Body Mass Index and Associations of Sodium and Potassium With Blood Pressure in INTERSALT

Transcription

1 729 Body Mass Index and Associations of Sodium and Potassium With Blood Pressure in INTERSALT Alan R. Dyer, Paul Elliott, Martin Shipley, Rose Stamler, Jeremiah Stamler, on behalf of the INTERSALT Cooperative Research Group Abstract This report further examines the relation of body mass index (BMI) to associations of 24-hour urinary sodium, potassium, and sodium-potassium ratio with blood pressure in INTERSALT, a 52-center international study of electrolytes and blood pressure. Analyses without adjustment for BMI indicated average systolic pressure greater by 6.00 mm Hg per 100 mmol higher sodium and diastolic by 2.52 mm Hg. With adjustment for BMI, these values were reduced to 3.14 and 0.14 mm Hg, respectively. For the sodium-potassium ratio, blood pressure associations were stronger when not adjusted for BMI, and for potassium, adjustment generally had little effect. To explore possible interactions of these variables with BMI in relation to blood pressure, the 52 centers were divided into two groups of 26 based on whether the center median for BMI was less than or greater than or equal to 24.5 kg/m 2, and individuals within each of the 52 centers were classified into lower- or higher-bmi groups based on individual BMI less than or greater than or equal to 24.1 kg/m 2. Sodium and the INTERSALT is an international study of the relations of electrolyte excretion and other factors to blood pressure carried out in 52 centers from 32 countries. Previous reports from INTERSALT indicated that body mass index (BMI), 24-hour urinary sodium and potassium excretions, and the sodiumpotassium ratio were all independently related to blood pressure of individuals 13 and that sodium and potassium excretions of individuals were both positively correlated with BMI. 2 In addition, pooled regression coefficients for sodium, potassium, and sodium-potassium ratio with blood pressure that were adjusted only for age and sex when compared with the coefficients also adjusted for alcohol intake and BMI (as well as potassium for sodium, and sodium for potassium) suggested that adjustment for BMI had a substantial effect on observed electrolyte-blood pressure associations in INTERSALT. 1 Received January 24, 1994; accepted in revised form March 30, From the Department of Preventive Medicine, Northwestern University Medical School, Chicago, 111 (A.R.D., R.S., J.S.); the Department of Public Health and Policy (P.E.) and Department of Epidemiology and Population Sciences (P.E., M.S.), London (UK) School of Hygiene and Tropical Medicine; and the Department of Epidemiology and Public Health, University College and Middlesex School of Medicine, University College, London (M.S.). Correspondence to Dr Alan R. Dyer, Department of Preventive Medicine, Northwestern University Medical School, 680 N Lake Shore Dr, Suite 1102, Chicago, IL sodium-potassium ratio were positively and significantly and potassium inversely and significantly related to systolic pressure in all four of these subgroups, and the sodium-potassium ratio and potassium were related to diastolic pressure in two and three subgroups, respectively. Electrolyte-blood pressure associations did not differ significantly between the two subgroups of centers or between the two subgroups based on individuals. Although these results indicate that adjustment for BMI has an important effect on INTERSALT associations of sodium and the sodium-potassium ratio with blood pressure and may represent an overadjustment, they also indicate that each of the three variables is related to blood pressure throughout the BMI range. Thus, they do not support the concept of important interactions of sodium and potassium with BMI in relation to blood pressure. (Hypertension. 1994;23 [part l]: ) Key Words blood pressure body mass index sodium potassium Intervention studies of the effects of weight loss, reduced sodium intake, and potassium supplementation on blood pressure have been inconsistent as to whether weight loss leads to lower blood pressure with or without reduced sodium intake 414 and in at least one study with or without increased potassium intake. 5 Furthermore, although sodium reduction lowers blood pressure, 15 it is unclear whether lower sodium intake reduces blood pressure equally in overweight and nonoverweight individuals 1617 ; also, combined effects of weight and sodium reduction on blood pressure have not yet been quantified. 14 This report uses INTERSALT data to further examine BMI and associations of sodium, potassium, and sodium-potassium ratio with blood pressure by quantifying the effects of adjustment for BMI on these associations and by exploring possible interactions of these variables with BMI in relation to blood pressure. Methods Details of the INTERSALT study design and methods have been published Briefly, each of the 52 centers recruited 200 men and women aged 20 to 59 years, with 25 in each of eight age-sex groups. Samples were selected randomly from population lists or by chunk sampling of defined populations. Local investigators were trained to implement a common standardized protocol at one of five training meetings, based on a detailed Manual of Operations. 19 Data forms were sent to the Coordinating Center in London for review, editing, coding, data entry, and analysis.

2 730 Hypertension Vol 23, No 6, Part 1 June 1994 Participants in each of the 52 centers provided written or verbal (if not literate) informed consent after approval of all study procedures by local institutional review boards. All procedures were conducted in accordance with institutional guidelines. Blood pressure (sitting) was measured twice with the use of a random-zero sphygmomanometer (Hawksley) after participants had emptied their bladders and sat quietly for 5 minutes. Systolic blood pressure was recorded as the appearance of sound (phase I) and diastolic as the disappearance of sound (phase V). Blood pressure of individual participants was the mean of the two readings. Both "spot" and 24-hour urine collections were made to estimate electrolyte excretion. The start and end of each 24-hour collection were supervised by clinic staff, and completeness was assessed by a standardized interview. Urine aliquots were stored locally at -20 C before being shipped frozen to the Central Laboratory at St Raphael University in Leuven, Belgium, where all urine analyses were performed with strict internal and external quality control. Sodium and potassium excretions were measured by atomic absorption flame photometry. 20 Individual sodium and potassium excretion values were the products of concentrations in the urine and urinary volume corrected to 24 hours. Height and weight were each measured twice using a stadiometer and beam balance scale where possible. BMI was calculated as weight divided by height squared (kilograms per meter squared). Daily intake of alcoholic drinks over the preceding 7 days was assessed by questionnaire and, based on local information, converted into volume (milliliters) of absolute alcohol. Repeat urine collections and blood pressure measurements were obtained at a second visit in a random 8% of participants to estimate intraindividual variability. To quantify the effect of adjustment for BMI on electrolyteblood pressure associations, center-by-center linear regressions of blood pressure on sodium, potassium, and sodiumpotassium ratio were calculated with and without adjustment for BMI. All regressions included adjustment for age, sex, and alcohol intake. The regressions of blood pressure on sodium excretion also included 24-hour potassium excretion, and those for potassium included sodium. To adjust for alcohol, 7-day intakes were stratified into three groups (0,1 to 299, and 300+ ml of absolute alcohol per week) with two 0-1 variables, ie, no alcohol intake versus 1 to 299 ml/wk (0,1) and no intake versus 300+ ml/wk (0,1). Five of people were excluded from these analyses because of missing data on alcohol intake. In addition, the results for diastolic blood pressure with adjustment for age, sex, alcohol intake, and BMI differ slightly from those previously reported 1 because of the exclusion from regression analyses for diastolic pressure of 17 people with diastolic blood pressures recorded as less than 30 mm Hg. Within-center regression coefficients were averaged (pooled) across centers to yield study-wide estimates of associations between blood pressure and each electrolyte variable with and without adjustment for BMI. Each center coefficient was weighted by the inverse of its variance. Regression coefficients for sodium are expressed as millimeters of mercury per 100 mmol (1 mmol = l meq=23 mg) sodium, coefficients for potassium as millimeters of mercury per 30 mmol (1 mmol = l meq=39 mg) potassium, and coefficients for the sodium-potassium ratio as millimeters of mercury per 2 U. These values for sodium, potassium, and the sodium-potassium ratio were chosen for two reasons. First, 100 mmol (meq) sodium, 30 mmol (meq) potassium, and 2 U of the sodium-potassium ratio are approximately the interquartile ranges for these variables in the INTERSALT sample of people used in these analyses, ie, 97 mmol (meq) for sodium, 32 mmol (meq) for potassium, and 2.13 U for the sodium-potassium ratio. Second, these values also generally reflect the changes that would be required in sodium and potassium intake in the US population and/or the INTER- SALT sample to achieve the goals for intake suggested by the National Research Council For sodium, the National Research Council recommends a reduction in daily intake from the current US consumption of 10 to 14.5 g salt (171 to 248 mmol sodium intake) to no more than 6.0 g salt (103 mmol sodium), with a goal of 4.5 g (77 mmol sodium). 21 Hence, the average reduction in sodium intake from current levels required to achieve this goal is approximately 100 mmol (meq). For potassium, the National Research Council recommends an intake of 3500 mg (90 mmol), 22 which represents an average increase of approximately 35 mmol (meq) over current IN- TERSALT levels of intake. Changes of this magnitude in sodium and potassium intake would reduce the average sodium-potassium ratio in INTERSALT by approximately 2 U. To explore possible interactions of electrolyte variables and BMI in relation to blood pressure, two sets of analyses were performed. First, the 52 centers were divided into two groups of 26 centers based on whether the median BMI at the center was less than 24.5 kg/m 2 (lower-bmi centers) or greater than or equal to 24.5 kg/m 2 (higher-bmi centers). There were 2518 men and 2492 women in the 26 lower-bmi centers and 2524 men and 2540 women in the 26 higher-bmi centers. Within each center, linear regressions of blood pressure on each electrolyte variable were calculated and adjusted for age, sex, alcohol intake, and BMI. Regressions for sodium also included potassium, and those for potassium included sodium. Withincenter regression coefficients, weighted by the inverse of their variances, were averaged across centers to yield estimates of the associations between blood pressure and each variable within each set of 26 centers. In the second set of analyses for interactions, individuals within the 52 centers were classified as lower-bmi if their BMI was less than 24.1 kg/m 2 and higher-bmi if their BMI was greater than or equal to 24.1 kg/m 2. (The cut point for determining higher- and lower-bmi individuals was chosen to approximately equalize the number of individuals in each subgroup after exclusion of any subgroup in a center with fewer than 20 individuals.) Within each center, linear regressions of blood pressure on each electrolyte variable were calculated for individuals in each group. Regressions were not run for any BMI subgroup in a center if that subgroup contained fewer than 20 people. This resulted in 4 centers and 57 people being excluded from analyses in the higher-bmi groups. There were 2349 men and 2661 women included in the analyses for the lower-bmi groups and 2658 men and 2349 women included in the analyses for the higher-bmi groups. In these analyses, adjustment was also made for age, sex, alcohol intake, and BMI within each BMI subgroup. Potassium excretion was also included in regressions of blood pressure on sodium, and sodium in those for potassium. Regression coefficients were averaged across centers for all higher-bmi groups and all lower-bmi groups, with weights equal to the inverse of the variance. BMI was included as a possible confounder in these two sets of analyses for two reasons. First, adjustment for BMI within BMI subgroups allows comparison of the results reported here with previously reported results. 1 Second, analyses in individuals classified as lower and higher BMI constitutes an adjustment for BMI that was not present in analyses in lower- and higher-bmi centers. Hence, to make these two sets of analyses comparable, BMI was included as an independent variable in all BMI subgroup analyses. Pooled regression coefficients for electrolyte variables were corrected for reliability ("regression dilution bias") according to methods previously described for INTERSALT The estimates of reliability used in these corrections were for 24-hour sodium excretion, for 24-hour potassium excretion, and for the sodium-potassium ratio. The corrections for models containing both sodium and potassium also require an estimate of the proportion of the covariance

3 Dyer et al BMI, Sodium, Potassium, and Blood Pressure 731 TABLE 1. Pooled Regression Coefficients Relating 24-Hour Urinary Sodium and Potassium Excretions and Sodium-Potassium Ratio to Blood Pressure With and Without Adjustment for Body Mass Index Systolic Pressure Dlastollc Pressure Variable 24-Hour urinary sodium Coefficient, mm Hg/100 mmol sodium With BMI Adjustment Without BMI Adjustment Corrected coefficient* 24-Hour urinary potassium Coefficient, mm Hg/30 mmol potassium Corrected coefficient* 24-Hour urinary sodium-potassium ratio Coefficient, mm Hg/2 U Corrected coefficientt ) H BMI indicates body mass index. *Corrected for reliability of both 24-hour sodium and potassium excretions and intraindividual correlation of sodium and potassium. tcorrected for reliability of 24-hour sodium-potassium ratio. tp<.05, P<.01, P< With BMI Adjustment Without BMI Adjustment between sodium and potassium that is due to true covariation between them. The estimate used in these analyses was (This estimate indicates that approximately 49% of the observed covariance between sodium and potassium excretions in INTERS ALT was due to their measurement on the same 24-hour urine collection.) Results Electrolyte-Blood Pressure Associations With and Without Adjustment for BMI Table 1 gives the pooled regression coefficients relating 24-hour sodium (millimeters of mercury per 100 mmol) and potassium (millimeters of mercury per 30 mmol) excretions and sodium-potassium ratio (millimeters of mercury per 2 U) to systolic and diastolic blood pressure with and without adjustment for BMI. The table also gives the standard error of each coefficient, the coefficient divided by its standard error (), and the coefficient corrected for reliability ("regression dilution bias"). Without adjustment for BMI, the pooled regression coefficient relating sodium to systolic blood pressure without correction for reliability was twice as large as the coefficient from the analysis that included BMI, ie, 2.05 mm Hg for sodium intake higher by 100 mmol (meq) versus 1.00 mm Hg per 100 mmol (meq). With correction for reliability, the coefficients were 6.00 and 3.14 mm Hg per 100 mmol (meq) sodium, respectively. For sodium and diastolic pressure, a regression coefficient that was positive and significant without adjustment for BMI became nonsignificant when BMI was added to the model. With correction for reliability, the regression coefficient relating sodium to diastolic blood pressure indicated average pressure greater by 2.52 mm Hg for sodium higher by 100 mmol (meq) without adjustment for BMI and 0.14 mm Hg per 100 mmol (meq) with adjustment. For potassium, unlike for sodium, the unconnected regression coefficients for both systolic and diastolic blood pressures were smaller without adjustment for BMI than with adjustment, and the coefficient for diastolic pressure from the model without BMI did not differ significantly from zero. However, when corrected for regression dilution bias, the difference due to inclusion of BMI in the model essentially disappeared for systolic pressure and was reduced for diastolic. For systolic blood pressure, the corrected regression coefficients indicated average pressure less by 2.02 and 1.97 mm Hg for potassium excretion higher by 30 mmol (meq) for the models with and without BMI adjustment, respectively. For diastolic blood pressure, the corrected coefficients indicated average pressure less by 1.12 and 0.99 mm Hg per 30 mmol (meq) higher potassium excretion, respectively. Adjustment for BMI also had a substantial effect on the regression coefficients for the sodium-potassium ratio and blood pressure. For systolic blood pressure, both the corrected and uncorrected regression coefficients were approximately 33% larger in the model without than with adjustment for BMI, whereas for diastolic pressure, coefficients were twice as large without adjustment for BMI. With correction for reliability and no adjustment for BMI, the regression coefficient for systolic blood pressure indicated average pressure greater by 4.94 mm Hg for the sodiumpotassium ratio higher by 2 U and average diastolic pressure greater by 2.06 mm Hg. With adjustment for

4 732 Hypertension Vol 23, No 6, Part 1 June 1994 TABLE 2. Medians (25th and 75th Percentlles) of Study Variables for Body Mass Index Subgroups Variable Hlgher-BMI Individuals from lower- and higher-bmi centers No. of centers No. of men and women Systolic pressure, mm Hg (104.0,124.0) (110.0,130.0) Diastolic pressure, mm Hg Urinary Na excretion, mmol Urinary K excretion, mmo) Urinary Na-K ratio BMI, kg/m 2 Moderate drinkers, %* Heavy drinkers, %t Lower- and higher-bmi individuals No. centers No. men and women Systolic pressure, mm Hg Diastolic pressure, mm Hg Urinary Na excretion, mmol Urinary K excretion, mmol Urinary Na-K ratio BMI, kg/m 2 Moderate drinkers, %* Heavy drinkers, %t 71.0 (63.0,79.0) (94.7,203.0) 49.1 (34.8,68.2) 2.98(1.89,4.46) 22.9 (20.8,25.2) (103.0,121.0) 69.0 (61.0, 76.0) (92.8,187.6) 47.5 (34.4,64.4) 2.91 (1.94,4.16) 21.7 (20.2,22.9) (67.0,82.0) (112.3,201.2) 53.6 (39.6,69.8) 2.90(2.18,3.83) 25.7 (23.0, 29.0) * (112.0,133.0) 77.0 (70.0,84.0) (118.8,213.7) 55.5 (40.9, 72.4) 2.95 (2.19,3.99) 27.2 (25.5,30.0) BMI indicates body mass index. Fifty-two centers were divided into 26 with median BMI <24.5 kg/m 2 (lower-bmi centers) and 26 with median BMI 2:24.5 kg/m 2 (higher-bmi centers). Individuals in each center were divided into those with BMI <24.1 kg/m 2 (lower-bmi) and those with BMI 2:24.1 kg/m 2 (higher-bmi). *1 to 299 ml absolute alcohol per week. t300+ ml absolute alcohol per week. tfifty-seven higher-bmi individuals from four centers were excluded (see "Methods"). BMI and correction for reliability, the regression coefficients indicated average pressure greater by 3.69 and 1.02 mm Hg per 2 U for systolic and diastolic blood pressures, respectively. Descriptive Statistics for BMI Subgroups Table 2 gives the medians and 25th and 75th percentiles for systolic and diastolic blood pressures, 24-hour urinary sodium and potassium excretions, sodium-potassium ratio, and BMI for participants from higherand lower-bmi centers as well as for all higher- and lower-bmi individuals. Table 2 also gives the percentages of participants in these BMI subgroups who reported consuming 1 to 299 ml absolute alcohol in the week before the exam and the percentage who reported consuming 300+ ml. Medians for blood pressure, sodium and potassium excretions, and BMI (as expected) were all greater in higher-bmi centers and individuals than in lower-bmi centers and individuals. However, the differences in blood pressure and electrolytes were larger for higher- and lower-bmi individuals than for higher- and lower-bmi centers, reflecting the inclusion of both higher- and lower-bmi individuals in the higher- and lower-bmi centers. Sodium and Blood Pressure in BMI Subgroups Table 3 gives the results of the regression analyses to assess possible interactions of sodium with BMI in relation to blood pressure. With multivariate adjustment, 24-hour sodium excretion was positively and significantly related to systolic pressure in both higherand lower-bmi centers and individuals. For analyses based on centers, the regression coefficient was larger for lower-bmi centers compared with higher-bmi centers. Without correction for reliability, the pooled regression coefficients indicated average systolic pressure greater by 1.32 mm Hg for each 100-mmol (meq) increment in sodium excretion in lower-bmi centers and 0.73 mm Hg greater in higher-bmi centers, a difference that was not statistically significant. With correction for reliability, the coefficients indicated average systolic pressure greater by 3.98 and 2.41 mm Hg, respectively, for sodium higher by 100 mmol (meq). For analyses based on classification of individuals, the regression coefficient was larger in higher-bmi compared with lower-bmi individuals. The pooled regression coefficients indicated average systolic pressure greater by 1.22 and 1.00 mm Hg, respectively, per 100 mmol (meq) sodium. This difference was also not

5 Dyer et al BMI, Sodium, Potassium, and Blood Pressure 733 TABLE 3. Pooled Regression Coefficients (mm Hg per 100 mmol) Relating 24-Hour Urinary Sodium Excretion to Blood Pressure in Body Mass Index Subgroups Systolic Pressure Dlastollc Pressure Variable Lower- and higher-bmi centers Corrected coefficient! Lower- and higher-bmi individuals Corrected coefficientt P } BMI indicates body mass index. Fifty-two centers were divided into 26 with median BMI <24.5 kg/m 2 (lower-bmi centers) and 26 with median BMI 2:24.5 kg/m 2 (higher-bmi centers). Individuals in each center were divided into those with BMI <24.1 kg/m 2 (lower-bmi) and those with BMI &24.1 kg/m 2 (higher-bmi). *Adjusted for age, sex, BMI, alcohol intake, and 24-hour potassium excretion. tcorrected for reliability of both 24-hour sodium and potassium excretions and Intraindividual con-elation of sodium and potassium. tp<.05, P<.01, P< statistically significant. With correction for reliability, the regression coefficients were 3.81 and 2.96 mm Hg per 100 mmol (meq). For diastolic pressure, none of the four regression coefficients were significantly different from zero. This finding is consistent with results in Table 1 for the full INTERSALT sample showing no significant association of sodium with diastolic pressure when adjustment was made for age, sex, alcohol intake, BMI, and potassium excretion. The coefficients for sodium and diastolic pressure also did not differ significantly between lowerand higher-bmi centers or individuals. Potassium and Blood Pressure in BMI Subgroups Table 4 gives the results of regression analyses relating 24-hour potassium excretion to blood pressure in higher- and lower-bmi centers and individuals. With multivariate adjustment, potassium was inversely and significantly related to systolic pressure in both higherand lower-bmi centers and higher- and lower-bmi individuals. Potassium was also significantly related to diastolic pressure in higher-bmi centers and higherand lower-bmi individuals. Although these inverse associations for higher-bmi centers and individuals were larger than those for lower-bmi centers and individuals, the differences were not statistically significant. Without correction for reliability, the pooled regression coefficients indicated average systolic pressure less by 0.54 mm Hg per 30-mmol (meq) increment in potassium excretion in lower-bmi centers and 1.13 mm Hg less for higher-bmi centers. With correction for reliability, the coefficients indicated average pressure less by 1.52 and TABLE 4. Pooled Regression Coefficients (mm Hg per 30 mmol) Relating 24-Hour Urinary Potassium Excretion to Blood Pressure In Body Mass Index Subgroups Variable Lower- and higher-bmi centers Corrected coefficienrt Lower- and higher-bmi individuals Corrected coefficientt Systolic Pressure Q * Dlastollc Pressure * Definitions are as in Table 3. 'Adjusted for age, sex, BMI, alcohol intake, and 24-hour sodium excretion. tcorrected for reliability of both 24-hour sodium and potassium excretions and intraindividual con-elation of sodium and potassium. #><.O5, P<.01, P<.001.

6 734 Hypertension Vol 23, No 6, Part 1 June 1994 TABLE 5. Pooled Regression Coefficients (mm Hg per 2 U) Relating 24-Hour Urinary Sodium-Potassium Ratio to Blood Pressure In Body Mass Index Subgroups Systolic Pressure Dlastollc Pressure Variable Lower- and higher-bmi centers * B Corrected coefficient)" Lower- and higher-bmi individuals f) Corrected coefficientt Definitions are as in Table 3. 'Adjusted for age, sex, BMI, and alcohol intake. tcorrected for reliability of 24-hour sodium-potassium ratio. *P<.05, P<.01, P< mm Hg, respectively, per 30 mmol (meq). The comparable estimates for diastolic pressure were 0.38 and 0.69 mm Hg per 30 mmol (meq) potassium without correction for reliability and 0.87 and 1.46 mm Hg with correction. The differences in regression coefficients for higherand lower-bmi individuals were smaller than those for higher- and lower-bmi centers. The regression coefficients for systolic pressure indicated average pressure less by 0.76 and 0.86 mm Hg, respectively, in lower- and higher-bmi individuals without correction for reliability and 1.92 and 2.39 mm Hg per 30 mmol (meq) with correction. For diastolic pressure, the coefficients without correction for reliability were 0.51 and 0.56 mm Hg and were 1.05 and 1.36 mm Hg with correction. Sodium-Potassium Ratio and Blood Pressure in BMI Subgroups Table 5 gives the results of regression analyses relating the 24-hour sodium-potassium ratio to blood pressure in BMI subgroups. With multivariate adjustment, the sodium-potassium ratio was significantly related to systolic pressure in both higher- and lower-bmi centers and individuals and to diastolic pressure in lower-bmi centers and higher-bmi individuals. Although the coefficients for both systolic and diastolic blood pressure were larger in lower-bmi compared with higher-bmi centers but were larger in higher-bmi than in lower- BMI individuals, none of these differences was statistically significant. Without correction for reliability, the pooled regression coefficients indicated average systolic pressure greater by 1.50 and 1.00 mm Hg for the sodium-potassium ratio higher by 2 U for lower- and higher-bmi centers and by 1.33 and 1.41 mm Hg for lower- and higher-bmi individuals. With correction for reliability, the coefficients indicated average systolic pressure greater by 4.48 and 2.96 mm Hg in lower- and higher-bmi centers and by 3.93 and 4.19 mm Hg for lower- and higher-bmi individuals. The uncorrected coefficients for diastolic pressure ranged from 0.22 mm Hg in lower-bmi individuals to 0.57 mm Hg in higher-bmi individuals and the corrected coefficients from 0.65 to 1.69 mm Hg. Discussion The purpose of these analyses was to further examine BMI and associations of sodium, potassium, and sodium-potassium ratio with blood pressure in INTER- SALT. The relation of BMI to these associations is of interest because of the suggested effect of adjustment for BMI on the sodium-blood pressure relations in INTERSALT 1 and other studies 2528 and because of the inconsistencies in the results of intervention trials on the independent effects of sodium reduction and weight loss on blood pressure These questions were addressed by quantifying the effect of adjustment for BMI on associations of sodium, potassium, and sodiumpotassium ratio with blood pressure and by exploring possible interactions of these variables with BMI in relation to blood pressure. Adjustment for BMI had a substantial effect on the size of regression coefficients relating sodium and the sodium-potassium ratio to blood pressure. Adjustment for BMI, in addition to adjustment for age, sex, alcohol intake, and potassium excretion, reduced the uncorrected regression coefficient for sodium with systolic blood pressure by more than 50% and the corrected coefficient by almost 50%. For diastolic pressure, adjustment for BMI made a significant association (P<.001) nonsignificant. For the sodium-potassium ratio, adjustment for BMI reduced both corrected and uncorrected regression coefficients for systolic pressure by approximately 25% and for diastolic by approximately 50%. For potassium, adjustment for BMI increased the size of uncorrected regression coefficients but left corrected coefficients little changed from the analyses that did not include BMI. An important issue in regard to these and other analyses of the sodium-blood pressure association is whether it is really appropriate to adjust for BMI. Although BMI is a variable measured essentially without error, this is not the case for 24-hour sodium

7 Dyer et al BMI, Sodium, Potassium, and Blood Pressure 735 excretion; indeed, for sodium excretion in INTER- SALT, the intraindividual variation was greater than the interindividual variation, as indicated by a coefficient of reliability that was less than Furthermore, given the positive correlation of BMI with sodium excretion, a portion of the association between BMI and blood pressure may be due to the higher sodium intake in overweight individuals. Hence, the inclusion of BMI as a potential confounder in previous INTERSALT analyses may have resulted in an overadjustment for that variable. Possible interactions of these electrolytes with BMI in relation to blood pressure were addressed by examining electrolyte-blood pressure associations in centers classified as lower- or higher-bmi based on the median BMI at the center being less than or greater than or equal to 24.5 kg/m 2 and by examining these same associations in individuals classified as lower- or higher- BMI based on individual BMI of less than or greater than or equal to 24.1 kg/m 2. In these analyses sodium and the sodium-potassium ratio were positively and significantly related to systolic pressure, and potassium was inversely and significantly related in all four BMI subgroups. The sodium-potassium ratio was also positively and significantly related to diastolic pressure in two subgroups and potassium inversely and significantly related in three. Pooled regression coefficients did not differ significantly between higher- and lower-bmi centers or individuals for any of these electrolyte variables. It has been suggested that the low blood pressures found in populations with low sodium intake might reflect lower body weight rather than low sodium intake, 29 thus implying that an association between sodium and blood pressure should be absent or reduced among lower-body-weight populations and individuals. As indicated above, no evidence to support this assertion was found in the present study. Thus, the findings of this report are concordant with other epidemiologic studies showing significant associations between sodium and blood pressure in lean Far Eastern populations 30 ' 31 and in Western populations with higher BMI The combined effects of weight and sodium reduction on blood pressure have not been quantified, 14 and it is unclear whether lower sodium intake reduces blood pressure equally in overweight and non-overweight individuals. 16 ' 17 In addition, it has been suggested that weight loss without a concomitant reduction in sodium intake does not lead to reductions in blood pressure. 5 Although the present analyses cannot address these issues directly, results from previous analyses showing that BMI was significantly related to both systolic and diastolic blood pressures, independent of sodium and potassium excretion, 1 and the present results showing that sodium and potassium excretions were significantly related to systolic pressure throughout the range of BMI indicate that the effects of sodium intake and overweight on blood pressure are additive. Thus, for most populations consuming salt in excess of 5.8 g/d (ie, >100 mmol/d sodium), the inference from the analyses presented here is that reduction in both salt intake and weight will have additive effects in lowering blood pressure. However, these cross-sectional observational data are limited in regard to resolving the issue posed by the apparently contradictory results of intervention trials; ie, with maintenance of high salt intake, is weight reduction in obese individuals ineffective in lowering blood pressure? Further well-designed and well-executed trials are needed to resolve this matter. The related question also remains as to whether a high salt intake is a primary, essential, necessary cause for a population rise of blood pressure with age from youth through middle and older age and for high prevalence rates of hypertension, and whether other factors (eg, obesity, high alcohol consumption, low potassium intake) are secondary, adjuvant, contributory causes operating only if salt intake is high. No "experiments of nature" in populations are available yielding data to resolve this issue definitively. Data on the four low sodium centers in INTERSALT were not consistent as to whether BMI is directly related to blood pressure in populations with habitual low salt intake. Thus, for men from all four of these samples the Yanomamo and Xingu Indians in Brazil, Papua, New Guinea, and Kenya with median 24-hour sodium excretion ranging from 0.2 to 51.3 mmol (meq) per 24 hours, 1 BMI was positively related to blood pressure, significantly so in the Yanomamo for both systolic and diastolic blood pressure and in Kenya for diastolic pressure. 2 However, in women there were no significant associations between BMI and blood pressure, and three of the eight within-center regression coefficients relating BMI to blood pressure were negative. Moreover, for men from these isolated low sodium samples, higher BMI could reflect muscularity rather than adiposity, so that the meaning of the direct BMI-blood pressure relation may be different than in most populations, for which greater BMI is predominantly due to greater adiposity. Hence, the INTERSALT data are not definitive in regard to this question. In any case, the overall results of INTERSALT indicate that for most populations, with high average salt intake, effects on blood pressure of sodium reduction and weight loss should be independent and additive. In conclusion, these INTERSALT results do not support the concept of important interactions of sodium and potassium with BMI in relation to blood pressure. Furthermore, they indicate that associations of sodium (positive) and potassium (inverse) with blood pressure occur across the BMI range. These findings have implications for the primary prevention of high blood pressure in populations; ie, they lend further support to recommendations for a combined approach, including lower sodium intake, higher potassium intake, reduction in the prevalence of heavy drinking, and prevention and control of obesity. 34 Acknowledgments This study was launched under the auspices of the Council on Epidemiology and Prevention of the International Society and Federation of Cardiology (ISFC). A full list of participating centers and investigators is given in Reference 1. The work was supported by grants from the Wellcome Trust (United Kingdom); the National Heart, Lung, and Blood Institute (2R01 HL-33387) (United States); the International Society of Hypertension; the World Health Organization; the Heart Foundations of Canada, Great Britain, Japan, and the Netherlands; the Chicago Health Research Foundation; the FWGO-FMRS (Belgian National Research Foundation); and the ALSK-CGER (Parastatal Insurance Co, Brussels).

8 736 Hypertension Vol 23, No 6, Part 1 June 1994 References 1. INTERSALT Cooperative Research Group. INTERSALT: an international study of electrolyte excretion and blood pressure. Results for 24-hour urinary sodium and potassium excretion. Br MedJ. 1988;297: Dyer A, Elliott P. Body mass index and blood pressure in the INTERSALT study of urinary electrolytes, other factors and blood pressure. J Hum Hypertens. 1989^3: Dyer AR, Elliott P, Shipley MJ. Body mass index versus height and weight in relation to blood pressure: findings for the 10,079 persons in the INTERSALT Study. Am J Epidemiol 1990;131: Dahl LK, Silver L, Christie RW. The role of salt in the fall of blood pressure accompanying reduction in obesity. N Engl J Med. 1958; 258: Heyden S, Tyroler HA, Hames CG, Bartel A, Thompson JW, Knshan I, Rosenthal T. Diet treatment of obese hypertensives. Clin Sci Mol Med. 1973;45: Reisin E, Abel R, Modan M, Silverberg DS, Eliahou H, Modan B. Effect of weight loss without salt restriction on the reduction of blood pressure in overweight hypertensive patients. N Engl J Med. 1978;298:l Fagerberg B, Andersson OK, Isaksson B, Bjorntorp P. Blood pressure control during weight reduction in obese hypertensive men: separate effects of sodium and energy restriction. Br Med J. 1984;288: Andersson OK, Fagerberg B, Hedner T. Importance of dietary salt in the hemodynamic adjustment to weight reduction in obese hypertensive men. Hypertension. 1984;6: Tuck ML, Sower L, Kiedzick G, Maxwell M. The effect of weight reduction on blood pressure, plasma renin activity and plasma aldosterone levels in obese patients. N Engl J Med ;304: Gillum RF, Prineas RJ, Jeffrey RW, Jacobs DR, Elmer PJ, Gomez O, Blackburn H. Nonpharmacologic therapy of hypertension: the independent effects of weight reduction and sodium restriction in overweight borderline hypertensive patients. Am Heart J. 1983;105: Haynes BR, Harper AC, Costley SR, Johnston M, Logan AG, Flanagan PT, Sackett DL. Failure of weight reduction to reduce mildly elevated blood pressure: a randomized trial. / Hypertens. 1984;2: The Trials of Hypertension Prevention Collaborative Research Group. The effects of nonpharmacologic interventions on blood pressure of persons with high normal levels: results of the Trials of Hypertension Prevention, Phase I. JAMA. 1992;267: Wassertheil-Smoller S, Blaufox MD, Oberman AS, Langford HC, Davis BR, Wylie-Rosset J. The Trial of Antihypertensive Interventions and Management (TAIM) Study: adequate weight loss, alone and combined with drug therapy in the treatment of mild hypertension. Arch Intern Med. 1992;152: Prineas RJ. Clinical interaction of salt and weight change on blood pressure level. Hypertension. 1991;17(suppl I):I-143-I Cutler JA, Follmann D, Elliott P, Suh I. An overview of randomized trials of sodium restriction. Hypertension. 1991;17(suppl I):I-27-I Cooper R, Van Horn L, Liu K, Trevisan M, Nanas S, Ueshima H, Larbi E, Yu CS, Sempos C, LeGrady D, Stamler J. A randomized trial on the effect of decreased dietary sodium intake on blood pressure in adolescents. J Hypertens. 1984;2: Rocchini AP, Key J, Bondie D, Chico R, Moorehead C, Katch V, Martin M. The effect of weight loss on the sensitivity of blood pressure to sodium in obese adolescents. N Engl J Med. 1989^21: INTERSALT Co-operative Research Group. INTERSALT Study: an international co-operative study on the relation of blood pressure to electrolyte excretion in populations, I: design and methods. / Hypertens. 1986;4: Elliott P, Stamler R. Manual of Operations for 'INTERSALT,' an international cooperative study on the relation of sodium and potassium to blood pressure. Controlled Clm Trials. 1988;9(suppl): ls-118s. 20. American Association of Clinical Chemists. Standard Methods of Clinical Chemistry, Volume 3. New York, NY: Academic Press; National Research Council, Committee on Diet and Health, Food and Nutrition Board, Committee on Life Sciences. Diet and Health: Implications for Reducing Chronic Disease. Washington, DC: National Academy Press; 1989: National Research Council, Subcommittee on the Tenth Edition of the RDAs, Food and Nutrition Board, Commission on Life Sciences. Recommended Dietary Allowances. 10th ed. Washington, DC: National Academy Press; 1989: Dyer AR, Shipley M, Elliott P, for the INTERSALT Cooperative Research Group. 24-hour urinary electrolyte excretion and blood pressure in the INTERSALT Study, I: estimates of reliability. Am J Epidemiol 1994;139: Dyer AR, Elliott P, Shipley M, for the INTERSALT Cooperative Research Group. 24-hour urinary electrolyte excretion and blood pressure in the INTERSALT Study, II: estimates of electrolyteblood pressure associations corrected for regression dilution bias. Ami Epidemiol 1994;139: Watson RL, Langford HG, Abernethy J, Barnes TY, Watson MJ. Urinary electrolytes, body weight, and blood pressure: pooled cross-sectional results among four groups of adolescent females. Hypertension. 1980;2(suppl I):I-93-I Connor SL, Connor WE, Henry H, Sexton G, Keenan EJ. The effects of familial relationships, age, body weight, and diet on blood pressure and the 24 hour urinary excretion of sodium, potassium, and creatinine in men, women, and children of randomly selected families. Circulation. 1984;70: M'Buyamba-Kabangu JR, Fagard R, Lijnen P, Mbuy wa Mbuy R, Staessen J, Amery A. Blood pressure and urinary cations in urban Bantu of Zaire. Am J Epidemiol 1986;124: Pietinen PI, Wong O, Altschul AM. Electrolyte output, blood pressure, and family history of hypertension. Am J Clin Nutr. 1979;32: Dustan HP. Obesity and hypertension. Ann Intern Med. 1985;103: Kesteloot H, Huang DX, Li Y, Geboers J, Joossens J. The relationship between cations and blood pressure in the People's Republic of China. Hypertension. 1987;9: Kesteloot H, Park BC, Lee CS, Brems-Heyns E, Claessens J, Joossens JV. A comparative study of blood pressure and sodium intake in Belgium and in Korea. EurJ Cardiol. 1980;ll: Strazzullo P, Trevisan M, Farinaro E, Cappuccio FP, Ferrara LA, De Campora E, Mancini M. Characteristics of the association between salt intake and blood pressure in a sample of a male working population in Southern Italy. Eur Heart J. 1983;4: Elliott P, Forrest RD, Jackson CA, Yudkin JS. Sodium and blood pressure: positive associations in a north London population with consideration of the methodological problems of within-population surveys. / Hum Hypertens. 1988;2: Stamler R. Implications of the INTERSALT Study. Hypertension. 1991;17(suppl I):I-16-I-2O.