DIAB-6314; No. of Pages 7 diabetes research and clinical practice xxx (2015) xxx–xxx

Contents available at ScienceDirect

Diabetes Research and Clinical Practice journ al h ome pa ge : www .elsevier.co m/lo cate/diabres

Incretin-based drugs and risk of acute pancreatitis: A nested-case control study within a healthcare database Davide Soranna a, Cristina Bosetti b, Manuela Casula c, Elena Tragni c, Alberico L. Catapano c,d, Carlo L.A. Vecchia e, Luca Merlino f, Giovanni Corrao a,* a

Dipartimento di Statistica e Metodi Quantitativi, Sezione di Biostatistica, Epidemiologia e Sanita` Pubblica, Universita` Milano-Bicocca, Milan, Italy b Dipartimento di Epidemiologia, IRCCS-Istituto di Ricerche Farmacologiche ‘‘Mario Negri’’, Milan, Italy c Dipartimento di Scienze Farmacologiche e Biomolecolari, Universita` di Milano, Milan, Italy d IRCSS Multimedica, Milan, Italy e Dipartimento di Scienze Cliniche e di Comunita`, Universita` Milano, Milan, Italy f Unita` Organizzativa Governo dei dati, delle strategie e piani del sistema sanitario, Regione Lombardia, Milan, Italy

article info

abstract

Article history:

To assess the association between use of incretin-based drugs for diabetes mellitus and the

Received 29 September 2014

occurrence of acute pancreatitis.

Received in revised form

A population-based, nested case-control study was performed within a cohort of 166,591

13 January 2015

patients from the Lombardy region (Italy) aged 40 years or older who were newly treated

Accepted 13 February 2015

with oral antihyperglycaemic agents between 2004 and 2007. Cases were 666 patients who

Available online xxx

experienced acute pancreatitis from April 1, 2008 until December 31, 2012. For each case

Keywords:

age at cohort entry, and date of index prescription. Conditional logistic regression was used

Acute pancreatitis

to model the risk of acute pancreatitis associated with use of incretin-based drugs within 30

patient, up to 20 controls were randomly selected from the cohort and matched on gender,

Antihyperglycaemic agents

days before hospitalization, after adjustment for several risk factors, including the use of

Drug safety

other antihyperglycaemic agents. Sensitivity analyses were performed in order to account

Incretin-based drugs

for possible sources of systematic uncertainty.

Healthcare databases Nested case-control study

Use of incretin-based drugs within 30 days was reported by 17 (2.6%) cases of acute pancreatitis versus 193 (1.5%) controls. The corresponding multivariate odds ratio was 1.75 (95% confidence interval, 1.02 to 2.99). Slightly lower and no significant excess risks were observed by shortening (15 days) and increasing (60 and 90 days) the time-window at risk. This study supports a possible increased risk of acute pancreatitis in relation to use of incretin-baseddrugs reported in a few previous studies. However, given the potential for bias and the inconsistency with other studies, additional investigations are needed to clarify the safety of incretin-based-drugs. # 2015 Elsevier Ireland Ltd. All rights reserved.

* Corresponding author. Tel.: +39 02 64485854; fax: +39 02 64485899. E-mail address: [email protected] (G. Corrao). http://dx.doi.org/10.1016/j.diabres.2015.02.013 0168-8227/# 2015 Elsevier Ireland Ltd. All rights reserved.

Please cite this article in press as: Soranna D, et al. Incretin-based drugs and risk of acute pancreatitis: A nested-case control study within a healthcare database. Diabetes Res Clin Pract (2015), http://dx.doi.org/10.1016/j.diabres.2015.02.013

DIAB-6314; No. of Pages 7

2

diabetes research and clinical practice xxx (2015) xxx–xxx

1.

Introduction

Glucagon-like peptide-1 (GLP-1) receptor agonists and dipeptidyl peptidase-4 (DPP-4) inhibitors are two classes of incretinbased treatments for type 2 diabetes mellitus [1]. They act either by mimicking the effects of GLP-1 (e.g., exenatide and liraglutide) or by inhibiting the enzyme DPP-4 that degrades endogenous GLP-1 (e.g., sitagliptin and linagliptin). These agents are effective in glycaemic control, do not increase weight [2], and may reduce major cardiovascular events [3,4]. However, safety concerns have arisen because of possible pancreatic adverse events [5,6]. In particular, several cases of acute pancreatitis have been reported following the use of exenatide, sitagliptin, and other incretin-based drugs [7–9]. A recent case-control study nested within a large US claim database also reported an almost 2-fold increased risk of acute pancreatitis in users of exenatide and sitagliptin [10]. However, other observational studies based on insurance claim or health databases did not confirm such association [11–17], in agreement with meta- and pooled-analyses of clinical trials (mainly sitagliptin) [18–20]. A recent meta-analysis of randomized and non-randomized studies also suggested that incretin-based drugs do not increase the risk of pancreatitis [21]. Due to methodological weaknesses, however, additional carefully designed and conducted observational studies are warranted as also suggested by the FDA and EMA [22]. To further investigate this issue, we have thus analysed the data from the healthcare databases of the Lombardy Region (Italy), providing information on a large unselected population representative of the real clinical practice.

2.

Methods

Data were retrieved from the healthcare utilization databases of the Italian Lombardy Region. In Italy, the population is covered by a National Health Service (NHS), and Lombardy provides an automated system of databases to collect a variety of information. Full details of the databases and the merging procedure have been reported elsewhere [23].

2.1.

Study population

The target population consists of all beneficiaries of the NHS, resident in Lombardy and aged 40 years or more. We identified those who received at least one prescription of drug for diabetes treatment (ATC code: A10) from January 1, 2004, until December 31, 2007 and we defined the first prescription as the index prescription. Because incretin-based drugs were registered in Italy at the end of 2007, the period of recruitment ensured that patients had not started antihyperglycaemic therapy with an incretin-based drug. Since in fact incretinbased drugs are usually prescribed to type 2 diabetic patients as a second choice antidiabetic, patients who start with incretin-based drugs may differ from those who switch from other antidiabetics in clinical and other features which we are not able to measure. Patients were excluded from data analysis if they: (i) had received any antihyperglycaemic agent and/or were hospitalized for diabetes within the four

years before the index prescription, to ensure inclusion of only newly treated individuals; (ii) were hospitalized for any pancreatic disease within four years before the index prescription, to ensure inclusion of only incident acute pancreatitis during follow-up; and/or (iii) were hospitalized for acute pancreatitis before April 1, 2008, to ensure a sufficient time span of potential exposure to the drugs of interest. Each member of the cohort accumulated person-months of followup from the date of the index prescription until the earliest among the dates of hospitalization for acute pancreatitis (see below), death, migration, or end of follow-up (December 31, 2012).

2.2.

Selection of cases and controls

A case-control study was nested within the cohort of incident antihyperglycaemic users. Nested case-control design is an efficient alternative to cohort design when the effect of timedependent exposure on rare events needs to be investigated using large databases [24,25]. Cases were members of the cohort who were hospitalized for acute pancreatitis (ICD-9 code 577.0). For each case patient, up to 20 controls at risk for the outcome at the time when the matched case had the event were randomly selected from the cohort and matched for gender, age at cohort entry, and date of index prescription. In this way, every set constituted by the index case and corresponding controls had the same period of observation.

2.3.

Exposure variable and covariates

We identified all prescriptions of antihyperglycaemic drugs, including incretin-based drugs (ATC codes: A10BD07, A10BH01, A10BH02, A10BH03, A10BX04, A10BX07), biguanides (A10BA), sulphonylurea (A10BB and A10BC), other oral antihyperglycaemic agents (A10BD, A10BF, A10BG, A10BH, A10BX, with exclusion of A10BX04, A10BX07, A10BH01, A10BD07, A10BH02, A10BH03), and insulin (A10A) dispensed to each case-controls set during the corresponding time-window of observation. Case patients and controls were classified according whether they were prescribed incretin-based drugs any time in the 30 days period before the date of hospital admission of the index case of each case-control set (time-window at risk). With the aim of avoiding the arbitrary nature of this timewindow, shorter (15 days) and longer (60 and 90 days) timewindows of exposure were considered in a sensitivity analysis. Information additionally considered were: (1) current exposure to other oral antihyperglycaemic, agents and/or to insulin in the 30 days before the date of hospital admission; (2) recent use (i.e., in the six months before the event date) of drugs known or suspected of increasing the risk of acute pancreatitis [26,27], such as ACE-inhibitors (C09A, C09B), angiotensin receptors blockers (C09C, C09D), furosemide (C03CA01, C03CB01, C03EB01), statins (C10AA), fibrates (C10AB), and valproic acid (N03AG01); (3) previous use (i.e., in the four years prior the date of the index prescription) of cardiovascular drugs (C) and antidepressants (N06A); (4) hospital discharge for gallstones, including diagnosis of either cholelithiasis (ICD-9-CM 574, 575.2) or cholangiography (51.10–51.11) or cholecystectomy (51.2), hypertriglyceridemia

Please cite this article in press as: Soranna D, et al. Incretin-based drugs and risk of acute pancreatitis: A nested-case control study within a healthcare database. Diabetes Res Clin Pract (2015), http://dx.doi.org/10.1016/j.diabres.2015.02.013

DIAB-6314; No. of Pages 7 diabetes research and clinical practice xxx (2015) xxx–xxx

(272), and any disease of the circulatory system (390–459) in the four years prior the date of the index prescription; and (5) the Charlson comorbidity index score [28] calculated using diagnostic information available from inpatient charts in the four years prior the date of the index prescription.

2.4.

Statistical analysis

Chi-square test, or its version for trend, was used to test differences between cases and controls for various variables of interest. Conditional logistic regression models were fitted to estimate the odds ratio (OR), and its 95% confidence interval (CI), of acute pancreatitis in relation to exposure to incretinbased drugs including the above listed covariates. The robustness of our estimates with regard to potential bias introduced by unmeasured confounders was investigated by using the rule-out approach described by Schneeweiss [29]. In applying this method, we allowed the possible unmeasured confounder: (i) to be present in the study population (i.e., the Italian patients with type 2 diabetes mellitus) with a prevalence of 10%, 20% or 40%; (ii) to be associated with the outcome with risk ratio varying from 1 to 10 (i.e., exposed patients to the confounder may have experienced hospitalization for acute pancreatitis up to 10-fold more than unexposed); (iii) to be associated with the exposure of interest with OR varying from 1 to 10 (i.e., current users of incretin-based drugs may be exposed to the confounder up to 10-fold more than patients who do not use incretin-based drugs). Rule-out approach aims to detect the extension of the confounding required to fully account for the observed exposure-outcome association, thus moving the observed point estimate to the null. All analyses were performed using the Statistical Analysis System Software (version 9.2; SAS Institute, Cary, NC, USA). Statistical significance was set at the 0.05 level. All p-values were two-sided.

3.

Results

Fig. 1 shows that among the 482,362 patients aged 40 years who had used antihyperglycaemic agents between 2004 and 2007, 315,771 were excluded because were prevalent users or

3

experienced hospitalization for diabetes or pancreatitis. The remaining 166,591 patients included into the cohort accumulated 1117,560 person-years of observation (on average 6.7 years of follow-up per patient) and generated 666 hospital admissions for acute pancreatitis, with an incidence rate of 6 cases per 10,000 person-year. The 666 case patients were matched to 12,946 controls (43 cases were matched with less than 20 controls). As shown in Table 1, mean age of cases and controls was 66 years (SD 11 years) and 55% of them were men (matching variables). Compared to controls, a higher proportion of case patients received incretin-based drugs within 30 days before the date of outcome onset (17, 2.55% vs. 193, 1.49%), while there was no evidence that cases and controls differed for exposure to other antihyperglycaemic agents during this 30day period. There was also evidence that cases more often used furosemide and fibrates and experienced hospitalizations for gallstones and cardiovascular disease than controls. The association between exposure to incretin-based drugs during various time-windows and the risk of acute pancreatitis is shown in Fig. 2. There was evidence that the risk of acute pancreatitis was increased among patients using incretinbased drugs within 30 days before event date (OR = 1.75, 95% CI 1.02–2.99), while weaker and no significant risk excesses were observed for shorter (OR = 1.60 for 15 days) and longer (OR = 1.63 and 1.52 for 60 and 120 days, respectively) timewindows. The results of possible residual confounding analysis obtained by means of the rule-out approach are presented in Fig. 3. For example, assuming a prevalence of 40% for the exposure to a hypothetical unmeasured factor in our target population, and that patients who currently use incretinbased drugs have a 6-fold higher odds of exposure to that factor than those who do not use them, at least a 8-fold excess risk of acute pancreatitis among patients who are exposed to that factor with respect to those who are not exposed to it would be required to nullify the observed association between current use of incretin-based drugs and the risk of acute pancreatitis. Weaker confounding-outcome associations are required for confounding factors with a lower prevalence, but a risk ratio of 4 is required for the lowest investigated prevalence (10%).

Fig. 1 – Flow-chart of subjects’ inclusion and exclusion criteria. Please cite this article in press as: Soranna D, et al. Incretin-based drugs and risk of acute pancreatitis: A nested-case control study within a healthcare database. Diabetes Res Clin Pract (2015), http://dx.doi.org/10.1016/j.diabres.2015.02.013

DIAB-6314; No. of Pages 7

4

diabetes research and clinical practice xxx (2015) xxx–xxx

Fig. 2 – Effect of current use of incretin-based drugs on the risk of acute pancreatitis according with the length of timewindow for current use, i.e., 30 days (main criterion) and 15, 60 and 120 days (secondary criteria).1

4.

Discussion

This study of new users antihyperglycaemic of drugs indicates that patients who had at least one prescription of incretinbased drugs within 30 days before the outcome onset had a

Table 1 – Selected characteristics of the 666 case patients hospitalized for acute pancreatitis and of the corresponding 12,946 controls included into the study. Case patients Age at cohort entry: mean (SD) Male gender Current use of antidiabetic drugsy Incretin-based drugs Biguanides Sulphonylurea Others oral antidiabetics Insulin

Controls

p-Value*

66 (11)

66 (11)

MV

368 (55%)

7198 (55%)

MV

17 154 106 102

193 3321 2040 1711

0.030 0.144 0.913 0.120

(2%) (23%) (16%) (15%)

(1%) (26%) (16%) (13%)

32 (5%)

575 (4%)

0.658

140 (21%) 121 (18%)

2460 (19%) 2021 (16%)

0.196 0.077

157 197 28 40

2151 4250 291 647