Preoperative Conversion of Blood Tumor Signals to Negative Outlines Recurrence Risk After Lung Cancer Immunotherapy

Preoperative imaging for lung cancer can show how much a tumor has shrunk, but may not reveal microscopic residual disease scattered throughout the body. A biomarker analysis from the phase 3 CheckMate 77T trial offers another window of observation: tracking circulating tumor DNA (ctDNA) in the blood and combining it with surgical pathology and clinical characteristics may enable more precise identification of which patients have a lower risk of recurrence and which still harbor warning signs.

CheckMate 77T enrolled patients with resectable stage IIA to IIIB non-small cell lung cancer and compared preoperative nivolumab plus chemotherapy followed by surgery and continued nivolumab with a placebo group undergoing the same treatment sequence. After a median follow-up of 41 months in the overall randomized population, the nivolumab group had a lower risk of recurrence, progression, or death, with a hazard ratio of 0.61; however, the overall survival difference at the first prespecified interim analysis had not crossed the threshold for statistical significance.

Using tumor tissue information, the research team customized ctDNA assays to detect each patient’s tumor-specific variants before and after treatment. Among 190 patients eligible for biomarker analysis, 140 had interpretable ctDNA results both before and after preoperative treatment; 66% of patients in the nivolumab group achieved ctDNA clearance before surgery, compared with 38% in the placebo group. In the nivolumab group, half of those with ctDNA clearance achieved a pathological complete response; none of those without clearance met this standard, indicating that blood signals and resection specimens provide mutually corroborating but not entirely overlapping information.

Postoperative monitoring also revealed a sharper warning sign. All 13 patients who were initially negative for molecular residual disease and later converted to positive during adjuvant treatment experienced recurrence; this conversion occurred in 4 patients in the nivolumab group and 9 in the placebo group. The sample was very small and cannot be used to estimate an individual patient’s outcome, but it supports understanding recurrence risk through longitudinal changes rather than a single blood draw.

To integrate these clues, researchers trained a random survival forest model using 80% of the data from the biomarker-evaluable population and tested it on the remaining 20%. Preoperative ctDNA clearance, absence of N2 mediastinal lymph node involvement, pathological complete response, squamous histology, and receipt of nivolumab were among the more important predictors of event-free survival. The model’s discrimination metric was 0.79 in the training data and fell to 0.65 in the test data, indicating that it captured some signal but still had limited generalizability. Moreover, the model predicted survival risk and did not demonstrate that it could accurately identify which patients would derive additional benefit from nivolumab.

The analysis also showed that patients with variants in KEAP1, STK11, CDKN2A, or SMARCA4, which have previously often been associated with poor prognosis, were not necessarily unable to benefit from perioperative immunotherapy. This suggests that treatment should not be ruled out clinically on the basis of a single gene variant, but the number of patients in each genetic subgroup was limited and the confidence intervals for some results were wide, leaving the findings insufficient as an independent basis for prescribing decisions.

The greatest limitation of this work is that only about 41% of randomized patients had adequate paired tumor and blood samples. Multiple analyses were post hoc and exploratory, and the model incorporated ctDNA clearance and pathological response, which are known only after treatment; it therefore cannot be used directly before therapy begins and has not yet been validated in an external population. At present, a more reasonable role is to regard ctDNA and pathological findings as tools for designing future risk-stratified trials to further test which patients can safely receive less treatment and which require intensified postoperative intervention, rather than immediately changing an individual patient’s established treatment course.