Prevalence of Transgender Depends on the “Case” Definition: A Systematic Review

Study Selection

Two reviewers (LC, MG) independently reviewed each citation. Non-English language articles were translated into English by LC. Each report was assessed according to the following a priori eligibility criteria: (1) reported results with prevalence estimates and measures of variance in the general adult population, (2) reported numerator for prevalence estimate, (3) reported identification of denominator or a description of how prevalence estimate was calculated. Publications that did not report (or did now allow estimating) prevalence in the general population were excluded from the analysis. Some studies were excluded because they reported prevalence of transgender in certain population subgroups such as twins¹² or men-having sex with men,¹³ but not in the general population. Some studies reported frequency of transgender as both prevalence and incidence measures.^{14, 15} In those instances, only prevalence estimates were used because incidence of transgender is difficult to interpret due to a lack of identifiable “date of onset.” To effectively study incidence of transgender in the general population, one needs to systematically collect data on gender identity from large population-based cohorts, preferably recruited at birth. To our knowledge, these data are not available. When two or more studies used the same or overlapping data, only the more recent prevalence estimate was used in the meta-analysis, but all studies were described in the systematic review.

Data Abstraction and Management

The main outcome measure of interest in this review was population prevalence of transgender. Secondary outcomes of interest included gender-specific prevalence estimates for male-to-female (MTF) or female to male (FTM) subgroups.

The two reviewers (LC, MG), abstracted the data from eligible publications. Information was tabulated to allow data management and descriptive analyses. Inconsistencies in the data entry were resolved by consensus. The author of one study¹⁶ was contacted and replied regarding additional information needed for the meta-analysis.

For each publication, data extracted included the size and the case definition of the numerator (e.g., diagnosis of gender dysphoria, or “gender identity disorder,” use of hormonal therapy, or history of gender affirmation surgery), the size and the type of denominator (e.g., general population of a given area, or a total number of participants in a survey), and the overall and gender-specific prevalence estimates. In order to apply meta-analytic techniques, the standard errors for each prevalence estimate were either extracted from the articles or calculated from the available information using OpenEpi on-line statistical software.¹⁷

Data Synthesis

As the language used to describe transgender has evolved considerably in the literature over the last few decades, the original terminology applied to the case definition in each publication was maintained so as not to infer meaning beyond what was stated by the authors.

Eligible publications reporting transgender population prevalence estimates were then grouped into three categories as shown in the Appendix. These categories included: 1) studies that reported prevalence of surgical or hormonal gender affirmation therapy; 2) studies that defined numerator based on the diagnosis of “transsexualism,” “gender identity disorder” or “gender dysphoria”; and 3) studies that assessed self-reported gender identity which differed from the assigned sex. All prevalence estimates were standardized and expressed per 100,000 to facilitate comparisons. Studies that used methods inconsistent with any of the above definitions were examined and discussed separately, but not included in the meta-analysis. Data extracted from publications were analyzed to calculate a pooled summary estimate of transgender prevalence weighted by sample size. Heterogeneity was assessed by a Q-test and its derivative an I² measure.¹⁸ The summary statistics were calculated using random effects models and the results were expressed as meta-prevalence (mP) estimates with the corresponding 95% confidence intervals (CI). In addition, subgroup and meta-regression analyses were performed in search for sources of heterogeneity. Variables that were considered as possible sources of heterogeneity included: year the study was conducted, geographic region, and the type of publication (peer-reviewed versus not peer-reviewed). Effect of outliers on the mP estimates was also assessed. All statistical analyses were performed using Stata software package (version 13.0, Stata Corporation, College Station, TX, USA).

Overview of studies

As shown in Figure 1, 1,470 references were screened by title and abstract, and of those 89 publications underwent full-text review. Of the 32 publications included in the final qualitative review, two studies^{19, 20} were based on non-independent data and could not be incorporated in a meta-analysis. Three additional studies^{15, 21,22} (two of those based on the same data) were not included in a meta-analysis because they used unique case definitions and could not be grouped with other studies. As a result, 27 studies were used in the meta-analysis.

Prevalence of Gender Affirmation Surgery and Hormone Therapy

As shown in Table 1, six publications have estimated prevalence of transgender by considering only those individuals who sought or received gender affirmation surgery, previously known as “sex reassignment surgery” (SRS). Only one of these reports was from the US, four were from Europe, and one from Singapore. The publication dates ranged from 1968 to 2014.

Based on the data from various specialized centers (most notably the Gender Identity Clinic at the Johns Hopkins Hospital), Pauly identified 2,000 MTF and 500 FTM people who requested SRS since 1953.²³ Using the total US population in 1968, the resulting prevalence was estimated at 1/100,000 for MTF and 0.25/100,000 for FTM. Unlike other similar studies, which calculated prevalence of MTF and FTM among natal males and females, respectively, Pauly divided each numerator by the total population.

In 2011, Esteva et al. sent out questionnaires to regions of Spain that have established Gender Identity Units.²⁴ In the questionnaires, the authors inquired about the number of applicants for SRS. Based on responses from all clinics, 3,303 individuals solicited genitoplastic reconstruction or other procedures such as breast augmentation and facial surgeries. Using the total Spanish population between ages 15 and 65 years as the denominator, the prevalence estimate was reported as 10/100,000.

A study in Belgium estimated the proportion of the population that had undergone SRS from 1985 until 2003 through retrospective collection of data on procedures performed by plastic surgeons and gender teams.²⁵ Among the gender teams and plastic surgeons that agreed to participate (response rate 24%), the study identified 412 Belgian-born “transsexuals” (292 MTF and 120 FTM) corresponding to prevalence estimates of 7.74/100,000 and 2.96/100,000, respectively.

Based on the information available from Italian clinics performing SRS procedures, Caldarera and Pfäfflin identified 424 MTF and 125 FTM surgeries performed between 1992 and 2008.²⁶ “SRS” procedures in that study referred to genital surgeries only. Using data from the National Institute of Statistics to estimate the denominator, the authors calculated prevalence of 1.5/100,000 and 0.4/100,000 for MTF and FTM, respectively. The authors acknowledged that their estimates were lower than those reported in other countries and attributed the difference to missing data or cultural factors.

A study in Sweden identified requests for SRS based on the data from the National Board of Health and Welfare Statistics. ²⁷ There were 767 applicants for the SRS (478 MTF and 289 FTM); of those 681 were approved and underwent the procedure (429 MTF and 252 FTM). Using the Swedish population as of December 2010, the prevalence estimates for applications for SRS were 12.9/100,000 for MTF and 7.5/100,000 for FTM. The corresponding prevalence estimates limited to those who underwent SRS were 11.6/100,000 for MTF and 6.6/100,000 for FTM.

Only one study assessing prevalence of SRS was conducted outside of North America or Europe. Tsoi calculated prevalence of “transsexualism” by identifying patients in Singapore who had sought SRS and who were subsequently diagnosed as “transsexuals” by psychiatrists.²⁸ The author reported that up until 1986, there were 458 (343 MTF and 115 FTM) Singapore-born “transsexuals” reported by the Department of Obstetrics and Gynecology and by private surgeons. Using the total male and female population of Singapore in 1986, the prevalence was reported as 35.2/100,000 and 12.0/100,000 for MTF and FTM, respectively. The author cites cultural acceptance of “transsexuals” and more established “sex reassignment” surgical procedures as possible reasons for the higher than previously reported prevalence estimates.

Two studies conducted in the Netherlands at the Free Amsterdam University clinic used hormonal therapy among “transsexuals” to define prevalence of transgender (Table 1). In 1976, the clinic established a gender team and based on data collected through 1986, 538 individuals had received hormonal therapy at that facility.¹⁶ Of those, 399 were MTF and 139 were FTM. Using the Dutch Bureau of Statistics data for population estimates, the prevalence was calculated as 5.6/100,000 for MTF and 1.9/100,000 for FTM.

In a more recent study based at the same clinic, the analysis was extended through the end of 1990.²⁹ By that time, the clinic was providing hormonal therapy to 713 Netherlands-born “transsexual” patients over the age of 15 years; 507 MTF and 206 FTM. The total population of the Netherlands from 1990 was used to calculate the prevalence estimates of 8.4/100,000 for MTF and 3.3/100,000 for FTM.

Prevalence of transgender-related diagnoses

Ten studies calculated prevalence of documented transgender-related diagnoses using International Classification of Disease (ICD) or the Diagnostic and Statistical Manual of Mental Disorders (DSM), but application of codes varied across the studies (Table 2). The terminology of transgender-related conditions and the diagnostic criteria have also evolved over the years.³⁰

Hoenig & Kenna aimed to identify all patients diagnosed as “transsexual” at the University Department of Psychiatry at the Royal Infirmary, Manchester between 1958 and 1968.³⁰ The study relied on referral of patients from other clinics. The clinic identified 66 individuals. Using the population of the Manchester Region on June 30, 1970, the prevalence estimates were calculated as 1.9/100,000 overall, 2.8/100,000 for MTF, and 0.9/100,000 for FTM.

In a study based in Northern Ireland, O’Gorman collected information from the Department of Mental Health, Queen’s University, Belfast.³¹ Twenty-eight individuals diagnosed as “transsexual” were identified (21 MTF and 7 FTM) over an unspecified fourteen year period. Using an approximation of the population in Northern Ireland, the overall prevalence was estimated at 1.9/100,000.

Esteva et al. estimated the prevalence of “gender identity disorder” based on clinic data from the Gender Identity Disorder Unit in Andalucía, Spain.³² The clinic reported that from its opening in 1999 until October 2004, they had seen a total of 391 individuals with this diagnosis, 243 MTF and 148 FTM. Using the total population of Andalucía in 2004, the corresponding prevalence estimates were reported as 10.3/100,000 for MTF and 6.5/100,000 for FTM.

In another Spanish study, Gomez Gil et al. estimated the prevalence of “transsexualism” in Catalonia using ICD-10 diagnostic code of F64.0 documented at the Hospital Clinic in Barcelona from 1996 through 2004.³³ The authors identified 161 patients (113 MTF and 48 FTM) who were living in Catalonia and based on population data from the regional Institute of Statistics, the prevalence estimates for MTF and FTM were calculated as 4.8/100,000 and 2.1/100,000, respectively. The authors also calculated the Barcelona-specific prevalence of 5.5/100,000 for MTF and 2.5/100,000 for FTM indicating a higher proportion of transgender people in the urban population.

In Japan, the prevalence of “gender identity disorder” was determined from the outpatient clinic of Okayama University hospital between April 1997 and October 2005.³⁴ Using the DSM-IV criteria, 579 patients (230 MTF and 349 FTM) constituted the numerator. The prevalence estimate for FTM was reported as 0.9/100,000.

Another Japanese study sought to assess the prevalence of “gender identity disorder” based on data collected at the specialized clinic of the Sapporo Medical University Hospital between December 2003 and January 2010.³⁵ Identification of cases was based on diagnoses using ICD-10 and DSM-IV codes. The study identified 343 patients (104 MTF and 238 FTM). Using the population of Hokkaido as the denominator, the MTF and FTM prevalence estimates were 3.97/100,000 and 8.20/100,000 respectively. The authors conclude that FTM may represent a greater proportion of transgender people in Japan then in other parts of the world, which is in agreement with the previous report from that country.³⁴

Ahmadzad-Asl et al. aimed to estimate the prevalence of “gender identity disorder” in Iran between 2002 and 2009 through review of records at the Tehran Psychiatric Institute and identification of subjects with a diagnosis of gender identity disorder according to the DSM-IV criteria.³⁶ A total of 281 individuals (138 MTF and 143 FTM) were identified yielding prevalence estimates of 0.7/100,000 total, 0.69/100,000 for MTF and 0.74/100,000 for FTM. The authors postulate that the patriarchal socio-cultural characteristics in Iran may explain the roughly equal numbers of MTF and FTM.

Using the DSM-IV and DSM-V criteria, Judge and colleagues collected information on patients diagnosed with “gender dysphoria” at the Department of Endocrinology of St. Columcille’s Hospital in Dublin between 2005 and 2014.³⁷ Among 218 patients referred to the clinic, 159 were MTF and 59 were FTM. Based on the 2011 census data, the prevalence estimates were reported as 6.8/100,000 total, 9.88/100,000 for MTF and 3.6/100,000 for FTM.

In a recent study conducted in the United States, Blosnich et al. used the Veterans Health Administration (VHA) electronic medical records from 2000 through 2011 to examine the prevalence of transgender-related diagnoses among veterans.²⁰ The numerator for the study included individuals who received ICD-9 diagnostic codes of 302.85 or 302.6 to define “gender identity disorder”. Using the VHA electronic record database to define the denominator for each year, the 2002 prevalence estimate was 12.5/100,000 and the prevalence reported in 2011 was 22.9/100,000, indicating an almost two-fold increase over the ten-year study period. Although these data did not distinguish MTF from FTM, it is important to note that the VHA population is 95% natal males.

In a more recent VHA-based publication, Kauth et al. used ICD-9 codes 302.85, 302.6, and 302.5 (transsexualism) identified between 2006 and 2013.³⁸ Prevalence in 2013 was reported as 32.9/100,000 (95% CI: 21.6, 44.1). This result is not directly comparable to those reported in the Blosnich et al. study due to expanded criteria for the case definition. In addition, the denominator in the Kauth et al study included all VHA enrollees while Blosnich et al. calculated the prevalence among VHA care utilizers.

Table 2 also includes three studies (from Sweden, Australia, and Scotland), which calculated the prevalence of transgender-related diagnoses by surveying clinics specializing in treatment of transgender patients. All three studies present relatively old data with publication dates between 1968 and 1999.

Walinder estimated the prevalence of “transsexuals” in Sweden in 1968 through a survey of psychiatric clinics.³⁹ Seventy-six percent of the psychiatrists reported providing treatment to a total of 67 “transsexuals”. After including 43 additional individuals known to the author, the numerator used for calculations was 110. Using census data from Sweden, the prevalence estimates were calculated as 1.9/100,000 total; 2.7/100,000 for MTF and 1/100,000 for FTM.

Prevalence of “transsexuals” in Australia between June 1976 and June 1978 was estimated through distribution of questionnaires to subscribers of the Australian and New Zealand Journal of Psychiatry.⁴⁰ According to the authors, all psychiatrists in Australia receive this journal. Of 904 questionnaires distributed, 263 were returned (29.1%). Based on the completed questionnaires, 243 “transsexual” individuals (209 MTF and 34 FTM) were identified. Prevalence estimates were calculated using the population of Australia 15 years of age or older as of June 31, 1978; and were determined to be 2.4/100,000 total, 4.2/100,000 for MTF and 0.7/100,000 for FTM.

Wilson et al. surveyed general medical practices in Scotland in 1998.⁴¹ The questionnaire obtained information regarding the number of patients registered to the practice as well as the number of patients with “gender dysphoria.” Just under three-quarters (73%) of the surveys were completed and returned, identifying 273 eligible patients. The denominator for calculation of the prevalence was based on the number of patients over the age of 15 years registered to the respondent’s practices standardized to match the age distribution of the general population in Scotland. The resulting population prevalence of gender dysphoria was 8.2/100,000.

Two studies (one in Iceland and one in Taiwan) ascertained prevalence of “transsexualism” using diagnostic interviews of the general population.^{42. 43} Both studies reported prevalence estimates that far exceeded those obtained from clinics or from surveys of health care providers (Table 2)

Prevalence of Transgender Identity

Six studies utilized survey-based data to estimate prevalence of gender identity that differed from the binary sex categories assigned at birth (Table 3). Three of these studies were based in the United States, one in the Netherlands, one in Belgium, and one in Taiwan.

Conron et al. analyzed data collected between 2007 and 2009 from the Massachusetts Behavioral Risk Factor Surveillance Study (MA-BRFSS).⁴⁴ The survey was administered to 28,662 adults ages 18 to 64 years. Each participant was asked: “Some people describe themselves as transgender when the experience a different gender identity from their sex at birth. For example, a person born into a male body, but who feels female or lives as a woman. Do you consider yourself to be transgender?” There were 131 participants who responded ‘yes’ to that question, corresponding to a prevalence of 500/100,000 persons. The authors acknowledge that their estimate may have been affected by misclassification bias due to the broad scope of the question. Another limitation of the data is a lack of information on natal sex and gender identity.

Gates combined reports from the 2003 California LGBT Tobacco Survey, and the 2009 California Health Interview survey in order to estimate the number transgender adults in the general populaiton.^{16, 44} The 2009 California Health Interview Survey reported that 3.2% of adult participants were LGBT. Based on the 2003 California LGBT Tobacco Survey, 2.4% of LGBT persons self-identify as transgender. Using these considerations Gates further estimated that the prevalence of transgender people among California adults is 0.1% or 100 per 100,000.

A 2010 survey used the Growing Up Today (GUTS) prospective cohort study of US young adults.⁴⁵ The survey implemented a two-step approach by first inquiring about sex assigned at birth, and then asking the participants about their self-described gender identity with the response options “Female,” “Male,” “Transgender, “or “Do not identify as female, male or transgender.” Among the 7,831 survey respondents, 26 (0.33%) identified as having gender identity that was different from their assigned (natal) sex. Of those, 7 (0.09%) were cross-sex identified, 5 (0.06%) self-described as transgender, and 14 (0.18%) did not identify as female, male, or transgender.

Using a 2012 internet-based sexual health survey of persons 15 to 70 years of age, Kuyper and Wijsen estimated the proportion of transgender people in the Netherlands.⁴⁶ The study aimed to identify individuals with “incongruent gender identities” and “gender dysphoric feelings.” The final sample included 8,064 participants, who were asked questions regarding gender identity and gender dysphoric feelings and the responses were recorded on a Likert Scale. About 20% of those invited to participate completed the questionnaire and met the eligibility criteria. Although the exact case definition is not clear, among natal males the reported prevalence was 600/100,000 and among natal females the corresponding estimate was 200/100,000.

Van Caenegem et al. estimated prevalence of “gender nonconformity” based on a population survey of adolescent and adult (age range 14–80 years) residents of the Flanders region, Belgium.⁴⁷ Survey participants were randomly selected from the Belgian National Register. Forty percent of eligible persons completed the survey. The denominator for prevalence calculations included 1,799 participants (864 natal males and 905 natal females). Questions pertaining to gender identity were assessed via a computer assisted personal interview. Based on a 5-point Likert scale (ranging from 1-totally disagree to 5-totally agree) the participants were asked to score statements “I feel like a woman” and “I feel like a man.” A person was considered “gender ambivalent” if the same answer (e.g., a score of 1 or a score of 2) was given to both statements. “Gender incongruence” was defined as a lower score assigned to the natal sex than to the opposite sex. Using these definitions, the prevalence for “gender incongruence” was estimated to be 700/100,000 natal males and 600/100,000 natal females. The corresponding estimates for “gender ambivalence” among natal males and females were 2,200/100,000 and 1,900/100,000, respectively.

A 2009 study of Taiwanese university students conducted interviews with 5,010 participants using the Adult Self-Report Inventory-4, which is a DSM-IV referenced scale.⁴⁸ Self-reported “gender dysphoria” was determined based on a response to the statement “I wish I was the opposite sex.” Responses “often” and “very often” were interpreted as evidence of gender dysphoria. The use of this rather loose definition of gender dysphoria resulted in very high prevalence estimates of 7,300/100,000 for FTM and 1,900/100,000 for MTF.

Prevalence of legal name or sex changes

Two studies, one from Germany and another from New Zealand, reported population prevalence of transgender based on documented administrative sex or name change.

Weitze and Osburg relied on the 1981 German Transsexuals’ Act, which allowed applicants to change their first name or their legal sex status. The study examined the decisions rendered during the first 10 years since implementation of the law.¹⁵ Information regarding the number of relevant applications and corresponding decisions was collected from the courts. During the study period, the courts issued decrees on 683 first name changes and 733 rulings on legal reestablishment of sex. These rulings concerned 1,199 individuals of which 1,047 received the approval. Using the adult population of West Germany before reunification, the prevalence was estimated as 2.1/100,000. A more recent report extended the work of Weitze and Osburg by evaluating change in legal sex status between 1991 and 2000 in all of Germany.²² The overall “transsexual” prevalence was estimated as 3.88/100,000 using the German population in 2000 as the denominator. The MTF and FTM prevalence estimates were 4.95/100,000 and 2.87/100,000, respectively.

In New Zealand, individuals may request a change of their gender code from ‘M’ or ‘F’ to ‘X’. To examine the frequency of this change, Veale contacted the New Zealand Department of Internal Affairs Passport Office.²¹ A total of 385 such changes were identified in 2008, and given the number of passport holders in New Zealand the prevalence was calculated as 16/100,000. Considering that 49% of passport holders in New Zealand were male and 51% were female, the corresponding prevalence estimates were 27/100,000 for MTF and 4.4/100,000 for FTM with a sex ratio of 6:1.

Meta-Analysis

Among studies reporting estimates for individuals who sought or received gender affirmation therapy, the summary analysis revealed a total mP of 9.2/100,000 (95% CI: 4.9, 13.6) with a range between 0.9 and 35.0 (Table 4). The studies of surgical gender affirmation were further sub-categorized based on whether SRS was received or requested with mP (95% CI) estimates of 5.5 (0.5, 10.5) and 9.2 (5.2, 13.3) per 100,000, respectively. Significant heterogeneity was present in all of the analyses (I² >99%). Meta regression demonstrated no significant time trend (regression coefficient = −0.33, p-value= 0.28).

A meta-analysis of studies that assessed prevalence of transgender-related diagnoses produced an overall mP estimate of 6.8/100,000 (95% CI: 4.6, 9.1); 5.8/100,000 (95% CI: 3.5, 8.1) for MTF and 2.5/100,000 (95% CI: 1.9, 3.1) for FTM (Table 5). Meta-regression demonstrated that study location, categorized as Europe, Asia, or Other (including US) was a predictor of prevalence (regression coefficient = 3.11, p-value<0.01). By contrast there was no association with year of publication (regression coefficient = −0.25, p-value=0.18).

As shown in Table 6, among studies that examined prevalence of transgender identity, the summary estimates were much higher than in the previous two groups. Although the results appeared heterogeneous in the overall analysis, the heterogeneity was primarily driven by one outlier study.⁴⁸ After this outlier study was removed from the analyses, the results for MTF and FTM were highly homogenous (I² <1%).