Reliability of Drug History to Verify Androgen Abuse in Men.

CONTEXT: Clinical evaluations that require excluding androgen abuse, a secretive, illicit activity, rely on the drug history, but its veracity for androgen abuse has neither been verified nor has any objective corroborating laboratory test been validated.
OBJECTIVE: In a high-risk population, to (a) validate the drug history of androgen abuse objectively using state-of-the-art World Anti-Doping Agency-accredited antidoping laboratory urine mass spectrometry tests and (b) to determine what biochemical tests best distinguish androgen abuse from nonuse in this population.
METHODS: Urine samples from current (n = 41) and past (n = 31) androgen abusers and nonusers (n = 21) were analyzed by comprehensive mass spectrometry-based detection tests for androgens and related drugs (ARD).
RESULTS: No prohibited ARDs were identified among nonusers. Current users had a median of 5 (range 1-13) drugs detected comprising 176 ARDs among 220 drug identifications. Past users had a median of 1 (range 0-9) drugs detected comprising 21 ARDs among 43 drugs. Negative predictive value was high (>0.8) for those denying drug usage while positive predictive value was good (>0.6) for both those reporting currently using (current) and not using (nonusers plus past users) ARD. Serum luteinizing hormone (LH) alone had high, but imperfect, discriminatory power (89%) to distinguish between current and noncurrent androgen use.
CONCLUSIONS: We demonstrates that a negative drug history in a high-risk group has high reliability and that even a single suppressed serum LH exhibits high discrimination for objectively detecting androgen abuse.

Androgen abuse (1) is a growing but poorly understood and little researched public health threat (2, 3) for young men (4). It refers to the use of testosterone and synthetic steroidal (anabolic steroids) or nonsteroidal (specific androgen receptor modulators) androgens together with drugs aiming to increase endogenous testosterone, all without a valid prescription. This pattern of drug use is herein referred to as androgen and related drugs (ARD). Originating in the early post–World War II decades in elite power sports where it aimed to exploit androgen-induced increases in muscle mass and strength for performance enhancement, androgen doping was banned by the International Olympic Committee for elite sports by the mid-1970s. This prohibition is currently enforced under the World Anti-Doping Code employing sanctions based on a strict liability policy, wherein a positive test makes the athlete at fault regardless of intention, negligence, knowing use, or otherwise. Testing of urine samples is conducted by World Anti-Doping Agency (WADA)-accredited antidoping laboratories that operate harmonized, highly sensitive urine detection tests for androgens and their metabolites with windows of detection that may extend for months from the last dose. Although imperfect, this regimen effectively deters androgen doping in elite sports, now detected at very low levels (<1% of tests) despite androgens remaining the most potent and popular (5) class of ergogenic doping agents (6). By the 1980s, androgen abuse crossed over into the general community for image rather than performance enhancement (7). Now most often used for bodybuilding (8), androgen abuse is an endemic variant of drug abuse among communities with sufficient affluence and leisure to sustain an illicit drug subculture.

Androgen abuse has wide-ranging side effects in men with prominent adverse effects on mood, leading to reckless, aggressive behavior, suppression of reproductive function, and harmful effects on other organs and tissues (9) but also sought-after effects on muscle mass and strength, mood, and body composition, which motivate its illicit use (10). The potentially dangerous adverse behavioral effects of high-dose androgens in men have been reported in well-designed placebo-controlled studies of healthy individuals (11) but are also confounded by other drug abuse and underlying personality factors. Androgen abuse universally suppresses the endogenous male reproductive system, causing markedly decreased sperm and testosterone production, leading to male infertility as well as lethargy and sexual dysfunction, respectively, which persist for variable periods after cessation of androgen administration (12, 13). The natural history of androgen abuse includes a slow, progressive recovery of male reproductive functions taking 6 to 18 or longer months, with less complete recovery following prolonged androgen abuse (12, 13), and similar effects are reported even after prolonged testosterone treatment at therapeutic doses (14). However, well-controlled analytical research into motivation and consequences of androgen abuse remains scarce (10, 15, 16).

A pivotal datum for investigation of androgen abuse is the self-reported drug history. As a secretive, illicit activity, androgen abusers may not provide a truthful drug history. This uncertainty is reinforced by their perception of doctors as antagonist gatekeepers of drug prescription and health monitoring, which they seek on request but are typically denied. However, among androgen abusers, objective verification of the reliability of drug history as well as any valid objective diagnostic biomarkers has not been reported systematically. The gold standard for verification of androgen abuse is the WADA-accredited urine drug screening tests in their certified antidoping laboratories. These tests are subject to rigorous ongoing accreditation to maintain a medicolegally defensible standard of objectively detecting the full range of drugs on the annually updated WADA Prohibited List (17) sufficient to preclude professional athletes from competing. Prudently, these laboratories do not offer open access to analytical services to avoid “gaming” by athletes with doping intentions seeking to determine sensitivity and windows of detection for prohibited drugs. The present study evaluated single blood, urine, and semen samples from current and past androgen abusers as well as nonuser controls (13). This allowed verification of the drug declarations obtained from a standardized, in-person drug history by comprehensive urine drug tests in a WADA antidoping laboratory. In addition, we investigate the utility of serum and urine hormonal profiling to verify androgen abuse.

Materials and Methods

Clinical Study and Participants

This study was a secondary analysis of a previously reported clinical trial (13). Briefly, male current users (defined as using androgens at the time of recruitment; n = 41), past users (defined as no androgen use for at least 3 months at the time of recruitment, with a median time since last use of 300 day; n = 31) and nonusers (regularly exercising men who never used androgens; n = 21) were recruited by social media, word of mouth, and conventional advertising. After providing written informed consent, each participant was interviewed by 1 endocrinologist (N.S.N.) using a standardized data collection format to identify use, recent or ever, of nonprescribed androgens or related prohibited substances. For each substance, the participant was asked if he ever used it, and if he had used it, was it used in his most recent use of androgens. Androgens were classified as substances covered by S1 of the WADA Prohibited List including clenbuterol, which is not an androgen, and finasteride, which is no longer prohibited. The clinical study was approved by the Sydney Local Health District Human Ethics Review Committee (Concord zone). None of the participants were involved in competitive sports. Approved recruiting material made clear the observational study would not provide payment, prescriptions, or health monitoring within or outside the study to participants. The laboratory urine antidoping profile analyses of coded samples was approved by Sport Integrity Australia, the WADA-affiliated national antidoping organization.

Steroid Analysis Methodology

The analysis of urine for steroids and related drugs was performed on never-thawed, frozen, stored urine samples from all 93 participants at the national WADA-accredited antidoping laboratory, the Australian Sports Drug Testing Laboratory (National Measurement Institute, North Ryde, Australia). The analyses, including negative and positive controls with internal stable isotopic standards (to detect contamination and estimate recovery) used gas or liquid chromatography-mass spectrometry methods in 5 runs employing harmonized analyses certified to cover all androgens and related substances on the contemporary WADA Prohibited List (17). The urinary endogenous steroid profiling and performance-enhancing drug screening procedure are validated for antidoping control according to the applicable World Anti-Doping Agency technical documents (WADA Technical Documents for Endogenous Anabolic Androgenic Steroids and Minimum Required Performance Levels, https://www.wada-ama.org). Where antidoping results would be considered to require isotope ratio mass spectrometry to verify an individual substance for defining an individual antidoping rule violation, for this study isotope ratio mass spectrometry was not performed, and the sample was classified as positive. Details of sample preparation, analysis and instrumentation are described in the Supplementary Information (18).

Data Analysis

Drug declarations were stratified according to recent and ever use. Specific androgens and related prohibited substances were categorized into testosterone, nandrolone, synthetic androgens including a wide range of marketed and nonmarketed illicit androgens (anabolic steroids) as well as nonsteroidal androgens (synthetic androgen receptor modulators) and estrogen inhibitors (comprising estrogen receptor antagonists such as tamoxifen and clomiphene (specific estrogen receptor modulators) as well as aromatase (estrogen synthetase) inhibitors such as anastrozole and letrozole.

For each participant, their declarations were compared with the similarly grouped laboratory detections to compute the positive predictive value (PPV) and negative predictive value (NPV), specificity, and sensitivity as well as chance-corrected kappa correlations with 95% confidence limits using standard definitions from 2 × 2 tables (19). Predictive values were considered high if they exceeded 0.8 and good if they exceeded 0.6, compared with a chance predictability of 0.5. The sum of sensitivity and specificity defines the Youden index (sensitivity + specificity − 1), which identifies the optimal cutpoint for correct categorical classification in a 2 × 2 table when equal weight is given to sensitivity and specificity (19).

In this setting of frequent multidrug use, we aimed to identify any common underlying patterns of multidrug androgen use using principal component analysis (PCA). This aimed to identify optimal combinations of drugs used together more than by chance. Using varimax rotation to define its major components, the analysis identifies a reduced set of conjoint patterns of use reported in canonical form.

To determine the optimal biochemical criteria to objectively discriminate current androgen abuse from those not currently using androgens, linear discriminant analysis was applied to current data on urine luteinizing hormone (LH) and testosterone (T)/LH ratio together with previously published data on serum LH, follicle-stimulating hormone (FSH), and anti-Müllerian hormone (AMH) (13) to distinguish between current users and those denying current usage (pooling nonusers and past users). The optimal linear discriminant function was 0-centered so that negative (suppressed) values of the discriminant function indicated high likelihood of androgen abuse whereas positive (nonsuppressed) values indicated low likelihood. Data analyses was undertaken in NCSS 2021 statistical software (NCSS, Kaysville, UT, USA)

Results

Pattern of Drug Declarations

None of the 21 nonuser controls declared any use of androgens or related prohibited substances. Numbers of drugs declared by current and past androgen abusers and nonusers are summarized in Table 1.

Table 1.

Declarations of recent and ever use of androgens by group

Declarations	Recent, n (%)	Ever
Nonusers (n = 21)
Testosterone	0 (0)	0 (0)
Nandrolone	0 (0)	0 (0)
Synthetic androgen	0 (0)	0 (0)
Estrogen inhibitors	0 (0)	0 (0)
hCG	0 (0)	0 (0)
Current users (n = 41)
Testosterone	37/ (90)	41 (100)
Nandrolone	14/ (34)	32 (78)
Synthetic androgen	33 (80)	38 (93)
Estrogen inhibitors	12 (29)	34 (3)
hCG	0 (0)	17 (41)
Past users (n = 31)
Testosterone	23 (74)	28 (90)
Nandrolone	12 (39)	20 (65)
Synthetic androgen	15 (48)	27 (87)
Estrogen inhibitors	0 (0)	23 (74)
hCG	0 (0)	14 (45)

Abbreviation: hCG, human chorionic gonadotropin.

Pattern of Urine Drug Detections

None of the 21 nonuser controls had any androgens or related substances detected in urine samples. Among nonusers, 16 had no drugs detected and 5 had a total of 8 drugs identified comprising 1 (3), 2 (1), or 3 (1) drugs identified in individuals (Table 2). The detected drugs were cocaine (3 instances) and single instances of finasteride, vilanterol, salbutamol, codeine, and cannabis (tetrahydrocannabinol).

Table 2.

Drugs detected on urine steroid and drug screen

	None	Testosterone	Nandrolone	Synthetic androgens	Estrogen Inhibitors	Other drugs
Current users (n = 41)	0	35	26	Trenbolone (20) Boldenone (19) Drostanolone (14) Methandienone (9) Oxandrolone (8) LGD4033 (6) Mesterolone (4) Ostarine (4) Stanozolol (4) Oxymetholone (3) Dehydrochloromethyl testosterone (1) RAD140 (1) S29 (1)	Tamoxifen (9) Anastrozole (6) Letrozole (1) Exemestane (1)	Cocaine (1) Cannabis (THC, 4) Amphetamines (4) Clenbuterol (3) GW1516 (3) Ibutamoren (3) Heptaminol (2) Higenamine (2) Methylhexanamine (2) Octodrine (2) Salbutamol (2) Alcohol (1) Amiloride (1) Buprenorphine (1) Chlorothiazide (1) Codeine (1) Hydrochlorothiazide (1) Modafinil (1)
Past users (n = 31)	10	2	4	LGD4033 (3) Boldenone (2) Ostarine (2) Andarine (1) Drostanolone (1) Oxymetholone (1) Stanozolol (1) RAD140 (1) hCG (1)	Clomiphene (1) Tamoxifen (1)	Cocaine (7) Cannabis (THC, 4) Amphetamines (4) Clenbuterol (2) GW1516 (2) Codeine (1) Tramadol (1) Ibutamoren (1) Salbutamol (1)
Nonusers (n = 21)	16	0	0	0	0	Cocaine (3) Finasteride (1) Vilanterol (1) Salbutamol (1) Codeine (1) Cannabis (THC, 1)

Abbreviation: THC, tetrahydrocannabinol.

Current users had a median of 5 drugs identified (range 1-13 drugs per person) with urine tests detecting 220 drugs (Table 2). Past users had a median of 1 drug (range 0-9 drugs) detected in urine samples at a time when they had reported a median of 300 days (range 90-4380 days) since last drug use. Ten of 31 past users had no drugs detected in urine, and the remaining men had a median of 2 drugs detected (range 1-9 drugs). There was a marginally significant (P = 0.06) shorter time since last drug use among those with a positive urine test [median 225 days (95% confidence limits 120, 365 days) vs 420 days (95% confidence limits 180, 730 days)].

The highly collinear patterns of androgen detections investigated by PCA identified 3 principal components. The first component had major loadings for nandrolone (0.936) and T (0.686), the second major component had a major loading for synthetic androgens (0.946) and the third component had a major loading for estrogen inhibitors (0.958). All other loadings were less than 0.340.

Linear discriminant analysis was used to evaluate potential generic markers that display a nonspecific suppression to any androgens. Using urine LH and T/LH ratio together with previously published data on serum LH, FSH, and AMH (13), serum LH alone could distinguish between current users (n = 41) and those denying current usage (nonusers plus past users, n = 52) with an overall accuracy of 89% (87/92; 3 false positive, 2 false negatives). The optimal discriminant function was

and was zero-centered so that negative values of C predict current androgen abuse whereas positive C values predict no current androgen usage. This formula indicates that a serum LH of 3.25 IU/L (C = 0) optimally demarcates suppressed from nonsuppressed serum LH values in this high-risk population.

The other potentially generic androgen-responsive markers displayed less discrimination—serum FSH (74% overall accuracy, 20% false positive, 7% false negative), serum sex hormone-binding globulin (57% overall accuracy, 20% false positive, 7% false negative), urine LH (55% overall accuracy, 33% false positive, 6% false negative), and urine T/LH ratio (55% overall accuracy, 0% false positive, 25% false negative). Adding serum FSH, sex hormone-binding globulin, AMH, urine LH, and T/LH ratio did not improve proportion of correct classification (data not shown).

Correlation of Declarations With Detections

The PPV and NPV values for the drug declarations are shown in Figure 1, which depicts the predictive values for the full data set (n = 93) as well as for those admitting current use (n = 41) and those denying current use (nonusers plus past users, n = 52). PPV was higher among current users (most exceeding 0.6) than among those not admitting current androgen use (none exceeding 0.6 PPV). NPV was very high among nonusers (all exceeding 0.8 NPV) and much lower in current androgen users (only 5/41 exceeding 0.8 NPV).

Figure 1.

Positive and negative predictive values for drug declarations compared with the gold standard of laboratory detections. Upward bars in solid fill represent positive predictive value and downward bars with grey fill represent negative predictive value. Large left panel includes all 93 participants in the study comprising 21 nonusers, 31 past users, and 41 current users. The smaller upper right panel includes only current users. The smaller lower right panel includes all noncurrent users including both nonusers and past users. For further including specificity, sensitivity, and kappa correlations details, see text.

The kappa values with 95% confidence limits as well as sensitivity and specificity and percent correct classification are shown in Table 3.

Table 3.

Reliability of drug history for androgen abuse

	Kappa statisticsb		Sensitivity and specificityc		Correct classificationd
Declarationsa	Recent	Ever	Recent	Ever	Recent, n (%)	Ever, n (%)
All participants (n = 93)
Testosterone	0.452 (0.297, 0.607)	0.374 (0.237. 0.511)	0.58/0.94 (0.52)	0.54/1.00 (0.54)	66 (71)	61 (66)
Nandrolone	0.337 (0.132, 0.542)	0.340 (0.173, 0.506)	0.58/0.78 (0.35)	0.48/0.88 (0.36)	67 (72)	61 (66)
Synthetic androgen	0.613 (0.453, 0.773)	0.499 (0.345, 0.652)	0.75/0.84 (0.59)	0.65/0.96 (0.61)	74 (80)	69 (74)
Estrogen blockers	0.391 (0.142, 0.639)	0.210 (0.091, 0.329)	0.58/0.88 (0.46)	0.28/0.97 (0.25)	79 (85)	53 (57)
Group 1 (n = 41)
Testosterone	0.320 (−0.092, 0.733)	— ^	0.89/0.50 (0.39)	0.85/—e (—)	35 (85)	35 (85)
Nandrolone	0.191 (−0.054, 0.436)	0.081 (−0.213. 0.376)	0.79/0.44 (0.23)	0.66/0.44 (0.10)	23 (80)	25 (61)
Synthetic androgen	0.266 (−0.093. 0.626)	0.554 (0.183, 0.926)	0.88/0.38 (0.25)	0.89/1.00 (0.89)	32 (78)	37 (90)
Estrogen blockers	0.286 (−0.017, 0.590)	0.129 (−0.041, 0.300)	0.58/0.72 (0.31)	0.41/0.86 (0.27)	28 (68)	20 (49)
Group 2 (n = 31)
Testosterone	0.047 (−0.023, 0.117)	0.015 (−0.011, 0.041)	0.09/1.00 (0.09)	0.07/1.00 (0.07)	10 (32)	5 (16)
Nandrolone	0.380 (0.090, 0.670)	0.151 (−0.002, 0.303)	0.33/1.00 (0.33)	0.20/1.00 (0.20)	23 (74)	15 (48)
Synthetic androgen	0.395 (0.084, 0.706)	0.016 (−0.142, 0.174)	0.47/0.88 (0.34)	0.30/0.75 (0.05)	22 (71)	11 (35)
Estrogen blockers	—e	0.047 (−0.023, 0.117)	—e/0.94 (—)	0.09/1.00 (0.09)	2 (6)	2 (11)

Bold text indicates statistically significant kappa values results.

aHuman chorionic gonadotropin is not tabulated as there was only a single positive recorded in 1 past user.

bAll kappa values are identical for unweighted, linear, or quadratic weighting with 95% CIs in parentheses and bold indicating statistically significant chance-corrected verification.

cThe Youden estimate is in in parentheses.

dCorrect classification of declarations relative to current urine testing. Relatively high rates of false negatives among past users reflect recollecting past use but without current positive urine detections.

eUndefined due to lack of any relevant declarations or positive urine tests.

Effects of Androgen Abuse on Urine Steroid Profiles

The urine steroid profiles showed highly significant differences between current and nonuser groups (signifying androgen effects) for 13 markers but not for 5 others (4-hydroxy androstenedione and human chorionic gonadotropin concentrations together with the A/E ratio, 5α/5B ratio and 5α A/A ratios) [Supplementary Table 1 (18)]. By contrast, past users did not differ overall from nonuser groups (signifying recovery from androgen effects) in 17 of the 18 markers except for epitestosterone, which remained significantly lower than in nonusers [Supplementary Table 1 (18)].

Using the absence of detected androgens among nonuser controls, an analysis of the apparent impact of detection of categories of androgens reflects the heavily overlapping patterns of androgen use [Supplementary Table 2 (18)], rendering aberrations of urine steroid profiles a nonspecific indicator of the multidrug androgen abuse patterns.

Discussion

The present study provides objective verification of the drug history of androgen abuse collected by a standardized in-person drug history compared with the gold standard of urine antidoping drug screening for which we have been unable to locate any prior comparable data. The major present findings are that the drug history of androgen abuse has variable reliability in predictive value for groups with different prior probability of androgen abuse. A striking finding was the very high reliability of NPV among nonusers (all exceeding kappa 0.8) although it was much lower among current androgen users [with only 5/32 classifications have good agreement (kappa > 0.8)]. Conversely, the PPV was higher among current users (most exceeding 0.6) than among those not admitting current androgen use (none exceeding 0.6). Such PPV estimates from androgen users is likely to be an underestimate because drugs with short duration of action may have disappeared from the urine despite relatively recent usage. As the reproducibility of such predictive findings are sensitive to the prevalence of androgen abuse in any sample (20), in addition to the overall analysis of the whole study group, we also analyzed predictive values (sensitivity, specificity, kappa values, correct classification) in subsets admitting (current users) and those denying androgen use (nonusers plus past users). These findings outline the cautions necessary in accepting the validity of even a careful in-person drug history of androgen abuse for clinical case management or analytical research without additional objective verification by objective analytical criteria. Whether the same validity extends to written questions instead of face-to-face response remains to be determined.

PCA showed that the multidrug use fell into 3 predominant, distinct patterns of conjoint usage. The use of testosterone and nandrolone together were particularly frequent with all synthetic androgens forming a separate distinct conjoint subset. The third group were estrogen inhibitors, which are typically used between abuse cycles as so-called postcycle therapy, used in an attempt to overcome suppressive effects of androgen abuse on the reproductive system so it is not surprising that they constitute a distinct subset from the active androgen abuse drugs. These findings highlight that androgen abuse is complex multidrug regimen and is not identifiable with any single androgen or androgen related drug(s). This is highly relevant since state-of-the-art urine detection tests are not, nor likely to become, widely available.

The present finding from the linear discriminant analysis showed that a single suppressed serum LH was the best criterion for generic identification of androgen abuse with 89% correct classification. Hence, serum LH, as a single analyte, provides a useful, convenient, and easily accessible diagnostic tool to verify or exclude the suspicion of androgen abuse where clinically or analytically relevant. In the absence of widely available state-of-the-art urine tests, a single suppressed serum LH measurement provides a powerful independent contribution to distinguishing between those reporting or denying current androgen use and serves to reinforce the validity of the drug history. The present findings are consistent with evidence that serum LH and FSH are important, simple biomarkers of recovery from androgen abuse (13) as well as in recovery of reproductive suppression following prolonged injectable testosterone treatment at therapeutic doses (14). While serum FSH was less influential alone in discriminating androgen abuse from nonusers, it is a temporally more stable hormone than LH, which characteristically is secreted in pulsatile mode, so is useful when coupled with serum LH measurements. In several studies we have shown that serum LH is more discriminating than urine LH, with or without urine testosterone (21-23), because renal hormone excretion adds noise to their interpretations compared with serum measurements with their direct physiological impact. A caveat on interpretating serum LH is that pathological testicular disorders causing defective Leydig cell testosterone production would produce a reflex rise in baseline and recovery serum LH (and FSH) concentrations.

Furthermore, we evaluated the impact of patterns of androgen abuse on the urine steroid and serum reproductive hormone profile. These indicate that among androgen abusers, aberrations of urine steroid profiles are nonspecific, due to the complex idiosyncratic patterns of androgen abuse rather than any specific androgen or prohibited substance alone. The urine steroid profiles demonstrate the near ubiquitous use of testosterone in androgen abuse regimens in this community. Effectively, the heavy overlapping patterns of androgens ingested means that the abnormal urine steroid profile provides some nonspecific verification of androgen abuse but not for specific androgens. This information differs from the objectives of standard antidoping tests, which, aligned with the Athlete’s Biological Passport (24), aims to detect with high precision the use of individual prohibited substances.

In clinical management of cases where a history of possibly undeclared androgen abuse is an important differential diagnosis, the major significance of a drug history is the utility of an NPV. For the present study, we pooled those denying current androgen use, creating a mixture of 31 past users and 21 nonusers, a composition that would reflect a common scenario of evaluating cases of potentially undisclosed androgen abuse without necessarily being universally applicable when a substantially different balance between these groupings would change the predictive values. The present findings suggest a negative drug history has relatively high reliability in this grouping. If further verification of a negative history of androgen abuse is sought, an additional simple, useful, and sensitive objective criterion is measurement of serum LH, which displays nearly 90% correct classification alone.

Strengths of this study include the complete data including drug history declarations and laboratory analyses, specifically urine steroid profiling by a WADA-accredited antidoping laboratory using its state-of-the-art androgen screening methods. Although these methods represent the most sensitive and specific tests available, it is theoretically possible that some androgen use could be missed among current users due to their short half-life of urinary excretion. Conversely, some positives among past users may represent a minority of untruthful declarations and/or the very long tail of detection months after last use. The latter corresponds to the relatively new problem arising from the high sensitivity of antidoping tests differentiating passive, accidental, or unintentional absorption of prohibited substances through unknown contamination of ingested materials (including food, drink, nutritional supplements, and drugs) from intentional use (25).

Limitations of the study include the relatively small sample size, which remains among the largest groups studied in comparable intensive individual details including obtaining blood, urine, and semen sampling, any of which deter participation. Extrapolating the present findings on predictive value of declarations must be cautious as when these data are applied to other populations, the predictive values may vary according to the fidelity of the declarations and the prevalence of androgen abuse in that population.