Pretreatment Severe RIL Risk Calculator | Rectal Cancer Chemoradiation

Patient Parameters

Baseline ALC (×10⁹/L)

Pre-treatment absolute lymphocyte count. Cohort range: 0.5–4.5

Age (years)

Age at start of chemoradiation

Sex

Male Female

Add week-2 ALC (on-treatment update, day 12–16)

Week-2 ALC (×10⁹/L)

Absolute lymphocyte count at day 12–16 of chemoradiation

Predicted probability of severe RIL

Enter values above

Baseline model (CV AUC 0.75)

Low (<32%) Intermediate High (>54%)

Model Details

Logistic regression with restricted cubic splines (rms::lrm), internally validated via bootstrap (B=200).

Term	Coefficient
Intercept	3.558
Baseline ALC (linear)	−1.587
Baseline ALC′ (RCS nonlinear)	0.185
Age (per year)	−0.022
Sex (female vs male)	1.236

RCS knots at ALC = 1.20, 1.88, 2.71 ×10⁹/L

Validation Metrics

Corrected C-index: 0.746
Corrected calibration slope: 0.874
Corrected Brier score: 0.208
Development cohort: N = 169, 73 events
Outcome: Severe RIL (ALC nadir < 0.5 ×10⁹/L)
Apparent AUC: 0.765
Cross-validated AUC: 0.747
Internal validation: Bootstrap B=200

Adding a week-2 ALC measurement (day 12–16) significantly improved discrimination (CV AUC 0.83, DeLong p=0.005; N=157). Toggle the week-2 option above to use the dynamic model.

Clinical Context

Developed in curative-intent locally advanced rectal cancer treated with pelvic chemoradiation (5-FU/capecitabine)
Severe RIL prevalence: 43.9% (76/173 evaluable patients)
Observed risk groups: 19% low, 43% intermediate, 68% high
Baseline ALC is the dominant predictor, explaining the steep nonlinear dose-response
Age and sex add modest gain over ALC alone; CRP and dosimetric DVH parameters do not improve out-of-sample performance

Interpretation Guide

Low risk (<32%): Low predicted severe-RIL probability in the manuscript tertile framework
Intermediate risk (32–54%): Consider closer haematological monitoring during chemoradiation
High risk (>54%): High predicted severe-RIL probability; consider proactive supportive planning

Research Tool Only. This calculator implements a prediction model from a single-centre retrospective cohort study and has not been externally validated. It is intended for research purposes and clinical hypothesis generation only. The nomogram predicts toxicity susceptibility, not treatment efficacy — it should not be used to modify chemoradiation intensity. Do not use as a sole basis for clinical decisions. All predictions carry uncertainty; internal validation yielded an optimism-corrected calibration slope of 0.874 and a corrected C-index of 0.746. Prospective external validation is required before clinical deployment.