Mortality in hemodialysis: Synchrony of biomarker variability indicates a critical transition

Critical transition theory suggests that complex systems should experience increased temporal variability just before abrupt change, such as increases in clinical biomarker variability before mortality. We tested this in the context of hemodialysis using 11 clinical biomarkers measured every two weeks in 763 patients over 2496 patient-years. We show that variability - measured by coefficients of variation - is more strongly predictive of mortality than biomarker levels. Further, variability is highly synchronized across all biomarkers, even those from unrelated systems: the first axis of a principal component analysis explains 49% of the variance. This axis then generates powerful predictions of all-cause mortality (HR95=9.7, p<0.0001, where HR95 is a scale-invariant metric of hazard ratio across the predictor range; AUC up to 0.82) and starts to increase markedly [~]3 months prior to death. Such an indicator could provide an early warning sign of physiological collapse and serve to either trigger intervention or initiate discussions around palliative care.