LAUSR

Only a few years ago, hypertonic saline had a reputation as a potentially evidence-based therapy for acute viral bronchiolitis in young children.1 However, as more trials were published, the evidence seemed to devolve. One assessment of the current state of the evidence on hypertonic saline treatment in bronchiolitis is that it conforms to the pattern described by Ioannidis2 in the famous article entitled, “Why Most Published Research Findings Are False.” Ioannidis2 and others3 have noted the tendency in medicine for the rapid uptake of new therapies based on positive early studies, which then requires reversal as additional trials inevitably result in regression to the mean. Publication bias—the fact that small positive studies are more likely to be published than small negative studies, which may simply be labeled underpowered and discarded—plays a major role in perpetuating this pattern. With respect to hypertonic saline, it is fair to say that early enthusiasm has evolved into increasing skepticism because of the addition of more recent negative studies and further consideration of significant heterogeneity in the populations under study, particularly in the inpatient literature.4-6 With this issue of JAMA Pediatrics, the literature on hypertonic saline must now incorporate a major new contribution to the debate in the emergency department (ED) setting. Angoulvant and colleagues7 report on a trial of 2 doses of nebulized saline, 3%, for the treatment of bronchiolitis in 777 infants between 6 weeks and 12 months of age with a primary outcome of hospitalization who were treated in the EDs of 24 French hospitals. This study (Efficacy of 3% Hypertonic Saline in Acute Viral Bronchiolitis [GUERANDE]) is the largest single trial published on the topic and was designed to enroll infants with moderate to severe bronchiolitis. It was powered to detect a 10% difference in hospitalization rates between the 2 treatment groups, and the investigators found no significant difference in these rates. Specifically, 48.1% of infants in the hypertonic saline group were admitted compared with 52.2% of infants in the normal saline group, with an adjusted risk difference of −3.2% (P = .25). Respiratory scores improved in both study arms, although more so in the hypertonic saline group (−3.1 vs −2.4 points on the Respiratory Distress Assessment Instrument). However, mild adverse events were also more common in the hypertonic saline group (8.9% vs 3.9%). As with all clinical research, there are potential weaknesses in this study. The range of admission rates among participating centers was remarkably broad (31.6%83.3%), thus calling into question the generalizability of admission criteria across study sites, although appropriate statistical adjustments were made. How does the current study compare with the existing literature on ED treatment? A prior meta-analysis1 of 7 studies of 951 patients reported a 20% relative reduction in risk of hospitalization with hypertonic saline. Of further interest, most of the studies in that meta-analysis were originally published as negative studies, meaning they did not meet their a priori threshold for clinical significance, similar to GUERANDE. However, when combined they provide a summary statistic that achieves a standard threshold for statistical significance, thus creating a positive meta-analysis of a group of studies that individually mostly failed to disprove the null hypothesis. As counterintuitive as it may sound, to accomplish such a feat is essentially one of the purposes of meta-analysis (eg, to overcome the problem of lack of power in any single study). However, as powerful a tool as meta-analysis may be, it still cannot absolve us of the necessity to reconcile statistical significance with clinical significance. One step in such a reconciliation for this particular question might be to examine a simple and direct quantitative comparison of the 2 data sets. The findings in the meta-analysis can be expressed as an absolute risk reduction for hospitalization of approximately 5%, with a number needed to treat (NNT) to prevent 1 hospitalization of 20. The GUERANDE study was powered on the assumption that a 10% absolute risk reduction in hospitalizations (NNT of 10) would indicate that the therapy was clinically useful. The current study, as performed, ultimately contradicted that assumption, but the question remains, is an NNT of 20 too many? Given that acute viral bronchiolitis is always one of the leading causes of children’s hospitalization, a therapy with a measurable effect, even a small one, might still be considered to be useful. However, there are obvious problems with the assertion that any quantifiable effect makes a therapy clinically useful. First, many studies, including GUERANDE, have a minimum threshold for illness to enroll; thus, the patients studied have a fairly high risk of hospitalization to begin with, whereas it is likely that any therapy, once accepted as useful, would be applied across a much broader range of patients and thus decrease the absolute risk reduction and increase the NNT even further. In addition, this medication requires time and effort to nebulize, and the cost of treating the many patients who will not benefit must be considered. Finally, the therapy has a nonnegligible rate of adverse effects, consistently higher than placebo across multiple studies. Thus, there must be some point at which the cost of the therapy and the attendant adverse effects outweigh an increasingly receding benefit, regardless of the outcome of testing for statistical significance. Ultimately, clinical significance is a value judgement. Professional thresholds for when the benefits of a therapy outweigh its costs and risks emerge organically out of a multiRelated article Opinion