Secondary Prevention of Cardiovascular Events With Long-Term Pravastatin in Patients With Diabetes or Impaired Fasting Glucose: Results from the LIPID trial

The study design has been described previously (14,15). People who were 31–75 years of age and had suffered a myocardial infarction or hospital admission for unstable angina 3–36 months before registration were eligible for the LIPID trial if plasma total cholesterol was 4.0–7.0 mmol/l and fasting triglycerides <5.0 mmol/l. Patients were screened in 87 centers in Australia and New Zealand. After an 8-week placebo run-in phase, 9,014 patients were randomly allocated to receive pravastatin 40 mg daily or matching placebo. Analysis of the results in the subgroup of patients with diabetes was prespecified in the protocol. It was later decided to also present the results of the treatment effects in the diabetic and IFG groups combined because both conditions represent different stages of the same disease process and the larger number of events allowed more reliable comparisons. The prespecified primary outcome for all LIPID trial subgroup analyses was death from CHD or nonfatal myocardial infarction (combined).

All patients were asked at the first screening whether they had ever been diagnosed with diabetes. The present American Diabetes Association cut points for the diagnosis of diabetes and for IFG were used to classify the subgroups according to a single fasting venous plasma glucose level measured at baseline (Fig. 1) (3). The availability of a single glucose measurement was considered adequate for epidemiological purposes at the time the study commenced, although not for the clinical diagnosis of diabetes. This may have resulted in some misclassification between groups, but this would have been nondifferential, resulting in more conservative estimates of differences between groups in risks and effects of treatment.

Patients who identified themselves as having diabetes (782 patients [8.7%]) and the 295 (3.3%) patients with probable undiagnosed diabetes (fasting plasma glucose ≥7.0 mmol/l) constituted the diabetic group (1,077 patients [11.9%]). Of the patients without diabetes, 940 (10.4%) were classified as having probable IFG (fasting plasma glucose 6.1–6.9 mmol/l) and 6,997 (77.6%) as having normal fasting glucose (NFG; fasting plasma glucose ≤6.0 mmol/l). Of those classified as having diabetes, 28 of 1,077 (2.6%) had type 1 diabetes. Metformin treatment was being taken by 136 patients and sulfonylureas by 358, of whom 84 were receiving both. Patients with no history of diabetes and who provided no baseline plasma glucose measurement (23 [0.3%]) were considered to have NFG to ensure conservative estimates in the analyses. In total, 2,017 patients (22.4%) showed glucose metabolism abnormalities at entry to the study; 1,001 were assigned to placebo and 1,016 to active treatment (Fig. 1).

Plasma cholesterol levels were measured by the core laboratory at baseline, 6 months later, yearly from baseline, and at study end. Fasting HDL cholesterol and triglyceride levels were measured at baseline, 1, 3, and 5 years later, and at study end. LDL cholesterol was estimated indirectly using the Friedewald formula (16). Apolipoprotein B levels were also measured at baseline and after 1, 3, and 5 years.

The study personnel and the patients remained blinded to the on-trial lipid results. The patients’ usual care, including other cholesterol-lowering treatment, continued to be under the direction of their own doctors. The results of the other large-scale trials of HMG-CoA reductase inhibitors (17–19) were communicated to the patients and their doctors, with further clarification that open-label cholesterol-lowering therapy could be commenced if this was found necessary. Compliance with study medication, any hospital admissions, serious adverse events, and study outcomes were monitored at routine 6-month clinic visits.

The Outcomes Assessment Committee or the Stroke Adjudication Committee, which was blinded to each patient’s study treatment, reviewed all deaths, myocardial infarctions, and strokes. An independent safety and data monitoring committee monitored progress. The trial was planned to continue until 700 CHD deaths had occurred, unless the trial was stopped early following an interim analysis showing a difference of at least 3 SDs in all-cause mortality or serious adverse events associated with pravastatin treatment (15). The trial was conceived, managed, and analyzed independently of the sponsor. All patients gave written informed consent, and the ethics committee for each participating center approved the trial.

All analyses were performed on an intention-to-treat basis. All P values were unadjusted for multiple comparisons. Statistical significance was determined at P < 0.05 (two tailed). Time-to-event analyses used the log-rank test stratified by qualifying event (20). Estimates of relative risk reduction (RRR) and 95% CIs were derived from the Cox proportional hazards model (21). In analyses, patients with diabetes or IFG were considered as separate groups and also as a single group with abnormal fasting glucose (AFG). The variation in the effect of pravastatin among the fasting glucose subgroups was evaluated using the interaction in a Cox model. The number needed to treat was calculated in all cases using the common RRR for the whole cohort and may be unreliable for some individual groups. The meta-analysis used published data with odds ratios and a fixed-effects model (22).

Patients with diabetes were older and more often obese than patients with NFG, and the diabetic group included more women and more patients with a history of hypertension, claudication, stroke, and transient ischemic attack (Table 1). Therapy with calcium-channel antagonists, ACE inhibitors, nitrates, and diuretics were all more common in the diabetic group. In the IFG group, these characteristics and the levels of drug use were intermediate. Balance in important prognostic variables between those allocated to pravastatin and those allocated to placebo was excellent (14).

The three groups differed in baseline lipid levels (all P < 0.001). Fasting levels of total cholesterol, and HDL and LDL cholesterol were lower in the diabetic group, whereas triglyceride levels were higher (Table 2). Lipid levels of those with IFG were intermediate.

The responses of blood lipids to pravastatin were very similar in all three groups with the exception of triglycerides, for which levels fell most in the diabetic group (Fig. 2). Apolipoprotein B levels, which reflect the total number of atherogenic lipoprotein particles and predict the risk of fatal myocardial infarction (23,24), fell by a similar proportion in the three groups.

The percentages of patients who permanently discontinued the study medication were significantly different in the three groups: 18% for NFG, 20% for IFG, and 23% for the diabetic group (P = 0.006). Prescription of active lipid-lowering therapy for those allocated to placebo increased progressively over time, which contributed to a 5% fall from baseline in mean blood cholesterol by study end. In the placebo group, 23% with diabetes, 22% with IFG, and 24% with NFG, respectively, dropped in to cholesterol-lowering medication. Patients withdrew from study medication over a similar time course; the observed difference in total cholesterol with pravastatin treatment fell from 21% at 1 year to 18% at 5 years.

In patients allocated to placebo, event rates were higher in those with diabetes than in the other groups (Figs. 3 and 4). For patients with diabetes compared with patients with NFG, the absolute excess risks were 8.9% for CHD death or nonfatal myocardial infarction and 6.3% for stroke, although their baseline cholesterol levels were similar. The risk of any cardiovascular event was 1.4 times (95% CI 1.3–1.5) the risk for those with NFG.

In placebo patients with IFG, the absolute excess risks were 3.3% for CHD death or nonfatal myocardial infarction and 1.8% for stroke. The risk of any cardiovascular event was 1.2 times (95% CI 1.1–1.3) the risk for those with NFG.

Allocation to pravastatin was associated with lower observed risks of all major cardiovascular end points for all three patient groups (Figs. 3 and 4). For most end points, RRRs were similar, but the measured reductions in absolute risk associated with pravastatin treatment were greater among those with diabetes and IFG than in those with NFG. Despite pravastatin treatment, the event rates for these outcomes among patients with diabetes were mostly higher than among those with NFG allocated to placebo. In contrast, for all end points, those with IFG allocated to pravastatin had lower event rates than patients with NFG allocated to placebo.

For the primary combined outcome of CHD death or nonfatal myocardial infarction, pravastatin therapy reduced the risk among all patients from 15.9 to 12.3% (RRR 24%, P < 0.001). In patients with diabetes, the RRR was 19% (23.4 vs. 19.6%, P = 0.11) compared with 23% (14.5 vs. 11.3%, P < 0.001) in the NFG group. The RRR in the IFG group was 36% (17.8 vs. 11.8%, P = 0.009); these treatment effects were not significantly different from one another (test for interaction, P = 0.53) (Fig. 3). The absolute risk was larger in the diabetic group (23.4%) than in the NFG group (14.5%), so the estimated number needed to treat to prevent one event, based on the common risk reduction, was lower in the diabetic group (n = 18) than in the NFG group (n = 29).

Pravastatin reduced the risk of stroke from 9.9 to 6.3% in the diabetic group (RRR 39%, CI 7–61%, P = 0.02) and from 5.4 to 3.4% in the IFG group (42%, CI −9 to 69%, P = 0.09). Pravastatin therapy thus reduced the absolute risk of stroke by 3.6% in the diabetic group and by 2.0% in the IFG group, compared with only 0.2% in the NFG group. A similar trend was seen for the expanded end point (any major cardiovascular event) and for coronary revascularization procedures.

Among all patients with AFG (diabetes and IFG combined), the increased numbers of events resulted in statistically significant reductions in major coronary events, stroke, and cardiovascular events, suggesting that separately nonsignificant results for these outcomes for the diabetic and IFG groups related to statistical power rather than an absence of benefit (Fig. 3).

Of the 6,997 patients with NFG at study entry, over the mean 6.0 years of follow-up, 264 patients (3.8%) developed diabetes, as evidenced by a fasting blood glucose level of ≥7 mmol/l or reported use of oral hypoglycemic medication or insulin. There was no apparent effect of pravastatin on the development of new diabetes: 138 (4.5%) of the survivors among those allocated to placebo and 126 (4.0%) of the survivors allocated to pravastatin developed diabetes (P = 0.32). In the IFG group, 9.2% of the placebo group survivors and 9.7% of the pravastatin group survivors developed diabetes.

A total of 2,782 major coronary events (coronary death or nonfatal myocardial infarction) occurred among 17,445 individuals with recorded glucose status during scheduled follow-up in LIPID, CARE, and 4S combined, and 499 strokes among 13,046 individuals with recorded glucose status in LIPID and CARE combined (stroke results for 4S by glucose status were not found in the published literature). Large, clinically important reductions were seen in major coronary events in all glucose groups. Long-term allocation to statin therapy reduced coronary events by 28% overall in diabetes (P = 0.001), 39% in IFG (P < 0.0001), and 29% in NFG (P < 0.0001) without evidence of statistical heterogeneity (P > 0.15) between individual study results within any category of glucose status (Fig. 5).

Statin therapy reduced stroke events by 32% (P = 0.04) overall in diabetes, 39% in IFG (P = 0.08), and 15% in NFG (P = 0.1) without evidence of statistical heterogeneity (P > 0.4) between study results. For both outcomes, the absolute event rate in the placebo groups was substantially higher among those with diabetes than those with IFG, with the lowest rate observed in those with NFG, which is consistent with the results of the LIPID study.

The LIPID study confirms that patients with a history of myocardial infarction or unstable angina and diabetes or IFG have a substantially higher risk of cardiovascular events. The risk of a major coronary event (CHD death or nonfatal myocardial infarction) for placebo patients with clinically diagnosed diabetes in the LIPID study was 1.6 times that of patients with NFG, compared with ∼1.8 times in CARE or 4S (12,13). For patients with IFG on placebo in the LIPID study, the risk of a major coronary event was 1.2 times that of patients with NFG (a significant difference) compared with nearly 1.2 times in 4S and nearly 1.3 times in CARE.

Treatment with pravastatin conferred clinical benefits across all groups, but in those with IFG or diabetes, the absolute risk reductions were greater. For example, over 6 years, preventing a single death from any cause would require treating 39 patients with NFG but only 23 patients with IFG or diabetes. Preventing one CHD death or myocardial infarction would require treating 29 patients with NFG compared with 20 patients with IFG or diabetes. Treating 33 patients with NFG or 29 patients with IFG or diabetes would prevent one revascularization procedure.

Together with the other major studies, LIPID provides clear evidence for the benefits of treatment with an HMG-CoA reductase inhibitor for secondary prevention of CHD in patients with diabetes or IFG. Meta-analysis of data from 4S, CARE, and LIPID supports the findings of LIPID alone with regard to coronary events and stroke. The baseline levels of plasma cholesterol were 4.0–7.0 mmol/l in LIPID (14), very similar in CARE (13), and 5.5–8.0 mmol/l in 4S (17), so the findings are applicable across a wide range of cholesterol levels, covering nearly all patients with diabetes and CHD seen in clinical practice.

The effects of pravastatin therapy in reducing stroke in diabetes are also especially important. The rates of stroke were substantially higher among those with diabetes and IFG, and treatment was more effective in these groups than in the group with NFG.

An unexpected outcome of the West of Scotland Coronary Prevention Study (WOSCOPS), a study of primary prevention with pravastatin, was a 30% reduction with pravastatin in the risk of developing diabetes during the study (25). Triglyceride-lowering or anti-inflammatory properties of pravastatin were postulated as a possible mechanism. In LIPID, the pravastatin and placebo groups did not differ in the rates at which they developed diabetes. The studies were of similar size, but LIPID had an older cohort with a median age of 62 years (14); the mean age in WOSCOPS was 55 years (18). In LIPID, 17% of the patients were women, whereas WOSCOPS only studied men. Finally, subjects in LIPID had clinical evidence of prior CHD, but those in WOSCOPS did not. However, the recently completed U.K. Heart Protection Study (HPS) (26) also reported no effect of statin therapy on the development of diabetes in a wide range of patients. Only 139 patients developed diabetes in WOSCOPS, suggesting that this finding could be plausibly related to random imbalances at baseline.

The worldwide prevalence of diabetes is expected to double between 1997 and 2010 (27). This prediction has led the World Health Organization to identify diabetes as one of the three largest health problems to be tackled globally at the start of this millennium. The meta-analysis (Fig. 5) presents strong evidence of the benefit of cholesterol-lowering therapy with statins for patients with prior CHD and diabetes or IFG. The HPS findings of a 17% reduction in major CHD events and a 27% reduction in major cardiovascular events with simvastatin in patients with prior CHD and diabetes support our results (26). The HPS further extends the findings to people with diabetes without prior cardiovascular disease, showing a similar proportional risk reduction in people at lower absolute risk. In the HPS, however, fasting blood glucose was not measured at baseline, so this study offers no additional information about patients with IFG, with or without vascular disease.

Two further studies that included people with diabetes have recently been published, but so far neither has reported results for people with IFG. The Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT-LLA) was terminated early after 3.3 of 5 years’ scheduled follow-up, reporting an overall 36% (CI 17–50%, P < 0.001) reduction in coronary events with 10 mg atorvastatin daily. There was, however, only a nonsignificant trend to fewer primary end point events among 2,532 people with diabetes (28). In the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT-LLT), allocation to pravastatin 40 mg daily resulted in an average difference in cholesterol levels of only 9.6%, owing to free use of statins in the control (usual-care arm), and resulted in no significant reduction in events overall or separately in the 3,638 people in the diabetic group (29). A third study, the Collaborative Atorvastatin Diabetes Study (CARDS) (30), recently concluded early, apparently because of significant benefits of atorvastatin 10 mg daily on a composite cardiovascular end point in >2,800 people with diabetes at high risk for, but with no prior, clinical coronary disease (H.A.W. Neil, personal communication).

Because patients with diabetes and established CHD have a high absolute risk of further events, substantial and prolonged reductions in their blood cholesterol levels are imperative. With pravastatin therapy, their RRR is similar to that of patients without diabetes but their absolute risk reduction is greater, so the number needed to treat to reduce cardiovascular events is less and, therefore, the cost-effectiveness of therapy is greater (31). The LIPID study reliably confirms that for patients with established CHD, the absolute benefits of statin treatment are substantially greater for those with AFG. Therapy is clearly indicated in this group.

The LIPID study was supported by grants from the Bristol-Myers Squibb Pharmaceutical Research Institute and the National Health and Medical Research Council and was independently coordinated by the National Health and Medical Research Council Clinical Trials Centre under the auspices of the National Heart Foundation of Australia. The research work of A.C. and R.J.S. is supported by research grants and personal fellowships from the National Health and Medical Research Council.