Author reply

Sullivan-Mee et al questioned whether our longitudinal results demonstrating a relationship between lower corneal hysteresis (CH) and faster progression in glaucoma could perhaps be confounded by a relationship between CH and disease severity. They suggest that the longitudinal, mixed-effects model presented in our study should have included an adjustment for disease severity as a baseline covariate. As stated in our study, an unstructured covariance between random effects was assumed, allowing for correlation between random intercepts (baseline) and slopes. This was done exactly for the fact that rates of visual field loss may depend on disease severity. Whether or not baseline disease severity should additionally be included as a fixed effect term in the mixed model is a very controversial issue in the statistics literature. A recent study by Glymour et al suggests that baseline adjustment in analyses of change may induce spurious statistical associations with inflated regression coefficients. Beside the statistical limitations of the adjustment proposed by Sullivan-Mee et al, one would also believe that, if lower CH is associated with worse disease severity, this would actually support our findings that lower CH is actually related to disease progression.