Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Clinical Trial
. 2024 Aug 10;15(1):6862.
doi: 10.1038/s41467-024-51316-7.

Identifying key circulating tumor DNA parameters for predicting clinical outcomes in metastatic non-squamous non-small cell lung cancer after first-line chemoimmunotherapy

Affiliations
Clinical Trial

Identifying key circulating tumor DNA parameters for predicting clinical outcomes in metastatic non-squamous non-small cell lung cancer after first-line chemoimmunotherapy

Haolun Ding et al. Nat Commun. .

Abstract

Circulating tumor DNA (ctDNA) provides valuable tumor-related information without invasive biopsies, yet consensus is lacking on optimal parameters for predicting clinical outcomes. Utilizing longitudinal ctDNA data from the large phase 3 IMpower150 study (NCT02366143) of atezolizumab in combination with chemotherapy with or without bevacizumab in patients with stage IV non-squamous Non-Small Cell Lung Cancer (NSCLC), here we report that post-treatment ctDNA response correlates significantly with radiographic response. However, only modest concordance is identified, revealing that ctDNA response is likely not a surrogate for radiographic response; both provide distinct information. Various ctDNA metrics, especially early ctDNA nadirs, emerge as primary predictors for progression-free survival and overall survival, potentially better assessing long-term benefits for chemoimmunotherapy in NSCLC. Integrating radiographic and ctDNA assessments enhances prediction of survival outcomes. We also identify optimal cutoff values for risk stratification and key assessment timepoints, notably Weeks 6-9, for insights into clinical outcomes. Overall, our identified optimal ctDNA parameters can enhance the prediction of clinical outcomes, refine trial designs, and inform therapeutic decisions for first-line NSCLC patients.

PubMed Disclaimer

Conflict of interest statement

X.S.X. is an employee of Genmab, Inc. Genmab did not provide funding for this study. The remaining authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Summary of plasma variants detected pretreatment in patients with metastatic non-squamous 1 L NSCLC.
The mutation count per sample is shown at the top, while alteration prevalence for each gene is listed on the right. Different colors represent the status of driver genes. Among a total of 262 patients, 48 patients exhibited undetectable tumor-derived mutations pretreatment. Source data are provided as a Source Data file.
Fig. 2
Fig. 2. Composite ranking of concordance between ctDNA features and radiographic response.
A barplot illustrating –log10 (P-values after FDR correction) of the association between Best Objective Response and the top 10 ctDNA longitudinal features based on logistic regression models on the training set (gray bars represents P-values ≥ 0.05 after FDR correction); P-values are calculated by a two-sided Z test using logistic regression models; forest plots depicting estimated odds ratio and barplots showing AUC on both training and test datasets based on the optimal cutoff point identified from the training set; error bars represent the exponential of estimated coefficients ± 1.96*SEM (N = 253). B Waterfall/distribution plots and confusion matrices for training and test sets utilizing optimal cutoff points identified from the training set for the top 3 ctDNA features (max % change of overall mutation numbers (n), max % change of n_kl (known or likely pathogenic mutation numbers), nadir of n_kl). Cut-points were determined by maximizing the Youden index (Youden = sensitivity + specificity – 1). Source data are provided as a Source Data file.
Fig. 3
Fig. 3. Composite ranking of predictive ctDNA features for progression-free survival (PFS).
A barplot illustrating –log10 (P-values after FDR correction) on the training set; P-values are calculated by a two-sided Z test using Cox proportional hazards models; forest plots depicting estimated hazard ratio and barplots showing c-index on both training and test datasets based on the optimal cutoff point identified from the training set (Top 10 ctDNA longitudinal features); error bars represent the exponential of estimated coefficients ± 1.96*SEM (N = 262). B distributions and Kaplan–Meier plots for training and test sets utilizing optimal cutoff points identified from the training set for the top 3 ctDNA features (nadir of median AF, median AFkl, and median TMPMP); P-values are determined using a two-sided log-rank test to compare Kaplan–Meier curves. The exact P-values, from left to right by row, are 4.17e-8 (Training set: 63 patients ≥ cutpoint; 74 patients <cutpoint), 1.44e-9 (Test set: 55 patients ≥ cutpoint; 70 patients <cutpoint), 1.92e-7 (Training set: 56 patients ≥ cutpoint; 81 patients <cutpoint), 3.51e-9 (Test set: 47 patients ≥ cutpoint; 78 patients <cutpoint), 2.91e-7 (Training set: 70 patients ≥ cutpoint; 67 patients <cutpoint), and 1.57e-11 (Test set: 63 patients ≥ cutpoint; 62 patients <cutpoint). C average ctDNA kinetics (median AF) for early progressors (progressive disease at Week 6, Week 12, and up to Week 12) and non-early progressors (CR/PR/SD). In the ‘Response at Week 6’ plot, 18, 18, 6, and 4 early progressors, and 235, 226, 213, and 205 non-early progressors were ctDNA evaluable at Weeks 0, 3, 6, and 9, respectively. In the “Response at Week 12” plot, 11, 11, 11, and 10 early progressors, and 208, 199, 193, and 191 non-early progressors were ctDNA evaluable at Week 0, 3, 6, and 9, respectively. In the “Response up to Week 12” plot, 25, 25, 13, and 11 early progressors, and 208, 199, 193, and 191 non-early progressors were ctDNA evaluable at Week 0, 3, 6, and 9, respectively. Blue shaded areas represent response evaluation periods. Source data are provided as a Source Data file.
Fig. 4
Fig. 4. Composite ranking of predictive ctDNA features for overall survival (OS).
A barplot illustrating –log10 (P-values after FDR correction) on the training set; P-values are calculated by a two-sided Z test using Cox proportional hazards models; forest plots depicting the estimated hazard ratio and barplots showing c-index on both training and test datasets based on the optimal cutoff point identified from the training set (Top 10 ctDNA longitudinal features); error bars represent the exponential of estimated coefficients ± 1.96*SEM (N = 262). B distributions and Kaplan–Meier plots for training and test sets utilizing optimal cutoff points identified from the training set for the top 3 ctDNA features (nadir of median TMPMP, median AFkl at Week 21, and nadir of median AFkl); P-values are determined using a two-sided log-rank test to compare Kaplan–Meier curves. The exact P-values, from left to right by row, are 1.80e-8 (Training set: 68 patients ≥ cutpoint; 69 patients <cutpoint), 6.37e-12 (Test set: 62 patients ≥ cutpoint; 63 patients <cutpoint), 1.17e-5 (Training set: 35 patients ≥ cutpoint; 53 patients <cutpoint), 1.93e-9 (Test set: 19 patients ≥ cutpoint; 58 patients <cutpoint), 2.22e-8 (Training set: 70 patients ≥ cutpoint; 67 patients <cutpoint) and 6.49e-8 (Test set: 63 patients ≥ cutpoint; 62 patients <cutpoint). Source data are provided as a Source Data file.
Fig. 5
Fig. 5. .
Landmark analysis for comparing prediction of overall survival (OS; N = 219) and progression-free survival (PFS; N = 214) using radiographic response or/and molecular risk stratification. A OS by radiographic response; B OS by molecular risk stratification; C OS by combining radiographic response and molecular risk stratification;  D PFS by radiographic response; E PFS by molecular risk stratification; and F PFS by combining radiographic response and molecular risk stratification. For OS analysis, patients were categorized into molecular low- and high-risk groups based on nadir of median TMPMP up to the landmark of Week 6. In the case of PFS analysis, molecular low- and high-risk patients were defined by a nadir of median AF up to the landmark of Week 6. Radiographic response at Week 6 was used in the landmark analysis; P-values are calculated using a two-sided log-rank test to compare Kaplan–Meier curves, yielding P-values of 2.7e-4 (A), 1.24e-10 (B), 7.55e-12 (C), 3.9e-4 (D), 3.57e-10 (E), and 1.83e-11 (F) respectively. Source data are provided as a Source Data file.

References

    1. Rossi, G. & Ignatiadis, M. Promises and pitfalls of using liquid biopsy for precision medicine. Cancer Res.79, 2798–2804 (2019). 10.1158/0008-5472.CAN-18-3402 - DOI - PubMed
    1. Wan, J. C. M. et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nat. Rev. Cancer17, 223–238 (2017). 10.1038/nrc.2017.7 - DOI - PubMed
    1. Narayan, P. et al. State of the science and future directions for liquid biopsies in drug development. Oncologist25, 730–732 (2020). 10.1634/theoncologist.2020-0246 - DOI - PMC - PubMed
    1. FDA draft guidance for industry: use of circulating tumor DNA for early stage solid tumor drug development; Available from: https://www.fda.gov/media/158072/download (2022).
    1. Vellanki, P. J. et al. Regulatory implications of ctDNA in immuno-oncology for solid tumors. J. Immunother. Cancer11, e005344 (2023). 10.1136/jitc-2022-005344 - DOI - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources