Colorectal cancer is among the most common cancers worldwide. For tumors that have grown into surrounding structures (locally advanced colon cancer, LACC), surgery alone is sometimes not enough. Increasingly, doctors are adding neoadjuvant (pre-surgery) radiation combined with chemotherapy and immunotherapy to shrink tumors before removal, improving the chances of a complete surgical resection.
The problem is that the colon is not a fixed organ. Unlike the prostate or liver, the colon moves and changes shape daily depending on gas, stool, and peristaltic muscle contractions. The position of a colon tumor can shift by more than 10 centimeters between treatment sessions, making it extremely difficult to reliably target with radiation while sparing the surrounding small bowel. Standard radiation planning assumes a relatively fixed anatomy, which may not match reality on any given day.
This study tested a solution: CT-guided online adaptive radiotherapy (ART), in which a new CT scan is taken immediately before each treatment session, the tumor contour is redrawn if it has moved, and a new radiation plan is computed on the spot before delivery. This approach, called the TORCH-C protocol (short-course radiotherapy + CAPOX chemotherapy + PD-1 immunotherapy blockade), was evaluated prospectively in 40 patients to determine whether it is feasible, safe, and effective for routine clinical use.
The study used a CT-linac, a machine that integrates a diagnostic-quality 16-slice helical CT scanner directly with a radiotherapy linear accelerator. Before each of the 5 treatment fractions (25 Gy total prescription), the patient was imaged on the CT-linac. A physician then reviewed whether the tumor was still adequately covered by the radiation plan. If coverage looked satisfactory, treatment proceeded as planned. If the tumor had deformed or shifted so that part of it lay outside the planned radiation field, an adaptive workflow was triggered.
During the adaptive workflow, the physician manually re-contoured the gross tumor volume (GTV) on that day's CT image, with the original planning CT, MRI, and PET-CT displayed for reference. A new planning target volume (PTV) was generated with a 5 to 10 mm safety margin, organs at risk were re-segmented automatically using a deep learning tool, and a new radiation plan was optimized in minutes using the same dose prescription and beam arrangement. A second verification CT was then acquired before beam delivery to confirm alignment.
Geometric accuracy was measured using three metrics: the Dice Similarity Coefficient (DSC), which measures volumetric overlap between the reference and daily contours (a score of 1.0 means perfect overlap); HD95, the 95th percentile Hausdorff distance measuring how far apart the most extreme surface points are; and ASSD, the average surface-to-surface distance. Real-time 3D in-vivo dose verification was performed during beam delivery by comparing measured doses with planned doses using gamma analysis.
Out of 200 total treatment fractions delivered to 40 patients, 48 fractions (24%) required online adaptation. That means about one in four treatment sessions needed a same-day plan revision before the radiation could be safely delivered. Nineteen of the 40 patients (47.5%) required at least one adaptive fraction during their treatment course.
The need for adaptation was highly dependent on which part of the colon the tumor was in. Tumors in the transverse colon required adaptation in 40% of fractions and showed the greatest geometric instability (median DSC = 0.70, HD95 = 37.9 mm, ASSD = 6.5 mm). Tumors in the ascending colon required adaptation in 51% of fractions. In contrast, tumors in the descending colon required adaptation in only 16% of fractions (DSC = 0.83, HD95 = 6.3 mm), and sigmoid colon tumors in only 12% (DSC = 0.85), reflecting greater anatomical stability in these fixed segments.
This segment-dependent instability is explained by anatomy: the transverse and ascending colon are suspended loosely in the peritoneal cavity and move freely with gut contents and peristalsis. The descending and sigmoid colon are more anchored to the posterior abdominal wall by connective tissue, making them inherently more stable targets for radiotherapy. This finding has direct implications for who benefits most from adaptive radiotherapy planning.
The dosimetric impact of adaptation was striking. Without adaptation, the initial reference plan would have delivered full prescription coverage to only 67.91% of the PTV on the day of treatment (after the tumor had moved). With adaptation, this improved to 98.25%, essentially ensuring the entire planned target volume received the full radiation dose. The dose covering 95% of the PTV (D95) increased from 16.96 Gy to 25.22 Gy, nearly reaching the full 25 Gy prescription across the entire target.
Critically, adaptation did not simply redirect dose back onto the tumor at the expense of surrounding healthy tissue. The adaptive plans actually reduced low-to-intermediate dose exposure to the bowel: small bowel V11Gy (the volume receiving at least 11 Gy, linked to gastrointestinal toxicity) decreased significantly, as did exposure of the uninvolved colon at dose levels associated with bowel injury. High-dose metrics for the small bowel were unchanged, confirming that adaptation preserved bowel safety while improving tumor coverage.
In-vivo 3D dose verification confirmed the accuracy of plan delivery. Median gamma passing rates (using a 3%/2mm criterion) were 95.6% for adaptive fractions and 97.9% for standard fractions, well above the clinical acceptability threshold. Only 3 fractions across all 200 treatments fell below 90%, demonstrating that the adaptive workflow achieved its dosimetric goals accurately in practice, not just on paper.
A major concern with adaptive radiotherapy is whether it can be completed quickly enough for routine clinical use. In this study, the median total adaptive session time was 24.6 minutes (range 16.5 to 30.3 minutes). The breakdown of time was: 1.1 minutes for the pre-treatment CT, 6.5 minutes for contour editing, 5.3 minutes for re-optimization, 1.1 minutes for a verification CT, and 3.2 minutes for beam delivery. This compares favorably to MRI-guided ART systems that typically require over 40 to 50 minutes per session.
Treatment was well tolerated. No patients required a treatment interruption due to side effects. Among the 40 patients, 22 (55%) experienced no acute gastrointestinal or urinary toxicity at all. The most common side effects were abdominal pain (20% of patients) and diarrhea (15%). There were 3 Grade 3 adverse events, 1 Grade 2, and 14 Grade 1 events, consistent with what is expected for this type of radiation treatment and suggesting no safety concerns attributable to the adaptive workflow itself.
In two cases, a second verification CT taken shortly after the adaptive plan was approved revealed that the tumor had shifted again during the adaptation window, making the newly approved plan invalid. These fractions were cancelled rather than delivered, illustrating the importance of a final check scan before beam delivery in highly mobile sites, a step notably absent from some MRI-guided ART workflows currently in clinical use.
This study establishes that CT-guided online adaptive radiotherapy is feasible and safe for routine use in locally advanced colon cancer. The finding that adaptation was needed in 24% of all fractions demonstrates that simply repositioning patients and using a fixed plan is not good enough for many colon tumors, particularly those in mobile segments of the proximal colon. A one-time-fits-all radiation plan ignores the daily biological and anatomical reality of the bowel.
The clinical implications are significant for patients treated with the TORCH-C protocol, which combines radiation with chemotherapy and PD-1 immunotherapy blockade. Achieving better tumor coverage by correcting for daily movement could enhance the immunogenic cell death triggered by radiation, potentially improving responses to the accompanying immunotherapy. Reducing unnecessary bowel irradiation could lower treatment toxicity and reduce the risk of delays in systemic therapy or surgery caused by GI side effects.
The study points toward a personalized adaptive radiotherapy strategy: patients with transverse or ascending colon tumors, who showed the highest geometric instability, may benefit most from routine adaptive treatment, while patients with sigmoid tumors may require adaptation less often. Future research should focus on long-term outcomes, including pathological response rates and survival, to confirm whether the observed dosimetric improvements translate into better clinical results.