Hazardous to Your Health: Kinetic Foundations of Risk Stratification and Therapeutic Triage

The conceptual independence of risk and hazard. The slope of the line representing the rate of change in risk (hazard) is greater for Jane (solid circles) than for Dick (open circles) (upper panel). As a result, although Jane’s risk is less than Dick’s at the present time (t), it will eventually become greater, at some future time. Risk stratification is therefore an incomplete basis for therapeutic triage as shown by the differential effects of treatment on risk and hazard (lower panel). See text for further discussion.

A hypothetic kinetic model based on a reversible transition between the stable (S) and unstable (U) states (k₁ = 0.012, k₂ = 0.058) and an irreversible transition between the unstable and event (E) state (k₃ = 0.029). Risk is quantified by the prevalence of events. The entire population begins from the stable state, [S]₀ = 1. Thereafter, the proportional prevalence for each state is plotted over 120 months of follow-up. The prevalence of the stable state (light line) falls as a simple exponential decay. The prevalence of the intermediate unstable state (medium line) rises to a maximum (at 36 months). As a result, the prevalence of events (heavy line) rises as a relatively flat curvilinear function over time.

Risk versus hazard for Figure 2. Note that hazard and risk are positively correlated at low levels of risk and are inversely correlated at higher levels of risk. Among patients with these higher levels of risk—typically selected for more aggressive therapy under a risk stratification algorithm—the rate of change in risk is actually decreasing (the risk is increasing more slowly). Risk stratification alone is therefore an incomplete basis for therapeutic triage.

Evolution of the kinetic model in Figure 2 when the entire population begins from an initial unstable state, [U]₀ = 1, rather than a stable state. Now, the prevalence of the unstable state (medium line) falls, whereas the prevalence of the stable state (light line) rises to a maximum (again at 36 months). As a result, the prevalence of events (heavy line) rises steeply over the initial period of follow-up and more slowly thereafter.

Therapeutic implications of kinetic modeling. The labeling is identical to that in Figure 4. See text for discussion.