Patients with diabetes run a high risk of cardiovascular mortality and the female advantage with fewer cardiovascular events than in men at younger ages, is attenuated once a woman has the diagnosis of diabetes.1 2 The risk is increased about twofold in men and up to four times in women.2 The declining trend in cardiovascular mortality seen among non-diabetic subjects in recent decades is not observed to the same extent in patients with diabetes, especially not in women.3 4 Back in 1988, Framingham data revealed that women with diabetes have a poorer outcome than diabetic men after a myocardial infarction.5 Knowledge relating to prognosis and treatment is, however, still limited, as many randomised trials in patients with myocardial infarction only recruited limited numbers of female patients. It has been shown repeatedly that patients with diabetes are prescribed various evidence-based treatments less frequently than their non-diabetic counterparts. This may, at least in part, explain their poorer prognosis.6 7 It is not yet clear whether the increased cardiovascular mortality in diabetic women is related to this under-use, to female gender or to an increased risk factor burden. Previous reports often analysed the impact of diabetes by comparing non-diabetic and diabetic women or men. The present study compares diabetic women with diabetic men in order to investigate whether there are any gender-related differences in prognosis, risk factors and the use of evidence-based management in a large cohort of unselected diabetic patients with myocardial infarction.

METHODS

Patients

The Register of Information and Knowledge about Swedish Heart Intensive Care Admission (RIKS-HIA) contains information on all patients admitted to Swedish coronary care units. (CCU). The number of participating CCUs rose from 19 in 1995 to 70 in 2002 (in all, 73 CCUs in Sweden). Data were collected in 1995–2002. Because of an increased risk of concomitant diseases at higher ages, patients above the age of 80 years were excluded. A total of 70 882 patients with a first registry-recorded acute myocardial infarction were included, of whom 14 873 (21%) had previously known diabetes mellitus and 22 396 (31%) were women. A more detailed analysis was performed in patients aged <65 years (n = 25 555).

Case record forms

Information on patient care is entered into the RIKS-HIA via case record forms including about 100 variables, as detailed elsewhere.6 The complete protocol is available at www.riks-hia.se. Source data verification is performed by an external monitor comparing the register information with actual hospital records in randomly selected patients from about 20 different hospitals annually. More than 40 000 measurements are checked annually and the agreement between the registered information and patient records has always been between 94–96% each year.

Follow-up

Information on the performance of coronary procedures before and after hospital admission was obtained by matching patient data with the national registries on coronary angiography, percutaneous coronary interventions (PCI) and coronary artery bypass grafting (CABG). Long-term mortality was obtained by merging the RIKS-HIA database with the National Cause of Death Register covering the vital status of all Swedish citizens between 1995 and 2003. The mean follow-up time was 4.4 years.

Definitions

Myocardial infarction

In 1995–2000, the criteria for the diagnosis of acute myocardial infarction (AMI) were based on the World Health Organization criteria from 19948 combining symptoms with the increase of a biochemical marker (mainly creatine kinase isoenzyme (CK-MB)) and typical ECG changes. From late 2001, the criteria for the diagnosis of myocardial infarction were adapted to the ESC/ACC/AHA consensus document, using troponin T or troponin I, together with typical symptoms and/or ECG changes.9 10

Diabetes mellitus

Diabetes was defined as a previously established diagnosis of this disease or the prescription of glucose-lowering treatment at the time of hospital discharge (oral drugs or insulin).

Statistical analysis

The estimated survival in different age groups was calculated using a Cox regression analysis for the time period 1995–2002 and it was stratified by gender and the presence or absence of diabetes. In the <65 years age group, baseline characteristics, treatment and complications were summarised by percentages and compared between groups by relative risks with 95% confidence intervals (CI). The mean age was compared between groups by calculating a 95% CI, assuming a normal distribution, for the difference in means. A propensity score method was used to compensate for the non-randomised study design. This method, which has been described in detail elsewhere,11 12 produces a summary score of all the background characteristics for each patient. The score is simply the estimated probability of being a female patient, given the baseline characteristics estimated from a logistic regression model, with gender as the dependent variable and the baseline variables as independent variables. The propensity score is a so-called “balancing score”,12 meaning that, at least theoretically, given the propensity score, the included baseline variables should be balanced. Checks were made to ensure that balance was obtained between the female and male groups in our study and that there was no need for the further inclusion of any of the baseline variables in the subsequent analyses. As we were only interested in the best possible prediction (the estimated score) and not in the interpretation of the model parameters, the propensity score model can be allowed to be very complex. As a result, the model included age (as a third-degree polynomial to improve the prediction), smoking habits, previous diseases (history of myocardial infarction, heart failure, PCI or CABG and hypertension), admission ECG characteristics and medication at admission (aspirin, β-blockers, ACE inhibitors, calcium antagonists, lipid-lowering drugs, nitrates, anticoagulants and diuretics). Several two-way interactions among these variables were also included in the model to improve the prediction.

To study differences between females and males as regards in-hospital or discharge treatment modalities, separate logistic regression models were fitted for diabetic and non-diabetic patients, respectively. The models included gender as the dependent variable and the propensity score to compensate for baseline differences together with each respective treatment modality as independent variables. The results are presented as odds ratios and 95% CI.

To compare the risk of mortality between men and women with diabetes aged <65 years, Cox regression analyses were performed. Three models were presented (see fig 3B–D). The first included the propensity score, adjusting for the baseline characteristics. The second included the propensity score, as well as treatment with diuretics at hospital discharge. The third model included the propensity score and several discharge treatments (aspirin, β-blockers, ACE inhibitors, calcium antagonists, nitrates, anticoagulation, diuretics, digitalis and heparins) and revascularisation within 14 days from admission. The results from the Cox regressions are presented as relative risks with 95% CIs. Model performance was evaluated with the C-index. All the analyses were performed with the statistical program R (version 2.6.1; R Development Core Team).13

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Estimated long-term survival curves in patients with diabetes and age <65 years. (A) Unadjusted survival curves for diabetic patients by gender. (B) Survival curves adjusted for differences in baseline characteristics and previous treatment by the propensity score. (C) Survival curves after adjustments for the propensity score and treatment with diuretics at discharge. (D) Survival curves after adjustments for the propensity score and all in-hospital and discharge treatments.

Ethical consideration

All the patients were informed of their participation in the RIKS-HIA registry and the long-term follow-up. The merging with other registries was approved by the Swedish National Board of Health and Welfare and the Swedish Data Inspection Board.

RESULTS

Survival (up to 8 years) in three different age groups, ⩽64, 65–74 and ⩾75 years, is presented in figure 1A–C. A survival disadvantage for diabetic women compared with diabetic men was found in the youngest age group (RR 1.34; 95% CI 1.16 to 1.55). There was no significant gender-related difference in mortality between diabetics in the two older patient groups (65–74 years: RR 0.99; 95% CI 0.91 to 1.08; >74 years: RR 0.96; 95% CI 0.89 to 1.04). Women without diabetes did not have a poorer outcome than men in any of the age groups.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Estimated survival probability by age, gender and diabetic state after myocardial infarction in patients admitted in 1995–2002. (A) ⩽65 years, (B) 65–74 years, (C) ⩾74 years.

Since gender-related survival differences were only found among diabetic men and women in the youngest age group, further analysis focused on these 25 555 patients where 23% were women. Previously diagnosed diabetes mellitus was present among 21% of the women and 16% of the men. Unadjusted baseline characteristics and treatments in relation to gender and the presence or absence of diabetes are shown in table 1. Women with diabetes were more frequently smokers and more frequently had previous hypertension and heart failure compared with men with diabetes and they were less frequently revascularised before admission for their first myocardial infarction. A similar pattern, but at lower rates, was seen among non-diabetic women and men, with the exception of previous heart failure, which was uncommon among patients without diabetes, and smoking, which was more common among non-diabetic women.

Compared with men, diabetic women were more frequently receiving treatment with diuretics and nitroglycerin at the time of hospital admission. During hospitalisation, they were less frequently given intravenous β-blockade and they developed heart failure more frequently than men with diabetes. At discharge, they were more frequently prescribed diuretics and nitroglycerine.

After adjustments for differences in baseline characteristics (including admission ECG) and previous treatment, diabetic women were less frequently given in-hospital treatment with intravenous β-blockade and ACE inhibitors at discharge (fig 2). There were no further significant gender-related differences in the investigated management modalities among diabetic patients. However, there was a non-significant trend (OR<0.90) towards less use of reperfusion therapy, acute revascularisation and β-blockers at discharge in diabetic women.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Adjusted differences in treatment during hospitalisation and at discharge between diabetic women and men aged <65 years after adjusting for the propensity score. The results are shown as odds ratios for diabetic women compared with diabetic men. ACE, angiotensin-converting enzyme.

Figure 3 shows long-term survival by gender in diabetic patients aged <65 years after adjustments in different steps. Following adjustments for discrepancies in baseline characteristics and previous treatment by means of the propensity score and also including the difference in diuretic treatment at discharge, no gender-related difference remained (RR 1.05, 95% CI 0.90 to 1.24) (fig 3B, C). With further adjustments, introducing all in-hospital and discharge treatments into the model, the relative mortality risk (RR 1.03, 95% CI 0.87 to 1.22) remained about the same fig 3D).

DISCUSSION

The most important finding in this report is the poorer outcome after myocardial infarction among young women with diabetes. The previously well-known finding of higher mortality among young women with myocardial infarction may therefore relate to those with diabetes.14^–18 Most of their increased mortality may relate to a risk factor burden that was already higher at admission than that among diabetic men.

Following a series of adjustments, the difference in survival disappeared, eliminating female gender as a separate causative factor. The fact that diabetic women already had a higher risk factor burden at admission may naturally be the result of gender-related vulnerability to risk factors. Further elaboration on this issue is, however, outside the objectives of this report. Compared with diabetic men, previous heart failure and hypertension in particular were more common among diabetic women and previous heart failure was the risk factor that particularly differentiated diabetic women from non-diabetic patients. The presence of heart failure could be explained by an increased prevalence of previous undiagnosed silent ischaemia among diabetic women. Owing to autonomic dysfunction, diabetic patients often have compromised pain perception, causing ischaemic heart disease in general and myocardial infarction in particular to present with atypical symptoms, such as tiredness and breathlessness.19 Atypical symptoms are in fact more common in women than men.20 21 This may increase the number of silent myocardial infarctions and cause misinterpretation regarding the presence of myocardial ischaemia among female diabetic patients. Another explanation may be the existence of diabetic cardiomyopathy.22 Heart failure is two to five times more common among diabetic patients, especially in females, even following the exclusion of those with known coronary artery disease.23

Smoking was surprisingly common among the young diabetic women. A high prevalence of smokers among young diabetic women has previously been described in the Swedish National Diabetes Register (NDR), which also revealed an increased risk of microalbuminuria and less effective glucose control in smokers.24 25 It has been suggested that smoking is a stronger risk factor for myocardial infarction in diabetic women than men.26 The finding that an accumulation of risk factors among young diabetic women may be an explanation for a substantial part of their higher mortality is of great clinical relevance. It must be taken into consideration when planning prevention strategies in both a primary and secondary setting. This position is further strengthened by a recent report from the DECODE study, which showed that risk factors have a higher impact on future cardiovascular mortality in women than in men with newly diagnosed diabetes.27

Differences in treatment during hospitalisation and at discharge were of minor importance for the poorer outcome among diabetic women. It was related to the increased use of diuretics at discharge, which is in turn explained by the higher prevalence of previous-onset and new-onset heart failure. Following adjustments for treatment during hospitalisation and at discharge and for other risk factors, only a few, albeit reasonably important, neglected treatments remained. It is possible to speculate about whether a somewhat more extensive use of a combination of several treatments, all somewhat but not significantly less frequently prescribed for females than males, might have had the potential to improve the prognosis among diabetic women. Considering the higher prevalence of hypertension and heart failure, greater use of ACE inhibitors in this group at hospital discharge might have been expected, for example. The present results are in accordance with the findings from a primary care setting where diabetic women were less likely than diabetic men to receive recommended treatments and to achieve treatment targets for several risk factors.28 Furthermore, in the light of our previous findings of less use of evidence-based treatment after myocardial infarction in patients with diabetes compared with those without,6 7 the present finding of even less use, albeit small, in diabetic women, might be of importance.

There are important study limitations. Several factors that might have influenced the selection of treatment, previous disease and outcome were not recorded in the RIKS-HIA such as serum creatinine, type of diabetes, left ventricle ejection fraction and glucose control and glucose-lowering treatment at admission and during follow-up. In spite of this, the RIKS-HIA has important strengths, including the high quality of the database and the inclusion of large numbers of unselected, consecutively collected patients representative of everyday practice. Furthermore, risk factors in patients dying from myocardial infarction before reaching hospital are missed in this registry analysis and this might introduce a gender-related selection bias. For instance, it has previously been reported that men tend to die more frequently than women before reaching hospital.29 The present report focused on diabetic patients and the observed gender difference among non-diabetic patients above the age of 65, with a survival disadvantage for men compared with women, was not further analysed in this report.

CONCLUSION

Compared with men, young women with diabetes run an increased risk of premature mortality if they had a myocardial infarction. This is explained primarily by an increased risk factor burden, which is already present at admission, and only to a minor extent by less use of evidence-based treatments or female gender. Increased attention to this unfavourable situation is clearly warranted. The present observation makes further study of the impact of improved risk factor management in this particular group of relatively young, easily identifiable, high-risk patients important, together with attempts to initiate treatment and cardiac investigations before their first myocardial infarction or the onset of heart failure.