Prospective Analysis of Mortality, Morbidity, and Risk Factors in Elderly Diabetic Subjects: Nagano Study Free

During April and August 1998, patients were registered at the outpatient clinic of Koshoku-Chuo Hospital, Asama General Hospital, Nagano-Chuo Hospital and Shinshu University Hospital, in Nagano-prefecture, Japan. A diagnosis of type 2 diabetes was based on the World Health Organization criteria (15). Those older than 65 years and attending each hospital for longer than 1 year were consecutively registered. The latter was included into the enrollment criteria to minimize the dropout during the study. The study was initiated on 1 September 1998 and closed on 31 August 2001. Patients on hemodialysis or with clinically overt, uncured malignancy were excluded. The study is community based; however, in Japan, there is no clear-cut distinction between a community hospital, a primary care unit, and a referral center. The patients were all ethnic Japanese. A total of 413 patients were registered: 206 at Koshoku-Chuo Hospital, 97 at Asama General Hospital, 66 at Nagano-Chuo Hospital, and 44 at Shinshu University Hospital. Of the patients, 23 (5.6%) dropped out during the follow-up; the data of 390 (94.4%) patients were used for analysis. Major clinical characteristics of the patients who dropped out and the analyzed patients were not significantly different (data not shown).

Upon registration, measurement of postprandial (2–4 h after breakfast) PG, HbA_1c, and serum chemistry including blood urea nitrogen, creatinine, total cholesterol, HDL cholesterol, triglycerides, and urinalysis was performed and an electrocardiogram taken. Blood samples for lipid measurement were obtained before breakfast. The serum LDL cholesterol was determined by Friedewald’s calculation (16). BMI was calculated as kg/m². Urinary excretion of albumin was determined by a quantitative method (17) using a random sample (>30 mg/g creatinine regarded as positive). The presence of micro- and macroangiopathies was judged as reported (18,19). In brief, the retinae were examined by ophthalmologists after dilation of the pupils, and nonproliferative and proliferative changes were separately recorded. When macroproteinuria (positive for protein by a conventional dipstick method) was present, diabetic nephropathy was judged to be present. Microalbuminuria alone was not taken as a basis for nephropathy because of low specificity of it as a marker of diabetic nephropathy in elderly diabetic subjects. A diagnosis of diabetic neuropathy was made on a clinical basis (18,19). A history of malignant disorders, myocardial infarction, and stroke was registered. For these events, only those necessitating hospitalization to medical facilities were taken into account. The duration of diabetes, family history of diabetes within third-degree relatives, smoking, and the past maximum body weight were also recorded. The data of the general population in Nagano-prefecture were obtained from the Nagano Municipal Health Center and the data of the general population of all of Japan from the official publication of the Ministry of Health and Welfare, Japan.

We treated the patients with the following guidelines. A recommendation to reduce caloric intake was made to those with a BMI ≥25 kg/m². A recommendation for mild exercise (such as walking for 30 min/day) was made to those without neuromuscular and arthritic problems. HbA_1c <7.0% was targeted if hypoglycemia necessitating third-party assistance or medical intervention did not develop. HbA_1c <6.5% was targeted if no hypoglycemia developed. A blood pressure <140/70 mmHg for those <69 and 145/80 for those >70 years of age was targeted if symptoms, such as dizziness related to the use of antihypertensive drugs, did not develop. The goal was total cholesterol <240 mg/dl. We explained to the patients that tight control of glycemia and blood pressure was recommended and most likely beneficial in the elderly but the value of it was not firmly established. Then, if the patient did not wish to receive treatment with the above-mentioned guidelines, the treatment targets were loosened. Also, when the patient could not expect third-party assistance if hypoglycemia developed (such as living alone), or the patient’s understanding of and/or reaction to hypoglycemia were considered insufficient, glycemic control by pharmacological agents was relaxed.

In our group, the above-mentioned guidelines were utilized with elderly patients with diabetes well before the study. Therefore, there was no implementation of new intervention at the beginning of the study. The patients were followed for 3 years with the same guidelines by M.K., M.N., T.K., and T.A. Outpatient clinic visits were every 4 weeks. The patients were directly seen and examined by the doctors at each visit, and GHbA_1c, postprandial PG, blood pressure, and body weight were measured.

Two end points were defined. One was death and the other was a diabetes-related aggregate end point included visual loss, end-stage renal failure necessitating hemodialysis, fatal and nonfatal myocardial infarction, angina pectoris, fatal and nonfatal stroke, amputation of an extremity or digit, and sudden death.

HbA_1c (normal range 4.5–5.8%) was determined by high-performance liquid chromatography in the clinical laboratory of each hospital. By comparison with the standard provided by the Japan Diabetes Society Glycohemoglobin Standardization Program (20), the maximum interlaboratory variation was 0.2%.

Cox proportional hazards model, the χ² analysis, logistic regression analysis, and one-way ANOVA were used (StatView; SAS, Cary, NC) and P < 0.05 was considered significant. Occurrence of the end points was also analyzed as a function of the quartile of the baseline variables as needed.

As shown in Table 1, the mean age, duration of diabetes, HbA_1c and BMI were 73.0 and 15.2 years, 6.8%, and 23.4 kg/m², respectively. A total of 332 patients (85%) were receiving oral hypoglycemic agents and/or insulin. In insulin-treated patients (n = 152, 39%), the mean dose was 25 units/day (0.44 units · kg⁻¹ body wt · day⁻¹). Of the patients, 294 (75%) and 108 (28%) had micro- and macroangiopathic complications, respectively. The mean blood pressure was 136/74 mmHg with 219 (56%) receiving antihypertensive agents, mostly Ca²⁺ channel blockers and ACE inhibitors (Table 1).

During the 3-year follow-up, 34 patients died. Major causes of death were malignant disorders (n = 12, 35%), strokes (n = 9, 26%), and myocardial infarctions (n = 4, 12%). The mortality rate of the study population was 8.7% per 3 years (2.9% per years), which was comparable to that of the age- and sex-matched population of Nagano-prefecture, 2.8% per year, or that of the age- and sex-matched total Japanese population, 3.2% per year. During follow-up, 42 patients experienced diabetes-related end points, of which 16 were fatal and 26 nonfatal. Two patients died due to sepsis and pneumonia, respectively, after full recovery from nonfatal diabetes-related events. Diabetes-related fatal events were nine strokes, four myocardial infarctions, two sudden deaths, and one cachexia associated with hemodialysis. Nonfatal events included 16 strokes, 3 myocardial infarctions, 1 angina pectoris, 2 incidents of visual loss, and 5 incidents of end-stage renal failure. One patient experienced nonfatal cerebral hemorrhage after placement of hemodialysis. Baseline characteristics of each subgroup of patients are also shown in Table 1.

Relative risk (RR) of each baseline characteristic for death and occurrence of the diabetes-related end point and its significance are shown in Table 2. By the univariate Cox proportional hazards model, male sex, age, high serum creatinine (Scr), greater ΔBMI (reduction from the maximum BMI), history of macroangiopathy and stroke, and current smoking status were significant risks. Significant risk reduction was observed in nonsmokers. Microalbuminuria and macroproteinuria, respectively, were progressively stronger risk factors. Among them, only high Scr and prior stroke were highly significant (P < 0.0001) and strong (RR >3.0) risks. As a quantitative index of renal dysfunction, high Scr was employed in the following analyses. Multivariate analysis by Cox proportional hazards model was performed with incorporation of 10 variables. The six variables listed above (sex, age, Scr, prior stroke, delta BMI, and current smoking status) were taken. Duration of diabetes, systolic blood pressure (SBP), PG, and HbA_1c were additionally incorporated into the analysis because these variables were identified as risks in previous studies of middle-aged or elderly patients with diabetes (10,11,21–26). In this analysis, only high Scr (RR 2.988, P = 0.0002) and prior stroke (RR 4.587, P < 0.0001) were identified as significant, independent risk factors. It should be noted that the number of events per variable was too small to avoid overadjustment (27). Therefore, the data of the multivariate analysis should be interpreted with a caution. Current smoking was associated with significantly higher mortality rate (3.5 times higher than in non- or exsmokers) in those with both nephropathy and prior stroke.

By the univariate Cox proportional hazards model, age, greater ΔBMI, high PG, high Scr, presence of microangiopathy, background retinopathy and neuropathy, and history of macroangiopathy and stroke were significant risks for the diabetes-related end point (Table 2). Microalbuminuria and macroproteinuria, respectively, were progressively stronger risk factors. Significant risk reduction was observed with greater BMI and a nonsmoking status. Here again, only high Scr and prior stroke were highly significant (P < 0.0001) and strong (RR >3.0) risks. By multivariate analysis, only age (RR 1.115, P = 0.002), high Scr (RR 3.278, P < 0.0001), and prior stroke (RR 3.784, P < 0.0001) were identified as risk factors. However, the small number of events per variable in the multivariate analysis precludes to draw a definitive conclusion. Hyperglycemia was not a risk by the univariate Cox proportional hazards model even when the microangiopathic events (visual loss and end-stage renal failure, n = 7) were employed as an aggregate end point.

History of stroke at baseline was an independent strong risk factor for the two end points (see above), and differential effects of glycemia or blood pressure in those with and without prior stroke were well anticipated. Therefore, the patients with and without prior stroke were separately analyzed. An important finding was obtained for distribution of the diabetes-related end point as a function of the SBP quartile. The distribution was J-shaped in those without prior stroke (Fig. 1A). The incidence was lowest at the 2nd quartile (SBP 126–136 mmHg) and highest at the 4th quartile (SBP ≥148 mmHg). Although a higher incidence of the events in the 1st quartile was not statistically significant in this J-shaped distribution, it was a sum of apparently heterogeneous distributions. Namely, in those receiving antihypertensive agents, the distribution was U-shaped (Fig. 1B). The nadir was at the 3rd quartile (SBP, 137–147 mmHg) and the peak at the 1st quartile (SBP ≤125 mmHg). The incidence at the 2nd quartile was almost the same as the 3rd quartile, and that in the 4th quartile was slightly lower than that in the 1st quartile. In this subgroup, the incidence of the diabetes-related end point was significantly higher in the 1st quartile than in the 3rd quartile. Whereas in those without prior stroke and not receiving antihypertensive agents, such U-shaped distribution was not found. (Fig. 1C). In the patients with prior stroke, incidence of the diabetes-related end point was lowest at the 4th SBP quartile (Fig. 1D). However, the four values in Fig. 1D were not significantly different from each other.

Mortality rates as a function of the SBP quartile were not significantly different irrespective of the presence or absence of a prior stroke (data not shown). Similarly, mortality rates and the incidence of diabetes-related end point as a function of other baseline characteristics (duration of diabetes, PG, HbA_1c, serum lipids, diastolic blood pressure, and mean blood pressure) were not significantly different (data not shown).

Hypoglycemia necessitating third-party assistance occurred 10 times in the entire cohort during the 3-year study period (0.9 per 100 person per year). None of the hypoglycemic events was directly related to mortality or a diabetes-related end point. In 235 randomly selected patients (60% of the study population) HbA_1c levels did not significantly change during follow-up. The mean (±SD) values upon entry and at the end were, in fact, identical (6.8 ± 1.1%).

Increased mortality in patients with diabetes in general (28–31) and in an elderly subgroup (21,32) has been well established. In contrast, the mortality rate of the current cohort was no more than that of an age- and sex-matched general population. It should be noted that the cohort consists of patients who were not particularly at low risk. We excluded only those on hemodialysis and uncured malignancy. Due to the observational nature of our study, we could not prove a cause-result relationship. More precisely, the breakdown was the same between the diabetic subjects and the general population for age and sex only. Nevertheless, we consider the normal mortality rate of the cohort to be attributable to the intensive treatment. In this study, risk factors were more intensively analyzed than outcome events to determine the residual risks after intensive glycemic and blood pressure control in elderly diabetic subjects. Under this condition, only renal dysfunction and prior stroke were highly significant and strong risks for mortality and a diabetes-related end point.

We found that Japanese patients with type 2 diabetes have clearly different sets of causes of mortality and morbidity compared to their Caucasian counterparts (2). Namely, stroke was much more common than myocardial infarction in Japanese patients. It may be due to genetic factors, but it also may be related to differences in weight in the two populations. The rate for malignancy (35% of the deaths) was higher than that seen in Caucasian patients with diabetes (2,33). This is most likely a reflection of a much lower rate for cardiac events (12%) in our cohort. In another study of the Japanese elderly population at large, the rate for malignancy was 38 and 24% of the deaths in men and women, respectively (34). Rosenthal et al. (33) performed a prospective analysis of risk factors for mortality in 137 elderly patients with diabetes and concluded that depression was the best predictor of mortality. They did not incorporate prior stroke into the analysis and we did not perform psychiatric evaluation of the patients. Those with prior stroke might have an increased prevalence of depression. If so, the results of our study and the study of Rosenthal et al. (33) may indicate a similar, if not identical, trend of risk factor for mortality.

In patients without prior stroke receiving antihypertensive agents, the incidence of diabetes-related end point as a function of SBP quartile was U-shaped. In those not receiving antihypertensive agents, such U-shaped distribution was not found. The finding suggests that a pharmacologically induced lower blood pressure (SBP ≤125 mmHg) is not a good sign. The high event rate, albeit not statistically significant, in the subjects with prior stroke who have lower SBP currently, is also worrisome. Antihypertensive treatment and the J-curve has been debated (35). Nevertheless, in the past, there was no prospective study evaluating benefit or risk of antihypertensive treatment in elderly diabetic subjects.

Glycemia in the cohort (the mean HbA_1c, 6.8%) may be sufficiently low so that it was not a risk any longer. The study was not to compare tight and poor control. Nonetheless, it was rather unexpected that hyperglycemia was not a risk for any end points in the face of a relatively large difference in the mean HbA_1c values between the highest and lowest quartiles, 8.3% and 5.6%, respectively. The patients at the 4th HbA_1c quartile had a longer duration of diabetes than those in the 1st quartile (18.0 ± 9.9 vs. 11.9 ± 9.3 years, P < 0.0001). Upon entry, the former group had a higher prevalence of diabetic microangiopathy than the latter (85% vs. 64%, P = 0.007), and the use of hypoglycemic agents and insulin was more frequent in the former than in the latter (all hypoglycemic agents 97% vs. 58%, insulin 48% vs. 14%, respectively, each with P < 0.0001). The fact that such a seemingly high-risk group did not experience an increased incidence of end points suggests that hyperglycemia, in the range and duration observed in the current study, contributes little to morbidity and mortality.

For attaining low glycemic control with fewer hypoglycemic events, we consider frequent contacts of the doctor and patient, such as once a month (as in the current study), important. However, to prove this hypothesis, a control study is clearly needed.

In conclusion, we found a normal mortality rate in intensively managed elderly diabetic subjects. A randomized control study should definitively decide the optimal levels of control and establish the risk, if any, of too much lowering of blood pressure.

We are indebted to Drs. Masao Tanaka, Yasunori Itakura, Miyuki Katai, and Tsuyoshi Inagaki for collecting the data and to Drs. Matthew C. Riddle and Monte A. Greer for editorial assistance.