The efficacy of biologic drugs in short-duration versus long-duration inflammatory bowel disease

Search strategy and study selectionEligible studies will be randomized placebo controlled IBD trials of biologics which are approved by the FDA for IBD treatment (anti-TNFs and/or anti-integrins), or RCTs comparing biologic-immunomodulator combination treatment versus immunomodulator alone (with placebo biologic). To be included, trials outcomes should include clinical efficacy measures as one of the end-points assessed and reported (i.e. purely pharmacokinetic or safety trials not assessing clinical efficacy will be excluded). CD trials for indications other than active luminal CD will be excluded. Discontinuation and post-surgical prophylaxis trials will be excluded. Duration of biologics therapy will have to be at least two doses of induction spanning = 2 weeks. Therefore, trials using a single infusion or injection of a biologic will be ineligible. A trial employing a single infusion/injection for induction, whether randomized or not, followed by randomization for maintenance phase will be eligible for analysis restricted to efficacy outcomes of the maintenance treatment phase. Studies investigating dosing which are not approved for clinical practice or exploring non-approved indications will be also excluded. Studies with high-risk of bias (see below) will also be excluded. As all studies included in the meta-analysis will be randomized clinical trials, inclusion and exclusion of individual patients will be based on the original inclusion/exclusion criteria of the respective trials. In all trials intention-to-treat population outcomes will be analyzed. Studies will be identified using existing Cochrane systematic reviews [Bickston CJ, Cochrane Database Syst Rev 2014, CD007571, Macdonald JK, Cochrane Database Syst Rev 2007, CD006097, Behm BW, Cochrane Database Syst Rev 2008 CD006893, Lawson MM Cochrane Database Syst Rev 2006 CD005112 ], AGA technical review [Dassopoulos T, Gastroenterology 2013] and ECCO guidelines [Dignass A, JCC 2012, Dignass A, JCC 2010]. These will be supplemented by a search of the medical literature conducted with MEDLINE (1976 to November 2015). Eligible studies will be identified using the term string: Crohn's disease or ulcerative colitis or inflammatory bowel disease or IBD combined with the set operator AND with studies identified by the term string: trial or randomized or placebo, and conjointly combined with the set operator AND with the term string: infliximab or CA2 or adalimumab or D2E7 or certolizumab or CDP870 or golimumab or CNTO148 or natalizumab or vedolizumab or MLN02 or MLN0002 or LDP-02 or anti-TNFa or anti tumor necrosis factor OR TNFa inhibitor or tumor necrosis factor inhibitor or antibody to tumor necrosis factor or tumor necrosis factor alpha or antibody to alpha4 or anti-alpha4 or antibody to integrin or anti-VLA4 or anti-VLA-4 or anti-integrin. This search will be further supplemented by search of the Cochrane CENTRAL register of controlled trials and the Cochrane IBD Group Specialized Trials Register using the terms 'Crohn's disease' and 'Ulcerative colitis'. In addition, clinical trials registry (Clinicaltrials.gov) will be searched using all developmental and generic drug names, as specified above, for the approved biologics. Abstracts of the papers identified by the initial search will be evaluated for appropriateness to the study question, and all potentially relevant papers will be obtained and evaluated in detail. Abstract proceedings from Digestive Diseases Week, United European Gastroenterology Week, and the European Crohn's and Colitis Organization between 2007 and November 2015 will be hand-searched to identify potentially eligible studies published only in abstract form. The citation lists of identified relevant studies will be used for performing a further cross-search of the literature. Articles will be assessed independently by two investigators according to the pre-defined eligibility criteria. Any disagreements between investigators will be resolved by discussion.Outcome assessmentThe primary outcome will be the rate of remission of the active arm over the placebo arm at the end of the trial induction (week 4-14) for patients with short disease duration compared to patients with long duration of disease. This choice of primary outcome stems from the hypothesis that primary response to induction more closely reflects the biology of underlying IBD and its modulation by duration of disease, compared to response to treatment during maintenance (which is confounded by factors unrelated to disease biology per se, such as immunogenicity). The definition of the primary outcome will be a CDAI<150 or Mayo=2 with no individual subscore>1 for CD and UC, respectively. If these are unavailable, the remission measures will be based on the specific clinical score and outcome definition employed by the respective clinical trial. A sensitivity analysis will examine primary outcome only in studies using the pre-specified scores.The rate of response at the end of induction will be a secondary response. Additional secondary outcomes will be rate of response at the end of the maintenance phase of the trial (when applicable) and rate of remission at the end of the maintenance phase of the study (when applicable). These maintenance outcomes will be exploratory given the confounding effect of other factors on maintenance outcomes as denoted above, and given the heterogeneity across studies in the feed-in cohorts entering maintenance studies. To account for the later co-variate, a sensitivity analysis will be performed for the maintenance outcome, only for patients who responded to induction. As noted, all outcomes will be analyzed separately for CD and UC. Thus, an additional outcome for UC patients will be the rate of colectomy for patients with short versus long-duration of disease at the end of the trial. Additional secondary exploratory outcome will be the rate of induction of remission (primary outcome) in long versus short disease duration for patients treated with anti-TNF class of drugs analyzed jointly and in patients treated by anti-integrins analyzed jointly. The definition of short-disease duration (‘early disease') is somewhat arbitrary given the absence of evidence or of biologic markers that are able to delineate the underlying biology of early disease and distinguish it from late disease. In the SONIC trial, for instance, a cut-off of 3 years was adopted in the stratified randomization to delineate short/long disease duration (Colombel, NEJM 2010). Other published trials and expert opinions definitions of ‘early disease' range between 1-5 years since onset [Schreiber S, AJG 2010, Schreiber S, N Engl J Med 2007, Ananthakrishnan AN, AJG 2010]. A recent IOIBD working group pragmatically defined early CD as one that is =18 months in duration (Peyrin-Biroulet L, AJG 2012). Thus, the present meta-analysis will employ the IOIBD disease duration time-point of 18 months to define early versus late disease. However, bearing in mind the chronic life-long course of IBD, the diverse definitions existing, the possibility that patients enrolled into clinical trials at such early stage of their disease are a unique patient group, and the paucity of patients with<18 months of disease in clinical trials hampering the statistical power of the analysis, we will also perform a second main analysis of the primary outcome, defining disease duration of <3 years as a short term disease. This will ensure that any putative benefit of early treatment and higher response rates will not escape detection by the present proposal, even if the window of opportunity is restricted to very early disease, yet also allow pragmatic assessment of treatment benefit if instituted before or after three years of disease. An additional analysis will be performed using 5 years as the cut-off value, for assessment of dose-response curve-like impact of progressively longer disease duration on response to therapy. Thus, data will be recorded using tables of nominal numbers of patients with early disease shorter than 18 months compared to those with disease longer than 18 months. Similar table will be used to obtain outcomes for all patients stratified by disease duration of three years or longer, and for patients stratified by disease duration of Five years or longerData extractionAll data available in published form will be extracted independently by two investigators to a Microsoft Excel spreadsheet to denote: type of study (Induction, maintenance, induction + maintenance), number of centers, country of origin, geographical region, disease type (UC or CD), Number of patients with long and with short disease duration, duration of therapy, clinical scores used for outcome measurements, time-point for induction and/or maintenance outcome determination, prior anti-TNFa exposure and percentage of concomitant immunomodulator. Data will be extracted as per intention-to-treat outcomes unless precluded by trial reportingAssessment of risk of biasTwo investigators will independently assess the risk of bias as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins JPT et al, Cochrane handbook for systematic reviews of interventions: version 5.0.0, 2008 ). Factors assessed will include:Sequence generation (i.e. was the allocation sequence adequately generated?)Allocation sequence concealment (i.e. was allocation adequately concealed?) Blinding (i.e. was knowledge of the allocated intervention adequately prevented during the study?) Incomplete outcome data (i.e. were incomplete outcome data adequately addressed?) Selective outcome reporting (i.e. are reports of the study free of suggestion of selective outcome reporting?)Other potential sources of bias (i.e. was the study apparently free of other problems that could put it at a high risk of bias?)A judgement of 'Yes' indicates low risk of bias, 'No' indicates high risk of bias, and 'Unclear' indicates unclear or unknown risk of bias. Disagreements will be resolved by consensus. Study authors will be contacted when insufficient information was provided to determine risk of bias.We will use the GRADE criteria to assess the overall quality of evidence used for specific outcomes. As we will include evidence from randomized controlled trials all data will begin as high quality evidence. However, the overall quality of evidence for outcome can then be downgraded due to: (1) risk of bias from the studies, (2) indirect evidence, (3) inconsistency (unexplained heterogeneity), (4) imprecision in data, and (5) publication bias. These will be assessed as described further below, and as detailed elsewhere by the GRADE consensus (Guyatt GH, J Clin Epidemiol 2011). The overall quality of evidence for each outcome will be determined and classified as high quality (i.e. further research is very unlikely to change the confidence in the estimate of effect); moderate quality (i.e. further research is likely to have an important impact on the confidence in the estimate of effect and may change the estimate); low quality (i.e. further research is very likely to have an important impact on the confidence in the estimate of effect and is likely to change the estimate); or very low quality (i.e. the estimate is very uncertain) (Schünemann 2011).Data synthesis and statistical analysisData will be pooled using a random effects model to give a more conservative estimate of the likelihood of benefit of biologics in short-term disease, allowing for any heterogeneity between studies [DerSimonian R, et al, Control Clin Trials 1986]. As the outcomes are dichotomous data, we will use the Odds Ratio (OR) and extract a single data set from each study for further analysis. Pooled results will be expressed as odds ratio for the respective outcome for biologics versus placebo arm, for patients with short disease duration compared to those long duration of disease, with 95% confidence intervals (CI). Co-variates will include exposure to prior anti-TNF, concomitant immunomodulators, difference in clinical scores used to measure the efficacy outcomes, and the class of the biologic (anti-TNF vs. anti-integrins). Co-variates will be handled by the below mentioned sub-analyses, and by meta-regression (if more than 10 studies are available). As described, disease type will be a-priori considered by performing the entire analysis separately for UC and CD. For maintenance outcomes, an additional sensitivity analysis will be performed including only patients who responded to induction. The results of individual studies can be diverse, and this inconsistency within a single meta-analysis can be quantified with a statistical test of heterogeneity, to assess whether the variation across trials is due to true heterogeneity, or chance. We will assess heterogeneity among studies using the Chi-square test (a P value of 0.10 will be considered statistically significant) and the I2 statistic. This I2 value ranges from 0% to 100%, with 0% representing no observed heterogeneity, and larger values indicating increasing heterogeneity. An I2 value of 25% indicates low heterogeneity, 50% indicates moderate heterogeneity and 75% indicates high heterogeneity [Higgins JPT et al, BMJ 2003]. We planned, a priori, to perform separate analyses according to disease type (UC or CD) according to the above specified primary and secondary outcomes. We also plan a priori to perform additional sub-analyses for class of drug used and exposure to previous anti-TNFa. In additional sensitivity analyses we will assess the primary outcome by: 1) including also trials with high risk of bias 2) including only the studies employing the same clinical score criterion for remission induction (CDAI<150 or Mayo=2 with no individual subscore>1 for CD and UC, respectively) 3) Using a fixed-effect model to pool data if heterogeneity assessment reveals I2 <50%. 4) including only trials in which all patients were anti-TNF naïve 5) including only trials with <40% of included population on concomitant immunomodulators (i.e, also excluding combination therapy trials) 5) For the maintenance outcome analysis, a sensitivity analysis including only patients who responded to induction before rolling over to maintenance phase. Stata version 14.0 will be used for all statistical analyses and to generate Forest plots of pooled RRs with 95% CIs, as well as funnel plots. The latter will be assessed for evidence of asymmetry and therefore possible publication bias or other small study effects, using the Egger test, if there will be sufficient (n=10) eligible studies included in the meta-analysis, in line with published recommendations [Higgins JPT et al, BMJ 2003].