Abstract
Introduction
Limited evidence exists on the comparative effectiveness of local treatments for prostate cancer (PCa) due to the lack of generalizability. Using granular national data, we sought to examine the association between radical prostatectomy (RP) and intensity-modulated radiation therapy (IMRT) treatment and survival.
Methods
Records were abstracted for localized PCa cases diagnosed in 2004 across seven state registries to identify patients undergoing RP (n=3019) or IMRT (n=667). Comorbidity was assessed by the Adult Comorbidity Evaluation-27 (ACE-27). Propensity score matching (PSM) was used to balance covariates between treatment groups. All-cause and PCa-specific mortality were primary endpoints. A subgroup analysis of patients with high-risk PCa (RP, n=89; IMRT, n=95) was conducted.
Results
Following PSM, matched patients (n=502 pairs) treated with either RP or IMRT were well-balanced with respect to covariates. With a median followup of 10.5 years (interquartile range [IQR] 9.9–11.0), the 11-year overall survival (OS) was 71.2% (95% confidence interval [CI] 66.9–75.8) for RP and 62.3% (95% CI 57.4–67.6) for IMRT. IMRT was associated with a 41% increased risk of all-cause mortality (hazard ratio [HR] 1.41, 95% CI 1.13–1.76) but not PCa-specific mortality (HR 1.75, 95% CI 0.84–3.64), as compared to RP. In patients with high-risk PCa, IMRT, as compared to RP, was not associated with a statistically significant difference in all-cause (HR 1.53, 95% CI 0.97–2.42) or PCa-specific mortality (HR 1.92, 95% CI 0.69–5.36).
Conclusions
Despite a low mortality rate at 10 years and possible residual confounding, we found a significantly increased risk of all-cause mortality but no PCa-specific mortality associated with IMRT as compared to RP in this population-based study.
Introduction
Radical prostatectomy (RP) and radiation therapy (RT) represent the most widely used and gold-standard definitive therapies for clinically localized prostate cancer (PCa). Randomized controlled trials comparing therapies for PCa are limited due to difficulties in patient accrual, leading to issues with generalizability and insufficient power to detect differences in mortality.1–4 Moreover, the evolution of treatment technology and practice patterns outpace maturation of trial data, further complicating the generalizability of results because the tested treatment approaches may be obsolete prior to publication.
In the absence of contemporary randomized controlled trials comparing the efficacy of RP and modern RT, comparative effectiveness studies using observational data provide the next line of evidence.5 This approach can be difficult since medical comorbidity is often unbalanced between patients receiving RT and those undergoing RP, with younger and healthier patients more likely to receive RP.6 Previous observational studies using population-level data have shown RP is associated with improved survival compared to RT; however, these studies had variable followup, and many did not employ granular or detailed measures of comorbidity.7–10 Importantly, RT has evolved since these studies were published and the impact of newer modalities remains unclear.
The Breast and Prostate Cancer Data Quality and Patterns of Care (PoC-BP) study, supported by the Center for Disease Control (CDC), facilitates examination of patterns of care for PCa treatment and the comparative effectiveness of RT and RP. This population-based cohort represents patients treated with contemporary external beam RT (EBRT) in the form of intensity-modulated RT (IMRT) as compared to earlier forms of RT examined in past observational studies.10 Using granular data, including detailed patient comorbidity information captured by the Adult Comorbidity Evaluation-27 (ACE-27)11 in the PoC-BP study, we sought to examine the association between RP and RT treatment and survival using a propensity score-matched analysis.
Methods
Study cohort
We used the CDC PoC-BP dataset of 8229 men diagnosed with histologically confirmed localized PCa in 2004 from seven state cancer registries funded by the CDC’s National Program of Cancer Registries. The sampling methods employed have been previously described and included data abstraction from hospitals and non-hospitals (e.g. office and radiation facility) between 2007 and 2009.12 State cancer registries provided vital status, date of death or last date of contact, and ICD codes for cause of death as of December 31, 2015.
Analysis was restricted to patients receiving definitive treatment within six months of diagnosis (57% of the cohort), as those receiving treatment after six months likely represents management with active surveillance. Further, we excluded patients treated with conservative therapies (active surveillance, expectant management, or primary androgen deprivation therapy [ADT]), combination brachytherapy ± IMRT, other forms of EBRT, ablation therapy, and other non-radical extirpative treatments. We identified 3019 patients who underwent RP (open and minimally invasive approaches, which have demonstrably comparable oncological outcomes13) and 667 patients treated with IMRT.
Patient, provider covariates
Patient demographics included cancer registry, age at diagnosis, race/ethnicity, marital status, socioeconomic status, insurance primary payer, and urban/rural location. Socioeconomic status and level of urbanization categorizations were based on patient’s residence at diagnosis based on the 2000 U.S. census tract-specific data. Patient comorbidity was measured using the ACE-27 instrument, a chart-based comorbidity index validated in oncological outcomes research that grades the severity of multiple medical conditions (none, mild, moderate, or severe) with regard to how activities of daily living (ADL) are impacted, and algorithmically creates an index score for comparing degree of comorbidity in patients.11 Conditions controlled with medications that do not limit ADLs and have not led to hospitalization are mild. Moderate conditions limit ADLs or require hospitalization or surgery. Severe comorbid conditions denote major complications, end-organ damage, uncontrollable symptoms, or debility requiring full ADL support. The ACE-27 comorbidity index has the most prognostic impact in patients with a high likelihood of cancer survival.14
Cancer clinical characteristics included Gleason score, serum prostate specific antigen (PSA) level, National Comprehensive Cancer Network (NCCN) risk group, and whether ADT was received. Practice and provider characteristics included physician medical school graduation year categorized by decade 1950–1990, practice type (solo vs. group), for-profit vs. non-profit, teaching status of facility, distance to treatment facility, and number of urologists per 100 000 men. We have previously described these provider variables, which are associated with the selection of RP or RT, and thus represent important variables to be included to reduce confounding.12
Prespecified subgroup analysis
We performed a prespecified subgroup analysis in patients with NCCN high-risk PCa treated with RP (n=89) and IMRT (n=95). This prespecified subgroup analysis was conducted given that prior studies have demonstrated heterogeneity in treatment effect with high-risk patients deriving greater survival benefit from surgery.10 NCCN high-risk PCa is defined as clinical ≥T3a (American Joint Committee on Cancer Clinical Staging System, 6th edition), serum PSA >20 ng/ml, or Gleason grade group 4 or 5 based on transrectal ultrasound-guided prostate biopsy.
Statistical analysis
The primary endpoints were overall survival (OS) and PCa-specific survival (CSS). Propensity score matching (PSM) was performed using a multivariable logistic regression model to predict RP treatment vs. IMRT with covariates using 1:1 nearest-neighbor matching with a greedy algorithm (caliper 0.2 × propensity-score).15 Standardized differences were within 0.1 between patients in RP and IMRT groups after PSM, ensuring well-balanced groups. PSM reduces selection bias by balancing covariates between groups16 and has previously been employed to balance baseline covariates between treatment groups in clinical studies.15 Any covariate unbalanced after PSM was adjusted for in our Cox proportional hazard models. Kaplan-Meier (KM) survival rates, with corresponding 95% confidence intervals (CI), were estimated from the propensity score-matched samples. Median followup was calculated using the reverse KM method. The reverse KM method to calculate median survival is the same calculation of KM with the event and censor indicator status switched. All-cause mortality was calculated from date of diagnosis to date of death or date of last followup. For PCa-specific mortality, deaths were identified from the ICD codes for PCa (C61, C619).
Marginal proportional hazard models were constructed to determine the association between OS or CSS and treatment modality, accounting for event clustering since matching inherently violates independence.17 The cumulative incidence function was used to estimate absolute risks with PCa and non-PCa deaths as competing risks. Cause-specific hazard models estimated relative effects of treatment in the competing risk setting. A Fine-Gray test that accounted for clustering was used to test for differences between cumulative incidence functions.18 Statistical significance was set at p<0.05 based on a two-tailed comparison. All analyses were performed using SAS Enterprise Guide 9.4 (SAS Institute Inc., Cary, NC, U.S.) and R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).
Results
A total of 3686 patients were analyzed, with 3019 receiving RP and 667 receiving IMRT (Table 1). Surgical patients in this unmatched cohort were more likely to be younger, with 45.5% of patients aged less than 60 years at the time of diagnosis compared to 58.8% being 70 years or older in the IMRT patients. Most patients were White (77.2% and 72.9% for surgery and IMRT, respectively) and married (79.7% and 75.4%, respectively). Patients who underwent RP were significantly more likely to be privately insured (74.0% compared to 45.5% in the IMRT patients). This is likely related to many RP patients being below the age to qualify for Medicare; 41.0% of patients receiving IMRT were covered by Medicare, compared with 15.5% of RP patients. Rural and urban locale of patients did not vary significantly between treatment groups. RP patients were healthier, with many having no comorbidities (42.0% vs. 26.7% for IMRT patients). Overall, patients were mostly healthy in both treatment groups, with 87.8% and 81.3% for RP and IMRT, respectively, having mild or no ACE-27 score. IMRT patients had higher PSA, and were also more likely to have received ADT (52.4% IMRT vs. 6.8% RP) and be in the NCCN high-risk group (19.7% IMRT vs. 12.6% RP). Practice type and facility ownership did not vary appreciably. IMRT patients tended to have a shorter distance to the treatment facility, while there was also a trend for surgery patients to have their care at teaching hospitals.
Table 1.
Sociodemographic, clinical, and provider characteristics of surgery and IMRT treatment groups with standardized differences before and after propensity score matching
| Before propensity score matching | After propensity score matching | |||||||
|---|---|---|---|---|---|---|---|---|
|
| ||||||||
| Variable | Total (%)a | Patients (n)b | Surgery n (%)c |
IMRT n (%)c |
Standardized difference | Surgery n (%)c |
IMRT n (%)c |
Standardized difference |
|
| ||||||||
| Total | 100 | 3686 (11 431) | 3019 (83.9) | 667 (16.1) | 502 (50.7) | 502 (49.3) | ||
| Registry | 0.45 | 0.14 | ||||||
| A | 23.4 | 645 (2675) | 558 (24.0) | 87 (20.2) | 81 (22.1) | 74 (22.5) | ||
| B | 14.1 | 674 (1613) | 462 (12.0) | 212 (25.2) | 139 (22.1) | 128 (19.8) | ||
| C | 7.8 | 242 (892) | 224 (8.6) | 18 (3.7) | 12 (3.1) | 15 (4.0) | ||
| D | 8.6 | 651 (986) | 537 (8.5) | 114 (9.3) | 94 (9.0) | 91 (9.5) | ||
| E | 13.9 | 485 (1588) | 402 (13.9) | 83 (13.8) | 51 (12.0) | 60 (12.9) | ||
| F | 18.8 | 456 (2151) | 377 (19.2) | 79 (16.8) | 76 (22.8) | 68 (18.3) | ||
| G | 13.4 | 533 (1526) | 459 (13.8) | 74 (11.1) | 49 (9.1) | 66 (13.0) | ||
| Age | 1.27 | 0.06 | ||||||
| <60 | 39.6 | 1471 (4528) | 1405 (45.5) | 66 (8.8) | 50 (9.8) | 66 (11.3) | ||
| 60–64 | 22.2 | 816 (2538) | 719 (24.0) | 97 (13.1) | 101 (18.8) | 96 (16.6) | ||
| 65–69 | 19.9 | 754 (2269) | 606 (20.0) | 148 (19.3) | 180 (33.3) | 132 (22.6) | ||
| 70–74 | 12.4 | 431 (1412) | 246 (8.8) | 185 (30.7) | 139 (30.1) | 141 (33.3) | ||
| 75+ | 6.0 | 214 (684) | 43 (1.7) | 171 (28.1) | 32 (8.0) | 67 (16.1) | ||
| Race/ethnicity | 0.13 | 0.09 | ||||||
| Caucasian | 76.5 | 2278 (8749) | 1887 (77.2) | 391 (72.9) | 276 (72.5) | 296 (72.8) | ||
| African American | 15.0 | 979 (1711) | 771 (14.2) | 208 (18.9) | 170 (19.4) | 152 (18.3) | ||
| Hispanic | 6.0 | 239 (683) | 204 (6.1) | 35 (5.4) | 29 (5.4) | 30 (6.2) | ||
| API/AI/AN | 2.5 | 190 (288) | 157 (2.5) | 33 (2.9) | 27 (2.7) | 24 (2.7) | ||
| Marital status | 0.20 | 0.04 | ||||||
| Married | 79.0 | 2873 (9030) | 2386 (79.7) | 487 (75.4) | 373 (76.4) | 381 (78.3) | ||
| Single/divorced/separated/ | ||||||||
| widowed | 16.9 | 671 (1928) | 508 (15.8) | 163 (22.5) | 117 (20.6) | 109 (19.7) | ||
| Unknown | 4.1 | 142 (474) | 125 (4.5) | 17 (2.1) | 12 (3.0) | 12 (2.0) | ||
| Socioeconomic status | 0.18 | 0.04 | ||||||
| Low | 11.4 | 645 (1298) | 496 (10.7) | 149 (14.7) | 105 (13.8) | 97 (13.0) | ||
| Mid | 17.0 | 680 (1941) | 543 (16.5) | 137 (19.7) | 107 (18.6) | 110 (20.5) | ||
| High | 71.6 | 2,350 (8165) | 1971 (72.7) | 379 (65.6) | 290 (67.6) | 295 (66.5) | ||
| Insurance | 0.68 | 0.06 | ||||||
| None | 1.5 | 64 (176) | 53 (1.6) | 11 (1.1) | 8 (0.8) | 9 (1.0) | ||
| Medicaid | 4.3 | 204 (492) | 150 (4.1) | 54 (5.6) | 35 (5.7) | 42 (5.8) | ||
| Medicare or other public | 19.6 | 765 (2237) | 486 (15.5) | 279 (41.0) | 187 (37.0) | 184 (37.3) | ||
| Private | 69.4 | 2503 (7935) | 2218 (74.0) | 285 (45.5) | 247 (49.5) | 245 (50.9) | ||
| Unknown | 5.2 | 150 (591) | 112 (4.9) | 38 (6.8) | 25 (6.9) | 22 (5.0) | ||
| Urban/rural | 0.07 | 0.04 | ||||||
| Rural | 12.5 | 522 (1430) | 417 (12.2) | 105 (14.2) | 78 (12.5) | 75 (13.2) | ||
| Urban | 50.6 | 1813 (5784) | 1502 (51.1) | 311 (48.1) | 246 (52.3) | 240 (49.2) | ||
| Rural-urban mix | 36.7 | 1340 (4191) | 1091 (36.5) | 249 (37.4) | 178 (35.2) | 187 (37.6) | ||
| Unknown | 0.2 | 11 (27) | 9 (0.2) | 2 (0.3) | ||||
| ACE-27 comorbidity score | 0.36 | 0.09 | ||||||
| None | 39.5 | 1367 (4514) | 1208 (42.0) | 159 (26.7) | 114 (22.5) | 133 (29.4) | ||
| Mild | 47.2 | 1828 (5398) | 1441 (45.8) | 387 (54.6) | 299 (58.3) | 290 (54.7) | ||
| Moderate | 9.1 | 338 (1038) | 253 (8.4) | 85 (12.4) | 60 (14.0) | 53 (10.2) | ||
| Severe | 2.4 | 93 (270) | 68 (2.2) | 25 (3.4) | 21 (4.0) | 18 (3.4) | ||
| Unknown | 1.9 | 60 (212) | 49 (1.6) | 11 (2.9) | 8 (1.2) | 8 (2.3) | ||
| Gleason | 0.13 | 0.02 | ||||||
| 2–6 | 47.2 | 1712 (5399) | 1396 (47.3) | 316 (47.1) | 246 (48.6) | 241 (47.5) | ||
| 7 | 42.6 | 1603 (4865) | 1339 (43.0) | 264 (40.5) | 197 (37.7) | 202 (40.9) | ||
| 8–10 | 10.2 | 369 (1164) | 283 (9.7) | 86 (12.5) | 59 (13.7) | 59 (11.7) | ||
| PSA | 0.46 | 0.06 | ||||||
| 0–3.9 | 14.7 | 500 (1645) | 452 (16.1) | 48 (7.6) | 45 (9.2) | 44 (9.3) | ||
| 4–9.9 | 67.0 | 2363 (7486) | 1977 (67.9) | 386 (62.4) | 317 (67.1) | 313 (65.4) | ||
| 10–19.9 | 12.9 | 502 (1440) | 364 (11.7) | 138 (18.9) | 99 (16.9) | 96 (16.8) | ||
| 20+ | 5.4 | 226 (608) | 136 (4.3) | 90 (11.1) | 41 (6.8) | 49 (8.5) | ||
| NCCN risk groupd | 0.25 | 0.03 | ||||||
| Low | 37.6 | 1316 (4295) | 1106 (38.5) | 210 (32.6) | 167 (35.9) | 166 (34.5) | ||
| Intermediate | 48.7 | 1842 (5565) | 1532 (48.9) | 310 (47.7) | 246 (45.9) | 241 (47.8) | ||
| High | 13.8 | 528 (1572) | 381 (12.6) | 147 (19.7) | 89 (18.2) | 95 (17.7) | ||
| Receipt of ADT | 1.22 | 1.12 | ||||||
| No | 85.5 | 2967 (9361) | 2666 (93.2) | 301 (47.6) | 443 (91.3) | 241 (50.1) | ||
| Yes | 14.5 | 560 (1585) | 194 (6.8) | 366 (52.4) | 35 (8.7) | 261 (49.9) | ||
| Graduation year | 0.25 | 0.21 | ||||||
| 1950–1969 | 10.4 | 279 (828) | 243 (10.6) | 36 (9.1) | 47 (11.4) | 26 (8.9) | ||
| 1970–1979 | 23.1 | 622 (1850) | 472 (22.1) | 150 (28.5) | 92 (26.3) | 110 (27.0) | ||
| 1980–1989 | 37.8 | 1008 (3020) | 817 (37.4) | 191 (39.8) | 137 (37.0) | 153 (40.6) | ||
| 1990+ | 28.7 | 791 (2293) | 672 (29.9) | 119 (22.6) | 104 (25.3) | 97 (23.5) | ||
| Practice type | 0.39 | 0.44 | ||||||
| Solo practice | 9.0 | 309 (720) | 295 (10.2) | 14 (3.0) | 57 (12.8) | 11 (3.3) | ||
| Group practice | 91.0 | 2391 (7271) | 1909 (89.8) | 482 (97.0) | 323 (87.2) | 375 (96.7) | ||
| Ownership of facility | −0.16 | −0.19 | ||||||
| For-profit | 9.8 | 325 (903) | 296 (10.2) | 29 (7.4) | 53 (10.8) | 21 (6.4) | ||
| Non-profit/ government | 90.2 | 2632 (8,296) | 2250 (89.8) | 382 (92.6) | 388 (89.2) | 295 (93.6) | ||
| Teaching status | −0.10 | 0.11 | ||||||
| Non-teaching | 43.9 | 1323 (4035) | 1121 (42.6) | 202 (53.0) | 237 (54.6) | 152 (50.9) | ||
| Teaching | 56.1 | 1634 (5,164) | 1425 (57.4) | 209 (47.0) | 204 (45.4) | 164 (49.1) | ||
| Distance to treatment facility | 0.10 | 0.11 | ||||||
| <5 | 29.2 | 884 (2760) | 712 (28.5) | 172 (32.9) | 121 (30.0) | 126 (31.1) | ||
| 5–9 | 23.3 | 702 (2199) | 592 (23.3) | 110 (23.1) | 112 (27.0) | 91 (24.9) | ||
| 10–14 | 12.5 | 370 (1179) | 308 (12.6) | 62 (11.6) | 45 (11.6) | 50 (11.8) | ||
| 15+ | 35.1 | 1092 (3315) | 902 (35.6) | 190 (32.3) | 153 (31.4) | 142 (32.1) | ||
| Number of urologists per 100 000 men | 0.19 | 0.05 | ||||||
| 0 | 13.7 | 559 (1347) | 444 (13.5) | 115 (14.9) | 86 (14.0) | 82 (13.8) | ||
| >0–6 | 45.4 | 1160 (4,461) | 988 (46.3) | 172 (40.5) | 154 (42.8) | 143 (44.4) | ||
| >6–10 | 23.7 | 765 (2325) | 621 (23.6) | 144 (23.8) | 106 (25.3) | 112 (22.7) | ||
| 10+ | 17.3 | 712 (1696) | 559 (16.6) | 153 (20.8) | 105 (17.9) | 105 (19.1) | ||
Column percentages based on weighted number of patients.
Unweighted number of patients (weighted number in parenthesis).
Unweighted number of patients (weighted column percentages in parenthesis).
Low (T1–2a AND Gleason score ≤6 AND PSA <10 ng/mL), intermediate (T2b–T2c OR Gleason score 7 OR PSA 10–20 ng/mL), and high (≥ T3a OR Gleason score 8–10 OR PSA >20 ng/mL). ACE-27: Adult Comorbidity Evaluation-27; ADT: androgen deprivation therapy; IMRT: intensity-modulated radiation therapy; NCCN: National Comprehensive Cancer Network; PSA: prostate-specific antigen.
PSM was used to create a well-balanced cohort, defined in terms of standardized differences being 0.10 or less for the covariates in Table 1. This conventional benchmark was achieved for all variables except state registry, receipt of ADT, graduation year, practice type, facility ownership, teaching status, and distance to treatment facility. We identified 502 RP and 502 IMRT patients that we regarded as sufficiently well-balanced with respect to standardized differences (Table 1). The maximum followup was 11.9 years. The overall median followup for this analytic cohort was 10.5 years (interquartile range [IQR] 9.9–11.0). The median followup was 10.6 years (IQR 10.0–11.1) and 10.4 years (IQR 9.6–10.9) for the RP and IMRT cohorts, respectively. Over the duration of followup, there were 31 (3.1%) PCa-related deaths in the analytic cohort.
IMRT was associated with a 41% increased risk of all-cause mortality, compared to RP (hazard ratio [HR] 1.41, 95% CI 1.13–1.76) (Table 2). There was no significant difference in PCa-specific mortality between IMRT and RP (HR 1.75, 95% CI 0.84–3.64). Consistent with these findings, KM analyses demonstrated a significantly worse OS, but not CSS, in patients treated with IMRT as compared to RP (Figures 1, 2). Table 2 demonstrates the 11-year OS benefit for RP at 71.2% (95% CI 66.9–75.8) as compared to IMRT at 62.3% (95% CI 57.4–67.6). There was no statistically significant difference in 11-year CSS (95.5%, 95% CI 93.5–97.6 and 96.3%, 95% CI 94.0–98.7% for IMRT and RP, respectively).
Table 2.
Eleven-year survival rates and hazard ratios of propensity score matched cohort and NCCN high-risk disease subgroup
| Treatment | n | Number of events | 11-year survival rate (%) (95% CI) | HR (95% CI) | p | |
|---|---|---|---|---|---|---|
| Propensity score matched cohort | ||||||
| Overall survival | Surgery | 502 | 128 | 71.2 (66.9–75.8) | 1.00 | 0.004 |
| IMRT | 502 | 164 | 62.3 (57.4–67.6) | 1.41 (1.13–1.76) | ||
| Prostate cancer survival | Surgery | 502 | 12 | 96.3 (94.0–98.7) | 1.00 | 0.12 |
| IMRT | 502 | 19 | 95.5 (93.5–97.6) | 1.75 (0.84–3.64) | ||
| NCCN high-risk disease subgroup | ||||||
| Overall survival | Surgery | 89 | 32 | 1.00 | 0.07 | |
| IMRT | 95 | 42 | 1.53 (0.97–2.42) | |||
| Prostate cancer survival | Surgery | 89 | 6 | 1.00 | 0.20 | |
| IMRT | 95 | 10 | 1.92 (0.69–5.36) | |||
CI: confidence interval; HR: hazard ratio; IMRT: intensity-modulated radiation therapy; NCCN: National Comprehensive Cancer Network.
Figure 1.
Overall survival of surgery and intensity-modulated radiation therapy (IMRT) treatment groups.
Figure 2.
Cumulative incidence of prostate cancer and non-prostate cancer deaths in surgery and intensity-modulated radiation therapy (IMRT) treatment groups.
We performed an a priori defined subgroup analysis of patients with NCCN high-risk PCa. IMRT, as compared to RP, was not associated with a statistically significant difference in all-cause mortality (HR 1.53, 95% CI 0.97–2.42) or PCa-specific mortality (HR 1.92, 95% CI 0.69–5.36) in the subset of patients with NCCN high-risk PCa (Table 2).
The 11-year cumulative incidence rates for overall, PCa-specific, and non-PCa specific mortality are shown in Table 3. The 11-year cumulative incidence rates were significantly higher for IMRT as compared to RP for overall and non-PCa-specific mortality, but not PCa-specific mortality (all-cause mortality 37.7%, 32.6–42.8 and 28.8%, 95% CI 24.4–33.3, p=0.003; non-PCa-specific mortality 33.8%, 28.9–38.9 and 25.7%, 21.6–29.9, p=0.011; and PCa-specific mortality 3.8%, 2.4–5.9) and 3.1%, 1.6–5.5, p=0.18, for IMRT and RP, respectively).
Table 3.
Eleven-year cumulative incidence rates for all-cause, prostate cancer, and non-prostate cancer mortality rates in surgery and IMRT treatment groups
| 11-year cumulative incidence (%) | |||
|---|---|---|---|
|
| |||
| Treatment | All-cause mortality (95% CI) | Prostate cancer mortality (95% CI) | Non-prostate cancer mortality (95% CI) |
| Surgery | 28.8 (24.4–33.3) | 3.1 (1.6–5.5) | 25.7 (21.6–29.9) |
| IMRT | 37.7 (32.6–42.8) | 3.8 (2.4–5.9) | 33.8 (28.9–38.9) |
| p | 0.003 | 0.18 | 0.011 |
CI: confidence interval; IMRT: intensity-modulated radiation therapy.
Discussion
In this observational study with granular comorbidity adjustment, we observed that among patients with localized PCa, IMRT was associated with a 41% increased risk of all-cause mortality as compared with RP, though there was no significant difference in PCa-specific mortality. In our a priori subgroup analysis of those patients with NCCN high-risk PCa, we did not observe statistically significant differences in all-cause or PCa-specific mortality.
In the absence of informative randomized data comparing modern approaches, comparative effectiveness studies with observational data are helpful, although the latter are controversial due to residual confounding by comorbidities. Prior observational cohort studies have demonstrated that patients receiving RT have a greater comorbidity burden than those treated with surgery.7,8 Comorbidity has been shown to be an independent prognostic factor in PCa outcomes,6 and is an important component of shared decision-making. The ACE-27 score used in our study to evaluate comorbidity burden, like the Deyo-Charlson comorbidity index, has been shown to provide unique but significant prognostic information regarding comorbid illnesses.19 To account for potential confounding from patient comorbidities, our study employed rigorous chart-based abstraction and assessment using the ACE-27 instrument to quantify risk.11 Furthermore, comorbidity as assessed by ACE-27 was used for PSM to reduce selection bias in PCa treatment whereby patients with greater comorbidity are more likely to be treated with radiation over surgery. Indeed, our study represents the best methods for comparing effectiveness of treatments used for patients with very disparate baseline comorbidity burden.
A recent systematic review and meta-analysis of clinically localized PCa found that patients treated with RT had a greater risk of PCa-specific mortality (HR 2.08, 95% CI 1.76–2.47, p<0.00001) and all-cause mortality (HR 1.63, 95% CI 1.54–1.73, p<0.00001), compared to those treated with RP.5 Our results are consistent with regard to all-cause mortality benefit for RP as compared to RT. However, we did not observe a significant difference in PCa-specific mortality between RP and RT. Many of the studies included in the previously mentioned systematic review and meta-analysis did not control for comorbidity; or, if they did, did not employ the type of rigorous statistical approach we adopted here. It has been previously demonstrated that comorbidity burden is greater in those men who undergo RT.6 As such, confounding by inconsistent control for comorbidities may explain the increased all cause and PCa-specific mortality risk associated with RT in this meta-analysis. Recognizing the differences in comorbidity burden between PCa patients treated with RP or EBRT, a prior retrospective cohort study compared treatment outcomes in a cohort of men with no comorbidities as assessed using ACE-27 and Charlson comorbidity index. In this study, they found that EBRT was associated with increased PCa-specific mortality and overall mortality, compared to RP.20 Unlike our study, this prior study did not attempt to adjust for age, thus their cohort RP patients were significantly younger than the RT patients (60 years vs. 66.8 years) which likely adds greater confounding possibly due to greater chance of developing comorbidity during followup.20
Limitations and strengths
Our study is limited by the retrospective, non-randomized nature of the cohort and challenges in inferring causality in a non-experimental setting. We considered as many covariates as possible to establish the propensity score and a well-balanced patient cohort. This reduced our small sample size after propensity score-matching and may have also adversely impacted statistical power. Although age and comorbidity burden at diagnosis were adjusted by PSM, we did not adjust for comorbidities that developed later. It is possible that advanced age or higher baseline comorbidity burden could be more likely to develop additional or worsening comorbidities during the followup period. Our study did not exclude those who were diagnosed with other primary cancers; therefore, the survival time may be affected by other cancers. Our study does not include data on cancer recurrence and subsequent treatments, which may also impact survival. This is particularly important since bias may exist if there is preferential use of salvage therapies between the RP and RT (e.g., salvage RP or adjuvant/salvage RT). The observed association of treatment with a significant difference in all-cause but not PCa-specific mortality may reflect residual confounding that persists despite our best effort to control covariates. Conversely, underpowering due to low PCa deaths may explain why the difference in PCa-specific mortality did not reach statistical significance. Lastly, the treatment of localized PCa has continued to evolve with high-dose rate brachytherapy becoming increasingly used, but not included in our analysis.21
Despite these limitations, the major strength of our study is its generalizability, as it represents a population-based cohort created from the CDC National Program of Cancer Registries, which seeks to include all diagnosed cancer patients, regardless of age or insurance status. Thus, our cohort reflects the diversity of patients with localized PCa regarding comorbidity, age, race, and insurance status, as well as practice and provider characteristics. Observational studies may represent a complementary opportunity to study treatment effect in localized PCa given the challenges of performing randomized trials and rapidly evolving practice patterns. In order to guide patient selection in an environment of shared decision-making, ongoing followup of observational cohorts and completed randomized controlled trials will be needed.
Conclusions
We observed that in a population-based observational cohort of men with localized PCa that was well-balanced based on multiple covariates, including a robust index of comorbidity, IMRT was associated with a statistically significant increased risk of all-cause mortality but not PCa-specific mortality.
Footnotes
Competing interests: The authors do not report any competing personal or financial interests related to this work.
This paper has been peer-reviewed.
Funding: Funding for the study was provided by the Hamill Foundation. The Breast and Prostate Cancer Data Quality and Patterns of Care study was supported by the Centers for Disease Control and Prevention through cooperative agreements with the California Cancer Registry (Public Health Institute) (1-U01-DP000260), Emory University (1-U01-DP000258), Louisiana State University Health Sciences Center (1-U01-DP000253), Minnesota Cancer Surveillance System (Minnesota Department of Health) (1-U01-DP000259), Medical College of Wisconsin (1-U01-DP000261), University of Kentucky (1-U01-DP000251), and Wake Forest University (1-U01-DP000264). The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
References
- 1.Hamdy FC, Donovan JL, Lane JA, et al. 10-year outcomes after monitoring, surgery, or radiotherapy for localized prostate cancer. N Engl J Med. 2016;375:1415–24. doi: 10.1056/NEJMoa1606220. [DOI] [PubMed] [Google Scholar]
- 2.Wilt TJ, Brawer MK, Jones KM, et al. Radical prostatectomy vs. observation for localized prostate cancer. N Engl J Med. 2012;367:203–13. doi: 10.1056/NEJMoa1113162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bill-Axelson A, Holmberg L, Ruutu M, et al. Radical Prostatectomy vs. watchful waiting in early prostate cancer. N Engl J Med. 2011;364:1708–17. doi: 10.1056/NEJMoa1011967. [DOI] [PubMed] [Google Scholar]
- 4.Lane JA, Donovan JL, Davis M, et al. Active monitoring, radical prostatectomy, or radiotherapy for localised prostate cancer: Study design and diagnostic and baseline results of the ProtecT randomised phase 3 trial. Lancet Oncol. 2014;15:1109–18. doi: 10.1016/S1470-2045(14)70361-4. [DOI] [PubMed] [Google Scholar]
- 5.Wallis CJD, Saskin R, Choo R, et al. Surgery versus radiotherapy for clinically-localized prostate cancer: A systematic review and meta-analysis. Eur Urol. 2016;70:21–30. doi: 10.1016/j.eururo.2015.11.010. [DOI] [PubMed] [Google Scholar]
- 6.Albertsen PC, Fryback DG, Storer BE, et al. The impact of co-morbidity on life expectancy among men with localized prostate cancer. J Urol. 1996;156:127–32. doi: 10.1016/S0022-5347(01)65964-0. [DOI] [PubMed] [Google Scholar]
- 7.Sooriakumaran P, Nyberg T, Akre O, et al. Comparative effectiveness of radical prostatectomy and radiotherapy in prostate cancer: Observational study of mortality outcomes. BMJ. 2014;348:1–13. doi: 10.1136/bmj.g1502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cooperberg MR, Vickers AJ, Broering JM, et al. Comparative risk-adjusted mortality outcomes after primary surgery, radiotherapy, or androgen-deprivation therapy for localized prostate cancer. Cancer. 2010;116:5226–34. doi: 10.1002/cncr.25456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wong YN, Mitra N, Hudes G, et al. Survival associated with treatment vs observation of localized prostate cancer in elderly men. J Am Med Assoc. 2006;296:2683–93. doi: 10.1001/jama.296.22.2683. [DOI] [PubMed] [Google Scholar]
- 10.Hoffman RM, Koyama T, Fan KH, et al. Mortality after radical prostatectomy or external beam radiotherapy for localized prostate cancer. J Natl Cancer Inst. 2013;105:711–8. doi: 10.1093/jnci/djt059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Piccirillo JF, Tierney RM, Costas I, et al. Prognostic importance of comorbidity in a hospital-based cancer registry. J Am Med Assoc. 2004;291:2441–7. doi: 10.1001/jama.291.20.2441. [DOI] [PubMed] [Google Scholar]
- 12.Satkunasivam R, Lo M, Stern M, et al. The role of provider characteristics in the selection of surgery or radiation for localized prostate cancer and association with quality of care indicators. Am J Clin Oncol Cancer Clin Trials. 2018;41:1076–82. doi: 10.1097/COC.0000000000000442. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Tewari A, Sooriakumaran P, Bloch DA, et al. Positive surgical margin and perioperative complication rates of primary surgical treatments for prostate cancer: A systematic review and meta-analysis comparing retropubic, laparoscopic, and robotic prostatectomy. Eur Urol. 2012;62:1–15. doi: 10.1016/j.eururo.2012.02.029. [DOI] [PubMed] [Google Scholar]
- 14.Read WL, Tierney RM, Page NC, et al. Differential prognostic impact of comorbidity. J Clin Oncol. 2004;22:3099–103. doi: 10.1200/JCO.2004.08.040. [DOI] [PubMed] [Google Scholar]
- 15.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28:3083–107. doi: 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41. doi: 10.1093/biomet/70.1.41. [DOI] [Google Scholar]
- 17.Stedman MR, Lew RA, Losina E, et al. A comparison of statistical approaches for physician-randomized trials with survival outcomes. Contemp Clin Trials. 2012;33:104–15. doi: 10.1016/j.cct.2011.08.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhou B, Fine J, Latouche A, et al. Competing risks regression for clustered data. Biostatistics. 2012;13:371–83. doi: 10.1093/biostatistics/kxr032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kallogjeri D, Gaynor S, Piccirillo M, et al. Comparison of comorbidity collection methods. J Am Coll Surg. 2014;219:245–55. doi: 10.1016/j.jamcollsurg.2014.01.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nepple KG, Stephenson AJ, Kallogjeri D, et al. Mortality after prostate cancer treatment with radical prostatectomy, external-beam radiation therapy, or brachytherapy in men without comorbidity. Eur Urol. 2013;64:372–8. doi: 10.1016/j.eururo.2013.03.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Mendez LC, Morton GC. High dose-rate brachytherapy in the treatment of prostate cancer. Transl Androl Urol. 2018;7:357–70. doi: 10.21037/tau.2017.12.08. [DOI] [PMC free article] [PubMed] [Google Scholar]