Severe intestinal malabsorption associated with ACE inhibitor or angiotensin receptor blocker treatment. An observational cohort study in Germany and Italy

Abstract Purpose The angiotensin II receptor blocker (ARB) olmesartan has been recently associated with sprue‐like enteropathy (SLE), a gastrointestinal condition characterized by intestinal malabsorption (IM) and severe diarrhea. Whether the increased risk of SLE is substance‐specific or a class effect involving all ARBs is uncertain. The aim of this study is to assess the risk of enteropathy associated with ARBs and angiotensin converting enzyme inhibitors (ACE‐i) by using data from large administrative and claim databases. Methods We obtained data from Italian local health‐care units and a large German claim database and included patients treated with olmesartan, other ARBs, and ACE‐i. In the absence of a specific diagnosis code for SLE, International Classification of Diseases codes for IM were used. Analysis implemented a Poisson regression with robust error variance procedure, which allowed accounting for different clusters (local health‐care units and countries) and correctly estimating the standard error for the relative risk of rare event occurrence. Results Patients were divided into 3 groups: olmesartan (25.591, 5.5%), other ARBs (104.901, 22.5%), and ACE‐i patients (334.951, 72.0%). Baseline characteristics were similar overall. The incidence of unspecified IM in ACE‐i patients was not different compared with that of olmesartan, whereas a higher rate ratio was observed when comparing ARB patients with the olmesartan group (RR: 2.50, 95% CI 1.21 to 5.19, P .01). When International Classification of Diseases codes for coeliac disease were included, no differences were observed. Conclusions We could not confirm previous findings of a higher risk of malabsorption in olmesartan‐only patients, and drug‐induced enteropathy should be considered the result of exposure to the class of ARBs rather than a specific drug‐related effect.

antibiotics, nonsteroidal antiinflammatory agents, chemotherapeutics, immunosuppressants, and, recently, olmesartan (an angiotensin-II receptor blocker, ARB) have been associated with SLE occurrence. 5 The association between severe SLE and the treatment with olmesartan medoxomil (OM) has initially been reported in a case series of 22 patients diagnosed with refractory celiac disease 6 followed by a small series of individual case reports. 4,7-10 A large observational cohort study, based on the French health insurance claim database and using International Classification of Diseases, Tenth Revision (ICD-10) codes for intestinal malabsorption (IM) and coeliac disease diagnosis, highlighted an increased risk of hospitalization in patients treated with OM when compared with other ARBs and angiotensin converting enzyme inhibitors (ACE-i). 11 Otherwise, it is not appropriate to rule out the class-effect hypothesis 12 because recent case reports suggested an association between severe enteropathy and other ARBs, including valsartan, irbesartan, telmisartan, eprosartan, losartan, and candesartan. 10,[13][14][15][16][17][18][19][20] A large population-based study of ARB-treated patients managed by the general practitioners in Italy and Germany showed similar low proportions of unspecified IM diagnosis among the different drugs belonging to the same class. This suggests the absence of differences among different ARBs and shows the very low incidence of IM in association with their intake. 21 These previous findings even more emphasize the need to clarify the role of hypertensive medications in the onset of severe forms of enteropathy.
The present study aimed at assessing the risk of enteropathy associated with ARBs and ACE-i in 2 European countries (Italy and Germany) and at assessing a potential increased risk for OM.

| Study design and population
This study was a noninterventional retrospective analysis based on longitudinal secondary data sources. Two different cohorts were constructed from administrative and claims databases in Italy and Germany, respectively. In Italy, all patients who initiated treatment with an ACE-i (Anatomical Therapeutical Classification [ATC] codes: C09A, C09B) or an ARB (ATC codes: C09C, C09D) between January 1, 2011 and December 31, 2015 were considered. Furthermore, according to the current availability of validated data per local health-care unit (LHU), 3 different selection periods were considered for the analysis to ensure the accuracy and efficiency of results produced (see Table   S1). For German patients, the selection period went from January 1, 2011 to December 31, 2014. The first prescription of ACE-i or ARB during this period constituted the entry date in the cohort (index date).
To limit the study to treatment-naïve patients for the studied drugs, patients who had at least 1 prescription containing ACE-i or ARB during the 12 months preceding the index date (preselection period) were excluded from the analysis. In addition, also patients that at the index date had prescriptions falling in more than 1 of the treatment groups of interest were excluded. Further exclusion criteria applied in the preselection period were hospitalization for IM (Italy: ICD-9 code: 579.x; Germany: K90.x), any exemption for coeliac disease (only available in Italy: code RI0060), any coeliac disease-specific serological testing (only available in Italy: 90.53.D, 90.49.5, 90.47.E codes), and any IM diagnosis (only available in Germany: ICD-10 codes K90.x). The 2 cohorts were then pooled into the final database used in the analysis.

| Exposure definition
The following 3 treatment exposures were investigated: exposure to OM, exposure to other ARBs, and exposure to ACE-i. Exposure to treatment was defined as both monotherapy and fixed combinations with diuretics and/or calcium channel blockers. The total number of days on therapy was analyzed by means of the defined daily dose.
Starting from the treatment at the index date (OM, other ARB, or ACE-i), the end of the exposure period was determined by treatment discontinuation or switch to a different antihypertensive therapy.
Discontinuation was defined as any gap of more than 90 days between refilling prescriptions.

| Outcomes
Given the lack of a specific diagnosis code defining SLE, ICD codes for IM were considered as proxies for the diagnosis of drug-associated sprue, both for primary and secondary outcomes, as previously used in the study by Basson et al. 11 The primary outcome was hospitaliza-

KEY POINTS
• A relatively low incidence rate of intestinal malabsorption hospitalizations in a large cohort of patients treated with ARBs or ACEIs in Italy and Germany is found.
• Previous findings about the higher risk of malabsorption linked exclusively with olmesatan-treated patients are not confirmed.
• ARBs other than olmesartan are associated with a significantly increased risk of unclassified malabsorption compared with the olmesartan group.
• ACEi patients present with a nonsignificant lower risk of malabsorption than all ARBs users.
• Drug-associated enteropathy should be considered as the potential consequence of the entire ARB class exposure rather than a specific drug-related effect.

| Data sources 2.4.1 | Italy
We obtained access to the administrative databases of 5 Italian LHUs:  The following databases were used: registration data (demography, time insured, and regional distribution), outpatient and inpatient cares, drug prescriptions, sick leaves, and sick benefits data. Using a personrelated pseudonym, it was possible to unambiguously identify patients in all datasets abovementioned.      abdominal discomfort, and diagnosis of coeliac disease or microscopic colitis when comparing olmesartan with control groups. 32,33 However, this result is in contrast with the findings by Basson et al, who reported a higher risk of malabsorption in patients treated with OM, even if confirming that olmesartan-induced enteropathy remains a rare condition. 11 A possible explanation of the difference between the 2 studies is that, in the present study, the primary outcome was defined by using ICD codes for unspecified IM based on the assumption that a clinician, in the absence of a specific code defining SLE diagnosis, would select a general and nonspecific diagnosis code. In fact, in the absence of a definitive etiology for villous atrophy, patients are most likely characterized as having unclassified sprue, a diagnosis of exclusion, for which the optimal management is still unknown. 8 Nevertheless, because SLE is an adverse drug reaction that mimics the appearance of celiac disease, in this study, we also assessed the risk of IM considering all ICD-10K90.x and ICD-9

| Statistical methods
579.x codes for Germany and Italy, respectively. We found crude incidence rates of malabsorption in our 2 cohorts almost 4-fold higher for Also, the use of national representative databases designed for administrative and billing purposes provides insight into otherwise difficult-tostudy, low-incidence clinical events and outcomes. 34 In conclusion, this study suggests that drug-induced enteropathy should be considered the result of exposure to the class of ARBs rather than a specific drug-related effect. Yet, the higher risk found in relation to the ARBs users is sensitive to the ICD codes considered in our study. Similarly, it has to be stressed that to better understand the causality of ARB-associated sprue and its magnitude, other studies accounting for different populations, different study designs, different time periods, and duration of drug exposure alongside the consideration of specific population-related lifestyles, individual susceptibility, and co-administered drugs should be encouraged. 35 Mechanisms of drug-induced enteropathy await further exploration.

ETHICS STATEMENT
Approval by the pertinent Ethical Committees and Competent Authorities were obtained in accordance with all the regulations in force and regulatory requirements.