CBCT-based dose accumulation improves toxicity prediction in hyperfractionated small-cell lung cancer

In 142 patients with limited-stage SCLC, CBCT-reconstructed dose improved pneumonitis and esophagitis prediction compared with planned dose, but discrimination remained modest.

KEY POINTS

  • This retrospective modeling study used data from 142 patients with limited-stage small-cell lung cancer enrolled in a randomized trial comparing twice-daily hyperfractionated thoracic radiotherapy schedules of 45 Gy in 30 fractions (1.5 Gy twice daily) and 54 Gy in 30 fractions (1.8 Gy twice daily). Patients received concurrent platinum–etoposide chemotherapy, and the analysis evaluated local progression-free survival, grade ≥2 radiation pneumonitis, and grade ≥3 radiation esophagitis.
  • To estimate delivered rather than planned dose, the investigators reconstructed treatment anatomy using CBCT and deformable image registration. Because daily imaging was not available for every fraction, one CBCT represented approximately five consecutive fractions, with the original plan recalculated on registered CT anatomy. The resulting accumulated dose was compared with conventional planning-CT dosimetry, incorporating biologically effective dose and incomplete repair between twice-daily fractions.
  • For grade ≥2 radiation pneumonitis, the CBCT-based Lyman–Kutcher–Burman NTCP model achieved an AUC of 0.67 versus 0.59 with planning-CT dose. Negative log-likelihood improved from 66.13 to 62.39. Although this represents better discrimination, the absolute performance remained modest, with overlapping uncertainty intervals (95% CI 0.49–0.81 versus 0.40–0.72).
  • For grade ≥3 radiation esophagitis, CBCT-based modeling similarly increased AUC from 0.58 to 0.62, while negative log-likelihood decreased from 56.64 to 55.55. Confidence intervals remained wide (0.43–0.79 versus 0.35–0.72), indicating that the improvement was small and does not establish clinically reliable individual risk prediction.
  • Conventional lung DVH metrics calculated on planning CT did not significantly distinguish patients who subsequently developed pneumonitis: V20 p=0.18, mean lung dose p=0.27, and biologically effective dose p=0.21. After CBCT-based dose reconstruction, all three metrics differed significantly between patients with and without grade ≥2 pneumonitis (all p<0.05). A similar pattern was observed for esophageal V35, mean dose, and biologically effective dose, suggesting that anatomical changes during treatment contained additional toxicity-related information.
  • The benefit for tumor-control prediction was substantially smaller. CBCT-based tumor control models achieved AUC values of 0.80 versus 0.79 for the 45-Gy schedule and 0.77 versus 0.75 for the 54-Gy schedule. In the underlying trial cohort, the 54-Gy regimen showed consistently higher local progression-free survival, including approximately 59.9% versus 32.7% at 72 months, while pneumonitis and esophagitis rates did not differ significantly between dose groups (p=0.52 and p=0.34, respectively).
  • The proposed framework combined a linear–quadratic model incorporating repair, redistribution, reoxygenation, repopulation, and Gompertzian tumor growth with a BED-based NTCP model. However, dose accumulation relied on sampled rather than daily anatomy, deformable registration introduced uncertainty, and the models were fitted and evaluated using the same retrospective dataset. Model-derived equivalent dose thresholds across fractionation schedules should not be interpreted as validated clinical OAR constraints.

CLINICAL TAKEAWAY

CBCT-based accumulated dose may provide a more informative basis for pneumonitis and esophagitis modeling than planned dose alone in twice-daily hyperfractionated SCLC radiotherapy. However, discrimination remains insufficient for routine individual risk prediction, and the small improvements in TCP modeling and lack of independent validation mean that this framework should currently be regarded as a research tool rather than a clinically actionable decision model.

SOURCE

International Journal of Radiation Oncology, Biology, Physics

Browse more research Suggest a correction