Arterialized Venous Bicarbonate Is Associated With Lower Bone Mineral Density and an Increased Rate of Bone Loss in Older Men and Women

Abstract

Context:

Higher dietary net acid loads have been associated with increased bone resorption, reduced bone mineral density (BMD), and increased fracture risk.

Objective:

The objective was to compare bicarbonate (HCO₃) measured in arterialized venous blood samples to skeletal outcomes.

Design:

Arterialized venous samples collected from participants in the Health, Aging and Body Composition (Health ABC) Study were compared to BMD and rate of bone loss.

Setting:

The setting was a community-based observational cohort.

Participants:

A total of 2287 men and women age 74 ± 3 years participated.

Intervention:

Arterialized venous blood was obtained at the year 3 study visit and analyzed for pH and pCO₂. HCO₃ was determined using the Henderson-Hasselbalch equation.

Main Outcome Measure:

BMD was measured at the hip by dual-energy x-ray absorptiometry at the year 1 (baseline) and year 3 study visits.

Results:

Plasma HCO₃ was positively associated with BMD at both year 1 (P = .001) and year 3 (P = .001) in models adjusted for age, race, sex, clinic site, smoking, weight, and estimated glomerular filtration rate. Plasma HCO₃ was inversely associated with rate of bone loss at the total hip over the 2.1 ± 0.3 (mean ± SD) years between the two bone density measurements (P < .001). Across quartiles of plasma HCO₃, the rate of change in BMD over the 2.1 years ranged from a loss of 0.72%/y in the lowest quartile to a gain of 0.15%/y in the highest quartile of HCO₃.

Conclusions:

Arterialized plasma HCO₃ was associated positively with cross-sectional BMD and inversely with the rate of bone loss, implying that systemic acid-base status is an important determinant of skeletal health during aging. Ongoing bone loss was linearly related to arterialized HCO₃, even after adjustment for age and renal function. Further research in this area may have major public health implications because reducing dietary net acid load is possible through dietary intervention or through supplementation with alkaline potassium compounds.

The skeletal effects of chronically ingesting a diet imbalanced in acid and base precursors remain in contention, although higher estimated dietary net acid loads have been associated with increased bone resorption, reduced bone mineral density (BMD), and increased fracture risk (1–4). The metabolism of dietary protein delivers noncarbonic acid to the systemic circulation from the oxidation of organic sulfur in sulfur-containing proteins, creating sulfuric acid (5–7). Dietary base in the form of bicarbonate results from the metabolism of organic potassium salts (eg, citrate, malate, gluconate), found predominantly in fruit and nongrain vegetable foods (8). Diets rich in protein and cereal grains but limited in vegetable/fruit intake, common among inhabitants of industrialized nations, tend to be rich in acid precursors and low in base precursors. Metabolism of such diets creates sufficient excess noncarbonic acid to induce a low-level chronic metabolic acidosis that correlates in severity with the dietary net acid load (5, 7, 9–11).

Renal acid excretory capacity becomes progressively impaired as renal function declines with aging (12–15). Consequently, individuals who habitually eat diets rich in protein/cereal and limited in vegetable/fruit develop a progressive increase in blood acidity and decrease in plasma bicarbonate with age (14, 16), even without increases in the dietary net acid load. Even with normal renal function, the kidneys do not excrete the entire daily acid load generated by a net acid-producing diet (10, 11). Nevertheless, arterial pH is maintained within a very narrow range inversely related to the magnitude of the daily dietary net acid load (11, 17).

A considerable body of data support the hypothesis that the skeleton functions as a continual base reservoir, providing for the titration of metabolic acid over a lifetime of dietary net acid loading (18), due predominantly to mobilization of alkali from bone (11, 19–21) that is used to titrate the persistent dietary acid load. Bone releases base, in the form of alkaline salts of calcium (phosphates, carbonates, hydroxides), in response to the extracellular acidity and hypobicarbonatemia that are induced by the net acid-producing diet and amplified by age-related decline in renal acid-base regulatory ability (17, 22–26). As base release persists, the liberated minerals accompanying the base are excreted in the urine, and bone mass progressively declines (27–29). In addition, extracellular acidification and hypobicarbonatemia in vitro increase the activity of osteoclasts (30–34) and suppress the activity of osteoblasts (32), suggesting that acidosis-induced loss of bone mass may be, in part, a cell-mediated process.

Whereas accumulating evidence suggests that the habitual ingestion of net acid-producing diets has adverse consequences on the skeleton, the degree of the attendant extracellular acidification has not been directly linked to the degree of bone loss. To investigate the relationship between systemic acid-base status and the skeleton, we used pCO₂-adjusted plasma bicarbonate measured in arterialized venous blood samples as an index of dietary net acid-induced systemic acid-base status collected from men and women participating in the Heath, Aging and Body Composition (Health ABC) study and examined the relationship with BMD and interval rate of bone loss.

Subjects and Methods

Health ABC is a prospective observational cohort study examining the relationship between changes in body composition and disability, morbidity, and mortality. The cohort is comprised of 3075 men and women who were between ages 70 and 79 years at recruitment in 1997–1998 at the University of Pittsburgh and the University of Tennessee at Memphis. Participants were recruited from all age-eligible Blacks residing in Pittsburgh, Pennsylvania, and Memphis, Tennessee, and a random sample of White Medicare beneficiaries. The cohort is approximately 40% Black and evenly divided between male and female participants. Potential participants were excluded if they reported the inability to independently perform activities of daily living, walk up 10 steps without resting, or walk a quarter of a mile. The appropriate institutional review boards approved the study procedures, and all participants provided informed consent.

Arterialized venous blood gas measurements

Arterialized venous blood was obtained at the year 3 clinic visit by inserting a butterfly catheter into a superficial vein on the dorsum of the participant's hand. The butterfly tubing was flushed with 3 mL heparin flush solution, and the forearm and hand were wrapped snugly in a water circulating heating pad at 42°C with the end of the butterfly tubing extending beyond the heating pad. After a minimum of 15 minutes, 3–5 mL of blood/heparin flush mixture was withdrawn from the butterfly; then 2–3 mL of blood were withdrawn in a preheparinized blood gas syringe, any air was expelled, the syringe was capped, and the sample was placed on ice immediately. The whole-blood samples were promptly analyzed in triplicate for pH, pCO₂, and pO₂ on a Radiometer ABL5 blood gas analyzer (Radiometer America, Inc). Standards were analyzed daily as well as immediately before and after each participant sample.

A two-point calibration using manufacturer standards was performed each morning and every 4 hours during machine operation. A quality control (QC) standard ampule was analyzed in duplicate before each participant's blood gas sample was analyzed, and a second, different QC ampule was analyzed in duplicate after each participant's sample was analyzed. On the rare occasions when the QC ampule values were not in the expected range, the machine underwent trouble shooting until the QC procedures were passed.

The pH and pCO₂ in arterialized venous samples correlate to arterial samples with r = 0.97 to r = 0.99 (35). Bicarbonate was determined using the Henderson-Hasselbalch equation by the blood gas analyzer from directly measured pH and pCO₂. To isolate the metabolic component of acid-base status, the resulting bicarbonate values were adjusted to the mean pCO₂ for the cohort according to the methods of Kurtz et al (11). After adjustment for pCO₂, the triplicate bicarbonate values for each individual were averaged to provide a single bicarbonate value for each participant.

Bone mineral density

BMD was measured at the proximal femur by dual-energy x-ray absorptiometry (DXA) on a Hologic QDR 4500A densitometer using software version 9.03 (Hologic, Inc). DXA quality assurance included daily phantom scanning and circulation between the two sites of a cross-calibration phantom. BMD of the proximal femur was obtained at year 1 and year 3 visits; there were 2.1 ± 0.3 (mean ± SD) years between the two DXA measurements. Annualized percentage change in bone density was calculated as the difference between the second bone density measurement and the baseline value divided by the baseline value, multiplied by 100, and then divided by the number of years between measurements.

Covariates

Covariates included in all models were assessed at year 1. Height was measured using a Harpenden stadiometer. Weight was measured with a calibrated balance beam scale. Demographic characteristics and health habits such as smoking were assessed by an interviewer-administered standardized questionnaire. Smoking was quantified as ever smoker yes/no. Serum was analyzed for creatinine using standard methods at the Laboratory of Clinical Biochemistry at the University of Vermont. Estimated glomerular filtration rate (eGFR) was calculated using the formula from the Modification of Diet in Renal Disease (MDRD) study (36). Protein and potassium intakes were measured using the Block dietary food frequency questionnaire (37).

Statistical analyses

χ² tests were used for categorical variables, and ANOVA was used for continuous variables to test for differences between individuals in whom bicarbonate measurements had been obtained and those in whom measurements could not be obtained because they had either withdrawn from the study or had technical issues such as difficult venous access. Linear regression was used to relate year 3 arterialized venous bicarbonate to year 1 and year 3 femoral neck and total hip BMD and annualized rate of bone loss at the femur between the two visits. Results for the femoral neck and total hip were similar; results for total hip are presented. Analyses were adjusted for age, race, sex, clinic site, weight, and smoking. Other covariates potentially related to bone density or acid-base balance, including protein intake, renal function, dietary potassium intake, calcium intake, angiotensin-converting enzyme inhibitor use, and angiotensin II receptor blocker use were added to the model using backward stepwise regression, retaining variables associated with the outcome variable at P < .05.

To examine for possible nonlinear effects, plasma bicarbonate was divided into quartiles, and BMD and rate of bone loss were examined across quartiles with adjustment for the same potential confounders.

Results

Arterialized venous plasma bicarbonate measurements were available in 2287 participants, representing 75% of the participants who had hip BMD measurements at year 1 and 89% of those who had hip BMD measurements at year 3. Participants in whom blood gas measurements could not be obtained (n = 788) were more likely to be Black (52 vs 38%; P < .001), slightly older (73.9 vs 73.5 y; P = .01), and have a smoking history (59 vs 55%; P = .04), but they did not differ in weight, proportion from each sex, estimated protein intake, or eGFR (Table 1). BMD was significantly higher in those with a blood gas measurement (P = .05).

Table 1.

Baseline (Year 1) Characteristics of the Health, Aging and Body Composition Cohort in Whom Arterialized Venous Blood Gas Measurements Were Obtained at the Year 3 Visit and Those in Whom Measurements Could Not Be Obtained

Characteristic	Have Bicarbonate Measurement	No Bicarbonate Measurement	P Value
n	2287	788
Age, y	73.5 ± 2.8	73.9 ± 3.0	.001
Race, %			.001
Black	38	52
White	62	48
Sex, %			.20
Male	49	47
Female	51	53
Weight, kg	76.0 ± 15.0	75.2 ± 15.3	.20
eGFR, mL/min	73.0 ± 15.8	71.7 ± 18.9	.07
Total hip BMD, g/cm2	0.890 ± 0.17	0.876 ± 0.18	.05
Total protein intake, g/d	66.9 ± 28.8	68.2 ± 31.2	.40
Ever smoked, %	55	59	.04

Values are expressed as mean ± SD or percentage.

The pCO₂-adjusted bicarbonate averaged 25.1 ± 1.4 mmol/L (mean ± SD) in the whole cohort and differed between the sexes and two races. White women had the highest values, averaging 25.5 ± 1.3 mmol/L whereas Black men had the lowest values, averaging 24.7 ± 1.5 mmol/L. In the whole cohort, plasma bicarbonate was positively associated with BMD at both the year 1 (r = 0.56; P = .001) and year 3 (r = 0.54; P = .001) measurements in models adjusted for age, race, sex, clinic site, weight, smoking, and eGFR. eGFR was significantly related to BMD at year 1 and at year 3; however, the addition of eGFR to the bone density models did not change the relationship between arterialized venous bicarbonate and bone density. Dietary intakes including total protein intake, potassium intake, and calcium intake were not significantly related to bone density in these models and were not retained in the final model. Models were further adjusted for angiotensin-converting enzyme inhibitor use and angiotensin II receptor blocker use; these medication variables were not significantly related to bone outcomes and did not change the relationship between bicarbonate measures and skeletal outcomes. Because fewer bicarbonate measurements were available in Black participants (n = 869) compared to White participants (n = 1418), models were also stratified by race. Within each race strata, the relationships between bicarbonate, BMD, and bone loss were the same as in the overall models adjusted for race. Subanalyses by sex revealed similar relationships in both subgroups. Each mmol/L higher level of bicarbonate was associated with 0.006 g/cm² (95% confidence interval [CI], 0.002–0.01) greater bone density at year 1 and 0.007 g/cm² (95% CI, 0.003–0.011) higher bone density at year 3 (Table 2).

Table 2.

Coefficients and β (Standardized Regression) Coefficients for Variables in Multivariate Model Relating Arterialized Venous Bicarbonate at Year 3 to BMD at Baseline (Year 1) and Year 3 Among the 2287 Members of the Health, Aging and Body Composition Cohort in Whom Arterialized Venous Blood Gas Measurements Were Obtained at the Year 3 Visit

Variable	Baseline (Year 1)				Year 3
	Coefficient	β Coefficient	SE	P Value	Coefficient	β Coefficient	SE	P Value
Arterialized venous bicarbonate, mmol/L	0.006	0.051	0.002	.001	0.007	0.057	0.002	.001
Race, Black/White	0.064	0.185	0.006	<.001	0.057	0.163	0.006	<.001
Gender, male/female	−0.124	−0.371	0.006	<.001	−0.114	−0.333	0.006	<.001
Clinic site, Memphis/Pittsburgh	0.028	0.084	0.005	<.001	0.031	0.092	0.005	<.001
Age, y	−0.003	−0.056	0.001	<.001	−0.004	−0.067	0.001	<.001
Weight, kg	0.005	0.418	0.0001	<.001	0.005	0.437	0.0002	<.001
Smoking status, yes/no	−0.002	−0.009	0.003	.56	−0.001	−0.003	0.003	.85
eGFR (MDRD), mL/min/1.73 m2	−0.001	−0.072	0.0002	<.001	−0.001	−0.067	0.0002	<.001

Abbreviation: MDRD, Modification of Diet in Renal Disease equation.

Plasma bicarbonate was inversely associated with rate of bone loss at the total hip over the 2.1 ± 0.3 (mean ± SD) years between BMD measurements in models adjusted for age, race, sex, clinic site, weight, and smoking (P < .001). The lower the plasma bicarbonate, the greater the rate of annual bone loss. Bone loss was 0.2%/y (95% CI, 0.09–0.3) higher for every 1 mmol/L lower level of bicarbonate (Table 3). MDRD-eGFR, baseline (year 1) hip bone density, total protein intake, potassium intake, total calcium intake, angiotensin-converting enzyme inhibitor use, and angiotensin II receptor blocker use were not related to bone loss and did not alter the relationship between bicarbonate and bone loss.

Table 3.

Coefficients and β (Standardized Regression) Coefficients for Variables in Multivariate Model Relating Arterialized Venous Bicarbonate Obtained at Year 3 to Percentage Change in BMD Over the 2.1 ± 0.3 (mean ± SD) Years Between Baseline (Year 1) and Year 3 Among the 2287 Members of the Health, Aging and Body Composition Cohort in Whom Arterialized Venous Blood Gas Measurements Were Obtained at the Year 3 Clinic Visit

	Coefficient	β Coefficient	SE	P Value
Arterialized venous bicarbonate, mmol/L	0.194	0.079	0.053	<.001
Race, Black/White	−0.729	−0.107	0.152	<.001
Gender, male/female	1.046	0.158	0.161	<.001
Clinic site, Memphis/Pittsburgh	0.160	0.024	0.139	.25
Age, y	−0.069	−0.059	0.025	.005
Weight, kg	0.032	0.142	0.005	<.001
Smoking status, yes/no	0.137	0.040	0.076	.07
eGFR (MDRD), mL/min/1.73 m2	0.003	0.013	0.005	.54

Abbreviation: MDRD, Modification of Diet in Renal Disease equation.

Across quartiles of plasma bicarbonate, the rate of change in bone density over the 2.1 years between year 1 and year 3 ranged from a loss of 0.72%/y in the lowest quartile to a gain of 0.15%/y in the highest quartile of bicarbonate (Figure 1).

Figure 1.

Annualized rate of change in BMD (percent per year) by quartile of arterialized venous bicarbonate adjusted for age, sex, race, clinic site, smoking, and eGFR.

Discussion

Our finding that arterialized plasma bicarbonate is associated positively with cross-sectional BMD and also negatively with the rate of bone loss in generally well-functioning men and women in their seventies supports the hypothesis that systemic acid-base status is an important determinant of skeletal health in the elderly. Our observation that ongoing bone loss was linearly related to arterialized bicarbonate, even after adjustments for age and renal function, suggests that the skeleton, at least in older individuals, functions as a base reservoir with potentially negative consequences for the skeleton. Further research and confirmation of these findings may have major public health implications because reducing dietary net acid load is possible through dietary intervention, by increasing dietary intake of fruit and vegetables while maintaining adequate protein intake, or through base supplementation with alkaline compounds such as potassium citrate or potassium bicarbonate.

Previous analyses have examined the relationship between the skeleton and acid-base metabolism using estimates of dietary net acid load from dietary intake data (1, 2, 38–41). Although dietary intake data provide useful estimates of dietary acid exposure (42, 43) and are practical for the large cohorts and long-term follow-up necessary for skeletal outcomes, direct physiological measurements of acid-base balance have been scarce. Although arterial blood provides the most direct measure of acid load, arterialized venous samples provide nearly identical values. Therefore, arterialized venous samples provide an accurate assessment of the acid load presented to the skeleton by the arterial system (35). This is the first study to compare in vivo arterialized venous plasma bicarbonate measurements to cross-sectional BMD and longitudinal changes in BMD.

The limitations of this study include the potential presence of unmeasured confounders, residual influence of confounders measured with misclassification, and uncertainties in the intra-individual variability of the arterialized bicarbonate measure. However, variability in the measurement, assuming it is nondifferential with respect to the outcome, would be most likely to attenuate any real association rather than introduce a spurious association. This was an observational study, and direct inferences about cause and effect cannot be made. Participants were recruited from just two centers, and only well-functioning older adults were recruited, which limits the generalizability of the observations. Additionally, it is possible that other causes of metabolic acidosis contributed to the plasma bicarbonate values among our participants in addition to their typical dietary acid load. We do not have data that would allow us to examine for aldosterone deficiency, pseudohypoaldosteronism type II, or other causes of metabolic acidosis. However, angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker use was not significantly related to BMD or bone loss in our models and did not significantly affect the coefficients or level of statistical significance for the relationship between bicarbonate and BMD or bone loss. In metabolic acidosis, increased serum chloride levels occur when the rate of production or excretion of other anions (that might otherwise manifest as an anion gap) is sufficiently slow or rapid, respectively. Thus, in otherwise healthy participants, when dietary acid load is the chief cause of metabolic acidosis, it is usually accompanied by increased blood levels of chloride. We do not have data on chloride levels in our study population.

We found that BMD was highest and bone loss was nonexistent in persons in the highest quartile of plasma bicarbonate concentration (mean [HCO₃], 26.8 ± 0.8 mmol/L) with no evidence of a threshold or plateau over the range of concentrations observed in our study. Bicarbonate concentrations in the highest quartile were well within the typical physiological range for adults. Based on published equations relating pCO₂-adjusted plasma bicarbonate concentration to dietary net acid load (11, 14), the latter might have ranged from as high as 150 mEq/d at the lowest quartile of plasma bicarbonate to slightly negative values (ie, a dietary net base load) at the highest quartile of plasma bicarbonate. These findings raise the possibility that even higher plasma bicarbonate concentrations, as would occur with habitual ingestion of net base-producing diets, could have even larger effects, potentially shifting the typical age-related bone mass decline to positive bone mass gain.

These data are supported by intervention results using intermediary markers and dietary manipulations. Previous short-term studies have demonstrated decreased urine calcium excretion and/or lower markers of bone resorption on diets supplemented with vegetables or alkaline potassium salts (3, 44–49). Longer-term data demonstrate sustained improvement in calcium balance over 6 months of supplementation with potassium citrate (50). In addition, supplementation with 60 mmol/d of potassium citrate was shown to improve bone density compared to placebo in older individuals with normal bone density and fairly low baseline net acid excretion (51).

The range of bicarbonate values observed in this cohort was relatively narrow, yet a significant association with both bone density and bone loss was evident. These results suggest that sufficient base supplementation to beneficially affect the skeleton could be achieved through dietary modifications or supplements of bicarbonate or bicarbonate precursors (eg, potassium citrate). Interventions to provide increased dietary base through dietary modification and/or specific base supplementation are inexpensive and have minimal risk. Sufficient dietary base precursors to neutralize the dietary acid generated by the consumption of an adequate daily protein intake could be achieved by concurrently consuming six or seven servings per day of potassium-rich fruit and vegetables, an intervention consistent with current recommended dietary guidelines. As such, this could be implemented on a population-wide basis and provide substantial protection against skeletal loss in the older population.

Our data provide direct evidence for the important role of acid-base status in skeletal health. We demonstrated that the degree of systemic acidity, as measured by arterialized blood gas samples, is associated with bone density and bone loss, further supporting a role for interventions that can alter acid-base status as a strategy to combat age-related declines in bone density.