In this issue of the Journal of Clinical Oncology, Brenner and Hakulinen examine absolute and relative survival in cohorts of Finnish breast cancer patients diagnosed before age 50 years [1]. Survival improved over the decades, with relative survival estimates at 10 years increasing from 49% in the 1953 through 1959 cohort to 68% in the 1983 through 1989 cohort, which was associated with similar increases in absolute survival. Long-term follow-up is not available for the most recent cohort, 1993 through 1999. Forty-year estimates for this group have been calculated using a method known as period analysis, which uses the most recent years with available complete information to estimate conditional survival (conditional on having survived the previous interval) for each successive interval without follow-up. The authors cite a previous article [2] to support their contention that this method provides accurate estimates. Based on the observation that relative survival estimates (relative to age-adjusted population estimates) remain less than 1.0 for each decade through 40 years, the authors conclude that breast cancer patients remain at higher risk of death over their entire lifetime. Although the observations are quite interesting, interpretation of them is problematic.Even if the period analysis method of estimation is an improvement over cohort analysis, it cannot be expected to adequately reflect expected survival of current breast cancer patients over 40 years. As noted in the article, conditional survival still must be estimated from patients diagnosed decades ago, at times when disease detection and management may have been quite different. The authors' previous article addresses only 10-year survival estimates. Although superior to cohort analysis, even for this relatively short interval, period analysis did not accurately predict breast cancer survival for that population. Estimates might rea-