Article Text
PDF
Abstract
Objectives Severe sprue-like enteropathy associated with olmesartan has been reported, but there has been no demonstration of an increased risk by epidemiological studies.
Aim To assess, in a nationwide patient cohort, the risk of hospitalisation for intestinal malabsorption associated with olmesartan compared with other angiotensin receptor blockers (ARB) and ACE inhibitors (ACEIs).
Design From the French National Health Insurance claim database, all adult patients initiating ARB or ACEI between 1 January 2007 and 31 December 2012 with no prior hospitalisation for intestinal malabsorption, no serology testing for coeliac disease and no prescription for a gluten-free diet product were included. Incidence of hospitalisation with a discharge diagnosis of intestinal malabsorption was the primary endpoint.
Results 4 546 680 patients (9 010 303 person-years) were included, and 218 events observed. Compared with ACEI, the adjusted rate ratio of hospitalisation with a discharge diagnosis of intestinal malabsorption was 2.49 (95% CI 1.73 to 3.57, p<0.0001) in olmesartan users. This adjusted rate ratio was 0.76 (95% CI 0.39 to 1.49, p=0.43) for treatment duration shorter than 1 year, 3.66 (95% CI 1.84 to 7.29, p<0.001) between 1 and 2 years and 10.65 (95% CI 5.05 to 22.46, p<0.0001) beyond 2 years of exposure. Median length of hospital stay for intestinal malabsorption was longer in the olmesartan group than in the other groups (p=0.02). Compared with ACEI, the adjusted rate ratio of hospitalisation for coeliac disease was 4.39 (95% CI 2.77 to 6.96, p<0.0001) in olmesartan users and increased with treatment duration.
Conclusions Olmesartan is associated with an increased risk of hospitalisation for intestinal malabsorption and coeliac disease.
- CELIAC DISEASE
- DRUG TOXICITY
- EPIDEMIOLOGY
- ADVERSE DRUG REACTIONS
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Cases of olmesartan-induced severe sprue-like enteropathy have been reported.
The reality of the association has been questioned.
It is also unknown whether there is an association between enteropathy and other angiotensin receptor blockers (ARBs).
What are the new findings?
In this large nationwide observational patient cohort, olmesartan exposure is associated with an increased risk of hospitalisation for intestinal malabsorption and coeliac disease.
This relative risk increases with treatment duration.
We found no such risk for other ARBs.
How might it impact on clinical practice in the foreseeable future?
Patients and physicians, including gastroenterologists, should be widely informed of this severe complication.
Introduction
Olmesartan is an angiotensin II receptor blocker (ARB); its prodrug, olmesartan medoxomil, has been first approved in 2002 in the USA and in 2003 in the European Union, for the treatment of hypertension. Severe sprue-like enteropathies associated with olmesartan have recently been reported.1 ,2 The first case series included 22 patients. These patients had severe, chronic diarrhoea and weight loss. Duodenal biopsies showed villous atrophy and inflammation. Coeliac disease serology was negative, and gluten-free diet was ineffective. All patients had taken olmesartan for several months or years. Olmesartan withdrawal was followed by clinical and, when assessed, histological improvement. Nine additional case reports and one literature review have been published3–8 and confirmed these findings. Olmesartan seems to account for a significant proportion of non-coeliac sprues. In a series of 72 adult patients with villous atrophy and negative coeliac disease serology, olmesartan was prescribed in 16 of these patients, and all but one obtained clinical improvement after olmesartan discontinuation.9 More recently, a new series of 39 patients with olmesartan-associated sprue has been reported.10 Interruptions and reintroductions could be studied in a subgroup of 12 patients. Interruptions were followed by remissions, and reintroductions were followed by relapses. These reports suggest that olmesartan may cause severe enteropathy. However, the level of evidence of case reports and small series is limited. The association between olmesartan and enteropathy has also been questioned, as the ROADMAP trial, a large randomised controlled trial with several years of follow-up, did not demonstrate any difference in diarrhoea or GI event rates between olmesartan and placebo.11–13 However, in 2011, the FDA requested a Mini-Sentinel modular programme report of risk assessment because the number of cases of coeliac disease among users of olmesartan was higher than expected in the FDA Adverse Event Report System. The incidence of coeliac disease was found to be similar among all ARBs, including olmesartan.14 ,15 Nevertheless, in July 2013, the FDA issued a ‘Drug Safety Communication’ approving a label change to include sprue-like enteropathy linked to olmesartan.
The association between olmesartan and enteropathy needs to be further investigated. The causality of the association remains uncertain, and its magnitude has not been determined. Moreover, it is unknown whether the association between enteropathy and ARB is limited to olmesartan or also includes other ARBs.
The objective of this study was to assess the risk of enteropathy associated with olmesartan. However, no specific diagnosis code is available for this disease, which was unknown prior to the first published case series. We, therefore, assessed the risk of intestinal malabsorption and coeliac disease associated with the prescription of olmesartan. For this purpose, we compared the rates of hospitalisation with a discharge diagnosis of intestinal malabsorption in patients who were prescribed olmesartan, other ARBs and ACE inhibitor (ACEI) in a large nationwide patient cohort.
Methods
Data sources
The SNIIRAM (Système National d'Information Interrégimes de l'Assurance Maladie) is the French National Health Insurance anonymised claim database. Claims from the general health insurance scheme were used in this study. They include 51.3 million of the 65.7 million inhabitants of France (2013 census), and are available since 2006. Anonymised patient-level records contain billable claims and sociodemographic data such as age and sex. Billable claims include dispensed drugs, laboratory tests (without their results), medical procedures and ambulatory medical care. This database has been previously described.16 ,17
The French hospital discharge database programme médicalisé des systèmes d'information (PMSI) contains information about each patient admitted to a public or private hospital in France, including inpatients and outpatients. This database contains information such as discharge diagnosis (recorded by International Classification of Diseases 10 (ICD-10) code), comorbidities, age, sex, diagnosis-related group, medical procedure performed and length of stay.16–18
These two databases were linked in the present study in order to correlate drug prescription with hospitalisation rates and diagnoses. This study was approved by the French data protection agency (Commission Nationale de l’Informatique et des Libertés). All databases used in this study only contained anonymous patient records.
Study population
A cohort was constructed from the SNIIRAM, including all adult patients who initiated treatment with an ACEI or ARB between 1 January 2007 and 31 December 2012. The first filled prescription of ACEI or ARB during this period of time constituted the entry date in the cohort (index date). Patients had to be enrolled in the database for at least 1 year before the index date to prevent left censoring. To ensure the absence of left censoring, patients were required to have at least one recorded claim of any type, 1–2 years before the index date. In order to limit the study to incident users of studied drugs, we excluded patients who had filled a prescription containing ACEI or ARB during the 12 months before the index date. Patients with at least one of the following criteria were also excluded: (1) hospitalisation with a discharge diagnosis of intestinal malabsorption (ICD-10 codes K90x) during the year before the index date, (2) any filled prescription containing a gluten-free diet product during the year before the index date, (3) any reimbursed coeliac disease-specific serological testing during the year before the index date.
The ICD codes of coeliac disease and malabsorption were considered as proxies for the diagnosis of olmesartan-associated sprue. It was, therefore, necessary to exclude patients with history of intestinal malabsorption before index date and/or patients with coeliac disease. Therefore, patients who had undergone serological testing or had received gluten-free diet or had been hospitalised with a discharge diagnosis of intestinal malabsorption before the index date were excluded.
Outcomes
The primary outcome was hospitalisation with a discharge diagnosis of intestinal malabsorption (ICD-10 codes K90x). The secondary outcome was hospitalisation with a discharge diagnosis of coeliac disease (ICD-10 code K90.0). Patients were censored at the first event, death or end of the study (31 December 2012 in the main analysis and 31 May 2012 in the sensitivity analysis to avoid information bias).
Exposure assessment
Three kinds of exposures were studied: exposure to olmesartan, exposure to other ARB and exposure to ACEI. Exposure was defined as follows for these three groups. It started from the date of a filled prescription containing a drug of interest (ie, olmesartan, other ARB or ACEI). The end of exposure was defined as the end of prescription duration plus a grace period of 30 days. Grace period is commonly used and recommended in pharmacoepidemiological studies based on claim databases in order to account for incomplete medication adherence and avoid underestimation of drug exposure or misattribution of events. Patient could simultaneously fall into several exposure categories (eg, ACEI+olmesartan). Such periods of overlapping exposure to different drug class were removed from the analysis to prevent misattribution of events. However, they were accounted for in the calculation of treatment duration to prevent classification bias.
Statistical methods
For the primary outcome, a Poisson regression model adjusted for the following potential confounders was used: age, sex, heart failure, dementia, diabetes, immune-mediated diseases (rheumatoid arthritis, Hashimoto thyroiditis, IgA deficiency, dermatitis herpetiformis, lupus, Sjogren, dermatopolymyositis, complement deficiency, angioedema, IBDs), transplantation, ongoing cancer and renal failure. The comorbidities were based on the diagnoses, medical procedures and drug prescriptions from the PMSI and the SNIIRAM.
For the secondary outcome (hospitalisation with a discharge diagnose of coeliac disease), the Poisson regression model was adjusted for age, sex and the following comorbidities: heart failure, diabetes, immune-mediated diseases (rheumatoid arthritis, Hashimoto thyroiditis, IgA deficiency, dermatitis herpetiformis, lupus, Sjogren, dermatopolymyositis, complement deficiency, angioedema, IBDs), active cancer and renal failure. Dementia and transplantation were removed because of a lack of events. We adjusted for these comorbidities for the following reasons. Patients with immune-mediated abnormalities are at increased risk for coeliac disease. Patients with cancer or allograft recipients are often prescribed drugs that may provoke diarrhoea and malabsorption. Patients with dementia are commonly treated differently from other patients regardless of the disease. Diabetes is a common cause of GI symptoms, including diarrhoea (autonomous neuropathy). Renal failure and heart failure may have influenced the choice of antihypertensive drug.
Poisson regression model fit was assessed by overdispersion analysis, using the deviance/number of degree of freedom ratio and the Pearson χ2 statistic. Medians were compared by the multisample median test (Brown–Mood test), which assigns 1 for observations greater than the median, and 0 otherwise, and produces χ2 statistics.19 Data management and statistical analyses were performed with SAS Enterprise Guide V.4.3.
Results
Study population
A total of 4 552 130 patients initiating ARB or ACEI treatment between 2007 and 2012 were selected from the database; 154 patients who had been hospitalised for intestinal malabsorption during the 12 months preceding inclusion and 4611 patients who had undergone coeliac disease serology testing during the past 12 months were excluded. Finally, 685 patients with a reimbursement claim for a gluten-free diet product in the past 12 months were also excluded. A total of 4 546 680 patients corresponding to 9 129 149 person-years (PY) were included: 118 846 PY of multiple exposures were excluded from the analysis and the remaining 9 010 303 PY of single treatment exposure were distributed as follows: 3 646 311 PY of ACEI exposure, 860 894 PY of olmesartan exposure and 4 503 098 PY of other ARB exposure. The inclusion flow chart is presented in figure 1.
Inclusion flow chart. ACEI, ACE inhibitor; ARB, angiotensin receptor blocker.
Baseline patient characteristics are presented in table 1. Mean age at inclusion was 63.9 years in the ACEI group, 61.3 years in the olmesartan group and 62.3 years in the other ARB group. The ACEI group comprised fewer women (45.6%) than the olmesartan group (53.9%) and the other ARB group (55.6%). The Poisson regression model was adjusted for both age and sex.
Population characteristics
Seventy-seven per cent of the PY in the olmesartan group were included during the 2010–2012 period compared with 72% in the ACEI group and 70% in the other ARB group. Median duration of treatment exposure varied from 326 days (ACEI) to 348 days (olmesartan) and 514 days (other ARBs).
No coding trend of hospital discharge diagnoses of intestinal malabsorption was observed over the study period (see online supplementary table S2).
Incidence of severe malabsorption and coeliac disease
Two hundred eighteen hospitalisations for intestinal malabsorption were observed, 87 in the ACEI group, 48 in the olmesartan group and 83 in the other ARB group, yielding crude incidence rate of 2.4 per 100 000 PY, 5.6 per 100 000 PY and 1.8 per 100 000 PY, respectively.
Olmesartan was associated with an adjusted rate ratio of 2.49 (95% CI 1.73 to 3.57, p<0.0001) of hospitalisation with a discharge diagnosis of intestinal malabsorption compared with ACEI and a rate ratio of 3.17 (95% CI 2.22 to 4.53, p<0.0001) compared with other ARBs. ARBs other than olmesartan were associated with a non-significant rate ratio of 0.78 (95% CI 0.58 to 1.07, p=0.12) of hospitalisation with a discharge diagnosis of intestinal malabsorption, compared with ACEI. Women had a higher rate ratio of hospitalisation with a discharge diagnosis of intestinal malabsorption (rate ratio 1.42, 95% CI 1.08 to 1.87, p=0.01). Inclusion of an interaction term between sex and treatment was added to the model, but was not significant, and was, therefore, not kept in the multivariate model. Gender-stratified results were also calculated. We found no difference between men and women (data not shown). Age had no influence on this rate ratio.
Median length of hospital stay was longer in the olmesartan group (9 days) than in the other ARB group (2 days) and the ACEI group (4 days) (p=0.02).
Hospitalisations with a discharge diagnosis of coeliac disease (ICD-10 code K90.0) were also studied, as olmesartan-associated enteropathy mimics coeliac disease. Adjusted rate ratio of hospitalisation with a discharge diagnosis of coeliac disease was 4.39 (95% CI 2.77 to 6.96, p<0.0001) in patients who were prescribed olmesartan compared with those who were prescribed ACEI and 4.82 (95% CI 3.12 to 7.45, p<0.0001) compared with other ARBs. This ratio was 0.91 (95% CI 0.58 to 1.42, p=0.68) in patients who were prescribed other ARBs compared with those who were prescribed ACEI. See table 4 for details.
The first case report linking olmesartan and enteropathy was published online on 25 June 2012. We, therefore, performed a sensitivity analysis in which the study period and follow-up ended on 31 May 2012, which gave very similar results (see online supplementary tables S3–S5).
Risk over time
Descriptive data were in favour of non-homogeneity of risk according to the duration of treatment exposure (table 2). To account for such changes in risk and to assess the kinetics of the risk of hospitalisation with a discharge diagnosis of intestinal malabsorption associated with olmesartan exposure, the model was stratified on treatment exposure. The following duration strata were used: less than 1 year, between 1 and 2 years, and 2 years or more. Compared with ACEI, the adjusted rate ratio of hospitalisation with a discharge diagnosis of intestinal malabsorption associated with olmesartan exposure was 0.76 (95% CI 0.39 to 1.49, p=0.43) for treatment duration shorter than 1 year, 3.66 (95% CI 1.84 to 7.29, p<0.001) between 1 and 2 years of treatment exposure and 10.65 (95% CI 5.05 to 22.46, p<0.0001) beyond 2 years of treatment exposure (table 3). Very similar results were obtained when follow-up ended on 31 May 2012 (see online supplementary tables S4 and S5). Compared with ACEI, the rate ratio of hospitalisation with a discharge diagnosis of coeliac disease was 1.98 (95% CI 0.85 to 4.61, p=0.11) for treatment shorter than 1 year; 4.36 (95% CI 2.04 to 9.34, p<0.001) for treatment between 1 and 2 years and 10.21 (95% CI 4.21 to 24.76, p<0.0001) for more than 2 years of olmesartan exposure (table 4). Details of discharge diagnoses by duration of treatment exposure in each group are presented in online supplementary table S1. No overdispersion was observed in any Poisson regression models.
Risk over time: descriptive data
Crude and adjusted rate ratios of hospitalisation with a discharge diagnosis of intestinal malabsorption over time (ref: ACEI)
Adjusted rate ratios of hospitalisation with a discharge diagnosis of coeliac disease (ref: ACEI)
Discussion
In this large nationwide cohort of patients, olmesartan users were found to have an increased risk of hospitalisation for intestinal malabsorption and coeliac disease compared with ACEI. These risks increased with duration of olmesartan exposure up to 10-fold beyond 2 years of exposure. Users of ARBs other than olmesartan did not exhibit an increased risk of hospitalisation for intestinal malabsorption or coeliac disease. These results were adjusted for potential confounders. During the first year of treatment, patients treated with other ARBs had a decreased rate of hospitalisation for intestinal malabsorption compared with patients treated with ACEI. There was an excess of diagnoses of malabsorption other than coeliac disease among ACEI users (ICD-10 codes K90.4, K90.8 and K90.9; see online supplementary table S1). However, no significant difference in terms of risk of hospitalisation for coeliac disease (ICD-10 code K90.0) was observed between users of ARBs other than olmesartan and ACEI users. The reason for this is unclear, but it does not affect the consistency of the results. It may have underestimated the rate ratio associated with olmesartan as compared with ACEI.
The strength of the association and the consistency with reported cases (including the long lag time between initiation of olmesartan and diagnosis of malabsorption) are strong arguments in favour of causality. In addition, the longer length of hospital stay in patients who were prescribed olmesartan suggests that their disease was distinct from and more severe than that observed in patients receiving ARBs or ACEI. Patients who obtained clinical improvement after stopping olmesartan and who experienced subsequent recurrence of symptoms on olmesartan rechallenge have also been described.6 ,7 In the aforementioned ROADMAP trial, no significant difference in the rate of GI adverse events or diarrhoea was observed between olmesartan and placebo.10–12 However, these adverse events are common in patients with diabetes (reported in 3.5% and 2.3% of patients in the olmesartan arm of this trial, respectively), and may have confounded the effect of olmesartan on the risk of severe enteropathy. This more specific risk was not assessed in this trial, which did not have sufficient statistical power to detect such an association. For the same reasons, a recent cohort study did not find any significant difference in the risk of GI disease-related hospitalisation among patients with diabetes treated by olmesartan compared with patients with diabetes treated by other ARBs.20
This study has several strengths. First, it was based on a large nationwide database. Second, we adjusted for potential confounders that may affect the outcome (hospitalisation with a discharge diagnosis of malabsorption) or the prescription of antihypertensive drugs. Finally, to prevent selection bias, we excluded those patients with malabsorption and those at risk for coeliac disease before the index date.
Several potential limitations of this study should also be discussed. First, this study was based on administrative data, which may result in information bias. There is no direct comparison between these data and chart review in France for the diagnosis of intestinal malabsorption or coeliac disease. However, the possible lack of sensitivity is unlikely to affect the three groups of the study differently; as such, it does not result in bias in the analysis, and could not refute the message of the study. Another issue raised by healthcare electronic records concerns trends in coding practice. However, in this study, no coding trend was observed for intestinal malabsorption among adult patients in France during the study period (see online supplementary table S2). Second, the potential indication bias should be discussed. However, ACEI and ARB share very similar therapeutic indications. Coeliac disease is more frequent in women and in younger subjects,21 but analyses were adjusted for age and sex. In addition, there is no reason why coeliac disease-predisposing HLA genotype would be overrepresented in patients who were prescribed olmesartan. Finally, it is unlikely that all cases of olmesartan-associated enteropathy were captured by hospital diagnoses of intestinal malabsorption and coeliac disease. It is likely that milder forms also exist. Overall number needed to harm was 31 350 patient-years of olmesartan exposure. Beyond 2 years of exposure, this number was 12 500 patient-years. However, caution is needed to interpret these values as this study was not aimed to measure the incidence of olmesartan-associated enteropathy, but rather to estimate the strength of the association between olmesartan and severe forms of enteropathy and malabsorption. As a consequence, this study underestimates the true incidence and only provides the incidence of the most severe forms of olmesartan-associated enteropathy.
In summary, this paper shows, with a higher level of evidence, the association between severe intestinal malabsorption and olmesartan exposure. These results have important practical consequences as olmesartan is widely prescribed worldwide. In France, olmesartan was prescribed to more than 800 000 patients in 2012. Patients treated with olmesartan should be informed about the risk of this complication, and should be advised to seek medical attention if they experience GI symptoms. This information should also be widely delivered to physicians of all disciplines, particularly to gastroenterologists who are faced to this new category of patients.
However, further studies are required to assess the frequency and clinical spectrum of milder forms. The pathophysiology of olmesartan-associated enteropathy also requires further investigation: the clinical and pathological features are remarkably similar to those of coeliac disease or refractory sprue, but the underlying cause and mechanisms are different. We expect such studies to shed new light on coeliac disease.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Footnotes
Contributors FC and HA had the idea for the study. MB conceived and planned the study and drafted the manuscript. MM performed data management and statistical analyses. All authors contributed to interpretation of the data and revised the manuscript. All authors approved the final manuscript.
Funding This research was funded by the French National Health Insurance Fund.
Competing interests None declared.
Ethics approval This research was authorised by the French Data Protection Agency (CNIL).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement FC had full access to all of the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis.