To examine clinical characteristics, treatment modalities and oncological outcomes of prostate cancer (PCa) according to young (≤50) vs. old age.
Of 407,599 men with primary adenocarcinoma of the prostate within the Surveillance, Epidemiology and End Results (SEER)-database (2004 to 2013), 18,387 were aged ≤50 years (4.5%). Time trends, cumulative incidence, and competing risks regression (CRR) analyses tested for differences between young and old patients. Multi-variable analyses were adjusted for year of diagnosis, race, marital status, Gleason Score, clinical tumor stage, and lymph node status.
Younger men had more favorable tumor characteristics: lower Gleason Score, lower median PSA, and lower rates of metastases at diagnosis compared to their older counterparts. Over time, no local treatment (NLT) rates increased, radical prostatectomy (RP), and brachytherapy (BT) rates decreased and external beam radiation (EBRT) rates remained unchanged. Moreover, the rate of de novo metastatic prostate cancer increased in young patients from 2% (2004) to 3.2% (2013) (p = 0.004). CRR models showed no difference in prostate cancer-specific mortality (PCSM) between young and old, across all local treatment types.
Young PCa patients have more favorable disease characteristics at presentation, are less frequently treated with RP or BT and more frequently benefit of NLT. PCSM did not differ between young and old patients. However, it is worrisome that recently more young PCa patients are diagnosed at a metastatic stage.
Although most prostate cancer (PCa) patients are aged 70 years and older, some men are diagnosed at a much younger age. Individuals diagnosed with PCa below 50 years have received less attention in urologic literature. Nonetheless, these individuals will be affected by their disease and its treatment the most and over the longest period of time .
Previous studies reported conflicting results regarding PCa in younger men [2–9]. Historical literature suggested that young PCa patients harbor more aggressive and more advanced disease at presentation, which may result in worse outcomes [2–4]. In contrast, contemporary studies reported better disease characteristics and more favorable cancer control outcomes in younger patients [5–9]. Patient selection and temporal changes in screening and early detection methods might explain these discrepancies . Nonetheless, a formal comparative analysis is warranted to substantiate that explanation. Moreover, it is of note that most studies, focusing on young patients, predominantly examined radical prostatectomy (RP) outcomes. Other treatment modalities and patients with more advanced disease at presentation were not considered [7, 9–12]. Such approach may have underestimated young PCa patients’ disease aggressiveness and failed to address the rates of metastatic PCa in this population, especially over time.
Therefore, the objective of this contemporary population-based study was to examine initial disease characteristics, treatment modalities, and oncological outcomes in PCa patients with special focus on individuals aged ≤50 years, across all stages at presentation and across all initial treatment modalities.
Materials and methods
Our study relied on the Surveillance, Epidemiology and End Results (SEER) database. This large population-based database covers approximately 28% of the Unites States population and includes 18 registries (SEER: Surveillance epidemiology and end results (http://seer.cancer.gov/about/)).
Overall, 498,207 patients diagnosed with primary adenocarcinoma of the prostate (site code C61.9, histological type according to ICD.0.3 codes: 8140, 8480, 8490, 8550) between 2004 and 2013 were identified. Exclusion criteria were missing data on patient and disease characteristics (n = 44,209) or on first course of treatment (n = 46,399). This resulted in 407,599 assessable patients. According to previous reports, patients aged ≤50 years were considered “young” (n = 18,387), whereas the remaining men represented the “older” cohort [7, 12, 13]. Primary treatment modalities consisted of: (1) RP (surgery site codes 50 or 70), (2) brachytherapy (BT) (radiation code Radioisotopes, Radioactive implants, or Combination of beam with implants or isotopes), (3) external beam radiation (EBRT) only (radiation code Beam radiation), and (4) no local treatment (NLT), which includes active surveillance (AS), watchful waiting (WW) and androgen deprivation therapy (ADT). Within RP and BT cohorts, treatment with additional EBRT was allowed. Age was categorized into the following groups: ≤50 years vs. >50 years as previously reported [8, 13]. For trend and survival analyses, only patients without de novo metastases were included. These criteria resulted in 384,306 patients including 17,651 men aged ≤50 years. Prostate cancer-specific mortality (PCSM) was defined as cancer-related deaths and PCSM-free survival (PCSM-FS) was measured from date of diagnosis to cancer-related death or last follow-up, whichever came first.
Descriptive statistics relied on tests of means and proportions using the χ2-test for categorical and the t-test for continuously coded variables. Annual rates of newly diagnosed PCa in young men aged ≤50 years, as well as tumor characteristics and treatments applied, were analyzed over time. The estimated annual percentage change (EAPC) was calculated using the squares log linear regression method, as previously reported . Cumulative incidence plots (Gray test) graphically displayed PCSM, adjusted for other cause mortality (OCM) and according to age . Multi-variable competing risks regression (CRR) analyses assessed the impact of age on PCSM . CRR models were adjusted for race, marital status, clinical tumor stage, Gleason Score, clinical lymph node status, and treatment modality. R software environment for statistical computing and graphics (version 3.3.0) was used for all statistical analyses. All tests were two sided with a level of significance set at p < 0.05.
Table 1 depicts initial patient and disease characteristics of all 407,599 PCa patients, according to age at diagnosis. Overall, 18,387 (4.5%) were aged ≤50 years, 506 (0.1%) ≤40 years, and 10 (<0.01%) ≤30 years. Median age of the entire cohort was 65 years (inter quartile range(IQR): 59–72) and median follow-up 52 months (IQR: 26–82). The majority of patients were white (69.1%) and married (66.7%).
Younger patients had significantly more favorable tumor characteristics compared to their older counterparts (Table 1). For example, 55.6% of men aged ≤50 years had Gleason scores ≤6 compared to 45.8% in men >50 years (p < 0.001). Conversely, older patients had higher rates of Gleason scores 8–10 (16.1 vs. 8.3%, p < 0.001). In addition, younger patients had lower median PSA (5.2 vs. 6.4 ng/ml, p < 0.001) and lower overall rates of metastatic stage at diagnosis compared to their older counterparts (2.5 vs. 3.2%, p < 0.001). Furthermore, men aged ≤50 years were more frequently of African American descent (25.1 vs. 14.0%, p < 0.001), whereas most men >50 years were Caucasian (69.5 vs. 59.7%, p < 0.001). In addition, younger patients were more likely to undergo RP (70.5 vs. 36.9%, p < 0.001) and less likely to receive EBRT as primary treatment (7.8 vs. 23.7%, p < 0.001).
Time trend analyses
Overall, rates of newly diagnosed PCa in men aged ≤50 years remained constant over the 9-year observation period and ranged between 4.6 and 3.8% (p = 0.7) (Fig. 1). However, rates of metastatic stage at diagnosis had increased in young patients from 2% in 2004 to 3.2% in 2013 (p = 0.004). In 17,651 non-metastatic PCa patients aged ≤50 years, no changes were recorded in rates of Gleason score 3 + 3 and 3 + 4 (p = 0.2), localized PCa ( ≤ cT2, p = 0.1) and clinical lymph node involvement (p = 0.5) at diagnosis. Nonetheless, the rates of initial treatments changed over time: RP (76.5%  to 62.4% , p = 0.008) and BT (7.4%  to 4.4% , p = 0.002) rates decreased, NLT rates increased (8.9%  to 25.2% , p < 0.001), and EBRT rates remained unchanged (7.2%  to 8.1% , p = 0.1) (Fig. 2).
In cumulative incidence analyses adjusted for OCM, PCSM was lower in patients aged ≤50 years vs. >50, when RP (0.5 vs. 0.7%, p < 0.001), BT (0.4 vs. 0.9%, p < 0.001), and NLT (3 vs. 5.4%, p < 0.001) were examined. No difference was recorded according to age after EBRT (2.8 vs. 2.5%, p = 0.7) (Fig. 3a–d).
In multi-variable CRR models (Table 2), age ≤50 was not an independent predictor of PCSM across all local treatment types (all p > 0.05). However, within the NLT cohort, age ≤50 years were an independent predictor for lower PCSM (p < 0.001). In addition, higher Gleason score, higher tumor stage, and lymph node involvement yielded higher PCSM across all treatment types. African American race reached independent predictor status for PCSM within the EBRT and RP cohort, while unmarried status was an independent predictor of PCSM across all treatment types except for BT. OCM was invariably higher in older patients, across all treatment types and in the NLT cohort (all p < 0.001).
In subgroup analyses, focusing on predictors of PCSM in exclusively young patients, a higher Gleason score, and lymph node involvement were independent predictors of PCSM within the RP, EBRT, and NLT cohort. Owing to the insufficient sample size, subgroup analyses could not be performed within the BT cohort.
We hypothesized that young patients may potentially harbor more aggressive disease characteristics at diagnosis and that this may translate into higher PCSM. Relying on the SEER database, we identified several noteworthy findings.
First, contrarily to our hypothesis, 18,387 younger patients, examined within our analysis, had more favorable clinical characteristics than their older counterparts: lower Gleason scores, lower PSA values, and lower rates of metastatic disease. This is in agreement with contemporary studies [8, 10, 13, 17]. Indeed, a recent multicenter study by Kinnear et al. , analyzing 7018 patients from South Australia including 182 aged ≤50 years, found that younger PCa patients had less aggressive clinical characteristics, and had significantly lower BCR and PCSM rates than their older counterparts. Similarly, Samadi et al.  described 2495 PCa patients (271 aged ≤50 years) that underwent robotic prostatectomy at a single North American institution. They observed higher rates of favorable Gleason scores in men ≤50 years old. Furthermore, in our study, more younger patients were of African American descent, which is in agreement with a previous multicenter study by Parker et al.  that compared race and oncological outcomes among 12,081 PCa patients including 543 aged <50 years between 1989 to 2009.
Second, we provided time trend analyses that focused on rates of de novo PCa diagnoses and de novo metastatic disease in young PCa patients. It is of note that incidence of de novo PCa is decreasing according to previous SEER analyses. While incidence of de novo PCa was 166.2 cases/100,000 men in 2004, incidence was 110.1 cases/100,000 men in 2013 . However, we showed that the amount of young patients, diagnosed with PCa remained constant over this observation period (~4%), whereas we identified an increase for de novo metastatic disease at presentation that ranged from 2% in 2004 to 3.2% in 2013 (p = 0.004). The increase coincides with the timing of the United States Preventative Service Task Force (USPSTF) recommendation against systematic PSA screening . Most recently, two studies examined the effect of the USPSTF recommendations on prostate biopsy use patterns in men from the United States and Canada. Both reported decreasing rates of de novo PCa diagnoses after the release of the USPSTF recommendations, in May 2012 [20, 21]. In both reports, the authors noted an encouraging trend towards lower rates of de novo low-risk PCa diagnoses. However, they also noted decreased detection rates of intermediate and high-risk PCa (Gleason 7–10). This finding may imply delays in diagnosis and treatment, as suggested by the authors. Moreover, our results corroborate those of Weiner et al. , who observed increasing trends of de novo metastatic disease at diagnosis in 767,550 men within the National Cancer Data Base. Specifically, the rates of metastatic disease at diagnosis increased by 72% over a period of ten years, across all ages. These observations require follow-up. Moreover, one can speculate that progress in imaging and more frequent use of, for example multiparametric magnetic resonance imaging (MRI) might also result in higher detection rates and therefore contribute to the higher rates of de novo metastatic disease at presentation.
Furthermore, we observed decreased rates of local treatments namely RP and BT in young patients with non-metastatic disease. Conversely, NLT use increased. These findings are in direct agreement with contemporary recommendations for AS. It also coincides with recent increased awareness for potential detrimental effect of over-diagnosing and over-treating PCa patients, especially at a younger age where consequences and complications have considerable impact on quality of life.
Finally, in cumulative incidence analysis, young patients appear to have more favorable outcomes relative to their older counterparts. However, in detailed multi-variable CRR analyses adjusted for several confounders, age did not reach independent predictor status, across all treatment modalities except for NLT, where young patients exhibited a survival benefit relative to their older counterparts. This apparent survival benefit may stem from the inability to distinguish between AS, ADT, and WW. According to this argument, ADT and WW might be more prevalent in older patients and may directly contribute to higher PCSM. Instead of younger age, we confirmed known predictors of PCSM such as higher Gleason score and lymph node involvement. Regardless of age those patients should be given more attention and might represent candidates for more aggressive treatments. Our observation, showing no difference in PCSM according to age for all local treatment types, corroborates the results of smaller scale series that examined shorter term outcomes, such as BCR. For example, Becker et al.  analyzed 13,268 patients including 443 aged <50 years that underwent RP in a single European institution. In their analyses, younger age did not affect BCR rates. Similarly, Burri et al.  studied 1665 EBRT or BT plus ADT patients, of whom 378 were aged ≤60 years. Here, 5 and 8-year BCR free survival did not differ between age groups.
To the best of our knowledge, our study relies on the largest sample size of young patients diagnosed with de novo PCa, within the United States. Moreover, our study relies on advanced statistical methodology. CRR that accounts for OCM has not been previously used in any other study examining the effect of age, where OCM plays a pivotal confounding role. Moreover, our study is based on a highly comprehensive design that focuses on disease characteristics and cancer control outcomes, across multiple treatment modalities, including NLT.
However, our study is not devoid of limitations. The SEER database restricts the amount of accessible details about some variables that are available in smaller studies, such as PSA prior to 2010 or the type of NLT (AS vs. ADT vs. WW). Similarly, information on type of EBRT and its dose were unavailable. Moreover, the SEER does not provide information on adjuvant or salvage therapies, biochemical recurrence or metastatic progression. Finally, individual comorbidity profiles could not be examined among recipients of different treatments. However, OCM was used to indirectly adjust for the effect of comorbidities that may contribute.
In conclusion, young PCa patients have more favorable primary disease characteristics at presentation, are less frequently treated with RP or BT and more frequently benefit of NLT. PCSM did not differ between young and old patients. However, it is worrisome that recently more young PCa patients are diagnosed at a metastatic stage.