Lead Concentrations in Relation to Multiple Biomarkers of Cardiovascular Disease: The Normative Aging Study

Background: Lead exposure has been associated with cardiovascular disease (CVD) in animal and human studies. However, the mechanisms of action have not been fully elucidated. We therefore examined the relationship between lead and multiple biomarkers of CVD. Methods: Participants were older men from the Normative Aging Study without preexisting coronary heart disease, diabetes, or active infection at baseline (n = 426). Serum biomarkers included lipid profile [total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides] and inflammatory markers [C-reactive protein, intercellular adhesion molecule-1, interleukin-6, and tumor necrosis factor receptor-2 (TNF-R2)]. We measured lead in blood and in bone by K-shell X-ray fluorescence. In this sample, 194 men (44.3%) had two or more repeated measures, resulting in 636 observations for analysis. We conducted analyses using mixed effects models with random subject intercepts. Results: Lead levels were associated with several CVD biomarkers, including levels of TNF-R2 and lipid markers. Specifically, in multivariable models, a 50% increase in blood lead level was associated with 26% increased odds of high TNF-R2 levels (> 5.52 ng/mL; odds ratio = 1.26; 95% confidence interval: 1.09, 1.45). There were positive associations of blood lead level with total cholesterol and HDL levels, and these associations were more evident when modeled as continuous outcomes than when categorized using clinically relevant cut points. In addition, longitudinal analyses indicated a significant increase in TNF-R2 levels over time to be associated with high blood lead level at the preceding visit. Conclusions: Blood lead level may be related with CVD in healthy older men through its association with TNF-R2 levels. In addition, the magnitude of the association of blood lead level with TNF-R2 level increased with age in the study population.


Research
There is increasing recognition of the role of environmental pollutants in relation to cardio vascular disease (CVD) (Bhatnagar 2006). Among the environmental factors, lead expo sure has been found in human studies to be associated with hypertension, peripheral arterial disease, and circulatory and cardio vascular mortality, and in animal studies, to promote hypertension and accelerate CVD (Bhatnagar 2006;NavasAcien et al. 2007;Weisskopf et al. 2009). In in vitro and in vivo studies, lead has been shown to decrease the bioavailability of nitric oxide partially by increasing the production of reactive oxygen species (ROS). It is postulated that the related increase in ROS and changes in cytokine pro duction may promote atherogenesis, although this has not been fully explored (Bhatnagar 2006), particularly in epidemiologic studies.
Atherogenesis is related to several biological processes, among them lipid metabolism and inflammation. Although interrelated in their effects on atherogenesis, these processes are typically considered separately in studies seek ing to identify factors that may increase risk of dysfunction in these domains. Experimental studies and human studies on lead poisoning or occupationally exposed workers have pro duced unexpected findings regarding the effect of lead on lipid metabolism. For example, Cocco et al. (1991) found that leadexposed patients had decreased cholesterol and low density lipoprotein (LDL) and increased high density lipoprotein (HDL) levels. Conversely, Ademuyiwa et al. (2005) reported a signifi cant positive correlation between blood lead and total cholesterol and LDL levels. Among studies of inflammation, in vitro studies and a study of occupationally exposed workers found a relationship between lead and tumor necrosis factorα (TNFα), but not between lead and interleukin6 (IL6) (Valentino et al. 2007). TNFα is involved in chronic cell mediated inflammation (Valentino et al. 2007) and is associated with endothelial dysfunction, ROS, and lipid metabolism. Findings for highsen sitivity Creactive protein (hsCRP) have been mixed. One occupational study found a sig nificant relation between lead exposure and CRP (Khan et al. 2008), but a population based study did not find a consistent associa tion (Songdej et al. 2010). We know of no study that has directly investigated the role of lead on endothelial dysfunction [i.e., adhe sion molecules such as intercellular adhesion molecule1 (ICAM1)].
We examined the relationship between lead levels and a comprehensive suite of cardiovascular risk biomarkers, including lipid profile and inflammatory markers, in a population of older men participating in the Normative Aging Study (NAS). We also investigated the relation over time. We tested the hypothesis that higher levels of blood or bone lead would be associated with higher total cholesterol, LDL, and triglycerides lev els, with lower HDL levels, and with higher hsCRP, ICAM1, IL6, and TNF receptor 2 (TNFR2) levels at specific timepoints and across time (i.e., associated with higher rates of change in biomarker levels).

Study population. NAS is a longitudinal study of aging that was established in 1963 by the Veterans Administration [now the Department
Background: Lead exposure has been associated with cardiovascular disease (CVD) in animal and human studies. However, the mechanisms of action have not been fully elucidated. We therefore examined the relationship between lead and multiple biomarkers of CVD. Methods: Participants were older men from the Normative Aging Study without preexisting coronary heart disease, diabetes, or active infection at baseline (n = 426). Serum biomarkers included lipid profile [total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides] and inflammatory markers [C-reactive protein, intercellular adhesion molecule-1, interleukin-6, and tumor necrosis factor receptor-2 (TNF-R2)]. We measured lead in blood and in bone by K-shell X-ray fluorescence. In this sample, 194 men (44.3%) had two or more repeated measures, resulting in 636 observations for analysis. We conducted analyses using mixed effects models with random subject intercepts. results: Lead levels were associated with several CVD biomarkers, including levels of TNF-R2 and lipid markers. Specifically, in multivariable models, a 50% increase in blood lead level was associated with 26% increased odds of high TNF-R2 levels (> 5.52 ng/mL; odds ratio = 1.26; 95% confidence interval: 1.09, 1.45). There were positive associations of blood lead level with total cholesterol and HDL levels, and these associations were more evident when modeled as continuous outcomes than when categorized using clinically relevant cut points. In addition, longitudinal analyses indicated a significant increase in TNF-R2 levels over time to be associated with high blood lead level at the preceding visit. conclusions: Blood lead level may be related with CVD in healthy older men through its association with TNF-R2 levels. In addition, the magnitude of the association of blood lead level with TNF-R2 level increased with age in the study population. of Veterans Affairs (VA)]. The subgroup of par ticipants used in the present analysis has been described elsewhere (Cheng et al. 2001;Hu et al. 1996). Briefly, NAS is a closed cohort of 2,280 male volunteers from the Greater Boston area who were screened at entry and enrolled if they were free of any known chronic medi cal conditions. Participants have been reevalu ated every 3-5 years using questionnaires and detailed onsite physical examinations.
By 1999, when measurements of inflam matory markers began, 668 original partici pants had died, and 104 others were no longer being followed primarily because they had moved out of the region after retirement. Further, some participants were lost during recruitment into the lead substudy, leaving 767 participants with blood lead measure ments taken during the 1999-2008 study period. Of the 767 men, we had serum mea surements of lipids or inflammatory markers on 707 during this period (see Figure 1 for participant flow chart).
At baseline (time of their first inflamma tory measure), 198 men with coronary heart disease (i.e., angina pectoris or history of myocardial infarction), 57 men with diabe tes mellitus (fasting glucose ≥ 126 mg/dL or medication use), and 26 men with hsCRP > 10 mg/L (indicating active infection) were excluded, resulting in a study population of 426 men. Exclusion by disease status at base line is based on the assumption that lead may affect the selected biomarkers in part through its effects on these diseases, which would make them causal intermediates. Among these 426 men, 16 (2.52%) had two additional visits, 178 (41.8%) had one additional visit, and 232 (54.5%) had only one visit over the course of the study period, yielding a total of 636 observations with biomarker and blood lead data available for analysis. However, the number of observations varied among bio markers because of differences in the starting year of the measurements (Figure 1).
Beginning in 1991, participants were invited to undergo bone lead measurements at the Ambulatory Clinical Research Center of the Brigham and Women's Hospital (Harvard School of Public Health, Boston, Massachusetts), and these measurements con tinued until 2002. Not all participants agreed to these extra visits. Of the 707 men with bio marker and blood lead data, a total of 405 men also had bone lead measurements ( Figure 1). After exclusion for baseline presence of heart disease, diabetes, or hsCRP > 10 mg/L, 230 men remained with a total of 371 observations available for bone lead analysis. For all analy ses, we estimated associations with the bone lead measurement taken closest in time to the biomarker measurement being analyzed. All participants provided written informed con sent, and the study protocol was approved by the human research committees of Brigham and Women's Hospital, and the VA Boston Healthcare System. Biomarker measurement. Serum samples were drawn after overnight fasting. Serum cholesterol was assayed over the course of the study period using standard enzymatic meth ods and reagents (SCALVO Diagnostics, Wayne, NJ). The HDL fraction was measured in the supernatant after precipitation of the LDL and very lowdensity lipoprotein fractions with dextran sulfate and magnesium, using the Abbott Biochromatic Analyzer 100 (Abbott Laboratories, South Pasadena, CA). Triglyceride levels were measured using the Dupont ACA discrete clinical analyzer (Dupont Company, Biomedical Products Department, Wilmington, DE). LDL was estimated using the Friedewald equation (Friedewald et al. 1972).
The inflammatory markers for all time points were analyzed on archived samples stored at -80°C. Samples from each participant were analyzed in duplicate and in a single batch to avoid betweenbatch analytical variation. The performance of the assays was monitored with standard quality control procedures, including the analysis of quality control samples in each batch. Serum hsCRP was measured using an immuno turbidi metric assay on the Hitachi 917 analyzer (Roche Diagnostics, Indianapolis, IN) with reagents and calibrators from DenkaSeiken (Niigata, Japan). Soluble ICAM1 was mea sured using a quantitative sandwich enzyme linked immunosorbent assay (R&D Systems, Minneapolis, MN). Sensitivity of the assay for ICAM1 was 0.35 ng/mL, and the daytoday variabilities of the assay at concentrations of 64.2, 117, 290, and 453 ng/mL were 10.1%, 7.4%, 6.0%, and 6.1%, respectively. IL6 and soluble TNFR2 were assayed with the Milliplex Map Human Cytokine/Chemokine Kit (Millipore Corp., St. Charles, MO) and detected with a multiplex detection plat form (Luminex® 100/200™ System; Luminex Corporation, Austin, TX). The intraassay coef ficients of variation for IL6 and TNFR2 were 10.6% and 5.9%, respectively.
Lead measurement. Blood lead was analyzed over the course of the study period by graph ite furnace atomic absorption with Zeeman background correction (ESA Laboratories, Chelmsford, MA). The instrument was cali brated after every 21 samples with National Institute of Standards and Technology (NIST; Gaithersburg, MD) standard reference mate rial (SRM 955a, lead in blood). Ten percent of the samples were run in duplicate, at least 10% as controls, and 10% as blanks. Compared with reference samples from the Centers for Disease Control and Prevention (CDC; Atlanta, GA), the precision ranged from 8% for concentrations < 30 µg/dL to 1% for higher concentrations.
Bone lead was measured for 30 min each at the midtibia shaft and at the patella using a Kshell Xray fluorescence instrument (ABIOMED Inc., Danvers, MA). The techni cal specifications and validity of this instrument are described in detail elsewhere Hu et al. 1990Hu et al. , 1994. The turnover is slower in cortical bone than in trabecular bone; thus, the halflife of lead in cortical bone ver sus trabecular bone is approximately 20 versus 8 years (Hu et al. 1991). The shaft of the tibia is mostly cortical bone, and the patella bone is mostly trabecular bone. Whereas bone lead lev els represent cumulative exposure, blood lead levels represent acute exposure. With blood lead measurements Statistical analysis. We evaluated mean baseline (time of the first biomarker measure) characteristics and biomarker levels according to blood lead quartiles using generalized linear models. Cut points for blood lead quartiles were based on the distribution in the largest sample. Given the variability in sample sizes for the biomarkers, we also performed sensitiv ity analyses for each biomarker by including only those participants who had all biomarkers measured (most participants excluded from the sensitivity analysis did not have a measure of TNFR2). Additionally, we determined unad justed Spearman rank correlations between baseline lead and biomarker levels. Then for all observations, we examined the relationship in ageadjusted models and multivariable models using generalized linear mixed models (with logit link) for dichotomized biomarkers (as described below) and with mixedeffects mod els for continuous biomarkers. We used the likelihood ratio test to select the compound symmetry covariance matrix with random intercept modeled in the following forms: Dichotomized outcome: Continuous outcome: where η ij = log(p ij /1 -p ij ) and represents the proportion of high biomarker outcomes (e.g., probability of being in the highrisk category for cholesterol) in subject i on visit j; Y ij is the biomarker level [e.g., log(total cholesterol level) in subject i on visit j]; β 0 is the overall intercept; u i is the separate random intercept representing the subjectspecific correlation among measurements; and β 1 through β p rep resent the fixed effects for up to p covariates. As a sensitivity analysis to adjust for the possibility that healthier men were more likely to return for subsequent exams, we weighted the findings by including an inverse prob ability of multiple exams (i.e., a revisit pro pensity score). The revisit propensity score was calculated from a logistic regression of the probability of not having two or more study visits, given all relevant factors including age, measurement of a particular marker (lipid profile or inflammatory marker), and total cholesterol level as well as diabetes, hyper tension, and smoking status.
To assess the potential modifying effects of change in age from baseline age for each biomarker over time, we ran models that included a crossproduct term for the inter action between change in age (time) and the preceding blood lead measure [lag blood lead (lag lead )], along with the main effects. For the first visit, we used the blood lead level prior to the inflammatory marker measurements. We also included the concurrent blood lead meas ure in the model because it is most likely more relevant to the acute biomarker levels, using the following form (e.g., for continuous outcome): Separate analyses considered biomarkers as continuous (log transformed to address issue of normality) or as categorical measures. We cate gorized the biomarkers to tie results to clini cally relevant thresholds where possible. The lipid measurements were dichotomized based on guidelines for recommended levels from the Third Report of the U.S. National Cholesterol Education Program Adult Treatment Panel (ATP III): total cholesterol ≥ 200 mg/dL, LDL ≥ 130 mg/dL, HDL < 40 mg/dL, and triglycerides ≥ 200 mg/dL (Expert Panel 2001). hsCRP was divided based on the American Heart Association (AHA)/CDC Scientific Statement (Pearson et al. 2003) at ≥ 3.1 mg/L (i.e., high risk for CVD). ICAM1, TNFR2, and IL6, which do not have clinical cutoffs, were divided at their medians (281 ng/mL, 5.52 ng/mL, and 28.0 pg/mL, respectively). For these analyses, we reported results as odds ratios (ORs) for a 50% increase in lead expo sure [i.e., exp(β log[lead] × ln[1.5])]. The results for continuous outcomes were reported as the ratio of the geometric means (GMs) for a 50% increase in lead exposure [i.e., 1.5^(β log [lead] )]. Lead measures were modeled as logtrans formed variables to address issues of normality. For analyses assessing the potential modifying effects of baseline age and change in age (inter action analyses), lead first was modeled contin uously and then dichotomized at the median for ease of interpretation. All analyses were per formed using SAS software, version 9.2 (SAS Institute Inc., Cary, NC).
Covariates of interest included baseline age (years), change in age from baseline over the followup period, and baseline measures of body mass index (BMI; kilograms per meter squared], education attainment level (< 12th grade, 12th grade, > 12th grade to < 4 years college/technical school, or ≥ 4 years of college), hypertension [physician diagnosis of hyper tension with treatment or systolic blood pres sure > 140 mmHg or diastolic blood pressure > 90 mmHg during examination (Chobanian et al. 2003)], statin use (yes or no), smoking status (current, former, or never), packyears of smoking, and alcohol consumption (< 2 vs. ≥ 2 drinks/day). Physical activity was mea sured as metabolic equivalent of task (MET). A pvalue of < 0.05 was considered significant, and a pvalue of < 0.10 was considered margin ally significant.

Results
Characteristics of the study participants by quartile of blood lead level are presented in Table 1. At baseline, men with higher blood lead levels were more likely to be current smokers, consume ≥ 2 drinks a day, and be less physically active. When the sample was limited to participants with TNFR2 mea surements, sample differences across covari ates were in the same direction but attenuated with regard to significance for alcohol con sumption and physical activity (data not shown). Table 2 shows the baseline CVD biomarker concentrations by quartile of blood lead level. Those with higher blood lead levels were more likely to have significantly higher total cholesterol, HDL, and ICAM1 levels. As with smoking and less physical activity, the results were the same when the sample was limited to participants with TNFR2 measurements. The mean ± SD concentra tion was 4.01 ± 2.30 µg/dL for blood lead, 26.2 ± 18.8 mg/g for patella lead, and 19.8 ± 13.2 mg/g for tibia lead.
Tibia lead levels did not correlate with any of the CVD biomarkers, and patella lead levels were significantly correlated only with HDL (Table 3). In contrast, blood lead was signifi cantly and positively correlated with total cho lesterol, LDL, HDL, hsCRP, ICAM1, and TNFR2. Therefore, we focused the remainder of our analysis on the relationship of blood lead levels with CVD biomarkers.
In ageadjusted models using the repeated measures, higher blood lead was significantly  (Table 4). After adjusting for the full set of covari ates, the associations with total cholesterol and HDL were no longer significant, and the association with hsCRP was attenuated, becoming marginally associated (p = 0.09). Associations with TNFR2 remained signifi cant and similar in magnitude. The relation ship was maintained after further including the revisit propensity score. A 50% increase in blood lead level was associated with a 26% increased likelihood of having high TNFR2 levels [OR = 1.26; 95% confidence interval (CI): 1.09, 1.45].
For continuous logtransformed outcomes in ageadjusted models, higher blood lead was significantly associated with higher total cholesterol, LDL, and higher HDL, as well as higher hsCRP and TNFR2 (Table 4). In multivariable and weighted multivariable analyses, the association was no longer main tained for hsCRP. The associations remained significant for total cholesterol, HDL, and TNFR2. For example, a 50% increase in blood lead level was associated with a 5% increase in the GM of TNFR2 (ratio of GMs = 1.05; 95% CI: 1.02, 1.08).
We also evaluated whether lag lead modified the odds of high TNFR2 (categorical) over time and the rate of change of TNFR2 (con tinuous) over time, controlling for concurrent blood lead (Table 5). We did not observe a significant estimated effect of lag lead in the odds of high TNFR2 across time, but there was an association with the rate of change of TNFR2 over time. Those with higher than median blood lead at the preceding visit had a discernibly greater increase in TNFR2 lev els over time than those with low blood lead. Specifically, compared with those with low blood lead, those with high blood lead had an additional 0.03 ng/mL increase in log TNFR2 (i.e., additional 1% increase in GM of TNFR2) across time (pinteraction = 0.01).

Discussion
Longitudinal analysis demonstrated a positive association of blood lead levels with TNFR2. The relationship with TNFR2 and blood lead significantly increased with change in age from baseline; that is, higher blood lead was associ ated with a steeper subsequent rate of increase in TNFR2 over time. The associations were unchanged even after adjusting for an array of medical conditions and biological and behavioral factors, as well as accounting for a higher propensity of those in better health to have repeated measures. Also, blood lead was positively associated with the lipid markers total cholesterol, LDL, and HDL; however, effect estimates were attenuated in multivari able models, particularly for outcomes catego rized using clinical cut points. In ageadjusted models, blood lead was significantly associated with hsCRP, but the association was not sig nificant after further adjustment.
Lead exposure has been associated with adverse effects on cardiovascular health. However, as far as we know, no other study has comprehensively evaluated the relation ship between lead and potential mediating biological processes involved in CVD or how these processes play out over time (i.e., with age). Potential mediating processes may be assessed with biomarkers that indicate lipid profile and levels of inflammation or endothe lial dysfunction (Pearson et al. 2003;Selhub 2006). Approximately 95% of lead in adults is stored in bone. It has been hypothesized that bone lead is continuously released into blood, with the rate of mobilization differing by age and bone type, and that the mobilized lead contributes to adverse health effects (Hu et al. 1998). We would expect that these acute biomarkers would be most closely related to the acute marker of lead (blood lead), which is consistent with our findings.
In the present study, our strongest find ing was that initial blood lead levels were associated with more rapid rise in inflamma tion as measured by TNFR2 over time. In the inflammatory cascade model proposed by Pearson et al. (2003), primary pro inflammatory cytokines such as TNFα can have a direct impact on circulation by increas ing the products of hepatic stimulation such as CRP or by influencing the levels of adhe sion molecules. TNFα functions as a media tor of inflammatory reactions and plays a part in the immunopathogenesis of hypertensive disease (Dorffel et al. 1999). Dysregulation of the expression of TNFα is associated with vascular aging and leads to vascular oxida tive stress and remodeling, thrombosis, cell infiltration, apoptosis, and vascular inflam mation (Zhang et al. 2009). Our results agree with those of Valentino et al. (2007), who found a positive relationship between TNFα and blood lead in lowlead-exposed work ers versus nonexposed workers. Lead appears to stimulate TNFα production by modify ing the expression of TNFR2 through its interaction with cysteinerich membrane proteins. The observed interaction of high baseline blood lead with changes in TNFR2 seems to indicate a modifying effect of blood lead on the normal agerelated increase in the  In an animal study, Guo et al. (1996) found that lead potentiated the lethality of LPS by increasing expression of both TNFα and TNFR2. This agerelated increase in response to environmental stimulus is thought to be due to alterations in Tolllike receptor func tion, a protective mechanism involved in lim iting inflammation (Larbi et al. 2008;May et al. 2009;van Duin and Shaw 2007). We found some evidence of an associa tion of blood lead with CRP and ICAM1 in unadjusted models, suggesting that effects on these inflammatory and endothelial functions are possible but might be an artifact of effects on other related processes. A small occupa tional study of 87 leadexposed workers and 61 non exposed workers matched on age and sex also found a relationship between blood lead and CRP in leadexposed workers in bivariate models (Khan et al. 2008). However, another populationbased study did not find a dosedependent relationship between blood lead level and CRP after adjusting for a broad array of potential confounders (Songdej et al. 2010), similar to our findings. Moreover, similar to reports by Valentino et al. (2007), we did not see any evidence of an effect of lead on IL6 levels.
We also observed an association between blood lead and total cholesterol based on an age adjusted model, but the association was attenu ated after controlling for additional covariates. In analyses of continuous (logtransformed) HDL (rather than HDL categorized by clinical guidelines), there was also a significant increase in HDL with higher blood lead levels. Similar findings have been reported for leadexposed patients by Cocco et al. (1991), but they also saw decreases in total cholesterol and LDL with lead exposure. However, another study of occupational lead exposure found a posi tive correlation between blood lead and total cholesterol and LDL, but not between blood lead and HDL (Ademuyiwa et al. 2005). In one study with rats, low doses of lead were associated with a decrease in total cholesterol and HDL and an increase in triglyceride levels (Skoczynska et al. 1993); however, in contrast to humans, rats transport the major proportion of plasma cholesterol by HDL (Hassan 1987;Skoczynska et al. 1993). Emerging research has found that although HDL is generally con sidered protective against CVD, under cer tain situations it may not be atheroprotective (Feng and Li 2009). Additionally, it is still not clear whether lead induces lipid abnormalities by altering lipid metabolism (e.g., increasing serum cholesterol levels) or by stimulating lipid peroxidation (Skoczynska et al. 1993).
This study has several limitations. The sam ple size was relatively small. Although we had approximately 636 observations, the available sample size varied by biomarker. In addition, we may have inadequately controlled for or missed potential confounders. Moreover, in this report, we focused on the acute measure of lead expo sure: blood lead. Although we chose this focus because most of the univariable relationships were observed with blood lead, the comparison with a marker of acute lead exposure (i.e., blood lead) may be appropriate when evaluating rela tionships with serum biomarkers. We also used lag lead measured at the previous visit (usually 3-4 years before the biomarker data) to look at change over time in biomarker levels. The anal ysis was modeled in this way because we wanted to be sure the exposure preceded the change in biomarker concentrations. We note that similar results were obtained whether we used concur rent or lag lead in the interaction term (data not shown). Finally, this study was of males only, and for some of these biomarkers, the literature suggests that their relationships with CVD may and on the AHA/CDC recommendation for C-reactive protein (hs-CRP) ≥ 3.1 mg/L. (Pearson et al. 2003). For biomarkers without recommendation based on the median: ICAM-1 > 281 ng/mL, TNF-R2 > 5.52 ng/mL, and IL-6 > 28.0 pg/mL. d Calculated from total cholesterol, HDL, and triglycerides for those with triglycerides < 400 mg/dL. vary between men and women (e.g., Jenny et al. 2006;Lim and Cassano 2002). However, this study was unique in that it allowed us to evalu ate acute and cumulative lead exposure on mul tiple cardiovascular risk factors and risk markers. To our knowledge, this is also the first study to examine changes in associations between lead and biomarkers over time. Although the comprehensive list of biomarkers allowed us to investigate pathways of action at different stages along the ischemic cascade, it may raise con cerns about multiple comparisons. Given mixed findings from prior studies, our results involv ing the lipidrelated variables may warrant fur ther investigation. However, for TNFR2, the magnitude of effect and consistency of findings crosssectionally and longitudinally support the possibility that the association between lead and TNFR2 is causal. Our primary findings are that blood lead levels are associated with TNFR2 levels. In addition, higher blood lead exposure at the preceding visit is associated with an increase in the change in TNFR2 with age. This indi cates that lead may play a role in increasing inflammation and may potentiate agerelated increases in inflammation.