Mental Health of Youths Who Use Puberty Blockers.

This retrospective cohort study used deidentified, individual-level data extracted from the Mount Sinai Institute for Health Equity Research Multi-Payer Claims Database (IHER MPCD), composed of administrative claims from the HealthVerity Private Source 20 database. The IHER MPCD includes multiple payer types, covering approximately 70 million commercial, 60 million Medicaid, and 16 million Medicare Advantage enrollees distributed across 150 payers from January 2016 through January 2025. We subset this dataset to the following mutually exclusive cohorts: Cohort A (transgender, receiving puberty blockers): youths aged 10 to 17 years who had at least 1 prescription for a puberty blocker used for the purposes of pubertal inhibition (see list in eTable 4 in Supplement 1 ), at least 1 International Statistical Classification of Diseases and Related Health Problems, Tenth Revision ( ICD-10 ) diagnosis code typically associated with being transgender, and no ICD-10 diagnosis codes indicating uses of the medication unrelated to puberty suppression (ie, no diagnosis related to precocious puberty, endometriosis, or similar) (see list of codes in eTables 2 and 3 in Supplement 1 ). Cohort B (cisgender, receiving puberty blockers): youths ages 10 to 17 years who had at least 1 prescription for a medication used for the purposes of pubertal inhibition, no ICD-10 diagnosis code typically associated with being transgender, and at least 1 ICD-10 diagnosis code related to precocious puberty. Cohort C (transgender, not receiving puberty blockers): youths aged 10 to 17 years who had no prescription for a puberty blocker and at least 1 ICD-10 diagnosis code typically associated with being transgender. Cohort D (cisgender, not receiving puberty blockers): youths aged 10 to 17 years who had no prescription for a puberty blocker, no ICD-10 diagnosis code typically associated with being transgender, and no ICD-10 diagnosis code indicating precocious puberty; to prevent class imbalance impacts on our models, we subset this cohort through random selection. Our study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology ( STROBE ) reporting guideline. It was reviewed and approved by the Icahn School of Medicine at Mount Sinai Institutional Review Board, which granted a waiver of authorization for the use and disclosure of protected health information and determination of exempt status under 45 CFR § 46.104(d)(4). ICD-10 diagnosis codes related to transgender status were identified from literature. Prescriptions were encoded in the database as National Drug Codes. We initially identified NDCs for puberty blockers via literature review and supplemented this based on mappings between medication names and codes present in the Observational Health Data Sciences and Informatics database, Athena. 15 All codes were then reviewed by experts with relevant clinical experience (T.G.G., S.T.J., D.S.H., and K.K.). Exclusion codes for precocious puberty, endometriosis, and other conditions for which puberty blockers could be used were based on literature and reviewed by the study team for completeness. All codes are included in eTables 2, 3, and 4 in Supplement 1 . For each cohort, we characterized mental health diagnoses using ICD-10 codes. The broad categories included conversion and dissociative disorders, eating disorders, emotionality, impulse disorders, mood disorders, obsessive-compulsive disorder, other nonpsychotic mental disorders (such as depersonalization-derealization syndrome and pseudobulbar affect), phobic disorders, sleep-related disorders, social determinants of health, somatoform disorders, substance use disorders, and suicidality. The full code list is available in eTable 1 in Supplement 2 . Sociodemographic covariates included age, patient sex or gender, race, ethnicity, region of longest residence, and longest-duration payer type in the respective year. In demographics, female gender marker is used to indicate that an F (female) has been applied as a marker for the individual by some entity (eg, a medical professional or government body). Participant race and ethnicity were as reported in the database from claim providers and mapped to the Observational Medical Outcomes Partnership Common Data Model; the ultimate provenance of these data is unknown and is likely a combination of self-reported and clinician-observed race and ethnicity. We used the Movement Advancement Project (MAP) Gender Identity Policy Tally percentage, finding the mean by year, as a proxy for transgender equality on a state and territory level (hereafter, regional transgender equality ), and the Franklin and Marshall Global Barometer of Transgender Rights (FM GBTR) as a proxy for transgender equality nationally (at the time our modeling was being performed, GBTR scores existed for only 2011 through 2023; for 2024 and 2025, we calculated a ratio between Equaldex Equality Index scores and GBTR scores and used that to estimate GBTR values), which we hereafter refer to as national transgender equality . We used the Social Deprivation Index (SDI), 16 a composite measure of area-level deprivation, as an estimate for demographic characteristics (eg, poverty rate, education level, unemployment rate, and crowded housing rate) not available in the IHER MPCD. The theoretical bases for the inclusion of each covariate are included in eTable 7 in Supplement 1 . Descriptive statistics were used to characterize the distribution of variables across the 4 cohorts. We also descriptively compared the rate of individuals with a transgender-associated ICD-10 diagnosis code to prior rates calculated by Hughes et al 9 (eTable 5 in Supplement 1 ). For quantitative analysis, inverse probability–weighted (IPW) generalized estimating equation (GEE) models were built with binary outcomes (yes or no) for mental health diagnoses that were significantly different between the 4 cohorts based on primary exposures of interest: gender modality (transgender or cisgender) and prescription of puberty blockers (yes or no). We performed χ 2 tests to assess associations between primary exposures and all covariates across mental health diagnosis categories and subcategories. We used IPW GEEs instead of traditional GEEs because data were determined to be missing at random. 17 , 18 This determination was based on the Little missing completely at random test 19 and multiple logistic regression models assessing missing-at-random assumptions using the missR package in R statistical software version 4.4.0 (2024-04-24 ucrt) (R Project for Statistical Computing) run within RStudio version 2026.01.2 (RStudio). 20 Missingness information broken down by attribute is shown in eTable 6 in Supplement 1 ; 28% of approximately 1.85 million patient-years were missing. Probability-based weights were calculated using gender modality, prescription of puberty blockers, age, reported patient sex or gender, race, ethnicity, and region of longest residence, all in the respective year. Weights were stabilized using marginal probability of completeness in the data and trimmed at first and 99th percentile values. The base model included primary exposures and the following covariates: age, year in study (scaled by dividing by 2), patient sex or gender (given the provenance of the data, it is unknown if this datum reflects gender identity, assigned gender at birth, legal gender, or something else), race (Asian, Black, White, other, and unknown), ethnicity (Hispanic, non-Hispanic, and unknown), days covered by insurance in respective year (scaled by dividing by 100), longest-duration payer type in respective year (commercial, Medicaid, and Medicare Advantage), region of longest residence in respective year (Northeast, Southeast, Midwest, West, and Pacific), and SDI based on first 3 digits of zip code of location with longest time in residence in the respective year weighted by population (scaled by dividing by 10). Patient sex or gender in the database was provided by the supplier and was ultimately decided by the most common value (between male or female) if present; it was listed as unknown if in equal proportion between male or female, if another value was provided, or if no value was provided. Race in this database was limited to Asian, Black, and White, and any other value was included as other. Interaction terms were selected using the quasilikelihood information criterion (QIC) as selection criteria. For each outcome, an autoregressive lag-1 correlation structure was fitted iteratively, with each iteration testing 1 candidate interaction term aside from base model terms. Interaction terms tested included transgender status with puberty blocker usage, transgender status with year in study (scaled), transgender status with race and ethnicity, transgender status with weighted SDI score (scaled), race with weighted SDI score (scaled), ethnicity with weighted SDI score (scaled), and transgender status with regional transgender equality (time varying and continuous, scaled by quartiles within each year, given that calculations between years are not necessarily comparable), and transgender status with national transgender equality. Each interaction was selected based on existing literature; for more information, see eTable 7 in Supplement 1 . An interaction was retained in the model if it reduced the QIC value, indicating improved model fit; otherwise, it was discarded. This process continued sequentially through all candidate interactions, with each iteration building on previously selected terms. An α level of .05 served as the criterion for statistical significance for all analyses. All P values and other tests were 2-sided. We also used the R package geepack version 1.3.13 (Søren Højsgaard, Ulrich Halekoh, Jun Yan, and Claus Thorn Ekstrøm) for our IPW GEE models.

The study population consisted of 231 783 youths, including 41 472 youths (17.89%) who were transgender (median [IQR] age at baseline, 9.00 [7.00-11.00] years; 31 246 with female gender marker [75.34%]; 814 Asian [2.3%], 2694 Black [7.7%], 29 257 White [83.9%], and 2088 other race [6.0%] among 34 853 with race data; 10 615 Hispanic [43.8%] among 24 208 with ethnicity data) and 190 311 youths (82.11%) who were cisgender (median [IQR] age at baseline, 8.00 [6.00-10.00] years; 128 942 with female gender marker [67.75%]; 5702 Asian [3.8%], 36 269 Black [23.9%], 97 341 White [64.2%], and 12 364 other race [8.2%] among 151 676 with race data; 56 289 Hispanic [49.9%] among 112 786 with ethnicity data). Among them, cohort A (transgender, puberty blockers) included 1260 individuals (median [IQR] age, 9.00 [8.00-11.00] years; 649 with female gender marker [51.51%]), cohort B (cisgender, puberty blockers) included 78 111 individuals (median [IQR] age, 9.00 [8.00-10.00] years; 74 576 with female gender marker [95.47%]), cohort C (transgender, no puberty blockers) included 40 212 individuals (median [IQR] age, 9.00 [7.00-11.00] years; 30 597 with female gender marker [76.09%), and cohort D (cisgender, no puberty blockers) included 22 449 855 individuals across the entire database before subsetting. Cohort D’ (D-prime), the subset of cohort D used in our analyses, included 112 200 individuals (median [IQR] age, 7.00 [5.00-9.00] years; 54 366 with female gender marker [48.45%]) ( Table 1 ). Overall (among the population including the original cohort D), there were 40 212 transgender youths (0.18%; cohorts A and C) and 22 527 966 cisgender youths (99.82%; cohorts B and D), among whom 1260 transgender youths (3.04%) were prescribed puberty blockers (cohort A) and 78 111 cisgender youths (0.35%) had a puberty blocker prescription (cohort C). While reported patient sex and gender was not definitively defined in our database among transgender patients, among cisgender patients using puberty blockers (cohort B), most patients (nearly 96%) were female. Among transgender patients using puberty blockers (cohort A), the patient sex or gender ratio was approximately even, and this was also true among cisgender patients not using puberty blockers (cohort D). Among transgender patients not using puberty blockers (cohort C), the sex or gender ratio was majority female (76.09%). Racial demographics skewed White among transgender subpopulations (981 of 1189 patients who were not missing race data who reported using puberty blockers [82.51%] and 28 276 of 36 411 patients who were not missing race data who reported not using puberty blockers [77.66%]). Identification as Hispanic remained relatively consistent across all 4 cohorts ( Table 1 ). Cohort D’ was the subset of cohort D used in analyses. Reported gender markers for each population are expanded upon in eTable 10 in Supplement 1 , and percentages and subsample sizes for each outcome type and subtype of interest by reported sex or gender category are shown in eTable 11 in Supplement 1 . Usage of different puberty blocker types by gender modality (cisgender or transgender) are shown in eTable 9 in Supplement 1 . The prevalence of transgender youths using puberty blockers between 2018 and 2022 was between 1 and 4 per 100 000 (0.001%-0.004%) (eTable 5 in Supplement 1 ). Mental health diagnoses were more likely among transgender youth patients across all categories ( Table 2 ). These differences were statistically significant across all categories except eating disorders, somatoform disorders, and suicidality. The largest observed differences were between mood disorders (13 452 of 79 371 patients using puberty blockers [16.95%] vs 19 826 of 152 412 patients who did not [13.01%]), sleep-related disorders (2979 patients using puberty blockers [3.75%] vs 3688 patients not using puberty blockers [2.42%]), and social determinant–related ICD-10 codes (1419 patients using puberty blockers [1.79%] vs 1790 patients using puberty blockers [1.17%]). Between transgender cohorts that did and did not use puberty blockers (A and C), the only significant difference observed was in suicidality (2273 of 40 212 transgender youths without puberty blockers [5.65%] vs 46 of 1260 with puberty blockers [3.65%]) and mood disorders categories (12 294 youths [30.57%] vs 426 youths [33.81%]) ( Table 2 ). All outcome type and subtype differences based on gender modality (cisgender or transgender) are shown in eTable 8 in Supplement 1 . Therefore, these were the outcomes we examined further in our GEE models ( Table 3 ). Comparing transgender vs cisgender youths, we observed the largest differences in mood disorders (12 720 of 41 472 transgender youths [30.67%] vs 20 558 of 190 311 cisgender youths [10.80%]), suicidality (1735 transgender youths [5.59%] vs 2319 cisgender youths [0.91%]), and sleep-related disorders (2010 transgender youths [4.85%] vs 4657 cisgender youths [2.45%]). Abbreviations: puberty blocker, puberty-inhibiting medication; suicidality, suicidal thoughts and behaviors. Using the Wald method, 95% binomial CIs were calculated. P values were calculated across mutually exclusive groups (transgender vs cisgender and transgender with puberty blocker prescription vs transgender with no puberty blocker prescription) using a 2-proportion z test. Cohort D’ was the subset of cohort D used in analyses. Abbreviations: aOR, adjusted odds ratio; GI, gender identity; MAP, Movement Advancement Project; puberty blocker, puberty-inhibiting medication; suicidality, suicidal thoughts and behaviors. Inverse probability–weighted generalized estimating equation model results are shown, with outcomes modeled as 1 (has diagnosis) vs 0 (does not have diagnosis). For conciseness, factors not significant across both models are not shown. The referent group was cisgender for gender, false for puberty blocker, male for reported patient sex or gender, White for race, not Hispanic for ethnicity, and Northeast for region. Scaled by quantile within year. MAP tallies were not selected in the mood disorder diagnosis model because they did not improve model quasilikelihood information criterion values. All factors found to be significant in our IPW GEE models are shown in Table 3 . Being transgender was associated with higher odds of mood disorder diagnosis (aOR, 4.42 [95% CI, 4.24-4.60]) and suicidal thoughts and behaviors (aOR, 12.21 [95% CI, 10.69-13.95]) compared with being cisgender. Being prescribed a puberty blocker was associated with higher odds of mood disorder diagnosis (aOR, 2.11 [95% CI, 2.02-2.19]) and suicidal thoughts and behaviors (aOR, 3.04 [95% CI, 2.65-3.49]) compared with no prescription. However, among transgender youths, prescription of a puberty blocker was associated with decreased adjusted odds of both outcomes (mood disorder diagnosis: aOR, 0.52 [95% CI, 0.45-0.60]; suicidality: aOR, 0.20 [95%CI, 0.14-0.29]) (eTables 12 and 13 in Supplement 1 ).

To our knowledge, this cohort study is the first study to use claims data in a longitudinal design to compare transgender and cisgender youths’ use of puberty blockers and the association with mental health disorder diagnoses. In this US nationwide study of more than 230 000 youths across 4 cohorts, transgender cohorts (A and C combined) had higher mental health–related diagnostic prevalence vs nontransgender cohorts (B and D combined), corroborating the extensive literature documenting elevated mental health burden among transgender people. 1 , 2 , 21 The largest observed differences in transgender youths vs cisgender youths were in mood disorders (30.67% vs 10.80%), suicidality (5.59% vs 0.91%), and sleep-related disorders (4.85% vs 2.45%). However, contrary to prior literature, 22 , 23 , 24 , 25 there were no statistically significant differences by gender modality in conversion and dissociative disorders, other nonpsychotic mental disorders, somatoform disorders, or substance use disorders. It has been suggested that much of the disparity in mental health burden is due to minority-stress processes, 26 , 27 such as bullying, discrimination, internalized stigma, familial or societal rejection, bullying, and related factors. 28 Results of our IPW GEE models indicated that transgender youths had a 4.42 higher adjusted odds of mood disorder and a 12.21 higher odds of suicidality diagnosis compared with cisgender youths. Use of puberty blockers was also associated with increased adjusted odds of a mood disorder (aOR, 2.11) and suicidality (aOR, 3.04). However, among transgender youths only, puberty blockers attenuated both outcomes (mood disorder: aOR, 0.52; suicidality: aOR, 0.20). Notably, this attenuation did not return odds to baseline (represented by an aOR of 1 among cisgender youths not using puberty blockers) but instead reduced them, resulting in a slightly higher odds of a mood disorder diagnosis and a markedly reduced odds for suicidality compared with transgender youths not using puberty blockers. This pattern may indicate that while puberty blockers may be associated with mitigation of some distress associated with undesired pubertal changes, their use in isolation may not fully address complex psychosocial stressors at play. One possible explanation for this remaining elevated adjusted odds is that both groups (transgender youths awaiting gender-affirming hormone therapy and cisgender youths experiencing early puberty) may experience a form of developmental asynchrony relative to their peers. For youths experiencing central precocious puberty, existing psychological and social burdens may impair coping mechanisms for dealing with early puberty, potentially exacerbating negative mental health outcomes (termed personal-accentuation and contextual-amplification hypotheses). 29 Likewise, in youths experiencing delayed puberty, studies have observed higher rates of low self-esteem, depression, peer stress, social isolation, and bullying. 30 Taken together, these findings suggest that it is possible that earlier initiation of gender-affirming hormone therapy for transgender youths, so that they may experience puberty in synchronicity with their peers, 29 , 30 may be associated with reduction or elimination of observed disparities compared with cisgender youths not using puberty blockers. Further research in this area is warranted. This study has several limitations. The potential explanation that early hormone therapy initiation may be associated with outcomes for transgender youth similar to those of their peers should be interpreted with caution. Observed similarities may reflect unmeasured confounding, differences in health care access or family support, selection bias inherent in retrospective cohort studies, or biases related to using claims datasets (eg, external validity concerns). 31 However, we were unable to determine specific levels of health care access for individuals in this sample or any enabling factors that may be associated with access and use of care. Additionally, if the experiences of transgender youths using puberty blockers mirrored those of cisgender youths using puberty blockers, we would have expected odds to approximate those observed in cisgender youths using puberty blockers (2.11 and 3.04), which was not the case. Prospective studies examining the timing, sequence, and precise psychosocial context of pubertal suppression and gender-affirming hormone therapy initiation are needed to clarify whether earlier initiation of gender-affirming hormone therapy may be associated with better developmental and mental health alignment for transgender youths with their peers. Similarly, given that administrative data do not contain clinical data beyond what care was billed, we were unable to assess potential treatments outside of traditional care contexts (eg, support groups) that may have influenced mental health outcomes. Additionally, claims data are merely observational and do not allow for determining definitive causes and can show only associations. Furthermore, our model accounted for only 2 axes: gender modality (cisgender or transgender) and puberty blocker prescription (yes or no). We did not test for central precocious puberty as an additional axis. Therefore, our results cannot be used to determine the association between puberty blocker prescription and central precocious puberty diagnosis. Additionally, our model used a greedy forward selection algorithm to select interaction terms, meaning that our selected model may not represent the global minimum QIC. Notably, we observed lower rates of prescribed puberty blocker use among transgender youths (0-4 per 100 000) than previously published estimates in Hughes et al (4-19 per 100 000). 9 This difference is likely due to a more generous set of inclusion criteria in Hughes et al. 9 For example, if a person had a transgender-related diagnosis (eg, gender dysphoria) and an exclusionary diagnosis (eg, precocious puberty or central precocious puberty) and a puberty blocker prescription, we excluded that individual from cohorts A and B given that it was not possible to determine whether puberty blockers were prescribed as part of a gender-affirming care regimen or for another diagnosis. Furthermore, Hughes et al 9 included only commercial insurance, whereas our IHER MPCD population included additional, noncommercial insurance types. However, the IHER MPCD may not be representative of the US population, and neither database is inclusive of all insurance types. Therefore, the true population estimate for puberty blocker use may lie somewhere between our lower estimate and the higher estimate of Hughes et al, 9 both of which were ultimately very low (<20 of 100 000 at the upper 95% CI bound).

Results of this cohort study broadly align with those of other studies to date, showcasing a generally higher mental health burden among transgender youths in comparison with cisgender youths 1 , 2 , 21 and increased rates of diagnoses related to suicidality among transgender youths whose claims did not include a puberty blocker prescription compared with transgender youths whose claims did include this prescription. Similar to Hughes et al, 9 we found that prescription of puberty blockers among transgender youths was rare, between 1 and 4 of 100 000 patients, and use among transgender youths was 3.04%. Importantly, puberty blocker prescription was associated with lower odds of mood disorder and suicidality diagnoses, but only among transgender youths. While this result should be studied further and interpreted with caution, it aligns with prior literature 10 , 32 , 33 and clinical insights. 34 , 35 , 36 More research is necessary to determine the mechanisms behind this association; although puberty blocker prescription appeared to attenuate both outcomes, it was not alone associated with diagnostic odds for transgender youths reduced to levels observed among their cisgender peers. Due to limitations of the administrative claims database being analyzed, we could not adjust for factors like psychosocial support, and policy environment measures (eg, regional and national transgender equality) did not appear to be associated with outcomes, although this may reflect measurement error (eg, scores were not available for the full study period or inconsistently measured from year to year). Ultimately, this analysis advances our understanding of the association between puberty blocker use and mental health diagnoses, provides further evidence countering misinformation implying negative mental health associations for transgender youths receiving this class of medications, and highlights needed future research into the mechanisms by which puberty blocker use may be associated with reduced but not fully eliminated mental health inequities for this population.

In the US, transgender youths face mental health disparities compared with cisgender youths, including mood disorders and suicidal thoughts and behaviors (hereafter referred to using the umbrella term suicidality ). 1 , 2 , 3 Politicization of gender-affirming care (GAC) for transgender youths in the US has led to restrictions on gender-affirming medical and surgical procedures at the state level. 4 A major focus of legislation is banning the use of puberty blockers (ie, puberty-inhibiting medications) for the purpose of GAC. Puberty blockers act as a reversible medical intervention for suppressing the development of undesired secondary sex characteristics. These medications have been used by cisgender children experiencing central precocious puberty since the 1960s. 5 , 6 These same medications were later adapted for use in GAC as part of the Dutch Protocol in the 1990s, specifically in scenarios wherein it is considered too early to begin gender-affirming hormone treatment. 7 , 8 Puberty blockers are used by a relatively small population of individuals. 9 When these medications are used to treat gender dysphoria, they are associated with decreased lifetime suicidal ideation among transgender youths who receive them. 10 Their use is supported by major medical organizations, including the American Academy of Pediatrics, 11 Endocrine Society, 12 Pediatric Endocrine Society, 13 and World Professional Association for Transgender Health. 14 However, the existing evidence base supporting use of these medications has been challenged in the current sociopolitical climate, and there have been calls for additional research on the association between use of these medications and mental health outcomes for transgender youths. To date and to our knowledge, there are no studies that use claims data to assess the population associations of puberty blockers among both transgender and cisgender youths. This study aimed to characterize a US nationwide sample of transgender youths who used puberty blockers (cohort A) and compared their rates of mental health diagnoses with those of transgender youths who did not use puberty blockers (cohort C), cisgender youths who used puberty blockers for treatment of precocious puberty (cohort B), and cisgender youths who did not use puberty blockers (cohort D). We examined rates of use of puberty blockers among transgender youths, whether mental health outcomes of interest were increased among transgender youths in comparison with cisgender youths, and whether mental health outcomes of interest were lower among transgender youths who used puberty blockers vs transgender youths who had not. We hypothesized that transgender youths receiving pubertal blockers would have decreased rates of mood disorders and suicidality than cisgender youths.