Calibrating effect-size for studies of pain treatment

Effect-size is a statistic often used in studies of pain treatment. Its attraction is that numerical values can be translated into adjectives, such as small, medium, and large, that describe the change in scores for outcome measures. However, these adjectives can be misused to imply that the effectiveness of the treatment is the equivalent of small, medium, or large. Consideration of examples reveals the fallacy of this practice. Treatments with little to no effectiveness can produce effect-sizes described as medium or large. Reciprocally, treatments with good but imperfect effectiveness generate effect-sizes described as very large or even huge. Effect-size was developed specifically to describe the magnitude of statistical changes, but has little proportional bearing on effectiveness of treatment. When evaluating treatments, readers should not be swayed by descriptors of effect-sizes. Instead, they should consult categorical data on success rates, upon which they can base their decision as to how well the treatment works.