What the KRYSTAL-12 Lancet Correspondence Actually Says About PRO Reporting in KRASG12C Trials

A correspondence published in The Lancet in 2026, authored in response to Barlesi et al.’s primary KRYSTAL-12 report, deserves more careful reading than the brief exchange of academic pleasantries it might appear to be. Egbert F. Smit and Anne-Marie C. Dingemans are not simply complimenting the investigators on a well-run trial. They are surfacing a structural tension in how the field measures treatment benefit when efficacy signals are modest and toxicity burdens are real. The correspondence runs short, but what it points at is not short at all.

“Smit and Dingemans note that, although an in-depth analysis of quality-of-life measures was not provided in the primary publication, EQ-5D-5L scores appeared similar between treatment groups despite higher gastrointestinal toxicity with adagrasib.”

Read that sentence twice. The EQ-5D-5L scores appeared similar between arms even though adagrasib carried higher gastrointestinal toxicity. That finding can be read two ways, and which reading you choose has real consequences for how you design the next KRASG12C trial. Either the GI toxicity was genuinely well-managed and did not meaningfully erode patient experience, in which case the instrument captured something important. Or the EQ-5D-5L, a five-domain generic health utility index built for population health economics, lacked the sensitivity to detect symptom-specific deterioration in a lung cancer population already navigating a heavy disease burden. The correspondence does not resolve that ambiguity. That is precisely why it matters.

The Instrument Selection Problem

The KRYSTAL-12 trial compared adagrasib against docetaxel in previously treated advanced or metastatic KRASG12C-mutated non-small-cell lung cancer. Median PFS with adagrasib was 5.5 months versus 3.8 months for docetaxel, a clinically meaningful but numerically modest separation. In that context, the patient experience story becomes load-bearing. A 1.7-month PFS gain justifies itself differently depending on whether patients feel better, feel the same, or feel worse during those additional months.

“We thank Egbert F Smit and Anne-Marie C Dingemans for their Comment on our Article that reported outcomes from the KRYSTAL-12 trial and welcome their emphasis on the importance of patient-reported outcomes in contextualising treatment benefit for patients with KRASG12C-mutated non-small-cell lung cancer.”

The phrase “contextualising treatment benefit” carries regulatory weight that the authors may not have intended to foreground quite so sharply. The FDA’s final guidance on Core Patient-Reported Outcomes in Cancer Clinical Trials asks sponsors to collect a core set of patient-reported clinical outcomes in registration trials for anti-cancer drugs, specifically because surrogate endpoints like PFS do not, by themselves, tell a patient or a clinician how a treatment will affect daily function. The authors of the primary KRYSTAL-12 publication acknowledge this contextualizing role but provided the EQ-5D-5L as the instrument of record. That choice invites scrutiny under the FDA’s framework, which pushes sponsors toward disease-specific or symptom-specific instruments capable of detecting changes in the domains most affected by the treatment under study.

The EQ-5D-5L was not designed for this job.

“EQ-5D-5L scores appeared similar between treatment groups despite higher gastrointestinal toxicity with adagrasib.”

This is the sentence that a trial design team needs to sit with before the next KRASG12C protocol is finalized. The EQ-5D-5L measures mobility, self-care, usual activities, pain/discomfort, and anxiety/depression across five severity levels. Nausea, diarrhea, and vomiting, the GI toxicities most associated with adagrasib, do not have a dedicated domain. They surface, if at all, in the pain/discomfort item, where they compete with disease-related symptoms for scoring space. The result is a floor effect: meaningful toxicity can occur without registering as a score change. Under ICH E9(R1)’s estimand framework, that is not a neutral measurement choice. It is a design decision with consequences for the interpretation of benefit-risk, and the ASCO reporting standards published in the Journal of Clinical Oncology are explicit that PROs should be reported alongside surrogate outcomes to provide a complete picture of how therapies affect patients’ lives.

What the eCOA Infrastructure Needs to Support

The correspondence raises a second, quieter problem beneath the instrument question: the operational infrastructure required to capture meaningful PRO data in a global randomized trial. Smit and Dingemans’ observation that an in-depth PRO analysis was not provided in the primary publication is not a criticism of the investigators’ intentions. It points at a systemic constraint. If the eCOA platform running patient-reported data collection was not configured at the protocol stage to generate the longitudinal, visit-by-visit PRO dataset that a GI-toxicity analysis would require, no amount of post-hoc effort recovers that granularity.

“Although an in-depth analysis of quality-of-life measures was not provided in the primary publication, EQ-5D-5L scores appeared similar between treatment groups despite higher gastrointestinal toxicity with adagrasib.”

The passive construction here is worth unpacking. “Was not provided” leaves open whether the data exist and were not reported, whether the analysis was planned but underpowered, or whether the collection framework was never designed to support it. Each scenario carries a different implication for eClinical design. The FDA’s November 2024 warning letter to a sponsor following a BioResearch Monitoring Program inspection cited violations under 21 CFR 312.58 related to electronic data records and audit trails, a reminder that the integrity of patient-reported data is not merely a scientific question but a regulatory one. When the FDA asks why a PRO analysis is absent from a primary publication, “it was not pre-specified with sufficient granularity” is an uncomfortable answer in a post-approval review.

Adagrasib received accelerated FDA approval on December 12, 2022, for previously treated KRASG12C-mutated NSCLC. KRYSTAL-12 was the confirmatory trial designed to convert that accelerated approval to a full approval. The PRO data embedded in it carry post-marketing weight, which makes the depth of their analysis a matter of regulatory record, not just academic completeness.

The Competitive Stakes in a $1.12 Billion Market

The KRASG12C inhibitor market does not stand still while this methodological debate unfolds. The global KRAS G12C inhibitor market was valued at USD 1.12 billion in 2024, with sotorasib (Lumakras, Amgen) holding a parallel position in the same indication. When two agents compete in a space where PFS differences between active treatment and docetaxel are measured in weeks rather than months, the PRO narrative becomes a commercial differentiator. Payers, oncologists, and patients comparing adagrasib and sotorasib will increasingly ask not just which drug extends progression-free survival longer, but which drug patients can tolerate over that survival period with the least erosion of daily function.

That question cannot be answered with EQ-5D-5L alone.

The correspondence between Smit, Dingemans, and the KRYSTAL-12 investigators is, read end-to-end, a field-level signal that the current PRO reporting standard in KRASG12C trials has not caught up with what prescribers and regulators actually need to know. The primary publication reported the number that mattered for approval: PFS favored adagrasib. The PRO data that would have allowed a clinician to look a patient in the eye and say “here is what your daily life is likely to look like on this drug, compared to the alternative,” that data was not assembled in a form that the primary publication could deliver. The Smit-Dingemans comment does not say this directly. It says it precisely.

Sponsors running confirmatory trials in KRASG12C NSCLC, and more broadly any trial where the active arm carries a differentiated toxicity profile against a modest PFS increment, should treat this correspondence as a protocol design checklist item. Pre-specify a disease-specific or symptom-specific PRO instrument alongside the EQ-5D-5L, with collection windows aligned to the expected onset of the toxicities of interest. Configure the eCOA platform to capture daily or weekly symptom diaries during the periods of highest GI toxicity exposure, not just at scheduled clinic visits. Build the longitudinal dataset at the outset rather than attempting to reconstruct it at the time of publication. The FDA’s Core PRO guidance provides the scaffold. The KRYSTAL-12 experience illustrates what happens when that scaffold is not erected early enough in the design process to bear the analytical weight the trial eventually needs to carry.

The next document to watch in this space is the FDA’s review memorandum for adagrasib’s full approval conversion. When the agency’s reviewers write about the PRO package submitted in support of that conversion, their language will tell you exactly how much the EQ-5D-5L gap cost the sponsor in the benefit-risk narrative.

References

1. The Lancet — “Correspondence: Patient-reported outcomes with adagrasib in KRYSTAL-12”
2. ASCO Post — “Adagrasib vs. Docetaxel in Previously Treated Advanced or Metastatic KRASG12C-Mutated NSCLC”
3. FDA — “Core Patient-Reported Outcomes in Cancer Clinical Trials: Final Guidance”
4. Oncology Practice Management — “KRAZATI Receives Accelerated FDA Approval for NSCLC with KRASG12C Mutation”
5. Clinical Pathways Research — “FDA Warning Letter to Sponsor: Records, Reports, and Audit Trails”
6. Journal of Clinical Oncology (ASCO) — “Patient-Reported Outcomes Reporting Standards in Cancer Clinical Trials”
7. Dataintelo — “KRAS G12C Inhibitors Market Report, 2024”