The impact of long‐term androgen deprivation therapy on cognitive function and socioeconomic decision making in prostate cancer patients

Androgen deprivation therapy (ADT) enhances survival of advanced prostate cancer patients and is therefore used as a concomitant therapy. However, ADT has been reported to cause negative side effects on cognition and emotional processing. So far, research referred to the effects of short‐term treatment. Since the brain may adapt to androgen deprivation, we were especially interested in the long‐term effects of ADT on cognitive and socioeconomic decision making.


| BACKGROUND
Androgen deprivation therapy (ADT) is a common method to treat aggressive testosterone-dependent prostate cancer in older men.
Pharmacological androgen deprivation can be induced by gonadotropin releasing hormone (GnRH) analogues that reduce the biosynthesis of testosterone through direct modulation of the hypothalamic-pituitary-gonadal axis, or, in contrast, nonsteroidal antiandrogens locally block the androgen receptor binding site and thus indirectly influence testosterone production by negative feedback. 1 ADT has been demonstrated to improve survival of prostate cancer patients and is therefore used as a common concomitant therapy for advanced prostate cancer patients. 2,3 However, several negative side effects on cognition, physical capability and socioemotional processing have been reported. For instance, testosterone deficiency in men was found to impair life quality, 4 memory, 5 and visuospatial performance, 6 as well as physical abilities such as walking. 7 The ADT-induced testosterone deficit further promotes depression 8 and emotional distress 9 and decreased sex drive up to a loss of libido and erectile dysfunction. 10 Although natural aging in men also leads to a decline of endogenous testosterone levels, this decline does not occur as abruptly as the chemical castration induced by ADT. Instead, it rather incorporates a constant and slow reduction over several years until old age, when the so called andropause is reached. Andropause can also be accompanied by some of the negative side effects shown for ADT.
These include mental impairments like cognitive decline, but also physical disabilities like erectile dysfunction, whereby the link to testosterone deficiency has been demonstrated (for an overview see Ref- erence [11]).
Cancer diagnosis as well is linked to cognitive decline and increased risk for depression. 12,13 Especially chemotherapy seems to provoke reversible cognitive deficits in cancer patients. These neuronal impairments were associated with a decline in estrogen and testosterone, supposedly further contributing to the decline in cognitive function (for review see Reference [14]).
Here, we measured the free testosterone concentrations of 46 prostate cancer patients, who had not been treated with chemotherapy, and 22 healthy controls. Twenty-four of the cancer patients received long-term ADT, whereas the remaining 22 patients did not receive any testosterone-effective treatment. In this sample, we wanted to evaluate the effects of long-term ADT on visuospatial cognition with reference to previous findings from studies assessing short-term ADT-treatment (3-9 months) that had demonstrated a consistent decline in visuospatial cognition in men. 6 Apart from that, we were especially interested in the effects of long-term chemical castration on two aspects of socioeconomic decision making that have previously been demonstrated as being highly testosterone sensitive, that is, fairness perception and intergroup biases in socioeconomic decision making. 15 Based on a rising number of findings that not only aging-related processes but also the decreasing habitual testosterone concentration may cause cognitive deficits in men (eg, References [16,17], for meta-analysis see Reference [6]), we hypothesized to find differences in visuospatial performance between the groups, with a marked visuospatial deficit in ADT patients. Furthermore, cancer patients without androgen suppression were expected to have lower self-reported depression scores, 18 (but see Reference [3]), a higher quality of life 19 and a better performance in the Mini Mental State examination (MMSE) than patients who receive long-term ADT (for meta-analysis see Reference [6]). Apart from that, another aim of this study was also to investigate, for the first time, whether testosterone deficiency may impair one aspect of social behavior that has repeatedly been associated with normal testosterone function. Previous research from our group agreed that higher endogenous testosterone may be associated with an intergroup bias in socioeconomic decision making, which may promote outgroup hostility, ingroup favoritism, or a combination of both. 15,20,21 Based on these findings, we hypothesized that untreated prostate cancer patients would show a stronger intergroup bias in a socioeconomic exchange task when being compared with long-term androgen-deprived patients. This hypothesis rests on the assumption that the brain might adapt to the long-term suppression of testosterone concentrations in patients with ADT, which may for instance result in a reduction of the androgen receptor density thus reducing overall androgen sensitivity in the brain (eg, in aging rats 22 ; but see Reference [23]). To our knowledge, this is the first study that assessed the influence of ADT on socioeconomic decision making in the UG and its influence on the intergroup bias.
Patient groups and the control group were matched for age, which may correlate with natural decline of testosterone 11

| Participants
In total, 68 men between 50 and 79 years were included in the study (mean age ± SD = 67 ± 7 years). Each subject gave written informed consent and ethical approval was obtained from the ethics committee of the Ärztekammer Hamburg (approval number: PV3948). Forty-six patients had a prostate cancer diagnose and were recruited from the Martini-Klinik at the University Hospital Hamburg-Eppendorf in Hamburg (Germany). Twenty-four of these patients received an ADT (HI group). Of these, 15 men received GnRH analogues (GA) and nine men took nonsteroidal antiandrogens (AA). We only included patients in the GA group that were treated for at least 5 months. The minimal intake period of AAs was 15 months. Another 22 patients did not receive antihormone treatment and thus participated as controls in the prostate cancer group (PCC group). In addition, 22 healthy men were recruited through advertisements (healthy control group; HC group). Groups were matched for age and education level and median time from prostate cancer diagnosis (33 months, range = 3-262) did not differ between the HI and PCC patients. The "Beck Depression Inventory II" (BDI) was used to evaluate depression severity, if any, in patients and controls. 24 Quality of life was examined with the "Nürnberger Life Quality" questionnaire (NLQ). 25 Further empathy score, 26 digit ratio 27 and trait impulsiveness 28 were used as matching parameters. For details see online supplemental materials.

| Free testosterone measuring procedure
The free, bioactive testosterone concentration was determined from saliva. The participants were instructed to collect three saliva samples in Eppendorf tubes (2 mL) at the morning of the test day. The samples were frozen at -20 C until being analyzed with a "testosterone free in saliva ELISA" from Demeditec Diagnostics GmbH (Kiel, Germany). For details see online supplemental materials.

| Visuospatial performance and cognition
The block-design test from the Wechsler Adult Intelligence Scale IV (WAIS IV) was used to measure visuospatial performance. 29 The block-design test is a core test to measure cognitive abilities like logical thinking and eye-hand coordination. 30 The test was carried out according to the instructions given in the manual. 30 The MMSE was further used to screen participants for cognitive impairment. 31 See online supplemental materials for details.

| Socioeconomic decision making
The ultimatum game (UG) is an established socioeconomic decisionmaking experiment. 32 Two players represent the proposer and the responder in a socioeconomic exchange game. The proposer offers a selected amount of points out of 10 points. The responder has the choice to accept or reject the offer. If the responder accepts the offer, both players will get the respective amount of points. If the responder rejects the offer, both will get zero points. Here we used an approved computer-based intergroup version of the UG 21 to compare the behavior between the two patient groups. In this game, men with high testosterone levels were previously found to show either higher ingroup cooperation or increased outgroup hostility or both, especially during an intergroup competition 15 . The intergroup UG was thereby specifically adapted to enable the comparison of the two patient groups. As an intergroup factor, we presented a pseudorandomized sequence of offers from fictitious prostate cancer patients of two different hospitals, one foreign hospital and the same hospital in which patients were treated for prostate cancer. Please see Figure 1 for a schematic description of an experimental trial. A more detailed description of the offers and the group rewards can be found in the online supplemental materials.

| Data analysis
IBM SPSS statistics 25 was used to analyze the data. We used F-tests to investigate the influence of the test group on the behavioral variables and the appropriate post hoc t-tests or correlations for further analysis. Statistical effects were considered significant at P < .05 (twotailed), if not otherwise indicated. Post hoc t-tests were used to further examine the group differences. P-values for post hoc tests were Bonferroni corrected, if necessary (see online supplemental materials).
The effects of the test group on matching parameters, life quality and depression state were analyzed with univariate ANOVAs and post hoc t-tests.
For analysis of the block-design test and MMSE scores, we used univariate ANCOVAs to investigate the influence of the test group (HI, PCC and HC) on the visuospatial performance and cognition. Ztestosterone was included as covariate. Post hoc Pearson correlations were used to analyze the correlations between Z-testosterone and the visuospatial performance.
With a 2 × 2 × 2 ANCOVA, we further examined how the treatment group (HI group or PCC group), the intergroup bias (ingroup or outgroup) and the experimental context (neutral or group competition) affected the overall rejection rates for unfair offers. Z-testosterone was included as covariate. An unpaired t-test was used to examine the interaction between the test group and the intergroup bias. For this purpose, the rejection rates for unfair proposals from outgroup vs ingroup members were calculated (Δ rejection rates = unfair offers of outgroup -rejection rates unfair offers of ingroup). A higher Δ rejection rate indicates a stronger intergroup bias. Thirteen PCC patients had a Gleason score of 3 + 4, five had a score of 4 + 3, one had a score of 3 + 3 and one had a score of 4 + 5 at the time of testing. For two PCC patients the Gleason score was not specified. In the HI group, four patients had a Gleason score of 3 + 4 (three of the GA group and one of the AA group), six had a score of 4 + 3 (four GA, two AA), three had a score of 4 + 5 (two GA, one AA), one had a score of 5 + 3 (GA), five had a score of 5 + 4 (two GA, three AA) and for five patients the Gleason score was not specified (three GA, two AA).

| Participant information
The staging (TNM-system) of prostate cancer was N0 in 13 patients of the PCC group, N1 in five patients of the PCC group and NX in two patients of the PCC group (the stage was not specified in two of the PCC patients). In the HI group four times N0 (one GA, three AA), 15 x N1 (11 GA, four AA) and five times "not specified" was documented.
Nine patients of the PCC group, 13 patients of the HI group and 13 healthy controls noted comorbidities/conditions (eg, diabetes or hypertension). For additional demographic information see Table 1.

| Matching parameters and descriptive values
We could not find any differences in empathy score, digit ratio and trait impulsiveness between the treatment and the two control groups. Depression severity and quality of life were affected by the treatment. Patients who received hormonal intervention showed higher depression scores than the PCC group and the HC group and a more compromised life quality than patients of the PCC group. For details, see online supplemental materials.

| Testosterone concentrations
Treatment significantly influenced the testosterone concentration (F 2, 65 = 13.28; P ≤ .001; η p 2 = 0.29). Post hoc t-tests showed that F I G U R E 1 Schematic description of a trial from the Ultimatum Game. The trial begins with a fixation cross, followed by the statement "New Game!" After that, the proposer makes his offer. The responder must than decide whether to accept or reject the offer. The responder is informed that he plays against real people that made their offers in a first part of the study

| Visuospatial performance and general cognition
We found a significant main effect of the Z-testosterone concentration on visual motor performance (F 1,62 = 4.29, P = .042, η p 2 = 0.07).
We further found a significant interaction between test group and Z-  Figure 2B for the group results.
The analysis of the MMSE further indicated that the "test group" was associated with differences in cognitive capacity of the participants (F 2,62 = 5.74; P = .005; η p 2 = 0.16). We could not find any
Most interestingly, we also noted an interaction between the test group and the intergroup bias (F 1,41 = 5.04; P = .030; η p 2 = 0.11). No other significant effects or interactions were detected (see online supplemental materials for detailed results of the ANCOVA).

| DISCUSSION
The present study investigated the influence of artificially decreased testosterone concentrations in elderly men on different aspects of emotional behaviors, cognition and life quality. In line with our predictions, we found small behavioral effects associated with the testosterone decline in patients with a long-term ADT. Treatment decreased testosterone concentrations in long-term ADT, with a significant difference between GA and AA treated patients. Similar to short-term treatment, long-term patients exhibited a reduced visuospatial performance in the block design test (for review see Reference [6]). This result F I G U R E 2 Correlation between groupwise standardized testosterone concentration (Z-testosterone) and visuospatial performance measured in the block-design test. A, In all participants. B, Separated by group (Pearson correlation, two-tailed; P-value is considered significant at P < .05) is in accordance with our hypothesis that decreasing habitual testosterone concentrations in men may cause cognitive deficits 16,17 . We also tested the influence of testosterone deprivation on behavior in a testosterone-associated socioeconomic decision-making task, the intergroup UG. We found a difference in rejection rates for unfair offers between the PCC and the HI group, in line with a higher intergroup bias in patients with nondeprived testosterone. Life quality and depression state were also negatively affected by ADT. These results are in concordance with the literature that reported impaired life quality and higher depression severity in patients receiving ADT 3,4,9,33 .

| The influence of hormonal intervention on the free testosterone levels
As expected, treatment significantly influenced free testosterone concentrations of the patients. Previous research corresponds with this finding, but in contrast to our study, most research measured serum testosterone, 16,34 but see Reference [35]. Measuring salivary testosterone offers the opportunity to detect the level of the free and bioactive testosterone. Since previous research on the modulatory effects of testosterone on behavior primarily focused on normal endogenous or exogenous testosterone concentrations, it was mandatory to investigate how the massive deprivation of testosterone, as experienced under ADT, was associated with behavioral changes.
Interestingly, the artificially depressed-free testosterone concentra-

| The influence of testosterone treatment on visuospatial performance and cognition
Classically, the block design test has been used to examine cognitive decline in aging populations 36 and its link to testosterone has also been examined. 37 In a large-scale meta-analysis from 14 original articles, patients with ADT showed significantly lower visuospatial performance ability (effect size, g = −0.67; P = .008), which led to the assumption, that testosterone deficiency may be responsible for this aspect of cognitive impairment. However, the meta-analysis did not consider the endogenous testosterone concentration. Therefore, a direct association between testosterone and visual spatial performance could only be assumed. 6 We found that the Z-testosterone and the interaction between test group and Z-testosterone were significantly associated with visuospatial performance. Across all groups, Z-testosterone was positively correlated with visuospatial performance, whereas, when considering the groups separately, the positive correlation was only found in HI patients and the HC group. We suppose that the small effects of treatment may be traced back to the range of testosterone within the groups  16,17 , and specifically so in the block design test (for review see Reference [38]).
Furthermore, the meta-analysis of McGinty et al 6 could demonstrate that visuospatial deficits were especially discovered in studies that examined patients with short-term treatment (65% up to 9 months), whereas these effects were assumed to disperse over time. 6 Here, we were able to supplement the findings from the meta-analysis by patients with long-term treatment (5-262 months), who nevertheless showed a persistent cognitive impairment.
F I G U R E 3 Direct comparison of the relative rejection rates for unfair offers of the outgroup in relation to the ingroup (Δ rejection rates) between the treatment groups. Bar graphs show the pooled mean value of both contexts (neutral and competition) and the line graphs show the two contexts separately (unpaired t-test; two-tailed; patients of the PCC group showed significantly higher mean relative rejection rates toward unfair offers than patients of the HI group, which were independent of context (t 42 = 2.267, P = .029, d = −3.

19)
A loss of cognitive functioning had previously not only been associated with increasing age, but also with the age-related loss of bioactive testosterone in men. 39 Nevertheless, the literature discussed the effects of free testosterone on cognitive capacity controversially. 16

| The influence of testosterone treatment on socioeconomic decision making
Socioeconomic decision making in the UG and related tasks has been associated with endogenous testosterone of healthy young men.
Especially in paradigms that highlighted the factor "group association," habitual testosterone seemed to be positively associated with ingroup cooperation and/or outgroup hostility 15

| Clinical implications
The findings of the present study reflect the expected, yet rather small effects of androgen deprivation on cognition and well-being in elderly men, even when considering the potential neuronal adaptation on testosterone decline after long-term treatment. It therefore needs to be emphasized that the negative side effects of ADT cannot outweigh the benefits of ADT on the course of advanced prostate cancer.
Nevertheless, the findings may support the recommendations of recent literature that showed that clinicians should be aware that some androgen-deprived prostate cancer patients develop neurocognitive impairments, yet fail to seek psychological help, 40 although it may be necessary in some cases.

| Conclusion
As expected, ADT significantly reduced free testosterone concentrations in prostate cancer patients. The analysis of the test battery further showed that androgen deprivation influenced different aspects of cognition. In accordance with the literature, testosterone concentrations were negatively associated with visuospatial performance and a reduced intergroup bias in socioeconomic choice was found in the patient group with lower testosterone concentrations. Furthermore, depression severity, cognitive capacity and quality of life were negatively affected by antihormone treatment.
would like to thank L. Izeti for her help with data collection and subject recruitment and all the anonymous participants of this research project for their contribution. This research was funded by resources provided by the Universität Hamburg (Germany).