ABSTRACT
Missing data are inevitable in any longitudinal study in which over an extended period of time patients experience morbidity or mortality due to disease or its treatment. This is especially true for trials involving patients with cancer. While discontinuation of treatment at the patient’s request is rare, therapy is often discontinued as a result of excessive toxicity, disease progression, and death. When the assessment of patient reported outcomes, such as symptoms and quality of life (QOL), is stopped after the discontinuation of treatment, it is highly likely that the outcome differs between those patients who have and have not been measured. Specifically, the causes of dropout will result in negative outcomes such as increased symptom
burden that will in turn impact the patient’s quality of life. We cannot ignore this missing data if we wish to obtain unbiased estimates of the longitudinal changes. The focus of this chapter is the interpretation of the results from various analytic strategies of studies with nonignorable dropout. Some missing data may also be unrelated to the patient’s QOL resulting from administrative problems such as staff forgetting to give the forms to the patient during a very busy time. This unrelated missing data is preventable and should be minimized in a well-conducted trial.
