SURVIVAL-CURVE FITTING USING THE GOMPERTZ FUNCTION - A METHODOLOGY FOR CONDUCTING COST-EFFECTIVENESS ANALYSES ON MORTALITY DATA

The analysis of published survival curves can be the basis for increme ntal cost-effectiveness evaluations in which two treatments are compar ed with each other in terms of cost per life-year saved. The typical c ase is when a new treatment becomes available which is more effective and more expensive than the corresponding standard treatment. When eff ectiveness is expressed using the end-point of mortality, cost-effecti veness analysis can compare the (incremental) cost associated with the new treatment with the (incremental) clinical;benefit measured in ter ms of number of life-years gained. The (incremental) cost-effectivenes s ratio is therefore quantified as cost per life-year gained. This pha rmacoeconomic methodology requires that the total patients years for t he treatment and the control groups are estimated from their respectiv e survival curves. We describe herein a survival-curve fitting method which carries our this estimation and a computer program implementing the entire procedure. Our method is based on a non-linear least-square s analysis in which the experimental points of the survival curve are fitted to the Gompertz Function. The availability of a commercial prog ram (PCNONLIN) is needed to carry out matrix handling calculations. Ou r procedure performs the estimation of the best-fit parameters from th e survival curve data and then integrates the Gompertz survival functi on from zero-time to infinity. This integration yields the value of th e area under the survival curve (AUG) which is an estimate of the numb er of patients years totalled in the population examined. If this AUC estimation is performed separately for the two survival curves of two treatments being compared, the difference between the two AUCs permits to determine the incremental number of patient years gained using the more effective of the two treatments as opposed to the other. The cos t-effectiveness analysis can consequently be carried out. An example o f application of this methodology is presented in detail. (C) 1997 Els evier Science Ireland Ltd.