Journal of

Critical Care
VOL VIII, NO 4 DECEMBER 1993

ORIGINAL INVJBTIGATIONS
A Physical Chemical Approach to the Analysis of Acid-Base
Balance in the Clinical Setting
Brian M. Gilfix, Mark Bique, and Sheldon Magder

We evaluated the clinical application of a model of [Cl-]; (3) protein concentration; and (4) “other species”
acid-base balance, which is based on quantitative (ie, anion and cations other than [Na+], [K+], and [Cl-]).
physical chemical principles (Stewart model). This The contribution of “other species” was obtained from
model postulates that acid-base balance is normally the difference between the SID measured and that
determined by the difference in concentration be- predicted from Stewart’s equation. It could also be
tween strong cations and anions (strong ion difference calculated from the difference between the standard
[SID]), Pcos, and weak acids (primarily proteins). We Siggaard-Anderson calculation of base excess and
measured electrolytes and blood gases in arterial base excess attributable to free water, [Cl-], and
blood samples from 21 patients in a medical or surgical proteins (ie, base-excess gap). Our results indicate
intensive care unit or emergency room of a tertiary that the SID gap, base excess gap, and anion gap
care hospital. The measured SID frequently differed reflect the presence of unmeasured ions, and both the
from SID calculated from the measured blood compo- anion-gap and base-excess gap provide readily avail-
nents, which indicates that unmeasured cations or able estimates of the SID gap. This provides a simple
anions are present; these could not be accounted for bedside approach for using the Stewart model to
by lactate, ketones, or other readily identifiable ions. analyze the nonrespiratory component of clinical acid-
We used an approach to acid-base analysis that is base disorders and indicates that, in addition to unmea-
based on changes in base excess or deficit due to sured anions, unmeasured cations can be present.
changes in: (1) free water as assessed by [Na+]; (2) in Copyright o 1993 by W.B. Saunders Company

H YDROGEN ION concentration is a cru-

cial determinant of the function of many
systems in the body, and therefore, analysis of
served clinicians well and provides a handy
bedside approach for classifying patients.
However, as discussed in detail by Stewart,
acid-base balance has long been at the center of this approach does not follow rigorous physical
clinical management. Disturbances of [H+] have chemical principle$J and, therefore, does not
traditionally been analyzed with the equations allow a proper analysis of mechanisms underly-
developed by Henderson and Hasselbalch at ing acid-base disorders. Briefly, Stewart ana-
the turn of the century.’ These equations relate lyzed the components of the solution that make
[H+] to Pcoz and [HC03-1. The definitions of up blood and identified four systems that deter-
alkalosis and acidosis are based on deviations of mine [H’]. The first is the difference in concen-
[H’] from the reference value of 40 nmol/L or a
deviation of the pH from 7.40. Changes in the From the Critical Cure Division und Clinical Rmhwrbt~
Pco, from 40 mm Hg are used to determine if Division, Royal Victoria Hospitul.
there is a respiratory component, and changes Received August 26, 1992; rrccepted Muy 2.7. 1993,
in the [HC03-] determine if there is a metabolic Address reptint requests to Sheldon Magder. MDC‘M. PhD.
Critical Care Divi,rion, Royul Victoria Hospital. 08 7 I’itw
component to the acid-base disturbance. Meta- Avr W, Montreal, Quebec H3G IAI.
bolic acidoses are then divided into those with a Copyniht (‘ 1993 by W.R. Suunder.s ~‘otrrpum
normal or wide anion gap. This approach has ma3-944119.~lOK04-0001$5.0010

Journalof Critical&we, Vol8, No 4 (December), 1993: pp 187-197 187

188
tration between strong anions and cations that
are defined as ions with dissociation constants
of > 1O-4 or < lo-12. The major ones in blood
are [Na+], [K+], and [Cl-]. Because of the law of
conservation of charges, the difference between
the concentrations of cations and anions has to
be balanced by other ions, and this has an
important effect on the dissociation of weaker
electrolytes including H20. The difference be-
tween the strong anions and cations is called the
strong ion difference (SID). The second factor
is Pcoz. When CO2 dissolves in water, it forms
the volatile weak acid carbonic acid, which
dissociates to [H+] and [HC03-1. The concentra-
tion of Pco2 in the blood is tightly regulated by
the ventilatory system. Thus, Pco, is an indepen-
dent variable that determines [HC03-1. Con-
sider, for example, how CO2 production can
increase more than 12-fold with exercise, and
pH does not change. The third factor he identi-
fied is the concentration of weak acids. The
major contributors to this group are serum
proteins4s5 and of these, albumin provides the
dominant anion. The side chain of histidine has
a logarithm of dissociation constant (pKa) close
to the pH of normal blood and, therefore, is an
important buffer at physiological PH.~ The fourth
factor is the presence of other acids or bases
(“other species”) that are not normally present
in significant quantities. These can be endog-
enously produced, such as lactic acid or ketoac-
ids, or come from exogenous precursors such as
methyl alcohol, ethylene glycol, or salicylates.
Based on this analysis, [H+] or [HCO,-1, as
well as the dissociation of any weak acids, are
dependent variables that are affected by the
SID, PcoZ, and total concentration of the weak
acids that, as noted above, mainly consist of
proteins, as well as the concentration of other
acids in the blood. It is changes in these four
variables that change the dissociation of water,
and it is this and not the loss or gain of [HC03-]
or [H+] that determines the concentration of
[H+]. Human blood is essentially an alkalotic
solution, so that [OH-] is much greater than
[H+]. Consequently, the addition of even a
strong acid such as HCl has a much smaller
effect on the final [H+] than might initially be
expected.
The most important immediate controller of
the acid-base balance is the respiratory system,
GILFIX, BIQUE, AND MAGDER
which can quickly change Pco2. The second
major control mechanism is the kidney through
its handling of [Na+] and [Cl-], which are the
major determinants of the SID. Other factors
include changes in total protein, which are
probably only of limited significance as regula-
tors, as well as shifts in electrolytes across the
cell membranes, such as the chloride shift into
erythrocytes. The latter increases the SID of
blood and, therefore, has an alkalinizing effect.
It has also been suggested that shifts in strong
ions between the blood and interstitial space
could also play a role in regulating [H+].2,3x7.8
Although Stewart’s analysis provides a rigor-
ous approach to acid-base balance and has
provided important insights into the metabolic
changes during exercise71y-” and the response to
salt loading,s,i2 its application at the bedside has
been difficult. Stewart provided a nomogram
that showed the effect of a change in SID on
[H+] at different levels of Pco2, assuming a
constant protein concentration2 but did not
provide a readily usable approach for nonrespi-
ratory problems. Jones pointed out that if no
unmeasured anions are present in the blood,
the arithmetic sum of [Na+], [K+], [Cl-],
[HC03-1, and the ionic equivalent of the disso-
ciated proteins in microequivalents per liter
should be zero.6,33This is based on the principle
of electrical neutrality. Any deviation from this
must be due to the presence of acidic or basic
elements that are not normally present in blood.
However, a major problem with this analysis is
that the ionic contribution of proteins cannot
easily be obtained, because proteins are not a
homogenous substance with an exact ionic
equivalent or fixed dissociation constant.4 Thus,
the ionic equivalent of proteins may vary with
changes in [H+], the type of protein, and possi-
bly even the Pco~.‘~,~~
To date, the Stewart model has only been
used in small selected series.5l16 Our purpose
was to determine whether it could be used to
characterize the abnormalities of acutely ill
patients. Based on the Stewart model, we di-
vided the metabolic component of acid-base
disturbances into four categories: (1) changes in
free water, which are assessed from changes in
the [Na+]; (2) changes in [Cl-]; (3) changes in
the protein concentration; and (4) changes in
the presence of unmeasured anions or cations.
ACID-BASE BALANCE 189

The latter are estimated from the difference in MATERIALS AND METHODS
the SID measured from [Na+], [K+], and [Cl-], Patients
and the SID predicted by Stewart’s polynomial Patients were selected from admission5 to a medical
equation, which uses the [H+], Pco2, and pro- intensive care unit, a surgical intensive care unit, and an
teins. The “equivalence” of proteins was as- emergency room over a I-month period. Table I lists the
sumed to be mainly due to albumin, as recently demographic information. principle diagnoses. and number
reported by Figge et al.‘j Finally, as will be of samples obtained per patient.

explained, the prediction of base excess from

the Siggaard-Anderson nomogram17 turned out Samples
to be a useful and simple means of assessing Samples in the intensive care units were drawn from
deviations from the predicted SID without hav- indwelling arterial catheters, sealed anaerobically and sent
simultaneously for blood gas, electrolyte, and protein mea-
ing to resort to Stewart’s complicated polyno-
surements; only the former were heparinized. Electrolyte
mial equation. This provides a simple bedside and protein measurements were performed on herum.
approach to acid-base balance, which is based Heparinized plasma was used for electrolyte ;Ind protein
on physical chemical principles. measurements in the 16 blood gas samples from the emer-

Table 1. Patient’s Characteristics

Anion
Gap
Age/Sex No.of Measured pH Range co2 HCO. BaseExcess
Pattent (Yd Determinants Diagnosis lntubated Range imEq/Ll (mmHg) (mEq!Ll (mEq/L)

1 72 F 1 Cardiogenic shock; pneumonia Yes 7.293 15.0 33 14 -8.7

2 54 M 1 Postoperative ACBP; noninsulin- Yes 7.537 3.0 25 22 0.6
dependent diabetes mellitus
3 63 M 3 Chronic obstructive pulmonary Yes 7.363-7.390 3.0-8.9 4749 24-29 3.7
disease
4 27M 1 Postoperative perforated colon No 7.434 4.0 40 28 0
5 78 F 15 Coronary artery dissection Yes 7.378-7.520 0.0-l .o 31-44 22-32 1.5.-10.8
6 55M 1 Lymphoma; pneumonia; chronic No 7.435 3.0 37 28 1.9
obstructive pulmonary
disease
7 26 M 2 Gasoline intoxication No 7.295-7.308 19.0-16.2 26-36 14-18 -74.-11.5
8 62 F 4 Sepsis; acute renal failure; No 7.335-7.430 4.0-1.0 37-43 23-25 -3.0-1.4
chronic respiratory acidosis
9 68 F 8 Septic syndrome; acute renal Yes 7.358-7.573 3.0-2.0 21-39 18-22 -6 l-I.0
failure; respiratory failure
IO 26 F 8 Perforated abscess; sepsis; Yes 7.392-7.561 3.0-10.3 37-45 28-33 2 8.-11
Crohn’s disease
II 33 M 2 Glomerulonephritis No 7.331-7.359 21.0 33 15-19 4.9-6.7
12 60 M 2 Pneumocystis pneumonia No 7.366-7.441 6.0-1.0 30-34 21 2.0-4.0
infection; sepsis
I3 70 M 1 Peptic ulcer disease No 7.409 9.0 41 25 1.7
I4 50 M 4 Respiratory failure; chronic Yes 7.443-7.505 3.0-7.0 34-38 23-27 0.7-5.1
hypoxia; arterio-venous
malformation
15 88 M 1 Congestive heart failure No 7.442 11.0 41 26 3.4
16 31 F I Anxiety No 7.445 9.0 33 22 0.3
I7 67 M 2 Cerebral hemorrhage Yes 7.394-7.435 6.0-6.0 38-42 19-21 1.7-2.0
18 56 M I Excision of lung mass Yes 7.267 7.0 57 29-37 1.1
19 40 F 1 Fever; arthritis; urinary tract No 7.439 7.0 33 0.4
infection; elevated liverfunc-
tion tests
20 85 M 2 Lower gastrointestinal hemor- No 7.355-7.366 6.0 33-36 51
rhage: noninsulin-dependent
diabetes mellitus; cystitis
21 63 F 12 Pneumonitis; acute respiratory Yes 7.329-7.526 2.0-6.0 37-60 1.5-12.4
distress syndrome
190 GILFIX. BIQUE, AND MAGDER

gency room. There were 57 sets of data on serum, and 16 Table 3. Steps for Assessing Acid-Base States
plasma samples from 21 patients for some patients had Is there an acidemia or alkalemia?
multiple samples over a period of days. Blood drawn for Is there a respiratory acidosis or alkalosis?
blood gas measurements was mixed with lithium heparin at Are there metabolic abnormalities in acid-base balance (ie,
100 U/L mL (final concentration), which did not add Nat, base excess or deficit)?
K+, Cl-, Ca2+, or HCOi- to the final measured value. In Free water effect = 0.3 (lNa+] - 140)
one of the two determinations from patient 7, electrolytes (Na+ in mEq/L)
and proteins measured from a venous sample were matched Corrected chloride (WC] = [Cl-] x 140/[Na+]
with the blood gas values drawn at the same time. (Cl- in mEq/L)
Blood gas analysis for pH, Pcoz, and base excess was Chloride effect = 102 - [Cl-c]
performed on a 1212 blood gas manager (General Instru- Albumin effect = (0.148 x pH ~ 0.818) (42 - [albumin])
ments Inc). Electrolytes (Na+, K+, Cl-, HC03-, Mg2+, (albumin in g/L)
Ca2+, phosphate), albumin, and total protein were mea- “Other species” = measured base excess -
sured on a SMAC analyzer (Technion Instrument Corp).
((a) + (b) + (c)) (in mEq/L)
The lactate in whole blood was determined on a model 23L
lactate analyzer (Yellow Spring Instrument Co, Inc).
[Na+], [K+], [Cl-], and [HCO3-] can be obtained from the
Data Analysis electrolyte measurements. [Atot] represents the equiva-
lence of the plasma proteins. Because [Atot] = [HA] +
Linear regression analysis with error analysis was per-
[A-] and HA = Ka [H+] [A-], [A-] can be derived from
formed by standard techniques. For calculation of the
[Atot] (grams per liter) using an estimated Ka for protein,
strong ion difference and anion gap and base excess gaps,
Ka = 3.0 x lo-‘, and the actual [H+]; [HA] is the
please refer to the Appendix and Tables 2 and 3.
undissociated protein, [A-] is the dissociated protein, and
[A-] equals [Atot]/(l + (lO-PH)/Ka).h Thus, as proposed
Theoretical Background by Jones, the SID can be calculated from [HCOj-] + [A-]
Relationship between measured and calculated strong ion and compared with the measured SID.6 If the two do not
diflerence. For each data set, the measured SID (SIDm) match, unmeasured ions (ie, SID gap) must be present.
was calculated as [Na+] + [K+] - [Cl-]; this gives the Figge et al concluded that [A-] is essentially due to albumin
inorganic SID. If [H+] and Pcoz and protein or albumin and used a complex polynomial equation to predict the ionic
concentration are known, SID can be calculated from the value of protein in solution.” However, in the biologically
formula developed by Stewart (Appendix)2,3 as modified by significant range of pH (ie, 6.80 to 7.80), the relationship of
Figge et al.15 This expression is derived from the dissocia- ionized albumin to total as a function of the pH is essentially
tion equations in a solution of strong ions, CO2, proteins, linear, such that (A-] = (0.148 x pH - 0.818) (albumin;
and other weak acids, and the conservation of mass and grams per liter) (see Fig 4 of Figge et alIs). We used this
charge. We have called this term SIDc or, more specifically, simplified term to assess the ionic contribution of albumin.
SIDc(H+,alb) when albumin is used in the calculation and Stewart used plasma values of electrolytes and total
SIDc(H+,TP) when total protein is used. SIDc(H+, alb) was protein. Because this is not practical in the routine clinical
calculated using the mathematical model developed by situation, and we wanted to examine the use of this
Figge et al,15 which regards albumin as the only significant approach in a clinical setting, 57 of the 73 electrolyte and
protein present (for details of derivation, see reference 3). protein values were obtained from serum. This introduces
We have also called the difference between the measured only a small error in the calculations. Plasma values are
and calculated SID the SID gap. greater than serum values by 0.9% for calcium, 0.2% for
By the law of conservation of mass and charge, [Na’] + chloride, and 4.0% for total protein.‘” Plasma values are less
[K+] - [Cl-] must equal [HC03-] + [A-], where [A-] is the than serum values by 1.3% for albumin, 1.8% for bicarbon-
equivalent of weak acids, which are mainly proteins, assum- ate, 7.0% for phosphorus, 8.4% for potassium, and 0.1% for
ing that all other ion species are present in negligible quantities. sodium.lx

Table 2. Definitions

Abbreviation

SID Strong ion difference

SlDm Measured SID = [Na+] + [K+] - [WI
SlDc Calculated SID from Appendix (Stewart*)
SIDc(H+,TP) SID calculated using total protein
SIDc(H+,alb) SID calculated using albumin
Anion gap [Na+] - [Cl-] - (HCO,-]
Caz+ (ionized) In mEq/L; 2 x (0.45 x (Ca2+] in mmol/L)
Pod”- Kl x (total Pi] 2 x K2 3 x K3 x K2
In mEq/L; I+--+
W+l + Kl) ([H+] + K2) (IH’I + K3)(lH+l + K2)
(Kl = 7.017 x 10-a; K2 = 7.99 x lo-*; K3 = 4.8 x lO-8 (251)
ACID-BASE BALANCE 191

We nexl tried to analyze the contributions of various and Na+ into red cells and consequent changes in the
type% of abnormalities to the SID gap. To do so, we used the plasma SID.” By identifying the effects of changes in [Na +J,
base-excess measurement developed by Sigaard-Anderson. [Cl-]. and albumin concentration, we account for differ-
This value is the amount of titratable acid that is required to ences from the reference values for NaT, Cl-. and proteins.
account for an acid-base change at a Pcoz of 40 mm Hg. He If there is still a residual base excess or deficit. the presence
ohtained this value from whole blood samples rather than of unmeasured basic or acidic elements is indicated. We
plasma or serum, which is the basis of Stewart’s analysis. tested the effectiveness of this assessment hy comparing this
However, only a small error is introduced by exchanging estimate of unmeasured elements with deviations in the
plasma for blood because the “buffering” effect of the measured SID from the SID predicted from Stewart’s
interstitial Huid has a far greater effect than red cells.?~‘” formula, ie, SIDc(H’). Because [Na’], [Cl 1, and protein
If the blood has a normal composition, the base excess are included in the calculated SID. any deviation must
should he zero. Any deviation in [Na’]. [Cl-], or proteins represent abnormal elements.
from normal levels will produce a base excess or deficit. This Dejinitiom of referwm vulues. We first classified all
allow\ an account of the “acidifying” or “alkalinizing” samples based on the standard approach. This required
equivalence. ie, base deficit or excess due to changes in free setting some arbitrary limits on reference values. We
water. [Cl-], and proteins from the reference values as defined a normal pH as 2 7.36 or _i 7.44. P(Y): as z 36 and
follows. Because [Na+] is the regulated variable that con- 244 mm Hg, and [HCO?-] as 222 or 526 mEq/L.
trols the volume of the body. we used [Na+] to assess the Specimens were then classified as acidotic if pH was I 7.36
effects of dilution. If one assumes that normal [Na+], [K+], and alkalotic if z 7.44. As in the standard nomenclature. if
and [Cl ] are the predominant strong ions and that [Na+] is the acidosis was associated with an elevated Pc‘o:, we called
I40 mEq/L. [K-l is 4 mEq/L. and [Cl-] is 102 mEq/L, then this a respiratory acidosis and if associated with a decrease
the normal SlD is 42 mEq/L. Deviations of the measured in [HCO? -1, a metabolic acidosis. Similarly. alkaloscs were
[Na. ] from the normal reference value for [Na+] of 140 classified as respiratory if the Pco: was decreased and as
mEq L can then be used to calculate the change in base metabolic alkaloses if the [HCO?~] was increased. Meta-
excess due to changes in free water by the following: bolic acidoses were then classified as those associated with
an increased anion gap if the anion gap was 2 I2 and a
hasc rlfect due to change in free water. normal anion gap if the anion gap was I 12 mEqiL.?” If the
pH was normal and Pcoz and HCOT were abnormal, or it
mEq/L = (71 x 42 = 0.3 ([Na+] - 140) both Pco: and HCOJ- were abnormal hut in the same
direction, the abnormality was called mixed.
Note that even if the SID is in the range of 35 to 45, the The steps used in the new approach are summarized in
effect on the contribution due to free water will be small. Table 3. We first classified all specimens as acidemic if the
The effect of changes in [Cl-] can be obtained by first pH < 7.36 or alkalemic > 7.44. We defined the specimen as
correcting the chloride concentration for the changes al- showing a respiratory acidosis if Pc.02 was :> -44 mm Hg and
ready accounted for due to changes in free water: alkalosis if P(‘o: was < 36 mm Hg, whether or not there was
an acidemia or alkalemia. We then determined the hasc
[Cl cJ* mEq/L = [CT] x 140/[Na+] excess due to water effects (ie, [Na’]), [Cl ] effects. and
albumin effects. as explained above. Finally. we calculated
then: the effects of “other species” from the difference hetween
the base excess from these three factors and the measured
base effect due to [Cl 1, mEqiL = 102 - [Cl-c]
base excess. We call this the base excess gap.
where 102 is the reference value for chloride.
Based on the previously noted analysis of Figge et al,” the RESULTS
effect of changes in total protein albumin can be calculated
Based on the standard nomenclature, 7
as: base effect due to albumin.
samples were classified as within the reference
mEq:L = (0.148 x pH - 0.818) (45 - albumin, g/L]). range. There were 9 samples classified as acido-
The sum of the above three equations can be positive ses; 2 were respiratory and 7 were metabolic
(base excess) or negative (base deficit). If no unusual (5 of the metabolic determinations had an
cations or anions are present in the blood, the sum of the increased anion gap). There were 27 samples
three equations should equal the measured base excess,
classified as alkalosis; 16 were respiratory and
because this is the amount of acid or base that the
Siggaard-Anderson nomogram predicts must be added to 11 were metabolic. The remaining 30 samples
blood with the PCO~ kept at 40 mm Hg to restore the pH to were classified as mixed. When classified on the
7.40. His analysis assumed normal electrolyte and protein basis of pH using our suggested approach, 31
concentrations. although this was not stated in the work. were within the reference range, 11 were acide-
because Stewart’s model predicts that variations in the mic, and 31 were alkalemic. There was no
electrolyte or protein concentrations will result in different
values. Indeed, Siggaard-Anderson had to revise his nomo-
respiratory disturbance in 31 determinations, a
gram, because he initially added NaF, which changed the respiratory acidosis in 14, and a respiratory
acid-base values presumably due to different shifts of F- alkalosis in 28.
192 GILFIX, BIQUE, AND MAGDER

The effects of base excess measurements due

to deviations in HzO, Cl-, total protein, and
“others species” are shown in Fig 1. The base
effects from changes in free water varied from
-3.6 to 3.0 mEq/L, [Cl-] from -14.8 to 11.9
mEq/L, albumin from -0.3 to 7.0 mEq/L, and
the “others species” from - 11.9 to 9.3 mEq/L.
SID (SIDc) could be estimated from the full 25 00
00
equation of Stewart” (Appendix) by using either
20 I
total protein equivalent@ or albumin equiva- 20 25 30 35 40 45 50 55 60
lents,15 the [H+], Pcoz, and electroneutrality SlDnl ,meq,L)

principle, which states that the sum of positive

and negative ions is zero. This calculated value
frequently differed from the inorganic SID
(SIDm) from the measured [Na+], [K+], and
[Cl-] (Fig 2). This indicates that other ionic
species were not accounted for. Both relations
also had positive y intercepts so that the SID
calculated was systematically greater than the
SID measured from [Na+], [K-J, and [Cl-]. This
indicates that some negative ions were not
accounted for in the simple “measured” SID.
Figge et ali5 showed that serum globulins play
a negligible role in acid-base equilibria, and it is
albumin that contributes the ionic equivalent of
serum proteins. The SIDc calculated using ei-
ther total protein equivalent@ or albumin equiva-
lent@ are strongly related, although there is a
systematic difference because of the difference
in calculating the ionic effect of protein (Fig 3).
For the rest of this analysis, unless noted other-
wise, albumin was used in the calculation of
SIDc, ie, SIDc(H+, alb), and base excess.
In Fig 4, the estimated base-excess due to
“other species” factors or the base-excess gap is
Fig 2. A plot of SlDc calculated from Stewart’s formula
based either on total protein or albumin as a function of SIDm.
See text for details.(O) For SlDc based on total protein, n = 65
and y = 12.7 + 0.66SIDm, r = .76.(O) For SlDc based on
albumin, n = 65 and y = 7.71 + 0.72SIDm. I = .76.
plotted against the SID gap. Recall that this was
obtained from the base excess calculated from
the blood gases and the sum of the base-excess
due to changes in free water, [Cl-], and protein
albumin. As can be seen, there was a tight
relationship between the two measurements
(u = .93) so that the base excess of “others
species” provides a useful and simple guide to
the presence of unmeasured anions.
In Fig 5, the standard anion gap is plotted as a
function of the SID gap. Similar to the relation-
ship between the base excess gap and SID gap,
there was a close relationship (r = .80) between
the SID gap and the anion gap. However, the
relationship was not as tight as with the base
excess gap.
DISCUSSION
A complete quantitative physical chemical
analysis of human blood is not possible because

Fig 1. A plot of the ranges of values of milliequivalent

effects from changes in free water, chloride, albumin, and Fig 3. A plot of SlDc calculated using total protein, SIDc(H+,
other species on base excess or deficit for all patients. The TP), as a function of SlDc calculated using albumin, SIDc(H+,
bars represent the mean values. alb). For n = 65, y = 6.86 + 0.86x. I = .g6.
ACID-BASE BALANCE
Fig 4. A plot of base excess gap as a function of SIDm-
SIDc(H+,alb) (ie, SID gap). The SID gap was very well predicted
by the base excess gap. For n = 62, y = 4.64 - 0.94x. r = .93.
of the complexity of the components of the
solution. Thus, it is not surprising that the SID
predicted by Stewart’s formula differed from
the measured value in most of our subjects
whom, by and large, had disorders of [H+]
homeostasis. Moreover, in the clinical setting,
the concentrations of all clinically relevant ionic
species are not always available. Furthermore,
the calculations are clumsier than the tradi-
tional Pco,/HC03- maps. Why, then, introduce
this approach to the analysis of variations in
[H+]? To begin with, the traditional analysis
treats [HC03-] and [H+] as independent vari-
ables when, in fact, they are actually dependent
variables in blood. By use of the base excess or
deficit, this approach allows a quantification of
the contribution of HzO, SID, proteins, and
other species to the metabolic acid base distur-
bance. The profiles that we have shown in Fig 1
can then be used to characterize the contribu-
25
Fig 5. A plot of anion gap as a function SIDm-SIDc(H+, alb)
(ie, SID gap). The SID gap was well predicted by the anion gap
but not as well es with the base excess gap. For n = 65, y =
5.11 + 0.65~. r = .60.
193
tions of the various components to the acid base
disturbance. Furthermore, the analysis reveals
that not only can there be unmeasured anions as
identified in a wide anion gap acidosis.1H but
there are also frequently unmeasured basic
elements that can influence [H+]. These have
been noted by others 21,22but are not obvious in
the usual analysis of acid base disturbances.
A number of technical factors of our study
must be discussed next. Because small differ-
ences in the concentrations of electrolytes can
have significant effects on the conclusions, great
care was taken in handling the samples and
ensuring that there was no delay in their process-
ing. Except for two determinations, all samples
were taken from an arterial source so that the
blood gas analysis and electrolytes reflected the
same environment. Stewart’s original analysis
was on plasma,’ but routine measurements are
on serum. Therefore, most of our determina-
tions were on serum, but as discussed under the
Materials and Methods section. this should
have only introduced a small error.
Another source of error is the variation in
concentrations of other strong electrolytes. The
major ones are [Mg”+] and [Ca2+]. but their
normal concentrations, even under extreme
pathophysiological conditions, are not large.
Total calcium concentrations were available for
60 of the determinations, but unfortunately, we
did not have measured ionized calcium. An
estimate of ionized calcium as well ah the
ionized phosphates resulted in an average
change in the calculated SID gap of less than 2
mEq/L. In renal failure, a large number of
organic acids can accumulate, but they are in
the order of 10mh mEq/L.“? The calculated SID
relationship was systematically less than the
simple measured SID. This indicates the pres-
ence of unmeasured negative ions. Similarly.
Figge et al could only account for 12 mEq/L of
the normal anion gap of 17 mEq/L,.” An as-
sumption in our work is that this is the baseline
condition, and we can assess deviations from it
in the pathological state.
The major limitation to the use of our ap-
proach is the lack of an accurate assessment of
the ionic equivalence of weak acids of which the
major components are proteins. It is most prob-
able that the dissociation constant for this
relationship varies with changes in the frac-
194
tional distribution of albumin to globulin4 the
presence of unusual proteins,14 and possibly
even the pH and PcoZ. l5 Therefore, the group
that we call “other species” could presumably
include changes in the behavior of proteins as
well as the presence of unmeasured ions.
A weak acid that contributes to the A- term
is phosphate. Normally, the concentration is too
low for it to play a major role. However, in renal
failure or rhabdomyolysis, phosphate levels can
be markedly elevated. One of our patients had a
serum phosphate concentration of 2.94 mmol/L.
Based on the dissociation constants of H3P04
(K1 = 7.017 x lo-‘, K2 = 7.99 x lo-*, Kj =
4.8 x 1O-8), the phosphate normality can be
calculated at 9.5 mEq/L.23 In this patient, the
SID gap was 18.3 mEq/L, and therefore, al-
though phosphate was important, something
else still had to be present.
Lactate and ketones did not explain a signifi-
cant proportion of the SID gap in our patients,
even those with a significant metabolic acidosis
from shock. This has also been observed by
others.24,25 In one patient with a serum lactate
concentration of 3.7 mEq/L, the SID gap was
actually positive indicating the presence of un-
measured basic elements. Of note, the anion
gap was also not increased in this patient, and
the pH was alkalemic. Unmeasured anions have
also been observed by Rossing et a1.14
Our results contrast with those of Kowal-
chack et al who applied a physical chemical
approach to the analysis of the acid base bal-
ance during exercise.‘“,” They found that SID
gap changes during exercise could all be ac-
counted for by an increase in lactate concentra-
tions. This can probably be explained by the
much more complex interaction of protein
breakdown, intravenous alimentation, and
changing renal function in patients in intensive
care units compared with normal exercising
subjects.
Jones has published a number of articles
advocating the physical chemical approach to
acid-base analysis6,” He suggested using the
principle of electrical neutrality to predict that
SID - HC03- - A- = 0, where A- is the
ionized component of proteins. A deviation of
the solution of this equation from 0 indicates
the presence of unmeasured anions (or cations
when the value is positive). We calculated this
GILFIX, BIQUE, AND MAGDER
value for all patients and called this difference
an SID gap because it represents the difference
between the SID predicted from the [HCOX-]
and A- and the measured value. We also
calculated an SID gap by inserting the [H+] and
Pco2 values in the equation developed by Stew-
art? This equation takes into account the
effect of changes in PcoZ, and therefore, the
value was systematically different from the sim-
pler calculation of Jones. However, Stewart’s
equation is cumbersome and requires a com-
puter to solve it. We were pleased, therefore, to
observe that the SID gap calculated from
Stewart’s formula was predicted very well by the
measurement of a contribution of “other
species” to the measured base excess. This
could be considered the base-excess gap. Thus,
the standard clinical values can provide a bed-
side estimate of unmeasured anions and cations
and obviate the need to resort to a computer for
each measurement.
Our SID gap also correlated well with the
traditional anion gap, although not as closely as
with the base-excess gap. One explanation is
that the anion gap does not take into account
the changes in protein concentration, although
this can be done, nor the effects of changes in
free water.
To understand why the base excess of “other
species” correlates so well with the SID gap, it is
necessary to review the meaning of base excess.
This term was introduced by Siggaard-Ander-
son” to account for deviations of pH from 7.40,
when Pco2 was controlled at 40 mm Hg. It was
obtained by adding known amounts of base or
acid to the blood to restore the pH to 7.40.
Although the electrolyte and protein concentra-
tions were not given, presumably they were
within the standard reference range. However,
any deviations of these electrolytes or proteins
from normal would have altered the results.
This is essentially what our approach does. It
removes the contribution of changes in [Na’],
[Cl-], and protein from their reference values,
and the remaining difference should corre-
spond to the contribution of acidic or basic
elements as observed by Siggaard-Anderson.
The tightness of the relationship suggests that
the contribution of these elements behave quite
consistently and, therefore, makes their assess-
ment possible. It needs to be emphasized that
ACID-BASE BALANCE 195

A o 1 4 6 DRYjvmb,, 10 12 14 16 -lo4
B o 2 4 6 Day ;..,., IC IP .i IF

Fig 6. Trend values over time of acid-base values in a ventilator-dependent patient following aortocoronary bypass procedure.
(A)., pH; 9, Pco2. (B)O, free H,O; El, Cl-; 0, albumin; x, other species. See text for discussion.

unmeasured basic elements were identified as
frequently as unmeasured acidic elements.
In our patients, changes in free water played
a small role in the [H+] homeostasis. Even a
[Na’] of 110 mEq/L would only have produced
an acidifying effect of -9 mEq/L, and none of
the patients had [Nat] in this range. The albu-
min concentrations were generally low, which is
to be expected in patients in the intensive care
unit. This produced a mild to moderate alkalin-
izing effect.4.rJ In most of our patients, a change
in [Cl-] was the major factor affecting acid-base
balance. The large effects of the swings in [Cl-],
as seen in the patients shown in Figs 6 and 7,
would not have been obvious in the traditional
analysis and point to important renal control
mechanisms. Finally, the contribution of “other
species” was of moderate importance. It is of
interest that -8.2 mEq/L of unmeasured an-
ions became evident in the example shown in
Fig 8 just before the patient’s death, even
though the lactate concentration was only 2.3
mEq/L. This indicates that some other factors
in the blood produced this acidifying effect. As
noted above, this has previously been observed
by Rossing et alI4 in patients with cirrhosis and
Mecher et al24 in septic patients.
Specific Examples
We have plotted the values for contributions
to base excess or deficit from free water, [Cl-],
proteins, and other factors against time as well
as the pH and PcoZ for 3 patients (Figs 6
through 8) in whom we had serial values.
The first patient (Fig 6) was ventilator depen-
dent following an aortocoronary bypass proce-
dure. Changes in free water had no major
effects. She started off with a marked acidifying
effect from [Cl-] that gradually normalized and
was counteracted by an initially high unmea-
sured cation that gradually decreased. She then
again developed an abrupt increase in [Cl-],
which produced a decrease in pH because it was
not “buffered” by unmeasured cations. This
occurred during a bout of diarrhea. As the ]Cl-]

L,

A
Fig 7. Trend values over time of acid-base values in a patient recovering from septic shock. (A)., pH; L-1, Pco,. fB)O, free H,O; ;L,
Cl--; /-I, albumin; x, other species. See text for discussion.
196 GILFIX, BIQUE, AND MAGDER

7 400

Fig 8. Values over time of a patient with acute respiratory distress syndrome and septic shock who died just after the last
measurement. (A)., pH; Cl, Pco,. (B)O, free H20; 0, Cl-; 0, albumin; x, other species. See text for discussion.

decreased, she developed a metabolic alkalosis, APPENDIX

and the pH drifted back to normal. Stewart’s [H+] equation for isolated plasma
The second patient started in septic shock using total protein3:
(Fig 7). He also started with a marked acidifying
effect of [Cl-], which gradually normalized. SID + 1000 x [H+] - 1000 x (Kc x Pco&
Because it was balanced by reciprocal changes
in unmeasured “other species,” the pH was [H+] - 1000 x (Ka x [Atot])/(Ka + [H’])
within the reference range during this period.
He then had a transient period of hyperventila- - 1000 x (K3 x Kc x Pco~)/[H+]~
tion, which produced an increase in pH without
any change in the metabolic factors. - 1000 x K’w/[H+] = 0
The third patient (Fig 8) suffered from acute
respiratory distress syndrome and septic syn- Stewart’s [H+] equation modified for isolated
drome. Despite having an initial blood lactate plasma using albumin using the estimate of the
level of 3.7 mEq/L, she was alkalotic. This was ionic equivalence?
primarily due to hypoproteinemia,14Js,1c with a SID + 1000 x [H+] - 1000 x (Kc x Pco2)/
small contribution from unmeasured basic ele-
ments and hypochloremia. At one point she had [H+] - 1000 x (K3 x Kc x Pco,)/[H+]’
a marked increase in [Cl-,] which was not
countered by the other factors. There was - 1000 x K’w/[H+] - ((0.148 x pH - 0.818)
partial compensation from increased respira-
tion but not complete, so that the pH decreased x [albumin, g/L])=0
considerably at one point. During the terminal
Abbreviations used in these equations:
period, [Cl-] decreased, and unmeasured an-
ions increased to -8.2 mEq/L, even though the SID Strong ion difference (mEq/L)
measured lactate was only 2.3 mEq/L. Ka Equilibrium dissociation constant for
In summary, the calculated base-excess or weak acids (plasma proteins)
deficit can be used to apply Stewart’s physical 3.0 x 1O-7 (Es/L)*
chemical approach to the clinical setting. This Kc Equilibrium dissociation constant for
allows contributions from deviations from nor- carbonic acid
mal of free water, SID, proteins, or other 2.58 x 10-l’ (Eq/L)2/mm Hg
species to be quantitated. This approach reveals K’w Equilibrium dissociation constant for
that acute changes in [Cl-] are a major contribu- water at 37°C
tor to changes in [H+] homeostasis, whereas 4.4 x lo-l4 (assuming [H,O] is constant)
changes in “other factors” occur more slowly. @q/L)’
Furthermore, unmeasured basic elements are K3 Equilibrium dissociation constant for
frequently present in the blood of critically ill bicarbonate
patients. 6.0 x 10-l’ (Eq/L)
ACID-BASE BALANCE 197

[H+] Hydrogen ion is (Eq/L) proteins, and the principle of electrical neutral-
PC@ Partial pressure of CO2 (mm Hg) ity.
[Atot] Concentration of proteins in Eq/L; 1 g/L ACKNOWLEDGMENT
= 2.4 x 10-j Eq/L The basic idea of using the Sigaard-Anderson concept of
This eauation with four unknown variables.
I
base excess to quantitate the various components of the
blood comes from unpublished work by Dr. V. Fencl.
ie, [H+], [Pco,], [Atot], and SID is derived from However, our interpretation does not necessarily coincide
the six equations that represent the dissociation with all of his ideas. His published work is represented by
of water, carbonic acid, bicarbonate, carbonate, references 4. 1.5. 16, and 17.

REFERENCES
1. Cohen JJ. Kassirer JP: Acid-Base. Boston, MA, Little, 13. Jones NL: Should we change our approach to acid-
Brown and Co, 1982 base physiology. Ann RSPSC 23:235-240, I990
2. Stewart PA: How to Understand Acid-Base. A Quanti- 14. Rossing TH. Boixeda D, Maffeo N, et al: Hypetventi-
tative Acid-Base Primer for Biology and Medicine. New lation with hypoproteinemia. J Lab Clin Med I 12:553-559.
York. NY. Elsevier North Holland, 1981 1986
3. Stewart PA: Modern quantitative acid-base chemistry. 1.5. Figge J. Rossing TH. Fencl V: The role of serum
Can J Physiol Pharmacol61:1444-1461, 1983 proteins in acid-base equilibria. J Lab Clin Med 117:453-
467, 1991
4. Rossing TH. Maffeo N, Fencl V: Acid-base effects of
16. Fencl V, Rossing TH: Acid-base disorders in crltical
altering plasma protein concentration in human blood in
care medicine. Ann Rev Med 40:17-29, 1989
vitro. J Appl Physiol 61:2260-2265, 1986
17. Siggaard-Andersen 0: The Acid-Base Status of the
5. McAuliffe JJ. Lind LJ, Leith DE, et al: Hypoprotein- Blood (ed 2). Baltimore. MD, Williams and Wilkins, 1964
emit alkalosis. Am J Med X1:86-90, 1986 18. Teitz NW: Textbook of Clinical Chemistry. Philadel-
6. Jones NL: A quantitative physiochemical approach to phia, PA, W.B. Saunders. lY86
acid-base physiology. Clin Biochem 23:189-195, 1990 19. Lindinger MI. Heigenhauser JF. Jones NL: Acid-
7. Forster IIV, Murphy CL, Brice AG, et al: Plasma [H’] base and respiratory properties of a buffered bovine erythro-
regulation and whole blood [CO11 in exercising ponies. J cyte perfusion medium. Can J Physiol Pharmacol 64:550-
Appl Physiol 6X( 1):309-31.5. 1990 555. 1985
X. Anderson JW. Jennings DB: Osmolality, NaCl dietary 20. Emmett M. Narins RG: Clinical use of the anion gap.
intake. and regulation of ventilation by CO2. Am J Physiol Medicine 56:38-54, 1977
25S:RlO5-RlI2, 19X8 21. Leverve XM, Guignier: Hyperlactataemia: acidosis
or alkalosis? J Clin Chem Clin Biochem 24273. 19X6
9. Forster I-IV. Murphy CL, Brice AG, et al: In vivo
22. Niwa T: Metabolic profiling with gas chromatography-
regulation of plasma [H*] in ponies during acute changes in
mass spectrometry and its application to clinical medicine. J
PCO2. J Appl Physiol68:316-321, 1990
Chromatogr 379:313-345. 1986
II). Kowalchuk JM, Heigenhauser GJF, Lindinger MI. et 23. Windholz M: The Merck Index. Rahway, NJ, Merck
al: Role of lungs and inactive muscle in acid-base control and Co, Inc. 1976
after maximal exercise. J Appl Physiol65:2090-2096, 1981 24. Mecher C, Rackow EC. Astiz ME. et al: Ilnac-
Il. Kow’alchuk JM. Heigenhauser GJF. Lindinger Ml, et counted for anion in metabolic acidosis during severt’ sepsis
al: Factors influencing hydrogen ion concentration in muscle in humans. Crit Care Med 19:705-71 I. I9Y I
after intense exercise. J Appl Physiol65:2080-2089, 1981 25. Wang F. Butler T, Rabbani GH, et al: The acidosis of
12. Anderson JW. Jennings DB: H+ homeostasis, osmo- cholera: contributions of hyperproteinemia, lactic acidemia
lality and body temperature during controlled NaCl and and hyperphosphatemia to an increased serum anion gap. N
Hz0 intake. Am J Physiol 2SS:R97-R105, 1986 Engl J Med 315:1591-1SYS. 1Y86