The literature review was limited to randomized, controlled trials of drug therapy that included patients with diabetes. Only studies that measured major clinical end points were included. Major clinical end points were defined as major cardiovascular events (for example, cardiovascular mortality, myocardial infarction, stroke), cardiovascular mortality, and total mortality. Of note, many of the trials reported somewhat different clinical end points in the patients with diabetes. All included cardiovascular mortality and myocardial infarction in their composite end point; some included stroke and revascularization, and one included unstable angina. We used the primary reported data directly from the published study in our review. We also subdivided the literature review into 2 categories. The first category evaluated the effects of lipid management in primary prevention (that is, patients without known cardiovascular disease); the second evaluated the effects in secondary prevention.
We used several sources to identify the relevant literature. We started with a search of the Cochrane Library. We then performed a MEDLINE search in September 2002. We used the keywords exp diabetes mellitus and exp lipids [therapy or prevention and control] and limited the search to randomized, controlled trials and human studies. The final search produced 919 results. Of these, most were discarded because they did not measure major clinical end points, did not report outcomes for patients with diabetes, were observational in nature, or were reviews or editorials. We then updated the search through consultation with experts and through references from the identified articles, meta-analyses, and review articles.
The primary author extracted data from the primary study reports. Accuracy and quality of the abstraction were confirmed through reabstraction and comparison with the original abstraction. The outcomes were broken into categories as described earlier, and data on absolute and relative risk reduction and numbers needed to treat for benefit were derived from the primary reports or were calculated in standard fashion (11).
The results of the studies were then combined by using meta-analytic techniques. We pooled data for both relative and absolute risks. A Mantel–Haenszel test was done to test for heterogeneity. In the analyses of secondary prevention, the data had substantial heterogeneity, so the pooled risk ratios and differences were calculated by using the DerSimonian and Laird method with a random-effects model. Sensitivity analyses were done by excluding studies that appeared to be outliers to ascertain the source of the heterogeneity. All analyses were done by using the statistical package Stata (Stata Corp., College Station, Texas).
There were no sources of direct funding for this manuscript. Dr. Vijan was a Veterans Affairs Career Development Awardee during preparation, and the American College of Physicians provided an honorarium to the authors.