ANALYSIS OF EARLY AND LATE DEATHS ON RTOG NONSMALL CELL-CARCINOMA OF THE LUNG TRIALS - COMPARISON WITH CALGB-8433

In a major study that showed a treatment advantage for induction chemo therapy followed by radiation therapy (CALGB 8433), there was a signif icantly (P = 0.02) lower proportion of patients dying within 105 days of registration in the chemotherapy/radiation arm than the radiation t herapy arm; without this difference, the overall survival was marginal ly better (P = 0.059) for the chemotherapy/radiation group. A retrospe ctive analysis of RTOG trials sought explanations for the phenomenon. Materials and methods: Patients who fit the CALGB eligibility criteria and received radiation therapy alone in four prospective trials of th e RTOG conducted between 1983 and 1989 were analyzed to determine fact ors that distinguished patients dying within 105 days from longer surv ivors. Two were trials of altered fractionation and two used standard fractionation. Of 683 patients identified, 107 (15.7%) died within 105 days after registration. The log linear model was used to evaluate re lationships between death within 105 days and known prognostic factors . Karnofsky performance status (KPS), <90 vs. greater-than-or-equal-to 90, was the only factor significantly related to death within 105 day s (P = 0.0052). A Cox model with the same factors plus fractionation a nd total dose found KPS and T-stage associated with overall survival ( P = 0.0005 and 0.025, respectively). The choice of the hyperfractionat ion arm (HFX) for Phase III study (69.6 Gy at 1.2 Gy b.i.d.) was based in part on comparison with standard fractionation (STD) from a concur rent RTOG protocol, 8321. Review of early deaths showed that this HFX arm had a lower proportion of patients dying within 105 days (7.9%) th an STD in 8321 (21.0%). After adjusting for KPS and T-stage, there was still a significant treatment difference (P = 0.0006) in favor of HFX . A landmark analysis was done to compare overall survival of patients alive at 106 days between 69.6 HFX and the 8321 STD: the median 1-, 2 -, and 3-year survival rates for HFX were 13.7 months, 61.0%, 31.6% an d 17.7%, respectively, compared with 10.7 months, 44.6%, 13.6% and 8.5 % for STD. These data suggest that stratification for known prognostic factors may not eliminate bias from the results of prospective Phase III trials. Analysis of the intergroup study that replicated the CALGB trial and added the 69.6-Gy arm must consider substrata of pretreatme nt factors that could lead to early deaths or otherwise reveal bias.