To our knowledge, this is the largest study in the US to assess the association between MC and medication use. The results of this multi-centre retrospective cohort study show that when compared to patients with chronic diarrhoea, patients with MC are less likely to use PPIs, H2 blockers and oral diabetes medications, and more likely to use NSAIDs.
Similar to our study, NSAIDs have been shown to have a positive association with MC in most other studies.[4–7,12,14,15,19] Only studies with small sample sizes showed no association.[8,10,11] Like most other studies, we included aspirin in the NSAID group. As both medications inhibit cyclooxygenase and cause disruptions in intestinal barriers, they both could be implicated in the development of MC by causing the translocation of toxins or bacteria in the gut. As aspirin is an irreversible binder of cyclooxygenase, it may confer a stronger risk of MC development. When examined separately, we found a positive trend for the association of MC with aspirin, but our study may have been underpowered to see a significant association. Nevertheless, patients with other types of colitis have been shown to have higher rates of NSAID use, so this association may not be specific to MC.
We also found an inverse association between MC and PPI use, unlike many prior studies that reported a positive association.[4,6,10,12,14,15] It has been theorised that PPIs may lead to MC by reducing stomach acidity and causing alterations in the intraluminal gut flora or electrolyte balance or by acting directly on the colonic mucosa to cause intraepithelial lymphocytosis. A review article by Law et al noted that the inclusion of healthy controls is a major limitation of current studies on this topic given the potential for reverse causality. Patients with MC may be more likely than healthy controls to have gastrointestinal symptoms prompting them to use or be prescribed PPIs. In fact, all prior studies that include a control group with diarrhoea do not show an association between PPIs and MC,[5,8,9,11] while the studies that emphasise a positive association all contain healthy controls.[4,6,10,12,14,15] Looking specifically at the US data, a non-significant trend for an inverse relationship between MC and PPIs was seen in one study which used a control group with and without diarrhea, and no association was seen in another where all controls had diarrhea.
While it is possible that PPIs may protect against MC by some as-yet unknown mechanism, it is more likely that the inverse association seen in our study is a feature of our control group being comprised of patients who used more PPIs, possibly related to their chronic gastrointestinal symptoms, including diarrhoea. The percentage of PPI use in our cohort (33%) was higher than that reported by a large US study which found PPI use among one in ten of all outpatient visits in 2009. It is possible that the patients with diarrhoea in our control group were prescribed PPIs at higher rates or took PPIs over the counter while undergoing further medical workup for the aetiology of their diarrhoea. While PPIs are not indicated for the treatment of diarrhoea, a large number of these patients may also have been experiencing dyspepsia or reflux thereby prompting their use of PPIs. Alternatively, our control group's increased use of PPIs could be due to a history of gastro-oesophageal disease or Barrett's oesophagus which would necessitate use of PPI. Furthermore, PPI use may cause diarrhoea as a side effect, especially in the elderly, independent of any putative effect on microscopic colitis. Unfortunately, we were unable to collect data regarding other symptoms or conditions at time of index colonoscopy given our study design. Interestingly, the control group from the Columbia University cohort used more PPIs than the MC group, while at the Mayo Clinic both controls and cases used similar amounts of PPIs. It is unclear why this discrepancy occurred as both Columbia University and the Mayo Clinic are tertiary care centres that serve similar patient populations with multiple medical comorbidities.
Since H2 blockers also reduce acid excretion in the stomach, it has been hypothesised that they may act to cause intestinal dysbiosis leading to MC through a similar mechanism as PPIs. A trend was seen for association between MC and H2 blockers in one study, while another study found no association. Our study found an inverse association between MC and H2 blockers, which could be due to an unknown protective mechanism that may be similar to that of PPIs. Alternatively, a larger number of patients in our control group may have used H2 blockers for reasons related to other indications such as dyspepsia or reflux.
We also found an inverse association between MC and oral diabetes medications as a group and also metformin alone. A nonsignificant trend for an inverse relationship with oral diabetes medications was seen in one study, while another found no association. Neither of these studies evaluated the association between MC and metformin individually. It is likely that oral diabetes medications are overrepresented in our control group because many of these medications, especially metformin, are known to cause diarrhoea. While MC may be associated with type 1 diabetes through a similar autoimmune contribution, it is unclear if type 2 diabetes and MC are related. More research is needed to determine if the use of oral diabetes medications used to treat type 2 diabetes or the presence of type 2 diabetes itself is associated with a decreased risk of MC.
Furthermore, many of these medication classes, such as PPIs, NSAIDs, statins, SSRIs and anti-diabetes medications, have been known to cause chronic diarrhoea through various mechanisms.[20,25,27] These medications may be the cause for the chronic diarrhoea experienced by our control group, and could explain the frequency of their use by our controls, who were identified based on their undergoing colonoscopy for the evaluation of diarrhoea. In our study, MC was not associated with many of these medications when compared to this type of control group, which suggests that medications may not be implicated in the development of MC, but rather may be contributing to the development of diarrhoea by other mechanisms. Future research is needed to determine such mechanisms by which many of these medications cause diarrhoea.
As MC is more common in patients with coeliac disease,[3,26,28] we also tested for interaction between coeliac disease and medication use with regard to their association with MC. We found that among patients with coeliac disease, there was no association with PPIs or H2 blockers and MC. This suggests that factors apart from medications, such as gluten or autoimmunity,[1,26] may be mediating the association between these two diseases. Since MC has also been linked to other lymphocytic-predominant gastrointestinal diseases, it is possible that there is some other common influence. Given the relatively small number of patients in our study, it is also possible that we are underpowered to adequately assess this association.
Our study has several limitations and strengths. First, due to our study design, data regarding medication dose or length of use were not collected. We were also not able to assess adherence to the gluten-free diet given the retrospective study design. Second, we had incomplete data regarding smoking history and autoimmune conditions other than coeliac disease which, because of their association with MC,[26,30] may have impacted our results. Thirdly, we were not able to collect data regarding indication for medication use or other symptoms present at index colonoscopy, which may have led to certain medications being over-represented in our control group. Additionally, since medication reconciliation with patients' pharmacies was not performed, it is possible that there were errors in patients' medication lists. Lastly, our study collected data from patients who were seen at two tertiary care centres, which limits the generalisability of the results. One strength of our study is that it includes a control group of patients who underwent colonoscopy for evaluation of diarrhoea, as opposed to healthy controls, which minimises reverse causality bias. Additional strengths include its multi-centre study design, selection of patients with biopsy-proven MC, inclusion of 13 medication classes and stratification by presence of coeliac disease.
In conclusion, in the largest US study on this topic, we found an inverse relationship between MC and H2 blockers, oral diabetes medications and PPIs, and a positive relationship with NSAIDs. Future studies should include control groups with chronic diarrhoea in order to further elucidate the relationship between MC and medication use.
BL: The Louis and Gloria Flanzer Philanthropic Trust.
Declaration of personal interests
Guarantor of the article
This analysis was submitted and approved by the Institutional Review Boards at Columbia University Medical Centre and Mayo Clinic.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Aliment Pharmacol Ther. 2021;53(11):1209-1215. © 2021 Blackwell Publishing