EURURO Vol. 71 No. 5

between 1992 and 2007 in order to determine which

parameters predicted competing mortality in this particular

population. Furthermore, we compared the prognostic

impact of individual parameters between patients selected

for radical prostatectomy aged 70 yr or older with their

younger counterparts. Potential prognostic parameters

were obtained from preoperative evaluation records and

discharge letters. Only parameters recorded in at least five

patients were included in the model calculation. Deaths in

the absence of uncontrolled prostate cancer progression or

from unknown causes (

= 2) were considered deaths from

competing causes. Demographic data of the study sample

are shown in Supplementary Table 1.

Besides age, five parameters were independent predictors

of competing mortality with hazard ratios of around 2 in

patients aged 70 yr or older

( Table 1

). With diabetes mellitus,

chronic lung disease, and other cancer, three common

comorbid conditions independently predicting competing

mortality in younger patients were not thoroughly associat-

ed with competing mortality beyond the 70th yr of age

(Supplementary Table 2). Based on the model obtained in the

latter population, patients were subdivided into quartiles

concerning the parametric part of model-predicted hazards

(Supplementary Fig. 1). Only patients above the fourth

quartile reached the 10-yr mortality rate (29.4%) expected in

the local normal population with the same age structure

(data source:

www.statistik.sachsen.de

Giving one point for each parameter of the model in

Table 1

except for age (ie, for peripheral vascular disease,

cerebrovascular disease, American Society of Anesthesiol-

ogists [ASA] physical status class 3, current smoker, and no

university degree or master craftsman or unknown level of

education as measures for possible occupational or life

style-related risk factors), we calculated a score and

compared it with five other measures of comorbidity for

which data were available in our database (Charlson score,

modified Charlson score, modified Lee mortality index,

prostate cancer specific comorbidity index, and unweighted

Charlson score)

[5–9]

. Age was excluded because not all

comparators contained age-related variables and the age

spectrum of patients selected for radical prostatectomy at

an age of 70 yr or older is limited. The score identified in this

study performed best concerning the separation of the

mortality curves (illustrated by the

values of comparisons

of neighboring curves) and the Akaike’s information

criterion figures, respectively

( Fig. 1

). The obtained simple

score reproduced remarkably well with the multivariable

model and reached narrowly the same Akaike’s information

criterion figure as the stratification into quartiles shown in

Supplementary Figure 1 (1425 vs 1422).

Besides accurate risk prediction, simplicity of use is of

concern with comorbidity classifications in men with early

prostate cancer

[9]

. In the suggested score, the five

identified parameters representing different health aspects

may rapidly be assessed during daily clinical practice.

The ability of clinicians to predict life expectancy has

been called into question and available models have not

been considered as better than government life tables

[1]

. This study provides arguments for the opposite. In the

subgroup of patients selected for radical prostatectomy

aged 70 yr or older, government life tables would predict a

10-yr mortality rate of 29.4% which corresponds with the

Table 1 – Optimal models predicting competing mortality in patients selected for radical prostatectomy at an age of <70 yr versus those who

underwent surgery at an age of 70 yr or older

70 yr (

= 2418)

70 yr (

= 543)