Since the turn of the twentieth century, reducing Na intake has been a key target for the prevention of hypertension, a major risk factor for stroke( Reference Esunge 1 , Reference Mayer, Kline and Cooke 2 ). Although Na occurs naturally in food, the excess Na ingested in Western cultures comes largely from processed food. Reduction of dietary Na is once again gaining momentum with new recommendations from the United States Department of Agriculture( 3 ) and state organisations introducing Na reduction campaigns( 4 , Reference Appel, Frohlich and Hall 5 ). Despite abundant research suggesting that lower dietary Na would decrease hypertension and stroke( Reference He, Ogden and Vupputuri 6 ), Na intake in the USA remains well above recommended levels, with only 5·5 % of Americans meeting United States Department of Agriculture targets( Reference Peralez-Gunn, Kuklina and Keena 7 ). Also, dietary Na may contribute to the higher rates of stroke and stroke mortality observed in black Americans compared with white Americans( Reference Strazzullo, D'Elia and Kandala 8 ).
In contrast with Na, increased dietary K may reduce the burden of hypertension and stroke. Increased K has been shown to lower blood pressure and this beneficial effect may be greater in the setting of a high-Na diet( Reference Ando, Fujita and Fujita 9 – Reference Matsui, Shimosawa and Uetake 12 ). In fact, US veterans switching from regular salt to a K-enriched salt experienced reduced CVD mortality( Reference Chang, Hu and Yue 13 ). Dietary K may also contribute to racial differences in stroke and stroke mortality because black Americans have a lower K intake than white Americans( Reference Newby, Noel and Grant 14 ).
Taken together, the dietary Na:K ratio may explain demographic differences in stroke and stroke mortality in the USA. Using the ratio provides a means of balancing the detrimental effects of Na and beneficial effects of K in one measure. Recent analyses using mortality follow-up from the National Health and Nutrition Examination Survey demonstrated that the dietary Na:K ratio was more strongly associated with CVD mortality than Na alone, but lacked power to examine this association by race( Reference Yang, Liu and Kuklina 15 ). Very little is known about the question whether the dietary Na:K ratio is associated with all-cause mortality and stroke incidence and whether these relationships differ across racial groups.
The primary hypothesis of this study was that an increased dietary Na:K ratio is associated with greater risk of death and incident stroke and that this association will be stronger in blacks than in whites. To test this hypothesis, we utilised the data from participants of the REasons for Geographic And Racial Differences in Stroke (REGARDS) study, a large population-based prospective cohort study of black and white US adults aged 45 years or older( Reference Howard, Cushman and Pulley 16 ).
The design and objectives of the REGARDS study have been described previously( Reference Howard, Cushman and Pulley 16 ). By design, the study oversampled blacks and included an approximately equal representation of men and women. Enrolment began in January 2003 and ended in October 2007. Approximately half of the participants were recruited from the ‘Stroke Belt’ (states of North Carolina, South Carolina, Georgia, Tennessee, Alabama, Mississippi, Arkansas and Louisiana). The REGARDS study participants were identified using a commercial, nation-wide list of over 250 million individuals in the USA (Genesys Incorporated). A trained interviewer contacted households by telephone, and one resident per household aged 45 years or older was randomly screened for eligibility. The response rate (the percentage agreeing to be interviewed among known eligible candidates contacted after adjustment for those of unknown eligibility) was 41 %, which is similar to other research studies that initially contacted participants by telephone( Reference Jackson, Chambless and Yang 17 , Reference Morton, Cahill and Hartge 18 ). Exclusion criteria were: race other than black or white; active treatment for cancer; medical conditions preventing long-term participation; cognitive impairment as subjectively judged by the telephone interviewer; staying in, or on a waiting list for, a nursing home; and inability to communicate in English. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the Institutional Review Boards at the University of Alabama Birmingham. Written informed consent was obtained from all subjects/patients.
Data collection and covariates
Data were collected during a telephone interview, an in-home visit and via self-administered questionnaires that were left for participants to complete following the in-home visit. Of relevance to the present analysis, the telephone interview included information on age, race, sex, education, household income, current smoking status and a prior diagnosis of hypertension, diabetes mellitus and myocardial infarction. During the in-home study visit, weight, height and waist circumference were measured by following a standardised protocol. BMI was calculated as weight in kilograms divided by height in metres squared. Using the mean of two blood pressure measurements, hypertension was defined as a systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg or self-reported use of anti-hypertensive medication. Participants were asked to fast for 10–12 h prior to the in-home study visit. A blood sample was collected, centrifuged, refrigerated and shipped overnight to the REGARDS central laboratory at the University of Vermont. Total cholesterol, HDL-cholesterol, TAG, serum glucose and serum creatinine were measured by colorimetric reflectance spectrophotometry using the Ortho Vitros Clinical Chemistry System 950IRC instrument (Johnson & Johnson Clinical Diagnostics) and LDL-cholesterol was calculated by the Friedewald equation for individuals with TAG <10 mmol/l. Dyslipidaemia was defined as a total cholesterol ≥6·2 mmol/l, LDL-cholesterol ≥4·1 mmol/l, HDL-cholesterol <1·0 mmol/l for men or <1·3 mmol/l for women, or lipid-lowering medication use. Diabetes mellitus was defined as fasting glucose ≥7·0 mmol/l, ≥11·0 mmol/l for participants who did not fast, or self-reported current use of anti-diabetes medication.
At the end of the in-home study visit, participants were asked to complete additional questionnaires which included the Block 98 FFQ (www.Nutritionquest.com). The Block FFQ has been validated for most nutrients, including Na and K, using multiple diet records( Reference Patterson, Kristal and Tinker 19 ). Different versions of this questionnaire have been studied extensively and validated in diverse populations( Reference Block, Woods and Potosky 20 ). The Block 98 version developed by Block Dietary Data Systems and distributed by NutritionQuest is an eight-page paper-and-pencil form with more than 150 multiple-choice questions based on 107 food items that can be completed in about 30–40 min. In the REGARDS study, the FFQ was self-administered and participants were asked to report on diet over the past year. Pictures were provided to help identify portion sizes. FFQ, compared with diary techniques, tend to underestimate micronutrient intake and overestimate total energy intake; however, this measurement error is usually uniformly distributed across study subpopulations( Reference Mares-Perlman, Klein and Klein 21 , Reference Caan, Slattery and Potter 22 ). Therefore, it is generally recommended that estimated nutrient intakes from FFQ should be reported and analysed as either ranked values (i.e. quintiles, etc.) or ratios of two different nutrients rather than absolute values per individual( Reference Caan, Slattery and Potter 22 ). Thus, race- and sex-specific quintiles of dietary Na (mg) and the Na:K ratio were used in all analyses.
Outcome assessment for incident stroke and death
Every 6 months, the REGARDS study ascertains all hospitalisations by participant or proxy interview using methods described elsewhere( Reference Howard, Kleindorfer and Judd 23 ). Personnel specifically pursue medical records for all hospitalisations related to stroke, transient ischaemic attack, stroke symptoms and deaths. Medical records for strokes that occurred before baseline are not pursued. For all participants who have died, study personnel interview the next of kin or other proxy to probe for information relating to the death, inquiring if a stroke not previously reported antedated the death; any hospitalisations prior to death are collected. Once the records are received, they are first reviewed by a trained neurological nurse to verify that the record is complete, and then the information is forwarded to a team of stroke neurologists for review. In the case of deaths, the interview with next of kin or proxy and death certificate or National Death Index data are included. A stroke was defined as focal neurological symptoms lasting greater than 24 h or imaging positive for a new stroke. For this analysis, adjudicated stroke events were available to 1 February 2011 and we included only incident strokes. The REGARDS study is notified of a participant's death by proxies either through the mail or on the telephone during the routine 6-month telephone calls and also uses the Social Security Death Index to look for participants who may have died and for whom we do not have a proxy report of death. For this analysis, death events were available to 1 April 2011. We used all-cause mortality as an outcome in addition to adjudicated stroke.
As the association of dietary Na with CVD has been reported to differ by race, we decided a priori to stratify analyses for black and white participants. The distribution of participant characteristics was determined for blacks and whites separately. The prevalence and multivariable-adjusted hazard ratios (HR) for incident stroke associated with race- and sex-specific quintiles of the Na:K ratio were calculated using Cox proportional hazards models. Lastly, the prevalence and multivariable-adjusted HR for all-cause mortality were calculated by sex-specific quintile of the Na:K ratio. Multivariable-adjusted HR were calculated in three steps. In the first step we adjusted for age, race, sex and total energy intake. In the second step we additionally adjusted for BMI, hypertension, diabetes, dyslipidaemia, history of heart disease, smoking, income and education. In the third step, which was only for analyses with stroke as an outcome, we further adjusted for bias in medical record retrieval and extraction( Reference Howard, McClure and Moy 24 ). For the mortality analysis, we performed a sensitivity analysis excluding those who died within 6 months of the baseline visit (as a proxy for terminal illness) and all participants who reported a history of stroke at baseline. As a secondary analysis, the multivariable-adjusted HR for incident stroke and all-cause mortality associated with quintiles of Na were calculated. Additionally, we considered the Na:K ratio as a continuous variable for secondary analyses and examined the HR associated with 1 sd increase in the Na:K ratio. Analyses were performed with SAS version 9.2 software (SAS Institute, Inc.).
The REGARDS cohort enrolled 30 239 participants. We excluded 8603 participants because they were missing FFQ data (n 6000) or answered less than 85 % of the FFQ (n 1557) or who had extreme energy intakes (<3347.2 kJ or >20 920 kJ/d for males (n 689) and <2092 kJ or >18 828 kJ/d for females (n 357) (4·184 kJ = 1 kcal)). Compared with blacks (58 %), whites (81 %) were more likely to return the FFQ (P < 0·001). College graduates (77 %) were also more likely than non-graduates (55 %) to return the FFQ (P < 0·001). Participants who did not return the FFQ had a slightly higher BMI (30·0 (sd 6·4) v. 29·1 (sd 3·1) kg/m2, P < 0·001). Age was not associated with the likelihood of returning the FFQ.
The median Na intake was higher in blacks (median = 0·92; interquartile range = 0·43) than in whites (median = 0·83; interquartile range = 0·32). Regardless of race, participants with higher dietary Na:K ratios were more likely to be hypertensives, younger, current smokers, obese and to reside in the Southeast (Table 1). Median energy intake increased across quintiles of Na:K ratio as did the total Na intake.
CAD, coronary artery disease; HTN, hypertension.
* Quintiles are race and sex specific.
Results on risk of death associated with increasing dietary sodium:potassium ratio
Compared with participants in the lowest quintile of Na:K intake, those in the highest quintile were at increased risk of death (Table 2). In age-, race-, sex- and energy-adjusted models, the HR for death comparing quintile 5 (Q5) to quintile 1 (Q1) of dietary Na:K intake was 1·52 (95% CI 1·31, 1·76). Adjustment for BMI, hypertension, diabetes, heart disease, smoking, income and education attenuated the association (HR 1·26; 95 % CI 1·08, 1·47). We also examined the dietary Na:K ratio as a continuous variable. For each standard deviation (0·28 mg Na per mg K) increase in the dietary Na:K ratio, the risk of death increased by 11 % (HR 1·11; 95 % CI = 1·05, 1·16). Although the P value for interaction was not statistically significant, we further examined this association stratified by race based on our a priori hypothesis. The risk of each standard deviation increase in dietary Na:K ratio was similar in blacks (HR 1·12; 95 % CI 1·04, 1·21) and whites (HR 1·09; 95 % CI 1·02, 1·16) (Table 2).
* Quintiles are race and sex specific.
† Model 1 – age, race, sex, total energy intake.
‡ Model 2 – age, race, sex, energy, BMI, hypertension, diabetes, history of heart disease, dyslipidaemia, smoking, income, education and use of statins.
§ For the models stratified by race, model 2 adjustments were used and race was not included in the model.
Results on risk of incident stroke associated with increasing dietary sodium:potassium ratio
Over a mean of 4·9 years of follow-up, we accumulated 363 stroke events. Although there was a suggestion for increased risk of stroke among individuals in Q5 compared with Q1 of dietary Na:K ratio, this association was not statistically significant (Table 3) in analysis before or after stratifying by race. In the fully adjusted model comparing Q5 with Q1 of dietary Na:K ratio, the HR of incident stroke was 1·29 (95 % CI 0·94, 1·78). For a 1 sd increase in dietary Na:K ratio, the HR of incident stroke was 1·08 (95 % CI 0·98, 1·20). Similar to the results for death, there was no statistically significant difference by race. In blacks, all quintiles of dietary Na:K intake had a higher risk of incident stroke compared with Q1. This association was not linear. In whites, the association between dietary Na:K intake and incident stroke was fairly flat across quintiles with a slight increase in Q5 v. Q1 (HR 1·29; 95 % CI 0·88, 1·90). In a sensitivity analysis limited to those who survived at least 6 months beyond the baseline visit and who did not have a history of stroke at baseline, the results were similar (data not shown).
* Quintiles are race and sex specific.
† Model 1 – age, race, sex, energy.
‡ Model 2 – age, race, sex, energy, BMI, hypertension, diabetes, history of heart disease, dyslipidaemia, smoking, income, education.
§ Model 3 – age, race, sex, energy, BMI, hypertension, diabetes, history of heart disease, dyslipidaemia, smoking, income, education, bias in retrieving and extraction of medical records.
║ For the models stratified by race, model 3 adjustments were used and race was not included.
Results on risk of death and risk of incident stroke associated with increasing dietary sodium
Increasing the dietary Na had a weak, non-linear association with risk of death that was attenuated after adjustment for BMI, hypertension, diabetes, dyslipidaemia, history of heart disease, smoking, income and education (Supplementary Table 1). Comparing Q5 with Q1 of dietary Na, the risk of death associated with increasing dietary Na was stronger for the black participants (HR 1·62; 95 % CI 1·03, 2·55) than for the white participants (HR 0·94; 95 % CI 0·67, 1·30). Dietary Na was associated with incident stroke in a pattern similar to that observed for the dietary Na:K ratio (Supplementary Table 2). Again, a weak non-linear association was observed. Comparing Q5 with Q1 of dietary Na, the risk of stroke associated with increasing dietary Na was stronger for the black participants (HR 2·13; 95 % CI 0·88, 5·13) than for the white participants (HR 1·30; 95 % CI 0·66, 2·57), though these estimates were not statistically different from one another.
Food groups and dietary sodium:potassium ratio
As we did not observe racial differences in the Na:K ratio, we also wanted to investigate whether there were differences in food choice contributing to both high and low Na:K ratios. We used Pearson's correlation to compare grams of food from each food group with the Na:K ratio. The foods were largely similar across races. In blacks and whites, high fruit and fruit juice intakes correlated with a lower Na:K ratio (Table 4). Processed meats, bread and non-fried chicken were each correlated with a higher Na:K ratio. In whites, high liquor consumption was correlated with a higher Na:K ratio, while coffee consumption was associated with a lower Na:K ratio. In blacks, added fats such as lard and gravy were correlated with a higher Na:K ratio, while cruciferous vegetables were associated with a lower Na:K ratio.
* For simplicity only the top five and bottom five correlations are shown. All P < 0·001.
Among middle-aged and older adults, a high dietary Na:K ratio was associated with an increased risk of death. This association was observed in blacks and whites and persisted after adjustment for several demographic, socioeconomic and other health factors. Although a similar trend was present for the risk of incident stroke, this association was not statistically significant, which most probably is due to the lower event rate for stroke compared with all-cause mortality. Our findings on higher risk of death with increasing dietary Na:K ratio are consistent with previous research showing that a higher dietary Na:K ratio, as measured in either urine or diet, is associated with a higher risk of death( Reference Cook, Cutler and Obarzanek 25 ). Compared with the results for dietary Na:K ratio, the direction of the association between Na intake and the outcomes was similar but the magnitude of the association was weaker.
Based on previous bench and population research, we hypothesised that the association of the Na:K ratio and both stroke and death would be higher in blacks than whites( Reference Fukui, Otani and Tsuzuki 26 , Reference Luft, Miller and Grim 27 ). Salt sensitivity has been found to be highly correlated with specific phenotypic and metabolic characteristics, particularly for blacks in the USA( Reference Campese, Parise and Karubian 28 , Reference Calhoun and Oparil 29 ). In contrast with whites, salt sensitivity is prevalent in both normotensive and hypertensive blacks( Reference Luft, Miller and Grim 27 , Reference Burnier 30 , Reference Morris, Sebastian and Forman 31 ). Diets that are high in K and low in Na (e.g. the Dietary Approaches to Stop Hypertension diet) effectively lower blood pressure in whites, but do so to a greater degree in blacks( Reference Svetkey, Simons-Morton and Vollmer 32 ). Therefore, we were surprised that we did not find racial differences in the effect of the Na:K ratio (or Na alone) on stroke or mortality in this study. Inclusion of hypertension as a covariate may have dampened the observed association as it may lie in the causal pathway between higher Na intake and either death or stroke. Our findings suggest that an increased Na:K ratio in the diet is harmful regardless of race.
Separating individual nutrients from overall diet quality is challenging. Therefore, it is possible that the association of a higher Na:K ratio with increased risk of all-cause mortality that we observed is due to an improvement in overall diet quality. In fact, we observed that lower Na:K ratio values were correlated with greater intakes of fruit, low-fat milk and vegetable intake. A lower dietary Na:K ratio is associated with a greater intake of whole grains, low-fat dairy products, fruits, vegetables and lean meats, all of which have been associated with better health( Reference Folsom, Parker and Harnack 33 ). However, in our study, the association between a higher Na:K ratio and increased risk of all-cause mortality persisted even after adjustment for alcohol, whole grain, fruit and vegetable intake. In addition to the fact that diet quality itself is associated with improved cardiovascular health, diet quality is also strongly correlated with socioeconomic status( Reference Darmon and Drewnowski 34 ). Although a possibility for residual confounding by socioeconomic status still exists, we controlled for two major indicators of socioeconomic status: income and education.
The present study has potential limitations. We are unable to assess causality due to the study's observational design. Na and K intake were estimated by assessing diet rather than multiple 24-h urine collections, which are known to more accurately reflect Na:K intake, and we did not have the ability to capture salt added to food by the participant. Although individuals may erroneously report nutritional intake on FFQ, dietary questionnaires provide an acceptable classification of nutrient intake on a population level and are widely accepted in cohort studies with large sample sizes such as the REGARDS study( Reference Dyer, Elliott and Chee 35 , Reference Toft, Kristoffersen and Ladelund 36 ). Participants may under-report their Na intake and inflate their intake of more healthful foods. This potential bias would reduce the ability to detect significant associations between the Na:K ratio and outcomes in the present study. Participants who returned the FFQ were more likely to be college educated than those who did not return the FFQ. Although we controlled for education and income in our analyses, individuals of lower socioeconomic status were less represented in our study. Although the REGARDS study has a strong surveillance system, we did not have complete ascertainment of death. Despite this potential bias, we found a significant association between dietary Na:K ratio and mortality. In addition, we did not know the cause of death and were unable to study CVD mortality. Previous studies have demonstrated stronger associations of the Na:K ratio with CVD mortality than all-cause mortality, so we would expect our findings to be strengthened with the addition of these data( Reference Yang, Liu and Kuklina 15 ). Despite these limitations, the REGARDS study also has several notable strengths. These strengths include the enrolment of a large sample of whites and blacks from all over continental USA, physician adjudication of strokes using medical records, and the collection of dietary data using a standardised and validated FFQ.
In the present study, individuals with a reported diet high in Na and low in K had a greater risk of death. Although the physiological mechanisms are not fully understood, diets high in K may mediate the adverse effects of dietary Na, resulting in lower blood pressures and a decreased risk of stroke and stroke mortality. Data from this study reinforce recommendations to lower dietary Na intake and increase K intake as a means of lowering mortality in the general population, but do not support specific recommendations for different race groups. Thus, our data suggest that the implementation of strategies to reduce Na in the US food supply has similar implications for whites and blacks.
This research project was supported by cooperative agreement number U01 NS041588 from the National Institute of Neurological Disorders and Stroke, National Institutes of Health, Department of Health and Human Services. In addition, an investigator-initiated grant from General Mills helped pay for the cost of obtaining nutrient data from the FFQ. The authors thank the investigators and staff of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at http://www.regardsstudy.org. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Neurological Disorders and Stroke or the National Institutes of Health. E. K. K. is supported by a National Scientist Development grant (no. 0635323N) from the American Heart Association. D. A. L. received research support from the National Institutes of Health (P30 DK092926 and K23 AG040278).
The supplementary material for this article can be found at http://www.journals.cambridge.org/jns