Treatment of small-cell lung cancer on protocol: potential bias of results.

During a 27-month period, 215 new cases of lung cancer were diagnosed at five McGill University hospitals. Only 44 patients (20%) so diagnosed were treated on available chemotherapy protocols. Six categories were used to define reasons for nonparticipation. The most important were medical reasons (MRs), 46%; non-medical reasons (NMRs), 20%; and physician preference (PP), 16%. The three remaining categories, representing 18% of exclusions, were death before diagnosis (DBD), surgical treatment (S), and a miscellaneous group (M). Median survival of patients on and off protocol was 10 and 7 months, respectively. Patients with limited disease treated off protocol for NMR and those treated with surgery did better than patients on protocol. Patients with extensive disease not enrolled because of MR did worse, and those excluded because of PP did better than patients treated on protocol. The implication of these findings for other cancer studies is that analysis of chemotherapy trials often represents treatment results in a small proportion of all patients with a given neoplasm. As such, caution must be exercised when extrapolating results to the group as a whole. We suggest that complete demographic data, including proportion of patients participating and reasons for exclusion, be included in all chemotherapeutic trial reports.