Browse the corpus

enrollment goal (target of 300 patients and actual recruitment of 810 patients) was a testament to the effectiveness of including a social media recruitment strategy. In our study, we believe this protocol promoted a deep sense of patient ownership in the study process and dramatically limited attrition (n = 35; 3.3%). Limitations This study is not without limitations. Because of the nonrandom treatment assignment, unmeasured confounding variables may have affected the surgical outcomes. Additionally, given the individualized nature of surgical therapy, issues of generalizability exist in our results. CTR requires special training, experience, and institutional infrastructure, which limits the generalizability of our findings. The degree to which outcomes in our cohort for CTR are transferable between centers requires further study. Despite its advantages, ERMT was performed at only 1 institution and involved several intraoperative techniques that differed substantially from how most surgeons endoscopically treat iSGS. In addition, the postoperative medical regimen was complex. These factors may affect the generalizability of the results.

r study. Despite its advantages, ERMT was performed at only 1 institution and involved several intraoperative techniques that differed substantially from how most surgeons endoscopically treat iSGS. In addition, the postoperative medical regimen was complex. These factors may affect the generalizability of the results. Conclusions We leveraged an engaged patient community on social media and collected patient-generated health data to study the natural history and outcomes of a rare airway disease. Our approach allowed nuanced comparison of the effectiveness of surgical treatments for iSGS. The most popular approach (ED) was associated with higher rates of recurrence compared with alternative treatments. Cricotracheal resection offered the most durable results but showed the greatest perioperative risk and worst long-term voice outcomes. Endoscopic resection with adjuvant medical therapy was associated with better disease control compared with ED, with minimal association with voicing. These results may be used to inform individual patient treatment decision-making and show the feasibility and effectiveness of integrating social media–based recruitment and patient-generated health data to drive the study of iSGS and other rare diseases. Supplement. eFigure. Study Patient Flow eMethods. Disease Specific Data, Multiple Imputation, and Propensity Score Matching eTable 1. Description of Propensity Score Matching Model Covariates in the Comparison (ED) and Treated (ERMT) Groups eTable 2. Propensity Score Matching: Cox Proportional Hazards Model eReferences. Click here for additional data file.

Introduction Tinnitus is one of the most prevalent and distressing symptoms associated with hearing loss. It is defined by the perception of sounds despite their physical absence (the phantom perception of sounds).1 This auditory condition affects more than 15% of the population worldwide (an estimated 70 million people in Europe). In 1 to 2 of 10 people, tinnitus becomes a chronic bothersome and incapacitating symptom,2,3 with highly unmet clinical needs.4 The psychological consequences of tinnitus affect the economy by influencing work and sleep and increasing the risk of sick leave and disability pension.5

timated 70 million people in Europe). In 1 to 2 of 10 people, tinnitus becomes a chronic bothersome and incapacitating symptom,2,3 with highly unmet clinical needs.4 The psychological consequences of tinnitus affect the economy by influencing work and sleep and increasing the risk of sick leave and disability pension.5 Tinnitus is a complex and heterogeneous auditory dysfunction with numerous causes and phenotypes.6 It is frequently associated with hearing problems and noise exposure,7 but it also occurs in humans with normal hearing.8 The current models stipulate that tinnitus mimics the processes of phantom limb perception,9 whereby the loss of sensory input (most often by sensory deafferentation) leads to compensatory mechanisms in the brain that cause the false sensation of a missing limb or sounds.8 This phenomenon of maladaptive plasticity in the presence of deafferentation appears as a common denominator of most forms of phantom percepts in the absence of sensory stimuli.10 In the context of tinnitus, this translates into greater neuronal activity (central gain) along the auditory pathway.11 Although depression covaries with tinnitus prevalence and severity over time,12 clinically significant anxiety and stress appear as a predominant emotional comorbidity.13 Limbic structures have been implicated in tinnitus in humans14,15 and in animal models,16 suggesting that central regions involved in emotional processing and cognition could contribute to tinnitus17 and more potentially to its severity.

ly significant anxiety and stress appear as a predominant emotional comorbidity.13 Limbic structures have been implicated in tinnitus in humans14,15 and in animal models,16 suggesting that central regions involved in emotional processing and cognition could contribute to tinnitus17 and more potentially to its severity. For decades, it has been widely believed that tinnitus is a consequence of environmental factors (as opposed to genetic factors, reviewed by Vona et al18). One familial aggregations study19 found no obvious correlation in siblings (0.16%); genotype candidate genes in patients with tinnitus failed to reveal positive associations,20,21,22,23,24 and a recent genome-wide association study25 with a small sample size found no significant associations. A twin study26 revealed a heritability of 0.40 based on self-reported tinnitus. It has been proposed that the lack of evidence on a significant association with genetic factors is attributable to the large heterogeneity of tinnitus and that tinnitus should not be considered a single entity but an ensemble of multiple subtypes.27 In support of this hypothesis, Maas et al28 recently found that specific forms of tinnitus had greater heritability in a sex-specific manner. When considering tinnitus perceived in 2 ears (bilateral), heritability reached 0.41 in women and 0.68 in men.28 In contrast, when tinnitus was heard in only 1 ear (unilateral), heritability decreased to near 0.27. However, 3 major limitations appear from these reports: (1) the sparse data could be attributable to tinnitus being self-reported and prevalence varying depending on how the question is formulated29; (2) self reporting tinnitus could be affected by shared-environment mechanisms whereby, for instance, a sibling with tinnitus may influence the awareness of tinnitus in the other or because both siblings live in a noisy environment; and (3) these studies did not consider tinnitus severity, which could also be associated with genetic factors, as is the case for other emotional processing disorders.30

reby, for instance, a sibling with tinnitus may influence the awareness of tinnitus in the other or because both siblings live in a noisy environment; and (3) these studies did not consider tinnitus severity, which could also be associated with genetic factors, as is the case for other emotional processing disorders.30 We addressed these issues by performing an adoption study using national medical registry data (ie, tinnitus has been diagnosed by a physician) to determine whether shared-environment mechanisms are associated with the co-occurrence of tinnitus within a family and thus constitute a bias in the estimates of heritability. Methods Data were collected on adoptees and their biological and adoptive parents from January 1, 1964, to December 31, 2015, to determine the heritability of tinnitus. Informed consent was waived as a requirement by the Ethics Committee at Lund University. Accordingly, all data were provided by Statistics Sweden and the National Board of Health and Welfare for research purposes. Data were coded according to European Union law. This study was approved by the Regional Ethical Review Board of Lund University.

ived as a requirement by the Ethics Committee at Lund University. Accordingly, all data were provided by Statistics Sweden and the National Board of Health and Welfare for research purposes. Data were coded according to European Union law. This study was approved by the Regional Ethical Review Board of Lund University. We used several Swedish nationwide registers as part of our analyses. Statistics Sweden and the National Board of Health and Welfare maintain the registers used in the present study.31,32,33,34 The Swedish personal identity number is issued to all residents in Sweden and was used to connect individual-level data from different registers.35 The personal identity numbers were replaced by Statistics Sweden with serial numbers to preserve anonymity. We used data from the Swedish Multi-Generation Register, the National Patient Register (NPR), the Total Population Register, and Small Area Market Statistics (SAMS).

ividual-level data from different registers.35 The personal identity numbers were replaced by Statistics Sweden with serial numbers to preserve anonymity. We used data from the Swedish Multi-Generation Register, the National Patient Register (NPR), the Total Population Register, and Small Area Market Statistics (SAMS). The Swedish Multi-Generation Register contains data on familial relationships, including adoptions. This register comprises data on index persons registered in Sweden after 1961 and born during and later than 1932.31 The NPR33 contains all hospital discharge diagnoses for all people in Sweden from 1964 to 2015. The hospital discharge register has nationwide coverage since 1987. The NPR also includes the Hospital Outpatient Register, which contains information on diagnoses from all hospital outpatient visits in Sweden between 2001 and 2015. The Primary Healthcare Register, which contains data from 1989 to 2016, was also used.36 The Total Population Register contains data on life events, including birth, death, name change, marital status, family relationships, educational attainment, and migration within Sweden as well as immigration to and emigration from other countries.32 Nearly 100% of births and deaths, 95% of immigrations, and 91% of emigrations are reported to the Total Population Register. Starting in 1991, SAMS data were used to define a municipal subarea and characterize a neighborhood; the code is composed of the county, municipality, and unique SAMS area (9200 in the whole of Sweden).

100% of births and deaths, 95% of immigrations, and 91% of emigrations are reported to the Total Population Register. Starting in 1991, SAMS data were used to define a municipal subarea and characterize a neighborhood; the code is composed of the county, municipality, and unique SAMS area (9200 in the whole of Sweden). Definition of Tinnitus and Comorbidities Patients with tinnitus in the Swedish Hospital Discharge Register (1964-2015), Outpatient Register (2001-2015), and Primary Healthcare Register (1997-2015) were identified by the following International Classification of Diseases (ICD) codes: ICD-7 code 781.32, ICD-8 code 781.31, ICD-9 code 388D, and ICD-10 code H931. There was no primary care code that is specific for tinnitus before 1997. The main and all secondary diagnoses were used. The validity in the Hospital Discharge Register is generally between 85% and 95%.33 Depression, anxiety, and hearing loss were identified by ICD codes any time during the 1964 to 2015 follow-up period using the Swedish Hospital Discharge Register (1964-2015), Outpatient Register (2001-2015), and Primary Healthcare Register (1997-2015). ICD codes are presented in the eTable in the Supplement.

Definition of Tinnitus and Comorbidities Patients with tinnitus in the Swedish Hospital Discharge Register (1964-2015), Outpatient Register (2001-2015), and Primary Healthcare Register (1997-2015) were identified by the following International Classification of Diseases (ICD) codes: ICD-7 code 781.32, ICD-8 code 781.31, ICD-9 code 388D, and ICD-10 code H931. There was no primary care code that is specific for tinnitus before 1997. The main and all secondary diagnoses were used. The validity in the Hospital Discharge Register is generally between 85% and 95%.33 Depression, anxiety, and hearing loss were identified by ICD codes any time during the 1964 to 2015 follow-up period using the Swedish Hospital Discharge Register (1964-2015), Outpatient Register (2001-2015), and Primary Healthcare Register (1997-2015). ICD codes are presented in the eTable in the Supplement. Sample The analyses were based on a data set that encompasses all Swedish-born adoptees (born between 1960 and 1990) and their biological and adoptive parents. Adoptees were excluded from the study if they had died before 16 years of age (ie, exclusion of adoptees with possible severe congenital diseases or confounding sociodemographic factors that cause children to be placed in adoptive homes), had migrated from Sweden before 16 years of age, had died before 1964 (ie, before start of follow-up), or were not linked to at least 1 biological and at least 1 adoptive parent.

doptees with possible severe congenital diseases or confounding sociodemographic factors that cause children to be placed in adoptive homes), had migrated from Sweden before 16 years of age, had died before 1964 (ie, before start of follow-up), or were not linked to at least 1 biological and at least 1 adoptive parent. All adoptees who had cohabited with a biological parent were excluded according to census data (every fifth year from 1960 to 1990) or SAMS data (yearly from 1991). Adoptees who had lived with their biological grandparent, aunt and/or uncle, and sibling or with stepparents together with their biological parent were also excluded. A total of 11 060 Swedish-born adoptees remained in the study after exclusions. They compose the study population in the cohort study. These adoptees could be linked to 19 015 adoptive parents and 17 025 biological parents. After exclusions, we identified 1029 patients (2.2%) with tinnitus among adoptees and their adoptive and biological parents. Of the 1029 cases of tinnitus, 214 were found in adoptees, 371 in biological parents, and 444 in adoptive parents. Of the 1029 patients with tinnitus, 525 (51.0%) were found in the Outpatient Register, 23 (2.2%) in the Hospital Discharge Register, and 481 (46.7%) in the Primary Healthcare Register through ICD codes. Seven tinnitus cases (0.7%) were identified with ICD-8 codes and 9 (0.9%) with ICD-9 codes. No tinnitus case was identified with ICD-7. A total of 1013 tinnitus cases (98.4%) were identified with ICD-10 codes.

.2%) in the Hospital Discharge Register, and 481 (46.7%) in the Primary Healthcare Register through ICD codes. Seven tinnitus cases (0.7%) were identified with ICD-8 codes and 9 (0.9%) with ICD-9 codes. No tinnitus case was identified with ICD-7. A total of 1013 tinnitus cases (98.4%) were identified with ICD-10 codes. Educational attainment was categorized into 4 groups: low (0–9 years), middle (10–11 years), high (≥12 years or more), and unknown. Statistical Analysis We used a cohort design and a case-control approach to study genetic and nongenetic factors associated with tinnitus among adoptees. We conducted 2 main analyses. Odds ratios (ORs) were determined with logistic regression for adoptees with an affected biological parent and for adoptees with an affected adoptive parent. We used a case-control matching method (1:4) for sex, educational attainment, county of birth, and ±1 year for birth year by drawing a sample of tinnitus-affected adoptees as patients and matched control groups of tinnitus-unaffected adoptees.37,38 In the case-control study, we connected both groups to their biological and adoptive parents, and ORs were calculated with conditional logistic regression.

of birth, and ±1 year for birth year by drawing a sample of tinnitus-affected adoptees as patients and matched control groups of tinnitus-unaffected adoptees.37,38 In the case-control study, we connected both groups to their biological and adoptive parents, and ORs were calculated with conditional logistic regression. In the cohort study, logistic regression was used to determine crude (crude [univariate] models for each variable; model 1) and multivariate (adjusted model; model 2) ORs for history of tinnitus in biological or adoptive parents. In the multivariate model (adjusted model 2), we used adoptees’ birth year, sex, educational attainment, and county of birth as covariates in the cohort study. The primary outcome was OR of tinnitus in adoptees with at least 1 affected biological parent compared with adoptees without any affected biological parent. The secondary outcome was OR in adoptees with at least 1 affected adoptive parent compared with adoptees without any affected adoptive parent. An important question in medicine is whether an observed variation in a particular disease is associated with environmental factors or biological factors (nature vs nurture debate). In genetics, heritability summarizes how heritable a disease of interest is, that is, the proportion of variance that emerges because of hereditary factors, especially with reference to the resemblance of offspring and parents.

e is associated with environmental factors or biological factors (nature vs nurture debate). In genetics, heritability summarizes how heritable a disease of interest is, that is, the proportion of variance that emerges because of hereditary factors, especially with reference to the resemblance of offspring and parents. Formally, heritability was defined as a ratio of variances, that is, the proportion of total variance that is associated with variation in additive genetic factors. According to classic quantitative genetics, the heritability of a binary trait (or disease) could be estimated by Falconer regression or with relatives’ tetrachoric correlation by presuming a liability threshold model of the disease in which everyone has a liability to develop the disease but only individuals above a threshold value do so.39,40,41

tive genetics, the heritability of a binary trait (or disease) could be estimated by Falconer regression or with relatives’ tetrachoric correlation by presuming a liability threshold model of the disease in which everyone has a liability to develop the disease but only individuals above a threshold value do so.39,40,41 To evaluate heritability for tinnitus, 2 different methods were used. First, we used Falconer regression, which is based on the liability of the threshold, to obtain heritability in adoptees of the biological parents. The method and its application are described in detail by Falconer and MacKay.40,41 With use of the prevalence rate of the relatives of the biological probands and the controls (ie, biological parents to affected and unaffected adoptees, respectively) from the case-control study, the mean (SE) heritability was calculated. Second, we used the approach described by Frisell et al,39 which used the tetrachoric correlation. This method allowed us to test the sensitivity of the calculated heritability to the assumed prevalence. The tetrachoric correlation is the inferred Pearson correlation from a 2 × 2 table with dichotomous normality being assumed. We used χ2 and Wald tests (in logistic regression). A 2-sided P < .05 was considered to be statistically significant. Statistical analysis was performed with SAS software, version 9.3 (SAS Institute Inc).

choric correlation is the inferred Pearson correlation from a 2 × 2 table with dichotomous normality being assumed. We used χ2 and Wald tests (in logistic regression). A 2-sided P < .05 was considered to be statistically significant. Statistical analysis was performed with SAS software, version 9.3 (SAS Institute Inc). Results We identified 1029 patients (2.2%) with tinnitus (440 [42.8%] male; mean [SD] age, 62 [14] years) in the study population during the study period (1964-2015). Table 1 gives the descriptive statistics of the adopted offspring, their biological parents, and their adoptive parents: age, sex, educational attainment, tinnitus, age at tinnitus diagnosis, and age at end of the study period (death, emigration, or end of the study period on December 31, 2015, whichever came first). The Figure shows the age distribution for Swedish-born (1960-1990) adoptees at first-time diagnosis of tinnitus. The adoptive parents, with a median age of 78 years (interquartile range [IQR], 70-83 years), were older than the biological parents, with a median age of 69 years (IQR, 59-74 years) at the end of the study period. Table 1 reports that the median birth years were 1965 (IQR, 1962-1970) for adoptees, 1942 (IQR, 1936-1947) for biological parents, and 1932 (IQR, 1927-1939) for adoptive parents. Tinnitus was found in 371 of 17 025 biological parents (2.18%) and 444 of 19 015 adoptive parents (2.33%). No statistically significant differences were found between these groups (x2 = 0.99, P = .32). Tinnitus in adoptees, biological parents, and adoptive parents was found to be associated with depression, anxiety, and hearing loss (Table 2). The ORs were 1.89 (95% CI, 1.40-2.55) for depression, 2.19 (95% CI, 1.67-2.88) for anxiety, and 24.38 (95% CI, 18.08-32.88) for hearing loss using a crude model.

.99, P = .32). Tinnitus in adoptees, biological parents, and adoptive parents was found to be associated with depression, anxiety, and hearing loss (Table 2). The ORs were 1.89 (95% CI, 1.40-2.55) for depression, 2.19 (95% CI, 1.67-2.88) for anxiety, and 24.38 (95% CI, 18.08-32.88) for hearing loss using a crude model. Table 1. Characteristics of the Study Population of Swedish-Born Adoptees Between 1960 and 1990 and Their Adoptive and Biological Parentsa Characteristic Adopted Offspring (n = 11 060) Adoptive Parents (n = 19 015) Biological Parents (n = 17 025) Female 5213 (47.13) 9505 (49.99) 10 358 (60.84) Age at end of follow-up, median (IQR), y 50 (44-52) 78 (70-83) 69 (59-74) Birth year Mean (SD) 1967 (7) 1933 (9) 1941 (10) Median (IQR) 1965 (1962-1970) 1932 (1927-1939) 1942 (1936-1947) Range 1960-1990 1888-1972 1898-1974 High educational attainment (≥12 y) 3450 (31.19) 4202 (22.10) 1452 (8.37) Tinnitus cases 214 (1.93) 444 (2.33) 371 (2.18) Female patients with tinnitus 113 (52.80) 244 (54.95) 232 (62.53) Age at tinnitus diagnosis, median (IQR), y 43 (38-47) 71 (63-77) 65 (60-70) Abbreviation: IQR, interquartile range. a Data are presented as number (percentage) of population unless otherwise indicated. The numbers of parents represent unique individuals. Some parents had several biological or adoptive children.

Table 1. Characteristics of the Study Population of Swedish-Born Adoptees Between 1960 and 1990 and Their Adoptive and Biological Parentsa Characteristic Adopted Offspring (n = 11 060) Adoptive Parents (n = 19 015) Biological Parents (n = 17 025) Female 5213 (47.13) 9505 (49.99) 10 358 (60.84) Age at end of follow-up, median (IQR), y 50 (44-52) 78 (70-83) 69 (59-74) Birth year Mean (SD) 1967 (7) 1933 (9) 1941 (10) Median (IQR) 1965 (1962-1970) 1932 (1927-1939) 1942 (1936-1947) Range 1960-1990 1888-1972 1898-1974 High educational attainment (≥12 y) 3450 (31.19) 4202 (22.10) 1452 (8.37) Tinnitus cases 214 (1.93) 444 (2.33) 371 (2.18) Female patients with tinnitus 113 (52.80) 244 (54.95) 232 (62.53) Age at tinnitus diagnosis, median (IQR), y 43 (38-47) 71 (63-77) 65 (60-70) Abbreviation: IQR, interquartile range. a Data are presented as number (percentage) of population unless otherwise indicated. The numbers of parents represent unique individuals. Some parents had several biological or adoptive children. Figure. Age Distribution for Swedish-Born (1960-1990) Adoptees When Tinnitus Was First Diagnosed Table 2. Prevalence of Hearing Loss, Depression, and Anxiety Among Study Participants With and Without Tinnitus Any Time During Follow-up (1964-2015) Variable No. (%) of Study Participants P Valuea No Tinnitus Tinnitus Adoptees No. 10 846 214 NA Depression 1927 (17.8) 62 (30.0) <.001 Anxiety 2938 (27.1) 97 (45.3) <.001 Hearing loss 310 (2.9) 97 (45.3) <.001 Adoptive Parents No. 18 571 444 NA Depression 1978 (10.6) 89 (20.0) <.001 Anxiety 1943 (10.5) 111 (25.0) <.001 Hearing loss 2 289 (12.3) 276 (62.2) <.001 Biological Parents No. 16 654 371 NA Depression 2194 (13.2) 92 (24.8) <.001 Anxiety 3116 (18.7) 120 (32.2) <.001 Hearing loss 1064 (6.4) 194 (52.3) <.001 Abbreviation: NA, not applicable.

18 571 444 NA Depression 1978 (10.6) 89 (20.0) <.001 Anxiety 1943 (10.5) 111 (25.0) <.001 Hearing loss 2 289 (12.3) 276 (62.2) <.001 Biological Parents No. 16 654 371 NA Depression 2194 (13.2) 92 (24.8) <.001 Anxiety 3116 (18.7) 120 (32.2) <.001 Hearing loss 1064 (6.4) 194 (52.3) <.001 Abbreviation: NA, not applicable. a χ2 Test. Cohort Study In crude model 1 (univariate), the OR for tinnitus in adoptees with at least 1 affected parent was increased (OR, 1.83; 95% CI, 1.04-3.24) (Table 3). In the fully adjusted model 2 (multivariate), which also included birth year, sex, county, educational attainment, depression, anxiety, and hearing loss, the familial OR for tinnitus was still significant at 2.01 (95% CI, 1.10-3.69). The estimated OR for tinnitus in adoptees with an affected adoptive parent was not statistically significant in the crude model 1 (univariate) (OR, 1.01; 95% CI, 0.53-1.91) or in the adjusted model 2 (multivariate) (OR, 1.04; 95% CI, 0.53-2.04). In a model that included the history of tinnitus in biological and adoptive parents, 10 179 individuals had no history of tinnitus in biological and adoptive parents, 498 individuals had adoptive parents only with tinnitus, 368 individuals had biological parents only with tinnitus, and 15 individuals had both biological and adoptive parents with history of tinnitus (interaction term Wald χ2 = 0.0011; SE = 351.6).

viduals had no history of tinnitus in biological and adoptive parents, 498 individuals had adoptive parents only with tinnitus, 368 individuals had biological parents only with tinnitus, and 15 individuals had both biological and adoptive parents with history of tinnitus (interaction term Wald χ2 = 0.0011; SE = 351.6). Table 3. Risk of Tinnitus in Adoptees as Determined by Affected Biological or Adoptive Parent (Cohort Study) Variable No. of Person-Years Cases/Persons at Risk Incidence Rate per 1000 Person-Years Incidence Ratio (95% CI) OR (95% CI)a Model 1 Model 2 Risk of tinnitus with at least 1 affected biological parent Biological parents not affectedb 495 278 201/10677 0.41 (0.35-0.46) NA NA NA Biological parent affected 18 122 13/383 0.72 (0.42-1.24) 1.77 (1.01-3.10) 1.83 (1.04-3.24) 2.01 (1.10-3.69) Risk of tinnitus with at least 1 affected adoptive parent Adoptive parents not affectedb 490 606 204/10 547 0.42 (0.36-0.48) NA NA NA Adoptive parent affected 22 794 10/513 0.44 (0.24-0.82) 1.06 (0.56-1.99) 1.01 (0.53-1.91) 1.04 (0.53-2.04) Abbreviations: NA, not applicable; OR, odds ratio. a The ORs were derived from unconditional logistic regression. Model 1 is a crude model (univariate). Model 2 is an adjusted model (multivariate), with adjustments for adoptees’ birth year, sex, county, educational attainment, depression, anxiety, and hearing loss. b Reference group.

Table 3. Risk of Tinnitus in Adoptees as Determined by Affected Biological or Adoptive Parent (Cohort Study) Variable No. of Person-Years Cases/Persons at Risk Incidence Rate per 1000 Person-Years Incidence Ratio (95% CI) OR (95% CI)a Model 1 Model 2 Risk of tinnitus with at least 1 affected biological parent Biological parents not affectedb 495 278 201/10677 0.41 (0.35-0.46) NA NA NA Biological parent affected 18 122 13/383 0.72 (0.42-1.24) 1.77 (1.01-3.10) 1.83 (1.04-3.24) 2.01 (1.10-3.69) Risk of tinnitus with at least 1 affected adoptive parent Adoptive parents not affectedb 490 606 204/10 547 0.42 (0.36-0.48) NA NA NA Adoptive parent affected 22 794 10/513 0.44 (0.24-0.82) 1.06 (0.56-1.99) 1.01 (0.53-1.91) 1.04 (0.53-2.04) Abbreviations: NA, not applicable; OR, odds ratio. a The ORs were derived from unconditional logistic regression. Model 1 is a crude model (univariate). Model 2 is an adjusted model (multivariate), with adjustments for adoptees’ birth year, sex, county, educational attainment, depression, anxiety, and hearing loss. b Reference group. Case-Control Study We further validated these findings using a case-control study. Tinnitus in adoptees was significantly associated with tinnitus in biological parents, with an OR of 2.22 (95% CI, 1.03–4.81) for the 10 adoptees with an affected biological parent compared with the 136 adoptees with unaffected biological parents. Tinnitus in an adoptive parent was not significantly associated with tinnitus in adoptees (7 adoptees with an affected adoptive parent vs 139 with no affected adoptive parent; OR, 1.00; 95% CI, 0.43–2.32). These findings suggest that genetic factors are associated with the transmission of clinically significant tinnitus and that there is no association of shared environment with the transmission of tinnitus.

7 adoptees with an affected adoptive parent vs 139 with no affected adoptive parent; OR, 1.00; 95% CI, 0.43–2.32). These findings suggest that genetic factors are associated with the transmission of clinically significant tinnitus and that there is no association of shared environment with the transmission of tinnitus. Heritability Heritability was determined in the case-control study with different estimates of the prevalence of tinnitus (Table 4). The prevalence in the particular source population is unknown; however, on the basis of a previous systematic review,29 a range of likely estimates was selected. The corresponding range of heritability estimates is presented in Table 4. The heritability varied from 19% in a population with 0.01% prevalence to 35% in a population with 5% prevalence. With a prevalence of 2.0% (Table 4) in the present population, the mean (SE) heritability was 31% (14%). This finding is similar to the heritability obtained using Falconer regression. The mean (SE) heritability determined using Falconer regression was 32.3% (15.7%). Table 4. Heritability of Tinnitus Based on Different Estimated Population Prevalence and Tetrachoric Correlation in the Case-Control Studya Prevalence Tetrachoric Correlation, Mean (SE) Heritability, Mean (SE), % 0.01 0.094 (0.04) 19 (8) 0.05 0.105 (0.05) 21 (10) 0.1 0.11 (0.05) 22 (10) 0.5 0.13 (0.06) 26 (12) 1.0 0.14 (0.06) 28 (12) 2.0 0.153 (0.07) 31 (14) 5.0 0.174 (0.08) 35 (16) Abbreviation: OR, odd ratio.

choric Correlation in the Case-Control Studya Prevalence Tetrachoric Correlation, Mean (SE) Heritability, Mean (SE), % 0.01 0.094 (0.04) 19 (8) 0.05 0.105 (0.05) 21 (10) 0.1 0.11 (0.05) 22 (10) 0.5 0.13 (0.06) 26 (12) 1.0 0.14 (0.06) 28 (12) 2.0 0.153 (0.07) 31 (14) 5.0 0.174 (0.08) 35 (16) Abbreviation: OR, odd ratio. a Based on Frisell et al.39 There were 10 exposed patients (adoptees with tinnitus and an affected biological parent) and 136 unexposed patients (adoptees with tinnitus without an affected adoptive parent). The OR was 2.22 (95% CI, 1.03-4.81) for all.

choric Correlation in the Case-Control Studya Prevalence Tetrachoric Correlation, Mean (SE) Heritability, Mean (SE), % 0.01 0.094 (0.04) 19 (8) 0.05 0.105 (0.05) 21 (10) 0.1 0.11 (0.05) 22 (10) 0.5 0.13 (0.06) 26 (12) 1.0 0.14 (0.06) 28 (12) 2.0 0.153 (0.07) 31 (14) 5.0 0.174 (0.08) 35 (16) Abbreviation: OR, odd ratio. a Based on Frisell et al.39 There were 10 exposed patients (adoptees with tinnitus and an affected biological parent) and 136 unexposed patients (adoptees with tinnitus without an affected adoptive parent). The OR was 2.22 (95% CI, 1.03-4.81) for all. Discussion This study identified an association between tinnitus and adoptees in relation to their biological parents but not to their adoptive parents. In other words, tinnitus in adoptive parents did not increase the odds of tinnitus among adoptees, a finding that suggests a limited association of family-related environmental factors with the heritability of tinnitus. Adoption studies are complementary to twin studies for a number of reasons. It is assumed in twin studies that concordance rates between monozygotic and dizygotic twins are comparable and can thus be used to estimate genetic contributions because the twins share the same environment. However, one study42 suggests that monozygotic twins are treated more equally than dizygotic twins, which in theory would inflate the heritability seen in twin studies. Whether violation of the equal environments assumption in twin studies for tinnitus affects estimates of heritability is not known. It is also not known whether the assumption of random mating affects the estimates of tinnitus heritability. Because adoptees do not grow up in their biological families, adoption studies offer an excellent model to investigate genetic influences of a given trait. Transmission from adoptive parents to nonbiological offspring would be mainly associated with environmental factors, whereas the transmission from biological parents to offspring would therefore be associated with genetic factors. However, although this notion is important to understand the transmission of tinnitus, the present work cannot completely rule out the contribution of shared environment in tinnitus development. In this study, most adoptees were diagnosed with tinnitus in adulthood, and whether familial environmental factors are weakened or not after adoptees become adults remains uncertain. Familial aggregation studies, including influence on spouses, may help complement the present study; however, one should consider that different mechanisms of development of severe tinnitus may occur between men and women, as recently suggested.13

tors are weakened or not after adoptees become adults remains uncertain. Familial aggregation studies, including influence on spouses, may help complement the present study; however, one should consider that different mechanisms of development of severe tinnitus may occur between men and women, as recently suggested.13 Strengths and Limitations A major strength in the present study is the use of nationwide registers, which are almost complete and have successfully been used to estimate familial risks for a number of diseases.31,32,33,34,36,43 Moreover, the study design eliminates the risk of recall bias, which is an important problem in case-control studies. We reveal a heritability of 32%, which is in the range of the values that 2 former Swedish twin studies reported (near 40% for any type of tinnitus).26,28 The study from Bogo et al26 used a sample of twins (n = 1084 individuals) whose pure-tone audiometric thresholds (up to 8 kHz) were available, whereas the study from Maas et al28 estimated heritability on a larger sample size (n = 10 464 twin pairs) in the absence of audiometric data. These 2 twin studies26,28 relied on self-reported tinnitus, which does not specify the degree of severity, the duration (acute vs chronic), and the time component (occasional vs permanent). In addition, the way that the question was formulated may cause substantial differences in the reported prevalence,29 which in turn can affect heritability values. In this study, heritability values were obtained using medical registries based on a diagnostic established by a physician, which may be seen as a more rigorous approach, although not free from biases. The prevalence of 2% appears on the low side of the mean prevalence of self-reported severe tinnitus in the literature.29 The number of people with tinnitus seeking help may be lower in this study because of the general lack of trust in the ability of physicians to provide a solution to their tinnitus.44 Because heritability values are influenced by prevalence, the values obtained in our study were potentially underestimated. Tinnitus assessment and treatment are not standardized at the national level in Sweden; however, the Stockholm County Council has developed a document to guide the assessment of tinnitus at the general practitioner level using the Tinnitus Handicap Inventory and proposed some rehabilitation options, such as cognitive behavioral therapy and a modified tinnitus retraining therapy.

national level in Sweden; however, the Stockholm County Council has developed a document to guide the assessment of tinnitus at the general practitioner level using the Tinnitus Handicap Inventory and proposed some rehabilitation options, such as cognitive behavioral therapy and a modified tinnitus retraining therapy. A patient is directed for special rehabilitation when the Tinnitus Handicap Inventory score is 58 or higher (meaning severe or catastrophic) or when clinically significant levels of anxiety or depression are found. How tinnitus assessment and rehabilitation are performed in other counties is unclear, and therefore the global picture of tinnitus care in Sweden is missing. Thus, the country would benefit from national tinnitus guidelines,45 which would harmonize the reporting procedures and national registry data. The problem of the clinical assessment of tinnitus is not restricted to Sweden because of the disparity in assessing tinnitus if found across all European Union countries (Cederroth et al, unpublished data, 2017), which emphasizes the need to standardize its clinical assessment until objective tools can be implemented in the health care system.

l assessment of tinnitus is not restricted to Sweden because of the disparity in assessing tinnitus if found across all European Union countries (Cederroth et al, unpublished data, 2017), which emphasizes the need to standardize its clinical assessment until objective tools can be implemented in the health care system. Assuming that occasional tinnitus precedes the development of permanent tinnitus and mild tinnitus also precedes the development of severe tinnitus would mean that the current study using medical registries, in which a diagnosis is reported, would underestimate the heritability values by focusing only on clinically significant tinnitus. A focus on severe tinnitus may reveal heritability values that would not be directly associated with the development of tinnitus per se; instead, these values would be more associated with the accompanying psychological distress. Whereas none of the investigated polymorphisms have been associated with tinnitus (using a group without tinnitus as controls), a variant in the promoter of the serotonin transporter gene (SLC6A4) (OMIM 182138) was associated with tinnitus severity when compared with patients with milder tinnitus.46 Similarly, a pilot genome-wide association study25 with a small sample size of 167 patients and 769 controls found no significant associations but revealed an enrichment in genes involved in serotonin receptor signaling. These studies are consistent with recent findings in mice whereby serotonin would excite fusiform cells from the dorsal cochlear nucleus,47 a structure from the auditory pathway known to be involved in tinnitus.35,48,49,50 Thus, the heritability values collected in our study may reflect the genetic heritability of severe tinnitus and not that of any type of tinnitus.

in mice whereby serotonin would excite fusiform cells from the dorsal cochlear nucleus,47 a structure from the auditory pathway known to be involved in tinnitus.35,48,49,50 Thus, the heritability values collected in our study may reflect the genetic heritability of severe tinnitus and not that of any type of tinnitus. A study from Martinez et al51 suggests that the existing ICD codes are of sufficient quality for research in a clinical setting. However, we believe that the existing ICD codes for tinnitus are obsolete, even when considering ICD codes for hearing loss, and would still benefit from a global revision. Laterality (meaning whether tinnitus is perceived unilaterally or bilaterally) appears as an important classification for genetic subtypes of tinnitus according to a recent twin study.28 Maas et al28 found that bilateral tinnitus is significantly influenced by genetic factors, whereas unilateral tinnitus is more subject to environmental influences. However, the degree of hearing loss in patients with unilateral and bilateral tinnitus appears to be similar (Cederroth et al, unpublished data, 2018), and thus the establishment of laterality using audiologic register data may not help in distinguishing patients with bilateral from those with unilateral tinnitus. Furthermore, although hearing loss and noise exposure are major etiologic factors for tinnitus, a nonnegligible proportion of patients experience tinnitus without hearing loss, supporting the notion that the current codes for tinnitus or hearing loss would not facilitate the replication of the study by Maas et al.8 A revision of the ICD codes for tinnitus is an endeavor that will require substantial evidence to classify tinnitus subtypes according to clear criteria. Fortunately, temporary ICD codes can be generated for research purposes at a national level and may help determine whether a newly proposed definition has clinical value before its implementation into the World Health Organization codes. Nevertheless, the consistency among these 3 studies in Sweden supports the idea that some forms of tinnitus are associated with genetics more than with shared environment. How this applies to other countries needs to be determined using similar approaches.

re its implementation into the World Health Organization codes. Nevertheless, the consistency among these 3 studies in Sweden supports the idea that some forms of tinnitus are associated with genetics more than with shared environment. How this applies to other countries needs to be determined using similar approaches. Although it is likely that most adoptions occurred in early childhood, we lack information regarding the age at which children were adopted. Perinatal factors and preplacement age could confound the genetic component. In previous studies,52 most children were adopted before 12 months of age. Another limitation is demographic differences between biological and adoptive parents. For instance, biological parents (including biological parents with tinnitus) were more often women than adoptive parents. Biological parents (including biological parents with tinnitus) were also younger than adoptive parents. In addition, the present adoption study included only adoptees who were born in Sweden; therefore, we cannot generalize the present study to a population of nonwhite origin.

e more often women than adoptive parents. Biological parents (including biological parents with tinnitus) were also younger than adoptive parents. In addition, the present adoption study included only adoptees who were born in Sweden; therefore, we cannot generalize the present study to a population of nonwhite origin. Conclusions This adoption study using national registry data suggests that clinically significant tinnitus is associated with genetic factors, with a heritability of 32% and that there is no association between shared-environment factors with the transmission of tinnitus. Because patients in Sweden with a diagnosis of tinnitus most often have severe tinnitus, the present data reveal that there could be an association between genetic factors and the transition from compensated to decompensated tinnitus. Thus, the association of genetics with tinnitus might be found on 2 levels: (1) for determining whether an individual develops tinnitus or not and (2) for determining whether an individual transitions from having nonbothersome tinnitus to bothersome tinnitus. The identification of genes involved in any of these 2 aspects may provide interesting insights into the molecular mechanisms that regulate phantom percepts and their treatment. Supplement. eTable. International Classification of Diseases (ICD) Codes Used to Define Depression, Anxiety, and Hearing Loss Click here for additional data file.

Introduction The paradox of rare diseases is that any single diagnosis affects a small number of individuals, but 6% to 8% of the world population is afflicted. This fact renders rare diseases both difficult to study and a significant public health concern. Even basic epidemiologic studies of rare diseases are challenging given the heterogeneous, progressive clinical course of the rare disease and the geographic dispersion of patients. Geographic distribution requires an extensive recruitment and monitoring infrastructure, dramatically increasing the cost and time required for participant accrual. These barriers, coupled with small markets, reduce incentives for the pharmaceutical industry and funders to support rare disease research, constrain evidence, and result in clinical practice variation and inconsistent patient outcomes.

ucture, dramatically increasing the cost and time required for participant accrual. These barriers, coupled with small markets, reduce incentives for the pharmaceutical industry and funders to support rare disease research, constrain evidence, and result in clinical practice variation and inconsistent patient outcomes. Fortunately, emerging technologies offer promise to reduce the barriers to research about rare disease. Patients have begun to cluster in online communities to provide mutual support and information about their conditions. These communities have become rich sources of knowledge about the lived experience of patients with rare disease. Researchers are increasingly interested in leveraging online communities to rapidly accrue sample sizes required to generate evidence and improve treatments. Paralleling this trend are advances in personal mobile computing to generate biomedically relevant data streams, including self-monitoring of disease symptoms and vital signs. Patient-generated health data may hold immense promise to improve treatment outcomes by providing clinicians and researchers a view of disease progression at home.

end are advances in personal mobile computing to generate biomedically relevant data streams, including self-monitoring of disease symptoms and vital signs. Patient-generated health data may hold immense promise to improve treatment outcomes by providing clinicians and researchers a view of disease progression at home. Idiopathic subglottic stenosis (iSGS) is a rare, recurrent, and devastating fibroinflammatory disease that leads to upper airway narrowing and severe dyspnea among adult white women. Because of high recurrence rates, more than half of patients with iSGS require repeated surgical procedures within 12 months of their initial diagnosis. Three treatments for iSGS predominate, and evidence has shown variability in outcomes. This variability has complicated patient decision-making as patients try to balance survival, symptoms, and quality-of-life considerations. We designed a prospective multicenter observational study comparing the effectiveness of the 3 most common treatments for iSGS using outcomes that matter most to patients: time to disease recurrence and treatment quality-of-life trade-offs. Harnessing emerging technologies, we directly recruited people with iSGS from an online community on Facebook in addition to traditional physician-led recruitment efforts. We used a novel approach to monitor treatment response that included clinical data from electronic health records, longitudinal physiologic data recorded in a smartphone app, and patient-reported outcome measures. By coupling patient-generated health data from the digital platform with clinical data from electronic health records, the burden of studying rare disease longitudinally across multiple sites was reduced. This approach facilitated rapid establishment of a study cohort and nuanced treatment comparisons.

utcome measures. By coupling patient-generated health data from the digital platform with clinical data from electronic health records, the burden of studying rare disease longitudinally across multiple sites was reduced. This approach facilitated rapid establishment of a study cohort and nuanced treatment comparisons. Methods Participants Adult patients (≥18 years) with untreated, newly diagnosed, or previously treated iSGS meeting established diagnostic criteria were enrollment candidates. Age, sex, and race/ethnicity were collected based on recorded electronic medical records and confirmed with the patients. Recruitment of patients took place from June 1, 2015, to June 1, 2017. Patients with obstructing subglottic stenosis not attributable to the 2 most common etiologies (granulomatosis with polyangiitis and intubation-related airway trauma) were included. Patients were excluded if their index operative date was not confirmed or they failed to complete required baseline surveys. The study was approved by the Vanderbilt University Medical Center institutional review board, Nashville, Tennessee, and written informed consent was obtained electronically from each participant.

atients were excluded if their index operative date was not confirmed or they failed to complete required baseline surveys. The study was approved by the Vanderbilt University Medical Center institutional review board, Nashville, Tennessee, and written informed consent was obtained electronically from each participant. Setting Patients were recruited using both traditional and novel methods with a goal of enrolling 300 participants. The traditional method involved patient identification and recruitment by clinicians participating in the North American Airway Collaborative (NoAAC) network. The NoAAC consists of 30 participating tertiary care centers across all regions of the United States as well as international sites in Australia, France, Iceland, Norway, and the United Kingdom. All NoAAC sites are referral centers for iSGS and thus have significant experience treating this rare disease.

oAAC) network. The NoAAC consists of 30 participating tertiary care centers across all regions of the United States as well as international sites in Australia, France, Iceland, Norway, and the United Kingdom. All NoAAC sites are referral centers for iSGS and thus have significant experience treating this rare disease. The novel recruitment method involved direct patient enrollment from a community of patients with iSGS on Facebook. The online community “Living with Idiopathic Subglottic Stenosis” currently has 3030 members (2636 are patients with iSGS); 42% of members visit the site daily, and 97% visit at least monthly. The online community has robust leadership, and the founder and chief moderator is a person with iSGS. Her stewardship engenders a positive culture of balanced information sharing. The NoAAC engaged the online community leader to solicit her formal involvement in study planning. In addition to aligning patient and clinician goals, this integration allowed information (both initial recruiting efforts and continuous project updates) to flow from the leader to the online community. Weekly conversations between the leader and principal investigators kept community members updated on recruitment. Annual in-person study meetings (with all online community members invited to attend) maintained participant engagement. These features allowed study information to be rapidly disseminated. Interested online community members with iSGS directly enrolled by contacting the study coordinator, who obtained permission for medical records release and who completed consent for the patients electronically. Patient information collected from each recruitment method was screened to confirm inclusion criteria and to allow data entry into the secure electronic data repository.

ed by contacting the study coordinator, who obtained permission for medical records release and who completed consent for the patients electronically. Patient information collected from each recruitment method was screened to confirm inclusion criteria and to allow data entry into the secure electronic data repository. Study Protocol Baseline The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and a prespecified protocol (eFigure in the Supplement). At enrollment, patients completed baseline surveys and a series of patient-reported outcome measures evaluating constructs affected by the disease and its treatment. In addition, disease-specific data, a mucosal atopy index, and comorbidity scores were abstracted from medical records. Index Intervention Each NoAAC center submitted its standard-of-care treatment algorithm before beginning enrollment. Symptomatic patients underwent standard-of-care treatment at their respective medical centers. The index, or most proximate, surgical procedure (if the last treatment predated study inception) was defined as time zero (T0).

ach NoAAC center submitted its standard-of-care treatment algorithm before beginning enrollment. Symptomatic patients underwent standard-of-care treatment at their respective medical centers. The index, or most proximate, surgical procedure (if the last treatment predated study inception) was defined as time zero (T0). Primary treatments were endoscopic dilation (ED), endoscopic resection with adjuvant medical therapy (ERMT), and open cricotracheal resection (CTR). All treatments are described in detail elsewhere. In brief, ED involves using a balloon or rigid bougie to expand the stenotic segment; ERMT uses a carbon dioxide laser to endoscopically resect quadrants of the stenotic airway followed by long-term adjuvant medication (eg, inhaled corticosteroid, a proton pump inhibitor, and trimethoprim/sulfamethoxazole); and CTR is an open surgical procedure that involves en bloc removal of the stenotic airway followed by anastomosis of the proximal and distal tracheal segments. The study nurse coordinator (C.K.) reviewed the specifics of the index surgical procedures to confirm the type of intervention and to ensure adherence to their standard of care.

surgical procedure that involves en bloc removal of the stenotic airway followed by anastomosis of the proximal and distal tracheal segments. The study nurse coordinator (C.K.) reviewed the specifics of the index surgical procedures to confirm the type of intervention and to ensure adherence to their standard of care. Longitudinal Surveillance Patients completed the electronic health status every 3 months after enrollment. The electronic data capture system generated a scheduled, automated query that solicited data on adjuvant medication use (eg, inhaled corticosteroids); whether patients had undergone any interventions to treat their iSGS (surgical or clinic-based) and, if so, whether they had experienced specific treatment-associated complications; and patient-reported outcome measures to track symptoms and quality of life (every 6 months). Reports of treatment or complications triggered notification of the study nurse coordinator, who investigated, verified, and documented these events, including the date of any recurrent surgical procedure. Patients were given the ability to self-monitor their disease status by daily recording of peak expiratory flow rate (PEFR) using an inexpensive portable handheld device and a free smartphone app designed by the authors. The PEFR was measured in liters per second during a single expiratory cycle and was reported as percentage of matched normative data (%PEFR).

nitor their disease status by daily recording of peak expiratory flow rate (PEFR) using an inexpensive portable handheld device and a free smartphone app designed by the authors. The PEFR was measured in liters per second during a single expiratory cycle and was reported as percentage of matched normative data (%PEFR). Outcomes The primary end point was need for and time to recurrent surgical procedure, defined as days from T0 to recurrent surgical procedure (TR). This end point was derived from semistructured interviews among patients with iSGS and participating clinicians, who agreed that it was the most appropriate surrogate for recurrence. Secondary end points included quality-of-life trade-offs and complications across treatment modalities. Statistical Analysis Primary Outcome Data was analyzed between July 1 and September 30, 2018. The Kaplan-Meier method assessed time to recurrent surgical procedure for the 3 primary treatments, and hazard ratios (HRs) and 95% CIs were calculated using weighted Cox proportional hazards regression models. Censoring occurred with a recurrent surgical procedure or a patient death. Propensity score matching and multiple imputation–adjusted HRs were used for known confounders and missing data (eMethods in the Supplement). Effect sizes (ESs) and 95% CIs were reported for all comparisons as indicated.

zards regression models. Censoring occurred with a recurrent surgical procedure or a patient death. Propensity score matching and multiple imputation–adjusted HRs were used for known confounders and missing data (eMethods in the Supplement). Effect sizes (ESs) and 95% CIs were reported for all comparisons as indicated. Indication for Recurrent Surgical Procedure The observational study design precluded prespecifying indications for repeat surgical procedure. In general, the primary indication for repeat surgical procedure was worsened breathing; however, the threshold for repeat surgical procedure may differ across centers. Physiologic (%PEFR) and patient-reported Clinical COPD (chronic obstructive pulmonary disease) Questionnaire (CCQ) metrics were used to assess whether this threshold systematically differed by treatment modality. Loess curves of mixed-effects model–fitted values for %PEFR assessed the stability of respiratory function for nonrecurring patients.

tient-reported Clinical COPD (chronic obstructive pulmonary disease) Questionnaire (CCQ) metrics were used to assess whether this threshold systematically differed by treatment modality. Loess curves of mixed-effects model–fitted values for %PEFR assessed the stability of respiratory function for nonrecurring patients. Results Recruitment and Study Population Of the 1056 patients consented, 383 patients (36.3%) were recruited via social media, and 673 (63.7%) were recruited via physician referral (Figure 1A). Patient recruitment via social media outpaced that via physician referral during the first 47 weeks of the 3-year study (Figure 1B). Patients recruited via social media were slightly younger (mean [SD] age of recruits: social media, 47 [10] years vs physician referral, 53 [12] years) but otherwise lacked differences in sociodemographic or baseline survey completion rates. Given their earlier enrollment in the study, patients recruited via social media had a longer mean (SD) follow-up (1.47 [0.88] years vs 1.27 [0.79] years). Overall, 35 patients (3.3%) withdrew from the study, and 211 (20.6%) consented to participate but were unable to produce treatment data associated with their disease and the index procedure. Exclusions were balanced between Facebook recruits (71 of 368; 19.3%) and physician referrals (140 of 653; 21.4%).

[0.79] years). Overall, 35 patients (3.3%) withdrew from the study, and 211 (20.6%) consented to participate but were unable to produce treatment data associated with their disease and the index procedure. Exclusions were balanced between Facebook recruits (71 of 368; 19.3%) and physician referrals (140 of 653; 21.4%). Figure 1. Flowchart and Absolute Number of Participants Recruited In all, 810 patients meeting inclusion criteria enrolled, of whom 798 (98.5%) were female, 787 (97.2%) were white, and 64.8% (487 of 752 reporting) held college or advanced degrees, with a median age of 50 years (interquartile range [IQR], 43-58 years). Index operations were ED (n = 603; 74.4%), ERMT (n = 121; 14.9%), and CTR (n = 86; 10.6%) (Figure 1A). Anatomically, patients with iSGS presented with a median (IQR) subglottic narrowing of 11 mm (8-17 mm) caudal to the vocal folds, median (IQR) craniocaudal length of 15 mm (10-20 mm), and median (IQR) airway obstruction at T0 of 60% (50%-75%).

n = 603; 74.4%), ERMT (n = 121; 14.9%), and CTR (n = 86; 10.6%) (Figure 1A). Anatomically, patients with iSGS presented with a median (IQR) subglottic narrowing of 11 mm (8-17 mm) caudal to the vocal folds, median (IQR) craniocaudal length of 15 mm (10-20 mm), and median (IQR) airway obstruction at T0 of 60% (50%-75%). Patient sociodemographic and clinical characteristics were similar across treatment modalities (Table). Patients who underwent ERMT were predominantly white (121 of 121; 100%), older (median, 56 years; range, 48-63 years), and had greater median percentage of luminal obstruction at T0 (75%; range, 64%-80%). Patients who underwent ED showed the shortest median segment of subglottal narrowing (12 mm; range, 10-17 mm). Stenosis observed in this group was the farthest from the vocal folds (15 mm; range, 10-20 mm). Patients who underwent ERMT experienced the longest median duration of disease (8.6 years; IQR, 3.4-13.0 years), and patients who underwent CTR had the most surgical procedures (n = 5) performed before open reconstruction at T0 (range, 3-7 surgical procedures).

he farthest from the vocal folds (15 mm; range, 10-20 mm). Patients who underwent ERMT experienced the longest median duration of disease (8.6 years; IQR, 3.4-13.0 years), and patients who underwent CTR had the most surgical procedures (n = 5) performed before open reconstruction at T0 (range, 3-7 surgical procedures). Table. Characteristics of Patients With Idiopathic Subglottic Stenosis Characteristic ED (n = 603) ERMT (n = 121) CTR (n = 86) Total (N = 810) Effect Sizea Test Typeb Age at index procedure, median (IQR), y 49 (42-57) 56 (48-63) 48 (39-55) 50 (43-58) 0.036 (0.015-0.063) 1 Female, No. (%) 593 (98.3) 121 (100) 84 (97.5) 798 (98.5) 0.032 (0-0.089) 2 Married, No. (%)c 426 (76.1) 86 (76.1) 70 (86.4) 582 (77.2) 0.076 (0-0.141) 2 White race, No. (%) 580 (95.9) 121 (100) 86 (100) 787 (97.2) 0.105 (0.017-0.171) 2 White (non-Hispanic or Latino) ethnicity, No. (%)c 503 (97.4) 87 (100) 60 (93.8) 650 (97.5) 0.089 (0-0.159) 2 Educational level, No. (%)c Graduate school 153 (27.4) 22 (19.5) 14 (17.3) 189 (25.1) 0.092 (0-0.127) 2 College graduate 218 (39.0) 42 (37.2) 38 (46.9) 298 (39.6) Some college 118 (21.0) 34 (30.1) 23 (28.4) 175 (23.3) High school or less 70 (12.5) 15 (13.3) 5 (6.2) 90 (12.0) Stenosis, median (IQR), % 50 (40-70) 75 (64-80) 60 (60-70) 60 (50-75) 0.06 (0.034-0.088) 1 Stenosis length, median (IQR), mm 12 (10-17) 15 (10-20) 17 (15-20) 15 (10-20) 0.02 (0.005-0.042) 1 Distance below glottis, median (IQR), mm 15 (10-20) 10 (5-15) 10 (5-15) 11 (8-17) 0.018 (0.005-0.04) 1 Disease duration, median (IQR), y 5.5 (2.5-9.9) 8.6 (3.4-13.0) 6.1 (3.9-10.2) 5.8 (2.6-10.7) 0.014 (0.002-0.034) 1 Surgical procedure, No. (range) 3 (2-7) 3 (2-6) 5 (3-7) 3 (2-7) 0.015 (0.004-0.031) 1 Charlson Comorbidity Index, median (IQR) 0 0 0 0 0.007 (0-0.018) 1 Gastroesophageal reflux disease, No. (%)c 206 (37.1) 43 (38.4) 31 (38.8) 280 (37.5) 0.013 (0-0.05) 2 Premenopausal, No. (%)c 204 (72.6) 20 (64.5) 25 (71.4) 249 (72.0) 0.051 (0-0.138) 2 Hormone replacement therapy, No. (%)c 5 (1.8) 1 (3.2) 0 6 (1.7) 0.054 (0-0.143) 2 Full-term pregnancy, No.

R) 0 0 0 0 0.007 (0-0.018) 1 Gastroesophageal reflux disease, No. (%)c 206 (37.1) 43 (38.4) 31 (38.8) 280 (37.5) 0.013 (0-0.05) 2 Premenopausal, No. (%)c 204 (72.6) 20 (64.5) 25 (71.4) 249 (72.0) 0.051 (0-0.138) 2 Hormone replacement therapy, No. (%)c 5 (1.8) 1 (3.2) 0 6 (1.7) 0.054 (0-0.143) 2 Full-term pregnancy, No. (%) 0 139 (25.3) 12 (10.9) 13 (16.5) 164 (22.2) 0.123 (0.041-0.157) 2 1 81 (14.7) 10 (9.1) 7 (8.9) 98 (13.3) 2 203 (36.9) 51 (46.4) 35 (43.8) 289 (39.1) 3 88 (16.0) 22 (20.0) 16 (20.0) 126 (17.1) >3 39 (7.1) 15 (13.6) 8 (10.0) 62 (8.4) Years of follow-up, median (IQR), y 1.3 (0.4-2.2) 1.5 (0.2-3.6) 4.3 (1.7-6.1) 1.4 (0.4-2.5) 0.011 (0.002-0.024) 1 Abbreviations: CTR, cricotracheal resection; ED, endoscopic dilation; ERMT, endoscopic resection with adjuvant medical therapy; IQR, interquartile range. a Epsilon-squared for Kruskal-Wallis test and Cramer V for Pearson χ2 test. The 95% CIs for epsilon-squared values are estimated by adjusted bootstrap percentile (bias-corrected and accelerated) interval with 1000 replications; 95% CIs for Cramer V values are estimated by noncentral χ2. b Test type: 1, Kruskal-Wallis test; 2, Pearson χ2 test. c Not reported for all participants.

a Epsilon-squared for Kruskal-Wallis test and Cramer V for Pearson χ2 test. The 95% CIs for epsilon-squared values are estimated by adjusted bootstrap percentile (bias-corrected and accelerated) interval with 1000 replications; 95% CIs for Cramer V values are estimated by noncentral χ2. b Test type: 1, Kruskal-Wallis test; 2, Pearson χ2 test. c Not reported for all participants. Time to Recurrence (Primary End Point) Median (IQR) follow-up after T0 was similar between patients who underwent ED (1.3 [0.4-2.2] years) vs those who underwent ERMT (1.5 [0.2-3.6] years) and was significantly longer for patients who underwent CTR (4.3 [1.7-6.1] years) (Table). Overall, 185 patients (22.8%) had a recurrent surgical procedure during the 3-year study period, but the rate differed across modalities. Open cricotracheal resection had a 1.2% recurrence rate (1 of 86 patients), followed by 12.4% (15 of 121) for ERMT and 28.0% (169 of 603) for ED. Kaplan-Meier analysis showed the need for and the time to recurrent surgical procedure between treatment modalities (Figure 2). With 1 recurrence within the CTR group, only ED and ERMT were comparable (ERMT vs ED: HR, 3.03; 95% CI, 1.78-5.17). Propensity score matching was used to minimize bias from nonrandomized treatment assignment. Weighted Cox proportional hazards regression models (eTable 1 in the Supplement) showed that ED was inferior compared with ERMT (HR, 2.77; 95% CI, 1.4-5.5), and this association persisted in propensity score–matching models accounting for missing data (ED vs ERMT: HR, 3.16; 95% CI, 1.82-5.51) (eTable 2 in the Supplement).

ghted Cox proportional hazards regression models (eTable 1 in the Supplement) showed that ED was inferior compared with ERMT (HR, 2.77; 95% CI, 1.4-5.5), and this association persisted in propensity score–matching models accounting for missing data (ED vs ERMT: HR, 3.16; 95% CI, 1.82-5.51) (eTable 2 in the Supplement). Figure 2. Kaplan-Meier Analysis of Disease Recurrence Among the 3 Treatment Groups Indication for Recurrent Operation In all, 496 (61.2%) patients recorded PEFR (ED: 383 [63.5%]; ERMT: 62 [51.2%]; and CTR: 51 [59.3%]); the only difference between reporters and nonreporters was that nonreporters were older at 52 years (95% CI, 43-59 years) vs 49 years (95% CI, 42-56 years). No difference in median %PEFR or CCQ score in the ED and ERMT groups was observed before recurrent surgical procedure (%PEFR: ED, 56.5% [IQR, 44%-69%] vs ERMT, 54% [IQR, 45%-59%] and CCQ: ED, 2.4 [IQR 1.6-3.4] vs ERMT, 2.6 [IQR, 1.6-3.3]). Thus, a similar threshold for repeat surgical procedure was exercised across modalities. Loess curves of mixed-effects model–fitted values for %PEFR confirmed that successfully treated patients without recurrence had stable respiratory function (Figure 3).

%] and CCQ: ED, 2.4 [IQR 1.6-3.4] vs ERMT, 2.6 [IQR, 1.6-3.3]). Thus, a similar threshold for repeat surgical procedure was exercised across modalities. Loess curves of mixed-effects model–fitted values for %PEFR confirmed that successfully treated patients without recurrence had stable respiratory function (Figure 3). Figure 3. Longitudinal Mean Peak Expiratory Flow Rate Among Patients Without Recurrence in the 3 Treatment Arms Loess smooth curve of mixed-effects model shows sustained peak expiratory flow rate (measured in liters per second during a single expiratory cycle and reported as percentage of matched normative data) among patients after successful treatment. Self-reported patient longitudinal peak expiratory flow rate was captured using an inexpensive portable handheld device and a free smartphone app created specifically for this study. Patient-Reported Outcome Measures Breathing Among patients who were successfully treated (ie, did not recur during the study period), those who underwent CTR had the best CCQ scores at 360 days after enrollment followed by those who underwent ERMT and ED treatment (ED, 1.80 vs ERMT, 1.25 vs CTR, 0.75). Treatment with CTR was associated with statistically and clinically better breathing at 360 days compared with either ERMT or ED (CTR vs ERMT: ES, 0.3; 95% CI, 0-0.8; CTR vs ED: ES, 0.8; 95% CI, 0.4-1.3). Treatment with ERMT showed better breathing outcomes compared with treatment using ED (ERMT vs ED: ES, 0.5; 95% CI, 0.1-0.8) (Figure 4A).

associated with statistically and clinically better breathing at 360 days compared with either ERMT or ED (CTR vs ERMT: ES, 0.3; 95% CI, 0-0.8; CTR vs ED: ES, 0.8; 95% CI, 0.4-1.3). Treatment with ERMT showed better breathing outcomes compared with treatment using ED (ERMT vs ED: ES, 0.5; 95% CI, 0.1-0.8) (Figure 4A). Figure 4. Secondary End Points of Patient-Reported Functional Outcome at 12 Months COPD indicates chronic obstructive pulmonary disease; ERMT, endoscopic resection with adjuvant medical therapy. Voice and Swallowing Outcomes among successfully treated patients differed 360 days after enrollment. In contrast with breathing, Voice Handicap Index-10 scores of patients who underwent CTR were both statistically and clinically worse compared with scores for patients who underwent ERMT (CTR, 13.0 vs ERMT, 3.5; ES, 6; 95% CI, 1.0-10.0) but not different compared with ED (CTR, 13.0 vs ED, 10.0; P = .07). Patients who underwent ED had worse scores compared with those who underwent ERMT (ED, 10.0 vs ERMT, 3.5; ES, 3; 95% CI, 0-6.0) (Figure 4B). The association with swallowing was minimal among all treatment groups, and although statistical differences were observed, no Eating Assessment Test-10 scores met the threshold to be considered clinically abnormal (all median scores <3) (Figure 4C).

rwent ERMT (ED, 10.0 vs ERMT, 3.5; ES, 3; 95% CI, 0-6.0) (Figure 4B). The association with swallowing was minimal among all treatment groups, and although statistical differences were observed, no Eating Assessment Test-10 scores met the threshold to be considered clinically abnormal (all median scores <3) (Figure 4C). Global Physical Health At 360 days, patients who underwent ED had lower median 12-Item Short-Form Version 2 scores compared with patients who underwent ERMT (ED, 49 vs ERMT, 53; ES, 2.3; 95% CI, −5.66 to 0.55) or CTR (ED, 49 vs CTR, 54; ES, 3.1; 95% CI, −6.9 to 0.29). These outcomes may suggest that global quality of life is statistically and clinically worse after treatment with ED compared with treatment of either ERMT or CTR (Figure 4D). Perioperative Complications and Death Cricotracheal resection was associated with a number of complications. In order of incidence, 8 of 86 patients (9.3%) required a temporary tracheostomy or T-tube and 8 (9.3%) required an unplanned return to the operating room during their initial hospitalization. Moreover, 7 patients (8.1%) developed a permanent unilateral vocal fold paralysis, and 3 (3.5%) had an anastomotic leak. Less common complications included permanent bilateral vocal fold paralysis in 1 patient (1.2%), urinary tract infections secondary to an indwelling catheter in 1 patient (1.2%), and postoperative pneumonia in 1 patient (1.2%).

developed a permanent unilateral vocal fold paralysis, and 3 (3.5%) had an anastomotic leak. Less common complications included permanent bilateral vocal fold paralysis in 1 patient (1.2%), urinary tract infections secondary to an indwelling catheter in 1 patient (1.2%), and postoperative pneumonia in 1 patient (1.2%). Endoscopic resection with adjuvant medical therapy was associated with adverse events during the surgical procedure and the postoperative medical regimen. In order of incidence, perioperative issues included 14 patients (11.6%) who developed temporary (≤4 weeks) tongue paresthesia associated with direct laryngoscopy and 4 patients (3.3%) who had a dental injury. Overall, 35 patients (28.9%) had an adverse reaction to trimethoprim/sulfamethoxazole received postoperatively, with 27 (22.3%) stopping the medication because of rash, nausea, or fever. Endoscopic dilation had a similar perioperative adverse event profile compared with ERMT. These similarities included 84 of 603 patients (13.9%) who had temporary tongue paresthesia, 34 (5.6%) who had a dental injury, and 3 (0.5%) who had transient postoperative subcutaneous emphysema. All occurrences of subcutaneous emphysema resolved with conservative management. Of 810 patients, 3 died (0.04%): 1 in the ERMT group and 2 in the ED group. Death was secondary to airway obstruction more than 30 days after the surgical procedure.

ury, and 3 (0.5%) who had transient postoperative subcutaneous emphysema. All occurrences of subcutaneous emphysema resolved with conservative management. Of 810 patients, 3 died (0.04%): 1 in the ERMT group and 2 in the ED group. Death was secondary to airway obstruction more than 30 days after the surgical procedure. Discussion Idiopathic subglottic stenosis is a recurrent, rare (1:400,000) fibroinflammatory disease that results in life-threatening blockage of the upper airway. Harnessing an engaged online community of patients coupled with innovative digital tools, we rapidly recruited a cohort of 810 patients with iSGS to compare the effectiveness of contemporary treatments with recurrent surgical procedure while assessing quality of life and perioperative risk trade-offs. Treatment effectiveness at recurrent surgical procedures and risk trade-offs differed by modality. Patients who underwent CTR had the lowest rate of recurrent surgical procedure (1.2%), followed by ERMT (12.4%) and ED (28.0%). Considering disease recurrence in the context of treatment trade-offs, CTR was associated with the greatest perioperative risk and the worst postoperative patient-reported voice outcomes. Regarding disease recurrence rate, ERMT was at a 15.6% lower rate compared with ED (the present standard-of-care treatment). Both endoscopic procedures (ERMT and ED) had a similar low rate of perioperative risks and modest differences in patient-reported breathing, voice, and quality-of-life score changes compared with CTR. Swallowing complaints were uncommon regardless of treatment modality.

with ED (the present standard-of-care treatment). Both endoscopic procedures (ERMT and ED) had a similar low rate of perioperative risks and modest differences in patient-reported breathing, voice, and quality-of-life score changes compared with CTR. Swallowing complaints were uncommon regardless of treatment modality. Treatment Advantages and Trade-offs Cricotracheal Resection Our results are consistent with published single-center series documenting excellent surgical procedure outcomes after CTR. The disease recurrence rate after ERMT also parallels a retrospective analysis. Interestingly, published case series reported that 80% of patients had recurrent stenosis after undergoing ED within 1000 days of the initial surgical procedure. This result is a significantly higher rate than we observed in our study (28.0%). This disparity may stem from our inclusion of patients from centers of excellence and from smaller centers recruited online. Previous retrospective studies reported outcomes from high-volume centers with patients who have recalcitrant disease cluster, which may explain a higher disease recurrence rate.

r study (28.0%). This disparity may stem from our inclusion of patients from centers of excellence and from smaller centers recruited online. Previous retrospective studies reported outcomes from high-volume centers with patients who have recalcitrant disease cluster, which may explain a higher disease recurrence rate. Although our results appear to support the effectiveness and durability of CTR, they must be interpreted in context. First, not all patients were candidates for CTR because a sufficient distance below the vocal folds was required to perform this surgical procedure. Second, previous reports identified a significant rate of recurrent disease (10%-30%) among patients who underwent CTR that occurred from 5 to 10 years after the surgical procedure. These outcomes were not observed in our cohort given the temporal scope of our study. Longer-term follow-up of patients who underwent CTR from our cohort will continue to address this question. Although reports of patient complications after CTR are variable within the literature, our data appeared comparable with the reported 10% to 20% rate of anastomotic complications and 5% to 10% rate of postoperative unilateral vocal fold paralysis.

o underwent CTR from our cohort will continue to address this question. Although reports of patient complications after CTR are variable within the literature, our data appeared comparable with the reported 10% to 20% rate of anastomotic complications and 5% to 10% rate of postoperative unilateral vocal fold paralysis. Endoscopic Treatments A notable outcome of our study was the findings for ERMT. In this 3-year study, ERMT offered significantly improved disease control compared with ED (the most common treatment) with minimal association with voice function, particularly when compared with CTR. Whether the reduced disease recurrence rate for ERMT was associated with surgical technique, postoperative medications, or a combination remains unclear. Moreover, whether ERMT outcomes will converge with ED during a longer follow-up is an area of continued study in our cohort.

ion, particularly when compared with CTR. Whether the reduced disease recurrence rate for ERMT was associated with surgical technique, postoperative medications, or a combination remains unclear. Moreover, whether ERMT outcomes will converge with ED during a longer follow-up is an area of continued study in our cohort. Effectiveness of Novel Recruitment Strategy For studies of rare disease, recruiting enough participants is challenging. Among clinical trials, 10% of investigators fail to recruit a single patient, and fewer than 20% of trials finish on time because of poor recruitment. Social media recruitment provides the ability to target a relevant and engaged audience of patients and allows for direct communication with participants. Although the issue of data privacy surrounding social media is clearly a salient topic, our study obtained no data from social media (ie, no content from social media profiles was queried or collected). We simply leveraged an online community of patients housed in social media for recruitment and to sustain patient engagement. Our ability to more than double our enrollment goal (target of 300 patients and actual recruitment of 810 patients) was a testament to the effectiveness of including a social media recruitment strategy. In our study, we believe this protocol promoted a deep sense of patient ownership in the study process and dramatically limited attrition (n = 35; 3.3%).