Michael Froehner, Department of Urology, University Hospital ‘Carl Gustav Carus’, Technical University of Dresden, Fetscherstrasse 74, D-01307 Dresden, Germany. e-mail: email@example.com
Study Type – Outcomes (cohort)
Level of Evidence 2b
What's known on the subject? and What does the study add?
Several comorbidity classifications have been investigated for their suitability to assist treatment decision-making in men with early prostate cancer. In unselected patients, some serious comorbidities have been shown to be associated with a 10-year competing mortality rate clearly superseding the 50% level.
The present study shows that it is hardly possible to discern meaningful subsets of patients with a 10-year risk of competing mortality of >50% by using comorbidity classifications. This finding suggests that the selecting clinicians did well in estimating the medium-term survival probability in men referred for radical prostatectomy.
• To identify subsets of patients who are most likely to die from competing causes ≤10 years after radical prostatectomy (RP).
PATIENTS AND METHODS
• In all, 2205 consecutive patients who underwent RP for clinically localized prostate cancer between 1992 and 2005 were studied. The 10-year cumulative competing mortality rates were determined in several worst-case scenarios formed by using comorbidity classifications and combinations of them.
• In this sample of men selected for RP, even those with the most severe comorbidity level had a competing mortality risk of <50% ≤10 years after RP.
• Depending on the comorbidity classification used, the 10-year cumulative competing mortality rates differed between 16 and 39% in the whole sample and between 18 and 48% in men aged ≥65 years.
• Clinicians do well in estimating the further life span in candidates for RP. Comorbidity classifications may assist treatment choice in this population but are not able to discern meaningful subsets to be excluded from curative treatment because of a life expectancy falling below a limit of 10 years.
the New York Heart Association classification of cardiac insufficiency
Canadian Cardiovascular Society classification of angina pectoris.
Among malignant diseases, prostate cancer is one of those in which the estimation of individual long-term life expectancy is of particular clinical concern. Traditionally, although not proven by controlled clinical trials, a life expectancy of >10 years has been required to consider men with early prostate cancer fit for curative treatment [1,2]. Therefore, candidates for curative treatment, especially for radical prostatectomy (RP), had undergone a selection process that eliminated those considered likely to die from competing causes ≤10 years after diagnosis. Even after this selection process a further stratification by the remaining life span is desirable to support clinical decision-making. This applies particularly to men aged ≥65 years in whom a demonstration of a superiority of immediate curative intervention is still pending  and who represent more than half of all incident prostate cancer cases .
The selection process may change the prognostic significance of comorbid conditions. A comorbidity level indicating unfitness for curative treatment in unselected patients  or patients who received no curative treatment ≤180 days after diagnosis  may be associated with a high long-term survival in a RP cohort . In the present study we looked for subsets of patients with the highest risk of dying from competing causes ≤10 years after RP.
PATIENTS AND METHODS
In all, 2205 consecutive patients who underwent RP for clinically localized prostate cancer between 1 December 1992 and 31 December 2005 at our institution constituted the study population. Institutional Review Board approval was obtained. The mean age was 64.2 years. Among patients without neoadjuvant hormonal treatment, the mean PSA level was 11 ng/mL. In all, 63% of patients had histopathologically organ-confined, node-negative disease. During follow-up 269 patients died, 84 from prostate cancer and 185 from competing causes. In the surviving patients, the mean follow-up was 7.9 years. Only one patient was lost to follow-up.
The following comorbidity classifications were used: the American Society of Anesthesiologists (ASA) physical status classification , the New York Heart Association (NYHA) classification of cardiac insufficiency , the classification of angina pectoris of the Canadian Cardiovascular Society (CCS)  and the Charlson score . Furthermore, a disease count was calculated by adding all concomitant diseases recorded in the database (angina pectoris, hypertension, history of thromboembolism, body mass index of ≥30 kg/m2, myocardial infarction, cardiac insufficiency, peripheral artery disease, cerebrovascular disease, lung disease, ulcer disease, mild liver disease, diabetes mellitus, connective tissue disease, hemiplegia, moderate or severe renal disease, solid tumour, leukaemia, lymphoma, moderate of severe liver disease, dementia, metastatic solid tumour) corresponding with an approach described by Houterman et al. . The comorbidity data were obtained from the preoperative cardiopulmonary risk assessment and the discharge records.
Several comorbidity classes and combinations of them were investigated as possible worst-case scenarios indicating severe potentially life-shortening comorbidity, if they were considered clinically plausible (combinations between the closely related classifications disease count and Charlson score were not used) and comprised at least 1% of patients (a limit below that a classification was considered clinically meaningless).
Competing (other than prostate cancer) and prostate cancer mortality were the study endpoints. Deaths were attributed to prostate cancer if uncontrolled progressive disease was present. Otherwise, deaths were attributed to competing causes. Survival data was collected from hospital files, private practice urologists and GPs, the patients themselves, their relatives, cancer registries or local authorities. Causes of death were reviewed for plausibility by a senior urologist (M.F.) and additional information was obtained if necessary.
The individual contribution of competing causes of death (prostate cancer vs other causes) was analysed by competing risk analysis resulting in an estimation of cumulative incidence rates for the different correlated risks .
The cumulative mortality curves for prostate cancer-specific and competing mortality in the whole sample are shown in Fig. 1. The 10-year prostate cancer-specific mortality rate was 5% (95% CI 4–6%), the 10-year competing mortality rate was 11% (95% CI 9–12%). The corresponding 15-year rates were 10% (95% CI 6–14%) and 23% (95% CI 18–28%).
Depending on the comorbidity classification used, in the investigated worst-case scenarios, the 10-year cumulative competing mortality rates ranged between 16 and 39% in the whole sample and between 18 and 48% in men aged ≥65 years (Table 1). A separate analysis of the 19 single comorbid conditions contributing to the Charlson score  revealed 10-year cumulative competing mortality rates between 8 and 26% (in patients aged <65 years) and between 4 and 39% (in patients aged ≥65 years), respectively (data not shown).
Table 1. The 10-year competing and prostate cancer-specific mortality rates with 95% CIs after stratification by several comorbidity classifications representing possible worst-case scenarios concerning the probability of premature competing mortality in the whole sample and in the subset of patients aged ≥ 65 years
NA, not available; *the three highest 10-year competing mortality rates considering the whole sample; †the three highest 10-year competing mortality rates considering the subgroup of men aged ≥ 65 years.
Cumulative mortality curves for the three scenarios with the highest 10-year competing mortality rates are shown in Fig. 1 for the whole sample (A–C) and for men aged ≥65 years (D–F). With rates of 0–12%, the10-year prostate cancer-specific mortality rates were low in all subgroups (Fig. 2).
Curative prostate cancer treatment would ideally be offered to men with a life expectancy of >10 years [1,2]. The present study shows that clinicians, although occasionally called into question [5,13–15], do well in estimating the life expectancy at least in candidates for RP, when a competing mortality rate <50% within 10 years is considered to represent a remaining life span of >10 years. Even in the subsets with the highest classifiable comorbidity risks, competing mortality did not reach this critical level.
Clinical guidelines recommend considering comorbidity in men with early prostate cancer to avoid overtreatment . Applying this, irrespective of the comorbidity measure used, clinicians need to know what the likelihood of long-term survival the patients with the highest risk of competing mortality in a certain population have to have a rational basis for making a treatment recommendation. Will most patients with a certain comorbidity prematurely die from competing causes requiring a reluctant recommendation of curative treatment or not? Considering the multitude of comorbidity measures and combinations of them tested (Table 1), the results of the present study suggest that it will be difficult if not impossible to develop a comorbidity classification that discerns a meaningful proportion of patients of whom most will die from competing causes ≤10 years after RP.
In unselected patients, a degree of comorbidity may suggest unfitness for curative treatment (a Charlson score of ≥3 with a 10-years competing mortality rate of 70% ), with that patients after having passed the filter ‘fit for RP’ have a high 10-year survival probability (in the present series with a 10-years competing mortality rate of only 27% ). Patients treated by external beam radiotherapy undergo a less strict selection than those referred for RP. A comorbidity level associated with a 70% 10-year competing mortality rate in the external beam radiotherapy setting may be associated with an only 30% rate in those selected for RP . In men aged ≥66 years selected for ‘receiving no curative treatment ≤180 days of diagnosis’, most with serious comorbidity indicated by a Charlson score of ≥2 died from competing causes ≤10 years after diagnosis . In the present study, only about a quarter of patients aged ≥65 years with a Charlson score of ≥2 died from competing causes ≤10 years after RP (Table 1). The discrepancies in the survival rates between this RP series and unselected men  or men selected for deferred or no curative treatment , underline the great importance of a careful consideration of the clinical setting and the degree of selection when treatment recommendations are based on comorbidity classifications. Considering aggressive therapy as overtreatment in men with a Charlson score of ≥3  is probably not generally justified . Recently published decision-aiding tables in elderly men  should be used, keeping in mind that the underlying population probably was selected by adverse comorbid risks by excluding those patients who received immediate curative treatment.
Do the results of the present study also apply to other samples? Compared with a large series recruited between 1987 and 2005 with a similar age profile (mean age 61  vs 65 years), the 10-year cumulative mortality rates from prostate cancer and competing causes (Fig. 1) were virtually the same. Compared with the Scandinavian randomised trial that recruited in the years 1989–1999 , both prostate cancer-specific and competing mortality rates were lower in the present series, possibly due to a more recent recruitment period and a corresponding more favourable tumour-related risk profile. Considering this data, it seems possible to draw the conclusion that for contemporary series application of the findings of the present study to earlier series like the Scandinavian randomised trial  should be done only with caution taking differences in patient selection into consideration.
Unlike in unselected patients [4,17,18] or those receiving no immediate curative treatment , few comparable data on the prognostic impact of serious comorbidity is available for men selected for RP. Two studies from the PSA era [14,15] provided 10-year competing mortality rates for patients undergoing RP with a Charlson score of ≥2. Their results were quite similar to the present series suggesting a wider applicability of the findings of the present study at least concerning the Charlson score.
The present study has several limitations. Although comprising a wide range of concomitant diseases and comorbidity classifications, it is possible that some meaningful parameters were not recorded in the database used. Some classifications relied on subjective clinical judgment. Validation studies in different samples would be desirable. A longer follow-up could provide narrower CIs. In some stratifications of patients aged ≥65 years (Table 1), the CIs were too wide to allow definite conclusions. Using the 10-year limit as a surrogate parameter for adequate life expectancy refers to a widely used clinical agreement but to data from controlled trials. Besides survival rates, functional and health-related quality-of-life outcomes (which were not assessed in the present study) determine treatment benefit. Recording of a wider spectrum of comorbid conditions than doable in large population-based samples is a possible strength of the present study.
In conclusion, the results of the present study suggest that discerning meaningful subsets of patients with a 10-year risk of competing mortality of >50% is hardly possible by stratifying men referred for RP by their comorbid risk factors. A careful consideration of the degree of selection a population has undergone is required when treatment recommendations in men with early prostate cancer are based on comorbidity classifications.