PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Ann Epidemiol. Author manuscript; available in PMC 2013 December 1.
Published in final edited form as:
PMCID: PMC3510749
NIHMSID: NIHMS418139

Relative Contributions of Biomarkers in Alzheimer’s Disease

Thaddeus J. Haight1 and William J. Jagust1,2,3, for the Alzheimer’s Disease Neuroimaging Initiativea

Abstract

Purpose

To assess relationships between biomarkers for Alzheimer’s Disease (AD) and their potential contributions to AD.

Methods

Biomarkers and cognitive evaluations were assessed longitudinally for 179 patients with mild cognitive impairment (MCI), from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) from 2003–2006, and were used to examine, at any given time, the joint contributions of hippocampal volume, whole brain volume, and brain glucose metabolism on clinical AD progression, using the Alzheimer’s Disease Assessment Scale-Cognitive subscale (ADAS-Cog). Marginal structural models (MSMs) were applied, and inverse-probability of treatment weight (IPTW) estimation was utilized to account for time-dependent confounding between study variables.

Results

At any given time, population-level differences (e.g. 1-standard deviation (SD) increase) in brain glucose metabolism (−1.036 95% CI: −1.608, −0.464) and hippocampal volume (−1.537 95% CI: −2.399, −0.674) independently reduced mean ADAS-Cog, whereas a 1-SD increase in whole brain volume did not (0.372 95% CI: −0.283, 1.027). Effects of brain glucose metabolism differed in subgroups defined by baseline covariates (e.g., age), but no subgroup effects were observed for hippocampal volume and brain volume.

Conclusions

Brain glucose metabolism and hippocampal volume represent relevant biological markers in subjects at risk for AD.

Medical Subject Headings (MeSH) biological markers, causality, dementia, longitudinal studies

INTRODUCTION

Alzheimer’s Disease (AD) represents a major public health concern given aging populations in the U.S. and worldwide (1, 2). Efforts to better understand the disease have included epidemiologic studies, some of which have utilized different biomarkers for AD (37). Developments in neuroimaging have led to the availability of additional markers, representative of anatomical, metabolic, and biochemical aspects of the disease (810). These different biomarkers could be used to clarify further the causal pathways that contribute to AD.

Relationships between various biomarkers, which measure structural and functional brain changes, and their relationship to AD are not well understood. Individual biomarkers have been studied extensively with respect to the disease (e.g., brain atrophy). Other studies have investigated multiple biomarkers and their interrelationships with regard to AD (1114). However, few studies, if any, have examined the joint relationships of multiple biomarkers and their respective contributions to AD over time.

Assessment of the individual effects of biomarkers that account for other biomarkers (e.g., contribution of whole brain volume independent of hippocampal volume) and other factors would improve understanding of the disease process. However, standard statistical methods may result in biased estimates of such effects given time-dependent confounders—i.e., variables that need to be adjusted but which occur on the causal pathway for the effect of interest (Figure 1). Causal inference methods have been developed and applied previously to evaluate effects given time-dependent confounders (1517).

FIGURE 1
Brain volume and hippocampal volume can directly affect ADAS-Cog as well as through reduction in brain glucose metabolism, as measured by FDG-PET. These measures are affected also by reduction in glucose metabolism which is the result of reduced neuronal/synaptic ...

Amnestic mild cognitive impairment (MCI) represents an intermediate clinical stage in AD (18). Subjects with MCI are classified as having memory impairment without additional cognitive and functional impairments that characterize individuals with AD (18). While MCI subjects are at increased risk of AD (i.e., 12% annual rate vs. 1–2% in non-MCI subjects), not all transition to AD (9). Levels of different biomarkers in MCI vary considerably, and a joint examination of these biomarkers would allow for investigation of their relative effects on changes in disease status in MCI.

We applied causal methods to examine the relative contributions of different biomarkers on AD progression in subjects with MCI. Moreover, we examined the contributions of these biomarkers in different subpopulations at variable risk for AD (e.g., age groups). A broader understanding of the relationships between individual biomarkers and AD could be informative with regard to AD progression and potential interventions to the disease.

METHODS

Subjects

The Alzheimer’s Disease Neuroimaging Initiative (ADNI) is a longitudinal study to develop and examine biomarkers of AD progression (9). Subjects were recruited from multiple clinical sites in the U.S. and Canada. Subjects were between ages 55 and 90, and were free of other significant neurologic diseases. Further details about inclusion criteria and other information are available (19, 20).

At baseline, all subjects received a clinical evaluation and a brain MRI scan. Subjects diagnosed with MCI (n=398) had evaluations repeated over follow-up at 6, 12, 18, 24, and 36 month intervals. In addition to clinical and MRI evaluations, half of the subjects received [18F]fluro-deoxyglucose position emission tomography (FDG-PET) scans, to assess brain glucose metabolism, at baseline and follow-up. Cerebrospinal fluid (CSF) samples were obtained for half of the subjects at baseline. Therefore, the number of subjects available for analysis differed depending on the set of biomarkers examined.

MCI subjects with baseline MRI and FDG-PET data (n=179) were selected for the study. Available follow-up data varied across subjects. In the 179 subjects, 739 observations from different subject-visit times were available for analysis (i.e., complete FDG-PET and MRI data). For 30 subjects with missing data at one of the time points, FDG-PET and MRI measures were imputed by taking the average of the extant values of these measures at the adjoining time points to the records with missing data. Similarly, values were imputed for one subject with missing values at two consecutive time points. Records where FDG-PET and MRI data were missing at a patient’s final follow-up visit were excluded. A subset of subjects with CSF samples (89 subjects, 389 total observations) were available for a separate analysis, which examined CSF beta-amyloid (CSF-Aβ) as an additional biomarker of AD.

Study Measures

Study outcome

The Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) is a measure of cognitive dysfunction in AD. It has primary functions similar to the Mini-Mental State Examination (MMSE) for evaluation of language, memory, and a variety of other cognitive functions (21). A 70-point version of the test was used—i.e., 11 item test that excluded number cancellation and delayed word recall(21). Total score ranges are 0 to 70, with a higher score indicative of poorer cognitive function.

Magnetic resonance imaging (MRI)

Structural magnetic resonance scans (1.5-T) were acquired at multiple ADNI sites based on a standardized protocol(22). Bilateral hippocampal volume and whole brain volume measures were obtained by scan processing using FreeSurfer software(23). Left and right hippocampal volume were averaged to create a summary measure for the analysis. Brain volume encompasses gray and white matter volume.

Brain glucose metabolism

[18F]Fluro-deoxyglucose positron emission tomography (FDG-PET) represents a quantitative in-vivo measure of brain glucose metabolism. Data were acquired at sites nationwide using a standardized protocol (24). Details of image processing of the PET scans are provided elsewhere (25, 26).

A composite measure of FDG-PET was used, based on an average of five brain regions that were found to differentiate AD patients and healthy controls in a meta-analysis (25).

Cerebrospinal fluid beta-amyloid(1-42) (CSF-Aβ)

Details of CSF collection and processing are given elsewhere (27). CSF-Aβ was used as a surrogate measure of brain amyloid pathology. Studies comparing CSF-Aβ with autopsy-based as well in-vivo based assessments of brain amyloid pathology, based on brain imaging, are available elsewhere (12, 28).

Other covariates

Different variables independently associated with ADAS-Cog were considered in the analysis as potential confounders and/or effect modifiers. Baseline variables(units) included: age (years); sex; height (cm); education (years); apolipoprotein e4 allele (ApoE4) status (categorized as 0,1, 2 alleles); modified Hachinski score; cardiovascular disease (yes/no (y/n)); diabetes (y/n); smoking (y/n); and hypertension (y/n). Time-dependent variables included: white matter hyperintensities (% total brain volume); systolic and diastolic blood pressure (mmHg); blood glucose (mg/dL); plasma homocysteine (μmol/L); geriatric depression score (range: 0–15); and neuropsychiatric inventory questionaire score (range: 0–5). In addition, FDG-PET, based on frontal-lateral brain regions, not included in the composite measure above, and previous ADAS-Cog, from t-1, were included as potential confounders.

Statistical Analysis

Descriptive analysis

Variables were plotted to evaluate their change over time. The relationship of brain volume and hippocampal volume was examined in plots, and Spearman correlations were generated to assess the respective variability of each variable and its potential contribution to the analysis. Associations of the different covariates described above (“Other covariates”) and ADAS-Cog were examined to assess these covariates as potential confounders.

Causal inference analysis

The analysis sought to examine the relative short-term effects of FDG-PET, hippocampal volume, and brain volume on expected ADAS-Cog, at given time t, as represented in Figure 1.

Based on Figure 1, changes in hippocampal volume and brain volume are assumed to produce immediate short-term effects on ADAS-Cog, both directly and indirectly (through FDG-PET) at any given time. Prior glucose metabolism (i.e., neuronal dysfunction and synaptic dysfunction reflected in reduced glucose utilization, measured by FDG-PET at t-1) is assumed to affect brain and hippocampal volume, which in turn are assumed to affect subsequent FDG-PET at t (e.g., brain atrophy), based on previous biological data and the most current model of AD biomarker change (10). The effect of FDG-PET on ADAS-Cog is assumed to be an average of FDG-PET effects at t and t-1 representative of short and longer-tem metabolic patterns on changes in ADAS-Cog. Other covariates described above and measured at t (not included in graph) are assumed to affect volumetric and FDG-PET variables at t, and could be affected potentially by prior levels of FDG-PET and volumetric measures at t-1 (e.g., depression score and various metabolic changes (e.g., blood glucose) that could occur indirectly from reduced function).

Marginal structural models (MSMs) were used to evaluate: 1) overall marginal effects of FDG-PET, hippocampal volume and brain volume on ADAS-Cog; and 2) independent (i.e., direct) effects of these biomarkers not mediated by other biomarkers in the causal pathway (e.g., effect of FDG-PET independent of volumetric effects). Additionally, MSMs were used to examine modification of these effects by baseline age, ApoE4 status, and CSF-Aβ.

Briefly, causal inference methodology can provide unbiased estimates of MSMs in the presence of time-dependent confounders as depicted in Figure 1—i.e., confounders of an exposure of interest that are affected by previous levels of that exposure. By comparison, standard analytical methods are likely to produce biased estimates of effects under these conditions(29). To identify and estimate effects from MSMs requires a set of assumptions and estimation procedures. Details are provided in the Appendix. Models were fit with linear regression using generalized estimating equations (GEE) that included weights that were derived for each subject (15). Given that short-term effects of different biomarkers were of interest, the available data for given subjects were pooled (e.g., up to 6 time points per subject) to improve precision. Analyses were implemented with standard software (SAS version 9.1.3 and R version 2.4.1).

RESULTS

Characteristics of the study group are presented in Table 1. Of the group, 65.9% were men. Subjects’ mean baseline age was 75, and ages ranged from 56 to 90 years. Subjects were well-educated: 95.0% and 65.4% reported completion of high school and college, respectively. Of the 179 subjects, 66 (36.9%) were diagnosed with AD within 3 years. Mean follow-up was 1.6 years, and 50% of the group completed at least 2 years of the study. Over half of the subjects were ApoE4 carriers. Mean ADAS-Cog at any given time was 11.7, and was greater (i.e., worse) in subjects who converted to AD. 89 of the 179 subjects had recorded baseline CSF-Aβ measures (mean (range)): 159.8 (48.3–281.8)), which differed significantly between subjects who did and did not convert to AD.

Table 1
Distributions of Baseline Characteristics in Study Participants

Individual and joint marginal effects of each of the biomarkers assessed in the study on ADAS-Cog are presented in Table 2. The individual effect of each biomarker (Table 2, left) represents the change in mean ADAS-Cog at any time t if everyone in the study population experienced an increase in one of the respective biomarkers (e.g., 1 standard deviation (SD) increase based on distribution of a particular biomarker in the study population). These effects indicate that a short-term increase in any one of the three biomarkers individually (FDG-PET, hippocampal volume, brain volume) would contribute to a significant reduction (i.e., improvement) in mean ADAS-cog. For example, an increase the equivalent of 1 SD (~520 mm3) in hippocampal volume in all subjects would result in a 15% reduction in mean ADAS-Cog from 11.8 to 10.0.

Table 2
Overall and Joint Marginal Effects of FDG-PET, Hippocampal Volume and Brain Volume on Mean ADAS-Cog

Joint models examined the relative, independent effects of the different biomarkers –i.e., relative to one another---on ADAS-Cog (Table 2, right). For any particular model, the effects represent the independent contributions of the correspondent biomarkers with respect to change in mean ADAS-Cog at any time t, if, contrary to fact, the levels of the other biomarkers in the model were fixed for everyone in the population. For example, based on the model of FDG-PET and hippocampal volume, the effect of FDG-PET (-1.193 95% CI: −1.795, −0.591) represents the reduction in mean ADAS-Cog for a 1 SD population-level increase in FDG-PET if, contrary to fact, everyone had the same hippocampal volume. An independent effect of hippocampal volume (−1.375 95% CI: −2.206, −0.544) was observed in the same model. In a separate model that examined all three biomarkers simultaneously (including brain volume), increases in FDG-PET (−1.036 95% CI: −1.608, −0.464) and hippocampal volume (−1.537 95% CI: −2.399, −0.674) reduced ADAS-Cog, but the same relative increase (i.e., 1 SD) in brain volume did not (0.372, 95% CI: −0.283, 1.027). These results indicate that higher FDG-PET and hippocampal volume confer benefits independently with respect to lower ADAS-Cog, at a population-level, whereas greater brain volume, separately from FDG-PET and/or hippocampal volume, does not.

Other models evaluated effects of the various biomarkers for subpopulations defined by differences in baseline age, Apoe4 status, and CSF-Aβ levels (Tables 35). Mean ADAS-Cog did not differ by these groups (Tables 35, left column). By contrast, the effects of FDG-PET on ADAS-Cog differed significantly for the different subpopulations based on the covariates considered (Tables 35, middle column, top). For example, if everyone in the population experienced higher FDG-PET (i.e. 1 SD), the effect was lower mean ADAS-Cog in older vs. younger subjects (for each year: −0.076 95% CI −0.152, 0.000); in subjects with one additional ApoE4 allele vs. subjects with one less E4 allele (−1.159 95% CI: −2.163, −0.156); and subjects with higher CSF-Aβ vs. subjects with lower CSF-Aβ (per 1 pg/mL: −0.014 95% CI: −0.026, −0.002).

Table 3
Overall and Joint Marginal Effects of FDG-PET, Hippocampal Volume and Brain Volume on Mean ADAS-Cog Examined With Age
Table 5
Overall and Joint Marginal Effects of FDG-PET, Hippocampal Volume and Brain Volume on Mean ADAS-Cog Examined with CSF-Aβ

Of the three biomarkers, only subgroup effects of FDG-PET were observed; there were no observed significant differences in effects of hippocampal volume or brain volume (Tables 35, middle column) in the same subpopulations examined with FDG-PET.

Estimates from joint models that included interactions of FDG-PET and the baseline covariates (Tables 35, right column) were attenuated slightly compared to the estimates in the univariable models with FDG-PET alone (Tables 35, middle column, top). Interactions between the volumetric biomarkers and the baseline covariates were not significant and are not presented.

DISCUSSION

We examined the causal relations of three biomarkers, representative of different pathways that could contribute to AD progression in subjects with MCI. FDG-PET, a marker of brain glucose utilization, and hippocampal volume were independently related to ADAS-Cog, which was used as a marker of AD progression. Brain volume was not independently related to ADAS-Cog after accounting for the other two markers. Moreover, FDG-PET was the only biomarker for which subgroup effects, based on age, apoE4 status, and CSF-Aβ, were observed.

Investigators have underscored the importance of examining multiple biomarkers for better assessment of the relationships of these biomarkers with AD (1214). The current study builds on previous work by employing methods which allowed for modeling the joint dependencies between different biomarkers over time and quantifying the marginal effects of these biomarkers to better clarify their individual relationships to disease progression.

The study is compelling for a number of reasons. First, FDG-PET appears to represent a pathway exclusive of hippocampal atrophy, which is strongly associated with cognitive decline in AD (3032). This result suggests that MCI subjects with reduced hippocampal volume, but with greater glucose metabolism, may experience less (or delayed) cognitive decline compared to if the same subjects experienced less glucose metabolism. An independent effect of FDG-PET could have implications for AD progression at a population-level given the wide variability of its distribution in MCI. Moreover, FDG-PET, in addition to MRI, could be utilized in studies for investigation of the relationships between these markers and other factors (e.g., diabetes, physical activity), for improved specification of pathways to AD.

Subgroup effects for FDG-PET were observed in groups at risk for AD. These results suggest that greater metabolic activity may be protective in such groups (i.e., older age, presence of ApoE4 allele), though the effects were relatively small. Effects of FDG-PET were observed in subjects with higher CSF-Aβ, who are at less risk of AD than those with lower CSF-Aβ. It is possible that CSF-Aβ is a more specific marker of AD than age or ApoE4, such that any protective effects of metabolic activity are outweighed by greater brain beta-amyloid, which lower CSF-Aβ represents.

Greater brain volume did not improve ADAS-Cog independently of the other biomarkers considered. Although brain volume represents a global measure of both disease progression (i.e., atrophy) and reserve, it may not be as specific to AD as the other biomarkers. Others have shown overall brain atrophy to be predictive of AD onset in MCI subjects after accounting for hippocampal volume (33), however, this may be related this study’s outcome measure (AD onset) compared to ADAS-Cog, which measures cognitive function. Also, it is possible that our methodology, which accounted for dependence between study variables over time, may explain in part the reduced effects of brain volume, as well the independent effects observed for the other biomarkers.

One of the strengths of the study includes its use of causal inference methods to summarize and relate effects of biomarkers assessed over time on ADAS-Cog. It is possible, however, that the MSM models used to examine these effects were not entirely accurate nor that our estimation of these MSMs effectively controlled for, or included, all potential confounders of the effects examined (See Appendix). We applied model selection tools that have been shown to provide improved model fits, over typical ad-hoc assumptions, to better specify models used in MSM estimation (34, 35). We assumed a temporal order between biomarkers (See Figure 1), which may have not been fully realized in the actual data—i.e., volume measures were obtained simultaneously, based on the same MRI scan, and in some instances were measured subsequently to FDG-PET. However, the assumption should be satisfied based on the biologic plausibility of the relationships between the variables (i.e., reduced volume affects FDG-PET, brain volume encompasses hippocampal volume). Still, it is possible that effect estimates may be biased due to violations of this assumption or others related to model misspecification or unmeasured confounding.

MCI subjects represent an increasingly important population for study given their transitory status with respect to further cognitive decline and AD, and their potential responsiveness to intervention (9). It is unclear whether the findings of the study are applicable to a wider population of MCI subjects at risk of AD. Subjects with stroke and other underlying neurologic abnormalities were excluded from ADNI. The rate of AD conversion in this MCI group (i.e., annual rate of 12%) is comparable to rates reported for different population studies, although these rates varied considerably (36, 37).

In summary, this study found that population-level increases in markers of brain metabolism and hippocampal volume, in subjects diagnosed with MCI, independently improved cognitive function at any given time, and appeared to reduce risk of progressive AD. These measures should be considered in future studies given their potential for specifying AD pathways and informing therapeutic and public health strategies against the disease.

Table 4
Overall and Joint Marginal Effects of FDG-PET, Hippocampal Volume and Brain Volume on Mean ADAS-Cog Examined With ApoE4 Status

Supplementary Material

02

Acknowledgments

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Abbott; Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Amorfix Life Sciences Ltd.; AstraZeneca; Bayer HealthCare; BioClinica, Inc.; Biogen Idec Inc.; Bristol-Myers Squibb Company; Eisai Inc.; Elan Pharmaceuticals Inc.; Eli Lilly and Company; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; GE Healthcare; Innogenetics, N.V.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Medpace, Inc.; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Servier; Synarc Inc.; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of California, Los Angeles. This research was also supported by NIH grants P30 AG010129, K01 AG030514, and the Dana Foundation.

The authors gratefully acknowledge Dr. Susan Landau for providing assistance in the navigation of the ADNI database, and Drs. Owen Carmichael, Charles DeCarli, and Christopher Schwarz for providing data from their work related to quantification of white matter hyperintensities, based on subject MRI scans from ADNI.

ABBREVIATIONS

AD
Alzheimer’s disease
ADAS-Cog
Alzheimer’s disease assessment scale-cognitive subscale
ADNI
Alzheimer’s disease neuroimaging initiative
ApoE4
Apolipoprotein e4 allele
CSF-AB
Cerebrospinal fluid beta-amyloid
CI
Confidence interval
FDG-PET
Fluro-deoxyglucose positron emission tomography
MRI
Magnetic resonance imaging
MCI
Mild cognitive impairment
MSM
Marginal structural model
SD
Standard Deviation

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

1. Alzheimer’s Association. Alzheimer’s disease facts and figures. Alzheimers Dement. 2012:131–168. [PubMed]
2. World Health Organization. Neurological Disorders-Public Health Challenges. World Health Organization; 2006.
3. Kuller LH, Shemanski L, Manolio T, Haan M, Fried L, Bryan N, et al. Relationship between ApoE, MRI findings, and cognitive function in the Cardiovascular Health Study. Stroke. 1998;29:388–398. [PubMed]
4. Vermeer SE, Prins ND, den Heijer T, Hofman A, Koudstaal PJ, Breteler MMB. Silent Brain Infarcts and the Risk of Dementia and Cognitive Decline. N Engl J Med. 2003;348(13):1215–1222. [PubMed]
5. Hofman A, Breteler MM, van Duijn CM, Krestin GP, Pols HA, Stricker BH, et al. The Rotterdam Study: objectives and design update. Eur J Epidemiol. 2007;22(11):819–829. [PMC free article] [PubMed]
6. Barnes DE, Haight TJ, Mehta KM, Carlson MC, Kuller LH, Tager IB. Secondhand Smoke, Vascular Disease, and Dementia Incidence: Findings From the Cardiovascular Health Cognition Study. Am J Epidemiol. 2010;171(3):292–302. [PMC free article] [PubMed]
7. Inaba M, White L, Bell C, Chen R, Petrovitch H, Launer L, et al. White matter lesions on brain magnetic resonance imaging scan and 5-year cognitive decline: the Honolulu-Asia aging study. J Am Geriatr Soc. 2011;59(8):1484–1489. [PubMed]
8. Craig-Schapiro R, Fagan AM, Holtzman DM. Biomarkers of Alzheimer’s disease. Neurobiol Dis. 2009;35(2):128–140. [PMC free article] [PubMed]
9. Weiner MW, Aisen PS, Jack CR, Jr, Jagust WJ, Trojanowski JQ, Shaw L, et al. The Alzheimer’s disease neuroimaging initiative: progress report and future plans. Alzheimers Dement. 2010;6(3):202–211 e7. [PMC free article] [PubMed]
10. Jack CR, Jr, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Weiner MW, et al. Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade. Lancet Neurol. 2010;9(1):119–128. [PMC free article] [PubMed]
11. Clark CM, Davatzikos C, Borthakur A, Newberg A, Leight S, Lee VM, et al. Biomarkers for early detection of Alzheimer pathology. Neurosignals. 2008;16(1):11–18. [PMC free article] [PubMed]
12. Jagust WJ, Landau SM, Shaw LM, Trojanowski JQ, Koeppe RA, Reiman EM, et al. Relationships between biomarkers in aging and dementia. Neurology. 2009;73(15):1193–1199. [PMC free article] [PubMed]
13. Landau SM, Harvey D, Madison CM, Reiman EM, Foster NL, Aisen PS, et al. Comparing predictors of conversion and decline in mild cognitive impairment. Neurology. 2010;75(3):230–238. [PMC free article] [PubMed]
14. Jack CR, Jr, Wiste HJ, Vemuri P, Weigand SD, Senjem ML, Zeng G, et al. Brain beta-amyloid measures and magnetic resonance imaging atrophy both predict time-to-progression from mild cognitive impairment to Alzheimer’s disease. Brain. 2010;133(11):3336–3348. [PMC free article] [PubMed]
15. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11:550–560. [PubMed]
16. Cook NR, Cole SR, Hennekens CH. Use of a marginal structural model to determine the effect of aspirin on cardiovascular mortality in the Physician’s Health Study. American Journal of Epidemiology. 2002;155:1045–1053. [PubMed]
17. Bodnar LM, Davidian M, Siega-Riz AM, Tsiatis AA. Marginal structural models for analyzing causal effects of time-dependent treatments: an application in perinatal epidemiology. Am J Epidemiol. 2004;159(10):926–934. [PubMed]
18. Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol. 1999;56(3):303–308. [PubMed]
19. Petersen RC, Aisen PS, Beckett LA, Donohue MC, Gamst AC, Harvey DJ, et al. Alzheimer’s Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology. 2010;74(3):201–209. [PMC free article] [PubMed]
20. ADNI Information Site. [accessed 12.03.12];Alzheimer’s Disease Neuroimaging Initiative. 2012 http://www.adni-info.org/
21. Mohs RC, Knopman D, Petersen RC, Ferris SH, Ernesto C, Grundman M, et al. Development of cognitive instruments for use in clinical trials of antidementia drugs: additions to the Alzheimer’s Disease Assessment Scale that broaden its scope. The Alzheimer’s Disease Cooperative Study Alzheimer Dis Assoc Disord. 1997;11(Suppl 2):S13–S21. [PubMed]
22. Jack CR, Jr, Bernstein MA, Borowski BJ, Gunter JL, Fox NC, Thompson PM, et al. Update on the magnetic resonance imaging core of the Alzheimer’s disease neuroimaging initiative. Alzheimers Dement. 2010;6(3):212–220. [PMC free article] [PubMed]
23. Freesurfer Software Site. [accessed 08.20.12];Cortical Reconstruction and volumetric segmentation. 2012 http://surfer.nmr.mgh.harvard.edu/
24. ADNI LONI Site. [accessed 08.20.12];Pet Image Processing. 2012 http://adni.loni.ucla.edu/research/protocols/pet-protocols/#acquisition.
25. Landau SM, Harvey D, Madison CM, Koeppe RA, Reiman EM, Foster NL, et al. Associations between cognitive, functional, and FDG-PET measures of decline in AD and MCI. Neurobiol Aging. 2009;32(7):1207–1218. [PMC free article] [PubMed]
26. Jagust WJ, Bandy D, Chen K, Foster NL, Landau SM, Mathis CA, et al. The Alzheimer’s Disease Neuroimaging Initiative positron emission tomography core. Alzheimers Dement. 2010;6(3):221–229. [PMC free article] [PubMed]
27. Trojanowski JQ, Vandeerstichele H, Korecka M, Clark CM, Aisen PS, Petersen RC, et al. Update on the biomarker core of the Alzheimer’s Disease Neuroimaging Initiative subjects. Alzheimers Dement. 2010;6(3):230–238. [PMC free article] [PubMed]
28. Fagan AM, Mintun MA, Shah AR, Aldea P, Roe CM, Mach RH, et al. Cerebrospinal fluid tau and ptau(181) increase with cortical amyloid deposition in cognitively normal individuals: implications for future clinical trials of Alzheimer’s disease. EMBO Mol Med. 2009;1(8–9):371–380. [PMC free article] [PubMed]
29. Robins JM. Association, causation, and marginal structural models. Synthese. 1999;121:151–179.
30. Jack CR, Jr, Dickson DW, Parisi JE, Xu YC, Cha RH, O’Brien PC, et al. Antemortem MRI findings correlate with hippocampal neuropathology in typical aging and dementia. Neurology. 2002;58(5):750–757. [PMC free article] [PubMed]
31. Korf ES, Wahlund LO, Visser PJ, Scheltens P. Medial temporal lobe atrophy on MRI predicts dementia in patients with mild cognitive impairment. Neurology. 2004;63(1):94–100. [PubMed]
32. Jack CR, Jr, Lowe VJ, Weigand SD, Wiste HJ, Senjem ML, Knopman DS, et al. Serial PIB and MRI in normal, mild cognitive impairment and Alzheimer’s disease: implications for sequence of pathological events in Alzheimer’s disease. Brain. 2009;132(Pt 5):1355–1365. [PMC free article] [PubMed]
33. Jack CR, Jr, Shiung MM, Weigand SD, O’Brien PC, Gunter JL, Boeve BF, et al. Brain atrophy rates predict subsequent clinical conversion in normal elderly and amnestic MCI. Neurology. 2005;65(8):1227–1231. [PMC free article] [PubMed]
34. Sinisi S, van der Laan M. Berkeley Electronic Press/UCB Biostatistics Division. 2004. Loss-Based Cross-Validated Deletion/Substitution/Addition Algorithms in Estimation; p. 143.
35. Haight TJ, Wang Y, van der Laan MJ, Tager I. Cross-validation deletion-substitution-addition model selection algorithm: Application to marginal structural models. Computational Statistics & Data Analysis. 2010;54(12):3080–3094.
36. Bruscoli M, Lovestone S. Is MCI really just early dementia? A systematic review of conversion studies. Int Psychogeriatr. 2004;16(2):129–40. [PubMed]
37. Panza F, D’Introno A, Colacicco AM, Capurso C, Del Parigi A, Caselli RJ, et al. Current epidemiology of mild cognitive impairment and other predementia syndromes. Am J Geriatr Psychiatry. 2005;13(8):633–644. [PubMed]
38. Greenland S, Robins J, Pearl J. Confounding and Collapsability in Causal Inference. Statistical Science. 1999;14(1):29–46.
39. van der Laan MJ, Robins J. Unified Methods for Censored Longitudinal Data and Causality. New York: Springer; 2002.
40. Hernan M, Brumback B, Robins J. Marginal Structural Models to Estimate Causal Effect of Zidovudine on the Survival of HIV-Positive Men. Epidemiology. 2000;11(5):561–570. [PubMed]
41. van der Laan M, Dudoit S, van der Vaart AW. The Cross-Validated Adaptive Epsilon-Net Estimator. Berkeley Electronic Press/UCB Biostatistics Division; 2004. p. 142.
42. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. New York: Chapman and Hall; 1993.