Braiddm.ca

Metformin and risk of cardiovascular events J. A. Johnson et al Reduced cardiovascular morbidity and mortality
associated with metformin use in subjects with Type 2
diabetes
J. A. Johnson*†, S. H. Simpson*, E. L. Toth*‡ and S. R. Majumdar*§ Abstract
*Institute of Health Economics, Edmonton, Alberta, Metformin therapy reduces microvascular complications in Type 2 dia- †Department of Public Health Sciences, ‡Division of betes; questions remain, however, regarding its impact on macrovascular events.
Endocrinology & Metabolism, Department of This study examined metformin use in relation to risk of cardiovascular-related Medicine and §Division of General Internal Medicine, Department of Medicine, University of We conducted a retrospective cohort analysis, using Saskatchewan Health administrative databases to identify new users of oral antidiabetic drugs.
Subject groups were defined by medication use during 1991–1999: sulphonylu-rea monotherapy, metformin monotherapy, or combination therapy. Deathsand non-fatal hospitalizations recorded during the study period were identifiedas cardiovascular-related from ICD-9 codes. The main outcome was a compos-ite of first non-fatal hospitalization or death. Standard multivariate techniques,including propensity scores, were used to adjust for potential confounding.
Multivariate Cox proportional hazard models were used to examine the rela-tionship between metformin use and the composite endpoint.
Metformin monotherapy was associated with a lower risk of the com- posite endpoint (adjusted hazard ratio 0.81; 95% confidence interval 0.68,0.97) compared with sulphonylurea monotherapy. Combination therapy withmeformin and a sulphonylurea was associated with lower mortality, but hadsimilar hospitalization rates, to sulphonylurea monotherapy.
Conclusions
Metformin monotherapy was associated with a lower risk of cardiovascular-related morbidity and mortality, and combination metformin andsulphonylurea therapy was associated with a reduced risk of fatal cardiovascu-lar events, when compared with sulphonylurea monotherapy.
Keywords
diabetes mellitus, metformin, morbidity and mortality, propensity score common condition in Type 2 diabetes [2]. Since hyperinsulinaemia Introduction
may aggravate other metabolic disorders, including hyper- Metformin, as monotherapy or in combination with sulpho- lipidaemia [2] and endothelial dysfunction [3], metformin may nylureas, improves glycaemic control in people with Type 2 provide protection from cardiovascular disease compared with diabetes mellitus [1]. It appears that metformin improves insulin sulphonylurea therapy, which acts by increasing insulin levels.
sensitivity, which in turn may reduce hyperinsulinaemia—a The effect of metformin on clinical outcomes has been evaluated in the United Kingdom Prospective Diabetes Study(UKPDS) [4] and in retrospective observational studies [5–7].
The observed association between metformin and macrovas- Correspondence to: Dr Jeffrey A. Johnson, #1200, 10405 Jasper Ave, Edmonton, AB, Canada T5J 3 N4. e-mail: jeff.johnson@ualberta.ca cular morbidity and mortality continues to be a topic of much 2005 Diabetes UK. Diabetic Medicine, 22, 497– 502
Metformin and risk of cardiovascular events • J. A. Johnson et al controversy [8,9]. The UKPDS demonstrated that metformin Subjects and methods
use in obese patients with Type 2 diabetes reduced all-causemortality and diabetes-related endpoints [4]. However, in Subjects
subjects failing after an average of 7 years of sulphonylureamonotherapy, addition of metformin was associated with an The Saskatchewan Prescription Drug Plan database was used to increased risk of all-cause and diabetes-related mortality com- identify new users of oral antidiabetic drugs. The adminis- pared with continuation of sulphonylurea monotherapy [4].
trative healthcare databases from Saskatchewan were used An increased risk of mortality with combination metformin because they provide comprehensive healthcare information and sulphonylurea therapy was also reported in several sub- for all registered beneficiaries in the province [13]. Essentially sequent observational studies [5–7]. These studies were rela- all (> 90%) residents of the province are beneficiaries of tively small, however, and limited by the inability to control Saskatchewan Health; exceptions include registered Indians, federal inmates and employees of the military and Royal Cana-dian Mounted Police, whose prescription benefits are adminis- We recently reported a population-based analysis demon- tered by federal programmes. Saskatchewan residents aged strating that metformin, either alone or in combination with 30 years or older were eligible for inclusion if they had continu- sulphonylurea, was associated with reduced all-cause and ous prescription drug coverage in Saskatchewan for at least cardiovascular mortality compared with sulphonylurea mono- 1 year prior to the index date (i.e. date of the first claim for an therapy [10]. We evaluated a cohort of new users of oral anti- oral antidiabetic drug in the index period). New users of oral diabetic drugs and therefore were able to control for duration antidiabetic drugs were identified as subjects with a prescrip- of therapy for diabetes. Several questions remained unan- tion claim for a sulphonylurea or metformin during the index swered, however. For example, were observations limited by period of January 1991 to December 1996, and no prescription selection bias (i.e. ‘sicker’ patients prescribed sulphonylureas, claims for antidiabetic drugs within the previous year. Subjects a form of confounding by indication), and do the benefits were excluded if they had any record of insulin use or if they of metformin therapy extend to non-fatal cardiovascular had less than 1 year between the first and last dispensationrecords for oral antidiabetic drugs. Exposure status was further events. In the absence of randomized controlled trials, refined by excluding subjects who did not obtain sufficient observational studies provide an alternative method for medication to provide the minimum recommended daily resolving clinically important questions [11,12]. The pur- dose during one or more 6-month interval of their follow-up pose of this study was to extend our mortality analyses by evaluating the association between metformin and rates of Subjects included in the study were grouped according to fatal and non-fatal cardiovascular events for those with Type their oral antidiabetic drug use recorded in the Saskatchewan Prescription Drug Plan database. Subjects who received either Figure 1 Flowchart of subject selection.
*Minimum recommended daily doses:
acetohexamide 250 mg; chlorpropamide
100 mg; glibenclamide 2.5 mg; tolbutamide
500 mg; tolazamide 100 mg; metformin
250 mg.
2005 Diabetes UK. Diabetic Medicine, 22, 497– 502
Table 1 International Classification of Diseases 9th Revision (ICD-9)
would have a minimum CDS of 2 because they had to be using oral antidiabetic drugs; use of lipid-lowering agents contributes an additional score of 1 and antihypertensives contribute a score of 2 (if an antihypertensive other than ACE inhibitor) or1 (for β-blockers and diuretics) [16]. Interaction terms between each variable and treatment group were also examined; how- ever, these terms were not included in the final models because none reached statistical significance (P < 0.1).
Survival curves for the time to first non-fatal hospitalization or death were generated using the corrected group prognosis method [17,18]. In this method, the first step is to estimate sur- vival curves for each unique combination of covariates. The final survival curve is the weighted average of all these indi- vidual curves, with the weighting based on number of individualsat each level of covariate [18]. To fit the data into the correctedgroup prognosis model, continuous variables (i.e. age, chronicdisease score) were collapsed to dichotomous variables.
sulphonylureas or metformin exclusively during follow-up Data used for this observational study reflect actual practice were categorized as sulphonylurea monotherapy or metformin in Saskatchewan during 1991–1999 in that the prescription for monotherapy, respectively. Subjects were categorized in the sulphonylureas and metformin was made at the discretion of combination therapy group if they received both metformin the clinician, rather than by randomization. Therefore, we were and sulphonylureas and if these drugs were used concurrently interested in adjusting for potential selection bias because for at least 50% of their follow-up. All study subjects were fol- group assignment in this study was based on medication use lowed prospectively from index date until death, study exit (i.e.
[19]. Propensity scores, intended to represent the likelihood of moved out of Saskatchewan), or 31 December 1999.
receiving one of the treatments given the individual’s charac-teristics, were calculated to control for potential differences Outcomes
between groups [20–22]. Separate propensity scores were gen-erated, using logistic regression, for metformin monotherapy The primary outcome was the composite endpoint of fatal vs. sulphonylurea monotherapy and combination therapy vs.
or non-fatal cardiovascular-related events (Table 1). Cause of sulphonylurea monotherapy. Propensity scores were generated death was ascertained through the computerized vital statistics using age, age2, sex, CDS, CDS2, current use of nitrates, as well file of Saskatchewan Health. Non-fatal events were defined as as interaction terms among all variables. The generated propens- any cardiovascular-related hospitalization in which the patient ity scores were then included as an additional covariate in the was discharged alive. Hospitalizations were classified as cardiovascular-related based on the primary diagnosis code(ICD-9) from hospital separation records, as it is the diagnosismost responsible for length of stay and resource utilization.
As a composite endpoint, only the first non-fatal or fatal A total of 12 188 Saskatchewan residents were identified as event was counted for each individual. The fatal and non-fatal new users of oral antidiabetic drugs during the index period event rates were analysed separately as secondary outcomes.
(Fig. 1). Of these, 1443 were excluded because they were dis-pensed insulin during their follow-up and 2009 had less than Statistical analysis
1 year of dispensation records for oral antidiabetic drugs. Afurther 3034 subjects were excluded because their average Comparisons between groups were evaluated using univariate daily supply of oral antidiabetic drug was below the minimum ANOVA for continuous variables and χ for categorical variables.
Cox proportional hazard models were used to evaluate the re- recommended daily dose. Of the 2641 subjects identified as lationship between metformin use (either as monotherapy or in taking both sulphonylurea and metformin, only 1081 sub- combination with sulphonylureas) and time to first event. In all jects used both drugs concurrently for at least 50% of their models, the sulphonylurea monotherapy group served as the follow-up. There were 2138 subjects in the sulphonylurea monotherapy group and 923 subjects in the metformin The following potential confounding variables were included monotherapy group. Table 2 provides demographic informa- in the adjusted analyses: age, sex, nitrate use, and a Chronic tion for all 4142 subjects included in the analysis.
Disease Score (CDS). Nitrate use was included because it is a A total of 381 subjects died from cardiovascular causes and marker for established coronary artery disease [14,15]. The 715 were hospitalized at least once for cardiovascular reasons CDS uses pharmacy dispensation information for specific drug during the follow-up period (Table 3). Subjects in the metformin classes to estimate a burden of comorbidities, and has provedvalid in predicting hospitalization and mortality [16]. The CDS monotherapy group had the lowest non-fatal hospitalization is the sum of all chronic diseases identified from drug therapy rate for cardiovascular causes (53.7 hospitalizations per 1000 during an observation period, including incremental therapies person-years) compared with sulphonylurea monotherapy for cardiovascular risk. For example, all subjects in this study users (75.3 per 1000 person-years; P < 0.05) and compared 2005 Diabetes UK. Diabetic Medicine, 22, 497– 502
Metformin and risk of cardiovascular events • J. A. Johnson et al Table 2 Subject demographics
*ANOVA (P < 0.01).
†χ2 (P < 0.01).
Figure 2 Adjusted (for age, sex, chronic disease score, nitrate use)
with combination therapy users (90.2 per 1000 person-years; survival curves illustrating the relationship between metformin use and P < 0.05). Non-fatal cardiovascular-related hospitalization time to first non-fatal hospitalization or death.
rates were similar for sulphonylurea monotherapy users andcombination therapy users (P = 0.08).
hospitalizations were similar for sulphonylurea monotherapy In the unadjusted analyses (Table 3), metformin, alone or in and combination therapy groups (P = 0.32).
combination with sulphonylurea, was associated with a lower Following adjustment for patient and treatment factors, risk of the composite endpoints [hazard ratio (HR) 0.67; 95% metformin monotherapy continued to be associated with a confidence interval (CI) 0.56, 0.80 and HR 0.80; 95% CI 0.69, lower risk of the composite endpoint (Fig. 2) as well as fatal 0.93, respectively]. Metformin, either alone or in combination, and non-fatal cardiovascular events. Metformin in combina- was also associated with a lower risk of mortality from cardio- tion with sulphonylurea continued to be associated with a vascular causes. Metformin monotherapy was associated lower risk of mortality from cardiovascular causes. Addition with a lower risk of non-fatal cardiovascular hospitalizations of a propensity score to the models did not make a substantial (HR 0.75; 95% CI 0.61, 0.93); event rates for cardiovascular change to observations from the adjusted model ( Table 3).
Table 3 Cardiovascular-related event rates
Non-fatal cardiovascular hospitalization‡ *Adjusted for age, sex, chronic disease score, nitrate use.
†Sulphonylurea monotherapy group served as reference group.
‡The first non-fatal cardiovascular hospitalization was used in the hazard ratio for subjects with multiple non-fatal hospitalizations.
§The first non-fatal cardiovascular hospitalization was used for subjects with both a non-fatal hospitalization and fatal event (i.e. these subjects were not double counted).
2005 Diabetes UK. Diabetic Medicine, 22, 497– 502
coronary artery disease. Additionally, metformin is likely to Discussion
be preferentially used in obese patients [2,4], who have an Randomized clinical trials have demonstrated the benefits increased cardiovascular risk. If this were the case, we would of metformin therapy for lowering elevated blood glucose have expected an increased event rate in the metformin group; and reducing the risk of microvascular complications [1,2,4].
therefore, had we been able to control for BMI, it is possible However, cardiovascular disease is the leading cause of mor- that the risk reduction would have been even greater. Further- bidity and mortality associated with Type 2 diabetes [23–26] more, administrative data of drug dispensations do not assure and the impact of oral antidiabetic therapy on macrovascular the exact level of drug consumption and therefore exposure.
complications and cardiovascular disease has been the subject We restricted our current analyses to individuals with evidence of much discussion [4 – 8]. We previously reported reductions of ongoing use of medication above minimum recommended in mortality associated with metformin use, either alone or in dosages, which provides a reasonable level of confidence in the combination with sulphonylureas using a retrospective cohort design, and administrative data from Saskatchewan Health Additionally, in our analyses, we identified new users of [10]. After controlling for age, sex, comorbidities, and oral antidiabetic agents; we could not, therefore, control for nitrate use (a proxy for established coronary artery disease), we duration of diabetes not treated with oral agents prior to our observed statistically significant and clinically important identification of cases. Nonetheless, we have no reason to reductions in all-cause and cardiovascular-related mortality.
believe that this duration would be different for individuals Using the same cohort as our previous study, but further treated initially with sulphonylureas vs. metformin, and thus, refining our drug exposure definitions and including a pro- it is unlikely to affect our risk estimates. Finally, we ascertained pensity score to minimize the role of unmeasured (residual) non-fatal outcomes as those resulting in hospitalization. We confounding, the results of this study are consistent with our recognize that this probably underestimates the rate of all non- previous report, and extend the observations to include non- fatal cardiovascular events by excluding milder cardiovascular fatal events. In multivariate analyses, the composite endpoint events. Again, however, there is no reason to expect that this of cardiovascular-related non-fatal hospitalizations and underestimation would be systematically different between the mortality occurred less frequently in metformin monotherapy cohorts under study, and therefore, it is unlikely to bias our users compared with sulphonylurea monotherapy users.
Combination sulphonylurea and metformin therapy users The association between hyperinsulinaemia, hyperlipidae- appeared to have a reduced risk of cardiovascular-related mia [2], endothelial dysfunction [3], and cardiovascular non-fatal and fatal events compared with sulphonylurea morbidity and mortality is well recognized [23 –26]. Pharma- cological therapy that increases insulin sensitivity should The observed event rates in the current analysis may also have a beneficial effect on cardiovascular outcomes in Type 2 address, to some extent, concerns regarding confounding diabetes in addition to its beneficial effects on diabetes out- by indication in our previous analysis [10]. For example, comes. In the present study, subjects receiving metformin sulphonylureas may have been prescribed for subjects with monotherapy were less likely to be admitted to hospital for or ‘advanced’ disease and therefore at higher risk of cardiovascular- die from cardiovascular events compared with sulphonylurea related hospitalization or death. Following this hypothesis, monotherapy users. Subjects receiving combination therapy subjects receiving combination therapy should be even ‘sicker’, were also less likely to die from cardiovascular causes, but had yet we observed a lower mortality rate than sulphonylurea a similar rate of non-fatal cardiovascular events. This finding monotherapy users, despite the former having the highest hos- would be consistent with the possibility that metformin therapy pitalization rate of our three study groups. Furthermore, it is may reduce the extent or severity of cardiovascular disease [2].
well recognized that glycaemic control in patients with Type 2 Our observations refute previous studies suggesting that the diabetes progressively worsens over time [27], and it is unlikely combination of metformin and sulphonylurea is associated that continued monotherapy provides sustained glycaemic with increased risk of morbidity and mortality in Type 2 dia- control [28]. Thus, our data suggest that even in the face of betes. Indeed, our results indicate that metformin is associated progressively deteriorating glycaemic control, metformin use with reduced cardiovascular morbidity and mortality in was associated with a reduction in fatal and non-fatal cardio- people with Type 2 diabetes. Until randomized trials are con- ducted to explore these issues further, these observational data As with other studies based solely on administrative data- provide the most valid information available for clinical bases, there are several limitations inherent in the available decision-making, and support the long-term safety and effect- data that must be recognized. Because our analysis was based iveness of metformin, alone or in combination, in patients entirely on administrative databases, and is limited by a lack of direct clinical information, we could not control for level ofglycaemic control, body mass index (BMI), or other modifi- Competing interests
able cardiovascular risk factors (e.g. smoking). However, wecontrolled for age, gender, comorbidity, and the presence of 2005 Diabetes UK. Diabetic Medicine, 22, 497– 502
Metformin and risk of cardiovascular events • J. A. Johnson et al Acknowledgements
9 Sasali A, Leahy JL. Is metformin cardioprotective? Diabetes Care 2003; 26: 243–244.
This study was funded in part by grants from the Alberta 10 Johnson JA, Majumdar SR, Simpson SH, Toth EL. Decreased mor- Heritage Foundation for Medical Research (AHFMR) and the tality associated with the use of metformin compared with sulfonylurea Institute of Health Economics. J.A.J. and S.R.M. hold Popula- monotherapy in type 2 diabetes. Diabetes Care 2002; 25: 2244 –2248.
11 Hornberger J, Wrone E. When to base clinical policies on observa- tion Health Investigator Awards through the Alberta Heritage tional versus randomized trial data. Ann Intern Med 1997; 127: 697–
Foundation for Medical Research. S.R.M. is a New Investig- ator with the Canadian Institutes for Health Research (CIHR).
12 Concato J, Shah N, Horwitz RI. Randomized, controlled trials, J.A.J. holds a Canada Research Chair in Diabetes Health observational studies, and the hierarchy of research designs. N Engl Outcomes. J.A.J. is a New Emerging Team (NET) leader and J Med 2000; 342: 1887–1892.
13 Downey W, Beck P, McNutt M, Stang M, Osei W, Nichol J. Health is supported by the Alliance for Canadian Health Outcomes Databases in Saskatchewan. In Strom B ed. Pharmacoepidemiology. Research in Diabetes (ACHORD) at the Institute of Health Chinchester: Wiley, 2000: 325–345.
Economics. The ACHORD NET grant is sponsored by the 14 Tsuyuki RT, Olson KL, Teo KK. Nitroglycerin as an indicator for the Canadian Diabetes Association, the Heart and Stroke Founda- presence of coronary heart disease. Can Pharmaceut J 1999; 132: 32–
tion of Canada, The Kidney Foundation of Canada, the 15 Connolly P, Cupples ME, Cuene-Grandidier H, Johnston D, CIHR—Institute of Nutrition, Metabolism and Diabetes and Passmore P. The importance of validating the diagnosis of coronary the CIHR—Institute of Circulatory and Respiratory Health.
heart disease when measuring secondary prevention: a cross-sectional The authors thank MaryRose Stang, PhD, from Saskatchewan study in general practice. Pharmacoepidemiol Drug Saf 2002; 11:
Health for her help in compiling the datasets and Dr William Ghali, University of Calgary for comments on an earlier draft 16 Von Korff M, Wagner EH, Saunders K. A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992; 45: 197–203.
of this manuscript. This study is based on de-identified data 17 Cox DR. Regression models and life tables. J R Stat Soc 1972; 34
provided by the Saskatchewan Department of Health. The interpretation and conclusions contained herein do not neces- 18 Ghali WA, Quan H, Brant R, van Melle G, Norris CM, Faris PD et al. sarily represent those of the Government of Saskatchewan or Comparison of 2 methods for calculating adjusted survival curves the Saskatchewan Department of Health.
from proportional hazards models. JAMA 2001; 286: 1494–1497.
19 Rubin DB. Estimating causal effects from large data sets using pro- pensity scores. Ann Intern Med 1997; 127: 757–763.
References
20 Rosenbaum P, Rubin D. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc 1984; 79:
1 DeFronzo RA, Goodman AM. Efficacy of metformin in patients with non-insulin-dependent diabetes mellitus. The Multicenter Met- 21 D’Agostino RB Jr. Propensity score methods for bias reduction in the formin Study Group. N Engl J Med 1995; 333: 541–549.
comparison of a treatment to a non-randomized control group. Stat 2 Dunn CJ, Peters DH. Metformin. A review of its pharmacological Med 1998; 17: 2265–2281.
properties and therapeutic use in non-insulin-dependent diabetes 22 Joffe MM, Rosenbaum PR. Invited commentary: propensity scores.
mellitus. Drugs 1995; 49: 721–749.
Am J Epidemiol 1999; 150: 327–333.
3 Mather KJ, Verma S, Anderson TJ. Improved endothelial function 23 Donnan PT, Leese GP, Morris AD. Hospitalizations for people with with metformin in type 2 diabetes mellitus. J Am Coll Cardiol 2001; type 1 and type 2 diabetes compared with the nondiabetic population 37: 1344–1350.
of Tayside, Scotland: a retrospective cohort study of resource use.
4 UK Prospective Diabetes Study Group. Effect of intensive blood- Diabetes Care 2000; 23: 1774–1779.
glucose control with metformin on complications in overweight 24 Gatling W, Tufail S, Mullee MA, Westacott TA, Hill RD. Mortality patients with type 2 diabetes (UKPDS 34). Lancet 1998; 352: 854–
rates in diabetic patients from a community-based population com- pared to local age/sex matched controls. Diabet Med 1997; 14: 316–
5 Fisman EZ, Tenenbaum A, Benderly M, Goldbourt U, Behar S, Motro M. Antihyperglycemic treatment in diabetics with coronary 25 Gu K, Cowie CC, Harris MI. Mortality in adults with and without disease: increased metformin-associated mortality over a 5-year diabetes in a national cohort of the U.S. population, 1971–1993.
follow-up. Cardiology 1999; 91: 195–202.
Diabetes Care 1998; 21: 1138–1145.
6 Olsson J, Lindberg G, Gottsater M, Lindwall K, Sjostrand A, Tisell 26 Bertoni AG, Krop JS, Anderson GF, Brancati FL. Diabetes-related A et al. Increased mortality in Type II diabetic patients using sulpho- morbidity and mortality in a national sample of U.S. elders. Diabetes nylurea and metformin in combination: a population-based observa- Care 2002; 25: 471–475.
tional study. Diabetologia 2000; 43: 558–560.
27 UK Prospective Diabetes Study Group. Intensive blood-glucose con- 7 Fisman EZ, Tenenbaum A, Boyko V, Benderly M, Adler Y, trol with sulphonylureas or insulin compared with conventional Friedensohn A et al. Oral antidiabetic treatment in patients with treatment and risk of complications in patients with type 2 diabetes coronary disease: time-related increased mortality on combined (UKPDS 33). Lancet 1998; 352: 837–853.
glyburide/metformin therapy over a 7.7-year follow-up. Clin Cardiol 28 Turner RC, Cull CA, Frighi V, Holman RR. Glycemic control with 2001; 24: 151–158.
diet, sulfonylurea, metformin, or insulin in patients with type 2 dia- 8 Turner RC, Holman RR. Metformin and risk of cardiovascular betes mellitus: progressive requirement for multiple therapies disease. Cardiology 1999; 91: 203–204.
(UKPDS 49). JAMA 1999; 281: 2005–2012.
2005 Diabetes UK. Diabetic Medicine, 22, 497– 502

Source: http://www.braiddm.ca/reduced_cvd_mortality.pdf