Abstract
Background and objectives
Low urine ammonium excretion is associated with ESRD in CKD. Few laboratories measure urine ammonium, limiting clinical application. We determined correlations between urine ammonium, the standard urine anion gap, and a modified urine anion gap that includes sulfate and phosphate and compared risks of ESRD or death between these ammonium estimates and directly measured ammonium.
Design, setting, participants, & measurements
We measured ammonium, sodium, potassium, chloride, phosphate, and sulfate from baseline 24-hour urine collections in 1044 African-American Study of Kidney Disease and Hypertension participants. We evaluated the cross-sectional correlations between urine ammonium, the standard urine anion gap (sodium + potassium − chloride), and a modified urine anion gap that includes urine phosphate and sulfate in the calculation. Multivariable-adjusted Cox models determined the associations of the standard urine anion gap and the modified urine anion gap with the composite end point of death or ESRD; these results were compared with results using urine ammonium as the predictor of interest.
Results
The standard urine anion gap had a weak and direct correlation with urine ammonium (r=0.18), whereas the modified urine anion gap had a modest inverse relationship with urine ammonium (r=−0.58). Compared with the highest tertile of urine ammonium, those in the lowest urine ammonium tertile had higher risk of ESRD or death (hazard ratio, 1.46; 95% confidence interval, 1.13 to 1.87) after adjusting for demographics, GFR, proteinuria, and other confounders. In comparison, participants in the corresponding standard urine anion gap tertile did not have higher risk of ESRD or death (hazard ratio, 0.82; 95% confidence interval, 0.64 to 1.07), whereas the risk for those in the corresponding modified urine anion gap tertile (hazard ratio, 1.32; 95% confidence interval, 1.03 to 1.68) approximated that of directly measured urine ammonium.
Conclusions
Urine anion gap is a poor surrogate of urine ammonium in CKD unless phosphate and sulfate are included in the calculation. Because the modified urine anion gap merely estimates urine ammonium and requires five measurements, direct measurements of urine ammonium are preferable in CKD.
Keywords: chronic metabolic acidosis, chronic kidney disease, AASK (African American Study of Kidney Disease and Hypertension), urine anion gap, outcomes, ESRD, mortality, ammonium, Acid-Base Equilibrium, Ammonium Compounds, Body Fluids, Kidney Diseases
Introduction
Clinical assessment of kidney function focuses almost exclusively on markers of glomerular function and proteinuria (1). However, tubulointerstitial disease is observed commonly in kidney biopsy specimens, is strongly associated with eGFR decline across different kidney disease etiologies (2–4), and is poorly captured by clinical markers of kidney function. We have been interested in identifying noninvasive markers of kidney tubule dysfunction and injury that assist in predicting CKD progression above and beyond eGFR and proteinuria (5–9). A critical function of kidney tubules is regulation of acid-base homeostasis, which involves production and excretion of ammonium and bicarbonate generation. Decrements in ammonium excretion are observed before low serum bicarbonate develops, and lower urine ammonium strongly associated with ESRD or death independent of measured GFR, proteinuria, and serum bicarbonate in the African-American Study of Kidney Disease and Hypertension (AASK) participants (10). Similar findings were observed in the NephroTest Cohort (11) and the Chronic Renal Insufficiency Cohort participants with diabetes (12). This suggests that lower urine ammonium identifies patients with impaired tubule function and high risk for CKD progression.
Few clinical laboratories measure urine ammonium, which represents an important barrier to translating these findings into clinical practice. However, the urine anion gap (UAG) has been advocated as a urine ammonium surrogate (13). In clinical practice, the UAG (sodium + potassium − chloride) is used to gauge robustness of ammonium excretion during metabolic acidosis. In metabolic acidosis secondary to diarrhea, ammonium excretion is enhanced, thereby adding “unmeasured cations” and inducing a more negative UAG (14). With renal tubular acidosis and the metabolic acidosis of CKD, robust increases in ammonium excretion do not occur, resulting in a more positive UAG (14,15). However, the ability of the UAG to predict urine ammonium concentration and assess clinical risk in patients with CKD is questionable. For example, in 30 individuals with impaired kidney function, the standard UAG had no correlation with urine ammonium concentration. However, the correlation between urine ammonium and the UAG was improved when urine phosphate and sulfate were included in the calculation (16). If confirmed, this finding suggests that the quantity of “unmeasured anions” critically influences the ability of the UAG to estimate ammonium concentrations in CKD. However, the relationship between these UAG measurements and urine ammonium concentration has not been evaluated in a large CKD cohort, and whether they predict clinical outcomes similarly to urine ammonium has not been evaluated.
We designed this study to address two aims. First, we determined the correlation between urine ammonium and the standard UAG and compared it with modified UAG calculations that include sulfate and phosphate in 1044 AASK participants overall and in 128 participants with acidosis. Second, we determined the association of these UAG calculations with risk of ESRD or death among the AASK participants and compared the strengths of these associations with those of directly measured urine ammonium. Our purpose was to determine whether the UAG gives reliable estimates of ammonium and associated outcomes in patients with CKD.
Materials and Methods
Study Participants
The details of the AASK have been published (17,18). Briefly, blacks ages 18–70 years old with hypertensive CKD (defined as measured GFR =20–65 ml/min per 1.73 m2 by clearance of iodine 125I iothalamate and diastolic BP >95 mm Hg) were eligible for the study. Between April of 1995 and September of 1998, 1094 participants were randomized to ramipril, metoprolol, or amlodipine as well as to one of two BP goals (usual mean arterial pressure goal of 102–107 mm Hg or a low mean arterial pressure goal of ≤92 mm Hg). At the end of the randomized trial phase, participants who did not reach ESRD were eligible to enroll in the nonrandomized AASK Cohort Study. Baseline urine samples were available for 1057 participants. We excluded 13 participants with daily urine ammonium excretion normalized by body weight ≥0.7 mEq/kg out of concern for falsely elevated urine ammonium excretion due to bacterial overgrowth. Hence, 1044 participants were evaluated, corresponding to 99% of participants with available samples and 95% of randomized participants. The AASK was overseen by the institutional review boards of the participating sites and performed under the principles embodied in the Declaration of Helsinki.
Measurements
Using standardized forms, trained personnel obtained data on baseline demographic, clinical, and laboratory data. Urine analytes were measured from aliquots of the baseline 24-hour urine collection. Urine ammonium was measured by the glutamate dehydrogenase method. Urine sodium, potassium, and chloride were measured using ion-selective electrodes. Urine phosphorous was measured using the photometric molybdate technique. Urine sulfate was measured using barium precipitation. Daily excretion of these ions was determined by multiplying concentrations with 24-hour urine volumes. The 24-hour urine samples were confirmed to have been collected according to the AASK protocol and were necessary before randomization. Urine collections were preserved with acetic acid at the time of collection, rendering urine pH measurements inaccurate. Thus, we calculated the quantity of phosphate in milliequivalents using the Henderson–Hasselbalch equation with pKa=6.8 for HPO42−/H2PO4−, assuming urine pH of 5.5 (16). Four parameters were evaluated in regard to their correlations with urine ammonium: the UAG, the urine anion gap with inclusion of urine phosphate (UAGPO4), the urine anion gap with inclusion of urine sulfate (UAGSO4), and the urine anion gap with inclusion of urine phosphate and sulfate (UAGPLUS) (Table 1).
Table 1.
Terminology and calculations of the four urine anion gap measurements in this study
| Term | Calculation, mEq/d |
|---|---|
| UAG | Urine (sodium + potassium) − (chloride) |
| UAGSO4 | Urine (sodium + potassium) − (chloride + sulfate) |
| UAGPO4 | Urine (sodium + potassium) − (chloride + phosphate) |
| UAGPLUS | Urine (sodium + potassium) − (chloride + phosphate + sulfate) |
Because urine pH was not available, the milliequivalents of phosphate were calculated using urine pH of 5.5 in all participants. UAG, urine anion gap; UAGSO4, urine anion gap with inclusion of urine sulfate; UAGPO4, urine anion gap with inclusion of urine phosphate; UAGPLUS, urine anion gap with inclusion of urine phosphate and sulfate.
Serum bicarbonate was measured using either the kinetic ultraviolet method (Roche Hitachi 747; Roche, Indianapolis, IN) or a CO2 electrode (Beckman CX3 Delta; Beckman, Brea, CA). Urinary protein excretion was expressed as a protein-to-creatinine ratio from the 24-hour urine collection. Daily dietary protein intake (grams per day) was calculated from 24-hour urine urea nitrogen excretion using the equation 6.25× [urine urea nitrogen in grams per day + (weight in kilograms ×0.031)] (19). Net endogenous acid production was calculated using the formula −10.2+(54.5× protein intake in grams per day)/urine potassium in milliequivalents per day (20).
Statistical Analyses
Participants were categorized by tertiles of daily UAG excretion and daily urine ammonium. Continuous variables are presented as means with SD, unless otherwise specified. Categorical variables are presented as percentages. Significance tests were performed using ANOVA for continuous variables and chi-squared tests for dichotomous variables. Pearson correlation coefficients between urine ammonium and each UAG measurement were determined in the total cohort and the acidosis subgroup, which was defined as serum bicarbonate <22 mEq/L. Bland–Altman plots quantified the level of agreement of urine ammonium with the additive inverse of UAGPLUS in the total cohort and the acidosis subgroup. The additive inverse of UAGPLUS was used, because higher urine ammonium correlates with more negative UAGPLUS; it would have been inappropriate to compare differences between positive and negative values in the Bland–Altman plot.
The longitudinal outcome of interest was the composite of death or ESRD, which was adjudicated by the outcomes committee. Follow-up time was censored at the administrative end date, at permanent loss to follow-up, or if the participant did not enroll in the cohort phase. Cox regression models related the composite outcome of death or ESRD to daily urine ammonium, UAG, UAGPO4, UAGSO4, and UAGPLUS. In analyses using urine ammonium as the independent variable, the highest tertile was the reference. For the UAG-based measurements, the lowest (most negative) tertile was used as the reference, which should correspond with higher urine ammonium. Cox models were adjusted for age, sex, randomized group, measured GFR, proteinuria, net endogenous acid production, serum potassium, and serum bicarbonate to maintain consistency with our prior report and facilitate comparisons across the independent variables (10). Proportional hazards assumptions were evaluated using a formal significance test on the basis of the unscaled and scaled Schoenfeld residuals and a graphical assessment of log-log survival curves using urine ammonium as the independent variable. Body mass index (BMI) violated the proportional hazards assumption and was included as a stratification variable (<25, 25 to <30, and ≥30 kg/m2) in the models.
The analyses were performed using Stata 14 (College Station, TX).
Results
Baseline Characteristics
Characteristics of the 1044 study participants are presented in Table 2. The mean age was 54 years old, 62% were men, mean measured GFR was 47±14 ml/min per 1.73 m2, and median urine protein-to-creatinine ratio was 80 mg/g. One hundred twenty-eight participants (12%) had metabolic acidosis at baseline, and the mean urine ammonium was 21±12 mEq/d. The mean 24-hour urine phosphate excretion was 21±10 mEq/d, and mean 24-hour urine sulfate excretion was 29±16 mEq/d. Compared with those with higher UAG, participants with lower UAG (and presumably, higher ammonium) had lower BMI, were more likely to smoke, had lower protein intake, and had higher net endogenous acid production, and a greater proportion were women. They also had slightly lower measured GFR and lower bicarbonate. Contrary to our expectation, urine ammonium was higher in individuals with higher (more positive) UAG.
Table 2.
Baseline characteristics presented as mean (SD) or number (percentage) unless otherwise specified
| Characteristics and Measurements | Total Population, n=1044 | Daily UAG Tertiles | ||
|---|---|---|---|---|
| −80 to <29 mEq/d n=348 | 29 to <50 mEq/d n=348 | 50–228 mEq/d n=348 | ||
| Demographics | ||||
| Age, yr | 54 (11) | 54 (10) | 55 (11) | 53 (11) |
| Men, no. (%) | 645 (62) | 178 (51) | 202 (58) | 265 (76) |
| Clinical characteristics | ||||
| Body mass index, kg/m2 | 30.6 (6.6) | 28.8 (6.4) | 30.8 (6.6) | 32.2 (6.5) |
| Cardiovascular disease, no. (%) | 538 (52) | 186 (53) | 177 (51) | 175 (50) |
| Current smoker, no. (%) | 305 (29) | 134 (39) | 85 (24) | 86 (24) |
| Past smoker, no. (%) | 298 (29) | 83 (24) | 102 (29) | 113 (32) |
| Never smoker, no. (%) | 441 (42) | 131 (38) | 161 (46) | 149 (43) |
| Systolic BP, mmHg | 150 (24) | 150 (25) | 150 (24) | 151 (22) |
| Protein intake, g/d | 69 (25) | 51 (16) | 68 (18) | 88 (25) |
| Net endogenous acid production, mEq/d | 83 (37) | 96 (42) | 83 (34) | 69 (27) |
| ACE-i/ARB use, no. (%) | 397 (39) | 125 (37) | 135 (40) | 137 (40) |
| Diuretic use, no. (%) | 649 (64) | 206 (61) | 212 (62) | 231 (68) |
| Measured GFR, ml/min per 1.73 m2 | 47 (14) | 46 (14) | 46 (14) | 48 (13) |
| Measured GFR <30 ml/min per 1.73 m2, no. (%) | 163 (16) | 60 (17) | 62 (18) | 41 (12) |
| Serum measurements | ||||
| Bicarbonate, mEq/L | 25 (3) | 25 (3) | 25 (3) | 25 (3) |
| Bicarbonate <22 mEq/L, no. (%) | 128 (12) | 55 (16) | 35 (10) | 38 (11) |
| Potassium, mEq/L | 4.2 (0.6) | 4.2 (0.7) | 4.2 (0.6) | 4.2 (0.6) |
| Anion gap, mEq/L | 10 (2) | 10 (3) | 10 (2) | 10 (2) |
| Urine measurements | ||||
| Protein-to-creatinine ratio, mg/ga | 80 (29–342) | 77 (31–298) | 68 (27–332) | 100 (34–373) |
| Urea nitrogen, g/d | 8.0 (3.7) | 5.4 (2.2) | 8.0 (2.6) | 10.8 (3.7) |
| Sodium, mEq/d | 152 (81) | 111 (69) | 147 (67) | 197 (82) |
| Potassium, mEq/d | 43 (24) | 27 (16) | 40 (13) | 63 (25) |
| Chloride, mEq/d | 153 (82) | 120 (79) | 148 (70) | 191 (82) |
| Ammonium, mEq/d | 21 (12) | 19 (11) | 22 (11) | 24 (13) |
| Sulfate, mEq/d | 29 (16) | 18 (10) | 29 (12) | 40 (17) |
| Phosphate, mEq/d | 21 (10) | 14 (6) | 21 (7) | 29 (10) |
UAG, urine anion gap; ACE-i, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker.
Presented as median (interquartile range).
By comparison, higher urine ammonium was associated with men and higher measured GFR, bicarbonate, BMI, protein intake, and net endogenous acid production as previously reported (10). Thus, the baseline characteristics of the cohort were markedly different when UAG was used as an estimate of ammonium excretion compared with directly measured ammonium.
Correlations of Urine Ammonium with the UAG Measurements
Table 3 presents values of urine ammonium and the four UAG measurements in the study sample overall and across tertiles of urine ammonium. We expected that a more negative UAG would reflect higher urine ammonium, and therefore, an inverse correlation was hypothesized. By contrast, we observed a slightly positive correlation between UAG and urine ammonium. Whereas UAG was higher with higher urine ammonium, UAGSO4, UAGPO4, and UAGPLUS were lower with higher urine ammonium and thus, in the expected direction. Figure 1 shows that UAGPLUS had the strongest inverse correlation with urine ammonium in the entire study sample (r=−0.58). The correlations between the four UAG measurements and urine ammonium were similar in the subgroup of 128 individuals with acidosis. Among participants with acidosis, all but two had a positive UAG, and there was no correlation of the UAG with urine ammonium in the acidosis subgroup (r=0.07).
Table 3.
Values of urine ammonium and the four urine anion gap measurements in the total population and by tertiles of urine ammonium
| Measurement, mEq/d | Total Population, n=1044 | Urine Ammonium Tertile 1 (Range, 0–15 mEq/d), n=348 | Urine Ammonium Tertile 2 (Range, 15–24 mEq/d), n=348 | Urine Ammonium Tertile 3 (Range, 24–81 mEq/d), n=348 | P Value |
|---|---|---|---|---|---|
| Urine ammonium | 21 (12) | 10 (3) | 19 (3) | 35 (9) | — |
| UAG | 42 (25) | 37 (25) | 42 (24) | 47 (26) | <0.001 |
| UAGSO4 | 13 (19) | 18 (18) | 14 (19) | 6 (19) | <0.001 |
| UAGPO4 | 20 (20) | 22 (20) | 21 (20) | 18 (19) | 0.05 |
| UAGPLUS | −9 (19) | 3 (15) | −7 (16) | −22 (18) | <0.001 |
Values shown as mean (SD). —, not applicable; UAG, urine anion gap; UAGSO4, urine anion gap with inclusion of urine sulfate; UAGPO4, urine anion gap with inclusion of urine phosphate; UAGPLUS, urine anion gap with inclusion of urine phosphate and sulfate.
Figure 1.
Whereas UAG was higher with higher urine ammonium, UAGSO4, UAGPO4, and UAGPLUS were lower with higher urine ammonium. Pairwise correlation between measured urine ammonium (abscissa) and the estimates of urine ammonium (1) urine anion gap (UAG), (2) urine anion gap with inclusion of urine phosphate (UAGPO4), (3) urine anion gap with inclusion of urine sulfate (UAGSO4), and (4) urine anion gap with inclusion of urine phosphate and sulfate (UAGPLUS) in (A) the total cohort and (B) the subgroup with acidosis at baseline.
Bland–Altman plots (Figure 2) show the agreement between urine ammonium with the additive inverse of UAGPLUS in the entire cohort and the subgroup with acidosis. Urine ammonium was, on average, 13 mEq/d higher than UAGPLUS, indicating positive bias in the cohort. More importantly, the 95% limits of agreement were broad (−18–43 mEq/d). The results were similar in the subgroup of participants with acidosis.
Figure 2.
The limits of agreement between urine ammonium and the additive inverse of UAGPLUS were broad. Bland–Altman plots showing mean difference and limits of agreement between urine ammonium and the additive inverse of urine anion gap with inclusion of urine phosphate and sulfate (UAGPLUS) in (A) the total cohort and (B) the acidosis subgroup. In each case, there is positive bias. That is, mean urine ammonium is higher than the additive inverse of UAGPLUS. (A) Mean difference =13 mEq/d, limits of agreement =−18, 43 mEq/d. (B) Mean difference =11 mEq/d, limits of agreement =−18, 39 mEq/d.
Association of UAG Measurements with Clinical Outcomes
Next, we determined the associations of each UAG measurement with the composite outcome of ESRD or death. During 7862 patient-years of follow-up, there were 296 ESRD events and 168 deaths. After adjusting for demographics, measured GFR, proteinuria, and other risk factors, the lowest urine ammonium tertile was associated with 46% higher risk of ESRD or death compared with the highest tertile, as previously reported (10). We used this hazard ratio as an indicator of the reliability of the UAG measures to predict the same clinical outcomes. Although the reference group for the urine ammonium analysis was those in the highest urine ammonium tertile, the reference group for the UAG-based measures was the lowest tertile (most negative), which is anticipated to correspond with higher urine ammonium. In comparison with the hazard ratio for the lowest urine ammonium tertile, the highest UAG tertile was not associated with death or ESRD when adjusted for identical covariates. Similarly, neither the highest UAGSO4 tertile nor the highest UAGPO4 tertile were associated with ESRD or death. By contrast, the highest UAGPLUS tertile was associated with ESRD or death; an association that was comparable in strength with that observed for the lowest tertile of urine ammonium (Table 4).
Table 4.
Associations between urine ammonium and each urine anion gap measurement with the composite outcome of dialysis or death before dialysis
| Composite Outcome (Death or ESRD) | Urine Ammonium | UAG | UAGSO4 | UAGPO4 | UAGPLUS |
|---|---|---|---|---|---|
| Tertile corresponding with | |||||
| Low urine ammonium or high UAG | |||||
| Hazard ratio (95% CI) | 1.46 (1.13 to 1.87) | 0.82 (0.64 to 1.07) | 0.94 (0.72 to 1.21) | 0.95 (0.74 to 1.21) | 1.32 (1.03 to 1.68) |
| Mean (SD) urine ammonium, mEq/d | 10.2 (3.3) | 23.9 (12.9) | 18.8 (10.4) | 20.4 (12.4) | 14.9 (9.0) |
| No. of events | 188 | 151 | 63 | 153 | 174 |
| Intermediate urine ammonium or UAG | |||||
| Hazard ratio (95% CI) | 1.14 (0.89 to 1.46) | 0.82 (0.65 to 1.03) | 0.93 (0.73 to 1.18) | 0.82 (0.65 to 1.04) | 1.23 (0.97 to 1.56) |
| Mean (SD) urine ammonium, mEq/d | 19.5 (2.6) | 21.6 (10.9) | 19.6 (10.3) | 21.3 (11.4) | 18.8 (8.3) |
| No. of events | 146 | 150 | 163 | 156 | 159 |
| High urine ammonium or low UAG | |||||
| Hazard ratio (95% CI) | Reference | Reference | Reference | Reference | Reference |
| Mean (SD) urine ammonium, mEq/d | 34.6 (9.3) | 18.7 (10.5) | 25.8 (12.8) | 22.5 (12.4) | 30.4 (11.3) |
| No. of events | 130 | 163 | 138 | 155 | 131 |
Models were adjusted for age, sex, randomized group, measured GFR, proteinuria, net endogenous acid production, serum potassium, and serum bicarbonate and stratified by body mass index. Also shown are the mean (SD) values of urine ammonium within each tertile of UAG and the numbers of events within each tertile (number at risk =348 in each). The concentrations of urine ammonium are in the opposite direction expected in the UAG group. Inclusion of sulfate (UAGSO4) and phosphate (UAGPO4) yielded the expected pattern of higher urine ammonium concentration with lower (more negative) UAG. UAGPLUS showed the best separation of urine ammonium in the expected direction across the tertiles. Only UAGPLUS recapitulated the hazard ratios of the composite outcome observed with urine ammonium. UAG, urine anion gap; UAGSO4, urine anion gap with inclusion of urine sulfate; UAGPO4, urine anion gap with inclusion of urine phosphate; UAGPLUS, urine anion gap with inclusion of urine phosphate and sulfate; 95% CI, 95% confidence interval.
Discussion
Lower urine ammonium excretion is a risk factor for poor outcomes in CKD (10–12); however, urine ammonium is uncommonly measured by clinical laboratories. We evaluated whether UAG reasonably estimates urine ammonium and recapitulates the association between low urine ammonium and clinical outcomes in the AASK. UAG had a poor and direct correlation with urine ammonium excretion and hence, was not associated with clinical outcomes. The correlation between UAG and urine ammonium was substantially improved and in the expected inverse direction when urine phosphate and sulfate were included in the UAG calculation as UAGPLUS. Furthermore, UAGPLUS approximated the risk of ESRD or death observed when urine ammonium was the independent variable of interest. Thus, UAG is a poor surrogate for urine ammonium in CKD, unless sulfate and phosphate are included in the calculation.
In their seminal paper, Goldstein et al. (13) showed a strong inverse correlation (r=−0.97) between the standard UAG and urine ammonium in patients with metabolic acidosis and normal GFR. Batlle et al. (14) also reported a strong inverse correlation between UAG and urine ammonium in 53 patients with distal renal tubular acidosis, diarrhea, or healthy controls receiving ammonium chloride infusions. Clinically, the UAG is used to gauge whether robust urinary ammonium excretion is present in the setting of metabolic acidosis, signaled by a negative UAG. In CKD, ammonium excretion is impaired in acidosis. This change, and potentially other changes, in unmeasured urine anions and cations rendered the UAG positive in the majority of the AASK participants with acidosis. In terms of estimating ammonium more precisely in patients with CKD, our results confirm and meaningfully extend those of Kirschbaum et al. (16). Specifically, they found that the poor correlation between UAG and urine ammonium was markedly improved when urine phosphate and sulfate were accounted for in individuals with CKD. As in this study, Kirschbaum et al. (16) included individuals with and without acidosis. Although they did not report whether these estimates of ammonium excretion vary by acidosis status in CKD, we show that the correlations between the four UAGs assessed in this study and ammonium are similar irrespective of acidosis status. Because UAG had poor correlation with urine ammonium, it was not surprising that there was no association of UAG with clinical outcomes. However, we show that inclusion of urine phosphate and sulfate in the UAG calculation (UAGPLUS) recapitulated these associations with outcomes similar to but weaker than those observed with directly measured urine ammonium.
Goldstein et al. (13) recognized that the presence of anions other than chloride, such as phosphate and sulfate, would influence the UAG, and in this study, sulfate (mean =29 mEq/d) and phosphate (mean =21 mEq/d) accounted for about 25% of the urine anions. Their inclusion in UAGPLUS substantially improved the correlation with urine ammonium and rendered it in the expected inverse direction. Differences of urine phosphate excretion occur in persons with and without CKD (21,22), and dietary factors, GFR, and use of intestinal phosphate binders in CKD are factors that account for interindividual urine phosphate variability (23). Across-individual differences in urinary sulfate, largely determined by animal protein intake, have been shown in prior studies also (24,25). Thus, individual variability of urine phosphate and sulfate materially influences the relationship of the UAG with urine ammonium and if not accounted for, renders UAG a poor surrogate of urine ammonium in patients with CKD.
Given the poor performance of the UAG, we conclude that it should not be used as a surrogate for ammonium when estimating CKD progression risk. Although our results show that accounting for urine phosphate and sulfate better approximates urinary ammonium in CKD, the UAGPLUS calculation requires measurement of five variables, each increasing the likelihood of measurement error. Furthermore, the level of agreement between daily urine ammonium excretion and UAGPLUS was biased and broad as shown in Figure 2. In large-scale studies, such as the AASK, UAGPLUS may be useful to rank order individuals to evaluate its association with clinical end points. However, the use of UAGPLUS does not seem to be feasible in clinical practice. Instead, we believe that the time has come to directly quantify urine ammonium, which clinical laboratories could perform using diluted urine samples and the standard plasma ammonia assay (26).
A limitation of this study is that urine samples in the AASK were preserved with acetic acid at the time of collection; thus, we could not measure urine pH. We assumed urine pH of 5.5 in study participants, consistent with mean levels reported in patients with CKD (16), which corresponds to 5% of phosphate as HPO42− and 95% as H2PO4−. The mean milliequivalents of phosphate and its distribution in the population would not be significantly different with lower urine pH. However, higher urine pH would increase the milliequivalents of phosphate in the HPO42− state, generate a more negative UAGPLUS, and strengthen the inverse correlation with urine ammonium. In sensitivity analyses, the Cox model results were similar, irrespective of whether a urine pH of 5.5, 6.0, or 6.8 was imputed for our study sample. Nevertheless, urine pH measurements could have narrowed the mean difference and levels of agreement between urine ammonium and UAGPLUS. Hence, the agreement between UAGPLUS and urine ammonium may have been underestimated. Because urine phosphate is not included in the UAG, this would not influence the poor correlation between UAG and urine ammonium. We did not quantify urine bicarbonate, which could influence the UAG. In general, bicarbonaturia is <10 mEq/L if the urine pH is below 6.8, and urine pH above this is uncommon in CKD (16,27). Our findings suggest that excluding urine sulfate and phosphate from the UAG calculation is more important than excluding urine bicarbonate. The urine osmolal gap is an alternative urine ammonium estimate that has previously been reported to have stronger correlation with ammonium than the UAG (16). Urine osmolality was not measured in the AASK; hence, we could not compare the performance of the urine osmolal gap with that of UAG, in particular UAGPLUS, in this cohort; future studies are required to evaluate this approach. Systemic pH and pCO2 were not measured in the AASK; hence, some with low bicarbonate may have had chronic respiratory alkalosis, which also manifests with a positive UAG, rather than metabolic acidosis. Nevertheless, we suspect that the majority of these patients with CKD and low bicarbonate had metabolic acidosis. Strengths of this study include the well characterized, large study sample with carefully collected data. Furthermore, over 95% of randomized participants were included in these analyses. GFR was directly measured by 125I iothalamate clearance, which is important considering the strong relationship between kidney function, urine ammonium excretion, and long-term outcomes.
In conclusion, urine phosphate and sulfate contribute significant anionic milliequivalents to the urine in CKD and are variable from one individual to the next. These anions are not included in the typical UAG calculation and render it a poor surrogate for urine ammonium concentration in CKD. Although lower urine ammonium excretion was independently associated with risk of ESRD or death, there was no association of UAG with these end points. Including urine sulfate and phosphate in the UAG improves the correlation with urine ammonium and approximates the association of urine ammonium with clinical outcomes. However, a modified UAG that accounts for phosphate and sulfate still leads to significant bias and broad levels of agreement, and it requires measurement of five variables, which increases the risk of measurement error. Direct measurements of urine ammonium by clinical laboratories will help clinicians better assess kidney tubule function as it relates to acid-base regulation and CKD progression risk.
Disclosures
None.
Acknowledgments
Financial support for this study was provided by the Robert Wood Johnson Foundation Harold Amos Medical Faculty Development award (to K.L.R.). K.L.R. also receives support from Career Development award IK2 CX000537 from the US Department of Veterans Affairs Clinical Sciences Research and Development Service and National Institutes of Diabetes and Digestive and Kidney Diseases grant U01DK099933. J.H.I. received support from grants R01DK098234 and K24DK11042 from the National Institutes of Diabetes and Digestive and Kidney Diseases.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
See related editorial, “The Urine Anion Gap in Context,” on pages 195–197.
References
- 1.Levey AS, de Jong PE, Coresh J, El Nahas M, Astor BC, Matsushita K, Gansevoort RT, Kasiske BL, Eckardt KU: The definition, classification, and prognosis of chronic kidney disease: A KDIGO controversies conference report. Kidney Int 80: 17–28, 2011 [DOI] [PubMed] [Google Scholar]
- 2.Howie AJ, Ferreira MA, Adu D: Prognostic value of simple measurement of chronic damage in renal biopsy specimens. Nephrol Dial Transplant 16: 1163–1169, 2001 [DOI] [PubMed] [Google Scholar]
- 3.Nath KA: Tubulointerstitial changes as a major determinant in the progression of renal damage. Am J Kidney Dis 20: 1–17, 1992 [DOI] [PubMed] [Google Scholar]
- 4.Takebayashi S, Kiyoshi Y, Hisano S, Uesugi N, Sasatomi Y, Meng J, Sakata N: Benign nephrosclerosis: Incidence, morphology and prognosis. Clin Nephrol 55: 349–356, 2001 [PubMed] [Google Scholar]
- 5.Goldenstein L, Driver TH, Fried LF, Rifkin DE, Patel KV, Yenchek RH, Harris TB, Kritchevsky SB, Newman AB, Sarnak MJ, Shlipak MG, Ix JH; Health ABC Study Investigators : Serum bicarbonate concentrations and kidney disease progression in community-living elders: The Health, Aging, and Body Composition (Health ABC) Study. Am J Kidney Dis 64: 542–549, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Ix JH, Biggs ML, Mukamal K, Djousse L, Siscovick D, Tracy R, Katz R, Delaney JA, Chaves P, Rifkin DE, Hughes-Austin JM, Garimella PS, Sarnak MJ, Shlipak MG, Kizer JR: Urine collagen fragments and CKD progression-The Cardiovascular Health Study. J Am Soc Nephrol 26: 2494–2503, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ix JH, Katz R, Bansal N, Foster M, Weiner DE, Tracy R, Jotwani V, Hughes-Austin J, McKay D, Gabbai F, Hsu CY, Bostom A, Levey AS, Shlipak MG: Urine fibrosis markers and risk of allograft failure in kidney transplant recipients: A Case-Cohort Ancillary Study of the FAVORIT Trial. Am J Kidney Dis 69: 410–419, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Raphael KL, Wei G, Baird BC, Greene T, Beddhu S: Higher serum bicarbonate levels within the normal range are associated with better survival and renal outcomes in African Americans. Kidney Int 79: 356–362, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Raphael KL, Murphy RA, Shlipak MG, Satterfield S, Huston HK, Sebastian A, Sellmeyer DE, Patel KV, Newman AB, Sarnak MJ, Ix JH, Fried LF; Health ABC Study : Bicarbonate concentration, acid-base status, and mortality in the health, aging, and body composition study. Clin J Am Soc Nephrol 11: 308–316, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Raphael KL, Carroll DJ, Murray J, Greene T, Beddhu S: Urine ammonium predicts clinical outcomes in hypertensive kidney disease. J Am Soc Nephrol 28: 2483–2490, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Vallet M, Metzger M, Haymann JP, Flamant M, Gauci C, Thervet E, Boffa JJ, Vrtovsnik F, Froissart M, Stengel B, Houillier P; NephroTest Cohort Study group : Urinary ammonia and long-term outcomes in chronic kidney disease. Kidney Int 88: 137–145, 2015 [DOI] [PubMed] [Google Scholar]
- 12.Scialla JJ, Asplin J, Dobre M, Chang AR, Lash J, Hsu CY, Kallem RR, Hamm LL, Feldman HI, Chen J, Appel LJ, Anderson CA, Wolf M; Chronic Renal Insufficiency Cohort Study Investigators : Higher net acid excretion is associated with a lower risk of kidney disease progression in patients with diabetes. Kidney Int 91: 204–215, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Goldstein MB, Bear R, Richardson RM, Marsden PA, Halperin ML: The urine anion gap: A clinically useful index of ammonium excretion. Am J Med Sci 292: 198–202, 1986 [DOI] [PubMed] [Google Scholar]
- 14.Batlle DC, Hizon M, Cohen E, Gutterman C, Gupta R: The use of the urinary anion gap in the diagnosis of hyperchloremic metabolic acidosis. N Engl J Med 318: 594–599, 1988 [DOI] [PubMed] [Google Scholar]
- 15.Batlle DC, Chen S, Haque SK: Physiological principles in the clinical evaluation of electrolyte, water, and acid-base disorders. In: Seldin and Giebisch’s The Kidney: Physiology & Pathophysiology, 5th Ed., edited by Alpern RJ, Caplan MJ, Moe OW, London, Academic Press, 2013, pp 2477–2511 [Google Scholar]
- 16.Kirschbaum B, Sica D, Anderson FP: Urine electrolytes and the urine anion and osmolar gaps. J Lab Clin Med 133: 597–604, 1999 [DOI] [PubMed] [Google Scholar]
- 17.Appel LJ, Middleton J, Miller ER 3rd, Lipkowitz M, Norris K, Agodoa LY, Bakris G, Douglas JG, Charleston J, Gassman J, Greene T, Jamerson K, Kusek JW, Lewis JA, Phillips RA, Rostand SG, Wright JT: The rationale and design of the AASK cohort study. J Am Soc Nephrol 14[Suppl 2]: S166–S172, 2003 [DOI] [PubMed] [Google Scholar]
- 18.Wright JT Jr., Bakris G, Greene T, Agodoa LY, Appel LJ, Charleston J, Cheek D, Douglas-Baltimore JG, Gassman J, Glassock R, Hebert L, Jamerson K, Lewis J, Phillips RA, Toto RD, Middleton JP, Rostand SG; African American Study of Kidney Disease and Hypertension Study Group : Effect of blood pressure lowering and antihypertensive drug class on progression of hypertensive kidney disease: Results from the AASK trial. JAMA 288: 2421–2431, 2002 [DOI] [PubMed] [Google Scholar]
- 19.Maroni BJ, Steinman TI, Mitch WE: A method for estimating nitrogen intake of patients with chronic renal failure. Kidney Int 27: 58–65, 1985 [DOI] [PubMed] [Google Scholar]
- 20.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 68: 576–583, 1998 [DOI] [PubMed] [Google Scholar]
- 21.Palomino HL, Rifkin DE, Anderson C, Criqui MH, Whooley MA, Ix JH: 24-hour urine phosphorus excretion and mortality and cardiovascular events. Clin J Am Soc Nephrol 8: 1202–1210, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Selamet U, Tighiouart H, Sarnak MJ, Beck G, Levey AS, Block G, Ix JH: Relationship of dietary phosphate intake with risk of end-stage renal disease and mortality in chronic kidney disease stages 3-5: The Modification of Diet in Renal Disease Study. Kidney Int 89: 176–184, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hill KM, Martin BR, Wastney ME, McCabe GP, Moe SM, Weaver CM, Peacock M: Oral calcium carbonate affects calcium but not phosphorus balance in stage 3-4 chronic kidney disease. Kidney Int 83: 959–966, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.van den Berg E, Pasch A, Westendorp WH, Navis G, Brink EJ, Gans RO, van Goor H, Bakker SJ: Urinary sulfur metabolites associate with a favorable cardiovascular risk profile and survival benefit in renal transplant recipients. J Am Soc Nephrol 25: 1303–1312, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.van den Born JC, Frenay AR, Bakker SJ, Pasch A, Hillebrands JL, Lambers Heerspink HJ, van Goor H: High urinary sulfate concentration is associated with reduced risk of renal disease progression in type 2 diabetes. Nitric Oxide 55-56: 18–24, 2016 [DOI] [PubMed] [Google Scholar]
- 26.Ha LY, Chiu WW, Davidson JS: Direct urine ammonium measurement: Time to discard urine anion and osmolar gaps. Ann Clin Biochem 49: 606–608, 2012 [DOI] [PubMed] [Google Scholar]
- 27.Halperin ML, Richardson RM, Bear RA, Magner PO, Kamel K, Ethier J: Urine ammonium: The key to the diagnosis of distal renal tubular acidosis. Nephron 50: 1–4, 1988 [DOI] [PubMed] [Google Scholar]