Nonlinear Association Between Cerebrospinal Fluid and Florbetapir F-18 β-Amyloid Measures Across the Spectrum of Alzheimer Disease

Abstract

Importance: Cerebrospinal fluid (CSF) and positron emission tomographic (PET) amyloid biomarkers have been proposed for the detection of Alzheimer disease (AD) pathology in living patients and for the tracking of longitudinal changes, but the relation between biomarkers needs further study.

Objective: To determine the association between CSF and PET amyloid biomarkers (cross-sectional and longitudinal measures) and compare the cutoffs for these measures.

Design, setting, and participants: Longitudinal clinical cohort study from 2005 to 2014 including 820 participants with at least 1 florbetapir F-18 (hereafter referred to as simply florbetapir)-PET scan and at least 1 CSF β-amyloid 1-42 (Aβ1-42) sample obtained within 30 days of each other (501 participants had a second PET scan after 2 years, including 150 participants with CSF Aβ1-42 measurements). Data were obtained from the Alzheimer's Disease Neuroimaging Initiative database.

Main outcomes and measures: Four different PET scans processing pipelines from 2 different laboratories were compared. The PET cutoff values were established using a mixture-modeling approach, and different mathematical models were applied to define the association between CSF and PET amyloid measures.

Results: The values of the CSF Aβ1-42 samples and florbetapir-PET scans showed a nonlinear association (R2 = 0.48-0.66), with the strongest association for values in the middle range. The presence of a larger dynamic range of florbetapir-PET scan values in the higher range compared with the CSF Aβ1-42 plateau explained the differences in correlation with cognition (R2 = 0.36 and R2 = 0.25, respectively). The APOE genotype significantly modified the association between both biomarkers. The PET cutoff values derived from an unsupervised classifier converged with previous PET cutoff values and the established CSF Aβ1-42 cutoff levels. There was no association between longitudinal Aβ1-42 levels and standardized uptake value ratios during follow-up.

Conclusions and relevance: The association between both biomarkers is limited to a middle range of values, is modified by the APOE genotype, and is absent for longitudinal changes; 4 different approaches in 2 different platforms converge on similar pathological Aβ cutoff levels; and different pipelines to process PET scans showed correlated but not identical results. Our findings suggest that both biomarkers measure different aspects of AD Aβ pathology.

Figure 1. Association Between CSF Aβ1–42 Levels and Florbetapir F-18 PET SUVRs

A model for participants with 2 APOE ε4 copies is not included. The solid gray areas indicate disagreement in the classification based on the pair of Aβ measures. Aβ indicates β-amyloid; AD, Alzheimer disease; CB, cerebellum; CSF, cerebrospinal fluid; MARS, multivariate adaptive regression spline; MCI, mild cognitive impairment; PET, positron emission tomographic; SUVR, standardized uptake value ratio; and WM, white matter.

Figure 2. Clinical Associations and Longitudinal Changes of PET SUVRs

Scatterplots showing the association between Alzheimer’s Disease Assessment Scale–cognitive subscale (ADAS-cog) scores on the y-axis and the summary composite standardized uptake value ratios (SUVRs) (A) and the cerebrospinal fluid (CSF) β-amyloid 1–42 (Aβ1–42) levels (B) on the x-axis. The blue continuous line represents the multivariate adaptive regression spline for a 65-year-old participant. The red dashed line represents the cutoff for the biomarker represented in the plot, and the green dashed line represents the value at which the CSF Aβ1–42 level plateaus. Longitudinal SUVR yearly changes (after a 2-year follow-up) for the average cerebellum (CB) (C), the average white matter (WM) (D), the summary CB (E), and the summary composite (F) are shown on the y-axes, with the x-axes representing baseline SURVs. G, Longitudinal changes in CSF Aβ1–42 level after a 2-year follow-up are shown. The red dashed line represents the value that corresponds to the CSF Aβ1–42 level of 192 pg/mL. Yearly changes in the summary CB (H) and the summary composite (I) values during follow-up (y-axis) are based on the presence of normal or abnormal baseline CSF Aβ1–42 levels and florbetapir F-18–positron emission tomographic (PET) measures. The horizontal line in each box indicates the median, while the top and bottom borders of the box mark the 75th and 25th percentiles, respectively. The whiskers above and below the box mark the 90th and 10th percentiles, respectively. The points beyond the whiskers are outliers beyond the 90th and 10th percentiles.

Figure 3. Association Between CSF Aβ1–42 Level and Florbetapir F-18–PET Measure Longitudinal Changes

Matrix showing the individual scatterplots depicting the association between cerebrospinal fluid (CSF) β-amyloid 1–42 (Aβ1–42) level (x-axis) and florbetapir F-18–positron emission tomographic (PET) standardized uptake value ratio (y-axis) changes during a 2-year follow-up (below the diagonal) and the corresponding Pearson correlation coefficient and P value (above the diagonal). The panels in the diagonal direction depict histograms showing the distribution of CSF Aβ1–42 levels and PET SUVRs. Average_CB indicates average cerebellum; Average_WM, average white matter; Sum_CB, summary cerebellum; and Sum_Comp, summary composite.