Search tips
Search criteria 


Logo of ageingLink to Publisher's site
Age Ageing. Jul 2011; 40(4): 501–507.
PMCID: PMC3290328
Occurrence of medical co-morbidity in mild cognitive impairment: implications for generalisation of MCI research
Blossom C. M. Stephan,1 Carol Brayne,1 George M. Savva,1 and Fiona E. Matthews2, The Medical Research Council Cognitive Function and Ageing Study3
1Department of Public Health and Primary Care, Institute of Public Health, Cambridge University, Forvie Site, Robinson Way, Cambridge CB2 0SR, UK
2MRC Biostatistics Unit, Institute of Public Health, Cambridge CB2 0SR, UK
3MRC CFAS, Institute of Public Health, University Forvie Site, Robinson Way, Cambridge CB2 0SR, UK
Address correspondence to: B. C. M. Stephan. Tel: (+44) 1223 763837; Fax: (+44) 1223 330330. Email: bcms2/at/
Received October 19, 2010; Accepted March 28, 2011.
Background: diagnosis of mild cognitive impairment (MCI) typically excludes individuals with medical co-morbidity. Interest in MCI screening raises the questions of what are the best criteria to identify a representative sample and what factors are associated with MCI progression to dementia.
Objectives: to compare the pattern of disease co-morbidity across different cognitive groups and to examine the role of health co-morbidity as a risk factor for dementia progression from MCI.
Methods: individuals from the MRC Cognitive Function and Ageing Study were classified as having no cognitive impairment (NCI), MCI, other cognitive impairment no dementia (OCIND) or dementia. At 2 years dementia status was assessed.
Findings: over 50% of individuals in each group reported one or more medical condition. The pattern of disease prevalence was similar in the NCI, MCI and OCIND groups. Anaemia was the only health factor associated with an increased risk of dementia progression from MCI.
Conclusion: classification of MCI using medical exclusions would exclude the majority of the population from a MCI diagnosis. This has implications for treatment decisions and clinical trial recruitment. This could not only make recruitment more difficult but also limit the generalisability of trial results. Medical co-morbidity does not help to distinguish progressive from non-progressive MCI.
Keywords: mild cognitive impairment (MCI), health-related co-morbidity, dementia risk, elderly
Poor health has been associated with an increased risk of cognitive decline and may also be a risk factor for Alzheimer's disease (AD) and vascular dementia. Examining health status in cognitively impaired older adults may be important in identifying the underlying pathogenesis of cognitive decline, and if helpful in predicting dementia risk might be relevant to the development of strategies to delay the onset of dementia. Accounting for health and its risk factors in cognitively impaired older individuals is also important for clinical trial enrolment and decision-making about treatment, particularly in the presence of co-morbidity given the risks associated with polypharmacy.
The term mild cognitive impairment (MCI) is widely used to describe a state of increased dementia risk between ageing without impaired cognitive function and dementia. Although not always explicitly stated a MCI diagnosis is often made following medical exclusion [13]. In screening protocols however, the exclusion of co-morbid conditions may adversely affect selection of the population at risk. Interest in screening for MCI raises the question of what are the best eligibility criteria to identify a representative MCI sample. While clinically significant medical abnormalities can compromise participant safety, expose individuals to undue risk or significantly interfere with trial procedures and outcomes, too strict inclusion criteria can affect sample representativeness (to the population at risk) and trial validity. The implication of medical exclusion on case diagnosis is not known and determining the representativeness of participants selected for trials requires knowledge of disease prevalence in the general population and in individuals with MCI. Furthermore determining those individuals in whom cognitive impairment can be ameliorated through alternative channels such as uncompensated diabetes, hyperthyroidism or hyponatraemia is important for exclusion from clinical trials.
The annual conversion rate from MCI to dementia varies from 5 to 10% according to the definition used. Not all individuals with MCI progress to dementia, with some individuals remaining stable or reverting to normal [4, 5]. Markers are now being sought to distinguish progressive from non-progressive forms of MCI. Longitudinal studies have found that poor health in mid-life can increase risk of cognitive decline and dementia in late life [6, 7]. In some studies, poor health is found to elevate risk of dementia progression from MCI [810]. Associations are not always robust, however, and may depend on the timing of health measurement (mid-life versus later life), study sample (clinic versus population based) and MCI diagnostic criteria (amnestic versus non-amnestic). Whether poor health status can be used to inform the subdivision between progressive and non-progressive MCI may be of value in defining dementia risk in MCI cohorts.
The aim of this study is to compare the pattern of major disease co-morbidity across the spectrum of age-associated cognitive changes including normal, MCI, OCIND and demented groups. We also examined the role of health co-morbidity as a risk factor for 2-year incident dementia across the different cognitive states.
Participants were from MRC CFAS (see [11]. Baseline interviewing began in 1991. Individuals aged 65 years or older were randomly selected from the Family Health Service Authority lists in five areas of the UK including centres in Cambridgeshire, Gwynedd, Newcastle, Nottingham and Oxford. In total, 13,004 participants completed the screening interview undertaken by a trained interviewer at the participants' place of residence. Information on demographic and socio-demographic status, functional ability, self report health status and cognitive performance measured using the Mini Mental State Examination (MMSE) [12], and selected items from the Geriatric Mental State (GMS) Automated Geriatric Examination for Computer-Assisted Taxonomy (AGECAT) [13] were collected. A sub-sample of approximately 20% (= 2,640) of respondents were selected based on age, centre and cognitive ability to complete a more detailed assessment. This included the organicity sections of the AGECAT and the Cambridge Cognitive Examination (CAMCOG) [14]. All individuals were also asked to provide details of an informant, and the History and Aetiology Schedule was completed for as many as possible. The study was approved by the local ethics committee for each centre and informed consent obtained from all participants before interview. Data from the initial prevalence screen, first assessment and 2-year follow-up interviews (Data Version 8.2) were used in this analysis.
Health co-morbidity index
Health status was determined based on self or an informant report of the presence of a condition. From the baseline interview 10 conditions common in the older population were selected including: pernicious anaemia, Parkinson's disease (PD), breathing difficulties, angina, hypertension, diabetes mellitus, peripheral vascular disease (PVD: intermittent claudication), transient ischaemic attack (TIA), self-reported history of stroke or heart attack. Angina and PVD were derived from the Rose Angina Scale [15]. A co-morbidity score (range 0–10) was defined as the number of health conditions present.
Dementia status was derived using the full AGECAT diagnostic algorithm [13], defined as an AGECAT organicity rating case level of 3 or above. This is reported to be equivalent to dementia as diagnosed using the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R) [16, 17]. In total, 8.8% (9.8% age standardised rate) had a case level diagnosis of dementia with a proportion of this group with other psychiatric co-morbidity.
MCI definition
A diagnosis of MCI was made based on revised Mayo Clinic criteria for MCI which combines the A-MCI, non-amnestic MCI (N-MCI) and multiple MCI (M-MCI) subtypes [18, 19]. Full details of the criteria for mapping each subtype in CFAS have been reported previously [20], although in the present study mapping was modified in that no medical exclusions were applied. At first assessment individuals without dementia were classified into three groups including: (i) MCI; (ii) NCI or (iii) OCIND. For classification of MCI individuals had to be non-demented and fulfil the following criteria: (i) subjective/informant complaint of memory loss; (ii) normal general cognitive function (MMSE ≥22); (iii) no severe functional impairment (defined as requiring help at least several times per week with washing, cooking and dressing, or if the individual was housebound) and, (4) objective memory and/or non-memory impairment (based on age-adjusted CAMCOG cut-off values derived from composite memory and non-memory scores). To be classified as NCI, individuals had to fulfil the following criteria: (i) normal general cognitive functioning; (ii) no severe functional impairment; (iii) normal memory and non-memory test performance and (iv) not demented. The OCIND group included all non-normal individuals who failed to fulfil MCI criteria [21]. At first assessment interview 608 individuals fulfilled criteria for NCI, 319 MCI and 580 OCIND.
Summary statistics (means and medians, weighted for study design) were used to compare baseline demographic characteristics across cognitive groups. ANOVA was used to test for group differences in age, cognitive status (MMSE) and educational attainment. The population prevalence (95% confidence interval: 95% CI) of each health condition stratified by cognitive status was calculated. Group differences in disease prevalence were tested using the Chi-squared statistic. Post hoc comparisons of prevalence were tested using logistic regression weighted for study design, with the MCI group as the referent. The pattern of disease and medical co-morbidity was also examined in the A-MCI, N-MCI and M-MCI groups to test whether collapsing across each subtype to form the MCI group was driving the results. Age standardised prevalence estimates were also calculated using 5-year age groups (including 64–69, 70–74, 75–79, 80–84 and 85 + years). This controlled for the baseline differences in the age structure across the different cognitive groups. The association between each health condition and dementia progression was evaluated using univariate and multivariate (age, gender and education) logistic regression models (weighted for attrition) in the NCI, MCI and OCIND groups. Analyses were undertaken using Stata Version 10.0.
Socio-demographic characteristics
Baseline prevalence and socio-demographic characteristics for each cognitive group by disease co-morbidity status are shown in Table 1. Most of the sample (50.1%, back-weighted) was not cognitively impaired. Overall, the dementia group was the oldest in both the disease and no-disease co-morbidity conditions (all P < 0.0001). There were no other group differences in age. In both the disease and no-disease co-morbidity conditions the NCI group had a higher level of education than all other groups.
Table 1.
Table 1.
Baseline demographic characteristics (weighted for study design) of the different cognitive groups stratified by disease status (including those individuals who do not report any of the health conditions and individuals who report one or more of the health (more ...)
Health co-morbidity across cognitive groups
Table 2 shows the prevalence of each health condition in the NCI, MCI, OCIND and dementia groups. Age standardisation did not substantially change prevalence estimates (data not shown) and unstandardised estimates are used for all further analysis. Overall, hypertension was the most common condition, being present in over 20% of individuals in each cognitive group. PD was the least common condition being present in less than 5% of individuals in each cognitive group. There was a high degree of similarity in disease prevalence across the different cognitive groups, except that stroke and PD were significantly more prevalent in the demented group. The MCI, NCI and OCIND groups had a similar pattern of disease prevalence for all conditions. Across the different subtypes of MCI, anaemia (t = 2.91, < 0.005) and TIA (t = 2.65, < 0.005) were more prevalent in the M-MCI compared with the N-MCI group. Otherwise, all other conditions were similarly prevalent across MCI subtypes.
Table 2.
Table 2.
Health profile by cognitive group (weighted for study design)
Multi co-morbidity
Over 50% of individuals in each cognitive group had one or more medical conditions as shown in Table 3. The pattern of co-morbidity was similar across all groups [χ2(18) = 35.3, > 0.05].
Table 3.
Table 3.
Percent (weighted for study design) of medical co-morbidities in each cognitive group
Health and dementia risk associations across cognitive groups
At 2 years, of the 2,640 individuals interviewed at the first assessment, 451 had died (219 of who were classified as demented at the baseline interview). Of the sample of 1,641 individuals seen at the 2-year follow-up, 381 persons were demented. In the MCI group, anaemia was the only significant predictor of 2-year dementia progression (unadjusted OR = 13.5, 95% CI: 2.6–71.3; adjusted for age education and gender OR = 10.6, 95% CI: 2.3–48.7). No health factor increased risk of dementia progression in the NCI or OCIND groups.
The findings confirm previous reports of a high prevalence of disease co-morbidity in the older population [22]. In all groups, fewer than 50% reported no medical conditions with a large number of individuals reporting between one and three conditions. Specific disease prevalence, however, differed between the demented and non-demented groups. Generally there was no association between single-risk factors and incident dementia in MCI. Disease status was predictive of 2-year incident dementia in the MCI group only when co-morbidity was associated with anaemia.
While the pattern of disease prevalence was similar in the NCI, MCI and OCIND groups, individuals with dementia were more likely to have medical co-morbidity related to stroke and PD. This is not unexpected as each of these conditions is known to increase in incidence with age (i.e. the dementia group was older than all other groups), and each is a risk factor for incident dementia [23, 24]. It is possible, given that medical conditions were acquired using self-report, that prevalence estimates may reflect bias in self-reporting or selective survival across groups. However, self-report and objective disease status have been found to be in high agreement for most of the diseases included here [25, 26].
With regard to co-morbidity, the results confirm that medical co-morbidity commonly occurs in both cognitively preserved and cognitively impaired older individuals [27, 28]. As the specific pattern of disease prevalence was similar in all non-demented groups, the results suggest that the selected conditions and their pattern of co-morbidity alone do not strongly influence the impairment captured in the MCI and OCIND states. This is in contrast to findings from clinical and observational studies that suggest that disease co-morbidity associated with, for example, stroke, hypertension, heart attack, diabetes and their co-occurrence, increase risk for major adverse outcomes associated with health and cognition in older adults. The individuals studied here, however, are representative of the general population and are not a selected group of patients. Published research on progression from secondary and tertiary level health-care systems is likely to vary due to differences in participant's profile and reason for referral.
Most published MCI case definitions do not explicitly state medical exclusion criteria although these are routinely employed in the diagnosis of MCI and tend to relate to psychiatric and vascular co-morbidity [3]. In clinical trials, strict eligibility criteria are typically necessary when the isolated effect of an agent is to be tested and for safety reasons. However, the results suggest that strict eligibility criteria could result in insufficient or biased case identification as less than 50% of individuals across any group report no medical co-morbidities. As such clinical trial samples will not reflect the population for whom the treatment is being sought. Trials must be designed with careful consideration and general agreement of inclusion/exclusion criteria for MCI, and should only be restricted to those individuals who are most likely to be affected favourably by the treatment [3].
An important focus of MCI research is the identification of improved methods of classification to differentiate MCI cases at risk of dementia progression from those who are not [29]. Medical co-morbidity may affect dementia progression in MCI cohorts. It has been hypothesised that heterogeneity in the outcome of MCI may be linked to the presence of co-morbid disease [30]. Previous studies have, however, been inconclusive in identifying medical risk factors for incident dementia in MCI groups. Only atrial fibrillation and low folate levels have been associated with an increased risk of dementia progression in MCI [31]. Disparate results are possibly due to differences in MCI case selection criteria, sample age (young-old versus oldest-old), disease severity levels, measurement of risk factors and study populations (clinic samples with ranging emphasis versus population based). In this study, only anaemia increased risk of dementia in individuals with MCI. Anaemia has not, however, been consistently associated with cognitive decline and dementia risk [32, 33]: not all studies find an association, including a previous study using the CFAS cohort. One possible reason for a lack of an association in otherwise non-demented samples may be due to a wash-out effect on the relationship when the different cognitive groups (i.e. normal, MCI and OCIND) are combined for analysis. Given that the prevalence of anaemia was similar across all groups, the results suggest that one mechanism by which anaemia impacts dementia risk in MCI could be due to compromised compensatory processes in the MCI state. Indeed, in MCI the pathology underlying cognitive symptoms (e.g. AD or vascular pathology) may make it difficult for the brain to compensate for the additional insult (i.e. low-oxygen supply) caused by the anaemia. In contrast, in the NCI and OCIND states neuropathology would be unlikely to be as great and, therefore, compensatory mechanisms would be expected to be more efficient. This needs to be tested.
Medical co-morbidity is an important aspect of ageing. There is relatively little difference across the different cognitive groups apart from increased stroke and PD in dementia. Trends of co-morbidity and treatment need to be updated, especially as patterns of disease and medication use are important considerations for clinical trial design with regard to issues of polypharmacy and sample representativeness. Most published MCI case definitions do not explicitly state medical exclusion criteria although these are routinely employed in the diagnosis of MCI and tend to relate to psychiatric and vascular co-morbidity [3]. For clinical trial enrolment although exclusion criteria are necessary to minimise confounding, too strict inclusion may result in a sample not reflective of the population for whom the treatment is being sought. This will compromise the generalisability of results. Trials should be designed for real populations with careful consideration and general agreement of inclusion/exclusion criteria for MCI.
Key points
  • Clinical trials in older aged populations typically exclude individuals with common medical co-morbidities.
  • MCI sample characteristics will change depending on the medical exclusions applied for a case diagnosis.
  • Strict eligibility criteria could result in insufficient or biased MCI case identification, especially in clinical trials, restricting the generalisability of results.
  • Medical co-morbidity does not appear to increase the risk of dementia progression in individuals with MCI.
Author contributions
B.S., F.M. and C.B. developed and designed the project and obtained funding. F.M. and G.S. provided statistical expertise. B.S. completed the literature review. All authors took part in preparation of the manuscript and provided critical intellectual interpretation and manuscript revision. All authors read and approved the final manuscript. The statistical analysis was carried out by B.S., G.S. and F.M.
Conflicts of interest
None declared.
We would like to thank all the respondents, their families and primary care teams, from across the country for their participation in CFAS. CFAS has been funded by the Medical Research Council (MRC), and the Department of Health. The MRC continues to fund CFAS. The study is part of the Cambridge and Peterborough CLAHRC. B.S. is funded by the Joint European Post-Doctoral Programme: The European Research Area in Ageing (ERA-AGE) Network FLARE Programme. F.M. is funded by UK.MRC.U.1052.00.013. The Funding Bodies have no role in the design, methods, subject recruitment, data collections, analysis and preparation of paper.
1. Grundman M, Petersen RC, Ferris SH, et al. Mild cognitive impairment can be distinguished from Alzheimer disease and normal aging for clinical trials. Arch Neurol. 2004;61:59–66. [PubMed]
2. Montgomery SA, Thal LJ, Amrein R. Meta-analysis of double blind randomized controlled clinical trials of acetyl-L-carnitine versus placebo in the treatment of mild cognitive impairment and mild Alzheimer's disease. Int Clin Psychopharmacol. 2003;18:61–71. [PubMed]
3. Visser PJ, Scheltens P, Verhey FRJ. Do MCI criteria in drug trials accurately identify subjects with predementia Alzheimer's disease? J Neurol Neurosurg Psychiatry. 2005;76:1348–54. [PMC free article] [PubMed]
4. Matthews FE, Stephan BCM, McKeith I, Bond J, Brayne C, MRC CFAS. Two-year progression from mild cognitive impairment to dementia: to what extent do different definitions agree? J Am Geriatr Soc. 2008;56:1424–33. [PubMed]
5. Mitchell AJ, Shiri-Feshki M. Rate of progression of mild cognitive impairment to dementia–meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand. 2009;119:252–65. [PubMed]
6. Mitnitski A, Skoog I, Song X, et al. A vascular risk factor index in relation to mortality and incident dementia. Eur J Neurol. 2006;13:514–21. [PubMed]
7. Kivipelto M, Ngandu T, Laatikainen T, Winblad B, Soininen H, Tuomilehto J. Risk score for the prediction of dementia risk in 20 years among middle aged people: a longitudinal, population-based study. Lancet Neurol. 2006;5:735–41. [PubMed]
8. Siuda J, Gorzkowska A, Patalong-Ogiewa M, et al. From mild cognitive impairment to Alzheimer's disease—influence of homocysteine, vitamin B12 and folate on cognition over time: results from one-year follow-up. Neurol Neurochir Pol. 2009;43:321–9. [PubMed]
9. Solfrizzi V, Panza F, Colacicco AM, et al. Vascular risk factors, incidence of MCI, and rates of progression to dementia. Neurology. 2004;63:1882–91. [PubMed]
10. Kivipelto M, Helkala EL, Hanninen T, et al. Midlife vascular risk factors and late-life mild cognitive impairment: a population-based study. Neurology. 2001;56:1683–9. [PubMed]
11. Brayne C, McCracken C, Matthews FE. Cohort profile: the Medical Research Council Cognitive Function and Ageing Study (CFAS) Int J Epidemiol. 2006;35:1140–5. [PubMed]
12. Folstein M, Folstein S, McHugh PR. Mini-Mental State: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–98. [PubMed]
13. Copeland JR, Dewey ME, Griffiths-Jones HM. A computerized psychiatric diagnostic system and case nomenclature for elderly subjects: GMS and AGECAT. Psychol Med. 1986;16:89–99. [PubMed]
14. Roth M, Tym E, Mountjoy CQ, et al. A standarized instrument for the diagnoses of mental disorder in the elderly with special reference to the early detection of dementia. Br J Psychiatry. 1986;149:698–709. [PubMed]
15. Rose GA. The diagnosis of ischaemic heart pain and intermittent claudication in field surveys. Bull World Health Organ. 1962;27:645–58. [PubMed]
16. Copeland JR, Kelleher MJ, Kellett JM, et al. A semi-structured clinical interview for the assessment of diagnosis and mental state in the elderly: the Geriatric Mental State Schedule. I. Development and reliability. Psychol Med. 1976;6:439–49. [PubMed]
17. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R) 3rd edition. Washington, DC: APA; 1987.
18. Petersen RC, Stevens JC, Ganguli M, Tangalos EG, Cummings JL, DeKosky ST. Practice parameter: early detection of dementia: mild cognitive impairment (an evidence-based review). Report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2001;56:1133–42. [PubMed]
19. Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol. 1999;56:303–8. [PubMed]
20. Matthews FE, Stephan BCM, Bond J, McKeith I, Brayne C, MRC CFAS. Operationalisation of mild cognitive impairment: a graphical approach. PLoS Med. 2007;4:e304. doi:10.1371/journal.pmed.0040304. [PMC free article] [PubMed]
21. Stephan BCM, Brayne C, McKeith I, Bond J, Matthews FE, MRC CFAS. Mild cognitive impairment in the older population: who is missed in classifications and does it matter? Int J Geriatric Psychiatry. 2008;23:863–71. [PubMed]
22. Wolff JL, Starfield B, Anderson G. Prevalence, expenditures, and complications of multiple chronic conditions in the elderly. Arch Intern Med. 2002;162:2269–76. [PubMed]
23. Yip AG, Brayne C, Matthews FE. Risk factors for incident dementia in England and Wales: the Medical Research Council Cognitive Function and Ageing Study. A population-based nested case-control study. Age Ageing. 2006;35:154–60. [PubMed]
24. Grodstein F. Cardiovascular risk factors and cognitive function. Alzheimers Dement. 2007;3(Suppl. 1):S16–22. [PubMed]
25. Corser W, Sikorskii A, Olomu A, Stommel M, Proden C, Holmes-Rovner M. Concordance between comorbidity data from patient self-report interviews and medical record documentation. BMC Health Serv Res. 2008;8:85. [PMC free article] [PubMed]
26. Simpson CF, Boyd CM, Carlson MC, Griswold ME, Guralnik JM, Fried LP. Agreement between self-report of disease diagnoses and medical record validation in disabled older women: factors that modify agreement. J Am Geriatr Soc. 2004;52:123–7. [PubMed]
27. Zekry D, Herrmann FR, Grandjean R, et al. Demented versus non-demented very old inpatients: the same comorbidities but poorer functional and nutritional status. Age Ageing. 2008;37:83–9. [PubMed]
28. Lyketsos CG, Toone L, Tschanz J, et al. Population-Based Study of Medical Comorbidity in Early Dementia and “Cognitive Impairment, No Dementia (CIND)”: Association With Functional and Cognitive Impairment: The Cache County Study. Am J Geriatr Psychiatry. 2005;13:656–64. [PubMed]
29. Stephan BC, Kurth T, Matthews FE, Brayne C, Dufouil C. Dementia risk prediction in the population: are screening models accurate? Nat Rev Neurol. 2010;6:318–26. [PubMed]
30. Tervo S, Kivipelto M, Hanninen T, et al. Incidence and risk factors for mild cognitive impairment: a population-based three-year follow-up study of cognitively healthy elderly subjects. Dement Geriatr Cogn Disord. 2004;17:196–203. [PubMed]
31. Ravaglia G, Forti P, Maioli F, et al. Conversion of mild cognitive impairment to dementia: predictive role of mild cognitive impairment subtypes and vascular risk factors. Dement Geriatr Cogn Disord. 2006;21:51–8. [PubMed]
32. Deal JA, Carlson MC, Xue Q, Fried LP, Chaves PHM. Anemia and 9-year domain-specific cognitive decline in community-dwelling older women: the Women's Health and Aging Study II. J Am Geriatr Soc. 2009;57:1604–11. [PMC free article] [PubMed]
33. Peters R, Burch L, Warner J, Beckett N, Poulter R, Bulpitt C. Haemoglobin, anaemia, dementia and cognitive decline in the elderly, a systematic review. BMC Geriatr. 2008;8:18. [PMC free article] [PubMed]
Articles from Age and Ageing are provided here courtesy of
Oxford University Press