Associations of Insulin Levels With Left Ventricular Structure and Function in American Indians : The Strong Heart Study

The SHS is a population-based survey of cardiovascular risk factors and cardiovascular disease in American Indians. As previously described (7–9), members aged 45–74 years of three American Indian communities in Arizona, seven tribes in Southwestern Oklahoma, and three tribes in South and North Dakota were recruited from all eligible individuals (overall participation rate 62%) for an initial examination in 1989–1992. Extensive characterization of subjects included standardized measurement of seated brachial blood pressure; aspects of body habitus including BMI, waist/hip ratio, and percentage of body fat by bioelectric impedance; fasting glucose, insulin, lipid, and lipoprotein concentration; and 2-h glucose tolerance test and glycosylated hemoglobin levels. Diabetes was diagnosed by World Health Organization (WHO) criteria (12) if fasting blood glucose was ≥140 mg/dl, glucose level after a 2-h challenge was ≥200 mg/dl, or subjects received hypoglycemic medication. For the present study, subjects with diabetes were excluded.

Insulin measurements used antibody 1012, WHO-traceable (1988) insulin standards, and supplies purchased from Linco Research (St. Louis, MO). The interassay and intra-assay coefficients of variation of the insulin assay at mid-range were 8.5 and 2.2%, respectively.

The second SHS examination began in August 1993 to assess changes over time in cohort members of body habitus, blood pressure, and most other baseline measures. In addition, echocardiograms were performed in 3,501 (97%) of the 3,630 phase II participants.

Imaging and Doppler echocardiography was performed as previously described (13–15). A standardized protocol was followed under which the parasternal and apical acoustic windows were used to visualize the LV internal diameter and wall thicknesses, assess LV wall motion, and search for mitral and aortic regurgitation.

Correct orientation of planes for imaging and Doppler recordings was verified as previously described (16). LV internal dimension and septal and posterior wall thicknesses were measured at end-diastole and end-systole by American Society of Echocardiography recommendations (17,18) on up to three cardiac cycles. The aortic anular diameter was measured in the long-axis view that maximized this dimension (13), and the Doppler flow velocity profile was used to calculate stroke volume by an invasively validated method (19).

End-diastolic LV dimensions were used to calculate LV mass by a formula shown to yield values closely related (r = 0.90, P < 0.001) to necropsy measurements (20), which also showed good reproducibility (ρ = 0.93, P < 0.001) between separate echocardiograms in 183 hypertensive patients (21). LV mass was primarily indexed for the power of the allometric (or growth) relationship between height and LV mass (height^2.7), as has been shown to detect expected deviations from normal in populations with BMI in the ranges 25–30, 30–40, and ≥40 kg/m² (22) and also for body surface area (21). Relative wall thickness was calculated as posterior wall thickness/internal radius; systolic fractional shortening in percentage of the ventricle’s internal dimension and end-systolic wall stress were calculated by standard methods (16).

Myocardial contractile efficiency was examined by relating LV systolic shortening to end-systolic stress (23). Primary reliance was placed on the relation of midwall fractional shortening to midwall circumferential end-systolic stress measured at the LV minor axis as previously described (23,24). Midwall fractional shortening was calculated taking into account the epicardial migration of the midwall. For evaluating LV performance taking circumferential end-systolic stress into account, midwall fractional shortening was expressed as a percentage of the value predicted from circumferential end-systolic stress using an equation derived in previously studied normal subjects (23), termed stress-corrected midwall fractional shortening (25).

SPSS (SPSS, Chicago, IL) software was used for data management and statistical analyses. Data are expressed as mean ± SD. Because the distribution of fasting plasma insulin was right skewed, it was log transformed before application of parametric statistical methods. Preliminary analyses examined the associations of log₁₀ insulin and LV mass with potential confounders, including age, sex, BMI, height, and systolic blood pressure. Univariate relations of log₁₀ insulin with measures of LV structure or function as dependent variables were assessed by Pearson correlations, followed by multiple linear regression analyses with the identified confounders as additional independent variables. A stepwise procedure was used with P to enter <0.05 and to remove <0.10. Statistical analyses were first performed separately in the 901 participants with normal glucose tolerance (51% women) and the 487 with impaired glucose tolerance (64% women); because results in both groups were directionally similar and were statistically significant in the larger subgroup with normal glucose tolerance for the same comparisons as in the entire population, data from the combined groups are presented.

Of the 3,501 participants in the second SHS examination who underwent echocardiography, glucose tolerance testing, plasma insulin evaluation, and clinical examination, 1,388 (45% men) were free of diabetes and had other data needed for the present report. A large proportion of the population was obese with a mean BMI of 30 ± 6 kg/m². Other clinical characteristics of the study group are shown in Table 1.

Among nondiabetic participants in the second SHS examination, preliminary analyses examined the associations of log₁₀ insulin and LV mass with potential confounders, including age, sex, BMI, height, and systolic blood pressure. In univariate analyses, LV mass was significantly related among women and men to older age, greater height, higher BMI, and higher systolic pressure (Tables 2 and 3). Regression analysis revealed that each of these variables plus male sex had highly significant statistically independent associations with LV mass. In univariate analyses, log insulin was significantly associated with age, height, BMI, and systolic pressure among men and with BMI and systolic pressure among women (Tables 4 and 5). In regression analysis, female sex, height, BMI, and systolic pressure—but not age—were independently associated with higher log fasting plasma insulin levels. On the basis of the finding of significant relations of age, height, BMI, and systolic blood pressure with both LV mass and the log plasma insulin level in one or both sexes, these variables were considered as confounders in multivariate analyses of the relations between insulin levels and LV mass as well as other measures of LV geometry, function, and systemic hemodynamics.

Among men (Table 6), univariate correlations were found between log₁₀ insulin and LV mass and its component parts, LV wall thicknesses, and chamber diameter. This association was preserved by indexation of LV mass for height^2.7, but it was eliminated by indexation of LV mass for body surface area, a variable strongly determined by body weight. Because of parallel increases in LV wall thicknesses and diameter, there was no association of log₁₀ insulin with LV relative wall thickness. Left atrial diameter was positively related to log₁₀ insulin, whereas no associations were observed between log₁₀ insulin and measures of LV systolic chamber or midwall function. Higher log₁₀ insulin was positively related to higher cardiac output but not cardiac output indexed for body surface area or total peripheral resistance.

Among women, univariate correlations were found between log₁₀ insulin and LV mass and its component parts, LV wall thicknesses, and chamber diameter (Table 6). This association was preserved by indexation of LV mass for height^2.7, but it was eliminated by indexation of LV mass for body surface area, a variable strongly determined by body weight. Because of stronger relations of LV wall thicknesses than LV diameter with log₁₀ insulin, the latter variable was weakly associated with higher LV relative wall thickness. Left atrial diameter was positively related to log₁₀ insulin. Log₁₀ insulin had a weak positive association with endocardial fractional shortening but not with measures of LV systolic midwall function. Higher log₁₀ insulin was positively related to higher cardiac output but not to cardiac output indexed for body surface area or total peripheral resistance.

Multiple linear regression analysis in women identified septal thickness and LV relative wall thicknesses but not the primary outcome measure of LV mass as being significantly related to log₁₀ insulin after adjusting for age, BMI, height, and systolic pressure (all P < 0.05). In men, LV structural parameters were not independently related to insulin level in similar multiple linear regression analyses, whereas cardiac output and peripheral resistance retained an independent positive relation with log₁₀ insulin.

That as much as half of the interindividual variability in LV mass remains unexplained after standard demographic and hemodynamic factors are taken into account (14) has stimulated investigation of potential cardiac effects of a variety of hormones and growth factors. A number of previous studies have evaluated relations between LV findings and fasting or postchallenge plasma insulin levels, with variably positive or negative results (6,26–41). One potential explanation for cardiac effects of insulin, hormonal stimulation of sodium retention, seems to occur under some but not other circumstances (2–4). The present study documents univariate associations between insulin levels and several measures of LV structure, thereby confirming previous positive reports, and also demonstrates that these univariate relations are markedly weakened or even completely attenuated in multivariate analyses, controlling for the strong confounding effects of obesity and the lesser ones of age and arterial pressure.

An important result of the present study is that associations between fasting plasma insulin levels and LV variables seem to differ by sex. Multivariate analysis revealed that higher insulin levels were associated, independent of covariates, with higher cardiac output in nondiabetic men and with higher LV wall thicknesses in women. Limited precedent for sex differences in cardiac effects of insulin is provided by a report from the Tecumseh Blood Pressure Study (6) in which insulin was positively related to LV hypertrophy in men but not women. Whether this represents a direct interaction between sex and trophic effects of insulin on the heart or, alternatively, whether this sex difference might be a nonspecific phenomenon paralleling the greater tendency of women than men to develop concentric LV geometry in response to pressure overload (42) is uncertain. Of note, other features of the insulin resistance syndrome were similar in both sexes, including ∼20% higher BMI and 20% lower HDL cholesterol levels in individuals with relatively high insulin levels.

A potential limitation of the present study is assessment of insulin levels in the fasting state but not in response to glucose loading. Several studies have found positive associations between postload insulin levels or areas under the postload insulin curve and LV structural variables (28,31–33,37,41), whereas several others have not (30,34,39). The lack of postload insulin measurements in SHS participants makes it impossible to determine whether there might have been stronger relations between LV variables and insulin responses to glucose loading than those observed with fasting insulin. However, one reason for stronger relations of postload than fasting insulin levels to other biologic variables in some studies—use of assays with poor sensitivity at low levels—does not apply to the SHS, which measured insulin by a sensitive radioimmunoassay. Another potential limitation of the present study is its use of fasting insulin rather than a more sensitive measure of insulin action; unfortunately, it was not feasible to perform direct assessments of insulin sensitivity in addition to a multifaceted examination that required at least 3 h of participant time.

A strength of the present study is the relatively large number of nondiabetic individuals from a population-based sample who were evaluated. Most previous studies have assessed relatively small groups of individuals (n = 26–120) from selected clinical samples (25,27,29,30,32–34,36,37,40,41). With one exception (6), most of the previous population-based studies of insulin-LV relations have also involved smaller populations (from 62 to 351) (27,29,33,36) than in the present study. An additional strength of the present study is the ability to measure adipose body mass and percentage of body fat by bioelectric impedance (43). However, substitution of the latter variables for BMI did not alter the results of the study; therefore, findings using the more easily measured BMI are presented.

In conclusion, strong univariate associations between fasting plasma insulin levels and abnormalities of LV structure, function, and systemic hemodynamics are attenuated but not completely eliminated when the effects of overweight and other covariates are taken into account. The mechanisms of the observed associations are uncertain but may include enhancement by insulin of increased distal tubular sodium reabsorption (3,44), with resultant increases in hemodynamic volume and pressure loads and possible direct myocardial trophic effects of insulin (45). One potential stimulus to increased LV wall thicknesses, elevated arterial stiffness, has been identified in women and men with type 2 diabetes in the SHS (15) but was not detected by the pulse pressure/stroke volume ratio in nondiabetic participants with elevated insulin levels in the present study. Additional research is needed to determine the generalizability of the present results to other population-based samples and to elucidate the mechanisms involved.

We thank the Indian Health Service, SHS participants, and participating tribal communities for extraordinary cooperation and involvement that made this study possible; Betty Jarvis, RN, Tauqeer Ali, MD, and Alan Crawford for coordination of the study centers; and Mary Paranicas, BA, and Dawn Fishman, BA, for data coordination and management of SHS data. We also thank Tauqeer Ali, MD, Helen Beatty, RDMS, Joanne Carter, RDMS, Michael Cyl, RDMS, and Neil Sikes, RDMS, for technical assistance and Virginia M. Burns for assistance in manuscript preparation.