BackgroundThere are no reports in the literature about patients with painfully restricted temporomandibular joints concerning the reliability of unidimensional and multidimensional pain scales on the basis of a generalizability and decision study. Generalizability and decision studies are designed to aid in reasoned decision-making and therefore are complementary to classic psychometric analyses, in which correlation coefficients express the reliability of a measurement design. The smallest detectable difference as an outcome of the decision study is the smallest statistically significant change that can be detected.PatientsTwenty-five consecutive patients (5 males and 20 females) with painfully restricted mandibular function participated in the current study.DesignRepeated measurements of three levels of unidimensional experienced pain (average, minimal, and maximal pain before the week of assessment; n = 200) and the multidimensional pain rating index of the MPQ (n = 100) were used. The measurement sessions were supervised by two observers on two separate measurement days, 1 week apart, with each measurement repeated twice.ResultsInconsistency in measurement results analyzed in terms of absolute error variance (i.e., the measurement facets plus all the interactions) represented at least 24% of total variance. The smallest detectable difference of visual analog scales varied from 43 to 15 mm, depending on the amount of repetitions, whereas for the pain rating index it varied from 22.7 to 14.4 units.ConclusionsFor statistically and clinically successful treatment of patients with painfully restricted temporomandibular joints, clinicians must overcome at least the smallest detectable difference and 38% of the initial average pain level.