European Urology

1. Introduction

Outcomes for men with prostate cancer (PCa) vary markedly. Men are increasingly diagnosed with low-risk disease, which may have little or no impact on life expectancy [1], [2], and [3]. Therefore, a man's risk of dying of other causes is an important consideration in PCa management.

Randomized trials evaluating PCa treatment are limited [4] and [5], leaving questions about treatment benefits and overtreatment risks. Observational studies have made major contributions to strategies for PCa risk stratification and treatment selection, but have limitations. Analyses from single centers typically include in-depth tumor characteristics, but mostly use biochemical recurrence as the end point, a relatively poor surrogate for PCa mortality [6] and [7]. A landmark study in Sweden of 223 men diagnosed with highly and intermediately differentiated localized PCa who received no curative therapy demonstrated that the majority of these men do not die of PCa [8]. However, because no data on serum levels of prostate-specific antigen (PSA) and Gleason score (GS) were available, the implications on current clinical practice are unclear.

Several studies have used register-based data from the linked Surveillance Epidemiology and End Results (SEER) study and US Medicare database to estimate mortality, specifically among men with localized PCa following conservative management [9] and [10], and according to comorbidity status [11]. However, contemporary risk categorization cannot be applied to SEER–Medicare linked data because PSA levels were not registered until recently and GS is aggregated. Moreover, SEER–Medicare linked data include only men >65 yr. A previous analysis of a subgroup of the present study population reported a 3% risk of PCa death at 10 yr after diagnosis among men with low- and intermediate-risk PCa who received surveillance as primary treatment [12].

To provide a broad depiction of long-term outcomes and better characterize men most likely to benefit from treatment and men most at risk of overtreatment, we now provide 15-yr mortality estimates for PCa death, cardiovascular death, and other causes of death among men with PCa managed noncuratively in a nationwide population-based cohort.

2. Methods

3. Results

In total, 117 328 men with PCa were registered in NPCR from 1991 through 2009 (Table 1). Median follow-up time was 4.4 yr (range: <0.1–20.0 yr). From 1991 through 2009, low-risk PCa increased and metastatic disease decreased substantially (Supplemental Fig. 1). The majority of men with PCa (78.6%) had no comorbidity according to CCI at diagnosis; the proportion of men having CCI score 0 was highest in the low-risk category and lowest in the metastases category (85.9% vs 72.9%; p < 0.001). To determine if variation in CCI score was due to PCa disease or confounded by age (median age between cases with low-risk disease and metastatic disease differed by 8 yr), we compared the proportions of CCI score 0 among PCa cases and matched PCa-free comparison men within risk categories. The proportions of comparison men with no comorbidities (84.2% of in the low-risk group and 75.1% of in the distant metastases group) were similar to cases within each risk category, indicating that the variability was likely age related.

More than half of the cohort was treated with noncurative intent. In general, noncuratively treated men were older and had higher CCI score. The proportion of men with localized disease treated with curative intent was similar among low- and intermediate-risk patients (48.0% and 47.6%, respectively), but lower among men with high-risk disease (27.1%). Overall, management with curative intent increased markedly over time (7.0% in 1991 vs 43.7% in 2009).

3.1. Cumulative mortality by risk and age category among noncuratively treated men

We restricted subsequent analyses to noncuratively treated men (N = 76 437). Mortality rates from PCa, cardiovascular disease, and other diseases varied almost 10-fold according to risk category and age (Fig. 1). We observed continuous increases in PCa mortality rates beyond 10 yr of follow up among men with localized PCa (Table 2). At 15 yr after diagnosis, the cumulative risk of PCa death was 8.9% (95% confidence interval [CI], 7.4–10.5) for low-risk disease; 19.6% (95% CI, 18.0–21.2) for intermediate-risk disease; 35.5% (95% CI, 34.2.2–36.8) for high-risk disease; 49.1% (95% CI, 47.1–51.2) for regionally metastatic disease; and 69.5% (95% CI, 68.2–70.7) for distant metastatic PCa.

Fig. 1 Cumulative mortality from prostate cancer, cardiovascular disease, and other causes according to age and risk category. Estimates are based on men in the National Prostate Cancer Register from 1991 through 2009. Included mortality events occurred from 2006 through 2010. A modified version of risk categories according to the National Comprehensive Cancer Network was used: low risk = T1–T2 tumor, prostate-specific antigen (PSA) level <10 ng/ml and Gleason score (GS) ≤6; intermediate risk = T1–T2 tumor and PSA level 10–<20 ng/ml or GS 7; high risk = T3 tumor or PSA level 20–<50 ng/ml or GS ≥8; regional metastases = T4 or N1 tumor or PSA level 50–100 ng/ml; and distant metastases = M1 tumor or PSA level ≥ 100 ng/ml.

Table 2 Percent cumulative 10- and 15-yr mortality (95% confidence interval) by age and prostate cancer stage

Disease-risk category by age group	10-yr mortality		15-yr mortality		p value1
	Prostate cancer	Other	Prostate cancer	Other	Prostate cancer	Other
All ages
Low risk	4.5 (3.8–5.2)	29.0 (27.5–30.5)	8.9 (7.4–10.5)	49.5 (46.5–52.4)
Intermediate risk	13.0 (11.9–14.0)	42.4 (40.9–43.9)	19.6 (18.0–21.2)	58.1 (56.0–60.2)
High risk	28.8 (27.7–29.8)	44.9 (43.7–46.0)	35.5 (34.2–36.8)	55.4 (53.9–56.8)	<0.001	<0.001
Regionally metastatic	41.3 (39.5–43.1)	37.5 (35.7–39.3)	49.1 (47.1–51.2)	43.5 (41.5–45.5)
Distant metastases	66.1 (64.9–67.3)	25.1 (24.0–26.3)	69.5 (68.2–70.7)	27.7 (26.5–28.9)
Age <65 yr
Low risk	1.2 (0.6–2.1)	11.0 (9.0–13.3)	5.5 (2.7–9.8)	20.6 (15.7–26.0)
Intermediate risk	8.3 (5.4–12.1)	17.0 (12.9–21.5)	17.4 (11.4–24.4)	22.6 (17.0–28.8)
High risk	30.8 (25.6–36.3)	17.6 (13.4–22.3)	40.2 (33.6–46.6)	27.7 (20.6–35.3)	<0.001	<0.001
Regionally metastatic	50.9 (45.6–55.9)	13.2 (9.8–17.0)	66.5 (59.5–72.7)	17.8 (13.1–23.1)
Distant metastases	76.1 (73.0–78.9)	9.9 (8.0–12.1)	81.0 (78.0–83.7)	11.1 (8.9–13.6)
Age 65–75 yr
Low risk	4.5 (3.6–5.5)	23.2 (21.3–25.2)	9.6 (7.4–12.1)	45.1 (41.0–49.2)
Intermediate risk	11.7 (10.3–13.2)	31.6 (29.5–33.8)	19.9 (17.6–22.4)	48.8 (45.4–52.0)
High risk	27.8 (26.0–29.6)	33.2 (31.3–35.1)	38.7 (36.3–41.0)	46.0 (43.5–48.5)	<0.001	<0.001
Regionally metastatic	42.1 (39.1–45.1)	23.9 (21.3–26.6)	54.7 (51.0–58.1)	31.2 (27.9–34.5)
Distant metastases	67.3 (65.2–69.4)	20.1 (18.3–22.0)	72.7 (70.6–74.8)	22.6 (20.7–24.5)
Age >75 yr
Low risk	7.2 (5.7–8.9)	55.5 (52.3–58.6)	10.1 (7.9–12.5)	78.0 (73.5–81.9)
Intermediate risk	14.9 (13.3–16.5)	56.3 (54.1–58.5)	19.6 (17.4–21.9)	74.4 (71.4–77.1)
High risk	29.4 (28.0–30.7)	53.4 (51.9–54.9)	33.4 (31.9–35.0)	63.4 (61.6–65.1)	<0.001	<0.001
Regionally metastatic	38.5 (36.1–40.9)	51.6 (49.1–54.1)	41.8 (39.2–44.4)	56.8 (54.1–59.3)
Distant metastases	62.7 (60.9–64.4)	32.2 (30.5–33.9)	64.3 (62.5–66.0)	34.9 (33.1–36.6)

¹p value for the null hypothesis of no difference between risk groups was calculated using Fine and Gray [21] competing-risk regression models.

The cumulative risk of death from other causes at 10 and 15 yr increased strongly with age (Table 2). Competing causes of death were distributed equally among cardiovascular deaths and all other causes of death regardless of risk group. Considerable increases in death from other causes occurred between 10 and 15 yr after PCa diagnosis except among those with distant metastases. The risk of dying of other causes for men with low-risk PCa was 29.0% (95% CI, 27.5–30.5) at 10 yr and 49.5% (95% CI, 46.5–52.4) at 15 yr. These relative increases attenuated with increases in risk category and age at diagnosis.

3.2. Cumulative mortality, competing causes of death, and comorbidity among noncuratively treated men

PCa mortality rates among noncuratively treated men with low- and intermediate-risk disease were similar regardless of comorbidity (Fig. 2). Comorbidity was a strong predictor of death from causes other than PCa, especially for the youngest men. Among men aged <65 yr, the subhazard ratios for other causes of death comparing CCI score ≥2 versus CCI score 0 were 6.1 (95% CI, 3.7–9.9) in the low-risk group, 9.3 (95% CI, 4.9–17.6) in the intermediate-risk group, and 5.1 (95% CI, 2.6–10.1) in the high-risk group. Mortality from other causes comparing CCI score ≥2 versus CCI score 0 were approximately three-fold greater for men aged 65–75 yr and approximately two-fold greater for men aged ≥75 yr, regardless of designation in the low-, intermediate-, or high-risk category (Supplemental Table 1).

Fig. 2 Cumulative 10-yr mortality from prostate cancer and other causes (95% confidence interval) by risk category, age, and Charlson Comorbidity Index (CCI). Estimates are based on men in the National Prostate Cancer Register from 1991 through 2009.

Given that comorbidities influence all-cause mortality, we restricted our investigation of observed versus expected mortality to PCa cases and the matched comparison cohort with CCI score 0 (Fig. 3). For men with low-risk PCa, the expected and observed all-cause mortality plots were superimposed during the entire 15-yr follow-up. Some difference was observed for intermediate-risk PCa after approximately 5 yr of follow up, and for men with high-risk PCa and metastatic PCa, the curves clearly diverged soon after diagnosis.

Fig. 3 Observed versus expected all-cause mortality overall and according to risk category. Analyses were restricted to cases and men in the comparison cohort with Charlson Comorbidity Index (CCI) score of 0.

4. Discussion

Our results from a large, contemporary, nationwide cohort in Sweden demonstrate the large variation in 15-yr PCa mortality for noncuratively treated men, ranging from 9% among men with low-risk PCa to 70% among men with distant metastases. There was a substantial impact of competing causes of death, particularly among men with localized PCa, and this was strongly related to existing comorbidities. Cardiovascular disease was responsible for approximately half of the non–PCa deaths regardless of risk group. Due to the low PCa mortality and because most noncuratively treated men with low-risk (80%) and intermediate-risk PCa (74%) had no major comorbidities, there was no difference between observed and expected cumulative all-cause mortality in low-risk disease, and only minor differences in intermediate-risk PCa. We note that men selected for noncurative treatment are generally less healthy than men pursuing curative therapy [12] and [22].

Two recent studies using SEER and SEER–Medicare linked databases allow us to assess generalizability in the context of US data. Our risk estimates are in the same range as in these US studies, even though risk categorization and the statistical approach were somewhat different. For instance, in a recent study by Albertsen et al. [11], PCa-specific mortality at 10 yr for men aged 66–74 yr treated conservatively with T1c/Gleason 5–7 tumors was 5% for CCI score 0, 2% for CCI score 1, and 5% for CCI score ≥2. PCa-specific mortality for men aged 65–74 yr in our low-risk category was 4% for CCI score 0, 6% for CCI score 1, and 4% for CCI score ≥2. For older men (≥75 yr) with Gleason 8–10 tumors in the Albertsen et al. study, PCa-specific mortality for T1c and T2 tumors, respectively, was 28% and 28% for CCI score 0, 24% and 20% for CCI score 1, and 19% and 16% for CCI score ≥2. For men aged >75 yr in our high-risk group, PCa-specific mortality across comorbidity categories was 31%, 27%, and 23% for CCI scores 0, 1, and ≥2, respectively. Overall cumulative mortality at 10 yr among the older high-risk men was similar in the cohorts.

Abdollah et al. [23] compared, among men of all ages with localized disease, PCa-specific mortality according to low/intermediate- versus high-risk category and management by observation, radiotherapy, or radical prostatectomy in the SEER database. Among younger men managed by surveillance, 10-yr PCa mortality in the Abdollah et al. study was comparable to our results in the low/intermediate-risk category (4% among men aged ≤59 yr and 5% among men aged 60–69 yr vs 3% for men <65 yr old in our cohort) and high-risk category (22% among men aged <59 yr and 60–69 yr vs 30.8% among men aged <65 yr in our cohort).

More men with localized PCa die from causes of death other than PCa, highlighting that PCa mortality cannot be interpreted in isolation. Although subgroup analyses from randomized trials should be interpreted cautiously, the benefit of surgical treatment in the Scandinavian Prostate Cancer Group 4 study of radical prostatectomy versus watchful waiting was confined to men aged <65 yr [4]. These data and our results indicate that achieving a balance between the benefits of improved survival and costs of adverse effects are especially pertinent for younger men, who have the most life-years to gain from treatment, but the most to lose in terms of quality of life.

Our study has several strengths: It was population-based and included 76 437 PCa cases treated noncuratively with data on GS, clinical stage, and serum PSA level. Follow-up was long and virtually complete for the entire cohort, and period analysis was used to base estimates on the most recent years. Limitations include the potential for bias in attribution of death following a cancer diagnosis, especially for more advanced disease. However, previous investigations have revealed the Swedish Cause of Death Register to be in good agreement with the cause of death as determined by chart reviews [24] and [25]. Some men had inadequate data for risk-category assignment and no data on tumor extent on biopsy were available. GS assignment was based on routine clinical evaluation in each participating hospital and the criteria for evaluating GS have changed over time. Thus, some risk-category misclassification may result from GS assignment.

5. Conclusions

There is a wide range in PCa outcomes according to risk categories. PCa mortality represents a relatively modest contributor to all-cause mortality among men diagnosed with localized low- and intermediate-risk PCa. In contrast, death rates from PCa are much higher in men with localized high-risk disease and, in particular, in with men with metastatic disease. A majority of men diagnosed with PCa do not have major comorbidity, but men who do have comorbidities have a much higher risk of death from cardiovascular disease and other causes.

Author contributions: Jennifer R. Rider had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Stattin.

Acquisition of data: Stattin, Wiklund, Andrén, Hugosson.

Analysis and interpretation of data: Rider, Sandin, Stattin.

Drafting of the manuscript: Rider.

Critical revision of the manuscript for important intellectual content: Andrén, Hugosson, Wiklund.

Statistical analysis: Sandin.

Obtaining funding: Stattin.

Administrative, technical, or material support: Stattin.

Supervision: Stattin.

Other (specify): None.

Financial disclosures: Jennifer R. Rider certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.

Funding/Support and role of the sponsor: This study was supported by the Swedish Research Council grant 825-2008-5910; Swedish Cancer Foundation grant 11 0471; Västerbotten County Council; and Lion's Cancer Research Foundation at Umeå University.