Oncologic impact of concomitant prostate cancer characteristics at the time of radical cystoprostatectomy for bladder cancer: a population-based analysis

Abstract Objective The aim of this study was to evaluate the prognostic impact of concomitant prostate cancer (PCa) of the cancer-specific mortality (CSM) in the aging patient’s papulation with bladder cancer (BCa) treated with radical cystoprostatectomy (RCP). Materials and Methods Within the SEER database (2004–2015), 1468 patients were treated with RCP for BCa harboring histopathological PCa findings. To account for other cause mortality (OCM), multivariable competing risk regression (CRR) tested for potential BCa-CSM differences according to PCa characteristics risk factors predicting CSM Results CRR analysis revealed that only following BCa characteristics, as high pathological tumor stages(Ta/Tis/T1 [REF.] vs. T2; HR 2.03, 95% CI: 1.16-3.57, p = 0.014 vs. T3; HR 4.32, 95% CI: 2.45-7.61, p < 0.001 vs. T4; HR 5.06, 95% CI: 2.77–9.22, p < 0.001), as well unfavorable BCa grade IV (Grade I–II [REF.] vs. Grade IV; HR 0.58, 95% CI: 0.35-0.98, p < 0.041) achieved independent predictor status of CSM. With regard to PCa characteristics, none of the covariates yielded independent predictor status of CSM. Conclusions Our study, based on the largest population cohort, demonstrates that even in organ-confined BCa patients, concomitant PCa as second malignancy does not represent a risk factor for survival.


Introduction
Radical cystoprostatectomy (RCP) represents the gold standard therapy for patients diagnosed with muscleinvasive or treatment-refractory non-muscle-invasive bladder cancer (BCa) [1]. Histopathological work-up often reveals concomitant prostate cancer (PCa) in male patients [2]. However, despite frequent co-prevalence of both malignancies, only a limited number of series examined if certain features of concomitant PCa affect survival outcomes in patients undergoing RCP [1, [3][4][5][6][7][8][9]. This is of particular importance, given that the majority of male BCa patients are of advanced age with a peak BCa incidence at 70 years [10]. Specifically, both malignancies, BCa and PCa, share advanced age as a common risk factor of respective unfavorable tumor characteristics [10,11]. This observation is suggestive for worsening of the oncological outcomes in patients in whom both malignancies are presented compared to those diagnosed with BCa exclusively [1,6-9, [12][13][14][15]. Moreover, multiple studies suggested that the demographic phenomenon of co-prevalent PCa and BCa could be explained through similarities in their respective carcinogenic pathways [12][13][14] Most series on this topic are limited by sample size, historical cohorts, and overall great heterogeneity of study findings [1,3-9, [15][16][17]. For example, many Surveillance, Epidemiology, and End Results (SEER) studies aimed to examine causal associations between irradiated PCa patients with subsequent occurrence of radiogenic BCa or rectal cancer [18,19], and thus, focused on metachronous malignancies following PCa treatment. Other studies evaluated the survival prediction of BCa according to treatment modality [20][21][22]. To date, no SEER series has examined the potential oncological effects of concomitant PCa at the time of RCP. Moreover, given that higher age is generally associated with greater comorbidity burden and with limited life expectancy, it is crucial to employ multivariable competing risk techniques for survival analyses in such BCa patients, which were rarely applied in previous series.
Specifically, the only existing population-based study to date showed no effect of concomitant PCa on 10-and 15-year cumulative overall mortality in a Danish RCP cohort [4]. Similarly, a French multicenter study with RCP patients from 25 institutions did not demonstrate any influence of PCa on overall survival [5]. Conversely, some series suggest that concomitant PCa represents a risk factor of overall survival [1,6-9]. For example, Buse et al. [6] reported PCa at the time of RCP to be associated with higher overall mortality after multivariable adjustment. Taken together, it remains unclear if certain unfavorable PCa characteristics adversely affect BCa specific survival in non-metastatic patients undergoing RCP. Therefore, we employed nationally representative SEER data to (1) evaluate the impact of concomitant PCa on survival outcomes after RCP for BCa and (2) to identify potential unfavorable PCa characteristics to adversely affect cancer-specific mortality.

Patient selection
Patient records, used in our epidemiological study, were retrieved within the SEER database, which includes patient demographics, cancer incidence, and survival data from 18 cancer registries, accounting for 26% of the U.S. population. Ethical approvals were obtained from participating institutions through their respective institutional review boards. The SEER database provides unidentified individual-level information and is totally available to the public via a formal application. The permission to gain access to the patient information is received.
Patients with a diagnosis of BCa (International Classification of Diseases for Oncology 67.0-67.9; histological codes: 8120, non-papillary urothelial carcinoma; 8130, papillary urothelial carcinoma; 8122, sarcomatoid urothelial carcinoma; 8070, squamous cell carcinoma; 8041, small cell neuroendocrine carcinoma) who underwent cystectomy (surgery site codes: 60, 61, 62,  63, 64, 70, 71, 72 and 80) between 2004 and 2015 were included [1]. The second inclusion criterion was a concomitant diagnosis of adenocarcinoma of the prostate (International Classification of Diseases for Oncology 61.9; histological code: 8140) and a surgical claim of prostatectomy (surgery site codes 50, 70, and 80; with or without radiation therapy) within the same year as aforementioned BCa (n ¼ 1807) or at least a surgical claim of RCP (BCa surgery code 71; n ¼ 142). These selection criteria were used as a proxy for BCa patients with evidence of concomitant PCa at RCP. Only BCa patients without metastatic disease (M0) and lymph node-positive disease (N0), according to the 6 th edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual were included. Tumor grades I (well differentiated), II (moderately differentiated), III (poorly differentiated), and IV (undifferentiated) were used according to the most recent update of the SEER database [20,23]. Patients who had undergone radiotherapy prior to surgery were excluded (n ¼ 4). Moreover, BCa patients with missing pathological BCa tumor stage or pTX (n ¼ 120), as well as, pT0 or missing pT3 (termed "NOS") PCa tumor stage (n ¼ 14) were excluded. The Gleason Score (GS) is derived from a composite variable based on prostatectomy or biopsy data: If applicable, the prostatectomy GS was used (n ¼ 1085; GS 6 vs. 7 vs. � 8), otherwise the biopsy GS was used (n ¼ 731). Moreover, only the patients who were diagnosed with BCa and PCa as their first of two or more malignancies and second of two or more malignancies, respectively, to further reduce the possibility of another first malignancy other than BCa or PCa as a confounder were included. Finally, these selection criteria resulted in a final cohort of 1468 patients. Prostate-specific antigen (PSA) values were available for selected patients diagnosed 2010-2013 only (n ¼ 416).

Statistical analyses
Descriptive statistics included frequencies and proportions for categorical and medians and interquartile ranges (IQRs) for continuously coded variables. The chi-squared test was used to compare the distribution of categorical variables and t test for continuously coded variables. To account for other cause mortality (OCM), multivariable competing risk regression (CRR) tested for potential bladder cancer-specific mortality (CSM) differences according to PCa characteristics. With regard to general characteristics, covariates consisted of age at BCa diagnosis (continuously coded), year at BCa diagnosis (continuously coded), race, marital status, and socioeconomic status. With regard to BCa characteristics, CRR were also adjusted for year of diagnosis, histological subtype, cancer grade, and pathological tumor stage. With regard to PCa characteristics, CRR were also adjusted for PSA, composite GS, pathological tumor stage, pathological nodal stage, and positive surgical margin status.
Similarly, CRR were repeated in a subgroup of patients who had complete PSA information (n ¼ 416). Here, CRR were adjusted for age (continuously coded), pathological BCa stage, PSA (continuously coded), composite, and pathological tumor stage.
All tests were two-sided, with a statistical significance set at p < 0.05. Analyses were performed with the statistical package for R (the R Foundation for Statistical Computing, version.

Results
Overall, the median age of the patients was 70 years (IQR 62-76; Table 1). Of all 1468 patients, 91% were of Caucasian, 4.4% of African American, and 4.4% of other descent. Regarding the marital status, 69% of the patients were married, 26% were not, and 5.5% had an unknown marital status. The proportion of patients with a low socioeconomic status was 52%.

Prostate cancer
Within those patients with available PSA levels, the median PSA was 2.1 (IQR 0.9-4.7) ng/ml. Proportions of GS 6, 7, and �8 were 71%, 26%, and 3.5%, respectively. According to the pathological T stage, 93% of the tumors were organ-confined (T2), and only 4.6% and 2.4% were considered as T3a and � T3b. An unfavorable pathological nodal stage of N1 was reported in 1.2% of the patients. The surgical margin status was positive in 4.8% of the patients. Taken together, the majority of patients, 64%, had low-risk pathological PCa characteristics of pT2 GS6 with the negative nodal stage (N0). In contrast, only 13% of the patients had unfavorable characteristics: pT3, and/or lymph node invasion, and/or positive surgical margin (R1).

Multivariable competing risk regression
Multivariable CRR tested risk factors of bladder CSM adjusted for BCa, as well as PCa characteristics ( With regard to PCa characteristics, none of the covariates yielded independent predictor status of CSM, albeit higher GS and pathological tumor stage did not meet conventional levels of statistical significance as risk factors. Finally, CRR tested the risk factors of CSM in a subgroup of patients who had complete PSA information (n ¼ 416) (Supplementary Table 1). Subgroup analysis was consistent with our main findings within the full cohort. Specifically, within the PSA subgroup, elevated PSA did not hold true as an independent risk factor (HR 0.98, 95% CI: 0.93-1.03, p ¼ 0.4).

Discussion
We retrospectively evaluated patients undergoing RCP for BCa who presented with concomitant PCa at the time of surgery. Tumor characteristics of both tumor entities were assessed and multivariable competing risk regression was employed to identify potential unfavorable PCa characteristics to affect BCa-specific mortality. Our study revealed several important findings.
First, a substantial proportion of RCP patients had advanced, non-organ confined tumor stage (39%) and most aggressive BCa tumor grade IV (65%). Such rates are consistent with the previous series and are often identified in those who undergo aggressive treatment such as RCP [22]. Moreover, unfavorable prognostic effects of adverse pathological stages in the BCa patient were assessed in the previous studies [20,22,24]. In the SEER-based analysis by Palumbo et al., 5-yr cancer-specific survival in the BCa patients with advanced disease was 63% for T3 and 53% for T4 disease [22]. Interestingly, localized PCa disease, diagnosed as a primarily malignancy, similar in its characteristics to cPCa, is associated with high 5-year survival rates of 99% [25]. This can possibly explain a mitigating effect of concomitant PCa findings on BCa specific mortality in the patients with advanced BCa.
Second, in contrast to rather unfavorable BCa characteristics, the vast majority of RCP patients had favorable PCa characteristics despite advanced median age   [12][13][14]. Molecular alterations, as tumor suppressor gene p53 mutations [14] or overexpression of fibroblast growth factor receptor 3 [12] and prostate stem cell antigen (PSCA) in BCa specimen [12,13] are deemed as potential risk factors for both cancer entities. In consequence, it cannot be ruled out that inclusion of novel non-morphological PCa factors, such as biochemical and genetic markers, might identify those individuals with truly added risk through PCa, which are not identifiable with conventional PCa characteristics. Fourth, unlike many previous studies, we accounted for other cause mortality (OCM) by employing CRR [1,3-9]. Due to the advanced age in our cohort, by which 71% of the patients were over 65 and 30% were over 75-year-old (data not shown), a corresponding but unknown proportion of advanced comorbidity burden must be assumed [20].
Fifth, despite a non-negligible amount of missing PCa data, such as PSA-values, within the existing SEERdataset, which is inherent to previous and current data revisions [27], the CRR model of the full cohort of 1468 patients remained consistent with the subgroup of 416 patients with complete PSA data. Moreover, implementation of stringent selection criteria, where BCa and/or cPCa where assigned primary and secondary malignancy, respectively, ruling out additional confounding malignancies to the best of our knowledge. Taken together, our very comprehensive and strict selection criteria to define our cohort and additional subgroup analyses confirmed the robustness of our results.
Despite their strengths, our study has some limitations. First, as a major limitation of our study, we assume de-centralized review of the histopathological findings, as well as the absence of a standardized sampling technique of RCP specimen, which leads to higher heterogeneity of pathological results. As addressed by Lopez-Beltran et al., variability in slice thicknesses, as well as the fact of whether or not the prostate was totally or partially embedded, could explain high variability within RCP PCa findings [2]. Moreover, in the case of missing pathological RCP GS data, correspondent biopsy GS data was used. The implementation of biopsy GS results, also adds traditional systematic biopsy limitations, like moderately low cancer-detection rates and possible upgrading of biopsy finding GS the prostatectomy [28]. However, the pathological prostate biopsy work-up is usually better standardized than RCP work-up, which may decrease the variability of pathological results [29]. Second, the vast majority of patients (91%) were Caucasian, which might be explained by demographic and geographic as well as age-related and racial differences [30]. This imbalance must be taken into account when analyzing the PCa characteristics as risk factors in our cohort, since previous series demonstrated that African-American men are at greater risk of PCa mortality [31]. In turn, our current findings do not necessarily translate to other race groups despite utilizing multivariable analyses. Third, PSA-values were available only in the minority of the patients and usually represent an independent risk factor in PCa patients. However, our subgroup analyses of those with complete PSA information were consistent with the main cohort. Fourth, lack of information on patient's comorbidities and performance status represents potential confounding factors, which are unavailable in SEER. However, we accounted for the patient's age and acknowledged the observation that age represents a proxy and a determinant of possible comorbidities [20,32]. Moreover, other variables, as e.g. radiographic findings, may also be considered as predictors of CSM, which were not recorded in the SEER dataset. Fifth, extremely comprehensive selection criteria were implemented to rule out additional malignancies as possible cofounders and to present the most contemporary SEER-based cohort. Finally, the SEER-database represents only North American patients, which limits generalizability of our findings, as well as all other limitations related to the population-based, retrospective nature of the database. Nonetheless, our study represents the largest recorded population-based cohort of the BCa patients with concomitant PCa findings treated with RCP [4,5].

Conclusion
In conclusion, unfavorable concomitant PCa characteristics do not seem to represent risk factors of BCa-specific mortality. Moreover, our study is the first population-based study on PCa findings in BCa patients treated with RCP conducted in North American patients.

Ethical approval
Patient records, used in our epidemiological study, were retrieved from the SEER database, which includes patient demographics, cancer incidence, and survival data from 18 cancer registries, accounting for 26% of the U.S. population. Ethical approvals were obtained from participating institutions through their respective institutional review boards. The SEER database provides unidentified individual-level information and is totally available to the public via a formal application. The permission to gain access to the patient information is received.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
Open Access funding enabled and organized by Projekt DEAL. We acknowledge support from the University of Muenster. No further specific grants from any funding agency in the public, commercial, or not-for-profit sectors were received.

Data availability statement
The data used in our epidemiological study were retrieved from openly available in the SEER database (https://seer.cancer.gov/).