Automatic search for optimal conditions in clinical studies.

Background: A program for an automatic search for the optimal conditions of important prognostic factors in clinical studies was developed. Methods: The program was developed for the following steps: (1) Input of the maximum and minimum values and of the interval of the variable to be investigated. Automatic calculation of the Cut Points. (2) Division of the patient data into two groups at every Cut Point and calculation of survival rates. (3) Sequential calculation of P-values and chisquare values for the two sets of survival rates. To determine the usefulness of this program, the optimal irradiation dose was searched for 537 patients with non-small cell lung cancer. Results: The P-value reached its minimum value and the chisquare value its maximum when the Cut Point was 5, 925cGy (0.0001 and 30.18). Between 5,925cGy and 6,900cGy, the P-value stayed at less than 0.05. Conclusion: Artificial errors in grouping by prognostic factors can be avoided and the search for optimal conditions can be conducted automatically and scientifically.