Performance status (PS) is a measurement of a patient’s functional capabilities and has been validated as a prognostic indicator in patients with cancer. This study was conducted to compare patient and physician-rated Eastern Cooperative Oncology Group (ECOG) PS to determine the incidence of inter-observer variability and identify determinants of PS disagreement among a cohort of patients with hematologic malignancy.
Newly diagnosed patients with leukemia and lymphoma were prospectively enrolled in the University of Iowa/Mayo Clinic SPORE Molecular Epidemiology Resource (MER) from 2002-2008. At diagnosis, patients and their physicians were asked to independently rate ECOG PS. Those who reported PS within 1 month of diagnosis were included in this analysis. Chi-squared and Wilcoxon rank-sum tests were used to assess the association between PS disagreement and prognostic or demographic factors; Cox proportional hazards models and c-statistics were used to evaluate the association between PS and overall survival.
1269 patients were included with a median age of 61 years (range 18-91). Of these, 58% were male; 275 had chronic lymphocytic leukemia (CLL), 127 had Hodgkin lymphoma (HL), and 867 had non-Hodgkin lymphoma (NHL). Overall, 829 (65%) patients and physicians rated PS the same. Age greater than 60 was a significant predictor of disagreement overall (p<0.001), as well as in CLL (p=0.014) and NHL (p=0.004), but not in HL (p=0.65). We found that across disease subtypes, the level of disagreement increased with more aggressive disease. This finding was significant in those with HL (p=0.027) and NHL (p<0.0001) as International Prognostic Score (IPS) and International Prognostic Index (IPI) increased respectively. In CLL, the trend for Rai stage was suggestive but not statistically significant (p=0.27). The incidence of disagreement was highest among patients with NHL (37%), with a higher percentage of disagreement among those with aggressive (45%) vs. non-aggressive (25%) subtypes (p<0.001). Disagreement was lower in patients with CLL (27%) and HL (33%). There were no significant associations between patient gender or education level and PS rating among subtypes or overall.
Patient and physician-rated PS were both significant predictors of overall survival in univariate models and also adjusted for subtype and subtype specific risk score. The prognostic ability for PS was similar for both patient rated and physician-rated assessment overall (c-statistic=0.76 for both patient and physician rated PS) and in those with HL (c-statistic=0.85 for both patient and physician rated PS) and NHL (c-statistic=0.76 for both patient and physician rated PS). However, patient-rated PS was better for prognostication in CLL (c-statistic=0.75; p<0.0001), compared to physician-rated PS (c-statistic=0.67; p=0.002).
Patients with hematological malignancies and their physicians do not always rate PS the same, particularly in patients who are older or have more advanced or aggressive disease. These findings suggest the need for physicians to communicate with patients when determining PS, as PS is a strong predictor of survival.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.