Comparative effectiveness of statins on non-high density lipoprotein cholesterol in people with diabetes and at risk of cardiovascular disease: systematic review and network meta-analysis

Abstract

Objective: To compare the efficacy of different statin treatments by intensity on levels of non-high density lipoprotein cholesterol (non-HDL-C) for the prevention of cardiovascular disease in people with diabetes.

Design: Systematic review and network meta-analysis.

Data sources: Medline, Cochrane Central Register of Controlled Trials, and Embase from inception to 1 December 2021.

Review methods: Randomised controlled trials comparing different types and intensities of statins, including placebo, in adults with type 1 or type 2 diabetes mellitus were included. The primary outcome was changes in levels of non-HDL-C, calculated from measures of total cholesterol and HDL-C. Secondary outcomes were changes in levels of low density lipoprotein cholesterol (LDL-C) and total cholesterol, three point major cardiovascular events (non-fatal stroke, non-fatal myocardial infarction, and death related to cardiovascular disease), and discontinuations because of adverse events. A bayesian network meta-analysis of statin intensity (low, moderate, or high) with random effects evaluated the treatment effect on non-HDL-C by mean differences and 95% credible intervals. Subgroup analysis of patients at greater risk of major cardiovascular events was compared with patients at low or moderate risk. The confidence in network meta-analysis (CINeMA) framework was applied to determine the certainty of evidence.

Results: In 42 randomised controlled trials involving 20 193 adults, 11 698 were included in the meta-analysis. Compared with placebo, the greatest reductions in levels of non-HDL-C were seen with rosuvastatin at high (-2.31 mmol/L, 95% credible interval -3.39 to -1.21) and moderate (-2.27, -3.00 to -1.49) intensities, and simvastatin (-2.26, -2.99 to -1.51) and atorvastatin (-2.20, -2.69 to -1.70) at high intensity. Atorvastatin and simvastatin at any intensity and pravastatin at low intensity were also effective in reducing levels of non-HDL-C. In 4670 patients at greater risk of a major cardiovascular events, atorvastatin at high intensity showed the largest reduction in levels of non-HDL-C (-1.98, -4.16 to 0.26, surface under the cumulative ranking curve 64%). Simvastatin (-1.93, -2.63 to -1.21) and rosuvastatin (-1.76, -2.37 to -1.15) at high intensity were the most effective treatment options for reducing LDL-C. Significant reductions in non-fatal myocardial infarction were found for atorvastatin at moderate intensity compared with placebo (relative risk=0.57, confidence interval 0.43 to 0.76, n=4 studies). No significant differences were found for discontinuations, non-fatal stroke, and cardiovascular deaths.

Conclusions: This network meta-analysis indicated that rosuvastatin, at moderate and high intensity doses, and simvastatin and atorvastatin, at high intensity doses, were most effective at moderately reducing levels of non-HDL-C in patients with diabetes. Given the potential improvement in accuracy in predicting cardiovascular disease when reduction in levels of non-HDL-C is used as the primary target, these findings provide guidance on which statin types and intensities are most effective by reducing non-HDL-C in patients with diabetes.

Systematic review registration: PROSPERO CRD42021258819.

Fig 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart

Fig 2

Network of available comparisons between statin intensities for non-high density lipoprotein cholesterol, and forest plot of network effect sizes of statin intensities compared with placebo. Size of node is proportional to number of trial participants, and thickness of line connecting nodes is proportional to number of trial participants randomised in trials directly comparing the two treatments. Numbers represent the number of trials contributing to each treatment comparison. Certainty of the evidence, according to the confidence in network meta-analysis (CINeMA) framework, is included in the forest plot and classified as *low, †moderate, and ‡high confidence of evidence. Appendix 11 shows the full CINeMA assessments

Fig 3

League table of direct comparisons for statin intensities with effect estimates as mean differences (mmol/L). Statin intensities are reported in order of most effective treatment based on surface under the cumulative ranking curve score compared with placebo. Data are standardised mean difference (95% credible interval) in the column defining treatment compared with the row defining treatment. Green=network meta-analysis estimates (105 comparisons); orange=direct pairwise meta-analysis estimates. Appendix 6 gives the numbers of patient and studies. ND=no direct evidence available. The certainty of the evidence, according to the confidence in network meta-analysis (CINeMA) framework, was classified as *very low, †low, ‡moderate, and §high confidence of evidence

Fig 4

Network of available comparisons between statin intensities for low density lipoprotein cholesterol, and forest plot of network effect sizes of the statin intensities compared with placebo. Size of node is proportional to number of trial participants, and thickness of line connecting nodes is proportional to number of trial participants randomised in trials directly comparing the two treatments

Fig 5

Network of available comparisons between statin intensities for non-high density lipoprotein cholesterol adjusted for patient risk, with forest plot of network effect sizes compared with placebo. Size of node is proportional to number of trial participants, and thickness of line connecting nodes is proportional to number of trial participants randomised in trials directly comparing the two treatments. Patient risk is classified as high (HR) and low to moderate (LR)