Journal of Applied Physiology

Hypoproteinemia, strong-ion difference, and acid-base status in critically ill patients

Peter Wilkes


The present study was a prospective, nonrandomized, observational examination of the relationship among hypoproteinemia and electrolyte and acid-base status in a critical care population of patients. A total of 219 arterial blood samples reviewed from 91 patients was analyzed for arterial blood gas, electrolytes, lactate, and total protein. Plasma strong-ion difference ([SID]) was calculated from [Na+] + [K+] − [Cl] − [La]. Total protein concentration was used to derive the total concentration of weak acid ([A]tot). [A]tot encompassed a range of 18.7 to 9.0 meq/l, whereas [SID] varied from 48.1 to 26.6 meq/l and was directly correlated with [A]tot. The decline in [SID] was primarily attributable to an increase in [Cl]. A direct correlation was also noted between PCO2 and [SID], but not between PCO2 and [A]tot. The decrease in [SID] and PCO2 was such that neither [H+] nor [ HCO3 ] changed significantly with [A]tot.

  • metabolic alkalosis
  • electrolytes
  • critical care

our understanding of acid-base chemistry, specifically the influence of proteins and strong ions on [H+] and [HCO3 ], has undergone a paradigm shift since the rediscovery of fundamental physicochemical principles by Stewart (23-25). In essence, the concentration of weak electrolytes {[H+], [ HCO3 ] and dissociated proteinate concentration ([A])} is shown to be determined by the magnitude of the respective dissociation constants and three independent variables: PCO2 , the total concentration of weak acids ([A]tot), and the charge difference between the sum of strong cations and strong anions ([Na+] + [K+] − [Cl] − [strong organic anions]), known as the strong-ion difference ([SID]). Strong ions are defined as electrolytes and strong organic acids that are completely dissociated in solution.

An independent variable is defined as one that influences the system but is not influenced by the system. Within the context of acid-base chemistry, the term “system” refers to any single aqueous compartment, i.e., plasma, interstitial, intercellular, or cerebrospinal fluid (CSF). Thus PCO2 directly affects weak electrolyte concentration via the CO2 hydration reaction while being itself a function of the rate of alveolar ventilation and CO2 production. Similarly, [A]tot, largely determined by total protein concentration ([Pro]tot) in plasma, influences weak electrolytes by virtue of the fact that the dissociation constant is somewhat less than physiological pH; therefore, the concentration of the dissociated fraction (the anion gap) exceeds that of the undissociated fraction. However, [A]tot, i.e., the sum of dissociated and undissociated fractions, is a function of protein metabolism and volume of distribution. Finally, [SID] influences the concentration of weak electrolytes in that, as a net positive charge, it must be balanced by the sum of all weak anions to maintain electrical neutrality. The magnitude of the dissociation constants for bicarbonate and weak acids compared with other dissociation constants is such that [SID] is closely approximated by the sum of [ HCO3 ] and [A]. [SID] is a function of the rates of input and output and volumes of distribution of the contributing strong ions. Thus PCO2 , [A]tot, and [SID] fulfill the criteria for independent variables in that they directly influence the dissociation reactions that generate weak electrolytes but are themselves determined by distinctly separate control mechanisms. The corollary of this statement is that weak electrolytes are dependent variables; their concentration can only be altered by a change in one or more of the three independent variables.

As discussed by Stewart (23-25) and reviewed by Jones (10, 11), acute acid-base disturbances result from change in PCO2 or [SID]. The compensatory response to a primary disturbance in either independent variable is an adjustment in the other such that change in [H+] is minimized. Although [A]tot does not change acutely, it does have a direct influence on the final concentration of [H+] and [ HCO3 ] for a given PCO2 and [SID]. In contrast to acute acid-base disturbances, low [A]tot resulting from hypoproteinemia can occur over several days. As suggested by the equation for electrical neutrality, predicted by physicochemical principles and shown both in vitro (21) and in vivo (15), the loss of weak acid has been associated with a decrease in [H+] and an elevation in [ HCO3 ]. Although similar changes in [H+] and [ HCO3 ] secondary to increased [SID] would be attenuated by increased PCO2 (2, 3, 10), the nature of the compensatory response to hypoproteinemia is not clear. Indeed, hypoproteinemia secondary to congestive heart failure or cirrhosis is associated with hyperventilation, which might be expected to further decrease [H+] (20). Moreover, clinical experience suggests that hypoproteinemic-associated alkalosis in the intensive care population is not as frequent an observation as is hypoproteinemia itself. This suggests the presence of a compensatory response mechanism(s) that serves to attenuate the decrease in [H+] and increase in [ HCO3 ] that would otherwise occur in patients with a reduction in [A]tot. Consequently, the presence of low [H+] and elevated [ HCO3 ] in the context of low [A]tot may indicate a breakdown in the normal compensatory response.

The present study was a prospective, nonrandomized observational examination of the relationship among hypoproteinemia and electrolyte and acid-base status in a critical care population of patients. The hypothesis of the present study was that the compensatory response to the loss of weak acid secondary to hypoproteinemia is a decrease in [SID] due to either an increase in [Cl] and/or a decrease in [Na+].


Patient population.

Approval for the study was obtained from the Ethics Review Board of Victoria Hospital, London, Ontario. The requirement for consent was waived because the study was only observational and patients would not be identified. Results of the study were not used in the management of patient care. The patient population consisted of surgical, medical, and trauma patients requiring critical care. The only criteria for inclusion were the presence of an arterial line and requirement for “routine” morning blood work as determined by the attending physician. There were no exclusion criteria. The duration of the study was 1 mo. A total of 91 patients was recruited.


Hospital procedure was to collect blood samples between 0500 and 0700 so as to have results available for morning rounds. All samples were drawn from an indwelling arterial line by nursing staff and handled according to standardized hospital-approved procedures. Arterial blood-gas samples were taken anaerobically and measured without delay. Lactate samples were immediately placed on ice and analyzed as soon as possible. Electrolytes and [Pro]tot were measured in plasma. All measurements were performed by qualified technicians by using techniques and equipment meeting standardized levels of quality control. Table 1 provides a summary of measurement methodology and coefficients of variation for each variable. Only one set of data from morning blood work (collected between 0500 and 0700) per patient per day was analyzed for this study. Each patient was followed for as long as the above inclusion criteria were fulfilled. The total number of samples was 223. Four samples were excluded because of incomplete data.

View this table:
Table 1.

Measurement methodology

Routine morning blood work within the Intensive Care Department comprised measurements of1) arterial blood gases;2) sodium, potassium, chloride, phosphate, calcium, and magnesium; and3) albumin. The only additional measurements required for the study were [Pro]tot and [La], for which an additional volume of blood was required. [Pro]tot was measured in the same sample used to determine albumin concentration ([Alb]).


[SID] was calculated from[SID]=[Na+]+[K+][Cl][La] Equation 1 Other strong cations and anions that could be included in the calculation of [SID] are calcium, magnesium, and sulfate. However, in view of the fact that measurement of ionized divalent cation concentration was not available at the time of the study and the lack of routine measurement of the sulfate concentration, these strong electrolytes were excluded from the [SID] calculation. Strong metabolic anions, such as formate and ketoacids, would have been included in the calculation of [SID] had their presence been suspected clinically.

The conversion coefficient for [Pro]tot, measured in units of grams per liter, to [A]tot, measured in units of milliequivalents per liter, was 0.24 (6, 11, 14). Because phosphate concentration accounted for <10% of the weak acid component, it was not included in the calculation of [A]tot.

[SID], [A]tot, and PCO2 were used in the calculation of [H+] from a fourth-order polynomial representing the simultaneous solution of four equilibrium reactions (the dissociation of water, the CO2 hydration reaction of carbonic acid to bicarbonate, bicarbonate to carbonate, and undissociated to dissociated weak acids), an equation of mass balance for [A]tot, i.e., one in which the sum of dissociated and undissociated weak acids must equal [A]tot, and an expression of electrical neutrality. The final polynomial and the iterative program (written in BASICA by the author) used to solve it were from Stewart (24). The ability of this equation to accurately predict measured [H+] is thus dependent on the values used for the dissociation constants (24) and the accuracy of six different measurements: [Na+], [K+], [Cl], [La], PCO2 , and [Pro]tot.

The dissociated proteinate component [A] of [A]tot, equivalent to the anion gap, was calculated from the solution of the two equations describing the dissociation reaction and mass balance for [A]tot [A]=(K3)[Atot]/(K3+[H+]) Equation 2whereK 3 is the equilibrium constant for the weak acid dissociation reaction. Electrical neutrality is approximated by[SID][HCO3][A]=0 Equation 3 Because the presence of electrical neutrality in an aqueous solution can be viewed as a gold standard, any deviation of the calculated estimate from zero suggests either the presence of unmeasured strong ions such as ketoacids or formate, which would be included in the calculation of [SID], and/or the accumulation of measurement variability (11). The magnitude of the deviation from electrical neutrality was termed the unmeasured ion concentration ([UMI]).

Statistical analysis.

Data pairs were analyzed by least-squares regression and analysis of variance (28). Statistical significance was determined atP < 0.05.


A profile of patient pathology is provided in Table2. For age, the mean ± SD was 59 ± 16 yr. The range of values and means for the three independent variables, [A]tot, [SID], and PCO2 , and three dependent variables, [A], [H+], and [ HCO3 ], are presented in Table 3. Also included are range and mean for PO2 , [La], [Pro]tot, [Alb], and [UMI]. A total of 22 patients had [La] levels between 2.5 and 13 meq/l for a period not exceeding 24 h. Regression analysis of the independent and dependent variables is summarized in Table 4. With respect to the independent variables [Eqs. 1–7 in Table 4], a statistically significant correlation was observed between [SID] and [A]tot(r = 0.236,P < 0.001). The decline in [SID] was secondary to an elevation in [Cl], as demonstrated by a weak, but statistically significant, inverse correlation between [Cl] and [A]tot(r = −0.147,P < 0.05) and a stronger correlation with [SID] (r = −0.538, P < 0.001). In contrast, [Na+] was not significantly related to either [A]tot or [SID]. Although PCO2 was found not to be correlated with [A]tot, it was significantly correlated with [SID] (r = 0.392,P < 0.001). A linear relationship was noted between plasma [Alb] and [Pro]tot(r = 0.548,P < 0.001).

View this table:
Table 2.

Summary of patient profiles categorized by organ system and pathophysiology

View this table:
Table 3.

Minimum, maximum, and mean ± SD values for independent and dependent variables and arterial Po 2

View this table:
Table 4.

Least squares regression analysis

The influence of independent variables on [H+], [HCO3 ], and [A] is also shown (Eqs. 9–17 in Table 4). Regression analysis demonstrates that neither [H+] nor [ HCO3 ] was correlated with [A]tot, whereas [A] was directly related to [A]tot(r = 0.982,P < 0.001). Although [H+] was not correlated with [SID], both [HCO3 ] and [A] were significantly correlated with [SID] (r = 0.477,P < 0.001 andr = 0.246,P < 0.001, respectively). Finally, both [H+] and [ HCO3 ], but not [A], were significantly correlated with PCO2 (r = 0.409,P < 0.001 andr = 0.508,P < 0.001, respectively).

A significant correlation was observed between [H+] from measured pH ([H+]meas) and [H+] calculated from [A]tot, [SID], and PCO2 ([H+]calc;r = 0.505,P < 0.001) (Fig.1 A). Those patients with elevated [La] levels are represented in Fig. 1 A by closed symbols. There was no significant difference in slope or elevation of the regression equations between the two patient populations. Much of the variation between [H+]calcand [H+]measwas a direct reflection of the magnitude of the [UMI] (Fig.1 B). The difference between [H+]measand [H+]calcwould be 1.5 neq/l if electrical neutrality were actually realized, i.e., substitution of an [UMI] = 0 in the regression equation.

Fig. 1.

A: calculated [H+] ([H+]calc) compared with directly measured [H+] ([H+]meas). Single outlying variable has been included in regression analysis because no technical errors could be identified in its calculation ([H+]calc= 0.51[H+]meas+ 20.03, r = 0.505,P < 0.001,n = 219). ▪, Patients in whom [La] levels exceeded 2.5 meq/l. B: difference between [H+]measand [H+]calcis directly related to magnitude of apparent deviation from electrical neutrality, unmeasured ion concentration ([UMI]) [H+]meas − calc = 1.20([UMI]) −1.52,r = 0.713,P < 0.001,n = 219. Electrical neutrality was estimated from strong-ion difference ([SID]) − [ Embedded Image ] − dissociated proteinate ([A]).

The possibility that an apparent discrepancy from electrical neutrality was masking the statistical relationship among [A]tot, [SID], and PCO2 was assessed by selecting data in which charge balance more closely reflected the reality of electrical neutrality. The average for accepted measurement variability for [Na+], [K+], and [Cl] was ±3.0 meq/l (Table 1). The use of a subset of data in which [UMI] was also within ±3.0 meq/l of electrical neutrality (n = 131) demonstrates improved agreement between [H+]calcand [H+]meas(r = 0.901,P < 0.001) (Fig.2). Patients with elevated [La] levels are represented by closed symbols in Fig. 2. The slope of the regression equation for this group was significantly decreased.

Fig. 2.

Selection of a subset of data in which charge balance was within ±3.0 meq/l of electrical neutrality demonstrates greater agreement between [H+]measand [H+]calc: [H+]calc= 1.16([H+]meas) − 3.39, r = 0.901,P < 0.001,n = 131. ▪, Patients in whom [La] levels exceeded 2.5 meq/l.

Submission of the selected subset of data to the same series of regression analysis as above demonstrated qualitatively similar patterns of relationships but with improved statistics. [SID] varied directly with [A]tot(r = 0.466,P < 0.001) (Fig.3) secondary to elevation of [Cl] (r = −0.2,P < 0.02) rather than to a decrease in [Na+] (P = not significant) (Fig. 3). As previously noted, PCO2 and [SID] were significantly correlated (r = 0.449,P < 0.001), and no correlation was observed between [A]tot and PCO2 (Fig.4). Patients with elevated [La] levels are represented by closed symbols in Figs. 3 and 4. Other than a significant decrease in elevation of the regression equation relating [SID] to PCO2 (Fig. 4), no differences were noted in the regression equations for the two patient populations.

Fig. 3.

[SID] (triangles) is directly related to total concentration of weak acid ([A]tot): [SID] = 0.99[A]tot + 24.7,r = 0.466,P < 0.001,n = 131. Decline in [SID] is attributable to an increase in [Cl] (circles) that occurs with a fall in [A]tot: [Cl] = −0.66([A]tot) + 110.7, r = −0.200,P < 0.05,n = 131. No correlation is observed between [Na+] (squares) and [A]tot: [Na+] = 0.18([A]tot) + 134.6, r = 0.060,P = not significant,n = 131. Closed symbols, patients in whom [La] levels exceeded 2.5 meq/l.

Fig. 4.

A strong correlation is observed between [SID](triangles) and Embedded Image : [SID] = 0.22( Embedded Image ) + 28.6,r = 0.449,P < 0.001,n = 131, but not between [A]tot (circles) and Embedded Image : [A]tot = 0.016( Embedded Image ) + 12.5,r = 0.070,P = not significant,n = 131. Closed symbols, patients in whom [La] levels exceeded 2.5 meq/l.

With respect to the dependent variables, neither [H+] nor [ HCO3 ] was correlated with [A]tot; however, [A] varied directly with [A]tot([A] = 0.85[A]tot − 0.11, r = −0.991,P < 0.001) (Fig.5 A). Regression equations describing [H+] and [ HCO3 ] vs. [A]tot in patients with elevated [La] levels (closed symbols) were, respectively, elevated and lower than in patients with normal lactate. There were no significant differences in the slopes of these regression equations. Although [H+] was inversely related to [SID] ([H+] = −0.352[SID] + 50.2,r = −0.277,P < 0.001), a direct relationship to [SID] was noted with both [A] and [ HCO3 ] ([A] = 0.2[SID] + 3.62, r = 0.488, P< 0.001 and [ HCO3 ] = 0.73[SID] − 0.62,r = 0.856,P < 0.001, respectively) (Fig.5 B). There were no significant differences in the regression equations for patients with elevated [La] levels (closed symbols). Finally, both [H+] and [ HCO3 ], but not [A], were related to PCO2 {[H+] = 0.38( PCO2 ) + 21.4,r = 0.605,P < 0.001 and [ HCO3 ] = 0.23( PCO2 ) + 17.4,r = 0.551,P < 0.001, respectively} (Fig. 5 C). Again, for those patients with increased [La] levels, the regression equations describing [ HCO3 ] and [H+] to PCO2 were not significantly different with regard to slope but had decreased and increased elevation, respectively.

Fig. 5.

Relationship of dependent variables [H+] (squares), [ Embedded Image ] (circles), and [A] (triangles) to independent variables [A]tot(A), [SID] (B), and Embedded Image (C). Closed symbols, data from patients in whom [La] levels exceeded 2.5 meq/l. No correlation with [A]tot is noted for [H+] and [ Embedded Image ], whereas a strong correlation is noted between [A] and [A]tot(A). [H+] and [SID] are inversely correlated, and a direct correlation with [SID] is noted for [ Embedded Image ] and [A] (B). Finally, both [H+] and [ Embedded Image ], but not [A], are correlated with Embedded Image (C). See equations in text.

Utilization of the regression equations generated above for the selected subset of data demonstrates that a decrease in [A]tot from 20 to 10 meq/l was associated with a decrease in [A] of 8.5 meq/l (from 16.9 to 8.4 meq/l). This quantity was closely approximated by a 9.8 meq/l decrease in [SID] from 44.4 to 34.6 meq/l at [A]tot of 20 and 10 meq/l, respectively. Sixty-seven percent of the decrease in [SID] could be attributed to a 6.6 meq/l increase in [Cl] (from 97.5 to 104.1 meq/l at [A]tot of 20 and 10 meq/l, respectively). Although a significant relationship between [Na+] with [A]tot could not be demonstrated, regression analysis suggested a downward trend of 1.9 meq/l (from 138.3 to 136.4 meq/l at [A]tot of 20 and 10 meq/l, respectively), which may account for 20% of the change in [SID]. Together, the increase in [Cl] and possible decrease in [Na+] account for 8.5 meq or 87% of the decrease in [SID]. The 9.8 meq/l decline in [SID] to 34.6 meq/l (predicted at [A]tot of 10 meq/l) is associated with a decrease in PCO2 of 8 Torr.

With the use of physicochemical principles, the influence of the changes in independent variables on [H+] and [ HCO3 ] can also be assessed. An isolated decrease in [A]tot from 20 to 10 meq/l would be expected to decrease [H+] from 43.8 to 32.9 neq/l and increase [HCO3 ] from 27 to 35 mmol/l ([SID] = 44 meq/l and PCO2 = 46 Torr calculated at [A]tot = 20 meq/l). Superimposing the observed 9.8 meq/l decrease in [SID] and 8 Torr decrease in PCO2 would result in little change in [A] but would increase [H+] to 38 neq/l and decrease [HCO3 ] to 26 mmol/l.


The patients followed in this study demonstrated a variety of pathologies, from relative stability after uncomplicated cardiac bypass surgery to serious illness because of septic shock. Twenty-two patients had elevated [La] levels between 2.5 and 13 meq/l for a period not exceeding 24 h. Accordingly, this intensive care population provided a continuum of disturbances in metabolic, electrolyte, and respiratory systems, and, as such, facilitated delineation of generalized patterns of response among [A]tot, [SID], and PCO2 . In view of the inherent assumption that physicochemical principles are equally valid in both health and disease (10, 24), no attempt was made to relate findings to individual disease and/or therapeutic modalities. However, it is equally possible that a compensatory response in one of the independent variables to a disturbance in [H+] imposed by a change in another of the independent variables would be effectively limited by medication, the presence of positive pressure ventilation, renal function, volume status, and underlying pathology. Nevertheless, despite the presence of these complicating factors, results support the hypothesis that the compensatory response to reduction in [A]tot(hypoproteinemia) is a decrease in [SID], secondary to increased [Cl] and more so than decreased [Na+]. Furthermore, it was shown that the decrease in [SID], but not [A]tot per se, is directly related to a reduction in PCO2 . The effect of decreased [SID] and PCO2 in association with low [A]tot is a reduction in dissociated proteinate fraction but no significant change in [ HCO3 ] or [H+]. Although enticing, this conclusion is weakened by the observed variability in [Cl], [Na+], and thus [SID], perhaps even to the point of questioning any biological relevance. However, it could equally be argued that observing even a statistically significant change in [Cl] and [SID] is surprising given the small change in [SID] relative to the SD for sodium and chloride. A decrease in [SID] of only 9.8 meq/l could easily occur without sodium or chloride levels deviating from the normal physiological range.

It is well established that [H+] and [ HCO3 ] are altered by an isolated change in [A]tot. Rossing et al. (21) demonstrated that, within an in vitro, single-compartment system in which [SID] and PCO2 are constant and [A]tot is lowered, both [H+] and [A] decrease, with the latter being closely approximated by an increase in [ HCO3 ]. Within the context of dynamic, multicompartmented physiological systems, however, a disturbance in [H+] secondary to a change in one independent variable is usually compensated for by a complementary change in another. For example, hypercapnia and associated increase in [H+] are compensated for by a decrease in [Cl] (18) and increased [SID]. Conversely, a decrease in [SID], secondary to either an electrolyte disturbance or metabolic production of a strong organic acid, causing an increased [H+], is compensated for by hyperventilation and decrease in PCO2 . The net result of these complementary changes in independent variables is a return of [H+] toward, but usually not an achievment of, a normal value (2, 3, 10, 11, 24). Thus [H+] and [ HCO3 ] reflect the influence of [SID] and PCO2 for a given [A]tot.

The nature of the compensatory response to a reduction in [A]tot secondary to hypoproteinemia is not clear. Indeed, Fencl and Leith (2) stated that there is no compensatory mechanism for the decreased [H+] associated with hypoproteinemia. Intuitively, the decrease in [H+] would be compensated for by hypoventilation, as occurs with increased [SID] (2, 3, 10). In fact, of the eight patients with hypoproteinemia described by McAuliffe et al. (15), all had [SID] closely approximating a normal value, and seven had levels of PCO2 >50 Torr. Consequently, [H+] was not appreciably perturbed (36 neq/l), but [ HCO3 ] was increased to 32.5 meq/l. In distinct contrast, Rossing et al. (20) reported paradoxical hyperventilation in spontaneously breathing, stable patients with low [A]tot secondary to chronic hepatic failure or hepatic cirrhosis, a response that would further exacerbate the decrease in [H+]. Results of Rossing et al. (20) are not directly comparable to those in the present study in that many of the patients in the present study had some degree of ventilatory support for a varying amount of time over the course of their inclusion in the study. Obviously, the ability of the respiratory system to partake in a compensatory response to changes in either [A]tot or [SID] would be a direct function of the degree to which ventilatory control was imposed. Nevertheless, a hyperventilatory response was noted with the decline in [SID]. The lack of correlation between PCO2 and [A]tot in the present study is perhaps somewhat surprising from a statistical perspective in view of the correlation between PCO2 and [SID] and between [SID] and [A]tot. Although no correlation was noted between [SID] and [A]tot by Rossing et al. (20), this may be a reflection of both a low ratio of anionic albumin to cationic globulin noted by these authors (see below) and the relatively narrow range and mild degree of hypoproteinemia (mean value = 61 ± 0.7 g/l). Interestingly, submitting this value of [A]tot intoEq. 4 in Table 4 predicts a [SID] of 39 meq/l, which closely approximates a value of 38 meq/l calculated from their data. Finally, a lack of correlation between [H+] and [A]tot and a significant correlation between PCO2 and [SID] are consistent with results of the present study.

The nature of the error signal and effector mechanism(s) responsible for integrating the three independent variables is not fully understood. There is a growing body of evidence to suggest that the goal of the regulatory response is not to maintain a given [H+] per se, but to optimize [H+]-to-imidazole pK of proteins so as to maintain the electrostatic state within the microenvironment and thus structural conformation and biological function (8, 9, 17, 19). Thus extracellular [H+] may only be a distant reflection of the actual regulated variable. Integration of the respective control systems would extend beyond those mechanisms regulating the three independent variables in plasma to include regulation of intracellular determinants of [H+] and those variables that influence imidazole pK, i.e., temperature, osmolality, and strong ions (22). Furthermore, hierarchal characteristics would be expected in that acid-base disturbances may not disrupt all physiological compartments to the same degree. Hence, control would ultimately be directed toward minimizing [H+]/pKdisturbances in those compartments least tolerable of change. For example, data from exercise physiology indicate that a decrease in systemic [SID] from generation of lactic acid is localized to the peripheral compartment with sparing of [SID] of the central compartment. Accordingly, ventilation is regulated to maintain central PCO2 to central [SID], and thus central [H+]/pK, at the expense of elevated systemic [H+] (9).

In contrast to the short-term acid-base disturbance of exercise, sustained alterations in systemic [SID] can be reflected in [SID] of CSF ([SID]CSF) (8, 9). Appropriate alteration in ventilation and PCO2 would be initiated so as to preserve central [SID]/ PCO2 . Similarly, the solubility of dissolved CO2ensures rapid equilibrium of PCO2 between central and peripheral compartments with respiratory pathology, leading to adjustment of both [SID]CSF and plasma [SID] (17). The preponderance of opinion is that movement of [Cl] is principally responsible for compensatory adjustments of [SID] in both CSF (7, 16, 17) and plasma (2, 6, 11). The mechanism of chloride regulation is also not well understood, but transcellular and -epithelial movement of chloride would be expected to involve change in electrochemical gradients affecting passive movement, as well as active enzymatic and hormonally regulated transport mechanisms.

The loss of weak acid from plasma is a unique acid-base disturbance in that low permeability of the blood-brain barrier to proteins would ensure that the central compartment is not directly affected. However, if the compensatory response to the alkalizing influence of hypoproteinemia is a decrease in systemic [SID] secondary to increased [Cl], [SID]CSF may also be expected to decline (9). This hypothesis provides an explanation for the hyperventilation associated with hypoproteinemia demonstrated by Rossing et al. (20) and suggested by the present study. The hypercapnia associated with hypoproteinemia reported by McAuliffe et al. (15) may thus reflect a primary respiratory acidosis rather than a compensatory response. The combination of low [A]tot with elevated PCO2 would represent conflicting signals to the regulation of [Cl] such that [SID], at least in the peripheral compartment, would remain relatively normal. However, the elevated PCO2 would be reflected in the central compartment and effect a compensatory response in which [SID]CSF would be expected to increase. Depending on the degree of interdependence between central and peripheral regulation of [SID], this particular combination of independent variables may represent significant disturbance of PCO2 /[SID] in CSF, if PCO2 is severely elevated and [A]tot very low. The consequences of imbalance of PCO2 /[SID] between central and peripheral compartments to ventilatory drive are unknown.

There are two, interconnected technical aspects of the present study deserving of discussion. The first relates to the difficulty in achieving charge balance so as to reflect the physical reality of electrical neutrality. The second is use of appropriate value for the coefficient required for conversion of [Pro]tot, in grams per liter, to [A]tot, in milliequivalents per liter, and the value for the weak acid dissociation constant. With respect to the former technical aspect, discrepancy in charge calculation from electrical neutrality could be due to an error of either omission or measurement, the two not being mutually exclusive. It would certainly seem reasonable that loss of physiological homeostasis might be associated with the appearance of, or unmasking of, ions not normally found in health and not measured in the present study. The problem of charge balance is not uncommon; discrepancy between measured [SID] and [SID] calculated so as to actually reflect electrical neutrality, i.e., an effective [SID], has been suggested to represent the presence of “unknown strong anions” in an in vitro animal model of sepsis (12) and in patients with sepsis (13, 6). However, the nature of these presumed organic acid anions is unknown. Although this possibility is not ruled out, there were no patients in the present study in whom the presence of an organic anion other than lactate was suspected by clinical diagnosis. Not including divalent cations in the calculation of [SID] in the present study, for reasons outlined above, could also have contributed to an error of omission. Waters et al. (27) demonstrated that incorporation of divalent cations to the calculation of [SID] improved agreement between measured and calculated pH at the extremes of acid-base disturbances. Review of the correlation between [H+]measand [H+]calcfrom the selected subset of data in which electrical neutrality was achieved to within ±3.0 meq/l does not demonstrate any increase in discrepancy between [H+]measand [H+]calcover the pH range of 7.24 to 7.6. In a similar vein, measurement of whole blood lactate, rather than plasma lactate, may contribute to [UMI] because a gradient of lactate exists across the red blood cell. However, the quantitative significance of this is difficult to assess as there were only 22 samples in which lactate was elevated above 2.5 meq/l, with the average of those being 3.85 ± 2.34 (SD) meq/l.

Technical and methodological approaches may have contributed to the observed variability in [Na+] and [Cl], thus the magnitude of [SID] and, by extension, of [UMI]. Accumulated measurement error is certainly a confounding factor in the demonstration of electrical neutrality (11), and accuracy would be improved if samples had been drawn in duplicate and measured with the same instrument by a single technician. However, the observation that mean [UMI] is not significantly different from zero (median = −0.5 meq/l, mode = −1.3 meq/l) and has approximately the same range (Table 3) and percent distribution (55% < 0, 45% > 0) on either side of zero argues against a consistent error of omission or technique. The gold standard of electrical neutrality in aqueous solutions refers to ionic activity rather than total concentration. In contrast to total concentration, the activity of any given electrolyte is a function of the ionic milieu in which it exists. It may be possible that the influence of those ionic species not measured by ion-selective electrodes, i.e., lactate, [ HCO3 ], total protein, and albumin, may be being under- and overestimated at the extremes of strong-ion concentration.

It can be concluded that difficulty in demonstrating electrical neutrality is not restricted to the present study; it is not possible to clearly distinguish between accumulation of measurement error and/or error of omission; and finally, no study to date has addressed the issue of a clinical condition(s) that could account for this observation, thus helping provide direction for future investigations. Equally true, however, the difficulty in demonstrating electrical neutrality does not negate the validity of physicochemical principles. The goal of the quantitative approach is not to replace the pH meter but to explain the results. Although attempting to minimize the possible influence of these sources of error by subdividing the data solely on the basis of approximation to electrical neutrality may be questionable, the only consequence of doing so was an improvement in the statistical relationships; i.e., the overall qualitative characteristics of the results were preserved.

The lingering discrepancy between [H+]measand [H+]calcat electrical neutrality suggests that the value of [A]tot (0.24 × [Pro]tot in g/l), milliequivalents per liter, and/or the value of the dissociation constant for [A]tot was incorrect. At present there is much debate within the literature regarding these two constants. As discussed by Jones (11) and Figge et al. (4, 5), Stewart’s (24) use of [A]tot can be viewed as a first approximation in that it not only combines both phosphates and proteins but also represents polyprotic macromolecules as comprising several independently dissociating bases with identical dissociation constants. After examination of this limitation in detail, Figge et al. (4) concluded that albumin was the only protein moiety that significantly influenced acid-base status. Nevertheless, a comparison between the equations of Figge et al. (4, 5) and those of Stewart (24) was performed by Kowalchuk and Scheuermann (14), and no difference was found between the two approaches.

Intriguingly, there is evidence to suggest that hyperproteinemia, although more rare, may also be responsible for primary acid-base disturbances (11). Wang et al. (26) suggested that hyperalbuminemia secondary to dehydration during cholera may in part be responsible for a metabolic acidosis. Similarly, the concentration of cationic globulins relative to albumin may also influence plasma acid-base status. De Troyer et al. (1) reported that sera from patients with IgG myeloma had a significantly lower anion gap than sera from patients with IgA myeloma. They hypothesized that IgG paraproteinemia would increase chloride binding because of its higher isoelectric point compared with IgA paraproteinemia. Thus pathological processes may alter both weak acid content and the anionic-to-cationic ratio.

In conclusion, the results of the present study demonstrate that the loss of weak acid secondary to hypoproteinemia is compensated for by an increase in [Cl] such that [SID] decreases. A direct correlation between systemic [SID] and PCO2 was also noted. The net affect of a decrease in [A]tot, [SID], and PCO2 is such that the concentration of dissociated weak acid declines, but changes in [H+] and [ HCO3 ] are minimized. Reflection of the low systemic [SID] in CSF may offer an explanation of the observed correlation between [SID] and PCO2 and suggest centrally mediated integration between the control systems regulating [SID] and PCO2 .


  • Address for reprint requests: P. Wilkes, Univ. of Ottawa Heart Institute, Ottawa Civic Hospital, 1500 Carling Ave., Ottawa, Ontario, Canada K1Y 4E9 (E-mail: pwilkes{at}


View Abstract