Article Text


OP0117 The effect of anakinra in slowing radiologic progression in rheumatoid arthritis patients ? evidence from a bootstrap analysis
  1. MW Cravets,
  2. E Ng,
  3. MB Bear
  1. Department of Biostatistics, Amgen Inc., Thousand Oaks, USA


Background Anakinra was studied in a randomised placebo-controlled study of 473 patients with rheumatoid arthritis (RA). Patients were randomised to one of 4 treatment groups (placebo, 30 mg, 75 mg or 150 mg anakinra) and treated for 24 weeks. Radiographic assessments of the hands and wrists based on Larsen score (Larsen 1977) were collected at baseline and 24 weeks. At the end of the study, about 27% of the patients withdrew. To assess the anakinra effect on reducing joint damage, the potential influence of dropouts on the study conclusions must be taken into account.

Objectives To demonstrate that the effect of anakinra in reducing the rate of radiologic disease progression is robust and insensitive to potential biases due to losses of followup.

Methods The 3 anakinra dose groups were combined into a single treatment group which was then compared to placebo. For each treatment group, x% of the missing Larsen score changes from baseline to week 24 were imputed by randomly selecting the observed change scores from the same treatment group, and the rest of the missing data were imputed similarly from the opposite treatment group. The resampled (bootstrapped) data were then analysed by an ANOVA model. This resampling procedure was repeated 1000 times, resulting in the estimation of 1000 p-values for testing the treatment effect. These 1000 p-values were then summarised to give an estimated picture of the IL-1ra treatment effect. By changing the proportion of sampling (x) systematically, the effect of anakinra on radiologic progression can be studied under different scenarios of missing data.

Results The results from all the analyses indicated that there was improvement in radiologic progression with anakinra treatment for 24 weeks. Statistical significance was maintained when up to 40% of the missing data were sampled from the opposite treatment group (mean and median p-values range from 0.021 to 0.078, and 0.011 to 0.05 respectively).

Conclusion Anakinra has a statistically significant effect on slowing the progression of disease for 24 weeks of treatment as measured by the Larsen Score, even when as much as 40% of the missing anakinra scores are imputed with placebo values.

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.