Age at diagnosis, cancer stage, and grade are among the most important factors used to determine the prostate cancer treatment modality such as prostatectomy, radiation, or active surveillance [8]. Since most of screen-detected prostate cancers are low risk and are unlikely to cause death, medical experts have agreed that active surveillance, a way to monitor disease periodically rather than treat it immediately, has emerged as a viable treatment option for patients with low risk prostate cancer (http://consensus.nih.gov/2011/prostate.htm).

We stratified cases by age group (<40, 40–49, 50–59, 60–69, 70–79, or 80 years or above), race (white, black, Asian/Pacific Islander (API), American Indian/Alaska Native (AIAN), or unknown), Hispanic ethnicity, US Census region of residence (Northeast, Midwest, South, and West), and cancer stage and grade (see Table 1). For cancer registry purposes, we used SEER Summary staging instead of AJCC staging system in this study. Prostate cancer stage was based on SEER Summary Stage 2000 rules (http://seer.cancer.gov/tools/ssm/) for diagnosis years 2001–2003 and on Collaborative Stage rules (http://web.facs.org/cstage/schemalist.htm) for diagnosis years 2004–2007. To characterize cancer stage at diagnosis across all study years, we then combined these two staging systems and used three categories: no invasion beyond prostatic capsule (localized stage), invasion beyond prostatic capsule to adjacent organs or regional lymph nodes (regional stage), and invasion to distant organs or distant lymph nodes (distant stage). Prostate cancer grade was determined by Gleason score. Briefly, Gleason scores 2–4 were classified as well differentiated prostate cancer, 5 and 6 as moderately differentiated, and 7 and above as poorly differentiated. More information on Gleason's system can be found at http://training.seer.cancer.gov/prostate/abstract-code-stage/morphology.html. During 2001–2003, prostate cancers with a Gleason score 7 were coded in the cancer registries as moderately differentiated; however from 2004 to 2007, these cases were coded as poorly differentiated [12]. To minimize the impact of this coding change on our estimates of incidence rates during the study period, 2001–2007, we collapsed grade categories into 2 groups: well differentiated; and moderately/poorly/undifferentiated. However, we used a standard (unclasped) cancer grade variable in the trend analysis from 2004 to 2007 (see Table 3).

Although the overall prostate cancer incidence remained stable from 2001 to 2007, there were statistically significant increases in the rate for men aged 0–39 years (APC = 6.6; 95% CI = 0.7, 12.7) and men aged 40–49 years (APC = 3.0; 95% CI = 0.6, 5.5), respectively. Statistically significant decreases in incidence were observed for men aged 70–79 years (APC = −2.3; 95% CI = −4.5, −0.1) and men aged 80 years and older (APC = −4.4; 95% CI = −5.8, −3.0), respectively. Significant declines in incidence also occurred among blacks (APC = −2.1; CI = −3.3, −0.9), AIANs (APC = −4.1; CI = −7.3, −0.7), APIs (APC = −3.2; CI = −4.8, −1.6), Hispanics (APC = −2.9; CI = −4.4, −1.4), and men residing in the West (APC = −1.8; CI = −3.3, −0.2). Incidence rates declined significantly over time for distant prostate cancer (APC = −3.1; CI = −4.4, −1.8) and well differentiated prostate cancer (APC = −26.0; CI = −27.3, −24.8), respectively, during the study years (Table 1).

As shown in Table 2, although the incidence of distant stage prostate cancer among men aged 40–49 years remained stable from 2001 to 2007, we observed statistically significant increases in the incidence of localized and regional diseases by 3.4% and 3.5% per year, respectively. Incidence of moderately/poorly/undifferentiated prostate cancer significantly increased by 3.8% per year from 2001 to 2007, accompanied by a significant decrease, 24.3% per year, among well differentiated cancer. For men aged 50–59 years, statistically significant declines in incidence occurred in both distant stage (APC = −1.7; CI = −2.7, −0.6) and well differentiated prostate cancers (APC = −24.2; CI = −27.5, −20.7). Likewise, significant declines were observed among men aged 60–69 years and men aged 70–79 years, respectively. For men aged 80 years or older, statistically significant declines in incidence occurred in each cancer stage and grade category.

Our findings that prostate cancer incidence trends differed dramatically between younger and older men are similar to a SEER study which showed that risks for being diagnosed with prostate cancer, for both blacks and whites, increased among men aged 30–49 years, but decreased among men aged 50 year and older from 2000 to 2007 [13]. It is well known that screening influences cancer incidence to some extent. The American Cancer Society recommends discussing prostate cancer screening with men who have at least a 10-year life expectancy at age 50 if they are at average risk [14]. Most major US medical organizations also recommend against prostate cancer screening among men who have a less than 10-year life expectancy [15–17]. Results form a National Health Interview Survey study showed that the prevalence of PSA screening among men aged 70 years or older was higher in 2005 than in 2000 [18]. Thus, prostate cancer screening might not account for the decreasing trend in prostate cancer incidence in this older population. Similarly, using Behavioral Risk Factor Surveillance System and NPCR/SEER data, we found discordance of prostate cancer incidence and screening among men aged 40–49 years [19]. Thus, screening alone might not explain the changes in prostate cancer incidence in certain age groups. Additional research is necessary on changes of physicians' and patients' knowledge and perspectives of prostate cancer screening, uptake of age-stratified PSA threshold for a diagnostic biopsy, and changes in environmental and behavioral risk factors of prostate cancer to better understand these discrepancies.

In 2002, Stephenson examined 1973–1997 SEER data and found that, after the introduction of PSA testing, localized and regional diseases substantially increased with a dramatic decrease in distant disease [7]. We found that from 2001 to 2007, incidence of distant stage prostate cancer declined only among men aged 50 or older, but not men aged 40–49 years. The reason for this age-specific decline is possibly because prostate cancer screening guidelines from major medical organizations during that time recommended men at average risk have a screening test at age 50. Our study demonstrated a dramatic decline in incidence of well differentiated cancers from 2001 to 2007 in each age group. One possible explanation for this decline is shift toward upgrading cancer using the Gleason Scoring system [6]. Recently, pathologists have tended to assign relatively high Gleason scores to biopsy specimens of prostate in order to align more closely with scores generated from reviews of the entire surgical specimen [20]. Another possible explanation is that, in 2005, the International Society of Urologic Pathologists (ISUP) and the World Health Organization recommended pathologists report all higher tertiary grade components of the cancer as part of a Gleason score [21].

Age at diagnosis, cancer stage, and grade are important factors in determining the prostate cancer treatment regimens that influence health outcomes. For instance, Lin et al. found that men aged 35–44 years with high-grade prostate cancer had worse overall survival and disease-specific survival compared with older men [22]. As shown in our study, over 80% of prostate cancer cases were localized stage. Treatment modalities for the localized prostate cancer markedly vary by cancer grade. According to the National Comprehensive Cancer Network (NCCN) guideline (http://www.nccn.com/cancer-guidelines.html#prostate), patients with a localized, low recurrence risk cancer (stage T1-T2a, Gleason scores ≤6, and PSA < 10ng/mL) could be managed with active surveillance or surgery or radiation alone. In particular, patients with a less than 20-year life expectancy and a very low recurrence risk prostate cancer (stage T1c), identified by needle biopsy after an elevated PSA, may be more appropriately followed without immediate intervention such as active surveillence. Our study shows that, from 2004–2007, about 56% of localized prostate cancers (stage T1 or T2) were well or moderately differentiated (Gleason scores ≤6). These cancers were more likely to be categorized as low recurrence risk if they were in stage T1or T2a (according to NCCN, stage T2b and T2c were in intermediate risk group). Using data from the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSUPE) registry, Cooperberg et al. found that only 9% of men with localized low recurrence risk prostate cancer did not undergo immediate treatment [23]. Thus, non preferability of active surveillance among localized low recurrence risk prostate cancer patients need to be further investigated.

Interestingly, we found that up to 43% of localized prostate cancers were poorly or undifferentiated and that the incidence of poorly differentiated cancer rose 8.0% and 6.1% per year from 2004 to 2007 for localized and regional prostate cancers, respectively. These increasing trends may be partially attributed to the acceptance and implementation of the 2005 ISUP recommendation, which called for reporting higher tertiary Gleason patterns and reporting different Gleason grades in multiple needle biopsies [21]. However, possible effects of new environmental or behavioral exposures to incur more aggressive prostate cancer cannot be completely ruled out. Continued monitoring of the pathological pattern in prostate cancer is greatly needed to better understand the increasing trend of the poorly differentiated cancer, especially with the US Preventive Service Task Force's draft recommendation against prostate cancer screening for men of all ages [24].

Using cancer registry data, covering over 93% of the US population enables our study to have broad generalizability. However, this study is subject to at least three limitations. First, because of changes in the grading system using Gleason score, we had to combine moderately and poorly differentiated, and undifferentiated tumors into one category to achieve a complete trend analysis from 2001 to 2007. Second, cancer registry collected the “best available grade.” Therefore, for men elect nonsurgically therapy, the Gleason score, which was based on biopsy, may be underrated compared with men treated surgically sooner after the diagnosis. However, urological pathologist tends to assign a higher Gleason score to biopsy specimen to address this issue. Our estimates of poorly differentiated cancer (42%) were conservative given the issue of underestimation of Gleason score. Third, we documented a substantial number of cancers that were unstaged or had an unknown grade. This problem diminished with time indicating better data quality in more recent years. Last, although NPCR and SEER data are the most geographically comprehensive data available, not all states were included in the analysis (from 84% in the South to 100% in the Northeast).