Prior-Cancer Diagnosis in Men with Nonmetastatic Prostate Cancer and the Risk of Prostate-Cancer-Specific and All-Cause Mortality

Purpose. We evaluated the impact a prior cancer diagnosis had on the risk of prostate-cancer-specific mortality (PCSM) and all-cause mortality (ACM) in men with PC. Methods. Using the SEER data registry, 166,104 men (median age: 66) diagnosed with PC between 2004 and 2007 comprised the study cohort. Competing risks and Cox regression were used to evaluate whether a prior cancer diagnosis impacted the risk of PCSM and ACM adjusting for known prognostic factors PSA level, age at and year of diagnosis, race, and whether PC treatment was curative, noncurative, or active surveillance (AS)/watchful waiting (WW). Results. At a median followup of 2.75 years, 12,453 men died: 3,809 (30.6%) from PC. Men with a prior cancer were followed longer, had GS 8 to 10 PC more often, and underwent WW/AS more frequently (P < 0.001). Despite these differences that should increase the risk of PCSM, the adjusted risk of PCSM was significantly decreased (AHR: 0.66 (95% CI: (0.45, 0.97); P = 0.033), while the risk of ACM was increased (AHR: 2.92 (95% CI: 2.64, 3.23); P < 0.001) in men with a prior cancer suggesting that competing risks accounted for the reduction in the risk of PCSM. Conclusion. An assessment of the impact that a prior cancer has on life expectancy is needed at the time of PC diagnosis to determine whether curative treatment for unfavorable-risk PC versus AS is appropriate.


Introduction
While favorable-risk (PSA ≤ 20; T2b category or less; Gleason score ≤ 7 [1]) prostate cancer (PC) can have a long natural history [2] and is often curable, unfavorable-risk PC (which comprises approximately 20% of cases) accounts for the majority of prostate cancer deaths [3]. Men of PC bearing age are also at risk for a metachronous cancer (i.e., history of or subsequent diagnosis of another cancer). When considering life expectancy in men with PC, competing risks are particularly relevant in men with favorable-risk disease [4][5][6][7][8][9][10], in order to avoid overtreatment of PC where the potential toxicities of treatment can be sustained with no prolongation in survival.
To our knowledge, no study has investigated the impact that the comorbidity of a prior cancer has on the risk of PCSM.
Therefore, we used a SEER population database registry to evaluate the impact that a prior cancer had on the risk of PCSM and all-cause mortality (ACM) in men with newly diagnosed, node negative, nonmetastatic PC, adjusting for age at and year of diagnosis, race, initial treatment (curative or noncurative) or active surveillance (AS) or watchful waiting (WW), and known PC prognostic factors.

Patients Selection and SEER Data Registry.
We used a population-based registry, SEER [11], in order to identify 166  preceding their PC diagnosis versus those without a prior cancer. The Wilcoxon rank-sum test [13] was used to compare the distribution of the continuous clinical characteristics and a Mantel-Haenszel chi square test [14] was used to compare the distribution of categorical covariates.

Risk of PCSM and ACM.
We used a Fine and Gray [12] competing risks and Cox regression [15] to assess whether there was an association between risk of PCSM and ACM, respectively, in men diagnosed with a prior cancer versus no prior cancer adjusting for known PC prognostic factors, curative, noncurative treatment or AS/WW, race, and age at and year of diagnosis of PC. Time zero was defined as the date of the PC diagnosis. Adjusted ( ) and unadjusted hazard ratios (HR) were calculated and are reported along with their 95% confidence intervals (CI) and associated values. A two-sided value <0.05 was considered statistically significant. was used for all calculations related to Fine and Gray and SAS version 9.3 (SAS Institute, Cary, North Carolina) was used for all remaining statistical analyses.

Estimates of PCSM and ACM.
The cumulative incidence method [16] was used to calculate estimates of PCSM stratified by whether the patient had a history of cancer or not at the time of his PC diagnosis. Age adjusted comparisons of the estimates were performed using Fine and Gray's regression. ACM was defined as one minus overall survival. Estimates of ACM were calculated using the method of Kaplan and Meier [17] and age adjusted estimates [18] stratified by whether the patient had a history of cancer or not at the time of his PC diagnosis were compared using an age adjusted log rank value.  also less likely to be African American (7.4% versus 12.3%, < 0.001) and they were more likely to have clinical tumor stage T1c (42.0 versus 37.5%, < 0.001) and were more likely to be diagnosed earlier in time ( = 0.035). Table 2, for men with a history of cancer at the time of the PC diagnosis versus not having the risk of PCSM was significantly decreased (AHR: 0.66 (95% CI: (0.45, 0.97); = 0.033), whereas the risk of ACM was significantly increased (AHR: 2.92 (95% CI: 2.64, 3.23); < 0.001) as shown in Table 3. Given the differences shown in Table 1 that should have led to an increase in the risk of PCSM amongst men with versus without a prior cancer, these results suggest competing risks and non curative PC treatment accounted for the observed reduction in the risk of PCSM.

Estimates of PCSM and ACM Stratified by Prior History of
Cancer. As shown in Figure 1(a), the age-adjusted cumulative incidence of PCSM was significantly lower for men with a prior cancer as compared to those without a history of cancer prior to the PC diagnosis ( = 0.012) However, age adjusted estimates of all-cause mortality were significantly higher ( < 0.001) for men with a prior cancer as compared to no prior cancer, as shown in Figure 1(b).

Discussion
Overtreatment remains an issue in the United States for men with low-risk PC; a target population for whom greater use of AS may be more appropriate particularly in men with significant comorbidity [19]. Our results indicate that men who had a history of cancer prior to the diagnosis of PC were followed longer, were older, and were more likely to have GS 8 to 10 PC and undergo WW or AS compared with those who did not have a history of cancer. However, despite the less frequent use of curative treatment in these men who had more aggressive PC that should have led to an increased risk of PCSM, these men had a significant decrease in the risk of PCSM while their risk of ACM increased significantly, suggesting that competing risks (prior cancer and other comorbidities) and not curative PC treatment accounted for the reduction of PCSM. The clinical significance of these findings is that AS should be more judiciously employed in men with competing risks. Specifically, while men with low risk PC are offered AS if their life expectancy is less than 10 years as per the 2013 NCCN guidelines [20]  AS would not be the preferred strategy by the existing NCCN guidelines, those who have had a prior cancer diagnosis should be considered for AS if the expected rate of cure of the prior CA is low. Several points require further discussion. First, a recent study has shown that men with low-risk PC are more likely to be offered AS when seen in concurrent multidisciplinary setting rather than sequentially [21]. These results suggest that when a multidisciplinary team of physicians consult on a patient AS is more likely to be recommended. Second, investigators have attempted to define novel assays such as circulating tumor cell burden [22] and gene profiling [23,24] that can stratify patients with castration resistant PC into cohorts with longer or shorter median survivals. Such tools are needed in men with low-risk PC in order to more appropriately select men for AS.
Third, men with a prior history of cancer as shown in Table 1 were more likely to be diagnosed with Gleason score 8 to 10 prostate cancer (19.2% versus 15.1%; < 0.001). This likely reflects the fact that they were also significantly older at prostate cancer diagnosis (median age: 72 versus 66; < 0.001) and advancing age has been shown to be associated with higher grade prostate cancer [25]. Also these men were more likely to be diagnosed at an earlier tumor stage (42% versus 37.5% T1c, < 0.001), which may reflect more active medical monitoring with PSA screening given the history of the prior malignancy.
Potential limitations of this study include the inherent limitations associated with a SEER analysis, including limited information about the biopsy specimen (number and extent of positive cores, tertiary Gleason grade 5, perineural invasion). Also, data on comorbidity other than a prior history of cancer is lacking. Nevertheless, the results of our analysis add to the ongoing dialogue about quantifying the impact competing risks can have on life expectancy when deciding on whether curative treatment is likely to benefit a patient with favorable-risk PC. Conversely, some men who are otherwise healthy with favorable-risk PC may require immediate curative treatment of PC as opposed to AS in order to avoid PCSM. In order to ascertain who these men are randomized controlled trials evaluating curative treatment compared to AS should employ a prerandomization stratification by comorbidity using a validated metric of comorbidity [9] in order to assess whether treatment compared to AS benefits all men or only those with no or minimal comorbidity.
In conclusion, while an attempt is being made to offer men AS, the degree to which this has been occurring in the United States between 2004 and 2007 in men with a prior cancer does not appear to be adequate to avoid overtreatment. Therefore, an assessment of the impact that the prior cancer has on life expectancy is needed at the time of PC diagnosis to determine whether curative treatment for unfavorable-risk PC versus AS is appropriate.