• Lars Peter Kloster Andersen, Ismail Gögenur, Henrik Torup, Jacob Rosenberg, and Mads Utke Werner.
• From the Departments of *Anesthesiology and †Surgery, Roskilde and Køge Hospital, University of Copenhagen, Roskilde, Denmark; Departments of ‡Anesthesiology and §Surgery D, Herlev Hospital, University of Copenhagen, Herlev, Denmark; and ‖Multidisciplinary Pain Center 7612, Neuroscience Center, Rigshospitalet, Copenhagen, Denmark.
• Anesth. Analg. 2017 Sep 1; 125 (3): 1008-1013.

BackgroundPain intensity ratings and opioid consumption (OC) are ubiquitous indicators of pain in postoperative trials of the efficacy of interventional procedures. Unfortunately, consensus on the appropriate statistical handling of these outcomes has not been reached. The aim of this article was, therefore, to reexamine original data obtained from a postoperative analgesic drug trial, applying a collection of standard statistical methods in analgesic outcome assessments. Furthermore, a modified integrated assessment method of these outcomes was evaluated.MethodsData from a randomized, double-blind, placebo-controlled study investigating the analgesic efficacy of a regional anesthetic block following a major elective surgical procedure were analyzed. The original data included measurements of pain intensity (visual analog scale [VAS]) at rest and during coughing (VAS0/2/4/6/12/18/24 h) and OC0-6/0-24 h administered by patient-controlled analgesia. The statistical analyses included comparisons of discrete pain intensity scores (VAS0/2/4/6/12/18/24 h), summary measures of pain intensity ratings (area under the curve [AUC]-VAS0-6/0-24 h; mean VAS0-6/0-24 h), and OC0-6/0-24 h. Finally, the analyses also included an integrated assessment of longitudinally measured pain intensity and opioid consumption (PIOC0-6/0-24 h). Also, estimation of effect size, generalized odds ratio of the individual analgesic outcome variables was performed.ResultsSixty-one patients were included in the final data analysis. Discrete pain intensity ratings differed significantly between the treatment groups at specific postoperative time points, but appropriate correction for multiple comparisons eliminated some of these differences. AUC-VAS0-6 h differed significantly at rest and during coughing, while no difference was found for AUC-VAS0-24 h. In contrast, mean VAS0-6 h and VAS0-24 h differed significantly between treatment groups at rest and during coughing. OC0-6/0-24 h differed significantly between the treatment groups. Finally, also PIOC0-6/0-24 h differed significantly at rest and during coughing.ConclusionsOur analyses demonstrate that the applied statistical method may alter the statistical significance and estimates of effect size of analgesic outcome variables in postoperative pain trials. Our findings underline the importance of defining valid statistical methods for future analgesic drug trials. We propose an integrated assessment of longitudinally measured pain intensity and opioid consumption (PIOC). The method combines two interdependent analgesic outcomes, lowers the risk of mass significance, and provides more accurate representation of the dynamic nature of postoperative pain and analgesic drug efficacy.

22 keywords...

hide…