Figure 5.

Serum neuron‐specific enolase (NSE) and serum tau for survival. Survival curves for tau, modeled by Cox proportional hazards regression, adjusted for age and NSE are shown. Tau and NSE were used after log10 transformation, and entered as continuous variables. Models were done using 24‐hour data (A, D), 48‐hour data (B, E), and 72‐hour data (C, F). Predicted survival curves (with 95% confidence intervals) are shown for 3 different levels of tau, corresponding to the median level of tau within the lowest, middle, and highest tertile of tau, respectively. Predicted survival curves are shown for 2 different scenarios, one in the context of low NSE (the median NSE level in the lower half of NSE; A–C) and one in the context of high NSE (the median NSE level in the higher half of NSE; D–F). Greater serum tau was significantly associated with shorter survival in all models.