Evidence-steered medicine in oncology: network-aware micro-combinations that safeguard standard of care and potentially improve benefit–risk
摘要
Standard oncology regimens often follow a dose-optimization paradigm: intensify a single agent until efficacy emerges or toxicity intervenes. Yet many targets sit in compensatory tumor–host networks that reroute around pathway pressure, while incremental efficacy may plateau as dose rises and toxicity escalates. Clinical reality is further complicated by heterogeneous patient states, including suppressed antigen presentation (e.g., β2-microglobulin loss or antigen-processing/presentation machinery defects), immature or poorly trafficking dendritic cells, and symptom/toxicity burdens that drive dose delays or reductions and erode relative dose intensity. These factors contribute to late-stage monotherapy failures and leave clinicians with a binary choice—more of the same therapy or a complete stop—when the actionable question is often whether a small, precisely timed change could restore the preconditions under which standard-of-care (SoC) therapy works best. Evidence-steered medicine (ESM) proposes a conservative adjunctive framework that adds only tiny, single-pulse exposures (“micro-adjuncts”) without delaying or displacing SoC. Each micro-adjunct is pre-specified with a directional mechanistic hypothesis and read out by short-horizon proxies at 48–96 h (e.g., HLA–ABC and TAP1/2 for antigen presentation; CD86/CCR7 for antigen-presenting-cell activation and trafficking readiness; hs-CRP for systemic inflammatory tone; HRV and symptom scores for host readiness). ESM enforces three safety-first gates: a mechanistic sign-consistency check; a transportability/overlap screen to avoid out-of-support exposure; and a banded decision rule that acts only when the pessimistic bound of the expected proxy effect exceeds a preset threshold under strict exposure caps, otherwise defaulting to NO-STEP. Governance features—drug–drug interaction matrices, order-set blocks, one-click rollback, and degrade-to-rules when calibration drifts—aim to convert combination experimentation into an auditable, risk-minimizing bedside layer. This review synthesizes the rationale for micro-dosed, order-aware adjuncts, summarizes implementable class-level candidates using approved agents and routine assays (with investigational items restricted to protocol settings), and outlines an evaluation plan embedded in care that prioritizes feasibility, safety, and mechanistic informativeness while protecting SoC delivery.