Systematic Review and Meta-Analysis: Clinical, Endoscopic, Histological and Safety Placebo Rates in Induction and Maintenance Trials of Ulcerative Colitis

Abstract

Background and Aims

Quantifying placebo rates and the factors influencing them are essential to inform trial design. We provide a contemporary summary of clinical, endoscopic, histological and safety placebo rates in induction and maintenance clinical trials of ulcerative colitis, and identify factors influencing them.

Methods

MEDLINE, EMBASE and the Cochrane library were searched from April 2014 to April 2020, updating a prior meta-analysis that searched from inception to April 2014. We included placebo-controlled trials of aminosalicylates, corticosteroids, immunosuppressives, small-molecules and biologics in adults with ulcerative colitis. Placebo rates were pooled using random-effects and mixed-effects meta-regression models to assess the associated study-level.

Results

In 119 trials [92 induction, 27 maintenance] clinical, endoscopic and histological remission placebo rates for induction trials were 11% (95% confidence interval [CI] 9–13%), 19% [95% CI 15–23%] and 15% [95% CI 11–19%], respectively; for maintenance trials, clinical and endoscopic placebo remission rates were 18% [95% CI 12–25%] and 20% [95% CI 15–25%], respectively. Higher endoscopic subscore and a higher rate of exposure to prior biologic therapy at enrolment were associated with lower clinical and endoscopic placebo remission rates. Absence of central reading was associated with an increase in placebo endoscopic response and remission rates. More follow-up visits and increasing trial duration were associated with higher clinical placebo rates.

Conclusions

Placebo rates in ulcerative colitis trials vary according to the endpoint assessed, whether it is for assessment of response or remission, and whether the trial is designed for induction or maintenance. These contemporary rates across different endpoints and drug classes will help to inform trial design.

1. Introduction

Treatment targets in inflammatory bowel disease [IBD] have evolved from achieving symptom control alone towards resolution of more objective measures such as endoscopic and/or histological inflammation.¹ Accordingly, several new therapies have been developed or are in late-stage development to achieve these goals.^2–4 Despite the routine availability of effective treatments in clinical practice as potential benchmarks, most pharmaceutical trials typically make use of a placebo to determine efficacy and to differentiate safety outcomes from the active intervention. Thus, reliable estimates of placebo rates of remission and response, as well as the factors that influence them, are essential for designing clinical trials in IBD. Factors that modulate the placebo response include concomitant medications, natural variation of the underlying disease, regression towards the mean, environmental and psychosocial factors, as well as factors related to clinical trial design and endpoint definitions.^5–8

The first meta-analysis quantifying placebo rates in randomized controlled trials [RCTs] in ulcerative colitis [UC] was conducted by Su et al. in 2007, which reported placebo clinical response and remission rates of 0–67% and 0–40%, respectively. However, most of the included trials were conducted before the era of biologic therapies.⁹ In 2017, Jairath et al. evaluated placebo response and remission rates in 51 induction and maintenance RCTs up to 2014, reporting pooled placebo clinical response and remission rates of 33% and 10% for induction trials, respectively, and for maintenance studies, 22% and 19%, respectively; studies enrolling patients with more severe clinical disease activity had lower placebo rates, whereas studies with more visits and longer follow-up had higher placebo rates.¹⁰ Importantly, the study by Jairath et al. included RCTs that were conducted before the routine incorporation of centralized reading of endoscopy, which has become a regulatory gold standard both for enrolment and assessment of the trial primary outcome, and is associated with lower endoscopic placebo rates.¹¹ Given this evolution in trial design and the publication of multiple new RCTs since 2014, we conducted an updated meta-analysis to estimate contemporary clinical, endoscopic, histological and safety outcomes in patients randomized to placebo in both induction and maintenance phases of UC trials. Secondly, a meta-regression was conducted to identify trial design features affecting placebo rates.

2. Methods

2.1. Search strategy

A search was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA]^12,13 recommendations searching in MEDLINE [Ovid, 1946], EMBASE [Ovid, 1947], the Cochrane Library [CENTRAL] and clinicaltrials.gov. The previous search by Jairath et al. was performed through to April 2014. We updated that search to April 2020. The search strategy is reported in Supplementary Table 1. Abstracts from conference proceedings (i.e. Digestive Disease Week and United European Gastroenterology Week [UEGW] 2012–2020), in addition to the bibliographies of relevant studies, review articles and meta-analyses, were hand-searched to identify additional, potentially eligible studies. During the conduct of this review, several clinical trials were presented at the UEGW 2020 conference, and thus we updated the hand-search of conference proceedings through to December 2020.

2.2. Study selection and eligibility criteria

The eligibility criteria were as reported previously by Jairath et al.¹⁰ RCTs were eligible for inclusion, provided they fulfilled the following criteria: [a] placebo-controlled trial in adult UC population investigating a biological agent, small molecule, corticosteroid, immunosuppressive or aminosalicylates; [b] use of the Disease Activity Index¹⁴ or the Mayo [or modified Mayo] Clinic Score^15,16 as enrolment criteria and for the assessment of clinical response and/or remission; [c] duration of more than 2 weeks for induction, and 4 months for maintenance of remission trials. Trials of probiotics, antibiotics, complementary therapies, devices or of hospitalized patients with severe UC were excluded.

Given the similarities between the Mayo Clinic score^15,16 and the UC Disease Activity Index [UCDAI]¹⁴ [both are 12-point scales incorporating four components of disease activity: stool frequency, rectal bleeding, mucosal appearance of the sigmoid colon on sigmoidoscopy, and physician’s global assessment], these indices were considered equivalent for analysis purposes.

2.3. Outcome assessment

Two authors [R.S. and J.C.] independently screened abstracts for eligibility. Full-text publications of all potentially eligible articles were reviewed by the same two authors to ensure that the inclusion criteria were met. Disagreements were resolved by a third author [J.K.M.] when consensus could not be reached. Quality control of the extracted data was performed by a fourth author [T.N.].

2.4. Data extraction and analysis

Data were extracted independently into a Microsoft Excel spreadsheet [XP professional edition; Microsoft]. The proportion of placebo patients with clinical response and remission, endoscopic response and remission, histological response and remission, adverse events [AEs] and serious adverse events [SAEs] were extracted on an intention-to-treat basis.

The following trial design characteristics were extracted: [i] trial design and participant characteristics [number of treatment arms, trial phase, publication year, study location, first author nationality, sample size, study duration, number and frequency of follow-up visits, number of participants analysed under treatment and placebo arms, post-randomization dropouts, age, gender ratio]; [ii] type of intervention [drug class, concomitant therapy, dosing, route and frequency of administration, ratio of active drug to placebo]; [iii] criteria for enrolment and outcome assessment [minimum Mayo/mMayo score for inclusion; Mayo/mMayo-based definitions of response and remission as prespecified in the primary trial], inclusion of endoscopy and minimum endoscopic score as enrolment and outcome criteria; [iv] disease severity and duration (baseline C-reactive protein [CRP], faecal calprotectin, disease distribution, disease duration, prior treatments exposure and rates of patients who have failed prior biological therapy); and [v] safety outcomes, specifically AEs, SAEs and withdrawals due to AEs.

2.5. Data synthesis and statistical methods

The statistical methodology was identical to that reported by Jairath et al.¹⁰ We estimated pooled placebo remission and response rates, for both induction and maintenance trials, using a random-effect model to account for between- and within-study variability. Mixed-effects meta-regression analyses with logits of event rates as outcome variables were used to assess association of each study-level characteristic with placebo rates, separately for induction and maintenance phase trials.¹⁷ We converted point estimates and 95% confidence intervals [CIs] back to the original scale. Covariates for inclusion in the meta-regression were selected based upon factors identified in the previous studies from Su et al.⁹ and Jairath et al.¹⁰ augmented with candidate clinical factors. Due to co-linearity between study-level features and limited number of studies, we did not perform multivariable meta-regression analysis.

To assess clinical heterogeneity, we constructed stratum-specific rates for several patient and trial-level covariates, for both placebo response and remission. Statistical heterogeneity within these strata was evaluated using I2 and Q-test statistics,^9,18 with a value below 50% representing lower heterogeneity.¹⁹ Potential publication bias was evaluated using funnel plots. A risk of bias analysis of included trials was performed using the Cochrane Collaboration’s tool.²⁰ Cumulative meta-analysis by date of publication was interpreted through a Bayesian approach whereby studies are sequentially used to update the prior distribution, such that the last posterior distribution becomes the next prior distribution.²¹ Analyses were conducted with the Metafor package²² for R version 3.1.1.²³

3. Results

3.1. Search results

The combined original search from Jairath et al.¹⁰ plus the updated search from April 2014 onwards resulted in a total of 14 189 citations, of which 6065 were removed as duplicates [Supplementary Figure 1]. For the remaining 8124 records screened, 342 full-text articles were selected and assessed for eligibility, of which 119 trials were eligible for data extraction [92 induction and 27 maintenance phase studies]. For the 92 induction trials [n = 6084 participants randomized to placebo; range 5–331], we were able to evaluate clinical response rates in 79 trials and clinical remission rates in 84 trials. For the 27 maintenance phase trials [n = 2560 participants randomized to placebo, range 31–260], we were able to evaluate clinical response rates in 15 trials and clinical remission rates in 21 trials.

3.2. Description of included studies

Study baseline characteristics are summarized in Table 1 and Supplementary Table 2, with risk of bias assessment in Supplementary Table 3. For induction trials, the mean placebo sample size was 66.1 [range 5–331], the number of centres ranged from one to 372 and length of follow-up ranged from 2 to 60 weeks. Most induction trials were designed as stand-alone studies [68/92; 74%], were phase 2 [44/92; 47.8%], multicentre/multinational [63/92; 68.5%] and investigated a biological agent [48/92; 52.7%]. For maintenance trials, the mean placebo sample size was 102.4 [range 31–260], the number of centres ranged from one to 372, and length of follow-up ranged from 12 to 96 weeks. Most maintenance trials investigated biological therapies [17/27; 65.3%]. From the 27 maintenance studies, only 16 were part of an integrated induction and maintenance trial; of these trials, 4/16 re-randomized their patients in a 1:1 ratio to active treatment vs placebo after obtaining clinical response or remission during induction, and 12/16 were straight-through trials.

3.3. Placebo response and remission rates for induction and maintenance trials

3.3.1. Clinical response and remission rates

In total, 79 induction trials evaluated clinical response and 84 evaluated clinical remission; 15 maintenance trials evaluated clinical response and 21 evaluated clinical remission. The pooled placebo clinical response rate for induction trials was 32% [95% CI 30–35%; range 6–92%] [Figure 1] and the pooled placebo clinical remission rate was 11% [95% CI 9–13%; range 1–49%] [Figure 2]. For maintenance trials, the pooled placebo clinical response rate was 26% [95% CI 22–31%; range 16–45%] [Figure 3] and the pooled placebo clinical remission rate was 18% [95% CI 12–25%; range 5–68%] [Figure 4]. Significant heterogeneity was also demonstrated for both induction trials [p < 0.001] and maintenance trials [p < 0.001].

3.3.2. Endoscopic response and remission rates

In total, 24 induction trials evaluated endoscopic response and 65 induction trials evaluated endoscopic remission; for maintenance trials, three evaluated endoscopic response and 17 endoscopic remission. For induction trials, the pooled placebo endoscopic response rate was 26% [95% CI 20–34%; range 1–63%] [Figure 5] and the pooled placebo endoscopic remission rate was 19% [95% CI 15–23%; range 1–68%] [Figure 6]. Statistically significant heterogeneity was observed among trials [p < 0.001]. For maintenance trials, the pooled placebo endoscopic response rate was 18% [95% CI 12–26%; range 13–23%] [Figure 7] and the pooled placebo endoscopic remission rate was 20% [95% CI 15–25%; range 4–37%] [Figure 8]. As some trials defined endoscopic response the same way others defined endoscopic remission, when endoscopic assessment was stratified by use of Mayo endoscopic score [MES] = 0 or MES ≤ 1, the pooled placebo rates were 9% [95% CI 0.05–0.16] and 23% [95% CI 0.20–0.27], respectively [Figure 9]. Statistically significant heterogeneity was observed among trials for both endoscopic remission [p < 0.001] and endoscopic response [p = 0.008].

3.3.3. Histological response and remission rates

In total, three induction trials reported histological response and 18 induction trials reported histological remission; for maintenance trials, one evaluated histological response and five evaluated histological remission. For induction trials, the pooled placebo histological response rate was 14% [95% CI 5–33%; range 7–30%] [Supplementary Figure 2] and the pooled placebo histological remission rate was 15% [95% CI 11–19%; range 4–41%] [Supplementary Figure 3]. Statistically significant heterogeneity was observed for both histological response [p = 0.007] and histological remission [p < 0.001]. For maintenance trials, the pooled placebo histological remission rate was 15% [95% CI 7–29%; range 2–33%] [Supplementary Figure 4]. Statistically significant heterogeneity was observed for histological remission [p < 0.008]. There was only one maintenance trial reporting placebo histological response, and thus pooled data are not available.

3.4. Factors affecting placebo response and remission rate for induction trials

3.4.1. Disease and patient related characteristics

Clinical response/remission

Disease duration >5 years before enrolment was associated with a significantly lower placebo clinical response rate compared with disease duration ≤5 years (32% vs 46%, respectively; odds ratio [OR] 0.56, 95% CI 0.4–0.78, p = 0.001] [Tables 2 and 4]. The same pattern was also observed for the placebo clinical remission rate [18% vs 10%, respectively: OR 0.59, 95% CI 0.36–0.99, p = 0.045] [Tables 2 and 4]. Trials that required an endoscopy subscore ≥1 at enrolment were associated with a higher placebo response rate compared to trials requiring a minimum endoscopy subscore ≥2 [46% vs 33%; OR 1.76, 95% CI 1.08–2.88, p = 0.024] [Tables 2 and 4]. The same pattern was also observed for placebo remission rate [25% vs 11% for endoscopy subscore ≥1 vs ≥2 respectively: OR 2.74, 95% CI 1.34–5.6, p = 0.006] [Tables 2 and 4]. No differences were observed for the pooled placebo response and remission rates according to the requirement for a minimum rectal bleeding subscore for study entry [required vs not required]. Trials with higher rates of prior exposure to biological therapies were associated with a lower placebo clinical response rate (OR 0.93, 95% CI 0.9–0.97, p < 0.001 [per 10% increase in biologics] and OR 0.78, 95% CI 0.62–0.99 p = 0.045 [≥50% vs <50% patients bio-exposed]). The same was true for the pooled clinical remission rate (OR 0.88, 95% CI 0.78–0.99, p = 0.028 [per 10% increase in biologics] and OR 0.46, 95% CI 0.24–0.86, p = 0.015 [≥50% vs <50% patients bio-exposed]).

Endoscopic response/remission

Studies using an endoscopy subscore ≥1 for study entry were associated with a higher pooled placebo endoscopic remission rate compared with studies using a more stringent criterion of endoscopy subscore ≥2 [56% vs 20%; OR 5.24, 95% CI 1.64–16.74, p = 0.005] [Supplementary Tables 4 and 6]. Trials with higher rates of prior exposure to biological therapies were associated with a lower endoscopic placebo response rate (OR 0.72, 95% CI 0.6–0.87, p < 0.001 [per 10% increase in biologics] and OR 0.18, 95% CI 0.06–0.55, p = 0.002 [≥50% vs <50% patients bio-exposed]). The same was true for the pooled endoscopic remission rate (OR 0.88, 95% CI 0.76–1.01, p = 0.074 [per 10% increase in biologics] and OR 0.44, 95% CI 0.2–0.94, p = 0.034 [≥50% vs <50% patients bio-exposed]). No other disease- or patient-related characteristics had significant impact on the pooled placebo endoscopic response and remission rates aside from minimum endoscopic severity entry score.

Histological remission

A minimum composite UCDAI score <6 was associated with a significantly lower placebo histological remission rate compared with a minimum score ≥6 [7% vs 18%, respectively; OR 0.37, 95% CI 0.16–0.83, p = 0.016] [Supplementary Tables 5 and 7]. No other disease- or patient-related characteristics were associated with histological remission rates.

3.4.2. Trial design features

Clinical response/remission

For response, neither the duration of follow-up, number of follow-up visits, setting, continent of origin, trial design, primary timepoint to measure endpoint, number of centres, use of central reading for enrolment nor year of publication [before 2005 vs after 2005] had any significant impact on the pooled placebo clinical response rate. For remission, more follow-up visits were associated with a higher pooled placebo rate [OR 1.12, 95% CI 1.01–1.23, p = 0.029; for every one-visit increment] [Table 4]. The primary endpoint measurement at a later time point was associated with a higher placebo remission rate [OR 1.05, 95% CI 1.01–1.1, p = 0.014; per 1-week increment] [Table 4]. The number of trial centres was associated with a lower pooled placebo clinical remission rate [OR 0.93, 95% CI 0.9–0.99, p = 0.041; per 20-centres increment] [Table 4].

Endoscopic response/remission

The number of trial centres was associated with a lower pooled placebo endoscopic response rate [OR 0.8, 95% CI 0.66–0.98, p = 0.028; per 20-centres increment]. A later publication year was also associated with a higher pooled placebo endoscopic response rate [OR 0.92, 95% CI 0.88–0.97, p = 0.002; per 1-year increment] [Supplementary Table 6]. The absence of central reading was associated with a statistically significant increase in the pooled placebo endoscopic response [OR 3.13, 95% CI 1.59–6.17, p = 0.001] and remission rate [OR 2.61, 95% CI 1.62–4.2, p < 0.001] [Supplementary Table 6]. Defining endoscopic remission as MES ≤ 1 was associated with higher rates of endoscopic remission [OR 4.4, 95% CI 2.58–7.49, p < 0.001] compared to when it was defined as MES = 0.

Histological remission

A greater number of follow-up visits was associated with a higher pooled placebo histological remission rate [OR 1.36, 95% CI 1.15–1.61, p < 0.001; per one-visit increment] [Supplementary Table 7]. The remaining class of drug characteristics had no significant impact on the pooled placebo histological remission rates.

3.4.3. Therapeutic class

Clinical response/remission

For response, pooled placebo rates ranged from 19% to 26% [Table 2]. The highest placebo response rates were for biologicals [34%; 95% CI 31–36%; p < 0.001], and the lowest for immunosuppressives [19%; 95% CI 10–31%; p = 0.113]. Studies using an oral route of administration were associated with a lower placebo response rate compared with studies using a topical route of administration [29% vs 37%; OR 0.66, 95% CI 0.43–1, p = 0.047] [Tables 2 and 4]. Studies using a subcutaneous route of drug administration were associated with a lower placebo remission rate compared with studies using a topical route of administration [8% vs 18%; OR 0.38, 95% CI 0.19–0.77, p = 0.007] [Tables 2 and 4]. For remission, pooled placebo rates according to drug class ranged from 10% to 17% [Table 2]. The highest placebo response rates were seen for aminosalicylates [17%; 95% CI 12–24%; p = 0.006], and the lowest for biologicals [10%; 95% CI 8–13%; p < 0.001].

Endoscopic response/remission

The absence of concurrent use of immunosuppressives was associated with a higher pooled placebo endoscopic remission rate [OR 0.57, 95% CI 0.34–0.96; p = 0.036] [Supplementary Table 6]. The remaining drug classes had no significant impact on the pooled placebo endoscopic response and remission rates.

Histological remission

None of the drug classes had any significant impact on the pooled placebo histological remission rates.

3.5. Factors affecting placebo clinical response and remission rate for maintenance trials

Given the small number of evaluable maintenance trials for endoscopic and histological response and remission, meta-regression to identify factors modulating the endoscopic and histological placebo response and remission was only feasible for induction trials.

3.5.1. Disease- and patient-related characteristics

Studies defining disease severity as moderate–severe were associated with a significantly lower pooled placebo clinical remission rate compared with studies defining disease severity as mild–moderate [14% vs 42%, respectively: OR 0.19, 95% CI 0.09–0.42, p < 0.001] [Tables 3 and 5]. A minimum UCDAI score <6 was associated with a significantly higher placebo remission rate compared to ≥6 [36% vs 15%, respectively: OR 3.94, 95% CI 1.52–10.26, p = 0.005] [Tables 3 and 5]. Studies only requiring an endoscopy subscore ≥1 at enrolment were associated with a higher pooled placebo remission rate as compared with studies requiring a subscore ≥2 [51% vs 15%; OR 6.33, 95% CI 2.54–15.78, p < 0.001] [Tables 3 and 5]. Studies that did not require a minimum rectal bleeding subscore for study entry were associated with a lower pooled placebo remission rate than those that did [15% vs 35%: OR 0.31, 95% CI 0.12–0.8, p = 0.016]. There were no patient- or disease-related characteristics that had a significant impact on the pooled placebo response rate.

3.5.2. Trial design features

For response, more follow-up visits [>7 vs ≤7] were associated with a higher pooled placebo rate [30% vs 21%; OR 1.63, 95% CI 1.02–2.59, p = 0.042] [Tables 3 and 5]. A significant association was also found when using number of follow-up visits as a continuous outcome measure [OR 1.07, 95% CI 1–1.13, p = 0.036; per one-visit increment] [Table 5]. A later publication year was also associated with a higher pooled placebo response rate [OR 1.05, 95% CI 1–1.1, p = 0.047; per 1-year increment] [Table 5]. For remission, more follow-up visits were associated with a lower pooled placebo rate [OR 0.96, 95% CI 0.94–0.99, p = 0.006; per 1-week increment] [Table 5]. In contrast to induction trials, the timing of primary endpoint measurement was associated with a lower placebo remission rate [OR 0.97, 95% CI 0.94–0.99, p = 0.017; per 1-week increment] [Table 5]. No significant impact on placebo response or remission was seen for trial setting, continent of origin, trial design, number of trial centres or the presence of central readers to confirm active disease.

3.5.3. Therapeutic class

For response, there were only two drug classes in which the pooled placebo rate was evaluable: biologicals and other. The pooled placebo rate was 26% [CI 21–31%; p < 0.001] [Table 3] for biologicals and 28% [CI 10–57%; p < 0.001] for other. For remission, pooled placebo rates according to drug class ranged from 13% to 41% [Table 3]. The highest placebo response rates were observed for aminosalicylates [41%; 95% CI 10–81%; p < 0.001], and the lowest for biologicals [13%; 95% CI 10–17%; p < 0.001].

3.6. Safety outcomes

For induction trials, the pooled proportion of placebo patients who experienced an AE/SAE or withdrew due to an adverse event was 49% [95% CI 44–55%; range 4–95%], 6% [95% CI 5–8%; range 1–50%] and 7% [95% CI 5–9%; range 1–47%], respectively [Supplementary Figures 5–7]. Significant heterogeneity was demonstrated amongst these trials [p < 0.001]. For maintenance trials, the pooled proportion of patients of who experienced an AE, SAE or withdrew due to an adverse event was 69% [95% CI 61–76%; range 15–90%], 9% [95% CI 7–12%; range 1–26%] and 8% [95% CI 6–11%; range 1–19%], respectively [Supplementary Figures 8–10]. Significant heterogeneity was demonstrated amongst these trials [p < 0.001].

3.7. Observed time trends for placebo rates

3.7.1. Induction trials

Clinical response/remission

The cumulative meta-analysis showed an increase in the placebo rate from 1987 to 2007 [from 13% to 33%] for response, which remained constant [32% to 34%] through 2020 [Supplementary Figure 11]. For remission, a consistent increase in the placebo rate was seen from 1987 to 2007 [from 5% to 14%] and was followed by a decrease from 2007 to 2020 [14% to 11%] [Supplementary Figure 12].

Endoscopic response/remission

For response, cumulative meta-analysis demonstrated a consistent decrease in placebo rate from 2003 to 2020 [from 46% to 26%] [Supplementary Figure 13]. For remission, a consistent increase in placebo rate from 1996 to 2016 [from 11% to 27%] and then a decrease from 2016 to 2020 [27% to 19%] was observed [Supplementary Figure 14].

Histological remission

Cumulative meta-analysis demonstrated a consistent increase in the placebo remission rate from 1996 to 2014 [from 5% to 12%] and a steadier increase from 2014 to 2020 [from 12% to 15%] [Supplementary Figure 15].

3.7.2. Maintenance trials

For response, there was a consistent increase in the placebo response rate from 2005 to 2020 [from 20% to 26%] [Supplementary Figure 16]. For remission, a decrease in the placebo remission rate from 1996 to 2017 [from 27% to 17%] was observed, followed by constancy through 2020 [17% to 20%] [Supplementary Figure 17].

3.8. Publication bias from funnel plots

3.8.1. Clinical response/ remission

The risk of publication bias was not statistically significant [z = −1.8102, p = 0.0703] for induction trials reporting response [Supplementary Figure 18], and significant [z = −4.79, p < 0.001] for those trials reporting remission [Supplementary Figure 19]. For maintenance trials, the risk of publication bias was significant [z = −2, p = 0.046] for trials reporting response [Supplementary Figure 20] and also significant [z = −2.36, p = 0.0182] for trials reporting remission [Supplementary Figure 21].

3.8.2. Endoscopic response/remission

The risk of publication bias was significant [z = −3.64, p < 0.001] for induction trials reporting response [Supplementary Figure 22] and significant [z = −6.54, p < 0.001] for those trials reporting remission [Supplementary Figure 23].

3.8.3. Histological remission

The risk of publication bias was insignificant [z = −1.7, p = 0.09] for induction trials reporting remission [Supplementary Figure 24].

4. Discussion

In this contemporary meta-analysis including 119 studies of pharmacological therapies for UC, we found that placebo rates varied according to whether trials were designed as induction or maintenance trials, the type of endpoint evaluated, and whether the assessment was for response or remission. For induction trials, pooled placebo rates for remission were lowest for clinical remission [11%], followed by histological [15%] and then endoscopic remission [19%], whereas for maintenance trials placebo rates were lowest for histological [15%] followed by clinical [18%] and then endoscopic remission [20%].

A key observation was that trials requiring enrolment of patients with a higher endoscopic subscore [i.e. more severe inflammation] were associated with statistically significantly lower placebo rates of clinical response and remission, as well as lower placebo endoscopic rates of remission. This underlines the importance of ensuring that patients enrolled into trials have sufficient objective disease activity in order to control excessively high placebo rates and enable the conduct of more efficient trials.²⁴ For UC trials today, this is typically achieved through centrally read endoscopy, or through a combination of a well-trained local endoscopist and central reader, with an adjudication process in the case of disagreement. This is also an important consideration for designing trials of mild-to-moderate UC, where eligibility criteria may require a lower endoscopic entry score and potentially result in higher placebo rates. The cumulative meta-analysis for endoscopic response and remission demonstrates a continual decline in placebo rates since 2014 [pooled endoscopic placebo remission rate 25% before 2014; 19% after 2014], when central reading evolved as a regulatory standard. Furthermore, studies that did not use central reading to confirm endoscopic disease activity were associated with higher rates of placebo endoscopic response [OR 3.13, 95% CI 1.59–6.17, p = 0.001] and remission [OR 2.61, 95% CI 1.62–4.2, p < 0.001], lending support to the concept that central reading has been associated with the observed reductions in placebo rates in UC trials. Notably, no differences in placebo clinical response and remission rates were observed according to rectal bleeding subscore, implying that endoscopy should be the dominant criterion in defining disease activity at enrolment.

Disease and trial design characteristics also influenced placebo rates. Disease duration >5 years—potentially representing a harder to treat population—before enrolment was associated with a significantly lower placebo clinical response and remission rate compared with disease duration ≤5 years. However, no such association was observed with the more objective endpoints of endoscopy or histology. This discordance could be influenced by additional disease processes in patients with UC and longstanding disease, aside from active inflammation, such as impaired colonic motility and rectal incapacity.²⁵ In induction trials, a higher rate of patients with prior exposure to biological therapies was associated with significantly lower placebo clinical and endoscopic response and remission rates, probably representing a more refractory treatment population that is less likely to achieve spontaneous improvement.

For trial design factors, it has been well documented that placebo effects are influenced by the positive interaction between patient and healthcare providers.^5,7 We found that a greater number of follow-up visits was associated with a higher pooled placebo clinical remission rate, with a 12% increase in the odds of remission with each extra trial visit for induction trials and 7% for maintenance trials. Similarly, each 1-week increase in the timing of primary endpoint measurement was associated with a 5% increase in the odds of placebo remission for induction trials. Thus, critical attention should be paid to schedules of events when designing studies to minimize unnecessary visits, which ironically receives greater attention today due to the move towards decentralized trials accelerated by the COVID-19 pandemic.

It was notable that some discordances were observed between response and remission results. First, pooled placebo rates for clinical response were highest for biologicals compared to other drug classes, whereas pooled placebo rates for clinical remission were lowest for biologicals. Second, an increasing number of trial visits [in maintenance studies] was associated with a higher pooled placebo response rate, but a lower pooled placebo remission rate. These observations are probably related to the concept of response-expectancy, whereby taking part in a biological trial and more frequent interactions with caregivers may generate a response expectation, which is less likely to be observed with the more stringent endpoint of remission. Third, for maintenance studies, there was a numerically higher pooled placebo endoscopic remission (20% [95% CI 15–25%]) than response rate (18% [95% CI 12–26%]). This could be explained in part by the lack of standardized definitions for endoscopic remission and response. For instance, from the 18 maintenance studies that defined endoscopic remission, 4/18 [22.2%] used a definition of Mayo score = 0, and 14/18 [77.8%] used a Mayo score ≤1; which was the same definition used to define endoscopic response in 3/3 maintenance studies that assessed this outcome. Furthermore, there were only three maintenance trials assessing endoscopic response, and thus the estimates may be imprecise. Nevertheless, when the pooled placebo rate was stratified by endoscopic assessment definition used [MES = 0 or MES ≤ 1], the pooled placebo rate for MES = 0 was 9% [95% CI 0.05–25%] and for MES ≤ 1 was 23% [95% CI 0.20–0.27], consistent with the concept that a less stringent endoscopic definition would be associated with higher rates of endoscopic response/remission.

Histology has emerged as an increasingly important endpoint in drug development for the treatment of UC.²⁶ Whilst currently configured as a secondary endpoint, it is noteworthy that the first agent [ustekinumab] has received a label for the combined endpoint of histo-endoscopic mucosal healing.²⁷ It is relevant that for both induction and maintenance trials, placebo rates for histological remission were lower than for endoscopic remission, implying that the former could potentially be a more efficient endpoint. Furthermore, there are fully validated indices for use of histology in clinical trials.²⁸ Concurrent clinical development programmes will provide more data to inform the operating properties and configuration of histology as an endpoint, relative to endoscopic and clinical remission, and further inform sample size estimates.

We observed highly statistically significant heterogeneity in induction trials for clinical and endoscopic response and remission [p < 0.001, for all outcomes], and histological response [p = 0.0074] and remission [p < 0.001]; there were similar findings in maintenance trials for clinical response and remission [p < 0.001], endoscopic response [p = 0.0083] and remission [p < 0.001], and histological remission [p < 0.0083]. Despite stratification across multiple covariates, this heterogeneity remained, but was less evident in maintenance studies with <40 weeks of follow-up, trials published before 2005 and those with fewer than seven follow-up visits. Despite homogenous trial-level inclusion criteria, there are multiple known and unknown confounders which contribute to heterogeneity, and access to patient-level data would be needed to further explore the influence of individual patient characteristics on heterogeneity.

There are several key differences between the current analysis and the prior study from Jairath et al. in 2017.¹⁰ First, this updated meta-analysis provided an additional 61 studies between April 2014 and December 2020, compared to a total of 51 studies in total prior to 2014, reflecting the massive growth in drug development and completion of programmes in the past 7 years. Second, we were able to include endoscopic and histological outcomes for induction and maintenance trials, where the prior study only reported clinical response and remission outcomes. Third, we were able to assess placebo rates in prior biologic-exposed vs non-exposed patients, which is highly relevant to trials being designed and conducted today. Fourth, we also included safety outcomes. Aside from size and large number of trials included across multiple classes, the strengths of this study include provision for the first time, contemporary pooled estimates of placebo rates for clinical, endoscopic and histological remission for induction and maintenance trials, as well as detailed meta-regression to estimate the influence of covariates on these rates. Limitations include the high level of heterogeneity and inability to access patient-level data. Furthermore, the effect of study design on maintenance trials, according to whether patients were re-randomized after achieving clinical response/remission in the induction phase, was not feasible due to the small sample size.

In conclusion, placebo rates vary according to the endpoint assessed, whether it is for assessment of response or remission, and whether the trial is designed for induction or maintenance. Trials enrolling patients with a higher endoscopic subscore and a greater proportion of patients with prior exposure to biological therapies at enrolment were associated with significantly lower placebo rates of clinical response and remission, as well as lower placebo endoscopic rates of remission. Use of central reading to confirm disease activity was associated with a significant decrease in pooled placebo endoscopic response and remission rates. More follow-up visits and increase length in the timing of primary endpoint measurement were associated with higher pooled placebo rates. These contemporary rates across different endpoints and drug classes will help to inform design of clinical trials in UC, sample size calculation and selection of efficient endpoints.

Acknowledgments

None.

Funding

None.

Conflict of Interests

R.S. received consulting fees from Alimentiv Inc. M.H. is an employee of Alimentiv Inc. T.N. is an employee of Alimentiv Inc. J.C. is an employee of Alimentiv Inc. G.Y.Z. is a consultant to Alimentiv Inc J.K.M. is an employee of Alimentiv Inc. N.V.C. received research grants and personal fees from R-Biopharm, Takeda and UCB; and personal fees from Alimentiv, Inc., Celltrion and Prometheus. J.H. received speaker’s fees from Abbvie, Janssen, and Takeda; consulting fees from Alimentiv Inc. E.C. received an educational grant and speaker’s fees from Abbvie and consulting fees from Alimentiv Inc. R.B. received research support from the Fund for the Future award and Jill Roberts Funds at the Department of Medicine, Weill Cornell Medicine. P.S.D. received a Research Scholar Award from the American Gastroenterological Association, research support from Pfizer, Takeda, Janssen, Abbvie; and consulting fees from Takeda, Janssen, Abbvie. S.S. has received research grants from Janssen and AbbVie, and personal fees from Pfizer [for ad hoc grant review]. G.R.D’H. has received research grants from Abbvie, Alimentiv, Alphabiomics, Eli Lilly, Johnson and Johnson, Pfizer, Progenity, Procise Diagnostics, Prometheus, Roche, Takeda; has served as advisor for Abbvie, Ablynx, Active Biotech AB, Agomab Therapeutics, Alimentiv, Allergan, Alphabiomics, Amakem, Amgen, AM Pharma, Applied Molecular Therapeutics; Arena Pharmaceuticals, AstraZeneca, Avaxia, Biogen, Bristol Meiers Squibb/Celgene, Boehringer Ingelheim, Celltrion, Cosmo, DSM Pharma; Echo Pharmaceuticals, Eli Lilly, Engene, Exeliom Biosciences, Ferring, DrFALK Pharma, Galapagos, Genentech/Roche, Gilead, Glaxo Smith Kline, Gossamerbio, Pfizer, Immunic, Johnson and Johnson, Kintai Therapeutics, Lycera, Medimetrics, Medtronic, Mitsubishi Pharma, Merck Sharp Dome, Mundipharma, Nextbiotics, Novonordisk, Otsuka, Photopill, ProciseDx, Prodigest, Prometheus laboratories/Nestle, Progenity, Protagonist, RedHill, Salix, Samsung Bioepis, Sandoz, Seres/Nestec/Nestle, Setpoint, Takeda, Teva, Tigenix, Tillotts, Topivert, Versant and Vifor; received speaker fees from Abbvie, Biogen, Ferring, Galapagos/Gilead, Johnson and Johnson, Merck Sharp Dome, Mundipharma, Norgine, Pfizer, Samsung Bioepis, Shire, Millenium/Takeda, Tillotts and Vifor. W.J.S. has received research grants from Abbvie, Abivax, Arena Pharmaceuticals, Boehringer Ingelheim, Celgene, Genentech, Gilead Sciences, Glaxo Smith Kline, Janssen, Lilly, Pfizer, Prometheus Biosciences, Seres Therapeutics, Shire, Takeda, Theravance Biopharma; consulting fees from Abbvie, Abivax, Admirx, Alfasigma, Alimentiv [previously Robarts Clinical Trials, owned by Alimentiv Health Trust], Alivio Therapeutics, Allakos, Amgen, Applied Molecular Transport, Arena Pharmaceuticals, Bausch Health [Salix], Beigene, Bellatrix Pharmaceuticals, Boehringer Ingelheim, Boston Pharmaceuticals, Bristol Meyers Squibb, Celgene, Celltrion, Cellularity, Cosmo Pharmaceuticals, Escalier Biosciences, Equillium, Forbion, Genentech/Roche, Gilead Sciences, Glenmark Pharmaceuticals, Gossamer Bio, Immunic [Vital Therapies], Index Pharmaceuticals, Intact Therapeutics, Janssen, Kyverna Therapeutics, Landos Biopharma, Lilly, Oppilan Pharma, Otsuka, Pandion Therapeutics, Pfizer, Progenity, Prometheus Biosciences, Protagonists Therapeutics, Provention Bio, Reistone Biopharma, Seres Therapeutics, Shanghai Pharma Biotherapeutics, Shire, Shoreline Biosciences, Sublimity Therapeutics, Surrozen, Takeda, Theravance Biopharma, Thetis Pharmaceuticals, Tillotts Pharma, UCB, Vendata Biosciences, Ventyx Biosciences, Vimalan Biosciences, Vivelix Pharmaceuticals, Vivreon Biosciences, Zealand Pharma; and stock or stock options from Allakos, BeiGene, Gossamer Bio, Oppilan Pharma, Prometheus Biosciences, Progenity, Shoreline Biosciences, Ventyx Biosciences, Vimalan Biosciences, Vivreon Biosciences. Spouse: Iveric Bio - consultant, stock options; Progenity - stock; Oppilan Pharma - consultant, stock options; Prometheus Biosciences - employee, stock, stock options; Ventyx Biosciences – stock, stock options; Vimalan Biosciences - stock, stock options. B.G.F. has received grant/research support from AbbVie Inc., Amgen Inc., AstraZeneca/MedImmune Ltd., Atlantic Pharmaceuticals Ltd., Boehringer-Ingelheim, Celgene Corporation, Celltech, Genentech Inc/Hoffmann-La Roche Ltd., Gilead Sciences Inc., GlaxoSmithKline [GSK], Janssen Research & Development LLC., Pfizer Inc., Receptos Inc./Celgene International, Sanofi, Santarus Inc., Takeda Development Center Americas Inc., Tillotts Pharma AG, and UCB; consulting fees from Abbott/AbbVie, Akebia Therapeutics, Allergan, Amgen, Applied Molecular Transport Inc., Aptevo Therapeutics, Astra Zeneca, Atlantic Pharma, Avir Pharma, Biogen Idec, BioMx Israel, Boehringer-Ingelheim, Bristol-Myers Squibb, Calypso Biotech, Celgene, Elan/Biogen, EnGene, Ferring Pharma, Roche/Genentech, Galapagos, GiCare Pharma, Gilead, Gossamer Pharma, GSK, Inception IBD Inc, JnJ/Janssen, Kyowa Kakko Kirin Co Ltd., Lexicon, Lilly, Lycera BioTech, Merck, Mesoblast Pharma, Millennium, Nestle, Nextbiotix, Novonordisk, Pfizer, Prometheus Therapeutics and Diagnostics, Progenity, Protagonist, Receptos, Salix Pharma, Shire, Sienna Biologics, Sigmoid Pharma, Sterna Biologicals, Synergy Pharma Inc., Takeda, Teva Pharma, TiGenix, Tillotts, UCB Pharma, Vertex Pharma, Vivelix Pharma, VHsquared Ltd., and Zyngenia; speakers bureau fees from Abbott/AbbVie, JnJ/Janssen, Lilly, Takeda, Tillotts, and UCB Pharma; is a scientific advisory board member for Abbott/AbbVie, Allergan, Amgen, Astra Zeneca, Atlantic Pharma, Avaxia Biologics Inc., Boehringer-Ingelheim, Bristol-Myers Squibb, Celgene, Centocor Inc., Elan/Biogen, Galapagos, Genentech/Roche, JnJ/Janssen, Merck, Nestle, Novartis, Novonordisk, Pfizer, Prometheus Laboratories, Protagonist, Salix Pharma, Sterna Biologicals, Takeda, Teva, TiGenix, Tillotts Pharma AG, and UCB Pharma; and is the Senior Scientific Officer of Alimentiv Inc. C.M. has received consulting fees from AbbVie, Amgen, AVIR Pharma Inc, Bristol Myers Squibb, Ferring, Fresenius Kabi, Janssen, McKesson, Mylan, Takeda, Pfizer, Roche, Alimentiv [formerly Robarts Clinical Trials Inc.]; speaker’s fees from AbbVie, AVIR Pharma Inc, Janssen, Takeda, and Pfizer; research support from Pfizer. V.J. has received consulting fees from AbbVie, Alimentiv Inc [formerly Robarts Clinical Trials], Arena pharmaceuticals, Bristol Myers Squibb, Celltrion, Eli Lilly, Ferring, Fresenius Kabi, GlaxoSmithKline, Genetech, Gilead, Janssen, Merck, Mylan, Pendopharm, Pfizer, Roche, Sandoz, Takeda, Topivert; speaker’s fees from, Abbvie, Ferring, Janssen Pfizer Shire, Takeda, Alimentiv, Inc. [formerly Robarts Clinical Trials, Inc.] is an academic gastrointestinal contract research organization [CRO], operating under the Alimentiv Health Trust. Alimentiv, Inc. provides centralized imaging management solutions in clinical trials, including endoscopy, histopathology and magnetic resonance imaging. Alimentiv, Inc. provides full service CRO capabilities as well as precision medicine services. Authors M.H./T.N./J.C./J.K.M. are employees. Authors G.Y.Z./N.V.C./J.H./E.C./G.R.D’H./B.G.F./W.J.S./C.M./V.J. are consultants and have neither equity positions nor shares in the corporation.

Author’s Contributions

Guarantors of the article: R.S., V.J. Development of study concept and design: R.S., C.M., V.J., G.D.H., W.J.S., B.G.F. Study supervision: R.S., C.M., V.J., W.J.S., B.G.F., T.N., J.K.M. Acquisition, analysis, and interpretation of the data: R.S., J.C., T.N., M.H., V.J., C.M., G.D.H., W.J.S., B.G.F. Statistical analysis: M.H., G.Y.Z. Drafting of the manuscript: R.S., V.J. Critical revision of the manuscript for important intellectual content: R.S., M.H., T.N., J.C., G.Y.Z., J.K.M., N.V.C., J.H., E.C., R.B., P.S.D., S.S., G.D.H., W.J.S., B.G.F., C.M., V.J. All authors approved the final version of the manuscript including the authorship list. This manuscript, including related data, figures and tables has not been previously published and is not under consideration elsewhere.

Data Availability Statement

The data underlying this article are available in the article and in its online supplementary material.

References