Association of Low Plasma Aβ42/Aβ40 Ratios With Increased Imminent Risk for Mild Cognitive Impairment and Alzheimer Disease FREE

There is a growing consensus that the best way to manage Alzheimer disease (AD) will ultimately be through preventive therapy similar to that used for atherosclerotic heart disease. To facilitate preventive therapy, it is important to develop AD-related biomarkers that can be used to identify at-risk individuals in the same way that cholesterol levels are used to identify those at risk for atherosclerotic heart disease. In 1998, the Consensus Report of the Working Group on Molecular and Biochemical Markers of Alzheimer Disease¹ proposed that an ideal diagnostic marker for AD should be reliable, able to detect a fundamental feature of AD neuropathology, and validated through confirmed cases. In addition, it should be noninvasive, simple to perform, and inexpensive. These criteria apply equally well to premorbid and diagnostic biomarkers. In this study, we evaluated plasma amyloid β protein (Aβ40 and Aβ42) and the Aβ42/Aβ40 ratio as premorbid biomarkers for AD and the amnestic type of mild cognitive impairment (MCI),² which usually precedes AD.

The deposition of Aβ in senile plaques is a pathological hallmark of AD. Measurement of plasma Aβ40 and Aβ42 levels is noninvasive, and it could be developed into a test that is as simple to perform and inexpensive as the test for high- and low-density lipoprotein levels. Thus, plasma Aβ40, Aβ42, and the Aβ42/Aβ40 ratio could be highly useful premorbid biomarkers if they are reliable predictors of MCI and AD.

The amyloid β protein is a secreted peptide, normally present in plasma and cerebrospinal fluid (CSF), that is derived from a large precursor protein through the sequential action of 2 proteases referred to as β and γ secretase.³ Most secreted Aβ has 40 amino acids (Aβ40), but a small percentage has 42 (Aβ42).³ In all patients with AD, Aβ42 aggregates and deposits in the brain, forming senile plaques that are one of the pathological hallmarks of AD.⁴ Studies of autopsied brains have shown that the Aβ deposited in AD has a ragged amino terminus and that, in approximately one third of patients, Aβ42 is virtually the only form deposited.⁴ In another third, Aβ42 is the major form deposited and, in the remaining third, large amounts of Aβ40 and Aβ42 are deposited.⁴

The level of Aβ42 in CSF is significantly reduced both in patients with AD and in those with MCI, which precedes AD.⁵ In the Tg2576 mouse model of AD, the CSF Aβ level also declines as Aβ is deposited in the brain.⁶ Importantly, plasma Aβ levels decline in parallel.⁶

Other observations suggest that, in many patients with AD, deposition of Aβ42 may be preceded and fostered by an elevated level of Aβ that can be detected in plasma. Patients with trisomy 21, who invariably develop AD pathology if they live past 40 years of age, have higher plasma Aβ40 and Aβ42 levels compared with controls from birth.^7- 8 Although Aβ40 and Aβ42 levels are elevated in trisomy 21, the study of trisomy 21 brains obtained at autopsy has shown that Aβ42 is deposited before Aβ40. It is well established that virtually all of the genetic mutations that cause early-onset familial AD increase the Aβ42 level in a way that is detectable in plasma.⁹ More recently, we¹⁰ and others¹¹ have shown that plasma Aβ40 and Aβ42 levels increase with aging, one of the most important risk factors for AD. Finally, Ertekin-Taner et al¹² have shown that plasma Aβ40 and Aβ42 levels are elevated in asymptomatic first-degree relatives of patients with AD, who are known to be at increased risk for AD.

These summarized results indicate that Aβ42 levels are reduced in the CSF of patients with MCI or AD in association with the selective deposition of Aβ42 that occurs in the brain. They suggest that the plasma Aβ42 level and Aβ42/Aβ40 ratio may decline in parallel, but in many at-risk individuals plasma Aβ40 and Aβ42 levels appear to be elevated initially; thus, the plasma Aβ42 level may often decline into the normal range as Aβ42 deposition commences. This makes the Aβ42/Aβ40 ratio a potential biomarker for selective Aβ42 deposition. The results of the longitudinal study presented herein are, to our knowledge, the first to indicate that the plasma Aβ42/Aβ40 ratio may be useful for identifying elderly white subjects who are at imminent risk for MCI or AD.

This study was performed under the supervision of the Mayo Institutional Review Board and all participants signed informed consent.

Participants were 563 cognitively normal older adults forming a complete subset of those who entered the Mayo Clinic Alzheimer Disease Patient Registry (ADPR) as normal controls from 1992 through 2003 and had at least 2 stored plasma specimens. For purposes of this and all ADPR studies, cognitively normal adults are defined as community-dwelling, independently functioning individuals who were examined by their primary care physician and met the following selection criteria:

Individuals with histories of medical conditions or disorders that may affect cognition (eg, head injury) were included only if the condition was no longer active and there was no evidence of persistent or residual cognitive impairment. Individuals with current, chronic medical conditions (eg, hypertension or diabetes) were included if their physician judged their condition to be well controlled and not affecting cognition.

All participants underwent baseline neurologic and neuropsychological evaluations. In addition, a study nurse collected collateral information, including family history of dementia, current medications, the Clinical Dementia Rating,¹³ and the Neuropsychiatric Inventory¹⁴ from an informant. Activities of daily living were evaluated using parts A and B of the Record of Independent Living.¹⁵ Data from informants were available for 98% of the study participants. Family members provided data 84% of the time and close friends provided information 14% of the time. When an informant was not available, information was solicited directly from the participant.

After these data were collected, a neurologist reviewed the Clinical Dementia Rating score with the participant and informant and performed a neurologic examination, including the Kokmen Short Test of Mental Status,¹⁶ Hachinski Ischemic Index,¹⁷ and Unified Parkinson's Disease Rating Scale.¹⁸ An experienced psychometrist then administered a baseline battery of neuropsychological tests, including the Dementia Rating Scale (DRS),¹⁹ Auditory Verbal Learning Test,²⁰ Wechsler Memory Scale–Revised,²¹ and Wechsler Adult Intelligence Scale–Revised.²²

Study participants were contacted annually for reevaluation. Medication lists and information regarding family history of dementia were updated at each follow-up visit. In addition, the Clinical Dementia Rating, Record of Independent Living, Neuropsychiatric Inventory, Hachinski Ischemic Index, Unified Parkinson's Disease Rating Scale, Kokmen Short Test of Mental Status, and neuropsychological test battery were repeated.

All baseline and follow-up examinations were reviewed at a weekly consensus conference, staffed by Mayo Alzheimer's Disease Research Center neurologists, neuropsychologists, nurses, psychometrists, a geriatrician, and a social worker. At baseline, entry criteria were reviewed and a Clinical Dementia Rating score of 0 was confirmed for all individuals enrolled in the ADPR as normal. Neuropsychological data were used to make the consensus diagnosis. Persons not meeting these criteria were assigned a clinical diagnosis and offered follow-up through alternate study mechanisms, if appropriate.

At the annual follow-up, all diagnostic information, including neuropsychological data, was reviewed at a consensus conference. Data were compared with those of baseline studies to evaluate progression to MCI or dementia. Diagnoses of amnestic and nonamnestic MCI were determined using Mayo criteria.²³ Standard published criteria were used to diagnose possible and probable AD.²⁴ Participants with abnormal clinical findings on follow-up but who did not meet the established criteria for MCI or dementia were coded as having cognitive impairment of undetermined origin. In the follow-up of this series, we have reported that more than 85% of amnestic MCI cases convert to AD by 7 years.²⁵ For this reason and because of limited numbers of patients converting to AD, we combined our end point as incident cases of amnestic MCI and AD (MCI/AD).

For inclusion in the analysis, we required that participants be originally recruited as cognitively normal, have at least 12 months of follow-up beyond the first sample, and be older than 55 years. Table 1 shows the characteristics of the 563 participants who met these criteria. Median follow-up for this study was 3.7 (25%-75% range, 2.6-5.9) years. Within 15 months of the abstraction for this study, 444 of these 563 ADPR participants had follow-up visits and were still active. Six participants had only 1 follow-up visit, 171 had a total of 3 visits, 155 had 4 visits, 69 had 5 visits, and the remaining 162 had 6 or more visits. Fifty-three participants (9.4%) were diagnosed as having MCI/AD during their follow-up. Of the 53 converters, 36 developed MCI, and 7 of these went on to develop AD (5 probable and 2 possible). The other 17 developed AD (12 probable and 5 possible) without being observed with MCI. Median age at the first plasma sample was 78 (25%-75% range, 73-83) years, and median age of onset of MCI/AD was 88 (25%-75% range, 85-92) years.

Recent data indicate that medications and supplements may influence Aβ levels.²⁶ We elected not to adjust for this potential confounder because the list of medications that may affect Aβ is evolving and uncertain and because the effect of suggested Aβ modifiers in the doses normally given to elderly subjects has yet to be firmly established in large follow-up studies.

To evaluate the relationship of cognitive change to plasma Aβ levels, we identified a subgroup of individuals who had 2 DRS evaluations approximately 5 years apart and a plasma sample at the time of the first evaluation. They also had to be cognitively normal at the first time point. A window of 5 (±2) years was used for the time between DRS evaluations, and a window of ±6 months was used for the time between the initial visit and the date the plasma sample was obtained. The cohort who met these criteria consisted of 379 persons, of whom 237 (63%) were female, with a median age of 77 years and a median education of 13 years.

At each visit, nonfasting patients had blood drawn into an EDTA tube. The plasma was spun off, divided into 1-mL aliquots, and stored at −70°C. One aliquot was used to measure the plasma Aβ levels. Plasma Aβ measures were made using an enzyme-linked immunosorbent assay method with the well-characterized antibodies BAN-50/BA27 specific for Aβ40 and BAN-50/BC05 specific for Aβ42.⁹Figure 1 shows the distributions of raw plasma Aβ levels.

The Kaplan-Meier method was used to estimate the distribution of time to development of MCI/AD, with the time of the first plasma sample considered as the start of follow-up. These times were censored at the date of the last follow-up for patients who never developed MCI/AD. Cox proportional hazards models were used to investigate associations with Aβ measures. Because of the extremely skewed distributions of the Aβ measures and a number of outliers (Figure 1), in this analysis the Aβ42 and Aβ40 levels and Aβ42/Aβ40 ratio were included in models in a number of different ways so that sensitivity of results could be explored. The primary approach was to include each Aβ measure as a numerical variable defined by the quartiles of the distribution. Quartiles (labeled as Q1-Q4) were also included as categorical variables to enable estimation of hazard ratios relative to the highest quartile and to aid in exploring whether a linear association was reasonable. Other approaches were to include each Aβ measure as a continuous variable on the logarithm scale and as a binary category variable defined by the median. Sensitivity of results to the removal of the outlying Aβ measures (the upper and lower 2%) was considered in the continuous variable models. Age, sex, presence of a 3/4 or 4/4 genotype of apolipoprotein E (APOE), years of education, age-adjusted DRS score, and family history of dementia were considered to be potential confounding variables and, of these, age and APOE4 allele showed evidence of an association with the Aβ40 level, Aβ42/Aβ40 ratio, as well as with time to MCI/AD at the 0.20 significance level (Table 2 and Table 3). We designated these to be likely confounding variables and adjusted for them in the primary analyses addressing the association of Aβ measures with time to MCI/AD. No adjustments were made for multiple testing in these exploratory analyses.

Further regression models were used to explore the relationship between Aβ measures and change in cognition during a subsequent period of approximately 5 years in a subset of 379 participants who were administered the DRS within 6 months of an Aβ measure. The total DRS raw score was converted to an age-adjusted scaled score on the basis of data from the Mayo Older American Normative Study. Because follow-up intervals varied, annualized rates of cognitive change were used as the outcome variable in the least-squares regression analyses that were conducted similarly to the time-to-conversion analyses.

Fifty-three of the 563 subjects developed MCI/AD. The estimated cumulative incidence in this cohort of patients (Figure 2) with a median age of 78 years at the start of follow-up was 5% at 4 years, with 95% confidence interval (CI) ranging from 3% to 8%. It reached 11% (95% CI, 7%-14%) at 6 years, 18% (95% CI, 12%-24%) at 8 years, and 30% (95% CI, 19%-39%) at 10 years of follow-up. Advancing age and APOE genotype, 2 well-established risk factors for AD, showed evidence of association with incidence of MCI/AD (Table 3), whereas family history did not.

The plasma Aβ42/Aβ40 ratio showed evidence of association with conversion to MCI/AD (Table 4), but the Aβ42 and Aβ40 levels did not (Table 4). The risk of MCI/AD for patients with an Aβ42/Aβ40 ratio in the lowest quartile was estimated to be 3 times the risk for subjects with a ratio in the highest quartile (P = .01), and this association persisted (P = .04) after adjusting for age and APOE genotype. The relationship between the Aβ42/Aβ40 ratio and cumulative incidence of MCI/AD is shown graphically in Figure 3. Subjects whose Aβ42/Aβ40 ratio was in the lowest quartile (Q1, <0.05) reached a 10% incidence by 3 years, followed by those in Q2 who took approximately 5 years, and those in Q3 and Q4 who took approximately 8 years to reach 10% cumulative incidence. This pattern was suggestive of a deviation from the proportional hazards assumption of our models, but formal significance tests indicated no significant deviation.

Figure 4 shows the cumulative incidence of MCI/AD stratified according to whether the Aβ42/Aβ40 ratio was above or below the median separately for patients with the APOE 3/3 and 3/4 genotypes. Figure 5 similarly groups patients by the Aβ42/Aβ40 ratio and age. Figure 6 shows the estimated proportions of individuals with MCI/AD after 5 years of follow-up depending on the combined effects of age (<80 or ≥80 years), Aβ42/Aβ40 ratio (above or below the median), and specific APOE genotype (3/3 or 3/4). Among older subjects with the 3/4 genotype, the estimated incidence at 5 years was 32% (95% CI, 9%-50%) for those with an Aβ42/Aβ40 ratio below the median, compared with only 6% (95% CI, 0%-16%) for those with a ratio above the median. For older subjects with the 3/3 genotype, the incidences were 17% (95% CI, 6%-26%) and 7% (95% CI, 0%-15%) for those with ratios below and above the median, respectively. Among younger individuals with the 3/4 genotype, the 5-year incidence was 10% (95% CI, 0%-23%) for those with a low Aβ42/Aβ40 ratio, compared with 3% (95% CI, 0%-9%) for those with a higher ratio. In the lowest risk group according to age and APOE genotype (<80 years with the 3/3 genotype), the estimated incidence was only 2%, regardless of whether the Aβ42/Aβ40 ratio was below or above the median (95% CIs, 0%-7% and 0%-6%, respectively).

The regression analysis using changes in DRS scores as the outcome variable in a subset of patients resulted in confirmatory findings. After adjustment for age and APOE genotype, the Aβ42/Aβ40 ratio showed similar levels of evidence of association with decline in cognition, with lower ratios being associated with greater declines (Table 5).

Our findings indicate that cognitively normal elderly white subjects with low Aβ42/Aβ40 ratios may be at increased risk for MCI/AD. This association is best explained by postulating that the time course of cerebral Aβ aggregation and deposition is extended, typically occurring over a period of 10 to 15 years before MCI/AD develops in the elderly white population that we studied. Amyloid β42 deposits selectively in the AD brain,⁴ and it is well established that the CSF Aβ42 level is decreased in subjects with MCI or AD. In the Tg2576 model of AD, the plasma Aβ levels decline in parallel with the CSF Aβ level as Aβ deposits in the brain. We postulate, therefore, that the CSF and plasma Aβ42 levels decline in parallel as Aβ42 deposits selectively in the human AD brain. Because Aβ42 aggregation and deposition precede MCI/AD, most subjects who develop MCI/AD have a low Aβ42/Aβ40 ratio several years before MCI/AD can be diagnosed. Thus, we postulate that the plasma Aβ42/Aβ40 ratio is not a trait variable like the APOE4 allele or family history. It is rather a state variable that reflects the extent to which there has been selective aggregation and deposition of Aβ42 in the brain.

It is well established that the risk for AD increases with aging and that it increases in first-degree relatives up to the age of approximately 80 years. Both of these factors are associated with increases in plasma Aβ40 and Aβ42 levels. If the time course of aggregation and deposition is extended as our data suggest, then elderly white subjects with elevated Aβ levels will be at increased risk for MCI/AD but not at increased immediate risk that could be detected in a 5-year period. In subjects who begin with elevated Aβ40 and Aβ42 levels, the plasma Aβ42 level may decline into the normal range as selective Aβ42 deposition occurs and the Aβ42/Aβ40 ratio falls into the lowest quartile. This could explain why imminent MCI/AD is predicted better by a low Aβ42/Aβ40 ratio than by a low Aβ42 level alone.

Two studies^27- 28 from North Manhattan, NY, have been reported in which the plasma Aβ level was analyzed as a premorbid biomarker for AD. In both studies, an elevated Aβ42 level was associated with an increased risk of conversion to AD. There were, however, important differences between those studies and the one we report herein. The population assessed in the studies reported from North Manhattan differed from our population in 2 important ways: the incidence rate was much higher (16% at 5 years for AD vs 9% at 5 years for MCI/AD in this study), despite a slightly lower mean age at baseline (76 vs 78 years), and the population was mixed (one third consisted of African American subjects; one third, Caribbean Hispanic first-generation immigrants; and one third, white subjects) with 16.8% of the African American, 19.7% of the Caribbean Hispanic, and 7.6% of the white subjects developing AD in those studies. The African American and Caribbean Hispanic subjects not only had a higher incident rate than the white population, but also differed in important ways such as having different APOE4 frequencies. The apparent conflict between our results and those reported by Mayeux et al^27- 28 can be resolved by postulating that elevated Aβ42 levels cause AD relatively quickly in the North Manhattan population but more slowly in the our elderly white population. In the North Manhattan population, the relatively short time from the elevation of the Aβ42 level to the onset of AD would make an elevated Aβ42 level a good predictor for imminent AD. If the time from the Aβ42 level elevation to onset of MCI/AD was greater in our white population, then the plasma Aβ42 level would decline into the normal range in the premorbid period as Aβ42 deposits selectively in the brain and the Aβ42/Aβ40 ratio would fall to low levels, making the Aβ42/Aβ40 ratio a better predictor of imminent MCI/AD in this population.

The relative risk of AD in first-degree relatives of subjects with AD declines with aging, as does the relative risk in those with an APOE4 allele. In our series, which was composed of many subjects older than 80 years, family history (defined as being a first-degree relative of a person with AD) showed no significant association with progression to MCI/AD. After correcting for age and APOE genotype, the relative risk for family history was only 1.1, with a 95% CI of 0.7 to 2.0 and a P value of .67. After correcting for age, the relative risk associated with a higher-risk APOE genotype (3/4 or 4/4) was 2.1, with a 95% CI of 1.2 to 3.7 and a P value of .01.

The estimated relative risk of 3.1 that we obtained for the Aβ42/Aβ40 ratio in our exploratory evaluation of Aβ levels in an elderly white population suggests that, in this population, the Aβ42/Aβ40 ratio may be as useful as APOE4 and more useful than family history as a premorbid biomarker for identifying those at imminent risk for MCI/AD. The results obtained when we assessed the combined effect of the Aβ42/Aβ40 ratio and aging were particularly striking. In those older than 80 years whose Aβ42/Aβ40 ratio was below the median, the 5-year cumulative incidence of MCI/AD was approximately 20%. In all other subjects, it was approximately 5%. If these findings are confirmed in additional large longitudinal studies, the plasma Aβ42/Aβ40 ratio could become an important biomarker for developing and implementing a preventive approach to AD therapy in elderly white subjects.

—————————————————————

Since the manuscript was submitted, van Oijen et al²⁹ published a study analyzing plasma Aβ (Aβ42, Aβ40, and the Aβ40/Aβ42 ratio) in a large case-cohort study embedded in the prospective, population-based Rotterdam Study. Their results are similar to those reported in our study.

Correspondence: Neill R. Graff-Radford, MBBCh, FRCP, Department of Neurology, or Steven G. Younkin, MD, PhD, Department of Neuroscience, Mayo College of Medicine, Mayo Clinic, 4500 San Pablo Rd, Jacksonville, FL 32224 (graffradford.neill@mayo.edu or younkin.steven@mayo.edu).

Accepted for Publication: July 6, 2006.

Author Contributions: Drs Graff-Radford and S. G. Younkin had full access to the data and take responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: Graff-Radford, L. H. Younkin, and S. G. Younkin. Acquisition of data: Lucas, Boeve, Knopman, Ivnik, L. H. Younkin, and S. G. Younkin. Analysis and interpretation of data: Graff-Radford, Crook, Ivnik, Smith, Petersen, and S. G. Younkin. Drafting of the manuscript: Graff-Radford, Crook, L. H. Younkin, Petersen, and S. G. Younkin. Critical revision of the manuscript for important intellectual content: Graff-Radford, Crook, Lucas, Boeve, Knopman, Ivnik, Smith, Petersen, and S. G. Younkin. Statistical analysis: Crook, Smith, and S. G. Younkin. Obtained funding: Graff-Radford, Petersen, and S. G. Younkin. Administrative, technical, and material support: Lucas, Boeve, Smith, L. H. Younkin, Petersen, and S. G. Younkin. Study supervision: Graff-Radford and S. G. Younkin.

Financial Disclosure: Drs Graff-Radford, Crook, Lucas, L. H. Younkin, and S. G. Younkin have been included as inventors in a patent application filed by the Mayo Clinic on the Aβ42/Aβ40 ratio as a predictive test for MCI and AD.

Funding/Support: This study was supported by grants R01 AG06656, P50 AG16574, and U01 AG06786 from the National Institute on Aging, and by the Robert and Clarice Smith and Abigail Van Buren Alzheimer's Disease Research Program.

Role of the Sponsors: The funding agencies had no role in the design or conduct of the study; the collection, management, and analysis of data; or the preparation, review, and approval of the manuscript.

Acknowledgment: We thank the following research assistants, of Jacksonville, Fla, for their assistance: Francine Parfitt, MA, Josie Pagdanganan, Jennifer Adamson, and Michael Heckman, MS.