© 2003 American Thoracic Society
Questions about ARDSNetwork trial of low tidal volumeTo the Editor:We share the concerns of Eichacker and colleagues (1) in their meta-analysis that the "traditional" tidal volumes received by patients in the control arm of the ARDSNetwork low tidal volume trial (2) produced plateau pressures that were substantially greater than routine care at the time. In addition to this concern, we have other concerns that make interpretation of the results difficult. Review of the original report from the ARDSNetwork trial demonstrates that many important pieces of data are missing (2). These omissions have never been adequately explained. It is still not possible to determine the tidal volume or mode of ventilation that 14% of patients were receiving before enrollment (2–4). After randomization, on Day 1, based on the number of patients originally enrolled, 8% do not have tidal volumes reported and 17% do not have blood gas data. Of greater concern, the percentage of patients with missing data is significantly greater in the group receiving low tidal volumes than in the group receiving traditional tidal volumes (Table 1) . Since more patients receiving the "traditional" tidal volume died, one might have expected the opposite. Our worry is that this imbalance may have arisen as a result of early withdrawal by clinicians of patients who were unable to tolerate the low tidal volume. Consistent with this viewpoint, patients receiving low tidal volumes had higher levels of positive end-expiratory pressure and significantly decreased levels of oxygenation compared with the "traditional" tidal volume group (2). It needs to be explained why only the percentages for survival and nonsurvival were provided. The investigators did not report the actual numbers of survivors and nonsurvivors upon which the analysis of the trial was based. Accounting for the withdrawal of patients and its influence on the results of the trial is critical for interpretation.
We believe that there are not only problems with the design of the ARDSNetwork low tidal volume trial, as pointed out by Eichacker and colleagues, but also with the rigor of the presentation of the results. Large numbers of values were reported as missing without comment or discussion as to the bias that these omissions may have introduced.
a Centre Hospitalier de Versailles Le Chesnay, France REFERENCES
From the Authors:Several aspects of the design of the ARDSNetwork lower tidal volume trial (1) were addressed in our response (2) to the recent analysis of Eichacker and colleagues (3). The traditional tidal volume protocol in our trial used tidal volumes and inspiratory pressures that were consistent with routine care while giving higher priority to the traditional gas exchange and breathing comfort goals. The mean (± SD) traditional group plateau pressure on study Day 1 was 33 ± 9 cm H2O (1). Mean routine care plateau pressure (before patients were enrolled) was 30 ± 8 cm H2O (4). Thus, Day 1 traditional group plateau pressures were higher by only 3 cm H2O, and there was considerable overlap in the ranges of these plateau pressures. In fact, 33% of the routine care plateau pressures exceeded 33 cm H2O, and 11% exceeded 40 cm H2O. Traditional group plateau pressure on study Day 7 was 37 ± 9 cm H2O. At this time, many patients were weaning or had been extubated. Plateau pressures in these patients, which would have been lower than those who continued on volume-cycled ventilation, were not measured. Therefore, the higher Day 7 mean plateau pressure represents a subset of patients with more severe disease and cannot be compared directly to the mean routine care plateau pressure.There are many reasons why some physiologic data points may be missing in a trial such as ours. For example, tidal volumes and plateau pressures were not available when patients were weaning or breathing spontaneously. There may be significant differences in the proportions of missing values if there are systematic differences in the progress of the patient groups. Some clinicians deviated from the protocol rules in both study groups if they were concerned about the safety of their patients. When this occurred we continued to record the same parameters when they were available, and these values are included in the mean values shown in Table 3 of our original article (1). Most importantly, analyses of mortality and other clinical outcomes were conducted on an "intention to treat" basis. These analyses included all randomized patients, regardless of whether trial procedures were completed as planned. Therefore, the missing physiologic data points do not introduce bias in the analysis of clinical outcomes, which demonstrated significantly lower mortality in the lower tidal volume study group.
For the NIH NHLBI ARDS Network:
a Johns Hopkins University Baltimore, Maryland REFERENCES
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
