Systematic Review With Meta-analysis

Loss of Response and Requirement of Ustekinumab Dose Escalation in Inflammatory Bowel Diseases

Hongsheng Yang; Bingyang Li; Qin Guo; Jian Tang; Bo Peng; Ni Ding; Miao Li; Qingfang Yang; Zicheng Huang; Na Diao; Xia Zhu; Jun Deng; Huili Guo; Pinjin Hu; Kang Chao; Xiang Gao


Aliment Pharmacol Ther. 2022;55(7):764-777. 

In This Article


This study was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines and the proposal for reporting Meta-analysis of Observational Studies in Epidemiology (MOOSE) (Tables S1 and S2).[7,8]

Bibliographic Search

The PubMed, Embase, Web of Science and Cochrane Library databases were systematically searched for studies up to July 2021. The search terms were "inflammatory bowel disease," "Crohn's disease," "ulcerative colitis," "ustekinumab," "anti-IL-12/23p40" and their synonyms. All relevant MeSH terms were also used in PubMed. The search strategy is outlined in Data S1. Major conference proceedings (European Crohn's and Colitis Organization, United European Gastroenterology Week and Digestive Disease Week) were manually searched between 2015 and 2021. We also searched the reference lists of selected articles or related reviews of potential studies.

Eligibility Criteria

We included clinical trials (randomised or non-randomised) and cohort studies (prospective or retrospective) investigating ustekinumab in adult CD or UC patients. Eligible studies reported or provided adequate data to calculate incidence rates of LOR or dose escalation among CD or UC patients who responded to ustekinumab or the proportion of patients who regained response/remission among secondary non-responders after ustekinumab dose escalation. We excluded studies in which (a) data on CD or UC patients could not be separated; (b) the LOR rates, need for dose escalation, or efficacy of dose escalation were not systematically documented; (c) LOR or dose escalation was investigated in primary non-responders or a mixed population with primary responders and primary non-responders and (d) LOR or dose escalation was investigated in a paediatric population. We considered all articles regardless of publication type (full-text publications or conference abstracts).

The primary outcomes were (a) the annual risk of LOR/dose escalation among ustekinumab-treated responders, (b) the rate of LOR/dose escalation among ustekinumab-treated responders and (c) the efficacy of ustekinumab dose escalation in patients with secondary LOR. The secondary outcome included predictors of LOR or dose escalation.

Study Selection

Two reviewers (HY and BL) independently reviewed the titles, abstracts or full texts of the identified studies. All articles were screened according to inclusion and exclusion criteria. The most comprehensive articles were chosen for the overlapping reports. Disagreements were resolved through discussion to achieve a consensus.

Data Extraction and Risk of Bias Assessment

Two investigators (HY and BL) independently retrieved the data using a standardised data extraction form. The risk of bias in the included studies was assessed using the Institute of Health Economics of Canada (IHE) quality appraisal checklist.[9]

Data Analysis

Patients with CD and UC were analysed separately. For each study, the percentage of events (LOR or dose escalation) was calculated using the event number divided by the number of primary responders. The event number was used as the numerator to calculate the annual risk of events, and the number of induction responders multiplied by the follow-up time was used as the denominator. For studies including patients with various follow-up times, the mean/median follow-up time was used. Proportions of regained responses after dose escalation were also determined.

The variance-stabilising double arcsine transformation was used for incidence rates and proportions because the transformation stabilised the variance, and a value closer to the limits of the 0–1 range could be pooled using the inverse variance method.[10] Summary estimates were synthesised by a random effects model and fixed effect model using the DerSimonian-Laird approach.[11,12] The random effects model was chosen a priori for all analyses to yield a more conservative estimate.

Cochran's Q-test was used to assess the heterogeneity caused by study variation, with a significance level of 0.10.[13] The extent of heterogeneity among the studies was quantified by calculating the I2 statistic.[14] We used cut-offs of <40%, 30%–60%, 50%–90% and >75% to suggest low, moderate, substantial and considerable heterogeneity respectively. Subgroup analyses were then conducted to investigate possible sources of heterogeneity.[15] Subgroup analyses were performed based on publication type, study design, induction regimen, maintenance regimen, follow-up time, the definition of LOR and dose escalation modality. The subgroup difference was assessed by a Q-test for heterogeneity in the random effects model.[16] Begg's and Egger's tests were performed to evaluate potential publication bias if the number of included studies was not <10.[17,18] All analyses were performed using R software (version 3.6.1) and the "meta" package for R.[19,20] All statistical tests were two-sided, and statistical significance was defined as P < 0.05, except for heterogeneity tests.