Rate and Extent of Recovery from Reproductive and Cardiac Dysfunction Due to Androgen Abuse in Men

Abstract

Context

Androgen abuse impairs male reproductive and cardiac function, but the rate, extent, and determinants of recovery are not understood.

Objective

To investigate recovery of male reproductive and cardiac function after ceasing androgen intake in current and past androgen abusers compared with healthy non-users.

Methods

Cross-sectional, observational study recruited via social media 41 current and 31 past users (≥3 months since last use, median 300 days since last use) with 21 healthy, eugonadal non-users. Each provided a history, examination, and serum and semen sample and underwent testicular ultrasound, body composition analysis, and cardiac function evaluation.

Results

Current abusers had suppressed reproductive function and impaired cardiac systolic function and lipoprotein parameters compared with non- or past users. Past users did not differ from non-users, suggesting full recovery of suppressed reproductive and cardiac functions after ceasing androgen abuse, other than residual reduced testicular volume. Mean time to recovery was faster for reproductive hormones (anti-Mullerian hormone [AMH], 7.3 months; luteinizing hormone [LH], 10.7 months) than for sperm variables (output, 14.1 months) whereas spermatogenesis (serum follicle-stimulating hormone [FSH], inhibin B, inhibin) took longer. The duration of androgen abuse was the only other variable associated with slower recovery of sperm output (but not hormones).

Conclusion

Suppressed testicular and cardiac function due to androgen abuse is effectively fully reversible (apart from testis volume and serum sex hormone binding globulin) with recovery taking between 6 to 18 months after ceasing androgen intake with possible cumulative effects on spermatogenesis. Suppressed serum AMH, LH, and FSH represent convenient, useful, and underutilized markers of recovery from androgen abuse.

androgen abuse, testosterone, drug abuse, anabolic steroid, reproductive function, cardiac function

Androgen abuse, defined as androgen use without prescription or medical indication, originated during the Cold War among Eastern European elite athletes seeking illicit performance enhancement in power sports. During the 1980s, androgen abuse crossed over into the general community for image rather than performance enhancement for bodybuilding. Androgen abuse in elite sport is subject to stringent antidoping surveillance coupled with deterrence by bans from competition, the critical step to an elite athlete’s pursuit of fame and fortune. By contrast, the community-based demand for illicit androgens, supplied chemical manufacturing plants with unregulated Internet advertising and sales, makes androgens easily available free from medical scrutiny. This virtually unregulated marketing has made androgen abuse relatively frequent, with a prevalence of 6.4% among young men and even higher among recreational sports (18.4%), athletes (13.4%), prisoners (12.4%), drug users (8.0%), and high school students (2.3%) (1). Androgen abuse is associated with violent criminality (2–4) and psychiatric (5–8) and medical disorders (9) including excess morbidity and mortality (10).

Exogenous androgens suppress the hypothalamic–pituitary–testicular axis by negative feedback effects that inhibit testicular endocrine (steroidogenesis) and exocrine (spermatogenesis) functions (11). This leads to reduced spermatogenesis and sperm output, testicular atrophy, and subfertility during abuse as well as reduced endogenous testosterone and androgen effects following cessation. Although steroidal negative feedback is theoretically reversible (11, 12), few studies have investigated the rate, extent, and determinants of recovery from androgen abuse. One study suggested only partial reversibility of hormonal and cardiac effects but did not investigate sperm output or testicular size (13). Androgen abuse is also associated with increased cardiovascular risk with impaired cardiac function and adverse clinical effects (arrhythmia, thromboembolism, left ventricular hypertrophy, cardiomyopathy, heart failure, accelerated atherosclerosis) (14–18) including sudden cardiac death (19–22). However, only limited and conflicting evidence is available on reversibility of androgen-induced cardiac dysfunction and failure (14, 23–27). Two recent studies have reported cardiac systolic dysfunction among current users (17, 28) that persisted among past users after a median of 30 months since cessation (28). In epidemiological studies, low high-density lipoprotein (HDL) cholesterol increases cardiovascular risk (29), and androgen abuse lowers levels of HDL cholesterol in a reversible manner (30). HDL function, as assessed by cholesterol efflux capacity (CEC), is prognostically important (31), but whether this alters reversibly with androgen abuse and how this relates to change in HDL concentration is unclear.

As the growing prevalence of androgen abuse is a serious but neglected public health problem (6) with limited systematic study of its determinants and consequences (9), the present study aimed to investigate the rate, extent, and determinants of recovery from androgen abuse-induced reproductive and cardiac dysfunction after cessation of androgen intake in a cross-sectional study of men who are current, past, and non-users. Such information could form a basis for supportive management of men seeking to, or who have stopped, androgen abuse.

Materials and Methods

Study design

In a cross-sectional observational design, the study compared current and past androgen abusers with healthy, regularly exercising non-user controls to estimate the rate and extent of recovery from androgen abuse effects on reproductive and cardiac function. Recruitment advertising stated that participants would neither be paid nor prescribed drugs and volunteers were advised their health data would be protected. The study refrained from collecting potentially incriminating evidence on illicit sourcing of illicit drugs. The study was approved by the Sydney Local Health District Human Ethics Committee (Concord Hospital) within NHMRC/Australian Health Ethics Committee guidelines for Human Experimentation (Australian Code for the Responsible Conduct of Research), which are consistent with the Declaration of Helsinki.

Participants

All men were aged 18 to 55 years, provided written informed consent, and were regularly involved in recreational exercise (including gym and/or weight training, cycling, running, rowing) at least 3 times per week. Recruitment advertising was via social media, needle and syringe exchange program centers, gyms, and word-of-mouth, and users were classified into current and past users according to their time since last use. Participants were stratified into (i) current androgen users; (ii) past androgen users, defined as those who ceased using androgens for at least 3 months; and (iii) non-user controls, defined as healthy regularly exercising men who never used androgens.

Study procedures

During a single morning visit (before 10 am), participants provided a standardized medical (including reproductive and drug) history, physical examination, and a blood and semen sample and completed quality-of-life questionnaires. They underwent testicular ultrasound using standard methods (32) by a single experienced operator, body composition analysis by dual energy absorptiometry analysis (33), and cardiac function evaluation by echocardiography and computed tomography calcium score in a core cardiac function laboratory.

Laboratory analyses

Laboratory analyses were blinded to the group. Serum testosterone, dihydrotestosterone, estradiol, estrone, dehydroepiandrosterone (DHEA), 3α and 3β androstanediols (3α-diol, 3β-diol, respectively) were measured in a single batch by the Andrology laboratory, ANZAC Research Institute using a validated liquid chromatography-mass spectrometry (LC-MS) method (34, 35) and with within-assay coefficient of variability of <10% and extraction recovery >90% for all steroids. Serum anti-Mullerian hormone (AMH), inhibin B, total inhibin and activin were measured by specific immunoassays in serum diluted 1:15 for serum AMH and otherwise undiluted serum by Dr Ajay Kumar (Ansh Lab, Webster, Texas, US) (36) within a single batch with coefficient of variation (CV) <7.4% for all analytes. Inhibin B was measured in a two-site immunoassay that sandwiches the α subunit and the βb subunits while the total inhibin immunoassay is a two-site immunoassay directed to the α subunit. Serum luteinizing hormone (LH), follicle-stimulating hormone (FSH), and sex hormone binding globulin (SHBG) were measured by commercial immunoassays (Roche) with biochemistry and hematology profiles by routine autoanalyzer methods at the Diagnostic Pathology Unit, Concord Repatriation General Hospital, all subject to regular external quality control and with a within-assay CV <10%. Semen analysis was performed by World Health Organization Semen Manual methods (37) at the Clinical Andrology Laboratory, Concord Repatriation General Hospital. Body composition was evaluated on a Lunar bone densitometer.

Serum CEC was determined using Chinese hamster ovary cells, stably expressing ABCA1 under a tetracycline inducible promoter exposed to 1% apoB-depleted serum (v/v) for 4 h (38, 39). Total and basal efflux are those measured from cells with and without tetracycline, respectively. ABCA1-specific efflux is the difference in efflux between total and basal efflux. Efflux was normalized to a standard apoB-depleted serum that was taken along in each experiment. Intra- and inter-assay CVs were 3.9 and 2.9% for basal and 2.9 and 2.3% for ABCA1-specific efflux, respectively.

Cardiac parameters

Cardiac size and function were reported by a single investigator (CY) blind to participants group and with a CV <5.8% for all variables. Participants underwent 2-dimensional (2D) transthoracic echocardiography (EPIQ & Affinity, Philips Medical Systems) as per current guidelines (40) with images stored digitally and analyzed offline (Tomtech, Munich, Germany). Left ventricular ejection fraction (LVEF) was measured by the modified Simpson’s biplane method, and normal LVEF was defined as ≥52% (40). Diastolic function was assessed by pulse wave doppler of the mitral inflow early velocity (E), and early left ventricular (LV) relaxation velocity (e ́) of the medial and lateral mitral valve annulus. Echocardiographic variables were indexed according to body surface area (BSA), LV mass was calculated using the area-length method and global longitude strain was calculated using the best-quality 2-dimensional apical 2-, 3-, and 4-chamber views (41). Coronary artery calcium (Agatson) score was derived using a 256-slice computer tomography scanner (Revolution CT) and analyzed using the AW workstation (both GE Healthcare, Milwaukee, WI, US) with men <40 years old having a median score of 1 with interquartile range 0 to 3 (42) and 60% of men aged 35 to 40 years old at this center having a score of 0 (Ridley, personal communication).

Data analysis

Continuous variables were analyzed by 1-way analysis of variance with the main effect being group (current and past users and non-user controls) and outcomes being reproductive (steroids and peptide reproductive hormones, semen analysis, testis size, body composition), lipoprotein, and cardiac (echocardiography, coronary calcium score) variables. Covariance analysis was undertaken for potentially confounding variables that independently influenced the reproductive or cardiac outcomes. Planned post-hoc tests were used when 1-way analysis of variance showed significant differences between groups. These comprised linear contrasts to estimate whether androgen effects were significant (current users vs non-users) and, if so, whether the androgen effects were reversible (past users vs non-users). For categorical variables, analysis was by contingency table analysis (extended Fishers exact test) with analogous post-hoc tests between the user groups. Sperm output and concentration were cube-root, sperm motility square-root, and times (time since last use, total duration of use) log transformed prior to analysis to normalize skewed distributions according to power transforms optimized by Box–Cox analysis (43). Vasectomized men were excluded from analyses of sperm output. Mean time for recovery for sperm output, testis volume, and serum gonadotrophins were calculated as the time taken to reach the mean value of the control group according to a linear regression of the variable over time since cessation of exogenous androgen intake. Undetectable hormone values were imputed using a validated substitution methodology for left censoring of hormone variables (44). Stepwise linear discriminant analysis was performed to identify the minimal set of blood test variables that optimally distinguished current from past users with the efficiency defined as the proportion of accurate classification. Where multiple testing was required, a Bonferroni adjustment to P-values for significance was employed. Body mass index (BMI) was calculated as kg/m² and BSA was calculated by the Gehan–George formula (45). Statistical analysis was performed by NCSS 2019 (NCSS, Kaysville, UT, US), StatXact (Cytel, Boston, MA, US) and SPSS (IBM, Chicago, IL, US) software.

Results

The current users (n = 41), past users (n = 31), and healthy non-user controls (n = 21) were well matched for age, marital status (33% single), sexual orientation (92% heterosexual, 6.5% gay or bisexual, 1.5% unstated), smoking (18% current smokers), alcohol intake (72% regular consumption), height, bone mineral density or content (absolute or standardized), fat, and bone mass (Table 1). The proportion of non-heterosexuals was higher than expected for the Australian male population (8% vs 3%, P = .016) (46). Current users displayed higher body weight, BMI, BSA, lean mass, pulse rate, and blood pressures (Table 1) as well as higher blood hemoglobin, leukocytes, and serum creatinine but lower serum albumin, but there were no differences between groups in blood platelets, serum urea, bilirubin, or serum transaminases (Table 2).

Table 1.

Descriptive features of the 3 groups

Variable	Current users	Past users	Non-users	P  a	P  b	P  c
n	41	31	21
Age at study (years)	34 ± 1	33 ± 2	32 ± 2
Age at first use (years)	27 ± 1	23 ± 1	—	.014	—	—
Total use (weeks)	137 (72, 338)	115 (48, 168)	—		—	—
Height (cm)	179 ± 1	176 ± 1	180 ± 1
Weight (kg)	95.3 ± 2.3	87.0 ± 2.6	83.8 ± 3.2	.013	.004
Body mass index (cm/kg2)	29.5 ± 0.6	28.0 ± 0.7	25.9 ± 0.9		.001
Body surface area (m2)	2.19 ± 0.03	2.07 ± 0.03	2.05 ± 0.04	.008	.010
Heart rate (beats/min)	75 ± 2	67 ± 2	66 ± 3	.009	.016
Systolic blood pressure (mm)	130 ± 3	121 ± 3	120 ± 4	.031	.041
Diastolic blood pressure (mm)	77 ± 1	73 ± 1	70 ± 2	.037	.006
Fat mass (kg)	15.7 ± 1.2	19.8 ± 1.5	17.6 ± 1.7
Lean mass (kg)	75.6 ± 1.7	64.8 ± 2.0	63.1 ± 2.3	<.001	<.001
Bone mass (kg)	3.4 ± 0.4	4.0 ± 0.5	3.2 ± 0.6
Hip BMD (gm/cm2)	1.18 ± 0.02	1.13 ± 0.03	1.14 ± 0.03
Hip z score	0.32 ± 0.31	0.24 ± 0.37	0.36 ± 0.44
Spine BMD (gm/cm2)	1.34 ± 0.03	1.27 ± 0.03	1.31 ± 0.04
Spine z score	0.51 ± 0.26	0.22 ± 0.31	0.44 ± 0.36

All data are shown as mean ± standard error of the mean except for total use as median (interquartile range). P-values for linear contrasts shown only when overall analysis of variance showed significant differences.

Abbreviation: BMD, bone mineral density.

^aCurrent vs past users.

^bCurrent vs non-users.

^cPast vs non-users.

Table 2.

  Hematology and biochemistry

Variable	Current users	Past users	Non-user controls	P  a	P  b	P  c
n	41	31	21
Hemoglobin	164 ± 2	154 ± 2	151 ± 2	<.001	<.001	.26
Leukocytes	7.4 ± 0.3	5.7 ± 0.4	5.9 ± 0.5	.002	<.001	.81
Platelets	250 ± 12	232 ± 14	228 ± 17
Serum urea	6.6 ± 0.3	6.6 ± 0.3	6.0 ± 0.3
Serum creatinine	97 ± 2	90 ± 2	90 ± 3	.038
Serum bilirubin	8.3 ± 0.5	9.6 ± 0.6	9.6 ± 0.7
Serum albumin	40.6 ± 0.6	42.6 ± 0.6	46.4 ± 0.8	.021	<.001	<.001
Serum ALP	62 ± 2	71 ± 2	69 ± 3
Serum GGT	18.7 ± 7.2	35.6 ± 8.2	19.6 ± 10
Serum ALT	52 ± 5	45 ± 5	36 ± 15
Serum AST	43 ± 3	36 ± 5	35 ± 5

All data are shown as mean ± standard error of the mean. P-values for linear contrasts shown only when overall analysis of variance showed significant differences.

Abbreviations: ALP, alkaline phosphatase; ALT, alanine aminotransferase AST, aspartate aminotransferase; GGT, gamma glutamyl transferase.

^aCurrent vs past users.

^bCurrent vs non-users.

^cPast vs non-users.

Patterns of drug use

The mean age at first androgen use was 25 years with 12/72 (16%) starting under the age of 18 years, and the median duration of androgen abuse was 2.4 years for both current and past users. Fewer current compared with past (51% vs 81%, P = .008) users reported a cyclical pattern of androgen use, typically 12 weeks on followed by 12 weeks off, while the remainder were continuous users. All current and past users reported ever use of injectable androgens with 79% also reporting ever using oral androgens, but there was no difference between user groups. Current users were mostly using injectable androgens (30/41, 73%) with the remainder using oral (3/41, 7%) or both (8/41, 20%). A wide variety of androgens were used (Table 3), often simultaneously, comprising testosterone esters and endogenous pro-androgens (androstenedione, DHEA) as well as 17α alkylated (methandienone, oxandrolone, oxymetholone, oxandrolone, stanozolol, chlorodehydromethyltestosterone, danazol, methylepitiostanol, chlorodehydromethylandrostenediol, fluxymesterone), non-17α alkylated (methenolone, mesterolone, boldenone, drostanolone, trenbolone) and nonsteroidal (specific androgen receptor modulators) synthetic androgens.

Table 3.

Patterns of ever use of selected androgens and other drugs

Variable	Current users	Past users	Non-user controls	P  a
n	41	31	21
Number of drugs used#
Any androgen	7 (5, 8)	5 (3, 8)	—	.16
Any synthetic androgen	4 (3.5, 6)	3 (2, 5)	—	.07
Any alkylated androgen	2 (1, 3.5)	2 (1, 4)	—	.70
Any testosterone	2 (1,3)	2 (1,3)	—	.69
Any anti-estrogen	2 (1, 3)	2 (0, 3)	—	.47
Ever use of drugs
Injectable androgen	41	31	0	1.00
Oral androgen	34	23	0	.39
hCG	17	14	0	.81
Aromatase inhibitors	23	14	0	.48
Tamoxifen	28	23	0	.61
Clomiphene	21	18	0	.64
Growth hormone	16	9	0	.46
Insulin	9	2	0	.10
Thyroxine	15	5	0	.07
Diuretics	11	5	0	.39
SARM	2	3	0	.55
PDE5 inhibitors	20	11	0	.34
Cocaine	12	12	0	.45
Amphetamines	24	21	0	.47
Marijuana	24	25	10	.12
Opiates	8	3	1	.34
Hallucinogens	18	16	3	.54
Multivitamins	22	16	6	1.00
Creatine	18	15	3	.81
Protein supplements	34	25	11	1.00
Fish oil	12	4	0	.15
Liver detox	7	4	0	.75

Abbreviations: hCG, human chorionic gonadotropin; PDE5, phosphodiesterase 5; SARM, specific androgen receptor modulator.

^#Shown as median (interquartile range).

^a  P for comparison between current and past users for number of drugs used (Mann–Whitney) or for prevalence of ever use (Fishers exact test).

Ever use of other reproductive-active drugs was reported to be similar by current and past users for human chorionic gonadotropin (hCG; 32/72, 44%) and anti-estrogens (estrogen blockers tamoxifen: 51/72, 71%; clomiphene: 39/72, 54%; aromatase inhibitors: 37/72, 51%) as well as non-reproductive hormones (growth hormone: 25/72, 35%; thyroxine: 20/72, 28%; insulin: 11/72, 15%). None of the controls reported using these drugs. Compared with non-user controls, many current and past users reported use of other nonprescribed pharmaceutical drugs (amphetamines 63%, phosphodiesterase type 5 inhibitors 30%, diuretics 22%, L-dopa 2%, cabergoline 2%) but did not differ from non-user controls for reported use of antidepressant (5%) or hypnotics (4%). Current and past users together also displayed higher rates than non-user controls for use of cocaine (33%), proteins (58%), and creatine (32%), but they did not differ for use of other nonprescription nutritional products (multivitamins 47%, fat burners 7%, fish oil 15%, “liver detox” 10%, laxatives 5%) or other illicit drugs (marijuana 65%, opiates 13%, hallucinogens 31%).

Reproductive function

Current users had lower testis volume, sperm output and motility, higher serum testosterone, dihydrotestosterone, estradiol, estrone, 3α-diol and 3β-diol but lower serum DHEA, LH, FSH, SHBG, AMH, inhibin B, and total inhibin than non-users (Table 4). Variables that differed between current and non-user controls were not different between past and non-users consistent with full reversibility after cessation of androgen abuse except for mild residual reduction in testis volume and serum SHBG.

Table 4.

Reproductive function

Variable	Current users	Past users	Non-users controls	P  a	P  b	P  c
n	41	31	21
History of infertility	8	0	0	.009	.009
Acne	18	5	0	.02	<.001
Gynecomastia	12	7	0		.005
Temporal hair loss	13	10	2
Testicular volume (mL)	14 ± 2	19 ± 2	28 ± 2			.004
Semen volume (mL)	3.0 ± 0.2	3.4 ± 0.3	3.7 ± 0.3
Sperm output (million/ejaculate)#	4 (0, 39)	210 (80, 359)	203 (92, 340)	<.001	<.001
Motility (%)	29 ± 4	50 ± 4	57 ± 5	<.001	<.001
Serum LH (IU/L)	0.5 ± 0.3	5.5 ± 0.3	5.2 ± 0.4	<.001	<.001
Serum FSH (IU/L)	0.5 ± 0.3	4.7 ± 0.4	4.9 ± 0.4	<.001	<.001
Serum SHBG (IU/L)	17 ± 2	34 ± 2	42 ± 3	<.001	<.001
Serum testosterone (ng/mL)	36.5 ± 3.3	6.2 ± 3.8	8.7 ± 4.6	<.001	<.001
Serum DHT (ng/mL)	1.5 ± 0.1	0.5 ± 0.1	0.7 ± 0.2	<.001	<.001
Serum estradiol (pg/mL)	146 ± 17	41 ± 19	48 ± 23	<.001	<.001
Serum estrone (pg/mL)	65 ± 5	32 ± 6	38 ± 8	<.001	.005
Serum DHEA (ng/mL)	4.8 ± 0.5	6.6 ± 0.6	7.2 ± 0.7	.024	.008
Serum 3α-androstanediol (ng/mL)	2.2 ± 0.2	0.4 ± 0.2	0.6 ± 0.3	<.001	<.001
Serum 3β-androstanediol (ng/mL)	0.35 ± 0.03	0.16 ± 0.04	0.11 ± 0.06	<.001	<.001
Serum AMH (pg/mL)	3862 ± 621	7562 ± 599	7052 ± 728	<.001	<.001
Serum inhibin B (pg/mL)	97 ± 9	160 ± 10	173 ± 56	<.001	<.001
Serum total inhibin (pg/mL)	45 ± 4	85 ± 5	97 ± 6	<.001	<.001
Serum activin B (pg/mL)	88 ± 12	97 ± 14	76 ± 16

All data are shown as mean ± standard error of the mean. P-values for linear contrasts shown only when overall analysis of variance showed significant differences

Abbreviations: AMH, anti-Mullerian hormone; DHEA, dehydroepiandrosterone; DHT, dihydrotestosterone; FSH, follicle-stimulating hormone; LH, luteinizing hormone; SHBG, sex hormone binding globulin

^#Sperm output as median and interquartile range in brackets (excluding 6 vasectomized men).

^aCurrent vs past users.

^bCurrent vs non-users.

^cPast vs non-users.

Using blood test variables in a linear discriminant analysis, the optimal distinction between current and past users was achieved with 5 variables (serum LH, AMH, hematocrit, FSH, and total inhibin), which provided 96% (68/71) correct classification overall with correct classification of 100% of past users and 3 (7.5%) current users misclassified as past users. The discriminatory power of the individual components, expressed as standardized canonical coefficients, were LH (0.60), AMH (0.42), FSH (0.38), inhibin total (0.27), and hematocrit (0.24). Restricting the predictor variables to widely available blood tests (serum LH, FSH, hematocrit) reduces correct classification to 91.5% (65/71) while adding serum AMH increases it to 94% (67/71). The addition of other blood test (serum SHBG, hemoglobin, urea, creatinine), clinical (acne, gynecomastia), or sperm output variables did not further improve discrimination.

The average time to recovery, estimated from regression of the variable on time since cessation of androgen intake (Fig. 1) was faster for circulating reproductive hormone (serum AMH, 7.3 months; serum LH, 10.7 months) than for sperm variables (output, 14.1 months; concentration, 10.4 months) whereas biomarkers of spermatogenesis (serum FSH, inhibin B, total inhibin) took longer still to recover (Table 5). Total duration of androgen abuse was associated with a slower recovery of sperm variables (output, concentration, motility) and serum inhibin B, a Sertoli cell marker, but no other variables. For these sperm variables, the impact of time since cessation had greater impact on rate of recovery compared with total duration of use based on higher standardized regression coefficients (data not shown). Neither age nor any anthropometric variables (height, weight, BMI, BSA) were associated with rate of recovery of sperm or hormonal variables. Restricting the analysis to current users or to past users alone, who had recovered fully at some unknown time in the past, shows no significant relationship with time since cessation.

Figure 1.

Plot of sperm output (top left), serum FSH (top right), serum AMH (lower left), and serum LH (lower right) against time since last use (days, log scale). In each panel, the non-user controls are in green circle symbols, current users in red star symbols, and past users in blue square symbols. The solid line is the regression of the y axis variable on time since last use for current and past users together (excluding the non-user controls). The horizontal dashed line represents the mean of that variable for non-user controls and the vertical dashed line is the interpolation from the regression onto the x axis representing the average time taken to reach the mean for the non-users.

Table 5.

Rate of recovery for fully reversible androgen suppressed functions

Variable	Non-usersa	Mean time to recovery (months)
Sperm output (million)	189	14.1
Sperm concentration (million/mL)	54.4	10.4
Sperm motility (%)	60	37.6
Serum LH (IU/L)	5.2	10.7
Serum FSH (IU/L)	4.9	19.6
Serum AMH (ng/mL)	7.1	7.3
Serum inhibin B (pg/mL)	173	31.7
Serum inhibin total (pg/mL)	97	56.2
Fat-free mass (kg)	66	26
Serum HDL cholesterol (mM)	1.2	51
Basal cholesterol efflux	0.99	10.3
ABCA-1 specific cholesterol efflux	0.59	5.7

AMH, anti-Mullerian hormone; FSH, follicle-stimulating hormone; HDL high density lipoprotein; LH, luteinizing hormone

^aMean for all variables except median for sperm output, concentration and motility.

The self-administration of exogenous testosterone made it impossible to estimate directly the time to recovery of suppressed endogenous testosterone or its metabolites by this approach. Age, usage pattern (cyclical vs continuous), stacking (concurrent use of multiple androgens), or “post-cycle therapy” (current or ever use of hCG or antiestrogens to stimulate testicular function) had no significant impact on rate of recovery for sperm variables, testis volume or hormones.

Current users had a higher prevalence of a history of infertility, acne, gynecomastia, and temporal hair loss compared with past users and non-user controls (Table 4), and current and past users together had significantly higher prevalence than among non-users. There were no differences in reproductive hormone profiles among men with and without acne, gynecomastia, or temporal hair loss after adjustment for multiple testing.

Lipoprotein parameters and cholesterol efflux capacity

Current users displayed lower serum HDL cholesterol and higher serum triglycerides but no differences from non-user controls in serum total or low-density lipoprotein cholesterol. Basal and ABCA1-specific cholesterol efflux capacity were both significantly reduced in current users, but past users were no different from non-user controls indicating reversibility. Covariance analysis indicated that the differences in CEC were attributable to the concomitant changes in serum HDL cholesterol (data not shown). Interestingly, the rate of recovery of both basal and ABCA1-specific cholesterol efflux was faster than that of HDL concentration (Table 6).

Table 6.

Lipoprotein parameters

Variable	Current users	Past users	Non-user controls	P  a	P  b	P  c
n	41	31	21
Serum total cholesterol (mM)	3.93 ± 0.15	4.01 ± 0.17	4.04 ± 0.21
Serum HDL cholesterol (mM)	0.64 ± 0.05	1.10 ± 0.06	1.21 ± 0.07	<.001	<.001
Serum LDL cholesterol (mM)	2.69 ± 0.14	2.46 ± 0.16	2.41 ± 0.19
Serum triglycerides (mM)	1.32 ± 0.09	1.02 ± 0.11	0.91 ± 0.13	.041	.013
Serum basal cholesterol efflux	0.83 ± 0.03	1.00 ± 0.03	1.00 ± 0.04	<.001	<.001
Serum ABCA1-specific efflux	0.49 ± 0.02	0.61 ± 0.02	0.59 ± 0.02	<.001	<.001

All data are shown as mean ± standard error of the mean. P-values for linear contrasts shown only when overall analysis of variance showed significant differences.

Abbreviations: HDL, high density lipoprotein; LDL, low density lipoprotein.

^aCurrent vs past users.

^bCurrent vs non-users.

^cPast vs non-users.

Cardiac parameters

Left ventricular interventricular septum thickness and posterior wall thickness were within the normal range (≤11 mm) in all groups, but both were significantly greater in current users than in non-user controls (Table 7). Past users did not differ from non-user controls in both variables. LV mass, whether in absolute or BSA-indexed terms, was increased in current users relative to non-user controls, but there was no significant difference between non-users and past users. Right ventricular dimensions were normal and not significantly different between groups. Tricuspid annular plane systolic excursion was normal in all groups with no difference between current and non-users.

Table 7.

Echocardiographic and CT parameters

Variable	Current users	Past users	Non-users	P  a	P  b	P  c
n	41	31	21
Structure
Interventricular septum (mm)	10.7 ± 0.4	9.1 ± 0.4	8.7 ± 0.5	.007	.003
Posterior wall thickness (mm)	10.6 ± 0.3	9.3 ± 0.4	8.9 ± 0.5
LV EDD (mm)	52.3 ± 0.8	51.9 ± 1.0	49.3 ± 1.2
LV ESD (mm)	36.5 ± 1.3	36.2 ± 1.4	33.7 ± 1.7
Left ventricular mass (g)	232 ± 10	190 ± 11	167 ± 14	.006	<.001
LVM/BSA (g/m2)	106 ± 4	92 ± 4	81 ± 5	.016	<.001
End diastolic volume (ml)	137.4 ± 4.1	134.4 ± 4.8	127.4 ± 5.7
End systolic volume (ml)	60.0 ± 2.1	57.8 ± 2.4	52.9 ± 2.9
Left atrial volume (ml)	61.6 ± 2.5	58.7 ± 3.0	53.2 ± 3.5
Left atrial volume indexed (mL/m2)	29.1 ± 1.2	28.5 ± 1.4	26.4 ± 1.6
Right ventricle basal diameter (mm)	38.3 ± 1.0	37.4 ± 1.1	38.0 ± 1.3
Right ventricle mid diameter (mm)	27.72 ± 0.9	27.4 ± 1.0	29.2 ± 1.2
Right ventricle length (mm)	74.1 ± 1.8	70.7 ± 2.1	70.4 ± 2.5
Systolic function
Left ventricular ejection fraction (%)	56.7 ± 0.5	57.3 ± 0.6	58.2 ± 0.74
Global longitudinal strain (%)	–17.6 ± 0.4	–19.4 ± 0.5	–19.7 ± 0.5	.002	.001
TAPSE (mm)	2.4 ± 0.1	2.4 ± 0.1	2.2 ± 0.1
Diastolic function
E-wave (cm/s)	75.3 ± 2.3	73.4 ± 2.8	81.9 ± 3.2
A-wave (cm/s)	50.6 ± 2.0	49.4 ± 2.3	55.7 ± 2.7
E/A ratio	1.56 ± 0.09	1.66 ± 0.11	1.52 ± 0.12
Early septal ventricular E’ (cm/s)	9.5 ± 0.5	10.7 ± 0.6	12.1 ± 0.7		.006
Early lateral ventricular E’ (cm/s)	13.3 ± 0.7	14.9 ± 0.8	16.6 ± 1.0		.006
Average E/e’ ratio	7.25 ± 0.28	6.34 ± 0.33	6.53 ± 0.39
CT calcium score
Calcium score (Agaston units)	0.0 (0, 0)	0.0 (0, 0)	0.0 (0, 0)

Data are shown as mean ± standard error of the mean except for calcium score as median (interquartile range). P-values for linear contrasts shown only when overall analysis of variance showed significant differences

Abbreviations: BSA, body surface area; EDD, end diastolic diameter; ESD, end systolic diameter; LV, left ventricular; LVM, left ventricular mass; TAPSE, tricuspid annular plane systolic excursion.

^aCurrent vs past users.

^bCurrent vs non-users.

^cPast vs non-users.

LV ejection fraction was within the normal range in all groups. However, global longitudinal strain, a more sensitive marker of LV function, was reduced among current users compared with non-user controls. Again, past users did not differ from non-user controls. Early LV relaxation velocity (E’), a marker of diastolic function, was reduced in current users relative to non-user controls for both septal and lateral E’ (Table 7). Again, past users did not differ from non-user controls in each of these variables. Left ventricular end diastolic diameter was mildly increased in current users relative to control; otherwise, there were no other significant differences in echocardiographic parameters noted.

Coronary artery calcification was not different between groups with a median score of 0 (IQR 0–0) in all groups. A score of >1 Agatson units was detected in 10 (11%) participants including 2 controls (136, 164 units), 5 current (2–9 units) and 3 past (17, 41, 420 units) users.

Discussion

Androgen abuse has deleterious effects on male reproductive and cardiac function, but there is little systematic study reported of the rate and extent of recovery after cessation of androgen abuse. The present cross-sectional, observational study design is the best available as prospective study of this illicit activity would be difficult to conduct.

This study shows that suppression of testicular function due to androgen abuse by young men undertaking mainly cyclical use of nonprescribed androgens for an average of over 2 years appears to be fully but slowly reversible, apart from residual reduction in testis volume and serum SHBG, as the effects among past users were mostly no different from non-users. Variables with complete reversibility allowed us to estimate the average time to recovery of reproductive functions by interpolating the time taken to reach the average of non-user controls from the regression of time since cessation. The time to recovery varies from 7 to 9 months for testicular steroidogenesis with slower recovery (10–14 months) of spermatogenesis (Table 5). Although intake of exogenous testosterone made it impossible to estimate the time to recovery of endogenous testosterone, the time to recovery of serum AMH and LH provide realistic surrogate estimates of recovering endogenous testosterone as LH is the principal driver of testicular testosterone secretion. The present estimates are a little shorter than an estimate from a meta-analysis, suggesting normalization of serum LH and FSH between 13 and 24 weeks (47). For sperm variables and inhibin B, a marker of Sertoli cell function, but not other hormonal variables reflecting Leydig cell function, the duration of prior androgen abuse was also associated with a slower recovery time whereas age and anthropometric variable did not influence rate of recovery. Overall, these findings are most consistent with a reversible effect on testicular hormone secretion but an additional cumulative detrimental effect on sperm production with prolonged androgen-induced suppression of spermatogenesis as evident in the residual effect on testis volume. An assumption of this novel approach to estimating the time to recovery is that the current and past users form a single cohort on a continuum of time since cessation of androgen intake. This is consistent with the profiles of current and past users who did not differ in pattern of drug use and differed only slightly in age of commencing androgen abuse, a variable not linked to any study outcome.

The present findings differ from previous studies reporting persistent reduction in serum testosterone long after reported cessation of androgen abuse (13, 48). These discrepancies may be due to either ongoing surreptitious androgen abuse (to alleviate androgen deficiency withdrawal symptoms and/or loss of muscle mass gain that motivated androgen abuse) or to prolonged impairment of testicular function perhaps from greater net androgen exposure. Objective evidence for the former may come from urine drug screening to detect exogenous androgen intake although studies so far have not reported such testing, and failure to detect use at the time of study may not exclude persistent effects of prior androgen intake. For the latter possibility, whether more prolonged and/or intense androgen exposure may have greater detrimental effects is difficult to assess due to the unreliability of uncorroborated recall of drug history from users and inability to equate potency of different synthetic androgens. The relatively full recovery in the present study argues against undisclosed ongoing androgen abuse compared with Rasmussen et al (13), which reported persistent decreases in blood testosterone (but did not measure sperm output) over longer follow-up period although the duration of androgen use and number of androgens used were comparable in both studies. Future studies should investigate these alternatives.

To our knowledge, there are no previous estimates of sperm output recovery. We observed the biomarkers of testicular spermatogenesis were slower to recover than sperm output. Although previous studies show a persistent reduction in sperm output and quality after cessation of androgen intake (49–51), the degree and tempo of recovery was not defined. Recovery to normal sperm output within 12 months (52, 53) as well as failure of recovery for over a year (54–56) are reported. Prospective studies using lower, standard testosterone replacement doses for male contraception show a median time to recovery to normal sperm output of 4 to 5 months with 95% recovering within 12 months (57, 58). However, androgen abuse involves much higher, often massive doses of multiple androgens, so recovery may take longer and/or be less complete. Furthermore, the low requirement of spermatogenesis for intratesticular testosterone (59, 60) means that sperm output recovery may be faster than endogenous testosterone. Delayed recovery of sperm output has prompted uncontrolled empirical trials of hCG treatment to improve sperm output for infertile couples assuming that post-cessation of androgen abuse represents a hypogonadotropic state (54, 55). Yet no studies have investigated this hypothesis and claimed success in uncontrolled reports coincides with the likely time for natural recovery of sperm output. Our findings suggest that persistent azoospermia after ceasing androgen abuse may reflect undiagnosed prior reproductive pathology and/or persistent surreptitious use of androgens to relieve withdrawal symptoms although irreversible androgen-induced spermatogenic damage with prolonged androgen abuse cannot be excluded (11, 12).

The study provided information on determinants of androgen abuse and its consequences. The higher than expected proportion of non-heterosexuals (46) is consistent with the higher prevalence of androgen abuse and body building among gay or bisexual men (61) but only accounted for a small proportion of the study cohort. There was no significant difference in sperm suppression or recovery between cyclical or continuous regimens or stacking, the concurrent use of multiple androgens. This does not support the folkloric rationale that intermittent “holidays” from androgen abuse maintain or restores sensitivity to high-dose androgens or that within supraphysiological dosing regimens multiple androgens achieve different results from a single androgen. Furthermore, the use of HCG or anti-estrogens, known colloquially as “post-cycle therapy,” were not associated with alleviation of suppression or improved recovery of sperm output but well-controlled studies are lacking. In 1 study of 18 power athletes using massive androgen doses, concurrent use of hCG was associated with better sperm output but also with detrimental changes in sperm morphology and quality suggestive of poor sperm function (51). Previous studies of androgen abusers have reported reduced testicular size (48, 50, 62), but the degree and tempo of reversibility has not been reported in detail. Our findings of persistent reduction in testis volume at a median time of nearly a year after ceasing androgen intake are consistent with a previous report where reduction in testicular size remained at a median 2.5 years after ceasing androgen intake (13). Acne and gynecomastia reported frequently among current and past users is consistent with previous studies (62–64). Delayed recovery from androgen abuse-induced hypogonadism has been reported previously case reports and small studies (48), but to our knowledge, there are no prior systematic estimates of rate and extent of recovery of suppressed endogenous testosterone production.

Androgen use was associated with reversible reduction in serum HDL, elevation in serum triglycerides and decreased CEC, a prognostically important measure of HDL function (65). Analysis by analysis of covariance indicated that changes to CEC were attributable to the concomitant changes in serum HDL cholesterol, thus similar to findings of 1 observational study (66), but differing from a recent study of androgen replacement after chemical castration which showed changes in HDL without change in CEC (67). This suggests that decreased CEC is a novel and temporally sensitive adverse consequence of androgen abuse. That changes in ABC1-specific efflux were seen implies that specific subpopulations of HDL particles may especially sensitive to androgens (39).

Structural (LV mass) and functional (global myocardial strain) echocardiographic parameters were adversely affected by current androgen abuse, and all showed reversibility after cessation. Numerical differences between past users and non-users in some parameters allude to different rates of recovery of cardiac parameters. Subclinical impairment of LV global longitudinal strain we observed may convey long-term adverse prognostic cardiovascular risk (68). The increased LV mass and diastolic dysfunction with current androgen use is consistent with recent reports (28). Why Rasmussen et al (28) showed persisting abnormalities in global strain much later after cessation than our group which displayed complete recovery is unclear but may relate to differences in net androgen exposure, which is difficult to document retrospectively in this population or other unknown factors. There was no evidence of premature coronary calcification in this young population (mean age 34 years), which is consistent with Baggish et al (17), who studied men a decade older (median age 42 years) with much longer duration of androgen use (median 7.4 years). These findings differ from a small uncontrolled study of older professional bodybuilders who used high doses of androgens for a mean of 12.6 years (69).

Strengths of this study include the comprehensive analysis of reproductive function including sperm output, testicular volume measurement, and steroid hormone measurement by liquid chromatography–mass spectrometry for current and past androgen abusers with healthy non-users matched for age and regular exercising. The limitations include the observational nature, which was unavoidable for an illicit, secretive, and detrimental self-medication with nonapproved drugs. As it would not be ethical or feasible to study such behavior prospectively, the present design was the best alternative. Although the sample size was relatively small, we obtained sufficient data to make plausible, evidence-based interpretations about the rate, extent, and determinants including the impact of patterns of abuse and additional drugs or nutritional supplements ingested. Larger studies that include androgen abusers with longer duration of use and follow-up might clarify the proportions, if any, of ex-androgen abusers who have persistent and significant impairment of reproductive function. Other limitations are that this study did not investigate sperm function tests (deoxyribonucleic acid fragmentation, acrosome function, etc.), which would evaluate if recovery of normal sperm output was also accompanied by complete restoration of sperm fertilizing ability. Furthermore, did we did not test past and non-users with urine antidoping tests to evaluate whether they were free from current androgen intake, although recovery of normal reproductive hormones and sperm output suggested ongoing surreptitious androgen intake was rare or nonexistent in those groups. As this study only included male androgen abusers, the impact of androgen abuse on females requires further study, noting the paucity of studies (70).

We conclude that suppressed testicular function due to androgen abuse in men of similar age and lifetime androgen exposure in this study is virtually completely reversible within 7 to 18 months after ceasing androgen intake. The duration of prior use retarded recovery of sperm but not of hormonal variables. Whether these findings generalize to older men with longer duration of androgen exposure warrants further study. For current androgen abusers, suppressed serum AMH, LH, and FSH represent convenient, useful, and underutilized markers of androgen abuse and recovery that directly reflect testicular suppression. By providing a read-out of the brain’s state of recovery from androgen abuse, they may provide useful biomarkers for clinicians in as supportive management of patients during the slow recovery as well as forming surrogate biomarkers for studies of recovery from androgen abuse. Androgen abuse is a neglected public health issue, and further studies to support prevention and recovery from androgen abuse, notably for adolescent and young men, are needed together with public and professional education.

Acknowledgments

We thank Dr Yogeshwar Makanji and Dr Ajay Kumar (Ansh Labs) for their assistance in peptide hormone assays and Dr Eva Jackson (Nepean Blue Mountains Sexual Health Clinic) for helpful assistance.

Additional Information

Disclosure Summary: There was no external funding for this study. DJH has received institutional grants without personal income for investigator-initiated androgen pharmacology studies and has provided expert testimony to anti-doping, professional standards and tort litigation hearings. No other authors have any relevant declarations.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.