Draw me a zebra: Study of an item-drawing protocol to assess semantic memory in a memory clinic

BackgroundSemantic memory is often impaired in neurodegenerative diseases such as Alzheimer's disease and frontotemporal lobar degeneration. Evaluating it requires cross-referencing the results of different tests in several modalities. However, most of the available tests only assess verbal production, which may be altered by conditions other than semantic deficits, making interpretation more difficult.ObjectiveTo evaluate the Item Drawing Test (IDT), a non-verbal semantic production task, and examine its reliability and preliminary discriminative ability for detecting semantic impairment in a memory clinic population.MethodsForty-one patients with suspected neurodegenerative disease were assessed using an algorithm combining two semantic memory tests from GREMOTs, the Pyramids and Palm Trees Test, and the IDT, which included 10 drawings for a total score of 49 points. Diagnostic performance was explored using receiver operating characteristic analysis. Inter-rater reliability and associations with global cognitive status (MMSE) were also examined.ResultsThe IDT distinguished 16 patients with semantic impairment (SI) from those without. The optimal threshold score was 42/49 (sensitivity = 88%, specificity = 81.3%; AUC = 0.839; 95% CI = 0.699-0.978). Inter-rater reliability was excellent (intraclass correlation coefficient = 0.972; 95% CI = 0.953-0.984). IDT score correlated moderately with MMSE (r = 0.53, p = 0.0004). In multivariable logistic regression including MMSE, global cognitive status remained significantly associated with SI (p = 0.008). Qualitative analysis of the IDT results provided detailed information about patients' semantic knowledge. Patients with SI produced semantic errors, semantic feature substitutions, and showed visuospatial errors or a tendency toward prototype.ConclusionsThe IDT enables a rapid detection of semantic impairment in patients at risk of neurodegenerative disease. This tool may be useful in clinical practice to limit the influence of verbal production in semantic assessment.