Psychometric Properties and Interpretability of PRO-CTCAE® Average Composite Scores as a Summary Metric of Symptomatic Adverse Event Burden

Figures

• 

  Figure A1. Pairwise Spearman’s Correlations of Composite Scores.

• 

  Figure A2. Residual Correlations from the CFA Model.

• 

  Figure 1. Stacked bar plots illustrating the distribution of composite score categories for selected terms across cohorts. Note. Stacked bar plots illustrate the percentage distribution of composite score categories (0–3) for selected PRO-CTCAE terms across three cancer cohorts: (a) Lung, (b) Breast, and (c) Head and Neck. Higher composite scores reflect greater symptom severity.

• 

  Figure 2. Scree Plots of Principal Components by Cohort: Lung, Breast, and Head and Neck Cancer. Note. Scree plots display the proportion of variance explained by each principal component for each cancer cohort. These plots help identify the number of components that capture the most variation in PRO-CTCAE responses across tumor types, supporting dimensionality reduction and informing subsequent analyses.

• 

  Figure 3. Distribution of Average Composite Score (ACS) Across 3 Disease Sites. Note. Because composite scores range from 0 to 3, the ACS also theoretically ranges from 0 to 3. However, in practice, the maximum score of 3 on ACS is rarely observed, as it is uncommon for individuals to have a score of 3 on composite scores across all AE terms measured. As a result, the ACS often does not reach its theoretical maximum, despite having a defined range of 0 to 3.

• 

  Figure 4. Scatter plots of composite averages vs. the scores derived from principal component analysis (PC1) and confirmatory factor analysis (CFA) across three cohorts. Note. Each cohort—Lung (top row), Breast (middle row), and Head and Neck (bottom row)—includes two plots: PC1 vs. ACS (left) and Factor score vs. ACS (right). High Pearson correlation coefficients (ranging from 0.97 to 0.999) indicate strong linear associations, supporting the validity of ACS as a summary indicator of symptomatic AE burden.

• 

  Figure 5. Symptomatic AE profiles across four latent classes in the lung cohort (N = 183). Note. Four distinct symptomatic AE profiles were observed in the lung cohort. Each latent profile was characterized by a distinct symptom expression pattern. Class-specific mean AE-level composite scores (range 0–3) are depicted on the Y axis. While all classes exhibited a distinct symptomatic AE profile, only latent classes 1 and 2 (high on all symptoms versus low on all symptoms) demonstrated significantly different mean ACS scores. The mean ACS scores for classes 3 and 4 were nearly identical, yet the profiles themselves were qualitatively distinct. Individuals in latent class 3 were experiencing significant appetite loss, fatigue and shortness of breath, whereas those in latent class 4 were experiencing significant constipation, fatigue and pain. The mean (SD) of ACS was 0.6 (0.3) in class 1 (low on all symptoms), 2.1 (0.3) in class 2 (high on all symptoms), 1.4 (0.3) in class 3 (elevated appetite loss, fatigue, and shortness of breath), and 1.5 (0.2) in class 4 (elevated constipation, fatigue, and pain).