Neuropsychological prediction of conversion to dementia from questionable dementia: statistically significant but not yet clinically useful
- 1Department of Care of the Elderly, Beijing University of Chinese Medicine, Dongzhimen Hospital, Beijing, China
- 2Department of Psychology, University of Southampton, Southampton, UK
- 3Department of Care of the Elderly, University of Bristol, Frenchay Hospital, Bristol, UK
- Correspondence to: Professor G Wilcock, Department of Care of the Elderly, Frenchay Hospital, Bristol BS16 1LE, UK;
- Received 13 March 2002
- Revised 21 November 2002
Background: Verbal memory impairment, one of the earliest signs of Alzheimer’s disease (AD), may help identify people with cognitive impairment, insufficient for a diagnosis of dementia (questionable dementia: QD), at risk of developing AD. Other cognitive parameters have been found that may indicate which people with QD will go on to develop dementia. Nevertheless, some researchers have reported only partial success in differentiating between mild AD and age related cognitive impairment.
Objectives: To discover if there are early, pre-clinical cognitive markers that could help identify patients attending our memory clinic who were at risk of developing dementia.
Methods: Multidisciplinary assessment of a consecutive sample of 195 patients with QD seen in a National Health Service hospital outpatient clinic; 135 seen for a mean follow up of 24.5 months.
Results: Conversion rate to dementia was 27.4% (37 of 135). A diagnosis of probable or possible AD was made in 15.6% (21 of 135) of cases. Despite statistically significant differences in some cognitive tasks between those who did and those who did not go on to dement, Cox regression analyses failed to improve prediction rates markedly above base rates and were unstable.
Conclusion: A large number of studies claim good prediction of conversion to dementia using cognitive test scores. Although this study produced similarly good sensitivity and specificity values, proper consideration of the statistical analyses and their clinical significance suggested that these prediction methods are currently too imprecise for clinical use. Use of cognitive indicators combined with neuroradiological, neuropathological, and genetic factors for predicting conversion to dementia might prove more reliable but may be beyond the scope of many geriatric services.
↵* The index quantifies the mean proportion of 12 words learned on trials 2 and 3 as a function of how much information the participant has left to learn after trials 1 and 2. The equation for this index is
Normative data were generated from a sample of healthy controls n = 37 aged 57–69 and n = 46 aged 70+ and are available from the second author.
Competing interests: none declared.
See Editorial Commentary 413