HIV Articles  
Insulin Resistance a Possible Causal and Treatable Risk Factor for Ischemic Stroke - editorial - pdf attached - full text from original study below
  Download the PDF
from Jules: in HIV patients it appears have higher rates of diabetes & insulin resistance but the prevalence of insulin resistance has not been the focus of much research. As HIV+ individuals age insulin resistance and diabetes are likely to increase in prevalence. This study finds increased risk for stroke for patients with insulin resistance which is statistically significant, and increased risk for MI (myocardial infarction) but this is not statistically significant, they discuss reasons in Comments.
Graeme J. Hankey, MD, FRCP, FRCP(Edin), FRACP; Tan Ze Feng, MD
Arch Neurol. October 2010;67(10):1177-1178. doi:10.1001/archneurol.2010.252
The world is facing a looming epidemic of stroke.1 Although the incidence of stroke in high-income countries declined by about 42% (1%/y) between 1970 and 2008, concurrent with a reduction in the prevalence and level of causal risk factors in the population, the incidence of stroke in low-income countries increased by about 225% (5.6%/y) in the same period, concurrent with increasing exposure of the population to raised blood pressure, smoking, physical inactivity, and Westernized diets that are low in fruit and vegetables but high in salt and fat, as well as an increasing average age of the population.1-2
A role of physicians and neurologists in the global effort to prevent stroke is to identify and treat high-risk individuals with targeted risk reduction strategies such as lowering blood pressure and cholesterol, antithrombotic therapy, and in appropriate patients, carotid revascularization. Despite these endeavors, however, stroke occurrence and recurrence continues at unsustainable rates.3 Additional complementary strategies of identifying and treating high-risk individuals that are effective, affordable, and widely accessible are needed.
In this issue of the Archives, Rundek et al hypothesize that measuring insulin resistance, by means of the Homeostasis Model Assessment (HOMA) of Insulin Resistance (fasting blood glucose [mmol/L] x fasting insulin [µU/mL]/22.5), may improve our ability to identify individuals without diabetes who are at high risk of stroke and more accurately quantify their risk.4 Among a multiethnic population of 1509 individuals without diabetes or stroke in Northern Manhattan, baseline evidence of insulin resistance, as defined by a score in the top quartile of the HOMA for insulin resistance, was associated with an increased risk of first-ever ischemic stroke after a mean follow-up of 8.5 years (adjusted hazard ratio, 2.8; 95% confidence interval, 1.3-6.0) after adjusting for sex, race/ethnicity, known vascular risk factors (including glucose concentration), and the metabolic syndrome and its components.4 This finding is consistent with other observational epidemiological studies5-7 and suggests, but does not prove, that insulin resistance may be a significant causal risk factor for stroke and other cardiovascular events, independent of other risk factors. The implications of these findings are exciting if insulin resistance can be proven to be a causal risk factor for stroke (rather than a marker of increased risk) because insulin resistance cannot only be measured but also treated.
Insulin resistance is a metabolic disorder in which a normal amount of insulin produces a subnormal physiological response because of diminished sensitivity (resistance) of target cells, such as muscle, to the actions of insulin. After insulin binds to its receptors on target cells, such as muscle, the transport of glucose into the cell is defective, most likely because of defective intracellular signaling and activation of transport proteins, particularly glucose transporter type 4. This leads to defective synthesis of muscle glycogen.8
The cause of insulin resistance is uncertain but it arises when genetically predisposed individuals are exposed to lipid oversupply associated with environmental factors such as a sedentary lifestyle and central obesity. High rates of insulin resistance are observed in elderly persons, African Americans, individuals with obesity, physical deconditioning, hypertension, lipoprotein abnormalities, and vascular disease of the heart and brain, and almost all patients with type 2 diabetes.8 Among patients with transient ischemic attack and ischemic stroke, about 50% (95% confidence interval, 38%-62%) have insulin resistance.9 However, a significant proportion of apparently healthy subjects are also insulin resistant. Throughout the world, hundreds of millions of people, perhaps more than1 billion people, are estimated to have insulin resistance.5
How is insulin resistance measured? The definitive technique to quantify insulin resistance is the euglycemic hyperinsulininemic clamp. During a constant infusion of insulin, glucose is infused at a rate that is adjusted to maintain glucose homeostasis, which is equivalent to the rate of glucose use by peripheral tissues. Insulin resistance is diagnosed when the glucose infusion rate is low (reflecting reduced glucose use). Despite the lack of a universally accepted threshold that defines insulin resistance, this test is so cumbersome that it is not suitable for routine clinical use. Less invasive methods include the quantitative insulin sensitivity check index and the HOMA. The method used by Rundel et al,4 HOMA, is perhaps the most widely used for indirectly estimating insulin sensitivity and insulin secretion but it is only a surrogate measure of insulin resistance and not a definitive technique.
Insulin resistance causes defective muscle glycogen synthesis and a compensatory increase in insulin production by the pancreas. In people without insulin resistance, insulin has a variety of antiatherogenic properties. For example, it dilates resistance arterioles in skeletal muscle (primarily by stimulation of nitric oxide release by endothelial cells), reduces platelet aggregation of fibrinogen synthesis, and possesses anti-inflammatory and antioxidant properties.5, 8
Among people with insulin resistance, the antiatherogenic effect of insulin appears to be attenuated or blunted, presumably because of the same type of impaired intracellular signaling and activation of transport proteins in other cells targeted by insulin as occurs in muscle cells in the presence of insulin resistance.
The hyperinsulinemia that accompanies insulin resistance may also elevate blood pressure (by raising renal sodium reabsorption, sympathetic activity, and endothelin production), triglycerides, and platelet aggregation, impair fibrinolysis, alter low-density lipoprotein particle size, and reduce high-density lipoprotein.8 If the hyperinsulinemia fails to maintain glucose homeostasis, the blood glucose concentration rises and type 2 diabetes develops.
Strategies shown to improve insulin sensitivity include aerobic exercise (increased daily energy expenditure), weight loss, and insulin-sensitizing drugs (metformin, rosiglitazone, pioglitazone).8 Other possible strategies being evaluated include dietary modifications including reducing saturated and trans fats and alcohol consumption and supplementing the diet with fish oil, whole grains, fruits, vegetables, and vitamin D.
Rosiglitazone and pioglitazone are thiazolidinedione drugs that, through modulation of the transcription factor peroxisome proliferator-activated receptor {gamma}, have beneficial effects on insulin action (and pancreatic β cell function) and result in more durable glycemic control than other antidiabetic drugs (especially sulfonylureas).10 Furthermore, the thiazolidinediones have other beneficial cardiometabolic effects including reduction of visceral fat mass, decreased systemic inflammation, and improvement in biomarkers and surrogate outcomes associated with atherosclerosis. Adverse effects include weight gain, increased incidence of fractures, fluid retention, and a 2-fold increased risk of heart failure. Rosiglitazone is also associated with a 30% to 40% increase in risk of myocardial infarction. However, a meta-analysis of 19 randomized, double-blinded, trials comparing pioglitazone with placebo or an active comparator in a total of 16 390 patients with diabetes revealed that, after treatment for 4 months to 3.5 years in duration, death, myocardial infarction, or stroke occurred in 375 of 8554 patients (4.4%) who received pioglitazone and 450 of 7836 patients (5.7%) who received control therapy (hazard ratio, 0.82; 95% confidence interval, 0.72-0.94; P = .005). Progressive separation of time-to-event curves became apparent after approximately 1 year of therapy. Individual components of the primary end point were all reduced by a similar magnitude with pioglitazone treatment, with hazard ratios ranging from 0.80 to 0.92. Serious heart failure was reported in 200 (2.3%) of the pioglitazone-treated patients and 139 (1.8%) of the control patients (hazard ratio, 1.41; 95% confidence interval, 1.14-1.76; P = .002).11
Although pioglitazone appears beneficial in reducing major vascular events in diabetes, it is uncertain whether the suggested benefit of pioglitazone is mediated by insulin sensitization or by additional observed reductions in risk factors (hypertension, dyslipidemia, and hyperglycemia) that are known to also improve with pioglitazone therapy. Thus, the role of insulin resistance as a causal risk factor for major vascular events remains uncertain, as does the effect of pioglitazone in nondiabetic patients with stroke.
To establish a causal relationship between insulin resistance and stroke, it is necessary to not only show that insulin resistance is an independent, significant risk factor for stroke, as reported by Rundel et al,4 but also that removing or minimizing exposure to insulin resistance is associated with a reduction in the incidence of stroke. This question, and that of the effect of pioglitazone in nondiabetic patients with stroke, will be addressed by the Insulin Resistance Intervention after Stroke (IRIS) trial ( Identifier: NCT00091949). The IRIS trial is a randomized, double-blind, placebo-controlled, parallel assignment trial that aims to determine whether pioglitazone is effective in preventing recurrent stroke or myocardial infarction during at least 3 years among 3136 nondiabetic persons who are insulin resistant and have had a recent transient ischemic attack or ischemic stroke. It began in February 2005 and continues to recruit and follow up patients.
While awaiting the results of the IRIS trial, we can take confidence from the study of Rundel et al4 that measuring insulin resistance may help refine prognostic estimates of future risk of stroke obtained by means of traditional risk stratification schemes. Although it is premature to widely screen for insulin resistance as a means to prevent stroke, its measurement may have a role in particular cases in which traditional risk stratification schemes suggest that the patient is at intermediate risk of stroke (rather than high or low risk) and in whom an additional finding of insulin resistance may be sufficiently compelling to supplement lifestyle advice with pharmacological interventions to lower stroke risk.
Original Contribution
Insulin Resistance and Risk of Ischemic Stroke Among Nondiabetic Individuals From the Northern Manhattan Study
Tatjana Rundek, MD, PhD; Hannah Gardener, ScD; Qiang xu, PhD; Ronald B. Goldberg, MD; Clinton B. Wright, MD, MS; Bernadette Boden-Albala, DrPH; Norbelina Disla, BS; Myunghee
C. Paik, PhD; Mitchell S. V. Elkind, MD, MS; Ralph L. Sacco, MD, MS Arch Neurol. Oct 2010
Background - Whether insulin resistance predicts ischemic stroke (IS) is still a matter of debate.
Objective - To determine the association between insulin resistance (IR) and risk of first ischemic stroke in a large, multiethnic, stroke-free cohort without diabetes.
Design - Prospective, population-based cohort study.
Setting - Longitudinal epidemiologic study.
Participants - A cohort of 1509 nondiabetic participants from the Northern Manhattan Study (mean [SD] age, 11 [10] years; 64.2% women; 58.9% Hispanics).
Main Outcome Measures - Insulin sensitivity, expressed by the homeostasis model assessment (HOMA) of insulin sensitivity (HOMA index = [fasting insulin x fasting glucose] / 22.5). Insulin resistance was defined by a HOMA-IR index in the top quartile (Q4). Cox proportional hazards models were used to determine the effect of HOMA-IR on the risk of incident IS, myocardial infarction (MI), vascular death, and combined outcomes (IS, MI, and vascular death).
Results - The mean (SD) HOMA-IR was 2.3 (2.1), and Q4 was at least 2.8. During mean follow-up of 8.5 years, vascular events occurred in 180 participants; 46 had fatal or nonfatal IS, 45 had fatal or nonfatal MI, and 121 died of vascular causes. The HOMA-IR Q4 vs less than Q4 significantly predicted the risk of IS only (adjusted hazard ratio, 2.83; 95% confidence interval, 1.34-5.99) but not other vascular events. This effect was independent of sex, race/ethnicity, traditional vascular risk factors, and metabolic syndrome and its components.
Conclusions - Insulin resistance estimated using the HOMA is a marker of increased risk of incident stroke in nondiabetic individuals. These findings emphasize the need to better characterize individuals at increased risk for IS and the potential role of primary preventive therapies targeted at IR.
Insulin resistance (IR) is a metabolic disorder characterized by diminished tissue sensitivity to insulin that originates from environmental factors, such as a sedentary lifestyle, central obesity, and genetic predisposition.1 Insulin resistance is a pivotal pathophysiologic contributor to the increased risk of cardiovascular disease.2-5 Whether IR predicts ischemic stroke (IS) is still a matter of debate.4, 6-10 We have yet to clarify whether IR is a risk factor for incident IS in the general population after accounting for traditional vascular and metabolic risk factors. The gold standard for direct measurement of insulin sensitivity and secretion are the euglycemic hyperinsulinemic clamp methods, which are cumbersome and unsuitable for epidemiologic studies. The homeostasis model assessment (HOMA) is a widely used clinical and epidemiologic tool for indirect estimates of insulin sensitivity and insulin secretion.11-12
The aim of this study was to determine whether baseline IR estimated using HOMA increases the risk of incident IS, myocardial infarction (MI), and vascular death in a large, multiethnic, population-based, stroke-free cohort without a diagnosis of diabetes. We13-14 previously observed that impaired fasting glucose levels and metabolic syndrome are powerful predictors of incident IS. The present study extends these results to the predictive effect of IR.
In 1509 individuals free of stroke, MI, and the diagnosis of diabetes, the mean (SD) age was 68 (11) years; 36% were male, and 59% were Hispanic, 21% black, and 21% white. The mean (SD) HOMA-IR index was 2.3 (2.1). A value of 2.8 was the cutoff point of the HOMA-IR index Q4 distribution (the IR group). The percentage of individuals in HOMA-IR Q4 was similar for men (23%) and women (26%) but varied across race/ethnic groups (15% for whites, 22% for blacks, and 30% for Hispanics, P < .05). A HOMA-IR value greater than 3 was present in 23% of participants (13% white, 18% black, and 69% Hispanic, P < .05).
Vascular risk factor characteristics stratified by IR (HOMA-IR Q4 and Q1-Q3) are presented in Table 1. Individuals in HOMA-IR Q4 were younger; were more likely to be Hispanic; had higher blood pressure, waist circumference, body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), triglyceride levels, and fasting glucose levels; and had lower HDL cholesterol levels. They were less likely to be moderate alcohol users and to be physically active. Table 2 provides baseline characteristics stratified bysex. Women were older and had a greater BMI and a higher low-density lipoprotein cholesterol level; men were predominantly smokers and moderate alcohol users and had higher diastolic blood pressures and waist circumferences. Although more women used statins, the overall use of statins was low. There was no difference in the use of antiplatelet agents.
Overall, 35% of the participants had metabolic syndrome (61% of those in HOMA-IR Q4 and 27% in the first 3 quartiles). The proportion of individuals in HOMA-IR Q4 was greater in those with vs without metabolic syndrome (43% vs 15%). In addition, the proportion of those in HOMA Q4 was also greater in individuals with vs without each component of metabolic syndrome (P < .05), particularly for elevated fasting glucose level (60% vs 23%) and waist circumference (40% vs 15%) (data not shown).
During mean follow-up of 8.5 years, 180 participants experienced 1 or more symptomatic vascular events. We observed 46 cases of fatal or nonfatal IS, 45 cases of fatal or nonfatal MI, and 121 vascular deaths. The incidence rate (per 1000 person-years) of IS was 3.5; MI, 3.5; and combined vascular events, 14.0.
Preliminary analyses using HOMA-IR as a continuous variable did not indicate a significant association with the risk of IS (age-adjusted relative risk [RR], 1.04; 95% confidence interval [CI], 0.90-1.19) or combined vascular events (age-adjusted RR, 1.03; 95% CI, 0.97-1.10). Analyses of HOMA-IR quartiles also did not show a dose-response relationship (in Table 3 for IS only). After adjustment for covariates, a clear dose-response relationship was not apparent for IS (model 4, trend test P = .08) or for other events (eg, for combined vascular events in model 4, trend test P = .74) (data not shown). A threshold effect was observed in individuals in HOMA-IR Q4, where elevated risks of IS (HOMA-IR Q4 vs Q1 age-adjusted RR, 3.11; 95% CI, 1.25-7.76) and combined vascular events (HOMA-IR Q4 vs Q1 age-adjusted RR, 1.35; 95% CI, 0.91-2.00) (data not shown) were observed.
The HOMA-IR Q4 vs Q1-Q3 was associated with a significant 2.5-fold increased risk of IS in the age-adjusted analysis (RR, 2.47; 95% CI, 1.28-4.77) (Table 3). The association persisted in the model controlling for sociodemographic factors and metabolic syndrome (model 3: multivariate-adjusted RR, 2.43; 95% CI, 1.21-4.91) and in the model controlling for vascular risk factors (model 4: 2.83; 95% CI, 1.34-5.99).
The association between IR and risk of MI and vascular death as secondary outcomes was not significant (model 4: multivariate-adjusted RR for MI, 1.77; 95% CI, 0.88-3.58; for vascular death, 1.10; 0.69-1.74), suggesting that the effect of IR may not be as strong for MI and vascular death as for IS risk (Table 3).
The association between IR and risk of combined vascular events showed that individuals in HOMA-IR Q4 had a 45% increased risk of vascular events in the age-adjusted model (RR, 1.45; 95% CI, 1.04-2.03) (Table 3). This association persisted after controlling for demographic factors but was attenuated and no longer significant after controlling for metabolic syndrome status (model 3: RR, 1.37; 95% CI, 0.96-1.96) or after adjustment for vascular risk factors (model 4: 1.25; 0.86-1.82).
Effect modification by sex was suggested for the association between HOMA-IR and IS (an interaction term between the HOMA-IR Q4 and sex was significant in the multivariate-adjusted model, P < .05). A significant association between IR and risk of IS was observed in men (HOMA-IR Q4 vs Q1-Q3 in model 4: multivariate-adjusted RR, 10.86; 95% CI, 3.04-38.82) but not in women (model 4: 1.27; 0.41-3.92) (data not shown). No effect modification by race/ethnicity was observed.
To test for potential inclusion bias of standard Cox regression parameter estimates, we conducted the analysis by weighting each record in the stroke risk set by the inverse of the probability of observation, which suggested that even after correcting for potential bias due to inclusion of those with available HOMA, the conclusions remained unchanged because the association between IR (HOMA Q4) and risk of IS remained statistically significant.

In this multiethnic, prospective, population-based cohort study of nondiabetic individuals, we report that IR estimated using HOMA in Q4 (vs Q1-Q3) is associated with a 2.8-fold increased risk of first IS but not with other vascular events. Adjustment for established cardiovascular risk factors, including glucose level, obesity, and metabolic syndrome, did not attenuate the association with IS. We also observed a stronger association between IR and first IS in men than in women but not in any specific race/ethnic groups. The potential effect modification of this relationship by sex and race/ethnicity deserves further exploration in larger, ethnically diverse prospective cohorts.
There are several possible reasons for the stronger effect of IR on the risk of IS than of MI in the present study compared within the other studies. First, racial/ethnic disparities in the different effects of IR on the risk of stroke or MI may be one of the possible explanations. This population-based study consisted of predominantly Hispanic participants compared with the predominantly white population of the Framingham Offspring Study and the Multi-Ethnic Study of Atherosclerosis.6-7 Hispanic and black participants in the present population are at particularly greater risk for stroke than are white participants19 and have a higher prevalence of metabolic syndrome and a greater effect of the metabolic syndrome on the risk of stroke than of MI compared with white participants.16 Hispanic and black participants also had a higher prevalence of IR than did white participants in this study, which might have had a greater effect on the risk of stroke than of MI. The present study, however, had limited power to explore predictive effect modification of IR by race/ethnicity. Second, we excluded individuals with a history of MI, which may be another explanation for the smaller predictive effect of IR on incident MI compared with other studies that included individuals with prevalent MI. Third, IR has been associated with subclinical atherosclerosis and, therefore, may be more likely linked to IS due to small- or large-vessel atherosclerosis than due to cardioembolism. Small- and large-vessel atherosclerotic strokes were more frequent IS subtypes in Hispanic patients than in white participants in a previous study.19 Fourth, IR is associated with hypertension, hypertriglyceridemia, and low HDL cholesterol levels (as noted in Table 1), which may be more closely and specifically linked to IS than to MI. This may be especially important in black and Hispanic participants. Because IR drives high triglyceride levels and low HDL cholesterol levels, IR may have a relatively greater effect on stroke than on MI given that low-density lipoprotein cholesterol is less strongly related to stroke. In addition, HDL cholesterol levels are lower in men than in women, which also might have contributed to the stroke sex-specific differential effect of IR in the present study.
The risk of incident IS in the present population (adjusted hazard ratio of 2.8) is higher than that reported in other population-based studies even after adjusting for metabolic syndrome. In large population-based studies2, 4, 6, 20-21 of nondiabetic participants in the highest 20th to 30th percentile of IR, the adjusted hazard ratio of stroke ranged from 1.5 to 2.6, but in most studies, this effect was lost after adjusting for components of metabolic syndrome. Although a resistance to insulin action may provide the unifying mechanism of metabolic syndrome, the results of the present study suggest that metabolic syndrome (as defined by the Third Report of the National Cholesterol Education Program Adult Treatment Panel) may not capture all the vascular risk associated with IR, raising the possibility that other pathways affected by IR, such as inflammation, may be important.22
We observed a greater effect of IR on risk of first IS in men than in women, although the prevalence of IR was similar. Reasons for the sex-specific differential effect remain unclear. Complex interactions between IR and sex-specific vascular risk factor profiles (eg, smoking or higher blood pressure in men),23-24 sex differences in insulin action,25 and its biological effects on the atherosclerotic process26 may have accounted for the observed difference in the risk of stroke between men and women.
The strengths of the present study include a prospective population-based design with thorough case ascertainment, confirmation of diagnosis, a well-documented baseline exposure, and comprehensive prospectively collected data on established risk factors for cardiovascular disease. The aggressive follow-up strategies resulted in less than 1% loss to follow-up. Study participants were seen in person at study enrollment and follow-up, whenever possible, to document outcome events. The inclusion of a large multiethnic, elderly, heterogeneous cohort with similar geographic access to the medical center is generalizable to other multiethnic urban populations and allows for more valid comparisons across race/ethnic categories. However, the power to detect effect modification by race/ethnicity may be limited in the present study. Additional limitations include the potential for residual confounding, the 1-time exposure measurement, and the limited statistical power of the stroke analysis. These results, therefore, must be interpreted with caution. Owing to the lack of a dose-response relation and the possibility of chance findings, further exploration with larger data sets and more end points is necessary. It is likely that because of the small number of end points we did not observe an effect of HOMA-IR on MI. Another possible explanation may be the fact that HOMA-IR is a relatively blunt instrument for estimating reduced sensitivity to insulin, although, in general, HOMA-IR values correlate reasonably well with clamp-derived gold standard values.27
The present study provides evidence that IR as measured using HOMA is independently associated with an increased risk of first IS. Insulin resistance may be a novel therapeutic target for stroke prevention. Clinical trials, such as the Insulin Resistance Intervention after Stroke trial20 in participants with stroke and transient ischemic attack and the Veterans Affairs High-Density Lipoprotein Intervention Trial5 in patients with coronary heart disease, have shown improved insulin sensitivity and β-cell function after treatment with certain classes of drugs, such as a peroxisome proliferator-activated receptor {gamma}-agonist or cholesterol-lowering drugs. In addition to secondary stroke prevention, future studies are needed to determine whether the treatment of IR can reduce the risk of incident stroke and cardiovascular disease.

The Northern Manhattan Study (NOMAS) is a prospective, population-based cohort study of stroke incidence, risk factors, and prognosis in a multiethnic urban community. The methods for NOMAS have been extensively described elsewhere.15-16 A total of 3298 stroke-free individuals were enrolled in NOMAS between 1993 and 2001. After excluding individuals with previously diagnosed diabetes or fasting glucose levels of at least 126 mg/dL (to convert to millimoles per liter, multiply by 0.0555) (n = 705; 21%), those with previous MI (n = 244; 7%), those without blood samples available for fasting insulin and glucose analysis (n = 1464; 44%), and those of a race/ethnicity other than black, white, or Hispanic (n = 79; 2%), a sample of 1509 stroke-free individuals was included in this study.
Participants were screened annually by telephone to determine any change in vital status, to detect neurologic and cardiac symptoms and events, and to review interval hospitalizations, risk factor status, medications, and changes in functional status. Persons with positive telephone interview screens were examined in person by the study neurologists and cardiologists. Incident IS was the primary outcome. The secondary outcomes were incident MI, vascular death, and any vascular event, defined as incident IS, MI, and vascular death combined. Follow-up procedures and outcome classifications have been detailed previously.13-14
Baseline blood samples, collected after at least 12 hours of fasting, were assayed for insulin level using an analyzer (Immulite 2000; Diagnostic Products Corp, Los Angeles, California) and the manufacturer's reagents and a solid-phase, 2-site, chemiluminescent enzyme-labeled immunometric assay.11-12 The HOMA-IR index was calculated as follows: [fasting insulin x fasting glucose] / 22.5.
Race/ethnicity was defined by self-identification based on a series of interview questions modeled after the US census.15 The race/ethnic categorizations included Hispanic, non-Hispanic white, and non-Hispanic black. Individuals of another race/ethnicity were excluded from the analyses. Hypertension was defined as a self-reported history of hypertension or a measured systolic blood pressure of at least 140 mm Hg or diastolic blood pressure of at least 90 mm Hg. Smoking was categorized as never, former, and current (within 1 year). Moderate alcohol use was defined as current drinking of more than 1 drink per month and 2 drinks or fewer per day. Moderate to heavy physical activity was defined as engaging in recreational activities in a typical 14-day period. Waist circumference, low-density lipoprotein cholesterol level, high-density lipoprotein (HDL) cholesterol level, and systolic and diastolic blood pressures were examined as continuous variables. Metabolic syndrome was defined by the Third Report of the National Cholesterol Education Program Adult Treatment Panel.17
The HOMA-IR index was examined continuously and as quartiles to investigate a potential dose-response relationship or a threshold effect with vascular outcomes. The prevalence of sociodemographic characteristics (age, sex, race/ethnicity, and education), traditional vascular risk factors, and other baseline variables was stratified by the HOMA-IR status at the cutoff level of the top quartile (Q4) of the HOMA-IR distribution (Q4 = 2.8 or defined as the IR group).
Person-time of follow-up was accrued from baseline to the end of follow-up (January 2008), the time of the outcome event, death, or loss to follow-up, whichever came first. Cox proportional hazards models were used to determine the effect of IR on the risk of incident IS as a primary outcome and on the risk of MI, vascular death, and any vascular event as secondary outcomes. Univariate age-adjusted Cox models were conducted (model 1), as were multivariate-adjusted models controlling for sociodemographic factors (age, sex, race/ethnicity, and high school completion [model 2]), sociodemographic factors and metabolic syndrome (model 3), and sociodemographic factors and risk factors, including waist circumference, systolic and diastolic blood pressure, moderate alcohol consumption, HDL cholesterol level, moderate to heavy physical activity, and current/former smoking (model 4). Last, the 2 interaction terms, sex x HOMA-IR and race/ethnicity x HOMA-IR, were added to model 4 to examine potential effect modification by sex and race/ethnicity. Statistical analyses were conducted using a software program (SAS version 9.1; SAS Institute Inc, Cary, North Carolina).
In addition, we examined the potential for selection bias that can result from the use of a subset of the overall study population if inclusion is jointly affected by IR and risk of cardiovascular events. The population analyzed in this study (n = 1509) represented a subsample of the full NOMAS cohort (N = 3298) with available information on HOMA and meeting all the inclusion criteria. The potential for selection bias was examined by fitting logistic regression models using the observation indicators as response and follow-up times, uncensoring indicators, and risk factors as covariates. Based on this result, we conducted an inverse probability weighting method that adjusts potential selection bias of standard Cox proportional hazards regression variable estimates by weighting each record in the risk set by the inverse of the probability of observation.18
Correspondence: Tatjana Rundek, MD, PhD, Department of Neurology, Miller School of Medicine, University of Miami, Clinical Research Bldg, Ste 1348, 1120 NW 14th St, Miami, FL 33136 (
Accepted for Publication: March 17, 2010.
Author Contributions: All authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Rundek, Boden-Albala, and Sacco. Acquisition of data: Boden-Albala, Disla, and Elkind. Analysis and interpretation of data: Gardener, xu, Goldberg, Wright, Paik, and Elkind. Drafting of the manuscript: Rundek and Gardener. Critical revision of the manuscript for important intellectual content: Rundek, Gardener, xu, Goldberg, Wright, Boden-Albala, Disla, Paik, Elkind, and Sacco. Statistical analysis: Gardener, xu, Boden-Albala, and Paik. Obtained funding: Boden-Albala and Elkind. Administrative, technical, and material support: Rundek, Gardener, Goldberg, Wright, Disla, and Elkind. Study supervision: Rundek and Sacco.
Financial Disclosure: None reported.
Funding/Support: This study was supported by the Gilbert Baum Memorial Grant and the Goddess Fund for Stroke Research in Women (Dr Rundek); by grants R37/R01 29993 (Drs Rundek, Gardener, Goldberg, Wright, Boden-Albala, Paik, Elkind, and Sacco) and K23 NS42912 (Dr Elkind) from the National Institute of Neurological Disorders and Stroke; by the Kathleen Scott Research Fellowship from the American Heart Association (Dr Elkind); and by grant 2 M01 RR00645 from the General Clinical Research Center.
Author Affiliations: Departments of Neurology (Drs Rundek, Gardener, Wright, and Sacco), Medicine (Dr Goldberg), and Epidemiology, Human Genetics (Dr Sacco), Miller School of Medicine, University of Miami, Miami, Florida; and Departments of Biostatistics (Drs xu and Paik) and Sociomedical Sciences (Dr Boden-Albala), Joseph Mailman School of Public Health, and Department of Neurology (Drs Boden-Albala and Elkind and Ms Disla), Columbia University, New York, New York.
  iconpaperstack view older Articles   Back to Top