Background

Cardiovascular disease is the leading cause of mortality and a primary contributor to the burden of disease worldwide (Lopez et al. 2006). Environmental toxicants, including lead and other metals, are potentially preventable exposures that may explain population variation in cardiovascular disease rates (Bhatnagar 2006; Weinhold 2004). However, after more than 100 years since initial reports suggested a link between lead exposure and cardiovascular outcomes (Lancéreaux 1881; Lorimer 1886), the contribution of lead to cardiovascular disease is still incompletely understood.

Population research on the cardiovascular effects of lead has focused largely on the association with blood pressure and hypertension. Several reviews and metaanalyses combining data from more than 30 original studies and around 60,000 participants have examined the evidence relating blood lead to blood pressure or hypertension [Hertz-Picciotto and Croft 1993; Nawrot et al. 2002; Schwartz 1995; Sharp et al. 1987; Staessen et al. 1994, 1995; U.S. Environmental Protection Agency (U.S. EPA) 2006]. All these reviews concluded that there was a positive association between blood lead levels and blood pressure (Table 1). The estimated increase in systolic blood pressure associated with a 2-fold increase in blood lead levels (e.g., from 5 to 10 µg/dL) ranged across reviews from 0.6 to 1.25 mmHg. This epidemiologic relationship is also supported by a large body of experimental and mechanistic evidence (U.S. EPA 2006). Because lead exposure is widespread, even a modest effect would imply that lead exposure is an important determinant of blood pressure levels and hypertension in human populations.

Table 1.

The cardiovascular effects of lead, however, are not limited to increased blood pressure and hypertension. Lead exposure has also been associated with an increased incidence of clinical cardiovascular end points such as coronary heart disease, stroke, and peripheral arterial disease (Lustberg and Silbergeld 2002; Menke et al. 2006; Navas-Acien et al. 2004; Schober et al. 2006), and with other cardiovascular function abnormalities such as left ventricular hypertrophy and alterations in cardiac rhythm (Cheng et al. 1998; Schwartz 1991).

In the present article, our objective was to perform a systematic review of the epidemiologic evidence on the association of lead exposure with cardiovascular disease end points. Because previous reviews have examined the connection between lead and blood pressure in depth (Table 1), our systematic review emphasizes other clinical and intermediate cardiovascular outcomes to obtain a broader picture of the impact of lead on cardiovascular disease. Finally, we assessed the causal role of lead on blood pressure and cardiovascular disease by applying the criteria and terminology of the 2004 Surgeon General Report The Health Consequences of Smoking [U.S. Department of Health and Human Services (U.S. DHHS) 2004] to the available information.

Methods

Search strategy and data abstraction. We aimed to identify all observational studies assessing the association between lead exposure and cardiovascular end points. Using free text and key words (Appendix A), we searched PubMed (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed), EMBASE (http://www.embase.com/), and TOXLINE (http://toxnet.nlm.nih.gov/) through August 2006 with no language restrictions. In addition we manually reviewed the reference lists from relevant original research and review articles and documents.

For lead exposure, we included studies that used biomarkers (lead levels in blood, bone, or other specimens), environmental measures (airborne lead levels), or indirect measures (job titles, job exposure matrices, living in lead-contaminated areas). For cardiovascular end points, we included studies that reported clinical cardiovascular end points (cardiovascular disease, coronary heart disease, stroke, or peripheral arterial disease) and intermediate cardiovascular end points (left ventricular mass, heart rate, heart rate variability, or electrocardiographic abnormalities) other than blood pressure levels or hypertension.

We excluded publications containing no original research, studies not carried out in humans, case reports, case series, ecologic studies, studies lacking a cardiovascular outcome, and studies lacking data on lead exposure (Figure 1). For studies with multiple publications on the same population, we selected the publication with the longest follow-up. For studies with equivalent follow-up periods, we selected the study with the largest number of cases or the most recent publication. We excluded autopsy studies measuring lead in arterial tissue and studies based on polycardiography and ballistocardiograpy, techniques no longer in use. For consistency, blood lead levels were converted to micrograms per deciliter.

Figure 1. Flow diagram of study selection process. Databases: PubMed (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed); EMBASE (http://www.embase.com/); TOXLINE (http://toxnet.nlm.nih.gov/).

We adapted the criteria used by Longnecker et al. (1988) to assess study quality for studies of clinical end points and the criteria used by Appel et al. (2002) to assess study quality for studies of intermediate end points (Appendices B and C).

Statistical methods. Measures of association (odds ratios, prevalence ratios, standardized mortality ratios, relative risks, relative hazards, comparisons of means, linear regression coefficients, correlation coefficients) and their standard errors were abstracted or derived from published data (Greenland 1987). For studies reporting measures of association for population subgroups (Cooper et al. 1985; Malcolm 1971), we pooled the measures of association using an inverse-variance weighted random-effects model (Egger et al. 2001).

Because of substantial heterogeneity and methodologic limitations of the original studies, we considered that quantitative pooling was inappropriate. We thus present a qualitative systematic review of the available evidence.

Results

Lead and clinical cardiovascular disease in general populations. Twelve studies met our inclusion criteria (Table 2). Lead was measured in blood in all the prospective cohort studies (Kromhout 1988; Lustberg and Silbergeld 2002; Menke et al. 2006; Mřller and Kristensen 1992; Pocock et al. 1988) and in the only cross-sectional study available (Muntner et al. 2005). Blood lead levels were substantially lower in more recent compared with older studies. Case–control studies assessed lead exposure on the basis of lead levels in blood (Kosmala et al. 2004), plasma (Mansoor et al. 2000), and urine (Pan et al. 1993; Tsai et al. 2004), on a job exposure matrix (Gustavsson et al. 2001), and on lead levels in the air of the residential neighborhood of study participants (Dulskiene 2003). None of these studies determined lead in bone. Although cohort studies and the cross-sectional study tended to fulfill prespecified quality criteria, case–control studies failed to fulfill some important quality criteria (Appendix B).

Table 2.

Table 3.

Table 4.

Table 4. Continued.

Lead exposure was positively associated with clinical cardiovascular end points in all studies (Table 2). Among prospective studies, the relative risks for coronary heart disease ranged between 1.1 comparing blood lead levels > 24.8 µg/dL versus < 12.4 µg/dL in the British Regional Heart Study (Pocock et al. 1988) and 1.89 comparing blood lead levels ≥ 3.63 µg/dL versus < 1.93 µg/dL in the National Health and Nutrition Examination Survey (NHANES) III Mortality Follow-up Study (Menke et al. 2006). The relative risk for stroke in the NHANES III Mortality Follow-up Study was 2.51. There were no prospective studies on the association of blood lead with peripheral arterial disease. However, the relative risk for peripheral arterial disease comparing blood lead levels ≥ 2.47 µg/dL versus < 1.03 µg/dL in a cross-sectional analysis of NHANES 1999–2002 was 1.92 (Muntner et al. 2005).

Lead and cardiovascular mortality in occupational populations. Eighteen studies from the United States (Cooper et al. 1985; Michaels et al. 1991; Robinson 1974; Sheffet et al. 1982; Steenland et al. 1992; Tollestrup et al. 1995), Europe (Alexieva et al. 1981; Belli et al. 1989; Carta et al. 2003; Cocco et al. 1997, 1994; Davies 1984; Dingwall-Fordyce and Lane 1963; Gerhardsson et al. 1995; Lundstrom et al. 1997; Malcolm 1971; Wilczynska et al. 1998), and Australia (McMichael and Johnson 1982) met our inclusion criteria (Table 3). Battery, ceramic, pigment, refinery, and smelter industries were studied. All studies used job titles to ascertain exposure and death certificates to identify coronary heart disease (12 studies), stroke (15 studies) and overall cardiovascular mortality (9 studies). Most were retrospective cohort studies and used external comparisons to the general population to derive standardized mortality ratios. The exceptions were the study by Dingwall-Fordyce and Lane (1963), two proportional mortality studies (Alexieva et al. 1981; McMichael and Johnson 1982) and two prospective cohort studies (Robinson 1974; Tollestrup et al. 1995). Occupational studies failed to fulfill most prespecified quality criteria (Appendix B).

Relative risk estimates across occupational studies varied widely, with positive, inverse, and null associations (Table 3). Several studies reported the associations among workers with the heaviest exposure (Dingwall-Fordyce and Lane 1963; Lundstrom et al. 1997; Malcolm and Barnett 1982; Steenland et al. 1992), by year of hire (Cooper et al. 1985; Lundstrom et al. 1997), and incorporating a latency period (Lundstrom et al. 1997). In two of the three studies that reported associations by duration of employment, coronary heart disease (Steenland et al. 1992) and stroke (Michaels et al. 1991) mortality were higher among workers with the highest number of years of employment.

Lead and intermediate cardiovascular outcomes. Five studies evaluated ventricular wall dimensional and functional parameters (Beck and Steinmetz-Beck 2005; Kasperczyk et al. 2005; Schwartz 1991; Tepper et al. 2001; Zou et al. 1995) (Table 4). Increased blood lead levels were associated with an increased prevalence of left ventricular hypertrophy in U.S. adults (Schwartz 1991) and with a nonstatistically significant increase in left ventricular mass in U.S. battery workers (Tepper et al. 2001). Similarly, Polish steel workers had higher left ventricular mass and lower ejection fraction compared to administrative workers from the same factory (Kasperczyk et al. 2005), and lead-exposed Polish workers had impaired diastolic function compared with nonexposed controls (Beck and Steinmetz-Beck 2005). Chinese refinery workers with blood lead levels > 50 µg/dL had similar interventricular septum and left ventricular wall thickness compared to workers < 50 µg/dL (Zou et al. 1995), although lead levels in the reference category are unknown.

Ten studies measured heart rate variability among lead-exposed workers (Andrzejak et al. 2004; Böckelmann et al. 2002; Gajek et al. 2004; Gennart et al. 1992; Ishida et al. 1996; Murata et al. 1995; Murata and Araki 1991; Muzi et al. 2005; Niu et al. 1998; Teruya et al. 1991), and one study measured heart rate variability in Seoul, Korea, public officials not occupationally exposed to lead (Jhun et al. 2005) (Table 4). Most of these studies had limitations in terms of sample size, methods of lead assessment, and lack of adjustment for potential confounders (Table 4; Appendix C). The conditions for electrocardiographic ascertainment and the heart rate variability indices differed widely across studies, making comparisons difficult. The coefficient of variation of the R-R interval was lower in lead-exposed workers compared with other workers in two of five studies in which the coefficient of variation was measured under normal breathing, and in one of three studies in which it was assessed during deep breathing. Among Seoul public officials (Jhun et al. 2005), increased lead levels were inversely associated with measures of low frequency, high frequency, and total power spectrum in univariate analyses, but adjusted results were not presented because lead exposure was dropped from the stepwise regression models used.

Fifteen studies reported the association of lead with other electrocardiographic parameters (Cheng et al. 1998; Gatagonova 1995a, 1995c; Kirkby and Gyntelberg 1985; Kosmider 1968; Kosmider and Petelenz 1961, 1962; Kosmider et al. 1965; Kromhout et al. 1985; Krotkiewski et al. 1964; Saric 1981; Shcherbak 1988; Sroczynski et al. 1990, 1985; Stozinic and Colakovic 1980) and one study with other vascular abnormalities (Aiba et al. 1999). All studies, except the Normative Aging Study (Cheng et al. 1998), were conducted in occupational populations in Europe. These types of outcome, including rhythm disorders, ischemic changes and cycle duration, varied widely across studies, and the findings were inconsistent. The Normative Aging Study measured lead in blood, tibia, and patella and identified associations between tibia lead and intraventricular conduction defects (QRS duration) and increased QT duration in subjects < 65 years of age (Cheng et al. 1998).

Finally, heart rate was evaluated using different methods in five studies, four in lead-exposed workers (Böckelmann et al. 2002; Kosmider and Petelenz 1961; Murata et al. 1995; Zou et al. 1995) and one in elderly men from the Netherlands (Kromhout et al. 1985), with inconsistent findings.

Discussion

Lead exposure and hypertension—sufficient evidence to infer a causal relationship. Chronic lead poisoning was connected to hypertension in the 19th century (Lorimer 1886). With rare exceptions (Vigdortchik 1935), a major limitation of early reports was the lack of a comparison group (Sharp et al. 1987). The hypertensive effects of lead have been extensively documented in experimental animals chronically exposed to high lead concentrations and in workers chronically exposed to high lead levels (Agency for Toxic Substances and Disease Registry 1999; U.S. EPA 2006). Generally, the development of hypertension in subjects chronically exposed to high lead levels has been interpreted as a possible consequence of lead nephropathy. At environmental levels of exposure, however, the effect of lead on blood pressure has been controversial. Numerous studies have addressed this question. All reviews have concluded that there is an association between lead and blood pressure, although the strength of this association is modest (Table 1). Substantial evidence, however, implies that this relationship is causal.

Consistency.The association between lead exposure and blood pressure has been found in populations with different geographic, ethnic, and socioeconomic background. While residual confounding by socioeconomic status is a concern, studies in homogenous samples and studies that have adjusted for a variety of socioeconomic indicators have still identified an association between lead exposure and blood pressure (Martin et al. 2006; Pocock et al. 1984).

Temporality.The association between blood lead and elevated blood pressure has been identified not only in cross-sectional but also in prospective studies that showed that new cases of hypertension and within-person elevations in blood pressure levels over follow-up were related to baseline lead exposure (Glenn et al. 2003; Mřller and Kristensen 1992; Weiss et al. 1986).

Strength of the association.While the strength of the association between lead and blood pressure is modest, it may have been substantially underestimated because of measurement error in both lead and blood pressure determinations. Most studies used single blood lead measurements to assess lead exposure. When bone lead was used as a biomarker of long-term exposure (Hu et al. 2007), lead in cortical or trabecular bone was positively associated with increased systolic blood pressure or hypertension in all prospective (Cheng et al. 2001; Glenn et al. 2003) and cross-sectional studies (Gerr et al. 2002; Hu et al. 1996; Korrick et al. 1999; Lee et al. 2001; Martin et al. 2006; Rothenberg et al. 2002; Schwartz and Stewart 2000). Furthermore, even bone lead is subject to error derived from the sampling site and from the technical difficulties of the measurement. In addition, blood pressure measurements were often conducted using nonstandardized protocols, without repeated measures, or in samples including hypertensive subjects.

Biologic gradient (dose response).Some studies have demonstrated a progressive dose–response relationship between lead exposure and blood pressure (Pocock et al. 1984; Schwartz 1988; Weiss et al. 1986). However, the shape of the dose–response relationship is not completely characterized, particularly at low levels of exposure. It is not known what is the lowest level of lead exposure not associated with blood pressure, although in the available studies there seems to be no evidence of a threshold effect (Hertz-Picciotto and Croft 1993; Schwartz et al. 2001).

Biologic plausibility and experimental data. Numerous experimental studies in animals have shown irrefutable evidence that chronic exposure to low lead levels results in arterial hypertension that persists long after the cessation of lead exposure (U.S. EPA 2006). The precise mechanisms explaining a hypertensive effect of low chronic exposure to environmental lead are unknown. An inverse association between estimated glomerular filtration rate and blood lead has been observed at blood lead levels < 5 µg/dL in general population studies (Ekong et al. 2006; Muntner et al. 2005), indicating that lead-induced reductions in renal function could play a major role in hypertension. Other potential mechanisms include enhanced oxidative stress (Stohs and Bagchi 1995; Vaziri et al. 2001), stimulation of the renin-angiotensin system (Carmignani et al. 1999; Rodriguez-Iturbe et al. 2005), and down-regulation of nitric oxide (Ding et al. 1998; Dursun et al. 2005) and soluble guanylate cyclase (Farmand et al. 2005). These mechanisms could result in increased vascular tone and peripheral vascular resistance (U.S. EPA 2006).

Causal inference.We conclude that the evidence is sufficient to infer a causal relationship between lead exposure and high blood pressure. Further research is still needed to determine the precise dose–response relationship, the relative importance of short-term versus chronic lead effects, the relevant mechanisms at environmental levels of exposure, and whether the magnitude of the association is different in children or in other vulnerable population subgroups.

Clinical cardiovascular end points in general populations.Consistency and temporality.Few cohort studies have evaluated the prospective association of lead with clinical cardiovascular outcomes in general population settings. The findings of the NHANES II and NHANES III Mortality Follow-up studies are remarkable. NHANES are periodic, standardized surveys designed to provide representative health data from the U.S. noninstitutionalized population. Despite a marked decline in lead levels in U.S. adults, both surveys showed statistically significant increases in cardiovascular mortality with increasing blood lead (Lustberg and Silbergeld 2002; Schober et al. 2006). In addition a cross-sectional analysis of NHANES 1999–2002 data identified an association of blood lead with the prevalence of peripheral arterial disease (Muntner et al. 2005; Navas-Acien et al. 2004). The British Regional Heart Study (Pocock et al. 1988) and two other small cohort studies (Kromhout 1988; Mřller and Kristensen 1992) showed positive but nonstatistically significant associations of coronary heart disease or stroke incidence with higher lead levels. The confidence intervals from these studies were wide but included the point estimates of the NHANES studies. Additional studies are needed to determine the consistency of the evidence in diverse populations.

Strength of the association and dose response.The associations of blood lead with clinical cardiovascular end points in the NHANES studies were moderately strong, with a clear dose–response gradient. An unresolved issue is the impact of uncontrolled confounding and measurement error on the relative risk estimates in studies of lead and clinical cardiovascular end points. NHANES studies adjusted for race, education, income, and urban versus rural location, which reduces potential confounding by socioeconomic status. Studies with more detailed information on the determinants of lead exposure may contribute to a better understanding of this issue. Similarly, evaluating lead effects using a single blood lead measure may result in measurement error with substantial underestimation of the magnitude of the association. This is particularly problematic when there are marked temporal trends in lead levels, as this source of error adds to within-person variability in blood lead levels to increase regression-dilution bias.

Biologic plausibility and experimental data.Lead levels of 0.8 ppm (Revis et al. 1981) and 0.1 ppm (Minaii et al. 2002) in drinking water induced atherosclerosis in animal models, and lead levels of 0.5–10 µM induced the proliferation of vascular smooth cells and fibroblasts in in vitro models (Fujiwara et al. 1995). Lead-related atherosclerosis could be explained by several mechanisms, including increases in blood pressure, impairment of renal function (Ekong et al. 2006), and induction of oxidative stress (Stohs and Bagchi 1995; Vaziri et al. 2001), inflammation (Heo et al. 1996), and endothelial dysfunction (Vaziri et al. 2001).

Causal inference.Because of the scarce number of prospective studies and the lack of information on incident nonfatal events, we conclude that the evidence is suggestive but not sufficient to infer a causal relationship with clinical cardiovascular end points. Prospective studies are required to characterize fully the impact of lead on cardiovascular morbidity and mortality. These studies need detailed and repeated assessment of lead exposure and its determinants, standardized assessment of traditional cardiovascular risk factors, and long-term follow-up to identify incident cardiovascular events and trends in subclinical markers of atherosclerosis. Although elevated blood pressure and impaired renal function are proposed mechanisms that mediate the effects of lead on clinical cardiovascular outcomes, other mechanisms are likely to be involved. Future epidemiologic studies should explore in detail the magnitude of the contribution of specific mediators of clinical cardiovascular lead effects.

Cardiovascular mortality in occupational populations.Adequacy of the evidence.The validity of occupational studies of lead and cardiovascular mortality is limited by several methodologic problems. A major limitation is the healthy worker effect (Arrighi and Hertz-Picciotto 1994). The comparison of exposed workers with the general population is particularly inappropriate for cardiovascular mortality because workers are healthier and their lifestyles and cardiovascular risk factors are likely to differ widely from those of the general population (Choi 1992). In addition, cardiovascular diseases are associated with prolonged disability and changes in employment status. Even in studies based on comparisons with unexposed workers, the selection of healthier individuals at time of hire or for specific jobs within an industry may have resulted in biased estimates of the association. Correcting the bias introduced by the healthy worker survivor effect is extremely challenging, and stratifying by duration of employment or time since hire is unlikely to completely account for this source of bias (Arrighi and Hertz-Picciotto 1994; Howe et al. 1988).

Additional limitations include the assignment of lead exposure based on job titles and of cardiovascular deaths based on death certificates. Misclassification of exposure and outcome may have resulted in further underestimation of the association of lead and cardiovascular end points. Finally, the lack of determinations of established cardiovascular risk factors and of other occupational exposures may have contributed to uncontrolled confounding.

Causal inference.As a result of these methodologic limitations, and despite many occupational cohort studies published in the literature (Table 3), available information on occupational lead exposure and cardiovascular mortality is inadequate to infer the presence or absence of a causal relationship. Because studies of environmental lead exposure provide evidence of an association between lead and cardiovascular mortality at lower exposures than those experienced by occupationally exposed workers, we expect the impact of lead in exposed workers to be at least as important as in environmentally exposed subjects.

Lead exposure and heart rate variability. Consistency, temporality, and strength of the association.Several studies, mostly cross-sectional, found an association between increased lead exposure and decreased heart rate variability. The diversity in the methods and conditions used for measuring heart rate variability makes it difficult to compare the association of lead exposure and heart rate variability across studies. In addition, the validity and precision of these studies are often limited by small sample sizes, limitations in the assessment of lead exposure, and lack of control for established cardiovascular risk factors and other confounders.

Appendix A.

Appendix B.

Appendix C.

Biologic plausibility and experimental data. Lead, a well-established neurotoxicant, could affect heart rate variability by interfering in autonomic nervous control of the heart (Chang et al. 2005). Heart rate variability measures the fluctuation of the heart rate around the mean heart rate (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology 1996). Because the basis of normal cardiac autonomic functioning is the shift from parasympathetic to sympathetic modulation, decreased heart rate variability is a marker of cardiac autonomic dysfunction. Indeed, decreased heart rate variability in supine position and in response to postural change has been associated with increased incident coronary heart disease and all-cause mortality in large prospective cohort studies in populations free of cardiovascular disease (Liao et al. 1997; Tsuji et al. 1996).

Causal inference.We conclude that the evidence is suggestive of but not sufficient to infer a causal relationship of lead exposure with heart rate variability. Large studies with adequate measures of lead exposure and standardized assessment of heart rate variability are needed to better characterize the association between lead exposure and autonomic cardiac control.

Public health implications. The evidence in this systematic review is sufficient to infer a causal relationship of lead exposure with elevated blood pressure, and it is suggestive of but not sufficient to infer a causal relationship of lead with clinical cardiovascular outcomes and cardiovascular function tests. These associations have been observed at blood lead levels well below 5 µg/dL (Menke et al. 2006; Nawrot and Staessen 2006). Indeed, no lower threshold has been established for any lead-cardiovascular association.

Although future research will contribute to characterize fully the impact of lead exposure on cardiovascular health, these findings have several important public health implications. First, there is an immediate need to lower the current safety standard of the World Health Organization and the U.S. Occupational Safety and Health Administration for blood lead in workers (currently established at 40 µg/dL). Second, a criterion for elevated blood lead levels in adults needs to be established and screened for in preventive services. In fact, the cardiovascular end points described above plus the substantial evidence that chronic lead exposure affects cognitive function (Shih et al. 2007) and renal function (Ekong et al. 2006) at levels < 5 µg/dL indicate that the U.S. Centers for Disease Control and Prevention criterion for elevated blood levels in children (10 µg/dL) is too high for adults. Third, the hypertensive effects of lead exposure and its impact on cardiovascular mortality need to be included in risk assessment and in economic analyses of lead exposure impact. Finally, regulatory and public health interventions must be developed and implemented to prevent and reduce lead exposure beyond current levels in adults.