Optimizing the use of patient data to improve outcomes for patients: narcotics for chronic noncancer pain.

Randomized trials can provide important direction to clinical decision-making; however, their strength of inferences may be weakened by methodological limitations, the extent that their reported outcomes fail to address patient-important end points and by failing to report results that provide interpretable estimates of magnitude of effect. Strategies that investigators can use to address interpretability include reporting mean differences between groups in relation to the minimal important difference and reporting the proportion of patients who benefit from treatment and the associated number needed to treat. These strategies also apply to reporting pooled estimates from meta-analyses, even when studies use different instruments to measure the same construct. We illustrate these techniques using, as an example, current evidence for the use of opioids in chronic noncancer pain.