Lower modeled blood dose with protons does not translate into less lymphopenia in NSCLC

Across 134 NSCLC patients, proton therapy reduced modeled blood dose by 12% without significantly changing lymphocyte nadir versus photons.

KEY POINTS

  • This retrospective analysis evaluated 134 patients with NSCLC receiving concurrent chemoradiotherapy, comprising a 109-patient model-development cohort and an independent 25-patient external validation cohort. The development cohort included 38 patients treated with passive-scattering proton therapy and 71 receiving photon IMRT, predominantly at a prescription of 74 Gy, while the validation cohort consisted exclusively of photon-treated patients receiving approximately 66–70 Gy. Radiation-induced lymphopenia was assessed using the lowest absolute lymphocyte count (ALC) within 60 days after radiotherapy initiation.
  • Radiation exposure to circulating blood was modeled using HEDOS, a stochastic compartmental blood-flow model combining anatomical dose distributions with estimated blood transit and treatment-delivery times. Two approaches were compared: a basic model incorporating blood exposure within the heart, lungs, and tumor, and an expanded model also incorporating large vessels, liver, and spleen. Proliferating bone marrow was assessed separately using age- and sex-weighted anatomical distributions involving the thoracic spine, sternum, ribs, clavicles, and scapulae.
  • Including major vessels approximately doubled estimated circulating blood dose without improving prediction. Mean modeled blood dose per 2-Gy fraction increased from 0.078 ± 0.034 Gy using the basic model to 0.153 ± 0.059 Gy with expanded anatomical coverage. Nevertheless, incorporating these additional compartments did not improve discrimination for grade 4 lymphopenia. The optimal generalized equivalent uniform dose (gEUD) parameter favored greater weighting of low-dose blood exposure, although discrimination differences across parameter values were small.
  • Blood dose and baseline lymphocyte count independently predicted grade 4 lymphopenia. In the development cohort, 31/109 patients (28.4%) developed grade 4 lymphopenia, defined as ALC <0.2 × 10⁹/L, compared with 5/25 (20.0%) in the validation cohort. Each additional 1 Gy of modeled blood gEUD was associated with greater odds of grade 4 lymphopenia (OR 1.98, 95% CI 1.32–3.07; p=0.001), while higher baseline ALC was protective (OR 0.35, 95% CI 0.14–0.77; p=0.014). Age and sex were not independent predictors.
  • Model discrimination was moderate, with no clear advantage over conventional organ-dose metrics. The combined blood-dose and clinical model achieved an AUC above 0.74 during development and 0.73 (95% CI 0.63–0.83) in the external validation cohort. Validation Brier score was 0.163, compared with 0.170 in development. Models using mean heart dose, mean lung dose, or integral body dose showed broadly similar predictive performance, suggesting that the more complex blood-flow model did not provide a clinically meaningful incremental advantage.
  • Proton therapy reduced modeled blood dose but not the observed severity of lymphopenia. Estimated mean blood dose was 12.1% lower with proton therapy than photon IMRT (p=0.006) using the basic model, with a similar 12.5% reduction (p=0.008) using the expanded model. Despite this dosimetric difference, median lymphocyte nadir was 0.275 × 10⁹/L with protons versus 0.295 × 10⁹/L with photons (p=0.673). These nonrandomized comparisons suggest that reducing modeled circulating blood exposure alone may be insufficient to preserve lymphocyte counts.
  • Bone marrow dose showed weak associations with lymphocyte depletion and none with recovery. Composite proliferating marrow V3–V5 correlated weakly with lymphocyte nadir (r=−0.20; p=0.027 for all three thresholds), whereas mean marrow dose and most thoracic vertebral dose metrics were not significant predictors. Neither marrow V3 nor thoracic vertebral V3 predicted persistent grade ≥2 lymphopenia during follow-up (p=0.628 and p=0.774, respectively). The study used simplified modality-specific beam times, lacked comprehensive lymphoid-organ dose modeling, included only 25 external validation patients, and did not assess survival or immunotherapy outcomes.

CLINICAL TAKEAWAY

Modeled circulating blood dose and baseline lymphocyte count were associated with severe lymphopenia during chemoradiotherapy for NSCLC, but complex blood-dose modeling did not outperform simpler organ-dose metrics. Proton therapy reduced estimated blood exposure without a corresponding improvement in lymphocyte nadir. These findings suggest that successful immune-sparing planning may require consideration of lymphoid tissues and broader treatment-related factors rather than blood dose alone.

SOURCE

Radiotherapy and Oncology

Browse more research Suggest a correction