Patients with mPC scored lower in all neurocognitive tests and had more frequent neurocognitive deficits than reference samples. The results suggest that treatment for mPC is correlated with mild neurocognitive deficits.

The frequencies of mild cognitive deficits were significantly above the expected frequencies. Almost all patients (92%) showed at least a slight cognitive deficit in at least one test. In contrast, moderate and severe deficits were relatively rare in our sample. These frequencies are also greater than those described by the ICCTF. Notably, our study included not only the three recommended tests of learning and memory, processing speed, and executive function but also the NBV test of verbal working memory. Excluding the NBV test results reduced the frequency of patients with a cross-test neurocognitive deficit from 13 to 10, a frequency not significantly higher than that expected in a healthy reference group (7).

Compared to the reference sample, patients treated for mPC had the most significant deficits in word fluency (in the WIWO) and visuomotor processing speed (in the TMT-A). Word fluency is also the only domain in which significantly more severe neurocognitive deficits were observed. Regarding verbal long-term memory (measured with the AWLT short-term delayed recall) and executive function (measured with the TMT-B and NBV test), significantly lower performance and increased mild cognitive impairments were observed in patients treated for mPC. Verbal learning ability, assessed with immediate recall/learning sum in the AWLT, is the only subtest in which patients did not significantly differ from the reference sample. These results are consistent with a recent review of men with advanced prostate cancer and suggest that treatment with ADT may increase the risk of cognitive impairment [6]. Overall, the available data are unclear. In two long-term studies, no consistent trend in terms of the affected neurocognitive domain was found [13, 14]. In addition, two reviews did not obtain consistent results concerning ADT-induced cognitive impairment [15, 16].

In our sample, we found no evidence that a single type of treatment predicted cognitive performance. Treatment duration had a few partial correlations that supported our hypothesis. However, these analyses are only exploratory. Due to the small number of cases, multivariate analyses are limited in their ability to determine correlations. The significant association between the GDS and duration of abiraterone acetate use does not confirm a causal effect, although it supports our hypothesis. Nevertheless, longer durations of treatment with abiraterone acetate in our sample of patients were associated with neurocognitive deficits. This correlation was not found for enzalutamide and is the opposite of that found in a disproportionality analysis of a pharmacovigilance database [17]. The neurotoxicity risk was higher with enzalutamide than with abiraterone acetate. However, these authors also note the exploratory nature of their study [17].

One limitation of our study is the nature of our cohort. This incidental sample was evaluated during an outpatient appointment and is relatively small. Moreover, no randomized control group is available. Nevertheless, age- and sex-specific reference scores are available for these neurocognitive tests, which enabled good classification of test results. Our results are important, because studies on cancer- and treatment-related effects in patients with mPC are scarce [18]. As our test battery facilitates cross-comparability of results according to ICCTF guidelines, it can be easily included in future meta-analyses.

Assigning single affected neurocognitive domains to individual treatment-specific characteristics is difficult for several reasons. First, the individual treatment histories of mPC patients are often so different and complex that clear distinctions are not possible. Second, cognitive domains do not function independently, especially in the case of mild impairment. Third, comparability of study results is challenging due to changes in treatment plans and the use of new agents. Fourth, the large number of possible associations between treatments and cognitive domains necessitates very large sample sizes to achieve sufficient statistical power.

In addition to these difficulties, determining the cause of neurocognitive deficits is further complicated by the possible influence of hypogonadism. In patients treated with ADT there are metabolic changes involving the glycemic control and lipid metabolism, increased thrombotic risk, an increased risk of myocardial infarction, severe arrhythmia and sudden cardiac death. Still, these adverse effects can be also due to the subsequent hypogonadism [19]. Therefore, our results should be interpreted only as measures of associations, not as causal relationships.

Nevertheless, in the context of the available evidence, our results indicate that the combination of ADT with a cytostatic or a second-generation antiandrogen is correlated with mild general cognitive impairments. In addition, moderate and severe cognitive impairments were found in a few cases. The identification of single agents responsible for impairment in specific cognitive domains among individuals remains unclear [15].