Demographic Predictors of Symptom Prevalence in Advanced Cancer (702)

Abstract

1.Recognize the influence of patient demographics on symptom profiles in an advanced cancer population.2.Discuss the strengths and weaknesses of using a novel statistical method to evaluate large symptom databases obtained from palliative care patients.3.Review the importance of demographic predictors in conducting epidemiological studies related to cancer. Background. Patient and disease characteristics may affect symptom prevalence. Knowledge of the demographic predictors with the strongest impact on symptom prevalence can facilitate symptom management. Research objectives. We aimed to identify the dominant predictor of symptoms from age, gender, performance status (PS), and primary site in advanced cancer. Method. Recursive Partitioning Analysis (RPA) identified the dominant predictors of 38 symptoms in 948 consecutive patients. RPA split data into two categories. It assessed all possible data splits for the four variables and selected the one that maximized the prevalence difference between the two categories. Category difference p values all had to be significant (p < 0.05) to determine the dominant predictor. Results. Median age was 65 years (range 12–94 years); 55% were male; 65% had ECOG PS 3–4. Most common cancers were lung, genitourinary, and gastrointestinal. Gender was not a dominant predictor for any symptom. Age was the dominant predictor for sleep problems, depression, nausea, anxiety, vomiting, headache, tremors, and blackout. Symptom prevalence declined with age. PS was the dominant predictor for pain, easy fatigue, weakness, anorexia, lack of energy, constipation, early satiety, taste changes, confusion, memory problems, sedation, hiccough, hallucinations, and mucositis. Various cancer primary sites were the dominant predictor for dry mouth, dyspnea, weight loss, cough, edema, hoarseness, dizziness, dyspepsia, dysphagia, belching, bloating, wheezing, itching, and diarrhea. Head/neck and pancreas cancers individually were both dominant predictors for dysphagia and belching, respectively. Only two symptoms (aches/pains, dreams) had no dominant predictor. Conclusion. Thirty-six symptoms had a dominant demographic predictor. Age was the dominant predictor for 8; the influence was unidirectional. Gender did not predict any symptom. PS was the dominant predictor for 14 symptoms; the influence was bidirectional. Head/neck and pancreas cancers had clinically and statistically significant influence over symptom prevalence. Implications for research, policy, or practice. Symptom profiles based on dominant demographic predictors may be present in advanced cancer.