- To describe survival and cause of death in a nationwide cohort of Danish patients with prostate cancer undergoing radical prostatectomy (RP).
- To describe risk factors associated with prostate cancer mortality.
absolute risk reduction
prostate cancer-specific mortality
Prostate Cancer Intervention Versus Observation Trial
Scandinavian Prostate Cancer Group Study Number 4
Surgical treatment of localised prostate cancer remains controversial. The benefit of radical prostatectomy (RP) compared with watchful waiting, i.e. absolute reduction of risk of prostate cancer death, is estimated to be between zero and 25% depending on age, biopsy Gleason score and clinical T-category . PSA screening has resulted in a lead-time bias and stage migration in prostate cancer [2, 3]. Consequently, survival rates among contemporary post-PSA era RP series has increased compared with historic RP series, a phenomenon known as the Will-Rogers effect [3, 4]. Denmark has never implemented a national PSA screening programme. However, a recent sharp increase in incidence is indicative of opportunistic PSA testing [5, 6]. Still, Denmark has one of the World's largest mortality/incidence ratios and mortality from prostate cancer has remained constant for >35 years [7, 8]. Further, relative survival of prostate cancer (i.e. survival compared with an age-matched background population) has only increased slightly in the past decade and remains ≈80% lower compared with the four neighbouring Nordic countries, which indicates that prostate cancer diagnosis is not only driven by regular PSA screening . In 1995, as the last Nordic country, Denmark initiated surgical treatment for localised prostate cancer . To our knowledge, only two European countries have reported survival after RP in a nationwide cohort, both limited by missing information on preoperative parameters [11, 12].
The objective of the present study was to describe survival and causes of death in a complete nationwide cohort of Danish patients with prostate cancer undergoing RP from 1995 to 2011. We also compared our results with contemporary and historic RP series.
Between 1995 and 31 December 2011, 6489 patients underwent RP for localised prostate cancer at six hospitals in Denmark. Data collection has been approved by the Danish Data Protection Agency (file#2011-41-6926) and the National Board of Health (file#6-8011-916).
Information on the current status of all patients (dead, alive or emigrated) was retrieved from the National Danish Central Person Registry, which contains day-to-day updated information on all Danish residents. The limit for follow-up was 31 July 2012, resulting in a minimum follow-up after RP of 6 months. The duration of follow-up was calculated from the date of RP. Emigrated patients were censored at the date of emigration.
The date and cause of death were retrieved from The Danish Register of Causes of Death. Death ≤ 30 days of RP was recorded as prostate cancer death. In all cases of death, national electronic patient files, discharge letters and The National Danish Pathology Register were reviewed to confirm the cause of death.
Patients with clinically localised prostate cancer (T1–cT2) and a life-expectancy of ≥10 years at diagnosis have been offered RP. Some young patients with clinical T3 disease were offered RP and included in this analysis.
Before RP, the patients underwent staging procedures according to local and/or national guidelines to exclude metastatic disease including: CT, bone scans and/or MRI. Patients were staged according to the Union Internationale Contre le Cancer (UICC) TNM classification 
Clinical data included in the present analysis were: age, clinical T-category, preoperative PSA level and biopsy Gleason score. Most biopsies (estimated 75%) were evaluated according to the International Society of Urological Pathology (ISUP) 2005 guidelines . We did not re-evaluate the biopsy Gleason score for patients included before 2005. However, we performed sensitivity analysis to account for any time-dependency.
Most RPs were performed as open retropubic procedures. No separate analysis accounting for differences in effect depending on other types of surgery, i.e. laparoscopic or robotic were performed.
The primary endpoint was survival. Overall (OS) and cancer-specific survival (CSS) probabilities were estimated using the Kaplan–Meier method. Nelson–Aalen estimation was used to analyse the cumulative incidence of prostate cancer and any cause of death at 5, 10 and 15 years after RP. Univariate analysis for prostate cancer death was performed with the cohort stratified by the D'Amico risk classification . Multivariate analysis was done using Cox proportional hazard model to estimate the risk of prostate cancer death for the chosen covariates. Age was scored by the actual age with hazard ratio (HR) for a 10-year difference and baseline PSA level was entered on a log scale (base 2, therefore HR was for two-fold difference in PSA). Biopsy Gleason score was treated as a categorical variable and stratified into four categories, ≤6, 3 + 4, 4 + 3 and ≥8. Results are presented as HRs with 95% CIs.
Tests for interaction between covariates were performed. Sensitivity analysis was done as the patients were accrued over a substantial time span. Model assumptions for Cox proportional hazard model were assessed using Martingale and Schoenfeld residuals. A P < 0.05 was considered to indicate statistical significance.
Of the 6489 patients, 22 were excluded because of established metastatic disease (four patients) or insufficient follow-up. Thus, the final cohort consisted of 6467 patients. None of these patients were lost to follow-up. The median (range) follow-up was 4 (0.5–16) years. In all, 2293 and 411 patients were followed for more than 5 and 10 years, respectively. The median follow-up varied between institutions due to the individual institutions initiating RP in different years. In all, 33 (0.5%) patients emigrated during follow-up and these patients were censored at the date of emigration.
The patients' characteristics are listed in Table 1. The median (interquartile range, IQR) age was 64 (60–69) years. The median (IQR) preoperative PSA level was 8.8 (6.1–13) ng/mL. In 185 (2.9%) patients information on cT-category was missing. Biopsy Gleason score was not reported in 176 (2.7%). Most patients had impalpable (T1) disease (54%) and 139 patients (2.1%) were operated with clinically locally advanced disease (cT3).
|Median (range; IQR):|
|Age, years||64 (36–77; 60–68)|
|PSA level, ng/mL (n = 6432)||8.8 (0.2–201; 6.1–13)|
|Tx (missing)||185 (2.9)|
|Biopsy Gleason score|
|3 + 4||2001 (30.9)|
|4 + 3||657 (10.2)|
|Not scored||176 (2.7)|
|D'Amico risk classification:|
|Not classified||272 (4.2)|
|Number of cases per year:|
|Total number of cases per institution:|
When stratified according to the D'Amico risk classification, 19.5% patients were low-risk, 48.2% were intermediate-risk and 28.1% were high-risk. A few patients (4.2%) were not classified as a result of missing information (Table 1).
During follow-up, 328 patients died, 109 (33.2%) had prostate cancer as the cause of death, and 219 (66.8%) of other causes. Six patients (0.09%) died ≤ 30 days of RP. A complete list of causes of death is listed in Table 2.
|Cause of death||N (%)|
|Other-cause mortality||219 (66.8)|
|Stratified by organ system:|
|Urothelial carcinoma of the pelvis||3|
|Chronic kidney failure||4|
|Complications after gastrointestinal surgery||2|
|Bleeding after biopsy of the liver||1|
|Cancer of the brain||2|
|Bleeding after thrombolytic therapy for brain infarction||1|
|Amyotrophic lateral sclerosis||1|
|Restrictive lung disease||2|
|Chronic obstructive lung disease||6|
|Acute myocardial infarction||20|
|Sudden cardiac death||13|
|Ruptured aortic aneurism||1|
|Chronic lymphocytic leukaemia||1|
|Acute myeloid leukaemia||2|
|To the head||3|
|Carbon monoxide poisoning||1|
|Sepsis, unknown pathogen||5|
|Chronic alcohol abuse||6|
|Cancer of unknown primary||2|
The estimated 10- and 15-year OS was 85.6% (95% CI 83.7–87.7) and 71.6% (95% CI 59.7–83.5), respectively. The estimated 10- and 15-year CSS was 93.6% (95% CI 92.1–95.2) and 81.6% (95% CI 68.3–94.9), respectively. Stratified on D'Amico criteria, the 10-year CSS for low-risk cancers was 97.8% (95% CI 92.9–99.3), for intermediate-risk 96.4% (95% CI 94.6–97.7) and 85.7% (95% CI 80.5–89.6) for high-risk cancers, Figs 1 and 2.
Univariate analysis showed a significant impact of the D'Amico risk classification to predict PCSM (P < 0.001). Patients with high-risk disease had a significantly higher risk of dying from prostate cancer than low-risk patients (HR 5.7, 95% CI 2.7–11.8, P < 0.001), Table 3.
|Analysis||HR (95% CI)||P|
|D'Amico risk classification:|
|Multivariate analysis, Cox proportional hazard model|
|Age, per 10-year increase||1.23 (0.84–1.80)||0.29|
|PSA, for every doubling||1.35 (1.12–1.62)||<0.001|
|Biopsy Gleason score:|
|3 + 4||1.50 (0.89–2.54)||0.13|
|4 + 3||2.34 (1.25–4.39)||0.008|
In multivariate analysis, cT-category was a predictor of prostate cancer death (P < 0.001). Compared with T1 disease, both cT2c (HR 2.2) and cT3 (HR 7.2) significantly increased the risk of prostate cancer death. For every doubling of PSA the risk of prostate cancer death was increased by 34.8%. Biopsy Gleason score also influenced the risk of prostate cancer death (P = 0.003). Biopsy Gleason score 4 + 3 and ≥8 were associated with a 2.3 and 2.7 increased risk of prostate cancer death compared with biopsy Gleason score ≤ 6, respectively. There was no effect of institution on the risk of prostate cancer death (Table 3). Also, there was no differences in 30-days or overall mortality between institutions (data not shown).
The cumulative incidences of prostate cancer-specific mortality (PCSM) and all-cause mortality (Nelson–Aalen) stratified by the D'Amico classification after 5, 10 and 15 years, respectively, are listed in Table 4. The cumulative incidence of all-cause mortality after 10 years was 15.4% (95% CI 13.2–17.7) and 6.6% (95% CI 4.9–8.2) for PCSM, Fig. 2.
|% (95% CI) at:|
|5 years||10 years||15 years|
|Death from any cause:||4.9 (4.2–5.7)||15.4 (13.2–17.7)||33 (17.3–48.8)|
|Low-risk||2.8 (1.6–4)||8.4 (4.9–11.9)||19 (4.2–33.7)|
|Intermediate-risk||5.4 (4.3–6.5)||12.6 (9.9–15.3)||18.1 (12.4–23.7)|
|High-risk||6 (4.5–7.5)||21.6 (16.2–27.1)||41.9 (21.2–62.6)|
|Death from prostate cancer:||1.3 (0.9–1.9)||6.6 (4.9–8.2)||20 (4.5–35.4)|
|Low-risk||0.4 (0–0.8)||2.2 (0–4.8)||8.9 (0–22.2)|
|Intermediate-risk||0.8 (0.4–1.3)||3.6 (2.1–5.2)||5.1 (1.8–8.4)|
|High-risk||2.7 (1.6–4.3)||15.4 (10.2–20.6)||29 (9.6–48.4)|
The Danish public healthcare system offers all Danish residents free and equal access to healthcare services, including prostate cancer treatment. Only a very limited number of RPs are performed in private hospitals and these were not included in the present study. RP was introduced in Denmark (population 5.6 million) in 1995 and first performed in a single institution. Since then the number of RPs performed has increased rapidly, and RP is presently performed in six different hospitals. The present study represents a nationwide experience with RP in a population that not been subjected to systematic PSA screening, as this was never implemented in Denmark. Nonetheless, a sharpincrease in incidence indicates that opportunistic PSA testing is on-going [5, 6].
In the present cohort, ≈15% of the RPs were performed as either laparoscopic or robot-assisted procedures. Currently, no evidence suggests any differences in outcome with regard to mortality compared with open surgery, and in the absence of randomisation, no attempts to compare surgical methods was done .
Reflecting the long natural history of prostate cancer, biochemical recurrence (BCR) is the most frequently used endpoint when reporting outcome after RP . We have previously described the risk of BCR in a subset of 1200 patients included in the present cohort and found it to be similar to other surgical series . However, time to BCR per se is a poor predictor of PCSM, and other parameters, e.g. PSA doubling time, have been included for more accurate assessment of risk for PCSM .
Several publications have reported survival after RP. Selection of patients is crucial when comparing outcomes in surgical series. One of the first results of retropubic RP of 159 patients who were not PSA tested, reported a perioperative mortality of 2.5% and a 10- and 15-year CSS of 55% and 45%, respectively . The introduction of PSA testing has resulted in stage migration, which is reflected when comparing early and late surgical series. A USA multi-institutional surgical series of 2758 patients operated in the early PSA era, reported a CSS of 77–94% after 10 years depending on WHO-grade . Recently, Qi et al.  published a series of 3621 PSA-screened men and found a 10-year CSS of 98–100% for those with preoperative PSA levels of <4 ng/mL. Most (≈90%) had clinical T1 disease. In the present nationwide population of patients detected outside a population-based screening programme with a median PSA level of 8.8 ng/mL and with 43% palpable tumours, the estimated 10-year CSS of 93.6% seems to compare well with the series above.
Data presented in Table 2 confirm that mortality from causes other than prostate cancer exceeds prostate cancer deaths significantly. The National Registry of Causes of Death made it possible to retrieve ‘cause of death’ in all deceased patients in the present cohort. The distribution of causes of death is as expected in an age-matched male population, except for a remarkable high number of deaths from pancreatic cancer. In a study based on 135 713 Swedish patients with prostate cancer , pancreatic cancer as a second primary cancer along with other adenocarcinomas were registered slightly more frequently than expected .
For PCSM, RP is the only treatment proven superior to observation/watchful waiting in randomised controlled studies for patients with localised prostate cancer. The Scandinavian Prostate Cancer Group Study Number 4 (SPCG-4) from the pre- and early PSA era found a benefit of RP compared with watchful waiting . The study primarily included patients with intermediate-risk prostate cancer. The overall absolute risk reduction (ARR) of prostate cancer death was 6.1%. The estimated 15-year cumulative incidence of PCSM was 14.6% (95% CI 11.2–19.1) and 46.1% (95% CI 40.8–52) for all-cause mortality within the RP cohort. Recent in depth analysis of the SPCG-4 data suggest a strong impact of age on the ARR. In a competing risk model, the ARR varied between 0–25% depending on biopsy Gleason score, clinical T-category and age. Patients aged > 70 years had no benefit of RP, regardless of the cT-category and/or biopsy Gleason score. The results by Vickers et al.  suggest that patients with high-risk prostate cancer have the largest reduction in ARR if undergoing RP instead of watchful waiting.
The USA Prostate Cancer Intervention Versus Observation Trial (PIVOT) included patients primarily diagnosed by PSA screening . Overall, the study could not show a benefit of RP compared with observation. The cumulative incidence of PCSM was 4.4% (95% CI 2.7–7.0) and 40.9% (95% CI 36–46.1) for all-cause mortality at 12 years. However, subgroup analysis indicated that patients with D'Amico high-risk disease were likely to profit from RP with a relative reduction in the risk of PCSM of 40% (95%CI 16–100).
PCSM and survival figures in the present national cohort are very similar those reported from the treatment arms of the SPCG-4 and PIVOT trials. Interestingly, the present figures are in-between the two randomised trials, and it might be speculated that this merely is a reflection of differences in the use of PSA-driven diagnosis.
Our multivariate model showed that prostate cancer death affects patients with high PSA levels, cT-category ≥cT2c and biopsy Gleason score ≥4 + 3. We did not find an effect of age on risk of prostate cancer death. Furthermore, time-dependency was not found for any variables, including biopsy Gleason score and we did not find interactions between covariates. These findings correspond to the results of several large surgical series, including Hull et al. . There was no effect of on institution on the risk of prostate cancer death or preoperative mortality. Whether this is truly an effect of treatment equality needs further analysis. Longer follow-up is needed to confirm this finding. Annual case load and surgical experience has previously been reported to be associated with in-hospital mortality and other surgical parameters [27, 28]. However, treatment paradigms are very similar between institutions and all institutions follow national guidelines based on evidence-based medicine. Due to national registries, electronically available national patient files, and electronic communication with GPs, patients are less likely to be lost to follow-up compared with many other nations.
D'Amico et al.  included pre-treatment PSA level, cT-category, and biopsy Gleason score in their risk classification. As reported by others, we found thiis classification to be highly predictive of prostate cancer death, although the stratification originally intended to classify patients according to their risk for PSA recurrence after curatively intended therapy [19, 29]. In a series of 7591 men undergoing RP, Boorjian et al.  evaluated the D'Amico classification for 10-year CSS and found it to be 99.7%, 97% and 95% for low-, intermediate- and high-risk disease, respectively. We found CSS to be 98%, 96% and 86% at 10 years for the three risk groups. While the figures for low- and intermediate-risk patients are remarkably similar, the apparent difference for high-risk patients may be related to selection. Boorjian et al. included patients with median PSA levels of 6.5 ng/mL, no cT3 disease and biopsy Gleason score ≥ 8 in only 5% of the cases. In the present analysis the median PSA level was 8.8 ng/mL, cT3 disease were present (2.1%) and biopsy Gleason score was ≥8 in 8.4% of the cases. The available data does not allow further in-depth analysis of possible differences in selection.
The main limitation of the present study is the relatively short follow-up, which is reflected in the few events. The main implication of this is reflected in the CIs of the 10- and 15-year estimates of survival. Further, the few events, especially few prostate cancer deaths do not allow for in-depth analysis of time-dependent sampling bias between institutions, impact of surgical experience on survival, and effect of possible lead-time in prostate cancer diagnosis over time. However, these analyses are planned for the future.
In conclusion, the present study presents a survival and mortality analysis of all patients who have undergone RP for clinically localised prostate cancer in Denmark since the introduction of this treatment strategy in 1995. It is noteworthy that prostate cancer-specific survival and PCSM in this nationwide cohort of 6467 patients are similar to reported outcomes of RP in both uncontrolled series and randomised trials, even in a nation that has not recommended PSA-based screening and where mortality/incidence ratio remains as one of the highest in Europe. RP for localised prostate cancer remains controversial. Although Denmark has recommended this conservative strategy for diagnosis and treatment of localised prostate cancer throughout the past 15 years, the present findings indicate that RP is now being performed in growing numbers, probably leading to increasing over-treatment as documented in randomised trials. Longer follow-up is needed to confirm these findings.