Hypercapnic Acidosis and Mortality in Acute Lung Injury

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

Hypercapnic acidosis (HA),more commonly referred toas permissive hypercapnia,has been viewed as an ac-ceptable side effect of lung-protectiveventilation that can be tolerated in aneffort to avoid ventilator-associated lunginjury (16). Evidence is growing, how-ever, that HA has favorable anti-inflam-matory and antioxidative effects at thesubcellular, cellular, whole organ, andwhole organism levels in hypoxic and in-flammatory conditions (715). This hasled to speculation that HA may provideadditional benefit to patients with a vari-ety of injuries and illnesses (1618). To

date the only data available on HA inhumans with acute lung injury (ALI)have been limited to trials of lung-protective ventilation in which the effectsof HA have not been separated from theeffects of changes in mechanical ventila-tion.

In 2000, the National Institutes ofHealth Acute Respiratory Distress Syn-drome (ARDS) Network published thelargest multiple-center randomized trialof patients with ALI (6). The ARDS Net-work trial compared 12 vs. 6 mL/kg pre-dicted body weight tidal volumes anddemonstrated a 9% absolute and a 22%relative reduction in 28-day mortalityrate through the use of 6 mL/kg predictedbody weight tidal volumes. Therefore, weused data from the ARDS Network trial toseparate the effect of HA from lung-protective ventilation in patients withALI.

MATERIALS AND METHODS

Study Design. Secondary analysis of ran-domized clinical trial data.

Patients and Variables. The details of theinclusion and exclusion criteria for the ARDSNetwork clinical trial on which this analysis isbased have been published elsewhere (6). Pa-

tients were enrolled from March 1996 toMarch 1999 at ten university centers. The en-rolled patients met the American-EuropeanConsensus definition of ALI and ARDS includ-ing acute hypoxic respiratory failure requiringmechanical ventilation, bilateral patchy infil-trates on chest radiograph, and no evidence ofleft atrial hypertension. The ARDS Networktrial compared mechanical ventilation usingvolume-cycled assist control with a tidal vol-ume of 12 mL/kg predicted body weight vs. 6mL/kg predicted body weight. Hypercapniawas not a goal of the ventilator strategy in thisstudy and was not a randomized intervention.Minute ventilation was used to keep 7.30 arterial pH 7.45. At a maximum respiratoryrate of 35 and pH 7.15, management ofacidosis with sodium bicarbonate was at thediscretion of the investigator.

The ARDS Network and the Human Sub-jects Divisions of the University of Washingtonand the VA Puget Sound Health Care Systemapproved this study. Unless stated otherwise,mortality refers to 28-day mortality rate andtidal volume in mL/kg refers to mL/kg pre-dicted body weight as calculated in the parentclinical trial. Partial pressures of oxygen werecorrected for altitude as in the parent clinicaltrial.

Based on the experimental evidence (1921), we examined an early (or acute) HA alongwith other data available early in a patientscourse of treatment. We defined HA based on

*See also p. 229.From the Division of Pulmonary and Critical Care

Medicine, Department of Medicine, University of Wash-ington, Seattle, WA (DAK, GDR, LDH, ERS); and theSection of Pulmonary and Critical Care, Department ofMedicine, 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 NetContract No. NO1 HR46055.

None of the authors has any financial interests todisclose.

Copyright 2005 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/01.CCM.0000194533.75481.03

Objective: We tested the hypothesis that hypercapnic acidosisis associated with reduced mortality rate in patients with acutelung injury independent of changes in mechanical ventilation.

Design: Secondary analysis of randomized clinical trial datausing hypothesis-driven multivariate logistic regression.

Setting: Randomized, multiple-center trial (n 861) compar-ing 12 mL/kg to 6 mL/kg predicted body weight tidal volumespreviously published by the National Institutes of Health AcuteRespiratory 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 associatedwith hypercapnic acidosis defined as day 1 pH 45 mm Hg were 0.14 (95% CI 0.030.70, p .016) in the 12mL/kg predicted body weight tidal volume group and 1.18 (95% CI0.592.35, p .639) in the 6 mL/kg predicted body weight tidal

volume group. Other definitions of hypercapnic acidosis spanninga range of magnitudes suggest a dose-response associationbetween hypercapnic acidosis and 28-day mortality in the 12mL/kg predicted body weight tidal volume group. None of ourdefinitions of hypercapnic acidosis were associated with reduc-tion in 28-day mortality in the 6 mL/kg predicted body weight tidalvolume group.

Conclusions: Hypercapnic acidosis was associated with re-duced 28-day mortality in the 12 mL/kg predicted body weighttidal volume group after controlling for comorbidities and severityof lung injury. These results are consistent with a protective effectof hypercapnic acidosis against ventilator-associated lung injurythat was not found when the further ongoing injury was reducedby 6 mL/kg predicted body weight tidal volumes. (Crit Care Med2006; 34:17)

KEY WORDS: hypercapnia; acidosis; respiratory; respiratory dis-tress syndrome; adult; respiration; artificial; hypoventilation; mor-tality

1Crit Care Med 2006 Vol. 34, No. 1

arterial blood gas data collected on the morn-ing following randomization and institution ofmechanical ventilation according to the studyprotocol (day 1). Although they reflect a singlepoint in time, day 1 data have the advantage ofbeing the most complete set of blood gas val-ues and avoid the survival bias of requiringserial measurements to define HA. The datacollection protocol provided day 1 blood gasvalues from 7 hrs to as much as 31 hrs afterinstitution of protocolized mechanical venti-lation. The day 1 blood gas values and defini-tions of HA and respiratory alkalosis areshown in Figure 1. Because there is no con-sensus definition of therapeutic HA, our anal-ysis plan included assessing the effect of dif-ferent degrees of HA. These were defined apriori. HA1 was defined as a pH 7.40 andPaCO2 40 mm to assess the effect of anydegree of HA. HA2 was defined as a pH 7.35and PaCO2 45 mm Hg to exclude patientswhose blood gas values would otherwise havebeen considered within normal limits and toassess a more severe degree of HA. There weretoo few patients with greater extremes of re-spiratory acidosis (e.g., pH 7.30 and PaCO250 mm Hg) to perform reliable analyses.

Finally, we hypothesized a priori that ex-tremes of acidosis, even if associated with el-evated PaCO2, would indicate significant met-abolic acidosis, comorbidities, and/or probableinevitable irreversibility of the underlying dis-ease. These patients would likely have suchoverwhelming cellular dysfunction that HAwould not afford protection (10) or would bepoorly tolerated. Furthermore, patients with apH 7.15 may have been given sodium bicar-bonate as part of the trial at the discretion ofthe providers, but this information was notrecorded. Thus, a third definition of HA (HA3)with a lower limit of pH and upper limit ofPaCO2 consistent with a simple but moderatelysevere acute respiratory acidosis was examined(7.15 pH 7.35 and 45 mm Hg PaCO2 65 mm Hg). Respiratory alkalosis was definedas pH 7.45 and PaCO2 35 mm Hg. Again,there were too few patients with more extremerespiratory alkalosis (e.g., pH7.50 and PaCO230 mm Hg) to perform reliable multivariateanalysis.

Since no direct measure of a metabolicacidosis was recorded for the majority of pa-tients on day 1, arterial HCO3

values werecalculated from the arterial pH and PaCO2 val-

ues based on the Henderson-Hasselbalchequation. Calculated arterial HCO3

valueswere compared with measured venous HCO3

values in those patients with serum chemis-tries recorded on day 1 using the method ofBland and Altman (22). Base excess was thencalculated according to the formula base ex-cess (HCO3

10 [pH 7.40]) 24.Statistical Methods and Modeling Strat-

egy. The primary research question was toassess the independent effect of HA on mor-tality rate. Because HA was not a randomizedintervention, patients who achieved HA mightdiffer in many ways from those who did notachieve HA. To account for these factors andassess the independent effect of HA, we per-formed hypothesis-driven multivariate logisticregression including confounding variablesassociated with both HA and mortality. To dothis we built a logistic regression model ac-counting for the patients severity of illnessand the severity of his or her ALI and analyzedeach tidal volume group separately. We usedAcute Physiology and Chronic Health Evalua-tion (APACHE) III, a validated intensive careunit severity of illness measure, to measurethe severity of critical illness and account forage and comorbid illness (23). We used pri-mary risk factor for ALI, a known predictor ofoutcome in this patient population, to accountfor different mechanisms of ALI. We used day1 PaO2/FIO2 and plateau pressure (Pplat) tomeasure the severity of ALI. Bivariate analysesof these latter variables confirmed that theywere related similarly to mortality and HA anddid not identify other important confounders.The final logistic regression model predictedmortality as a function of APACHE III score,risk factor for ALI, PaO2/FIO2, and Pplat. Toanswer the research question, we added thedifferent definitions of HA as a binary covari-ate to this model.

Several secondary analyses were per-formed. To assess the effect of gender, tidalvolume, and positive end-expiratory pressure(PEEP) on day 1, these variables were added tothe final model separately and together. Con-founding by metabolic acidosis was examinedby adding base excess to the model to deter-mine whether metabolic acidosis affected theassociation between HA and mortality. To con-firm the finding that HA acts differently de-pending on whether patients received lung-protective ventilation, we tested an additionalmodel that included all patients plus tidal vol-ume randomization and an interaction termbetween HA and tidal volume.

Statistical analyses were performed usingSTATA version 7.0 (Stata Corporation, CollegeStation TX). Results are reported as mean SD unless otherwise indicated. Where indicatedin the tables, continuous variables were com-pared with the two-sample Students t-testwith equal variances, and categorical variableswere compared with the Pearson chi-squaretest for independence.

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

2 Crit Care Med 2006 Vol. 34, No. 1

RESULTS

A total of 861 patients were studied inthe parent clinical trial; 141 were ex-cluded due to missing arterial blood gason day 1, and 16 were excluded due tomissing data in other variables. The de-mographics of the study population areshown in Table 1 and of the excludedpatients in Table 2. Patients excluded dueto missing data had lower mortality rateand statistically significantly lower Pplat,FIO2, and, in the 6 mL/kg tidal volumegroup, lower PEEP compared with thepatients included in the analysis. The re-maining patients in each tidal volumegroup were characterized as hypercapnicor nonhypercapnic based on the defini-tions of HA as described. The character-istics of these four groups for HA1 areshown in Table 3. Similar data for HA2and HA3 are shown in Tables 4 and 5,respectively. Patients from each of theten ARDS Network centers were repre-sented in the 12 mL/kg HA1 group.

The results of the logistic regressionmodels for each tidal volume group andeach definition of HA are shown in Figure2. The odds ratio (OR) of mortality and95% confidence intervals (CI) of the 12mL/kg and 6 mL/kg tidal volume groupsbased on the presence of HA1 were 0.32(CI 0.130.79, p .013) and 1.07 (CI0.611.90, p .808), respectively. In-

cluding more severe and restrictive defi-nitions of HA in the model led to loweradjusted ORs in the 12 mL/kg tidal vol-ume group (Fig. 2). For HA2, the adjustedOR of mortality in the 12 mL/kg tidalvolume group was 0.14 (CI 0.030.70, p .016). Finally, HA3 was associated withthe greatest reduction in adjusted OR ofmortality in the 12 mL/kg tidal volumegroup (OR 0.06, CI 0.010.47, p .008). None of the adjusted ORs for mor-tality related to HA in the 6 mL/kg tidalvolume group were 1 or reached statis-tical significance (Fig. 2). Lastly, respira-tory alkalosis was not statistically signif-icantly associated with mortality in eithertidal volume group.

The additions of gender, day 1 tidalvolume, and/or PEEP did not alter theassociation between HA and mortality.For example, the OR of mortality associ-ated with PEEP in the 12 mL/kg tidalvolume group with HA2 was 1.09 (95% CI0.951.25, p .229). HCO3

values cal-culated from blood gas data were highlycorrelated with venous values for those inwhom data were available. (According tothe method of Bland and Altman, thecalculated arterial HCO3

concentrationswere 0.87 mmol/L lower on average thanthe measured venous values with an SD of2.14 mmol/L. The venous-arterial differ-ence and SD are consistent with changes

in HCO3 concentration produced by the

loss of CO2 through the lungs and stan-dard measurement imprecision, respec-tively.) The addition of base excess to themodel did not alter the association be-tween HA and mortality. When combinedinto a single model incorporating bothtidal volume groups and an interactionbetween tidal volume group and HA1,HA1 and tidal volume interacted signifi-cantly confirming a different effect of HA1in the different ventilatory strategies (p.049).

Since excluded patients had a lowermortality than study patients, we per-formed a conservative analysis assumingthat the excluded patients did not haveHA and assigning them the average PaO2/FIO2 of the non-HA patients. This is aconservative analysis because it wouldtend to exaggerate any harmful effects ofHA and minimize any protective effects.This did not alter the relationship be-tween HA and mortality.

DISCUSSION

This study demonstrates that HA isassociated with reduced mortality in pa-tients receiving 12 mL/kg tidal volumesand is not associated with an effect onmortality in patients receiving 6 mL/kgtidal volumes in the ARDS Network clin-ical trial. These results support the hy-pothesis that HA is protective, at least insome patients, rather than simply a tol-erated side effect in the management ofpatients with ALI. These findings are con-sistent with the theory that ventilator-associated lung injury is occurring to agreater extent in the 12 mL/kg tidal vol-ume group and that HA mitigates thisinjury. They are also consistent with thetheory that the lung-protective ventila-tion strategy considerably reduces thisinjury to the point that a protective effectof HA is not detectable (20, 21).

The acutely injured lung in the midstof an inflammatory response has provenrefractory to targeted molecular inter-ventions designed to reduce inflamma-tion. The only intervention that hasproven effective in reducing mortality isthe reduction in tidal volume to 6mL/kg vs. 12 mL/kg demonstrated in theARDS Network Trial (6). Our findings areconsistent with the growing basic scienceevidence supporting the protective effectsof HA in states of lung injury. The origi-nal concept of salutary acidosis, the pHparadox, arose from models of ischemiaand reperfusion. It has been observed re-

Table 1. Characteristics of the study population (6)

Variable 12 mL/kg 6 mL/kg

No. 429 432Age, years 52 18 51 17Female 0.41 0.40APACHE III score 84 28 81 28Risk factor for ALI, n (%)

Trauma 37 (8) 54 (12)Sepsis 111 (26) 117 (27)Multiple transfusions 15 (3) 10 (2)Aspiration 62 (14) 65 (15)Pneumonia 156 (36) 144 (33)Other 48 (11) 42 (10)

Day 1 physiologic valuesMinute ventilation, L/min 12.6 4.6 12.9 3.6Respiratory rate, min1 16 6 28 7Tidal volume, mL/kg 11.8 0.8 6.2 0.9Pplat, cm H2O 33 9 25 7pH 7.41 0.07 7.38 0.08PaCO2, mm Hg 35 8 40 10Base excess, mEq/L 2.1 5.7 1.6 5.4FIO2 0.51 0.18 0.57 0.19PEEP, cm H2O 8.6 3.7 9.6 3.7PaO2, mm Hg 77 19 76 23PaO2/FIO2 176 76 158 73

Mortality rate, % 40 31

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


peatedly that cells and organs near deathfrom ischemia and/or anoxia survive andfunction better if reoxygenated in anacidic milieu so that the intracellular ac-idosis resolves less abruptly. This hasbeen demonstrated in the liver, heart,kidney, brain, and lung (715, 1921).These effects are lost when the HA isbuffered by NaHCO3 (13, 24, 25). Broc-card et al. (20) reported protection fromventilator-induced lung injury by HA inisolated perfused rabbit lungs. Sinclair etal. (21) extended this finding by demon-strating that HA was protective in an invivo rabbit model of ventilator-inducedlung injury. Furthermore, Laffey et al.(26) found that HA with low tidal vol-umes in vivo can protect the lung injuredby aerosolized endotoxin even when ap-plied after the onset of the inflammatoryinjury.

There are many potential mechanismsby which HA might reduce lung injury. Aunifying mechanism stems from the gen-erally suppressive effects of acidosis onnearly all cellular and molecular pro-cesses (for review, see Ref. 17). Specificmechanisms that have been elucidatedinclude reduced tumor necrosis factor-release by alveolar macrophages (27), re-duced neutrophil-endothelial cell adhe-sion (28), reduced activity of xanthineoxidase leading to reduced free radicalgeneration (29), reduced nuclear fac-tor-B (30), reduced interleukin-8 andfree radical production from activatedneutrophils (24), and suppression of ni-tric oxide production (31) by inhibition ofinducible nitric oxide synthase. AlthoughCO2 and acidosis are known to stimulatesurfactant synthesis and secretion, Laffeyet al. (32) found no changes in surfactantchemistry with 12% inspired CO2 in arabbit model of high tidal volume-induced lung injury. The mechanism bywhich 6 mL/kg tidal volumes are associ-ated with reduced mortality comparedwith 12 mL/kg is not known. It is specu-lated that 6 mL/kg tidal volumes produceless local and systemic inflammation dueto ventilator-associated lung injury. Thereduced levels of serum interleukin-6found in the patients ventilated with 6mL/kg tidal volumes in the ARDS Net-work trial support this idea (5, 6). How-ever, the exact mechanism by which tidalvolume affects mortality in ALI and themanner in which HA modifies this rela-tionship can only be speculated on at thistime.

There are several potential limitationsof evaluating the effects of HA using a

Table 2. Characteristics of patients lacking day 1 blood gas data and therefore excluded from theanalysis

Variable

12 mL/kg 6 mL/kg

Included Excluded Included Excluded

No. 369 60 351 81Age, years 52 18 54 17 51 17 52 17Female, % 42 35 40 38APACHE III score 85 28 79 31 82 28 78 17Risk factor for ALI/ARDS, n (%)

Trauma 31 (8) 6 (10) 41 (12) 13 (16)Sepsis 95 (26) 16 (27) 96 (27) 21 (26)Multiple transfusions 15 (4) 0 (0) 7 (2) 3 (4)Aspiration 52 (14) 10 (17) 50 (14) 15 (19)Pneumonia 134 (36) 22 (37) 124 (35) 20 (25)Other 42 (11) 6 (10) 33 (9) 9 (11)

Day 1 physiologic valuesMinute ventilation, L/min 12.6 4.6 12.2 3.5 12.9 3.6 13.1 3.6Respiratory rate, min1 16 6 17 7 29 7 26 9Tidal volume, mL/kg 11.8 0.8 11.9 0.8 6.2 0.9 6.3 0.6Pplat, cm H2O 33 9 29 6

a 25 7 22 5a

pH 7.41 0.07 7.38 0.08PaCO2, mm Hg 35 8 40 10Base excess, mEq/L 2.1 5.6 1.6 5.4FIO2 0.51 0.18 0.47 0.13 0.57 0.19 0.49 0.14

a

PEEP, cm H2O 9 4 8 3 10 4 8 3a

PaO2, mm Hg 76 19 76 23PaO2/FIO2 175 76 157 73

Mortality rate, % 40 33 32 26

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

ap .05 compared with patients included in the model.

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

Variable

12 mL/kg 6 mL/kg

No HA1 HA1 No HA1 HA1

No. 332 37 247 104Age, years 52 18 50 20 51 16 49 18Female, % 41 54 41 38APACHE III score 85 27 84 35 82 28 81 27Risk factor for ALI/ARDS, n (%)


Day 1 physiologic valuesMinute ventilation, L/min 12.8 4.7 10.9 3.6a 13.2 3.8 12.3 3.1a

Respiratory rate, min1 16 6 16 7 29 7 30 7Tidal volume, mL/kg 11.9 0.7 11.4 1.5a 6.3 0.9 6.0 0.9a

Pplat, cm H2O 32 8 40 12a 24 6 28 7a

pH 7.42 0.06 7.34 0.08a 7.41 0.07 7.31 0.08a

PaCO2, mm Hg 34 7 48 5a 35 6 51 9a

Base excess, mEq/L 2.3 5.7 0.2 4.4a 2.3 5.4 0.0 4.8a

FIO2 0.50 0.17 0.66 0.20a 0.53 0.17 0.65 0.21a

PEEP, cm H2O 8 3 12 5a 9 3 11 4a

PaO2, mm Hg 77 19 70 15a 77 24 73 17

PaO2/FIO2 181 76 126 53a 168 75 132 63a


APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acuterespiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA1 isdefined as pH 7.40 and PaCO2 40 mm Hg.

ap .05 compared with patients without HA1.


secondary analysis of clinical trial data.These include bias due to patient selec-tion, center effect, exposure measure-ment, and unmeasured confounders. It ispossible that excluding patients enrolledin the trial but lacking day 1 blood gasdata biased the study results. This couldoccur if excluded patients had HA butdied before having a blood gas measuredon day 1, or they lacked HA but had lowermortality rate. Our conservative analysisassigning the excluded patients withlower mortality to the non-HA group sug-gests that this bias does not explain ourresults. There was no evidence of a centereffect as all ARDS Network centers con-tributed HA patients to this analysis. Ofcourse, it is possible that physicians whomanage patients with HA also use othertreatments that themselves are benefi-cial. This seems unlikely given that pa-tients were cared for by many physiciansat multiple levels of training in a multi-ple-center study. In the ARDS Networkstudy analyzed here, the magnitude andrange of HA were limited (Fig. 1) whencompared with greater hypercapnia em-ployed in animal studies in which protec-tion was afforded by HA even with 57mL/kg tidal volumes (26). We based ourexposure measure of HA on a single arte-rial blood gas on day 1. We consideredassessing sustained HA as an exposure;however, there were only three patientswith sustained HA defined as pH 7.35and PaCO245 mm Hg on days 1 and 3 inthe 12 mL/kg tidal volume group. Thiswas insufficient to perform reliable mul-tivariate logistic regression. The exposuremeasure we used has the potential toinclude patients with only transient HA

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

Variable

12 mL/kg 6 mL/kg


No. 356 13 298 53Age, yrs 52 18 44 17 52 17 46 16a

Female, % 41 69 40 45APACHE III score 85 27 91 47 82 28 81 27Risk factor for ALI/ARDS


Day 1 physiologic valuesMinute ventilation, L/min 12.6 4.6 12.1 4.5 13.0 3.7 12.2 2.7Respiratory rate, min1 16 6 20 10a 29 7 32 6a

Tidal volume, mL/kg 11.8 0.8 10.8 2.0a 6.3 0.9 5.9 0.9a

Pplat, cm H2O 32 8 47 15a 24 6 29 6a

pH 7.42 0.06 7.27 0.10a 7.40 0.07 7.28 0.09a

PaCO2, mm Hg 34 7 52 5a 37 7 56 10a

Base excess, mEq/L 2.1 5.7 1.5 6.0 2.0 5.3 0.4 5.5a

FIO2 0.50 0.17 0.75 0.21a 0.55 0.18 0.69 0.20a

PEEP, cm H2O 8 3 14 6a 9 4 12 4a

PaO2, mm Hg 77 19 63 13a 77 23 71 15

PaO2/FIO2 178 75 100 38a 164 75 119 51a


APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acuterespiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA2 isdefined as pH 7.35 and PaCO2 45 mm Hg.


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

Variable

12 mL/kg 6 mL/kg


No. 358 11 306 45Age, years 52 18 44 18 51 17 45 16a

Female, % 41 82a 39 44APACHE 85 27 82 52 82 29 78 24Risk factor for ALI/ARDS, n (%)


Day 1 physiologic valuesMinute ventilation, L/min 12.7 4.6 10.7 3.1 13.0 3.7 12.1 2.8Respiratory rate, min1 16 6 17 8 29 7 31 7a

Tidal volume, mL/kg 11.8 0.8 10.7 3.1 6.2 0.9 5.9 0.8a

Pplat, cm H2O 32 8 44 14a 25 7 28 6a

pH 7.42 0.07 7.31 0.04a 7.39 0.08 7.30 0.05a

PaCO2, mm Hg 35 8 51 6a 38 9 53 5a

Base excess, mEq/L 2.2 5.7 0.4 4.1 1.9 5.6 0.2 3.6a

FIO2 0.51 0.17 0.71 0.20a 0.55 0.19 0.66 0.19a

PEEP, cm H2O 8 3 13 6a 9 4 11 3a

PaO2, mm Hg 77 19 62 13a 77 23 72 16

PaO2/FIO2 178 75 103 40a 162 75 125 52a


APACHE, Acute Physiology and Chronic Health Evaluation; ALI, acute lung injury; ARDS, acuterespiratory distress syndrome; Pplat, plateau pressure; PEEP, positive end-expiratory pressure; HA3 isdefined as 7.15 pH 7.35 and 45 mm Hg PaCO2 65 mm Hg.


Figure 2. Adjusted odds ratios for mortality. OR,odds ratio; CI, confidence interval; *predictedbody weight. Adjusted for Acute Physiology andChronic Health Evaluation III score, risk factorfor acute lung injury, day 1 PaO2/FIO2, and Pplat.


while also excluding patients who tran-siently failed to meet these criteria. Nev-ertheless, the finding that HA on day 1 inthe 12 mL/kg arm of the study was asso-ciated with better survival and the evi-dence of increasing effect with increasingdegrees of HA suggests that some protec-tion by HA is afforded early in the courseof ALI even if it is not sustained. Furthersupport for a protective effect of even ashort duration of HA is evident in manyanimal studies, where a benefit from ac-idosis accrues in only several hours (1322, 24, 2632). Unmeasured confoundingvariables are a universal problem withmultivariate analyses of cohort studies.We did not have data on the use of so-dium bicarbonate, accurate estimates ofCO2 production, or deadspace ventilation.Since there were limitations in the toler-able limits of low pH in the ARDS Net-work trial, our ability to test the hypoth-esis that HA is beneficial is limited sinceextremes were avoided. If greater ex-tremes of HA had been allowed in thisstudy, it is possible that that either morebenefit or more harm may have beendetected. Most patients do not have apure respiratory acidosis, so some pa-tients with HA had a combined respira-tory and metabolic acidosis. In addition,although we treated it as a single entity,patients may develop HA by differentmechanisms including reduced minuteventilation, increased CO2 production,increased deadspace ventilation, and in-creased shunt fraction. Each of thesemechanisms may have varying degrees ofpathophysiological significance. Thesevariables need to be considered and quan-tified in future studies where HA exists inlow tidal volume groups. Nevertheless,although inability to control for these

confounders is an important limitation ofthis study, we believe that the likely di-rection of the bias introduced by the lackof these variables actually strengthensour results. For an unmeasured variableto account for the observed protectiveeffect of HA, it would have to be associ-ated with HA and with reduced mortalityrate. Many common clinical variablesmight be associated with our definition ofHA, such as increased deadspace (33),CO2 production, shunt fraction, and met-abolic acidosis (23), but these are all gen-erally associated with increased mortalityrate. Therefore, failing to account forthese variables biases our assessment ofthe effect of HA toward harm, strength-ens the conclusions about a protectiveeffect in patients receiving 12 mL/kg, andraises the possibility that a beneficial ef-fect of HA was obscured in the 6 mL/kggroup. We considered including respira-tory rate and minute ventilation in themodel; however, because of the volumecontrol ventilator protocol used in thetrial, they were inextricably linked to HAand their effects cannot be separated.Last, although Hough et al. (34) foundslightly higher intrinsic PEEP in the 6mL/kg tidal volume group, the small dif-ference between 6 and 12 mL/kg tidalvolume groups (median 1.3 vs. 0.5 cmH2O) is likely clinically insignificantgiven the lack of effect on mortality ofhigher PEEP (8 vs. 13 cm H2O) testeddirectly in patients with ALI (35).

CONCLUSION

This secondary analysis of data fromthe ARDS Network trial of lower tidalvolume provides evidence in support ofthe theory that HA exerts a protectiveeffect in ALI. This effect appears to bemodulated by the ventilatory approach.Because of the study limitations, it isimportant to appreciate that, althoughthese clinical observations support a bodyof basic science on the beneficial effectsof HA, they do not confirm them. Confir-mation awaits further clinical studies andan appropriately designed randomizedtrial of HA coupled with lung-protectiveventilation in patients with ALI. Based onthe data available in this and other stud-ies, we propose that greater extremes ofHA than those found in the ARDS Net-work trial of 12 vs. 6 mL/kg tidal volumeswould likely be needed to measure anyeffect of HA on outcome in ALI patientsreceiving lung protective ventilation. Theoptimal dose, duration, and type (lowered

minute ventilation vs. inspired CO2) ofHA remain unanswered by this analysis.In addition, measurement of variablessuch as deadspace fraction and CO2 pro-duction will be important in trials inves-tigating HA in ALI.

REFERENCES

1. Hickling KG, Walsh J, Henderson S, et al:Low mortality rate in adult respiratory dis-tress syndrome using low- volume, pressure-limited ventilation with permissive hyper-capnia: A prospective study. Crit Care Med1994; 22:15681578

2. Amato MB, Barbas CS, Medeiros DM, et al:Beneficial effects of the open lung ap-proach with low distending pressures inARDS: A prospective randomized study onmechanical ventilation. Am J Respir CritCare Med 1995; 152:18351846

3. Gentilello LM, Anardi D, Mock C, et al: Per-missive hypercapnia in trauma patients.J Trauma 1995; 39:84652

4. Amato MB, Barbas CS, Medeiros DM, et al:Effect of a protective-ventilation strategy onmortality in the acute respiratory distresssyndrome. N Engl J Med 1998; 338:347354

5. Ranieri VM, Suter PM, Tortorella C, et al:Effect of mechanical ventilation on inflam-matory mediators in patients with acute re-spiratory distress syndrome: A randomizedcontrolled trial. JAMA 1999; 282:5461

6. Ventilation with lower tidal volumes as com-pared with traditional tidal volumes for acutelung injury and the acute respiratory distresssyndrome. The Acute Respiratory DistressSyndrome Network. N Engl J Med 2000; 342:13011308

7. Kitakaze M, Weisfeldt ML, Marban E: Acido-sis during early reperfusion prevents myo-cardial stunning in perfused ferret hearts.J Clin Invest 1988; 82:920927

8. Orchard C, Kentish J: Effects of changes ofpH on the contractile function of cardiacmuscle. Am J Physiol 1990; 258:C967C981

9. Currin RT, Gores GJ, Thurman RG, et al:Protection by acidotic pH against anoxic cellkilling in perfused rat liver: evidence for a pHparadox. FASEB J 1991; 5:207210

10. Nomura F, Aoki M, Forbess JM, et al: Effectsof hypercarbic acidotic reperfusion on recov-ery of myocardial function after cardioplegicischemia in neonatal lambs. Circulation1994; 90:1132111327

11. Caldwell-Kenkel JC, Currin RT, Coote A, etal: Reperfusion injury to endothelial cellsafter cold storage of rat livers: protection bymildly acidic pH and lack of protection byantioxidants. Transpl Int 1995; 8:7785

12. Moore TM, Khimenko PL, Taylor AE: Resto-ration of normal pH triggers ischemia-reperfusion injury in lung by Na/H ex-change activation. Am J Physiol 1995; 269:H1501H1505

13. Cardenas VJ Jr, Zwischenberger JB, Tao W, etal: Correction of blood pH attenuates

H ypercapnic aci-dosis was asso-ciated with re-duced 28-day mortality in

the 12 mL/kg predicted body

weight tidal volume group

after controlling for comor-

bidities and severity of lung

injury.


changes in hemodynamics and organ bloodflow during permissive hypercapnia. CritCare Med 1996; 25:827834

14. DArmini AM, Lemasters JJ, Egan TM, et al:Effect of perfusion pH on rat lung viability onnon-heart beating donors. Eur J Cardiotho-rac Surg 1997; 12:787791

15. Hurtado C, Pierce GN: Inhibition of Na/H

exchange at the beginning of reperfusion iscardioprotective in isolated, beating adultcardiomyocytes. J Mol Cell Cardiol 2000; 32:18971907

16. Laffey JG, Kavanagh BP: Carbon dioxide andthe critically illToo little of a good thing?Lancet, 1999; 354:12831286

17. Kregenow DA, Swenson ER: Hypercapnic ac-idosis: Implications for permissive and ther-apeutic hypercapnia. Eur Respir J 2000; 20:611

18. Swenson ER: Therapeutic hypercapnic acido-sis: Pushing the envelope. Am J Respir CritCare Med 2004; 169:89

19. Laffey JG, Tanaka M, Engelberts D, et al:Therapeutic hypercapnia reduces pulmonaryand systemic injury following in vivo lungreperfusion. Am J Respir Crit Care Med 2000;162:22872294

20. Broccard AF, Hotchkiss JR, Vannay C, et al:Protective effects of hypercapnic acidosis onventilator-induced lung injury. Am J RespirCrit Care Med 2001; 164:802806

21. Sinclair SE, Kregenow DA, Lamm WJ, et al:Hypercapnic acidosis is protective in an in

vivo model of ventilator-induced lung injury.Am J Respir Crit Care Med 2002; 163:403408

22. Bland JM, Altman DG: Statistical methodsfor assessing agreement between two meth-ods of clinical measurement. Lancet 1986;1(8476):307310

23. Knaus WA, Wagner DP, Draper EA, et al: TheAPACHE III prognostic system. Risk predic-tion of hospital mortality for critically illhospitalized adults. Chest 1991; 100:16191636

24. Coakley RJ, Taggart C, Greene C, et al: Am-bient pCO2 modulates intracellular pH, in-tracellular oxidant generation, and interleu-kin-8 secretion in human neutrophils.J Leukoc Biol 2002; 71:603610

25. Laffey JG, Engelberts D, Kavanagh BP: Buff-ering hypercapnic acidosis worsens acutelung injury. Am J Respir Crit Care Med 2000;161:141146

26. Laffey JG, Honan D, Hopkins N, et al: Hyper-capnic acidosis attenuates endotoxin inducedacute lung injury. Am J Respir Crit Care Med2004; 169:4656

27. Bidani A, Wang CZ, Saggi SJ, et al: Evidencefor pH sensitivity of tumor necrosis factor-alpha release by alveolar macrophages. Lung1998; 176:111121

28. Serrano CV Jr, Fraticelli A, Paniccia R, et al:pH dependence of neutrophil-endothelial celladhesion and adhesion molecule expression.Am J Physiol 1996; 271:C962C970

29. Shibata K, Cregg N, Engelberts D, et al: Hy-percapnic acidosis may attenuate acute lunginjury by inhibition of endogenous xanthineoxidase. Am J Respir Crit Care Med 1998;158:15781584

30. Takeshita K, Suzuki Y, Nishio K, et al: Hy-percapnic acidosis attenuates endotoxin-induced nuclear factor-[kappa] B activation.Am J Respir Cell Mol Biol 2003; 29:124132

31. Adding LC, Agvald P, Persson MG, et al: Reg-ulation of pulmonary nitric oxide by carbondioxide is intrinsic to the lung. Acta PhysiolScand 1999; 167:167174

32. Laffey JG, Engelberts D, Duggan M, et al:Carbon dioxide attenuates pulmonary im-pairment resulting from hyperventilation.Crit Care Med 2003; 31:26342640

33. Nuckton TJ, Alonso JA, Kallet RH, et al: Pul-monary dead space fraction as a risk factorfor death in the acute respiratory distresssyndrome. N Engl J Med 2002; 346:12811286

34. Hough CL, Kallet RH, Ranieri VM, et al:Intrinsic positive end-expiratory pressure inAcute 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 InstituteARDS Clinical Trials Network. Higher versuslower positive end-expiratory pressures in pa-tients with the acute respiratory distress syn-drome. N Engl J Med 2004; 351:327336


Documents

Hypercapnic Acidosis and Mortality in Acute Lung Injury