The Urine Anion Gap: Common Misconceptions

Abstract

Two papers, one in 1986 and another one in 1988, reported a strong inverse correlation between urinary anion gap (UAG) and urine ammonia excretion (UNH₄) in patients with metabolic acidosis and postulated that UAG could be used as an indirect measure of UNH₄. This postulation has persisted until now and is widely accepted. In this review, we discuss factors regulating UAG and examine published evidence to uncover errors in the postulate and the design of the original studies. The essential fact is that, in the steady state, UAG reflects intake of Na, K, and Cl. Discrepancy between intake and urinary output of these electrolytes (i.e., UAG) indicates selective extrarenal loss of these electrolytes or nonsteady state. UNH₄ excretion, which depends, in the absence of renal dysfunction, mainly on the daily acid load, has no consistent relationship to UAG either theoretically or in reality. Any correlation between UAG and UNH₄, when observed, was a fortuitous correlation and cannot be extrapolated to other situations. Furthermore, the normal value of UAG has greatly increased over the past few decades, mainly due to increases in dietary intake of potassium and widespread use of sodium salts with anions other than chloride as food additives. The higher normal values of UAG must be taken into consideration in interpreting UAG.

In a seminal paper in 1986, Goldstein et al. ¹ showed a strong inverse correlation between urinary anion gap (UAG) and urine ammonia excretion (UNH₄) in patients with metabolic acidosis and normal renal function, and they suggested that UAG could be used as an indirect measure of UNH₄. Two years later, in 1988, Batlle et al. ² confirmed the observation of Goldstein et al. ¹ and also reported a strong inverse correlation between UAG and UNH₄ in patients with distal renal tubular acidosis, diarrhea, and healthy controls given oral ammonium chloride. These two papers cemented the idea of the validity of UAG as a surrogate marker of UNH₄, a concept now taught in renal conferences and textbooks as an accepted dogma without any attempt to analyze the validity of the original arguments.^3,4 Although some arguments have been raised against the utility of UAG as a marker of UNH₄, no systematic quantitative analysis has been made.^5–7 Herein, we wish to present the argument, with precise quantitative analysis, that the original data that led to the concept of UAG were inappropriately interpreted, resulting in faulty conclusions. Furthermore, we will present evidence that the numerical normal value of UAG has greatly increased over the past few decades, which would further affect its clinical interpretation.

What Is a Normal UAG, and What Factors Determine It?

By convention, the UAG is calculated from the formula UAG = (UNa + UK) – UCl. On the basis of the well-known principle that electrical balance must be maintained between total cations and total anions, Oh and Carroll⁷ developed the following formula: urine Na + urine K + urine unmeasured cations (UUCs) = urine Cl + urine unmeasured anions (UUAs), where UUCs are all cations of urine other than Na and K and UUAs are all anions of urine other than Cl. Rearranging the above terms, UAG = urine Na + urine K – urine Cl = UUA – UUC.

It has been shown that daily urine excretion rates of the components of urine anion gap, Na, K, and Cl, are nearly equal to their dietary intake.⁸ This can be attributed to the fact that these three electrolytes are nearly completely absorbed from the gastrointestinal tract and that they are neither metabolized nor produced in the body. Although the long-term intake and long-term urinary excretion rates of Na and Cl are equal, there is some daily discrepancy between Na intake and Na excretion; however, variations are matched by nearly identical variations in Cl excretion, and hence, they would have no effect on the estimation of UAG.⁷ The amount of K excreted in the urine is about 10 mEq/d less than the amount ingested because about 10 mEq/d is excreted normally in the stool.⁸ Some evidence exists that Na may be selectively stored in the extracellular matrix balanced by sulfate anions in chondroitin sulfate,⁹ but this has to be a slow process, simply because the daily availability of sulfate for production of chondroitin sulfate is rather limited.⁹ Therefore, in the steady state, UAG must reflect the dietary intake of Na, K, and Cl, corrected for the minor fecal loss of K.⁷

Discrepancy between intake and urinary excretion occurs in one of two conditions: (1) selective loss of any one of the three components of the UAG most commonly through the gastrointestinal tract, such as vomiting and diarrhea, or (2) temporary delay in urinary excretion of a component of UAG. Examples of the latter include the following. (1) During treatment of K depletion with KCl, Cl would be excreted in the urine as K is retained in the cell in exchange for H, which would in turn stimulate renal ammonia production. During K depletion, the reverse would occur in the UAG. (2) During acute respiratory alkalosis, Na would be excreted with HCO₃ while Cl is being retained, causing increased UAG, whereas during the recovery phase of respiratory alkalosis, the reverse would occur as HCO₃ is generated with excretion of NH₄Cl and UAG decreases. (3) During acute respiratory acidosis, UAG would decrease as urinary H excretion, accompanied by Cl, increases, and the reverse would occur during the recovery phase of respiratory acidosis. (4) Administration of lithium chloride would not be matched by a similar reduction in UAG initially as some of the lithium is retained in the cell in exchange for Na, and part of the administered chloride is excreted with Na. After a steady state is achieved, continuous administration of lithium chloride would reduce UAG, exactly matching the amount of excess Cl administered.

Above principles will apply to all subjects, including those with impaired kidney function and those with renal tubular acidosis, as long as they are in a steady state. It has been suggested that stool loss of K may be increased in CKD as a compensation mechanism for the reduced renal excretion of potassium,^10,11 but balance studies do not show significantly higher stool K excretion in patients with CKD compared with healthy controls.¹² Any discrepancy in UAG and dietary intake in patients with distal renal tubular acidosis (RTA) must be temporary and transient, depending on the phase of the patient’s illness at the time when the data for UAG were obtained.¹³ A negative K balance occurs commonly in these patients, and if the data were collected while the patient was developing a negative K balance, UAG will be greater than predicted from the dietary intake of Na, K, and Cl, as some of K lost in urine comes from the cell as protons enter the cell. If the data were obtained while the patient is recovering from K depletion, the reverse would be the case.

Important considerations in the estimation of the UAG are the circadian and day-to-day variations in urinary excretion as well as the influence of the urine pH in the ionization of anions, such as phosphate or the presence of bicarbonate. Urine pH in humans can vary between four and seven. When urine pH is seven, the concentration of bicarbonate would be about 10 mEq/L (assuming pCO₂ 40 mm Hg), which will increase UAG by the same magnitude replacing chloride. A change of urinary pH from four to about seven could increase charges of phosphate by as much as 17.5 mEq/L for a person excreting 30 mmol of phosphate in 24 hours.

The composition of diet has a large influence on the numerical value of the UAG and also in establishing the relationship between UAG and UNH₄. In meat eaters, increased dietary intake of K together with phosphate and sulfur-containing amino acids, with the latter metabolized to sulfuric acid, would cause increased UAG accompanied by increased UNH₄. The same amount of ingested K accompanied by organic anions, such as citrate, would also produce a higher UAG; in this case, however, metabolism of organic anions into bicarbonate would lead to reduced UNH₄, resulting in a negative correlation between UAG and UNH_4.

Table 1 describes chronologically a series of published studies on UAG in normal subjects, and the last column shows a significant increase of its value over the years. The original data by Goldstein et al. ¹ showed UAG of about 41 mEq/d, and now, the value is >70 mEq/d. These changes in UAG over time can be attributed to two changes: increases in urine K excretion and change in urine Na-Cl ratio from a value <1.0 to a value >1.0.^{1–7,14–18} These changes must reflect changes in dietary intake of these three electrolytes. The quantity of K intake has the greatest effect on UAG because K in food is mostly accompanied by an anion other Cl. K ingested as KCl is usually as a medication (K replacement) and occasionally, as a salt substitute or food additive. People are clearly eating more K nowadays, perhaps because of the public awareness of benefits of high-K intake and increasing popularity of plant-based diets.¹⁹ Changes in Na-Cl ratio cannot be due to changes in intake of salt (sodium chloride), the main source of Na and Cl in urine, because both are contained in equal amounts. The higher UAG is in part explained by increasing dietary intake of Na accompanied by anions other than Cl, mainly in the form of food additives, such as sodium bicarbonate, sodium phosphate, sodium benzoate, and sodium nitrate.²⁰ In contrast, all natural cereals, such as rice, wheat, barley, and rye, contain more chloride than sodium.¹⁴

Table 1.

Data on UAG in published studies on healthy subjects

Study (Reference)	Population	UNa, mEq/d	UK, mEq/d	UCl, mEq/d	UAG, mEq/d
Goldstein et al. 1	United States; normal subjects	NA	NA	NA	41, a 58
Documenta Geigy Scientific Tables 14	Normal men	177	57	184	50 b
	Normal women	128	47	132	43
Luft et al. 15	United States; normal subjects	176.7	60.3	183.7	53.3
Inase et al. 5	Japan; 30 normal subjects	NA	NA	NA	Estimated at 55 c
Cogswell et al. 16	United States; 827 adults from NHANES 2014 study	156.9	55.3	141.6	70.6
Castle et al. 17	United States; 777 adults for metabolic stone risk factors done twice	173.2, 167.5	67.3, 66.7	167.8, 162.7	72.7, 71.5
van der Leeuw et al. 18 (PREVEND study)	The Netherlands; 5673 adults from general population	145	69	135	79

NA, not available in the published paper; NHANES, National Health and Nutrition Examination Survey; PREVEND, Prevention of Renal and Vascular Endstage Disease.

^aThe value of 41 mEq/d was taken from a 1966 reference.

^bValues cited here for both men and women were from older references (1959 and 1961).

^cThis value of 55 mEq/d was estimated from visual inspection of figure 1 in ref. 5.

What Further Research on the Relationship between UAG and UNH₄ Has Been Done after the Original Papers by Goldstein et al. ¹ and Batlle et al.?²

After the original papers by Goldstein et al. ¹ and Batlle et al.,² there have been four published studies looking at the relationship between UAG and UNH₄ (Table 2). Two of these studies have been in healthy subjects, and two have been in patients with CKD. In contrast to the findings of Goldstein et al. ¹ and Batlle et al.,² two of the studies showed no association, and two showed a positive association between UAG and UNH_4. The studies in Table 2 also demonstrate a very wide range of UAG values in healthy subjects. In patients with CKD, correlation between UAG and UNH₄ was either nonexistent or positive.

Table 2.

Studies after the studies by Goldstein et al. ¹ and Batlle et al. ² describing the association between UAG and UNH₄

Study (Reference)	Population	UNa	UK	UCl	UAG	UNH4	Comment
6	Patients with CKD, Virginia, United States	81±32 mEq/L	30±21 mEq/L	102±47 mEq/L	8±29 (−59 to +62) mEq/L	10.8±9.4 mEq/L	No correlation between UAG and UNH4
21	47 urine samples from 32 normal subjects, Canada	NA	NA	NA	NA	NA	Positive correlation between UAG and UNH4
5	30 normal subjects, Japan	NA	NA	NA	Estimated at 55 mEq/d from figure	Estimated at 35 mEq/d from figure	No correlation between UAG and UNH4
22	1044 United States participants from AASK Study (all Blacks with hypertension and CKD)	152±81 mEq/d	43±24 mEq/d	153±82 mEq/d	42 (−80 to +228) mEq/d	21±12 mEq/d	Positive correlation between UAG and UNH4

NA, not available in the published paper; AASK, African American Study of Kidney Disease and Hypertension.

What Was the Source of Misinterpretation?

The initial publications about the UAG mentioned above^1,2 were an attempt to find ways to differentiate between gastrointestinal losses of bicarbonate versus renal tubular acidosis as causes of hyperchloremic metabolic acidosis. It was understood that in extrarenal acidosis, UNH₄ would be increased, whereas it would remain subnormal in patients with acidosis of renal origin. Therefore, the measurement of UNH₄ in a clinical scenario of hyperchloremic metabolic acidosis was thought be a good way to differentiate extrarenal from renal acidosis. Realizing that clinical laboratory usually did not measure UNH₄, these authors proposed to use UAG as a surrogate measure of UNH4. In the study by Batlle et al.,² eight patients with diarrhea-induced acidosis, except one, had a negative UAG, whereas all of those with RTA had a positive UAG. They also administered NH₄Cl orally to induce metabolic acidosis in healthy volunteers. The authors attributed the reduction in UAG to the increase in UNH₄. The truth is that reduced UAG in this situation is due to intake of Cl without Na or K. In fact, Cl administration with any cation other than Na or K (for example, lithium chloride or choline chloride) would have reduced UAG. Furthermore, increased UNH₄ stimulated by any acid that does not contain Cl would not have reduced UAG. For example, healthy subjects given an oral load of methionine, a sulfur-containing amino acid, showed increased UNH₄ excretion without major changes in UAG.²³ Similarly, patients with ketoacidosis²⁴ and respiratory acidosis²⁵ have stimulated renal ammonia production because of the acidosis, resulting in increased UNH₄ but without reduction in UAG.

In renal tubular acidosis in general, no abnormality in UAG would be expected in the absence of dietary modification. The fact is that, however, patients with renal tubular acidosis reported in the literature tend to have a reduced UAG.^2,26 Thus, UAG of 11 patients with classic distal RTA reported by Batlle et al. ² was only 26 mEq/d, which is much lower than normal values by either the old standards or the new standards, despite reduced UNH₄ (Table 1). If UAG and UNH₄ were inversely correlated as predicted by Goldstein et al. ¹ and Batlle, et al.,² one would have expected a higher than normal UAG in these patients. The most likely explanation is that these patients were sick, eating poorly, and in negative K balance prior to their hospitalization, and during their evaluation in the hospital, the patients were probably treated with KCl but had not yet achieved a steady state. As already explained above, during this period, K is retained intracellularly in exchange for protons, which would lead to excretion of NH₄ Cl in the urine, resulting in reduction in UAG. Unless the diet is altered permanently, this would cause only a temporary change in UAG.

The only determinant of the UAG is dietary content of three electrolytes, namely Na, K, and Cl, because urinary excretions of these three electrolytes are nearly equal to their intake. Deviation of UAG from prediction of dietary intake indicates selective extrarenal loss, such as diarrhea or vomiting, or an unsteady state, which is always transient. Usefulness of UAG in estimating the magnitude of extrarenal loss of these electrolytes in diarrhea and vomiting is reduced by the wide variations in normal UAG values in the healthy population due mainly to wide variation in dietary intake of K, which is mostly ingested with anions other than Cl.

If one uses the UAG clinically, one must be aware of the fact that the normal value of UAG has sharply increased in recent years because of increase in dietary intake of K and increased use of food additives containing Na salts unaccompanied by Cl. Our analysis supports the opinion that the usefulness of UAG as a surrogate marker of UNH₄ has not been supported by data or by theory. Urine osmolal gap has been suggested as an alternative better way of estimating UNH₄, but it also has many limitations and requires more laboratory measurements without much improvement in accuracy. Let us stop guessing UNH₄ and measure it directly when indicated because most clinical laboratories have the technical capacity to measure urine ammonia directly.

Disclosures

J. Uribarri reports receiving consultation honoraria from Baxter. The remaining author has nothing to disclose.

Funding

None.