**Comparison of the SpO2/FIO2 Ratio and the PaO2/FIO2 Ratio in Patients With Acute Lung Injury or ARDS***

Todd W. Rice, MD, MSc; Arthur P. Wheeler, MD, FCCP; Gordon R. Bernard, MD, FCCP; Douglas L. Hayden, MA;

David A. Schoenfeld, PhD; and Lorraine B. Ware, MD, FCCP; for the National Institutes of Health, National Heart, Lung, and Blood Institute ARDS Network

Background: The diagnostic criteria for acute lung injury (ALI) and ARDS utilize the PaO2/ fraction of inspired oxygen (FIO2) [P/F] ratio measured by arterial blood gas analysis to assess the degree of hypoxemia. We hypothesized that the pulse oximetric saturation (SpO2)/FIO2 (S/F) ratio can be substituted for the P/F ratio in assessing the oxygenation criterion of ALI.

Methods: Corresponding measurements of SpO2 (values < 97%) and PaO2 from patients enrolled in the ARDS Network trial of a lower tidal volume ventilator strategy (n = 672) were compared to determine the relationship between S/F and P/F. S/F threshold values correlating with P/F ratios of 200 (ARDS) and 300 (ALI) were determined. Similar measurements from patients enrolled in the ARDS Network trial of lower vs higher positive end-expiratory pressure (n = 402) were utilized for validation.

Results: In the derivation data set (2,613 measurements), the relationship between S/F and P/F was described by the following equation: S/F = 64 + 0.84 × (P/F) [p < 0.0001; r = 0.89). An S/F ratio of 235 corresponded with a P/F ratio of 200, while an S/F ratio of 315 corresponded with a P/F ratio of 300. The validation database (2,031 measurements) produced a similar linear relationship. The S/F ratio threshold values of 235 and 315 resulted in 85% sensitivity with 85% specificity and 91% sensitivity with 56% specificity, respectively, for P/F ratios of 200 and 300. Conclusion: S/F ratios correlate with P/F ratios. S/F ratios of 235 and 315 correlate with P/F ratios of 200 and 300, respectively, for diagnosing and following up patients with ALI and ARDS.

(CHEST 2007; 132:410 – 417)

Key words: acute lung injury; ARDS; definition; Pao2/fraction of inspired oxygen ratio

Abbreviations: AECC = American European Consensus Conference; ALI = acute lung injury; AUC = area under the curve; CI = confidence interval; Fio2 = fraction of inspired oxygen; PBW = predicted body weight; PEEP = positive end-expiratory pressure; P/F = Pao2/fraction of inspired oxygen; ROC = receiver operator characteristic; S/F = pulse oximetric saturation/fraction of inspired oxygen; Spo2 = pulse oximetric saturation

A

cute lung injury (ALI) and the ARDS are devas- tating clinical syndromes with high morbidity and mortality.1,2 Acute hypoxic respiratory failure, as defined by the Pao2/fraction of inspired oxygen (Fio2) ratio (or P/F ratio) is one of the criteria for ALI/ARDS that was developed by an American European Consensus Conference (AECC) in 1994.3 A P/F ratio ≤ 300 and ≤ 200, respectively, are

utilized to define ALI and ARDS.3

Despite the straightforward nature of the AECC definition of ALI and ARDS, the requirement for

arterial blood gas sampling may contribute to the underdiagnosis of these syndromes.4 Concerns about anemia, excessive blood draws, and a movement to minimally invasive approaches have led to fewer arte- rial blood gas measurements in critically ill patients.5–7 In healthy subjects, changes in Pao2 correlate well with changes in pulse oximetric saturation (Spo2) for oxygen saturation in the range of 80 to 100%.8 –10 However, studies in critically ill patients, especially those with ALI/ARDS, are lacking. Furthermore, threshold values for Spo2/Fio2 (S/F) ratios could be used as noninvasive

criteria for diagnosing ALI/ARDS. In this study, we sought to derive and validate the relationship between S/F and P/F ratios in critically ill patients with ALI/ ARDS. We hypothesize that the continuously available S/F ratio can be used as a surrogate for the P/F ratio in the diagnosis of ALI/ARDS. The use of the S/F ratio may better facilitate the screening and rapid identifica- tion of patients with ALI/ARDS while avoiding the blood use and cost for blood gas determinations.

Materials and Methods

Derivation Data Set

Corresponding measurements of Spo2 and Pao2 from patients enrolled in the National Heart, Lung, and Blood Institute ARDS Network trial11 comparing tidal volumes of 6 mL/Kg predicted body weight (PBW) with those of 12 mL/kg were utilized to establish the relationship between S/F and P/F ratios. Each ARDS Network site received approval from local institutional review boards to conduct the studies. The inclusion and exclusion criteria for the ARDS Network tidal volume study11 have been reported elsewhere. All patients underwent measurements of Spo2 and Pao2 with documentation of inhaled concentrations of oxygen at study enrollment and as clinically indicated prior to study day 28 or achieving unassisted breathing. Research person- nel were instructed to document Spo2 values at the time of arterial blood gas sampling. In rare cases when this was not possible, the Pao2 measurement closest to the Spo2 value was utilized. The following measures were employed to improve the accuracy of the Spo2 measurements: optimal position and clean- liness of the sensor; satisfactory waveforms; no position changes or endobronchial suctioning for at least 10 min prior to the measurement; and no invasive procedures or ventilator changes for at least 30 min prior to the measurement.12 Spo2 was observed for a minimum of 1 min before the value was recorded. Because the P/F ratio cutoffs used to diagnosis ALI/ARDS differ

at lower barometric pressures, patients who were enrolled in the study at centers located > 1,000 m in altitude (eg, Salt Lake City and Denver) were excluded from the data sets. Measurements

*From the Division of Allergy, Pulmonary, and Critical Care Medicine (Drs. Rice, Wheeler, Bernard, and Ware), Department of Medicine, Vanderbilt University School of Medicine, Nash- ville, TN; ARDS Network Clinical Coordinating Center (Mr. Hayden and Dr. Schoenfeld), Massachusetts General Hospital, Boston, MA.

This research was funded by National Institutes of Health grants N01-HR-46054 (to Drs. Rice, Wheeler, and Bernard), N01-HR- 46064 (to Mr. Hayden and Dr. Schoenfeld), HL07123 (to Dr. Rice), HL70521 (to Dr. Ware), and HL81332 (to Dr. Ware) from the National Heart, Lung, and Blood Institute.

The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Manuscript received March 20, 2007; revision accepted May 2,

2007.

Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestjournal. org/misc/reprints.shtml).

Correspondence to: Todd W. Rice, MD, MSc, Division of Allergy, Pulmonary, and Critical Care Medicine, T-1218 MCN, Nashville, TN 37232-2650; e-mail: [email protected]

DOI: 10.1378/chest.07-0617

with Spo2 values of > 97% were also excluded from analysis because the oxyhemoglobin dissociation curve is flat above these levels.

Analysis of the Derivation Data Set

A scatterplot of S/F vs P/F ratios was utilized to determine the linear relationship between the two measurements. Generalized estimating equations13 were then utilized to quantify the best regression line. The equation for this regression line was em- ployed to determine threshold values for S/F ratios that correlate with P/F ratios of 300 and 200, respectively, for ALI and ARDS. The S/F ratio divided by the P/F ratio was plotted against Fio2, positive end-expiratory pressure (PEEP), and Spo2 to assess the effect that each had on the relationship. Linear mixed-effect modeling was undertaken to determine the effect that PEEP exerted on the S/F threshold values for defining either ALI or ARDS, as defined by the P/F ratio oxygenation criterion. Arterial pH and Paco2 might also affect the relationship between S/F and P/F ratios but were not included in the model because both require arterial blood sampling. In these situations, the availabil- ity of the P/F ratio would obviate the need for a noninvasive measurement defining ALI. An interaction term was included in the model to assess the effect modification by PEEP on the relationship between the P/F and S/F ratios. Receiver operator characteristic (ROC) curves were plotted to assess the degree of discrimination between S/F and P/F ratios and to slightly adjust the S/F ratio threshold values for both ALI and ARDS to optimize the sensitivity and specificity.

Validation Data Set

The relationship of S/F vs P/F ratio was externally validated using similarly matched data for S/F and P/F ratios from patients enrolled in another ARDS Network trial comparing lower vs higher PEEP.14 This study utilized inclusion and exclusion criteria that were similar to those of the derivation data set study.14 Arterial blood gas measurements and Spo2 data were collected at similar time points using methods that were similar to those of the derivation data set.

Analysis of the Validation Data Set

Generalized estimating equations13 was utilized to quantify the relationship between S/F and P/F ratios in the validation data set. ROC curves were plotted to determine the sensitivity and specificity of the threshold values derived from the derivation data set for both ALI and ARDS, with the area under the curve (AUC) calculated to assess the degree of discrimination between S/F and P/F ratios.

Statistical Analysis

Normally distributed continuous variables are expressed as the means and SD. Nonnormally distributed continuous variables are reported as the median and interquartile range. The correlation between P/F and S/F ratios was analyzed using Spearman correlation analysis. Linear regression modeling was utilized to compare the relationship between P/F and S/F ratios with adjustment for levels of PEEP as a potential confounder and effect modifier. A fixed effect with parametric compound sym- metry structure was utilized to account for multiple measure- ments obtained from the same patient. PEEP, P/F ratio, and the interaction term PEEP*P/F were included in the model as continuous variables. Statistical software packages (SPSS, version 14.0; SPSS; Chicago, IL; and SAS, version 9.1; SAS Institute Inc;

Cary, NC) were utilized to perform analyses, graph scatterplots and ROC curves, and calculate the AUC of the ROC. Likelihood

Table 1—Baseline Demographics of the Patients Enrolled in the Studies for Both Data Sets

ratios were calculated using appropriate software (Confidence

Interval Analysis, version 2.1.0) [available at: www.medschool. soton.ac.uk/cia]. Two thousand bootstrap samples were com- puted by resampling patients with replacement to determine 95% CI for the likelihood ratios.

Age, yr Female sex 50.8 ± 17.5

41 50.9 ± 17.4

44

Etiology of ALI

Pneumonia 34 41

Sepsis 26 22

Trauma 10 9

Aspiration 16 16

Variables

Derivation Data Set (n = 672)

Validation Data Set (n = 402)

Results

Of the 861 patients enrolled in the study compar- ing tidal volumes of 6 and 12 mL/kg PBW, 189 were enrolled at sites located at > 1,000 m in altitude (Fig 1). The remaining 672 patients provided 3,384 Pao2

Baseline P/F ratio 132 ± 61 152 ± 66

Minute ventilation, L/min 12.9 ± 4.1 11.7 ± 3.5

APACHE III score 85 ± 29 93 ± 31

and Spo2 measurements at known Fio2 values. Spo2 exceeded 97% in 711 patients, leaving 2,673 data

Nonpulmonary organ failures, No.

1.6 ± 0.8 1.0 ± 0.9

points for analysis in the derivation data set. Of the 549 patients who were enrolled in the trial compar- ing high and low PEEP, 146 were enrolled at centers that were at altitudes of > 1,000 m, and 1 patient had no matched measurements for Spo2 and Pao2. The remaining 402 patients provided 2,031 measure- ments with Spo2 values of ≤ 97% for the validation data set (Fig 1).

Patients enrolled in both studies had similar base- line demographics, which have been previously de- tailed elsewhere11,14 and are briefly summarized in

*Values are given as the mean ± SD or %. APACHE = acute physiology and chronic health evaluation.43

Table 1. The respiratory parameters from the time of the measurements for both data sets are depicted in Table 2. In the derivation data set, the minimum Spo2 measurement was 56%, with 94% of the mea- surements being between 88% and 97%. Likewise, the minimum Spo2 measurement in the validation data set was 62%, with 95% of the measurements

Figure 1. Flow diagram for the data points utilized in the derivation and validation data sets. The derivation set was derived from the ARDS Network trial11 of 6 vs 12 mL/kg tidal volume ventilation. The validation set was derived from the ARDS Network trial14 of higher vs lower PEEP.

being between 88% and 97%. The majority of P/F ratio measurements met the AECC oxygenation criterion for ARDS (P/F ratio ≤ 200) in both the derivation data set (2,130 of 2,673 measurements; 79.7%) and the validation data set (1,475 of 2,031

measurements; 72.6%), while 96.9% of measure- ments for the derivation data set (2,590 of 2,673 measurements) and 96.1% of measurements for the validation data set (n = 1952/2031) met the criterion for ALI (P/F ratio ≤ 300).

Derivation Data Set

S/F and P/F ratios demonstrated a linear correla- tion. This relationship, which did not differ between the two tidal volume strategies (ie, 6 vs 12 mL/kg PBW) is described by the following regression equa- tion: S/F = 64 + 0.84 × (P/F) [95% CI, S/F =

(58 — 70) + (0.79 — 0.88) × P/F] (p < 0.0001;

r = 0.89) [Fig 2]. The relationship between S/F and P/F ratios did not change across varying levels of Fio2 (Fig 3, top, A) or PEEP (Fig 3, middle, B). Since the

Fio2 delivered to patients was protocol-driven with a goal Spo2 between 88% and 92%, the inverse of the Fio2 correlates similarly with P/F ratio (r = 0.83) [Fig 3, bottom, C]. ROC curves (Fig 4) demonstrated that S/F ratios had excellent ability to discriminate between patients with and without ARDS (ie, P/F ratio ≤ 200; AUC = 0.929) and ALI (P/F ≤ 300; AUC = 0.920).

Linear mixed-effect analysis of the derivation data

set demonstrated that PEEP had a significant effect on S/F ratios (p < 0.001) and slightly modified the effect of the P/F ratio on S/F ratios (p = 0.001) as described by the following equation: S/F = 129 + 0.72 × (P/F) — 4.0 × (PEEP) — 0.008 × (PEEP) ×

(P/F) [95% CI: S/F = (121 — 137) + (0.68 — 0.76)

× P/F — (3.3 — 4.7) × PEEP — (0.004 — 0.013) ×

(PEEP) × (P/F)] (p < 0.001; r = 0.87). The linear re- gression equation utilized in conjunction with the mixed-effect model and ROC curves predicted that S/F

ratios of 235 and 315 would correspond with P/F ratios of 200 (ARDS) and 300 (ALI), respectively.

Validation Data Set

S/F and P/F ratios demonstrated a similar linear relationship in the validation data set, described by the following equation: S/F = 68 + 0.84 × (P/F) [95% CI, S/F = (60 — 77) + (0.78 — 0.89) × P/F]

(p < 0.0001; r = 0.82) [Fig 5]. S/F ratios also dem-

onstrated discriminatory ability for P/F ratio values of both 200 and 300 in the validation data set as shown by AUC values of 0.928 and 0.878, respec- tively, for ROC curves. The S/F ratio threshold of 235 from the derivation data set accurately identified 1,257 of the 1,475 cases of ARDS in the validation data set (P/F ratio ≤ 200), yielding a sensitivity of 85%. The same threshold value also correctly dis- criminated 472 of the 556 cases in which the P/F ratio was > 200, for a specificity of 85%. The positive and negative likelihood ratios for the S/F ratio value of 235 discriminating P/F ratio values of ≤ 200 (ie, the oxygenation criterion for ARDS in the AECC definition) were 5.64 (95% CI, 4.69 to 7.08) and 0.17 (95% CI, 0.15 to 0.20), respectively. Similarly, the

S/F threshold of 315 demonstrated 91% sensitivity (accurately identifying 1,778 of the 1,952 cases) for discriminating ALI (P/F ratio ≤ 300) with 56% specificity (correctly discriminating 44 of the 79 cases in which the P/F ratio was > 300). The positive and negative likelihood ratios for the S/F ratio of 315 for ALI (P/F ratio ≤ 300) were 2.06 (95% CI, 1.64 to 2.76) and 0.16 (95% CI, 0.12 to 0.21), respectively.

Discussion

We hypothesized that the continuously available S/F ratio can serve as a surrogate for P/F ratio in the diagnostic criteria for ALI/ARDS. Using data from patients with ALI and ARDS who were enrolled in two large clinical trials,11,14 we found that S/F ratio correlates well with a simultaneously obtained P/F ratio. The correlation improves slightly if PEEP is included in the regression model. S/F ratios of 235

Table 2—Respiratory Parameters at the Time of the Corresponding S/F and P/F Ratio Measurements in Both Data Sets*

Derivation Data Set Measurements Validation Data Set Measurements

Variables (n = 2,673) (n = 2,031)

S/F ratio 194 ± 65 (188; 137–235) 208 ± 69 (194; 155–242.5)

P/F ratio 155 ± 66 (146; 106–190) 166 ± 68 (157.5; 115–207.5)

Tidal volume, mL/kg PBW 9.0 ± 2.9 6.4 ± 1.7

Total respiratory rate, breaths/min 23.3 ± 8.7 28.1 ± 7.7

Minute ventilation, L/min 13.2 ± 4.2 (12.7; 10.4–15.5) 12.1 ± 3.5 (11.8; 9.6–14.1)

PEEP, cm H2O 8.7 ± 3.9 10.7 ± 4.6

Paco2, mm Hg 40.4 ± 10.7 (39; 33–45.8) 42.9 ± 13.0 (41; 35–47)

Arterial pH 7.39 ± 0.07 7.38 ± 0.08

*Values are given as the mean ± SD (median; interquartile range [for variables that are not normally distributed]).

Figure 2. S/F ratio vs P/F ratio scatterplot for the derivation data set. The line represents the best fit linear relationship (S/F ratio = 64 + 0.84 × [P/F]) [p < 0.0001; r = 0.89].

and 315, were found to correspond to P/F ratios of 200 and 300, respectively, which are the oxygenation criteria defining ARDS and ALI, respectively.3 These threshold S/F ratios demonstrated excellent sensitivity and good specificity in predicting the corresponding P/F ratios in a validation data set. To our knowledge, these findings represent the first large study of the relationship between Spo2 and Pao2 in critically ill patients.

The noninvasive and continuously available Spo2 is standard monitoring in most ICUs.15 Although Spo2 reliably predicts Pao2 measured by blood gas analysis in healthy subjects,8,9,15–17 patient race, oximeter location, and disease states, like low cardiac output or methemoglobinemia, may reduce the ac- curacy.8,10,16,18 Despite the ubiquity of Spo2, Pao2 is the accepted “gold standard” for determining arterial oxygenation. The measurement of Pao2 may also significantly vary in patients over short periods of time despite constant Fio2 due to factors such as positioning, agitation, and endotracheal suction- ing.19,20 Institutions, in an effort to contain costs, conserve blood, and reduce inappropriate use,5–7 have vastly reduced the number of arterial blood gas samples obtained in mechanically ventilated pa- tients.21

The sensitivity and specificity of the threshold S/F ratios of 235 and 315 derived in this study suggest that they are appropriate surrogates for P/F ratios of

200 and 300. The use of the S/F ratio in the diagnostic definitions for ALI/ARDS has several potential clinical applications. First, the use of these

values will allow the recognition of patients who likely have ALI/ARDS but have not undergone arterial blood gas sampling, facilitating early enroll- ment into clinical trials and early diagnosis and treatment in clinical practice. Second, the S/F ratio threshold of 315 can be utilized as a continuously available screening tool to identify which patients should undergo arterial blood gas analysis to deter- mine whether they meet the oxygenation criterion for ALI. For example, a ventilated patient receiving 30% Fio2 with 94% Spo2 (S/F ratio 313) who meets the other criteria for ALI has a high likelihood of also meeting the P/F ratio oxygenation criterion.

Utilizing S/F ratios to facilitate the clinical diag- nosis of ALI/ARDS should help to address the underdiagnosis of these syndromes. A volume-lim- ited and pressure-limited ventilation strategy is the only therapeutic intervention that has been shown to significantly reduce mortality in patients with ALI.11 Despite being inexpensive and easy to use, this intervention has not been widely adopted.22–26 One explanation may be that ALI and ARDS are often not recognized,4,26 likely contributing to the failure to implement treatment strategies such as lung-protec- tive ventilation and conservative fluid manage- ment.11,27

S/F ratios may be useful in other important clinical

applications. Many organ failure scores, such as lung injury score,28 sequential organ failure assessment,29 simplified acute physiology score II,30 or multiorgan dysfunction score,31 utilize P/F ratios to quantify hypoxemia. In instances in which these scores are calculated frequently, the respiratory component is often omitted due to the lack of repeated arterial blood gas analyses. Using S/F ratio as a surrogate measure of hypoxemia would allow these scores to be calculated in the absence of arterial blood gas sam- pling. It should be noted, however, that, except for the lung injury score, these scoring systems are often used for widely heterogeneous groups of critically ill patients and not just those with ALI/ARDS requiring mechanical ventilation. Since a diagnosis of ALI or ARDS was required for enrollment in both of the trials utilized in our analysis, > 95% of the measure-

ments in both the derivation and validation data sets

met the oxygenation criteria for ALI. The relatively few patients with a P/F ratio of > 300 in our data sets and the exclusion of patients enrolled at sites located at altitudes of > 1,000 m renders extrapolation of our results to these populations uncertain. Further- more, the majority of the patients were nonsurgical, with medical conditions causing ALI/ARDS. Our results should be prospectively validated in other patient populations, including patients not requiring mechanical ventilation and patients without lung

Figure 3. Relationship of S/F vs P/F ratio across varying levels of (top, A) Fio2 (r = 0.14), (middle, B) PEEP (r = 0.18), and (bottom, C) P/F vs 1/F ratio (P/F = —13 + 81.2/Fio2; r = 0.83) for the derivation data set. The line in all panels represents the best-fit linear relationship.

Figure 4. ROC curves for (top, A) S/F vs P/F ratios of ≤ 200 and (bottom, B) S/F vs P/F ratios of ≤ 300 for the derivation data set.

injury, to ensure that they remain accurate in these heterogeneous populations.

There are some additional limitations of this study. First, although the vast majority of the Spo2 and Pao2 measurements were made simultaneously, the protocols allowed separation by a few hours, which could contribute to discrepancies between measure-

Figure 5. S/F vs P/F ratio scatterplot for the validation data set. The line in both panels represents the best-fit linear relationship (S/F = 68 + 0.84 × [P/F]) [p < 0.0001; r = 0.82].

ments. Despite this, S/F and P/F ratios remained highly correlated. Second, measurements made with an Spo2 of > 97% were excluded from analysis. At these saturations, the slope of the relationship be- tween Spo2 and Pao2 becomes almost zero, and large changes in Pao2 may result in little or no change in Spo2. We believe that this limitation is acceptable because routine ICU care titrates Fio2 to maintain saturations of 92 to 95%.16 Finally, numer- ous studies have reported low specificity and sensi- tivity of the AECC definitions for ALI and ARDS,4,32–38 with many studies criticizing the defi- nition of hypoxemia.38–42 Our proposal to utilize S/F ratio as a surrogate is not meant to alleviate these concerns. The S/F ratio thresholds determined in this study were based on P/F ratios proposed by the

AECC.3 Although using an S/F ratio will allow the

degree of hypoxemia to be assessed noninvasively, the optimal definition of hypoxemia for the diagnosis of ALI/ARDS and whether measurements should be obtained on standardized PEEP and ventilator set- tings remains controversial.4,39 – 42 Furthermore, S/F ratio does not allow the evaluation of acid-base status or Paco2 levels, two other potentially important clinical reasons for performing blood gas analysis.

In summary, we have derived and validated threshold values for S/F ratio that can be used as surrogates to diagnose ALI/ARDS when a P/F ratio is unavailable. Utilizing the noninvasive and contin- uously available S/F ratio may facilitate an earlier diagnosis of ALI/ARDS, allowing the application of

appropriate therapies such as lung-protective venti- lation and conservative fluid management strategies. Future studies are needed to validate the relation- ship between S/F and P/F ratio in more heteroge- neous populations of critically ill patients.

References

⦁ Ware LB, Matthay MA. The acute respiratory distress syn- drome. N Engl J Med 2000; 342:1334 –1349

⦁ Rubenfeld GD, Caldwell E, Peabody E, et al. Incidence and outcomes of acute lung injury. N Engl J Med 2005; 353:1685– 1693

⦁ Bernard GR, Artigas A, Brigham KL, et al. The American- European Consensus Conference on ARDS: definitions, mechanisms, relevant outcomes, and clinical trial coordina- tion. Am J Respir Crit Care Med 1994; 149:818 – 824

⦁ Ferguson ND, Frutos-Vivar F, Esteban A, et al. Acute respiratory distress syndrome: underrecognition by clinicians and diagnostic accuracy of three clinical definitions. Crit Care Med 2005; 33:2228 –2234

⦁ Merlani P, Garnerin P, Diby M, et al. Quality improvement report: linking guideline to regular feedback to increase appropriate requests for clinical test; blood gas analysis in intensive care. BMJ 2001; 323:620 – 624

⦁ Pilon CS, Leathley M, London R, et al. Practice guideline for arterial blood gas measurement in the intensive care unit decreases numbers and increases appropriateness of tests. Crit Care Med 1997; 25:1308 –1313

⦁ Roberts D, Ostryzniuk P, Loewen E, et al. Control of blood gas measurements in intensive-care units. Lancet 1991; 337: 1580 –1582

⦁ Jensen LA, Onyskiw JE, Prasad NG. Meta-analysis of arterial oxygen saturation monitoring by pulse oximetry in adults. Heart Lung 1998; 27:387– 408

⦁ Perkins GD, McAuley DF, Giles S, et al. Do changes in pulse oximeter oxygen saturation predict equivalent changes in arterial oxygen saturation? Crit Care 2003; 7:R67

⦁ Webb RK, Ralston AC, Runciman WB. Potential errors in pulse oximetry: II. Effects of changes in saturation and signal quality. Anaesthesia 1991; 46:207–212

⦁ Ventilation with lower tidal volumes as compared with tradi- tional tidal volumes for acute lung injury and the acute respiratory distress syndrome: the Acute Respiratory Distress Syndrome Network. N Engl J Med 2000; 342:1301–1308

⦁ Sasse SA, Jaffe MB, Chen PA, et al. Arterial oxygenation time after an Fio2 increase in mechanically ventilated patients. Am J Respir Crit Care Med 1995; 152:148 –152

⦁ Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130

⦁ Brower RG, Lanken PN, MacIntyre N, et al. Higher versus lower positive end-expiratory pressures in patients with the acute respiratory distress syndrome. N Engl J Med 2004; 351:327–336

⦁ Jubran A. Pulse oximetry. Intensive Care Med 2004; 30:2017–

2020

⦁ Jubran A, Tobin MJ. Reliability of pulse oximetry in titrating supplemental oxygen therapy in ventilator-dependent pa- tients. Chest 1990; 97:1420 –1425

⦁ Yamaya Y, Bogaard HJ, Wagner PD, et al. Validity of pulse oximetry during maximal exercise in normoxia, hypoxia, and hyperoxia. J Appl Physiol 2002; 92:162–168

⦁ Smatlak P, Knebel AR. Clinical evaluation of noninvasive monitoring of oxygen saturation in critically ill patients. Am J Crit Care 1998; 7:370 –373

⦁ Thorson SH, Marini JJ, Pierson DJ, et al. Variability of arterial blood gas values in stable patients in the ICU. Chest 1983; 84:14 –18

⦁ Sasse SA, Chen PA, Mahutte CK. Variability of arterial blood gas values over time in stable medical ICU patients. Chest 1994; 106:187–193

⦁ Jubran A. Advances in respiratory monitoring during mechan-

ical ventilation. Chest 1999; 116:1416 –1425

⦁ Young MP, Manning HL, Wilson DL, et al. Ventilation of patients with acute lung injury and acute respiratory distress syndrome: has new evidence changed clinical practice? Crit Care Med 2004; 32:1260 –1265

⦁ Rubenfeld GD, Cooper C, Carter G, et al. Barriers to providing lung-protective ventilation to patients with acute lung injury. Crit Care Med 2004; 32:1289 –1293

⦁ Weinert CR, Gross CR, Marinelli WA. Impact of randomized trial results on acute lung injury ventilator therapy in teaching hospitals. Am J Respir Crit Care Med 2003; 167:1304 –1309

⦁ Meade MO, Jacka MJ, Cook DJ, et al. Survey of interventions for the prevention and treatment of acute respiratory distress syndrome. Crit Care Med 2004; 32:946 –954

⦁ Kalhan R, Mikkelsen M, Dedhiya P, et al. Underuse of lung protective ventilation: analysis of potential factors to explain physician behavior. Crit Care Med 2006; 34:300 –306

⦁ Wiedemann HP, Wheeler AP, Bernard GR, et al. Compari- son of two fluid-management strategies in acute lung injury. N Engl J Med 2006; 354:2564 –2575

⦁ Murray JF, Matthay MA, Luce JM, et al. An expanded definition of the adult respiratory distress syndrome. Am Rev Respir Dis 1988; 138:720 –723

⦁ 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 Inten- sive Care Medicine. Intensive Care Med 1996; 22:707–710

⦁ Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study. JAMA 1993; 270:2957–2963

⦁ Marshall JC, Cook DJ, Christou NV, et al. Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit Care Med 1995; 23:1638 –1652

⦁ Abraham E. Toward new definitions of acute respiratory distress syndrome. Crit Care Med 1999; 27:237–238

⦁ Abraham E, Matthay MA, Dinarello CA, et al. Consensus conference definitions for sepsis, septic shock, acute lung injury, and acute respiratory distress syndrome: time for a reevaluation. Crit Care Med 2000; 28:232–235

⦁ Esteban A, Fernandez-Segoviano P, Frutos-Vivar F, et al. Comparison of clinical criteria for the acute respiratory distress syndrome with autopsy findings. Ann Intern Med 2004; 141:440 – 445

⦁ Meade MO, Cook RJ, Guyatt GH, et al. Interobserver variation in interpreting chest radiographs for the diagnosis of acute respiratory distress syndrome. Am J Respir Crit Care Med 2000; 161:85–90

⦁ Meade MO, Guyatt GH, Cook RJ, et al. Agreement between alternative classifications of acute respiratory distress syn- drome. Am J Respir Crit Care Med 2001; 163:490 – 493

⦁ Moss M, Goodman PL, Heinig M, et al. Establishing the relative accuracy of three new definitions of the adult respi- ratory distress syndrome. Crit Care Med 1995; 23:1629 –1637

⦁ Ferguson ND, Davis AM, Slutsky AS, et al. Development of a clinical definition for acute respiratory distress syndrome using the Delphi technique. J Crit Care 2005; 20:147–154

⦁ Estenssoro E, Dubin A, Laffaire E, et al. Impact of positive end-expiratory pressure on the definition of acute respiratory distress syndrome. Intensive Care Med 2003; 29:1936 –1942

⦁ Ferguson ND, Kacmarek RM, Chiche JD, et al. Screening of ARDS patients using standardized ventilator settings: influ- ence on enrollment in a clinical trial. Intensive Care Med 2004; 30:1111–1116

⦁ Villar J, Perez-Mendez L, Kacmarek RM. Current definitions of acute lung injury and the acute respiratory distress syn- drome do not reflect their true severity and outcome. Inten- sive Care Med 1999; 25:930 –935

⦁ Aboab J, Louis B, Jonson B, et al. Relation between Pao2/ Fio2 ratio and Fio2: a mathematical description. Intensive Care Med 2006; 32:1494 –1497

⦁ 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 PF-07104091