Relationship between salt excretion and blood pressure in various regions of China*

Bulletin ofthe WorldHealth Organization,62(2):255-260(1984) ,In - / © World Health Organization 1984 Relationship between salt excretion and bloocd pressure in various regions of China* L. S. Liu,' S. C. TAO,1 & S. H. LAI2 Overnight urine samples were collected on three consecutive daysfrom 3105 persons in 12 regions of China, and analysed for levels of sodium and potassium. The mean 9-hour overnight urine sodium level rangedfrom 49.51 to 139.12 mmol, and urinepotassiumfrom 6.32 to 18.43 mmol. Univariate regression, simple correlation analysis, and multivariate ridge regression analysis were carried out on the data. A positive correlation between blood pressure and urine sodium or sodium/potassium ratio was found in each of the twelve regions. Urine potassium showed a negative correlation with blood pressure in three regions. During 1979-80, a nationwide survey of blood pressure was carried out in China. It was found that the northern and northeastern provinces had a higher prevalence of hypertension than other areas of the country. Geological, socioeconomic and racial fac- tors may partly account for this difference; however, differences in food consumption and cooking habits, especially salt intake, may also contribute to the observed variation. A collaborative study was carried out in 1981 to investigate the relationship between hypertension and salt intake in different parts of the country, and fur- ther, to determine whether sodium itself has an effect on blood pressure. Twelve regions were selected for study, comprising 4 urban and 6 rural areas, 1 fishing area, and 1 salt- producing area (Table 1). Guangdong and Lhasa were included because they had the lowest and highest rates of hypertension, respectively, in the 1979 survey. Other regions were selected mostly because they had an existing community control programme, so that urine collection could be easily organized. Patients already undergoing treatment for hypertension were excluded from the study. METHODS Random population samples of men and women aged between 25 and 65 years were taken from the populations that had been screened in the 1979-80 blood pressure survey. A total of 3105 persons partici- pated in the study (Table 2). * From the Cardiovascular Institute and Fu Wai Hospital, Chinese Academy of Medical Sciences, Beijing, China. ' Department of Cardiology. 2 Department of Epidemiology. Table 1. Age-adjusted prevalence of hypertension in 12 regions of China in 1979-80 Age-adjusted Region No. No. with prevalence screened hypertension (%) Lhasa, Tibet 9 672 1 850 17.76 (urban) Jiuxianchiao Beijing 20 412 2 332 10.56 (urban) Shijingshan 12 805 1 162 8.57 Beijing (rural) Shanxi 14 381 953 6.38 (urban) Lanzhou 8 461 187 4.91 (urban) Shaanxi 21 342 702 2.75 (rural) Jiangsu 27 640 947 3.01 (rural) Shanghai 15 847 509 2.21 (rural) Zhejiang 9 681 316 2.92 (salt-producing) Zhejiang 9 720 344 3.45 (fishing area) Guangdong 12 686 283 1.93 (rural) Guangxi 49 645 1 688 3.4 (rural) Overnight urine samples were collected from each participant on three consecutive days. The urine was analysed for sodium and potassium by flame phot- ometry at a local laboratory. The samples from Lhasa and Jiangsu were sent by post to the laboratory of the 4399 -255- L. S. LIU ET AL. Table 2. Characteristics of the participants in the urine sodium survey, China 1981 No. of No. with Region partici- hyper- Sex Mean Mean Mean Weight/ pants tension age a height' weighta height3 a M F Lhasa 200 56 74 126 46.29 1.58 56.78 14.38 Tibet (5.66) (0.08) (8.31) (1.89) Jiuxianchiao 334 27 181 153 43.19 1.63 62.84 14.46 Beijing (5.03) (0.08) (9.62) (2.09) Shijingshan 216 18 100 116 43.41 1.61 59.74 14.50 Beijing (5.53) (0.08) (8.96) (2.27) Shanxi 201 0 105 96 25.39 1.64 55.84 12.57 (10.28) (0.08) (8.45) (1.59) Lanzhou 209 17 209 44.15 1.69 63.52 13.56 (4.84) (0.06) (8.59) (5.65) Hanzhung 107 47 56 51 55.34 1.56 51.01 16.27 Shaanxi (8.95) (0.08) (6.85) (1.91) Zhenjiang 242 18 95 147 65.20 1.56 50.69 13.49 Jiangsu (11.85) (0.08) (6.88) (1.88) Shanghai 232 7 53 179 44.17 1.57 50.98 12.62 (17.65) (0.08) (7.49) (8.08) Diashan 226 7 34 192 43.92 1.58 50.62 12.73 Zhejiang (14.65) (0.07) (8.95) (1.96) Zhoushan, 200 1 200 33.92 1.66 55.92 12.21 Zhejiang (14.72) (0.06) (6.32) (6.27) Guangdong 215 5 112 103 44.43 1.59 50.78 12.70 (5.82) (0.08) (6.56) (1.37) Guangxi 723 9 667 56 43.55 1.65 58.07 13.04 (5.58) (0.07) (8.27) (2.06) a Figures in parentheses give standard deviation. Shanghai Hypertension Institute for examination. There was no standardization of methods in the various laboratories. All the data were included in the analysis; completeness was ascertained primarily by careful questioning of the participants about the possibility of missed or extra specimens. The height and weight of each participant were measured with the subject in light clothes without shoes. Blood pressure was measured in the right radial artery after 15 minutes' rest in a quiet room, with the subject in a sitting position and his or her arm resting on a table at heart level. Pressures were recorded at the appearance of systolic and fifth-phase diastolic Korotkoff sounds. The measurements were made in the morning of each of the three days when urine was collected. The average of the two lowest readings was taken as the blood pressure for that person. Of 3105 persons, 212 (6.8%o) had definite hypertension (sys- tolic pressure > 160 mmHg (21.3 kPa) or diastolic pressure > 95 mmHg (12.7 kPa)). Statistical analysis Systolic pressure (SP) and diastolic pressure (DP) are the dependent variables; Na/K, Na, K, 1/(1 + e- Na/OK), 1/(1 +e- Na/00), and 1/(1 + eU- l-O) were taken as the independent variables (Na is urine sodium level, and K is urine potassium). The stat- istical techniques used included unifactorial re- gression and simple correlation; multiple stepwise regression was also tried using sex, age, height, weight, weight/height3, Na, K, log Na, log K, and Na/K as the independent variables. Unfortunately, the intercorrelations among the independent vari- ables led to unstable coefficient estimates and aliasing, so the ridge regression method was used for analysis (see Annex 1). RESULTS The means, standard deviations, and coefficients of variance for blood pressure, urine sodium, and urine potassium in the 12 regions are presented in Table 3. The mean 9-hour overnight sodium level ranged from 49.51 to 139.12 mmol, and the potas- sium level from 6.32 to 18.43 mmol, which corres- ponds to 154.03-432.82 mmol of sodium and 19.68- 57.34 mmol of potassium per 24 hours, or 9.01- 25.32 g of NaCI per day. 256 SALT EXCRETION AND BLOOD PRESSURE IN CHINA Table 3. Means, standard deviations and coefficients of variance of pressure in twelve regions of China urine sodium, urine potassium, and blood Urine sodium Urine potassium Systolic pressure Diastolic pressureRegion (mmol) Mmo) a (mmHg) a,b (mmHg) a.b Lhasa, 68.15 ± 36.55 7.44± 4.02 132.46 ± 24.59 88.94± 12.88 Tibet (53.63%) (54.01%) (18.57%) (14.49%) Jiuxianchiao 110.48 ± 62.23 13.82 ± 5.42 117.64± 16.74 77.94± 11.58 Beijing (56.33%) (39.22%) (14.23%) (14.85%) Shijingshan 87.36 ± 36.29 10.00 ± 4.67 120.85 ± 15.40 78.78 ± 10.25 Beijing (41.54%) 146.70%) (12.74%) (12.95%) Shanxi 75.07 ± 28.91 9.19 ±4.12 112.15 ± 9.35 71.69 ± 7.23 (38.52%) (44.82%) (8.34%) (10.09%) Lanzhou 69.78 ± 27.03 13.56 ± 6.33 113.17 ± 14.78 77.53± 11.89 (38.74%) (46.67%) (13.06%) (15.32%) Hanzhung 139.12 ± 33.55 11.62 ± 5.95 148.19 ± 32.97 89.80± 16.69 Shaanxi (24.12%) (51.18%) (22.25%) (18.59%) Zhenjiang 61.94 ± 29.68 6.32 ± 3.31 128.14± 19.92 77.90± 8.47 Jiangsu (47.91%) (52.30%) (15.55%) (10.88%) Shanghai 75.20± 33.09 7.62 ± 3.69 113.59 ± 18.33 69.38 ± 9.31 (44.00%) (48.37%) (16.14%) (13.42%) Daishan 101.64±47.28 7.40± 3.63 114.82 ± 22.64 70.06 ± 10.92 Zhejiang (46.52%) (49.07%) (19.72%) (15.59%) Zhoushan 89.38 ± 37.40 8.74± 3.8 109.92 ± 9.43 66.38 ± 6.71 Zhejiang (41.84%) (43.02%) (8.58%) (10.11 %) Guangdong 112.95 ± 24.49 18.43 ± 8.98 112.13 ± 13.55 71.61 ± 8.02 (21.68%) (48.74%) (12.08%) (11.20%) Guangxi 62.88±24.32 8.20±4.14 112.62±34.19 71.28±8.70 (38.68%) (50.49%) (30.36%) (12.21%) a Mean ± standard deviation. Figure in parentheses gives coefficient of variation. b Blood pressure is given in mmHg: 100 mmHg = 13.3 kPa. The results of the simple regression and ridge re- gression analyses for the 12 populations are shown in Tables 4 and 5. It was found that age, body weight, and weight/height3 were strongly correlated with blood pressure; in this age range, height had no add- itional effect. Blood pressure was positively related to the sodium/potassium ratio in 2 regions, to sodium level in 6, and to I/(I +eeNa/l°) in 6 regions, as shown by the unifactorial regression and simple correlation analyses. In the multifactorial ridge regression analysis, a positive correlation was found between sodium level and blood pressure in 10 of the 12 populations, while in the remaining 2 populations positive correlations were also found between blood pressure and Na3 and Na2/K. The standardized coefficients showed that the effect of sodium level on blood pressure was lower than that of age, and similar to that of body weight. The ratio of sodium to potassium was not indepen- dently related to blood pressure, and potassium level was found to have a negative correlation with blood pressure only in Tibet, Guangdong, and Lanzhou.- DISCUSSION Several cross-cultural and regional comparative studies have supported the hypothesis that a high sodium intake is a risk factor for essential hyper- tension. The present study, however, did not demon- strate any correlation between prevalence of hyper- tension and sodium excretion levels. However, since the salt intake within the survey population was not homogeneous, the small group (100-200 persons) who gave urine samples is probably not representative of the population (10 000-20 000 in most regions). In addition, the two surveys were not done at the same time. There may also be genetic factors that contri- bute to the differences in the prevalence of hyper- tension (the population in 11 regions was mainly Han nationality, while those in Lhasa were Tibetan). Urine collections from the entire population in the blood pressure survey may help further to elucidate these factors. Within populations, attempts to correlate indi- vidual levels of blood pressure and salt intake have 257 L. S. LIU ET AL. Table 4. Significant simple correlation coefficients (P < 0.1) between blood pressure and urine sodium and potassium levels Region Na/K Na K 1/(1 +e- Na/10K) 1/(1 +e Na/100) 1/(1 +e K/10) Lhasa spa 0.11 -0.1 0.11 0.10 -0.11 Tibet DP -0.1 -0.11 Jiuxianchiao SP 0.08 0.08 0.07 0.15 b o.09 Beijing Shijingshan SP 0.27h 0.14 0.26b 0.13' Beijing DP 0.280 0.200 0.26 0.18b Shanxi SP 0.17c 0.17' 0.16' 0.17c DP 0.21 b 0.25b 0.20b 0.26b Lanzhou DP 0.11 0.11 Hanzhung SP 0.24 b 0.24c Shaanxi DP 0.16 0.25b 0.26b Zhenjiang SP 0.11 0.14' 0.12 0.14' Jiangsu DP 0.11 0.12 0.11 0.13' Shanghai DP 0.10 0.09 Daishan SP 0.21b 0.1 3 0.1 9 b 0.12 Zhejiang DP 0.18b 0.14' 0.17b 0.13 Zhoushan SP 0.11 0.11 0.10 0.11 Zhejiang DP 0.10 Guangdong SP 0.16 c 0.1 5c DP 0.13' 0.10 Guangxi DP 0.06 0.07 0.05 a SP = systolic pressure; DP= diastolic pressure. P< 0.01. P <0.05. given conflicting results. Some authors believe that the range of salt intake in a given population is not likely to be large enough to demonstrate a clear cor- relation with blood pressure. In some societies, the salt intake greatly exceeds the physiological require- ment, and the resulting saturation may produce a negative result. However, our data have shown cer- tain positive correlations between blood pressure and urine sodium in each of the twelve regions. Such a finding is probably related to the relatively large vari- ation in the amount of sodium excreted within each population, as shown in Table 3 by the coefficients of variance. Since potassium is said to have some protective effect against the adverse effects of high salt intake on blood pressure, very high sodium and very low potassium intake might enhance the positive correlation between sodium excretion and blood pressure and make it more easily detected. Our results support the concept that chronic ex;cess sodium leads to an increased risk of hypertension; reduction of salt intake as a measure in the primary prevention of hypertension is therefore worthy of further study. ACKNOWLEDGEMENTS The collaborative group taking part in the study included the Shanghai Institute of Hypertension, the Guangdong Provincial Institute of Cardiovascular Diseases, the Beijing Jiuxianchiao Hospital, the Second People's Hospital, Lanzhou Medical College, the Tibet Academy of Medical Sciences, the Sanxi Provincial People's Hospital, the Hanzhung Municipal People's Hospital, the Zhengjiang'Region Peoplets Hospital, the Guangxi Medical College, the Cardiovascular Disease Institute, Chinese Academy of Medical Sciences, and the Capital Iron and Steel Complex Hospital. 258 SALT EXCRETION AND BLOOD PRESSURE IN CHINA 0 ) o6 a (N 6 (I 6 6 0 6 0 o LO - 6 667 Ot - 14 t N CY CY n-- 0 6o 666 C N- 66 666 6 600 00 0o0o0 0 C1 0 - 0(N '-0( N66 66 66 N - - - 66N 66-CC5 .6 6 c 6 6 .0 0I 6-67 7 oo0 C' (C')m I 0 00C') 6 66 6 6 0 CD 0 - - - - - - - - 0 - - -;z & ;:; & N _- -- 0 - C1 I 66 66 oo 66 66 66 II I1 0.0.dC/ _3 eL CL(n aj a.a. 00L L L Q. a. 0 0na (n 0 (n 0 (n (n 0 0 . U) .C -J (1) -c (1 0 N c -j ~C 0) CD.C 0a,.xC n c vrs-or v I N 0 (1) N () (. 259 6 _Q 6 (4 6 0 6'7 6 6 0 0 6 6 E .2 (n 0 Q E .2 0. U) .0 c (U E V ._ 0 C,, a C VD 0 C._ Q U) a co co C) C.) 0._ 0 0._ a) C,, Cl) 0._ . C -a) a, -c4. 0) -j z U, 0 -J z a, . - .2)U, a, I x a, U, n a)(1) 0 a) CV - - 0 - 0 D0 o or- * 04 -* C" - 04 4 04 - - c, -a o-w coo or'- 00 . . 11 11 tl 11 11 11 CL a 0 OOa ()a 0.0 COO 11 11 Q- C/ aM 0e- 6 it 11 C/a 00 o. k 0. (n a a, 0. a.) -cnZ -'000 Cl .0 Z c c 0.2 .2 .2 0- O 0 o cn uuX 11 -a Q L. S. LIU ET AL. RESUME RELATION ENTRE LA NATRIURIE ET LA TENSION ARTtRIELLE DANS DIFFtRENTES RtGIONS DE CHINE Les differences traditionnelles qui existent entre le Nord et le Sud de la Chine en matiere de consommation alimen- taire et d'habitudes culinaires, ration sod6e en particulier, sont depuis longtemps considerees comme expliquant large- ment les ecarts significatifs de prevalence de l'hypertension dans ces deux grandes regions. Un projet collectif a e ex6cut6 en 1981 afin d'6tudier le lien entre prevalence de l'hypertension et ration sod6e. Des &chantillons d'urine ont e pr6lev6s le matin A la premiere miction pendant trois jours cons&cutifs chez 3105 habitants de 12 des regions du pays. Sur 24 heures, les taux urinaires se sont echelonnes de 154,03 i 432,82 mmol pour le sodium et de 19,68 i 57,34 mmol pour le potassium. Les resultats ont ete soumis A une analyse statistique de regression simple, de correlation simple et de regression multiple selon une technique particu- liere (ridge regression). Une corr6lation positive a 't6 mise en evidence entre la tension arterielle d'une part et le sodium ou le rapport sodium/potassium urinaires d'autre part dans chacune des douze regions; dans deux regions, on a en revanche observe une correlation negative entre le taux de potassium urinaire et la tension art6rielle. Ces r6sultats sont probablement lies aux variations relativement importantes de la natriurie au sein de chacune des populations. Les conclusions obtenues corroborent la these selon laquelle une exces chronique de sodium se solde par un risque accru d'hypertension. La reduction de l'apport sod6 en tant que mesure de prevention primaire de l'hypertension m6rite par consequent d'etre 6tudi&e plus A fond. Annex I RIDGE REGRESSION ANALYSIS Ridge regression analysis provides a better es- timation method than simple regression when the independent variables are highly correlated. Ridge regression analysis controls the variance resulting from high colinearity by increasing the elements of the diagonal of the normal equation matrix in small steps, producing slightly biased, but more stable, co- efficient estimates. The ridge regression model is: Y = X: + e where: Y is an n x 1 vector of observations on a dependent variable; X is an n xp matrix of observations on p indepen- dent variables; ,B is the p x 1 vector of regression coefficients; and e is an n x 1 vector of residuals, which is a vector of random disturbance and satisfies E(e) = 0, E(,E')= or2L It is assumed thatX and Y have been scaled so that X'X and X' Y are matrices of correlation co- efficients. The least square estimator for ,3 is (X'X) 1X' Y, while the ridge regression estimator is f3(k) = (X'X + kI) 'X' Y (for a series of given values of k). The idea of ridge regression analysis is to determine the smallest value of k for which ,8(k) is stable. 260