Two independent reasons to combine
Combination therapy widens the window for two separate reasons. Mechanistically, requiring a cell to defeat two unrelated killing mechanisms at once makes the resistant phenotype rare. Toxicologically, choosing agents whose dose-limiting toxicities fall on different normal tissues lets each drug be given near its own maximum tolerated dose instead of forcing a shared reduction.
The design rule
A rational combination pairs agents with non-overlapping mechanisms of action and non-overlapping dose-limiting toxicities. Overlap in either dimension narrows the window instead of widening it.
Why timing is a selectivity tool
Normal proliferating tissues recover faster than tumor tissue between cytotoxic insults, so the normal-tissue dose-response curve returns toward baseline while the tumor curve does not. Dose-dense scheduling shortens the interval to exploit that difference. For phase-specific agents, which kill only cells in a particular cell-cycle phase, prolonged or repeated exposure is required so that a larger fraction of the tumor passes through the vulnerable phase while the drug is present.
Reading a regimen as a window argument
Consider a regimen combining a DNA-damaging alkylator with a microtubule poison. The two agents act at different points in the drug-to-death chain, so a cell that effluxes the alkylator well is not automatically protected against the microtubule poison, and a cell with enhanced DNA repair is not protected against mitotic disruption. Their dose-limiting toxicities also differ: the alkylator is limited mainly by bone-marrow suppression, while the microtubule poison is limited by peripheral neuropathy. Because the two normal-tissue curves peak on different tissues, each drug can be pushed close to its own ceiling. If instead both agents were myelosuppressive, the shared ceiling would force dose reduction of both, and the tumor curve would move right while the normal-tissue curve stayed put.