Association of blood lead to blood pressure in men aged 55 to 75 years: effect of selected social and biochemical confounders. NFR Study Group.

The association of blood lead (B-Pb) concentration to blood pressure was investigated in men aged 55 to 75 years living in the Rome area, who had no history of exposure to lead in the workplace and who participated between 1989 and 1990 in an epidemiologic survey for coronary heart disease (New Risk Factor Project). Of the 1856 individuals eligible for the study, 59 were excluded from analyses because not all relevant data were available; and 478 were excluded because they were treated for hypertension. In the remaining subjects (n = 1319) the median B-Pb concentration was 113 micrograms/l (range: 40-442 micrograms/l). Systolic blood pressure (SBP) averaged 140 +/- 18 (standard deviation) mm Hg (range 98-220) and diastolic blood pressure (DBP) 84 +/- 9 mm Hg (range 56-118). Median B-Pb values increased significantly from 111 micrograms/l in subjects with normal blood pressure (n = 668) to 113.5 micrograms/l in subjects with borderline high blood pressure (n = 373) and to 120 micrograms/l in subjects with increased blood pressure (n = 278). After log-normal conversion of B-Pb, the linear correlation coefficient between In[B-Pb(ug/l)] and both SBP and DBP was statistically significant (r = 0.1332, p < 0.001 and r = 0.0737, p = 0.007, respectively). The linear regression coefficient was 6.8 mm Hg/In(micrograms/l) for SBP and 1.8 mm Hg/In(microgram/l) for DBP. Multiple regression analyses revealed that, after correction for body mass index (BMI), age, heart rate, skinfold thickness, serum lipids, and glucose levels; blood lead was still a significant predictor of increased SBP and DBP.(ABSTRACT TRUNCATED AT 250 WORDS)


Introduction
Low to moderate increases in blood lead (B-Pb) concentrations, as those occurring in subjects living in industrialized areas even in the absence of an occupational exposure, have been related to a small or moderate increase in blood pressure (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). This positive relationship has not been observed in a large survey conducted in Wales (11).
A number of confounding factors, whose impact has not always been analyzed, have been reported to affect the entity of the blood pressure/B-Pb relation. Grandjean et al. (12) observed that the positive relationship between B-Pb and blood pressure was no longer significant when alcohol intake and hemoglobin were included in a multiple regression analysis. Staessen et al. (13) reported that the addition of gender and age as covariates to models predicting blood pressure reduced the partial correlations of systolic and diastolic blood pressure with B-Pb to a nonsignificant level. Sharp et al. (6) observed that a positive relationship between blood pressure and B-Pb was present in black males but not among nonblack males; the inclusion of alcohol use indicators in multiple regression models did not alter the relation-ship between blood pressure and B-Pb by more than 10%. In black subjects adjusting for tobacco use markedly increase the association between B-Pb and blood pressure. In addition, the relationship was particularly strong in blacks who were infrequent users of caffeine (6).
The aim of the present study was to examine the relationship between B-Pb and blood pressure in middle-aged and elderly men living within the Rome area, taking into account the potential confounding effect of selected social and biochemical factors associated to blood pressure and B-Pb.

Materials and Methods
Between June 1989 and December 1990, a total of 1856 men aged 55 to 75 years participating in the New Risk Factors (NRF) Survey underwent a comprehensive examination. The primary aim of the project was to identify risk factors which can predict coronary heart disease and total mortality.
Initially between 1979 and 1981, a total of 3395 subjects were seen. They represented the 76.5% of 4438 men enrolled from four occupational groups belonging to defined public companies or agencies located in Rome where the screening was offered at the worksite. At the time of the follow-up (1989)(1990)) 410 subjects had died; and of the remaining 2985 subjects, 1856 (62.2%) agreed to participate. The examination included a medical history; a physical examination; and measurement of height, weight, various skinfold thicknesses, and blood pressure. Furthermore, a venous blood sample was drawn after overnight fasting.
Diseases, reported by the subjects themselves, were coded according to the International Classification ofDiseases (ICD), 9th Revision.
Blood pressure measurements were performed according to World Health Organization Guidelines (14). With the subject seated, systolic blood pressure (SBP) and fifth phase diastolic blood pressure (DBP) were measured twice in the left arm with an ordinary sphygmomanometer to the nearest 2 mm Hg. Increased blood pressure was defined as a SBP at or above 160 mm Hg and/or a DBP at or above 95 mm Hg. Borderline high blood pressure was defined as a SBP below 160 mm Hg but above 140 mm Hg and/or a DBP below 95 mm Hg but above 90 mm Hg. Heart rate (HR, beats/min) was calculated from electrocardiographic traces as an average of rates in lead I and lead V6. Information about alcohol consumption and smoking habits was collected for each subject by direct interview. Alcohol consumption (ALC) was calculated for wine, beer, and spirits separately and the amount of alcohol intake was expressed in grams of absolute ethanol consumed per day. Daily cigarette consumption (CIG/D) was recorded as a measure of tobacco use. Height and weight were compacted into a body mass index (BMI, kg/M2). Triceps, biceps, subscapular, suprascapular, and suprailiac skinfold thicknesses were measured by Harpenden calipers to the nearest millimeter. The variable skinfold thickness (SKF) was defined as the sum of the five measurements.
Venous blood samples for B-Pb determination were drawn into polypropylene tubes containing K3EDTA as anticoagulant and stored at -20°C until analysis. All the equipment used was free from lead. B-Pb (pg/1) was measured by Atomic Absorption Spectrophotometry (AAS) using a Perkin-Elmer model Zeeman 5000, HGA 500. The method for B-Pb (15) determination has been previously described. The analytical variability of the method has been checked by the adoption of appropriate internal quality controls. Control charts were prepared with measurements carried out on control samples with certified titles, provided by the Community Bureau of Reference, Commission of the European Community. Between-day precision, obtained from B-Pb measurements carried out on our subjects, was 2.7%. Analytic precision and accuracy for B-Pb were also monitored by means of external quality control samples, according to procedures previously described (16,17). Serum levels of cholesterol (CHOL, mg/dl), high-density lipoprotein cholesterol (HDL, mg/dl), triglycerides (TRIG, mg/dl), and glucose (GLU, mg/dl) were determined by automated methods within 12 hr of collection, the samples having been stored at +40C until analysis. The amount of non-HDL cholesterol (N-HDL) was calculated as the difference between CHOL and HDL.
Subjects were excluded from the present study when not all relevant data were available (n = 59) or if they were treated for hypertension (n = 478). After these exclusions a sample of 1319 subjects was obtained. In further analyses subjects were divided according to alcohol consumption in drinkers (n = 1068) and nondrinkers (n = 251). Log-normal transformation of the B-Pb approached a Gaussian distribution; the use of ln [B-Pb] enabled us to perform parametric analyses of data, i.e. linear cor-relation and multiple stepwise regression. All statistical analyses of data were carried out using the BMDP Statistical Software package.
In the 1319 subjects the linear correlation coefficients between ln[B-Pb] and both SBP and DBP were statistically significant (r = 0.1332,p<0.001 and r = 0.0737, p = 0.007, respectively). The linear regression coefficient was 6.8 mm Hg/ln(pg/l) for SBP and 1.8 mm Hg/ln(jig/l) for DBP. The entity of the association determined by means of Pearson's correlation coefficient of the other considered variables to ln[B-Pb] and to both SBP and DBP, is reported in Table 2. The variables ALC and N-HDL were positively related to ln[B-Pb] and both SBP and DBP. HDL was related directly to ln[B-Pb] and SBP. TRIG and SKF were related directly to SBP and DBP and inversely to ln[B-Pb]. AGE was related directly to SBP and inversely to both DBP and ln[B-Pb]. CIG/D was related directly to ln[B-Pb] and inversely to both SBP and DBP. As some of these associations might be a possible source of positive or negative confounding bias, multivariate analyses were performed with the aim of estimating the entity and the independence of the relationship between B-Pb and blood pressure. A first set of multiple stepwise regressions analyzed SBP and DBP as dependent variables for the 1319 subjects, with a model including ln[B-Pb], AGE, BMI, HR, N-HDL, HDL, TRIG, GLU, CIG/D, ALC, and SKF as possible predictors. In these models, 16 addition of covariates to the multiple stepwise regression models predicting SBP and DBP is showed in Figure 1; after the last step of the stepwise analysis the regression coefficients were 5.6 mm Hg/ln(pg/l) for SBP and 1.7 mm Hg/ln(pg/l) for DBP.
In further analyses subjects were divided, according to alcohol consumption, into drinkers and nondrinkers.
After correction for BMI, AGE, HR, SKF, serum lipids, CIG/D, and GLU the adjusted regression coefficients were 5.6 mm Hg/ln(ig/l) for SBP and 2.5 mm Hg/ln(pg/l) for DBP.

Discussion
In our sample (n = 1319) B-Pb concentrations were significantly higher among subjects with borderline or increased blood pressure than in subjects with normal blood pressure. This positive association was confirmed by a significant direct linear correlation between B-Pb concentration and both systolic and diastolic blood pres- sure. The linear regression coefficient was 6.8 mm Hg/ln(ig/l) for SBP and 1.8 mm Hg/ln(jig/l) for DBP. The entity and the independence of these relationships were studied by means of multiple regression analyses. The partial correlation coefficients of systolic and diastolic blood pressure with B-Pb concentration remained statistically significant. Adjusted regression coefficient, after controlling for age, BMI, heart rate, serum lipids, glucose levels, smoking, and skinfold thickness was 5.6 mm Hg/ln(pig/l) for systolic and 1.7 mm Hg/ln(pg/l) for diastolic blood pressure.
The highest negative confounding was due to cigarette consumption. A similar effect of tobacco use on the blood pressure/B-Pb relationship has been reported by Sharp et al. (6).
A negative relationship between tobacco use and blood pressure (18) and a positive one between smoking and B-Pb concentration have been previously described (19). For the cohort under study, we previously reported (20) that the median B-Pb concentration was 29.1% higher in subjects smoking more than 20 cigarettes daily than in nonsmokers. Moreover, the reported smoking status, which explained 2.12% of the total variance, was the second major predictor of B-Pb concentration (20).
Volume 102, Supplement 9, November 1994 A positive confounding from alcohol consumption has been reported by other authors (12). Various studies have shown a direct relation between alcohol consumption and both blood pressure (21) and B-Pb concentration (19,22,23). A highly significant increase in B-Pb with rising alcohol consumption has been previously reported for the subjects examined in the present survey (20). Alcohol consumption, explaining 14.71% of the total variability, was by far the most important predictor of B-Pb concentration; the odds ratio on the risk of having a B-Pb higher than 180 jig/l was proportional to the level of alcohol consumption, with a 27.7-fold increased risk for heavy drinkers (alcohol intake >100 g/day); B-Pb was more specific and sensitive than HDL-cholesterol and y-glutamyltransferase, commonly considered indexes of alcohol consumption (24,25), in identifying moderate or heavy drinkers (alcohol intake >50 g/day), and also showed the highest positive predictive value.
In spite of these findings, our subjects alleged alcohol consumption, in contrast with results obtained in Denmark (12), failed to alter in a significant manner the relation between B-Pb and blood pressure.
In our subjects, the highest positive confounding was due to HDL-cholesterol. This biochemical indicator of alcohol consumption (24) decreased the systolic blood pressure/B-Pb and diastolic blood pres-sure/B-Pb regression coefficients by 15 and 23%, respectively.
The strong association of B-Pb concentration to alcohol consumption prompted us to analyze the B-Pb/blood pressure relation separately in drinkers and nondrinkers. Among non-drinkers (n = 251) -for whom 2.5 and 97.5 centiles of blood lead distribution were 55 and 181 pg/l, respectivelyno significant relationship between B-Pb and both systolic (r= 0.0014) and diastolic blood pressure (r= -0.0826) was found.
Among drinkers (n = 1068), for whom 2.5 and 97.5 centiles of B-Pb distribution 63 and 252 pg/l, respectively, the linear correlation between B-Pb and both systolic (r = 0.1449) and diastolic blood pressure (r = 0.1042) was statistically significant (p <0.001). Linear regression coefficient was 7.5 mm Hg/ln(pg/l) for SBP and 2.6 mm Hg/ln(pg/l) for DBP. After correction for body mass index, heart rate, age, HDLcholesterol, glucose and skinfold thickness the adjusted regression coefficients for SBP was 5.6 mm Hg/ln(pg/l). The adjusted regression coefficients for DBP, after correction for body mass index, heart rate, age, and smoking, was 2.5 mm Hg/ln(pg/l).
In considering our results two main comments can be made. The first comment is that B-Pb as a biochemical index of alcohol intake, at least in the absence of occupational exposure, may be a better indicator of actual alcohol use than alleged alcohol consumption (for which voluntary under-reporting cannot be excluded). According to this B-Pb would be only an indicator of alcohol intake and the effect on blood pressure would be actually due to alcohol consumption.
The second comment is that low levels of lead exposure, such as those, mainly associated to alcohol consumption and occurring in our subjects (20), might cause an increase in blood pressure. Animal data seems to support the hypothesis of a causal link. In fact, both in vitro and in vivo experimental results have strongly suggested a causal relationship between low to moderate levels of lead exposure and increases in blood pressure (26)(27)(28).
In conclusion, the adjusted regression coefficient between ln[B-Pb] and blood pressure we have found among drinkers represents an increment of about 6 and 3 mm Hg in systolic and diastolic blood pressure over the range of observed B-Pb levels (i.e., 40-442 pg/1). In terms of a slope of 2 mm Hg for a 100 pg/l change in B-Pb, the decline in B-Pb concentration of about 100 pg/l we have observed in male subjects living in the Rome area since 1979 (20), although not producing significant changes in the relative risk of cardiovascular disease, might have some importance from a public health perspective.