www.thelancet.com Vol 380 October 20, 2012 1441
Lancet 2012; 380: 1441–43
Department of Public Health
and Primary Care, University of
Cambridge, Cambridge, UK
(C Brayne MD, D Davis MRCP)
Prof Carol Brayne, Institute of
Public Health, University of
Cambridge, Robinson Way,
Cambridge CB2 0SR, UK
Making Alzheimer’s and dementia research fi t for populations
Carol Brayne, Daniel Davis
Two decades ago, the established view was that ageing
and Alzheimer’s disease arise from distinct patho-
physiological processes. Tremendous scientifi c eff orts
were underway to separate and characterise their
biological bases. At the time, refl ective contributions
were made by the corresponding author that raised
concerns about this approach, arguing from an epi-
demiological perspective that what cognitive decline
actually constituted in the unselected ageing population
should be carefully examined.1–3 The view was expressed
that investigations into the biological underpinnings of
ageing and dementia were being biased by assumptions
about causality, through re ifi cation of particular diag-
nostic entities without the necessary empirical evidence
from relevant populations.
We propose that the accepted framework has generally
inhibited the discipline from making advances that
might benefi t the many people with usual dementia. The
little progress that has been made in dementia prevention
calls for refl ection on successes and failures of the
substantial investment over the past decades by govern-
ments, the commercial sector, and charities.4 The poor
understanding of ageing and prostate cancer, and the
application of an inadequate screening test that has led,
many believe, to over treatment, harm, and expense, has
been called the “great prostate mistake”.5 Is it too late to
avoid the “great Alzheimer’s mistake”?
A need for standardised defi nitions of dementia led to
the development of diagnostic criteria based on clinical
measures of symptoms and observations. However, even
in fairly homogeneous cultural settings, these criteria
had the potential to be interpreted diff erently by indi-
vidual clinicians and researchers. The diagnosis of
dementia has remained an entity with indistinct bound-
aries. Although few disagreed about what constituted
moderate and severe dementia, studies including milder
cognitive defi cits led to much less consistent estimates of
dementia prevalence.6 To some extent, the variability in
reported prevalence has been mitigated by use of more
uniform approaches to operationalised defi nitions, but
huge variation remains.7,8 Establishment of defi nitions
for this intermediate cognitive state is currently of great
interest but attempts have led to some degree of
unnecessary duplication of eff ort. Contemporary def-
initions look remarkably similar to those proposed more
than 20 years ago for minimal or questionable dementia.9
Because of the absence of clear-cut identifi ers to predict
dementia (or its subtypes) for any given indi vidual,
attention has turned to the value of the addition of
biomarkers (eg, amyloid β on imaging, or peptide ratios
in cerebrospinal fl uid) to neuropsychological domains or
even the consideration of such markers as diagnostic
entities in their own right.10
Major diffi culties exist for the validation of proposed
biomarkers in relation to dementia as an outcome in
relevant populations and in appropriate timeframes. The
sheer volume of evidence from huge population-based
studies and randomised trials that contributed so much to
our understanding of blood pressure and cholesterol—
and their association with future cardio vascular disease—
does not exist for dementia of any type, particularly in the
age groups at greatest risk. This evidence took many
decades to establish, and inves tigators of enormous
longitudinal studies and meta-analyses are still looking at
emerging risk factors.11 We do not have anything on a
similar scale for the specifi c risk factors of interest or
biopathological profi les in dementia. Understanding of
the meaning of these biomarkers and their natural history
at particular ages, in particular cohorts, and over time is
crucial before they can be applied widely,12 with the
recognition that this approach has the potential to identify
most of the middle-aged population as having a disorder.13
Furthermore, the eff ect of age on total dementia risk
has so far over shadowed any other associations under
investigation,14,15 a fact often downplayed in discussions of
the potential eff ect of modifi able risk.
Even though large studies are called for, funds to
establish truly representative populations are diffi cult
to raise. Imaging is expensive, as is blood taking and
appropriate storage. Response rates are decreasing, and
attrition and dropout are substantial in older populations,
with little ability to address potential bias. However, if
vast amounts of money are to be spent on long-term,
regular investigation, detection, and treatment (along
with any adverse eff ects) of preclinical syndromes from
middle age, this is the type of evidence that would be
needed by any body with responsibility for recom-
mendations and fi nancing on health-care spending
(such as the National Institute for Health and Clinical
Excellence in the UK). There is, as yet, no systematic or
coordinated eff ort to support researchers exploiting the
opportunities from existing cohort studies.
Much of the research and rhetoric ignores the diffi culty
inherent in confl ating the dementia syndrome with the
notion of Alzheimer’s disease.4 In the ageing brain,
pathologies associated with frank dementia are often seen
in people who die without dementia. As already noted,
most prostate cancer is in situ, and will not aff ect the
lifespan or quality of life in the older male population. The
understanding of why some men’s prostate cancers become
aggressive is incomplete, so screening interventions that
merely detect the mass of prostate tissue (prostate-specifi c
antigen testing) lead to more harm through unnecessary
biopsies and consequent interventions than reduced
cancer-specifi c mortality. Similarly, why specifi c neuro-
pathologies in some older people are associated with
www.thelancet.com Vol 380 October 20, 2012
expression of dementia whereas in others these same
pathologies are not, remains unexplained.16 Without better
understanding of these factors, pursuit of the defi nition of
pre-dementia according to biological measures will lead us
down the same route as prostate cancer screening. What
older people want to know is whether they will get
dementia and how to plan for it—not whether they will fi t
some biopathological profi le.
Despite the media hype, the promise of many risk
factors and therapies has yielded little in terms of tangible
outcomes for dementia. The best present intervention
for Alzheimer’s disease is still based on symptomatic
improvement. Why so little progress? Perhaps there has
been too much focus on fi ndings that might be the
product of a selection bias for people with dementia who
have been admitted into various clinical services. There
is the real danger of residual confounding or inadequate
attention to the fact that apparent protective factors
might only be markers for some other less measurable,
more fundamental, mechanism for risk or protection.
This mistake has been made for other disorders: most
apparently, hormone replacement therapy seemed to
have a protective eff ect against heart disease in obser-
vational studies, but in trials was shown to cause harm.17
The same has also happened for Alzheimer’s disease—
eg, in trials of anti-infl ammatory drugs, and vitamin-
based and endocrine-based inter ventions.
Selection bias is a major limitation for the gen-
eralisability of research fi ndings. Much observational
epidemiology has been done on volunteer samples.
Selective participation and attrition are mostly unad-
dressed,18 and these have the potential to systematically
aff ect the direction of fi ndings. Recruitment into
treatment trials from memory clinics tends not to be
representative of those with most dementia and
comorbidity,19 and so interpretation should be informed
by detailed scrutiny of how generalisable fi ndings from
such a setting would be to the entirety of people with
dementia. By only studying groups in specialised clinical
settings in which a suffi cient proportion will develop
frank dementia over a relatively short time,20 there is
substantial risk that the same measures applied in the
population (ie, unselected settings) will have lower
positive predictive values, leading to more false positives.
This eff ect has been shown for mild cognitive
impairment in many studies, with the Medical Research
Council Cognitive Function and Ageing Study providing
the most systematic assessment of the performance of
diff erent defi nitions.21
Population prevalence of the actual ages of people with
dementia shows the largest proportion is aged 80 years and
older, and this proportion will further increase substantially
over coming decades (appendix p 1). People with dementia
at every stage and age will have greatly diff erent needs,
including pathways through dependence to death. Clinical
and epidemiological research should refl ect these
variations if it is to have appropriate generalisability.
Most dementia research is not done in those aged
85 years and older. Appendix p 2 shows schematically
why—as in the Buddhist parable of blind men gaining
vastly diff erent impressions of an elephant from
examining only one part each—diff erent sectors in the
research community have such varied perceptions of
dementia, and why research fi ndings seem to produce
confl icting results. Each clinical setting and specialist
group dealing with people with dementia has diff erent
fi lters from the whole population, but each group tends to
generalise its experience to the entirety. Thus we have the
call for screening from old-age psychiatrists in memory
clinics, but those in primary and secondary care who deal
with the bulk of the older population know that this
approach would overwhelm our systems. The provenance
of the research should be more clearly framed in the
context of its meaning for the whole dementia population,
with limitations for generalisation clearly articulated.
An important but under-researched component of
dementia epidemiology is the recognition of terminal
decline. The prevalence of dementia and cognitive
decline is common at varying intervals before death.22
Attention has been focused so strongly on prevention,
detection, and cure that terminal decline, through which
so many of us will pass, has been relatively neglected.
In the appendix we attempt to explain why, if we take
dementia as one entity, to make sense of the apparently
divergent and diverse fi ndings about dementia from
diff erent research disciplines is not possible. However, if
a longitudinal and population perspective is taken, we
can see the many diff erent eff ects on who is seen by
whom and ends up where, and who is being represented
by the research from these varied settings. Of crucial
importance is that approaches and fi ndings are anchored
to the reality of dementia in the true population, if we are
not to continue to drain public and commercial resources
on the basis of overextended claims.
One approach would be to review the existing and
emerging evidence to assess its place and likely relevance
for the whole population with dementia syndrome, and
over what timescale. This approach would emphasise
the exact nature of any investment needed to take
fi ndings forward and provide some estimate of the likely
eff ects. We advocate an approach to dementia that is
driven by true-population awareness, not merely
restating the size of the problem, and for researchers to
argue the true place of their research with respect to
real populations. Additionally, we suggest that research
mapping in a systematic way for primary, secondary, and
tertiary prevention of dementia and then simulating
population eff ects will help to inform the scope and
direction of future research.
CB developed the ideas and prepared the fi rst draft. DD and CB worked
together to produce the fi nal draft.
Confl icts of interest
We declare that we have no confl icts of interest.
See Online for appendix
Viewpoint Download full-text
www.thelancet.com Vol 380 October 20, 2012 1443
CB thanks all her colleagues over the years, and those participants,
families, and the wider community who have contributed so much to
our population studies. DD is supported by the Wellcome Trust as a
Research Training Fellow.
1 Brayne C, Calloway P. Is Alzheimer’s disease distinct from normal
ageing? Lancet 1988; 332: 514–15.
2 Brayne C. Clinicopathological studies of the dementias from an
epidemiological viewpoint. Br J Psychiatry 1993; 162: 439–46.
3 Brayne C. Research and Alzheimer’s disease: an epidemiological
perspective. Psychol Med 1993; 23: 287–96.
4 Richards M, Brayne C. What do we mean by Alzheimer’s disease?
BMJ 2010; 341: c4670.
5 Ablin RA. The great prostate mistake. The New York Times
(New York), March 10, 2010: A27.
6 Matthews FE, Stephan BC, Bond J, McKeith I, Brayne C.
Operationalization of mild cognitive impairment: a graphical
approach. PLoS Med 2007; 4: 1615–19.
7 Ferri CP, Prince M, Brayne C, et al, for Alzheimer’s Disease
International. Global prevalence of dementia: a Delphi consensus
study. Lancet 2005; 366: 2112–17.
8 Erkinjuntti T, Ostbye T, Steenhuis R, Hachinski V. The eff ect of
diff erent diagnostic criteria on the prevalence of dementia.
N Engl J Med 1997; 337: 1667–74.
9 Roth M, Tym E, Mountjoy CQ, et al. CAMDEX. A standardised
instrument for the diagnosis of mental disorder in the elderly with
special reference to the early detection of dementia. Br J Psychiatry
1986; 149: 698–709.
10 Dubois B, Feldman HH, Jacova C, et al. Research criteria for the
diagnosis of Alzheimer’s disease: revising the NINCDS-ADRDA
criteria. Lancet Neurol 2007; 6: 734–46.
11 Danesh J, Erqou S, Walker M, et al. The Emerging Risk Factors
Collaboration: analysis of individual data on lipid, infl ammatory
and other markers in over 1.1 million participants in 104 prospective
studies of cardiovascular diseases. Eur J Epidemiol 2007; 22: 839–69.
12 Zimmern RL. Testing challenges: evaluation of novel diagnostics
and molecular biomarkers. Clin Med 2009; 9: 68–73.
13 Moynihan R, Doust J, Henry D. Preventing overdiagnosis: how to
stop harming the healthy. BMJ 2012; 344: e3502.
14 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.
15 Daviglus ML, Plassman BL, Pirzada A, et al. Risk factors and
preventive interventions for Alzheimer disease: state of the science.
Arch Neurol 2011; 68: 1185–90.
16 Bennett DA, Schneider JA, Arvanitakis Z, et al. Neuropathology
of older persons without cognitive impairment from two
community-based studies. Neurology 2006; 66: 1837–44.
Rossouw JE, Anderson GL, Prentice RL, et al. Risks and benefi ts
of estrogen plus progestin in healthy postmenopausal women:
principal results from the Women’s Health Initiative randomized
controlled trial. JAMA 2002; 288: 321–33.
18 Bhamra S, Tinker A, Mein G, Ashcroft R, Askham J. The retention
of older people in longitudinal studies: a review of the literature.
Qual Ageing 2008; 9: 27–35.
19 Schoenmaker N, Van Gool WA. The age gap between patients in
clinical studies and in the general population: a pitfall for dementia
research. Lancet Neurol 2004; 3: 627–30.
20 Petersen RC, Morris JC. Mild cognitive impairment as a clinical
entity and treatment target. Arch Neurol 2005; 62: 1160–63.
21 Matthews FE, Stephan BC, McKeith IG, Bond J, Brayne C. Two-year
progression from mild cognitive impairment to dementia: to what
extent do diff erent defi nitions agree? J Am Geriatr Soc 2008;
22 Muniz-Terrera G, Matthews FE, Stephan B, Brayne C. Are terminal
decline and its potential indicators detectable in population studies
of the oldest old? Int J Geriatr Psychiatry 2011; 26: 584–92.