Twenty-one studies were eligible for inclusion (n=3,501, range 30 to 1,256): 19 RCTs (n=3,367) and two non-RCTs (n=134). Most studies failed to report their methods in adequate detail. Only eight described their randomisation method (considered adequate), only three reported blinded assessment and most failed to describe dropouts or report analysis by intention to treat. In nine studies, groups differed at baseline.
Patient education versus usual care or placebo/attention control: The intervention significantly improved knowledge and attitudes towards cancer pain and analgesia, the difference being half a point on a 0 to 5 scale (WMD 0.52, 95% CI 0.04 to 1.00, p=0.03; six RCTs). There was significant heterogeneity (I2=91%). Pooling of all nine studies with data suitable for meta-analysis also showed a significant improvement for this outcome (SMD 0.54, 95% CI 0.22 to 0.87). Three studies that reported dichotomous outcomes all reported a statistically significant benefit from the intervention.
Sixteen studies reported pain intensity using a 1 to 10 scale: The intervention significantly reduced pain intensity (WMD -1.10, 95% CI -1.80 to -0.41, p=0.002; eight RCTs). There was significant heterogeneity (I2=88%). The intervention also significantly reduced maximum pain intensity (WMD -0.78, 95% CI -1.21 to -0.35, p=0.0004; eight studies, I2=38%), least pain (WMD -0.98, 95% CI -1.68 to -0.28, p=0.006; two RCTs, I2=34%) and current pain (WMD -0.65, 95% CI -1.21 to -0.09, p=0.02; four RCTs, I2=13%).
Findings for self-efficacy (six studies) were inconsistent. No statistically significant benefit for the intervention was found in two out of three studies that reported medication adherence. No statistically significant benefit for the intervention was found in seven out of eight studies that reported interference by pain with daily activities.
Both single and multiple exposure interventions were beneficial. Effects were more marked when compared with usual care than when compared with a placebo or attention control. Results did not vary substantially by study design.