Response to Letter to the Editor: "Rate and Extent of Recovery from Reproductive and Cardiac Dysfunction Due to Androgen Abuse in Men"

Abstract

We were surprised to read this comment on our paper ( 1) from a group that produced an important study ( 2). Having learned about their study during its publication, it prompted ours with its new quantitative approach to estimating the time to recovery from androgen abuse. We are pleased to now clear up any misunderstandings or misinterpretations. Our statistical methods were fully stated. Firstly, each standard deviation (SD) can be calculated from the tabulated SEMs and numbers using the formula SD = SEM x √ (n-1). Secondly, the covariates tested were all described under the “Results” section and comprised age, height, weight, body mass index, body surface area total duration of abuse, time since cessation, usage pattern, “postcycle therapy,” acne, gynecomastia, and temporal hair loss. As stated, in the covariance analyses only did the total duration of abuse and time since cessation have any significant impact on rate of recovery. Thirdly, our data analysis initially follows a global analysis approach, with prespecified linear contrasts. This avoids arbitrary, post hoc testing of selected findings, a process that violates nominal significance levels and introduces multiple, ad hoc comparisons. When justified, we normalize data by power transforms optimized via Box-Cox analysis ( 3), as we did for sperm variables using a cube-root transformation. We prefer that approach over nonparametric testing, which sacrifices power as well as versatility for valid post hoc testing, including prespecified linear contrasts. However, for serum testosterone concentrations in this study, it would uniquely be a mistake to try homogenizing the 3 groups into a unimodal population by any transformation. Current users taking exogenous testosterone distort the measured testosterone concentrations, which creates an inherently bi- or multimodal distribution. Whether analyzed globally in the natural scale parametrically (Table 4) or using a lower powered nonparametric alternative, serum testosterone was significantly higher among current users than among past users and nonusers, with latter groups not differing statistically. More importantly, the mean serum testosterone concentrations in past users and nonusers are in the upper half of the reference ranges established in Australian men using the same liquid chromatography-mass spectrometry method ( 4, 5) as well as in aggregated international data ( 6, 7). Interpreting these data must incorporate the additional facts that past users and nonuser groups had similar serum LH and FSH with lower SHBG in both our study and Rasmussen’s. This indicates that the past users are eugonadal with a lower SHBG, the latter creating a lower serum testosterone as a numerical artefact rather than implying androgen deficiency. As usual, viewing serum testosterone in artificial isolation of other reproductive hormones is misleading, just as it is in obesity. The persistence of lower SHBG might reflect either a long-term effect of androgen abuse in the liver, the source of SHBG secretion, or possibly ongoing but undisclosed androgen intake. The latter point is an important limitation of both studies, as neither study monitored the urine of past users for undisclosed androgen intake, a key issue for future studies of past androgen abusers.