Our model suggests that ctDNA could improve clinical decision-making for stage I NSCLC. Prospectively identified high-risk patients could benefit from systemic therapy. We assumed that occult micrometastases and ctDNA had clinical significance, which is supported in published literature.
Julie Koenig, Ben Durkee, Erqi Pollom, Iris Gibbs, Max Diehn; Stanford University School of Medicine; Department of Radiation Oncology, Stanford University Medical Center; Institute for Stem Cell Biology and Regenerative Medicine, Stanford University
BACKGROUND: A nontrivial fraction of stage I non–small-cell lung cancer (NSCLC) is occult metastatic disease [Rusch, JCO 2011]. These patients have worse disease-free survival and overall survival (OS) [Rusch, JCO 2011]. The phase III Cancer and Leukemia Group B (CALGB) 9633 trial found a small benefit for adjuvant chemotherapy in stage I patients with high-risk disease based on size criterion [Strauss, JCO 2008]. However, the authors could not reproducibly identify these high-risk patients [Strauss, JCO 2008].
Circulating tumor DNA (ctDNA) is highly predictive of residual disease after definitive therapy [Newman, Nat Med 2014]. We hypothesize that ctDNA could identify high-risk patients with stage I NSCLC who could benefit from early systemic therapy.
METHODS: We created a Bayesian model to simulate posterior probabilities after screening for occult metastatic disease by ctDNA. We used prevalence data from the American College of Surgeons Oncology Group (ACOSOG) Z0040 trial, and recently published receiver operating characteristics for ctDNA [Newman, Nat Med 2014].
Next, we built a two-state Markov model, with the assumption that occult micrometastases conferred worse survival (hazard ratio [HR] = 1.82), based on ACOSOG. Patients receiving chemotherapy were assigned a survival benefit (HR = 0.69, high risk) and detriment (HR = 1.12, low risk) by extrapolation from Cancer and Leukemia Group B (CALGB) trial 9633. The model was run with no patient receiving chemotherapy and rerun with patients who had detectable ctDNA receiving adjuvant chemotherapy.
RESULTS: The 3-year OS rate was 71% and 73% for high- and low-risk patients, respectively, with a total payoff of 2.56 life-years. These modeled estimates correlate well with CALGB data. The positive likelihood ratio for ctDNA was 10.0. After screening for ctDNA, the posterior probability of occult disease for a positive test improved to 67.0%, from a baseline prevalence of 16.9%. Treating high-risk patients with adjuvant chemotherapy improved survival to 79%. The number needed to treat was 12.5.
CONCLUSION: Our model suggests that ctDNA could improve clinical decision-making for stage I NSCLC. Prospectively identified high-risk patients could benefit from systemic therapy. There are limitations to this study: a model does not substitute for a clinical trial. We did not do a sensitivity analysis. Our time horizon (3 yr) is short. We assumed that our high-risk patients (defined by ctDNA) would derive the same benefit as high-risk patients in the CALGB trial (defined by size). To support this, the data show that ctDNA levels correlate with tumor volume [Newman, Nat Med 2014]. We assumed that occult micrometastases and ctDNA had clinical significance, which is supported in published literature.
Proceedings of the 97th Annual Meeting of the American Radium Society - americanradiumsociety.org