The Incidence of Congestive Heart Failure in Type 2 Diabetes: An update Free

The study population consisted of 9,591 individuals in whom type 2 diabetes was diagnosed before 1 January 1997 and an age- and a sex-matched comparison group of the same number of individuals without diagnosed diabetes. All subjects were members of Kaiser Permanente Northwest (KPNW), a not-for-profit group-model health maintenance organization serving ∼450,000 members in the vicinity of Portland, Oregon. Individuals in whom CHF was diagnosed as of 1 January 1997 were excluded from the analysis (1,131 subjects with diabetes and 435 subjects without diabetes). We also excluded 229 diabetic subjects who were determined not to have diabetes after review by an endocrinologist and 311 nondiabetic subjects in whom diabetes developed during the follow-up period. These exclusions were applied after the initial one-to-one age and sex matching, so the final study population included 8,231 diabetic patients and 8,845 nondiabetic patients. These 17,076 patients were followed from 1 January 1997 through 31 December 2002 or until they experienced CHF, died, or left the health plan, whichever came earlier.

KPNW maintains comprehensive electronic health care utilization data on all its members. The electronic medical record, in use since 1996, allows the attending clinician to record as many as 20 International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) coded diagnoses at each ambulatory patient contact and up to nine discharge diagnoses for inpatient hospital admissions. An electronic problem list, also coded in ICD-9-CM, is available to the clinician at each contact. In addition, all KPNW laboratory tests are performed by a single regional laboratory, and the results are stored in a searchable electronic database. A pharmacy is located in each medical office, and most members have a pharmacy benefit, helping to ensure complete capture of pharmaceutical dispenses in the KPNW pharmacy database.

We defined CHF as the first occurring inpatient or outpatient diagnosis of CHF. However, if the diagnosis occurred solely in an outpatient setting, we required that the diagnosis be coded in at least two separate occurrences. KPNW clinicians have access to online guidelines for assessment of heart failure to assist in making the diagnosis. The guidelines state that when symptoms (e.g., orthopnea, dyspnea, paroxysmal nocturnal dyspnea, jugular venous distension, or edema) suggest heart failure, a careful examination in conjunction with a chest x-ray and electrocardiography is indicated. Heart failure is diagnosed when there is evidence of valvular dysfunction, systolic or diastolic dysfunction, or pericardial disease. More than 90% of subjects with CHF had undergone echocardiography, electrocardiography, or chest X-ray just before the diagnosis, ensuring that most of our patients probably had definitive CHF (3). The closed group-model structure of KPNW also encourages primary care clinicians to consult both formally and informally with cardiac specialists. Approximately 35% of patients with CHF were seen by cardiologists.

We also used the medical records to determine the presence of ischemic heart disease (ICD-9-CM codes 410.xx–414.xx). Blood pressure and body weight, measured at most outpatient contacts, were averaged over the observation period.

Antihyperglycemic therapy was identified at baseline from pharmacy dispensings. From laboratory data, we identified subjects with microalbuminuria, defined as an albumin excretion rate between 30 and 300 mg/day, or gross proteinuria (albumin excretion rate >300 mg/day or serum creatinine >1.5 mg/dl or 24-h urine protein >165 mg/day). End-stage renal disease (ESRD) was defined as referral for kidney dialysis or transplantation. We did not allow subjects to appear in more than one of the four stages of renal disease (none, microalbuminuria, proteinuria, or ESRD). We also extracted and averaged all HbA_1c values (Diamat high-performance liquid chromatography assay; Bio-Rad, Hercules, CA) available during the observation period (1997–2002 or when patients experienced CHF, died, or disenrolled).

All analyses were performed using SAS statistical software (SAS Institute, Cary, NC). Bivariate associations were tested using Student’s t test for continuous measures and the χ² statistic for dichotomous variables. We used the Kaplan-Meier procedure to estimate the survival curves and Cox regression analysis to calculate the hazard ratios for predictors of incident CHF. To estimate the potential effect of obesity on CHF incidence, we limited the Cox regression analysis to the 79% of subjects (n = 6,496) for whom BMI values were available. Because of the high correlation between systolic and diastolic blood pressures, we used only diastolic blood pressure in the model because of its greater predictive power.

The demographic characteristics of the diabetic cohort and the nondiabetic comparison group stratified by incident CHF are shown in Table 1. There were no statistically significant differences between the two groups in age or sex distribution. In both groups, subjects in whom CHF developed were significantly older than those who did not develop CHF. Patients with diabetes in whom CHF developed were, on average, 5.5 years younger than the nondiabetic subjects in whom CHF developed (P < 0.001). Because follow-up was terminated after CHF was diagnosed, length of follow-up is longer for those who remained free of CHF but was very similar overall for the diabetic and nondiabetic groups.

Figure 1 displays Kaplan-Meier curves of CHF incidence for the diabetic and nondiabetic cohorts. The slope of the curve for the diabetic cohort is steeper, indicating an increased rate of CHF development compared with that for the nondiabetic cohort.

Table 2 displays the overall CHF incident rates per 1,000 person-years of follow-up for the diabetic cohort compared with the nondiabetic cohort, stratified by 10-year age-groups. Overall, diabetic patients experienced CHF at 2.5 times the rate of nondiabetic comparison subjects (30.9 vs. 12.4 cases per 1,000 person-years, P < 0.001). The absolute incidence rate of CHF increased steadily with age for both cohorts. However, the relative rate declined with increasing age, from a high of 11.0 (95% CI 5.6–21.8) for those younger than 45 years to 1.8 (1.6–2.2) for those aged 75–84 years.

Diabetic patients who developed CHF differed from patients who did not on a number of clinical characteristics (Table 3). Patients with incident CHF had longer duration of diabetes at baseline (6.1 vs. 4.4 years, P < 0.001) and were more likely to have ischemic heart disease (38.0 vs. 16.4%, P < 0.001) and more advanced renal disease (e.g., 29.8 vs. 18.2% had gross proteinuria, P < 0.001). Glycemic control (HbA_1c) and blood pressure were statistically but not clinically significantly different between those who did and those who did not develop CHF. BMI did not differ. Patients who experienced CHF were more likely to be using insulin (33.2 vs. 18.8%, P < 0.001) and metformin (15.9 vs. 13.3%, P = 0.020) at the start of follow-up.

The results of using Cox proportional hazards modeling to predict CHF incidence among the diabetic patients are shown in Table 4. Age at baseline was the strongest predictor of time to CHF, increasing the hazard by 40% for every 5 years of age (95% CI 1.35–1.45). The presence of ischemic heart disease more than doubled the risk of CHF (hazard ratio 2.36, 95% CI 2.06–2.69). Greater BMI (1.12 per 2.5 units of kg/m², 1.09–1.15) and poorer glycemic control (1.32 per HbA_1c percentage point, 1.23–1.41) were also strong predictors of more rapid development of CHF. Microalbuminuria seemed to reduce the risk of CHF (0.78, 0.65–0.93), but more severe renal disease increased the risk of CHF. Longer duration of diabetes and greater diastolic blood pressure also increased the hazard of experiencing CHF. Use of insulin was associated with an increased risk of CHF, but use of sulfonylureas or metformin was not. Female sex was not a significant predictor of CHF incidence.

Our results, based on 8,231 patients of all ages and stages of diabetes followed for up to 6 years, indicate that the incidence of CHF in diabetes is 3–15 times greater than the previously reported 2–10 cases per 1,000 subjects (7,10–15). Published estimates of the incidence of CHF in patients with diabetes have relied on relatively small or unrepresentative samples, have been based on only patients with more severe CHF, or have been limited by relatively short follow-up (3,7,10–15). These limitations may have led to underestimations of CHF in patients with diabetes. In addition, our results may arise, in part, from increasing rates of CHF over time that are related to increased survival of patients with diabetes. The most recent Framingham data contain follow-up through 1996 (14), and other estimates of CHF are based on data from similar or earlier time periods. The current results include follow-up through 2002—data that are more contemporary than previous studies (7,13,14). Finally, comparisons of studies with differing mean ages are especially difficult in age-dependent diseases such as CHF, which may account for some of the difference we report.

Although CHF incidence rates increase with age and are higher in diabetic patients than in nondiabetic subjects across all age-groups, the difference in CHF rate ratios decreases with increasing age. Therefore, it seems that diabetes not only increases the risk of CHF but also accelerates its occurrence. These results are consistent with a similar finding we previously reported across all types of cardiovascular disease (16) and with a recent study demonstrating that patients in whom diabetes is diagnosed at younger ages are at much greater risk for cardiovascular disease (17).

Our previously published multivariate model of CHF incidence over a 30-month follow-up period included both lower baseline HbA_1c and a reduction in HbA_1c at follow-up as predictor variables (3). In the current study, using the same population but with a longer, 72-month follow-up period, higher average HbA_1c emerged as a strong predictor of CHF incidence. Although we again found that patients with CHF had a lower average HbA_1c value before CHF than at baseline (data not shown), it was not a significant predictor in the new multivariate model. However, when HbA_1c measurements were averaged over the entire observation period, higher values were predictive of CHF. These results suggest that cumulative glycemic burden, rather than glycemic control at a given point in time, is the more important CHF risk factor, underscoring the importance of maintaining good glycemic control throughout the course of diabetes. This is also consistent with results from the UKPDS, which used an updated mean of annual measurements to demonstrate a similar relationship between HbA_1c and risk of heart failure (11).

The relationships between diabetes, heart disease, and obesity are well established. However, even within a study population that had an average BMI >32 kg/m², we found that each additional 2.5 BMI units increased CHF risk by 12%. This is probably due to the increased insulin resistance and hypertension associated with obesity and the corresponding risk of cardiovascular disease (18–20). Obesity rates in the U.S. are continuing to increase (21). Our results suggest that a corresponding increase in CHF can be expected even if other risk factors are well managed.

Microalbuminuria is recognized as a risk factor for cardiovascular disease in diabetes (22,23) and has been shown to be a risk factor for CHF as well (12). In our model, microalbuminuria seemed to reduce the risk of CHF. However, because patients with microalbuminuria are immediately treated with ACE inhibitors in the KPNW system, it may merely be serving as a marker for more aggressive treatment, and the apparent reduced risk may be representing the beneficial effects of ACE inhibitors on CHF. As renal dysfunction became more severe, the risk of CHF increased considerably (25% for gross proteinuria and 54% for ESRD). Whether impaired renal function is an early manifestation of preclinical CHF or is itself a cause of CHF cannot be established from these data. Nevertheless, impaired renal function seems to be an “early warning sign” for possible impending CHF development.

The UKPDS demonstrated that tight blood pressure control reduces macrovascular disease risk in type 2 diabetes (24). In our data, elevated blood pressure was predictive of incident CHF, suggesting that the benefits of blood pressure control might also extend to risk of CHF.

One important limitation of the current study is the use of electronic data to define CHF, which may result in our overstating the true incidence of CHF. However, we used methods that mirrored those recently reported by McCullough et al. (25), in which 83% of automated ICD-9-CM diagnoses of CHF were validated with medical record review. We previously found that 92.5% of our CHF patients were diagnosed after echocardiography, electrocardiography, or chest X-ray (3). Additionally, 82% of our CHF cases were found in the inpatient medical record, where diagnoses are more likely to be objectively confirmed. Unlike another study in a similar setting (13), we allowed the inpatient CHF discharge diagnosis to occur in any position, not just the first listed. Finally, just 18% of our cases were found only in the outpatient medical record, alleviating any concern that primary care clinicians may be coding CHF too aggressively in the medical office.

In our previous report, use of insulin and oral agents was strongly predictive of CHF incidence (3). In line with our previous findings, use of insulin at the start of follow-up was predictive of CHF. However, we did not find use of sulfonylurea or metformin to be predictive of CHF in the present study. Controversy surrounding the negative cardiac effects of sulfonylureas remains, although our findings are consistent with the UKPDS, which did not find increased mortality in the sulfonylurea-treated group (15). Studies specifically designed to estimate the potential effects of antidiabetic agents on CHF incidence, accounting for the high rate of antidiabetic medication switching and addition in the diabetic population, are needed. The choice of antidiabetic medication is strongly correlated with glycemic control, making it very hard to separate the independent effects of each of these factors on the risk of CHF development.

The high incidence of CHF among the diabetic population emphasizes the need for early recognition and aggressive treatment of modifiable risk factors for CHF. Our results confirm that tight glycemic and blood pressure control as well as weight loss and treatment of renal disease are essential to the long-term health of people with diabetes. Given the substantially larger CHF rate ratios and the greater life expectancy of patients younger than 65 years, younger persons may benefit most from intensive risk factor modification.

This research was sponsored by GlaxoSmithKline, Inc.