SEARCH

SEARCH BY CITATION

Keywords:

  • competing-risks regression;
  • prostatic neoplasms/mortality;
  • prostatic neoplasms/therapy;
  • radiotherapy/statistics and numerical data;
  • United States/epidemiology

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

Objectives:  To compare the mortality outcomes of radical prostatectomy and radiotherapy as treatment modalities for patients with localized prostate cancer.

Methods:  Our cohort consisted of 68 665 patients with localized prostate cancer, treated with radical prostatectomy or radiotherapy, between 1992 and 2005. Propensity-score matching was used to minimize potential bias related to treatment assignment. Competing-risks analyses tested the effect of treatment type on cancer-specific mortality, after accounting for other-cause mortality. All analyses were stratified according to prostate cancer risk groups, baseline Charlson Comorbidity Index and age.

Results:  For patients treated with radical prostatectomy versus radiotherapy, the 10-year cancer-specific mortality rates were 1.4 versus 3.9% in low-intermediate risk prostate cancer and 6.8 versus 11.5% in high-risk prostate cancer, respectively. Rates were 2.4 versus 5.9% in patients with Charlson Comorbidity Index of 0, 2.4 versus 5.1% in patients with Charlson Comorbidity Index of 1, and 2.9 versus 5.2% in patients with Charlson Comorbidity Index of ≥2. Rates were 2.1 versus 5.0% in patients aged 65–69 years, 2.8 versus 5.5% in patients aged 70–74 years, and 2.9 versus 7.6% in patients aged 75–80 years (all P < 0.001). At multivariable analyses, radiotherapy was associated with less favorable cancer-specific mortality in all categories (all P < 0.001).

Conclusions:  Patients treated with radical prostatectomy fare substantially better than those treated with radiotherapy. Patients with high-risk prostate cancer benefit the most from radical prostatectomy. Conversely, the lowest benefit was observed in patients with low-intermediate risk prostate cancer and/or multiple comorbidities. An intermediate benefit was observed in the other examined categories.


Abbreviations & Acronyms
CaPSURE =

Cancer of the Prostate Strategic Urologic Research Endeavor

CCI =

Charlson Comorbidity Index

CSM =

cancer-specific mortality

ICD =

International Classification of Disease

NNT =

number needed to treat

OCM =

other-cause mortality

PCa =

prostate cancer

RP =

radical prostatectomy

RT =

radiotherapy

SEER =

Surveillance, Epidemiology, and End Results

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

PCa is the most common non-cutaneous malignancy in North American men, and is the second cause of cancer death in men.1 The majority of contemporary patients are treated with either RP or RT.2 However, there are no randomized data that compare the efficacy of these two treatments modalities. Furthermore, there is a scarcity of contemporary observational data that examine the association between treatment type and survival.

To address the paucity of data, we decided to test the relationship between RP versus RT and CSM in a population-based cohort of North American men. We relied on competing-risks analyses3 to estimate CSM, after accounting for OCM. Furthermore, we stratified our analyses according to PCa risk characteristics, baseline comorbidity status and age. This allows a more specific quantification and comparison of PCa risk-, comorbidity-, and age-specific CSM estimates of RP and RT.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

Data source

We relied on data from the SEER registries-Medicare insurance program linked database, which is 98% complete for case ascertainment. The SEER regions represented approximately 14% of the USA population before 2000, and 26% thereafter. The Medicare insurance program encompasses approximately 97% of USA persons aged 65 years or older. Linkage to the SEER database is complete for approximately 93% of the patients.4

Study population

Between 1992 and 2005, we identified 141 155 individuals aged ≥65 years diagnosed with non-metastatic PCa as their first malignant disease, who had both Medicare Part A and Part B claims available, and who were not enrolled in a health maintenance organization throughout the duration of the study. Patients were not included if PCa was diagnosed at autopsy or on death certificate only, or if their original or current reason for Medicare entitlement was listed as disability or Medicare status code, including disability.

Patients were also excluded if they harbored T3/T4 tumors (8556), had anaplastic or unknown grade (5657), had unknown stage (227), were aged more than 80 years at diagnosis (18 880) or had missing socioeconomic information (971). Patients were also excluded if they were treated with observation as initial management or received neoadjuvant therapy before definitive treatment. These selection criteria yielded 68 665 patients treated with either RP or RT. Those represented the focus of the study.

Variable definition

Information on patient age was obtained from the Medicare file. SEER data were used to abstract patient race (white vs black vs others), marital status (married vs unmarried), population density (metropolitan vs non metropolitan) and SEER registry. The percentage of the 2000 census tract with a 4-year college education and the median income per census tract were used as proxies for socioeconomic status. Medicare claims were used to derive patients' baseline CCI, using a previously validated algorithm.5 SEER data were also used to abstract tumor grade (Gleason score 2–4 vs 5–7 vs 8–10). Clinical extension information was used to define tumor clinical stage.6

Treatment type was identified by searching Medicare files for the appropriate ICD Ninth revision, and Healthcare Common Procedure Coding System codes during the 6 months after the date of diagnosis, as described by Wong et al.6 Initial therapy (RP vs RT), as well as adjuvant hormonal therapy status, were extracted.

Outcomes

The SEER data provides cause of death information, with cause-specific mortality information available through the end of 2007. Deaths as a result of PCa (ICD-9185.9 or ICD-10 C619) were classified as CSM. All other death events were classified as OCM.

Statistical analyses

The independent sample t-test and χ2-test were used to compare the statistical significance of differences in means and proportions, respectively.

In the first step of our analyses, we attempted to adjust for the inherent selection bias within observational data using propensity-score matching.7 Propensity scores attempt to statistically reproduce randomized trials by balancing the characteristics of different treatment groups.8 The propensity to undergo RP was calculated using a multivariable logistic regression model that was adjusted for age at diagnosis, race, marital status, annual median income quartiles, percentage of 4-year college education quartiles, CCI, population density, clinical stage, tumor grade, year of diagnosis and SEER registry. We used the nearest neighbor method matching, with a caliper width of 0.2 of the standard deviation of the logit to match cases. This methodology results in a modest residual bias and the highest precision.9 The standardized difference measure was used to test how well the radiotherapy patients matched to the RP patients.10

In the second step of our analyses, patients were stratified according to PCa risk characteristics (high-risk: clinical stage T2c or Gleason score 8–10 vs low-intermediate risk: all other patients), CCI (0 vs 1 vs≥2), and age (65–69 vs 70–74 vs 75–80 years). For each category, cumulative incidence plots were generated to graphically depict CSM and OCM rates. The Gray test was used to assess the statistical significance of differences in survival rates.11 Thereafter, the NNT was calculated.

Multivariable competing-risks regression analyses tested the effect of treatment type on CSM, after accounting for OCM.3 This type of analyses was used to avoid overestimation of CSM, as censoring due to OCM might artificially reduce the pool of individuals at risk of CSM events.12 Finally, to limit the analysis to men who received radiotherapy under relatively contemporary standards, we repeated the multivariable analyses in a subset of men who received treatment after 1998, as in a previously described methodology.13 We also carried out a sensitivity analysis to measure the potential effect of an unmeasured confounder on the relationship between treatment type and CSM.14,15

All statistical analyses were carried out using the R statistical package system (R Foundation for Statistical Computing, Vienna, Austria), with a two-sided significance level set at P < 0.05.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

Between 1992 and 2005, 22 144 (32.2%) patients were treated with RP versus 46 521 (57.8%) patients with RT (Table 1). RP and RT patients differed with respect to several characteristics. Specifically, RP patients were younger (aged 65–69 years: 52.6 vs 24.1%, P < 0.001), more frequently white (89.6 vs 86.9%, P < 0.001) and married (82.8 vs 75.5%, P < 0.001) than RT patients. RP patients had lower annual median income (4th quartile: 24.5 vs 25.2%, P < 0.001), a higher education level (4th quartile: 25.9 vs 24.6%, P < 0.001) and more often resided in non-metropolitan areas (85.5 vs 86.2%, P = 0.03). Differences with respect to clinical characteristics were also detected. Specifically, RP patients were healthier (CCI 0: 50.6 vs 43.2%, P < 0.001), but harbored more advanced stage (cT2c: 15.6 vs 11.7%, P < 0.001) and grade at presentation (Gleason score 2–5: 4.9 vs 5.5%, 6–7: 68.2 vs 67.8% and 8–10: 26.9 vs 26.7%, P < 0.001) relative to RT patients. In the most contemporary year of the study (2005), patients were treated less frequently with RP in comparison with RT (7.6 vs 10%, P < 0.001).

Table 1.  Descriptive characteristics of 68 665 patients treated with RP versus RT for PCa between 1992 and 2005 within the SEER-Medicare linked database, and 40 890 treatment type-propensity score matched patients
CharacteristicsEntire cohort (n = 68 665)Propensity score matched cohort (n = 40 890)
Radical prostatectomy n = 22 144 (32.2%)Radiotherapy n = 46 521 (67.8%) P-valueRadical prostatectomy n = 20 445 (50%)Radiotherapy n = 20 445 (50%)Standardized difference
Age (years)  <0.001   
 6511 658 (52.6)11 209 (24.1)10 143 (49.6)9 228 (45.1)
 708 614 (38.9)19 279 (41.4)8 430 (41.2)9 312 (45.5)–4.3
 751 872 (8.5)16 033 (34.5)1 872 (9.2)1 905 (9.3)–0.1
Race  <0.001   
 White19 840 (89.6)40 437 (86.9)18 230 (89.2)18 097 (88.5)
 Black1 326 (6)3 716 (8)1 281 (6.3)1 370 (6.7)–0.4
 Other978 (4.4)2 368 (5.1)934 (4.6)978 (4.8)–0.2
Marital status  <0.001   
 Married18 332 (82.8)35 103 (75.5)16 748 (81.9)16 527 (80.8)
 Unmarried3 812 (17.2)11 418 (24.5)3 697 (18.1)3 918 (19.2)–1.0
Annual median income ($)  <0.001   
 ≤37 724.005 285 (23.9)11 883 (25.5)4 982 (24.4)5 135 (25.1)
 37 724.01–49 625.005 841 (26.4)11 326 (24.3)5 367 (26.3)5 307 (26)0.2
 49 625.01–66 049.005 582 (25.2)11 583 (24.9)5 142 (25.2)5 072 (24.8)0.3
 ≥66 049.015 436 (24.5)11 729 (25.2)4 954 (24.2)4 931 (24.1)0.1
College education (%)  <0.001   
 ≤15.035 274 (23.8)11 908 (25.6)4 998 (24.4)5 047 (24.7)
 15.04–25.635 592 (25.3)11 570 (24.9)5 160 (25.2)5 244 (25.6)–0.4
 25.64-42.045 536 (25)11 619 (25)5 099 (24.9)5 077 (24.8)0.1
 ≥42.055 742 (25.9)11 424 (24.6)5 188 (25.4)5 077 (24.8)0.5
CCI  <0.001   
 011 206 (50.6)20 100 (43.2)10 154 (49.7)9 984 (48.8)
 16 552 (29.6)13 835 (29.7)6 070 (29.7)6 089 (29.8)0
 22 700 (12.2)6 866 (14.8)2 566 (12.6)2 624 (12.8)–0.2
 ≥31 686 (7.6)5 720 (12.3)1 655 (8.1)1 748 (8.5)–0.4
Population density  0.03   
 Metropolitan18 943 (85.5)40 078 (86.2)17 443 (85.3)17 343 (84.8)
 Nonmetropolitan3 201 (14.5)6 443 (13.8)3 002 (14.7)3 102 (15.2)–0.4
Clinical stage  <0.001   
 T17 421 (33.5)18 946 (40.7)6 944 (34)7 187 (35.2)
 T2a/b11 264 (50.9)22 127 (47.6)10 360 (50.7)10 342 (50.6)0
 T2c3 459 (15.6)5 448 (11.7)3 141 (15.4)2 916 (14.3)1.1
Tumor grade  0.6   
 Gleason score 216 187 (73.1)34 099 (73.3)14 911 (72.9)14 930 (73.0)
 Gleason score 85 957 (26.9)12 422 (26.7)5 534 (27.1)5 515 (27.0)–0.3
Adjuvant hormonal therapy    
 No22 144 (100)41 912 (90.1)20 445 (100)18 656 (91.2)
 Yes0 (0)4 609 (9.9)0 (0)1 789 (8.8)
Year of diagnosis  <0.001   
 19921 917 (8.7)2 673 (5.7)1 664 (8.1)1 479 (7.2)
 19931 346 (6.1)2 102 (4.5)1 186 (5.8)1 129 (5.5)0.2
 19941 261 (5.7)1 867 (4)1 118 (5.5)1 052 (5.1)0.3
 19951 372 (6.2)1 738 (3.7)1 201 (5.9)1 065 (5.2)0.6
 19961 277 (5.8)1 850 (4)1 141 (5.6)1 046 (5.1)0.4
 19971 317 (5.9)2 024 (4.4)1 151 (5.6)1 109 (5.4)0.2
 19981 091 (4.9)2 042 (4.4)1 014 (5)1 029 (5)0
 19991 159 (5.2)2 264 (4.9)1 074 (5.3)1 087 (5.3)0
 20001 854 (8.4)4 882 (10.5)1 781 (8.7)1 856 (9.1)–0.3
 20011 989 (9)5 141 (11.1)1 885 (9.2)1 989 (9.7)–0.5
 20022 056 (9.3)5 397 (11.6)1 964 (9.6)2 118 (10.4)–0.7
 20031 991 (9)4 956 (10.7)1 891 (9.2)2 004 (9.8)–0.5
 20041 835 (8.3)4 951 (10.6)1 750 (8.6)1 812 (8.9)–0.3
 20051 679 (7.6)4 634 (10)1 625 (7.9)1 670 (8.2)–0.2
SEER registry  <0.001   
 San Francesco1 256 (5.7)1 923 (4.1)1 128 (5.5)1 107 (5.4)
 Connecticut1 198 (5.4)5 023 (10.8)1 197 (5.9)1 325 (6.5)–0.6
 Detroit2 312 (10.4)6 533 (14)2 254 (11)2 467 (12.1)–1.0
 Hawaii308 (1.4)1 095 (2.4)305 (1.5)330 (1.6)–0.1
 Iowa2 177 (9.8)3 892 (8.4)2 043 (10)2 102 (10.3)–0.2
 New Mexico970 (4.4)1 352 (2.9)863 (4.2)851 (4.2)0
 Seattle2 350 (10.6)3 343 (7.2)2 119 (10.4)2 096 (10.3)0.1
 Utah1 662 (7.5)1 784 (3.8)1 432 (7)1 255 (6.1)0.8
 Atlanta798 (3.6)2 139 (4.6)775 (3.8)842 (4.1)–0.3
 San Jose672 (3)1 162 (2.5)624 (3.1)609 (3)0
 Los Angeles3 235 (14.6)3 334 (7.2)2 747 (13.4)2 244 (11)2.4
 Rural Georgia49 (0.2)137 (0.3)48 (0.2)51 (0.2)0
 Greater California2 606 (11.8)4 510 (9.7)2 413 (11.8)2 555 (12.5)–0.6
 Kentucky651 (2.9)2 293 (4.9)651 (3.2)715 (3.5)–0.3
 Louisiana979 (4.4)1 963 (4.2)925 (4.5)962 (4.7)–0.1
 New Jersey921 (4.2)6 038 (13)921 (4.5)934 (4.6)0

Propensity-score matching resulted in 40 890 matched patients (Table 1). The standardized difference between the matched RP and RT groups in all the examined covariates was less than 10%, which indicates a high degree of similarity in the distribution of prognostic variables.10,16

Amongst the matched cohort, 47.4, 43.4 and 9.2%, respectively, were aged 65–69, 70–74 and 75–80 years. CCI was 0 versus 1 versus ≥2 in 49.2 versus 29.7 versus 21.0%, respectively. The rate of high-risk PCa was 36.4%, and it was 34.4, 37.3 and 42.5% (P < 0.001) in patients aged 65–69, 70–74 and 75–80 years, respectively. The rate of high-risk PCa was 35.4, 36.7 and 38.3% (P < 0.001) in patients with a CCI of 0, 1 and ≥2, respectively. RP patients received no adjuvant hormonal therapy within 6 months from diagnosis, whereas 8.8% of RT patients received adjuvant hormonal therapy.

The 10-year CSM rates in low-intermediate PCa (Fig. 1a) were 1.4 versus 3.9% (P < 0.001) for patients treated with RP versus RT, respectively. Conversely, it was 6.8 versus 11.5% (P < 0.001) in high-risk PCa (Fig. 1b) for the same respective groups. The absolute rate differences were 2.5 and 4.7% (NNT 40 and 21, respectively). In the multivariable analyses, RT independently predicted higher CSM rates both in low-intermediate and high-risk PCa patients (Table 2).

image

Figure 1. Competing-risks plots showing CSM and OCM in patients with (a) low-intermediate risk and (b) high-risk prostate cancer treated with radical prostatectomy versus radiotherapy. Results are represented as rates with the corresponding 95% confidence intervals. inline image, Radical prostatectomy; inline image, Radiotherapy.

Download figure to PowerPoint

Table 2.  Univariable and multivariable competing risks regression models for prediction of CSM (after accounting for non-cancer related mortality) and OCM (after accounting for CSM) in 40 890 patients with organ confined prostate disease, treated with RP versus RT, between 1992 and 2005, within the SEER-Medicare linked database
Patient categoriesCSMOCM
HR (95% CI) P-valueHR (95% CI) P-value
  1. †Adjusted for age at diagnosis, race, marital status, annual median income quartiles, percentage of 4-year college education quartiles, CCI, population density, year of diagnosis, adjuvant hormonal therapy status and SEER registry. ‡Adjusted for age at diagnosis, race, marital status, annual median income quartiles, percentage of 4-year college education quartiles, population density, clinical stage, tumor grade, year of diagnosis, adjuvant hormonal therapy status and SEER registry. §Adjusted for race, marital status, annual median income quartiles, percentage of 4-year college education quartiles, CCI, population density, clinical stage, tumor grade, year of diagnosis, adjuvant hormonal therapy status and SEER registry. ¶Adjusted for age at diagnosis, race, marital status, annual median income quartiles, percentage of 4-year college education quartiles, CCI, population density, clinical stage, tumor grade, year of diagnosis, adjuvant hormonal therapy status and SEER registry. For all analyses, radical prostatectomy was considered as the reference category.

Low-intermediate risk prostate cancer2.61 (2.16–3.15)<0.0011.57 (1.49–1.66)<0.001
High risk prostate cancer1.85 (1.57–2.17)<0.0011.60 (1.48–1.73)<0.001
CCI of 02.38 (2.00–2.83)<0.0011.48 (1.38–1.59)<0.001
CCI of 11.98 (1.57–2.51)<0.0011.68 (1.55–1.82)<0.001
CCI of 2 or more1.71 (1.30–2.24)<0.0011.58 (1.46–1.71)<0.001
Patient age 65–69 years§2.56 (2.10–3.13)<0.0011.72 (1.59–1.85)<0.001
Patient age 70–74 years§1.83 (1.54–2.19)<0.0011.67 (1.56–1.77)<0.001
Patient age 75–80 years§2.09 (1.45–2.99)<0.0011.37 (1.21–1.54)<0.001
Patients operated after 19981.83 (1.43–2.22)<0.0011.67 (1.52–1.85)<0.001

After stratification of patients with a CCI of 0, 1 and ≥2, the 10-year CSM rates were 2.4 versus 5.9%, 2.4 versus 5.1% and 2.9 versus 5.2% for RP versus RT, respectively (all P < 0.001; Fig. 2). The corresponding absolute rate differences were 3.5, 2.7 and 2.3% (NNT 28, 37 and 43, respectively). In multivariable analyses, RT patients had a statistically significant higher rate of CSM across all CCI groups (Table 2).

image

Figure 2. Competing-risks plots showing CSM and OCM in prostate cancer patients with (a) CCI of 0, (b) CCI of 1 and (c) CCI of ≥2 aged (d) 65–69 years, (e) 70–74 years and (f) 75–80 years treated with radical prostatectomy versus radiotherapy. Results are represented as rates with the corresponding 95% confidence intervals; inline image, Radical prostatectomy; inline image, Radiotherapy.

Download figure to PowerPoint

Finally, in patients aged 65–69, 70–74 and 75–80 years, the 10-year CSM rates for RP versus RT were 2.1 versus 5.0%, 2.8 versus 5.5% and 2.9 versus 7.6%, respectively (all P < 0.001; Fig. 2). The corresponding absolute rate differences were 2.9, 2.7 and 4.7% (NNT 34, 37 and 21, respectively). Again, RT was an independent predictor of less favorable CSM rates across all age categories in multivariable analyses. It is noteworthy that when the analyses were restricted to patients who underwent treatment after 1998, RT patients showed a 1.8-fold higher rate of dying from PCa than their RP counterparts (P < 0.001, Table 2).

Our sensitivity analysis (Table 3) showed that the observed relationship between treatment type and CSM was so robust, and that even major differences in the prevalence of an unobserved confounder with a hazard ratio of 1.5, 2.0 or 3.0 between the two treatment groups will not alter the significant beneficial impact of RP on CSM.

Table 3.  Sensitivity analyses estimating the effect of an unmeasured confounder on the hazard ratio of cancer-specific mortality
Prevalence in radical prostatectomy patients (%)Prevalence in radiotherapy patients (%)CSM
Hazard ratio 1.50Hazard ratio 2.0Hazard ratio 3.0
10706.4 (5.7–7.3)7.9 (7.0–9.0)11.7 (10.4–13.3)
807.1 (6.3–8.0)8.9 (7.8–10.0)13.3 (11.7–15.1)
907.7 (6.8–8.8)9.8 (8.6–11.1)14.9 (13.1–16.9)
20705.1 (4.5–5.8)5.7 (5.0–6.5)7.1 (6.2–8.0)
805.6 (5.0–6.4)6.4 (5.6–7.2)8.0 (7.1–9.1)
906.1 (5.4–6.9)7.0 (6.2–8.0)9.0 (8.7–11.2)
30703.6 (3.2–4.1)4.4 (3.9–5.0)5.1 (4.5–5.7)
804.0 (3.5–4.5)5.0 (4.4–5.6)5.7 (5.1–6.5)
904.3 (3.8–4.9)5.5 (4.8–6.2)6.4 (5.7–7.3)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

We sought to quantify the benefit of RP versus RT on CSM, after accounting for OCM. Furthermore, we provided specific CSM estimates according to different PCa risk, comorbidity and age groups.

When patients were stratified according to PCa characteristics, the absolute benefit of RP on CSM at 10 years ranged from 2.5% to 4.7%, and the NNT ranged from 21 for patients harboring a high-risk PCa to 40 for patients harboring a low-intermediate risk PCa. When patients were stratified according to baseline comorbidity, the absolute CSM benefit was 3.5, 2.7 and 2.3% for patients with 0, 1 and 2 or more CCI, respectively. The corresponding NNT were 28, 37 and 43, respectively. Finally, when patients were stratified according to age, the absolute CSM benefit was 2.9, 2.7 and 4.7% for patients aged 65–69, 70–74 and 75–80 years, respectively. The corresponding NNT were 34, 37 and 21, respectively. In all subcategories, RT was associated with a higher CSM rate in multivariable analysis (Table 2).

The most substantial RP versus RT benefit on CSM was recorded in patients with high-risk PCa, no baseline comorbidity and in the oldest age category. The benefits recorded according to PCa risk and baseline comorbidity are intuitive. Few RP instead of RT need to be carried out to benefit one patient with high-risk PCa. In consequence, RT in high-risk individuals is associated with less favorable cancer control outcomes. Similarly, when the benefit was stratified according to baseline comorbidity, the fewest patients needed to be treated with RP instead of RT when healthy patients with no comorbidity were examined. These patients are more likely to succumb to PCa than individuals with several comorbidities. Conversely, patients with multiple baseline comorbidity are less likely to benefit of RP, as they frequently succumb to OCM.

The stratification according to age showed counterintuitive findings. Here, the oldest age category appeared to benefit the most of RP. Intuitively, older patients are more likely to succumb to OCM. As such, the opposite findings should have been recorded, where the NNT is the lowest in the youngest age category. The observed paradox might result from the higher proportion of individuals with high-risk PCa in the oldest age category relative to other age strata. Specifically, the proportion of high-risk patients was 34.4, 37.3 and 42.5% for the 65–69, 70–74 and 75–80 years age strata, respectively. This distribution might explain the seemingly paradoxical benefit of RP versus RT, which is actually justified and consistent with other results, based on PCa-risk characteristics stratification.

Our findings corroborated previous reports. Wong et al. showed that the protective effect of RP (HR 0.50, 95% CI 0.47–0.53) was higher than the protective effect of RT (HR 0.81, 95% CI 0.78–0.85), when both active treatment types were compared with observation.6 However, no direct comparison between RP and RT was reported in that study. Furthermore, Wong et al. examined the rate of overall mortality as an end-point.6 Given the protracted and frequently “indolent” nature of PCa, overall mortality rate might be largely affected by baseline morbidity. In consequence, it might not correctly reflect the benefit of treatment type. Recently, Cooperberg et al. examined the benefit of PCa treatment in 7538 patients included in the CaPSURE database.13 They reported that RT increases the risk of CSM by 1.5-fold in comparison with RP. However, no propensity-score adjustment for treatment assignment was carried out in their analyses. In consequence, bias related to treatment assignment might have affected their results. Furthermore, they based their analyses on the CaPSURE database, which is considered less generalizable than the SEER-Medicare. Finally, our analyses examined a considerably larger sample than the one reported by Cooperberg et al.13

The present findings have several clinical implications. Our results clearly confirm the survival benefit of RP versus RT in individuals with high-risk PCa. Second, our results also show that a survival benefit exists even in individuals with low-intermediate risk PCa. However, under these conditions, a relatively high number of RP (40) need to be carried out to benefit one patient. That being said, both active treatment modalities are associated with a comparable rate of postoperative complications and quality of life.17 In consequence, it might be argued that RP improves CSM without further deterioration of patient quality of life in comparison with RT. Third, the present findings corroborate the importance of baseline comorbidity in the treatment decision-making. Patients with two or more comorbidities showed a lower benefit of RP relative to RT. As many as 43 individuals are required to be surgically treated to benefit one patient in comparison with RT. Patients with multiple comorbidities might mainly succumb to OCM. In consequence, it can be argued that these patients might have little benefit from any type of active treatment.

The present study was not devoid of limitations. All our findings originate from observational data. Lack of randomization between RP and RT groups according to baseline characteristics might have affected the observed outcomes. However, the propensity-score matching used in our analyses reduces the chance that a treatment assignment bias attributes to our results. This was further demonstrated by the robustness of our sensitivity analyses. It can be argued that not all of our patients received RT of contemporary standards. Indeed, there have been several technological developments in the field of radiotherapy for PCa in the recent years. For example, the development of image-guided radiotherapy, which allows imaging and tracking the prostate during a treatment course, has led the way toward the development of more accurate treatment methods including dose-guided and adaptive strategies. To account for this limitation, we carried out a subanalysis that was restricted to patients who were operated on after 1998, as in previously reported methodology.13 This analysis confirmed the beneficial effect of RP on CSM in comparison with RT. Nonetheless, it is possible that the most contemporary techniques of RT will result in better cancer control outcomes in most contemporary patients. This might warrant examination in future studies. In the SEER registries, exclusive clinical stage data were available only for patients diagnosed with PCa from 1994 and onwards. Between 1992 and 1993, pathological stage rather than clinical stage was recorded when it was higher. In consequence, RP patients in these 2 years might have had their pathological stage recorded rather than their clinical stage. This limitation was shared by several previous reports.6,18 Furthermore, our subanalysis that focused exclusively on patients operated on after 1998 showed no significant changes in the results. The observed difference in OCM between RP and RT patients might be attributed to a residual bias in patient selection that was not accounted for adequately by adjusting for age and CCI (comorbidity). However, these factors might have an impact on OCM, but not on CSM, which represent the main outcome of the current report. Prostate-specific antigen and the number of positive cores were not available for the majority of patients. In consequence, it was not possible to further stratify patients according to these variables. Instead, we stratified patients according to tumor stage and grade, which are considered important proxies of disease extent. Last, but not least, unfortunately we had no information on the quality of life of the patients included in this cohort.

The present results showed that the 10-year CSM rates were more favorable in patients treated with RP in comparison with those treated with RT, even after adjusting for inherent bias in treatment assignment. Individuals with high-risk PCa benefit the most from RP. Conversely, the lowest benefit was observed in patients with low-intermediate risk PCa and/or multiple comorbidities. An intermediate benefit was observed in the other examined categories.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References

Pierre I Karakiewicz is partially supported by the University of Montreal Health Centre Urology Specialists, Fonds de la Recherche en Sante du Quebec, the University of Montreal Department of Surgery and the University of Montreal Health Centre (CHUM) Foundation.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Conflict of interest
  9. References