Discriminant analysis of lung function test results in the selection of patients for bronchial carcinoma surgery

Discriminant analysis was applied to the results of lung function tests carried out on patients prior to thoracotomy for carcinoma of the bronchus. A group of 64 patients who subsequently suffered post-operative complications were compared with one of 78 who were complication free. The largest differences between the mean discriminants of the groups using from 2 to 9 attributes was with combinations which included the residual volume and maximum voluntary ventilation values in every case. The differences increased little from the 2- to the 9-attribute analyses. Using the group mean discriminate as limits, two sets of 240 patients and one of 116 patients were divided into 3 risk categories. The findings were compared with those using a previously published method based upon empirically determined limits for the lung function test results. Consistent findings indicated that it was necessary to add the unweighted maximal mid-expiratory flow rate and whether the operation was a right pneumonectomy to the 2-attribute discriminant procedure to give one that was comparable to the empirical method. It was concluded that little improvement can be brought about using discriminant analysis on the test results although the new procedure may prove more convenient in practice.