Evolution of bias and sample size in postoperative pain management trials after hip and knee arthroplasty.

Abstract

Bias (systematic error) and small trial sample size (random error) may induce imprecise and exaggerated treatment effects in randomised controlled trials (RCTs). To avoid this, SPIRIT- and CONSORT-guidelines, and Cochrane Collaboration bias recommendations were developed. We investigated risk of bias and trial sample size development over time in postoperative pain trials. This study was based on data from two systematic reviews regarding pain management after total hip arthroplasty (THA) or total knee arthroplasty (TKA). RCTs of analgesic interventions with a comparator control group were included. Primary outcomes were risk of bias and trial sample size developments over time. We calculated cumulated bias scores ranging from 0 to 14 based on Cochrane's seven bias domains (0 = low; 1 = unclear, 2 = high). Developments were evaluated with run and control charts. Further, we compared data from trials published between 1990-1999 and 2010-2016. We included 171 trials published between 1989 and 2016. Overall, the summarised risk of bias decreased, mainly due to better randomization and allocation concealment. Visual inspection suggested an on-going improvement that started around 2007. Trial sample size did not change significantly. For trials published between 1990-1999 and 2010-2016 adequate reporting increased from 36% to 75% for random sequence generation, from 9% to 38% for allocation concealment and from 27% to 52% for blinding of participants/personnel. Risk of bias for RCTs regarding postoperative pain management after THA and TKA has decreased from 2007 to 2016, mainly due to better randomization and allocation concealment. Deficiencies remain. Thus, reporting according to validated guidelines is essential. Sample sizes did not change significantly.