Keywords
INTRODUCTION
Type 2 diabetes mellitus (DM2) is a well-known risk factor for cardiovascular disease (CVD).1,2 A cohort study carried out in 1998 found that patients with DM2 but no history of myocardial infarction (MI) had a similar CHD mortality to that of persons without diabetes mellitus but with a prior history of MI,3 leading to the conclusion that DM2 was a coronary risk equivalent. This concept was included in the Third Report of the National Cholesterol Education Program4 and in the European guidelines on the prevention of CVD.5,6 These guidelines focus on the prevention of CVD rather than the prevention of coronary heart disease (CHD), and classify persons with DM2 in the same category of a high risk for a fatal cardiovascular event as persons with established CVD.5 However, over recent years several studies have examined this subject, but with contradicting results. Thus, some studies have found that CHD mortality was the same7-9 or even greater10,11 in persons with diabetes mellitus without CHD (DM+CHD) as in persons without DM but with CHD (DMCHD+), whereas other studies have found that DM confers a lower risk for complications than CHD.12-17 Furthermore, yet other studies indicate that men and women could react differently in this aspect.18-20
Determining whether DM2 is or is not a coronary risk equivalent may be important in order to improve the current recommendations on the treatment of cardiovascular risk factors, although to date no systematic review has been undertaken. In this study we carry out a systematic review whose main aim is to assess whether DM+CHD persons have the same risk for CHD and CVD mortality as DMCHD+ persons, although we recognize that the terms DM and CHD cover a wide spectrum of clinical situations that may differ greatly.
METHODS
Inclusion Criteria
Studies were included if they fulfilled the following criteria: a) they were prospective and had a follow-up of at least 5 years; b) they included at least the following 3 groups of subjects: DM+CHD, DMCHD+, and DMCHD (subjects without DM and without CHD); and c) they presented data on CHD or CVD death separately for men and women. If a study led to publication of more than one article, then the most recent article was analyzed.
Bibliographic Search and Identification of the Primary Studies
We undertook a search in PubMed for studies in human subjects, with no restriction on language, up to February 28, 2006. The following search strategy was used, with these medical subject headings: ("diabetes mellitus") AND ("follow-up studies" OR "longitudinal studies" OR "prospective studies" OR "case-control studies" OR "cohort studies," OR "comparative study") AND ("coronary disease/complications" OR "coronary disease/mortality"), OR ("cardiovascular diseases/complications" OR "cardiovascular diseases/mortality") OR ("myocardial infarction/complications" OR "myocardial infarction/ mortality"). In order to detect possible clinical trials that had not been found with the previous strategy, we also carried out the following search (studies in humans, with no restriction on language up to February 28, 2006): ("diabetes mellitus") AND ("clinical trials") AND ("coronary disease/complications" OR "coronary disease/mortality") OR ("cardiovascular diseases/ complications" OR "cardiovascular diseases/mortality"), OR ("myocardial infarction/complications" OR "myocardial infarction/mortality"). Two independent reviewers obtained the full-text article of all those considered to be potentially relevant for the review. In order to detect any other articles that might fulfill the inclusion criteria we also analyzed all the references of the full-text articles retrieved and consulted with experts in the subject. All possible discrepancies between the 2 reviewers were resolved by consensus agreement between them.
Data Extraction
The following information was collected for each of the articles selected: design, characteristics of the study population (number of persons, age, sex, race), country, years of follow-up, losses during the follow-up, type of DM, duration of the DM, diagnostic criteria for DM and CHD, method used to determine death, study groups with their estimations of adjusted odds ratios (OR) or hazard ratios (HR), with their respective 95% confidence intervals (CI), for each of the events recorded in each study. None of the authors of any of the studies was contacted concerning the data analyzed. When the same article referring to the same event included different estimations of OR or HR, we always used that which had been adjusted for more confounding factors. As many studies report the HR instead of the OR and considering that both terms represent very similar concepts, we shall use the term HR from here on. Two independent reviewers analyzed the quality of the studies according to the criteria of the STROBE report (Strengthening the Reporting of Observational studies in Epidemiology).21 Possible discrepancies were resolved by mutual agreement between the reviewers.
Data Analysis
Subjects in the DMCHD group were used as the reference group (this was done in most of the studies analyzed) and compared against the DM+CHD and the DMCHD+ groups and the group of patients with DM and with CHD (DM+CHD+). The estimations of the HR were combined, weighting for the inverse of the variance with a random effects model.22 This model was chosen because it functions better than the fixed effects model when the number of studies to be combined is less than 20,23,24 and it also takes into account the variation between the studies.25,26 The HR of the different categories (DM+CHD vs DMCHD+, etc) were compared using their standard errors.27 The possible publication bias was studied in comparison with the greatest number of studies (ie, CHD mortality in men) by means of the method of Egger et al28 and the regression of the funnel plot.29 All the analyses were performed with the statistical program Stata 8-SE (College Station, Texas, United States).
RESULTS
The bibliographic search retrieved 4233 articles with the first strategy and 655 with the second (designed to specifically detect clinical trials). After reviewing the titles and abstracts of all these, 22 full-text articles were selected. A further 2 articles were obtained after reviewing the references for these 22 articles.18,30 Of these 24 articles, 11 were excluded for the following reasons: a) they did not include persons from the DM+CHD or DMCHD+ groups31; b) they lacked data on CHD or CVD death stratified according to sex12,15,32-34; c) data were presented in a more recent article3; and d) similar data were presented in another publication but with no relevant new data.35
Thus, 12 studies finally fulfilled the inclusion criteria totally.7-11,13,14,16,17,19,20,30 An additional study was included because, although it grouped subjects according to CVD instead of CHD,18 we considered CHD to be the most important clinical manifestation of CVD, bearing in mind the age and origin of the persons studied. Analysis of the impact of this study on the final results of the meta-analysis was done by a sensitivity analysis, excluding this study from the analysis. Notably, the study by Haffner et al3 was not included in the meta-analysis because it included that of Juutilainen et al,11 which is based on the same cohort of subjects, and was published more recently with data from a longer-term follow-up (13 years rather than 7).
The 13 articles analyzed included data from over 15.000 persons in the DM+CHD group and over 21 000 in the DMCHD+ group, and who were followed up from 5 to more than 20 years. Table 1 summarizes the main characteristics of these studies, ordered by year of publication. They were all published over the last 6 years and carried out in Europe or America; 7 of them provide information about the race of the participants.8,13,16-19,30 With regard to the design of the studies, only 1 of them was a randomized clinical trial,13 11 were cohort studies,7,8,10,11,14,16-20,30 and the remainder were derived from the results of 2 population-based registries.9
Twelve of the studies provided data on men and just 9 on women. Most of the studies provided no information on the type of DM or its duration, and diagnosis of persons with CHD, CVD, or DM was done in the most part from information provided by the patients themselves. All the studies except one18 give information about CHD mortality, whereas 7 give information about CVD mortality. Additionally, 2 studies provide information about mortality due to stroke.17,20 We considered all those studies that had an acceptable quality in accordance with the STROBE report.
Table 2 shows the estimations of the HR with their respective 95% CI resulting from the combination of the corresponding HR of the studies evaluated. These estimations were calculated for CHD and CVD mortality, with the DMCHD group used as the reference group and stratifying the results by sex. Figure 1 shows the different HR in men in relation to CHD and CVD mortality, comparing the DM+CHD- and DM-CHD+ groups with the reference group (DMCHD). After combining the results of the 11 studies, the men in the DM+CHD group had a lower CHD mortality than in the DMCHD+ group, though the difference was not statistically significant (P=.066) (Table 2). For the women, 8 studies were combined to provide data on CHD mortality. The women in the DM+CHD group experienced greater CHD mortality than those in the DMCHD+ group, though again the difference was not statistically significant (P=.42) (Table 2).
Figure 1. A: funnel plot comparing coronary heart disease mortality between men without diabetes mellitus and without coronary heart disease (DMCHD; reference group) with men without diabetes mellitus but with coronary heart disease (DM+CHD). B: funnel plot comparing coronary heart disease mortality between men in the DMCHD group with men in the DM+CHD group. C: funnel plot comparing cardiovascular disease mortality between men in the DMCHD group with men in the DM+CHD group. D: funnel plot comparing cardiovascular disease mortality between men in the DMCHD group with men in the DMCHD+ group.
Concerning the men, after combining the results of the 5 studies that provided information about CVD mortality, this was found to be lower in persons from the DM+CHD group than those in the DMCHD+ group. These results were at the limit of statistical significance (P=.051) (Table 2). The results did not change after excluding the study by Becker et al,18 which grouped persons according to CVD rather than CHD (HRDM+CHD=2.67 [95% CI, 2.11-3.40], HRDMCHD+=3.83 [95% CI, 2.95-4.97], HRDM+CHD+=6.92 [95% CI, 5.33-8.99]). For women, it was possible to combine 3 studies to obtain estimations of the HR for mortality due to CVD. No significant differences were found between the DM+CHD and DMCHD+ groups (P=.592) (Table 2).
These results also remained the same after eliminating the study by Becker et al18 (HRDM+CHD=4.69 [95% CI, 4.22-5.22], HRDMCHD+=4.27 [95% CI, 1.42- 12.82], HRDM+CHD+=11.61 [95% CI, 3.82-35.28]).
Comparison of the HR between men and women showed significant differences between the 2 groups of DM+CHD in CHD mortality (HR in men=3.06 [95% CI, 2.45-3.83]; HR in women=4.68 [95% CI, 3.40-6.45]; P=.033), and CVD mortality (HR in men=2.55 [95% CI, 2-3.26]; HR in women=4.70 [95% CI, 4.23-5.22]; P<.001). However, no differences were found between men and women for either of the types of mortality (CHD and CVD) or in the DMCHD+ group or in the DM+CHD+ group (data not shown).
Combination of the 2 studies in men which provided information on mortality due to stroke17,20 showed that this was significantly greater in the DM+CHD group as compared with the DMCHD+ group (HR=3.07 [95% CI, 2.64-3.57] vs 1.70 [95% CI, 1.40-2.07], respectively, P<.001).
In order to study the publication bias we used the method of Egger in the comparison with the greatest number of studies (CHD mortality in men).
The P value of the ordinate in the origin was .778 in the DM+CHD group. Similar results were obtained in the DMCHD+ group (P=.352) and the DM+CHD+ group (P=.617). The results of the regression in the funnel plot were also non-significant.
DISCUSSION
This review confirms that both DM and CHD significantly increase the risk for CHD and CVD mortality. Additionally, it shows that comparison of persons who are DM+CHD with those who are DMCHD+ gives a relative risk of CHD and CVD mortality that varies between men and women. The men in the DM+CHD group had a non-significant trend towards lower CHD and CVD mortality as compared with the men in the DMCHD+ group. No differences were found for the women in CHD or CVD mortality between the DM+CHD and the DMCHD+ groups. It would probably have been better to compare directly persons who were DM+CHD with those who were DMCHD+, but all the studies except for one use DM-CHD- as the reference group.
Coronary Heart Disease Mortality and Cardiovascular Disease Mortality in DM+CHD Men as Compared With DMCHD+ Men
Coronary heart disease mortality in the men was lower in the DM+CHD group than the DMCHD+ group, though the difference was not significant.
Although this finding in itself is not sufficient to consider that DM is not a coronary risk equivalent, other results from this systematic review suggest that this could be the case. Firstly, CVD mortality, which basically includes CHD mortality and mortality due to cerebrovascular disease, was also lower in men from the DM+CHD group than men from the DMCHD+ group, in this case reaching the limit of statistical significance (P=.051). Moreover, it should be recalled that several studies adjusted for cholesterol concentrations, which could have led to over-adjusting the risk in the CHD+ persons whilst having no effect on the DM+ persons. It could thus be speculated that without this adjustment for cholesterol figures, the differences would have been even greater than those found between the DM+CHD and the DMCHD+ groups.
Coronary Heart Disease Mortality and Cardiovascular Disease Mortality in DM+CHD Women as Compared With DMCHD+ Women
As opposed to what was found for the men, the DM+CHD women had a greater risk for CHD and CVD mortality than the DMCHD+ women, although the difference was not statistically significant. In this case the following reflections can be made. Firstly, the statistical power of the sample of women was lower than that for the sample of men, as fewer studies were combined and, additionally, the total number of women was much lower (the cohort in the MRFIT study is much larger than any of the other studies included, and this study only included men). Secondly, if the data had not been adjusted for cholesterol concentrations, the differences between the 2 groups, DM+CHD and DMCHD+, might have been even greater. These facts support the idea that DM in women confers a similar risk for CHD and CVD mortality to CHD itself, although the information that we were able to obtain from the data with women was less than that for the men.
These results, therefore, all indicate that men and women with DM could behave differently concerning the risk for CHD and CVD mortality. In fact, comparison of the DM+CHD and the DMCHD+ groups in relation to the risks for CHD and CVD mortality showed that the risks were significantly greater in the women as compared to the men. To this extent, our results are in agreement with earlier studies that show that DM increases the risk for CVD to a greater degree in women than in men.36-38 This increased risk for women conferred by DM may even occur in pre-diabetes states and is only partly explained by an increase in the traditional cardiovascular risk factors.39
Limitations
This review has certain limitations. Firstly, most of the studies involved in our review included the diagnosis of CHD/CVD and DM based on information provided by the patients themselves, which may have led to errors in the classification of subjects within the various groups. Moreover, it should be recalled that the terms "CHD" and "DM" each include diverse nosologic situations. Generally speaking, one can consider that using the information provided by the patients in order to classify them as having CHD/CVD may be a relatively reliable way of doing so. However, in the case of DM this strategy may have led to "milder cases" of DM being excluded from those persons with the disease.9 Furthermore, none of the studies used the current criteria for the diagnosis of DM, which use lower blood glucose figures than the previous classification in order to define the presence of DM. It therefore seems quite likely that the persons with DM included in most of the studies analyzed correspond to what could be considered a well-established clinical disease. Thus, some persons in the groups classified as not having DM could have been reclassified as having DM, according to the current criteria for the diagnosis of DM. The extrapolation of our results to all persons who currently fulfill the criteria for DM is therefore complex.
Secondly, this review did not examine the effects of the duration of DM on the risk for CHD and CVD mortality, because these data were lacking in most of the studies included. Nevertheless, some of these studies found that as the duration of DM increases, the greater the risk of CHD mortality, both in men16 and in women.8,14 In fact, the duration of DM could be a factor that explains the differences in CHD mortality between the various studies carried out in order to determine whether DM is or is not a coronary risk equivalent.15 Finally, the HR for CHD and CVD mortality may vary considerably according to race,40 and even within similar ethnic populations.41 In this case, the lack of sufficient relevant data in the studies analyzed prevents us undertaking an analysis of the impact of race on CHD and CVD mortality.
CONCLUSIONS
This is the first systematic review to evaluate studies comparing CHD and CVD mortality in persons who were DM+CHD and DMCHD+. The men in the DM+CHD group tended to have a lower CVD mortality than the men in the DMCHD+ group, whereas the DM+CHD women tended to have greater CHD and CVD mortality. Nonetheless, with the studies available at the present time, neither of these two trends was statistically significant. Further studies are required to determine whether these trends can in fact be confirmed.
ACKNOWLEDGEMENTS
The authors are grateful to Neil Hossack for the English language version of the manuscript.
ABBREVIATIONS
CHD: coronary heart disease
CI: confidence interval
CVD: cardiovascular disease
DM: diabetes mellitus
DM2: type 2 diabetes mellitus
HR: hazard ratio
MI: myocardial infarction
STROBE: Strengthening the Reporting of Observational studies in Epidemiology
The authors have received grants from the Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III, Fondo de Investigaciones Sanitarias: RGDM 03/212 and PI 050626.
Correspondence: Dr. J.M. González-Clemente.
Servicio de Diabetes, Endocrinología y Nutrición. Hospital de Sabadell.
Parc Taulí, s/n. 08208 Sabadell. Barcelona. España.
E-mail: gonclem@yahoo.es
Received February 2, 2007.
Accepted for publication July 11, 2007.