Diagnosing clinical dementia is based on an assessment of different variables, such as the patient's medical history, known risk factors, and biochemical features. Partial least squares discriminant analysis was used to evaluate variables of importance for diagnosing dementia in a clinical dementia population. Polymorphism for genotypes of glutathione S-transferase (GST) and sulfotransferase 1A1, hypothetically of importance in dementia disorders, was also included in the analysis. The study population consisted of 73 patients with Alzheimer's disease (AD), 14 with mixed dementia, 75 patients with vascular dementia, and 28 control cases. We found that several of the variables, such as the presence of ApoE4 allele, high cerebrospinal fluid levels of total tau protein, low levels of beta-amyloid((1-42)), and a low score on the Mini-Mental State Examination, facilitated a discrimination between the diagnoses compared with the controls. The different diagnoses overlapped. There were indications that genotypes of GSTs contributed to a subgrouping within AD.