We identified 344 patients with MC and 668 matched controls. Demographic information is shown in Table 1. The mean (SD) age of both groups was 63.1 (15.6) years and a majority (71.4%) were women. Patients with MC were more likely to have coeliac disease (OR 12.70; 95% CI 5.73–28.13). Despite more patients with MC currently smoking (10.5%) compared to controls (4.9%), the overall association between smoking history and MC was not significant (OR 1.07; 95% CI 0.89–1.28). Of the 344 patients with MC, 131 (38.1%) had CC, 183 (53.2%) had LC and 19 (5.5%) had mixed phenotypes, with 11 (3.2%) having unknown types.
Table 2 shows medication use in patients with MC and controls. MC was inversely associated with H2 blockers (OR 0.46; 95% CI 0.25–0.87), oral diabetes medications (OR 0.48; 95% CI 0.28–0.83) and PPIs (OR 0.68; 95% CI 0.51–0.92). There was a positive association between MC and NSAID use (OR1.64; 95% CI 1.14–2.37). After adjusting for smoking, an inverse association was still seen between MC and H2 blockers (OR 0.46; 95% CI 0.24–0.88), oral diabetes medications (OR 0.47; 95% CI 0.27–0.81) and PPIs (OR 0.64; 95% CI 0.47–0.87), and a positive association was seen with NSAIDs (OR 1.63; 95% CI 1.12–2.38). There were no significant differences between patients with MC and controls in their use of other medications, on both univariate and multivariate analysis.
The medications that were significant on multivariate analysis were then assessed by histologic subtype (Table 3). LC was inversely associated with PPI use (P = 0.02), oral diabetes medication (P = 0.007) and H2 blocker use (0.05), although not the other subtypes. NSAID use was positively associated with LC (P = 0.02), and showed a trend for association with CC (P = 0.08).
We then performed a multivariate analysis, adjusting for use of PPIs, NSAIDs, SSRIs/SNRIs, H2 blockers, oral diabetes medication and smoking history. Inverse relationships were still seen between the risk of MC and the following medications: H2 blockers (OR 0.47; 95% CI 0.24–0.91), oral diabetes medications (OR 0.49; 95% CI 0.28–0.86) and PPIs (OR 0.69; 95% CI 0.50–0.94). NSAID use still showed a positive association with MC (OR 1.71; 95% CI 1.16–2.52).
We then tested for interaction between coeliac disease and medication class in our multivariate model. The presence of coeliac disease was an effect modifier for the association with H2 blockers (P for interaction = 0.007) and PPIs (P for interaction = 0.004). Among coeliac disease patients, the association between H2 blocker use and MC was not significant, although the confidence interval was wide (OR 6.8; 95% CI 0.21–226), while among noncoeliac disease patients the inverse association remained significant (OR 0.38; 95% CI 0.18–0.81). Likewise, among coeliac disease patients the association between PPI use and MC was not significant (OR 0.47; 95% CI 0.35–6.37), while among noncoeliac disease patients the inverse association remained significant (OR 0.71; 95% CI 0.52–0.98).
When examining individual PPIs (Table 4), we found no significant differences between specific PPIs used in patients with MC compared to controls (P > 0.05). Next, we tested the association between PPI use and MC, stratifying by site. At Columbia University Medical centre 15.1% of patients with MC used PPIs compared to 28% of controls (OR 0.45; 95% CI 0.28–0.75), while at the Mayo Clinic 35.2% of patients with MC used PPIs compared to 37.4% of controls (OR 0.93; 95% CI 0.63–1.38).
We also examined other individual medications. When aspirin and other NSAIDs were examined separately, there were no associations seen with MC on univariate analyses (other NSAIDs: OR 0.84; 95% CI 0.53–1.35; aspirin: OR 1.33; 95% CI 0.93–1.92) or multivariate analyses (adjusted other NSAIDs: OR 0.80; 95% CI 0.49–1.29; adjusted aspirin: OR 1.31; 95% CI 0.90–1.89). We found an inverse association between MC and metformin on both univariate (OR 0.45; 95% CI 0.23–0.85) and multivariate analysis (OR 0.43; 95% CI 0.22–0.82).
Aliment Pharmacol Ther. 2021;53(11):1209-1215. © 2021 Blackwell Publishing