Effect of chemotherapy in metastatic prostate cancer according to race/ethnicity groups

No North‐American study tested the survival benefit of chemotherapy in de novo metastatic prostate cancer according to race/ethnicity. We addressed this void.

vs Asian. Race/ethnicity specific propensity score matching was applied. Here, additional landmark analysis was performed.
Results: Of 4232 de novo metastatic prostate cancer patients, 2690 (63.3%) were Caucasian versus 783 (18.5%) African-American versus 504 (11.8%) Hispanic/Latino versus 257 (6.1%) Asian. Chemotherapy rates were: 21.3% versus 20.8% versus 21.0% versus 20.2% for Caucasians versus African-Americans versus Hispanic/Latinos versus Asians, respectively. At 30 months of follow-up, overall survival rates between chemotherapy-exposed versus chemotherapy- promising new systemic therapies, remains low. [1][2][3][4][5][6] To the best of our knowledge, no large scale, sufficiently-powered, North-American study examined the effect of race/ethnicity on overall survival benefit from chemotherapy in de novo mPCa. 2,[7][8][9][10][11] The effect of race/ethnicity was examined in few post hoc analyses of prospective randomized trials. 12 However, no prospective randomized trials examining the effect of systemic therapies on overall survival in de novo mPCa, relied on preplanned race/ethnicity stratification schemes. [13][14][15] Moreover, the proportions of race/ ethnicity groups other than Caucasian, were small, and so were the actual numbers of included patients from race/ethnicity groups other than Caucasians. 12 Taken together, it is unknown, whether North-American race/ethnicity groups other than Caucasians, benefit of systemic chemotherapy. We addressed these knowledge gaps in the current analysis. Specifically, we hypothesized that chemotherapy in de novo mPCa is associated with similar benefits across all race/ethnicity groups. We tested this hypothesis relying on the Surveillance, Epidemiology, and End Results (SEER) database (2014-2015).

| Study population
The current SEER database samples 34.6% of the US population and approximates it in demographic composition and cancer incidence. 16 Within the SEER database (2014-2015), we identified patients

| Statistical analyses
The statistical analyses consisted of four steps. First, we addressed overall survival in separate and distinct analyses that addressed four  Second within each race/ethnicity group, we separately relied on propensity score matching to address potential differences between chemotherapy-exposed versus chemotherapy-naïve patients. Propensity   interval), PSA (≤97, >97 ng/ml), GGG (≤IV, V, unknown), cT-stage (≤cT2, cT3/4, cTx), cM-stage (cM1a/b, cM1c, cM1unspecific) and local treatment (yes, no, unknown). Each chemotherapy exposed patient was matched with one chemotherapy naïve patient within each race group.
The exception consisted of Caucasians, in which one chemotherapyexposed patient was matched with two chemotherapy-naïve patients.
Third, after propensity score matching, all Kaplan-Meier plots and Cox regression models were separately refitted within the four separate and distinct race/ethnicity groups: Caucasian, African-American, Hispanic/Latino, Asian. The same covariates as above were used.
Finally, survival analyses were repeated in propensity score matched cohorts after landmark analysis (3 months) was applied.

| Descriptive characteristics
Between 2014 and 2015, we identified 4234 de novo mPCa patients.

| Overall survival in Caucasian race/ethnicity
Unmatched analyses compared 573 chemotherapy-exposed versus

| Overall survival in Asian race/ethnicity
Unmatched analyses compared 52 chemotherapy-exposed versus 205 chemotherapy-naïve Asian patients. At 30 months of follow-up, overall survival rates were 77.7 versus 65.0%, favoring chemotherapyexposed patients ( Figure 4A

| DISCUSSION
We hypothesized that chemotherapy in mPCa is associated with similar benefits across all race/ethnicity groups. We tested this hypothesis within the SEER database between 2014 and 2015. We made several noteworthy findings.
T A B L E 2 Race/ethnicity group specific uni-and multivariable Cox regression models predicting overall mortality in metastatic prostate cancer patients according to chemotherapy exposure before and after propensity score matching  Hispanic and 8 Asian patients. 12 These numbers were clearly F I G U R E 2 Kaplan-Meier plots illustrating overall survival in 783 African-American de novo metastatic prostate cancer patients (A) before propensity score matching and (B) in 300 patients after propensity score matching, stratified by chemotherapy status insufficient to prospectively plan a stratification scheme that would allow specific race/ethnicity group analyses. 22 In consequence, few propensity score matched data to maximally reduce the differences in age and prostate cancer characteristics between chemotherapyexposed versus chemotherapy-naïve prostate cancer patients, within each specific race/ethnicity group.
Third, in both unmatched, as well propensity score matched analyses, we recorded statistically significantly lower overall mortality in Caucasians, who represent the largest race/ethnicity group. It is also noteworthy, that statistically significantly lower overall mortality was also recorded in Asians, who represent the smallest race/ethnicity group. These observations indicate two important facts.
Chemotherapy-exposure is associated with a potential survival benefit in Caucasian and Asian patients with or without adjustment for patient and tumor characteristics differences. Specifically, the magnitude of overall mortality reduction was even stronger in matched Asian cohorts. Additionally, it is of utmost importance to note, that lower overall mortality, that was recorded in chemotherapy-exposed patients, was documented in equally important degree in the numerically smallest, Asian race/ethnicity group. In consequence, despite most important sample size and power limitations that apply to Asian race/ethnicity group, the effect of chemotherapy and lower F I G U R E 4 Kaplan-Meier plot illustrating overall survival in 257 Asian de novo metastatic prostate cancer patients (A) before propensity score matching and (B) in 80 patients after propensity score matching, stratified by chemotherapy status overall mortality rates in chemotherapy-exposed patients could be documented with high degree of statistically significance, in both matched and unmatched analyses. These observations regarding Asian mPCa patients are in agreement with previous reports. [23][24][25] It is of utmost importance to note, that the effect of chemotherapy was not associated with lower overall mortality in African-Americans (second largest race/ethnicity group), as well as in Hispanic/Latinos (third largest race/ethnicity group).
Taken together, our results indicate, that the recorded lack of overall survival benefit associated with chemotherapy administration in African-American and Hispanic/Latino patients cannot be solely attributed to sample size or power consideration, since a strong overall survival benefit was recorded in a much smaller patient subgroup (Asian). Clinical implications of the above findings clearly indicate that prospective analyses addressing chemotherapy benefit in African-American and Hispanic/Latino patients are urgently needed.
In absence of such data in the foreseeable future, additional retrospective epidemiological analyses mirroring the current study should be carried out. The NCDB represents an ideal data pool to carry out such analysis with the intent of validating or refusing our analysis.
Several limitations applied to our study. First, the rate of chemotherapy exposure is low in the current study. It is nonetheless very similar to the rate observed in other large-scale population-based studies. 18,26 Moreover, the nature of administered chemotherapy and adherence is unknown with respect to its type, number of lines, overall duration as well as, individual efficacy of used regimens.
Moreover, SEER does not provide information regarding concomitant medication such as antiandrogen deprivation therapy. This limitation is shared with other population-based data examining chemotherapy effects and should be interpretated accordingly. [26][27][28] Second, the retrospective nature of the study introduces a number of selection biases, that distinguish chemotherapy exposed patients from others.
Differences in patient and clinicopathological features that were not captured by SEER may have influenced the current findings. Moreover, as reported, chemotherapy-exposed patients tended to harbor more aggressive prostate cancer phenotypes. To address these biases multivariable analyses were complemented by propensity score matching, to more completely and strictly address these differences.
Third, patient numbers, especially for Asians were low. Finally, a number of established predictors of survival (lactate dehydrogenase, hemoglobin) for mPCa patients were unavailable in both the current as well as other population-based studies that include the NCDB. 29,30 5 | CONCLUSIONS De novo mPCa Caucasian and Asian race/ethnicity patients exhibit higher overall survival, when exposed to chemotherapy. Conversely, these observations could not be made for African-American and Hispanic/Latino patients. Since we observed significant improved overall survival in the largest and smallest race/ethnicity subgroup, lack of improved overall survival in the second and third subgroup is unlikely related to sample size limitations.

ACKNOWLEGMENT
Benedikt Hoeh was awarded a scholarship by the Giersch Stiftung.
Open access funding enabled and organized by Projekt DEAL.