Numerical Cognition in Neurodegenerative Disorders: A Systematic Review of Patterns of Preservation and Impairment

Numerical cognition comprises multiple processes supporting the representation, comprehension, production, and manipulation of numerical information. Although numerical impairments have been described in neurodegenerative diseases, findings remain fragmented across disorders and numerical domains. This systematic review aimed to characterize patterns of preserved and impaired core numerical abilities across neurodegenerative disorders and to examine their implications for cognitive models and clinical assessment.

PubMed, Web of Science, and PsycINFO were systematically searched for original empirical studies in English or French examining core numerical cognition in adults with neurodegenerative disorders or mild cognitive impairment. Reference lists were additionally screened. Methodological quality was appraised using design-specific Joanna Briggs Institute critical appraisal tools. Forty-seven studies published between 1989 and 2026 met the eligibility criteria. Findings were synthesized qualitatively because of substantial heterogeneity in study designs, clinical populations, and numerical measures.

Numerical profiles varied considerably across disorders, individuals, and disease stages, but several broad patterns emerged. Basic number comprehension and magnitude processing were often relatively preserved, whereas greater vulnerability was observed for processes requiring transformation, retrieval, or manipulation of numerical information. Arithmetic fact knowledge showed variable preservation, including dissociations between stored knowledge and its retrieval. Calculation procedures and complex calculation were among the most consistently affected abilities. Evidence was strongest for Alzheimer's disease and semantic variant primary progressive aphasia and remained limited for several other neurodegenerative disorders.

Numerical cognition is differentially affected across neurodegenerative diseases, supporting its multicomponent organization. These findings suggest the potential value of multidomain assessment rather than reliance on isolated calculation tasks. However, methodological heterogeneity, variation in methodological strength across studies, and the limited evidence available for several disorders reduce confidence in the generalizability of these patterns and currently preclude defining reliable disease-specific numerical signatures.

OSF, 10.17605/OSF.IO/DU2Y9.

None.