A structural equation modeling framework for estimating symptom burden based on symptom clusters in cancer survivors.

Symptoms after cancer and its treatment challenge the well-being of cancer survivors. Little is known about symptom co-occurrence and the burden experienced by cancer survivors. This study aims to estimate a symptom burden measure based on symptom clusters using self-reported symptoms in cancer survivors. The study utilized data on 40,766 cancer survivors (i.e., breast, prostate, colon, rectum, lung, melanoma, lymphoma, head/neck) diagnosed 2010-2019 from the Danish nationwide SEQUEL cohort. Twenty-two symptoms were assessed 2-12-years after diagnosis using validated self-reported instruments. Symptom severity was dichotomized into none/low versus moderate/severe. Exploratory factor analysis derived a structural equation model (SEM) describing the co-occurrence of symptoms in terms of latent variables (symptom clusters) from a subset of data (training data, n=28,076). Symptom scores for each cluster were estimated from the SEM and used as measures of symptom burden across cancer types in the remaining data (test data=9,366). The SEM included six symptom clusters: a pain-fatigue cluster, peripheral neuropathy and oedema cluster, gastrointestinal and urological symptoms cluster, psychological symptoms cluster, head and neck region symptoms cluster, and respiratory symptoms and infections cluster. Survivors of lymphoma, lung, and head/neck cancer had a higher symptom burden across all six clusters. Our study is the first to propose the use of SEM to quantify symptom burden in cancer survivors and demonstrate robust summaries of symptom burden across six symptom clusters. This approach may facilitate more nuanced comparisons of symptom burden across clinical and demographic groups and support targeted symptom management.