Pain assessment by patients is the rule in clinical trials but may not be in clinical practice. We examined studies comparing assessment of pain by patients and professionals in clinical practice using published studies (1990-2016; ≥20 patients), in English, in an institutional setting, comparing pain assessment within 24 hours by patients and health care professionals. A difference of at least 10% of the maximum score was considered significant. We judged quality on sampling method, blinding, and study size. Eighty studies (20,496 patients) provided data from a range of settings and locations; most (51%) used unbiased sampling, and most (68%) were blind or probably blind. Nine studies with ≥500 patients involved 58% of patients; 60 with <200 patients involved 25%. Large studies were more likely to use comprehensive or random sampling and blinding of patients and professionals. Underestimation of pain by professionals compared with patients was reported by 62/80 studies (78%); there was no difference in 17 (21%) and overestimation in 1 (1%). Underestimation was reported in 75% of large studies (>500 patients), 91% of mid-sized studies (200-400), and 78% of small studies (<200). High-quality studies (blind, comprehensive, or random sampling, >200 patients) consistently reported underestimation (10/11; 91%). The extent of underestimation tended to increase with pain severity. Professionals consistently tend to underestimate pain compared with assessment by patients. This tendency is more pronounced with more severe pain, and the extent of underestimation can be large. It is likely that this contributes to undertreatment of pain.