Which domains should be included in a cancer pain classification system? Analyses of longitudinal data.

The overall aim of the present study was to further develop an evidence-based platform for the content of an international cancer pain classification system. Data from a multicentre, observational longitudinal study of cancer patients were analysed. Analyses were carried out in 2 samples: (A) Cross-sectional data of patients on opioids at inclusion, and (B) patients just admitted to palliative care. Outcome measures in the models we investigated were pain on average, worst pain, and pain relief at inclusion, and at day 14, respectively. Uni- and multivariate regression models were applied to test the explicative power on pain outcomes of a series of known pain domains, including incident pain, psychological distress, neuropathic pain, pain localisation, sleep disturbances, total morphine equivalent daily dose (MEDD), and cancer diagnosis. In the 2 analyses, 1529 (A) and 352 (B) patients were included, respectively. Incident pain, pain localisation, MEDD, use of nonsteroidal antiinflammatory drugs, and sleep were associated with one or more of the pain outcomes in analysis A, while initial pain intensity, initial pain relief, incident pain, localisation of pain, cancer diagnosis, and age were predictors in the longitudinal analysis. Identified domains explained 16% to 24% of the variability of the pain outcome. Initial pain intensity emerged as the strongest predictor of pain outcome after 2 weeks, and incident pain was confirmed to be a relevant domain. The regression models explained only a minor part of the variability of pain outcomes.