Dr. Martin is concerned about the validity of extending our conclusions to "most of the world's sport divers," that is, recreational divers. We acknowledge that most of the divers we studied are technical divers in his eyes, and therefore the extrapolation of our results to other groups of divers can be questioned. Contrary to his explanation about technical diving, all of our divers used exclusively compressed air for breathing. With use of compressed air for diving, endogenous bubble formation can be observed after decompression from a steady-state exposure of only 135 kPa (corresponding to a depth of 3.5 m [12 ft]) (1). Therefore, paradoxical embolism of nitrogenous bubbles, with its clinical consequences, can also take place after ascent from dives undertaken with recreational intentions. As we mentioned in our article, the reported incidence of serious decompression events is very low (about 3 events per 10 000 dives), even in divers with patent foramen ovale (2). For this reason, we recruited only divers with a large number of dives (>200), who, of course, sometimes dived deeper than purely recreational divers. Nevertheless, our data do not at all support Dr. Martin's assumption that primarily deep dives (beyond 40 m [133 ft]) may be responsible for ischemic brain lesions in divers. The increased frequency of ischemic brain lesions observed in divers with patent foramen ovale (1.23 ± 2.0 lesions per person) compared with divers without patent foramen ovale (0.64 ± 1.22) (P = 0.07) supports the theory that arterial gas bubbles can cause neurologic injuries in divers by a direct embolic effect or by acting as nuclei for bubble growth in supersaturated tissue (3). The direct embolic effect, at least, must not be restricted to deep dives.

Dr. Bollen's statistical comments are misleading. In our article, we correctly reported the odds of finding ischemic brain lesions in divers compared with nondivers (41 lesions in 52 divers vs. 7 lesions in 52 controls). As mentioned in the statistical analysis section, the P values for the number of ischemic brain lesions in the different study groups were calculated with a Poisson regression model, using patent foramen ovale and diver or control as either grouping variable or predictor. This model fitted well, with no evidence of overdispersion. In a multivariate regression model, as a general rule of thumb, there should be 10 outcome events per covariable (4). Although 48 ischemic brain lesions were observed in our cohort, considering the small number of participants with ischemic brain lesions (19 divers and 6 controls), we did not adjust for potential confounders.