Abstract
Background
It is important to minimize placebo rates in randomised controlled trials (RCTs) to efficiently detect treatment differences between interventions. Historically, high placebo rates have been observed in clinical trials of ulcerative colitis (UC). A better understanding of factors influencing placebo rates may lead to more informed clinical trial design.
Objectives
A systematic reviewandmeta-analysiswas conducted to evaluate placebo response and remission rates in RCTs evaluating UC treatments in adult patients.
Search methods
Electronic databases (i.e. MEDLINE, EMBASE, and CENTRAL) were searched from inception to 1 March 2017 with no language restrictions applied. Reference lists and conference proceedings of major gastroenterology meetings were also handsearched to identify additional studies.
Selection criteria
Placebo-controlled RCTs of adult patients with UC treated with corticosteroids, aminosalicylates, immunosuppressives or biologics were eligible, provided enrolment and outcome assessment was conducted using the Ulcerative Colitis Disease Activity Index (UCDAI) or the Mayo Clinic Score. The minimum trial duration was two weeks for induction trials and four months maintenance trials.
Data collection and analysis
Pairs of authors independently determined study eligibility and extracted data with any disagreements resolved through consensus. Outcomes of interest included the proportion of patients with clinical response and remission. Trial characteristics such as the design, participant demographics and disease history, interventions, and enrolment and assessment criteria were also recorded. The methodological quality of the included studies was evaluated using the Cochrane risk of bias tool. Pooled placebo response and remission rates and 95% confidence intervals (95% CI) were calculated using a binomial normal model for proportions. Induction of remission and maintenance studies were pooled separately. The impact of study-level characteristics on placebo response and remission rates was investigated using mixed-effects meta-regression analyses with logits of event rates as the outcome variables. An assessment of pooled placebo rates over time was conducted using a cumulative meta-analysis based on date of publication. Publication bias was examined using funnel plots.
Main results
The screening process identified 61 included studies which encompass 58 induction phases (5111 patients randomised to placebo) and 12 maintenance phases (1579 patients randomised to placebo). For induction trials, the pooled estimate of placebo response was 33% (95% CI 30% to 36%) while the pooled estimate of placebo remission was 12% (95% CI 9% to 15%). For maintenance trials, the pooled estimate of placebo response was 23% (95% CI 19% to 28%) while the pooled estimate of placebo remission was 17% (95% CI 10% to 27%).
Studies enrolling patients with more active disease confirmed objectively by endoscopy were associated with significantly lower placebo remission and response rates than trials enrolling patients with less active disease (27% versus 4%, OR 2.60, 95% CI 1.25 to 5.42, P = 0.01 for UCDAI endoscopy sub score >= 1 versus >= 2 for remission; and 27% versus 4%, OR 1.70, 95% CI 1.02 to 2.82, P = 0.02 for UCDAI endoscopy sub score greater than or equal to one versus greater than or equal to two for response). With respect to drug class, the lowest placebo response and remission rates were observed in trials evaluating corticosteroids (23%; 95% CI 19 to 29%, and 5%; 95% CI 2 to 11%, respectively). Trials of biologics had the highest placebo response rate (35%; 95% CI 30 to 41%), while trials evaluating aminosalicylates had the highest placebo remission rate (18%; 95% CI 12 to 24%). Disease duration of greater than five years prior to enrolment was associated with a significantly lower placebo response rate compared to disease duration of less than or equal to five years (29% versus 47%, respectively; OR 0.54, 95% CI 0.32 to 0.92, P = 0.02). The requirement of a minimum rectal bleeding score for study eligibility was associated with an increased placebo response rate compared to studies that did not use rectal bleeding for trial eligibility (37% versus 32%, respectively; OR 1.70, 95% CI 1.02 to 2.82, P = 0.02). Finally, the time point of primary outcome assessment was found to be significantly associated with placebo remission rates such that every one week increment in endpoint assessment was associated with a 6% increase in the placebo remission rate (OR 1.06, 95% CI 1.02 to 1.10, P = 0.01).
Cumulative meta-analysis indicated a consistent increase in the placebo response rate from 1987 to 2007 (from 13% to 33%), although rates have remained constant from 2008 to 2015 (32% to 34%). Similarly, placebo remission rates increased from 1987 to 2007 (5% to 14%) but have remained constant from 2008 to 2015 (12 to 14%). On meta-regression, there were no statistically significant differences between the 1987-2007 and 2008-2015 point estimates for both response (P = 0.81) and remission (P = 0.32).
Authors' conclusions
Placebo response and remission rates vary according to endoscopic disease severity and rectal bleeding score at trial entry, class of agent, disease duration, and the time point at which the primary outcome was measured. These observations have important implications for the design and conduct of future clinical trials in UC and will help researchers design trials, determine required sample sizes and also provide useful information about trial design features which should be considered when planning new trials.