Higher thymic radiation exposure was associated with an increased risk of distant metastasis among patients with locally advanced non–small cell lung cancer (NSCLC), particularly among patients whose thymus was well preserved before treatment, according to research published in Annals of Oncology.

“The thymus is currently not routinely treated as an important organ during radiation planning in adults,” said senior study author Hugo J. W. L. Aerts, PhD, Professor at Harvard University and Director of the Artificial Intelligence in Medicine (AIM) Program at Mass General Brigham.

“Our findings raise the possibility that preserving immune health during medical treatment may become a new principle of medicine. Modern oncology increasingly depends on the patient’s immune system, particularly as immunotherapies become part of standard treatment across many cancers. As we continue to improve cancer treatment, we may need to think not only about protecting organs such as the heart, lungs and spinal cord, but also about preserving the immune system itself,” Dr. Aerts added. “If confirmed in prospective studies, the thymus could become one of the first immune organs routinely protected during treatment planning.”

Prior Related Research

The study builds on the team’s earlier work establishing AI-derived thymic health as an imaging biomarker. In two papers published in Nature, the researchers applied deep learning to CT scans from more than 25,000 adults in a national lung cancer screening trial and more than 2,500 participants in the Framingham Heart Study. Higher thymic health was associated with lower all-cause mortality and a lower risk of developing lung cancer. In a separate cohort of more than 1,200 patients treated with immunotherapy, higher thymic health was associated with a 37% lower risk of disease progression and a 44% lower risk of death.

This work resulted in the development and validation of a deep learning model, called Thymic Health, for estimating thymic function.

“The thymus is a small immune organ located in the upper chest that helps produce and train T cells, which are essential for recognizing infections and fighting cancer. It is most active early in life and gradually shrinks with age when functional tissue is replaced by fat. For many years, people interpreted this as meaning the adult thymus was essentially irrelevant,” Dr. Aerts said.

Study and Model Methods

In this study, researchers used a deep learning model on routine CT scans to estimate thymic health in patients with NSCLC, then paired those estimates with radiation treatment plans to quantify how much incidental dose the thymus received.

The analysis included three cohorts of patients, comprising 460 patients from the randomized phase III RTOG-0617 trial, who were treated with 60 Gy or 74 Gy of thoracic radiation plus carboplatin-paclitaxel–based chemotherapy, with or without cetuximab, plus two real world cohorts of patients treated with platinum-based chemoradiotherapy (n = 422) or chemoradiotherapy followed by consolidation durvalumab (n = 225).

The Thymic Health model evaluates the size, shape, density, and tissue composition of the thymus and produces a thymic health score, quantified on a scale between 0 and 1, with higher scores representing well-preserved thymic tissue.

The team then used each patient’s radiation treatment plan to calculate mean thymic dose. Associations between mean thymic dose and distant metastasis and overall survival were assessed with multivariable models adjusted for clinical confounders. The researchers also tested whether pretreatment thymic health modified these associations. Follow-up CT scans taken about 1 year after treatment were analyzed with the same model to track changes in thymic health.

Additionally, an exploratory feasibility study in 16 patients tested whether thymic dose could be reduced without compromising the treatment plan.

Study Findings

Incidental thymic irradiation was associated with a higher risk of distant metastasis in all three cohorts, even after accounting for confounding factors. Each 1-Gy increase in mean thymic dose was associated with a 1.6% to 4.2% increase in the risk of distant metastasis, depending on the cohort. Adjusted hazard ratios (aHRs) were 1.29 in RTOG-0617 (P = .0028), 1.33 in the real-world chemoradiotherapy group (P = .011), and 1.95 in the real-world durvalumab group (P = .007). The strongest signal came from the cohort that received immunotherapy.

“In all three patient groups, we found that higher radiation exposure to the thymus was consistently associated with a greater risk of the cancer spreading to distant parts of the body,” Dr. Aerts said.

Pretreatment thymic health, as measured by the model, determined who carried this risk. Patients with preserved thymic health had the best outcomes overall. However, they were also the only group in which higher thymic dose was associated with distant metastasis.

Among patients with preserved thymic health, higher thymic dose was also associated with worse overall survival, with aHRs of 1.25 in RTOG-0617 and 1.23 in the chemoradiotherapy alone group. No such association was seen in patients whose thymic health was already impaired before radiotherapy.

In an exploratory analysis, a mean thymic dose above 35 Gy was associated with a 40% to 96% higher risk of distant metastasis, depending on the cohort. In RTOG-0617, the HR was 1.40 (95% confidence interval [CI] = 1.01–1.93). The investigators stressed that this threshold is population-specific and should not be treated as a validated clinical dose constraint.

On follow-up imaging 1 year after treatment, the model detected dose-dependent declines in thymic health. In a smaller subgroup, higher thymic dose was also associated with lower circulating lymphocyte counts. The researchers said these findings suggest a possible biological link between thymic radiation and loss of immune competence.

In the 16-patient replanning study, thymic dose could often be substantially reduced with existing planning techniques while maintaining tumor coverage and cardiopulmonary dose constraints.

“Perhaps most encouragingly, we found that protecting the thymus would not require a new treatment or new technology,” Dr. Aerts commented. “Using existing radiotherapy planning techniques, radiation exposure to the thymus could often be substantially reduced without compromising treatment of the cancer or increasing radiation to other important organs. If future studies confirm that this improves outcomes, thymus-sparing could be implemented relatively rapidly in routine cancer care.”

The study was retrospective, and the real-world cohorts may carry differences in treatment and patient selection that introduce bias. The authors also noted that the thymic health model will need further validation across different scanners, institutions, and countries before it can be used in clinical practice. The 35-Gy threshold was exploratory and requires testing in prospective studies.

“Over the past several decades, radiation oncology has become increasingly successful at protecting healthy organs in the chest like the heart and lungs, while maintaining excellent cancer control. Our findings suggest the thymus may deserve consideration alongside those organs. Importantly, this can often be achieved using existing treatment-planning techniques, meaning it could potentially be adopted relatively quickly if confirmed in prospective studies,” added co-senior study author Raymond H. Mak, MD, a radiation oncologist at Mass General Brigham Cancer Institute and Professor at Harvard Medical School.

DISCLOSURES: This work was supported by the National Institutes of Health National Cancer Institute, the European Union–European Research Council, the Deutsche Forschungsgemeinschaft, the Lundbeck Foundation, the Novo Nordisk Foundation, and a Savvaerksejer Jeppe Juhl og Hustru Ovita Juhl research stipend. For full disclosures of the study authors, visit annalsofoncology.org.

ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO^®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.