egresos Hospitalarios).16 Hospital discharge was defined as the exit of a patient from inpatient services (at least one night in the facility). All facilities included in this dataset are operated by national and state ministries of health (Secretaría de Salud, SSA). These facilities largely serve individuals without insurance or covered by Seguro Popular, a public insurance programme for the unemployed, self-employed, or those without access to employment-based social security.17 These government facilities accounted for an estimated 76% of the country’s total inpatient post-abortion cases in 2009.12

gely serve individuals without insurance or covered by Seguro Popular, a public insurance programme for the unemployed, self-employed, or those without access to employment-based social security.17 These government facilities accounted for an estimated 76% of the country’s total inpatient post-abortion cases in 2009.12 We built a dataset of obstetric events (births and abortions). Each discharge record includes up to six international Classification of Diagnosis Codes Version 10 (ICD-10) codes for reason for admission. We identified abortive events using ICD-10 codes O02-O08 (online supplementary web appendix table A1); we excluded ectopic and molar pregnancies with an abortive outcome, ICD-10 codes O00 and O01. We identified all types of abortion care through these hospital records; although the codes theoretically provide a way to differentiate between induced and spontaneous abortions, nearly 70% of second-trimester abortions were coded as 'O06 Unspecified abortion', making it difficult to differentiate between type of abortion. Therefore, we classified all abortive events (ICD codes O02-O08) as abortion for this analysis. In order to comprehensively capture second-trimester abortions, we also included women who had abortion codes for up to six discharge diagnosis codes (N=894; 0.6% of total sample (online supplementary web appendix table A2)). Additional abortions that may not have been identified using ICD-10 diagnostic codes were identified using a 'type of attention' checkbox in the SAEH records, which indicates whether an abortion or a delivery was performed (N=3974; 2.7% of total sample).

6% of total sample (online supplementary web appendix table A2)). Additional abortions that may not have been identified using ICD-10 diagnostic codes were identified using a 'type of attention' checkbox in the SAEH records, which indicates whether an abortion or a delivery was performed (N=3974; 2.7% of total sample). 10.1136/bmjsrh-2018-200300.supp1Supplementary data Among the records between 2007–2015, 22% (238 972) were missing GA (online supplementary web appendix table A3). Records missing GA were disproportionately from women in less marginalised urban areas, such as Mexico City. There were few abortions identified after 20 weeks; under international and national definitions, spontaneous or induced abortions after 20 weeks are classified as 'stillbirths'.18 The SAEH records also include the following variables: age, municipality and state of residence, parity, hospital of admission, and length of hospital stay. We defined reproductive age as women aged 15–44 years.

and national definitions, spontaneous or induced abortions after 20 weeks are classified as 'stillbirths'.18 The SAEH records also include the following variables: age, municipality and state of residence, parity, hospital of admission, and length of hospital stay. We defined reproductive age as women aged 15–44 years. We used 2010 population estimates of women aged 15–44 years from the Mexican government’s population agency, Consejo Nacional de Población (CONAPO),19 at the state and national level to calculate rates. We also used the 2010 municipality-level marginalisation index, a measure of community-level socioeconomic status.20 The index includes information about education, housing, income, and population density and is generated using principal components analysis.21 We classified municipalities into five groups: very low, low, medium, high, and very high marginalisation. We merged the marginalisation index into the individual-level discharge data using the residence municipality of the woman. We also categorised states into four groups based on the percentage of the population living in poverty in 2014 as reported by The National Council for the Evaluation of Social Development Policy (CONEVAL)21 and previous literature.22

to the individual-level discharge data using the residence municipality of the woman. We also categorised states into four groups based on the percentage of the population living in poverty in 2014 as reported by The National Council for the Evaluation of Social Development Policy (CONEVAL)21 and previous literature.22 We included information on state-level legal exceptions for abortion categorised as binary variables based on whether a state had exceptions for health of the woman or fetal anomaly in their penal code.10 Finally, we merged in hospital tier of specialisation (primary, secondary and tertiary).23 24 All data sources were publicly available and downloadable (online supplementary web appendix table A4). Analysis We calculated second-trimester abortion hospitalisation rates per year per 1000 women aged 15–44 years at the national and state levels. We mapped second-trimester abortion hospitalisation rates across Mexico by state, and by the presence of state-level health or fetal anomaly exception laws. We next examined changes in mean and median GA of second-trimester abortions over time. We examined trends nationally and in Mexico City versus other Mexican states. Since Mexico City has legal first-trimester abortion services, we hypothesised that trends in second-trimester GA would be lower in Mexico City due to availability of first-trimester abortion.

edian GA of second-trimester abortions over time. We examined trends nationally and in Mexico City versus other Mexican states. Since Mexico City has legal first-trimester abortion services, we hypothesised that trends in second-trimester GA would be lower in Mexico City due to availability of first-trimester abortion. Finally, we identified factors associated with utilisation of hospital-based second-trimester (vs first-trimester) abortion using a logistic regression model. We included age, parity, municipality-level marginalisation, state-level health and fetal anomaly exceptions, and hospital specialisation. We assessed model robustness by running models with different covariates, such as hospital bed size, women’s education, and maternal mortality rate; results were robust and we show only the final model. We estimated robust standard errors to account for non-independence of observations within hospitals using the 'cluster' option in STATA version 15 (StataCorp, College Station, TX, USA).25 Patient and public involvement There was no patient involvement in this secondary data analysis.

Finally, we identified factors associated with utilisation of hospital-based second-trimester (vs first-trimester) abortion using a logistic regression model. We included age, parity, municipality-level marginalisation, state-level health and fetal anomaly exceptions, and hospital specialisation. We assessed model robustness by running models with different covariates, such as hospital bed size, women’s education, and maternal mortality rate; results were robust and we show only the final model. We estimated robust standard errors to account for non-independence of observations within hospitals using the 'cluster' option in STATA version 15 (StataCorp, College Station, TX, USA).25 Patient and public involvement There was no patient involvement in this secondary data analysis. Results Among 1 083 803 induced and spontaneous abortions documented in the SAEH hospitalisation records between 2007 and 2015, we identified 145 956 (13.4%) second-trimester abortions. The majority of second-trimester abortions (70%) were classified as ICD-10 code O06 ('unspecified'). Nearly 60% of second-trimester abortions were among women aged ≤25 years, and only 2.4% occurred among women aged >40 years (table 1). Most (70%) women who had a second-trimester abortion were hospitalised for 1 day. Over 50% of reported second-trimester abortions occurred in the least marginalised municipalities (table 1). Overall, 20.2% of second-trimester abortions were performed in tertiary care hospitals, 78.4% in secondary care hospitals, and 1.4% in primary health facilities. However, in Mexico City, 51.2% of second-trimester abortions were performed in specialised hospitals, while in the poorer states of Guerrero, Oaxaca and Chiapas, only 1.6% of second-trimester abortions were performed in specialised hospitals (online supplementary web appendix table A5).

4% in primary health facilities. However, in Mexico City, 51.2% of second-trimester abortions were performed in specialised hospitals, while in the poorer states of Guerrero, Oaxaca and Chiapas, only 1.6% of second-trimester abortions were performed in specialised hospitals (online supplementary web appendix table A5). Table 1 Characteristics of women receiving in-hospital second-trimester abortions, Mexico 2007–2015 Characteristic Frequency (n) (N=145 956) Percentage (%) Age (years) <18 18 397 12.60 18–25 66 924 45.42 26–30 27 626 18.93 31–40 30 208 20.70 >40 3421 2.35 Length of hospital stay (days) 0 2576 1.76 1 102 609 70.30 2 27 124 18.58 3+ 13 647 9.35 Municipality-level marginalisation index* Very low 73 660 50.55 Low 23 719 16.28 Medium 33 250 22.82 High 8139 5.59 Very high 6947 4.77 *Very high marginalisation represents the poorest group, while very low marginalisation represents the wealthiest group.

days) 0 2576 1.76 1 102 609 70.30 2 27 124 18.58 3+ 13 647 9.35 Municipality-level marginalisation index* Very low 73 660 50.55 Low 23 719 16.28 Medium 33 250 22.82 High 8139 5.59 Very high 6947 4.77 *Very high marginalisation represents the poorest group, while very low marginalisation represents the wealthiest group. The annual nationwide hospitalisation rate of second-trimester abortion services remained stable, between 0.5 to 0.6 per 1000 women of reproductive age (15–44 years) over the study period. In our spatial analysis of second-trimester induced and spontaneous abortion hospitalisations, there was clustering of higher rates of hospitalisation in the central and southern parts of Mexico (figure 1). The states of Zacatecas and Durango had the highest rates of utilisation of hospital-based, second-trimester abortion services (1.1 and 0.89 per 1000 women, respectively) followed by predominantly southern states, including Tlaxcala and Guerrero. The northeastern states, including Coahuila, Nuevo León and Tamaulipas, had some of the lowest rates of utilisation of second-trimester abortion services. Figure 1 Second-trimester abortion utilisation rate per 1000 women aged 15–44 years in 2015 for individual Mexican states

The annual nationwide hospitalisation rate of second-trimester abortion services remained stable, between 0.5 to 0.6 per 1000 women of reproductive age (15–44 years) over the study period. In our spatial analysis of second-trimester induced and spontaneous abortion hospitalisations, there was clustering of higher rates of hospitalisation in the central and southern parts of Mexico (figure 1). The states of Zacatecas and Durango had the highest rates of utilisation of hospital-based, second-trimester abortion services (1.1 and 0.89 per 1000 women, respectively) followed by predominantly southern states, including Tlaxcala and Guerrero. The northeastern states, including Coahuila, Nuevo León and Tamaulipas, had some of the lowest rates of utilisation of second-trimester abortion services. Figure 1 Second-trimester abortion utilisation rate per 1000 women aged 15–44 years in 2015 for individual Mexican states No relationship between legal exceptions that permit abortion and documented utilisation of public, hospital-based abortion services was detected. In states with a health exception (figure 2), utilisation rates of hospital-based second-trimester abortion were 0.30 to 1.10 per 1000 women; rates were similar in states without a health exception (0.2 to 0.9). Among states with fetal anomaly exceptions, the hospitalisation rater varied between 0.21 and 0.77 per 1000 women, and did not differ based on region (figure 2). Figure 2 Second-trimester abortion utilisation rate per 1000 women aged 15–44 years in 2015 in Mexican states with health exceptions and fetal anomaly exceptions.

No relationship between legal exceptions that permit abortion and documented utilisation of public, hospital-based abortion services was detected. In states with a health exception (figure 2), utilisation rates of hospital-based second-trimester abortion were 0.30 to 1.10 per 1000 women; rates were similar in states without a health exception (0.2 to 0.9). Among states with fetal anomaly exceptions, the hospitalisation rater varied between 0.21 and 0.77 per 1000 women, and did not differ based on region (figure 2). Figure 2 Second-trimester abortion utilisation rate per 1000 women aged 15–44 years in 2015 in Mexican states with health exceptions and fetal anomaly exceptions. The majority (62%) of second-trimester abortions were performed between 13 and 16 weeks, 38% occurred between 17 and 20 weeks, and <1% from 21 to 24 weeks. In 2007, mean GA of second-trimester abortion patients in Mexico City was 18.11 weeks (95% CI 18.01 to 18.21), 1.82 weeks higher than the other Mexican states at 16.29 weeks (95% CI 16.25 to 16.34). In 2009, this difference narrowed to 0.12 weeks, and remained constant for the rest of the study period. However, overall mean GA remained marginally higher in Mexico City (16.6) compared with the other states (15.9) over the study period (p<0.0001).

than the other Mexican states at 16.29 weeks (95% CI 16.25 to 16.34). In 2009, this difference narrowed to 0.12 weeks, and remained constant for the rest of the study period. However, overall mean GA remained marginally higher in Mexico City (16.6) compared with the other states (15.9) over the study period (p<0.0001). In multivariable analyses, adolescents were more likely than older women to present for abortion services in the second trimester compared with the first trimester (odds ratio (OR) 1.07, 95% CI 1.04 to 1.11) (figure 3). Women living in municipalities with high levels of marginalisation had higher odds of utilising abortion services in their second trimester versus first trimester compared with women living in municipalities with lower marginalisation (OR 1.43, 95% CI 1.18 to 1.73). Living in a state with a health or fetal anomaly exception was not associated with utilisation of second-trimester versus first-trimester services. Figure 3 Factors associated with second-trimester abortion versus first-trimester abortion care. OR, odds ratio; Ref., reference.

In multivariable analyses, adolescents were more likely than older women to present for abortion services in the second trimester compared with the first trimester (odds ratio (OR) 1.07, 95% CI 1.04 to 1.11) (figure 3). Women living in municipalities with high levels of marginalisation had higher odds of utilising abortion services in their second trimester versus first trimester compared with women living in municipalities with lower marginalisation (OR 1.43, 95% CI 1.18 to 1.73). Living in a state with a health or fetal anomaly exception was not associated with utilisation of second-trimester versus first-trimester services. Figure 3 Factors associated with second-trimester abortion versus first-trimester abortion care. OR, odds ratio; Ref., reference. Discussion In this study, we present novel data on the utilisation of second-trimester abortion services in the public sector in Mexico. We found that women seek care in health facilities for second-trimester abortion care, including spontaneous, induced, and post-abortion services, even in places where abortion is highly restricted, although not at the rates that might be expected.2 Although second-trimester induced abortion is largely illegal throughout Mexico, some women are able to access abortion services under the available exceptions; many others present for and receive post-abortion care.

even in places where abortion is highly restricted, although not at the rates that might be expected.2 Although second-trimester induced abortion is largely illegal throughout Mexico, some women are able to access abortion services under the available exceptions; many others present for and receive post-abortion care. We found the average hospitalisation rate for second-trimester abortions was 0.57 per 1000 women, and 13.4% of total hospital-based abortions occurred in the second trimester. Existing estimates of utilisation of abortion services in Mexico range from 7.3 per 1000 women26 to 38 per 1000 women aged 15–44 years.12 Under the assumption second-trimester induced and spontaneous abortions account for ~15% of total abortions,2 27 we might expect the utilisation of second-trimester abortion to range from 1.1 to 5.7 per 1000 women. We found consistent utilisation of second-trimester abortion services during the study period, including in Mexico City, even following liberalisation of first-trimester abortion in the capital in 2007. However, we did detect a slight decline in GA at the time of second-trimester abortion among women in Mexico City after 2007, which may be due to improved access to first-trimester abortion. Our results demonstrate the need for second-trimester abortion services remains present, even with expansion of first-trimester abortion services.

etect a slight decline in GA at the time of second-trimester abortion among women in Mexico City after 2007, which may be due to improved access to first-trimester abortion. Our results demonstrate the need for second-trimester abortion services remains present, even with expansion of first-trimester abortion services. Furthermore, we found health and fetal anomaly exceptions in the state laws were not associated with higher utilisation of second-trimester abortion services. The potential under-utilisation of second-trimester abortion services may be due to poor implementation of all legal exceptions.7 For example, while Jalisco passed a law in 2009 that mandated institutions to provide abortion in cases of rape, there have only been 16 cases documented.28 This type of low utilisation was reflected in our study; women living in states with legal exceptions for health or fetal anomaly were no more likely to utilise second-trimester services than women living in states without these exceptions. There remains widespread under-utilisation of legal exceptions, likely due to lack of information and training among providers, poor dissemination of information to women, and a continued environment of criminalisation.7 29 30

ly to utilise second-trimester services than women living in states without these exceptions. There remains widespread under-utilisation of legal exceptions, likely due to lack of information and training among providers, poor dissemination of information to women, and a continued environment of criminalisation.7 29 30 Our results support other evidence that Mexico’s most vulnerable women are often disproportionately affected by obstacles to obstetric or prenatal care.14 This finding has been reflected in other settings, where younger, poorer and more disenfranchised women were more likely to seek second-trimester abortions.4 14 31–33 In our study, women in less marginalised municipalities comprised half of our sample of second-trimester abortions, but had lower odds of presenting for second-trimester (versus first-trimester) abortion services. Adolescents had higher odds of presenting for second-trimester (versus first-trimester) abortion services when compared with older women.

ginalised municipalities comprised half of our sample of second-trimester abortions, but had lower odds of presenting for second-trimester (versus first-trimester) abortion services. Adolescents had higher odds of presenting for second-trimester (versus first-trimester) abortion services when compared with older women. Our study has limitations; Mexico is among a few low- and middle-income countries with robust health information systems, but secondary health services data are limited. We included all abortions – spontaneous and induced – and were not able to differentiate between them. Second, this is not a study of second-trimester abortion incidence, but of documented utilisation in the public sector. Our results may not be generalisable to private or employment-based health sectors or abortions that occur out-of-facility.34–36 The SAEH data contain limited individual-level sociodemographic information; we leveraged other data sources with facility, municipality and state-level characteristics to partially address this limitation of the data. Fourth, in our comparison of women seeking first-trimester versus second-trimester abortions, our sample of first-trimester abortions is limited by the fact that it does not include any outpatient services from Mexico City’s legal first-trimester abortion programme. However, we found similar results when running our model with and without data from Mexico City. In addition, 22% of records were missing GA; improvements in the documentation of GA would increase our understanding of the timing of abortion in Mexico.

ervices from Mexico City’s legal first-trimester abortion programme. However, we found similar results when running our model with and without data from Mexico City. In addition, 22% of records were missing GA; improvements in the documentation of GA would increase our understanding of the timing of abortion in Mexico. In our study, we leveraged objective clinical data to report on in-facility second-trimester spontaneous and induced abortion across the public sector in a low- and middle-income country where abortion is restricted. We found the average hospitalisation rate for second-trimester abortions was 0.57 per 1000 women, and 13.4% of total hospital-based abortions occurred in the second trimester. We highlight there is a need for second-trimester abortion services in Mexico. To improve health outcomes for Mexican women, especially the most vulnerable, access to safe second-trimester abortion services must be ensured through the implementation of current legal exceptions and renewed attention to provider training. Funding: BGD was supported by Society of Family Planning awards SFPRF11-02, SFPRF10-II2-2, R01HS025155 (Cottrell, PI), and grant number K12HS022981 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: BGD was supported by Society of Family Planning awards SFPRF11-02, SFPRF10-II2-2, R01HS025155 (Cottrell, PI), and grant number K12HS022981 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data availability statement: All data relevant to the study are included in the article or uploaded as supplementary information.

Key messages Four in ten women of childbearing age planned to have a second child during the first 2 years of China’s universal two-child policy. Economic, childrearing, and health barriers affected fertility intent. The age range of 25-39 years, living in rural areas, being a farmer, possessing a higher household income, having a history of abortion, and already having one child were associated with intent for a second child. Introduction The total fertility rate in China declined from an estimated 2.9 births per woman in 1979 to between 1.5 and 1.7 births per woman currently.1 The government of China announced the one-child policy in 1979, considering population containment as essential to elevating China out of the severe poverty caused by decades of economic mismanagement.2 Each couple in China was thereby allowed to give birth to one child. After three decades of the one-child policy, China had to face the new challenges of accelerated population ageing, a skewed sex ratio, and the decline in the working-age population.1 2 In order to optimise population structure and cope with accelerated population ageing, the Chinese government changed its policy in 2013, allowing couples (in which at least one of the marital partners was an only child) to have two children.1 In October 2015, the Chinese government announced that all couples would be allowed to have two children, which marked the official end of the one-child policy and the beginning of the new, universal two-child policy.1

wing couples (in which at least one of the marital partners was an only child) to have two children.1 In October 2015, the Chinese government announced that all couples would be allowed to have two children, which marked the official end of the one-child policy and the beginning of the new, universal two-child policy.1 This political change was presumably designed to stimulate flagging birth rates in China, which threaten to undermine the economic growth tthat the country has enjoyed in recent decades.3 However, only 1.45 million (13.2%) of the 11 million eligible couples applied for permission to have a second child by May 2015.1 4 Fertility intent is regarded as a predictor of fertility behaviour;5 and in addition to policy, intent to procreate is also influenced by individual, social, economic and cultural factors.4 6 Several investigators have reported on the intent to have a second child in China (which varies from 13.2% to 69.3%) and its associated factors, including age, occupation, educational level, and gender of the first child.4–8 However, these studies were principally conducted in particular populations or settings, such as a rural population, women attending outpatient gynaecology clinics, or women from one specific province. Studies that focus on fertility intent among Chinese women in the general population at the national level are still limited. Thus, we conducted a cross-sectional study to assess the intention to have a second child among Chinese women of childbearing age after the enactment of China’s new universal two-child policy, so as to determine the short-term effects of this policy on fertility plans and to provide suggestions on policymaking.

imited. Thus, we conducted a cross-sectional study to assess the intention to have a second child among Chinese women of childbearing age after the enactment of China’s new universal two-child policy, so as to determine the short-term effects of this policy on fertility plans and to provide suggestions on policymaking. Methods Study design and study population We conducted a cross-sectional study among Chinese women of childbearing age between 2016 and 2017. Participants were selected using a two-stage, stratified sampling method. First, according to the population proportions of the Eastern, Central and Western regions (approximately 5:4:2) in China, we randomly selected provinces in each region (five in the Eastern, 4 in the Central and two in the Western regions). Eleven out of 31 provinces were selected, including five provinces/cities (Beijing, Shanghai, Zhejiang, Fujian and Guangdong) in the Eastern region, four provinces (Henan, Hfourbei, Jiangxi and Hunan) in the Central region and two provinces (Sichuan and Shaanxi) in the Western region. We then selected 1200 participants randomly from each province via an online survey platform (Wen Juan Xing) that was based on their registered information by province. The online survey platform (Wen Juan Xing) was established in 2005 and has 20.29 million Chinese internet users. The inclusion criteria were women 18–49 years of age from 11 provinces in China between 2016 and 2017. The exclusion criteria were women who declined to participate in the survey. A total of 13 200 women aged 18–49 years were selected via the online survey platform and invited to participate in this survey by web, email or text message. Web-based questionnaires were then sent to the selected individuals who agreed to participate in the survey, and participants completed the questionnaires online by mobile phone or computer. Finally, 11 991 of 13 200 women from 11 provinces responded and completed the questionnaires, a response rate of 90.8%. This study was approved by the Institutional Review Board of the Chinese Association of Maternal and Child Health Studies.

participants completed the questionnaires online by mobile phone or computer. Finally, 11 991 of 13 200 women from 11 provinces responded and completed the questionnaires, a response rate of 90.8%. This study was approved by the Institutional Review Board of the Chinese Association of Maternal and Child Health Studies. Data collection We used a self-designed questionnaire to collect information from the participants, covering sociodemographic information (age, ethnicity, educational level, occupation, marriage, residence, and annual household income) and reproductive status (current number of children, gender of children, and history of abortion). Fertility intent was investigated with the questions “Do you want to have more children in the future?" (answer Yes or No) and “If you want to have more children, how many more children do you want to have?”. We also investigated gender preferences and barriers to having more children, including economic, childrearing, health, educational, and other barriers. Economic barriers referred to the cost of raising more children. Childrearing barriers referred to the shortage of time or labour in raising more children. Health barriers referred to reasons related to personal health status. Educational barriers referred to the stress of fulfilling the educational needs of children. In the present study, our primary outcome was the prevalence of the intent to have a second child, which was defined as the proportion of women who had the intent to have two or more children among all the women surveyed.

atus. Educational barriers referred to the stress of fulfilling the educational needs of children. In the present study, our primary outcome was the prevalence of the intent to have a second child, which was defined as the proportion of women who had the intent to have two or more children among all the women surveyed. Patient and public involvement Patients and members of the public were not involved in the development of the study. Statistical analysis Categorical variables are shown as proportions. We calculated the prevalence of intent to have a second child and its 95% CI and used the χ² test to compare the rates of second-child intent in different characteristic groups. Univariate and multivariate logistic regression models were used to examine the relative factors associated with intent to have a second child. All factors (independent variables) in the univariate analysis were added to the multivariate model; these were age group, year, ethnicity, region, area of residence, educational level, marital status, occupation, annual household income, history of abortion, having a first child, and barriers. The dependent variable was the intent to have a second child or not. Adjusted odds ratios (aORs) and their 95% CIs were calculated by a backward stepwise method. We performed sensitivity analyses that adjusted for a continuous age variable instead of a five-level age-group variable. We analysed subgroups that were stratified by parity. Two-sided p values <0.05 were considered statistically significant. All analyses were performed with SPSS version 18.0.

ackward stepwise method. We performed sensitivity analyses that adjusted for a continuous age variable instead of a five-level age-group variable. We analysed subgroups that were stratified by parity. Two-sided p values <0.05 were considered statistically significant. All analyses were performed with SPSS version 18.0. Results Characteristics and fertility intent of the study population Of the 11 991 women included in the study, 50.8% were 30 years of age or older, 16.0% had a history of abortion, 43.9% had no children, and 43.7% had one child. Approximately 1/3 (1459, 32.9%) of the unmarried women and 2855 (37.8%) of the married women intended to conceive again in the future (online supplementary table 1). Fertility intent was higher in 2017 than in 2016 among unmarried women, married women without any children, and married women with two or more children (all p<0.05). 10.1136/bmjsrh-2018-200197.supp1Supplementary data

Results Characteristics and fertility intent of the study population Of the 11 991 women included in the study, 50.8% were 30 years of age or older, 16.0% had a history of abortion, 43.9% had no children, and 43.7% had one child. Approximately 1/3 (1459, 32.9%) of the unmarried women and 2855 (37.8%) of the married women intended to conceive again in the future (online supplementary table 1). Fertility intent was higher in 2017 than in 2016 among unmarried women, married women without any children, and married women with two or more children (all p<0.05). 10.1136/bmjsrh-2018-200197.supp1Supplementary data Intent to have a second child and related barriers Of the 11 991 women, the overall rate of second-child intent was 39.4% (95% CI 38.5% to 40.3%; table 1). Among women with one child, the second-child intent rate was 39.5% (95% CI 38.1% to 40.8%). Among women without second-child intent, economic barriers constituted the largest factor (47.7%) affecting fertility intent, followed by childrearing barriers (17.1%) and health barriers (15.6%). The proportion of health barriers increased from 14.5% in 2016 to 16.1% in 2017 (table 2), and women over 35 years of age noted greater health barriers than did women under 35 years, but fewer economic barriers than women aged <35 years (all p<0.001). Table 1 Factors associated with the intent to have a second child among 11 991 women of childbearing age

Intent to have a second child and related barriers Of the 11 991 women, the overall rate of second-child intent was 39.4% (95% CI 38.5% to 40.3%; table 1). Among women with one child, the second-child intent rate was 39.5% (95% CI 38.1% to 40.8%). Among women without second-child intent, economic barriers constituted the largest factor (47.7%) affecting fertility intent, followed by childrearing barriers (17.1%) and health barriers (15.6%). The proportion of health barriers increased from 14.5% in 2016 to 16.1% in 2017 (table 2), and women over 35 years of age noted greater health barriers than did women under 35 years, but fewer economic barriers than women aged <35 years (all p<0.001). Table 1 Factors associated with the intent to have a second child among 11 991 women of childbearing age Factor n (%) Intent Univariate model Multivariate model To have a second child (n (%)) cORs (95% CI) P value aORs (95% CI) P value Total 11 991 4725 (39.4) Year 2016 3647 (30.4) 1316 (36.1) 1 1 2017 8344 (69.6) 3409 (40.9) 1.22 (1.13 to 1.33) <0.001 1.31 (1.17 to 1.48) <0.001 Age (years) <25 2748 (22.9) 556 (20.2) 1 1 25–29 3151 (26.3) 1308 (41.5) 2.80 (2.49 to 3.15) <0.001 1.28 (1.10 to 1.50) 0.002 30–34 2588 (21.6) 1390 (53.7) 4.57 (4.05 to 5.16) <0.001 1.30 (1.09 to 1.55) 0.004 35–39 1812 (15.1) 900 (49.7) 3.89 (3.41 to 4.43) <0.001 1.29 (1.15 to 1.44) 0.016 ≥40 1692 (14.1) 571 (33.7) 2.01 (1.75 to 2.30) <0.001 0.99 (0.82 to 1.19) 0.902 Region Eastern 8422 (70.2) 3275 (38.9) 1.23 (1.07 to 1.41) 0.004 1.28 (1.10 to 1.49) 0.001 Central 2582 (21.5) 1113 (43.1) 1.46 (1.25 to 1.70) <0.001 1.50 (1.27 to 1.78) <0.001 Western 987 (8.2) 337 (34.1) 1 1 Ethnicity Han 11 651 (97.2) 4595 (39.4) 1 Other 340 (2.8) 130 (38.2) 0.95 (0.76 to 1.19) 0.654 Residence Rural 4595 (38.3) 1974 (43.0) 1.27 (1.18 to 1.37) <0.001 1.62 (1.47 to 1.78) <0.001 Urban 7396 (61.7) 2751 (37.2) 1 1 Educational level Primary school or below 109 (0.9) 36 (33.0) 1 Junior high school 1632 (13.6) 746 (45.7) 1.71 (1.13 to 2.57) 0.011 Senior high school 1088 (9.1) 493 (45.3) 1.68 (1.11 to 2.55) 0.015 College or higher 9162 (76.4) 3450 (37.7) 1.22 (0.82 to 1.83) 0.322 Marriage Single 4441 (37.0) 941 (21.2) 1 1 Married 7550 (63.0) 3784 (50.1) 3.74 (3.43 to 4.07) <0.001 1.74 (1.50 to 2.02) <0.001 Occupation Factory worker 1614 (13.5) 710 (44.0) 1 1 Farmer 300 (2.5) 176 (58.7) 1.81 (1.41 to 2.32) <0.001 1.43 (1.09 to 1.88) 0.010 Manager 3323 (27.7) 1432 (43.1) 0.96 (0.86 to 1.09) 0.551 1.07 (0.94 to 1.22) 0.323 Rural migrant worker 1392 (11.6) 681 (48.9) 1.22 (1.06 to 1.41) 0.007 1.11 (0.95 to 1.30) 0.200 Clerk 1714 (14.3) 594 (34.7) 0.68 (0.59 to 0.78) <0.001 0.86 (0.71 to 1.03) 0.103 Student 1592 (13.3) 240 (15.1) 0.23 (0.19 to 0.27) <0.001 0.63 (0.51 to 0.79) <0.001 Other 2056 (17.1) 892 (43.4) 0.98 (0.86 to 1.11) 0.714 1.02 (0.88 to 1.18) 0.806 Annual household income (RMB) <30 000 3650 (30.4) 1151 (31.5) 1 1 30 000–80 000 4406 (36.7) 1847 (41.9) 1.57 (1.43 to 1.72) <0.001 0.99 (0.88 to 1.10) 0.798 >80

1592 (13.3) 240 (15.1) 0.23 (0.19 to 0.27) <0.001 0.63 (0.51 to 0.79) <0.001 Other 2056 (17.1) 892 (43.4) 0.98 (0.86 to 1.11) 0.714 1.02 (0.88 to 1.18) 0.806 Annual household income (RMB) <30 000 3650 (30.4) 1151 (31.5) 1 1 30 000–80 000 4406 (36.7) 1847 (41.9) 1.57 (1.43 to 1.72) <0.001 0.99 (0.88 to 1.10) 0.798 >80  000 3935 (32.8) 1727 (43.9) 1.70 (1.55 to 1.87) <0.001 1.13 (1.00 to 1.28) 0.047 History of abortion No 10 073 (84.0) 3758 (37.3) 1 1 Yes 1918 (16.0) 967 (50.4) 1.71 (1.55 to 1.88) <0.001 1.18 (1.05 to 1.31) 0.004 First child Boy 3639 (30.3) 1811 (49.8) 3.47 (3.16 to 3.80) <0.001 2.70 (2.33 to 3.13) <0.001 Girl 3091 (25.8) 1745 (56.5) 4.54 (4.12 to 5.00) <0.001 3.65 (3.14 to 4.24) <0.001 None 5261 (43.9) 1169 (22.2) 1 1 Barriers to having more children Economic barrier 5576 (46.5) 2110 (37.8) 0.65 (0.56 to 0.76) <0.001 0.37 (0.31 to 0.43) <0.001 Health barrier 1862 (15.5) 731 (39.3) 0.69 (0.59 to 0.82) <0.001 0.39 (0.32 to 0.47) <0.001 Childrearing barrier 2040 (17.0) 794 (38.9) 0.68 (0.58 to 0.81) <0.001 0.32 (0.27 to 0.39) <0.001 Educational barrier 1752 (14.6) 723 (41.3) 0.75 (0.64 to 0.90) <0.001 0.38 (0.32 to 0.46) <0.001 Other 761 (6.3) 367 (48.2) 1 1 aOR, adjusted odds ratio; cOR, crude odds ratio;RMB, renminbi. Table 2 Barriers to the intent to have a second child among women without an intent for a second child

000 3935 (32.8) 1727 (43.9) 1.70 (1.55 to 1.87) <0.001 1.13 (1.00 to 1.28) 0.047 History of abortion No 10 073 (84.0) 3758 (37.3) 1 1 Yes 1918 (16.0) 967 (50.4) 1.71 (1.55 to 1.88) <0.001 1.18 (1.05 to 1.31) 0.004 First child Boy 3639 (30.3) 1811 (49.8) 3.47 (3.16 to 3.80) <0.001 2.70 (2.33 to 3.13) <0.001 Girl 3091 (25.8) 1745 (56.5) 4.54 (4.12 to 5.00) <0.001 3.65 (3.14 to 4.24) <0.001 None 5261 (43.9) 1169 (22.2) 1 1 Barriers to having more children Economic barrier 5576 (46.5) 2110 (37.8) 0.65 (0.56 to 0.76) <0.001 0.37 (0.31 to 0.43) <0.001 Health barrier 1862 (15.5) 731 (39.3) 0.69 (0.59 to 0.82) <0.001 0.39 (0.32 to 0.47) <0.001 Childrearing barrier 2040 (17.0) 794 (38.9) 0.68 (0.58 to 0.81) <0.001 0.32 (0.27 to 0.39) <0.001 Educational barrier 1752 (14.6) 723 (41.3) 0.75 (0.64 to 0.90) <0.001 0.38 (0.32 to 0.46) <0.001 Other 761 (6.3) 367 (48.2) 1 1 aOR, adjusted odds ratio; cOR, crude odds ratio;RMB, renminbi. Table 2 Barriers to the intent to have a second child among women without an intent for a second child Factor* Total 2016 2017 n % n % n % Economic barrier 3466 47.7 1138 48.8 2328 47.2 Childrearing barrier 1246 17.1 436 18.7 810 16.4 Health barrier 1131 15.6 337 14.5 794 16.1 Educational barrier 1029 14.2 326 14.0 703 14.2 Other 394 5.4 94 4.0 300 6.1 *χ² test, p<0.001.

Table 2 Barriers to the intent to have a second child among women without an intent for a second child Factor* Total 2016 2017 n % n % n % Economic barrier 3466 47.7 1138 48.8 2328 47.2 Childrearing barrier 1246 17.1 436 18.7 810 16.4 Health barrier 1131 15.6 337 14.5 794 16.1 Educational barrier 1029 14.2 326 14.0 703 14.2 Other 394 5.4 94 4.0 300 6.1 *χ² test, p<0.001. Factors related to fertility intent for a second child The intent for a second child in 2017 (40.9%) was higher than in 2016 (36.1%, p<0.001), and its prevalence varied from a nadir of 20.2% in women aged <25 years to a peak (53.7%) in women aged 30–34 years and then declined (table 1). The trends were similar between 2016 and 2017 (figure 1). Intent for a second child was higher in Central China (43.1%) relative to Eastern (38.9%) and Western China (34.1%, p<0.001), higher in rural areas (43.0%) compared with urban areas (37.2%, p<0.001), and higher in women with a history of abortion (50.4%) relative to women without such a history (37.3%, p<0.001). The rate increased commensurately with increasing annual household income (χtrend 2=118.82, p<0.001). Women whose first child was a girl also displayed a greater intent to have a second child (56.5%) compared with women whose first child was a boy (49.8%, p<0.001). Figure 1 Age-specified prevalence of the intent for a second child among Chinese women in 2016 and 2017.

Factors related to fertility intent for a second child The intent for a second child in 2017 (40.9%) was higher than in 2016 (36.1%, p<0.001), and its prevalence varied from a nadir of 20.2% in women aged <25 years to a peak (53.7%) in women aged 30–34 years and then declined (table 1). The trends were similar between 2016 and 2017 (figure 1). Intent for a second child was higher in Central China (43.1%) relative to Eastern (38.9%) and Western China (34.1%, p<0.001), higher in rural areas (43.0%) compared with urban areas (37.2%, p<0.001), and higher in women with a history of abortion (50.4%) relative to women without such a history (37.3%, p<0.001). The rate increased commensurately with increasing annual household income (χtrend 2=118.82, p<0.001). Women whose first child was a girl also displayed a greater intent to have a second child (56.5%) compared with women whose first child was a boy (49.8%, p<0.001). Figure 1 Age-specified prevalence of the intent for a second child among Chinese women in 2016 and 2017. Multivariate models showed that the age range of 25–39 years (aOR 1.28–1.30), living in Eastern (aOR 1.28, 95% CI 1.10 to 1.50) or Central China (aOR 1.52, 95% CI 1.27 to 1.78), living in a rural area (aOR 1.62; 95% CI 1.47 to 1.78), being a farmer (aOR 1.43, 95% CI 1.09 to 1.88), having a higher household income (aOR 1.13, 95% CI 1.00 to 1.28), having a history of abortion (aOR 1.18, 95% CI 1.05 to 1.31), and already having one child (aOR 3.65 for girls, aOR 2.70 for boys) were associated with a higher intent to have a second child (all p<0.05; table 1). In contrast, economic, health, childrearing, and educational barriers were associated with a reduced intent to have a second child (aOR 0.65–0.75). In the sensitivity analyses, the associations between factors and the intent to have a second child did not change appreciably after adjusting for the continuous variable of age instead of a five-level age group variable. In the subgroup analyses, factors associated with the intent to have a second child were stable (table 3).

ensitivity analyses, the associations between factors and the intent to have a second child did not change appreciably after adjusting for the continuous variable of age instead of a five-level age group variable. In the subgroup analyses, factors associated with the intent to have a second child were stable (table 3). Table 3 Factors associated with the intent to have a second child in the multivariate models stratified by parity

ensitivity analyses, the associations between factors and the intent to have a second child did not change appreciably after adjusting for the continuous variable of age instead of a five-level age group variable. In the subgroup analyses, factors associated with the intent to have a second child were stable (table 3). Table 3 Factors associated with the intent to have a second child in the multivariate models stratified by parity Characteristics Multivariate models Parity=0 Parity≥1 aOR (95% CI) P value aOR (95% CI) P value Year 2016 1 1 2017 1.26 (1.03 to 1.54) 0.023 1.36 (1.18 to 1.57) <0.001 Age (years) <25 1 1 25–29 1.26 (1.06 to 1.50) 0.008 0.74 (0.48 to 1.13) 0.163 30–34 1.16 (0.90 to 1.49) 0.263 2.22 (1.41 to 3.45) 0.001 35–39 0.70 (0.47 to 1.04) 0.075 1.96 (1.25 to 3.13) 0.003 ≥40 0.50 (0.30 to 0.84) 0.009 1.33 (1.00 to 1.79) 0.050 Region Eastern 1.22 (1.06 to 1.42) 0.007 1.31 (1.16 to 1.49) <0.001 Central 1.28 (0.98 to 1.67) 0.070 1.29 (1.07 to 1.55) 0.008 Western Ethnicity Han 1 1 Other 0.94 (0.67 to 1.34) 0.743 0.73 (0.51 to 1.04) 0.080 Residence Rural 1.22 (1.06 to 1.41) 0.007 1.99 (2.27 to 1.74) <0.001 Urban 1 1 Educational level Primary school or below 1 1 Junior high school 1.20 (0.78 to 1.85) 0.397 1.20 (0.78 to 1.84) 0.415 Senior high school 1.30 (0.84 to 2.01) 0.246 1.29 (0.83 to 2.00) 0.252 College or higher 0.93 (0.61 to 1.41) 0.723 0.91 (0.60 to 1.39) 0.674 Marriage Single 1 1 Married 1.43 (1.20 to 1.71) <0.001 1.56 (1.17 to 2.08) 0.002 Occupation Factory worker 1 1 Farmer 1.10 (0.60 to 2.01) 0.762 1.36 (0.98 to 1.89) 0.068 Manager 1.31 (1.01 to 1.70) 0.039 1.03 (0.88 to 1.21) 0.734 Rural migrant worker 1.04 (0.77 to 1.40) 0.803 1.07 (0.88 to 1.30) 0.487 Clerk 0.95 (0.68 to 1.34) 0.783 0.85 (0.68 to 1.07) 0.165 Student 0.75 (0.56 to 1.00) 0.050 0.21 (0.04 to 1.18) 0.076 Other 1.22 (0.94 to 1.58) 0.134 1.02 (0.86 to 1.22) 0.809 Annual household income (RMB) <30 000 1 1 30 000–80 000 1.01 (0.85 to 1.21) 0.888 0.95 (0.82 to 1.10) 0.485 >80 000 1.17 (0.95 to 1.44) 0.140 1.12 (0.95 to 1.31) 0.171 History of abortion No 1 1 Yes 1.81 (1.33 to 2.45) <0.001 1.11 (0.99 to 1.25) 0.076 Barriers to having more children Economic barrier 0.68 (0.50 to 0.93) 0.016 0.69 (0.58 to 0.82) <0.001 Health barrier 0.96 (0.78 to 1.19) 0.714 0.83 (0.72 to 0.95) 0.007 Childrearing barrier 0.83 (0.68 to 0.99) 0.049 0.97 (0.84 to 1.14) 0.727 Education barrier 0.92 (0.75 to 1.13) 0.431 0.68 (0.50 to 0.93) 0.016 Other 1 1 aOR, adjusted odds ratio; RMB, renminbi.

arrier 0.68 (0.50 to 0.93) 0.016 0.69 (0.58 to 0.82) <0.001 Health barrier 0.96 (0.78 to 1.19) 0.714 0.83 (0.72 to 0.95) 0.007 Childrearing barrier 0.83 (0.68 to 0.99) 0.049 0.97 (0.84 to 1.14) 0.727 Education barrier 0.92 (0.75 to 1.13) 0.431 0.68 (0.50 to 0.93) 0.016 Other 1 1 aOR, adjusted odds ratio; RMB, renminbi. Discussion In this cross-sectional study we found that the rate of intent for a second child was 39.4% of 11 991 Chinese women of childbearing age. This was much lower than the rate (61.25%) recorded under the one-child policy6 and was similar to rates reported by others during the two-child policy.9 10 According to the report by the China Women’s Federation in 2017, only 20.5% of parents wanted to have a second child.9 However, Zhao and colleagues,10 in a hospital-based cross-sectional study of 814 nulliparous pregnant women in 2017, reported that 51.2% of women intended to have a second child. The discrepancy between these studies might be related to the different characteristics of the study population, as we found that economic, childrearing, and health factors were the major barriers to fertility intent. Liu and colleagues11 also reported that couples avoided a second conception because of the cost of childrearing and educational and occupational pressures. Most parents regarded a family’s socioeconomic status as a key factor in raising a second child.9 We found that young women considered childraising barriers to be the second most important factor, while women of advanced age considered health to be the second most important factor. It is worth noting that the proportion of health barriers increased from 2016 to 2017, which indicated that parents worried more about maternal and perinatal safety. The quality of women’s eggs declines dramatically with increasing age, leading to an increased risk of pregnancy-related complications for women of advanced age.12 Thus, women of advanced age might give more consideration to health barriers than do younger women.

arents worried more about maternal and perinatal safety. The quality of women’s eggs declines dramatically with increasing age, leading to an increased risk of pregnancy-related complications for women of advanced age.12 Thus, women of advanced age might give more consideration to health barriers than do younger women. We found that fertility intent was associated with socioeconomic status, similar to results from previous studies.6 13 14 In our study, the prevalence of intent to have a second child peaked in the 30–34 years age group and gradually declined with increasing age. This trend was consistent with data from North American and European countries13 and might relate to the awareness of women of advanced maternal age (over 35 years) of the higher risks for adverse pregnancy outcomes. Previous studies have shown that women of advanced age exhibited a significantly increased risk of adverse maternal and infant outcomes.15–18 Considerable attention will therefore need to be provided by obstetricians to these older pregnant women.

age (over 35 years) of the higher risks for adverse pregnancy outcomes. Previous studies have shown that women of advanced age exhibited a significantly increased risk of adverse maternal and infant outcomes.15–18 Considerable attention will therefore need to be provided by obstetricians to these older pregnant women. Our results showed that farmers manifested a greater intent to have a second child relative to workers, which were in accordance with results reported by Zheng and colleagues.6 The intent by farmers to have more children might be related to their need for more labourers within the family, especially in rural areas. We also found household income to be significantly positively associated with fertility intent after adjustment for potential confounders, which was consistent with previous studies.19 20 This might be due to the fact that more affluent people might possess greater means to raise and educate more children. However, some studies showed economic level to be negatively associated with fertility intent.13 14 For example, Kim and colleagues14 found that fertility intent was lower in the group with higher monthly household income. The discrepancy might be related to disparities in economic pressures in different geographical areas. Besides socioeconomic factors, we found that a history of abortion was also associated with higher fertility intent; that is, women with a history of abortion were at higher risk for adverse pregnancy outcomes,21 22 which might prompt the wish for a healthier baby. We also found that women with one child had a higher fertility intent compared with women with no children, and the OR was even higher in women whose first child was a girl compared with women whose first child was a boy. This could be partially explained by the traditional preference for boys.1 7

for a healthier baby. We also found that women with one child had a higher fertility intent compared with women with no children, and the OR was even higher in women whose first child was a girl compared with women whose first child was a boy. This could be partially explained by the traditional preference for boys.1 7 The main strength of our study is that it is the first to estimate the intent to have a second child and to evaluate related factors in the general population during the new two-child policy in China; this could provide useful information for policymakers in providing better family planning services in the future. Our study also has several limitations. First, we collected the data via the web, so that women who were not internet users were not included in the study. Nevertheless, in China there were as many as 751 million internet users in 2017, accounting for one in five of the world's internet users; and 96.3% of them surf the internet by mobile phone.23 Additionally, internet use was more prevalent in women of childbearing age than in other age groups. Second, women from this study were selected via people listed on Wen Juan Xing, and thus selection biases might limit the generalisability of the results. We compared the distribution of age groups in each region in our survey to that in the statistical yearbook and found that the proportion of women under 25 years of age was higher in the Western region, which might lead to the underestimation of fertility intent. Moreover, using the backward stepwise method for model selection might lead to parameter estimates far from zero.24 Third, other risk factors such as fertility-related knowledge and attitude, and the intention of family members, were not investigated in the present study.

ead to the underestimation of fertility intent. Moreover, using the backward stepwise method for model selection might lead to parameter estimates far from zero.24 Third, other risk factors such as fertility-related knowledge and attitude, and the intention of family members, were not investigated in the present study. In conclusion, four out of 10 women of childbearing age planned to have a second child during China’s universal two-child policy, with economic, childrearing, and health barriers affecting fertility intent. The age range of 25–39 years, living in Eastern or Central China or in rural areas, being a farmer, possessing a higher household income, having a history of abortion, and already having one child were associated with a more pronounced intent to have a second child. We therefore assert that fertility-related factors be given full consideration when promoting the implementation of the new two-child policy. Contributors: JL searched the literature, analysed the data, interpreted the results, and drafted the manuscript. QM collected the data. SZ and QW revised the manuscript. ML conceived the study, designed the study, supervised the study, interpreted the results, and revised the manuscript. Funding: This study was funded by the National Natural Science Foundation of China (Grant No. 81703240) and the Chinese Association of Maternal and Child Health Studies (Grant No. AMCHS-2016-1). Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: This study was funded by the National Natural Science Foundation of China (Grant No. 81703240) and the Chinese Association of Maternal and Child Health Studies (Grant No. AMCHS-2016-1). Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data availability statement: Data are available upon reasonable request.

Key messages This study explored pharmacy staff knowledge and sales practices related to misoprostol in Mexico City, where first-trimester abortion was legalised in 2007. Our study found that misoprostol was available for sale at 61% of pharmacies surveyed, with 44% requiring a prescription. Women continue to seek access in pharmacies in the setting of legal abortion; improving pharmacy staff knowledge and practices can help reduce unsafe abortion. Introduction Mexico City, one of Mexico’s 32 states, legalised first-trimester abortion on request in 2007. Abortion is regulated at the state level, and first-trimester abortion is legal only under specific circumstances in other Mexican states. The majority of evidence about abortion in Mexico focuses on the public abortion programme, known as the Interrupcion Legal del Embarazo (ILE) programme, in Mexico City.1–4 Women also have access to services in the private sector, including private clinics,5 or by procuring misoprostol from a pharmacy with or without a prescription. While misoprostol was initially developed for treatment of ulcers, it is used in obstetrics and gynaecology to induce labour, treat haemorrhage, and induce pregnancy termination, which is sometimes also referred to as ‘initiating return of menses’.6 7 Misoprostol was first made available in Mexican pharmacies in 1985 by prescription only.8 However, pharmacies do not have to register prescriptions (as they do for antibiotics, for example) and available data suggest that women across Latin America obtain misoprostol without a prescription.7 8

ing return of menses’.6 7 Misoprostol was first made available in Mexican pharmacies in 1985 by prescription only.8 However, pharmacies do not have to register prescriptions (as they do for antibiotics, for example) and available data suggest that women across Latin America obtain misoprostol without a prescription.7 8 To date, research suggests a general lack of pharmacy staff knowledge about the use of misoprostol for abortion, including side effects, what the patient should expect, and effective dosing.7 8 This is an issue not only with misoprostol provision but also with other medical treatments provided directly in pharmacies in low- and middle-income countries worldwide.9 Yet women continue to seek access to medical abortion outside the clinic setting, even where abortion is legal, due to perceived ease of access, cost of services, and anonymity.2 Improving access and safe use of misoprostol through pharmacies can help reduce unsafe abortion.10

w- and middle-income countries worldwide.9 Yet women continue to seek access to medical abortion outside the clinic setting, even where abortion is legal, due to perceived ease of access, cost of services, and anonymity.2 Improving access and safe use of misoprostol through pharmacies can help reduce unsafe abortion.10 There is limited evidence about the knowledge and sale habits of pharmacy staff in Mexico City following legalisation of first-trimester abortion. The purpose of this study was to describe availability of misoprostol, whether or not the pharmacy requires a prescription, knowledge of indications for misoprostol use, and legal status of abortion among pharmacy employees in Mexico City. We hypothesised that even after abortion became legal, women continue to seek misoprostol outside of clinics, as evidenced by sales habits of pharmacies. We also explored differences in knowledge and sales habits between small independent and chain pharmacies, and by gender of pharmacy staff. Methods This study was a secondary analysis in collaboration with Ipas, a global non-profit focused on expanding access to safe abortion. Anonymous data were collected as part of an internal monitoring and evaluation project conducted by Ipas. No local institutional review board (IRB) approval was sought at the time of data collection. No Protected Health Information/Health Insurance Portability and Accountability Act (PHI/HIPAA) identifiers were collected.

nymous data were collected as part of an internal monitoring and evaluation project conducted by Ipas. No local institutional review board (IRB) approval was sought at the time of data collection. No Protected Health Information/Health Insurance Portability and Accountability Act (PHI/HIPAA) identifiers were collected. Pharmacies were randomly sampled from each of Mexico City’s 16 delegaciones (similar to a zip code area) and stratified by large chain pharmacies (eg, Ahorro, Guadalajara) and small independent pharmacies ('mom-and-pop' establishments) within each delegacion. In June 2015 when the Ipas assessment was started, there were 1305 pharmacies in Mexico City, 35% of which were chains and 65% small independent establishments. The distribution of chain and independent pharmacies across Mexico City as well as the density of pharmacies in each delegacion guided the study sample composition. Some smaller delegaciones did not have enough small independent pharmacies to make up 65% of the sample, in which case those willing to participate were included. The final sample was 174 pharmacies. Data were collected July–August 2015.

the density of pharmacies in each delegacion guided the study sample composition. Some smaller delegaciones did not have enough small independent pharmacies to make up 65% of the sample, in which case those willing to participate were included. The final sample was 174 pharmacies. Data were collected July–August 2015. Research assistants approached one employee at each selected pharmacy and asked them to participate in the study. Following verbal informed consent the staff person answered questions about their sociodemographic and pharmacy training characteristics. They also answered questions about their practices related to sale of misoprostol, including need for prescription, sale frequency, and knowledge of indications for use. The survey did not specifically ask if they sold misoprostol for medical abortions. Dosing recommendations were not discussed since the focus was on availability not pharmacist recommendations. Data did not contain any identifiers; this secondary analysis was deemed non-human subjects research by the Oregon Health & Science University IRB. Our primary outcome was availability of misoprostol in the pharmacy. Secondary outcomes were requirement of a prescription for purchase, knowledge of indications for misoprostol use, knowledge of the legal status of abortion in Mexico City, and estimated frequency of requests for misoprostol. Other variables included demographic, employment, and educational information about the participating pharmacy staff.

e requirement of a prescription for purchase, knowledge of indications for misoprostol use, knowledge of the legal status of abortion in Mexico City, and estimated frequency of requests for misoprostol. Other variables included demographic, employment, and educational information about the participating pharmacy staff. We used descriptive and bivariate statistics. We tested for differences using Chi-square tests to compare our outcomes by pharmacy type (chain vs small independent) and gender of the respondent. Results Of the 174 pharmacies, 65 (37%) were chain and 109 (63%) small independent, very similar to the known distribution of pharmacies in Mexico City. Respondents included 86 (49%) male and 88 (51%) female employees, with a roughly equal distribution between pharmacy types (table 1). The majority of interviewed employees at both types of pharmacies had completed at least some technical school (82% in chain and 66% in small independent pharmacies) and the sample included 21 licensed pharmacists, with no significant difference in education level of interviewees between pharmacy types (p=0.735) (table 1). Table 1 Sample characteristics

Results Of the 174 pharmacies, 65 (37%) were chain and 109 (63%) small independent, very similar to the known distribution of pharmacies in Mexico City. Respondents included 86 (49%) male and 88 (51%) female employees, with a roughly equal distribution between pharmacy types (table 1). The majority of interviewed employees at both types of pharmacies had completed at least some technical school (82% in chain and 66% in small independent pharmacies) and the sample included 21 licensed pharmacists, with no significant difference in education level of interviewees between pharmacy types (p=0.735) (table 1). Table 1 Sample characteristics Characteristic Overall (n=174) Chain pharmacies (n=65) Small independent pharmacies (n=109) P value % (n) % (95% CI) % (95% CI) Male respondent 49.5 (86) 46 (34 to 58) 51 (42 to 61) 0.505 Formal pharmacy training 28 (48) 29 (18 to 40) 27 (18 to 35) 0.735 Mean age (years) (SD) 37.6 (18–72) (0.98) 34 (1.2) 40 (1.4) 0.635 According to interviewed pharmacy employees, misoprostol was available at 61% of sites. Only 49% of independent pharmacies reported selling misoprostol, compared with 81.5% of chain pharmacies (p<0.05) (table 2). Of the small independent pharmacies selling misoprostol, 72% sold Cytotec, 21% generic, and 7% had both types. Among chain pharmacies 36% offered Cytotec, 38% generic, and 26% stocked both (p<0.001). Table 2 Survey results

Characteristic Overall (n=174) Chain pharmacies (n=65) Small independent pharmacies (n=109) P value % (n) % (95% CI) % (95% CI) Male respondent 49.5 (86) 46 (34 to 58) 51 (42 to 61) 0.505 Formal pharmacy training 28 (48) 29 (18 to 40) 27 (18 to 35) 0.735 Mean age (years) (SD) 37.6 (18–72) (0.98) 34 (1.2) 40 (1.4) 0.635 According to interviewed pharmacy employees, misoprostol was available at 61% of sites. Only 49% of independent pharmacies reported selling misoprostol, compared with 81.5% of chain pharmacies (p<0.05) (table 2). Of the small independent pharmacies selling misoprostol, 72% sold Cytotec, 21% generic, and 7% had both types. Among chain pharmacies 36% offered Cytotec, 38% generic, and 26% stocked both (p<0.001). Table 2 Survey results Question Overall (n=174) Percentage (%) reporting yes (n) Chain pharmacies (n=65) Percentage (%) reporting yes (95% CI) Small independent pharmacies (n=109) Percentage (%) reporting yes (95% CI) P value Is misoprostol sold? 61 (106) 82 (72 to 91) 49 (39 to 58) <0.001 Is a prescription required? 44 (76) 42 (30 to 54) 45 (36 to 54) 0.66 Is abortion legal? 80 (139) 80 (70 to 90) 80 (72 to 97) 0.91 Fewer than half (44%) of pharmacy staff reported requiring a prescription for sale of misoprostol. Among chain pharmacy employees, 41.5% reported requiring a prescription, while 50% of small independent pharmacy staff reported doing so (table 2). Nearly half (48.8%) of male pharmacy staff reported requiring a prescription compared with 38.6% of females, but this difference was not statistically significant (p=0.18).

Among chain pharmacy employees, 41.5% reported requiring a prescription, while 50% of small independent pharmacy staff reported doing so (table 2). Nearly half (48.8%) of male pharmacy staff reported requiring a prescription compared with 38.6% of females, but this difference was not statistically significant (p=0.18). Eighty percent of respondents knew that abortion was legal in Mexico City, with no difference between pharmacy types (80% of chain and 80% of small independent pharmacy employees) or gender of pharmacy worker (80.2% of males, 79.6% of females)(table 2). Knowledge of indications for misoprostol use was similar between independent and chain pharmacy staff. Most individuals were able to identify the outpatient indications (ulcers and abortion) but not the inpatient uses (haemorrhage, induction of labour). Though only 13.8% of chain and 8.2% of independent pharmacy employees correctly identified at least three out of five of the indications, the vast majority knew that it can be used to treat ulcers (74% chain, 66% small independent; 69% overall; p=0.66) and induce abortion (71% chain, 61.5% small independent; 65% overall; p=0.33) with no significant difference by pharmacy type (figure 1). Figure 1 Knowledge of indications for misoprostol by pharmacy type. Ind, independent. A larger proportion of respondents at pharmacies where misoprostol is sold reported knowing that misoprostol can be used for medical abortion (71% vs 56%; p=0.045). Knowledge of abortion as an indication for misoprostol does not vary by requirement for a prescription (p=0.3)

Figure 1 Knowledge of indications for misoprostol by pharmacy type. Ind, independent. A larger proportion of respondents at pharmacies where misoprostol is sold reported knowing that misoprostol can be used for medical abortion (71% vs 56%; p=0.045). Knowledge of abortion as an indication for misoprostol does not vary by requirement for a prescription (p=0.3) Most interviewees (75.4%) reported 0–2 requests for misoprostol per month, with no significant difference between pharmacy types. However, interviewees from six pharmacies reported 30 or more requests per month, all in different delegations, and one pharmacy reported 180 requests a month. Discussion This study shows that misoprostol is widely available across pharmacies in Mexico City, but more available at large chain pharmacies than small independent ones, contrary to our hypothesis. Just under half report requesting a prescription, with female employees slightly less likely to require one. The difference by gender did not reach statistical significance, though it may be meaningful in practice, because it could impact women’s access to misoprostol at the population level. Eighty percent of interviewed pharmacy workers know abortion is legal in Mexico City and the majority know that misoprostol can be used to induce abortion.

nder did not reach statistical significance, though it may be meaningful in practice, because it could impact women’s access to misoprostol at the population level. Eighty percent of interviewed pharmacy workers know abortion is legal in Mexico City and the majority know that misoprostol can be used to induce abortion. Our study found that misoprostol was available for sale at 61% of pharmacies surveyed, with 44% reporting that they required a prescription. This is comparable to other studies that have been done in Mexico and Latin America,7 8 11 though rates of prescription requirement were reportedly as low as 11% in the mystery client portion of one study.11 Misoprostol is expensive to keep in stock, which can be burdensome to small pharmacies who sell the medication infrequently, which may partially account for chain pharmacies being more likely to stock misoprostol than small independent pharmacies. Worldwide, there is significant variation in misoprostol availability, as well as how often misoprostol is recommended compared with other medications when an abortifacient medication is requested at a pharmacy.10 12–15 The studies previously performed in Mexico and Latin America did not differentiate availability based on pharmacy type, though one did note that misoprostol was more likely to be sold by the pill rather than the bottle in independent pharmacies and low socioeconomic areas, which was used as a surrogate marker to indicate sale for use in medical abortion.8 It was also significantly more available in Mexico City, where abortion is presently legal, than in other areas of Mexico, where abortion is still legally restricted.8 Our study, undertaken after the decriminalisation of first-trimester abortion in Mexico City, suggests that women continue to access misoprostol outside of clinics, with some pharmacies requiring a prescription.

ere abortion is presently legal, than in other areas of Mexico, where abortion is still legally restricted.8 Our study, undertaken after the decriminalisation of first-trimester abortion in Mexico City, suggests that women continue to access misoprostol outside of clinics, with some pharmacies requiring a prescription. Mystery client studies previously performed in the region have focused on what medication is recommended when an abortifacient is requested and whether or not it is an effective method, as well as whether or not safety information and follow-up precautions are given. However, these studies were performed in areas where abortion is legally restricted, so are not directly comparable with our findings. Our data show that the majority of pharmacy staff know that misoprostol can be used for ulcers and to induce abortion, but are less familiar with other indications such as management of postpartum haemorrhage. This may be related to the fact that we surveyed only outpatient pharmacies. Prior studies around the world show wide variation in medication and dose recommendations given to women requesting medication to interrupt a pregnancy in settings with variable legal restrictions on abortion,10 demonstrating an ongoing need for additional education on safe and effective dosing, though this was not part of the present study.

e world show wide variation in medication and dose recommendations given to women requesting medication to interrupt a pregnancy in settings with variable legal restrictions on abortion,10 demonstrating an ongoing need for additional education on safe and effective dosing, though this was not part of the present study. We found that female pharmacy staff were slightly less likely than their male counterparts (39% vs 49%) to require a prescription, although this difference did not reach statistical significance. To our knowledge, only one previous study looked at differential prescription requirement by gender of the pharmacy worker, which also found that female staff were slightly more likely to require a prescription and sell by the pill, though the findings were not significant.8

ence did not reach statistical significance. To our knowledge, only one previous study looked at differential prescription requirement by gender of the pharmacy worker, which also found that female staff were slightly more likely to require a prescription and sell by the pill, though the findings were not significant.8 Our results can be used to tailor interventions for pharmacy workers to improve the information women receive when requesting misoprostol, with or without a prescription. Educational interventions targeting pharmacy workers have been shown to be effective to improve referrals, information about medical abortion regimens and follow-up.16 17 Training would obviously need to be region-specific, based on factors of legality, medication accessibility, and availability of follow-up. Even in a context of legal abortion and good access to services in the public and private sectors, women may choose to obtain misoprostol directly from pharmacies; this was not known in Mexico City. Our results suggest that there is a need for educating pharmacy employees, since they are providing misoprostol directly to women, both with and without a prescription.

cess to services in the public and private sectors, women may choose to obtain misoprostol directly from pharmacies; this was not known in Mexico City. Our results suggest that there is a need for educating pharmacy employees, since they are providing misoprostol directly to women, both with and without a prescription. Our findings must be interpreted in light of several limitations. First, we relied on self-reported sales practices. Second, we did not ask about dosing recommendations. This study focused on availability, requirement of a prescription, and knowledge of indications. Our sample of pharmacies was designed to be representative of the distribution of chain and small independent pharmacies in Mexico City’s delegaciones, yet there were localities where the target number of pharmacies could not be reached due to limited overall pharmacy numbers and willingness to participate. We may be underpowered for some comparisons; however, this descriptive study presents novel data on pharmacy practices in Mexico City after legalisation of abortion. The overall representation of pharmacies in all delegaciones of Mexico City, and inclusion of gender and educational diversity among staff, are strengths of the study. Additionally, social and religious biases, as well as utilising a survey method rather than the mystery client approach taken by other studies, could have contributed to underreporting of misoprostol availability and sales practices, even in the legal setting. However, our data do not allow us to assess whether underreporting occurred.

ial and religious biases, as well as utilising a survey method rather than the mystery client approach taken by other studies, could have contributed to underreporting of misoprostol availability and sales practices, even in the legal setting. However, our data do not allow us to assess whether underreporting occurred. First-trimester medical abortion is legal and widely available in public and private sector clinics in Mexico City. Despite this, women still access misoprostol directly from pharmacies, with or without a prescription. Most pharmacy staff know that abortion is legal and that misoprostol can be used to induce abortion, but availability, knowledge of indications for use of misoprostol, and requirement of a prescription varies widely among pharmacies. This may result in variable access to misoprostol depending on the type pf pharmacy a woman walks into. Pharmacies may be a good place to target education for pharmacy employees and women about safe and effective use of misoprostol for abortion, whether through expanding knowledge of indications and appropriate dosing regimens, or improving access to information within pharmacies about abortion services available in Mexico City. Women will continue to seek access to medical abortion in pharmacies; therefore, improving pharmacy staff knowledge and practices can help reduce unsafe abortion. Contributors: RS and BGD conceived the analysis. MEC and RS led data collection. GW and BGD conducted the analysis. GW drafted the manuscript, BGD, RS, MEC and SK provided substantive revisions.

First-trimester medical abortion is legal and widely available in public and private sector clinics in Mexico City. Despite this, women still access misoprostol directly from pharmacies, with or without a prescription. Most pharmacy staff know that abortion is legal and that misoprostol can be used to induce abortion, but availability, knowledge of indications for use of misoprostol, and requirement of a prescription varies widely among pharmacies. This may result in variable access to misoprostol depending on the type pf pharmacy a woman walks into. Pharmacies may be a good place to target education for pharmacy employees and women about safe and effective use of misoprostol for abortion, whether through expanding knowledge of indications and appropriate dosing regimens, or improving access to information within pharmacies about abortion services available in Mexico City. Women will continue to seek access to medical abortion in pharmacies; therefore, improving pharmacy staff knowledge and practices can help reduce unsafe abortion. Contributors: RS and BGD conceived the analysis. MEC and RS led data collection. GW and BGD conducted the analysis. GW drafted the manuscript, BGD, RS, MEC and SK provided substantive revisions. Funding: Dr Blair Darney was supported by the Society of Family Planning (SFPRF11-02) and grant number K12HS022981 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. Competing interests: None declared.

Funding: Dr Blair Darney was supported by the Society of Family Planning (SFPRF11-02) and grant number K12HS022981 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. Competing interests: None declared. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed. Data availability statement: Data are available upon reasonable request.

Key messages Previous data from observational studies provided inconsistent evidence about the association between depot medroxyprogesterone acetate (DMPA) and HIV acquisition. New data from a high-quality randomised clinical trial found no statistically significant differences in HIV acquisition among DMPA users compared with implant users or copper intrauterine device users. Observational data continue to suggest no association between other hormonal contraception methods and HIV acquisition. The World Health Organization has updated recommendations on contraceptive use for women at high risk of HIV, based in part on this new evidence.

New data from a high-quality randomised clinical trial found no statistically significant differences in HIV acquisition among DMPA users compared with implant users or copper intrauterine device users. Observational data continue to suggest no association between other hormonal contraception methods and HIV acquisition. The World Health Organization has updated recommendations on contraceptive use for women at high risk of HIV, based in part on this new evidence. Introduction Observational epidemiological studies since 1991 have provided inconsistent evidence about a link between hormonal contraception use and HIV acquisition among women.1 2 Of particular concern has been the potential for increased risk of HIV acquisition associated with injectable depot medroxyprogesterone acetate (DMPA), a widely used contraceptive method in some areas of the world where HIV incidence rates, and maternal morbidity and mortality rates, are high.2 3 The World Health Organization (WHO) has held several consultations to review accumulating evidence and inform recommendations for safe contraception use among women at high risk of HIV, published in WHO’s Medical Eligibility Criteria for Contraceptive Use (WHOMEC).4 In 2016, a systematic review of epidemiological evidence on the association between hormonal contraception use and HIV acquisition was conducted in preparation for a WHO consultation.2 That review included relevant evidence published to 15 January 2016 from 31 observational studies. Most studies were considered “unlikely to inform the primary question” due to methodological limitations, and the review’s appraisal focused on 14 higher-quality studies deemed “informative but with important limitations”. The systematic review concluded that the evidence for most hormonal contraceptive methods did not suggest an association with HIV acquisition, but that new information had increased concerns about a possible association between DMPA and HIV risk. A key uncertainty when interpreting the findings of these observational data was the possibility of unmeasured, residual confounding distorting the true association between DMPA use and HIV acquisition, particularly in relation to condom use and other sexual behaviours.

a possible association between DMPA and HIV risk. A key uncertainty when interpreting the findings of these observational data was the possibility of unmeasured, residual confounding distorting the true association between DMPA use and HIV acquisition, particularly in relation to condom use and other sexual behaviours. In June 2019, results from the Evidence for Contraceptive Options and HIV Outcomes (ECHO) trial were published.5 The ECHO trial was a randomised clinical trial (RCT) specifically designed to compare HIV incidence among users of three common contraceptives: intramuscular DMPA (DMPA-IM), levonorgestrel implant (LNG implant) and the copper intrauterine device (Cu-IUD), with the goal of minimising confounding through randomisation. We provide an update of the previous systematic review, including results from the ECHO trial and other new epidemiological evidence reported since the last publication. Methods This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.6 We followed the same methodology that was used in the previous systematic review,2 except where noted below. The review question was: “Among women at risk of HIV, does use of a hormonal contraceptive method compared with non-use of a hormonal contraceptive method (or use of another specific hormonal contraceptive method) increase the risk of HIV acquisition?”.

used in the previous systematic review,2 except where noted below. The review question was: “Among women at risk of HIV, does use of a hormonal contraceptive method compared with non-use of a hormonal contraceptive method (or use of another specific hormonal contraceptive method) increase the risk of HIV acquisition?”. Inclusion and exclusion criteria We included studies of women of reproductive age at risk of HIV infection (ie, HIV-seronegative at baseline) that compared incident, laboratory-confirmed HIV infection among women using a hormonal contraceptive method (all types of injectable, oral, implant, patch, ring, or the levonorgestrel-containing intrauterine device (LNG-IUD)), compared with women using a non-hormonal method (eg, condom, non-hormonal IUD, sterilisation, withdrawal) or no contraception. We also included studies comparing incident HIV infection among users of one specific hormonal contraceptive method compared with another specific hormonal contraceptive method (eg, DMPA users compared with norethisterone enanthate (NET-EN) users). We included longitudinal studies (observational studies, randomised trials, or meta-analyses containing data not previously published). Given the timing and location of the studies, we assumed that most, if not all, DMPA use in these studies was DMPA-IM rather than subcutaneous DMPA (DMPA-SC).

nanthate (NET-EN) users). We included longitudinal studies (observational studies, randomised trials, or meta-analyses containing data not previously published). Given the timing and location of the studies, we assumed that most, if not all, DMPA use in these studies was DMPA-IM rather than subcutaneous DMPA (DMPA-SC). Search strategy We searched Pubmed and Embase for the dates 15 January 2016 (the cut-off date for the last systematic review) to 26 June 2019 (online supplementary appendix 1). We hand-searched reference lists of included studies. Three authors (TC, KC, PH) screened the abstracts and full-text articles using Covidence,7 and conflicts were resolved through discussion. 10.1136/bmjsrh-2019-200509.supp1Supplementary data Data extraction and quality assessment We used a data abstraction form to abstract data from any newly identified studies. The previous review used a study quality assessment framework that focused on observational studies.2 Briefly, studies that failed to adjust for condom use or that had unclear measurement of hormonal contraception exposure were considered “unlikely to inform the primary question”; all other studies were considered “informative but with important limitations” to emphasise the potential for residual or uncontrolled confounding in these observational studies. For this review, we expanded the previous framework to include criteria for assessing RCTs and to define criteria for an additional higher-quality level of studies that were “informative with few limitations” (online supplementary appendix 2).8

al for residual or uncontrolled confounding in these observational studies. For this review, we expanded the previous framework to include criteria for assessing RCTs and to define criteria for an additional higher-quality level of studies that were “informative with few limitations” (online supplementary appendix 2).8 Graphical summaries and meta-analysis We updated the forest plots with all newly identified reports that were considered “informative but with important limitations” or “informative with few limitations”. The forest plots present adjusted point estimates and confidence intervals (CIs) for each contraceptive method compared with a non-hormonal method or no contraception, or head-to-head comparisons between two specific hormonal methods. For reports in which authors presented estimates from more than one model, we used the most fully adjusted estimates.

d point estimates and confidence intervals (CIs) for each contraceptive method compared with a non-hormonal method or no contraception, or head-to-head comparisons between two specific hormonal methods. For reports in which authors presented estimates from more than one model, we used the most fully adjusted estimates. We also updated the meta-analysis of observational studies assessing the association between DMPA (compared with no hormonal contraception) and risk of HIV. We determined a priori that it was not appropriate to combine estimates from observational studies and RCTs into a single meta-analysis, given differences in potential sources of bias.9 Meta-analyses were conducted in Stata version 17 (StataCorp, College Station, TX, USA). We included the most fully adjusted Cox hazards ratio (HR) estimates from each study. These point estimates and 95% CIs were log-transformed to calculate standard errors, using a random effects model.10 We evaluated statistical heterogeneity using the I2 statistic; a value of less than 25% was considered indicative of low heterogeneity.11 Patient and public involvement As this analysis was based on published data, patients and the public were not involved in the design of the study.

We also updated the meta-analysis of observational studies assessing the association between DMPA (compared with no hormonal contraception) and risk of HIV. We determined a priori that it was not appropriate to combine estimates from observational studies and RCTs into a single meta-analysis, given differences in potential sources of bias.9 Meta-analyses were conducted in Stata version 17 (StataCorp, College Station, TX, USA). We included the most fully adjusted Cox hazards ratio (HR) estimates from each study. These point estimates and 95% CIs were log-transformed to calculate standard errors, using a random effects model.10 We evaluated statistical heterogeneity using the I2 statistic; a value of less than 25% was considered indicative of low heterogeneity.11 Patient and public involvement As this analysis was based on published data, patients and the public were not involved in the design of the study. Results Description of included studies We screened 769 abstracts and 17 full-text reports (online supplementary figure 1). We excluded 13 full-text reports – four reported on data already included in the 2016 review, six were letters or conference abstracts, and three had an irrelevant exposure, outcome or comparison group.12–24 From a conference abstract,13 we identified an additional 'in press' study, which has now been published.25 Therefore, five new reports met our inclusion criteria (online supplementary table 1).5 25–28 Two reports provided the first data from RCTs,5 26 another provided data from a new observational analysis,25 and two provided additional analyses from previously published observational studies.27 28 Two RCTs5 26 and two observational studies25 28 provided direct estimates for the association between hormonal contraception and HIV acquisition, while three studies (one RCT and two observational) examined various factors that might modify any association between hormonal contraceptive use and HIV risk.5 27 28 One RCT was considered “informative with few limitations”,5 and the other “unlikely to inform the primary question”.26 The three observational studies were rated “informative but with important limitations”.25 27 28 Therefore, three new reports (from one RCT and two observational studies) provided higher-quality (either “informative with few limitations” or “informative but with important limitations”) direct evidence about the association between hormonal contraceptive use and HIV acquisition.5 25 28

portant limitations”.25 27 28 Therefore, three new reports (from one RCT and two observational studies) provided higher-quality (either “informative with few limitations” or “informative but with important limitations”) direct evidence about the association between hormonal contraceptive use and HIV acquisition.5 25 28 Hormonal contraception and risk of HIV infection: new RCTs The ECHO trial provided the first evidence from an RCT that was specifically designed to compare HIV incidence rates among women using DMPA-IM, LNG implants or Cu-IUDs.5 This study did not include a group using no contraception. Investigators randomised 7829 women from 12 sites in four African countries to the three contraceptive groups. Participants were aged 16–35 years, HIV-seronegative, and seeking contraception. Follow-up visits occurred every 3 months for up to 18 months and included HIV testing, contraceptive counselling, behavioural assessment, and HIV prevention services. Modified intention-to-treat analyses were conducted, along with pre-specified analyses restricted to participants with continuous use of the assigned contraceptive method and adjusted for potential confounders including condom use and sexual behaviours. Ninety-nine percent of women accepted their assigned method, and participants used their assigned method for 92% of the follow-up time. More than 91% of participants attended each scheduled follow-up visit, 99% had at least one follow-up HIV test, and the study accumulated 10 409 woman-years of follow-up. Overall HIV incidence was 3.81 per 100 woman-years. No statistically significant differences in HIV incidence were observed among the three contraceptive groups in the main modified intention-to–treat analysis: HR 1.04 (96% CI 0.82 to 1.33) for DMPA-IM compared with Cu-IUD; HR 1.23 (96% CI 0.95 to 1.59) for DMPA-IM compared with LNG implant, and HR 1.18 (96% CI 0.91 to 1.53) for Cu-IUD compared with LNG implant. The continuous contraceptive use analysis, which adjusted for baseline and time-varying covariates, yielded similar results. The ECHO trial was considered to be “informative with few limitations” because of robust randomisation and allocation procedures, high follow-up and contraceptive continuation rates, assessment of condom use and sexual behaviour as possible confounders, and appropriate intention-to-treat and pre-specified sensitivity analyses.

trial was considered to be “informative with few limitations” because of robust randomisation and allocation procedures, high follow-up and contraceptive continuation rates, assessment of condom use and sexual behaviour as possible confounders, and appropriate intention-to-treat and pre-specified sensitivity analyses. Women and providers were not blinded to contraceptive group allocation; however, study personnel involved in HIV testing were blinded to contraception assignment, and the study team worked to ensure that counselling messages were consistent across the three groups.

trial was considered to be “informative with few limitations” because of robust randomisation and allocation procedures, high follow-up and contraceptive continuation rates, assessment of condom use and sexual behaviour as possible confounders, and appropriate intention-to-treat and pre-specified sensitivity analyses. Women and providers were not blinded to contraceptive group allocation; however, study personnel involved in HIV testing were blinded to contraception assignment, and the study team worked to ensure that counselling messages were consistent across the three groups. Another open-label RCT was initiated in 2009 but stopped early because of plans to conduct the ECHO trial.26 This trial randomised women in South Africa to receive a Cu-IUD or progestogen-only injectable (mostly DMPA, some NET-EN).26 Follow-up was planned for 12 months. The analysis included 656 injectable users and 634 Cu-IUD users who had a follow-up HIV test. All of the results presented were unadjusted. No increase in HIV seroconversion was observed between injectable users and Cu-IUD users in the intent-to-treat analysis (risk ratio (RR) 0.88, 95% CI 0.48 to 1.59) or in the per-protocol analysis (RR 0.94, 95% CI 0.52 to 1.71). Separate per-protocol risk estimates for DMPA and NET-EN, each compared with Cu-IUD, also indicated no association. Although this study used appropriate procedures for randomisation and allocation concealment, it was considered “unlikely to inform the primary question” because of no measurement of condom use, failure to include a time-varying analysis of hormonal contraceptive exposure or description of contraceptive discontinuation or switching, and the long inter-survey interval for assessing HIV seroconversion (median 19–20 months).

idered “unlikely to inform the primary question” because of no measurement of condom use, failure to include a time-varying analysis of hormonal contraceptive exposure or description of contraceptive discontinuation or switching, and the long inter-survey interval for assessing HIV seroconversion (median 19–20 months). Hormonal contraception and risk of HIV infection: new observational studies A secondary analysis of an RCT of the dapivirine ring to prevent HIV assessed HIV incidence among 1136 South African study participants using DMPA-IM, NET-EN, implants (type not specified) or Cu-IUDs.25 Contraceptive use, sexual behaviour and other covariates, and HIV infection status were assessed during monthly visits. Ninety-five participants seroconverted over 1771 woman-years of follow-up. In contrast to previous observational studies in which the non-hormonal contraceptive comparison group included a mix of non-hormonal contraceptive use and no contraceptive use, this study used Cu-IUD users as the non-hormonal contraceptive referent group. Overall, there were no statistically significant differences in HIV incidence among users of the different contraceptive methods. When Cu-IUD users formed the referent group, the adjusted hazards ratio (adj HR) was 0.91 (95% CI 0.44 to 1.87) for DMPA-IM, adj HR 1.02 (95% CI 0.49 to 2.12) for NET-EN, adj HR 0.46 (95% CI 0.13 to 1.70) for implants and adj HR 0.90 (95% CI 0.45 to 1.76) for all three hormonal methods considered together. When NET-EN users formed the referent group, the adj HR was 0.89 (95% CI 0.55 to 1.44) for DMPA-IM, adj HR 0.45 (95% CI 0.13 to 1.53) for implants and adj HR 0.98 (95% CI 0.47 to 2.03) for Cu-IUD. This study was considered “informative but with important limitations”. Strengths included time-varying analysis of hormonal contraception exposure, condom use, and other relevant confounders; control for trial allocation group (dapivirine or placebo); clear exposure and comparison groups; and a short inter-survey interval (monthly). Limitations included no information on implant type (levonorgestrel or etonogestrel), no information about attrition within each contraceptive group, and the potential for residual or unmeasured confounding.

roup (dapivirine or placebo); clear exposure and comparison groups; and a short inter-survey interval (monthly). Limitations included no information on implant type (levonorgestrel or etonogestrel), no information about attrition within each contraceptive group, and the potential for residual or unmeasured confounding. Another paper provided additional analyses with more participants and further follow-up time from the Mombasa Cohort, a long-term observational study of HIV acquisition among female sex workers in Kenya.28 Previous analyses had observed that both oral contraceptives (type not specified) and DMPA were associated with an increased HIV risk compared with no contraceptive use.29 LNG implant use was associated with an elevated but not statistically significant relative risk.30 The new analysis included 1985 women with 7127 woman-years of follow-up; hormonal contraception, HIV, and relevant confounders were measured monthly. Results for oral contraceptives (adj HR 1.48, 95% CI 1.05 to 2.09) and DMPA (adj HR 1.72, 95% CI 1.34 to 2.20), both compared with no contraception, were almost identical to the previous results. LNG implants were not associated with HIV seroconversion (adj HR 0.99, 95% CI 0.40 to 2.45). This study was considered “informative with important limitations”, primarily because of the potential for unmeasured or residual confounding.

both compared with no contraception, were almost identical to the previous results. LNG implants were not associated with HIV seroconversion (adj HR 0.99, 95% CI 0.40 to 2.45). This study was considered “informative with important limitations”, primarily because of the potential for unmeasured or residual confounding. Hormonal contraceptive use versus non-use in studies considered “informative but with important limitations” or “informative with few limitations” Previously, we included 31 observational studies, and focused on 14 observational studies considered “informative but with important limitations”.2 Results from three newly identified publications were of sufficient quality to be added to the forest plots from the previous review (figures 1–3 and online supplementary figure 2).5 25 28 A new estimate for the head-to-head comparison between DMPA-IM and NET-EN use was less than unity and not statistically significant,25 unlike the statistically significant, above unity, estimates from two studies in the previous review; the 95% CIs for all three estimates overlapped (figure 4). Finally, new head-to-head estimates for implants versus NET-EN and DMPA versus LNG implants were added (figure 4). The revised meta-analysis of observational studies of DMPA compared with non-hormonal or no contraception and risk of HIV acquisition resulted in a pooled OR of 1.40 (95% CI 1.24 to 1.58) with an I2 of 0%, indicating low heterogeneity (online supplementary figure 3); this was almost identical to the pooled estimate in the 2016 review.2

analysis of observational studies of DMPA compared with non-hormonal or no contraception and risk of HIV acquisition resulted in a pooled OR of 1.40 (95% CI 1.24 to 1.58) with an I2 of 0%, indicating low heterogeneity (online supplementary figure 3); this was almost identical to the pooled estimate in the 2016 review.2 Figure 1 Use of depot medroxyprogesterone acetate (DMPA) (or unspecified injectable) versus non-hormonal or no contraception and HIV acquisition, among 14 studies considered “informative but with important limitations” or “informative with few limitations”. Error bars show 95% confidence intervals (CIs), with the exception of the ECHO trial which uses 96% CI. Observational studies arranged in order of decreasing magnitude of risk estimate. Graph does not display estimates from marginal structural models, except where use of such models resulted in a different conclusions regarding statistical significance; in such cases, estimates from both models are displayed on a single line (also identified by bracket signs). Note: restricted to all data on DMPA or unspecified injectables, as estimates of unspecified injectables are likely comprised largely of DMPA. *Analysis showed significant findings at p=0.05. †Estimate for Cox model taken from slightly updated analysis that controlled for total number of unprotected sex acts. ¶Updated estimate from Baeten et al, 2007. ˆUnpublished estimates from a sub-analysis of Morrison 2015 meta-analysis, restricted to pooled analysis using databases not previously used to publish estimates on hormonal contraceptive methods and HIV acquisition risk. ¥Different statistical models adjusted for slightly different confounders. §Unpublished estimates disaggregated by injectable type. adjHR, adjusted HR; adjIRR, adjusted incidence risk ratio; CI, confidence interval; DMPA, depot medroxyprogesterone acetate; ECHO, Evidence for Contraceptive Options and HIV Outcomes Trial; HR, hazard ratio; ITT, intention-to-treat; MSM, marginal structural models.

npublished estimates disaggregated by injectable type. adjHR, adjusted HR; adjIRR, adjusted incidence risk ratio; CI, confidence interval; DMPA, depot medroxyprogesterone acetate; ECHO, Evidence for Contraceptive Options and HIV Outcomes Trial; HR, hazard ratio; ITT, intention-to-treat; MSM, marginal structural models. Figure 2 Use of norethisterone enanthate versus non-hormonal or no contraception and HIV acquisition, among seven studies considered informative but with important limitations. Error bars show 95% confidence intervals. Studies arranged in order of decreasing magnitude of risk estimate. Graph does not display estimates from marginal structural models. ˆUnpublished estimates from a sub-analysis of Morrison et al. 2015 meta-analysis, restricted to pooled analysis using databases not previously used to publish estimates on hormonal contraceptive methods and HIV acquisition risk. §Unpublished estimates disaggregated by injectable type. adjHR, adjusted HR; adjIRR, adjusted incidence risk ratio; CI, confidence interval; HR, hazard ratio; NET-EN, norethisterone enanthate.

d analysis using databases not previously used to publish estimates on hormonal contraceptive methods and HIV acquisition risk. §Unpublished estimates disaggregated by injectable type. adjHR, adjusted HR; adjIRR, adjusted incidence risk ratio; CI, confidence interval; HR, hazard ratio; NET-EN, norethisterone enanthate. Figure 3 Use of implants* (versus non-hormonal or no contraception) and HIV acquisition, among four studies considered “informative but with important limitations” or “informative with few limitations”. Error bars show 95% confidence intervals (CIs), with the exception of the ECHO trial which uses 96% CI. Studies arranged in order of decreasing magnitude of risk estimate. *Sabo et al, Wall et al and the ECHO trial included levonorgestrel implants; Palanee-Phillips et al did not specify implant type. **The ECHO trial reported an adjHR 1.18 (96% CI 0.91 to 1.53) for the comparison of copper intrauterine devices (Cu-IUDs) versus levonorgestrel (LNG) implant. The inverse association is represented on this figure for LNG implant versus Cu-IUD. adjHR, adjusted HR; CI, confidence interval; ECHO, Evidence for Contraceptive Options and HIV Outcomes Trial; HR, hazard ratio; ITT, intention-to-treat.

ison of copper intrauterine devices (Cu-IUDs) versus levonorgestrel (LNG) implant. The inverse association is represented on this figure for LNG implant versus Cu-IUD. adjHR, adjusted HR; CI, confidence interval; ECHO, Evidence for Contraceptive Options and HIV Outcomes Trial; HR, hazard ratio; ITT, intention-to-treat. Figure 4 Hormonal contraceptive methods and HIV acquisition in head-to-head studies, among four studies considered “informative but with important limitations” or “informative with few limitations”. Error bars show 95% confidence intervals (CIs), with the exception of the ECHO trial which uses 96% CI. Within each comparison group, studies are arranged in order of decreasing magnitude of risk estimate. *Analysis showed significant findings at p=0.05. **Results from randomised clinical trial; all other results from observational studies. adjHR, adjusted HR; CI, confidence interval; COC, combined oral contraception; ECHO, Evidence for Contraceptive Options and HIV Outcomes Trial; DMPA, depot medroxyprogesterone acetate; HR, hazard ratio; LNG, levonorgestrel; NET-EN, norethisterone enanthate. Effect modification The previous systematic review did not identify consistent factors that modified any association between hormonal contraception and HIV acquisition (eg, age, herpes simplex virus type 2 (HSV-2), study site).2 New data from the ECHO trial found that age and HSV-2 status did not modify results comparing HIV incidence among the different contraceptive methods.5

dentify consistent factors that modified any association between hormonal contraception and HIV acquisition (eg, age, herpes simplex virus type 2 (HSV-2), study site).2 New data from the ECHO trial found that age and HSV-2 status did not modify results comparing HIV incidence among the different contraceptive methods.5 Two analyses newly assessed bacterial vaginosis (BV) as an effect modifier. A cohort study in Zambia27 had previously reported no statistically significant associations for oral contraceptives, DMPA or LNG implant use with HIV acquisition.31 The new analysis examined whether the relationship between hormonal contraception and HIV acquisition was modified by a BV diagnosis among 564 women in serodiscordant couples. In the time-varying analysis, both DMPA (adj HR 6.55, 95% CI 1.14 to 37.77) and oral contraceptives (adj HR 5.20, 95% CI 1.68 to 16.06) were associated with HIV seroconversion during periods when BV was diagnosed, each compared with non-hormonal or no contraceptive use. No statistically significant associations were observed for DMPA (adj HR 1.35, 95% CI 0.64 to 2.85) or oral contraceptives (adj HR 1.36, 95% CI 0.76 to 2.42) among women not diagnosed with BV. Results were similar when variables for sperm on wet-prep and condomless sex in the past 3 months were added to the model. No incident HIV infections occurred among LNG implant users.

for DMPA (adj HR 1.35, 95% CI 0.64 to 2.85) or oral contraceptives (adj HR 1.36, 95% CI 0.76 to 2.42) among women not diagnosed with BV. Results were similar when variables for sperm on wet-prep and condomless sex in the past 3 months were added to the model. No incident HIV infections occurred among LNG implant users. In contrast, the updated analysis from the Mombasa Cohort Study did not find that BV was an effect modifier.28 DMPA use was associated with increased risk of HIV among women with BV (adj HR 1.56, 95% CI 1.08 to 2.25) and without BV (adj HR 2.08, 1.46–2.97), each compared with no contraceptive use. Oral contraceptive users had an elevated but not statistically significant relative risk among women with BV (adj HR 1.50, 95% CI 0.94 to 2.39) and without BV (adj HR 1.61, 95% CI 0.99 to 2.64). LNG implants were not associated with HIV seroconversion in either group.

.46–2.97), each compared with no contraceptive use. Oral contraceptive users had an elevated but not statistically significant relative risk among women with BV (adj HR 1.50, 95% CI 0.94 to 2.39) and without BV (adj HR 1.61, 95% CI 0.99 to 2.64). LNG implants were not associated with HIV seroconversion in either group. Discussion The previous 2016 systematic review concluded that new observational evidence at that time increased concern about a possible causal association between DMPA use and HIV acquisition, although there was uncertainty about the potential for unmeasured and residual confounding particularly from differential condom use or other measures of sexual behaviour.2 In this update, the addition of results from one new observational study,25 and updated results from a previously reported observational study,28 did not materially change the pattern of risk seen in the forest plots or the meta-analysis of observational studies suggesting an increased risk of HIV acquisition with DMPA use.2 Data from a new RCT, however, do not support such a conclusion. 5 The ECHO trial found no statistically significant differences in HIV incidence among DMPA-IM users when compared with users of a non-hormonal method (Cu-IUDs) or LNG implant users. This trial provided high-quality, direct evidence of HIV risk among users of three contraceptive methods. The ECHO trial randomised women to contraceptive methods, had low attrition and high rates of continuation with the assigned contraceptive method, and found similar results of no association in both the main modified intention-to-treat analyses and secondary sensitivity analyses that controlled for several baseline and time-varying factors including condomless sex. Therefore, residual confounding is unlikely to have affected results from the ECHO trial, and so it provides higher-quality evidence about whether DMPA increases the risk of HIV acquisition in women than that from observational studies.

controlled for several baseline and time-varying factors including condomless sex. Therefore, residual confounding is unlikely to have affected results from the ECHO trial, and so it provides higher-quality evidence about whether DMPA increases the risk of HIV acquisition in women than that from observational studies. The ECHO trial observed no increased risk of HIV acquisition among LNG implant users when compared with Cu-IUD users or DMPA-IM users.5 These findings are consistent with three observational studies (two on LNG implants and one on unspecified implants) that found no increased risk of HIV acquisition among implant users compared with non-hormonal contraceptive users.25 28 31 The study of unspecified implants also observed a statistically non-significant reduced risk of HIV among implant users compared with NET-EN users, which had not been assessed previously.25 The ECHO trial did not examine NET-EN, and therefore information on NET-EN and HIV acquisition came solely from observational studies. Previously, six studies examined HIV acquisition among NET-EN users compared with non-hormonal users, with none finding a statistically significant association; a newly identified seventh study reported an estimate consistent with those findings.25 Two observational studies had previously found an increased risk of HIV acquisition among DMPA users directly compared with NET-EN users,2 an association not observed in a newly identified study.25

istically significant association; a newly identified seventh study reported an estimate consistent with those findings.25 Two observational studies had previously found an increased risk of HIV acquisition among DMPA users directly compared with NET-EN users,2 an association not observed in a newly identified study.25 The assessment of oral contraceptive use and HIV risk was also based only on evidence from observational studies. The overall body of evidence continues to suggest no increased risk of HIV among oral contraceptive users; the only new evidence was an updated estimate that was nearly identical to the previously included estimate from the same study.28 The previous review did not identify any consistent factors that modify the relationship between hormonal contraceptive use and HIV acquisition.2 Neither age nor HSV-2 status were modifiers in the ECHO trial.5 Two studies assessed for the first time whether BV is an effect modifier,27 28 with conflicting results. Differences between the two studies included the populations studied (women in serodiscordant couples27 and female sex workers28), BV diagnostic techniques, measurement of potential confounders, and overall findings (ie, one study observed an association between hormonal contraception and HIV but did not find effect modification by BV,28 while the study that did not find an overall association suggested effect modification).27

male sex workers28), BV diagnostic techniques, measurement of potential confounders, and overall findings (ie, one study observed an association between hormonal contraception and HIV but did not find effect modification by BV,28 while the study that did not find an overall association suggested effect modification).27 Although a new RCT has provided high-quality evidence about the incidence of HIV among users of two hormonal contraceptive methods, most of the body of evidence remains based on observational studies. The limitations of the observational data on this topic have been described in detail.2 32 33 Key concerns are unmeasured and residual confounding, as well as unclear exposure and outcome measurement, and lack of time-varying measures in analysis. We continue to urge caution when interpreting the meta-analysis results from observational studies of DMPA and HIV acquisition, given important uncertainties about each study included in the meta-analysis and the potential for spurious precision. For RCTs, proper randomisation should equalise confounding factors among groups at baseline. However, unblinded follow-up of participants provides opportunities for confounding and other biases to occur, for example through changes in contraceptives used or differential condom use by contraceptive group because of knowledge of the initial contraceptive allocated. Such changes need to be captured and allowed for in time-varying analyses, as was done in the ECHO trial.5 Although the ECHO trial was not designed to answer the question of whether any of the contraceptive methods assessed increased the risk of HIV acquisition when compared with no contraceptive method use, the high rates of HIV infection observed among the study groups were similar to the underlying incidence used when designing the trial. This finding provides indirect evidence of no increased risk of HIV acquisition among DMPA-IM, Cu-IUD or LNG implant users, compared with women using no contraception. Finally, unanswered questions remain about hormonal contraceptive methods for which there is currently no evidence, including DMPA-SC, etonogestrel implants, LNG-IUDs and non-oral forms of combined hormonal methods.

HIV acquisition among DMPA-IM, Cu-IUD or LNG implant users, compared with women using no contraception. Finally, unanswered questions remain about hormonal contraceptive methods for which there is currently no evidence, including DMPA-SC, etonogestrel implants, LNG-IUDs and non-oral forms of combined hormonal methods. An earlier version of this review, which included observational studies and RCTs published before the ECHO trial, was considered during a 2019 WHO Guideline Development Group Meeting to assess recommendations on contraception for women at high risk of HIV acquisition. WHO’s updated guidance states that women at high risk of acquiring HIV infection are eligible to use all hormonal contraceptive methods and IUDs without restriction.34 Conclusions New high-quality evidence from an RCT (the ECHO trial) did not find an increased risk of HIV among women using DMPA compared with LNG implant or Cu-IUD users.5 These RCT data provided high-quality evidence about whether DMPA use increases HIV risk, as they are less likely to be confounded than findings from observational studies. As concluded in previous reviews, the body of evidence suggests no increased risk of HIV acquisition associated with use of other hormonal contraceptive methods. Additional educational resources WHO Guidance on Hormonal Contraceptive Eligibility for Women at High Risk of HIV https://www.who.int/reproductivehealth/publications/contraceptive-eligibility-women-at-high-risk-of-HIV/en/

Conclusions New high-quality evidence from an RCT (the ECHO trial) did not find an increased risk of HIV among women using DMPA compared with LNG implant or Cu-IUD users.5 These RCT data provided high-quality evidence about whether DMPA use increases HIV risk, as they are less likely to be confounded than findings from observational studies. As concluded in previous reviews, the body of evidence suggests no increased risk of HIV acquisition associated with use of other hormonal contraceptive methods. Additional educational resources WHO Guidance on Hormonal Contraceptive Eligibility for Women at High Risk of HIV https://www.who.int/reproductivehealth/publications/contraceptive-eligibility-women-at-high-risk-of-HIV/en/ WHO Medical Eligibility Criteria for Contraceptive Use https://www.who.int/reproductivehealth/publications/family_planning/Ex-Summ-MEC-5/en/ WHO Selected Practice Recommendations for Contraceptive Use https://www.who.int/reproductivehealth/publications/family_planning/SPR-3/en/ Hannaford PC, Chipato T, Curtis KM. Copper intrauterine device use and HIV acquisition in women: a systematic review. BMJ Sex Reprod Health 2020;46:17–25. The authors would like to acknowledge the tremendous amount of work conducted by Chelsea Polis, PhD, Sharon Phillips, MD, MPH, and Daniel Westreich, PhD, as authors of the 2016 systematic review on which this update is based. They would also like to acknowledge the assistance of Joanna Taliano, MA, MLS, Reference Librarian at CDC, for running the search strategies.

ndous amount of work conducted by Chelsea Polis, PhD, Sharon Phillips, MD, MPH, and Daniel Westreich, PhD, as authors of the 2016 systematic review on which this update is based. They would also like to acknowledge the assistance of Joanna Taliano, MA, MLS, Reference Librarian at CDC, for running the search strategies. Contributors: KC, PH, TC, PS and JK were coauthors on the previous systematic review and contributed to the planning of this updated review. KC, PH and TC conducted the literature search, screening, and risk of bias assessment. MIR conducted the meta-analysis. KC wrote the first draft of the manuscript. All authors contributed to subsequent drafts and approved of the final manuscript. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Disclaimer: The findings and conclusions of this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention, the World Health Organization, or other institutions with which the authors are affiliated. Competing interests: MIR reports personal fees from Merck (contraceptive implant trainer) and Bayer (IUD trainer), outside the submitted work. TC, JK and PS were members of the ECHO trial consortium, and JK and PS were part of the writing group for the ECHO trial results. All of the authors participated in the 2019 WHO Guideline Development Group (GDG) process which assessed recommendations on contraception for women at high risk of HIV.

e the submitted work. TC, JK and PS were members of the ECHO trial consortium, and JK and PS were part of the writing group for the ECHO trial results. All of the authors participated in the 2019 WHO Guideline Development Group (GDG) process which assessed recommendations on contraception for women at high risk of HIV. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages Seven studies have investigated whether copper intrauterine devices (Cu-IUDs) are associated with an increased risk of women acquiring HIV, including a high-quality randomised controlled trial. The collective evidence does not indicate an increased risk of HIV acquisition among users of Cu-IUDs.

Key messages Seven studies have investigated whether copper intrauterine devices (Cu-IUDs) are associated with an increased risk of women acquiring HIV, including a high-quality randomised controlled trial. The collective evidence does not indicate an increased risk of HIV acquisition among users of Cu-IUDs. Introduction For nearly 30 years, inconsistent evidence from epidemiological observational research has been accumulating that hormonal contraceptives (particularly depot medroxyprogesterone acetate (DMPA)-containing injectables) may be associated with an increased risk of women acquiring HIV. In June 2019, the Evidence for Contraceptive Options and HIV Outcomes (ECHO) trial reported HIV incidence rates in users of three common contraceptives: the copper intrauterine device (Cu-IUD), intramuscular DMPA (DMPA-IM) and the levonorgestrel (LNG) implant.1 The ECHO trial sought to address, through randomisation, a key concern with previous observational studies; namely that of possible confounding (the distortion of the true relationship between contraceptive use and HIV acquisition because of a third factor linked to both exposure and outcome), especially in relation to condom use and other sexual behaviours. Cu-IUD users have often been used as a reference group in the observational studies of hormonal contraceptives and HIV risk because of an assumption that Cu-IUDs do not increase the risk. This assumption is supported by a small biological study of genital tract immune cell changes after insertion of a Cu-IUD which did not indicate increased susceptibility to HIV infection.2 Other biological studies have found that copper as a solution3 or filter4 deactivates HIV. It is important to know if Cu-IUDs affect, positively or negatively, the risk of HIV acquisition among users. We report here a systematic review of longitudinal epidemiological studies which have examined the risk of HIV acquisition among users of the Cu-IUD.

found that copper as a solution3 or filter4 deactivates HIV. It is important to know if Cu-IUDs affect, positively or negatively, the risk of HIV acquisition among users. We report here a systematic review of longitudinal epidemiological studies which have examined the risk of HIV acquisition among users of the Cu-IUD. Methods This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.5 We examined two related but slightly different review questions as follows. (1) Among women at risk of HIV (ie, women who are HIV-seronegative at baseline), does use of a Cu-IUD compared with use of another non-hormonal contraceptive method or no contraceptive method increase the risk of HIV acquisition? (2) Among women at risk of HIV (ie, women who are HIV-seronegative at baseline), does use of a Cu-IUD compared with use of a specific hormonal contraceptive method increase the risk of HIV acquisition?

with use of another non-hormonal contraceptive method or no contraceptive method increase the risk of HIV acquisition? (2) Among women at risk of HIV (ie, women who are HIV-seronegative at baseline), does use of a Cu-IUD compared with use of a specific hormonal contraceptive method increase the risk of HIV acquisition? Inclusion and exclusion criteria We included studies of women of reproductive age at risk of HIV infection (ie, seronegative for HIV at baseline) that compared incident, laboratory-confirmed HIV infection among women using a Cu-IUD compared with women using a non-hormonal method (eg, condom, sterilisation, fertility awareness-based methods, withdrawal) or no contraception. We also included studies comparing incident HIV infections among users of Cu-IUDs with another specified hormonal contraceptive (eg, DMPA or other injectable, oral contraceptive (OC), ring, patch or implant) users. We included longitudinal studies (observational studies, randomised controlled trials (RCTs) or meta-analyses); cross-sectional studies, letters, narrative reviews, conference abstracts and other non-primary research reports were excluded. We contacted the chief investigators of studies that did not specify the type of IUD (ie, device was simply referred to throughout as 'IUD') used by participants to ascertain whether the usage related to Cu-IUDs rather than LNG-IUDs or inert IUDs (ie, IUDs with no bioactive component)

ry research reports were excluded. We contacted the chief investigators of studies that did not specify the type of IUD (ie, device was simply referred to throughout as 'IUD') used by participants to ascertain whether the usage related to Cu-IUDs rather than LNG-IUDs or inert IUDs (ie, IUDs with no bioactive component) Search strategy We searched Pubmed, Embase and the Cochrane Library from database inception to 26 June 2019 (online supplementary appendix 1). We hand-searched reference lists of included studies. All authors screened abstracts and full-text articles using Covidence,6 and conflicts were resolved through discussion. Two authors (KC and AT) also screened all the articles included in a companion review of hormonal contraceptive use and risk of HIV acquisition7 to identify any additional studies that included an IUD-only arm. 10.1136/bmjsrh-2019-200512.supp1Supplementary data

Search strategy We searched Pubmed, Embase and the Cochrane Library from database inception to 26 June 2019 (online supplementary appendix 1). We hand-searched reference lists of included studies. All authors screened abstracts and full-text articles using Covidence,6 and conflicts were resolved through discussion. Two authors (KC and AT) also screened all the articles included in a companion review of hormonal contraceptive use and risk of HIV acquisition7 to identify any additional studies that included an IUD-only arm. 10.1136/bmjsrh-2019-200512.supp1Supplementary data Data extraction, quality assessment and data synthesis We used a study quality assessment framework modified from that used for a previous review of hormonal contraceptive use and risk of HIV.8 Briefly, studies that failed to adjust for condom use or had unclear exposure measurement were judged to be “unlikely to inform the primary question”. All other studies could be considered to be either “informative but with important limitations” or “informative with few limitations”, depending on the potential for residual or uncontrolled confounding in the studies (online supplementary appendix 2). We abstracted data from all included studies into an evidence table that included information on the study design, population, exposure and comparison groups, results, strengths and weaknesses, and quality. We synthesised the study results narratively; the small number of point estimates for specific comparisons precluded meta-analysis. For studies that reported risk estimates for a hormonal contraceptive method compared with a Cu-IUD, we report the inverse of the risk estimate so that the Cu-IUD is the exposure and the hormonal contraceptive method is the referent group.

mall number of point estimates for specific comparisons precluded meta-analysis. For studies that reported risk estimates for a hormonal contraceptive method compared with a Cu-IUD, we report the inverse of the risk estimate so that the Cu-IUD is the exposure and the hormonal contraceptive method is the referent group. 10.1136/bmjsrh-2019-200512.supp2Supplementary data Patient and public involvement As this analysis was based on published data, patients and the public were not involved in the design of the study.

mall number of point estimates for specific comparisons precluded meta-analysis. For studies that reported risk estimates for a hormonal contraceptive method compared with a Cu-IUD, we report the inverse of the risk estimate so that the Cu-IUD is the exposure and the hormonal contraceptive method is the referent group. 10.1136/bmjsrh-2019-200512.supp2Supplementary data Patient and public involvement As this analysis was based on published data, patients and the public were not involved in the design of the study. Results Description of included studies We screened 2494 abstracts and 29 full-text reports (figure 1). We excluded 23 full-text reports – one reported data subsequently updated in a later report, 10 were letters or conference abstracts, and 12 had an irrelevant exposure, outcome, study design or participant group. An 'in press' full report of a new study that has now been published9 was identified from a conference abstract.10 Seven studies met our inclusion criteria (table 1):1 9 11–15 five cohort studies9 11–14 and two RCTs.1 15 The studies were conducted in six countries: Eswatini, Italy, Kenya (three studies), Tanzania, South Africa (three studies) and Zambia. Between 343 and 7830 women were enrolled into the different studies, conducted in a variety of clinical settings: hospital or outpatient clinics serving HIV-infected people, family planning clinics, a clinic for sex workers, hospital-based pregnancy termination services or clinical trial sites for a novel vaginal ring containing dapivirine. Three studies did not specify the type of IUDs that were used.11 12 14 Two of these studies were conducted in Kenya and the study authors confirmed that Cu-IUDs were the type of IUD available during the study.12 14 The third study included two IUD users in Italy;11 the study authors were unable to provide information on the type of IUD used. One study was deemed to be “informative with few limitations”,1 two “informative but with important limitations”,9 14 and four “unlikely to inform the primary question”.11–13 15

e study.12 14 The third study included two IUD users in Italy;11 the study authors were unable to provide information on the type of IUD used. One study was deemed to be “informative with few limitations”,1 two “informative but with important limitations”,9 14 and four “unlikely to inform the primary question”.11–13 15 Figure 1 Identification of included studies. Table 1 Summary of evidence on intrauterine device use and risk of HIV acquisition in women

e study.12 14 The third study included two IUD users in Italy;11 the study authors were unable to provide information on the type of IUD used. One study was deemed to be “informative with few limitations”,1 two “informative but with important limitations”,9 14 and four “unlikely to inform the primary question”.11–13 15 Figure 1 Identification of included studies. Table 1 Summary of evidence on intrauterine device use and risk of HIV acquisition in women Author, year, location, funding Study design, purpose, period of data collection Population Number seroconverted/number analysed, number of seroconverters by exposure group, overall HIV incidence Exposure Comparison Results (point estimate, CI) Strengths Weaknesses Quality Saracco 1993,11 Italy Ministry of Health and National Research Council of Italy Cohort; determine incidence and risk factors for male to female sexual HIV transmission, 1987–1991 343 seronegative women with HIV-infected male partners, recruited from 16 hospital and outpatient clinic locations; 529.6 person-years follow-up 19/343 after mean 14.1 months 1/1 IUD and no condom use seroconverted 0/1 IUD and condom use seroconverted 0/22 OC users seroconverted (11 reported always using condom) 3.6 per 100 person-years IUD (type not specified) No clear comparison group; seroconversions reported for OC users No comparative estimates given Serodiscordant couples Study interval 6 months No comparative estimates provided and no clear comparison group No adjustment for condom use No time-varying analyses Follow-up rates not reported or unclear Unlikely to inform the primary question Sinei 1996,12 Kenya US NIH and USAID, WHO, Ortho Pharmaceutical Corp.

given Serodiscordant couples Study interval 6 months No comparative estimates provided and no clear comparison group No adjustment for condom use No time-varying analyses Follow-up rates not reported or unclear Unlikely to inform the primary question Sinei 1996,12 Kenya US NIH and USAID, WHO, Ortho Pharmaceutical Corp. Cohort; pilot study to examine relationship between contraceptive methods and HIV transmission, 1990–1992 1537 seronegative women recruited from a family planning clinic 16 seroconversions by 12 months: IUD users: 7 DMPA users: 2 OC users: 5 12-month cumulative incidence rates (95% CI) per 100: Overall: 2.1 (95% CI 1.1 to 3.2): IUD: 1.66 (95% CI 0.4 to 2.9) DMPA: 1.92 (95% CI 0.0 to 4.6) OC: 4.45 (95% CI 0.5 to 8.4) IUD (type not specified) No clear comparison group, but separate seroconversion rates provided for DMPA and OC users No comparative estimates given Study interval of 3 months No comparative estimates provided and no clear comparison group No adjustment for condom use No time-varying analyses Low follow-up rates (29% overall) Unlikely to inform the primary question Kapiga 1998,13 Tanzania Rockefeller Foundation, US NIH Cohort, to study the incidence of HIV associated with contraceptive methods, 1992–1995 1370 (of an original cohort of 2471) women recruited from family planning clinics; 2236 person-years follow-up 75/1370 seroconverted 3.4 per 100 person-years (95% CI 2.6 to 4.1) IUD: 8/162 IUD (copper T loop specified); ever-use during the follow-up period No IUD use (included those using hormonal methods, non-hormonal methods or no method) IUD use vs no use: adjRR 0.80 (95% CI 0.38 to 1.69) Duration of IUD use 1–12 months vs 0 months: adjRR 1.22 (95% CI 0.44 to 3.37)

years (95% CI 2.6 to 4.1) IUD: 8/162 IUD (copper T loop specified); ever-use during the follow-up period No IUD use (included those using hormonal methods, non-hormonal methods or no method) IUD use vs no use: adjRR 0.80 (95% CI 0.38 to 1.69) Duration of IUD use 1–12 months vs 0 months: adjRR 1.22 (95% CI 0.44 to 3.37) >13 months vs 0 months: adjRR 0.59 (95% CI 0.21 to 1.66)   No adjustment for condom use in the main model; authors state in text that results were not materially altered when excluding condom users or adjusting for condom use No time-varying analysis of IUD exposure (exposure measured as ever-use during the follow-up period) Comparison group included all non-IUD users Study interval unclear; HIV testing only occurred at baseline and final interview 44.5% loss to follow-up Unlikely to inform the primary question Lavreys 2004,14 Kenya US NIH Cohort, to assess the association between contraceptive use and HIV acquisition, 1993–2003 1498 seronegative women, recruited from a clinic for sex workers, Mombasa, 1272 returned for follow-up; providing 2931 person-years follow-up; 15 428 follow-up visits 248/1272 Number seroconverted by group not reported IUD (type not specified) No contraceptive method or tubal ligation adjHR 1.1 (95% CI 0.4 to 3.0) Adjusted for condom use Time-varying analysis of contraception exposure and confounders Clearly defined comparison group Study interval 1 month Follow-up rates not reported or unclear Potential for residual confounding Informative but with important limitations Hofmeyr 2017,15 South Africa Effective Care Research Unit, East London, South Africa RCT; to compare rates of incident HIV among women using progestogen-only injectables and those using Cu-IUDs; 2009–2012 2493 enrolled, 1246 in injectable arm and 1247 in IUD arm, recruited from women attending pregnancy termination services at two hospitals in South Africa ITT analysis 20/656 (3%) injectable arm 22/634 (3.5%) IUD arm Median follow-up time: 19–20 months Cu-IUD DMPA and NET-EN together and separately ITT analysis Injectables vs IUD RR 0.88 (95% CI 0.48 to 1.59) Separate ITT results for different injectables not given Per protocol analysis Injectables vs IUD RR 0.94 (95% CI 0.52 to 1.71) DMPA vs IUD RR 1.01 (95% CI 0.55 to 1.86) NET-EN vs IUD RR 0.58 (95% CI 0.14 to 2.42) Appropriate randomisation procedures, with good concealment of allocations at point of assignment Clearly defined comparison group No measurement or adjustment for condom use (or other potential confounder

R 0.94 (95% CI 0.52 to 1.71) DMPA vs IUD RR 1.01 (95% CI 0.55 to 1.86) NET-EN vs IUD RR 0.58 (95% CI 0.14 to 2.42) Appropriate randomisation procedures, with good concealment of allocations at point of assignment Clearly defined comparison group No measurement or adjustment for condom use (or other potential confounder s) No information on discontinuation or switching, or time-varying analysis of contraceptive use 19–20 month median interval for HIV testing 37% of follow-up HIV test results by self-report Unlikely to inform the primary question Palanee-Phillips 2019,9 South Africa NIH Cohort; from RCT that examined effectiveness of dapivirine ring to prevent HIV; 2012–2015 1136 HIV-seronegative women from South African clinical study sites; 1771 person-years follow-up 95/1136 seroconverted Median follow-up: 1.6 years HIV incidence/100 woman-years: Overall: 5.6 DMPA: 5.79 NET-EN: 6.22 Implant: 1.93 Cu-IUD: 4.48 Cu-IUD DMPA NET-EN Implant Cu-IUD vs DMPA:adjHR 1.10 (95% CI 0.53 to 2.27)* Cu-IUD vs NET-EN: adjHR 0.98 (95% CI 0.47 to 2.04)* Cu-IUD vs implant: adjHR 2.17 (95% CI 0.59 to 7.69)* Cu-IUD vs all three hormonal methods together: adjHR 1.11 (95% CI 0.57 to 2.22)* Adjusted for condom use. Time-varying analysis of contraception exposure, condom use, and other relevant confounders Clearly defined comparison groups Monthly intersurvey interval No information on follow-up time or attrition by study group Potential for residual/ unmeasured confounding Informative but with important limitations Evidence for Contraceptive Options and HIV Outcomes (ECHO) Trial Consortium, 2019,1

r relevant confounders Clearly defined comparison groups Monthly intersurvey interval No information on follow-up time or attrition by study group Potential for residual/ unmeasured confounding Informative but with important limitations Evidence for Contraceptive Options and HIV Outcomes (ECHO) Trial Consortium, 2019,1 12 sites in Eswatini, Kenya, South Africa, Zambia Bill & Melinda Gates Foundation, USAID and PEPFAR, Swedish International Development Cooperation Agency, South African Medical Research Council, UNFPA, Government of South Africa RCT, 2015–2019 7830 HIV-seronegative women seeking contraception (7829 randomised) 397/7715 seroconverted HIV incidence/100 woman-years Overall: 3.81 DMPA-IM: 4.19 LNG implant: 3.31 Cu-IUD: 3.94 Follow-up time: up to 18 months Cu-IUD DMPA-IM LNG-implant ITT analysis, Cu-IUD vs DMPA-IM: HR 0.96 (96% CI 0.75 to 1.22)* Cu-IUD vs LNG implant: HR1.18 (96% CI 0.91 to 1.53) Continuous use analysis, Cu-IUD vs DMPA-IM: adjHR 0.91 (96% CI 0.69 to 1.19)* Cu-IUD vs LNG implant: adjHR 1.18 (96% CI 0.90 to 1.55) No effect modification by age or HSV-2 status, and no substantial differences by several other factors Publication of full trial protocol16

D vs LNG implant: HR1.18 (96% CI 0.91 to 1.53) Continuous use analysis, Cu-IUD vs DMPA-IM: adjHR 0.91 (96% CI 0.69 to 1.19)* Cu-IUD vs LNG implant: adjHR 1.18 (96% CI 0.90 to 1.55) No effect modification by age or HSV-2 status, and no substantial differences by several other factors Publication of full trial protocol16 Clear randomisation and allocation procedures Good adherence to allocated treatments (99% uptake; continuation 92% of follow-up time), with limited discontinuation or change of contraceptive method Patients and clinicians not blinded; however, study team made concerted efforts to not provide different information/counselling to women in DMPA-IM group vs other groups Informative with few limitations *Estimates are for the inverse associations of those reported by the authors. adjHR, adjusted hazard ratio; adjRR, adjusted relative risk; CI, confidence interval; Cu, copper; DMPA, depot medroxyprogesterone acetate; HIV, human immunodeficiency virus;HSV, herpes simplex virus; IM, intramuscular; ITT, intention-to-treat;IUD, intrauterine device; LNG, levonorgestrel; NET-EN, norethisterone ethanate; OC, oral contraceptive;PEPFAR, President's Emergency Plan for AIDS Relief; RCT, randomised controlled trial; UNFPA, United Nations Population Fund; USAID, United States Agency for International Development; US NIH, United States National Institutes of Health; WHO, World Health Organization.

EN, norethisterone ethanate; OC, oral contraceptive;PEPFAR, President's Emergency Plan for AIDS Relief; RCT, randomised controlled trial; UNFPA, United Nations Population Fund; USAID, United States Agency for International Development; US NIH, United States National Institutes of Health; WHO, World Health Organization. Four studies11–14 addressed our first review question, namely among women at risk of HIV (ie, women who are HIV-seronegative at baseline), does use of a Cu-IUD compared with use of another non-hormonal contraceptive method or no contraceptive method increase the risk of HIV acquisition? The only study classified as “informative but with important limitations” reported on 10-year follow-up of 1498 seronegative women recruited from a clinic for female sex workers in Mombasa, Kenya conducted between 1993 and 2003.14 A total of 15 428 follow-up visits accumulated; median number 6 (interquartile range, IQR 2–15), median interval between visits 35 (IQR 28–55) days. Thirty-one women in the study used an IUD. Overall, 248 women seroconverted, although the number in each contraceptive group was not reported. Multivariate Cox proportional hazards models compared time-varying exposure to an IUD with no contraception or tubal ligation, with adjustments including number of sexual partners each week, condom use, vaginal douching with soap products, and presence of a sexually transmitted infection or other genital infection before acquisition of HIV. The adjusted hazards ratio (adjHR) for IUD use compared with no contraception or tubal ligation was 1.1 (95% confidence interval, CI 0.4 to 3.0).

al partners each week, condom use, vaginal douching with soap products, and presence of a sexually transmitted infection or other genital infection before acquisition of HIV. The adjusted hazards ratio (adjHR) for IUD use compared with no contraception or tubal ligation was 1.1 (95% confidence interval, CI 0.4 to 3.0). The second study followed 343 women living in Italy with an HIV-infected male partner.11 This longitudinal study conducted between 1987 and 1991 followed women enrolled in an earlier cross-sectional study that had advised women with an IUD to discontinue use because the baseline data suggested a three-fold increased risk of HIV prevalence among IUD users.16 Despite this counselling, two women continued to use an IUD; the rest of the women did not use IUDs. One of the IUD users, who did not use condoms, seroconverted during longitudinal observation.11 The very small number of exposed women, and long interval (at least 6 months) between study visits, contributed to an assessment that the study was “unlikely to inform the primary question”.

st of the women did not use IUDs. One of the IUD users, who did not use condoms, seroconverted during longitudinal observation.11 The very small number of exposed women, and long interval (at least 6 months) between study visits, contributed to an assessment that the study was “unlikely to inform the primary question”. In the third longitudinal study, 1537 seronegative women attending a family planning clinic in Nairobi, Kenya were followed every 3 months between 1990 and 1992 as a pilot to test the feasibility of a larger study of contraception and HIV acquisition in women.12 Sixteen women in the entire cohort, seven of whom were IUD users, seroconverted within 12 months of study entry. The cumulative life table HIV incidence rate at 12 months per 100 IUD users at admission was 1.66 per 100 (95% CI 0.4 to 2.9) compared with 1.92 per 100 (95% CI 0.0 to 4.6) among DMPA users and 4.45 per 100 (95% CI 0.5 to 8.4) in OC users. This study was deemed “unlikely to inform the primary question”, partly because of the failure to report time-varying analyses of contraceptive exposure or relative risk estimates.

r 100 (95% CI 0.4 to 2.9) compared with 1.92 per 100 (95% CI 0.0 to 4.6) among DMPA users and 4.45 per 100 (95% CI 0.5 to 8.4) in OC users. This study was deemed “unlikely to inform the primary question”, partly because of the failure to report time-varying analyses of contraceptive exposure or relative risk estimates. Another cohort study of women attending three family planning clinics in Dar es Salaam, Tanzania followed 1370 seronegative service users between 1992 and 1995.13 Few women were using a Cu-IUD at baseline (4.1%), compared with OCs (63.4%), DMPA (4.2%), other methods (8.0%) and no methods (20.3%). Study follow-up visits coincided with routine family planning visits, with the length of interval between visits undefined. Limited resources meant that blood for HIV testing could not be collected at every follow-up visit. Eight of the 75 women who seroconverted had ever used a Cu-IUD during follow-up, giving an adjusted relative risk (adjRR) of 0.80 (95% CI 0.38 to 1.69) compared with never-users of a Cu-IUD. There was no relationship with duration of Cu-IUD use. The adjustments included number of sexual partners but not condom use, although the authors reported that inclusion of ever-use of condom in the model did not materially change the results. The ill-defined follow-up interval, limited HIV testing, and failure to use time-varying exposure information for the risk estimates meant that this study was judged to be “unlikely to inform the primary question”.

authors reported that inclusion of ever-use of condom in the model did not materially change the results. The ill-defined follow-up interval, limited HIV testing, and failure to use time-varying exposure information for the risk estimates meant that this study was judged to be “unlikely to inform the primary question”. Three studies1 9 15 addressed our second review question, namely among women at risk of HIV (ie, women who are HIV-seronegative at baseline), does use of a Cu-IUD compared with use of a specific hormonal contraceptive method increase the risk of HIV acquisition? One study was judged to be “informative with few limitations”.1 The ECHO trial was an RCT specifically designed to compare HIV incidence among women using Cu-IUDs, DMPA-IM or LNG implants; a comparison group of women not using any contraception was not included. It was conducted in 12 sites in Eswatini, Kenya, South Africa and Zambia, recruiting HIV-seronegative women aged 16–35 years and seeking contraception, with follow-up for up to 18 months. After randomisation to one of the three contraceptive methods, study visits occurred every 3 months and included HIV testing, contraceptive counselling, behavioural assessment, and HIV prevention services. Outcome assessment for HIV was conducted without knowledge of the contraception assignment, and the study worked to ensure that information and counselling provided did not differ between the study groups. Ninety-nine percent of women accepted their assigned method. Participants used their assigned method for most (92%) of the 10 409 woman-years of follow-up. More than 91% of participants attended each scheduled follow-up visit and 99% had at least one follow-up HIV test. The overall incidence of HIV was 3.81 per 100 woman-years, with no statistically significant differences observed in rates between the three contraceptive groups. The main modified intention-to–treat analysis found a HR 0.96 (96% CI 0.75 to 1.22) for the Cu-IUD compared to DMPA-IM (note that this estimate is the inverse of that reported by the authors) and HR 1.18 (96% CI 0.91 to 1.53) for Cu-IUD compared with the LNG implant. A prespecified sensitivity analysis of participants with continuous use of the contraceptive method assigned, adjusting for baseline and time-varying covariates including condom use, yielded similar results.

rse of that reported by the authors) and HR 1.18 (96% CI 0.91 to 1.53) for Cu-IUD compared with the LNG implant. A prespecified sensitivity analysis of participants with continuous use of the contraceptive method assigned, adjusting for baseline and time-varying covariates including condom use, yielded similar results. Strengths of the ECHO trial included robust randomisation and allocation procedures, high rates of follow-up and continuation with assigned contraceptive methods, assessment of condom use and sexual behaviour as possible confounders, and appropriate intention-to-treat and prespecified sensitivity analyses. Although women and providers of services in the ECHO trial could not be blinded to contraceptive group allocation, HIV testing was conducted without knowledge of allocation, and there was no evidence that this led to the different groups of women acting, or being managed, differently with respect to important issues such as HIV prevention counselling.

ervices in the ECHO trial could not be blinded to contraceptive group allocation, HIV testing was conducted without knowledge of allocation, and there was no evidence that this led to the different groups of women acting, or being managed, differently with respect to important issues such as HIV prevention counselling. A cohort study, judged to be “informative but with important limitations”, included 1136 HIV-seronegative women from South African sites originally recruited for a double-blind, placebo-controlled RCT of a monthly vaginal ring containing dapivirine for HIV-1 prevention.9 Trial participants attended monthly visits between 2012 and 2015, where standardised information about sexual behaviour and contraceptive use was collected, and pregnancy and HIV infection status were tested. Cu-IUD use comprised 15.1% of the 1771 women-years of follow-up included in the analysis, with DMPA use 50.7%, norethisterone enanthate (NET-EN) use 25.4% and implant use 8.8%. A total of 95 women seroconverted. The adjusted Cox proportional hazards model included time-varying condom use and number of sexual partners. The Cu-IUD was not associated with a significantly altered risk of HIV acquisition when compared with any of the hormonal contraceptive methods (note that these estimates are the inverse of those reported by the authors): adjHR compared with DMPA 1.10 (95% CI 0.53 to 2.27); NET-EN 0.98 (95 % CI 0.47 to 2.04); implant 2.17 (95% CI 0.59 to 7.69); all three hormonal methods together 1.11 (95% CI 0.57 to 2.22).

mpared with any of the hormonal contraceptive methods (note that these estimates are the inverse of those reported by the authors): adjHR compared with DMPA 1.10 (95% CI 0.53 to 2.27); NET-EN 0.98 (95 % CI 0.47 to 2.04); implant 2.17 (95% CI 0.59 to 7.69); all three hormonal methods together 1.11 (95% CI 0.57 to 2.22). The third study was a pragmatic, open-label, parallel-arm RCT conducted at termination of pregnancy services within two hospitals in South Africa between 2009 and 2012.15 Participants were randomised to receive an injectable progestogen-only contraceptive (mostly DMPA, but sometimes NET-EN at the provider's discretion) or a Cu-IUD, and were followed up every 3 months. Forty-two women seroconverted, 22 of those allocated to use a Cu-IUD. The intention-to-treat analysis included 656 injectable users and 634 Cu-IUD users with a follow-up HIV test. The intention-to-treat unadjusted RR between injectable and Cu-IUD users was 0.88 (95% CI 0.48 to 1.59); per-protocol unadjusted RR 0.94 (95% CI 0.52 to 1.71). Separate per-protocol unadjusted risk estimates for the two injectables were: DMPA 1.01 (95% CI 0.55 to 1.86) and NET-EN 0.58 (95% CI 0.14 to 2.42); each compared with Cu-IUD users. The trial was considered “unlikely to inform the primary question” as it did not report on switching between groups after study entry or discontinuation rates, nor did it provide adjusted risk estimates.

tables were: DMPA 1.01 (95% CI 0.55 to 1.86) and NET-EN 0.58 (95% CI 0.14 to 2.42); each compared with Cu-IUD users. The trial was considered “unlikely to inform the primary question” as it did not report on switching between groups after study entry or discontinuation rates, nor did it provide adjusted risk estimates. IUD use versus non-hormonal contraceptive use in studies considered “informative but with important limitations” or “informative with few limitations” The only “informative but with important limitations” study able to examine whether IUDs are associated with an increased risk of HIV when compared with non-hormonal or no contraception did not find such an effect (figure 2).14 No studies were considered “informative with few limitations” for this question. Figure 2 Intrauterine device use and risk of HIV acquisition in “informative with few limitations” or “informative but with important limitations” studies. *Estimates are the inverse associations of those reported by the authors. adjHR, adjusted hazard ratio; CI, confidence interval; Cu-IUD, copper intrauterine device; DMPA, depot medroxyprogesterone acetate; HR, hazard ratio; ITT, intention-to-treat; LNG, levonorgestrel; NET-EN, norethisterone enanthate

tations” studies. *Estimates are the inverse associations of those reported by the authors. adjHR, adjusted hazard ratio; CI, confidence interval; Cu-IUD, copper intrauterine device; DMPA, depot medroxyprogesterone acetate; HR, hazard ratio; ITT, intention-to-treat; LNG, levonorgestrel; NET-EN, norethisterone enanthate IUD use versus a specified hormonal contraceptive use in studies considered “informative but with important limitations” or “informative with few limitations” One “informative with few limitations” study provided high-quality evidence that Cu-IUD users had a similar risk of acquiring HIV to that seen in users of DMPA-IM or the LNG implant (figure 2).1 This evidence was consistent with another “informative but with important limitations” study that found no evidence of differences in HIV risk between IUD users and users of three hormonal contraceptives, regardless of whether each hormonal method was examined separately or together.9

IM or the LNG implant (figure 2).1 This evidence was consistent with another “informative but with important limitations” study that found no evidence of differences in HIV risk between IUD users and users of three hormonal contraceptives, regardless of whether each hormonal method was examined separately or together.9 Discussion This systematic search of the literature from bibliographic database inception to 26 June 2019 identified five cohort studies and two RCTs that met our inclusion criteria; one was deemed sufficiently robust to meet our “informative with few limitations” criteria,1 and two were deemed “informative but with important limitations”.9 14 None of these studies indicated an increased risk of HIV acquisition among IUD users, irrespective of whether the comparison group included non-hormonal or no contraceptive users, or users of a specific hormonal contraceptive method. None of the lower-quality “unlikely to inform the primary question” studies11–13 15 provided evidence inconsistent with the more robust studies. Our search identified a large number of potential papers and we cross-checked the reference lists of included studies. It is unlikely, therefore, that we missed additional published data that could inform our research questions. We excluded cross-sectional studies since it is impossible to determine in such studies the time relationship between contraceptive use and HIV infection acquisition. While some studies did not specify IUD type, given the countries and years studied, the IUDs were most likely copper-bearing.

orm our research questions. We excluded cross-sectional studies since it is impossible to determine in such studies the time relationship between contraceptive use and HIV infection acquisition. While some studies did not specify IUD type, given the countries and years studied, the IUDs were most likely copper-bearing. The ECHO trial was not designed to determine whether Cu-IUD use increases the risk of HIV acquisition compared with using no contraception.1 Instead, it provided direct, high-quality evidence of an absence of increased risk of acquiring HIV among Cu-IUD users, compared with women using DMPA-IM injectables or LNG implants. There was no evidence that the different contraceptive groups in the ECHO trial acted, or were managed, differently with respect to key issues such as HIV prevention counselling. Furthermore, secondary sensitivity analyses controlled for important time-varying confounders such as condomless sex. It is highly unlikely, therefore, that the ECHO trial results were affected by residual confounding (a key concern when interpreting observational studies). The high incidence of HIV in the ECHO trial, and low losses to follow-up, meant that it had sufficient statistical power to detect a 30% increase in risk of HIV between contraceptive groups assessed. Thus, it addressed another important limitation of previous observational studies, namely low power because of limited IUD usage or small number of women seroconverting.

low losses to follow-up, meant that it had sufficient statistical power to detect a 30% increase in risk of HIV between contraceptive groups assessed. Thus, it addressed another important limitation of previous observational studies, namely low power because of limited IUD usage or small number of women seroconverting. Information from an early draft of this review, which considered only observational studies and RCTs published before the ECHO trial, was presented at a 2019 WHO Guideline Development Group meeting to assess recommendations on contraception for women at high risk of HIV. 17 Conclusion The collective evidence from this review, which now includes information from a large high-quality RCT, does not indicate evidence of an increased risk of HIV acquisition among users of Cu-IUDs. Additional educational resources WHO Contraceptive Eligibility for Women at High Risk of HIV https://www.who.int/reproductivehealth/publications/contraceptive-eligibility-women-at-high-risk-of-HIV/en/ WHO Medical Eligibility Criteria for Contraceptive Use https://www.who.int/reproductivehealth/publications/family_planning/Ex-Summ-MEC-5/en/ WHO Selected Practice Recommendations for Contraceptive Use https://www.who.int/reproductivehealth/publications/family_planning/SPR-3/en/ Curtis KM, Hannaford PC, Rodriguez MI, et al. Hormonal contraception and HIV acquisition among women: an updated systematic review. BMJ Sex Reprod Health 2020;46:8–16 The authors thank Joanna Taliano, MA MLS, Reference Librarian at CDC for running the search strategies.

WHO Selected Practice Recommendations for Contraceptive Use https://www.who.int/reproductivehealth/publications/family_planning/SPR-3/en/ Curtis KM, Hannaford PC, Rodriguez MI, et al. Hormonal contraception and HIV acquisition among women: an updated systematic review. BMJ Sex Reprod Health 2020;46:8–16 The authors thank Joanna Taliano, MA MLS, Reference Librarian at CDC for running the search strategies. Contributors: PH, AT, TC and KC contributed to the planning of this review. PH, AT, TC and KC conducted the literature search, screening, and risk of bias assessment. PH wrote the first draft of the manuscript. All authors contributed to subsequent drafts and approved the final manuscript. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Disclaimer: The findings and conclusions of this report are those of the authors and do not necessarily represent the official postion of the Centers for Disease Control and Prevention, the World Health Organization, or other institutions with which the authors are affiliated. Competing interests: TC was a member of the ECHO trial consortium. PCH, KMC, TC participated in the 2019 WHO Guideline Development Group (GDG) process which assessed recommendations on contraception for women at high risk of HIV. Patient consent for publication: Not required. Provenance and peer review: Not commissioned; externally peer reviewed.