Most of the research on Alzheimer’s biomarkers to date has been done in predominantly white populations, but a few small studies have hinted at possible racial differences. A paper in the January 7 JAMA Neurology supports this idea. Researchers led by John Morris at Washington University in St. Louis found less cerebrospinal fluid tau but more hippocampal atrophy in at-risk African-Americans than in white counterparts. The differences only cropped up in subgroups, however, and the small number of participants render any conclusions tentative. Do the findings represent a true biological difference, or merely reflect recruitment biases and social factors that the study could not account for?

  • The largest AD biomarker study of African-Americans to date finds some racial differences.
  • African-American ApoE4 carriers had lower CSF tau than white E4 carriers.
  • It is unclear if the findings reflect biology, or are due to environmental factors or recruitment bias.

Other researchers said that larger cohorts will be needed to answer this, but called the WashU findings a useful step. “This study represents the largest sample size of African-American individuals with positron emission tomography and CSF data to date,” Lisa Barnes at Rush University Medical Center, Chicago, wrote in an accompanying editorial. Adam Brickman at Columbia University in New York City noted, “This certainly provides more evidence that race, ethnicity, and likely other demographic variables need to be considered explicitly in future work related to Alzheimer’s biomarkers.”

Efforts to study AD biomarkers in diverse populations have been hampered by poor recruitment of minorities. In St. Louis, African-Americans make up about 18 percent of the population, but initially comprised only 3 percent of the observational cohort at the Knight Alzheimer’s Disease Research Center at WashU. Over the last two decades, researchers have increased participation to 18 percent, and last October hosted a conference to discuss strategies for boosting African-American enrollment (Oct 2018 conference news). There, Morris and William Hu from Emory University, Atlanta, reported independent preliminary data suggesting CSF tau runs lower in African-Americans than in whites (Oct 2018 conference news). 

Morris and colleagues now publish the cross-sectional data from this observational cohort. Out of 1,255 participants with biomarker data, 173, or 14 percent, were African-American. For specific biomarkers, the numbers were smaller, however, because not all participants consented to all procedures. A total of 143 African-Americans underwent MRI scanning, 87 donated CSF, and 65 had PiB PET scans. All participants had clinical assessments within six months of biomarker measurement. Two-thirds of both African-American and white participants had normal cognition, with the remainder mostly CDR 0.5 or 1.

The researchers saw no differences in amyloid load by race, whether measured by PET or CSF. On MRI, however, they recorded smaller hippocampal volumes in African-Americans with a family history of dementia than in non-Hispanic whites with such history, but no differences among those without a family history. CSF total tau was lower among the African-Americans than the whites, averaging 294 pg/ml versus 443 pg/ml, and p-tau181 likewise ran low, averaging 53 pg/ml in African-Americans versus 71 pg/ml in whites. Total tau and p-tau are considered markers of neurodegeneration and toxic pathological forms of tau, respectively. The tau differences were almost entirely driven by ApoE4 carriers. Noncarriers only trended toward lower t-tau compared with whites. However, the subgroup sizes were small. Only 37 African-American carried the ApoE4 allele.

The tau findings agree with Hu and colleagues, who likewise reported lower CSF t-tau and p-tau in African-Americans (Howell et al., 2017Oct 2018 conference news). “This is an important replication study because we can now say this phenomenon is not restricted to African-Americans in the Atlanta area,” Hu wrote to Alzforum.

Intriguingly, other studies have found that the ApoE4 allele confers less AD risk on African-Americans than whites, and that seems to jibe with the lower tau levels (Farrer et al., 1997). It hints that ApoE4, which has been shown to worsen tau pathology, exerts less sway over tau in African-Americans (Sep 2017 news). “We are accumulating data that point to African-Americans responding differently to the early brain changes of Alzheimer's disease,” Hu noted.

Other researchers said it is too early to draw conclusions, pointing to limitations in the data set. There were several differences between the African-American and white cohorts, with the former consisting of a higher percentage of women and having fewer years of education, less family history of dementia, a higher average body mass index, and higher blood sugar. While adjusting for all of these factors did not change the findings, WashU researchers did not correct for other possible environmental differences or sources of bias. Jennifer Manly at Columbia University in New York noted that the study did not examine measures of socioeconomic status, access to healthcare, or race-specific stressors. “Thus, the results are subject to residual confounding,” she wrote to Alzforum.

In addition, studies like this often suffer from recruitment bias, where people who volunteer to participate are not typical of the general population. One sign this may be the case here, Manly noted, is that the WashU researchers saw no differences in cerebral ischemic lesions by race. In most population studies, these indicators of cerebrovascular disease are higher in African-Americans than age-matched whites. The limited numbers of people who chose to participate in biomarker substudies may have introduced further recruitment bias, Manly added. Brickman concurs. “It is too soon to determine whether the reported differences are due to true underlying differences, to environmental exposures, to systematic sampling bias, or to some combination of these factors,” he wrote.

Morris agrees these are valid concerns. “We will need larger samples of African-Americans to fully understand whether our findings are confirmed or refuted,” he told Alzforum. “Increasing the diversity of our research cohorts is imperative.”

If the finding of lower CSF tau holds up, it would imply the need for distinct diagnostic cutoffs in African-Americans and whites. In practice, that might be difficult to apply, because many people in the U.S. have mixed genetic heritage, Hu noted. “One solution would be to identify genetic factors that influence CSF tau values,” he suggested.

Barnes advocates for large collaborative efforts to increase minority representation in research studies. “As the field moves toward a biological definition of AD, the under-inclusion of minority populations in AD research will significantly hinder our progress as a field,” she predicted.—Madolyn Bowman Rogers


  1. Inclusion of participants from differing racial, ethnic, and cultural backgrounds in Alzheimer’s biomarker studies is critical considering the demographic make-up of the aging population. The majority of our knowledge base about Alzheimer’s biomarkers was generated from studies that included predominantly White, non-Hispanic samples. This study is important because it is among the first and largest Alzheimer’s biomarker studies to include meaningful numbers of African-Americans and explicitly examines effects of race. The findings reported certainly raise the possibility of differing biomarker profiles across racial groups but it is too soon to determine whether reported differences are due to true underlying differences, to environmental exposures, to systematic sampling bias, or to some combination of these factors. One concern I have is that recruitment strategies and participation rates were different between the two racial groups represented, which may have introduced systematic differences in biomarker profiles that are unrelated to underlying disease. It was also surprising that measures of cerebrovascular disease did not differ between racial groups, as most studies that have examined this question have reported greater prevalence in African-Americans, and it is unfortunate that the authors did not include measures of small vessel cerebrovascular disease, like white-matter hyperintensities. Future studies should be designed prospectively to examine race- and ethnicity-related differences in biomarker profiles and include analytic strategies that address potential sources of sampling bias explicitly.

    Regarding cutoff scores for diagnostic purposes, again the authors should be commended for shining more light on the possibility that the factors that contribute to Alzheimer’s disease and its clinical presentation may not be “one size fits all” for all groups and all individuals. But it is really too soon and there are certainly not enough data yet collected from community-based samples to generate true normative data that can be used to establish race-specific diagnostic cutpoints in diagnosis. The study certainly provides more evidence that race, ethnicity, and likely other demographic variables need to be considered explicitly in future work related to Alzheimer’s biomarkers.

  2. This work led by Dr. Morris is certainly very consistent with our finding, and is an important replication study because we can now say this phenomenon is not restricted to African-Americans in the Atlanta area.

    From my perspective, there are two significant implications. The first has to do with why: Do African-Americans have the same Alzheimer's disease as Caucasians (which can be true if both WashU and we shared the same sampling bias)? A different manifestation of Alzheimer's disease in the CSF? Or a different brain disease altogether? We are accumulating data that point to African-Americans responding differently to the early brain changes of Alzheimer's disease. If so, it would be pretty hard to come up with a singular trajectory for people of every genetic ancestry in this disease.

    The second has to do with the "cutoffs.” We can't easily come up with a race-specific threshold value for CSF tau because mixed genetic ancestry is common in the U.S., and I suspect that we would see the same trends in cerebral tau PET imaging. One solution is to identify genetic factors that influence CSF tau values. Alternatively, we can use a downstream marker of neuronal injury not influenced by race, but none of the candidates has very good sensitivity in Alzheimer's disease so far.

  3. I have some concerns about this study. I do not think the conclusions are supported by the results. The authors touch on several limitations in the study at the end of their paper, but in my view, these are too severe to draw conclusions about racial differences in the molecular signature of AD. I am making these comments in recognition of the significant efforts that the authors and their colleagues at the Knight ADRC have made to successfully engage and recruit African-Americans into AD research.

    Multiple potential sources of bias threaten studies on racial differences in hippocampal volume or AD biomarkers. In this study, there were substantial differences in recruitment between Whites and African-Americans. We don’t know why African-Americans were less likely to report family history of AD—is that because the White cohort was enriched for family history (by design) or because, in general, African-Americans are less likely to report family history of AD, even when it is present? Either way, this raises serious concerns that the interactions reported between race and marker levels are unfounded. One alarming sign that the recruitment bias in this sample is significant is that they found no racial differences in cerebral ischemic lesions across race. This is inconsistent with the widely known/accepted finding of a higher prevalence of cerebrovascular disease among African-Americans compared to age-matched Whites, even among those without a clinical stroke. Another alarming sign that recruitment differed across race is that APOE E4 positivity is more than twice as high among Whites in the Knight ADRC sample (41.7 percent) than in the general population of older U.S. Whites (20 percent or less), whereas the prevalence of E4 in African-American participants (45.6 percent) is not that different from other studies (e.g. 36.8 percent in Dr. Barnes’s MARS study in Chicago). Another major source of bias is that once people were recruited into the Knight ADRC, there was very low participation in CSF and PET. It is not sufficient to compare the percentage of African-Americans who agreed for each procedure to the percentage of Whites who agreed, because there may be different reasons why Blacks and Whites did or did not participate and thus the results may be biased even if the refusal rate was identical across race. In summary, the authors need to demonstrate that the White and Black participants in their study are representative of White and Black older adults who live in St. Louis, where recruitment took place.

    Beyond bias, the sample size of African-Americans is too small. Small sample sizes raise concerns about results that are non-robust and thus are less likely to be replicated. The sample sizes get even smaller when African-Americans and Whites are broken down into those with AD biomarkers, and broken down further between those with and without family history and with and without an E4 allele. No hypotheses are presented in this paper, so it is unclear whether the analyses looking at interactions between race and E4 status, and race and family history, were planned or post hoc (performed after an examination of patterns in the data).

    Even if the selection bias and sample size issues could be resolved, the investigators did not measure social forces that may account for racial differences in AD biomarkers, and thus the results are subject to residual confounding. For their main analysis, depicted in the figure, they compared African-Americans and Whites on AD biomarkers only after adjusting for E4, sex, education, CDR, BMI, and family history. It is unclear why these variables were chosen; some are confounds and some are potential mediators of race effects on health. They covaried for years of education but did not account for the fact that the school experience of African-Americans is very different than that of Whites, even if they are matched by years of education attained. They did not examine any other measures of socioeconomic status, such as occupation, income, or wealth. They did not take into account that African-Americans in their study may have grown up in very different areas of the country and thus have different social and environmental exposures. They did not measure access to health care, or whether participants could afford (or are taking prescribed) blood-pressure medications (they did not measure or report blood pressure). They did not measure financial, social, or race-specific stressors such as exposure to racial profiling or police brutality. And they did not measure or report on current neighborhood conditions that differ across race due to residential segregation.

    The concept of race for this paper was not comprehensive. The authors seem to frame their paper and their discussion around the common assumption that race is a biologically informative category, and that race differences can be attributed to differences in genetic make-up. Their assumptions are reflected in the way they used covariates and in their discussion of the results. I think that scientific inquiry focusing on racial disparities in AD would improve if researchers approached race as a socially constructed classification that is converted to biology via racism and inequality.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Alzheimer’s Researchers Seek Advice on How to Include African-Americans
  2. Do African-Americans Have More, or Different, Alzheimer’s Disease? Too Little Data to Tell
  3. ApoE4 Makes All Things Tau Worse, From Beginning to End

Paper Citations

  1. . Race modifies the relationship between cognition and Alzheimer's disease cerebrospinal fluid biomarkers. Alzheimers Res Ther. 2017 Nov 2;9(1):88. PubMed.
  2. . Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. JAMA. 1997 Oct 22-29;278(16):1349-56. PubMed.

Further Reading

Primary Papers

  1. . Assessment of Racial Disparities in Biomarkers for Alzheimer Disease. JAMA Neurol. 2019 Jan 7; PubMed.