Survival by treatment method in patients with localized prostate cancer, derived from the Cox proportional hazards regression model adjusted for age, sector of care, period of diagnosis, method of detection, lymph node status, clinical tumor stage, differentiation, and prostate-specific antigen. Only deaths from prostate cancer were considered.
Merglen A, Schmidlin F, Fioretta G, Verkooijen HM, Rapiti E, Zanetti R, Miralbell R, Bouchardy C. Short- and Long-term Mortality With Localized Prostate Cancer. Arch Intern Med. 2007;167(18):1944-1950. doi:10.1001/archinte.167.18.1944
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
No clear guidelines exist for managing localized prostate cancer because clinical studies have not yet established which treatment provides the best long-term outcome. We assessed the effect of treatment on prostate cancer–specific mortality considering the determinants of treatment and prognosis.
The population-based cohort included all 844 patients having a diagnosis of localized prostate cancer between January 1, 1989, and December 31, 1998, in Geneva, Switzerland. Treatments included prostatectomy (n = 158), radiotherapy (n = 205), watchful waiting (n = 378), hormone therapy (n = 72), and other types of therapy (n = 31). We compared survival curves using the log-rank test. With multivariate Cox proportional hazards analysis and propensity score methods, we evaluated the independent effect of treatments on prostate cancer–specific mortality.
Treatment options only slightly influenced 5-year prostate cancer–specific mortality but had an important effect on long-term mortality. Ten-year specific survival was 83% (95% confidence interval [CI], 73%-93%), 75% (95% CI, 67%-83%), and 72% (95% CI, 66%-80%) for patients who underwent surgery, radiotherapy, and watchful waiting, respectively (P < .001). At 10 years, patients treated with radiotherapy or watchful waiting had a significantly increased risk of death from prostate cancer compared with patients who underwent prostatectomy (multiadjusted hazard ratio, 2.3 [95% CI, 1.2-4.3] and 2.0 [95% CI, 1.1-3.8], respectively). The increased mortality associated with radiotherapy and watchful waiting was primarily observed in patients younger than 70 years and in patients with poorly differentiated tumors (Gleason score ≥ 7; reference, 1 [best]-10 [worst]). Patients who received hormone therapy alone already had an increased risk of prostate cancer–specific mortality at 5 years (hazard ratio, 3.5 [95% CI, 1.4-8.7]).
Our study results suggest that surgery offers the best chance of long-term prostate cancer–specific survival, in particular for younger patients and patients with poorly differentiated tumors. Until clinical trials provide conclusive evidence, physicians and patients should be informed of these results and their limitations.
Strategies for management of localized prostate cancer are still being debated because randomized trials have not yet established which treatment offers the best chance of survival.1,2 Therefore, treatment choice is strongly influenced by patient and physician personal preferences and experience.3
In the absence of conclusive evidence from clinical trials, population-based observational studies are the only way to evaluate the effect of various treatment methods on mortality. However, observational studies are prone to selection bias because treatment choice is determined by factors that may also affect prostate cancer–specific mortality. Therefore, it is important in observational studies to adjust for the main determinants of treatment. Few observational studies have evaluated the effect of treatment on prostate cancer survival while controlling for the effect of other confounders.4- 9 To our knowledge, none of these studies have compared the effect of all treatment methods for prostate cancer (ie, prostatectomy, radiotherapy, watchful waiting, and hormone therapy) on long-term prostate cancer–specific mortality after adjusting for treatment determinants and prognostic factors. This was the purpose of the present study.
We used data from the Geneva Cancer Registry (Institute for Social and Preventive Medicine, Geneva University, Geneva, Switzerland), which records all incident cancers occurring in the population of the canton (approximately 435 000 inhabitants in 2004). All hospitals, pathology laboratories, and practitioners are requested to report all cancer cases. Trained registrars systematically abstract data from medical and laboratory records. Physicians regularly receive questionnaires to obtain missing data.
Recorded data include sociodemographic variables; method of detection; prostate-specific antigen (PSA) value at diagnosis; and tumor characteristics (including histology, differentiation based on tumor grade, and Gleason score)10; stage at diagnosis according to TNM classification10 lymph node status; treatment; survival status; and cause of death.
The registry regularly assesses survival data. The index date was the date of confirmation of diagnosis or the date of hospitalization if it preceded the diagnosis and pertained to the disease. Active follow-up is performed yearly using the files of the Cantonal Population Office (Geneva, Switzerland), which is in charge of registration of the resident population. Trained registrars establish cause of death by systematically consulting clinical records, interpreting questionnaires completed by the patient's physician, or both.
Between January 1, 1989, and December 31, 1998, 1740 men were diagnosed with prostate cancer in the resident population. We excluded 127 men with prostate cancer diagnosed at death and 118 men with previous invasive cancer except nonmelanoma skin cancer. We limited the study to the 844 patients with clinically localized prostate cancer, that is, clinical stage T1, T2, T3, and M0.
Treatments given during the first 6 months after having a diagnosis of prostate cancer included prostatectomy (radical, retropubic, or perineal), hormone treatment (surgical or hormonal castration), and external radiotherapy; brachytherapy was not used during the study. Watchful waiting consisted of active follow-up of the patient and invasive treatment with disease progression.
We considered as confounders all factors linked to both treatment choice and prognosis. We identified factors linked to treatment choice using the χ2 test. We evaluated factors related to prognosis by multivariate Cox proportional hazards analysis. Age, period of diagnosis, method of detection, lymph node status, tumor stage, differentiation, and PSA value were the variables considered as confounders when examining the effect of treatment on death from prostate cancer. For death from other causes, confounders included in the model were age, civil status, social class, sector of care, period of diagnosis, and method of detection.
We calculated survival using the Kaplan-Meier method and compared survival curves using the log-rank test. We evaluated the independent effect of treatments on mortality using the Cox proportional hazards model, which accounted for all identified confounders. In confirmatory analyses, we used propensity scores to minimize bias from nonrandom assignment of treatments.11 Using logistic regression, we identified sociodemographic and clinical predictors of treatment use. From these logistical models, we derived (in continuous variables, predicted) individual probabilities of receiving a given treatment compared with prostatectomy. Again using Cox proportional hazards models, we adjusted for these probabilities for each treatment method. We considered prostate cancer–specific mortality, mortality from other causes, and overall mortality. To evaluate whether treatment effect varied according to age at diagnosis, tumor characteristics, or period, we a priori performed subgroup analyses and interaction tests between treatment and age (<70 vs ≥70 years), differentiation (Gleason score <7 vs ≥7), tumor stage (T1-T2 vs T3), and period (1989-1993 vs 1994-1998). Data were analyzed with SPSS software (version 14; SPSS Inc, Chicago, Illinois). P < .05 was considered statistically significant.
Among the 844 patients with localized prostate cancer, 797 men (94%) had histologically confirmed disease and 31 (4%) had cytologically confirmed disease, and in 16 (2%), the diagnosis was based on an elevated PSA level, findings at imaging (ie, radiographs, computed tomographic scan, and transrectal ultrasound), or both. The patients' mean age was 71 years (range, 44-97 years) (Table 1). Approximately one-third of the prostate cancers were detected by screening (PSA level, digital rectal examination, or both). Lymph node metastases were not systematically investigated, and 311 patients (37%) had unknown lymph node status. Pathologists in Geneva, Switzerland, did not systematically report Gleason score and grade at the beginning of the study period, and 132 patients (16%) had unknown differentiation. We had no PSA values at diagnosis for 231 patients (27%). One hundred fifty-eight (19%) patients underwent prostatectomy, 205 (24%) underwent radiotherapy (radiotherapy alone in 152 and radiotherapy plus hormone therapy in 53), 378 (45%) were managed with watchful waiting, 72 (9%) received hormone therapy alone, and 31 (4%) underwent other treatment combinations (Table 1).
Compared with patients who underwent surgery, those who received radiotherapy were older, had slightly higher PSA values at diagnosis, and, when first seen, more often had unknown tumor grade and less often had unknown PSA values. Patients managed with watchful waiting were older than those who underwent surgery, and their tumors were less often detected at screening. Compared with patients in the surgery group, patients in the treatment group had early-stage and well-differentiated tumors at diagnosis. Patients who received hormone therapy were older, and more often had stage T3 tumors and PSA values greater than 30 μg/L. Patients managed with watchful waiting and patients who received hormone therapy often had more missing information on lymph node status and tumor grade (Table 1). Sector of care, period of diagnosis, civil status, and social class also differed by treatment.
The mean duration of follow-up was 6.7 years (median, 6.8 years; range, 0-15.8 years). During the study, 47 patients (5.6%) left Geneva and were lost to follow-up. At 5 years (Table 2), patients who underwent surgery or received radiotherapy had similar overall survival rates (respectively, 86%, 95% confidence interval [CI], 80%-92%; and 87%, 95% CI, 83%-91%), which were higher than those in patients managed with watchful waiting (61%, 95% CI, 55%-66%). Five-year prostate cancer–specific survival was 94% (95% CI, 90%-98%) in patients who underwent surgery, 93% (95% CI, 89%-97%) in those who received radiotherapy, 86% (95% CI, 82%-90%) in patients managed with watchful waiting, 71% (95% CI, 59%-83%) in patients who received hormone therapy alone, and 71% (95% CI, 53%-89%) in patients who received other combinations of therapy.
Compared with patients who underwent prostatectomy, the risk of death from prostate cancer at 5 years was not different for patients who received radiotherapy and patients managed with watchful waiting but was significantly higher in patients who received hormone therapy (adjusted hazard ratio [HRadj], 3.5; 95% CI, 1.4-8.7) or other treatments (HRadj, 5.8; 95% CI, 2.1-16.0). At 10 years (Table 2), however, patients who underwent prostatectomy had a higher overall survival rate (69%, 95% CI, 59%-79%) than did patients managed with radiotherapy or watchful waiting (respectively, 54%, 95% CI, 50%-62%; and 35%, 95% CI, 29%-41%; log-rank test, P < .001). This pattern was similar for disease-specific survival. Compared with patients who underwent prostatectomy, we observed increased mortality risk in patients who received radiotherapy (HRadj, 2.3; 95% CI, 1.2- 4.3), patients managed with watchful waiting (HRadj, 2.0; 95% CI, 1.1-3.8), hormone therapy (HRadj, 4.4; 95% CI, 2.2-8.8), and patients who received other treatments (HRadj, 3.1; 95% CI,1.3-7.5) (Table 2). The mortality from other causes did not change significantly between treatment methods after adjustment for confounders (Table 2). Exclusion of patients with no histologic confirmation of prostate cancer provided similar results (data not shown). The Figure shows survival curves according to treatment options derived from Cox proportional hazards models, accounting for other prognostic factors.
In confirmatory analyses, adjustments for propensity scores yielded no substantial differences relative to traditional multivariate Cox proportional hazards analyses. For example, the risk of death from prostate cancer at 10 years in patients who received radiotherapy and patients managed with watchful waiting remained unchanged (propensity score HRadj, 2.3; 95% CI, 1.2-4.4) compared with patients who underwent prostatectomy (propensity score HRadj, 2.0; 95% CI, 1.0-3.7).
Within subgroups, increased long-term mortality associated with radiotherapy and watchful waiting was limited to younger men (< 70 years) (Table 3) and patients with poorly differentiated tumors (Gleason score ≥ 7) (Table 4). In particular, younger patients who received radiotherapy had a 6.7-fold (95% CI, 2.2-20.7-fold) increased risk of death from prostate cancer at 10 years compared with patients who underwent prostatectomy. Also, patients with poorly differentiated tumors who received radiotherapy had a 5.2-fold (95% CI, 1.3-20.6-fold) increased risk of death from the disease compared with patients who underwent surgery. However, none of the interaction tests between treatment and age, differentiation, tumor stage, or period was significant.
Results of this population-based study show that treatment option may only slightly influence short-term survival after prostate cancer but may have an important effect on long-term mortality. Compared with patients who underwent prostatectomy, at 10 years, those who received radiotherapy and patients managed with watchful waiting had an approximately 2-fold increased risk of dying of the disease. The differences in specific mortality were most pronounced in patients younger than 70 years and in those with poorly differentiated tumors. Patients who received hormone therapy only or other combined treatments already had a strongly impaired prognosis at 5 years.
To our knowledge, this is the first study that compares the effect of all therapeutic methods used in routine health care practice (ie, surgery, radiotherapy, watchful waiting, hormone treatment, and other combined treatments) on prostate cancer–specific mortality after adjustment for potential confounders in a population-based study. This study has several limitations. As with observational studies in general, treatments were not randomly allocated in our population. This study is also limited by the lack of information on comorbidities. Although differences in mortality from other causes were not statistically significant between treatment groups, the watchful waiting and hormone therapy groups showed an almost 50% increase in mortality from causes other than prostate cancer both at 5 and 10 years, which suggests the presence of baseline differences in prognostic factors. To control for this bias, we used 2 methods considered appropriate to reduce the effect of confounding by indication: Cox proportional hazards regression analysis adjusted for major prognostic determinants and propensity score method based on adjustment for the determinants of treatment allocation. Both methods produced similar results, which suggests that these results are robust although bias may not be completely captured. Moreover, even though comorbidities can be important determinants of treatment choice, their effect on prostate cancer–specific mortality per se is low.5,12
Differences in diagnosis assessment also exist between therapeutic groups because tumor stage and grade are based only on clinical status and biopsy samples in patients who did not undergo surgery. Furthermore, inasmuch as screening prevalence, diagnostic assessment, treatment, and surveillance have probably changed during the study period, these results may not be generalizable to the present situation. We acknowledge that these results should be discussed in the context of quality of life, an issue not considered in this study.
Three randomized trials compared prostatectomy with watchful waiting.13- 17 The Veterans Administration Cooperative Urological Research Group randomized trial showed no survival differences between patients who underwent surgery and patients managed with watchful waiting.13- 15 However, this study included a limited number of patients and had a high rate of a posteriori exclusion (22%), mainly owing to treatment refusal or incorrect staging. A Scandinavian study reported higher prostate cancer–specific mortality after 10 years of follow-up in patients managed with watchful waiting compared with patients who underwent surgery.16 However, this study included few patients with cancers detected at screening; therefore, these results may not apply to the present situation. The ongoing Prostate Cancer Intervention vs Observation Trial includes a more representative sample of clinically localized prostate cancers, but results are not yet available.17
Two clinical trials compared prostatectomy with radiotherapy.18,19 Paulson18 reported a significantly lower risk of disease progression in patients who underwent surgery. Akakura et al19 showed lower prostate cancer–specific mortality in patients who underwent surgery but a better quality of life in patients who received radiotherapy. Other clinical trials failed to meet recruitment criteria and were ended prematurely.1,20
While comparing the effects of watchful waiting and surgery on mortality from prostate cancer, we found similar results as those reported in, to our knowledge, the only published clinical randomized trial. In this trial, the risk of prostate cancer mortality at 10 years was 0.56 (range, 0.36-0.88) in patients who underwent prostatectomy compared with patients managed with watchful waiting.16 The mean follow-up was 8.2 years, slightly longer than in our study. Using the same reference categories, we found a comparable HRadj of 0.66 (95% CI, 0.39-1.10) in patients who underwent surgery. The absolute benefit in terms of specific mortality at 10 years was also close in the clinical trial vs our study (respectively, 5% [95% CI, −0.3% to 11%] vs 11% [95% CI, 3.1%-18.9%]).16
Clinical trials to date have yielded only limited information on treatment efficacy; thus, knowledge is basically provided by a few observational studies.21- 25 Three observational studies evaluated the effect of surgery, radiotherapy, or both, and watchful waiting on long-term prognosis using a population-based cohort of patients with localized prostate cancer, tconsidering important prognostic factors.4- 6 Lu-Yao and Yao4 compared 10-year specific survival rates for prostatectomy, radiotherapy, and watchful waiting stratified by histologic grade and found that patients with poorly differentiated tumors received the highest benefit from prostatectomy.4 The results of our stratified analyses suggest that young patients and patients with higher Gleason scores may especially benefit from surgery. However, because treatment effects were not statistically different across age groups and tumor grades, these subgroup analyses should be interpreted with caution. To draw a conclusion on the different effects in patient subgroups, larger studies are needed. Barry et al5 reported 10-year overall and specific survival by treatment groups stratified by the primary prognostic factors but made no comparison of outcome between treatment groups. However, they reported a particularly poor prognosis in patients managed with watchful waiting even in the short term, and in particular in patients with poorly differentiated tumors. Wong et al6 compared active treatment vs watchful waiting on 10-year overall mortality and used propensity score to adjust for treatment selection bias. They reported a survival benefit associated with active treatment in patients with prostate cancer at low and intermediate risk but did not differentiate between surgery and radiotherapy considered together in the active treatment group.6 All of these results are in agreement with our findings. Three observational studies compared prognosis in patients who underwent prostatectomy vs those who received radiotherapy.7- 9 However, these studies used biochemical failure as outcome and did not consider prostate cancer–specific mortality.
There is growing evidence from observational studies that prostatectomy offers the best chance of long-term specific survival in men with localized prostate cancer, in particular, in younger patients and those with poorly differentiated tumors. Until clinical trials provide conclusive evidence, physicians and patients should be informed of these results and their limitations.
Correspondence: Christine Bouchardy, MD, MPH, Geneva Cancer Registry, Institute of Social and Preventive Medicine, Geneva University, 55 Boulevard de la Cluse, 1205 Geneva, Switzerland (firstname.lastname@example.org).
Accepted for Publication: May 18, 2007.
Author Contributions:Study concept and design: Merglen, Schmidlin, Verkooijen, Zanetti, and Bouchardy. Acquisition of data: Fioretta and Verkooijen. Analysis and interpretation of data: Merglen, Schmidlin, Fioretta, Verkooijen, Rapiti, and Miralbell. Drafting of the manuscript: Merglen, Rapiti, Miralbell, and Bouchardy. Critical revision of the manuscript for important intellectual content: Merglen, Schmidlin, Fioretta, Verkooijen, Rapiti, Zanetti, Miralbell, and Bouchardy. Statistical analysis: Merglen, Fioretta, Verkooijen, and Rapiti. Study supervision: Zanetti and Bouchardy.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by PROSPER (Program for Social Medicine, Preventive and Epidemiological Research) grant 3233-069350 from the Swiss National Science Foundation (Dr Verkooijen).
Previous Presentation: This study was presented in part as a poster at the 21st Annual Congress of the European Association of Urology; April 7, 2006; Paris, France.
Additional Contributions: Stina Blagojevic, BSci, provided technical and editorial assistance. Isabelle Neyroud-Caspar, MD, Hyma Schubert, MA, Thomas Agoritsas, MD, Ana Caratsch, MSci, David James from the Association Prosca (Prostate Cancer Patient Association) reviewed the manuscript and provided helpful comments. The registry team provided data and support.