Randomized trials with quality of life endpoints: are doctors' ratings of patients' physical symptoms interchangeable with patients' self-ratings?

The assessment of physical symptoms is a key component of quality of life studies in palliative care, but is often hampered by missing data from patient-completed questionnaires. In two large multicentre randomized trials of palliative treatment conducted by the Medical Research Council Lung Cancer Working Party, Involving over 700 patients, patients completed Rotterdam Symptom Checklists and doctors reported on eleven of the same physical symptoms at each assessment, using the same 4-point severity scale. Ratings by doctors and patients were compared with respect to compliance, severity, and outcomes for the respective trials. Doctors provided more data than patients: 66% vs. 52% in the first 6 months in one trial, 58% vs. 61% in the other. Comparisons of over 33,000 symptom assessments showed 78% complete agreement between doctor and patient, 18% disagreement by one, 4% two, and 1% three grades (complete disagreement). There was no change in levels of agreement over time, but increasing disagreement with increasing symptom severity, and a consistent bias towards doctors underestimating severity. Nevertheless, the two methods of data collection resulted in similar between-treatment conclusions. Therefore, in randomized trials the doctors' assessments of key physical symptoms may be sufficient for the between-treatment comparison. However, the fact that doctors underestimate symptom severity 15% of the time has important implications for palliative interventions.