Venous gas emboli detected by two-dimensional echocardiography are an imperfect surrogate endpoint for decompression sickness.

ABSTRACT

INTRODUCTION: In studies of decompression procedures, ultrasonically detected venous gas emboli (VGE) are commonly used as a surrogate endpoint for decompression sickness (DCS). However, VGE have not been rigorously validated as a surrogate endpoint for DCS. METHODS: A data set for validation of VGE as a surrogate endpoint for DCS was retrospectively assembled comprising maximum VGE grades measured using two-dimensional echocardiography and DCS outcome following 868 laboratory man-dives. Dives were conducted according to only ten different experimental interventions such that the ten cumulative incidences of DCS (0-22%) provide relatively precise point estimates of the probability of DCS, P(DCS). Logistic models relating the P(DCS) to VGE grade and intervention were fitted to these validation data. Assessment of the models was used to evaluate the Prentice criteria for validating a surrogate endpoint. RESULTS: The P(DCS)) increased with increasing VGE grade. However, the difference in the P(DCS) between interventions was larger than explained by differences in VGE grades. Therefore, VGE grades did not largely capture the intervention effect on the true endpoint (DCS) in accord with the Prentice definition of a surrogate endpoint. CONCLUSIONS: VGE can be used for comparisons of decompression procedures in samples of subjects but must be interpreted cautiously. A significant difference in VGE grade probably indicates a difference in the P(DCS). However, failure to find a significant difference in VGE grades does not necessarily indicate no difference in P(DCS).