|| Checking for direct PDF access through Ovid
The distinctive features of individual patients, here termed individual differences, are inescapable aspects of day-to-day patient pain management, but classically designed research studies ignore such differences. This paper introduces statistical pattern visualization methodology for the study of complex individual differences in clinical settings. We demonstrate the application of such methods in patients undergoing bone marrow transplantation (BMT) and suffering severe oral mucositis as a consequence of the aggressive BMT preparative regimen. Oral mucositis produces severe pain and patients often require parenteral opioid medication for several weeks. Unfortunately, the opioid can cause side-effects that limit drug use for pain control. Patients differ in severity and duration of oral mucositis, analgesic response to opioids, and side-effects. We identified and classified individual differences in patterns of drug use, pain control and side-effects in 33 BMT patients who received opioid drug via patient-controlled analgesia (PCA) systems for 7 days or more. These systems allowed bolus dosing and also provided a basic level of analgesic protection through continuous drug infusion. Continuous infusion levels increased or decreased in response to patient bolus self-administration. We employed statistical smoothing (moving average) techniques to remove random variation from the individual data sets and created three-way (trivariate) plots of change over time in drug use, pain and an opioid side-effect (impairment of concentration). The patterns apparent in these plots indicated that 24.2% of patients used PCA optimally (increases in drug use associated with reductions in pain and little or no side-effect), an additional 30.3% manifested a potentially optimal pattern limited by side-effect that worsened with dosing, and 36.4% used PCA suboptimally (modest pain control plus side-effects). In addition, for each subject we created a summary measure for the simultaneous change in three variables: the distance of each day's trivariate score from the origin of a three dimensional plot. This summary measure correlated significantly with the changing severity of patients' oral mucositis over time (r=0.502). This study demonstrates how interactive graphic techniques can provide a basis for examining changes over time among multiple, correlated variables associated with a single individual. It illustrates the application of such techniques and demonstrates that individual subject data sets merit examination in cases where clinical data reflect human performance.