2522 Diabetes Care Volume 37, September 2014 Metformin Does Not Affect Cancer Konstantinos K. Tsilidis,1,2 Despoina Capothanassi,1 Naomi E. Allen,3 Risk: A Cohort Study in the U.K. Evangelos C. Rizos,4 David S. Lopez,5 Karin van Veldhoven,6–8 Clinical Practice Research Carlotta Sacerdote,7 Deborah Ashby,9 Paolo Vineis,6,7 Ioanna Tzoulaki,1,6 and Datalink Analyzed Like an John P.A. Ioannidis10 Intention-to-Treat Trial Diabetes Care 2014;37:2522–2532 | DOI: 10.2337/dc14-0584 OBJECTIVE Meta-analyses of epidemiologic studies have suggested that metformin may re- 1Department of Hygiene and Epidemiology, Uni- duce cancer incidence, but randomized controlled trials did not support this versity of Ioannina School of Medicine, Ioannina, hypothesis. Greece 2Cancer Epidemiology Unit, University of Oxford, EPIDEMIOLOGY/HEALTH SERVICES RESEARCH RESEARCH DESIGN AND METHODS Oxford, U.K. 3 A retrospective cohort study, Clinical Practice Research Datalink, was designed to Clinical Trial Service Unit, University of Oxford, Oxford, U.K. investigate the association between use of metformin compared with other anti- 4Lipid Disorders Clinic, Department of Internal diabetes medications and cancer risk by emulating an intention-to-treat analysis Medicine, University Hospital of Ioannina, Ioan- as in a trial. A total of 95,820 participants with type 2 diabetes who started taking nina, Greece 5 metformin and other oral antidiabetes medications within 12 months of their Division of Epidemiology, University of Texas School of Public Health, Houston, TX diagnosis (initiators) were followed up for first incident cancer diagnosis without 6Department of Epidemiology and Biostatistics, regard to any subsequent changes in pharmacotherapy. Cox proportional hazards School of Public Health, Imperial College London, models were used to estimate multivariable-adjusted hazard ratios (HR) and St Mary’s Campus, London, U.K. 7 95% CI. Human Genetics Foundation (HuGeF), Turin, Italy 8University College London Institute of Child RESULTS Health, Centre for Paediatric Epidemiology and A total of 51,484 individuals (54%) were metformin initiators and 18,264 (19%) Biostatistics, London, U.K. were sulfonylurea initiators, and 3,805 firstincidentcancerswerediagnosed 9Imperial Clinical Trials Unit, School of Public ’ during a median follow-up time of 5.1 years. Compared with initiators of sulfo- Health, Imperial College London, St Mary s Hos- pital, London, U.K. nylurea, initiators of metformin had a similar incidence of total cancer (HR 0.96; 10Stanford Prevention Research Center, Depart- 95% CI 0.89–1.04) and colorectal (HR 0.92; 95% CI 0.76–1.13), prostate (HR 1.02; ment of Medicine; Department of Health Re- 95% CI 0.83–1.25), lung (HR 0.85; 95% CI 0.68–1.07), or postmenopausal breast (HR search and Policy, Stanford University School of – Medicine; and Department of Statistics, Stanford 1.03; 95% CI 0.82 1.31) cancer or any other cancer. University School of Humanities and Sciences, Stanford, CA CONCLUSIONS Corresponding author: Konstantinos K. Tsilidis, In this large study, individuals with diabetes who used metformin had a similar risk [email protected], [email protected]. of developing cancer compared with those who used sulfonylureas. Received 6 March 2014 and accepted 24 April 2014. The effects of antidiabetes medications on cancer risk have recently attracted This article contains Supplementary Data online significant public interest. In particular, meta-analyses of observational studies at http://care.diabetesjournals.org/lookup/ suppl/doi:10.2337/dc14-0584/-/DC1. have found that the biguanide metformin, which is an insulin sensitizer and the fi © 2014 by the American Diabetes Association. most commonly used rst-line therapy for type 2 diabetes, may reduce cancer Readers may use this article as long as the work incidence (1,2). However, a meta-analysis of randomized controlled trials did not is properly cited, the use is educational and not support this hypothesis (3). for profit, and the work is not altered. care.diabetesjournals.org Tsilidis and Associates 2523 Many of the published pharmacoepi- medical records, with 95% of neoplasms (acarbose), meglitinide analogs demiologic studies are difficult to inter- and 88% of endocrine and metabolic dis- (nateglinide, repaglinide), GLP-1 pret, however, because they may be orders identified in the CRPD confirmed receptor antagonists (exenatide), subject to several biases (4). The pro- with alternate data sources (8,9). dipeptidyl peptidase 4 inhibitors gressive nature of type 2 diabetes re- (sitagliptin, vildagliptin); quires changes in pharmacotherapy Study Population 5. monotherapy with insulin; We extracted data on all participants over time and makes very difficult the 6. combination therapies with metfor- with type 2 diabetes aged 35 to 90 years assessment of the independent associa- min if the participants were simulta- who were prescribed at least one anti- tion of a specific medication and cancer neously exposed to two or more diabetes agent between 1 January 1987 risk. The lack of randomization makes OHAs including metformin during and 31 December 2010, provided that observational effect estimates vulnera- the initial 12-month treatment the first prescription was dated at least ble to confounding by indication due to period; 6 months after the CPRD registration the different prognoses of individuals 7. combination therapies with insulin; date of the participant. This was to en- between treatment groups. Moreover, and sure that most of the included partici- most randomized clinical trials are not 8. other combination therapies. pants with diabetes would be new users designed or sufficiently powered to ex- of antidiabetes drugs and to exclude amine cancer outcomes due to short The main analysis was based on a com- prevalent users with unknown type follow-up periods and very few cancer parison of those who started using met- and duration of treatment. In addition, events. formin compared with those who used eligible participants had never a history To overcome these limitations, we sulfonylureas because these two medi- of cancer. We also excluded the initial emulated the design and analysis of a cations were the most common and be- 12 months of follow-up after the first trial in a large retrospective cohort study cause sulfonylureas had similar clinical antidiabetes prescription, because any within the U.K. Clinical Practice Re- indications to metformin. cancer diagnoses occurring within that search Datalink (CPRD), one of the time are unlikely to be attributable to Outcome Assessment world’s largest electronic medical re- the medications. Of the initial 113,301 First incident cancer cases were defined cord databases, to investigate asso- participants with type 2 diabetes identi- by the presence of National Health Ser- ciations of cancer risk among users fied, 95,820 were available for statistical vice Read codes in the CPRD. The Read ofmetformincomparedwithusersof analysis (Fig. 1). code dictionary was first searched to sulfonylurea or other first-line oral hy- identify malignant neoplasms, excluding poglycemic agents (OHAs) among indi- Exposure Assessment nonmelanoma skin cancers, using sev- viduals with newly diagnosed diabetes. Exposure to antidiabetes medications eral sensitive and specific algorithms This “incident diabetes drug” cohort de- was assessed by the presence of antidia- and code lists. Two researchers (K.K.T. sign and intention-to-treat (ITT) analysis betes prescription records in the CPRD and D.C.) manually reviewed and con- can be regarded as the equivalent of a using British National Formulary codes. firmed the codes identified by the initial nonrandomized “trial” that avoids many Treatment-naive participants with dia- searches and excluded codes for border- of the biases of traditional observational betes at CPRD enrollment who first line, in situ, or suspected malignancies. studies (5). started using metformin (metformin ini- A similar process was followed to define tiators) were compared with initiators type 2 diabetes. RESEARCH DESIGN AND METHODS of other OHAs. Exposure to certain anti- Data Source diabetes drugs was classified based on Covariate Assessment This study was conducted using the U.K. the initial 12-month treatment period, Information on sociodemographic data, CPRD, previously known as General and an individual’s treatment pattern lifestyle characteristics, anthropometric Practice Research Database (6), and was categorized into one of the follow- variables, medical conditions, and treat- was approved by the Medicines and ing mutually exclusive groups: ments was extracted from the CPRD at Healthcare Products Regulatory Agency approximately the time of the first anti- Independent Scientific Advisory Com- 1. monotherapy with metformin, if diabetes prescription (index date). Age mittee. The CPRD was established in they were only exposed to metfor- and sex were recorded at the time of 1987 and currently encompasses more min during their initial 12-month the first prescription. BMI measure- than 5 million people enrolled from over treatment period; ments and smoking status (never, for- 600 general practitioners nationwide. 2. monotherapy with sulfonylureas, mer, current) were retrieved for the The people enrolled in the CPRD are rep- either first generation (tolbuta- 1 year before the index date. Alcohol resentative of the U.K. population with mide, chlorpropamide, tolazamide, consumption (never, former, current) regard to age, sex, and geographical dis- acetohexamide) or second genera- was defined from
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages11 Page
-
File Size-