Reliable data on 5- and 10-year survival in early-stage breast cancer provides accurate estimates of 15-year survival

Abstract #4076 Background: Evidence-based medicine encourages utilization of empirical evidence from controlled clinical trials in decision making both at the bedside and in health policy. However, the applicability of this approach may be limited by the lack of long-term results from efficacy studies. Decision analytic methods that model outcomes beyond empirically available evidence, typically over the lifetime horizon, have been proposed. There have been few studies of the accuracy of model-based survival projections utilizing reliable but early empiric results for estimating long-term survival. Such results, however, would have direct clinical and health policy implications. Methods: We took advantage of recently published individual-patient data meta-analysis by Early Breast Cancer Trialists' Collaborative Group (EBCTCG) (Lancet 2005;365:1687-1717) which regularly updates their results at 5-year intervals. A Markov model was built to compare empirical results with those modeled beyond the time horizon for which data was available. We first modeled breast cancer mortality rates comparing the effects of 5 years of tamoxifen versus no treatment. We then modeled the 15-year probability of death from breast cancer using the 5- and 10-year mortality data reported in the overview. Finally, we estimated the effect of breast cancer on life expectancy in women ranging in age from 40 to 70 years by decade. Results: Verification of the model produced 5-, 10-, and 15-year survival data that were identical to the empirical data reported in the EBCTCG overview. Modeling 15-year survival based on 5- and 10-year empirical data resulted in an average error estimate in breast cancer mortality of 0.75% [range: -0.83 to 2.19%]. Although modeling of differences in life expectancies ranged from an underestimate of -7.93% to an overestimate of 12.64%, in absolute terms over the span of 15 years this corresponded to a loss of 18 days or a gain of 40 days of life, respectively. Regardless of the assumptions used in the model, the decision remained stable: tamoxifen remained the better choice compared to no therapy. Conclusions: We conclude that reliable data generated at 5 or 10 years may be used to generate models that estimate 15-year survival. Whether 5-year data can be employed over the lifetime horizon remains to be demonstrated. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 4076.