Observations on survival curve analysis with particular reference to breast cancer treatment.

The proper analysis of survival curves of patients treated for cancer is critical to advances in clinical investigation. In this article, the presentation of such survival curves is reviewed and the various possible effects of treatment on these curves are discussed. The results of breast cancer treatment are used as examples. It is argued that the use of a semilogarithmic presentation is preferred in displaying the long-term force of mortality of the disease. In many human cancers, such as carcinoma of the breast, the demonstration of a "cured" subgroup requires follow-up of greater than 10 years and cannot be inferred from data evaluated earlier. Beneficial effects noted on the initial portion of the survival curve are not necessarily associated with an improvement in the likelihood of long-term survival, but are rather a reflection of delay or reduction on the initial force of mortality. Since both hosts and tumors are heterogeneous, it is also possible that there are different effects on various subgroups, making curve interpretation complex. In some cases, an early detrimental effect may obscure a benefit in the long-term outcome. These considerations emphasize that effects seen on the early portion of survival curves may not only be premature, but may also be misleading.