Article Text

PDF
PTU-188 Statistical approaches to missing data in trials of irritable bowel syndrome with constipation: experience with linaclotide (constella®)
  1. M Falques1,
  2. C Díaz1,
  3. D Vilardell1,
  4. J Fortea1,
  5. JM Johnston2
  1. 1Almirall S. A., Barcelona, Spain
  2. 2Ironwood Pharmaceuticals, Cambridge, USA

Abstract

Introduction Missing data may affect validity and interpretation of clinical trials, as different statistical approaches for handling missing data may lead to different conclusions. As patterns of missing data are often unknown until unblinding, study design should be guided by similar trials. Information on handling missing data in trials in irritable bowel syndrome with constipation (IBS-C) is limited.

Method This post-hoc analysis compared imputation methods for missing data in 2 pivotal randomised, double-blind, placebo (PBO)-controlled, multicentre Phase 3 clinical trials (Trials 31 and 302) of linaclotide 290 μg once-daily (Constella®; LIN) in IBS-C. In both trials, pre-specified European Medicines Agency (EMA)-recommended co-primary endpoints were: 1) 12-week abdominal pain/discomfort responder rate and 2) 12-week IBS degree-of-relief responder rate. These endpoints were analysed using observed cases (OC), last observation carried forward (LOCF), baseline observation carried forward (BOCF), drop-out as non-responder and multiple imputation approaches.

Results In Trial 302, the different imputation methods for missing data yielded results consistent with the initial OC approach for both co-primary endpoints (Table 1). Results were similar for Trial 31, with statistically significant treatment differences for all imputation methods for both endpoints (P < 0.0001 for all except for 12-week abdominal pain/discomfort responder: OC, P = 0.0002 and drop-out as non-responder, P = 0.0056).

Abstract PTU-188 Table 1

Imputation methods for Trial 302

Conclusion All five imputation methods yielded non-biassed results in both LIN Phase 3 IBS-C trials, supporting the robustness of LIN treatment effects on EMA endpoints. The ‘drop-out as non-responder’ approach could be considered to be the most conservative, as using this method the treatment-effect estimates for both co-primary endpoints were lower than those obtained using other methods.

Disclosure of interest M. Falques Employee of: Almirall S. A., C. Díaz Employee of: Almirall S. A., D. Vilardell Employee of: Almirall S. A., J. Fortea Employee of: Almirall S. A., J. Johnston Employee of: Ironwood Pharmaceuticals.

Statistics from Altmetric.com

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.