Effect of sex steroid use on cardiovascular risk in transsexual individuals: a systematic review and meta-analyses

Summary

Objective To summarize the available evidence on the cardiovascular effects of cross-sex steroid use in transsexuals.

Methods We searched relevant electronic databases and sought additional references from experts. Eligible studies reported on cardiovascular events, venous thromboembolism, blood pressure and fasting serum lipids. Data were extracted in duplicate. We used the random-effects model to estimate the pooled weighted mean difference (WMD) and 95% confidence intervals (CIs).

Results We found 16 eligible studies, mostly uncontrolled cohorts of varied follow-up durations (1471 male-to-female (MF) and 651 female-to-male (FM) individuals). In the MF individuals, cross-sex hormone use was associated with a statistically significant increase in fasting serum triglycerides without changes in the other parameters (WMD = 23·39 mg/dl; 95% CI = 4·82–41·95). In the FM individuals, there was a similar increase of triglycerides (WMD = 31·35 mg/dl; 95% CI = 7·53–55·17) and a reduction of high density lipoprotein (HDL)-cholesterol (WMD = −6·09 mg/dl; 95% CI = −11·44 to −0·73). There was a statistically significant but clinically trivial increase in systolic blood pressure (WMD = 1·74 mmHg; 95% CI = 0·21–3·27). Analyses were associated with significant heterogeneity across studies. There were very few reported cardiovascular events (deaths, strokes, myocardial infarctions or venous thromboembolism), more commonly among MF individuals.

Conclusions Very low quality evidence, downgraded due to methodological limitations of included studies, imprecision and heterogeneity, suggests that cross-sex hormone therapies increase serum triglycerides in MF and FM and have a trivial effect on HDL-cholesterol and systolic blood pressure in FM. Data about patient important outcomes are sparse and inconclusive.

Introduction

Gender identity disorder (GID) affects individuals preoccupied with their wish to live as members of the opposite sex. Such individuals intensely desire to adopt the social role of the other sex or to acquire the physical appearance of the other sex through hormonal or surgical manipulation.¹ Sex reassignment therapy seeks to achieve this transition using a multi-modality approach that often includes psychological, hormonal and surgical interventions.² Men seeking transition to the female sex (MF) generally use oestrogen, antiandrogens (cyproterone acetate, spironolactone) or a gonadotropin-releasing hormone agonist (GnRH agonists). Women seeking transition to the male sex (FM) generally use testosterone.³

It is plausible that sex steroid use may be associated with potential adverse effects such as acne, venous thromboembolism, atherosclerosis, hypertension, hyperlipidemia, prostate hyperplasia; and may cause or exacerbate neoplasia of the prostate, breast and ovaries.^{3, 4} Two large randomized trials characterized the effect of oestrogen-containing hormonal use on cardiovascular risk in women,^{5, 6} and the Coronary Drug Project evaluated this therapy in men post-myocardial infarction.⁷ A recent review reported on the weak available evidence linking testosterone use with cardiovascular risk in hypogonadal and eugonadal men,⁸ a finding that was echoed in the recently published Endocrine Society guidelines for androgen use in women, in which the panel described limited evidence regarding the cardiovascular safety of low-dose testosterone use in women with presumed androgen deficiency.⁹ The different characteristics of the patients and of the hormone schedule in these trials mean these studies apply only indirectly to sexual steroid use in transsexual individuals.

In this systematic review, we sought to summarize the available evidence of the effects of cross-sex hormone use on the cardiovascular risk of transsexual individuals. Outcomes of interest were cardiovascular events, venous thromboembolism, fasting serum lipid fractions and blood pressure.

Methods

This report adheres to the standards of reporting of Meta-analysis Of Observational Studies in Epidemiology.¹⁰

Eligibility criteria

We sought to include both randomized trials and observational studies of transsexual individuals who used cross-sex steroids regardless of whether they had sex reassignment surgery or not. Eligible studies exposed MF to oestrogen, antiandrogens (cyproterone acetate, spironolactone) or GnRH agonists and FM to testosterone. Studies must have clearly stated that individuals used sex steroids for at least 3 months and those participants were followed up for at least 3 months. In order to avoid selection bias and ascertain a well-documented exposure, eligible studies excluded individuals who had received sex steroids, even if self-prescribed, before the initiation of the study. In the case of a study having two transsexual groups, one of them previously exposed to sex steroids, we included data only from the group not previously exposed.

Outcomes of interest were cardiovascular events, e.g. death, stroke, myocardial infarction, venous thromboembolism. We were also interested in levels of blood pressure and lipid fractions. Eligible studies provided comparison between intervention and controls groups or a pre–post intervention comparison. We included studies regardless of their publication status, language or size. Review articles, commentaries and letters that did not contain original data were excluded.

Study identification

An expert reference librarian designed and conducted the electronic search strategy with input from study investigators with expertise in conducting systematic reviews. To identify eligible studies, we searched electronic databases (Ovid MEDLINE, Ovid Embase, Ovid PsycInfo, Thomson Scientific Web of Science and Elsevier Scopus) and bibliographies of eligible studies through February 2008 and sought additional references from the experts. The detailed search strategy is available upon request.

Reviewers, working independently and in duplicate, reviewed all abstracts and titles and, upon retrieval of potentially eligible studies, the full text publications for eligibility. Disagreements were resolved by consensus (the two reviewers discussed the discrepancy and reached a decision) or arbitration (if disagreement was not resolved by the two reviewers, a third reviewer adjudicated the difference). We estimated chance-adjusted agreement among reviewers using the kappa statistic.

Quality assessment

Reviewers, working independently and in duplicate, analysed the eligible articles to assess the reported quality of each study, i.e. the confidence in the accuracy of the estimates. For each study, we assessed whether investigators were able to ensure that participants had started with similar prognosis (through randomization with allocation concealment or through careful matching among individuals free of the outcome at baseline, or through adjustment for confounders at baseline), and maintained such similarity during the study (by blinding of participants and investigators, and by careful accounting and adjusting for co-interventions), and through the end (by minimizing loss to follow-up, and similar assessment of outcome measures in an intention-to-treat manner). In addition, we looked at length and adequacy of follow-up and how the outcome was assessed (self-report and interview vs. medical records). We assessed our chance-adjusted agreement on study quality using the kappa statistic with disagreements resolved by consensus or arbitration.

Data extraction

Reviewers, working independently and in pairs, used a standardized form to extract full description of study participants, including age, whether MF or FM, type, dose and duration of hormonal therapy, and data related to blood lipid fractions and blood pressure levels. We also noted the number of venous thromboembolic events, myocardial infarctions, strokes or deaths reported.

Author contact

We sent letters to the corresponding authors (or any other author if we were not able to reach the corresponding author) of each of the eligible studies by electronic mail. We asked these authors to verify the data we extracted and to complete data missing from the published record. In case of no response, we repeated the request three weeks later.

Statistical analysis

Analysis was conducted using StatsDirect version 2.6.7 (StatsDirect Ltd., UK, 2005). Using DerSimonian-Laird random effects meta-analyses,¹¹ we pooled the weighted mean differences (WMD) for continuous outcomes and their associated 95% confidence interval (CI). When measures of variability were not reported, we estimated standard errors from P and t values.¹² Longitudinal and cross-sectional studies were pooled separately. We measured inconsistency across studies that cannot be explained by chance alone, but rather by differences in participants, interventions, outcomes or design, using the I² statistic.¹³ We also planned to estimate the pooled cumulative incidence of cardiovascular events; however, the varied follow-up duration across studies and the limited number of events precluded pooling.

Subgroup and sensitivity analyses

A priori hypotheses to explore potential causes of heterogeneity included possible differences in: population age, e.g. adolescents (who have brief exposure to sex hormones prior to transition) vs. adults; different treatment regimens (e.g. oestrogens alone, vs. oestrogen + GnRH agonists vs. oestrogen + antiandrogens), oral vs. transdermal oestrogen, high vs. physiological dose, agonist alone vs. agonist alone + pituitary suppression (e.g. oestrogen alone vs. oestrogen + goserelin); outcome characteristics, e.g. symptomatic vs. all events; study design, e.g. controlled study vs. single cohort; and study quality, e.g. blinded vs. open outcome assessment, patients lost to follow-up and follow-up duration (arbitrarily chosen durations of 1 year or shorter vs. longer). We also sought to explore how results of the meta-analyses would change when borderline eligible articles are excluded.

Results

Study identification

Initial search of the literature yielded 341 publications, of which 64 were potentially relevant to this review based on titles and abstracts (Fig. 1). After full-text review, we found 15 eligible studies,^14-28 selected with near-perfect agreement across reviewers (kappa statistic = 0·9; 95% CI = 0·8–1·0). Data from an unpublished study was kindly offered by one of the primary authors contacted (Schneiders et al., unpub. obs.) Four studies, fulfilling our inclusion criteria, had overlapping patient populations with other included studies,^29-32 which was confirmed by author contact, and were not included to avoid duplicating individuals. We also excluded studies in which participants used sex steroids for less than 3 months or measured cardiovascular risk factors of unclear significance, e.g. leptin, homocysteine and vascular reactivity.

Figure 1

The process of study selection.

Study characteristics

Table 1 summarizes the characteristics of eligible studies. Twelve studies included groups of MF and 10 studies had FM groups; overall, the literature reviewed included 1471 MF and 651 FM individuals. Patients averaged 31 years of age. MF individuals used various regimens that mainly included oral, intramuscular or transdermal oestrogens with less frequent use of cyproterone acetate, goserelin acetate or spironolactone. FM individuals used various regimens of testosterone that were mainly administered intramuscularly with infrequent use of oral preparations. No transdermal testosterone was used. Progestins were added if menses did not cease.

Table 1. Baseline characteristics of included studies

Author, yr	Mean age, yr	Patients	No. intervention group	Description of intervention	Duration of exposure	Control group	Presence of CV risk factors
Asscheman, 1989^14	32	MF	303	100 mg cyproterone acetate and 100 μg ethinylestradiol qd (n = 258). Until 1980, diethylstillbestrol (5–15 mg/day) was prescribed for a few patients. Some patients insisted on IM oestrogen. They procured the oestrogen outside the clinic and these were self-administered in doses of 200–800 mg oestradiol17-undecanoate/ IM / mo (n = 45). Recommended dose in post-menopausal women was 20–100 mg/mo	53 mo	NA	No
	25·4	FM	122	Long-acting testosterone esters 250 mg IM/2 weeks (n = 69), or testosterone undecanoate 120–160 mg/day/PO (n = 19), or both, but not simultaneously (n = 34). If menstrual activity did not cease within 3 mo after start of the hormone administration, lynestrol 5 mg/PO/day was added (n = 3)	43 mo	NA	Smoking, HTN and family history of heart disease
Berra, 2006^15	30	FM	16	Testosterone depot (100 mg testosterone enanthate + 25 mg testosterone propionate)/IM/q 10 day	6 mo	NA	No
Damewood, 1989^16	NR	MF	18	Premarin (oestrone sulphate 48%, equilin sulphate 26%, 17alpha dehydroequilin sulphate 15%), PO, 1·25–2·5 mg/day	> 3 yrs	28healthy volunteers, No Rx	Smoking
De Cuypere, 2005^17	42	MF	32	Multiple regimens	Unclear	NA	More smokers, older
	33	FM	23	Multiple regimens	Unclear	NA	No
Dittrich, 2005^18	38	MF	60	3·8 goserelin acetate, SQ/ q 4 weeks + 6 mg oestradiol-17b valerate, PO/qd	24 mo	NA	3 patients had thrombophilia
Elbers, 2003^19	26	MF	20	Ethinyl oestradiol 100 μg/day + cyproterone acetate 100 mg/day	12 mo	NA	No
	23	FM	17	Testosterone esters 250 mg/IM/2 weeks	12 mo	NA	No
Giltay, 2003^20	32	MF	15	Oral ethinyl oestradiol 100 μg/day + CA 100 mg/day	4 mo	NA	Smoking
	32	MF	15	Transdermal 17b-oestradiol 100 μg 2 × /week + CA 100 mg/day	4 mo	NA	Smoking
Jacobeit, 2007^21	33	FM	12	Testosterone undecanoate 1,000 mg/IM at weeks 0, 6 and q 12–14	12 mo	NA	No
Mueller, 2006^22	37	MF	40	3·8 mg goserelin acetate/4 weeks + 6 mg PO 17Boestradiol valerate qd	12 mo	NA	No
Mueller, 2007^23	30	FM	37	Testosterone undecanoate 1000 mg/IM/q 12 weeks	12 mo	NA	Smoking
Prior, 1989^24	31	MF	23	Spironolactone (less than 100 up to 200 mg qd) and conjugated oestrogens (PO, 0·625–2·5 mg qd) and progesterone (PO, 10 mg qd)	12 mo	NA	5 of 23 had HTN
Schlatterer, 1998^25	NR	MF	46	Cyproterone acetate 50–100 mg/ PO/qd + depot oestrogens IM at varying intervals	Variable	NA	NR
	NR	FM	42	Depot testosterone 250 mg IM every 2–4 weeks	Variable	NA	NR
Toorians, 2003^26	35	MF	8	Cyproterone acetate 50 mg/PO/bid	4 mo	NA	No
	32	MF	14	Ethinyl oestradiol 50 mg/PO/bid + cyproterone acetate 50 mg/PO/bid	4 mo	NA	No
	30	MF	20	E2-valerate 2 mg/PO/bid + cyproterone acetate 50 mg/PO/bid	4 mo	NA	No
	26	FM	14	Testosterone esters 250 mg/IM/q 2 weeks	4 mo	NA	No
	30	MF	14	17-b-oestradiol 100 μg E2/TD/qd + cyproterone acetate 50 mg/PO/bid	4 mo	NA	No
Van Kesteren, 1997^27	41	MF	816	Cyproterone acetate 100 mg + ethinyl oestradiol 100 μg/PO/qd	2 mo–41 yrs	NA	1 had a previous VTE
	34	FM	293	Testosterone esters 250 mg IM/2 weeks or testosterone undecanoate/PO/qd	2 mo–41 yrs	NA	No
Wilson, 2006^28	38	MF	25	Oestrogen 1·5 to 5·0 mg/PO/qd	18 mo	13 MF, No Rx	No
Schneiders, unpub. obs.	51	FM	75	Testosterone esters 250 mg/IM/q 2 weeks (n = 46) or testosterone undecanoate 160–240 mg/PO/qd (n = 29), depending on the patient’s preference.	13 yrs	NA	NR

• No, number; NA, not applicable; NR, not reported; FM, female-to-male; MF, male-to-female; Rx, treatment; VTE, venous thromboembolism; mo, month; yr, year; PO, orally; IM, intramuscularly; SQ, subcutaneous; TD, transdermal; qd, daily; bid, twice daily; HTN, hypertension.

Author contact

We successfully contacted all of the corresponding authors (another author in two studies) by electronic mail. All authors either contributed missing data (where these data had been collected but not reported in the format we needed for analyses) or confirmed study characteristics, quality assessments and data as collected.

Study quality

Table 2 summarizes the methodological quality of the 16 included studies. Studies were uncontrolled or self-controlled observational studies, one of which included a nested trial (randomizing MF individuals to oral or transdermal oestrogen preparations). Medical records were the most frequent method of ascertaining exposure to hormonal therapy and assessing outcomes.

Table 2. Quality assessment of included studies

Author	Study design	Only for randomized trials		Selection of control group	Ascertainment of exposure	Ascertainment of outcome status at baseline	Factors controlled for	Assessment of outcome	Length of follow-up	% lost to follow-up
		Allocation concealment	Blinding
Asscheman, 1989^14	Single cohort, retrospective, compared with age-adjusted data	NA	NA	Different source	Medical records	No	Age, sex, Sex-steroid, exposure time	Medical records	1–17 yrs	15%
Berra, 2006^15	Single cohort, prospective, before and after comparison	NA	NA	NA	Medical records	Yes	None	Medical records	6 mo	Unclear
Damewood, 1989^16	Cross-sectional	NA	NA	Different source	Unclear	Unclear	Age, Smoking status, Alcohol intake, exercise, dietary habits	Interview	NA	0%
De Cuypere, 2005^17	Cross-sectional	NA	NA	NA	Interview/Questionnaire	No	None	Blinded assessment	Unclear	51·3%
Dittrich, 2005^18	Single cohort, prospective, before and after comparison	NA	NA	NA	Medical records	Yes	None	Medical records	24 mo	0%
Elbers, 2003^19	Prospective cohorts, before and after comparison	NA	NA	NA	Interview/Questionnaire	Yes	NA	Interview	12 mo	Unclear
Giltay, 2003^2^0	RCT	No	No	No	Interview/Questionnaire	Yes	NA	Interview	4 mo	0%
Jacobeit, 2007^21	Single cohort, retrospective, before and after comparison	NA	NA	NA	Medical records	Yes	None	Medical records	12 mo	Unclear
Mueller, 2006^22	Single cohort, prospective, before and after comparison	NA	NA	NA	Interview/Questionnaire	No	NA	Interview	24 mo	NR
Mueller, 2007^23	Single cohort, prospective, before and after comparison	NA	NA	NA	Unclear	No	None	Medical records	12 mo	0%
Prior, 1989^24	Prospective cohorts, before and after comparison	NA	NA	NA	Investigator	No	None	Investigator	12 mo	0%
Schlatterer, 1998^25	Cross-sectional	NA	NA	NA	Medical records	No	None	Medical records	Variable	0%
Toorians, 2003^26	5 arms; 2 were randomized, all exposed	No	No	NA	Investigator	Yes	NA	Investigator	4 mo	10%
Van Kesteren, 1997^27	Single cohort, retrospective	NA	NA	NA	Medical records	No	Age, gender, period of time on sex-steroid	Medical records	2–19 yrs	11%
Wilson, 2006^28	Cross-sectional	NA	NA	Same community	Unclear	No	None	Medical records	NA	Unclear
Schneiders, unpub. obs.	Single cohort, retrospective, before and after comparison	NA	NA	NA	Medical records	No	None	Medical records	13 yrs	31%

• NA, not applicable; NR, not reported; mo, month; yr, year.

Meta-analyses

Patient important outcomes. There were very few reported cardiovascular events across the trials, with varied length of follow-up. Figure 2 shows the proportion of participants in each eligible study who had reported cardiovascular events.

Figure 2

Proportions of cardiovascular outcomes in included studies. Each study is represented by a circle. F/U, follow-up; FM, female-to-male individuals; MF, male-to-female individuals; MI, myocardial infarction; VTE, venous thromboembolism.

Serum lipids and blood pressure. Tables 3 and 4 represent pooled data for the lipid parameters and blood pressure measurements in MF and FM individuals after sex steroid use, respectively.

Table 3. Meta-analysis of serum lipids and blood pressure measurements after hormonal therapy in male-to-female individuals

	Pooled weighted mean difference* (95% CI)	P-value	Heterogeneity across studies (I^2)
Lipids (mg/dl)
Cholesterol^18, 19, 22, 24	−1·31 (−13·42 to 10·79)	0·83	84%
Triglycerides^18-20, 22, 24	23·39 (4·82 to 41·95)	0·01	84%
Low-density lipoproteins^19	–	–	NA
High-density lipoproteins^19, 20, 24	3·70 (−2·3 to 9·69)	0·23	93%
Blood pressure (mmHg)
Systolic^19, 24	0·16 (−14·04 to 14·37)	0·98	NA
Diastolic^19, 24	1·42 (−6·71 to 9·55)	0·73	NA

• Results expressed as post–pre intervention difference. Positive values indicate an increase of the tested parameter after sexual steroid use.
• CI, confidence interval; NA, I² incalculable when pooling fewer than three studies.
• *Random-effect meta-analyses.

Table 4. Meta-analysis of serum lipids and blood pressure measurements after hormonal therapy in female-to-male individuals

	Pooled weighted mean difference* (95% CI)	P-value	Heterogeneity across studies (I^2) (%)
Lipids (mg/dl)
Total cholesterol^15, 19, 21, 23 unpub. obs.	1·19 (−10·92 to 13·31)	0·85	69
Triglycerides^15, 19, 23 unpub. obs.	31·35 (7·53 to 55·17)	0·01	85
Low-density lipoproteins^15, 19, 21, 23 unpub. obs.	2·07 (−8·49 to 12·63)	0·70	81
High-density lipoproteins^15, 19, 21, 23 unpub. obs.	−6·09 (−11·44 to −0·73)	0·03	95
Blood pressure (mmHg)
Systolic^15, 19, 23 unpub. obs.	1·74 (0·21 to 3·27)	0·03	27
Diastolic^15, 19, 23 unpub. obs.	1·45 (−0·57 to 3·48)	0·16	73

• Results expressed as post–pre intervention difference. Positive values indicate an increase of the tested parameter after hormonal therapy.
• CI, confidence interval.
• *Random-effect meta-analyses.

Subgroup and sensitivity analyses

All feasible subgroup interaction analyses are shown in Tables 5 and 6. In MF individuals, significant interaction existed in HDL level-route of hormonal administration, showing a higher serum level after oral administration than the transdermal route. The same was observed with triglycerides levels. In FM individuals, we found insufficient data to compare the route of administration of hormonal therapy (intramuscularly vs. orally). Subgroups defined by individuals followed up for more than 1 year vs. less than 1 year, showed significant increase in cholesterol and triglycerides in studies with follow-up period of more than 1 year. All other subgroup analyses were nonsignificant.

Table 5. Male-to-female subgroup analyses for lipid and blood pressure parameters

Subgroup	Subgroup description	Difference from baseline (95% CI)	P-value
Cholesterol: Hormonal preparation used	Oestrogen+antiandrogen^19, 24	−3·62 (−17·51 to 10·27)	0·41
	Oestrogen + GnRH agonists^18, 22	7·90 (−15·75 to 31·54)
Cholesterol: Patients follow-up time	1 yr or less^19, 24	−3·62 (−17·51 to 10·27)	0·41
	More than 1 yr^18, 22	7·90 (−15·75 to 31·54)
HDL: Route of hormonal administration	TD^20	−4·68 (−2·01 to −7·35)	0·00
	PO^19, 20, 24	6·45 (2·68 to 10·23)
Triglycerides: Hormonal preparation used	Oestrogen+antiandrogen^19, 20, 24	23·76 (1·97 to 45·55)	0·83
	Oestrogen + GnRH agonists^18, 22	19·36 (−13·72 to 52·45)
Triglycerides: Route of hormonal administration	TD^20	−9·79, (6·56 to −26·14)	0·00
	PO^18-20, 22, 24	31·98 (12·79 to 51·16)
Triglycerides: Patients follow-up time	1 yr or less^19, 20, 24	23·76 (1·97 to 45·55)	0·83
	More than 1 yr^18, 22	19·36 (−13·72 to 52·45)

• Subgroup analyses of outcomes not reported in this table were unfeasible due to sparse data.
• CI, confidence interval; TD, transdermal; PO, oral.

Table 6. Female-to-male subgroup analyses for lipid and blood pressure parameters

Outcome/subgroup	Subgroup description	Mean difference from baseline (95% CI), SE	P-value
Cholesterol: Route of hormonal administration	IM^15, 19, 21, 23	−3·25 (−15·19 to 8·70)	0·02
	Mixed routes/Oral unpub. obs.	15·60 (4·92 to 26·28)
Cholesterol: Individuals follow-up time	1 yr or less^15, 19, 21, 23	−3·25 (−15·19 to 8·70)	0·02
	More than 1 yr unpub. obs.	15·60 (4·92 to 26·28)
Cholesterol: Individuals lost to follow-up	None reported^23	3·74 (−7·80 to 15·28)	0·14
	Any loss of patients unpub. obs.	15·60 (4·92 to 26·28)
LDL: Route of hormonal administration	IM^15, 19, 21, 23	0·51 (−13·01 to 14·04)	0·38
	Mixed routes unpub. obs.	7·80 (−1·33 to 16·93)
LDL: Individuals follow-up time	1 yr or less^15, 19, 21, 23	0·51 (−13·01 to 14·04)	0·38
	More than 1 yr unpub. obs.	7·80 (−1·33 to 16·93)
LDL: Individuals lost to follow-up	None reported^23	6·89 (−3·91 to 17·69)	0·90
	Any loss of patients unpub. obs.	7·80 (−1·33 to 16·93)
HDL: Route of hormonal administration	IM^15, 19, 21, 23	−6·67 (13·38 to 0·04)	0·47
	Mixed routes unpub. obs.	−3·90 (−0·35 to −7·45)
HDL: Individuals follow-up time	1 yr or less^15, 19, 21, 23	−6·67 (13·38 to 0·04)	0·47
	More than 1 yr unpub. obs.	−3·90 (−0·35 to −7·45)
HDL: Individuals lost to follow-up	None reported^23	−10·70 (−5·98 to −15·42)	0·02
	Any loss of patients unpub. obs.	−3·90 (−0·35 to −7·45)
Triglycerides: Route of hormonal administration	IM^15, 19, 23	14·40 (4·51 to 24·30)	0·00
	Mixed routes unpub. obs.	97·90 (57·01 to 138·79)
Triglycerides: Individuals follow-up time	1 yr or less^15, 19, 23	14·40 (4·51 to 24·30)	0·00
	More than 1 yr unpub. obs.	97·90 (57·01 to 138·79)
Triglycerides: Individuals lost to follow-up	None reported^23	30·29 (5·97 to 54·61)	0·01
	Any loss of patients unpub. obs.	97·90 (57·01 to 138·79)
SBP: Route of hormonal administration	IM^15, 19, 23	1·39 (−0·04 to 2·82)	0·26
	Mixed routes unpub. obs.	4·00 (−0·33 to 8·33)
SBP: Individuals follow-up time	1 yr or less^15, 19, 23	1·39 (−0·04 to 2·82)	0·26
	More than 1 yr unpub. obs.	4·00 (−0·33 to 8·33)
SBP: Individuals lost to follow-up	None reported^23	4·28 (0·36 to 8·20)	0·93
	Any loss of patients unpub. obs.	4·00 (−0·33 to 8·33)
DBP: Route of hormonal administration	IM^15, 19, 23	1·73 (−1·04 to 4·51)	0·75
	Mixed routes unpub. obs.	1·10 (−1·59 to 3·79)
DBP: Individuals follow-up time	1 yr or less^15, 19, 23	1·73 (−1·04 to 4·51)	0·75
	More than 1 yr unpub. obs.	1·10 (−1·59 to 3·79)
DBP: Individuals lost to follow-up	None reported^23	2·86 (0·57 to 5·15)	0·33
	Any loss of patients unpub. obs.	1·10 (−1·59 to 3·79)

• Subgroup analyses of outcomes not reported in this table were unfeasible due to sparse data
• CI, confidence interval; IM, intramuscular.

In terms of sensitivity analyses, pooling studies with a cross-over study design separately did not show any significant change in measured lipid fractions. The exclusion of unpublished data decreased between study heterogeneity and drove the statistically significant decrease in HDL and increase in systolic blood pressure towards the null, although a strong trend continued to exist (P = 0·05 and 0·06; respectively).

Discussion

Summary of findings

We conducted a systematic review and meta-analyses of studies that enrolled transsexual individuals who used sex steroids as part of a sex reassignment programme. Overall, we found no significant effect of hormones on cardiovascular outcomes. As expected, HDL-cholesterol decreased in FM receiving androgens and increased in MF receiving oestrogens, although these changes were statistically significant in FM only. Reciprocal increase of serum triglycerides was noted in both FM and MF. The effects of cross-sex steroids on other lipid fractions and on blood pressure were imprecisely measured and thus remain uncertain. Data were insufficient to allow meaningful assessment of patient important outcomes like death, stroke, MI or venous thromboembolism, although Fig. 2 suggests a higher incidence of these events among MF individuals. Of note, most of the individuals in these studies are from one centre and received a fairly high dose of oestrogens. Therefore, the only identifiable effects of cross-sex hormones appear to be on surrogate outcomes of less importance to patients.³³ The quality of evidence is very low. This is due to the uncontrolled and observational nature of included studies, small number of events leading to wide CIs and imprecision of estimates, brief and varied duration of follow-up, heterogeneity of treatment regimens, and inconsistency of results across studies that was unexplained by subgroup analyses.^{34, 35}

Comparison with published literature

To our knowledge, this is the first meta-analysis evaluating cardiovascular outcomes in transsexual individuals using sex steroids. Several randomized trials of sex steroid use in different populations can be considered in the light of the paucity of direct evidence in transsexual individuals. In the Coronary Drug Programme, men 30–64 years of age who had experienced a myocardial infarction were randomly allocated to either 2·5 mg or 5 mg of conjugated oestrogens. The 5-mg arm was stopped earlier than planned because it was associated with increased cardiovascular mortality. The 2·5-mg arm did not affect cardiovascular risk but was associated with a trend towards increased risk of thromboembolism (4·7% in the oestrogen group vs. 2·9% in the placebo group).⁷ How these results apply to MF individuals, who are usually at lower cardiovascular risk than the coronary patients participating in this trial and who may use higher doses of oestrogen with different administration route, remains unclear. Nonetheless this evidence, along with randomized trial evidence of the cardiovascular effects of oestrogen use in postmenopausal women,^{5, 6} raises concern about the extent to which oestrogen preparations could cause harmful cardiovascular events. Uncertainty also affects the applicability of the findings of studies of androgens in women with low libido or presumed androgen deficiency, who use smaller doses of androgens, to the care of FM individuals.⁹

Limitations and strengths

The main limitation of this review stems from the limited methodological quality of the primary studies identified and summarized here and the potential for biased reporting of these cohorts. Data were sparse and total sample size is limited. In addition, study-level analysis did not allow proper evaluation of the effect of some important patient-level characteristics that impact cardiovascular risk, e.g. smoking.

The strengths stem from the focussed review question, the comprehensive literature search that included multiple databases without language restrictions, the explicit eligibility criteria, the rigorous protocol-driven methodology of executing the review with protection against bias (e.g. use of independent reviewer pairs) and the parsimonious analysis plan.

Implications for research and practice

The uncertainty in relation to cardiovascular events bears on the recommendations for using sex steroids to achieve the desired sex. Future research is needed to ascertain harms of hormonal therapies in this context. It is possible to conduct randomized trials nested within these cohorts to test the relative efficacy and safety of different sex steroid administration approaches. It is also possible to use registries of transsexual individuals to characterize individuals with and without cardiovascular events at a given point in time (sufficiently long after sex steroid use starts) and to identify what risk factors contributed to this situation. High-quality observational studies in which baseline risk of cardiovascular disease is assessed and balanced between study arms and proper ascertainment of exposure and outcome are also feasible and desirable. For clinicians involved in prescribing cross-sex hormones, this review highlights the very low quality of evidence and encourages them to convey this uncertainty to their patients. Treatment decisions in the light of low-quality evidence should be made based on patients’ values, preferences, resources, cultural and social factors.^{34, 35}

Conclusions

Very low-quality evidence, downgraded due to methodological limitations of included studies, imprecision and heterogeneity, suggests that cross-sex hormone therapies increase serum triglycerides in MF and FM and have a trivial effect on HDL-cholesterol and systolic blood pressure in FM. Data about patient important outcomes are sparse and inconclusive.

Acknowledgements

We are grateful to the authors of primary studies who responded to our requests for data confirmation and missing data (Gooren LJ, Giltay EJ, Müller A, Meriggiola MC, Prior JC, T’Sjoen G, Stalla GK and Damewood M). This systematic review was funded by a contract from the Endocrine Society.

Competing interests/financial disclosure

MBE, MZG, MHM, PJE and VMM have nothing to declare.