Clinical significance of diabetes likely induced by statins: Evidence from a large population-based cohort

Introduction. A large number of studies has shown that use of statins is accompanied by an increased risk of developing type 2 diabetes [1-5], which is thus currently listed as an inconvenience of these drugs that may attenuate in some patients their protective effect. However, several aspects of the statin-induced diabetes have not been adequately clarified. For example, albeit several hypotheses have been advanced, the mechanisms through which statins favour the alteration of glucose metabolism that leads to the appearance of hyperglycemia and diabetes remain unclear [4,6]. Furthermore, although statin-induced diabetes is generally believed not to offset the protective lipid-lowering effect of statins on the cardiovascular (CV) system [7-9], limited information exists on whether statin-induced diabetes has the same adverse prognostic significance of native diabetes, i.e., whether it is associated with a similar increasing risk of diabetes-related macrovascular complications. This information is of fundamental importance to reliably quantify the impact of statin-induced diabetes on the role played by statins on primary and secondary CV prevention [10,11]. Aims. Since we have previously shown that at the population level an increasing adherence with statin treatment is accompanied by a clear-cut progressive increase in the risk of new onset type 2 diabetes [12], thus the purpose of the present study was to provide information on the extent to which type 2 diabetes more likely induced by statins affects the risk of macrovascular complications to a similar or different degree compared to diabetes of a more likely native nature. Methods. The data used for this study were retrieved from the healthcare utilization databases of Lombardy, a Region of Italy which accounts for about 16% (almost ten million) of its population, by means of a record-linkage procedure among databases. In Italy, the whole population is covered by the National Health Service and in Lombardy this has been associated since 1997 with an automated system of databases to collect a variety of information. All the residents in Lombardy, with an age between 40 and 80 years and at least one prescription of statins between 2003 and 2005, were identified. The date of the first dispensation was considered as the step-1 index date. Four patient categories were excluded: (i) patients who received one or more statin prescriptions within three years prior the step-1 index date; (ii) patients who received at least one antidiabetic agent, or were hospitalized with a diagnosis of diabetes, within the three years before the step-1 index date; (iii) patients who were hospitalized for CV disease or received prescriptions of CV drugs such as nitrates or digitalis within the three years before the step-1 index date; and (iv) patients who did not renew the initial prescription of statins and/or did not reach at least one year of follow-up. The remaining 84,828 patients newly treated with statins between 2003 and 2005 were enrolled in the study and represented the step-1 cohort, each of its members accumulating person-years of follow-up from the step-1 index date (i.e., index statin prescription) until the earliest among the dates of starting antidiabetic drug therapy (step-1 outcome, i.e., the appearance of diabetes as diagnosed by the prescription of antidiabetic drugs) or censoring, e.g. death from any cause, emigration or step-1 phase stopping (i.e., December 31th 2009). The 4,391 step-1 cohort members who exhibited new onset diabetes and the 77,893 statin-treated patients who did not have any antidiabetic drug dispensation were considered eligible for inclusion in the step-2 cohort. Cohort members who experienced diabetes were 1:1 randomly matched with those who did not developed diabetes for gender, age at cohort entry (±1 year), step-1 index date (±30 days) and previous adherence with statin therapy. A patient without diabetes had also to be at risk of developing diabetes when this was identified in the matched patient with diabetes. The 1,070 couples (formed by a patient with diabetes and his/her match without diabetes) who experienced a hospitalization for CV disease during the step-1 follow-up were excluded. The remaining 3,321 diabetic and non-diabetic matched pairs represented the step-2 cohort, each of its members accumulating personyears of follow-up from the date of the first antidiabetic drug prescription (i.e., the step-2 index date), until the earliest among the dates of step-2 outcome (first admission to public or private hospitals for macrovascular complications), death from any cause, emigration, or December 31th 2012. Macrovascular complications included myocardial infarction, peripheral vascular disease, myocardial revascularization, heart failure and cerebrovascular disease, as reported by the diagnosis at discharge from hospital. All prescriptions of statins dispensed to the cohort members during the step-1 follow-up were identified. Adherence with therapy was quantified as the cumulative number of days during which the medication was available divided by the number of days of follow-up, i.e., the “proportion of days covered” (PDC) by treatment. A two-stage data analysis was carried out. In the first stage, we looked for replication of our previous findings [12] that increasing the level of adherence with statins increases the risk of developing diabetes (the step-1 outcome), using the step-1 cohort. Cox proportional hazard regression was used to estimate the hazard ratio (HR), and the 95% CI, for patients to develop diabetes in relation to the exposure to statin treatment. The exposure to statin treatment was quantified by four PDC categories, i.e. a very low (PDC < 10%), low (PDC from 11% to 50%), intermediate (PDC from 51% to 89%) and high (PDC ≥ 90%) adherence with statins. Because drug exposure may vary over time, adherence categories were included in the model as time-dependent variables, thereby accounting for their cumulative and varying nature. Data were also adjusted for 1) the type of the index prescribed statin; 2) previous use of antihypertensive, antiarrhythmic, antithrombotic and antidepressant agents, as well as of no-steroidal anti-inflammatory drugs and drugs used for chronic obstructive lung disease; and 3) the Charlson comorbidity index score, calculated in the three years prior the step-1 index date. The second-stage data analysis focused on whether the greater or lesser chance for diabetes to be induced by statins translated into a different risk of macrovascular complications. To this aim, the 3,321 matched pairs forming the step-2 cohort were stratified according to the four categories of adherence with statins used for the step-1 follow-up. The overall risk of macrovascular complications associated with the diabetic condition was calculated for the entire cohort, as well as separately in each category of adherence with statins, using the HR, and 95% CI, derived from the Cox proportional hazard model. The risk of each specific macrovascular complication was also. Data were adjusted for the aforementioned covariates (excluding matching variables) as time-fixed. At the step-2 index date, covariates also included adherence with statins during the step-1 follow-up as well as with antidiabetic agents during the step-2 follow-up. Two ancillary analyses were performed to check whether interpretation of the risk of macrovascular complications in diabetic patients with a different adherence to statin might be affected by (i) a delayed protective effect of statin dispensed during the step-1 follow-up and/or (ii) the exclusion of patients with CV events during the step-1 follow-up or before the inclusion in the study, with thus a selection of those less susceptible to the protective effect of the drug. The Statistical Analysis System Software (version 9.4; SAS Institute, Cary, North Carolina, USA) was used for the analyses. For all hypotheses tested, two-tailed p-values less than 0.05 were considered to be significant. Results. The 84,828 patients belonging to the step-1 cohort accumulated 467,317 person-years of follow-up, on average 5.5 years per patient. During this period, 6,935 patients started antidiabetic therapy (the step-1 outcome) with an incidence of 14.8 cases every 1,000 person-years. The risk of developing diabetes raised progressively and significantly as adherence with statin increased. Compared to patients with very low adherence, the increase was 24% (95% CI: 12% to 37%), 72% (95% CI: 56% to 90%), and 95% (95% CI: 60% to 139%) for patients with low, intermediate and high adherence, respectively. The 3,321 matched pairs of diabetic and non-diabetic patients belonging to the step-2 cohort accumulated 33,623 person-years of follow-up, on average 5.0 years per patient. During this period, 376 and 272 diabetic and non-diabetic patients respectively experienced the step-2 outcome, the corresponding incidence being 22.9 and 16.1 hospitalizations every 1,000 person-years. The diabetes - macrovascular disease association experienced during the step-2 follow-up from the entire cohort of diabetic and non-diabetic patients, as well as within each category of step-1 adherence with statins, is shown in Figure 1. In the entire cohort, patients with diabetes had a risk excess of macrovascular complications 35% (15% to 58%) higher than that of patients without diabetes. As shown in the top panel of Figure 1, the potential of diabetes for generating macrovascular complications decreased with the increasing adherence to statins, the risk excess being 88% (95% CI: 17% to 204%), 51% (18% to 94%), 16% (-8% to 46%) and 4% (-64% to 197%) for very low, low, intermediate and high adherence, respectively. P-value for heterogeneity (p-value=0.0217) raises the significance. This trend was similar for the specific events separately considered, for all of which the diabetes-related risk was much less in patients with a high as compared to low adherence: 111% vs 9% for myocardial infarction, 112% vs 16% for myocardial revascularization and 59% vs 8% for cerebrovascular disease (Figure 1, bottom panel). The results of the ancillary analysis further confirmed the results obtained. Conclusions. In summary, our data confirm that there is a definite increase in the development of diabetes with statin therapy (as observed during the step-1 follow-up). Type 2 diabetes lost its association with increasing macrovascular risk when previous adherence with statin treatment was very high, and thus the chance of its induction by the drug greater. Statin-dependent type 2 diabetes might be prognostically less adverse than native diabetes.Trials reflecting the clinical relevance of treatment-induced diabetes mellitus compared to naïve diabetes mellitus regarding macrovascular complications are required to confirm this finding