Colorectal cancer (CRC) is the second leading cause of cancer-related deaths in the United States. After curative removal of the primary tumor, cancer recurrence remains the dominant cause of death. Among patients whose cancer spreads beyond the colon or rectum, the lung is the most common site of distant spread outside the abdomen.
When CRC spreads only to the lungs and nowhere else, pulmonary metastasectomy (surgical removal of lung tumors) has been associated with improved survival and has become widely accepted for appropriately selected patients. However, critical questions have remained unanswered: which patients benefit most, and does adding chemotherapy around the time of surgery improve outcomes compared to surgery alone?
Prior studies identified two key factors that predict outcomes after lung surgery: a longer disease-free interval (DFI), meaning the time between removal of the primary colon tumor and detection of lung spread, and fewer lung metastases. Patients with a DFI of 12 months or more and three or fewer lung tumors tend to do well; those with a DFI under 6 months or more than 6 lung tumors face a more difficult prognosis. All prior evidence for this had come from retrospective reviews rather than randomized controlled trials.
The Thoracic Surgery Oncology Group (TSOG) 103 trial was a prospective, multi-institutional, randomized controlled trial conducted across three academic cancer centers from July 2018 to September 2023. Its goal was to fill the evidence gap by formally testing two separate clinical questions within a single protocol, using a risk-stratified design that assigned patients to different treatment pathways based on their individual disease characteristics.
All patients had histologically confirmed colorectal adenocarcinoma with lung-limited metastatic disease and were eligible for both surgical resection and chemotherapy. They were classified as low-risk (DFI of 12 months or more and three or fewer metastases) or high-risk (DFI under 6 months or more than 6 metastases) before randomization. Patients with intermediate-risk features were not enrolled, to maintain treatment clarity within each cohort.
Low-risk patients were randomized to pulmonary metastasectomy with or without perioperative chemotherapy (three months of chemotherapy before and three months after surgery versus surgery alone). High-risk patients all received three months of chemotherapy first, then were re-imaged and randomized to surgical resection or continued chemotherapy, provided their disease had not progressed. Patients whose disease worsened during initial chemotherapy were removed from the trial.
Randomization was stratified by the side of the primary colon tumor (left versus right) and KRAS mutation status, both of which are known to affect outcomes after pulmonary metastasectomy. All operations were performed with curative intent, requiring complete removal of all visible disease with clear surgical margins (R0 resection). Permitted procedures included open thoracotomy, minimally invasive thoracoscopy, and robotic-assisted surgery, at the surgeon's discretion.
A total of 48 patients were enrolled: 22 in the low-risk group and 26 in the high-risk group. The majority of patients came from MD Anderson Cancer Center (86 to 89% per cohort), with smaller contributions from Thomas Jefferson University and the University of Toronto. The median age at cancer diagnosis was approximately 57 to 59 years in the low-risk group and 52 to 59 years in the high-risk group, and both groups had roughly equal numbers of men and women.
In the low-risk cohort, 20 of 22 patients were randomized: 8 received perioperative chemotherapy plus surgery and 12 underwent surgery alone. All patients in both groups had an ECOG performance status of 0, meaning they were fully active and capable of normal physical activity. Wedge resection, the least extensive surgical approach removing only the tumor and a small surrounding rim of lung tissue, was the most common extent of operation (71 to 75% of patients). All patients achieved complete tumor removal with clear surgical margins, and there were no postoperative complications.
In the high-risk cohort, 4 of 26 patients were excluded after disease progressed during initial chemotherapy and could not be randomized. Of the remaining 22, 8 received chemotherapy only and 14 received chemotherapy followed by surgery. Initial chemotherapy produced a complete or partial response in about 43 to 50% of patients in both groups, indicating that a meaningful proportion had their lung tumors shrink before any surgery was considered.
In the low-risk cohort, cancer recurred in 5 of 8 patients (62.5%) who received chemotherapy plus surgery, compared to 4 of 12 (33.3%) who underwent surgery alone, though this difference did not reach statistical significance (P = .20). The median recurrence-free survival (RFS) was 21.8 months in the chemotherapy-plus-surgery group, while RFS for surgery-alone patients had not yet been reached at the time of analysis. Three-year RFS was markedly better in the surgery-alone arm (61.9% versus 17.9%), though again the study was too small to draw firm conclusions from this difference.
For overall survival in the low-risk group, one death occurred in each treatment group. Three-year overall survival was similar between arms: 85.7% with chemotherapy plus surgery and 88.9% with surgery alone (P = .68). Importantly, no clear survival benefit was seen from adding perioperative chemotherapy for low-risk patients, suggesting that carefully selected low-risk patients may do well with surgery alone.
In the high-risk cohort, recurrence occurred in 4 of 8 chemotherapy-only patients (50%) and 6 of 14 chemotherapy-plus-surgery patients (42.9%). The median RFS was 33.4 months with chemotherapy alone and 55.8 months with surgery added, a numerically encouraging difference that did not reach statistical significance (P = .95). No deaths occurred in either high-risk group during the study period, so overall survival could not be calculated.
The trial's most significant limitation was insufficient patient enrollment. The low-risk cohort aimed to enroll 150 patients but recruited only 22; the high-risk cohort targeted 215 but enrolled just 26. As a result, the study is substantially underpowered: it cannot definitively confirm or rule out clinically important differences between treatment groups. All results must be interpreted as descriptive and hypothesis-generating rather than conclusive.
Slow enrollment was driven by two major forces. First, the growing adoption of circulating tumor DNA (ctDNA) testing as a guide for clinical decision-making shifted how physicians managed these patients outside of the trial. Second, clinicians and patients often lacked equipoise, meaning many believed surgery was the better option and were reluctant to accept randomization to chemotherapy only. The difficulty of presenting surgical and non-surgical options to patients who perceive surgery as potentially curative is a recurring challenge in oncology trials.
The therapeutic landscape for metastatic CRC also changed substantially during the trial's enrollment period. Systemic chemotherapy regimens improved, new targeted therapies became available, and evidence supporting stereotactic body radiotherapy and ablative techniques for lung metastases expanded. These advances may have further reduced enthusiasm for participating in a trial that did not incorporate these options.
Despite these limitations, the trial provides important operational insights for future studies: the rates of crossover between arms, the proportion of patients lost before randomization, and the challenges of multi-institution coordination will help future investigators design more successful trials of surgical management for lung-limited metastatic CRC.
For patients with low-risk lung-limited metastatic CRC, meaning those with a disease-free interval of a year or more and three or fewer lung metastases, the TSOG 103 findings suggest that up-front surgery without additional chemotherapy may be sufficient. The surgery-alone group showed numerically better recurrence control at three years, and no survival advantage was demonstrated from adding perioperative chemotherapy. This aligns with a common clinical intuition that adding toxic systemic therapy may not benefit patients whose disease biology is already favorable.
For patients with high-risk disease, the question of whether to add surgery after chemotherapy remains unresolved. The trend toward longer recurrence-free survival in patients who had surgery added to chemotherapy (55.8 versus 33.4 months) is encouraging, but the study enrolled too few patients to confirm a real benefit. The finding that no deaths occurred in either high-risk group during the study period reflects the relatively good outcomes achievable with modern chemotherapy even without surgery.
The growing role of ctDNA testing may reshape how decisions about lung surgery are made in future. The study authors note that lung-limited metastatic CRC is frequently ctDNA-negative in blood, which may indicate more favorable tumor biology. Using ctDNA status to guide whether a patient needs additional chemotherapy before or after surgery is an emerging approach that future trials should incorporate.
Decisions about lung metastasectomy for CRC should always involve a multidisciplinary team including medical oncologists, thoracic surgeons, and radiologists. Given the complexity of this setting, patients at centers with dedicated expertise in both thoracic surgery and colorectal oncology are likely to receive the most carefully tailored treatment plans. Larger, more definitive trials such as the PUCC-Trial are ongoing and will eventually provide clearer guidance.
TSOG 103 was the first prospective, randomized trial to directly test the role of perioperative chemotherapy and surgical resection within a risk-stratified framework for patients with lung-limited metastatic colorectal cancer. While it was unable to provide definitive conclusions due to insufficient enrollment, its preliminary findings are consistent with the hypothesis that surgery alone may be adequate for well-selected low-risk patients without the added burden of chemotherapy.
The trial underscores a fundamental challenge in surgical oncology: randomizing patients to non-surgical treatment when surgery is perceived as curative is extremely difficult. Future trials in this space will likely need to incorporate ctDNA-guided decision-making, allow for modern local therapies including stereotactic body radiotherapy, and perhaps focus on molecular subtypes of CRC, such as KRAS mutation status and primary tumor sidedness, to enroll more homogeneous patient groups.
The broader PUCC-Trial (DRKS00024727), which requires patients to have three or more lung metastases and allows standard non-surgical local therapies, is attempting to address some of these enrollment challenges. Findings from that and other forthcoming studies will be essential for establishing evidence-based recommendations about the role of pulmonary metastasectomy across the full spectrum of metastatic CRC presentations.