Tseng C. Exogenous Insulin Use and Hypertension in Adult Patients With Type 2 Diabetes Mellitus. Arch Intern Med. 2006;166(11):1184-1189. doi:10.1001/archinte.166.11.1184
Endogenous hyperinsulinemia, along with insulin resistance, is associated with hypertension. This study investigated the link between exogenous insulin use and hypertension in patients with type 2 diabetes mellitus.
Using national health insurance records in Taiwan, data from 87 850 adult patients with type 2 diabetes mellitus were collected cross-sectionally with retrospective recall for onset of events. Data were analyzed to evaluate the strength of association, consistency, dose-response relationship, and temporality between exogenous insulin use and hypertension.
There were 5927 insulin users, who were characterized by being 1 year older in age, female preponderance, longer duration of diabetes, slightly lower body mass index, and less smoking but higher prevalence of hypertension with higher blood pressure. After adjustment for age, sex, body mass index, duration of diabetes, smoking, and parental hypertension, the odds ratios (95% confidence interval [CI]) for hypertension for patients using insulin for less than 5 years, 5 to 9 years, and 10 years or more were 1.14 (95% CI, 1.06-1.23), 1.35 (95% CI, 1.18-1.54), and 1.46 (95% CI, 1.24-1.74), respectively, compared with nonusers. In insulin users who did not have hypertension at the start of insulin use, the respective odds ratios for those using insulin for 5 to 9 years and 10 years or more were 1.5 (95% CI, 1.25-1.80) and 2.15 (95% CI, 1.72-2.70), respectively, compared with those using insulin for less than 5 years.
In patients with type 2 diabetes mellitus, insulin users are at higher risk for development of hypertension. These observational data suggest the need for further study of the relationship between exogenous insulin use and hypertension.
Endogenous hyperinsulinemia is an early marker of insulin resistance and has been demonstrated to be a risk factor for hypertension,1,2 decreased responsiveness to antihypertensive medications,3 and morbidity and mortality associated with atherosclerotic disease.4- 7 Although not all studies consistently show a positive association between endogenous hyperinsulinemia and cardiovascular disease, a recent meta-analysis based on 17 prospective studies concluded that endogenous hyperinsulinemia is a significant, though weak, predictor for cardiovascular disease and that the relationship seems to be stronger in white populations than in nonwhite populations.8
The use of insulin injection is necessary for good control of blood glucose levels in some patients with type 2 diabetes mellitus (T2DM), especially when diabetes progresses to a later stage with pancreatic failure. However, the benefit of lowering blood glucose levels with exogenous insulin may be counteracted by the potential risk for insulin-induced atherosclerosis. Exogenous insulin treatment in patients with T2DM has been shown to be associated with cardiovascular morbidity and mortality in some observational studies9,10 but not in others.11,12 Randomized controlled trials evaluating the risk for cardiovascular disease or hypertension after long-term exogenous insulin treatment are lacking. However, some clinical trials, such as the University Group Diabetes Program13 and the United Kingdom Prospective Diabetes Study,14 addressing the effect of glycemic control on vascular morbidity and mortality in patients with T2DM, neither support nor reject the hypothesis that insulin therapy is associated with atherosclerosis. It is possible that the benefit of rigid blood glucose control by insulin might have been counteracted by the atherosclerotic effect of insulin. Furthermore, atherosclerotic disease may take decades to develop, and most clinical trials or observational studies might not have enough power to detect a significant risk associated with the use of exogenous insulin during the period of study.
Hypertension is a well-recognized risk factor for atherosclerosis and is expected to be present years before the clinical manifestations of vascular disease. Therefore, the association between insulin treatment and hypertension is expected to be more easily detected than the association between insulin treatment and atherosclerotic disease. Endogenous hyperinsulinemia resulting from insulin resistance can be a risk factor for the development of hypertension, and insulin has been shown to have a variety of hemodynamic effects15- 18 and effects as a growth factor on the vasculature,19 favoring the development of hypertension. However, whether the common use of exogenous insulin in the control of blood glucose levels in millions of patients with diabetes can cause hypertension is an interesting concept, with clinical importance, but has not been well investigated. Therefore, the purpose of this study was to investigate whether exogenous insulin use for control of blood glucose levels can be linked to the development of hypertension in patients with T2DM by analyzing data obtained from a national sample derived from national health insurance (NHI) records.
The study was approved and supported by the Department of Health of Taiwan. Because more than 96% of the population of Taiwan is covered by the compulsory and universal NHI (exceptions are those involved in military service and those incarcerated), almost all patients with diabetes have been using NHI.20,21 Therefore, use of the clinical setting databases of claims against NHI is appropriate to derive a national sample of patients with diabetes. The assembly of such a national sample has been described in detail.20,21 As a result, 256 036 patients with diabetes using NHI from March 1, 1995, through December 31, 1998, were identified from 66 hospitals and clinics throughout Taiwan. The clinical settings were selected under the consideration of homogeneous geographic distribution and inclusion of all levels of medical settings. To create a cohort of 90 000 patients for long-term follow-up, 128 572 of 256 036 patients were randomly selected, supposing a response rate of 70%.
From March 1, 1995, to April 30, 2002, well-trained interviewers conducted a telephone survey using a structured questionnaire. Researchers tried as many as 3 times to reach subjects before excluding them. To obtain complete information from the patients and to keep missing data to a minimum, the interviewers handed in the completed questionnaires every week. All returned questionnaires were checked by an assistant and double-checked by the investigator (C.-H.T.).
Data were collected cross-sectionally with retrospective recalls to the onset of events. The information obtained from questionnaire interview for analyses in this study included patient age, sex, height, and weight; parental history of hypertension; symptoms at onset of diabetes; treatment method for the differentiation of types of diabetes; the year of diabetes diagnosis; personal history of hypertension and smoking; and usual systolic blood pressure (SBP) and diastolic blood pressure (DBP). If a patient reported a diagnosis of hypertension or was using insulin for blood glucose level control, further questions about the time of hypertension diagnosis or the start of insulin use were asked.
The classification of type 1 diabetes mellitus (T1DM) was based on 1 of the following 2 criteria: diabetic ketoacidosis at the onset of diabetes mellitus or requirement of insulin injection within 1 year of diagnosis of diabetes. Patients not diagnosed as having T1DM were considered to have T2DM. Patients identified as having T1DM, being younger than 18 years, or not reporting usual blood pressure were excluded from the study. A total of 93 484 patients (response rate, 72.7%) completed the interview. After excluding 3528 patients with T1DM, 89 956 patients were identified as having T2DM. Exclusion of those younger than 18 years yielded 89 857 patients. Further exclusion of patients who did not report their usual blood pressure yielded 87 850 patients for analysis. Among them, 5927 patients reported that they were using insulin injections for blood glucose level control.
Duration of diabetes was calculated as the patient's age at the time of the interview minus age at onset of diabetes. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Hypertension was defined by a positive history or by elevated blood pressure (SBP ≥140 mm Hg or DBP ≥90 mm Hg) in patients without a history of hypertension.
The baseline characteristics between users and nonusers of insulin were compared using the t test for continuous variables and the χ2 test for categorical variables. In the analyses of the relationship between insulin use and hypertension, the other variables were treated as potential confounders. In users of insulin, the duration of insulin use was subdivided into less than 5 years, 5 to 9 years, and 10 years or more. Age, duration of diabetes, and BMI were classified into 2 subgroups using the following cutoff points, respectively: 65 years for age, 20 years for duration of diabetes, and 25 for BMI. A BMI of 25 or more was defined as obesity according to the recommendation for Asian populations.22 Hypertension prevalence for nonusers and users of insulin for duration of insulin use was calculated for the different subgroups of potential confounders, and the χ2 test and linear test for trend were performed. Blood pressure was categorized as SBP less than 130 mm Hg, 130 to 139 mm Hg, and 140 mm Hg or higher; and DBP as less than 80 mm Hg, 80 to 89 mm Hg, and 90 mm Hg or higher. The distributions of SBP and DBP for insulin use were tested using the χ2 test. Logistic regression was used to estimate the odds ratios and their 95% confidence intervals for hypertension. Models were created unadjusted, adjusted for 1 potential confounder one at a time, and adjusted for all confounders simultaneously. To ensure correct temporality of insulin use before the development of hypertension, all analyses were also performed separately in patients who used insulin but did not have hypertension at the start of insulin use.
A Cox proportional hazards regression model was created to compare the relative risk for development of hypertension between 2 groups of patients: those who did not use insulin and were free from hypertension at the time of diabetes diagnosis and those who used insulin but did not have hypertension at the start of insulin therapy. The other independent variables entered into the model simultaneously for adjustment were age, sex, BMI, duration of diabetes, smoking, and parental hypertension. The age and duration of diabetes entered into this model were calculated to the respective values at the beginning of follow-up.
Table 1 compares the characteristics of those patients using and not using insulin. Except for parental hypertension, all of the variables differed significantly. Users were characterized by being 1 year older in age, preponderance of women, longer duration of diabetes, slightly lower BMI, fewer smokers, higher SBP and DBP, and higher prevalence of hypertension.
Because of the study design and the manner of data collection, blood pressure at the time of starting insulin treatment were unavailable for all insulin users. One might suspect that patients using insulin might be in a prehypertensive state (even if the blood pressure was within normal limits) and might have a higher prevalence of hypertension and higher SBP or DBP at the start of insulin treatment than those not using insulin. Therefore, secondary analyses were carried out to compare the baseline characteristics between patients not using insulin and insulin users with a duration of insulin therapy of less than 1 month (n = 489). Results showed that in insulin users compared with nonusers, except for a significantly higher preponderance of women (58.3% vs 53.6%) and longer duration of diabetes (10.7 ± 7.3 years vs 6.6 ± 6.1 years), none of the other characteristics, including age, BMI, smoking, hypertension, SBP, DBP, and parental hypertension, differed significantly.
Table 2 gives the prevalence of hypertension in insulin users and nonusers for different subgroups of potential confounders in all patients and in patients using insulin but without hypertension at the start of insulin use. A significant P value (<.001) for the trend test or χ2 test was observed for each subgroup comparison.
Table 3 gives the blood pressure control status for insulin use in all patients and in patients using insulin but without hypertension at the start of insulin use. With increasing duration of insulin use, a higher proportion of patients would have poorer blood pressure control. Table 4 gives the odds ratios derived from the various logistic models. A dose-response relationship between duration of insulin use and hypertension was observed. Table 5 gives the relative risk for the development of hypertension for the various risk factors. The use of insulin was significantly associated with a 1.45-fold higher risk.
The findings consistently favored the hypothesis that use of exogenous insulin might induce hypertension, independent of the effect of potential confounders, based on the strength of association, dose-response relationship, consistency in different analyses, correctness of temporality, and biological plausibility. The strength of association showed an approximately 2-fold increase in odds ratios in patients using insulin for more than 10 years compared with nonusers when analyzed in all patients in most of the logistic models. It improved to an approximately 3-fold increase in odds ratios in patients using insulin for 10 years or more compared with those using insulin for less than 5 years when analyses were performed in patients using insulin and the potential confounding effect of hypertension at the start of insulin use was excluded (Table 4). Furthermore, a dose-response relationship between duration of insulin use and prevalence of hypertension (Table 2), blood pressure control (Table 3), and the odds ratios (Table 4) was consistently observed in different analyses. Temporal correctness could be demonstrated with the Cox proportional hazards regression model, which showed a 1.45-fold higher risk for development of hypertension in patients using insulin but without hypertension at the start of insulin use vs patients not using insulin and without hypertension at the time of diagnosis of diabetes (Table 5).
Patients receiving exogenous pharmacologic dosages of insulin for the control of blood glucose levels may be repetitively exposed to high concentrations of insulin. This excess of insulin may exert detrimental effects on the vascular system, leading to elevated blood pressure. Insulin is vasoactive in the peripheral vasculature and has been suggested to be atherogenic by virtue of its growth factor–like activity.19 It has been shown in earlier studies in animals that insulin treatment can induce atheroma23 and that insulin deficiency may protect against the development of atherosclerosis, which effect is lost with insulin treatment.24 Although this could not be similarly demonstrated in a later animal study,25 it should be noted that findings from animal studies might not be readily extrapolated to human beings.
Although insulin may have a vasodilatory effect, this effect seems to be active in the microvasculature of the skeletal muscle, which is an important determinant of insulin-mediated glucose uptake.15,16 Recent in vivo studies showed that insulin exerts a different effect on larger arteries18 and can cause elevation in blood pressure17 in healthy humans. Arcaro et al18 demonstrated that exogenous insulin infusion in healthy subjects could cause endothelial dysfunction and abrogate endothelium-dependent vasodilation in the large conduit arteries, probably through mechanisms involving the induction of oxidative stress. In the presence of insulin resistance, as in patients with T2DM, the effect of insulin on the microvasculature in the skeletal muscle is blunted.15,16 Therefore, the effect of excessive insulin from exogenous sources can be shifted from the nutritive to the nonnutritive vessels such as the large conduit arteries, as observed by Arcaro et al,18 leading to elevated blood pressure and hypertension.
The study by Kern et al17 provided direct evidence for the development of insulin-induced hypertension. They clearly demonstrated that insulin could significantly increase SBP at either a low insulin infusion rate (1.5 mU/kg per minute) or a high insulin infusion rate (15 mU/kg per minute). Although the plasma catecholamine levels significantly increased with the high rate of infusion, they were not changed significantly at the low rate of infusion. Therefore, insulin can directly exert effects on the development of hypertension in humans. However, other mechanisms cannot be excluded. For example, insulin has been shown to exert the effects of sodium retention,26 ion transport,27 stimulation of sympathetic nerves,28 and induction of oxidative stress.29 All of these could also contribute to the pathogenesis of hypertension after prolonged use of insulin.
Whether insulin can be atherogenic independent of insulin resistance remains to be confirmed. A recent study suggested that insulin is atherogenic. Muis et al30 demonstrated that the cumulative dosage of insulin in patients with T1DM with a mean duration of diabetes of 20.7 years is closely associated with carotid intima-media thickness, a marker of generalized atherosclerosis. This association was not materially changed even after adjustment for potential confounders related to insulin resistance, suggesting that prolonged treatment with exogenous insulin can be atherogenic. Therefore, the hypothesis linking insulin, hypertension, and atherosclerosis is probably applicable to both T1DM and T2DM, with or without the presence of insulin resistance.
Selection bias could have resulted from telephone interviews because patients without a telephone would be excluded from the study. This problem is probably not of great effect in Taiwan because the rate of telephone ownership is high. According to the Taiwan Statistical Data Book, published by the Council of Economic Planning and Development of Taiwan, the telephone subscription rate is as high as 58.2 per 100 persons and the mobile telephone subscription rate is more than 100 per 100 persons.31
The study has other limitations. First, it considered only the duration of insulin use as a surrogate marker for elevated exposure to insulin in the blood without considering the dosage of insulin used or measuring the blood insulin concentration. Second, although no statistical differences in hypertension prevalence or SBP and DBP were found between patients who had used insulin for less than 1 month and nonusers in secondary analyses, it is possible that prehypertension or higher blood pressure rates were present at the start of insulin treatment in insulin users compared with nonusers, owing to more severe insulin resistance in users. However, even if this were the case, it would probably not change the conclusions because a dose-response relationship was observed in users of insulin for different durations (Tables 2, 3, and 4). Third, recall bias is an inherent problem associated with questionnaire interviews. However, since the bias was probably not differential, this would only underestimate the risk. Fourth, blood pressure was not measured. However, this limitation did not change the conclusions because the results were similar if hypertension was defined by a positive history without considering the reported blood pressure in secondary analyses (data not shown). Fifth, it should be recognized that the study was conducted in patients with diabetes and that the results may not be extended to subjects without diabetes with endogenous hyperinsulinemia without further confirmation.
In summary, this study clearly demonstrates a strong link between exogenous insulin use and the development of hypertension in patients with T2DM. In consideration of the strength of association, the dose-response relationship, the consistency in different analyses, the correctness of temporality, and the biological plausibility, the causation of hypertension by exogenous insulin use is highly possible. This study has merit because of its large sample size. However, because of the retrospective follow-up nature of the study, future confirmation by studies with a prospective, longitudinal, follow-up design is necessary.
Correspondence: Chin-Hsiao Tseng, MD, PhD, Department of Internal Medicine, National Taiwan University Hospital, 7 Chung-Shan S Rd, Taipei, Taiwan (firstname.lastname@example.org).
Accepted for Publication: January 6, 2006.
Financial Disclosure: None.
Funding/Support: This study was supported in part by grant DOH89-TD-1035 from the Department of Health and grants NSC-90-2320-B-002-197, NSC-92-2320-B-002-156, NSC-93-2320-B-002-071, and NSC-94-2314-B-002-142 from the National Science Council.
Acknowledgment: I thank the Department of Medical Research at National Taiwan University Hospital for providing facilities and space, and support for related studies.