Document 99NVvqEGpwzDBOz665x0zEoE5

UNITED STATES ENVIRONMENTAL PROTECTION AGENCY THE RELATIONSHIP BETWEEN BLOOD LEAD LEVELS AND BLOOD PRESSURE Joel Schwartz \ Concerns about the health effects of ambient exposure to lead traditionally have focused on children. Although lead has a variety of adverse effects on the health of adults, most of them appear not to be of substantial concern except at very high bloodlead levels. Recently, however, two new and extensive analyses of the NHANES II data set have shown a strong and robust relation ship between blood lead levels and blood pressure. That finding has important implications for the benefits of reducing lead in gasoline, because high blood pressure, in turn, is linked to a variety of cardiovascular diseases. This paper analyzes the statistical relationship between blood lead and blood pressure. The first part provides a brief overview of earlier studies on the subject, while the second part provides a detailed discussion of a recently completed statistical analysis of the NHANES IT data. A.1. Earlier Studies Lead has long been associated with effects on blood pressure and the cardiovascular system, including a paper in the British Medical Journal by Lorimer in 1B86 that found that higher bloodlead levels increased the risk of hypertension. Most of the N31040 TEH 0412806 2- - studies have focused.only cn hypertension and relatively high lead-exposure levels, and have not looked for a continuous effect of lead on blood pressure. Investigators reporting ^uch ah effect include Beevers et al. (1980), Morgan (1976), Richet et el. (1966), and Dingwal1-Fordyce and Lane (1963). Others have failed to find effects of lead on hypertension that were signifi cant at the 95 percent confidence level, although most of them did find a positive association. These include Ramirez-Cervantes et al. (1978) and Fouts and Page (1942). More recently, Batuman et al. (1983) found an association between chelatable body-lead levels and hypertension in veterans, and several recent general population studies and lower leadexposure studies (Beevers et al., 1976; Kromhout and Coulande, 1984 } have f-'und a significant association with blood lead. Moreau et al. (1982) also found a significant relationship (p < 0.001) between blood lead levels and a continuous measure of blood pressure in 431 French policemen, after controlling for age, body mass index, smoking, and drinking. An even more recent British study (Pocock et al., in press) found blood lead significantly related to blood pressure at the 99 percent confidence level, but the authors felt that the small size of their correlation coefficient suggested no noticeable effect. However, that conclusion appears to reflect a misunder standing of statistics. It is the regression coefficient that indicates the size of an effect. A correlation coefficient t e H 0412807 i DUP050453863 3- cor.foun.is that measurement with the variances of the dependent and independent variables. While their full data set was .not available to us, Pocock et al. presented their grouped data, and we were able to perform a regression of blood pressure versus \ the log of blood lead on their group averages, both before and after adjustment for confounders. The regressions were weighted by the inverse of the variance of each group, and con firmed their finding that blood lead was a significant predictor of blood pressure in their data, both before and after adjusting for covariates. Moreover, the regression coefficient indicated that the size of the effect was significant, suggesting a change of 3 mm Hg {millimeters of mercury, the standard measure of blood pressure) as blood lead goes from 5 to 15 ug/dl. Weeden (1975) found lead associated with the vascular renal changes linked to essential hypertension, indicating a possible causal pathway. Cooper and Gaffey (1974) analyzed mortality data from 1,267 death certificates for 7,032 lead workers employed between 1900 and 1969, and found a significant l excess of deaths from hypertension disease and renal disease. A later analysis of similar data from 1971 to 1975 also found an increase in cardiovascular and renal disease, but it was no longer significant at the 95 percent confidence level (Cooper, 1981). Animal data also link lead to hypertension. Victery (1982) found lead associated with a significant elevation of blood pres_r- ir rat. with Mood lead levels of 41 ug/dl. Importantly, TEH 0412808 DUP050453864 this study confirms Beevers et al.'s finding of a sex differen tia.!, with male but not female rats becoming hypertensive. Webb (1981) ex-arined the vascular responsiveness of tail arteries in rats exposed to blood lead levels in the 40 ug/dl range that hac suffered increases in systolic blood pressure, and found that the arteries in exposed rats had increased contraction in response to stimulation by neurotransmitters. lannaccone et al. (1981) also reported increased blood pressure (p < 0.001) in rats at blood lead levels of 38.4 ug/dl, as well as significant increases in the blood pressure response to noradrenalin. Perry and Erlanger (1979) found that low level exposure of rats to lead produced increases of 15-20 mm Hg in systolic blood pressure. Kopp (1980) repeated those findings and found electrocardiogram changes, indicating an effect on the heart itself. Subsequent tissue analysis of the heart showed reduced levels of ATP in the heart muscle, indicating that heme synthesis inhibition by lead was affecting energy availability in the heart itself. % The direct cardiological effects of lead are also indicated by electrocardiogram changes in lead-poisoned children, which are reversed by chelation therapy (Freeman, 1965; Silver and Rodrigues-Torres, 1968). Williams (1977, 1978, 1979) has shown persistant increased susceptibility to norepinephrine- induced arrhythmias in rats exposed to lead in the first free weeks of i;fe. TEH 0412809 DUP050453865 A.2. Analysis of NHANES II Data `-In light of these indications of potential effects of lead on the cardiovascular system, the relationship between blood lead levels and blood pressure has recently been explored (Harlan et al., 1985; Pirkle et al., 1985) using the NHANES II data. The NHANES II i an excellent data base for this analysis because of the care given to accurate measurements, the great range of information on possible confounding factors, and because it is a representative sample of the U.S. population. As such it avoids the problems of selection bias, healthy-worker effect, other occupational exposures, and the choice of controls that confound many occupational studies. Harlan et al. found blood lead related t'-> pressure for males aged 12 to 74 after controlling for the traditional variables associated with blood pressure (age, age-squared, body mass index, race) as well as alcohol consumption, socio-economic factors, and all nutritional variables suspected of affecting blood pressure. Moreover, this relationship held in each year of the NHANES II sample, when analyzed separately, and this relationship held for both blacks and whites. Pirkle et el. found that blood lead levels were a statis tically significant predictor of blood pressure in adult, males. This relationship held not only when blood lead was evaluated in a regression with all known factors that have previously been established as correlated with blood pressure, but also when that relationship subsequently was tested against 87 additional TEH 0412810 DUP050453866 -6- variables representing linear and nonlinear functions of"every dietary and serologic variable in the NHANES II survey. ; A.2.a. Blood Pressure Measurements Three blood pressure measurements were taken during NHANES II. A seated measurement was taken as soon as the examinee entered. Later,, a recumbent measurement was taken. A second seated measurement was taken just before the end of the examina tion. It is standard medical practice to prefer the second seated measurement, because nervousness on just entering a medical examination center makes the first seated measurement less stable. All of the results presented are for the second seated measurement. However, almost all of the regressions and robustness tests described were performed on all three measure ments, and on the average of the first and third seated measure ments; all of the conclusions concerning lead's significance held for all eight regressions (four diastolic, four systolic). fc A.2.b. Initial Analysis After replicating the Harlan et al. results for all adult males, the first goal was to determine if blood lead levels were related significantly to blood pressure in white males, 40 to 59 years old. This subgroup was chosen because at lower ages both blood pressure and blood lead vary with age. This collinearity could artificially mask or enhance the correlation between blood TEH 0412811 DUP050453867 7- - lead and bleed pressure. Between 40 and 59 years of age, however, blood pressure is essentially independent of age. Choosing ithis subgroup avoids any collinearity problems. We focused on'whites because data relating cardiovascular disease to blood pressure are less extensive for nonwhites. The established correlates of blood pressure are age, sex, race, and one of the indices of relative height-to-weight. Body mass index (BMI * weight/height^) was used in this- analysis. By limiting attention to 40 to 59 year old white males, there was no need to control for race, sex, or, to a large degree, age. Although aye was only occasionally signi ficant in the stepwise analysis, both age and age-sguared were forced into each multiple regression model to be certain any effect of lead was independent of age. The natural log of blood lead was more normally distributed, was more statistically significant, and gave a higher r 2 than untransforned blood lead, blood-lead-squared, blood lead plus blood-lead-squared, the square root of blood lead, or blood lead to other fractional powers {0.15, 0.2, 0.3, 0.4). All of the results reported here are for the natural log of blood lead, but regressions using lead on the untransformed scale gave very similar results. The initial regressions analyzed systolic and diastolic blood pressures for white males, 40 tc 59 years old, with a model consisting of age, age-squared, BMI, and blood lead. These regressions were done to determine whether blood lead levels were significantly associated with systolic and diastolic blood pres sures after controlling for age, sex, race, and BMI, which are r TEH 0412812 DUP050453868 8- - wel l-docu'-ented correlates of blood pressure. Lead wasstatistially significant (p < 0.01) for both systolic and dia-Sto-lic blood pressures in all the regressions (unweighted, weighted, and weighted "with design effects). The regressions also tested whether this relationship held up when other potentially confound ing variables were considered. A.3. Tests of Robustness The regression models were expanded to incorporate additional variables, with particular attention directed to the stability and significance of the lead coefficient in the presence of nutri tional factors and blood biochemistries. A large set of nutritional and biochemical variables from NHANES II was included in the stepwise regressions. Additional regression analyses considered potential problems of interaction terms. In ensure the robustness of the relationships, further analyses were done to address marginally insignificant variables and nonnutrition variables. Although our analysis focused on males aged 40 to 59, additional regressions were #lso performed considering all males over age 20. A.3.a. Nutritional and Biochemical Variables To provide an unusually rigorous test of the independent significance of blood lead, almost all of the nutritional and bio chemical variables in the NHANES II were included in stepwise regressions. In addition, to account for possible curvilinear relationships, squared and natural logarithmic transformations TEH 0*12813 i D UP050453869 9- - of almost all of these variables were also included. The vari ables are listed in Table 1. The objective was not to evaluate the_ possible association of nutritional or biochemical measure ments' with blood pressure, but rather to conservatively estimate the strength and independence of the relationship between blood pressure and blood lead. Including these additional 87 variables increases the proba bility of variables being found statistically significant due to chance alone. This complicates the interpretation of nutritional and biochemical factors, but not the interpretation of the lead variable; it only makes it more difficult for lead to maintain its significance. The general procedure for variable selection was as follows. First, weighted stepwise multiple linear regression was used to determine which variables were significantly related (p < 0.05) tbl->c? pressure 'u:r_: the Stepwise and NAXR options of the SAS procedure, STEPWISE). The MAXR procedure was the principle one used; it determines for any given model size (i.e., number of * variables) the variables that explain the greatest amount of the variance (i.e., maximize R^}. We chose the largest model with all variables significantly related to blood pressure (p < 0.05). The Stepwise option, which uses forward selection with backwards elimination, chose very similar models, and also always chose blood lead. From the 87 nutritional and biochemical variables, the weighted stepwise regression selected five additional variables for diastolic pressure and six additional variables for systolic pressure using a 5 percent significance test. These were used TEH 0412814 DUP050453870 TABLE 1. Variables Included in the Stepwise Repression Analyses age * age-squared * body mass index dietary sodium t salt shaker sodium dietary sodium X salt shaker sodium * dietary potassium t dietary sodium - potassium ratio dietary calcium t dietary phosphorus t dietary pr^te jn + dietary fat t dietary carbohydrate t dietary cholesterol t dietary saturated fatty acids t dietary oleic acid t dietary linoleic acid t dietary iron t - dietary vitamin A + ; dietary vitamin C t dietary thiamine + dietary riboflavin t dietary niacin t serum cholesterol* serum vitamin C t serum iron + serum transferrin saturation serum zinc t serum copper t serum albumin t hemoglobin + red Mood cell count ethanol consumption / week t cigarettes smoked / day total dietary grams t total dietary calories + cigar or pipe smoking * forced into each regression to remove any possible age effects on blood pressure. + the natural log and squared transformation of these variables were also included in the stepwise regression. TEH 0412815 DUP050453871 as the starting model for the SAS procedure SURREGR, which addi tionally incorporated the survey design effects. For both sys_t.olic and diastolic blood pressures, one variable from the weighted-stepwise regression failed to maintain significance at the 5 percent level after the design effects were incorporated. The final regression results for systolic and diastolic pressures, after accounting for the weighting and design effects, are given in Table 2. The multiple logistic regressions (of the probability of hypertension) were also performed using programs from SAS. The procedure LOG 1ST was used for the stepwise unweighted multiple logistic regression, and the procedure NLIN (nonlinear regression) was used for the weighted logistic regression calculations. The selection process again chose the largest significant model that explained the greatest amount of the variance. Calculations of threshold levels for effects were mace using the procedure NLIN on segmented regression models, which finds the threshold point that minimizes the sum of the squares of the error terms. The * results of the logistic regression on hypertension are shown in Table 3. Note that the logistic regressions included blacks as well as whites, because these regressions were used only to predict the effect of lead on the probability of having hyperten sion and were not used to estimate the number of cardiovascular diseases and deaths. As noted earlier, blacks were not included in the linear regressions because the best available coefficient frr gr-dirting cardiovascular risks included insufficient numbers of blacks. TEH 0412816 DUP050453872  12- XABLE 2. Regression of Diastolic and Systolic Blood Pressures in White Males Aged 40 to 59 Variable Diastolic Coefficient t-statistic Probability Age Age2 Body Mass Index Log(blood lead) Dietary Potassium Hemoglobin Albumin Log(dietary ' vitamin C) 0.2768 -0.0014 1.131 3.954 -0.0018 1.548 3.587 1.838 0.17 0.10 8.55 2.85 4.92 3.90 2.50 4.65 0.8636 ; 0.9321 (0.0080\> 070001 0.0005 0.0179 0.0001 Systolic Age Age2 Body Mass Index Leg(blood lead) Albumin Log(dietary Vitamin C) Log (dietary riboflavin) Log(dietary oleic acid) Log(serur vitamin C) 1.311 -0.0068 1.736 8.436 7.088 2.411 -5.509 3.992 -3.472 0.57 0.30 9.42 3.24 2.50 3.84 3.07 2.49 2.47 0.5720 0.7706 <3.00283 0.017^ 0.0005 0.0044 0.0183 0.0184 TEH 0412817 DUP050453873 -13- TASIE 3. weighted Logistic Regression on Probability of Diastolic Blood Pressure Greater Than or Equal to 90 'pin Hg in Men Aged 40 to 59 Variable Coefficient t-statistic p-Value Constant Log(Blood Lead) Albumin Body Mass Index Hemoglobin Log(Vitamin C) Dietary Potassium Total Carbohydrates -16.41 0.693 0.0873 1.700 0.0329 0.3585 -0.00058 0.00246 10.13 y 3.96 3.70 9.34 5.25 5.98 7.47 3.09 ` 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0010 TEH 0412818 1 DUP050453874 14- After including the nutritional variables, the blood analytes, and their curvilinear transformations, lead remained significantly associated (p < 0.01) with both systolic and diastolic blood pressures. The magnitude of this relationship, adjusted for the other significant variables, is shown graphically in -Figures 1 and 2'. Furthermore, segmented regression analyses indicated there was no threshold blood lead level in the data. These segmented "hockey stick" regressions fit two regression lines to the data. One, below the putative blood lead threshold T, depends on all the variables except lead. The other, for blood lead levels above T, includes lead. An iterative technique is used to find the value of T that minimizes the sum of the squares of the error terms over the full range of both regression lines. In this case, the error in the regression was minimized at a threshold of zero; that is, lead was significantly related to blood pressure at all levels down to zero. A.3.b. Interaction Terms In multiple regression analysis, another consideration is I the possibility of significant interaction terms. To evaluate this possibility, an additional weighted stepwise regression analysis was done for systolic and diastolic blood pressures. The variables consisted of the linear interaction terms between the final variables in the model (shown in Table 2) and the linear form of all the other variables originally selected for the initial stepwise regression, including their log and square transforms (Table 1). This meant running a stepwise regression with 162 interaction terms added to the final regression models TEH 04T2819 DUP050453875 -17- for systolic end diastolic pressures. Using such a large set of variables gave a high probability that some variables" wou'ld enter at the 5 percent level by chance. However, the purpose was riot to determine if those variables were independently significant, but, rather, to further test the significance and independence of the relationship between blood pressure and blood lead. As expected, several interaction variables entered the systolic and diastolic regressions, but in each regression the lead coefficient varied less than 10 percent and remained significant (p < 0.015). A.3.c. Marginally Insignificant Variables Three other analyses were done to ensure that this relationship was robust. First, the original weighted stepwise regression was extended to include variables significant through the 15 percent level to see if marginally insignificant variables influenced the significance of lead. For both systolic and dia stolic pressures, lead remained significant and there was little change in the magnitude of the coefficient. Second, for diastolic blood pressure, all the variables were included that were significant between the p = 0.05 and the 0.15 levels, and every possible combination of those variables was considered. All 255 combinations were added to the variables that were statistically significant, and a regression was performed on each one. The coefficient of the log of blood lead varied by only plus cr minus 10 percent from the value we obtained when we included only significant variables, and the highest p-value for lead was still less than 0.01. TEH 0412820 DUP050453876 -18- The last analysis was the r-ost demanding test of the indepen dence of the relationship between blood pressure and blood lead. Models for diastolic and systolic blood pressures were fit by weighted stepwise regression to the original model variables {Table, 1), excluding lead. This gave all of the other variables and their curvilinear transformations the maximum opportunity to explain variation that could also be explained by lead. After obtaining this new final model without lead, a single regression was run adding the lead variable to the variables of this new final model. For both systolic and diastolic pressures, lead was still statistically significant {p < 0.016} and the magnitude of the lead coefficient changed less than 10 percent from those ob tained in the original analysis. The results of all these analyses indicated that the strength and independence of the relationship between blood pressure and Mood lead were remarkably stable. Because some people have found small amounts of ethanol associated with reduced blood pressure, ethanol was also modeled as a quadratic function of consumption, and with two dummy variables for light and heavy drinking. The stepwise regression was repeated, with no change. A.3.d. Konnutrition Variables Pirkle et al. then considered nonnutrition variables that might be associated with blood pressure. In additional runs completed since then, we have added several other variables; The complete set is shown in Table 4. Socio-economic and demo graphic factors as well as additional medical history variables were included. TEH 0412821 I DUP050453877 19- TABLE 4. Nonnutrition Variables Tested in the Stepwise. Regression Demographic Variables Family Income Poverty I-ndex Regiorv of the Country Season of the Year Degree of Urbanization Residence Inside*Central City Educational Level Other Personal-History Variables Tricep Skinfold Subscapular Skinfold Recreational Exercrise Work-Related Exercise Recent Weight Loss Family History of Hypertension Kidney Disease Serum Creatinine Hypertension Variables Hypertensive Medication Low Salt Diet t TEH 0412822 DUP050453878 -20- Hypertension medication and low salt diet were tested -not for inclusion in a final model, as they are essentially 4 indicators of high blood pressure, but rather to see if the response.to lead differed in those groups. The coefficient of lead did not change appreciably, and lead interaction terms with the two variables were insignificant. The other variables in Table 4 were tested in two ways. First, the stepwise regression procedure was repeated with them using all nutritional and serum measurements that were significant at the p * 0.15 level. The nutritional factors were limited to these significant at the 0.15 level to give the nonnutritional factors a greater chance to enter the model. Again, lead was selected (p < 0.005) and its coefficient changed by less than 10 percent from the original model that included only age, age2, : ~~.j -ass index. 7 -t variables in Taile A were then added to al1 of those on Table V-l (including their nonlinear transforms) and the step wise process was repeated -- with the same results. Finally, the stepwise procedure was rerun using all of the*variables in Tables 1 and 4 except lead; lead was then inserted into the model resulting from this procedure. It was still significant (p < 0.006), with less than a 10 percent change in its coefficient. Because the presence of two terms to describe the curvilinear dependence of blood pressure on age might reduce the chances of variable correlated with age achieving significance, age was modeled as a single curvilinear function (sine of age), and the stepwise TEH 0412823 DUP050453879 -21- regressicn repeated; the results were the same. In addition, smoking and drinking were forced into the regression, and le.ad was still significant (p < 0.01), with only a 3 percent change in its coefficient. Our previous studies have shown that about half, of the lead in people in the NHANES II sample came from gasoline. Tetraethyl lead has very little cadmium in it, so confounding with cadmium {which is also suspected of affecting blood pressure) is unlikely. However, we repeated the regression excluding occupationally exposed workers, who may also have cadmium exposure. Lead remained significant (p < 0.01), and its coefficient increased somewhat. We also regr.essed gasoline lead directly on blood pressure, and it was significant. Although all of these analyses make it clear that collinearity is not a problem in these regressions, variance inflation factors were co'-jruted; no significant variable had a variance inflation factor above 1.4. (Variance inflation factors below 4 are considered acceptable in multiple regression analyses.) To ensure that the significance of lead in the regression was not due to the presence of a few influential observations, influence diagnostic procedures were run. Studentized residuals were plotted for all the observations, and the largest residuals were clustered near the middle of the data, where their influence is slight. Cook's D statistics also were computed for each obser vation. The highest Cook's D was 0.029, and the second highest was 0.023, both of which are very small. Moreover, of the 10 observations with the largest Cook's D statistics, six had posi tive residuals and four had negative residuals, indicating that TEH 0412824 DUP050453880 -2 2- the most influential observations split almost evenly on which way they would influence the lead regression coefficient* A. 3.e. Other Age Groups The.40 to 59 year old age group represents about one-third of adult males, and is the only one where the confounding of age and blood lead can be eliminated unambiguously. Additional regressions were performed, however, to confirm the Harlan et al. finding of an ef'fect in all adult males. Tables 1 and 4 contain several variables that Harlan et al. did not consider in their analysis. Therefore, the stepwise regression analysis was repeated using all of the variables in both tables, and their square and natural log transforms as indicated. All males over the age of 20 were considered. Lead was selected by the regres sion, with a p-value less than 0.01. To check whether the rela tionship might be substantially different for different age gr u; s, dummy variables for each 10-year age group (between 20 and 70 years), and interaction terms between lead and those dummy variables, were inserted in the stepwise regression. Such interaction terms check for differences in the lead/blood pressure relationship without having to subdivide the sample. None of the interaction terms was significant at even the p * 0.15 level. A.4. Summary of Blood Lead - Blood Pressure Results The final models for blood pressure, including all statistically significant variables, are shown in Table 5. The final logistic model for the probability of hypertension is shown in Table 6. TEH 0412825 ______ .... & DUP050453881  23- TABLE 5. Regression of Diastolic and Systolic Blood Press-res in White Males Aged 40 to 59 Variable Diastolic Coefficient F-Statistic Probabil ity * ` Age Age-squared Body Mass Index Blood leadt Potassium Hemoglobin Albumin Dietary Vitamin C+ Family history of hypertension Recreational ' exercise -0.210 0.003 \ 4.609 \ -6.002 0.151 0.354 1.886 2.085 -1.851 0.02 0.04 67.88 12.19 25.30 16.81 7.42 23.67 4.37 9.48 0.8960 0.8373 0.0000 < 6.0014 0.0000 0.0003 0.0104 0.0000 0.0446 0.0042 Systolic Age Age-squared Body Mass Index Blood leadt Albumin Dietary Vitamin C+ Dietary Riboflavin* Dietary Oleic Acid* Serum Vitamin Ct ' v.iei -e hypertension 1.142 -0.005 r87510 0.695 2.458 -5.101 3.650 3.365 3.6;3 I 0.25 0.05 85.90 10.54 6.09 13.78 8.14 5.34 5.81 ,2 0.6226 0.8208 0.0000 0.0027 0.0192 0.0008 0.0075 0.0275 0.0218 M."30Q 1 log transform TEH 0412826 DUP050453882 -24- TABLE 6. Logistic Regression on Probablity of Blood Pressure Greater Than or Equal to 90 mm Hg in Hen Aged 40 to 59________________________________ Variable Coefficient t-Statistic p-Value Constant , Log(Blood Lead} Albumin Body Mass Index Hemoglobin Log(Vitarr-in C) Dietary Potassium Total Carbohydrates Recreational Exercise -15.40 0.793 0.650 0.1571 0.0265 0.3593 -0.00053 0.00286 0.3864 7.0 3.20 2.06 6.57 3.19 4.22 5.33 2.86 0.128 0.0000 0.0014 0.0399 0.0000 0.0015 0.0000 0.0000 0.0080 0.0026 % TEH 0412827 DUP050453883 It is noteworthy that the logarithmic form of the doseresponse relationship suggests a large initial effeet., leveling off'at higher blood-lead levels. This may explain why only about 60 percent of the occupational studies (i.e., high lead-exposure studies) have found an effect that was significant at the 95 percent confidence level, while almost all of the studies of lower lead levels have found the relationship to be significant. The other low-exposure studies, the animal data, and the robustness of these results suggest that the relationship is causal. Moreover,' specific analyses to determine whether there is a lower threshold below which lead has no effect on blood pres sure showed that the data were fit best with a threshold of zero. V TEH 0412828 DUP050453884