Survival patterns of cancer patients

Optimal means of modeling death rates of large populations with a specific disease have not been described in the literature.

Statistical modeling was used on archival data.

In the authors' prior publications describing the survival of untreated cancer patients, data that were adequately fitted by an exponential curve were found to be much better fitted by an inverse Gompertz curve (R(2) = 99.7% for untreated breast carcinoma, 99.9% for untreated cervical carcinoma). However, when data from treated patients are examined, fits show that successive stages begin at successive positions on the inverse Gompertz curve. Breast carcinoma data showed that treatment begun at an early stage raised survival to a linear decline; at an intermediate stage led to a modified inverse Gompertz, the earlier the stage at which therapy was begun, the greater the survival rate; and at a late stage exhibited an exponential decline showing a negligible effect of late treatment. Confidence in this approach was enhanced by its applicability using published data from the National Cancer Database for breast, pancreatic, bone, and skin cancers.

The use of data modeling allowed us to assess realistically the value of intervention in cancer populations and to optimize staging schemas. It strongly reinforced the concept that early detection provides a far greater impact on a population's subsequent survival than therapy of advanced disease.