Three out of 4 serum biomarkers for Alzheimer disease (AD) were inconsistent among Black patients, according to results of a study published in Neurology.
Data for this study were collected at the Charles F. and Joanne Knight Alzheimer Disease Research Center. Community-dwelling older adults with or without cognitive impairment underwent cognitive evaluation, provided a blood sample, and had cerebrospinal fluid (CSF) collection and/or amyloid positron emission tomography (PET) scan.b
Individuals with CSF biomarker data and sufficient plasma aliquots available were matched 1:1 between cohorts of self-identified African American (AA; n=76) and non-Hispanic White (NHW; n=76) individuals. Cohorts were evaluated by plasma biomarkers of amyloid (and Aβ42/Aβ40 ratio), tau (p-tau181 and p-tau231) and neuroaxonal injury (neurofilament light chain [NfL]) using high-performance immunoprecipitation-mass spectrometry or Simoa immunoassays. Results of plasma biomarkers were compared with results from CSF biomarkers or PET scans.
The AA and NHW cohorts were aged median 68.4 (interquartile range [IQR], 64.9-73.2) and 68.4 (IQR, 64.1-73.1) years, 58% and 51% were women, and 42% and 42% were carriers of apolipoprotein E (APOE) e4, respectively.
According to the amyloid PET scan (median centiloids, 2.3 vs 10.1; adjusted P =.02) and CSF biomarkers (22% vs 43%; P =.003), the AA cohort had lower average levels of brain amyloidosis.
In the plasma biomarker analysis, the AA cohort had significantly higher concentration of Aβ42 (median, 41.9 vs 40.9 pg/ml; adjusted P =.03) and a higher ratio of Aβ42/Aβ40 (median, 0.1047 vs 0.0963; adjusted P <.0001). The AA and NHW cohorts did not differ for plasma concentrations of Aβ40 (median, 409 vs 425 pg/ml), p-tau181 (median, 12.3 vs 14.2 pg/ml), p-tau231 (median, 8.2 vs 9.1 pg/ml), or NfL (median, 11.1 vs 11.8 pg/ml), respectively.
Compared with CSF biomarkers, results from the plasma biomarkers corresponded for Aβ40 and the ratio of Aβ42/Aβ40. The other biomarker results differed in the CSF evaluation, in which the AA cohort was associated with similar Aβ42 (median, 735 vs 682 pg/ml) and decreased p-tau181 (median, 31 vs 38.0 pg/ml; adjusted P =.0008) and NfL (median, 644 vs 736 pg/mL; adjusted P =.08) compared with the NHW cohort.
Among the plasma biomarkers, the ratio of Aβ42/Aβ40 was most strongly correlated with CSF Aβ42/Aβ40 (r, 0.52) and PET centiloids (r, -0.30). Similarly, plasma Aβ42/Aβ40 had the highest area under the receiver operating characteristic curve (ROC AUC) for predicting CSF (ROC AUC 0.86 vs 0.64-0.76) and PET (ROC AUC 0.86 vs 0.55-0.74) amyloid status.
In a model which accounted for covariates, the plasma Aβ42/Aβ40 model (ROC AUC, 0.90) outperformed the model with only covariates (ROC AUC, 0.82) for predicting CSF Aβ42/Aβ40. In this combined covariate-plasma Aβ42/Aβ40 model, individuals who had cognitive impairment (odds ratio [OR], 9.2; 95% CI, 1.9-46; P =.007), were APOE e4 carriers (odds ratio [OR], 5.7; 95% CI, 2.3-14; P =.0002), or were older (OR, 1.2; 95% CI, 1.03-1.21; P =.007) and were more likely to have CSF Aβ42/Aβ40 positivity.
In a model-based solely on covariates, AA individuals were associated with decreased risk for CSF Aβ42/Aβ40 positivity (OR, 0.27; 95% CI, 0.12-0.64; P =.003). Similarly, AA individuals were at a decreased risk for positivity in models based on plasma p-tau181, p-tau231, and NfL (all P ≤.007). Similar findings were observed for predicting amyloid PET positivity.
This study may have been limited by the small number of individuals (n=7) in each cohort who had cognitive impairment.
These data indicated that plasma Aβ42/Aβ40, which the C2N Diagnostics PrecivityAD plasma Aβ42/Aβ40 assay is based, more accurately predicted CSF or PET amyloid status among both AA and NHW individuals compared with other plasma biomarkers. Altogether, these findings suggested that predicting AD risk using other plasma biomarkers would not be consistent among a diverse patient population.
We spoke with Suzanne E. Schindler, MD, PhD, associate professor of Neurology, Washington University School of Medicine in St. Louis, Missouri, and lead author of the study.
What was the motivation for your study?
Dr Schindler: We’re trying to develop blood tests for Alzheimer disease. There are other tests for Alzheimer disease, such as amyloid positron emission tomography (PET) scans and spinal fluid tests. But scans are very expensive and not many centers do them routinely. More centers can do spinal fluid tests, but many patients are not excited about getting a spinal tap. Basically, the other tests just aren’t practical.
Unfortunately, most tests for Alzheimer disease have only been validated in non-Hispanic White cohorts. There has been some evidence in the last several years that spinal fluid tests have a different meaning in African Americans and non-Hispanic Whites. As there are many researchers developing blood tests for Alzheimer disease, we wanted to evaluate how these blood tests perform in different groups, asking whether they give consistent results no matter if the patient is African American or White.
What did your study find?
Dr Schindler: One blood test, for Aβ42/Aβ40, gave consistent results between African American and non-Hispanic White patients. For 3 other blood tests, the results were inconsistent, and relying on these results could lead to the disproportionate misdiagnosis of African Americans.
What do these findings mean for clinical practice?
Dr Schindler: The only Alzheimer disease blood test that is currently available for clinical use is the test that performed well, so that is good news. But these other tests are in the pipeline. I think our findings demonstrate that the companies developing these assays need to evaluate them in diverse cohorts.
One potential reason why the Aβ42/Aβ40 assay worked better is because it’s a ratio of 2 proteins. Using ratios instead of the concentration of a single protein might help to account for variations between people.
Race norms are a big topic in medicine right now and there’s concern that using different cut-offs for different races has actually exacerbated health care disparities. I think a better solution is either to use a test that performs consistently across all groups, or to account for the underlying differences between groups instead of adjusting for race.
What do you think these findings mean for the genetics or the cause of AD in general?
Dr Schindler: One aspect that’s a bit puzzling about our findings, and other work, is that it looks like the levels of brain amyloid are lower in African Americans even when you adjust for age and sex. A recent study that came out in JAMA found that rates of dementia in Hispanic and African American are higher than in White individuals, so you would expect they’d have more amyloid, not less. We’re still trying figure out the reason for these differences.
Do you think the findings of this study are generalizable to other areas of neurology or medicine in general, or should diagnostic tools be evaluated critically about whether they are generalizable to all populations?
Dr Schindler: Nearly all research studies are based on largely non-Hispanic White populations. We have very poor representation of minoritized groups, particularly of African American individuals, in research studies. It’s a big problem that prevents generalizing research results to the general population.
I think when you’re diagnosing someone with a disease that has considerable implications for their health that you want to make sure that you get accurate results regardless of a person’s race or ethnic group. The United States Food and Drug Administration is coming out with guidelines saying that you need to have a diversity plan for clinical trials. They have not come out with guidance yet for diagnostics, but I could certainly imagine that happening.
Disclosure: Multiple authors declared affiliations with industry. Please refer to the original article for a full list of disclosures.
Schindler SE, Karikari TK, Ashton NJ, et al. Effect of race on prediction of brain amyloidosis by plasma Aβ42/Aβ40, phosphorylated tau, and neurofilament light. Neurology. 2022;10.1212/WNL.0000000000200358. doi:10.1212/WNL.0000000000200358