The Quality of Life Questionnaire Core 30 (QLQ-C30) and Functional Assessment of Cancer-General (FACT-G) differ in responsiveness, relative efficiency, and therefore required sample size (original) (raw)

2014, Journal of Clinical Epidemiology

Objective: QLQ-C30 and FACT-G are widely-used cancer-specific health-related quality of life (HRQOL) questionnaires. We aimed to compare their responsiveness to clinically important effects and statistical efficiency to detect such effects. Study design and setting: Secondary analysis of QLQ-C30 and FACT-G data from a randomised controlled trial of Medical Qigong (n=162 heterogeneous cancer patients). Difference in responsiveness (DR) and relative efficiency (RE) were calculated for five domains. Results: FACT-G total score was more efficient than QLQ-C30 global scale for detecting change within the intervention arm (RE=0.31 (0.083, 0.69)) and comparing change between trials arms (RE=0.17 (0.009, 0.58)). In the social domain, the QLQ-C30 scale was more responsive (DR=0.28 (0.024, 0.54)) and more efficient within arm only (RE=5.25 (1.21, 232.26)). In the physical, functional/role and emotional domains, neither questionnaire was more responsive or efficient. Conclusion: FACT-G would require about one third the sample of QLQ-C30 to detect a given change in overall HRQOL, while in the social domain it would require five times the sample size. FACT-G won advantage in overall HRQOL by reduced "noise" (smaller standard deviation achieved by summing across 27 items), while QLQ-C30 won advantage in the social domain via a larger "signal" (achieved through well-targeted item content). What is new? • As a measure of overall health-related quality of life (HRQOL), the FACT-G total score is more responsive to change over time than the QLQ-C30 global scale, and has greater statistical efficiency and hence power, for both change within a group and for comparing change between two groups. • In the social domain, the QLQ-C30 scale is more responsive than the FACT-G scale, and has greater statistical efficiency and hence power for the within-group change but not for comparing change between two groups. • A randomised trial which used the QLQ-C30 to assess overall HRQOL would require a sample size approximately five times greater than one which used the FACT-G to detect a given difference between trial arms as statistically significant. FACT-G would require about one third the sample of QLQ-C30 to detect a given change in overall HRQOL, while in the social domain it would require five times the sample size. • In the physical, emotional and role/functional domains, the FACT-G and QLQ-C30 have similar responsiveness, statistical efficiency and hence power and sample size requirements.