Am Fam Physician. 2022;105(6):590-592
Author disclosure: No relevant financial relationships.
Clinical Question
Evidence-Based Answer
Practice Pointers
Dementia causes difficulties with memory, language, and executive functioning that affect patients' abilities to perform activities of daily living. There are several different types of dementia, but Alzheimer disease is the most common. Dementia is estimated to affect 11% of Americans 65 years and older.5,6 Validated tools are needed to screen for dementia and memory problems to allow for further evaluation and intervention. Four Cochrane reviews examining the use of various outpatient cognitive screening tools were published in 2021.1–4
The first Cochrane review evaluated the two-part Mini-Cog cognitive screening test for dementia.1 The review included four randomized trials and 1,517 patients from the United States and Europe with a baseline prevalence of dementia that varied from 5% to 90%. All studies were performed in a primary care setting and varied widely in methods—which the authors suggest could have overestimated accuracy—and in clinical populations and results. Sensitivity of the Mini-Cog ranged from 76% to 100%, and specificity ranged from 27% to 85%. Only one (n = 383) of the four studies was found to be at low risk of bias, and it demonstrated a sensitivity of 76% and a specificity of 73%. Because the review included only a single high-quality study and demonstrated significant heterogeneity across the four studies, the authors did not recommend the routine use of the Mini-Cog for dementia screening.
A second Cochrane review examined the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), available as a 16- or 26-item test for cognitive impairment.2 This questionnaire asks an acquaintance to judge on a scale of 1 to 5 how the patient's memory has changed over the course of 10 years in a variety of parameters (1 = much improved, 2 = a bit improved, 3 = unchanged, 4 = a bit worse, and 5 = much worse). The Cochrane review identified only one study that took place at a primary care clinic site in Hawaii and included 230 participants of self-identified Japanese-American descent. The prevalence of dementia in this cohort was 16 cases among 230 participants (7%). The study was at high risk of bias in multiple areas, including patient selection, unclear processes for test administration and results reporting, and use of a reference standard for clinical diagnosis of dementia that the reviewers did not consider standard practice. The authors of the study reported a negative predictive value of greater than 90% at all studied thresholds (reported as 3.2 to 3.7 on a scale of 1 to 5) but a positive predictive value of less than 50% for cutoff points below 3.6, indicating limited clinical utility of a positive result and the potential for large numbers of patients who do not have dementia to be referred for specialist evaluation.
A third Cochrane review sought to determine the accuracy of the Montreal Cognitive Assessment (MoCA) for the detection of dementia.3 The authors included seven studies of 9,422 patients from multiple developed countries (the prevalence of dementia ranged from 5% to 54%). A single study in China accounted for 8,411 patients and had a baseline prevalence of dementia of 5%. Significant heterogeneity among the seven studies made direct comparison of the findings difficult. Four studies used the standard threshold for mild cognitive impairment (a score of less than 26), and three studies used lower scores. As noted by the developers of the MoCA, although a score of 18 may be considered the cutoff for Alzheimer disease, a standard cutoff score for dementia has not been established.7 Of the four studies using a score of less than 26 on the MoCA, there was at least a 94% sensitivity in detecting patients with dementia; however, specificity for these diagnoses was poor (60% and lower). None of the studies evaluating MoCA in this review were performed in a primary care setting, which would limit the use of this test in a typical screening population.
A fourth Cochrane review on the Mini-Mental State Examination (MMSE) included 11 studies of 1,569 patients in the United States, Europe, and Japan.4 The authors of the studies administered the MMSE to a group of patients already assessed as having mild cognitive impairment (using a variety of methods) and observed them for the development of dementia (diagnosed by accepted structured clinical criteria). Ultimately, 36.5% of the overall cohort developed dementia. The studies were heterogeneous and used a variety of threshold scores for diagnosis, which prevented the authors from pooling the results in the usual fashion. The sensitivity and specificity of the MMSE ranged from 23% to 76% and 40% to 94%, respectively, for the development of all-cause dementia. For Alzheimer disease, the sensitivity and specificity of the MMSE ranged from 27% to 89% and 33% to 90%, respectively. The single study that examined the prediction of vascular dementia found a sensitivity of 36% and specificity of 80%. No studies were found on the diagnosis of frontotemporal or Lewy body dementia. The low sensitivity for the MMSE makes it a poor choice for primary care–based screening.
A 2020 comparative effectiveness review on the diagnosis and treatment of Alzheimer disease was more favorable toward use of the MoCA.8 This review differed from the Cochrane reviews in that it included only studies of patients in whom cognitive impairment was suspected (the baseline prevalence of Alzheimer dementia varied from 50% to 71%), studied only Alzheimer disease (compared with other forms of dementia), and found fewer studies for each of the tests evaluated. The locations of the included studies were not reported. In addition, the review used the medians of sensitivity and specificity for the individual tests, which may give a false sense of precision about these tests because the range of results was so large.
The evidence from the Cochrane reviews described here supports the conclusion of the 2020 U.S. Preventive Services Task Force statement on screening for cognitive impairment: The evidence is lacking, and the balance of benefits and harms of screening for cognitive impairment cannot be determined.9
The practice recommendations in this activity are available at https://www.cochrane.org/CD011415, https://www.cochrane.org/CD010771, https://www.cochrane.org/CD010775, and https://www.cochrane.org/CD010783.
