Introduction

Adult-onset testosterone deficiency (TD) in men is a syndrome classified by a combination of low serum testosterone and symptoms attributable to the reduced levels of the hormone, following exclusion of primary hypogonadism and hypothalamic-pituitary-gonadal axis pathology [1]. Prevalence of this syndrome appears to range between 0.6–12% in men > 50 years of age [1, 2] with this figure greater (40–70%) in men with type 2 diabetes mellitus (T2DM) [3]. The clinical phenomena of adult-onset TD can be neatly categorised into cardiovascular, metabolic, psychological, physical, and sexual as shown in Figure 1, and can affect quality of life (QoL) [1, 4]. Many studies including the T-trials have shown that following testosterone therapy (TTh) some of the above-mentioned symptoms improved with restoration of serum hormone levels [5–7].

Display full size

Figure 1. 

Some clinical features associated with low serum testosterone

The ageing male symptom scale (AMSS), a validated internationally used scoring system, designed as a self-administered questionnaire has been demonstrated to be useful in evaluating symptoms (somatic, psychological, and sexual as shown in Figure 1) associated with health-related QoL of ageing in males as well as monitoring the changes following TTh [8–11].

Elliott et al. [12] in 2017 assessed 23 randomised controlled trials (RCTs) with 3,090 men, including the BLAST (Burntwood, Lichfield, Atherstone, Sutton Coldfield, and Tamworth) study, and demonstrated that TTh was associated with improved QoL, although this effect appeared weakened with coexistent moderate/severe depression [13, 14]. Behre et al. [15] in 2012 carried out a multicentre RCT assessing efficacy regarding QoL associated with TTh (1% testosterone gel) in hypogonadal men aged 50–80 years. Inclusion criteria included serum total testosterone (TT) < 15 nmol/L, bioavailable testosterone < 6.68 nmol/L, and an AMSS value > 36. QoL was evaluated via changes (∆) in AMSS (∆AMSS); total AMSS and the somatic, psychological, and somatic subscales [15]. Baseline AMSS values in the TTh and placebo (P) arms were 47.5 ± 0.7 (183 men) and 47.1 ± 0.6 (179 men), respectively. Interestingly, although AMSS improved in both trial arms, the TTh arm demonstrated greater improvement (p = 0.002); mean ± standard deviation (SD) ∆AMSS in TTh and P arms were –10.54 ± 0.7 and –7.19 ± 0.78, respectively. TTh was associated with significant or near significant improvement in ∆AMSS (somatic: p = 0.053, psychological: p = 0.049, and sexual: p < 0.001) compared with P. Mean subscale ∆AMSS ± SD values in the TTh arm were –4.30 ± 0.33 (somatic), –2.89 ± 0.26 (psychological), and –3.37 ± 0.31 (sexual) and P arm were –3.35 ± 0.36 (somatic), –2.12 ± 0.29 (psychological), and –1.69 ± 0.27 (sexual) [15]. Lee et al. [16] in 2013 conducted a single centre cross-sectional study of 176 men in a psychiatric outpatient setting and showed significant correlations between AMSS (total and subscales) and depression/anxiety using the Hospital Anxiety and Depression Scale (HADS) questionnaire. The HADS was developed in 1983 to estimate anxiety [HADS-anxiety (HADS-A)] and depression [HADS-depression (HADS-D)], which often coexist [14]. In both, HADS-A and HADS-D, scores of 8–10, 11–15, and 16–22 indicate mild, moderate, and severe anxiety and depression, respectively [14]. Diagnostic thresholds demonstrated that a score of ≥ 8 using HADS-A and HADS-D had a specificity of 0.78 and a sensitivity of 0.9 for anxiety, and specificity of 0.79 and a sensitivity of 0.83 for depression, respectively [14, 17].

Many studies have shown TTh to be associated with significant benefits in depression when compared with P [18–21]. However, cohort characteristics differed between these studies. A meta-analysis of 27 RCTs (1,890 men) by Walther et al. [22] in 2019 showed that TTh compared to P was associated with a significant decrease in symptoms of depression, with the association especially evident at higher doses of TTh. They speculated that the efficacy was comparable to that of a moderate antidepressant, although acknowledging significant between study heterogeneity [22].

A recent meta-analysis of 17 trials with participant data by Hudson et al. [23] in 2023 including the BLAST study evaluated benefits associated with TTh in subgroups of men with low serum testosterone levels. Although AMSS values improved, this was not significant in men with T2DM, and it was speculated that lower QoL in men with T2DM may have demonstrated a blunted response to TTh [23]. The mean ± SD of AMSS values in the 482 men on TTh and 456 men on P were 32.19 ± 10.23 and 34.22 ± 11.10, respectively.

Materials and methods

Results

The baseline data of the total cohort and men of TU and P are shown in Table 1. No significant inter-group differences (rank-sum) were evident regarding age, serum TT, SHBG, cFT, and AMSS values. Not surprisingly serum TT, cFT concentrations at study end (30 weeks) were significantly higher than baseline in men treated with TU whilst serum SHBG levels decreased, however no significant changes were observed in the men on P. Table 1 also demonstrates that AMSS values improved significantly (sign-rank) in both groups; median [interquartile range (IQR)] ∆AMSS in men on TU and P were –6 (–10, 0), p < 0.0001 and –2 (–7, –3), p = 0.010, respectively. The decrease in AMSS values in men on TU was significantly greater (p = 0.014, rank-sum) than the men on P. TU did not lead to significant improvement in AMSS after 6-weeks (p = 0.52, rank-sum) and 18-weeks (p = 0.077, rank-sum) compared to P. Table 1 also shows significant improvements in median (IQR) HADS-D and HADS-A values only in men on TU, –1 (–3, 0), p = 0.0083 and –1 (–3, 0), p = 0.0010, respectively. Significant intergroup differences regarding ∆HADS-D (p = 0.072) and ∆HADS-A (p = 0.072) were not evident.

Table 2 shows AMSS data (baseline, 30-week, and ∆AMSS) in the men for whom the subscale AMSS was recalculated from the questionnaires stored in secondary care. Intra-group statistics (sign-rank) showed that somatic, psychological, and sexual subscale AMSS values improved in the men on TU whilst only somatic subscale AMSS values improved in the men randomised to P as evident in Table 2.

Separate linear regression analyses showed that ∆AMSS (dependent variable) was inversely associated with baseline AMSS values (independent variable) in men on TU [coefficient: –0.28, 95% confidence intervals (CI): –0.47, –0.079, p = 0.007] and P (coefficient: –0.20, 95% CI: –0.33, –0.079, p = 0.002). Thus, improvement in AMSS was greater in men with higher baseline values in both arms of the RCT. To further investigate the above-mentioned association between baseline and ∆AMSS, the cohort was stratified by baseline values as seen in Table 3; < 27 (no significant symptoms consistent with a low testosterone level), 27–36 (mild symptoms consistent with a low testosterone level), 37–49 (moderate symptoms consistent with a low testosterone level), and > 49 (severe symptoms consistent with a low testosterone level). There was no difference between the TU and P groups regarding the distribution of men stratified by baseline AMSS (p = 0.48, chi-square). Table 3 shows that no significant change in AMSS was evident in men with baseline AMSS values < 27 in both TU and P study arms. TU was associated with improvement in AMSS in the other 3 groups stratified by baseline AMSS (men with mild, moderate, and severe symptoms associated with low serum TT concentrations). P was only associated with improvements in men with severe symptoms associated with low serum TT concentrations (AMSS > 49).

Table 4 shows AMSS values (baseline, 30-week, and change) stratified (< 8 and ≥ 8) by HADS-D and HADS-A values as well as the association between TU (P: reference) and ∆AMSS (dependent variable) with the models adjusted for baseline AMSS values (Table 5). TU was significantly associated with ∆AMSS (model 1) in the total study cohort and this was independent of baseline AMSS (model 3); baseline AMSS was also seen to be associated with ∆AMSS (model 2). Importantly, baseline serum TT and ∆serum TT [the regression model also included TU (P: reference) and baseline AMSS as independent variables, both these remained associated] were not significantly associated with ∆AMSS. Baseline AMSS values were significantly higher in men with baseline HADS-D (p < 0.0001, rank-sum) and HADS-A (p < 0.0001, rank-sum) ≥ 8. Models 4–7 show the association between ∆AMSS and TU (P: reference) with the analyses adjusted for baseline AMSS in the patient cohort stratified by HADS-D and HADS-A values (< 8 and ≥ 8). Interestingly TU was associated with improvement in AMSS in men with HADS-D ≥ 8 and HADS-A < 8. Thus, the benefit in AMSS following TU (compared to P) appeared to be restricted to men with mild-severe symptoms (≥ 8) of depression (model 4) and no symptoms (< 8) of anxiety (model 7) on HADS-D and HADS-A values.

Table 5 (models 8, 9, and footnote) shows the cohort stratified by HADS-D of 11; it was evident that TU was significantly associated with ∆AMSS in the men with mild depression (HADS-D: 8–11) with no such association in men with HADS-D: 11–21 (moderate and severe depression) observed (this group was not further stratified as there were only 7 men with severe depression (HADS-D: 15–21).

Discussion

In response to the recent publication by Hudson et al. [23], we re-analysed data from the BLAST study in men with T2DM and adult-onset TD to establish ∆AMSS in the total cohort and ∆AMSS in the subscales following TTh with TU compared to P. This was followed by evaluating whether TU (P: reference) and baseline AMSS were associated with ∆AMSS in the total cohort and when stratified by HADS-D and HADS-A.

Interestingly AMSS improved significantly in both the TU and P arms, although the improvement was significantly greater in the men on TU (Table 1). The previous report by Hackett et al. [13] did not show significant differences between the TU and P arms (p = 0.092, ANCOVA test) in the original publication. A repeat of the analysis (ANCOVA) on the amended database showed that TU was associated with greater improvement in AMSS (p = 0.024). This discrepancy could have been due to reclassification of the database carried out in 2015. Similarly, we showed improvements in HADS-D and HADS-A values in men on TU, although there was no significant intergroup difference (the original report [13] only reported intergroup differences). Analyses with the restricted cohort with subscale data showed that whilst AMSS decreased with TU in all subscales, reduction in AMSS following P was only observed in the somatic AMSS subscale. We can only speculate that the improvement of somatic subscale AMSS could be related to either the lifestyle advice provided at study initiation or added motivation of being selected for the study. Baseline AMSS values also appeared to predict ∆AMSS in both men on TU and P. Further, it also appears in our cohort of men that depression and anxiety as measured by HADS-D and HADS-A had an impact on ∆AMSS; interestingly this benefit was only evident in men reporting symptoms of mild depression and those not reporting symptoms of anxiety.

In the original report of the BLAST study [13] stratifying the cohort by HADS-D value of 11 (HADS-D ≥ 11 has been considered clinically significant depression—moderate/severe depression) suggested that AMSS only improved in men with HADS-D value < 11. The results from models 4 and 5 in Table 5 appear to be at odds with the previous finding [13]. This extended analysis by selecting different HADS-D thresholds refined this finding and established that the benefit was evident only in men with HADS-D of 8–10 (mild depression), more details are found in the footnote of Table 5. In the original analysis, the stratifying categories included mild depression/no depression and moderate/severe depression [13]. Thus, the category where mild depression was placed led to seemingly different outcomes in ∆AMSS. As lower HADS-D values were associated with lower baseline AMSS values [due to the inverse association between baseline AMSS and ∆AMSS (Table 5, model 2)] we would expect a more modest improvement in men with lower HADS-D values; this was evident (Table 5, model 5). This argument also suggests greater improvement in AMSS values in men with moderate/severe depression (Table 5, model 9); this was not evident, perhaps due to TTh over 30 weeks being insufficient.

Our findings of improvement in AMSS are only of clinical value if it is a useful marker of QoL. Lunenfeld et al. [7] in 2013 recommended that despite low specificity, AMSS was possibly useful in monitoring clinical response following TTh. Heinemann et al. [8] in 2003 in an open post-marketing study of 1,670 men with low serum TT provided testosterone gel analysed AMSS before and 3 months after TTh. Significant improvements in AMSS were observed in the summation of AMSS values as well as the individual subscale values. It also appeared that higher baseline AMSS values were associated with greater improvement: 9%, 24%, 32%, and 39% in men with absent, mild, moderate, and severe symptoms at baseline, respectively [8]. This is in agreement with our findings and the Wilder principle [24]. The RCT by Behre et al. [15] in 2013 was confined to men with baseline AMSS > 36 and like our study showed improvements in AMSS values in both study arms with the TTh arm demonstrating greater benefit. Our findings do not fit with the speculation reached by Hudson et al. [23] that the higher baseline AMSS values in men with diabetes may have led to a non-significant ∆AMSS.

We have previously hinted at the presence of heterogeneity in adult-onset TD and this could lead to variation in response following TTh [30]. Thus, it is essential that the cohort characteristics and baseline presentation patterns are heeded. In our study, it is evident that the baseline AMSS value and scale of depression/anxiety altered the outcomes. Thus, longer RCTs with larger cohorts are required to tease out potential interactions that will exist in chronic pathological states such as adult-onset TD. The classification by including serum testosterone status and associated symptoms (as evident in Figure 1) ensures heterogeneity.

Our results have shown that QoL measured using AMSS has improved in men with T2DM. In view of the high prevalence of adult-onset TD in men with diabetes, it is tempting to speculate that even a modest improvement in QoL could lead to improved outcomes, including treatment adherence. Associations have been demonstrated between QoL measures and treatment compliance in chronic conditions such as hypertension and diabetes [31, 32]. Uchmanowicz et al. [31] in 2018 assessed QoL using the World Health Organization QoL Scale Brief version questionnaire and showed a positive correlation with antihypertensive treatment compliance. Some studies have shown a similar correlation between QoL and therapeutic compliance in patients with diabetes. Of the 6 studies considered in a systematic review by Gusmai et al. [32] in 2015, 4 studies suggested that better QoL improved treatment adherence (3 of them used biochemical outcome markers such as HbA1c, whilst the remaining study used information gathered from questionnaires as the outcome).

We also demonstrated that TU was associated with improvement in both depression and anxiety (Table 1), evaluated via HADS. Korenman et al. [33] in 2018 reported a significant association between depression and eugonadotrophic hypogonadism (serum TT < 10.4 nmol/L) in men (compared with matched population controls) aged between 18 and 40 years. A review by Indirli et al. [34] in 2023 hinted at a bidirectional association between hypogonadism and depression. They highlighted that the impact of TTh on depression and the reciprocal association between antidepressants on hypogonadism have not been clarified [34]. Thus, our findings are of potential importance. Interestingly (via multivariate regression models) our data also shows that psychological subscale values were associated with the somatic subscale values (p < 0.001), but not with the sexual subscale values (p = 0.24). Thus, to validate our observations and to unravel the intricate associations between depression, anxiety, AMSS subscales, and TTh, an RCT with AMSS and HADS as primary endpoints is required. In the event of our findings being validated, it would be interesting to create a further study to see if improvement in depression and anxiety via prior interventions by psychologists and psychiatrists leads to better AMSS outcomes following TTh. If this was proven a case for a multifaceted approach including psychologists/psychiatrists in the management of men with adult-onset TD is strengthened. Adult-onset TD is defined as a combination of low serum testosterone levels and associated symptoms [1]; hence, it is unlikely that men with AMSS values < 27 would be offered TTh based on symptoms evaluated via the AMSS questionnaire. It would be interesting whether further research would refine this approach by suggesting using AMSS subscale values (as an association was not established between psychological and sexual subscales).