Feature Articles

Hypercapnic acidosis and mortality in acute lung injury*

David A. Kregenow, MD; Gordon D. Rubenfeld, MD; Leonard D. Hudson, MD; Erik R. Swenson, MD

Objective: We tested the hypothesis that hypercapnic acidosis

is associated with reduced mortality rate in patients with acute
lung injury independent of changes in mechanical ventilation.
Design: Secondary analysis of randomized clinical trial data
using hypothesis-driven multivariate logistic regression.
Setting: Randomized, multiple-center trial (n 861) comparing 12 mL/kg to 6 mL/kg predicted body weight tidal volumes
previously published by the National Institutes of Health Acute
Respiratory Distress Syndrome (ARDS) Network.
Patients: Acute lung injury patients enrolled in a randomized,
multiple-center trial (n 861).
Interventions: None.
Measurements and Main Results: The adjusted odds ratio and
95% confidence intervals (CI) for 28-day mortality rate associated
with hypercapnic acidosis defined as day 1 pH <7.35 and PaCO2
>45 mm Hg were 0.14 (95% CI 0.03 0.70, p .016) in the 12
mL/kg predicted body weight tidal volume group and 1.18 (95% CI
0.59 2.35, p .639) in the 6 mL/kg predicted body weight tidal

ypercapnic acidosis (HA),

more commonly referred to
as permissive hypercapnia,
has been viewed as an acceptable side effect of lung-protective
ventilation that can be tolerated in an
effort to avoid ventilator-associated lung
injury (1 6). Evidence is growing, however, that HA has favorable anti-inflammatory and antioxidative effects at the
subcellular, cellular, whole organ, and
whole organism levels in hypoxic and inflammatory conditions (715). This has
led to speculation that HA may provide
additional benefit to patients with a variety of injuries and illnesses (16 18). To

*See also p. 229.

From the Division of Pulmonary and Critical Care
Medicine, Department of Medicine, University of Washington, Seattle, WA (DAK, GDR, LDH, ERS); and the
Section of Pulmonary and Critical Care, Department of
Medicine, Virginia Mason Medical Center, Seattle, WA
(DAK).
Supported, in part, by grants 5 T32 HL07287-24,
1 F32 HL070510-01, NIH HL 24163, and ARDS Net
Contract No. NO1 HR46055.
None of the authors has any financial interests to
disclose.
Copyright 2005 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000194533.75481.03

Crit Care Med 2006 Vol. 34, No. 1

volume group. Other definitions of hypercapnic acidosis spanning

a range of magnitudes suggest a dose-response association
between hypercapnic acidosis and 28-day mortality in the 12
mL/kg predicted body weight tidal volume group. None of our
definitions of hypercapnic acidosis were associated with reduction in 28-day mortality in the 6 mL/kg predicted body weight tidal
volume group.
Conclusions: Hypercapnic acidosis was associated with reduced 28-day mortality in the 12 mL/kg predicted body weight
tidal volume group after controlling for comorbidities and severity
of lung injury. These results are consistent with a protective effect
of hypercapnic acidosis against ventilator-associated lung injury
that was not found when the further ongoing injury was reduced
by 6 mL/kg predicted body weight tidal volumes. (Crit Care Med
2006; 34:17)
KEY WORDS: hypercapnia; acidosis; respiratory; respiratory distress syndrome; adult; respiration; artificial; hypoventilation; mortality

date the only data available on HA in

humans with acute lung injury (ALI)
have been limited to trials of lungprotective ventilation in which the effects
of HA have not been separated from the
effects of changes in mechanical ventilation.
In 2000, the National Institutes of
Health Acute Respiratory Distress Syndrome (ARDS) Network published the
largest multiple-center randomized trial
of patients with ALI (6). The ARDS Network trial compared 12 vs. 6 mL/kg predicted body weight tidal volumes and
demonstrated a 9% absolute and a 22%
relative reduction in 28-day mortality
rate through the use of 6 mL/kg predicted
body weight tidal volumes. Therefore, we
used data from the ARDS Network trial to
separate the effect of HA from lungprotective ventilation in patients with
ALI.

MATERIALS AND METHODS

Study Design. Secondary analysis of randomized clinical trial data.
Patients and Variables. The details of the
inclusion and exclusion criteria for the ARDS
Network clinical trial on which this analysis is
based have been published elsewhere (6). Pa-

tients were enrolled from March 1996 to

March 1999 at ten university centers. The enrolled patients met the American-European
Consensus definition of ALI and ARDS including acute hypoxic respiratory failure requiring
mechanical ventilation, bilateral patchy infiltrates on chest radiograph, and no evidence of
left atrial hypertension. The ARDS Network
trial compared mechanical ventilation using
volume-cycled assist control with a tidal volume of 12 mL/kg predicted body weight vs. 6
mL/kg predicted body weight. Hypercapnia
was not a goal of the ventilator strategy in this
study and was not a randomized intervention.
Minute ventilation was used to keep 7.30
arterial pH 7.45. At a maximum respiratory
rate of 35 and pH 7.15, management of
acidosis with sodium bicarbonate was at the
discretion of the investigator.
The ARDS Network and the Human Subjects Divisions of the University of Washington
and the VA Puget Sound Health Care System
approved this study. Unless stated otherwise,
mortality refers to 28-day mortality rate and
tidal volume in mL/kg refers to mL/kg predicted body weight as calculated in the parent
clinical trial. Partial pressures of oxygen were
corrected for altitude as in the parent clinical
trial.
Based on the experimental evidence (19
21), we examined an early (or acute) HA along
with other data available early in a patients
course of treatment. We defined HA based on

Figure 1. Day 1 pH and PaCO2 values of the study population. HA, hypercapnic acidosis; RA, respiratory
alkalosis; PBW, predicted body weight.

arterial blood gas data collected on the morning following randomization and institution of
mechanical ventilation according to the study
protocol (day 1). Although they reflect a single
point in time, day 1 data have the advantage of
being the most complete set of blood gas values and avoid the survival bias of requiring
serial measurements to define HA. The data
collection protocol provided day 1 blood gas
values from 7 hrs to as much as 31 hrs after
institution of protocolized mechanical ventilation. The day 1 blood gas values and definitions of HA and respiratory alkalosis are
shown in Figure 1. Because there is no consensus definition of therapeutic HA, our analysis plan included assessing the effect of different degrees of HA. These were defined a
priori. HA1 was defined as a pH 7.40 and
PaCO2 40 mm to assess the effect of any
degree of HA. HA2 was defined as a pH 7.35
and PaCO2 45 mm Hg to exclude patients
whose blood gas values would otherwise have
been considered within normal limits and to
assess a more severe degree of HA. There were
too few patients with greater extremes of respiratory acidosis (e.g., pH 7.30 and PaCO2
50 mm Hg) to perform reliable analyses.

Finally, we hypothesized a priori that extremes of acidosis, even if associated with elevated PaCO2, would indicate significant metabolic acidosis, comorbidities, and/or probable
inevitable irreversibility of the underlying disease. These patients would likely have such
overwhelming cellular dysfunction that HA
would not afford protection (10) or would be
poorly tolerated. Furthermore, patients with a
pH 7.15 may have been given sodium bicarbonate as part of the trial at the discretion of
the providers, but this information was not
recorded. Thus, a third definition of HA (HA3)
with a lower limit of pH and upper limit of
PaCO2 consistent with a simple but moderately
severe acute respiratory acidosis was examined
(7.15 pH 7.35 and 45 mm Hg PaCO2
65 mm Hg). Respiratory alkalosis was defined
as pH 7.45 and PaCO2 35 mm Hg. Again,
there were too few patients with more extreme
respiratory alkalosis (e.g., pH 7.50 and PaCO2
30 mm Hg) to perform reliable multivariate
analysis.
Since no direct measure of a metabolic
acidosis was recorded for the majority of patients on day 1, arterial HCO3 values were
calculated from the arterial pH and PaCO2 val-

ues based on the Henderson-Hasselbalch

equation. Calculated arterial HCO3 values
were compared with measured venous HCO3
values in those patients with serum chemistries recorded on day 1 using the method of
Bland and Altman (22). Base excess was then
calculated according to the formula base excess (HCO3 10 [pH 7.40]) 24.
Statistical Methods and Modeling Strategy. The primary research question was to
assess the independent effect of HA on mortality rate. Because HA was not a randomized
intervention, patients who achieved HA might
differ in many ways from those who did not
achieve HA. To account for these factors and
assess the independent effect of HA, we performed hypothesis-driven multivariate logistic
regression including confounding variables
associated with both HA and mortality. To do
this we built a logistic regression model accounting for the patients severity of illness
and the severity of his or her ALI and analyzed
each tidal volume group separately. We used
Acute Physiology and Chronic Health Evaluation (APACHE) III, a validated intensive care
unit severity of illness measure, to measure
the severity of critical illness and account for
age and comorbid illness (23). We used primary risk factor for ALI, a known predictor of
outcome in this patient population, to account
for different mechanisms of ALI. We used day
1 PaO2/FIO2 and plateau pressure (Pplat) to
measure the severity of ALI. Bivariate analyses
of these latter variables confirmed that they
were related similarly to mortality and HA and
did not identify other important confounders.
The final logistic regression model predicted
mortality as a function of APACHE III score,
risk factor for ALI, PaO2/FIO2, and Pplat. To
answer the research question, we added the
different definitions of HA as a binary covariate to this model.
Several secondary analyses were performed. To assess the effect of gender, tidal
volume, and positive end-expiratory pressure
(PEEP) on day 1, these variables were added to
the final model separately and together. Confounding by metabolic acidosis was examined
by adding base excess to the model to determine whether metabolic acidosis affected the
association between HA and mortality. To confirm the finding that HA acts differently depending on whether patients received lungprotective ventilation, we tested an additional
model that included all patients plus tidal volume randomization and an interaction term
between HA and tidal volume.
Statistical analyses were performed using
STATA version 7.0 (Stata Corporation, College
Station TX). Results are reported as mean
SD unless otherwise indicated. Where indicated
in the tables, continuous variables were compared with the two-sample Students t-test
with equal variances, and categorical variables
were compared with the Pearson chi-square
test for independence.

Crit Care Med 2006 Vol. 34, No. 1

Table 1. Characteristics of the study population (6)

Variable

12 mL/kg

6 mL/kg

No.
Age, years
Female
APACHE III score
Risk factor for ALI, n (%)
Trauma
Sepsis
Multiple transfusions
Aspiration
Pneumonia
Other
Day 1 physiologic values
Minute ventilation, L/min
Respiratory rate, min1
Tidal volume, mL/kg
Pplat, cm H2O
pH
PaCO2, mm Hg
Base excess, mEq/L
FIO2
PEEP, cm H2O
PaO2, mm Hg
PaO2/FIO2
Mortality rate, %

429
52 18
0.41
84 28

432
51 17
0.40
81 28

37 (8)
111 (26)
15 (3)
62 (14)
156 (36)
48 (11)

54 (12)
117 (27)
10 (2)
65 (15)
144 (33)
42 (10)

12.6 4.6
16 6
11.8 0.8
33 9
7.41 0.07
35 8
2.1 5.7
0.51 0.18
8.6 3.7
77 19
176 76
40

12.9 3.6
28 7
6.2 0.9
25 7
7.38 0.08
40 10
1.6 5.4
0.57 0.19
9.6 3.7
76 23
158 73
31

APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; Pplat, plateau
pressure; PEEP, positive end-expiratory pressure.

RESULTS
A total of 861 patients were studied in
the parent clinical trial; 141 were excluded due to missing arterial blood gas
on day 1, and 16 were excluded due to
missing data in other variables. The demographics of the study population are
shown in Table 1 and of the excluded
patients in Table 2. Patients excluded due
to missing data had lower mortality rate
and statistically significantly lower Pplat,
FIO2, and, in the 6 mL/kg tidal volume
group, lower PEEP compared with the
patients included in the analysis. The remaining patients in each tidal volume
group were characterized as hypercapnic
or nonhypercapnic based on the definitions of HA as described. The characteristics of these four groups for HA1 are
shown in Table 3. Similar data for HA2
and HA3 are shown in Tables 4 and 5,
respectively. Patients from each of the
ten ARDS Network centers were represented in the 12 mL/kg HA1 group.
The results of the logistic regression
models for each tidal volume group and
each definition of HA are shown in Figure
2. The odds ratio (OR) of mortality and
95% confidence intervals (CI) of the 12
mL/kg and 6 mL/kg tidal volume groups
based on the presence of HA1 were 0.32
(CI 0.13 0.79, p .013) and 1.07 (CI
0.611.90, p .808), respectively. InCrit Care Med 2006 Vol. 34, No. 1

cluding more severe and restrictive definitions of HA in the model led to lower
adjusted ORs in the 12 mL/kg tidal volume group (Fig. 2). For HA2, the adjusted
OR of mortality in the 12 mL/kg tidal
volume group was 0.14 (CI 0.03 0.70, p
.016). Finally, HA3 was associated with
the greatest reduction in adjusted OR of
mortality in the 12 mL/kg tidal volume
group (OR 0.06, CI 0.01 0.47, p
.008). None of the adjusted ORs for mortality related to HA in the 6 mL/kg tidal
volume group were 1 or reached statistical significance (Fig. 2). Lastly, respiratory alkalosis was not statistically significantly associated with mortality in either
tidal volume group.
The additions of gender, day 1 tidal
volume, and/or PEEP did not alter the
association between HA and mortality.
For example, the OR of mortality associated with PEEP in the 12 mL/kg tidal
volume group with HA2 was 1.09 (95% CI
0.951.25, p .229). HCO3 values calculated from blood gas data were highly
correlated with venous values for those in
whom data were available. (According to
the method of Bland and Altman, the
calculated arterial HCO3 concentrations
were 0.87 mmol/L lower on average than
the measured venous values with an SD of
2.14 mmol/L. The venous-arterial difference and SD are consistent with changes

in HCO3 concentration produced by the

loss of CO2 through the lungs and standard measurement imprecision, respectively.) The addition of base excess to the
model did not alter the association between HA and mortality. When combined
into a single model incorporating both
tidal volume groups and an interaction
between tidal volume group and HA1,
HA1 and tidal volume interacted significantly confirming a different effect of HA1
in the different ventilatory strategies (p
.049).
Since excluded patients had a lower
mortality than study patients, we performed a conservative analysis assuming
that the excluded patients did not have
HA and assigning them the average PaO2/
FIO2 of the non-HA patients. This is a
conservative analysis because it would
tend to exaggerate any harmful effects of
HA and minimize any protective effects.
This did not alter the relationship between HA and mortality.

DISCUSSION
This study demonstrates that HA is
associated with reduced mortality in patients receiving 12 mL/kg tidal volumes
and is not associated with an effect on
mortality in patients receiving 6 mL/kg
tidal volumes in the ARDS Network clinical trial. These results support the hypothesis that HA is protective, at least in
some patients, rather than simply a tolerated side effect in the management of
patients with ALI. These findings are consistent with the theory that ventilatorassociated lung injury is occurring to a
greater extent in the 12 mL/kg tidal volume group and that HA mitigates this
injury. They are also consistent with the
theory that the lung-protective ventilation strategy considerably reduces this
injury to the point that a protective effect
of HA is not detectable (20, 21).
The acutely injured lung in the midst
of an inflammatory response has proven
refractory to targeted molecular interventions designed to reduce inflammation. The only intervention that has
proven effective in reducing mortality is
the reduction in tidal volume to 6
mL/kg vs. 12 mL/kg demonstrated in the
ARDS Network Trial (6). Our findings are
consistent with the growing basic science
evidence supporting the protective effects
of HA in states of lung injury. The original concept of salutary acidosis, the pH
paradox, arose from models of ischemia
and reperfusion. It has been observed re3

Table 2. Characteristics of patients lacking day 1 blood gas data and therefore excluded from the
analysis
12 mL/kg

6 mL/kg

Variable

Included

Excluded

Included

Excluded

No.
Age, years
Female, %
APACHE III score
Risk factor for ALI/ARDS, n (%)
Trauma
Sepsis
Multiple transfusions
Aspiration
Pneumonia
Other
Day 1 physiologic values
Minute ventilation, L/min
Respiratory rate, min1
Tidal volume, mL/kg
Pplat, cm H2O
pH
PaCO2, mm Hg
Base excess, mEq/L
FIO2
PEEP, cm H2O
PaO2, mm Hg
PaO2/FIO2
Mortality rate, %

369
52 18
42
85 28

60
54 17
35
79 31

351
51 17
40
82 28

81
52 17
38
78 17

31 (8)
95 (26)
15 (4)
52 (14)
134 (36)
42 (11)

6 (10)
16 (27)
0 (0)
10 (17)
22 (37)
6 (10)

41 (12)
96 (27)
7 (2)
50 (14)
124 (35)
33 (9)

13 (16)
21 (26)
3 (4)
15 (19)
20 (25)
9 (11)

12.6 4.6
16 6
11.8 0.8
33 9
7.41 0.07
35 8
2.1 5.6
0.51 0.18
94
76 19
175 76
40

12.2 3.5
17 7
11.9 0.8
29 6a

12.9 3.6
29 7
6.2 0.9
25 7
7.38 0.08
40 10
1.6 5.4
0.57 0.19
10 4
76 23
157 73
32

0.47 0.13
83
33

13.1 3.6
26 9
6.3 0.6
22 5a

0.49 0.14a
8 3a
26

APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acute
respiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure.
a
p .05 compared with patients included in the model.

Table 3. Characteristics of model population by tidal volume and hypercapnic acidosis

12 mL/kg

6 mL/kg

Variable

No HA1

HA1

No HA1

HA1

No.
Age, years
Female, %
APACHE III score
Risk factor for ALI/ARDS, n (%)
Trauma
Sepsis
Multiple transfusions
Aspiration
Pneumonia
Other
Day 1 physiologic values
Minute ventilation, L/min
Respiratory rate, min1
Tidal volume, mL/kg
Pplat, cm H2O
pH
PaCO2, mm Hg
Base excess, mEq/L
FIO2
PEEP, cm H2O
PaO2, mm Hg
PaO2/FIO2
Mortality rate, %

332
52 18
41
85 27

37
50 20
54
84 35

247
51 16
41
82 28

104
49 18
38
81 27

27 (8)
83 (25)
11 (3)
51 (15)
122 (37)
38 (11)

4 (11)
12 (32)
4 (11)
1 (3)
12 (32)
4 (11)

24 (10)
71 (29)
5 (2)
42 (17)
80 (32)
25 (10)

17 (16)
25 (24)
2 (2)
8 (8)
44 (42)
8 (8)

12.8 4.7
16 6
11.9 0.7
32 8
7.42 0.06
34 7
2.3 5.7
0.50 0.17
83
77 19
181 76
41

10.9 3.6a
16 7
11.4 1.5a
40 12a
7.34 0.08a
48 5a
0.2 4.4a
0.66 0.20a
12 5a
70 15a
126 53a
35

13.2 3.8
29 7
6.3 0.9
24 6
7.41 0.07
35 6
2.3 5.4
0.53 0.17
93
77 24
168 75
30

12.3 3.1a
30 7
6.0 0.9a
28 7a
7.31 0.08a
51 9a
0.0 4.8a
0.65 0.21a
11 4a
73 17
132 63a
35

APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acute
respiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA1 is
defined as pH 7.40 and PaCO2 40 mm Hg.
a
p .05 compared with patients without HA1.

peatedly that cells and organs near death

from ischemia and/or anoxia survive and
function better if reoxygenated in an
acidic milieu so that the intracellular acidosis resolves less abruptly. This has
been demonstrated in the liver, heart,
kidney, brain, and lung (715, 19 21).
These effects are lost when the HA is
buffered by NaHCO3 (13, 24, 25). Broccard et al. (20) reported protection from
ventilator-induced lung injury by HA in
isolated perfused rabbit lungs. Sinclair et
al. (21) extended this finding by demonstrating that HA was protective in an in
vivo rabbit model of ventilator-induced
lung injury. Furthermore, Laffey et al.
(26) found that HA with low tidal volumes in vivo can protect the lung injured
by aerosolized endotoxin even when applied after the onset of the inflammatory
injury.
There are many potential mechanisms
by which HA might reduce lung injury. A
unifying mechanism stems from the generally suppressive effects of acidosis on
nearly all cellular and molecular processes (for review, see Ref. 17). Specific
mechanisms that have been elucidated
include reduced tumor necrosis factor-
release by alveolar macrophages (27), reduced neutrophil-endothelial cell adhesion (28), reduced activity of xanthine
oxidase leading to reduced free radical
generation (29), reduced nuclear factor-B (30), reduced interleukin-8 and
free radical production from activated
neutrophils (24), and suppression of nitric oxide production (31) by inhibition of
inducible nitric oxide synthase. Although
CO2 and acidosis are known to stimulate
surfactant synthesis and secretion, Laffey
et al. (32) found no changes in surfactant
chemistry with 12% inspired CO2 in a
rabbit model of high tidal volumeinduced lung injury. The mechanism by
which 6 mL/kg tidal volumes are associated with reduced mortality compared
with 12 mL/kg is not known. It is speculated that 6 mL/kg tidal volumes produce
less local and systemic inflammation due
to ventilator-associated lung injury. The
reduced levels of serum interleukin-6
found in the patients ventilated with 6
mL/kg tidal volumes in the ARDS Network trial support this idea (5, 6). However, the exact mechanism by which tidal
volume affects mortality in ALI and the
manner in which HA modifies this relationship can only be speculated on at this
time.
There are several potential limitations
of evaluating the effects of HA using a
Crit Care Med 2006 Vol. 34, No. 1

Table 4. Characteristics of model population by tidal volume and hypercapnic acidosis (HA)2
12 mL/kg

6 mL/kg

Variable

No HA2

HA2

No HA2

HA2

No.
Age, yrs
Female, %
APACHE III score
Risk factor for ALI/ARDS
Trauma
Sepsis
Multiple transfusions
Aspiration
Pneumonia
Other
Day 1 physiologic values
Minute ventilation, L/min
Respiratory rate, min1
Tidal volume, mL/kg
Pplat, cm H2O
pH
PaCO2, mm Hg
Base excess, mEq/L
FIO2
PEEP, cm H2O
PaO2, mm Hg
PaO2/FIO2
Mortality rate, %

356
52 18
41
85 27

13
44 17
69
91 47

298
52 17
40
82 28

53
46 16a
45
81 27

29 (8)
89 (25)
14 (4)
52 (15)
130 (37)
42 (12)

2 (15)
6 (46)
1 (8)
0 (0)
4 (31)
0 (0)

35 (12)
85 (29)
6 (2)
44 (15)
98 (33)
30 (10)

6 (11)
11 (21)
1 (2)
6 (11)
26 (49)
3 (6)

12.6 4.6
16 6
11.8 0.8
32 8
7.42 0.06
34 7
2.1 5.7
0.50 0.17
83
77 19
178 75
40

12.1 4.5
20 10a
10.8 2.0a
47 15a
7.27 0.10a
52 5a
1.5 6.0
0.75 0.21a
14 6a
63 13a
100 38a
31

13.0 3.7
29 7
6.3 0.9
24 6
7.40 0.07
37 7
2.0 5.3
0.55 0.18
94
77 23
164 75
31

12.2 2.7
32 6a
5.9 0.9a
29 6a
7.28 0.09a
56 10a
0.4 5.5a
0.69 0.20a
12 4a
71 15
119 51a
38

APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acute
respiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA2 is
defined as pH 7.35 and PaCO2 45 mm Hg.
a
p .05 compared with patients without HA2.

Table 5. Characteristics of model population by tidal volume and hypercapnic acidosis (HA)3
12 mL/kg

6 mL/kg

Variable

No HA3

HA3

No HA3

HA3

No.
Age, years
Female, %
APACHE
Risk factor for ALI/ARDS, n (%)
Trauma
Sepsis
Multiple transfusions
Aspiration
Pneumonia
Other
Day 1 physiologic values
Minute ventilation, L/min
Respiratory rate, min1
Tidal volume, mL/kg
Pplat, cm H2O
pH
PaCO2, mm Hg
Base excess, mEq/L
FIO2
PEEP, cm H2O
PaO2, mm Hg
PaO2/FIO2
Mortality rate, %

358
52 18
41
85 27

11
44 18
82a
82 52

306
51 17
39
82 29

45
45 16a
44
78 24

29 (8)
91 (25)
14 (4)
52 (15)
130 (36)
42 (12)

2 (18)
4 (36)
1 (9)
0 (0)
4 (36)
0 (0)

35 (11)
86 (28)
6 (2)
45 (15)
104 (34)
30 (10)

6 (13)
10 (22)
1 (2)
5 (11)
20 (44)
3 (7)

12.7 4.6
16 6
11.8 0.8
32 8
7.42 0.07
35 8
2.2 5.7
0.51 0.17
83
77 19
178 75
41

10.7 3.1
17 8
10.7 3.1
44 14a
7.31 0.04a
51 6a
0.4 4.1
0.71 0.20a
13 6a
62 13a
103 40a
18

13.0 3.7
29 7
6.2 0.9
25 7
7.39 0.08
38 9
1.9 5.6
0.55 0.19
94
77 23
162 75
31

12.1 2.8
31 7a
5.9 0.8a
28 6a
7.30 0.05a
53 5a
0.2 3.6a
0.66 0.19a
11 3a
72 16
125 52a
33

APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acute
respiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA3 is
defined as 7.15 pH 7.35 and 45 mm Hg PaCO2 65 mm Hg.
a
p .05 compared with patients without HA3.

Crit Care Med 2006 Vol. 34, No. 1

Figure 2. Adjusted odds ratios for mortality. OR,

odds ratio; CI, confidence interval; *predicted
body weight. Adjusted for Acute Physiology and
Chronic Health Evaluation III score, risk factor
for acute lung injury, day 1 PaO2/FIO2, and Pplat.

secondary analysis of clinical trial data.

These include bias due to patient selection, center effect, exposure measurement, and unmeasured confounders. It is
possible that excluding patients enrolled
in the trial but lacking day 1 blood gas
data biased the study results. This could
occur if excluded patients had HA but
died before having a blood gas measured
on day 1, or they lacked HA but had lower
mortality rate. Our conservative analysis
assigning the excluded patients with
lower mortality to the non-HA group suggests that this bias does not explain our
results. There was no evidence of a center
effect as all ARDS Network centers contributed HA patients to this analysis. Of
course, it is possible that physicians who
manage patients with HA also use other
treatments that themselves are beneficial. This seems unlikely given that patients were cared for by many physicians
at multiple levels of training in a multiple-center study. In the ARDS Network
study analyzed here, the magnitude and
range of HA were limited (Fig. 1) when
compared with greater hypercapnia employed in animal studies in which protection was afforded by HA even with 57
mL/kg tidal volumes (26). We based our
exposure measure of HA on a single arterial blood gas on day 1. We considered
assessing sustained HA as an exposure;
however, there were only three patients
with sustained HA defined as pH 7.35
and PaCO2 45 mm Hg on days 1 and 3 in
the 12 mL/kg tidal volume group. This
was insufficient to perform reliable multivariate logistic regression. The exposure
measure we used has the potential to
include patients with only transient HA
5

ypercapnic acidosis was associated with re-

duced 28-day mortality in

the 12 mL/kg predicted body
weight tidal volume group
after controlling for comorbidities and severity of lung
injury.

while also excluding patients who transiently failed to meet these criteria. Nevertheless, the finding that HA on day 1 in
the 12 mL/kg arm of the study was associated with better survival and the evidence of increasing effect with increasing
degrees of HA suggests that some protection by HA is afforded early in the course
of ALI even if it is not sustained. Further
support for a protective effect of even a
short duration of HA is evident in many
animal studies, where a benefit from acidosis accrues in only several hours (13
22, 24, 26 32). Unmeasured confounding
variables are a universal problem with
multivariate analyses of cohort studies.
We did not have data on the use of sodium bicarbonate, accurate estimates of
CO2 production, or deadspace ventilation.
Since there were limitations in the tolerable limits of low pH in the ARDS Network trial, our ability to test the hypothesis that HA is beneficial is limited since
extremes were avoided. If greater extremes of HA had been allowed in this
study, it is possible that that either more
benefit or more harm may have been
detected. Most patients do not have a
pure respiratory acidosis, so some patients with HA had a combined respiratory and metabolic acidosis. In addition,
although we treated it as a single entity,
patients may develop HA by different
mechanisms including reduced minute
ventilation, increased CO2 production,
increased deadspace ventilation, and increased shunt fraction. Each of these
mechanisms may have varying degrees of
pathophysiological significance. These
variables need to be considered and quantified in future studies where HA exists in
low tidal volume groups. Nevertheless,
although inability to control for these
6

confounders is an important limitation of

this study, we believe that the likely direction of the bias introduced by the lack
of these variables actually strengthens
our results. For an unmeasured variable
to account for the observed protective
effect of HA, it would have to be associated with HA and with reduced mortality
rate. Many common clinical variables
might be associated with our definition of
HA, such as increased deadspace (33),
CO2 production, shunt fraction, and metabolic acidosis (23), but these are all generally associated with increased mortality
rate. Therefore, failing to account for
these variables biases our assessment of
the effect of HA toward harm, strengthens the conclusions about a protective
effect in patients receiving 12 mL/kg, and
raises the possibility that a beneficial effect of HA was obscured in the 6 mL/kg
group. We considered including respiratory rate and minute ventilation in the
model; however, because of the volume
control ventilator protocol used in the
trial, they were inextricably linked to HA
and their effects cannot be separated.
Last, although Hough et al. (34) found
slightly higher intrinsic PEEP in the 6
mL/kg tidal volume group, the small difference between 6 and 12 mL/kg tidal
volume groups (median 1.3 vs. 0.5 cm
H2O) is likely clinically insignificant
given the lack of effect on mortality of
higher PEEP (8 vs. 13 cm H2O) tested
directly in patients with ALI (35).

CONCLUSION
This secondary analysis of data from
the ARDS Network trial of lower tidal
volume provides evidence in support of
the theory that HA exerts a protective
effect in ALI. This effect appears to be
modulated by the ventilatory approach.
Because of the study limitations, it is
important to appreciate that, although
these clinical observations support a body
of basic science on the beneficial effects
of HA, they do not confirm them. Confirmation awaits further clinical studies and
an appropriately designed randomized
trial of HA coupled with lung-protective
ventilation in patients with ALI. Based on
the data available in this and other studies, we propose that greater extremes of
HA than those found in the ARDS Network trial of 12 vs. 6 mL/kg tidal volumes
would likely be needed to measure any
effect of HA on outcome in ALI patients
receiving lung protective ventilation. The
optimal dose, duration, and type (lowered

minute ventilation vs. inspired CO2) of

HA remain unanswered by this analysis.
In addition, measurement of variables
such as deadspace fraction and CO2 production will be important in trials investigating HA in ALI.

REFERENCES
1. Hickling KG, Walsh J, Henderson S, et al:
Low mortality rate in adult respiratory distress syndrome using low- volume, pressurelimited ventilation with permissive hypercapnia: A prospective study. Crit Care Med
1994; 22:1568 1578
2. Amato MB, Barbas CS, Medeiros DM, et al:
Beneficial effects of the open lung approach with low distending pressures in
ARDS: A prospective randomized study on
mechanical ventilation. Am J Respir Crit
Care Med 1995; 152:18351846
3. Gentilello LM, Anardi D, Mock C, et al: Permissive hypercapnia in trauma patients.
J Trauma 1995; 39:846 52
4. Amato MB, Barbas CS, Medeiros DM, et al:
Effect of a protective-ventilation strategy on
mortality in the acute respiratory distress
syndrome. N Engl J Med 1998; 338:347354
5. Ranieri VM, Suter PM, Tortorella C, et al:
Effect of mechanical ventilation on inflammatory mediators in patients with acute respiratory distress syndrome: A randomized
controlled trial. JAMA 1999; 282:54 61
6. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute
lung injury and the acute respiratory distress
syndrome. The Acute Respiratory Distress
Syndrome Network. N Engl J Med 2000; 342:
13011308
7. Kitakaze M, Weisfeldt ML, Marban E: Acidosis during early reperfusion prevents myocardial stunning in perfused ferret hearts.
J Clin Invest 1988; 82:920 927
8. Orchard C, Kentish J: Effects of changes of
pH on the contractile function of cardiac
muscle. Am J Physiol 1990; 258:C967C981
9. Currin RT, Gores GJ, Thurman RG, et al:
Protection by acidotic pH against anoxic cell
killing in perfused rat liver: evidence for a pH
paradox. FASEB J 1991; 5:207210
10. Nomura F, Aoki M, Forbess JM, et al: Effects
of hypercarbic acidotic reperfusion on recovery of myocardial function after cardioplegic
ischemia in neonatal lambs. Circulation
1994; 90:1132111327
11. Caldwell-Kenkel JC, Currin RT, Coote A, et
al: Reperfusion injury to endothelial cells
after cold storage of rat livers: protection by
mildly acidic pH and lack of protection by
antioxidants. Transpl Int 1995; 8:77 85
12. Moore TM, Khimenko PL, Taylor AE: Restoration of normal pH triggers ischemiareperfusion injury in lung by Na/H exchange activation. Am J Physiol 1995; 269:
H1501H1505
13. Cardenas VJ Jr, Zwischenberger JB, Tao W, et
al: Correction of blood pH attenuates

Crit Care Med 2006 Vol. 34, No. 1

14.

15.

16.

17.

18.

19.

20.

21.

changes in hemodynamics and organ blood

flow during permissive hypercapnia. Crit
Care Med 1996; 25:827 834
DArmini AM, Lemasters JJ, Egan TM, et al:
Effect of perfusion pH on rat lung viability on
non-heart beating donors. Eur J Cardiothorac Surg 1997; 12:787791
Hurtado C, Pierce GN: Inhibition of Na/H
exchange at the beginning of reperfusion is
cardioprotective in isolated, beating adult
cardiomyocytes. J Mol Cell Cardiol 2000; 32:
18971907
Laffey JG, Kavanagh BP: Carbon dioxide and
the critically illToo little of a good thing?
Lancet, 1999; 354:12831286
Kregenow DA, Swenson ER: Hypercapnic acidosis: Implications for permissive and therapeutic hypercapnia. Eur Respir J 2000; 20:
6 11
Swenson ER: Therapeutic hypercapnic acidosis: Pushing the envelope. Am J Respir Crit
Care Med 2004; 169:8 9
Laffey JG, Tanaka M, Engelberts D, et al:
Therapeutic hypercapnia reduces pulmonary
and systemic injury following in vivo lung
reperfusion. Am J Respir Crit Care Med 2000;
162:22872294
Broccard AF, Hotchkiss JR, Vannay C, et al:
Protective effects of hypercapnic acidosis on
ventilator-induced lung injury. Am J Respir
Crit Care Med 2001; 164:802 806
Sinclair SE, Kregenow DA, Lamm WJ, et al:
Hypercapnic acidosis is protective in an in

Crit Care Med 2006 Vol. 34, No. 1

22.

23.

24.

25.

26.

27.

28.

vivo model of ventilator-induced lung injury.

Am J Respir Crit Care Med 2002; 163:
403 408
Bland JM, Altman DG: Statistical methods
for assessing agreement between two methods of clinical measurement. Lancet 1986;
1(8476):307310
Knaus WA, Wagner DP, Draper EA, et al: The
APACHE III prognostic system. Risk prediction of hospital mortality for critically ill
hospitalized adults. Chest 1991; 100:
1619 1636
Coakley RJ, Taggart C, Greene C, et al: Ambient pCO2 modulates intracellular pH, intracellular oxidant generation, and interleukin-8 secretion in human neutrophils.
J Leukoc Biol 2002; 71:603 610
Laffey JG, Engelberts D, Kavanagh BP: Buffering hypercapnic acidosis worsens acute
lung injury. Am J Respir Crit Care Med 2000;
161:141146
Laffey JG, Honan D, Hopkins N, et al: Hypercapnic acidosis attenuates endotoxin induced
acute lung injury. Am J Respir Crit Care Med
2004; 169:46 56
Bidani A, Wang CZ, Saggi SJ, et al: Evidence
for pH sensitivity of tumor necrosis factoralpha release by alveolar macrophages. Lung
1998; 176:111121
Serrano CV Jr, Fraticelli A, Paniccia R, et al:
pH dependence of neutrophil-endothelial cell
adhesion and adhesion molecule expression.
Am J Physiol 1996; 271:C962C970

29. Shibata K, Cregg N, Engelberts D, et al: Hypercapnic acidosis may attenuate acute lung
injury by inhibition of endogenous xanthine
oxidase. Am J Respir Crit Care Med 1998;
158:1578 1584
30. Takeshita K, Suzuki Y, Nishio K, et al: Hypercapnic acidosis attenuates endotoxininduced nuclear factor-[kappa] B activation.
Am J Respir Cell Mol Biol 2003; 29:124 132
31. Adding LC, Agvald P, Persson MG, et al: Regulation of pulmonary nitric oxide by carbon
dioxide is intrinsic to the lung. Acta Physiol
Scand 1999; 167:167174
32. Laffey JG, Engelberts D, Duggan M, et al:
Carbon dioxide attenuates pulmonary impairment resulting from hyperventilation.
Crit Care Med 2003; 31:2634 2640
33. Nuckton TJ, Alonso JA, Kallet RH, et al: Pulmonary dead space fraction as a risk factor
for death in the acute respiratory distress
syndrome. N Engl J Med 2002; 346:
12811286
34. Hough CL, Kallet RH, Ranieri VM, et al:
Intrinsic positive end-expiratory pressure in
Acute Respiratory Distress Syndrome (ARDS)
Network subjects. Crit Care Med 2005; 33:
527532
35. Brower RG, Lanken PN, MacIntyre N, et al:
National Heart, Lung, and Blood Institute
ARDS Clinical Trials Network. Higher versus
lower positive end-expiratory pressures in patients with the acute respiratory distress syndrome. N Engl J Med 2004; 351:327336