Ravi Varadhan

Advanced glycation end products and their circulating receptors predict cardiovascular disease mortality in older community-dwelling women

Aims: To characterize the relationship between advanced glycation end products (AGEs) and circulating receptors for AGEs (RAGE) with cardiovascular disease mortality. Methods: The relationships between serum AGEs, total RAGE (sRAGE), and endogenous secretory RAGE (esRAGE), and mortality were characterized in 559 community-dwelling women, ‡ 65 years, in Baltimore, Maryland. Results: During 4.5 years of follow-up, 123 (22%) women died, of whom 54 died with cardiovascular disease. The measure of serum AGEs was carboxymethyl-lysine (CML), a dominant AGE. Serum CML predicted cardiovascular disease mortality (Hazards Ratio [HR] for highest vs lower three quartiles, 1.94, 95% Confidence Interval [CI] 1.08–3.48, p=0.026), after adjusting for age, race, body mass index, and renal insufficiency. Serum sRAGE (ng/mL) and esRAGE (ng/mL) predicted cardiovascular disease mortality (HR per 1 Standard Deviation [SD] 1.27, 95% CI 0.98–1.65, p=0.07; HR 1.28, 95% CI 1.02–1.63, p=0.03), after adjusting for the same covariates. Among non-diabetic women, serum CML, sRAGE, and esRAGE, respectively, predicted cardiovascular disease mortality (HR for highest vs lower three quartiles, 2.29, 95% CI 1.21–4.34, p=0.01; HR per 1 SD, 1.24, 95% CI0.92–1.65, p=0.16; HR per 1 SD 1.45, 95% CI 1.08–1.93, p=0.01), after adjusting for the same covariates. Conclusions: High circulating AGEs and RAGE predict cardiovascular disease mortality among older community-dwelling women. AGEs are a potential target for interventions, as serum AGEs can be lowered by change in dietary pattern and pharmacological treatment.

Frailty, Inflammation, and Mortality Among Persons Aging With HIV Infection and Injection Drug Use

BACKGROUNDnSerum markers of inflammation increase with age and have been strongly associated with adverse clinical outcomes among both HIV-infected and uninfected adults. Yet, limited data exist on the predictive and clinical utility of aggregate measures of inflammation. This study sought to evaluate the relationship of a recently validated aggregate inflammatory index with frailty and mortality among aging HIV-infected and uninfected injection drug users.nnnMETHODSnFrailty was assessed among HIV-infected and uninfected participants in the AIDS Linked to the IntraVenous Experience (ALIVE) cohort study using the five Fried phenotypic criteria: weight loss, exhaustion, low physical activity, decreased grip strength, and slow gait. The aggregate inflammatory index was constructed from serum measures of interleukin-6 and soluble tumor necrosis factor-α receptor-1. Multinomial logistic regression was used to assess the relationship of frailty with inflammation. Cox proportional hazards models were used to estimate risk for all-cause mortality.nnnRESULTSnAmong 1,326 subjects, the median age was 48 years and 29% were HIV-infected. Adjusting for sociodemographics, comorbidity, and HIV status, frailty was significantly associated with each standard deviation increase in log interleukin-6 (odds ratio 1.33; 95% CI, 1.09-1.61), log tumor necrosis factor-α receptor-1 (odds ratio 1.25; 95% CI, 1.04-1.51) and inflammatory index score (odds ratio 1.39; 95% CI, 1.14-1.68). Adjusting for sociodemographics, comorbidity, HIV status, and frailty, the inflammatory index score was independently associated with increased mortality (HR 1.65; 95% CI, 1.44-1.89).nnnCONCLUSIONnA recently validated, simple, biologically informed inflammatory index is independently associated with frailty and mortality risk among aging HIV-infected and uninfected injection drug users.

A general binomial regression model to estimate standardized risk differences from binary response data

Estimates of absolute risks and risk differences are necessary for evaluating the clinical and population impact of biomedical research findings. We have developed a linear-expit regression model (LEXPIT) to incorporate linear and nonlinear risk effects to estimate absolute risk from studies of a binary outcome. The LEXPIT is a generalization of both the binomial linear and logistic regression models. The coefficients of the LEXPIT linear terms estimate adjusted risk differences, whereas the exponentiated nonlinear terms estimate residual odds ratios. The LEXPIT could be particularly useful for epidemiological studies of risk association, where adjustment for multiple confounding variables is common. We present a constrained maximum likelihood estimation algorithm that ensures the feasibility of risk estimates of the LEXPIT model and describe procedures for defining the feasible region of the parameter space, judging convergence, and evaluating boundary cases. Simulations demonstrate that the methodology is computationally robust and yields feasible, consistent estimators. We applied the LEXPIT model to estimate the absolute 5-year risk of cervical precancer or cancer associated with different Pap and human papillomavirus test results in 167,171 women undergoing screening at Kaiser Permanente Northern California. The LEXPIT model found an increased risk due to abnormal Pap test in human papillomavirus-negative that was not detected with logistic regression. Our R package blm provides free and easy-to-use software for fitting the LEXPIT model.

A regression model for risk difference estimation in population-based case-control studies clarifies gender differences in lung cancer risk of smokers and never smokers.

BackgroundAdditive risk models are necessary for understanding the joint effects of exposures on individual and population disease risk. Yet technical challenges have limited the consideration of additive risk models in case–control studies.MethodsUsing a flexible risk regression model that allows additive and multiplicative components to estimate absolute risks and risk differences, we report a new analysis of data from the population-based case–control Environment And Genetics in Lung cancer Etiology study, conducted in Northern Italy between 2002–2005. The analysis provides estimates of the gender-specific absolute risk (cumulative risk) for non-smoking- and smoking-associated lung cancer, adjusted for demographic, occupational, and smoking history variables.ResultsIn the multiple-variable lexpit regression, the adjusted 3-year absolute risk of lung cancer in never smokers was 4.6 per 100,000 persons higher in women than men. However, the absolute increase in 3-year risk of lung cancer for every 10 additional pack-years smoked was less for women than men, 13.6 versus 52.9 per 100,000 persons.ConclusionsIn a Northern Italian population, the absolute risk of lung cancer among never smokers is higher in women than men but among smokers is lower in women than men. Lexpit regression is a novel approach to additive-multiplicative risk modeling that can contribute to clearer interpretation of population-based case–control studies.

Endogenous Secretory Receptor for Advanced Glycation End Products and Chronic Kidney Disease in the Elderly Population

Background/Aims: The relationship of circulating endogenous secretory receptor for advanced glycation end products (esRAGE) and chronic kidney disease (CKD) has not been well characterized. The aim of the study was to determine whether plasma esRAGE is associated with CKD and is predictive of developing CKD in older adults. Methods: The relationship between plasma esRAGE and CKD (more than stage 3 of the National Kidney Foundation classification; estimated glomerular filtration rate <60 ml/min/1.73 m2) and CKD over 6 years of follow-up was examined in a cross-sectional and prospective study design in 1,016 men and women, ≧65 years, in the InCHIANTI study, a population-based cohort study of aging in Tuscany, Italy. Results: At enrollment, 158 (15.5%) had CKD. Mean (SD) plasma esRAGE was 0.45 (0.24) ng/ml. Plasma esRAGE (ng/ml) was associated with CKD (odds ratio per 1 SD = 1.30; 95% CI 1.1–1.6; p < 0.005) in a multivariable logistic regression model, adjusting for potential confounders. Plasma esRAGE was an independent predictor of incident CKD over 6 years of follow-up (hazard ratio per 1 SD = 1.37; 95% CI 1.1–1.7; p < 0.008) in a multivariable Cox proportional hazards model, adjusting for potential confounders. Conclusions: Elevated plasma esRAGE is independently associated with CKD and is an independent predictor of incident CKD in older community-dwelling adults.

Circulating selenium and carboxymethyl-lysine, an advanced glycation endproduct, are independent predictors of anemia in older community-dwelling adults

OBJECTIVEnTo assess whether selenium and carboxymethyl-lysine (CML), two biomarkers of oxidative stress, are independent predictors of anemia in older community-dwelling adults.nnnMETHODSnPlasma levels of selenium, CML, folate, vitamin B12, and testosterone and markers of iron status and inflammation were measured at baseline in 1036 adults at least 65 y old in the Invecchiare in Chianti Study, a population-based cohort study of aging in Tuscany, Italy, and examined in relation to prevalent anemia and incident anemia over 6 y of follow-up.nnnRESULTSnAt enrollment, 11.6% of participants were anemic. Of 472 participants who were non-anemic at enrollment, 72 (15.3%) developed anemia within 6 y of follow-up. At enrollment, plasma CML in the highest quartile (>425 ng/mL) and plasma selenium in the lowest quartile (<66.6 μg/L) predicted incident anemia (hazard ratio 1.67, 95% confidence interval 1.07-2.59, P = 0.02; hazard ratio 1.55, 95% confidence interval 1.01-2.38, P = 0.05, respectively) in a multivariate Cox proportional hazards model that adjusted for age, education, body mass index, cognition, inflammation, red blood cell distribution width, ferritin, vitamin B12, testosterone, and chronic diseases.nnnCONCLUSIONnElevated plasma CML and low plasma selenium are long-term independent predictors of anemia in older community-dwelling adults. These findings support the idea that oxidative stress contributes to the development of anemia.

Binomial Regression in R

Abstract Binomial regression is used to assess the relationship between a binary response variable and other explanatory variables. Popular instances of binomial regression include examination of the etiology of adverse health states using a case–control study and development of prediction algorithms for assessing the risk of adverse health outcomes (e.g., risk of a heart attack). In R, a binomial regression model can be fit using the glm() function. In this chapter, we demonstrate the following aspects of binomial regression, with R code, using real data examples: • To highlight the main components of a binomial model fitting using the glm() function • How to evaluate the modeling assumptions in binomial regression? • How to relax the assumptions when they are violated? • How to fit binomial models for non-independent data? • How to develop and evaluate prediction models for binary response? The chapter is meant to be a quick, practical guide to binomial regression using R. We particularly envision the accompanying task view to be a useful resource on all topics closely related to binomial regression.