KEY POINTS
- This retrospective secondary analysis used 2,485 weekly lymphocyte measurements from 498 patients across 73 centers, drawn from a randomized phase III trial of neoadjuvant chemoradiotherapy for locally advanced rectal cancer. All patients received 45 Gy in 25 fractions, combined with either capecitabine 900 mg/m² twice daily on RT days or capecitabine 650 mg/m² twice daily plus irinotecan 60 mg/m² weekly for four weeks, allowing investigators to examine the effects of both radiation and chemotherapy exposure.
- Investigators developed an interpretable exponential mixed-effects model to predict changes in absolute lymphocyte count during treatment. The model incorporated baseline lymphocyte counts, demographics, cumulative chemotherapy exposure, pelvic bone dose metrics and estimates of circulating blood irradiation obtained through a dynamic physiological blood-flow simulation. Relevant structures were delineated using TotalSegmentator, with automated contours reviewed by clinical oncologists.
- Lymphopenia was frequent in this population. Grade 3 lymphopenia occurred in 47.2% of patients and grade 4 in 4.0%, with a median relative reduction in absolute lymphocyte count of 73% from pretreatment baseline to nadir. This substantial treatment-associated immune-cell depletion provided the clinical basis for examining whether individual blood-dose and chemotherapy variables could predict lymphocyte trajectories.
- The resulting prediction model explained approximately 65% of the variability in weekly lymphocyte counts in independently held-out centers (R² 0.65, 95% CI 0.56–0.72), with a mean absolute prediction error of 0.24 × 10⁹/L. The model also achieved a mean concordance index of approximately 0.81 for lymphocyte prediction, indicating that combined treatment-related features explained a substantial, but incomplete, proportion of observed blood-count changes.
- Explainability analysis identified the dose received by 95% of the estimated circulating blood volume (blood D95) as the strongest predictor of lymphocyte depletion. Important additional variables included cumulative exposure to capecitabine and irinotecan, baseline lymphocyte count and pelvic bone dose, particularly V35 of the combined hip, sacral and lumbar bone structures. These findings support the importance of both direct irradiation of circulating lymphocytes and potential damage to marrow-containing tissues, although model associations do not establish independent causal pathways.
- A relative lymphocyte reduction threshold of approximately 75% from baseline to nadir was associated with subsequent progression-free survival. Patients with reductions <75% had three-year PFS of 72% (95% CI 67–76%), compared with 58% (95% CI 49–65%) among those with reductions ≥75% (log-rank p=0.002). The association persisted after multivariable adjustment (HR 1.24, 95% CI 1.08–1.43; p<0.005), although the optimized threshold was derived from the same clinical dataset.
- Crucially, replacing observed lymphocyte reduction with model-predicted lymphocyte reduction eliminated the significant association with PFS (p=0.51). This indicates that, despite reasonable prediction of weekly blood counts, the model is not sufficiently accurate to serve as a prognostic decision tool. Other limitations include retrospective modeling, simulated rather than directly measured circulating blood dose, incomplete biological covariates and the absence of prospective testing showing that reducing blood or marrow irradiation improves outcomes.
CLINICAL TAKEAWAY
Circulating blood exposure and concurrent chemotherapy appear to contribute substantially to lymphocyte depletion during rectal cancer chemoradiotherapy, while severe relative lymphocyte loss is associated with poorer progression-free survival. The results support further investigation of lymphocyte-sparing treatment planning, but the model's inability to predict survival from estimated lymphocyte changes means it is not ready for clinical decision-making or treatment modification.