Association of Second-generation Antiandrogens With Depression Among Patients With Prostate Cancer

Key Points Question Are second-generation antiandrogens (AAs) associated with increased risk of depression among older men diagnosed with prostate cancer? Findings In this cohort study of 30 069 men aged 66 years and older, there was a statistically significant 2-fold increase in depression among patients treated with second-generation AA compared with traditional forms of hormone therapy (HT) and no HT. Meaning These findings suggest that use of second-generation AAs is associated with a clinically significant increased risk of depression when compared with traditional HT alone or no HT.


Introduction
Prostate cancer accounts for more than 1 in 5 of all new cancers diagnosed in the United States. 1 Hormone therapy (HT), including androgen deprivation therapy (ADT), is frequently used in localized prostate cancer along with radiation and is a mainstay of treatment for metastatic, locoregional, and recurrent disease. 2,3 HT deprives prostate cancer cells of the androgen stimulation that promotes prostate cancer growth and progression, resulting in improvements in overall survival. 4 HT is typically achieved through medical castration, or androgen deprivation, with luteinizing hormone-releasing hormone (LHRH) agonists with or without antiandrogen (AA) therapy. Despite high initial response rates, nearly all men with advanced prostate cancer progress to castration-resistant disease, which can be treated with second-generation AAs. 5,6 Second-generation AAs are also used in combination with traditional forms of HT as first-line therapy in patients with hormone-naive metastatic prostate cancer. Second-generation AAs, which work by both inhibiting androgen production (abiraterone) and as androgen receptor antagonists (apalutamide, darolutamide, and enzalutamide), lead to a more profound decrease in androgen signaling than prior therapies. 6 This increased potency may carry a greater risk of adverse effects related to blocking and suppressing testosterone. 7 Prior studies have shown a consistent association between ADT and depression using both depression inventory testing and claims data. 8,9 More recently, depression instrument and patientreported outcome measures data have emerged that support a potential association of secondgeneration AAs with symptoms of depression. 7 Whether there is an association between secondgeneration AAs and a clinical diagnosis of depression is unknown.
We hypothesized that second-generation AAs would be associated with an increased risk of depression, including compared with traditional forms of HT. This may be particularly clinically relevant, as depression is associated with decreased overall survival in patients with prostate cancer. 10,11 Considering the prevalence of prostate cancer and the increasing use of secondgeneration AAs, an association between second-generation AAs and depression may have significant public health implications.

Cohort Selection
We used the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER)-Medicare and the Texas Cancer Registry (TCR)-Medicare linked databases. We included men with a first primary prostate cancer diagnosis of localized, regional, or distant stage diagnosed at age 66 years or older from January 2011 to December 2015 without a second cancer within 12 months. We included patients with continuous Medicare Parts A, B, and D coverage, without health maintenance organization enrollment, from 12 months before until 6 months after diagnosis. We excluded individuals who received any form of HT (LHRH agonists/antagonists, AAs) prior to prostate cancer diagnosis, those diagnosed with depression from 12 months before through 6 months after prostate cancer diagnosis, and those who did not survive at least 6 months following prostate cancer diagnosis (Figure 1). The study was deemed exempt from review by the MD Anderson Cancer Center institutional review board and followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. A waiver of informed consent was obtained because all data were received deidentified. exposure to a second-generation AA regardless of other HT exposure (second-generation AA group).
We used a time-varying exposure variable to categorize exposure to HT with individuals remaining in their given group once they met criteria. We stratified HT users by cumulative duration of HT use, defined as 1 to 6, 7 to 12, and more than 12 months.

Outcome and Variables
The follow-up period was 6 months after prostate cancer diagnosis to the end of Medicare coverage.
Patients were censored at date of last enrollment or end of data set follow-up, which was December 31, 2017. The primary outcome was depression, which was defined by the presence of any relevant International Classification of Diseases, Ninth Revision (ICD-9) or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnosis code (eTable 1 in the Supplement) using Medicare claims data, including physician, outpatient, and inpatient claims. 8 Demographic and clinical variables included year and age of diagnosis, race and ethnicity (as reported by the SEER and TCR registries; race and ethnicity were reported as Black, Hispanic, White, and other [American Indian or Alaska Native, Asian (ie, Chinese, Filipino, and Japanese), Native Hawaiian or Pacific Islander, 2 or more races, and other or unspecified]), marital status, education and poverty quartile, state buy-in (an indication that the beneficiary received Medicaid or other state assistance for those with low income), and residence area as well as T stage, N stage, stage, and grade extracted from the SEER or TCR patient enrollment file. Quartiles of percentage of non-high school education and poverty were estimated based on 2000 census tract. We calculated the Charlson Comorbidity Index score based on physician, inpatient, and outpatient claims in the 12 months before prostate cancer diagnosis. 12 We defined chemotherapy, radiation, and surgery based on ICD-9 and ICD-10

Statistical Analysis
We compared baseline demographic and clinical variables between groups using χ 2 tests. We plotted the 1-and 2-year cumulative incidences of depression treating death as a competing risk. 13 We examined the risk of depression in the second-generation AA, traditional HT, and no HT groups. We calculated the number needed to harm (NNH) to provide a measure of the absolute effect size of second-generation AAs by taking the inverse of the difference between the rates of depression between the no HT group and traditional HT groups compared with the second-generation AA group.
We implemented time-varying exposure multivariable Cox proportional hazards models to determine the association of ADT use with depression using the inverse probability treatment weighted (IPTW) method. 14 We estimated propensity scores for each group via logistic regression models including year and age of diagnosis, race, marital status, state buy-in, stage, and Charlson comorbidity score. We defined the IPTW as propensity score divided by 1 minus the propensity score for the traditional HT and second-generation AA groups, assigning the no HT group a weight of 1. We checked the postadjustment balance via the standardized difference, with a standardized difference less than 10% indicating good balance. We used a backward selection process and retained the variables in the final multivariable models based on both statistical and clinical significance. Our final model included year of diagnosis, age, race, marital status, state buy-in, stage, Charlson comorbidity score, and surgery within 6 months after diagnosis. We expressed the results in hazard ratios (HRs) or subdistribution hazard ratios (SHRs) and 95% CIs. Subgroup analyses were carried out within localized, regional, and distant disease to address potential confounding by indication. Unweighted analyses were also conducted. The index date for the primary analysis was 6 months after prostate cancer diagnosis (ie, start of exposure).
We conducted a sensitivity analysis using an alternative cohort design in which we defined the start of the follow-up period as 2 years after the diagnosis of prostate cancer, and rather than using a time-varying covariate, exposure groups were defined based only on exposure during that 2-year period. Individuals with a first exposure to HT after the start of the follow-up period were excluded (eTable 2 in the Supplement). The index date for the sensitivity analysis was 2 years after diagnosis of prostate cancer.
We considered P < .05 statistically significant, and all tests were 2-tailed. We used SAS version 9.4 (SAS Institute) and R version 4.0.0 (R Project for Statistical Computing).

Results
Our analytic cohort ( Figure 1 The 1-and 2-year cumulative incidence of depression was highest in the second-generation AA group vs the traditional HT and no HT groups, including with stratification by stage (Figure 2 and  Finally, we found a statistically significant association between depression and decreased overall survival in all 3 groups (ie, no HT, traditional HT, and second-generation AA) (eTable 4 in the Supplement).

Discussion
In this large retrospective cohort study, patients with prostate cancer who received a secondgeneration AA had a clinically relevant and statistically significant absolute increased risk of depression compared with patients who received traditional HT alone or never received HT.
Importantly, our results were consistent when examining only individuals with distant disease at diagnosis. This association remained statistically significant after adjusting for extensive demographic and clinical variables and when applying propensity score-based weighting.
Depression rates are high among patients with cancer, 15 and depression is associated with mortality in patients with cancer and specifically among men with prostate cancer. 10,11 In our analysis, we found that across all HT exposure groups, a post-prostate cancer diagnosis of depression was associated with worse overall survival. While past studies have supported an association between ADT and a diagnosis of depression, 8,[16][17][18] to our knowledge our data are the first to find higher rates of depression diagnoses with a second-generation AA.
Second-generation AAs have been shown to significantly increase survival when compared with no HT and primary ADT alone, especially in patients with metastatic prostate cancer. 19,20 Secondgeneration AAs have been proven to be beneficial in the treatment of prostate cancer, most importantly due to their ability to prolong life in patients with late-stage disease who have fewer safe and effective treatment options. However, their association with depression should not be

Limitations
Our study has limitations. First, our SEER-Medicare cohort was limited to men aged 66 years and older at diagnosis. While our results may not be generalizable to younger men, most patients diagnosed with prostate cancer are older. 1 Second, the design of our study is retrospective; thus, we are unable to fully assess causality. Third, our study is limited by the nature of large claims data, which rely on the accuracy of diagnostic codes and are not comprehensive of all patient characteristics.
Notably, administrative claims data have been shown to underestimate the true incidence of depression. 47 Fourth, the adjustment for patient characteristics based on Charlson Comorbidity Index was done prior to the diagnosis of prostate cancer. Therefore, it is possible that it does not reflect the burden of comorbidity at the time of treatment with second-generation AAs. Fifth, our second-generation AA group mostly included men who were also exposed to traditional forms of HT and therefore had a longer cumulative exposure to HT of any form. However, we did not observe a duration-dependent increase in depression risk with exposure to traditional HT in our full cohort, and therefore, a longer total duration of HT exposure does not appear to completely explain the magnitude of increased risk of depression observed in the second-generation AA group. Sixth, we excluded individuals with prevalent depression from our study given the limitations of examining the worsening of existing depression in claims-based analyses. However, the outcomes associated with second-generation AAs among men who carry a pre-AA therapy diagnosis of depression is critical and should be examined in future studies. Seventh, men receiving second-generation AAs may be more likely to receive care at research and academic centers and therefore may have more access to ancillary services that would increase their likelihood of being diagnosed with depression. However, we were unable to account for this in our analysis.
Furthermore, men who are receiving a second-generation AA may be more likely to have advanced disease and be receiving second-line therapies after the failure of first-line treatments.
Therefore, the second-generation AA group may have been more likely to experience depression secondary to prostate cancer severity and an adverse treatment course. Our results were consistent when examining only individuals with distant disease at diagnosis in our primary analysis. In our sensitivity analysis, in which we defined the start of the follow-up period as 2 years after the diagnosis of prostate cancer, we did not observe a statistically significantly increased risk when limiting our analysis to individuals with metastatic disease at diagnosis. However, it is difficult to interpret this finding given the limited power of this secondary subgroup analysis. Importantly, as our data set does not capture clinical progression (eg, transition to castrate-resistant disease and second-line therapy), our analysis does not fully account for confounding by indication (ie, individuals more likely to both have depression and more likely to receive second-generation AAs because of their disease course). Our results should be validated in other data sets that, ideally, contain information on patient and disease characteristics at the time of receipt of HT.

Conclusions
In our cohort of more than 30 000 US men aged 66 years with prostate cancer, we observed a clinically relevant and statistically significant association between the incidence of depression in men taking second-generation AAs when compared with men receiving traditional forms of HT or men not taking HT. While our results were consistent when examining only individuals with metastatic disease at diagnosis, our results should be validated in future studies designed to account for secondgeneration AA indication (eg, castration resistance). Considering that recipients of secondgeneration AAs have regular health care exposure due to the nature of their treatment, early depression screening and treatment are feasible interventions that could greatly improve their quality of life and clinical outcomes. The possible increased risk of depression with secondgeneration AA use should be discussed with patients, and depression screening should be considered in all recipients of HT, in particular those who receive second-generation AA therapies.