Systematic Review With Meta-analysis

Association Between Acetaminophen and Nonsteroidal Anti–inflammatory Drugs (NSAIDs) and Risk of Crohn's Disease and Ulcerative Colitis Exacerbation

O. O. Moninuola; W. Milligan; P. Lochhead; H. Khalili


Aliment Pharmacol Ther. 2018;47(11):1428-1439. 

In This Article


Search Strategy

We conducted a systematic literature search using a predetermined and registered protocol (Data S1) in accordance with the Preferred Reporting Items for Systematic review and Meta–Analysis (PRISMA) statement.[10,11] For electronic database search, we used medical subject headings (MeSH) and text words related to acetaminophen, NSAIDs, cyclooxygenase (COX–2) inhibitors, CD, and UC for MEDLINE (PubMed interface, 1974–March 2017) (Data S1) and Embase subject headings (Emtree) for EMBASE. Similar search strategies using related text words were employed for ongoing or recently completed trials and systematic reviews in Cochrane Central Register of Controlled Trials (Wiley interface, current issue), The International Clinical Trials Registry Platform Search Portal,, and PROSPERO. All the keywords were used to search the titles and abstracts. The reference lists of relevant articles were also reviewed for any additional articles not captured through the primary search method. Protocol Registration: PROSPERO: CRD42016051808.

Selection Criteria

Two reviewers (O. M. and W. M.) identified articles eligible for further review by performing an initial screen of identified abstracts and titles. Articles were eliminated in this initial screening if there were case series, case reports, animal studies, studies on aspirin or 5–Aminosalicylates derivatives, or studies not reported in English language. Full–texts of the remaining articles were retrieved and reviewed. Articles were considered for inclusion in the second screening if they reported original data from a study of eligible design, including randomised controlled trials (RCTs), nonrandomised clinical trials, prospective and retrospective cohort studies, and case–control or nested case–control studies of patients with established IBD. We also included studies that examined both adults and children if data for adults were reported separately. The observed agreement between reviewers for eligibility of articles on initial screening was 97.0%, corresponding to moderate agreement (k = 0.79) and in the second screening was 98.1%, corresponding to strong agreement (k = 0.87).[12] Disagreements between reviewers were resolved by consensus with third party experts (H. K. and P. L.).

Data Extraction

Our primary outcome was defined as IBD disease activity or complications while our primary exposure was use of acetaminophen, NSAIDs, or COX–2 inhibitors. There were significant variations in the outcome variable measures among studies, which included rates of hospitalisations and exacerbation, and changes in disease–specific activity index scores. Other variables extracted included study design, duration and frequency of exposure, publication year, and information on key indicators of study quality using Newcastle–Ottawa scale (NOS). We extracted reported odds ratios (ORs), hazard ratios (HRs), rate ratios (RRs) and incidence rate ratios (IRRs) with their respective 95% CIs or data enabling the calculation of these association measures. Both adjusted and unadjusted values were extracted; though, when available, the adjusted estimates were used.

Statistical Analysis

Meta–analysis summary estimates (RR) and 95% CIs were obtained by pooling effect estimates (RRs, IRRs, HRs and ORs) from all the eligible studies using fixed– or random–effects meta–analysis according to the method of Desimonian and Laird.[13] Given the relatively modest prevalence of IBD and its complications, the OR is a good approximate of the RR under the rare outcome assumption.[14] To assess for heterogeneity of RRs across studies, I2 statistics were calculated.[15] Given the small number of studies included in the meta–analysis, random effect models were used when there was moderate heterogeneity between studies, defined as I2 value ≥ 30%.[15] We assessed for publication bias using Egger's test, and funnel plot for analysis with 10 or more studies.[16] Subgroup analysis was performed for UC, CD and IBD when outcomes were reported separately by the primary authors. Consistent with prior studies, we used a cut–off score of 7 for NOS to define studies with low risk of bias,[17,18] and performed sensitivity analysis assessing the risk of disease exacerbation restricted to these studies. In addition, we examined the associations among cohort studies. All analyses were performed using stata 14 (StataCorp. 2015, College Station, TX, USA) statistical software package.