Determining the incidence and severity of ocular injuries and the protective vs harmful effect of air bag deployment is important. We commend Duma et al1 for raising the issue, but we found several reasons for concern.
Using the National Automotive Sampling System database may represent selection bias because it is not a random sampling. Is it not possible that, given the increased interest in injuries related to air bags, disproportionately more of these cases have been selected for analysis? How can we be certain that crashes involving air bag deployment were of equal severity to those without? Is the air bag not more likely to deploy in more severe crashes?