Prediction model for anabolic androgenic steroid positivity in forensic autopsy cases – a new tool to the autopsy room

Since the introduction of anabolic androgenic steroids (AAS) to the athlete community, the illicit use of these substances has been a perennial problem in society [1]. Every now and then, non-prescription AAS use is also encountered in the setting of forensic cause-of-death investigation. This is not surprising, since several factors have been identified that may increase the risk of sudden or premature death in connection with supraphysiological dosing these substances. These include pathological cardiac left ventricular and interventricular septum remodeling via hypertrophy and fibrosis [2‐5] with subsequent left ventricular dysfunction [6‐10]; post-exercise cardiac electrical instability [11]; hypertension [12]; acceleration of coronary plaque formation [9]; and hypercoagulability, predisposing to thrombotic events [13]. Other potentially fatal organ manifestations include acute cholestasis, peliosis hepatis, hepatic tumors [14]; and acute or chronic kidney injury [15]. Some individuals are prone to severe concomitant substance abuse [16‐18], and may exhibit aggressive or suicidal behavior, possibly mediated by antisocial personality or other behavioral disorders [19‐21]. Indeed, non-therapeutic AAS use seems to be an independent risk factor for premature death [22].

Based on previous forensic autopsy studies, the stereotype of a deceased AAS using person is a young male with overdeveloped musculature, pronounced cardiac hypertrophy, and polydrug use [3, 23‐26]. This description correlates with the global epidemiological estimates, indicating that the most typical user group is young to middle-aged males with athletic or personal appearance objectives [27]. On the grounds of these previous publications, Esposito and colleagues recently suggested an investigative protocol for forensic professionals in cases of suspected AAS-induced sudden deaths of young athletes [28]. However, there are multiple other, non-athletic populational subgroups using these substances [29]. Also, AAS use is often concealed from others, including medical professionals [30]. This may result in false negative AAS use anamnesis. The question is, do we see the whole picture?

The decision to request AAS analysis in an autopsy case often rests on the shoulders of an individual forensic pathologist. To ensure the correct targeting of these analyses, more sophisticated diagnostic methods are needed alongside the traditional approach, relying strongly on positive AAS use history. In this study, we aimed to develop further tools to manage this issue by testing, whether a multivariable model could predict the probability of AAS assay positivity in forensic autopsy cases. The model was based on a set of circumferential and macroscopic autopsy-derived variables, consistent with typical baseline data available for the forensic pathologist at the end of the macroscopic autopsy. The performance of the multivariable model was compared to the conventional method with lifetime documentation of AAS use as the only predictor of AAS positivity. The fundamental aim was to convert the multivariable model into a simple calculator to be applied to practical work in the autopsy room.

Previous studies have noted the importance of AAS analysis in connection with forensic cause-of-death investigations, especially in cases of sudden, cardiac death [25, 26, 28]. However, AAS-related deaths are not always due to cardiac causes [23, 24]. Currently, prompt international and national guidelines on AAS testing in various scenarios of cause-of-death investigation are lacking. This may cause inconsistency between individual forensic pathologists, lead to low number of requested analyses and considerable underdiagnosis, and in long term, impair scientific research in this field. For example, altogether 33 643 forensic autopsies were performed in Finland between 2016 and 2019 [31], but AAS assay was conducted in only 46 cases (0.1%). Of these, 34.8% turned out to be positive. Considering the global and national estimates on AAS use prevalence in modern society [27, 29, 41], the actual number of Finns using these hormones and undergoing forensic autopsy is most likely significantly higher.

The present study was designed to determine, whether a multivariable regression model could predict AAS assay positivity in forensic autopsy cases. The second objective was to evaluate the predictive value of the multivariable model, by comparing it to the conventional model with lifetime AAS use documentation as the sole predictor of AAS positivity. The key variables were chosen from a set of circumferential, external and internal examination data, available to the forensic pathologist at the end of the macroscopic autopsy – before making the decision to proceed with further toxicological analysis. We found that a nine-variable model predicted AAS positivity markedly better than documentation of lifetime AAS use alone. The calculator based on the multivariable model showed excellent discriminative ability in both the main sample and the independent validation sample and exceeded the AUC scores of the conventional model.

In the main sample, crude between-group differences were observed in lifetime AAS use documentation; description of the body habitus; PMI; weights of the heart, liver and kidneys; left ventricular and interventricular septum thickness; and the frequency of coronary and extra-coronary atherosclerotic changes. In both groups, the median weights of the organs listed above were at or above the upper quartile of the normal distribution in general adult population [37]. To some extent, this may be explained by the observed high BMI of the sample [37, 42]. However, high BMI itself may reflect obesity and abnormal metabolic state, in which case the organomegaly may not be purely physiological. In the case of AAS use, the high BMI seems to reflect overt muscularity, and the extent to which traditional organ weight reference tables can be used for comparison, remains obscure.

Nevertheless, the high median heart weight observed in the AAS positive group clearly indicates pathologic adaptation [43]. Our findings support the previously noticed associations with AAS use and cardiac hypertrophy, and abnormal left ventricular and interventricular septal wall measurements [2‐4]. Advanced subclinical coronary atherosclerosis has also been described before, with cumulative duration of AAS use increasing the plaque volume [9]. Young AAS using males have been shown to have disturbed shear rate in brachial artery along with heightened high-sensitivity C-reactive protein (hs-CRP) [44], which has been reported to associate with increased susceptibility to coronary and extra-coronary atherosclerosis [45]. However, comparative autopsy studies regarding the severity of atherosclerosis in different AAS using age groups are lacking.

Higher liver and kidney weight among the AAS positive warrants further investigations. This finding could be related to congestion, considering the rich vascular supply of these organs. Indeed, visceral congestion has been shown to increase renal weight, yet liver weight seems to be affected only by blood loss [42]. Congestion of these organs has previously been reported in case studies of deceased AAS using individuals, most likely reflecting the terminal stages associated with the cause and manner of death [3, 46]. However, a case report by Luke et al. described actual renal hypertrophy in a deceased AAS using weightlifter [47]. Experimental data suggests that several mechanisms may trigger renal growth, and this hypertrophy may be either physiological or pathological, as in diabetic nephropathy [48]. Some of the maladaptive pathways are stimulated by testosterone [15], the end of the route being fibrosis and renal atrophy. It should also be recognized that organ hypertrophy in connection with AAS use may be caused by other concomitantly used drugs, such as insulin-like growth factor (IGF) or growth hormone (GH) [15, 49, 50]. Due to their structural similarity to endogenous hormones, these drugs are not readily available in post-mortem toxicological screens.

The observed PMI medians in the main and validation samples were close to the average PMI of all Finnish forensic autopsies during the study period (8.5 days; personal communication with Aki Eklin, Senior Planning Officer at the Finnish Institute for Health and Welfare), which is mainly due to prolonged cold storage of the cadavers before the autopsy. The difference in PMI between the AAS positive and negative groups in the main sample is unlikely related to AAS use per se. One must consider the possibility of post-mortem drug degradation due to bacterial activity, which could cause true AAS positive cases to turn negative as the PMI increases. For this reason, PMI was initially considered as one explanatory variable in the multivariable model, but there was no additional benefit for model performance. This does not necessarily mean the issue is irrelevant, but a further evaluation of this perspective would require a larger study sample. Meanwhile, there is no reason to refrain from urinary AAS analysis based solely on prolonged PMI, since exogenous anabolic steroids and their metabolites are often found in forensic post-mortem urine samples in large quantities [51], and their overall stability in biological fluid samples is good [52].

In the validation sample, the calculator scores were mainly below 0.4 for the AAS negative, while there was more dispersion among the AAS positive. Some of this dispersion may be explained by the duration of AAS use and its timing related to death, since AAS are often used in cycles with breaks of variable length in between. It seems unlikely that a single cycle of AAS use would result in significant atherosclerosis or permanent change in organ size; yet it would likely be detected in the urine screen, if timed within a few weeks or a couple of months prior to death [53]. Some of the observed dispersion may also be explained by concomitant drug use or somatic disease. Further work and larger samples are thus needed to calibrate the calculator and to determine its applicability in cases with congenital and systemic disease.

Taking these limitations regarding implementation into account, the calculator still clearly exceeds the value of lifetime AAS use documentation when deciding on possible AAS analysis and can be recommended as an additional tool to the autopsy room. It is potentially beneficial in all forensic cases with no known somatic etiology for organomegaly, and especially in cases with obscure or lacking lifetime AAS use anamnesis. It may be also helpful in determining whether testing for AAS or for genetic disease would be relevant in cases with cardiac hypertrophy – the latter also concerning the next of kin. A high calculator score should prompt testing while the significance of a lower score is to be proportioned to the circumstances and possible juridical aspects of each death. In cases with high calculator score but negative urine assay, additional body hair analysis should be considered if available, as it may offer longer detection time (up to several months) than urine for previous AAS use [54].

We recognize the limitations of our study. First, this was a retrospective study and several forensic pathologists have performed the autopsies. The Finnish forensic autopsy reports are semi-structured and allow freedom of choice regarding certain practical matters. For example, the hearts may be weighed as a whole, after washing out all blood and blood clots; or after transecting the organ. Similarly, the liver may be weighed with or without gallbladder. As a result, there may be minor differences in organ measurements between operating pathologists, but this error is unlikely systematic in the study sample. Also, due to the retrospective nature of the study, only systematically documented organ measurements were available as potential variables in the model. These did not include all possibly relevant measurements, such as testicular size, which may decrease during active AAS use [55]. Second, qualitative variables of subjective origin were included in the predictors. Naturally, classifying the body habitus as “muscular” without objective measurements of the body composition, is prone to bias. Both the AAS positive and negative group demonstrated above normal median BMI, and despite the trend of the AAS positive having less abdominal subcutaneous fat, there were no statistically significant differences between the groups in either of these variables. Yet, the AAS positive individuals were significantly more often described as having a muscular body habitus vs. something else. This indicates that subtle differences between normal and abnormal muscularity have been observed by the forensic pathologists. Furthermore, there are currently no other easily implementable, quantitative measurements available in the autopsy room to replace this variable. Third, the study sample was rather small. This carries the risk of model overfitting. To reduce this risk, a penalized regression method was chosen. Last, the study sample included only males, so the results may not be applicable to females with AAS use. It should also be emphasized that our model is based on national samples, which may reduce the applicability of the calculator to other forensic autopsy environments.

This study introduced a prediction model for AAS urine assay positivity in forensic autopsy cases. The multivariable model performance was excellent in both the main and the validation sample. We believe that introducing the model based calculator to the autopsy room will increase knowledge on the issue and hopefully, encourage AAS testing especially in cases with obscure lifetime AAS use history. The calculator could also serve as a basis for developing a diagnostic tool for living patients suspected to have used AAS, as new imaging methods enabling more accurate organ size estimation emerge.