A dilemma in analysis: issues in the serial measurement of quality of life in patients with advanced lung cancer.

Despite the availability of several instruments to evaluate quality of life (QL) over time in patients with lung cancer, barriers in measurement remain. This methodological study used LCSS data (Lung Cancer Symptom Scale, a disease- and site-specific QL measure) to examine analysis methods to quantify QL where data needed for serial evaluation may be missing. Data from two large randomized trials, conducted at 30 centers, of a new combination chemotherapy regimen incorporating a new agent for patients (n = 673) with Stage III and IV non-small cell lung cancer were obtained for this study. QL had been prospectively measured at baseline, day 29, and every six weeks thereafter using the LCSS. For the slope analysis (SA) and area under the curve (AUC) analyses, an adjustment score of zero was used to indicate QL on the day of death (mortality adjustment) and each subsequent day until the end of the assessment period. Significant differences in QL, symptom scores and known prognostic factors at baseline were found in the attrition group. SA and AUC analysis allowed inclusion of 581 patients, giving an adequacy rate of 86%. By using a mortality adjustment, an additional 45 patients were included, increasing the inclusion rate to 93%. With the use of the mortality adjustment, QL was shown to decline over the interval, as opposed to rise if the adjustment had not been performed. The conclusions of the study were: (1) analysis for serial data using SA and AUC provides useful, but differing information; (2) when attrition (caused by death) is a factor, a mortality adjustment presented a more accurate assessment of QL as an endpoint; (3) more frequent evaluations of QL will capture rapid changes in patient status and reduce the attrition bias; (4) all patients should be followed until they die; and (5) QL should be given full consideration as a primary endpoint separate from survival.