- Email: [email protected]
Dropout and its effect on the AIBL study
Poster Presentations: P1
P1-151
CROSS-SECTIONAL CLINICAL, NEUROPSYCHOLOGICAL, NEUROIMAGING, NEUROPHYSIOLOGICAL AND BIOCHEMICAL CHARACTERIZATION OF PEOPLE WITH MILD COGNITIVE IMPAIRMENT IN WP5 PHARMACOG/ E-ADNI STUDY
Claudio Babiloni1, David Bartres-Faz2, Mira Didic3, Gianluigi Forloni4, Jorge Jovicich5, Flavio Nobili6, Pierre Payoux7, Peter Sch€onknecht8, Jens Wiltfang9, Olivier Blin10, Giovanni Frisoni11, 1University of Foggia, Foggia, Italy; 2Universitat de Barcelona and IDIBAPS, Barcelona, Spain; 3 Service de Neurologie et Neuropsychologie, Marseille, France; 4Istituto di Ricerche Farmacologiche, Milano, Italy; 5University of Trento, Trento, Italy; 6Clinical Neurophysiology, Department of Neursciences, Ophthalmology and Genetics, University of Genoa, Genoa, Italy; 7Institut National de la Sante et de la Recherche Medicale, Toulouse, France; 8 University of Leipzig, Leipzig, Germany; 9University of Duisburg-Essen, Essen, Germany; 10Universite Marseille, Marseille, France; 11IRCCS San Giovanni di Dio Fatebenefratelli, Brescia, Italy. Contact e-mail: gfrisoni@ fatebenefratelli.it Background: Workpackage5 of PharmaCog (E-ADNI) is a serial multicenter European study aimed to identify new biomarkers of disease progression in 150 patients with amnestic mild cognitive impairment (aMCI). E-ADNI uses core markers of the North American ADNI, and expands them with specific cognitive, neuroimaging, neurophysiological, and biochemical markers, harmonized with markers that are assessed in animal models in PharmaCog. Methods: We report preliminary cross-sectional data of the first 41 patients enrolled in 7 memory clinics in Italy (Brescia, Genoa), France (Marseille, Toulouse), Spain (Barcelona), and Germany (Essen, Leipzig). Patients underwent clinical and neuropsychological evaluation, high resolution 3T MRI with MPRAGE, T2, FLAIR, resting state, and DTI acquisitions, EEG with resting state and auditory P300 recording, lumbar punctures assessing Abeta42, tau and p-tau, and blood samples with PKC conformation and Abeta 1-42 binding on erythrocytes, amyloid precursor protein C-terminal fragments, plasma and lymphocytes biomarkers, and RNA splicing analyses. Each MPRAGE volume was analyzed in FreeSurfer, focusing on a subset of the automatically segmented regions which are of interest in neurodegenerative diseases. Patients were divided into Abeta positive (CSF-POS) and negative (CSF-NEG) based on CSF Abeta42 levels. Results: Clinical features are in agreement with what is expected for patients with aMCI in the absence of functional disability (mean age 69.0 + 6.5, Mini Mental State Exam 26.2 + 1.9, Functional Assessment Questionnaire 3.2 + 2.9). Concerning MR structural data, the segmentation measures of volumes and thickness are consistent across MRI sites and comparable with the US-ADNI data. CSF-POS MCI patients showed higher prevalence of family history of dementia than CSF-NEG, while neuropsychological, MR and EEG biomarkers showed no significant differences between the CSF-POS and CSF-NEG groups with this preliminary dataset. Conclusions: The patients enrolled in the European-ADNI have clinical and biomarker characteristics compatible with aMCI. Neuropsychological, MR and EEG differences between patients with high and low CSF Abeta42 levels should be explored in much more depth with larger group sizes. P1-152
DROPOUT AND ITS EFFECT ON THE AIBL STUDY
Petra Graham1, Bill Wilson2, Greg Savage3, Kathryn Ellis4, AIBL Research Group5, 1Macquarie University, Sydney, Australia; 2 CSIRO, Sydney, Australia; 3Macquarie University, Sydney, Australia; 4St Georges Hospital, Kew, Australia; 5Mental Health Research Institute, Perth, Australia. Contact e-mail: [email protected] Background: The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of ageing is a prospective study of 1112 individuals (211 with Alzheimer’s disease (AD), 133 with mild cognitive impairment (MCI) and 768 healthy controls (HCs)). These individuals undergo comprehensive cognitive and other health assessments every 18 months with the aim of determining risk of cognitive decline over time. By the third wave of the study 19% of individuals were either lost to follow-up (including refusal to partic-
P205
ipate) or were deceased. Interest in this study is in determining whether this dropout (from death or loss to follow-up) is informative and what the effect is on covariate estimates and estimates of cognitive change. Methods: Using CVLT II long delay free recall (CVLT LDFR) as the outcome, data for the three waves of AIBL were used in a Bayesian hierarchical joint regression model to adjust not only for subject-specific random intercepts and slopes over time but also for potentially informative dropout. This was achieved by jointly modelling change in CVLT LDFR score and time to dropout or death. Covariates included in the model were baseline status (HC, MCI or AD), variables known to be associated with cognitive change: age, sex, APoE 4 status as well as the baseline Ab1-42/Ab1-40 ratio, a biomarker thought to be associated with Alzheimer’s disease. Results: Results suggest evidence of informative dropout with individuals who only appeared in one or two waves having substantially lower baseline CVLT LDFR scores than those who appeared in all three waves. Those defined as MCI and AD at baseline had significantly faster rates of decline compared to baseline HC. Being male, older and being APOE4 positive were also associated with faster rates of decline however there was no evidence of an effect of baseline Ab1-42/Ab1-40 ratio on cognition. Coefficients for the fixed effects did not change a great deal after adjustment for informative dropout. Models ignoring informative dropout tended to underestimate the rate of decline. Conclusions: Whether or not to adjust for informative dropout appears to depend on the inference of interest. Those wishing to describe the rate of decline should adjust for informative missingness appropriately.
P1-153
INCREASING AGE AND CSF BIOMARKER LEVELS IN ALZHEIMER’S DISEASE: THE PLM STUDY
Julien Dumurgier1, Jacques Hugon2, Audrey Gabelle3, Olivier Vercruysse4, Stephanie Bombois5, Jean Louis Laplanche6, Katell Peoch6, Susanna Schraen7, Florence Pasquier8, Jacques Touchon9, Sylvain Lehmann10, Claire Paquet11, 1Lariboisiere Saint-Louis Hospital University Paris Diderot, Paris, France; 2Memory Center Lariboisiere Hospital Paris France, Paris, France; 3CHU Gui de Chauliac Neurology Department, Montpellier, France; 4Center for Memory Resources and Research, EA2691, Lille University Hospital, University of Lille Nor, Lille, France; 5Roger Salengro Hospital, Lille, France; 6Lariboisiere Hospital, Paris, France; 7INSERM, Lille, France; 8CHU Lille, Lille, France; 9CHU Montpellier, Montpellier, France; 10CHU Montpellier, Montpellier, France; 11 CMRR et Unite INSERM U839, Paris, France. Contact e-mail: jacques. [email protected] Background: Little is known about the relationship between increasing age and in-vivo levels of AD neuropathological lesions in brain. Therefore, considering that levels of CSF biomarkers may reflect the level of brain neuropathology, we studied the relationship between age and CSF biomarkers concentrations in a large and multicentric cohort of AD and non AD patients explored for cognitive disorders. Methods: 966 patients (AD, n¼528; non AD, n¼438) were included between January 2008 and December 2010 (mean age, 69.5 years; mean MMSE, 20.2) from three French memory centers. Multivariable l inear regression models were used to study the relationship between CSF biomarkers levels and age in AD and non-AD patients. The capacity of each CSF biomarker in discriminating patients was evaluated using the area under the receiver-operating characteristic (ROC) curves by quartile of distribution of age. Results: In AD patients, older age was associated with higher CSF Ab 1-42 and lower Tau levels. Conversely, in nonAD patients, age was associated with lower CSF Ab 1-42, higher Tau, and higher pTau-181 levels. In sex-stratified analysis, these relationships were significant only in women. Using ROC curve analysis, CSF AD biomarkers were more discriminant in younger patients than in older ones. Conclusions: We reported an association between age and the levels of CSF biomarkers in patients explored for cognitive disorders. The direction of the relationship is reversed depending whether the patient has been diagnosed with AD or not. Younger AD patients exhibited more severe CSF abnormalities compared to older AD patients and CSF biomarkers were more able to differentiate between AD and non-AD patients in younger subjects than in older ones.
P1-151
CROSS-SECTIONAL CLINICAL, NEUROPSYCHOLOGICAL, NEUROIMAGING, NEUROPHYSIOLOGICAL AND BIOCHEMICAL CHARACTERIZATION OF PEOPLE WITH MILD COGNITIVE IMPAIRMENT IN WP5 PHARMACOG/ E-ADNI STUDY
Claudio Babiloni1, David Bartres-Faz2, Mira Didic3, Gianluigi Forloni4, Jorge Jovicich5, Flavio Nobili6, Pierre Payoux7, Peter Sch€onknecht8, Jens Wiltfang9, Olivier Blin10, Giovanni Frisoni11, 1University of Foggia, Foggia, Italy; 2Universitat de Barcelona and IDIBAPS, Barcelona, Spain; 3 Service de Neurologie et Neuropsychologie, Marseille, France; 4Istituto di Ricerche Farmacologiche, Milano, Italy; 5University of Trento, Trento, Italy; 6Clinical Neurophysiology, Department of Neursciences, Ophthalmology and Genetics, University of Genoa, Genoa, Italy; 7Institut National de la Sante et de la Recherche Medicale, Toulouse, France; 8 University of Leipzig, Leipzig, Germany; 9University of Duisburg-Essen, Essen, Germany; 10Universite Marseille, Marseille, France; 11IRCCS San Giovanni di Dio Fatebenefratelli, Brescia, Italy. Contact e-mail: gfrisoni@ fatebenefratelli.it Background: Workpackage5 of PharmaCog (E-ADNI) is a serial multicenter European study aimed to identify new biomarkers of disease progression in 150 patients with amnestic mild cognitive impairment (aMCI). E-ADNI uses core markers of the North American ADNI, and expands them with specific cognitive, neuroimaging, neurophysiological, and biochemical markers, harmonized with markers that are assessed in animal models in PharmaCog. Methods: We report preliminary cross-sectional data of the first 41 patients enrolled in 7 memory clinics in Italy (Brescia, Genoa), France (Marseille, Toulouse), Spain (Barcelona), and Germany (Essen, Leipzig). Patients underwent clinical and neuropsychological evaluation, high resolution 3T MRI with MPRAGE, T2, FLAIR, resting state, and DTI acquisitions, EEG with resting state and auditory P300 recording, lumbar punctures assessing Abeta42, tau and p-tau, and blood samples with PKC conformation and Abeta 1-42 binding on erythrocytes, amyloid precursor protein C-terminal fragments, plasma and lymphocytes biomarkers, and RNA splicing analyses. Each MPRAGE volume was analyzed in FreeSurfer, focusing on a subset of the automatically segmented regions which are of interest in neurodegenerative diseases. Patients were divided into Abeta positive (CSF-POS) and negative (CSF-NEG) based on CSF Abeta42 levels. Results: Clinical features are in agreement with what is expected for patients with aMCI in the absence of functional disability (mean age 69.0 + 6.5, Mini Mental State Exam 26.2 + 1.9, Functional Assessment Questionnaire 3.2 + 2.9). Concerning MR structural data, the segmentation measures of volumes and thickness are consistent across MRI sites and comparable with the US-ADNI data. CSF-POS MCI patients showed higher prevalence of family history of dementia than CSF-NEG, while neuropsychological, MR and EEG biomarkers showed no significant differences between the CSF-POS and CSF-NEG groups with this preliminary dataset. Conclusions: The patients enrolled in the European-ADNI have clinical and biomarker characteristics compatible with aMCI. Neuropsychological, MR and EEG differences between patients with high and low CSF Abeta42 levels should be explored in much more depth with larger group sizes. P1-152
DROPOUT AND ITS EFFECT ON THE AIBL STUDY
Petra Graham1, Bill Wilson2, Greg Savage3, Kathryn Ellis4, AIBL Research Group5, 1Macquarie University, Sydney, Australia; 2 CSIRO, Sydney, Australia; 3Macquarie University, Sydney, Australia; 4St Georges Hospital, Kew, Australia; 5Mental Health Research Institute, Perth, Australia. Contact e-mail: [email protected] Background: The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of ageing is a prospective study of 1112 individuals (211 with Alzheimer’s disease (AD), 133 with mild cognitive impairment (MCI) and 768 healthy controls (HCs)). These individuals undergo comprehensive cognitive and other health assessments every 18 months with the aim of determining risk of cognitive decline over time. By the third wave of the study 19% of individuals were either lost to follow-up (including refusal to partic-
P205
ipate) or were deceased. Interest in this study is in determining whether this dropout (from death or loss to follow-up) is informative and what the effect is on covariate estimates and estimates of cognitive change. Methods: Using CVLT II long delay free recall (CVLT LDFR) as the outcome, data for the three waves of AIBL were used in a Bayesian hierarchical joint regression model to adjust not only for subject-specific random intercepts and slopes over time but also for potentially informative dropout. This was achieved by jointly modelling change in CVLT LDFR score and time to dropout or death. Covariates included in the model were baseline status (HC, MCI or AD), variables known to be associated with cognitive change: age, sex, APoE 4 status as well as the baseline Ab1-42/Ab1-40 ratio, a biomarker thought to be associated with Alzheimer’s disease. Results: Results suggest evidence of informative dropout with individuals who only appeared in one or two waves having substantially lower baseline CVLT LDFR scores than those who appeared in all three waves. Those defined as MCI and AD at baseline had significantly faster rates of decline compared to baseline HC. Being male, older and being APOE4 positive were also associated with faster rates of decline however there was no evidence of an effect of baseline Ab1-42/Ab1-40 ratio on cognition. Coefficients for the fixed effects did not change a great deal after adjustment for informative dropout. Models ignoring informative dropout tended to underestimate the rate of decline. Conclusions: Whether or not to adjust for informative dropout appears to depend on the inference of interest. Those wishing to describe the rate of decline should adjust for informative missingness appropriately.
P1-153
INCREASING AGE AND CSF BIOMARKER LEVELS IN ALZHEIMER’S DISEASE: THE PLM STUDY
Julien Dumurgier1, Jacques Hugon2, Audrey Gabelle3, Olivier Vercruysse4, Stephanie Bombois5, Jean Louis Laplanche6, Katell Peoch6, Susanna Schraen7, Florence Pasquier8, Jacques Touchon9, Sylvain Lehmann10, Claire Paquet11, 1Lariboisiere Saint-Louis Hospital University Paris Diderot, Paris, France; 2Memory Center Lariboisiere Hospital Paris France, Paris, France; 3CHU Gui de Chauliac Neurology Department, Montpellier, France; 4Center for Memory Resources and Research, EA2691, Lille University Hospital, University of Lille Nor, Lille, France; 5Roger Salengro Hospital, Lille, France; 6Lariboisiere Hospital, Paris, France; 7INSERM, Lille, France; 8CHU Lille, Lille, France; 9CHU Montpellier, Montpellier, France; 10CHU Montpellier, Montpellier, France; 11 CMRR et Unite INSERM U839, Paris, France. Contact e-mail: jacques. [email protected] Background: Little is known about the relationship between increasing age and in-vivo levels of AD neuropathological lesions in brain. Therefore, considering that levels of CSF biomarkers may reflect the level of brain neuropathology, we studied the relationship between age and CSF biomarkers concentrations in a large and multicentric cohort of AD and non AD patients explored for cognitive disorders. Methods: 966 patients (AD, n¼528; non AD, n¼438) were included between January 2008 and December 2010 (mean age, 69.5 years; mean MMSE, 20.2) from three French memory centers. Multivariable l inear regression models were used to study the relationship between CSF biomarkers levels and age in AD and non-AD patients. The capacity of each CSF biomarker in discriminating patients was evaluated using the area under the receiver-operating characteristic (ROC) curves by quartile of distribution of age. Results: In AD patients, older age was associated with higher CSF Ab 1-42 and lower Tau levels. Conversely, in nonAD patients, age was associated with lower CSF Ab 1-42, higher Tau, and higher pTau-181 levels. In sex-stratified analysis, these relationships were significant only in women. Using ROC curve analysis, CSF AD biomarkers were more discriminant in younger patients than in older ones. Conclusions: We reported an association between age and the levels of CSF biomarkers in patients explored for cognitive disorders. The direction of the relationship is reversed depending whether the patient has been diagnosed with AD or not. Younger AD patients exhibited more severe CSF abnormalities compared to older AD patients and CSF biomarkers were more able to differentiate between AD and non-AD patients in younger subjects than in older ones.