Helen M Macdonald, Susan A New, William D Fraser, Marion K Campbell, and David M Reid

Transcription

1 Low dietary potassium intakes and high dietary estimates of net endogenous acid production are associated with low bone mineral density in premenopausal women and increased markers of bone resorption in postmenopausal women 1 5 Helen M Macdonald, Susan A New, William D Fraser, Marion K Campbell, and David M Reid ABSTRACT Background: The Western diet may be a risk factor for osteoporosis. Excess acid generated from high protein intakes increases calcium excretion and bone resorption. Fruit and vegetable intake could balance this excess acidity by providing alkaline salts of potassium. Algorithms based on dietary intakes of key nutrients can be used to approximate net endogenous acid production (NEAP) and to explore the association between dietary acidity and bone health. Objective: We investigated the relation between dietary potassium and protein, NEAP (with an algorithm including the ratio of protein to potassium intake), and potential renal acid load (with an algorithm including dietary protein, phosphorous, potassium, magnesium, and calcium) and markers of bone health. Design: Measurements of bone mineral density (BMD) (n 3226) and urinary bone resorption markers (n 2929) at the lumbar spine and femoral neck were performed in perimenopausal and early postmenopausal women aged y(x SD) in BMD (g/cm 2 ), free pyridinoline (fpyd), and free deoxypyridinoline (fdpd) were expressed relative to creatinine. Dietary intake was assessed with a food-frequency questionnaire. Results: Comparison of the highest with the lowest quartile of potassium intake or the lowest with the highest NEAP showed a 6 8% increase in fpyd/creatinine and fdpd/creatinine. A difference of 8% in BMD was observed between the highest and lowest quartiles of potassium intake in the premenopausal group (n 337). Conclusions: Dietary potassium, an indicator of NEAP and fruit and vegetable intake, may exert a modest influence on markers of bone health, which over a lifetime may contribute to a decreased risk of osteoporosis. Am J Clin Nutr 2005;81: KEY WORDS Fruit, vegetables, net endogenous (noncarbonic) acid production, NEAP, potential renal acid load, PRAL, acid base balance, dietary potassium, bone resorption markers, bone mineral density, menopause INTRODUCTION Osteoporosis is a major public health problem in the Western World and is increasing in the developing world (1). Although diet is likely to play a role, most research has concentrated on calcium and vitamin D at the expense of other nutrients that could plausibly influence bone health. Recent population-based studies have shown beneficial effects of dietary potassium and fruit and vegetables on indexes of bone health (2 4). Also, an ancillary study to the Dietary Approaches to Stopping Hypertension (DASH) trial showed that a diet rich in fruit and vegetables and in low-fat dairy products reduced bone turnover compared with a control diet (5). It was suggested that osteoporosis may, in part, be caused by the continual release of alkaline salts from bone for acid-base balance (6). The Western diet, in particular, generates a large amount of acid. Without sufficient alkaline-forming foods in the diet, bone health may be compromised (7). Shortterm supplementation of postmenopausal women with potassium bicarbonate for 18 d (8) or potassium citrate for 3 mo (9) was shown to reduce bone turnover, and other studies have shown that calcium balance is improved with the administration of organic salts of potassium (10 12). Compared with a base-forming diet, an acid-forming diet was shown to increase urinary calcium excretion by 74% and C-telopeptide excretion by 19% (13). At the cellular level, metabolic acidosis causes calcium efflux from bone (14), stimulates osteoclastic action, and inhibits osteoblastic action (15), whereas the opposite is true of metabolic alkalosis (16). In vitro studies using cell cultures of rat osteoclasts have shown that even a small decrease in ph from 7.25 to 7.15 resulted in a 6-fold increase in the number of resorption pits (17) as a result of osteoclast stimulation; there was little or no resorption at ph values 7.3 (18). 1 From the Department of Medicine and Therapeutics, University of Aberdeen, Medical School Buildings, Aberdeen, United Kingdom (HMM and DMR); the Centre for Nutrition and Food Safety, School of Biomedical and Molecular Sciences, University of Surrey, Guildford, United Kingdom (SAN); the Department of Clinical Chemistry, Royal Liverpool University Hospital, Liverpool, United Kingdom (WDF); and the Health Services Research Unit, University of Aberdeen Medical School, Aberdeen, United Kingdom (MKC). 2 Presented in part at the annual Bone and Tooth Society Meeting, Cardiff, United Kingdom, The views expressed herein are those of the authors. 4 Supported by the UK Food Standards Agency and the UK Arthritis Research Campaign (DMR). The Health Services Research Unit is funded by the Chief Scientist Office of the Scottish Executive Health Department. 5 Reprints not available. Address correspondence to HM Macdonald, Osteoporosis Research Unit, Victoria Pavilion, Woolmanhill Hospital, Aberdeen, AB25 1LD. h.macdonald@abdn.ac.uk. Received May 31, Accepted for publication November 19, Am J Clin Nutr 2005;81: Printed in USA American Society for Clinical Nutrition 923

2 924 MACDONALD ET AL In large population studies, the measurement of acid-base balance is not practical, but algorithms based on the ratio of protein to potassium have been used to estimate net endogenous (noncarbonic) acid production (NEAP) (19). Reanalysis of baseline data from our group collected in 1993 from a subset of 1065 women showed lower BMD with higher dietary NEAP and a trend for higher bone resorption in a group of 62 women (20). The aim of this study was to determine whether there is an association between NEAP or potential renal acid load (PRAL) and indexes of bone health (bone resorption markers and BMD) in the complete population of 3000 women (now perimenopausal and early postmenopausal) 5 7 y from the first visit, and whether there is an independent association between dietary potassium or protein and markers of bone health. SUBJECTS AND METHODS Subjects The subjects were taken from a population-based screening program for osteoporotic fracture risk, involving 5119 women aged y, that took place between 1990 and The women were drawn at random from a 25-mile ( 40-km) radius of Aberdeen, a city with a population of in the northeast of Scotland, with the use of Community Health Index records (21, 22). All participants underwent bone densitometry and risk factor assessment by questionnaire, and the women were invited to undergo further assessment between 1997 and A total of 3883 women underwent a second assessment. There were no significant differences in age, height, weight, or baseline BMD between the women who returned for the second assessment and those who did not. However, women who were still menstruating were more likely to return for the second assessment (53.9% compared with 49.4%), and there were slightly fewer postmenopausal women (46.1% compared with 50.6%). At the second visit, 3510 women provided a second early morning, fasting urine sample for the measurement of bone resorption markers. Most of the women (n 3239) completed a diet and physical activity questionnaire at the time or shortly after the follow-up visit. So as not to bias the data excessively, only 6 women were excluded from the analysis, because they had excessive intakes of dietary potassium ( 8000 mg/d). From this group, there were 3226 measurements of BMD and 2929 measurements of bone resorption markers. Information on health, menopausal status, hormone replacement therapy (HRT), and other medication use was also collected. At the follow-up visit, for the women who had BMD measurements, a total of 336 women were still menstruating, 2877 women were postmenopausal (of whom 1180 women were receiving HRT at the time of the second visit), and 13 women were of unknown menopausal status. For the women who had bone marker measurements, 304 women were still menstruating, 2614 women were postmenopausal (of whom 1073 women were currently receiving HRT), and 11 women were of unknown menopausal status. A summary of the subject numbers is given in Figure 1. Menopausal status and use of HRT were considered independent variables in linear regression models. Self-reported osteoarthritis (n 597) and thiazide use (n 262) were also considered independent variables in the analysis. Subjects with other diseases and conditions, reported at the second visit, that could affect bone metabolism were included. However, the analysis was also repeated with the exclusion of women who FIGURE 1. Summary of the numbers of subjects who had bone mineral density (BMD) measurements (n 3226) and bone resorption markers analyzed (n 2929) at the follow-up visit. self-reported having a disease known to affect bone metabolism (eg, breast cancer, thyroid disease, malabsorption; n 334) and recent users of bisphosphonates (n 22), selective estrogen receptor modulators (n 2), tamoxifen (n 39), or oral glucocorticoids (n 92). The women were weighed at both visits with a set of balance scales (Seca, Hamburg, Germany) calibrated to 0.05 kg while wearing light clothing and no shoes. Height was measured with a stadiometer (Holtain Ltd, Crymych, United Kingdom). Written informed consent was obtained from all of the women, and the study was approved by the Grampian Research Ethics Committee. Bone mineral density measurements Bone mineral density (BMD) of the left femoral neck (FN) and lumbar spine (LS) (L2-L4) was measured by dual-energy X-ray absorptiometry (Norland XR26 and XR36; Cooper Surgical Inc, Trumbull, CT) as described previously (23). Most of the women were scanned by using the XR26, but 388 (14%) women were scanned by using the XR36. A comparison of 50 phantom measurements with both machines showed a small difference (1.258%) in mean BMD between the machines; therefore, a correction factor was used to convert the XR36 values to XR26- equivalent values. The same trends in results were seen whether or not this correction factor was used. Urinary bone resorption markers A second early morning fasting urine sample was used for the analysis of free pyridinoline (fpyd) and free deoxypyridinoline (fdpd) by using a modification of the HPLC method described by Black et al (24). Acidified urine was applied to microgranular cellulose (CC31) in butanol (1/4) and washed before elution with heptafluorobutyric acid (0.1%). The eluent was then analyzed by ion-pair reversed-phase HPLC using fluorescence detection. Acetylated PYD (Quidel/Metra Biosystems, Oxford, United Kingdom) was used as an internal standard. Creatinine was measured in urine with standard automated techniques (Roche, Lewes, United Kingdom), and the results were expressed as fpyd/cr and fdpd/cr (nmol/mmol). The interassay CV for both marker methods was 5.5% across the working concentration

3 DIETARY ACIDITY AND POTASSIUM AND BONE HEALTH 925 range for the assay, which was established by performing repeated analysis of a range of patient and quality-assurance samples (25). Diet and physical activity Dietary assessment was made at the follow-up visit with a food-frequency questionnaire (FFQ) that had been validated with the use of 7-d weighed intakes (26) and serum concentrations of antioxidants (27). A subgroup of women (n 898) also completed the same FFQ at baseline. For most of the women there was little change in nutrient intakes, although mean energy had decreased from a mean ( SD) of to MJ/d (P 0.01, paired t test). Protein had decreased by a mean of 2 to g/d (P 0.01), but there was no significant change in mean potassium intake ( mg/d). The database used for the nutrient analysis is based on McCance and Widdowson s Composition of Foods version 5 (28), which does not include the newer estimates of vitamin D contribution from meat (based on assumptions regarding the greater potency of vitamin D metabolites from meat sources). Physical activity levels (PALs) were obtained by using the same questions as used for the Scottish Heart Health Study (29). The PAL was calculated from the numbers of hours in a 24-h period doing heavy, moderate, or light activities and how many hours were spent sleeping or resting in bed. These questions were asked separately for working and nonworking days. PAL is normally defined as the ratio of energy expenditure divided by the basal metabolic rate (BMR), which is calculated from Schofield equations (30). These equations were derived from data collected from European women. For women aged y, the equation is BMR (MJ/d) weight (kg) (SEE 0.47.) Estimation of the acid-generating potential of the diet or NEAP was calculated according to the equations of Frassetto et al (19): NEAP 62.1 protein/potassium (meq)] 17.9 (1) in which protein and potassium were standardized to MJ, or alternatively NEAP 54.5 protein/potassium (meq)] 10.2 (2) in which the nutrients were not standardized. There was a linear relation between the NEAP values calculated from each equation with a correlation coefficient of 1. Because the energystandardization terms cancel each other out in Equation 1, estimated NEAP was also adjusted for energy with the residual method. PRAL was calculated by using the algorithm of Remer et al (31): PRAL protein 0.49 phosphorus potassium magnesium calcium (3) This was done for calcium from the diet only and also for total calcium including supplements. Because the difference between the 2 was not great, further analysis was carried out with PRAL calculated by using calcium from the diet only. The contribution of organic acids in giving an overall estimate of net acid excretion can be estimated from body surface area multiplied by 41 divided by 1.73 and added to PRAL (31). We used the following formula to calculate body surface area (32). Body surface area height (cm) weight (kg)/3600 (4) Statistical analysis The analyses were performed by using SPSS version 11.0 (2000; SPSS Inc, Chicago, IL). The categorization of nutrient intake was carried out because this is a particularly appropriate method for analyzing FFQ data. One-way analysis of variance (ANOVA) and analysis of covariance (ANCOVA), adjusted for confounding variables, were used to examine differences in characteristics and bone health indexes between different quartiles of dietary intake. Tukey s multiple comparison post hoc ANOVA test was used to identify significant differences between quartiles. A test for linearity was performed on the trend components, which were produced by partitioning the between-groups sums of squares. Chi-square tests were used for categorical variables. Stepwise linear multiple regression analysis was used to determine independent predictors (including dietary variables) of bone resorption markers and BMD. Dummy variables for nonuse and past use of HRT were used to account for HRT use (with current use as the reference). Potassium and protein intakes were adjusted for dietary energy intake by the residual method, which is generally preferred to the use of nutrient density (nutrient divided by energy intake) (33). NEAP (determined by using Equation 1) and PRAL were used unadjusted and adjusted for energy by saving the residuals of the regression with energy intake. Interaction between the dietary variables and menopausal status (women still menstruating compared with postmenopausal women, including HRT users) was tested for both BMD and bone resorption markers by using the general linear model in SPSS. RESULTS Subject characteristics in relation to dietary acidity Division of the women into 4 groups according to quartiles (Q) of estimated NEAP showed that those in the bottom quartile (Q1) produced a mean of 30 meq acid/d, whereas those in the top quartile (Q4) produced a mean of 52 meq/d (Table 1). Characteristics of the women according to each NEAP group showed no significant differences with the exception of height, BMI, FN BMD change, vegetarianism, and smoking. Women in the lowest quartile of NEAP were marginally taller on average and had a lower mean BMI than the women in Q4 (P 0.05, ANOVA with Tukey s test). Q1 contained significantly fewer smokers (P 0.03) and more women consuming vegetarian diets (P 0.001), although there were only 28 vegetarians and 2 vegans in this study. There appeared to be less loss of FN BMD in Q4 than in Q1 (P 0.05), but this difference was not significant after adjustment for confounders (age, weight, height, socioeconomic status, PAL, menopausal status, HRT use, and baseline FN BMD). Similar trends were seen across quartiles of energy adjusted NEAP (data not shown), with no difference observed in the distribution of vegetarians; however, the difference in FN BMD change between Q1 and Q4 of energy-adjusted NEAP was not significant (P 0.10).

4 926 MACDONALD ET AL TABLE 1 Subject characteristics according to quartile (Q) of estimated net endogenous acid production (NEAP) 1 NEAP Q1 (n 807) Q2 (n 809) Q3 (n 807) Q4 (n 803) P (ANOVA or chi-square test) NEAP Equation 1 (meq/d) NEAP Equation 2 (meq/d) Age at baseline (y) Age at follow-up (y) Weight at baseline (kg) Weight at follow-up (kg) Weight change (%/y) Height at baseline (cm) a a a b Height at follow-up (cm) a a,b a,b b BMI at baseline (kg/m 2 ) a b b b BMI at follow-up (kg/m 2 ) a b Physical activity level Baseline FN BMD (g/cm 2 ) Baseline LS BMD (g/cm 2 ) Follow-up FN BMD (g/cm 2 ) Follow-up LS BMD (g/cm 2 ) FN BMD change (%/y) a a,b a,b b LS BMD change (%/y) Baseline menopausal status (%) Premenopausal Perimenopausal Postmenopausal Follow-up menopausal status (%) Premenopausal Perimenopausal Postmenopausal Baseline HRT use (%) None Past Present Follow-up HRT use (%) None Past Present Socioeconomic status (%) I II III IV V No. of vegetarians or vegans Smokers (%) BMD, bone mineral density; FN, femoral neck; LS, lumbar spine; HRT, hormone replacement therapy. Means in a row with different superscript letters are significantly different, P 0.05 (ANOVA with Tukey s test). 2 NEAP [62.1 protein/potassium (meq)] 17.9, in which protein and potassium were standardized to MJ and energy was not adjusted for. 3 x SD (all such values). 4 NEAP [54.5 protein/potassium (meq)] 10.2, in which the nutrients were not standardized and energy was not adjusted for. 5 Test for linearity across quartiles was significant, P Measured at the follow-up visit only. 7 Based on postcode classification, where I represents the most affluent and V represents the most deprived. In terms of the other algorithm for estimating dietary NEAP, PRAL also increased from 9.3 meq/d for Q1 of NEAP to 15.5 meq/d for Q4 of NEAP (Table 2). The correlation between PRAL and NEAP was There was little difference in the quartile categorization of NEAP and PRAL; 76% women were classified in the same quartile and 0.3% women were grossly misclassified (a greater than one quartile difference). The difference between PRAL calculated by using calcium from the diet only and PRAL calculated by using total calcium including supplements was small and only the former was used for subsequent

5 DIETARY ACIDITY AND POTASSIUM AND BONE HEALTH 927 TABLE 2 Selected nutrient and food intakes according to quartiles of estimated net endogenous acid production (NEAP) 1 NEAP Q1 (n 807) Q2 (n 809) Q3 (n 807) Q4 (n 803) P 2 NEAP Equation 1 (meq/d) (9.4, 34.1) (34.1, 39.4) (39.4, 45.1) (45.1, 89.2) NEAP Equation 2 (meq/d) (13.8, 35.5) (35.5, 40.2) (40.2, 45.1) (45.1, 83.9) PRAL diet only (meq/d) a ( 44.1, 7.1) b ( 11.6, 14.7) c ( 3.2, 19.0) d ( 1.0, 58.2) PRAL including calcium from a ( 44.1, 7.1) b ( 11.6, 14.7) c ( 14.5, 19.0) d ( 3.6, 58.2) supplements (meq/d) NAE, PRAL plus contribution from ( 3.2, 51.2) (5.0, 63.2) (35.0, 68.8) (37.8, 109.8) organic acids (meq/d) Dietary potassium intake (mg/d) a (1387, 7419) a (1700, 6693) a (927, 7941) b (1369, 7175) Dietary protein (g/d) a (25.5, 145) b (37.1, 151.7) c (22.7, 198.4) b,c,d (40.3, 236.7) Dietary calcium intake (mg/d) a (245, 2395) b (233, 2456) c (276, 2806) d (276, 2683) Energy intake (MJ/d) a (2.8, 16.5) b (2.5, 16.1) c (1.9, 20.8) d (3.2, 23.4) Ratio of energy intake to calculated a (0.48, 2.85) b (0.43, 3.04) c (0.34, 3.20) d (0.57, 3.92) basal metabolic rate Zinc intake (mg/d) a (2.6, 16.8) b (4.1, 20.5) c (3.0, 20.9) d (4.4, 26.1) Vitamin D intake ( g/d) a (0.1, 14.2) a (0.2, 17.2) a (0.2, 21.8) a (0.4, 34.1) Food group Total fruit intake (g/d) a b c d Total vegetable (g/d) a b c d Total meat (g/wk) a b c d Processed meat (g/wk) a b c d Poultry (g/wk) a b c d Red meat including offal (g/wk) a b c d Offal only (g/wk) a b c d All values are x SD; minimum and maximum values in parentheses. PRAL, potential renal acid load; NAE, net acid excretion. Means in a row with different superscript letters are significantly different, P 0.05 (ANOVA with Tukey s test). 2 One-way ANOVA. Test for linearity across quartiles was significant (P 0.001) for all variables. 3 NEAP [62.1 protein/potassium (meq)] 17.9, in which protein and potassium were standardized to MJ and energy was not adjusted for. 4 NEAP [54.5 protein/potassium (meq)] 10.2, in which the nutrients were not standardized and energy was not adjusted for. analyses. Adding an estimate for organic acids to give an estimate of overall net acid excretion showed good agreement with NEAP; the correlation coefficient was Potassium, fruit, and vegetable intakes decreased with increasing quartile of NEAP (P 0.001, ANOVA). Conversely, there was a progressive increase in the intakes of protein, calcium, vitamin D, and meat (including processed meat, red meat, poultry, and offal) (P 0.001, ANOVA). Quartiles were significantly different from one another by Tukey s test (P 0.05), and tests for linearity were significant (P 0.001) (Table 2). Energy intake was lower in the lower dietary acidity groups (P between quartiles, ANOVA with Tukey s test). There were also more women who were low energy reporters [the ratio of energy intake to calculated basal metabolic rate (EIBMR) was 1.1]; 38% in Q1 compared with 18% in Q4. Similar trends were seen across quartiles of energy-adjusted NEAP (data not shown), with the exception of energy intake (8.1 MJ/d for all 4 groups; P 0.79), EIBMR ( ; P 0.75), and calcium intake, for which the mean intake was greater in Q1 ( mg/d) than in Q4 ( mg/d) (P 0.001). Also, there was no significant difference in the numbers of low energy reporters in each quartile (26%). Bone resorption Concentrations of bone resorption markers were significantly greater in the highest quartile of estimated NEAP than in the lowest quartile [P 0.01 (ANOVA) and P 0.01 (Tukey s test) for both fdpd/cr and fpyd/cr). Concentrations of the bone resorption markers fdpd/cr and fpyd/cr were significantly greater in the highest quartile of estimated PRAL than in the lowest 2 quartiles [Q1 compared with Q4: P 0.01 (ANOVA) and P 0.01 (Tukey s test); Q2 compared with Q4: P 0.02 (ANOVA with Tukey s test)]. The differences in NEAP and PRAL remained significant after adjustment for the confounding variables age, weight, height, socioeconomic status, PAL, menopausal status, and HRT use (P 0.01 for fdpd/cr and P 0.01 for fpyd/cr, ANCOVA), as shown in Figure 2 for fdpd/cr. According to categories of energy-adjusted potassium intake, bone resorption markers were significantly greater in the lowest quartile than in the other quartiles [Q1 compared with Q4: P and P 0.01 (Tukey s test) for both markers; Q1 compared with Q2 and Q3: P 0.05 (ANOVA with Tukey s test) for fdpd/cr; Q1 compared with Q3: P 0.05 (Tukey s test) for fpyd/cr]. For both bone resorption markers, the associations with potassium intake were still significant after adjustment for confounders (P 0.01, ANCOVA; Figure 2). Bone resorption marker concentrations were greater for Q1 of energy-adjusted protein intake than for Q3 (P 0.01 for fpyd/cr and P 0.09 for fdpd/cr, before adjustment for confounders; Tukey s test). The association between protein intake and bone resorption markers was significant after adjustment for confounders (P 0.01 for fpyd/cr and P 0.02 for fdpd/cr; Figure 2). Tests for linearity were significant for NEAP, PRAL, and potassium with both bone resorption markers (P 0.01) and for protein with

6 928 MACDONALD ET AL FIGURE 2. Mean ( 2 SEM) concentrations of the bone resorption marker free deoxypyridinoline expressed relative to creatinine (fdpd/cr) with increasing quartiles (Q) of energy-adjusted potassium intake, estimates of dietary acidity (net endogenous acid production, NEAP; potential renal acid load, PRAL), and energy-adjusted protein intake for all women who had bone resorption markers measured (n 2929). Means were adjusted for age, weight, height, socioeconomic status, physical activity level, menopausal status, and hormone replacement therapy status. Analysis of covariance: P 0.01 for potassium, P 0.01 for NEAP, P 0.01 for PRAL, and P 0.02 for protein. Quartiles within a variable with different lowercase letters are significantly different, P 0.05 (ANOVA with Tukey s test). The test for linearity across quartiles was significant (P 0.01) for potassium, NEAP, and PRAL but not for protein (P 0.06). Mean ( SD) values from Q1 to Q4 for dietary potassium were , , , and mg/d. Mean ( SD) values from Q1 to Q4 for dietary protein were , , , and g/d. Similar trends were seen for the bone resorption marker free pyridonoline; however, for this marker, Q3 was significantly different from Q1 (P 0.01, ANOVA with Tukey s test) for protein, and Q1 was significantly different from Q3 and Q4 (P 0.03 and P 0.01, respectively; ANOVA with Tukey s test), but not from Q2, for potassium. fpyd/cr (P 0.02) but not with fdpd/cr. Similarly, statistically significant results were obtained when these analyses were repeated with the exclusion of women who had diseases or who were taking medication that could affect bone metabolism. Using stepwise multiple linear regression, it was found that the continuous variables of estimated NEAP, PRAL, and dietary potassium were significant independent predictors of bone resorption (Table 3). We found no significant interaction between postmenopausal women (including past and present HRT users) and women who were still menstruating (and not receiving HRT) with dietary potassium or estimates of dietary acidity (NEAP, PRAL) on bone resorption markers. Bone mineral density For all women (n 3228), who were of mixed menopausal status, there was a small (2%) difference in BMD between the lowest and highest quartiles of potassium intake that was significant at the FN (P 0.02 ANOVA with Tukey s test) but not at the LS (P 0.10), although both sites were significant for linearity (FN: P 0.01; LS: P 0.02). There was a significant interaction for BMD with menopausal status (between those women still menstruating and those who were postmenopausal) and with dietary potassium (FN: P 0.01; LS: P 0.03) and a trend for interaction with NEAP (P 0.10) and energy-adjusted NEAP (FN: P 0.05 FN; LS: P 0.05). For the subgroup of women who were still menstruating (n 336), the difference in FN BMD between Q1 and Q4 of potassium intake was 8%. This difference was statistically significant by one-way ANOVA with Tukey s test (P 0.01), was significant for linearity (P 0.01), and remained significant after adjustment for confounders (P 0.01) (Figure 3). A similar trend was seen at the LS; there was a difference of 6% in BMD between the top and bottom quartiles of potassium intake, which was not statistically significant by ANOVA (P 0.06) or ANCOVA (P 0.20). There were no significant differences in BMD at either site between quartiles of estimated NEAP or PRAL or between quartiles of dietary protein intake. For BMD change, there were no differences at either site between quartiles of dietary protein or potassium intake. BMD loss at the highest quartile of NEAP and PRAL was lower than at the lowest quartile (P 0.05 for FN BMD and NEAP and P 0.03 for LS BMD and PRAL, ANOVA with Tukey s test), but these differences were not significant after adjustment for confounders nor were there differences between quartiles of energyadjusted NEAP and PRAL.

7 DIETARY ACIDITY AND POTASSIUM AND BONE HEALTH 929 TABLE 3 Results of multiple regression analyses to identify independent predictors of bone resorption for all women 1 Independent variable 2 Variation explained Unstandardized 95% CI for P % fdpd/cr Intercept , Weight (kg) , Height (cm) , Alcohol (quartiles) , Smoking , Dummy variable Premenopausal , Perimenopausal , Postmenopausal , Past HRT user , Energy-adjusted potassium (mg) , Estimated NEAP (meq) , Energy-adjusted NEAP (meq) , PRAL (meq) , fpyd/cr Intercept , Weight (kg) , Height (cm) , Alcohol (quartiles) , Smoking , Osteoarthritis , Dummy variable Premenopausal , Perimenopausal , Postmenopausal , Past HRT user , Energy-adjusted potassium (mg) , Estimated NEAP (meq) , Energy-adjusted NEAP (meq) , PRAL (meq) , n Independent variables selected from age, weight, height, smoking status, socioeconomic status, physical activity level, alcohol intake, presence of osteoarthritis (yes or no), thiazide use (yes or no), and HRT use. To account for menopausal status and HRT use, dummy variables were included in the model, with current HRT users as the reference. Energy-adjusted potassium, estimated NEAP, energy-adjusted NEAP, PRAL, and energy-adjusted protein were added separately to the model. Energy-adjusted PRAL and protein were not significant predictors of either bone resorption marker. The dependent variables were log transformed. fdpd/cr, free deoxypyridinoline expressed relative to creatinine; fpyd/cr, free pyridinoline expressed relative to creatinine; NEAP, net endogenous acid production; PRAL, potential renal acid load; HRT, hormone replacement therapy. Because we found no significant differences in dietary potassium intake at each visit in a subset of 898 women who had completed dietary questionnaires on both occasions, the association between potassium intake at follow-up and BMD at baseline was tested in women who were menstruating regularly at baseline (n 1541). There was a significant association between quartile of potassium intake and LS BMD [P 0.03 (ANOVA); BMD was significantly greater in Q4 than in Q1, P 0.05 (Tukey s test) and a significant trend for linearity, P 0.01)] but not between quartile of potassium intake and FN BMD (P 0.06, ANOVA). When regression analysis with BMD was used as the dependent variable, potassium intake was found to be a weak predictor of BMD in the full group of women, but it was not significantly different after adjustment for confounding variables. For the subgroup of women who were still menstruating, NEAP was inversely associated with FN BMD and LS BMD, and dietary potassium was a positive independent predictor of FN BMD and LS BMD after adjustment for confounders (Table 4), but neither protein intake nor PRAL added significantly to the model at either site. Similar statistically significant associations were found for BMD at the baseline visit (data not shown). None of the dietary variables was found to be a predictor of BMD change at either site. The exclusion of women who had conditions or who were taking medication likely to affect bone metabolism showed both potassium and dietary acidity (NEAP) to be significant predictors of BMD at both visits and both sites and accounted for a greater percentage (1 3%) of the variation in BMD. DISCUSSION Bone resorption This study showed that, for women around the time of menopause, NEAP was associated with increased concentrations of bone resorption markers. At a 6 8% increase in bone resorption markers between the lowest and highest quartiles of NEAP or potassium, the difference in mean potassium intake between these groups was only 17.1 mmol (for quartiles of NEAP) or 35

8 930 MACDONALD ET AL and height accounted for most of the variation in BMD. BMD change in the early postmenopausal period is dominated by hormonal influences (estrogen, progesterone), which may explain why we did not observe an association between estimated NEAP and BMD for the immediate postmenopausal group and why protein and potassium intakes were not associated with BMD change. The difference in BMD change observed for quartiles of NEAP and PRAL may have been a result of the women in the lowest group of NEAP being slightly taller and having a lower BMI, because this finding was not significant after adjustment for energy or after adjustment for confounders. FIGURE 3. Mean ( 2 SEM) bone mineral density (BMD) at the femoral neck (FN) and lumbar spine (LS) with increasing quartiles (Q) of energyadjusted potassium intake for premenopausal women (n 336); the values were adjusted for age, weight, height, socioeconomic status, physical activity level, and regularity of menstruation. Analysis of covariance: P 0.01 for FN BMD and P 0.20 for LS BMD. There was an interaction for BMD between premenopausal and postmenopausal women (including hormone replacement therapy users) with potassium intake (P 0.01 for FN BMD and P 0.03 for LS BMD). Quartiles within a variable with different lowercase letters are significantly different, P 0.01 (ANOVA with Tukey s test); P 0.06 for LS BMD (ANOVA with Tukey s test). The test for linearity across quartiles (ANOVA) was significant for FN BMD (P 0.01) and LS BMD (P 0.02). Mean ( SD) values from Q1 to Q4 for dietary potassium were , , , and mg/d. mmol (for quartiles of energy-adjusted potassium intake). In comparison, Buclin et al (13) found 19% less C-telopeptide excretion in men consuming a base-producing diet than in those consuming an acid-producing diet, in which the difference in potassium intake between the 2 diets was 89 mmol. Modest increases in bone resorption with increased NEAP are consistent with diet playing a long-term role in contributing a mild, lowgrade acidosis over a lifetime that is within physiologically normal limits and yet may still have a chronic effect on bone health. It is known that osteoclasts are stimulated to resorb bone at an acidic ph, being most sensitive to changes at ph values of 7.1 (34, 35). Models of metabolic acidosis in which bone was incubated in reduced bicarbonate medium showed stimulation of bone resorption by osteoclasts and inhibition of bone formation by osteoblasts (36). It would appear that even subtle chronic acidosis could be sufficient to cause considerable bone loss over time (37). Bone mineral density For women who were still menstruating, we found a difference of 8% BMD between those in Q1 and Q4 of dietary potassium intake, which can be considered to be biologically significant because it is equal to one-half the SD for this population. If maintained into old age, this could reflect a 30% decrease in fracture risk for those with higher intakes of potassium. Weight Dietary protein Diets that contain less animal protein and more fruit and vegetables have been suggested to be beneficial to bone health by virtue of their high potassium content (38). However, if foods are scored according to their PRAL (39), cereals, grains, and cheeses have high PRAL values; the PRAL for high-protein cheeses is more than twice that of meats or fish (39, 40). This may explain why some studies have not found a link between vegetarianism and increased BMD (41), although other factors (eg, weight, exercise) likely differentiate vegetarians from nonvegetarians (42). We found more vegetarians in the lowest quartile of NEAP, although the number of reported vegetarians in our study was very low ( 0.1%). The overall prevalence reported in the general UK population is 5%, but the proportion of vegetarian women decreases with age [from 11% of y-olds to 4% of y-olds (43)], and the mean age of our population was older. Our data suggest that a low protein intake may be detrimental to bone health because women in the lowest protein group had significantly greater concentrations of bone resorption markers. The Framingham Study found that a higher animal protein intake was associated with lower bone loss and that nonanimal protein sources were not related to BMD (44). The influence of protein may depend on whether the overall diet is balanced in terms of its acid-generating potential. A beneficial role for protein was noted in elderly subjects, provided that they were replete in calcium (45). Most of the women in our study had sufficient calcium in their diet. However, we cannot be certain that the higher concentration of bone resorption markers in the lowest quartile of protein intake was a result of low intakes of protein or to some other nutrient that was significantly lower in the lowest quartile (such as zinc or vitamin D). In the United Kingdom, meat and meat products provide 40% of the zinc in the diet (46), and zinc is associated with increased bone mass in premenopausal women (2). Because the bone resorption markers were expressed relative to creatinine, and creatinine concentrations increase with increasing protein intakes, this could be another reason why concentrations of the bone resorption markers were greater in the lowest quartile of protein intake. Similarly, if there is confounding by creatinine, we could be underestimating the association of NEAP with markers of bone resorption. Dietary potassium Potassium salts may benefit bone health by providing an anion that can be metabolized completely to carbon dioxide, or they may influence calcium excretion directly (10 12, 47). Potassium citrate has been shown to prevent bone resorption induced by dietary salt (sodium chloride) (48), and the DASH diet reduced bone turnover at low, medium, and high sodium intakes, although there was no difference in sodium excretion between the

9 DIETARY ACIDITY AND POTASSIUM AND BONE HEALTH 931 TABLE 4 Results of multiple regression analyses to identify independent predictors of bone mineral density (BMD) at follow-up for the subsets of premenopausal and perimenopausal women 1 Independent variable 2 Variation explained Unstandardized 95% CI for P % FN BMD 3 Intercept , Weight (kg) , Height (cm) , Energy-adjusted potassium (mg) , NEAP (meq) , Energy-adjusted NEAP (meq) , LS BMD 4 Intercept , Weight (kg) , Height (cm) , Energy-adjusted potassium (mg) , NEAP (meq) , Energy-adjusted NEAP (meq) , n 336. FN, femoral neck; LS, lumbar spine; NEAP, net endogenous acid production. 2 Independent variables were selected from age, weight, height, smoking status, socioeconomic status, physical activity level, alcohol intake, presence of osteoarthritis (yes or no), thiazide use (yes or no), and perimenopausal status (irregular periods). Energy-adjusted potassium, NEAP, and energy-adjusted NEAP were added separately to the model. Energy adjustment of NEAP was done by the residual method as for the normal dietary variables (potassium and protein). The dependent variables were log transformed. 3 Interactions between premenopausal and postmenopausal women (including hormone-replacement therapy users) were as follows: potassium (P 0.01), NEAP (P 0.09), and energy-adjusted NEAP (P 0.05). 4 Interactions between premenopausal and postmenopausal women (including hormone replacement therapy users) were as follows: potassium (P 0.03), NEAP (P 0.09), and energy-adjusted NEAP (P 0.05). DASH and the control diet groups at each level of salt intake (5). Barzel (49) suggests that the role of the anion chloride on saltinduced hypercalcuria is a special aspect of the general effect of acid-base imbalance on bone. We were unable to explore the role of sodium chloride in this study because the FFQ is not a reliable tool for estimating salt intake. Dietary potassium could simply be a marker of fruit and vegetable intake, and it may be other components in fruit and vegetables that have an influence on bone metabolism, eg, vitamin K (50), vitamin C (51), folate (52), and phytoestrogens (53). It has also been argued that the healthy kidney is able to cope with the demands of an acidic diet (54) and that the acid-base hypothesis is only relevant for persons with impaired renal function. However, in disputing this, it was claimed that small increases in blood ph that are within physiologically normal values can still virtually eradicate net renal acid excretion and affect bone metabolism (17, 55). Vegetables and herbs had a beneficial effect on bone resorption in young rats, over and above that of providing alkaline metabolites (56), but it is not known whether these diets are directly comparable with a normal human diet and, if so, whether similar effects on bone resorption would be observed, especially in an older population. This study involved a large population-based investigation of diet and bone health in women who were randomly selected from the community and for whom many important confounding variables were adjusted for. Some limitations of the study were that the data on diet and PAL were obtained by self-reported questionnaire (which could lead to possible overreporting on specific aspects of the diet, such as fruit and vegetable consumption), the data for the complete population were collected at the follow-up visit and may not fully represent the diet over the longer term, and the bone resorption data would reflect only recent bone turnover. Also, a limitation of dual-energy X-ray absorptiometry is that it only provides a measurement of areal BMD. To unravel the subtle changes of dietary acidity on bone health, a more detailed evaluation of bone status, which includes bone geometry and bone quality, may be required. This could be attained by using other techniques, such as computed tomography, quantitative ultrasound, and trabecular structural analysis. However, each of these methods has its own problems of interpretation. All women were included so as not to bias the sample. When women who had a condition or were taking medication that could affect bone metabolism were excluded, the magnitude of the associations was found to be stronger. These data suggest that it would be worthwhile to explore further the link between dietary potassium, dietary protein, and markers of bone health. A longer-term intervention study is required to fully evaluate whether fruit and vegetable intakes affect human bone metabolism through the provision of organic salts of potassium or other components. We are grateful to David Grubb of the Rowett Research Unit for running the ORACLE program to analyze our FFQ and physical activity questionnaire data and to the following persons, who assisted with the FFQ data: Fidelma Moore (data entry), Fiona Downie (estimation of fruit and vegetable intakes), and Yinka Mackay (estimation of meat intake). We are also extremely grateful for the hard work of the radiographers and research nurses at the Osteoporosis Research Unit and to all of the women who kindly participated in the study. HMM carried out the study and was responsible for the data analysis and for writing the manuscript. SAN was involved in the design of the baseline dietary study and reviewed the manuscript. WDF provided the bone marker results and critically reviewed the manuscript. MKC gave statistical advice and reviewed the manuscript. DMR was responsible for the study design of

10 932 MACDONALD ET AL APOSS and reviewed the manuscript. None of the authors had a financial or commercial interest in any company or organization sponsoring the research for this study. REFERENCES 1. Genant HK, Cooper C, Poor G, et al. Interim report and recommendations of the World Health Organization Task-Force for Osteoporosis. Osteoporos Int 1999;10: New SA, Bolton-Smith C, Grubb DA, Reid DM. Nutritional influences on bone mineral density: a cross-sectional study in menopausal women. Am J Clin Nutr 1997;65: Tucker KL, Hannan MT, Chen H, Cupples LA, Wilson PWF, Kiel DP. Potassium, magnesium, and fruit and vegetable intakes are associated with greater bone mineral density in elderly men and women. Am J Clin Nutr 1999;69: New SA, Robins SP, Campbell MK, et al. Dietary influences on bone metabolism further evidence of a positive link between fruit and vegetable consumption and bone health? Am J Clin Nutr 2000;71: Lin PH, Ginty F, Appel LJ, et al. The DASH diet and sodium reduction improve markers of bone turnover and calcium metabolism in adults. J Nutr 2003;133: Wachman A, Bernstein DS. Diet and osteoporosis. Lancet 1968; 1: Barzel US, Massey LK. Excess dietary protein can adversely affect bone. J Nutr 1998;128: Sebastian A, Harris ST, Ottaway JH, Todd KM, Morris RC. Improved mineral balance and skeletal metabolism in postmenopausal women treated with potassium bicarbonate. N Engl J Med 1994;330: Marangella M, Di Stefano M, Casalis S, Berutti S, D Amelio P, Isaia GC. Effects of potassium citrate supplementation on bone metabolism. Calcif Tissue Int 2004;74: Lemann J, Pleuss JA, Gray RW, Hoffmann RG. Potassium administration reduces and potassium deprivation increases urinary calcium excretion in healthy adults. Kidney Int 1991;39: Lemann J, Gray RW, Pleuss JA. Potassium bicarbonate, but not sodium bicarbonate, reduces urinary calcium excretion and improves calcium balance in healthy men. Kidney Int 1989;35: Lemann J, Pleuss JA, Gray RW. Potassium causes calcium retention in healthy adults. J Nutr 1993;123: Buclin T, Cosma M, Appenzeller M, et al. Diet acids and alkalis influence calcium retention in bone. Osteoporos Int 2001;12: Bushinsky DA. Net calcium efflux from live bone during chronic metabolic, but not respiratory, acidosis. Am J Physiol 1989;256:F Krieger NS, Sessler NE, Bushinsky DA. Acidosis inhibits osteoblastic and stimulates osteoclastic activity in vitro. Am J Physiol 1992;262: F Bushinsky DA. Metabolic alkalosis decreases bone calcium efflux by suppressing osteoclasts and stimulating osteoblasts. Am J Physiol 1996; 271:F Arnett TR, Spowage M. Modulation of the resorptive activity of rat osteoclasts by small changes in extracellular ph near the physiological range. Bone 1996;18: Meghji S, Morrison MS, Henderson B, Arnett TR. ph dependence of bone resorption: mouse calvarial osteoclasts are activated by acidosis. Am J Physiol Endocrinol Metab 2001;280:E Frassetto LA, Todd KM, Morris RC Jr, Sebastian A. Estimation of net endogenous noncarbonic acid production in humans from diet potassium and protein contents. Am J Clin Nutr 1998;68: New SA, Macdonald HM, Campbell MK, et al. Lower estimates of net endogenous non-carbonic acid production are positively associated with indexes of bone health in premenopausal and perimenopausal women. Am J Clin Nutr 2004;79: Torgerson DJ, Garton MJ, Donaldson C, Russell IT, Reid DM. Recruitment methods for screening programmes: trial of an improved method within a regional osteoporosis study. BMJ 1993;307: Garton MJ, Torgerson DJ, Donaldson C, Russell IT, Reid DM. Recruitment methods of screening programmes: trial of a new method within a regional osteoporosis study. BMJ 1992;305: Macdonald HM, New SA, Golden MH, Campbell MK, Reid DM. Nutritional associations with bone loss during the menopausal transition: evidence of a beneficial effect of calcium, alcohol, and fruit and vegetable nutrients and of a detrimental effect of fatty acids. Am J Clin Nutr 2004;79: Black D, Duncan A, Robins SP. Quantitative analysis of the pyridinium crosslinks of collagen in urine using ion-paired reversed-phase highperformance liquid chromatography. Anal Biochem 1988;169: MacDonald AG, Birkinshaw G, Durham B, Bucknall RC, Fraser WD. Biochemical markers of bone turnover in seronegative spondylarthropathy: relationship to disease activity. Br J Rheumatol 1997;36: New SA. An epidemiological investigation into the influence of nutritional factors on bone mineral density and bone metabolism. PhD thesis. University of Aberdeen, United Kingdom, Bodner CH, Soutar A, New SA, et al. Validation of a food frequency questionnaire for use in a Scottish population: correlation of antioxidant vitamin intakes with biochemical measures. J Hum Nutr Diet 1998;11: Holland B, Welch AA, Unwin ID, Buss DH, Paul AA, Southgate DAT. McCance and Widdowson s the composition of foods. Cambridge, United Kingdom: Royal Society of Chemistry, Tunstall-Pedoe H, Smith WCS, Crombie IK, Tavendale R. Coronary risk factor and lifestyle variation across Scotland: results from the Scottish Heart Health Study. Scot Med J 1989;34: Department of Health. Dietary reference values for food energy and nutrients for the United Kingdom. London: Her Majesty s Stationery Office, Remer T, Dimitriou T, Manz F. Dietary potential renal acid load and renal net acid excretion in healthy, free-living children and adolescents. Am J Clin Nutr 2003;77: Mosteller RD. Simplified calculation of body-surface area. N Engl J Med 1987;317: Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic studies. Am J Clin Nutr 1997;65(suppl):1220S 8S. 34. Arnett TR, Dempster DW. Effect of ph on bone resorption by rat osteoclasts in vitro. Endocrinology 1986;119: Goldhaber P, Rabadjija L. H stimulation of cell-mediated bone resorption in tissue culture. Am J Physiol 1987;253:E Bushinsky DA. Stimulated osteoclastic and suppressed osteoblastic activity in metabolic but not respiratory acidosis. Am J Physiol 1995;268: C Arnett T. Regulation of bone cell function by acid-base balance. Proc Nutr Soc 2003;62: Sellmeyer DE, Stone KL, Sebastian A, Cummings SR. A high ratio of dietary animal to vegetable protein increases the rate of bone loss and the risk of fracture in postmenopausal women. Study of Osteoporotic Fractures Research Group. Am J Clin Nutr 2001;73: Remer T, Manz F. Potential renal acid load of foods and its influence on urine ph. J Am Diet Assoc 1995;95: Remer T, Manz F. Estimation of the renal net acid excretion by adults consuming diets containing variable amounts of protein. Am J Clin Nutr 1994;59: Tesar R, Notelovitz M, Shim E, Kauwell G, Brown J. Axial and peripheral bone density and nutrient intakes of postmenopausal vegetarian and omnivorous women. Am J Clin Nutr 1992;56: Barr SI, Prior JC, Janelle KC, Lentle BC. Spinal bone mineral density in premenopausal vegetarian and nonvegetarian women: cross-sectional and prospective comparisons. J Am Diet Assoc 1998;98: Henderson L, Gregory J, Swan G. National Diet and Nutrition Survey: adults aged 19 to 64 years. Vol 1. Types and quantities of foods consumed. London: The Stationery Office, Hannan MT, Tucker KL, Dawson-Hughes B, Cupples A, Felson DT, Kiel DP. Effect of dietary protein on bone loss in elderly men and women: the Framingham Osteoporosis Study. J Bone Miner Res 2000; 15: Dawson-Hughes B, Harris SS. Calcium intake influences the association of protein intake with rates of bone loss in elderly men and women. Am J Clin Nutr 2002;75: Ministry of Agriculture Fisheries and Food (MAFF). The dietary and nutritional survey of British adults: further analysis. London: Her Majesty s Stationery Office, Brunette MG, Mailloux J, Lajeunesse D. Calcium transport through the luminal membrane of the distal tubule. I. Interrelationship with sodium. Kidney Int 1992;41: Sellmeyer DE, Schloetter M, Sebastian A. Potassium citrate prevents increased urine calcium excretion and bone resorption induced by a high sodium chloride diet. J Clin Endocrinol Metab 2002;87: Barzel US. Anion effects on calcium metabolism. J Bone Miner Res 1997;12:298 9.