AACR 2026: 3 shifts in cancer research oncologists can’t ignore
At this year’s annual American Association for Cancer Research (AACR) meeting, the signals aren't subtle.
Across biomarker science, adjuvant decision-making, and drug development, several trends are moving from academic curiosity into clinical friction—where oncologists have to decide what to trust, what to ignore, and what to start preparing for.
Here are three shifts reshaping the near-term landscape, and what they mean when you’re sitting across from a patient.
Related: Drug resistance is reshaping lung cancer care—AACR 2026 may reveal what’s nextAI is outperforming PD-L1 and TMB, but hasn’t earned a seat in decision-making
For years, PD-L1 and tumor mutational burden have anchored immunotherapy decisions despite well-known limitations. What’s changed is not just incremental improvement—it’s displacement.
Multiple models presented this year showed materially stronger predictive performance than PD-L1 and TMB across tumor types, using inputs that are already sitting in your workflow: histopathology slides, routine labs, genomics panels, and EMR data. In some cases, hazard ratios and AUCs weren’t even close.[]
The more important shift is structural. These are multimodal models—combining clinical, molecular, and imaging data into unified predictions. That’s a different paradigm from single-analyte biomarkers.
None of these models have prospective interventional validation. No trial has shown that acting on these predictions improves outcomes.
What this means for the clinic: Treat AI outputs as decision-support, not decision-makers. If you encounter these tools (especially embedded in commercial platforms), use them to pressure-test your clinical intuition.
ctDNA is making de-escalation feel real
Minimal residual disease testing with ctDNA has been “promising” for years. New data presented at AACR 2026 suggests it may start to feel actionable—at least in one direction.
Across colorectal, lung, hepatocellular, and uterine cancers, postoperative ctDNA-negative patients consistently showed very low recurrence risk.[][][][] In some analyses, projected recurrence rates dropped into ranges where oncologists already feel comfortable omitting adjuvant therapy.
ctDNA-positive patients carry higher risk, but no new randomized data showed that intensifying therapy improves overall survival. Regulatory acceptance of MRD as a surrogate endpoint remains unresolved.
So the field is now split: strong prognostic signal, incomplete therapeutic validation.
What this means for the clinic: When ctDNA is negative postoperatively, start thinking seriously about de-escalation discussions—especially in borderline cases. When it’s positive, resist the urge to reflexively intensify therapy outside trials unless other clinical factors clearly support it.
In solid tumors, delivery platforms—not just targets—are driving the next wave
For years, oncology pipelines were framed around targets: HER2, TROP2, KRAS. That framing is breaking down.
These are direct responses to known failure modes of earlier drugs.
In preclinical and early clinical data, some of these platforms are already outperforming first-generation agents—even against resistant disease.
What this means for the clinic: The sequencing question is getting more complicated. As these agents enter trials and early approvals, start tracking not just the target but the platform. When considering clinical trials or later-line options, prioritize matching mechanism of delivery (eg, dual payload, conditional activation) to the patient’s resistance pattern.