Accuracy of clinical diagnosis of dementia with Lewy bodies: a systematic review and meta-analysis

Background The diagnosis of dementia with Lewy bodies (DLB) is based on diagnostic clinical criteria, which were updated over the years. Objective To evaluate, through a systematic review, accuracy of the diagnostic criteria, testing a possible improvement over time. Methods We searched on MEDLINE and SCOPUS databases for studies reporting diagnostic parameters regarding the clinical diagnosis of DLB until October 2016. We performed meta-analysis, using a Bayesian approach, on those using pathological examination as gold standard, subclassified based on the different diagnostic criteria used. Results We selected 22 studies on 1585 patients. Pooled sensitivity, specificity and accuracy were 60.2%, 93.8%, 79.7%, respectively, for criteria antecedents to McKeith 1996. For McKeith 1996-possible, pooled sensitivity, specificity and accuracy were 65.6%, 80.6%, 77.9% in early stages and 72.3%, 64.3%, 66% in late stages, respectively. For McKeith 1996-probable, pooled sensitivity, specificity and accuracy were 19.4%, 95.1%, 77.7% in early stages and 48.6%, 88%, 79.2% in late stages, respectively. McKeith criteria 2005 were evaluated only in late stages: pooled sensitivity, specificity and accuracy were 91.3%, 66.7% and 81.6%, respectively, for possible diagnosis (only one study) and 88.3%, 80.8%, 90.7% for probable diagnosis, decreasing to 85.6%, 77.1% and 81.7% if only considering clinical settings focused on dementia diagnosis and care. Conclusions and relevance Diagnostic criteria have become more sensitive and less specific over time, without substantial change in the accuracy. Based on current data, about 20% of DLB diagnosis are incorrect. Future studies are needed to evaluate if the recently released revised consensus criteria will improve the diagnostic accuracy of DLB.