Of principles and protocols and weaning.

Half the patients in an intensive care unit receive mechanical ventilation and almost half of intubated time is devoted to weaning. Accordingly, weaning constitutes an enormous workload for intensive care staff (1). Two deliberate steps are involved in weaning. First, patient readiness is tested with physiological measurements, usually called weaning predictors (2, 3). Second, the patient is evaluated while ventilator support is decreased, gradually or abruptly (4, 5). This two-step approach has been found more dependable than allowing physicians wean in a desultory fashion (6). But findings reported in this issue of the Journal (pp. 673–678) appear to turn accepted wisdom on its head. In a prospective controlled trial, Krishnan and coworkers (7) compared protocolized weaning with usual care. Patients assigned to protocolized weaning were screened in the morning by a respiratory therapist and stable patients had frequencyto-tidal volume ratio measured. If the ratio was 105 or less, respiratory and nursing staff undertook a spontaneous breathing trial (without physician intervention). If the patient passed the one-hour trial, the physician was informed. Patients failing the assessment were rested until the following day. In the usual care arm, patients were managed at the discretion of their physicians. Patients did not undergo any scheduled screening, although physicians were free to make measurements at the bedside. Clinical outcomes did not differ between the two groups. Duration of mechanical ventilation, the primary outcome, was equivalent for protocolized weaning and usual care (60.4 versus 68.0 hours), as was the rate of successful weaning (74.7 versus 75.2%). At first glance, the findings of Krishnan and coworkers (7) suggest that previous research on weaning has not improved clinical practice. But interpretation of information depends on context. In research, special trouble is taken to ensure relevant context by studying a comparison group. Yet data from control groups have served mainly as backdrop, and the spotlight has been shone mostly on the treated group. Today, however, selection of control groups is moving more and more to center stage (8). It is no longer acceptable to compare a new bronchodilator agent against placebo (instead, the new agent must be compared against the best preexisting treatment) (8). Likewise, interpretation of research findings with a new ventilator strategy depends on how the control group is ventilated (9). Close examination of the control group of Krishnan and coworkers (7) tells us what questions we can and cannot answer with the new data. Do patients weaned according to the usual practice of Hopkins physicians do as well as patients subjected to protocolized weaning? Yes. Do the data indicate that the two-step approach to weaning (testing patient readiness followed by evaluation of the patient as ventilator assistance is deliberately decreased) is unhelpful? No—because that question was not tested. And it is doubtful whether the question could have been satisfactorily tested. To do so, the investigators would have had to compare their 2000–2001 usual care group with matched