Serial CT Response Score Predicts OS in Patients With Advanced NSCLC Receiving ICIs

Serial CT response score (CTRS) predicted overall survival (OS) more effectively than existing imaging-based measures in patients with advanced non–small cell lung cancer (NSCLC) receiving immune checkpoint inhibitors (ICIs), according to a study published in JAMA Network Open.¹

Addressing Gaps in Imaging-Based Prognosis for Advanced NSCLC

The researchers emphasized that accurately predicting long-term prognosis during ICI therapy remains a critical unmet need. Conventional imaging-based measures, such as tumor volume change (TVC) and Response Evaluation Criteria in Solid Tumors (RECIST), have limited the ability to predict long-term outcomes. As a result, advanced imaging-based biomarkers may improve decision-making in clinical practice and trials.

To address this gap, the researchers developed the externally validated Serial CTRS, a fully automated, deep-learning radiomics-based biomarker that uses pretreatment and early-treatment routine chest CT images to predict long-term OS. It was developed using a routine clinical practice (RCP) discovery cohort and validated using electronic health record data from adult patients with advanced NSCLC treated with ICIs, either as monotherapy or in combination therapy, across all lines of care at 10 institutions in the US and Europe between 2013 and 2023.

Additionally, independent validation was performed using data from the phase 1 GARNET trial (NCT02715284).² Overall, eligible participants included adults with advanced NSCLC who initiated ICIs between 2013 and 2021 in the RCP discovery cohort, 2013 and 2022 in the RCP test group, and 2017 and 2018 in the GARNET trial.¹ The researchers used Cox proportional hazards regression and receiver operating characteristic-area under the curve analyses to assess associations between Serial CTRS and OS.

Serial CTRS Outperforms Conventional Imaging Measures

The final study population included 1830 eligible patients with a median age of 67 years; the majority were male (55%; n = 1000). Of these patients, 1171 were from the RCP discovery cohort, 605 from the RCP test cohort, and 54 from the GARNET trial.

After adjusting for age, sex, PD-L1 expression, histologic profile, and tumor volume, Serial CTRS remained significantly associated with OS in multivariable analyses. A 10-percentage-point increase in the probability of 12-month OS was associated with HRs of 0.74 (95% CI, 0.70-0.79) in the RCP test cohort and 0.45 (95% CI, 0.32-0.65) in the GARNET trial group.

Serial CTRS also outperformed RECIST and TVC in OS risk discrimination. HRs distinguishing low- and high-survival groups were higher with serial CTRS in both the RCP test (HR, 6.19; 95% CI, 4.12-9.28) and GARNET (HR, 18.00; 95% CI, 5.40-59.97) cohorts.

Lastly, the researchers highlighted that the predictive value of Serial CTRS was consistent across PD-L1 expression levels and RECIST-defined subgroups, including patients with stable disease.

“These findings suggest that Serial CT response scores could improve clinical decision-making and enhance clinical trial designs for patients with NSCLC,” the authors wrote.

Serial CTRS Shows Promise, but Further Validation Is Needed

The researchers acknowledged several limitations, including that Serial CTRS currently offers limited interpretability of the imaging features driving predictions, a common issue with deep-learning approaches. However, they noted that ongoing work aims to improve model transparency.

Despite the limitations, the researchers emphasized the potential of Serial CTRS for predicting OS in patients with advanced NSCLC receiving ICIs.

“The automated design of Serial CTRS facilitates future integration into clinical practice and clinical trial workflows,” the authors concluded. “With further validation across therapeutic modalities, Serial CTRS has the potential to enable more accurate, early treatment readouts in both clinical practice and clinical trial settings.”

References

1. Sako C, Kurland BF, Schmidt TG, et al. Deep-learning serial CT prediction of survival in immunotherapy-treated non–small cell lung cancer. JAMA Netw Open. 2026;9(1):e2555759. doi:10.1001/jamanetworkopen.2025.55759
2. Study of TSR-042, an anti-programmed cell death-1 receptor (PD-1) monoclonal antibody, in participants with advanced solid tumors (GARNET). ClinicalTrials.gov. Updated January 13, 2026. Accessed February 5, 2026. https://clinicaltrials.gov/study/NCT02715284