Meta-Analysis of the Effect of Diabetes on Restenosis Rates Among Patients Receiving Coronary Angioplasty Stenting

RESEARCH DESIGN AND METHODS

Identification and selection of studies

We conducted a search of the medical literature for studies evaluating restenosis rates after coronary angioplasty in patients with and without diabetes. We searched the English language literature in Medline between 1990 and 2002, using the search phrases “stent,” “coronary artery angioplasty,” and “diabetes.” Ten articles containing primary reports were identified (7,11–19). Two authors of this meta-analysis independently reviewed each article (J.G. and J.R.). Studies were selected for analysis if they had angiographic follow-up at a minimum of 6 months; if the sex and ages of their population were included; and for patients with diabetes if the mode of therapy, namely diet, oral agents, or insulin, was documented. Trials were included in the meta-analysis if patients both with and without diabetes were studied and if rates of restenosis were calculated for each group. Studies were excluded if fewer than 150 people participated. Six articles met the inclusion criteria for the meta-analysis (7,12–17). The studies by Silva et al. (11) and Mehilli et al. (18) were excluded because of the small sample size, and Mehilli et al. only studied patients with diabetes. The studies by Abizaid et al. (7) and Huynh et al. (19) were excluded because they only assessed clinical outcome, without analyzing angiographic restenosis rates. This analysis incorporates one substudy (16) and five primary studies, including one case series (14), one prospective study (13), one study based on a prospective registry (17), and two retrospective studies (12,15). All studies used native coronary vessels, and one study also included saphenous vein graft stenting (14). Restenosis was defined in the studies as a ≥50% decrease in target vessel diameter at angiographic follow-up. Most of the patients in the studies had type 2 diabetes, and the proportion of patients treated with diet, oral agents, or insulin was similar among the studies. In particular, most patients were being treated with oral agents, and insulin was the therapy used in about one-quarter of the patients with diabetes.

Statistical analysis

Restenosis rates for patients with and without diabetes used in our analysis were obtained from the published reports for each of the six trials. In one instance (17), the restenosis rates were calculated from published data presented as numerators and denominators. In another instance (15), the restenosis rates among patients with diabetes were based on the weighted average among patients with diabetes receiving insulin versus oral therapy.

The homogeneity of the effect of diabetes on restenosis rates was tested using methods described by DerSimonian and Laird (20). Restenosis rates after coronary stenting among patients with diabetes and without diabetes from each study were plotted against average age, percentage of the study population that was male, and the percentage of the population with diabetes receiving insulin therapy.

Logistic random effects models were used to estimate the effect of diabetes on restenosis rates after controlling for the percentage of men and the average age of each study population. The outcome for each study was the percentage of patients developing restenosis in patients with and without diabetes after coronary angioplasty with stent placement. The effect of diabetes on restenosis rates was estimated with a random effect, whereas the effects of age and sex were estimated with fixed effects using restricted maximum likelihood estimates.

RESULTS

The restenosis results from six studies were available for 6,236 individuals—1,166 with diabetes and 5,070 without. After weighting studies by the number of participants, the average age was 64.8 years among patients with diabetes and 61.3 years among patients without diabetes. The percentage of patients with diabetes who were male was 71.7%, and among patients without diabetes, 78.6% were male.

The average restenosis rate among patients with diabetes was 36.7% (range 25–55%) and 25.9% (18–28.3%) among patients without diabetes. After weighting studies by the number of participants, the average difference in restenosis rates between patients with and without diabetes within a study was 10.8% (2–35%). Restenosis rates and 95% CIs are shown for patients with and without diabetes for each study in Table 1. The effect of diabetes on restenosis rates was not homogeneous across studies (P < 0.001). When restenosis rates in patients with and without diabetes were plotted against mean age by study and diabetes status (Fig. 1), restenosis rates were shown to increase with age. Restenosis rates also increased with the percentage of the diabetic population treated with insulin (Fig. 2) but were not affected by the percentage of the population that was male (graph not shown).

In a univariate random effects logistic regression model, the odds ratio of restenosis associated with diabetes was 1.61 (95% CI 1.21–2.14, P = 0.004) (Fig. 3). From this model, the estimated probabilities of restenosis were 0.36 among patients with diabetes and 0.26 among patients without diabetes. After controlling for age in a multivariate logistic regression model, the odds ratio of restenosis was 1.30 (0.99–1.70, P = 0.055). The odds ratio of restenosis associated with a 10-year increase in age was 2.01 (P = 0.016). From the multivariate model adjusting for age, the estimated probabilities of restenosis were 0.295 among patients with diabetes and 0.244 among patients without diabetes.

CONCLUSIONS

While the rates of restenosis varied among the studies, all but one study included in this review found diabetes to be a risk factor for restenosis after coronary stent placement. However, patients with diabetes were older on average than patients without diabetes in every study. This is important because diabetes and older age are known risk factors for restenosis (21). Adjusting for the difference in the ages of patients with and without diabetes in a multivariate model reduced the apparent effect of diabetes on restenosis rates by approximately half.

There is much debate in the literature about the affect of sex on restenosis rates after coronary stent placement. Initially, reports suggested that women have a less favorable outcome after angioplasty than men, possibly because of the greater age to which women live, the greater prevalence of other coronary risk factors among women, and the smaller vessel size in women (22). Watanabe et al. (23) showed that women receiving angioplasty have a twofold higher mortality than men with or without stent placement. They concluded that female sex is an independent predictor of mortality even after adjustment for age and a poorer cardiovascular profile among women. More recent data suggest this correlation may not be true. Jacobs et al. (24) concluded, based on the National Heart, Lung and Blood Institute Dynamic Registry, that despite the high cardiovascular risk factors found in women, outcomes in women have improved and sex differences in patients undergoing coronary angioplasty have decreased. In all but one study included in our meta-analysis, there were more women among the population with diabetes than among the population without diabetes. However, adjusting for sex in multivariate models did not significantly change the effect of diabetes on restenosis rates.

Among the diabetic populations of the studies included in our meta-analysis, rates of restenosis after coronary stenting were higher among populations with higher proportions of patients treated with insulin. Because most patients with diabetes had type 2 diabetes, insulin therapy may be a marker for disease duration and severity (25,26).

Because follow-up angiograms necessary to assess restenosis may not be available for patients experiencing clinical outcomes such as fatal myocardial infarction or sudden death, our results may be biased if patients with diabetes are more likely to have clinical outcomes after coronary angioplasty. We compared death rates between patients with and without diabetes in the studies included in the meta-analysis and found no significant differences. Limitations of the studies included in this meta-analysis include lack of data on diabetes control, diabetes duration, lipid levels, blood pressure, and inflammatory markers. In general, it must be recognized that a meta-analysis does not allow for a systematic evaluation of important biological confounders that may affect the primary outcome.

The advent of drug-eluting stents is a remarkable improvement in preventing restenosis in coronary arteries after coronary angioplasty. The RAVEL study group (27) showed a significant reduction in restenosis with sirolimus-eluting stents compared with standard stents (27 vs. 0%) among 238 patients with coronary artery disease. A subgroup analysis of patients with diabetes showed that among 19 patients who received sirolimus-eluting stents, the restenosis rate was 0% compared with 42% among 25 patients who received standard stents. The more recent SIRIUS trial (28) focused on patients with higher risks of restenosis. Among the patients with diabetes, the restenosis rate was 18% in 131 patients who received sirolimus-eluting stents compared with 51% among 148 patients who received standard stents. Thus, it appears that sirolimus-eluting stents further reduce the restenosis rate among patients with diabetes compared with standard stents. Further research addressing the impact of clinical factors, such as age, sex, and treatment with insulin, on restenosis rates after placement of drug-eluting stents in patients with diabetes will help patients and their physicians assess individual risks of restenosis after coronary angioplasty with stenting.

In summary, although the published literature suggests that diabetes appears to be a risk factor for higher rates of restenosis in patients receiving coronary angioplasty with stenting, this effect is overstated in that we found that a significant part of this effect appears to be related to the older age of the patients reported in these studies. It seems surprising that the population of patients with diabetes was older, since diabetes is an established risk factor for coronary artery disease and one would expect a younger age of onset of coronary artery disease in these individuals. This emphasizes the importance of recognizing that age is a major contributing factor to the higher restenosis rates found in patients with diabetes. As well, it is unclear if younger patients with diabetes would have increased restenosis rates compared with individuals without diabetes. While not statistically significant in our analyses, sex and treatment with insulin may also be related to the risk of restenosis and could possibly attenuate the perceived risks associated with diabetes even further. Thus, it is important to consider age and other clinical factors when evaluating whether coronary stenting procedures are effective therapeutic options in patients with diabetes.