Poster Presentations P3 P391

2University of South Florida, Tampa, FL, USA. Contact e-mail: arnold.

mitnitski@dal.ca

Background: On average, cognition declines with age, but more rapidly in

some groups than in others. In individuals, cognition can even improve, al-

though this is often viewed simply as diagnostic unreliability. Widely used

models focus on a small number of outcomes, (e.g. worse /not worse, dead

/ alive). By contrast, multistate transition models allow simultaneous consid-

eration of many outcomes - including mortality. Methods: We analyzed two,

four, six and eight-year cognitive changes and mortality with a novel multi-

state model. Its performance is illustrated with standard risk factors in a co-

hort of Japanese Americans living in King County, The Kame Project

(n¼1985). Cognitive states were defined by errors in the Cognitive Abilities

Screening Instrument score. Successive cognitive states - from high cogni-

tion/low errors to impaired cognition/high errors - errors were grouped by

2’s, an analytically detectable interval. Transition probabilities were modeled

using the Poisson model with the Poisson mean and mortality both dependent

on the cognitive state, and covariates: age, sex, and APOE genotype. Re-

sults: After two years, the probability of improvement was 29.3% (95%

CI¼27%-31.6%); the probability of remaining stable was 22.3% (20.1%-

24.4%); that of getting worse was 39.1% (36.6%-41.5%). The probability

of death was 8.8% (7.4%-10.2%). After 8 years, the chance of dying was

24.4% (22.3%-26.5%), whereas the chance of cognitive worsening, stability

and improvement did not change as much: 24.7% (22.5%-26.8%), 17.7%

(15.8%-19.6%), and 33.2% (30.8%-35.5%) respectively. In multivariable

analyses, age, sex and education were independently associated with cogni-

tive changes, improvements were more likely among younger, more educated

people and in women. Conclusions: A multistate transition model can esti-

mate the impact of common covariates on the individual probabilities of cog-

nitive improvement, stability decline and death. Especially with public health

initiatives aimed at preventing dementia, such an approach can reveal novel

patterns of change in response to interventions.

P3-161 CHANGES IN COGNITION IN RELATION TO

FRAILTY IN OLDER CANADIANS

Arnold B. Mitnitski, Nader Fallah, Kenneth Rockwood, Dalhousie Uni-

versity, Halifax, NS, Canada. Contact e-mail: arnold.mitnitski@dal.ca

Background: With age, on average, people show decline in both cognition

and in fitness, appreciated as increasing frailty. While frailty is recognized as

a risk for dementia, it is less clear how the accumulation of general health def-

icits relates to the accumulation of cognitive deficits, although each has been

identified separately as arising from structurally similar stochastic processes.

Methods: Here, we analyzed how five-year changes in cognition (defined as

the errors 0n the Modified Mini-Mental State Examination) are related to

general health status (defined by the Frailty Index) in older Canadians

(n ¼ 8,403). Cognitive change was analyzed using a novel multistate transi-

tion (stochastic) model the output of which was well fit (R-square >0.85) by

a modified Poisson distribution. Results: In multivariable analyses, both age

and frailty were independently associated with cognitive changes and the risk

of death. Risk estimates varied by sex. Women had both a higher chance of

staying alive (29% died, 95%CI¼27%-31%) and of showing improvement

or stabilization (34%, 95%CI¼32%-36%), whereas fewer men survived

(34% died, 95%CI¼32%-36%) and fewer showed stabilization or improve-

ment (30%, 95%CI¼28%-32%). Frail people less often showed cognitive im-

provement or stabilization (20%, 95% CI¼18%-22%) compared to non-frail

people, of whom 39% (95%¼37%-41%) did not deteriorate. Similarly, frail

people were more likely to die (48%, 95%CI¼46%-50%) versus 20%

(95% CI¼18%-22%) of non-frail people. By contrast, there was no difference

in mortality by education level, but a greater chance of cognitive stabilization/

improvement for people with higher education. Among more educated people

35% improved or remained stable (95% CI¼33%-37%) than did less edu-

cated people (29%, 95%CI¼27%-31%). The overall probability of dying

amongst higher educated people was 28% (95%CI¼26%-30%) versus 36%

(95%CI¼34%-38%) in people with lower levels of education. Conclusions:

Cognitive changes generally correlate with general health status. Men and

women each show deleterious effects from frailty, but the profiles of relative

risk expression differ between them.

P3-162 PLASMA F2-ISOPROSTANE LEVEL DOES NOT

PREDICT COGNITIVE FUNCTION IN NON-

DEMENTED ELDERLY: FINDINGS FROM THE

HEALTH ABC STUDY

Alexandra J. Fiocco1, Karla Lindquist1, Alka Kanaya1, Tamara B. Harris2,

Lew Kuller3, Caterina Rosano4, Suzanne Satterfield5,

Eleanor M. Simonsick6, Kristine Yaffe1, 1University of California SanFrancisco, San Francisco, CA, USA; 2National Institute on Aging, Bethesda,

MD, USA; 3University of Pittsburgh, Pittsburgh, PA, USA; 4Graduate

School of Public Health, Pittsburgh, PA, USA; 5University of Tennessee

Health Science Center, Memphis, TN, USA; 6National Institute on Aging,Baltimore, MD, USA. Contact e-mail: alexandra.fiocco@ucsf.edu

Background: Oxidative stress may play an important role in the develop-

ment of Alzheimer’s disease (AD). Recently, F2-isoprostanes (F2-IsoP)

have gained attention as a reliable oxidation biomarker of aging. Several

case-control studies suggest that patients with AD patients exhibit higher

F2-IsoPs compared to healthy controls in cerebral spinal fluid (CSF). How-

ever, there are no epidemiological studies that have assessed the relationship

between plasma F2-IsoP levels and longitudinal cognitive function in non-

demented elders. Methods: We studied a biracial cohort of 726 non-de-

mented elders between the ages of 70 and 79 years at baseline and followed

them for 8 years. Baseline plasma F2-IsoP levels were measured in unthawed

plasma samples by gas chromatography-mass spectrometry. Participants

were grouped according to F2-IsoP tertile. Cognitive function was measured

using the Digit Symbol Substitution test (DSST) at years 1, 5, and 8 and the

Modified Mini-Mental State Exam (3MS) at years 1, 3, and 8. The associa-

tion between F2-IsoP and cognitive function was analyzed using repeated

measures random effects models, adjusting for age, race, gender, education,

BMI and hypertension. Odds ratios were calculated (low F2-IsoP tertile as

reference) to assess the association between F2-IsoP tertile and cognitive im-

pairment (�1 sd decline on the 3MS). Results: There was no significant as-

sociation between F2-IsoP tertile and cognitive function, as measured by the

DSST and 3MS, at baseline or on repeated measure 7-year change (ps>0.05).

Further, no association was found between F2-IsoP tertile and risk of cogni-

tive impairment over years: Mid- F2-IsoP OR¼0.93 (95% CI 0.57, 1.53),

High-F2-IsoP OR¼0.96 (95% CI 0.57, 1.63) with low tertile as reference

group. Conclusions: This was the first epidemiological study to assess the

relationship between plasma F2-IsoP and cognitive decline in a diverse co-

hort of non-demented elders. Our findings do not support the use of F2-

IsoP plasma levels as a biomarker for cognitive aging.

P3-163 PREVALENCE OF COGNITIVE IMPAIRMENT AND

DEMENTIA IN A COHORT OF OLDEST OLD IN

BRAZIL: THE PIETA STUDY

Paulo Caramelli1, Antonio L. Teixeira1, Maira T. Barbosa1,

Ana Paula Santos1, Marcelo Pellizzaro1, Henrique C. Guimaraes1,

Rogerio G. Beato1, B. Machado Joao Carlos1, Hellen Marra1,

Etelvina L. Santos1, Patrıcia P. A. Fialho1, Thais H. Machado1,

Viviane A. Carvalho1, Anne M. Koenig1, Mariana A. Almeida1,

Simone R. Fonseca1, Cerise F. A. Coutinho1, Elisa Franca1,

Natali F. Dezontini1, Clarissa V. Moreira1, Debora P. Maia1,

Mauro Q. Cunningham1, Emılia Sakurai2, 1Faculty of Medicine, Federal

University of Minas Gerais, Belo Horizonte, Brazil; 2School of Mathematics

and Statistics, Federal University of Minas Gerais, Belo Horizonte, Brazil.

Contact e-mail: caramelp@usp.br

Background: Few data are available on the prevalence of cognitive impair-

ment among the oldest-old population of developing countries. The goal of

this study was to determine the prevalence of cognitive impairment no de-

mentia (CIND) and dementia in a cohort of Brazilian oldest-old individuals

and the associated sociodemographic and clinical variables. Methods: The

study was conducted in the rural and urban areas of the city of Caete, Minas

Gerais state, Brazil. The city has 1,155 subjects aged 75 years or more (old-

est-old) according to the 2007 Brazilian census. We aimed to evaluate at least

50% of these individuals. The selected random sample was submitted to

a general personal and health questionnaire, the Mini-Mental State Examina-

tion, the Pfeffer Functional Activities Questionnaire, a brief cognitive