Application of Multiple imputation in Analysis of missing data in a study of Health-related quality of life

When a new treatment has similar efficacy compared to standard therapy in medical or social studies, the health-related quality of life (HRQL) becomes the main concern of health care professionals and can be the basis for making a decision in patient management. National Surgical Adjuvant Breast and Bowel Protocol (NSABP) C-06 clinical trial compared two therapies: intravenous (IV) fluorouracil (FU) plus Leucovorin (LV) and oral uracil/ftorafur (UFT) plus LV, in treatment of colon cancer. However, there was a high proportion of missing values among the HRQL measurements that only 481 (59.8%) UFT patients and 421 (52.4%) FU patients submitted the forms at all time points. Ignoring the missing data issue often leads to inefficient and sometime biased estimates. The primary objective of this thesis is to evaluate the impact of missing data on the estimated the treatment effect. In this thesis, we analyzed the HRQL data with missing values by multiple imputation. Both model-based and nearest neighborhood hot-deck imputation methods were applied. Confidence intervals for the estimated treatment effect were generated based on the pooled imputation analysis. The results based on multiple imputation indicated that missing data did not introduce major bias in the earlier analyses. However, multiple imputation was worthwhile since the most estimation from the imputation datasets are more efficient than that from incomplete data. These findings have public health importance: they have implications for development of health policies and planning interventions to improve the health related quality of life for those patients with colon cancer