Phenytoin is primarily eliminated by hepatic metabolism (>95%).
Hepatic metabolism is mainly via the CYP2C9 enzyme system with a
smaller amount metabolized by CYP2C19. About 5% of a phenytoin dose is
recovered in the urine as unchanged drug. Phenytoin follows
Michaelis-Menten or saturable pharmacokinetics.
This is the type of nonlinear pharmacokinetics that occurs when the number of drug molecules overwhelms or saturates the enzyme’s ability to metabolize the drug. When this occurs, steady-state drug serum concentrations increase in a disproportionate manner after a dosage increase:

In this case the rate of drug removal is described by the classic Michaelis-Menten relationship that is used for all enzyme systems: rate of metabolism =$\ce{(Vmax ⋅ C) / (Km + C)}$, where Vmax is the maximum rate of metabolism in mg/d, C is the phenytoin concentration in mg/L, $\ce{Km}$ is the substrate concentration in mg/L, and where the rate of metabolism = $\ce{Vmax /2}$.
The clinical implication of Michaelis-Menten pharmacokinetics is that the clearance of phenytoin is not a constant as it is with linear pharmacokinetics, but is concentration- or dose-dependent. As the dose or concentration of phenytoin increases, the clearance rate (Cl) decreases as the enzyme approaches saturable conditions: $\ce{Cl = Vmax / (Km + C)}$. This is the reason concentrations increase disproportionately after a phenytoin dosage increase.
However, since clearance is dose or concentration-dependent, half-life also changes with phenytoin dosage or concentration changes. As doses or concentrations increase for a drug that follows Michaelis-Menten pharmacokinetics, clearance decreases and half-life becomes longer for the drug.
For drugs such as phenytoin with a low hepatic extraction ratio
(≤30%), the numeric value of liver blood flow is much greater than the
product of unbound fraction of drug in the blood and the intrinsic
clearance of the compound (LBF >> fB)
Plasma protein binding
displacement drug interactions cause major pharmacokinetic alterations
but are not clinically significant because the pharmacologic effect of
the drug does not change.
Because the clearance of the drug is
dependent on the fraction of unbound drug in the blood and intrinsic clearance for a low hepatic extraction ratio agent, a
decrease in plasma protein binding and increase in unbound fraction
will increase clearance.
Currently there is no substantive evidence of phenytoin auto-inducing its own metabolism (the theory of phenytoin displacing itself from plasma proteins really isn’t convincing at all but perhaps in a drug interaction with a drug which is predominantly plasma bound). The reason being that the information regarding phenytoin auto-induction is sparse and conflicting. The results of this study aren’t convincing enough when taken in perspective of its ability to auto-induce itself after an overdose.
References
- Applied Clinical Pharmacokinetics: A. Bauer, Phenytoin: Basic Clinical Pharmacokinetic Parameters