Exhaustive Search Outperforms a Genetic Algorithm for Formulary-Constrained Antiretroviral Regimen Selection

Antiretroviral regimen selection requires balancing efficacy, toxicity, cost and side-effect burden under patient-specific resistance and formulary constraints. Genetic algorithms have been applied to this task in prior work, generally without comparison against an exact baseline. We report such a comparison. We formulate WHO-compliant regimen selection as constrained combinatorial optimisation over a catalogue of 20 antiretrovirals, with drug availability modelled through three formulary tiers derived from WHO guidance. Evaluating a genetic algorithm and exhaustive search over a provably identical feasible set across 17 patient profiles, the feasible set contains 2–130 regimens per patient. The genetic algorithm recovers the exact optimum for every patient while requiring approximately 1,400 times more computation; for 14 of 17 patients the optimum is already present in the randomly initialised population. We conclude that metaheuristic search for this problem is warranted only under a sequencing formulation.