1 INTRODUCTION

Treatment-seeking transgender people who are not on hormone treatment have reported high levels of mental health problems, particularly anxiety, depression and self-harm, which are likely caused by a number of internal and external stressors.1-12 Studies examining mental health in transgender people have primarily focused on individuals attending transgender health services and hence those who are likely to experience a higher level of distress about their assigned sex at birth. These studies have primarily looked cross-sectionally at levels of anxiety,7, 13, 14 depression7, 15-17 and self-harm.18-20

With regard to anxiety, several studies have demonstrated high levels in transgender people before gender-affirming hormone treatment (GAHT).21, 22 For example, Bouman et al 13 found that levels of anxiety in transgender people were three times higher than those in a matched sample from the general population. This study also found that transgender males were more anxious than transgender females. Interestingly, the high scores on autistic traits found among this population have been suggested to be a product of the high levels of anxiety and low self-esteem often experienced by this group15 and not autism per se.23 However, a recent study has demonstrated stability in autistic traits following GAHT.24

Similar to anxiety, high levels of depression have also been reported in transgender individuals, prior to GAHT.16, 18, 22 Witcomb et al16 reported that transgender people prior to receiving GAHT had a fourfold increased risk of a probable depressive disorder compared to a matched control sample from the general population. Why this is the case is unclear, but social factors such as lack of general social support,9, 25-27 parental support28 and peer support have been found to be associated with depressive symptoms among transgender people.29, 30 Experiences of transphobic discrimination are associated with increased odds of suffering from depression31 independent of other types of discrimination, for example racism. This suggests that transgender people who are ethnic minorities are at even greater risk, because of the intersectional experience of discriminatory events. In addition, while unemployment increases the risk of depression in the general population32 and transgender people have been found to have a higher unemployment rate than cisgender people33—being in employment is associated with higher levels of experienced transphobia and fear of disclosing mental health problems in the transgender population.34, 35

Another factor that has been associated with mental health problems in treatment-seeking transgender people is age. Younger transgender people report high levels of bullying36 and very high levels of self-harm,19, 37 which have been associated with increased anxiety as well as effects on self-esteem, family relationships and social life, which all negatively influence mental well-being.

While these studies have provided valuable insight, the use of cross-sectional methodologies to examine the impact of the above factors, particularly the role of GAHT in mental health, is limited. Therefore, it is critical to explore this on a within-subject basis using a longitudinal design. This is the most effective approach to show the effects of GAHT on mental health as it provides the opportunity to examine individuals prior to and during GAHT.

A small number of longitudinal studies that focus on the effect of GAHT on mental health do exist. Colizzi et al38 reported significant reductions in mental health symptoms after the initiation of GAHT with anxiety reducing from 50% to 17% and depression from 24% to 11%. Heylens et al39 also showed significant reductions in symptoms of anxiety and depression after the initiation of GAHT to the point where they resemble those of the general population. These studies are, however, not without limitations. Heylens et al’s39 study has a small sample size (n = 57), while Colizzi et al’s38 study is limited by the lack of evaluation of factors that may have impacted on the mental health of their participants, such as social support. Both studies describe the need to replicate their findings. In contrast, Bränström and Pachankis40 using the Swedish population register showed no significant association between the likelihood of accessing mental health treatment and time since initiation of GAHT. The limitation of their study includes primarily that accessing mental health services does not necessarily reflect actual mental health and there is little additional information about the type of mental health treatment received by their participants. These limitations mean that this study cannot provide reliable evidence regarding the role of GAHT on the mental health symptoms of transgender people, and this information is vital in order to provide an evidence base of GAHT improving overall quality of life of transgender people.

While the available longitudinal studies have provided valuable evidence of the effect of GAHT on transgender people's mental health, there is a requirement to replicate these studies addressing their limitations. With this in mind, the primary aim of this study is to examine the effect of GAHT on anxiety and depression symptoms. The study will focus on those who have been on treatment for over 18 months as this allows for enough time for GAHT to produce physical, bodily changes but before surgical procedures have taken place, which could bias the results. As some physical changes can be quicker in assigned females at birth than in assigned males at birth (eg voice change with testosterone),2 which can affect mental health outcome following GAHT, the results of GAHT in anxiety and depression for both groups will be presented separately. It is hypothesized that an improvement in mental health will take place in those assigned male and female at birth following GAHT treatment. Unfortunately, because of the long waiting list for gender-affirming surgical treatment in the United Kingdom (UK), it is unlikely that people will have undergone these interventions before this time. The secondary aim of this study is to examine pre-GAHT factors which may be predicting changes in anxiety and depression following GAHT. The predictors selected for this study are based on the literature and include ethnicity, age, assigned sex at birth, civil status, employment, social support, and autistic traits. This study hypothesized that symptoms of depression and anxiety would be significantly decreased after 18 months of GAHT.

2 MATERIALS AND METHODS 2.1 Participants

Participants were invited to take part through a national transgender health service in Nottingham, UK. This service is part of the National Health Service (NHS) and offers assessment for suitability of GAHT as well as chest and genital reconstructive surgery. The service also offers GAHT and speech and language therapy. The service accepts referrals from people aged 17 and over who are seeking, or considering, medical transition.

2.2 Procedures

The sample consisted of individuals who attended an assessment at the transgender health service from November 2014 to March 2018, who agreed participation and who were not on GAHT prior to the assessment. Prior to the clinical assessment, every patient was invited to participate in the study. If agreed, they were invited to complete a baseline questionnaire pack (T0). The pack included a socio-demographics questionnaire (age, sex assigned at birth, gender identity, ethnicity, employment status, relationship status and whether participants were taking cross-sex hormones and/or blockers pre-assessment—as a significant proportion of young people are referred from the only existing child and adolescent transgender health services in the United Kingdom). Validated questionnaires regarding anxiety and depression (HADS), social support (MSPSS) and autistic traits (AQ-Short) were also included in the information pack. Data were only included if participants returned a signed consent form with the study questionnaires.

Participants who consented and returned T0 questionnaires were invited to complete a T1 questionnaire 18 months after commencing GAHT. The T1 questionnaire pack consisted of a HADS questionnaire. This allowed a comparison of changes in depression and anxiety symptoms before and after GAHT. Data were collected in October 2019. Except for the data analysis, the study was primarily unfunded and set up in a busy clinic.

2.3 Tools 2.3.1 The Hospital Anxiety and Depression Scale (HADS)

The HADS is a 14-item self-report screening scale originally developed to indicate the possible presence of anxiety and depression states in medical non-psychiatric outpatient clinics.41 The HADS consists of two subscales, HADS—Depression (HADS-D) and HADS—Anxiety (HADS-A). Each subscale has seven items that are rated on a 4-point Likert scale that ranges from 0 to 4 with some items being reverse-scored. A maximum total of 21 can be obtained on each subscale. A score of 0-7 on both scales implies a non-clinical range, while a score of 8-10 suggests the possible presence of a depressive or anxiety disorder. A score of 11 or higher suggests the probable presence of a depressive or anxiety disorder. Caseness of depression and anxiety has been suggested for scores above 8.42 The HADS has previously been used with transgender individuals.13, 43 For depression (HADS-D), this gave a specificity of 0.7 and a sensitivity of 0.9. For anxiety (HADS-A), this gave a specificity of 0.78 and a sensitivity of 0.9. In this study; Cronbach's alpha for depression was 0.76 and for anxiety 0.68.

2.3.2 The multidimensional scale of perceived social support (MSPSS)

The MSPSS is a 12-item self-report scale to record levels of social support from family, friends and significant others.44 The measure consists of three subscales to measure the three different types of support. Items are rated on a Likert scale that ranges from 1 (‘very strongly agree’) to 7 (‘very strongly disagree’). To calculate subscale scores, items from each subscale are added together and divided by 4. A total score is calculated by adding together all 12 items and dividing by 12. The mean and total scores range from 1 to 7 with a higher score indicating a higher level of perceived social support. A mean total scale score ranging from 1 to 2.9 can be considered low support; a score of 3 to 5 can be considered moderate support; and a score from 5.1 to 7 can be considered high support. The MSPSS has previously been used with transgender individuals.26, 45 In this study, Cronbach's alpha was 0.89.

2.3.3 Autism spectrum quotient—short version (AQ-Short)

The AQ-Short is a 28-item self-report questionnaire designed to measure autistic traits to give an indication of where the person lies on the continuum of the spectrum, ranging from healthy to autistic.46, 47 It is a shortened version of the validated AQ-50.48 It consists of two higher order factors related to autistic traits, including numbers and patterns (which assess the extent to which people are fascinated by numbers, dates, patterns and categories) and social behaviours. The AQ-Short is a 4-point Likert scale ranging from ‘definitely agree’ to ‘definitely disagree’, with some items being reverse-scored. Total scores range between 28 and 112. The AQ-Short has previously been used with transgender populations.23, 24, 49 Higher scores represent higher levels of autistic traits. Although not intended to be a diagnostic tool, a cut-off of ≥ 70 was found to have a sensitivity of 0.94 and specificity of 0.91 to discriminate between an autism sample and a community sample. Cronbach's alpha was 0.86.

2.3.4 Data analysis

Data analyses were performed using the Statistical Software Package Stata 16.50 Stata 16 was used to conduct power analysis. Only those participants not on GAHT at assessment (T0) were included in the regression analysis. All missingness was imputed using analytical model with 20 imputed data sets generated for each model. Paired sample t tests were used to determine whether there had been a significant change in the HADS-D and HADS-A subscales from T0 to T1. Multiple regression was conducted to explore ethnicity, employment status, relationship status, age, assigned sex, MSPSS and AQ. The hypothesis regarding whether the specific factors were predictive of changes in anxiety and depression was tested via a moderator analysis, entering only the subscales found to be significant in the linear regressions and a product of their combined centred scores. This was tested via a multiple regression. Bonferroni corrections were used to correct multiplicity issue if needed. Although data are not normally distributed, Allison51) states that normality is the least important assumption of regression and as data met the assumptions for linearity, homoscedasticity and absence of multicollinearity or extreme outliers, a multiple regression analysis was conducted. The socio-demographic categories were split into two distinct groups for each category as seen in Table 1. Assigned sex at birth instead of gender identity was used in the socio-demographics in view of the many different gendered identities described, as an analysis based on gender identities would have made the analyses too complex to interpret; this followed previous study approaches.23, 24 The MSPSS and AQ-28 Short at T0 were significant factors predicting change in HADS-D and HADS-A at T1. To check the robustness of regression estimates sensitive to missingness, all regression models were re-run on observed data only and the results were examined against the results from imputed data set.

Table 1. Socio-demographic characteristics (n,%) and mean (SD) scores on HADS, MSPSS and AQ-Short score of all responders assigned male and female at birth Responders Assigned male at birth (2) Assigned female at birth (1) n = 178 n = 95 n = 83 n% n% n% Age, median (range) years 23 (1774) 28 (1779) 21 (1764) Ethnic origin n (%) White 167 94 92 97 75 91 Other 11 6 3 3 8 9 Not known - - - - - - Employment status n (%) Employed (1) 75 41 45 47 28 33 Student (2) 58 32 20 21 37 44 Housewife/househusband (4) - - - - - - Voluntary work (3) 7 4 5 5 2 2 Retired (6) 5 3 6 6 - - Disabled (5) 7 5 5 5 3 3 Unemployed (0) 26 14 14 14 12 14 Not known (9) 1 1 1 1 Civil status n (%) Single (1) 120 69 55 57 65 78 Married (2) 15 7 12 12 1 1 Civil partner ( 3) 5 3 - - 5 6 In a relationship (7) 4 3 1 1 4 4 Divorced/separated (4) 20 11 18 18 2 2 Widowed (5) 2 2 2 2 1 1 Other (6) - - - - - - Not known (9 + Blanks) 12 7 7 7 5 6 HADSD 7.24 (4.03) 7.03 (4.11) 7.48 (3.94) HADSA 8.07 (4.34) 7.54 (4.31) 8.69 (4.32) MSPSS 4.85 (1.29) 4.64 (1.35) 5.1 (1.16) AQ 64.77 (11.86) 62.83 (11.94) 66.97 (11.44)

Ethical approval for the study was received from the NHS Ethics Committee (14/EM/0092) and the Research and Development Department at Nottinghamshire Healthcare NHS Foundation Trust in line with Health Research Authority guidance,52 which included approval for individuals aged 17 and over to sign giving their consent without the need for additional parental consent.

3 RESULTS 3.1 Socio-demographic characteristics of the participants

A total of 1,271 participants were assessed between November 2014 and March 2018, completed T0 questionnaires and agreed participation in the study. Seventy-one per cent (N = 906) could be included in the analysis as they had not received hormones prior to assessment. Of these, 178 (20%) went on to complete a T1 questionnaire after 18 months of GAHT, indicating a response rate of 20%. Responders did not differ from non-responders in terms of either demographic characteristics or baseline AQ-Short scores, but they were significantly less anxious at baseline than non-responders (median 9 vs 8, P = .001; z = 3.225) (see Table 1).

The age range from the 178 participants that completed T0 and T1 questionnaires ranged from 17 to 79 years with a median age of 23 years. More than half of the participants (n = 95; 53.3%) were assigned male sex at birth, and 83 (46.7%) were assigned female sex at birth. The large majority of participants classified themselves as white (n = 167; 94%), were single (n = 120; 69%) and were in employment (n = 75; 41%) or students (n = 58; 32%). Participants who were assigned male sex at birth were more likely to be in employment compared to participants assigned female sex at birth (47% vs 33%), while more participants assigned female sex at birth were single at the time of assessment (78% vs 57%) and a higher percentage of participants assigned male sex at birth were divorced/separated (18% vs 2%) (see Table 1).

3.2 Anxiety and depression scores

The mean score for the total group for anxiety was 8.07 (SD: 4.34). It was higher in those assigned female (8.69 (SD: 4.32)) versus those assigned male at birth (7.54 (SD: 4.31). The mean score for depression was 7.24 (SD: 4.03), also higher in assigned females (7.48 (SD: 3.94)) than assigned males at birth (7.03 (SD: 4.11)).

At T0 (before hormone treatment), 51.13% of participants scored 8 or over on the HADS-A subscale, and in the case of the HADS-D subscale, 47.75% of participants scored above 8 placing these participants within the categories of possible to probable presence of an anxiety or depression disorder. At T1, 47.19% of participants scored 8 or above in the HADS-A subscale showing a reduction of 3.94%, and in the HADS-D subscale, 25.84% of people scored 8 or above showing a reduction of 21.91%.

3.3 Change in anxiety and depression scores between T0 and T1

There was a statistically significant reduction in mean scores of HADS-D from T0 to T1 (mean change difference, −2.05; 95% CI, −2.72 to −1.38; P = .00). This indicated a reduction in depression following 18 months of GAHT. There was also a reduction in the HADS-A score from T0 to T1, but this was not statistically significant (mean change difference, −0.31; 95% CI, −0.97 to 0.36, P = .37). The same findings (a significant reduction in HADS-D and a non-significant reduction in HADS-A) were found when comparing T0 and T1 according to sex assigned at birth (see Table 2).

Table 2. Means (SD) of HADS-D and HADS-A scores of responders at T0 and T1 All responders Assigned male at birth Assigned female at birth N Mean (SD) Mean change from T0 to T1 (95% CI), P-value n Mean (SD) Mean change from T0 to T1 (95% CI), P-value n Mean (SD) Mean change from T0 to T1 (95% CI), P-value HADS-D T0 178 7.24 (4.03)

−2.05

(−2.72 to − 1.38)

P = .00

95 7.03 (4.11)

−1.91

(−2.80 to − 1.01)

P = .00

83 7.48 (3.94)

−2.21

(to 3.23 to − 1.20)

P = .00

T1 178 5.19 (3.73) 95 5.13 (3.92) 83 5.26 (3.52) HADS-A T0 178 8.07 (4.34)

−0.31

(−0.97 to 0.36),

P = .37

95 7.54 (4.31)

−1.16

(−1.50 to 0.39)

P = .25

83 8.69 (4.32)

−0.55

(−0.97 to 0.92) P = .97

T1 178 7.77 (3.90) 95 6.98 (3.96) 83 8.66 (3.65) 3.4 Predictors of anxiety and depression change after hormone treatment

Two multiple regressions with seven predictor variables were conducted to explore the predictors of change, from T0 to T1, in scores on HADS-D and HADS-A. The predictors for each were ethnicity, employment status, relationship status, assigned sex at birth, and age. MSPSS and AQ-Short at T0 were used as independent variables for both regressions.

The results for the first regression showed that overall, the model was significant (F(7,152)=2.09, P = .04) and explained 8.8% (R2 = 0.088) of the total variance in depression scores. The model also showed that mean MSPSS scores at T0 were the only significant predictor of HADS-D change between T0 and T1 (β = 0.81, P = .006). The second regression showed that overall, the model was significant (F(7,152)=2.09, P = .048) and explained 8.8% (R2 = 0.088) of the total variance in anxiety scores. The model also showed mean AQ-Short scores at T0 was a significant predictor of HADS-A change between T0 and T1 (β = −0.069, P = .034). The findings suggest that only levels of social support (MSPSS scores) and autistic spectrum traits (AQ scores) were able to predict changes in anxiety and depression following 18 months of GAHT. Having higher levels of social support (higher scores of MSPSS) predicted a reduction in depression scores following 18 months of GAHT (P = .006) and having lower levels of autistic spectrums traits (lower AQ scores) predicted a reduction of anxiety symptoms following 18 months of GAHT (P = .03), although this reduction was statistically non-significant (see Table 3).

Table 3. Predictive role of ethnicity, employment status, relationship status, assigned gender at birth, age, MSPSS and AQ-Short for change in HADS-D and HADS-A from T0 to T1 HADS − D HADS − A Coef. 95% CI P Coef. 95% CI P Ethnicity (grouped as White and all other at T0) −0.37 −3.19 to 2.45 .794 −0.67 −3.62 to 2.28 .652 Employment (grouped as unemployed and disabled and all other at T0) −0.97 −2.75 to 0.81 .284 −0.51 −2.37 to 1.35 .591 Relationship (grouped as single, widowed, divorced/separated and other at T0 and all other) 0.31 −1.82 to 2.43 .776 −0.61 −2.83 to 1.62 .590 Assigned sex at T0 0.36 −1.21 to 1.93 .651 −0.17 −1.81 to 1.47 .841 Age at T0 0.06 −0.50 to 0.61 .843 −0.03 −0.08 to 0.03 .409 Mean MSPSS at T0 0.81 0.24 to 1.39 .006 0.56 −0.04 - 1.16 .065 Mean AQ − Short at T0 −0.04 −0.096 to 0.025 .250 −0.07 −0.13 to 0.05 .034 Constant −1.93 −4.76 to 0.90 .179 0.41 −2.34 to 3.16 .787 4 DISCUSSION

This prospective longitudinal study aimed to explore whether 18 months of GAHT reduces symptoms of anxiety and depression in transgender people, while addressing the limitations of previous studies, by recruiting a large sample of participants within the same setting. The study found a significant reduction in symptoms of depression in transgender individuals after 18 months of starting GAHT, with a more than one-fifth decrease in the number of participants who scores reflected a possible or probable depressive disorder. A statistical reduction of anxiety was not found. While reductions in depression, and to a lesser extent anxiety, were seen, a significant proportion of participants still present, post-treatment, with a possible or probable depressive disorder (25.84%) or anxiety disorder (47.19%). Data from previous studies in the field were used to compare these findings with the general population.