Additional File 1:

Figure S1. The relevant question from the clinical reporting form Shows the question patients were asked to grade severity of LS from 1 to 4. Figure S2. Maximum spinal cord dose in patients with unilateral and bilateral neck radiation Box and whisker plot showing no difference in Dmax for patients with unilateral and bilateral neck radiation. Figure S3. Dose parameters in patients with no LS symptoms (unshaded, n = 75), and with LS (shaded, n = 42) A – Dose to spinal cord. B – Volume of spinal cord receiving 10, 20, 30, and 40 Gy. C – Percentage of spinal cord receiving 10, 20, 30, and 40 Gy. Figure S4. Dose parameter multi-collinearity plots. A – V20% vs V30%. B – V20% vs V40% Scatter plots showing significant multicollinearity between V20% and V30%, but less collinearity between V20% and V40%. Table S1. Collinearity statistics for models containing V20%, V30%, and V40% Tables showing variance inflation factor and tolerance statistics for logistic regression models containing A V20%, V30%, and V40% (high collinearity); and B V20% and V40% (low collinearity). Table S2. Ordinal logistic regression with highest reported grade of LS as the dependent variable Logistic regression output showing younger age and absence of diabetes are significantly associated with higher grade LS. Table S3. Binary logistic regression with LS vs Non-LS as the dependent variable, and absolute dose volumes (in cc). Logistic regression output showing that using absolute volume or percentage volume makes little difference to the predictive power of the model or the odds ratio for variables in the refined model. Table S4. Ordinal logistic regression with highest reported grade of LS as the dependent variable, and absolute dose volumes (in cc). Logistic regression output showing that using absolute volume or percentage volume makes little difference to the predictive power of the model or the odds ratio for variables in the refined model. (PDF 234 kb)