Optimised comorbidity indices can reflect patient performance status in register based studies of prostate cancer

Eastern Cooperative Oncology Group Performance Status (ECOG-PS) is commonly used in cancer trials to select a study population with good performance, but ECOG-PS is rarely available in health-care registers. We assessed if patient age and comorbidity indices can substitute ECOG-PS when selecting men in register-based studies of advanced prostate cancer. ECOG-PS data for 3966 men on androgen deprivation therapy for prostate cancer were retrieved from Prostate Cancer data Base Sweden. Logistic regression models were used to discriminate between ECOG-PS 0-1 versus 2-4 based on age, Charlson comorbidity index (CCI), a novel Multidimensional Diagnosis-based Comorbidity Index (MDCI) based on ICD codes, and a Drug Comorbidity Index (DCI) based on filled prescriptions. The model based on age, MDCI, and DCI provided the best discrimination (AUC = 0.82; 95% CI 0.81-0.84). In a hypothetical cohort of 1000 men where 750 men had ECOG-PS 0-1, 600 men would be included when excluding those with high risk of ECOG-PS 2-4 using this model and 60 of these would have ECOG 2-4 instead of 250 men if all 1000 men had been included. Age and two new comorbidity indices can with reasonable precision substitute ECOG-PS and help identify subsets of study populations likely to have favourable ECOG-PS.