Decrease in PaCO2 with prone position is predictive of improved

outcome in acute respiratory distress syndrome*

Luciano Gattinoni, MD, FRCP; Federica Vagginelli, MD; Eleonora Carlesso, MSC; Paolo Taccone, MD;
Valeria Conte, MD; Davide Chiumello, MD; Franco Valenza, MD; Pietro Caironi, MD; Antonio Pesenti, MD;
for the Prone-Supine Study Group

Objective: To determine whether gas exchange improvement in

response to the prone position is associated with an improved
outcome in acute lung injury (ALI)/acute respiratory distress
syndrome (ARDS).
Design: Retrospective analysis of patients in the pronation arm
of a controlled randomized trial on prone positioning and patients
enrolled in a previous pilot study of the prone position.
Setting: Twenty-eight Italian and two Swiss intensive care
units.
Patients: We studied 225 patients meeting the criteria for ALI
or ARDS.
Interventions: Patients were in prone position for 10 days for 6
hrs/day if they met ALI/ARDS criteria when assessed each morning. Respiratory variables were recorded before and after 6 hrs of
pronation with unchanged ventilatory settings.
Measurements and Main Results: We measured arterial blood
gas alterations to the first pronation and the 28-day mortality rate.
The independent risk factors for death in the general population
were the PaO2/FIO2 ratio (odds ratio, 0.992; confidence interval,
0.986 0.998), the minute ventilation/PaCO2 ratio (odds ratio,
1.003; confidence interval, 1.000 1.006), and the concentration of

he prone position is widely

used in the treatment of patients with severe acute lung
injury (ALI)/acute respiratory
distress syndrome (ARDS), although an

*See also p. 2804.

From Istituto di Anestesia e Rianimazione (LG, FV,
EC, PT, VC, DC, FV, PC), Universit degli Studi di
Milano, Ospedale Policlinico, IRCCS, Milan, Italy; and
Istituto di Anestesia e Rianimazione (AP), Universit di
Milano-Bicocca, Ospedale S. Gerardo, Monza, Italy.
Supported, in part, by Hill-rom Italy, which supported investigators meetings and secretarial activities of the coordinating center.
Members of the Prone-Supine Study Group and
participating institutions are listed in the Appendix.
Address requests for reprints to: Luciano Gattinoni,
MD, FRCP, Istituto di Anestesia e Rianimazione, Ospedale Maggiore Policlinico, IRCCS, Via Francesco
Sforza, 35 20122 Milano, Italy. E-mail:
gattinon@policlinico.mi.it
Copyright 2003 by Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000098032.34052.F9

Crit Care Med 2003 Vol. 31, No. 12

plasma creatinine (odds ratio, 1.385; confidence interval, 1.116

1.720). PaO2 responders (defined as the patients who increased
their PaO2/FIO2 by >20 mm Hg, 150 patients, mean increase of
100.6 61.6 mm Hg [13.4 8.2 kPa]) had an outcome similar to
the nonresponders (59 patients, mean decrease 6.3 23.7 mm
Hg [0.8 3.2 kPa]; mortality rate 44% and 46%, respectively;
relative risk, 1.04; confidence interval, 0.74 1.45, p .65). The
PaCO2 responders (defined as patients whose PaCO2 decreased by
>1 mm Hg, 94 patients, mean decrease 6.0 6 mm Hg [0.8
0.8 kPa]) had an improved survival when compared with
nonresponders (115 patients, mean increase 6 6 mm Hg [0.8
0.8 kPa]; mortality rate 35.1% and 52.2%, respectively; relative
risk, 1.48; confidence interval, 1.072.05, p .01).
Conclusion: ALI/ARDS patients who respond to prone positioning with reduction of their PaCO2 show an increased survival at 28
days. Improved efficiency of alveolar ventilation (decreased
physiologic deadspace ratio) is an important marker of patients who will survive acute respiratory failure. (Crit Care Med
2003; 31:27272733)
KEY WORDS: acute lung injury; acute respiratory distress syndrome; prone position; carbon dioxide exchange

initial outcome study failed to show any

enhancing advantage in survival rates (1).
The main indication for prone positioning is severe hypoxemia: Indeed, prone
positioning is associated with a significant and lasting improvement in oxygenation in 60 80% of patients (2). However, we do not yet know whether the
improvements in gas exchange in the
prone position are associated with an enhanced outcome and, if so, which are the
possible salutary mechanisms (3).
We began to examine these questions
by retrospectively analyzing the database
of the prone/supine study (1), in which
152 ALI/ARDS patients were in prone position for 10 days. To these patients, we
added a group of 73 patients (unpublished data) enrolled in the pilot study
performed before we began the randomized trial. We wish to report our findings,
focusing on the association between im-

proved respiratory gas exchange in the

prone position and outcome.

MATERIALS AND METHODS

ALI/ARDS patients were recruited from 28
intensive care units in Italy and two in Switzerland (Appendix). One hundred and fifty-two
patients were randomized to the pronation
arm of a recently reported prone/supine trial
(1). To these, we added 73 patients who had
followed the same protocol during the pilot
study used to test the feasibility and optimize
the randomized trial. Because the enrollment
criteria and baseline characteristics of these
73 patients were similar to the reported trial,
we pooled both populations into a single
group of 225 patients. The records of 16 patients, however, lacked some important data
and were excluded from our analysis. The
study examined data from 209 patients.
The entry criteria for the study were ALI
(200 PaO2/FIO2 300 mm Hg and/or positive end-expiratory pressure 10 cm H2O) or

2727

ARDS (PaO2/FIO2 200 mm Hg with positive

end-expiratory pressure 5 cm H2O). Bilateral
chest radiograph infiltrates without evidence
of cardiac failure were required for enrollment. Exclusion criteria were age 16 yrs,
cardiogenic edema, intracranial hypertension,
or other clinical conditions contraindicating
the use of prone positioning, such as fractures
of the spine or severe hemodynamic instability. Informed consent procedures complied
with the European Guidelines for Good Clinical Practice (4), and the study was approved
by the Health Authority of the Regione Lombardia and by the local ethics committees of
the participating institutions.
Study Protocol. The entire study protocol
has already been published (1) and is only
summarized here. Briefly, the study lasted 10
days. Patients meeting the gas exchange criteria defined for enrollment at the morning
control (in the supine position) were in prone
position for 6 hr.
The physicians caring for the patients were
asked to comply with the guidelines of the
American-European Consensus Conference on
mechanical ventilation for routine ventilatory
management (5).
The information collected relative to ventilation were gas exchange variables (PaO2,
PaCO2, and pH), the total minute ventilation
(VE), respiratory rate, FIO2, and positive endexpiratory pressure, all at baseline and 6 hrs
after prone positioning. The physicians also
were asked to report the ventilator mode (control ventilation or assisted ventilation). We did
not collect more detailed information about
the precise ventilatory mode (volume controlled or pressure controlled, etc.). For the
purpose of this study, however, the ventilator
setting was left the same when the patient was
moved from supine to prone position for the
sake of standardization. Every morning we
recorded the Simplified Acute Physiology
Score II (6) and Sepsis-related Organ Failure
Assessment (7) as indexes of illness severity
and the respiratory and biochemical variables
indicative of organ system failure (renal, hepatic, circulatory, and coagulation failure), as
defined by the Acute Respiratory Distress Syndrome Network trial of the U.S. National
Heart, Lung and Blood Institute (8). Sedation
was assessed by the Ramsey scale (9).
The gas exchange response to the first pronation was analyzed in detail as an early predictor of outcome. To define the PaO2 responders, we used different cutoff values of
PaO2/FIO2 increase: those patients whose PaO2/
FIO2 ratio increased 1 mm Hg (0.133 kPa) at
the end of the first 6 hrs of pronation (3, 10),
those patients in whom the PaO2/FIO2 increased 20 mm Hg (2.66 kPa), and those in
whom the PaO2/FIO2 increased more than the
median value. We defined PaCO2 responders
those patients who decreased their PaCO2 1
mm Hg (0.133 kPa) after 6 hrs in the first
pronation and as PaCO2 nonresponders those

2728

patients who did not change or increased their

PaCO2 at 6 hrs.
Statistical Analysis. The outcome variable
was death at 28 days. This was used instead of
mortality rate at 6 months, which was not
available for the patients in the pilot study. A
paired or unpaired Students t-test was used to
compare means of variables for survivors and
nonsurvivors. A backward stepwise logistic regression model was used to identify the variables independently associated with death.
SAS computer software package (SAS Institute, Cary, NC) was used.

RESULTS
The main characteristics of the study
population are summarized in Table 1.
ARDS was diagnosed in 201 patients and
ALI in eight (11). At entry into the study,
survivors and nonsurvivors (at 28 days)
differed significantly in a number of vari-

ables (Table 1). These variables were introduced in a backward stepwise logistic
regression, and only three of them remained as independent risk factors of
death: the PaO2/FIO2 ratio as a marker of
the severity of oxygen exchange failure
(odds ratio [OR], 0.992; confidence interval [CI], 0.986 0.998), the VE/PaCO2 ratio, which is a rough surrogate for the
deadspace tidal volume ratio (VD/VT) (OR,
1.003; CI, 1.000 1.006), and the plasma
creatinine concentration (odds ratio,
1.385; CI, 1.116 1.720).
Time Course of the Pronation Response. Figure 1 summarizes for the entire group the time course of arterial oxygenation and PaCO2 response during the
10 days. The positive response to oxygenation was maintained over all 10 days,
although the trend was toward a slight

Table 1. Baseline characteristics

No.
Age, yrs
BMI, kgm2
SAPS II at enrollmenta
ALI/ARDSb
Days from intubation to first pronation
Cause of lung injury, %
Pneumonia
Aspiration
Other type of respiratory disease
Respiratory tract infection after surgery
Sepsis
Trauma
Other causes
Organ failure
Platelets, 109/L
Bilirubin, mol/L
Urea, mmol/L
Creatinine, mol/L
Ventilatory and gas exchange values
VT/kg ideal body weight, mL/kgc
Ve, L/min
PEEP, cm H2O
Mean Paw, cm H2O
PaO2/FIO2, mm Hg
PaCO2, mm Hg
pHa
Ve/PaCO2, mLmin1mm Hg1
VT, mL
Hemodynamics
Heart rate, beats/min
Mean arterial pressure, mm Hg
Systolic arterial pressure, mm Hg
Diastolic arterial pressure, mm Hg

Survivors

Nonsurvivors

116
56 17
24.8 3.9
39 13
5/111
3.8 6

93
59 18
24.6 16.1
42 12
3/90
4.5 5.6

.47
.25
.06
.73
.41

50.9
0.9
10.3
6.0
7.8
3.4
20.7

47.3
3.2
14.0
5.4
8.6
2.1
19.3

.71
.46
.55
.92
.97
.89
.95

205 122
27 50
27 24
123 106

167 140
60 120
35 28
177 150

.04
.01
.03
.002

10.4 2.7
10.0 2.4
9.9 2.7
16.9 4.0
135 44
45.2 11.3
7.39 0.09
237 90
667.8 174.8

10.4 2.7
11.3 3.4
9.9 2.9
17.7 4.8
115 47
45.3 12.6
7.38 0.09
269 110
639.8 148.2

.82
.001
.89
.18
.001
.97
.37
.02
.22

101 21
84 14
124.9 21.4
63.5 13.4

105 18
85 15
123 24.2
65.8 14

.29
.66
.56
.23

BMI, body mass index; SAPS, Simplified Acute Physiology Score; ALI, acute lung injury; ARDS,
acute respiratory distress syndrome; VT, tidal volume; Ve, total minute ventilation; PEEP, positive
end-expiratory pressure; Paw, airway pressure; pHa, arterial pH.
a
SAPS is used to assess the severity of illness and can range from 0 to 194, with higher scores
indicating a higher risk of death; bALI is characterized by a PaO2/FIO2 ratio of 300 mm Hg (39.9
kPa). ARDS is characterized by a PaO2/FIO2 ratio of 200 mm Hg (26.6 kPa); cthe predicted body
weight was calculated as 50 0.91 (height, 152.4 cm) for males and as 45.5 0.91 (height, 152.4 cm)
for females. Data are mean SD.

Crit Care Med 2003 Vol. 31, No. 12

decrease. Although there was not a clear

group trend for PaCO2 response over the
10 days, it is interesting to note that the
baseline PaCO2 level of the 46 patients
who remained in the study and were in
prone position to the 10th day did progressively and significantly increase (day
1 baseline PaCO2, 45 10 mm Hg [5.98
1.33 kPa]; day 10 baseline PaCO2, 53 17
mm Hg [7.05 2.26 kPa, p .001])
despite a similar VE (10.97 L/min 3.37
and 11.64 3.32, respectively, p .32).
Response to First Pronation. Table 2
reports the changes in respiratory variables during the first prone positioning
with unmodified ventilator settings.
Changes that were significantly different
between survivors and nonsurvivors were
introduced into a backward stepwise logistic regression model. Only the PaCO2
changes were independently associated
with outcome (OR, 1.051; CI, 1.013
1.091). Changes in hemodynamic variables were not shown as no statistical
difference between survivors and nonsurvivors was found.
Oxygenation Response. Considering
the entire population, the increase of arterial oxygenation at the end of the first
pronation (6 hrs) was dramatic with a
mean PaO2/FIO2 increase of 70 72 mm
Hg (9.31 9.58 kPa); median, 58 mm Hg
(7.71 kPa); range, 77 (10.24) to 318
(42.29 kPa) mm Hg. Seventy patients
shifted from ARDS to the less severe ALI
criteria just after prone positioning, and
21 no longer met the ALI criteria. The
frequency distribution of the Pa O 2
changes, provided as the difference between baseline and the end of 6 hrs of
pronation, is summarized in Figure 2,
upper panel. The prone position response
of arterial oxygenation, however, was not
associated with an improved 28-day outcome, as shown in Figure 2, lower panel.
The death rates of the PaO2 responders, as
classified according to the different cutoff
of PaO2/FIO2 (10), are reported in Table 3.
As shown, the mortality rate was similar
in responders and nonresponders, whatever criteria of classification were used.
Carbon Dioxide Response. Figure 3,
upper panel, shows the frequency distribution of the PaCO2 responses to the
prone positioning for the entire population. The average PaCO2 change was 0.4
8.2 mm Hg (0.05 1.09 kPa); median, 1
mm Hg (0.133 kPa); range, 30 to 39
mm Hg (from 3.99 to 5.18 kPa). The
PaCO2 decreased in 94 of 209 patients
(PaCO2 responders, mean PaCO2 decrease
6.0 5.6 mm Hg [0.79 0.74 kPa], p
Crit Care Med 2003 Vol. 31, No. 12

Figure 1. Mean values of PaO2/FIO2 ratio and PaCO2 immediately before prone positioning (open circles)
and at the end of the 6 hrs pronation (closed circles) during the 10-day study period. (Each calculation
includes only data for patients with all four measurements). The bars show the number of patients who
were in prone position each day for whom all four measurements were available.
Table 2. Ventilatory and gas exchanges changes after the first pronation

VT/kg ideal body weight, mL/kg

Ve, L/min
PEEP, cm H2O
Paw, cm H2O
PaO2/FIO2, mm Hg
PaCO2, mm Hg
pHa
Ve/PaCO2, mLmin1mm Hg1
Oxygenation index

Survivors

Nonsurvivors

0.1 1.2
0.04 1.60
0.1 1.0
0.2 1.7
70. 72
0.9 7.9
1.2 0.8
4 67
4.1 5.8

0.1 1.6
0.35 1.78
0.0 0.7
0.3 1.8
71 73
2.1 8.4
1.0 0.8
20 70
6.2 9.8

.51
.20
.28
.71
.93
.008
.10
.01
.07

VT, tidal volume; Ve, total minute ventilation; PEEP, positive end-expiratory pressure; Paw, airway
pressure. Data are mean SD.

.0001), whereas PaCO2 remained constant or increased in 115 (PaCO2 nonresponders, mean PaCO2 increase 5.6 6.1
mm Hg [0.74 0.81 kPa]). The survival
rate at 28 days was significantly different
for PaCO2 responders and PaCO2 nonresponders (log rank, p .01, Fig. 3, lower
panel). Moreover, it appears that there is
a quantitative relationship between the
PaCO2 response and outcome: In fact,
considering the PaCO2 not as a dichotomous but as a continuous variable, its
relationship with outcome is shown in

Figure 4, with a progressive increase of

mortality rate associated with the greatest PaCO2 increase after pronation.
The 28-day mortality rate for the
whole population, for the 153 patients
treated with controlled ventilation and
for the 56 patients treated with assisted
ventilation, is reported in Table 4.
Combined Oxygen and Carbon Dioxide Response. The mortality rate for patients grouped according to their combined PaCO2 and PaO2/FIO2 response to
prone position is summarized in Table 5.
2729

As shown, the outcome is only related to

the PaCO2 changes during the prone period.
Hemodynamic and Metabolic Response. The hemodynamic and metabolic
variables in the PaCO2 responders and
nonresponders at baseline, at 6 hrs of
prone positioning, and at 24 hrs (in supine) are summarized in Table 6.
As shown, nonsignificant differences
were found between the two groups, with
the exception of VE and PaO2/FIO2.

DISCUSSION

Figure 2. Upper panel, PaO2/FIO2 responses at the first pronation (209 patients). The PaO2 responders
are patients whose PaO2/FIO2 ratio increased 20 mm Hg (2.66 kPa). Similar curves are obtained,
however, using other cutoff criteria (Table 3). Lower panel, Kaplan-Meier estimates of survival at 28
days in PaO2 responders and PaO2 nonresponders (p .65 by the log-rank test).

Table 3. PaO2 responders, classified according to the different cutoff of PaO2/FIO2 increase and mortality
rate

No. responders/nonresponders
Mortality rate, %
Log-rank
Chi-square
Relative risk of death
95% Confidence interval
Data are mean

2730

SD.

PaO2/FIO2
1 mm Hg

PaO2/FIO2
20 mm Hg

PaO2/FIO2 median
(58 mm Hg)

182/27
45/41
0.9486
0.67
0.90
0.561.46

150/59
44/46
0.6541
0.81
1.04
0.741.45

104/105
46/43
0.7906
0.63
0.93
0.691.26

Over the past two decades, 150 articles have discussed the prone position in
ALI/ARDS. The majority deal with acute
physiologic effects, and all focus on oxygenation. Only a few have reported PaCO2
changes, which were generally insignificant and considered of little or no importance (10, 12, 13). In this study, however,
we found a key role for changes in PaCO2
as an early prognostic index in the setting
of ALI/ARDS during prone positioning.
That increases in PaCO2 over time are
associated with a worse outcome in lasting ARDS has been known for many
years. In late ARDS, these increases reflect, more than a deterioration of oxygenation, major structural changes of the
lung, such as the development of pneumatoceles or pseudocysts (14). As expected, in this study we found a progressive increase of PaCO2 with time (Fig. 1).
The prognostic importance of CO2 exchange in early ARDS, as reflected by
increased pulmonary deadspace fraction,
was noted only recently (15). Our findings are consistent with that report. We
also found that the VE/PaCO2 ratio, which
is a crude surrogate of the deadspace to
tidal volume ratio, is an independent risk
factor of death in early ALI/ARDS.
However, the most important findings
were the differing prognostic values of
the PaCO2 and PaO2/FIO2 responses to the
first pronation. We found a strong association between the PaCO2 response and
outcome but no association with the
PaO2/FIO2 response to pronation whatever
criteria we used to define the PaO2/FIO2
responders.
The alveolar P CO 2 (P A CO 2 ) which
closely reflects PaCO2 if the shunt fraction
is not taken in account, is proportional to
the ratio of the rates of CO2 elimination
(VCO2) and alveolar ventilation (VA
VE[1 VD/VT]) according to the following equation:
PaCO2 (VCO2/VA)863

[1]

Crit Care Med 2003 Vol. 31, No. 12

Figure 4. Progressive increase of mortality rate

associated with the greatest PaCO2 increase after
first pronation. This relationship is calculated
from the backward stepwise logistic regression
results.

Figure 3. Upper panel, PaCO2 responses at the first pronation (209 patients). Lower panel, KaplanMeier estimates of survival at 28 days in PaCO2 responders and PaCO2 nonresponders (p .01 by the
log-rank test).

The VCO2 and VA, in turn, depend on a

number of variables that may change
during the course of the disease such as
the metabolic rate, cardiac output, right
to left shunt, VE, VD/VT, and CO2 content
in mixed venous and lung capillary blood
(16). A key question is whether the different outcome associated with PaCO2
changes during the 6 hrs of pronation is
related to prone position per se or
whether there were other factors, during
that period, that changed the PaCO2, thus
selecting two populations with different
survival independently of prone position.
As an example, during the 6-hr period of
prone position, the PaCO2 in nonresponders could increase because of an
increase in VCO2 due to worsening of sepCrit Care Med 2003 Vol. 31, No. 12

sis or because of an increase of VD/VT due

to hypoperfusion of the lung capillaries in
underresuscitated patients. In this scenario, the association between P CO 2
changes and outcome would simply reflect the natural course of PCO2 within the
disease process, the association with
prone position being purely incidental.
If so, we should find some difference
between PaCO2 responders and nonresponders in some of the variables associated with disease severity or sepsis, such
as Simplified Acute Physiology Score,
Sepsis-related Organ Failure Assessment,
hemodynamics, level of sedation, and kidney, hepatic, or hematological function.
None of these variables, measured at 6
and 24 hrs from baseline, significantly

differed between PaCO2 responders and

nonresponders (Table 6). This suggests
that the natural course of the disease,
at least over the first 24 hrs, was similar
in the two groups.
To exclude the effects of time and
cointerventions, the ideal approach
would be to have a control group in supine. Unfortunately, in our randomized
trial, the PaCO2 was not measured 6 hrs
after baseline in the supine arm. However, we did measure it 24 hrs after baseline in both the supine and prone arms.
We could then identify the 24-hr PaCO2
responders and nonresponders using the
same criteria that we used to classify
the 6-hr response. Sixty-two patients in
the supine arm were 24-hr PaCO2 responders, and 76 were not. Mortality
rates were 40.3% vs. 43.4%, respectively
(p .71). In the prone arm (in which the
24-hr PaCO2 response was measured in
the supine position), 69 patients were responders and 73 were not, and mortality
rate was not significantly different (37.7
and 50.7%, respectively, p .11).
Most data strongly suggest that it is
not the natural course of the PaCO2, at
least over 24 hrs, that is predictive for
outcome, but rather the PaCO2 response
to prone maneuver. In our opinion, the
different responses reveal different underlying pathologies, and this is what is
associated with outcome, not the change
of PaCO2 per se.
As PaCO2 changes during prone positioning are unlikely to be due to variations of VCO2, they must be related to
changes in alveolar ventilation, which,
despite an unmodified ventilator setting,
may be due to changes of VE and/or VD/VT
2731

due to prone positioning. Different things

may happen, to different extents, during
prone positioning. These include increased elastance of the chest wall (17),
possible recruitment of dorsal regions associated with derecruitment of ventral regions (18), and, possibly, blood flow redistribution (19). The resulting gas

exchange depends on the interaction of

all these factors.
The PaO2 may increase through two
basic mechanisms and their various combinations. The first is the blood flow diversion from nonaerated to aerated regions of the lung. In this case, with
unmodified VA, the PaCO2 should de-

Table 4. 28-Day mortality rate according PaCO2 response

All patients
Responders
Nonresponders
Assisted ventilation
Responders
Nonresponders
Controlled ventilation
Responders
Nonresponders
Data are mean

No. of
Patients

No. of
Deaths

% of
Mortality

Relative
Risk

Confidence
Interval

94
115

33
60

35.1
52.2

1.48

1.072.05

.02

20
36

6
19

30.0
52.8

1.76

0.843.68

.10

74
79

27
41

36.5
51.9

1.42

0.952.05

.055

SD.

Table 5. Combined oxygen and carbon dioxide response

Gas mm Hg,
6 Hrs to Baseline

Group
1 PaCO2 responders
PaO2/FIO2 responders
2 PaCO2 responders
PaO2/FIO2 nonresponders
3 PaCO2 nonresponders
PaO2/FIO2 responders
4 PaCO2 nonresponders
PaO2/FIO2 nonresponders
Data are mean

PaCO2: 6.1 5.7

PaO2/FIO2: 109 68
PaCO2: 5.31 5.3
PaO2/FIO2: 5.17 21.7
PaCO2: 5.47 5.86
PaO2/FIO2: 91.2 54.3
PaCO2: 5.9 6.6
PaO2/FIO2: 7.4 24.9

28-Day
No. of No. of Mortality
Patients Deaths Rate, %
75

26

34.7

19

36.84

76

41

53.9

39

19

48.7

p
vs. 2 p .72
vs. 3 p .019

vs. 2 p .38
vs. 3 p .65

SD.

crease slightly because of the reduction of

shunted blood (i.e., the shunted mixed
venous PCO2 contributes less to the increase of PCO2 into the arterial blood) and
because more CO2 is available for exchange in aerated regions. However, if at
the same time the VE and VA decrease
because of the increased stiffness of the
thorax, the final result will be the association between increased PaO2 and unmodified or increased PaCO2.
The second basic mechanism by
which PaO2 can increase is the recruitment to aeration of previously collapsed
and perfused regions. In this case, the
PaO2 largely increases because of recruitment and the PaCO2 decreases primarily because of increased VA, as part
of the VE is delivered to the newly recruited and perfused lung regions.
However, the effects of recruitment in
prone position are partly obscured by
the increased chest wall elastance,
which may reduce VE and VA. However,
if the recruitment prevails on chest wall
impairment, the final gas exchange will
be an increased PaO2 associated with a
decrease of the PaCO2.
Indeed, one possible explanation for
our findings is that in the PaCO2 nonresponders, the primary mechanism of the
PaO2 increase is diversion of the blood
flow, whereas in the PaCO2 responders the
primary mechanism is recruitment. Indeed, in the PaCO2 nonresponders, the
chest wall impairment prevails over recruitment, whereas in the PaCO2 responders the recruitment prevails over the
chest wall impairment.

Table 6. Hemodynamic and metabolic variables in the PaCO2 responders and nonresponders at baseline, at 6 hrs of prone positioning, and at 24 hrs (in
supine)
Baseline

Patients
Systolic pressure, mm Hg
Diastolic pressure, mm Hg
Heart rate, beats/min
Minute ventilation, L/min
PaO2/FIO2, mm Hg
Creatinine, mg/100 mL
Platelets105, no./mm3
Bilirubin, mg/100 mL
SAPS
SOFAb,c
Ramsey scorec,d

6 Hrs Prone

24 Hrs Supine

Nonresponders

Responders

Nonresponders

Responders

Nonresponders

Responders

115
125 22
66 14
102 20
10.6 3.1
125.4 44.6
1.7 1.6
1.8 1.4
2.7 6.3
40.4 12.1
8.33 3.43
5.26 1.73

94
123 24
63 14
104 20
10.5 2.8
127.1 48.8
1.6 1.3
1.9 1.1
2.2 3.6
40.1 13.9
9.15 3.88
5.78 1.44

115
126 22
65 13
103 21
10.0 2.8a
183.2 80.2a

94
122 22
64 13
104 20
10.8 2.9
213.1 85.7

115
128 21
65 13
101 19
10.8 3.0
156.2 67.0
1.8 1.9
1.7 1.3
2.9 7.1

8.43 3.57
5.52 1.51

94
123 20
63 11
101 23
10.7 3.5
158.9 72.5
1.8 1.6
1.9 1.2
2.7 4.8

9.00 3.58
5.63 1.56

SAPS, Simplified Acute Physiology Score; SOFA, Sepsis-related Organ Failure Assessment. Data are mean SD.
p .05 vs. responders; bSOFA is used to describe organ dysfunction/failure and can range from 0 to 24, with higher scores indicating higher severity;
c
values reported refer to randomized trial patients only; dthe Ramsey scale is used to describe the sedation level; it can range from 1 to 6, where 1 indicates
anxious and agitated and 6 indicates nonresponsive.
a

2732

Crit Care Med 2003 Vol. 31, No. 12

ndependently of the
ventilator

setting,

which was maintained

unmodified, the patients

who reacted to prone positioning with a decrease in
PaCO2 had a better outcome
than the patients who did
not.

CONCLUSIONS
This study has several limitations. It is
retrospective, a number of physiologic
variables were not collected (because to
ensure that a multiple-center trial is feasible we have to keep variables to a minimum), and a control group at 6 hrs is
not available. However, a single fact remains true: Independently of the ventilator setting, which was maintained unmodified, the patients who reacted to
prone positioning with a decrease in
PaCO2 had a better outcome than the patients who did not. The PaCO2 change per
se does not seem to influence the outcome. However, the change during prone
positioning probably indicates some difference in the underlying pathology: a
lower potential for recruitment in PaCO2
nonresponders and a higher potential in
responders.

APPENDIX
Steering Committee: L. Gattinoni, G.
Tognoni. Scientific and Organizing Secretariat: L. Brazzi, R. Fumagalli, R.
Latini, R. Malacrida, D. Mascheroni, P.
Pelosi, A. Pesenti, G. Ronzoni. Nursing
Coordination: I. Adamini, D. Brambilla,
P. Di Giulio. Data Safety Monitoring
Board: B. Andreoni, P. Suter, M. G.
Valsecchi. Data Management and Analysis: L. Donati, V. Torri, V. Labarta. Participating Centers: Antella, S. M. Annunziata: U. Boncristiano, L. Manenti, A.
Orvieto; Arezzo, S. Donato: C. Boncompaghi, V. Capria, F. Magnanensi, C.
Recine; Asti: S. Cardellino, E. Costanzo,
M. Gavioso, A. Scotti; Como, SantAnna:
M. F. Magatti, P. Rossitto, A. Villa; Conegliano: U. Corbanese, M. De Zotti; Cuneo:
Crit Care Med 2003 Vol. 31, No. 12

V. Calleris, A. Ghigo; Faenza: P. P.

Casalini, S. Maitan, G. Pezzi, F. Sfogliaferri; Firenze, Careggi: R. Lippi, E. Pagni,
G. Pierotti; Firenze, Nuovo S. Giovanni di
Dio: S. Gori, V. Mangani, R. Oggioni, G.
Tulli; Legnano: P. Cortellazzi, O. Masiero,
D. Pinciroli; Locarno, Switzerland: P.
Delen, G. Domenighetti, G. Pennati; Lugano, Switzerland: A. Canonica, M.
Kralije, R. Malacrida, G. Vanini; Merate:
G. Casiraghi, S. Bonfanti, M. Marinari, G.
Rota; Milano, Centro S. Luigi: G. Cornaggia, N. Francavilla, F. Roveda; Milano, L.
Sacco: G. Donatiello, E. Malerba, R. Ravagnan, C. Zocchi; Milano, Maggiore Policlinico: I. Adamini, L. Gattinoni, D. Mascheroni; Milano, Niguarda C Granda: D.
Carugo, I. Fontana, E. Pannacciulli, A.
Ravizza; Milano, S. Paolo: R. Bestoso, G.
Iapichino, G. Zanforlin; Milano, S. Raffaele: E. Moizo, N. Monzani, S. Muttini,
G. Torri, Monza: D. Brambilla, S. Del
Cotto, A. Frigerio, L. Lampati, R. Fumagalli; Napoli, Universit Federico II: R.
Iaselli, G. Servillo, R. Tufano; Pavia, S.
Matteo: M. P. Buccellini, A. Braschi, G.
Rodi; Pistoia, Riuniti: L. Barontini, I.
Biagi, M. Covolo, V. Piacentino; Riccione,
Ceccarini: L. Bianchi, F. Cecchini, S.
Montanari, M. Sanseverino; Rimini, Infermi: P. Catorcini, A. Corsi; Roma, S.
Eugenio: A. Colantoni, A. Dauri, M. Stefanelli, E. Zupancich; Sondalo: D. Caldiroli, A. Vairetti; Torino, S. Giovanni Battista: R. Cavallo, E. Molinaro, D. Silengo,
R. Urciuoli; Treviglio: G. C. Aresi, M.
Borelli, P. DAdda, M. Morbelli; Vimercate: E. Basilico, C. Biffi, R. Citterio.

7.

8.

9.

10.

11.

12.

13.

14.

15.

REFERENCES
1. Gattinoni L, Tognoni G, Pesenti A, et al:
Effect of prone positioning on the survival of
patients with acute respiratory failure.
N Engl J Med 2001; 345:568 573
2. Gattinoni L, Pelosi P, Valenza F, et al: Patient
positioning in acute respiratory failure. In:
Principles and Practice of Mechanical Ventilation. Tobin MJ (Ed). New York, McGrawHill, 1994, pp 10671076
3. Chatte G, Sab JM, Dubois JM, et al: Prone
position in mechanically ventilated patients
with severe acute respiratory failure. Am J
Respir Crit Care Med 1997; 155:473 478
4. Human Medicines Evaluation Unit: ICH
Topic E6: Guideline for Good Clinical Practice. London, European Agency for the Evaluation of Medical Production, 1996, pp
1720
5. Slutsky AS: Mechanical ventilation. American College of Chest Physicians Consensus
Conference. Chest 1993; 104:18331859
6. Le GJ, Lemeshow S, Saulnier F: A new Simplified Acute Physiology Score (SAPS II)

16.

17.

18.

19.

based on a European/North American multicenter study. JAMA 1993; 270:29572963

Vincent JL, Moreno R, Takala J, et al: The
SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/
failure. On behalf of the Working Group on
Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive
Care Med 1996; 22:707710
ARDSnet: 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
Ramsey MAE, Savege TM, Simpson BRJ, et
al: Controlled sedation with alphaxalonealphadolone. BMJ 1974; 2:656 659
Jolliet P, Bulpa P, Chevrolet JC: Effects of the
prone position on gas exchange and hemodynamics in severe acute respiratory distress
syndrome. Crit Care Med 1998; 26:
19771985
Bernard GR, Artigas A, Brigham KL, et al:
The American-European Consensus Conference on ARDS. Definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994; 149:
818 824
Vollman KM, Bander JJ: Improved oxygenation utilizing a prone positioner in patients
with acute respiratory distress syndrome. Intensive Care Med 1996; 22:11051111
Blanch L, Mancebo J, Perez M, et al: Shortterm effects of prone position in critically ill
patients with acute respiratory distress syndrome. Intensive Care Med 1997; 23:
10331039
Gattinoni L, Bombino M, Pelosi P, et al: Lung
structure and function in different stages of
severe adult respiratory distress syndrome.
JAMA 1994; 271:17721779
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
Farhi LE: Ventilation-perfusion relationship.
In: Handbook of Physiology, Section 3: The
Respiratory System, Volume IV: Gas Exchange. American Physiological Society
(Eds). Baltimore, MD, Waverly Press, 1985,
pp 199 215
Pelosi P, Tubiolo D, Mascheroni D, et al:
Effects of the prone position on respiratory
mechanics and gas exchange during acute
lung injury. Am J Respir Crit Care Med 1998;
157:387393
Gattinoni L, Pelosi P, Vitale G, et al: Body
position changes redistribute lung computed-tomographic density in patients with
acute respiratory failure. Anesthesiology
1991; 74:1523
Richard JC, Decailliot F, Janier M, et al: Effects of positive end-expiratory pressure and
body position on pulmonary blood flow redistribution in mechanically ventilated normal pigs. Chest 2002; 122:998 1005

2733