Statistical Analysis of a Controlled Clinical Trial in Patients with Metastic Bone Pain

Abstract

The analgesic effect of 600 mg and 1500 mg of a pain killing drug to metastatic bone pain, and associated side effects, were assessed. The experimental design was a double-blind cross over clinical trial involving 44 patients known to suffer from metastatic bone pain. Each patient received the active drug in one of two dosages and the placebo in a random order, each lasting about 14 consecutive days. The data consisted of daily measurements of several pain and side effect variables. A few covariates were available. It was found that the patient and the investigator achieved a high degree of agreement on the blinded preference of the active drug to the placebo. A multivariate analysis of variance (MANOVA) on three different summary scores (mean, median, trmean) calculated on the daily measurements for which the patient received the active drug and on those for which the patient received the placebo was conducted. It was found that for the group of pain variables the order of application and the treatment do not have a significant effect marginally, but that they interact significantly. Variation between subjects was also significant. For the group of side effect variables, however, only significant variation between subjects was found. This suggests that the drug does not have noticeable overall side effects. To account for correlations among the response measurements within each patient, the methodology of generalized estimating equations was used to assess the significance of the effects of the predictors. Although the results are less reliable as they depend on the asymptotic behaviour of statistics, it was found that regardless of the level of correlation within patient response measurements, only the interaction of order of application with treatment has a significant effect on each of the pain variables. All the statistical analyses were carried out using Minitab, SAS, Matlab and Splus.