Generalizability of Tau and Amyloid Plasma Biomarkers in Alzheimer's Disease Cohorts of Diverse Genetic Ancestries

Introduction

Plasma phosphorylated threonine-181 of Tau and amyloid beta are biomarkers for differential diagnosis and preclinical detection of Alzheimer disease (AD). Given differences in AD risk across diverse populations, generalizability of existing biomarker data is not assured.

Methods

In 2,086 individuals of diverse genetic ancestries (African American, Caribbean Hispanic, and Peruvians) we measured plasma pTau-181 and Aβ42/Aβ40. Differences in biomarkers between cohorts and clinical diagnosis groups and the potential discriminative performance of the two biomarkers were assessed.

Results

pTau-181 and Aβ42/Aβ40 were consistent across cohorts. Higher levels of pTau181 were associated with AD while Aβ42/Aβ40 had minimal differences. Correspondingly, pTau-181 had greater predictive value than Aβ42/Aβ40, however, the area under the curve differed between cohorts.

Discussion

pTau-181 as a plasma biomarker for clinical AD is generalizable across genetic ancestries, but predictive value may differ. Combining genomic and biomarker data from diverse individuals will increase understanding of genetic risk and refine clinical diagnoses.

Genetic analyses in multiplex families confirms chromosome 5q35 as a risk locus for Alzheimer's Disease in individuals of African Ancestry

There is a paucity of genetic studies of Alzheimer Disease (AD) in individuals of African Ancestry, despite evidence suggesting increased risk of AD in the African American (AA) population. We performed whole-genome sequencing (WGS) and multipoint linkage analyses in 51 multi-generational AA AD families ascertained through the Research in African American Alzheimer Disease Initiative (REAAADI) and the National Institute on Aging Late Onset Alzheimer's disease (NIA-LOAD) Family Based Study. Variants were prioritized on minor allele frequency (<0.01), functional potential of coding and noncoding variants, co-segregation with AD and presence in multi-ancestry ADSP release 3 WGS data. We identified a significant linkage signal on chromosome 5q35 (HLOD=3.3) driven by nine families. Haplotype segregation analysis in the family with highest LOD score identified a 3'UTR variant in INSYN2B with the most functional evidence. Four other linked AA families harbor within-family shared variants located in INSYN2B's promoter or enhancer regions. This AA family-based finding shows the importance of diversifying population-level genetic data to better understand the genetic determinants of AD on a global scale.

African Ancestry Individuals with Higher Educational Attainment Are Resilient to Alzheimer's Disease Measured by pTau181

Background

Cognitive and functional abilities in individuals with Alzheimer's disease (AD) pathology (ADP) are highly variable. Factors contributing to this variability are not well understood. Previous research indicates that higher educational attainment (EA) correlates with reduced cognitive impairments among those with ADP. While cognitive and functional impairments are correlated, they are distinguishable in their manifestations.

Objective

To investigate whether levels of education are associated with functional impairments among those with ADP.

Methods

This research involved 410 African American (AA) individuals (Institutional Review Boards 20070307, 01/27/2023) to ascertain whether EA correlates with functional resilience and if this effect varies between APOE ɛ4 carriers and non-carriers. Utilizing EA as a cognitive reserve proxy, CDR-FUNC as a functional difficulties measure, and blood pTau181 as an ADP proxy, the non-parametric Mann-Whitney U test assessed the relationship between EA and CDR-FUNC in individuals with advanced pTau181 levels.

Results

The results showed that EA correlated with functional difficulties in AA individuals with high levels of pTau181, such that individuals with high EA are more likely to have better functional ability compared to those with lower EA (W = 730.5, p = 0.0007). Additionally, we found that the effect of high EA on functional resilience was stronger in ɛ4 non-carriers compared to ɛ4 carriers (W = 555.5, p = 0.022).

Conclusion

This study extends the role of cognitive reserve and EA to functional performance showing that cognitive reserve influences the association between ADP burden and functional difficulties. Interestingly, this protective effect seems less pronounced in carriers of the strong genetic risk allele ɛ4.

Attitudes and Perceptions about Brain Donation Among African Americans: Implications for Recruitment into Alzheimer's Disease Research

The objective of this study was to investigate attitudes toward brain donation and perceptions of medical research that influence brain donation among African Americans. Cross-sectional surveys were administered to African American community members (n = 227). Findings indicate that only 27% of respondents were willing to donate their brain. As medical mistrust was not found to be a significant barrier to research participation, there may be opportunity to increase brain donation by providing information about Alzheimer's disease and brain donation to potential donors and their families so that informed decisions about participating in research can be made.

Genetic analysis of cognitive preservation in the midwestern Amish reveals a novel locus on chromosome 2

INTRODUCTION

Alzheimer disease (AD) remains a debilitating condition with limited treatments and additional therapeutic targets needed. Identifying AD protective genetic loci may identify new targets and accelerate identification of therapeutic treatments. We examined a founder population to identify loci associated with cognitive preservation into advanced age.

METHODS

Genome-wide association and linkage analyses were performed on 946 examined and sampled Amish individuals, aged 76-95, who were either cognitively unimpaired (CU) or impaired (CI).

RESULTS

12 SNPs demonstrated suggestive association (P≤5×10-4) with cognitive preservation. Genetic linkage analyses identified >100 significant (LOD≥3.3) SNPs, some which overlapped with the association results. Only one locus on chromosome 2 retained significance across multiple analyses.

DISCUSSION

A novel significant result for cognitive preservation on chromosome 2 includes the genes LRRTM4 and CTNNA2. Additionally, the lead SNP, rs1402906, impacts the POU3F2 transcription factor binding affinity, which regulates LRRTM4 and CTNNA2.

An Alzheimer's disease risk variant in TTC3 modifies the actin cytoskeleton organization and the PI3K-Akt signaling pathway in iPSC-derived forebrain neurons

A missense variant in the tetratricopeptide repeat domain 3 (TTC3) gene (rs377155188, p.S1038C, NM_003316.4:c 0.3113C>G) was found to segregate with disease in a multigenerational family with late-onset Alzheimer's disease. This variant was introduced into induced pluripotent stem cells (iPSCs) derived from a cognitively intact individual using CRISPR genome editing, and the resulting isogenic pair of iPSC lines was differentiated into cortical neurons. Transcriptome analysis showed an enrichment for genes involved in axon guidance, regulation of actin cytoskeleton, and GABAergic synapse. Functional analysis showed that the TTC3 p.S1038C iPSC-derived neuronal progenitor cells had altered 3-dimensional morphology and increased migration, while the corresponding neurons had longer neurites, increased branch points, and altered expression levels of synaptic proteins. Pharmacological treatment with small molecules that target the actin cytoskeleton could revert many of these cellular phenotypes, suggesting a central role for actin in mediating the cellular phenotypes associated with the TTC3 p.S1038C variant.

Founder population-specific weights yield improvements in performance of polygenic risk scores for Alzheimer disease in the Midwestern Amish

Alzheimer disease (AD) is the most common type of dementia and is estimated to affect 6 million Americans. Risk for AD is multifactorial, including both genetic and environmental risk factors. AD genomic research has generally focused on identification of risk variants. Using this information, polygenic risk scores (PRSs) can be calculated to quantify an individual's relative disease risk due to genetic factors. The Amish are a founder population descended from German and Swiss Anabaptist immigrants. They experienced a genetic bottleneck after arrival in the United States, making their genetic architecture different from the broader European ancestry population. Prior work has demonstrated the lack of transferability of PRSs across populations. Here, we compared the performance of PRSs derived from genome-wide association studies (GWASs) of Amish individuals to those derived from a large European ancestry GWAS. Participants were screened for cognitive impairment with further evaluation for AD. Genotype data were imputed after collection via Illumina genotyping arrays. The Amish individuals were split into two groups based on the primary site of recruitment. For each group, GWAS was conducted with account for relatedness and adjustment for covariates. PRSs were then calculated using weights from the other Amish group. PRS models were evaluated with and without covariates. The Amish-derived PRSs distinguished between dementia status better than the European-derived PRS in our Amish populations and demonstrated performance improvements despite a smaller training sample size. This work highlighted considerations for AD PRS usage in populations that cannot be adequately described by basic race/ethnicity or ancestry classifications.

African ancestry APOE e4 non-carriers with higher educational attainment are resilient to Alzheimer disease pathology-specific blood biomarker pTau181

Cognitive and functional abilities in individuals with Alzheimer disease (AD) pathology (ADP) show greater than expected variability. While most individuals show substantial impairments in these abilities, a considerable number show little or no impairments. Factors contributing to this variability are not well understood. For instance, multiple studies have shown that higher levels of education are associated with reduced cognitive impairments among those with ADP. However, it remains unclear whether higher levels of education are associated with functional impairments among those with ADP. We studied 410 AA individuals with advanced levels of pTau181 (a biomarker for ADP; individuals as those having log 10 (pTau181) level greater than one standard deviation above the mean) to determine whether EA (categorized as low EA for individuals with ≤ 8 years of education and high EA for those with >8 years) promotes functional resilience and whether this effect varies between APOE ε4 carriers and non-carriers. We used the four non-memory components of the Clinical Dementia Rating (CDR) to create a composite score (CDR-FUNC) to evaluate functional difficulties (scored from 0=no impairment to 12=severe). We employed the non-parametric Mann-Whitney U test to assess the relationship between EA and CDR-FUNC in advanced levels of pTau181 individuals. The results showed that EA promotes resilience to functional problems in AA individuals with advanced levels of pTau181, such that individuals with high EA are more likely to have better functional ability compared to those with lower EA (W=730.5, p=0.0007). Additionally, we found that the effect of high EA on functional resilience was stronger in ε4 non-carriers compared to ε4 carriers (W=555.5, p=0.022). This study extends the role of cognitive reserve and EA to functional performance showing that cognitive reserve influences the association between ADP burden and functional difficulties. Interestingly, this protective effect seems less pronounced in carriers of the strong genetic risk allele ε4. The results highlight the intricate interplay of genetic and non-genetic factors in AD progression, suggesting a need for more personalized strategies to manage functional decline in AD.

An Alzheimer's disease risk variant in TTC3 modifies the actin cytoskeleton organization and the PI3K-Akt signaling pathway in iPSC-derived forebrain neurons

A missense variant in the tetratricopeptide repeat domain 3 ( TTC3 ) gene (rs377155188, p.S1038C, NM_003316.4:c.3113C>G) was found to segregate with disease in a multigenerational family with late onset Alzheimer's disease. This variant was introduced into induced pluripotent stem cells (iPSCs) derived from a cognitively intact individual using CRISPR genome editing and the resulting isogenic pair of iPSC lines were differentiated into cortical neurons. Transcriptome analysis showed an enrichment for genes involved in axon guidance, regulation of actin cytoskeleton, and GABAergic synapse. Functional analysis showed that the TTC3 p.S1038C iPSC-derived neuronal progenitor cells had altered 3D morphology and increased migration, while the corresponding neurons had longer neurites, increased branch points, and altered expression levels of synaptic proteins. Pharmacological treatment with small molecules that target the actin cytoskeleton could revert many of these cellular phenotypes, suggesting a central role for actin in mediating the cellular phenotypes associated with the TTC3 p.S1038C variant.

Highlights

The AD risk variant TTC3 p.S1038C reduces the expression levels of TTC3 The variant modifies the expression of AD specific genes BACE1 , INPP5F , and UNC5C Neurons with the variant are enriched for genes in the PI3K-Akt pathwayiPSC-derived neurons with the alteration have increased neurite length and branchingThe variant interferes with actin cytoskeleton and is ameliorated by Cytochalasin D.

Genetic variants in the SHISA6 gene are associated with delayed cognitive impairment in two family datasets

INTRODUCTION

Studies of cognitive impairment (CI) in Amish communities have identified sibships containing CI and cognitively unimpaired (CU) individuals. We hypothesize that CU individuals may carry protective alleles delaying age at onset (AAO) of CI.

METHODS

A total of 1522 individuals screened for CI were genotyped. The outcome studied was AAO for CI individuals or age at last normal exam for CU individuals. Cox mixed-effects models examined association between age and single nucleotide variants (SNVs).

RESULTS

Three SNVs were significantly associated (P < 5 × 10-8 ) with AAO on chromosomes 6 (rs14538074; hazard ratio [HR] = 3.35), 9 (rs534551495; HR = 2.82), and 17 (rs146729640; HR = 6.38). The chromosome 17 association was replicated in the independent National Institute on Aging Genetics Initiative for Late-Onset Alzheimer's Disease dataset.

DISCUSSION

The replicated genome-wide significant association with AAO on chromosome 17 is located in the SHISA6 gene, which is involved in post-synaptic transmission in the hippocampus and is a biologically plausible candidate gene for Alzheimer's disease.

Genetic architecture of RNA editing regulation in Alzheimer's disease across diverse ancestral populations

Most Alzheimer's disease (AD)-associated genetic variants do not change protein coding sequence and thus likely exert their effects through regulatory mechanisms. RNA editing, the post-transcriptional modification of RNA bases, is a regulatory feature that is altered in AD patients that differs across ancestral backgrounds. Editing QTLs (edQTLs) are DNA variants that influence the level of RNA editing at a specific site. To study the relationship of DNA variants genome-wide, and particularly in AD-associated loci, with RNA editing, we performed edQTL analyses in self-reported individuals of African American (AF) or White (EU) race with corresponding global genetic ancestry averaging 82.2% African ancestry (AF) and 96.8% European global ancestry (EU) in the two groups, respectively. We used whole-genome genotyping array and RNA sequencing data from peripheral blood of 216 AD cases and 212 age-matched, cognitively intact controls. We identified 2144 edQTLs in AF and 3579 in EU, of which 1236 were found in both groups. Among these, edQTLs in linkage disequilibrium (r2 > 0.5) with AD-associated genetic variants in the SORL1, SPI1 and HLA-DRB1 loci were associated with sites that were differentially edited between AD cases and controls. While there is some shared RNA editing regulatory architecture, most edQTLs had distinct effects on the rate of RNA editing in different ancestral populations suggesting a complex architecture of RNA editing regulation. Altered RNA editing may be one possible mechanism for the functional effect of AD-associated variants and may contribute to observed differences in the genetic etiology of AD between ancestries.

Consequences of a Rare Complement Factor H Variant for Age-Related Macular Degeneration in the Amish

Purpose

Genetic variants in the complement factor H gene (CFH) have been consistently implicated in age-related macular degeneration (AMD) risk. However, their functional effects are not fully characterized. We previously identified a rare, AMD-associated variant in CFH (P503A, rs570523689) in 19 Amish individuals, but its functional consequences were not investigated.

Methods

We performed genotyping for CFH P503A in 1326 Amish individuals to identify additional risk allele carriers. We examined differences for age at AMD diagnosis between carriers and noncarriers. In blood samples from risk allele carriers and noncarriers, we quantified (i) CFH RNA expression, (ii) CFH protein expression, and (iii) C-reactive protein (CRP) expression. Potential changes to the CFH protein structure were interrogated computationally with Phyre² and Chimera software programs.

Results

We identified 39 additional carriers from Amish communities in Ohio and Indiana. On average, carriers were younger than noncarriers at AMD diagnosis, but this difference was not significant. CFH transcript and protein levels in blood samples from Amish carriers and noncarriers were also not significantly different. CRP levels were also comparable in plasma samples from carriers and noncarriers. Computational protein modeling showed slight changes in the CFH protein conformation that were predicted to alter interactions between the CFH 503 residue and other neighboring residues.

Conclusions

In total, we have identified 58 risk allele carriers for CFH P503A in the Ohio and Indiana Amish. Although we did not detect significant differences in age at AMD diagnosis or expression levels of CFH in blood samples from carriers and noncarriers, we observed modest structural changes to the CFH protein through in silico modeling. Based on our functional and computational observations, we hypothesize that CFH P503A may affect CFH binding or function rather than expression, which would require additional research to confirm.

The genetic architecture of Alzheimer disease risk in the Ohio and Indiana Amish

Alzheimer disease (AD) is the most common type of dementia and is currently estimated to affect 6.2 million Americans. It ranks as the sixth leading cause of death in the United States, and the proportion of deaths due to AD has been increasing since 2000, while the proportion of many other leading causes of deaths have decreased or remained constant. The risk for AD is multifactorial, including genetic and environmental risk factors. Although APOE ε4 remains the largest genetic risk factor for AD, more than 26 other loci have been associated with AD risk. Here, we recruited Amish adults from Ohio and Indiana to investigate AD risk and protective genetic effects. As a founder population that typically practices endogamy, variants that are rare in the general population may be of a higher frequency in the Amish population. Since the Amish have a slightly lower incidence and later age of onset of disease, they represent an excellent and unique population for research on protective genetic variants. We compared AD risk in the Amish and to a non-Amish population through APOE genotype, a non-APOE genetic risk score of genome-wide significant variants, and a non-APOE polygenic risk score considering all of the variants. Our results highlight the lesser relative impact of APOE and differing genetic architecture of AD risk in the Amish compared to a non-Amish, general European ancestry population.

Leveraging videoconferencing supports the continuity of Alzheimer research during the COVID‐19 pandemic and beyond

Background

The COVID‐19 pandemic has placed a demand on researchers to limit in‐person contact with participants, greatly impacting Alzheimer Disease (AD) research. To address this problem, we describe here an approach to using digital technology to continue nationwide clinical recruitment and ascertainment of biological samples while adhering to COVID‐19 guidelines and travel restrictions.

Method

To accomplish this, we considered a videoconferencing approach for remote delivery of cognitive assessments. A multi‐site panel of neurologists and clinical psychologists and a detailed literature review ensured a protocol that captures the best‐practices for administering assessments through videoconferencing while ensuring consistency between remote and in‐person administration. Clinical coordinators underwent training to ensure good agreement with in‐person administration. Most aspects of the cognitive assessments easily transferred to videoconferencing, though Trail Making A and B, and Digit Symbol‐Coding were removed to protect the integrity of the evaluation. Additionally, we coordinated the collection of biological samples with a national company, Quest Diagnostics, to provide phlebotomy services at the participant’s residence. All protocols were developed under the guidance of the Human Subjects Research Office at the University of Miami and with approval of local IRB.

Result

Our clinical coordinators completed over two dozen remote assessments using these protocols. Both cases and controls were enrolled, across various ethnic populations within our study. The distributions of age and 3MS were similar between in‐person and remote assessments. The uptake of videoconferencing enrollment varied among the age groups, level of impairment, at‐home support system and telemedicine readiness. For example, earlier‐onset groups had the best uptake, while older‐onset groups showed the least uptake due to a higher prevalence of telemedicine unreadiness (Lam et al., 2020; Bossen et al.,2015).

Conclusion

This study demonstrates that remote enrollment and ascertainment of biological samples through videoconferencing and partnering with national mobile phlebotomy services is feasible. This approach allows researchers to continue ascertainment efforts while maintaining their participants’ autonomy through informed consent and privacy throughout the process and minimizing their exposure to COVID‐19.

Outreach and recruitment of African Americans for Alzheimer’s disease studies during the COVID‐19 pandemic

Background

Recruiting African Americans in AD studies remains a challenge, particularly during a pandemic, where major health disparities in this population are illuminated. The recruitment literature suggests myriad factors that contribute to the underrepresentation of AAs, including, but not limited to “mistrust” in researchers and their institutions. Maintaining a continuous presence in the AA community builds trust even when traditional outreach methods are not allowed. We continued to provide outreach and recruitment opportunities through COVID education and food for families as we educated them about AD, and opportunities for study participation.

Method

While our traditional outreach methods for recruiting AAs were interrupted, we continued to conduct AD outreach using virtual platforms, mobile phone calls, family conference calls and food distributions. We hosted nine webinars on COVID‐19 to maintain a presence in local and national AA communities and to remain connected to existing AD participants. We reached over 160,000 persons through webinars and social media. We established new relationship new faith leaders in the AA community who co‐hosted COVID‐19 webinars and also expressed interest in forming partnerships on AD education. In addition, we hosted food drives in AA communities that not only addressed food insecurity and COVID prevention, but also AD education and AD research opportunities. At the food drives we distributed bags with masks, hand sanitizers, AD brochures, booklets and study participation information.

Result

Between October and December of 2020, 64 AAs who attended food drives expressed interest in AD studies that required blood draws and cognitive testing. Fifteen enrolled in our genetic study, 15 requested additional follow‐up and 13 expressed interest in participating in more than one study.

Conclusion

Prior research suggests that recruiting AAs into AD studies requires continuous engagement. We used multiple strategies to maintain contact with the AA community and existing research participants, and successfully increased enrollment in the last quarter of the year. Maintaining consistent and continuous engagement facilitates trustworthiness with AAs and yields positive recruitment outcomes, even in a pandemic where traditional recruitment methods are limited.

Dissecting the role of Amerindian genetic ancestry and the ApoE ε4 allele on Alzheimer disease in an admixed Peruvian population

Alzheimer disease (AD) is the leading cause of dementia in the elderly and occurs in all ethnic and racial groups. The apolipoprotein E (ApoE) ε4 is the most significant genetic risk factor for late-onset AD and shows the strongest effect among East Asian populations followed by non-Hispanic white populations and has a relatively lower effect in African descent populations. Admixture analysis in the African American and Puerto Rican populations showed that the variation in ε4 risk is correlated with the genetic ancestral background local to the ApoE gene. Native American populations are substantially underrepresented in AD genetic studies. The Peruvian population with up to ~80 of Amerindian (AI) ancestry provides a unique opportunity to assess the role of AI ancestry in AD. In this study, we assess the effect of the ApoE ε4 allele on AD in the Peruvian population. A total of 79 AD cases and 128 unrelated cognitive healthy controls from Peruvian population were included in the study. Genome-wide genotyping was performed using the Illumina Global screening array v2.0. Global ancestry and local ancestry analyses were assessed. The effect of the ApoE ε4 allele on AD was tested using a logistic regression model by adjusting for age, gender, and population substructure (first 3 principal components). Results showed that the genetic ancestry surrounding the ApoE gene is predominantly AI (60.6%) and the ε4 allele is significantly associated with increased risk of AD in the Peruvian population (odds ratio = 5.02, confidence interval: 2.3-12.5, p-value = 2e-4). Our results showed that the risk for AD from ApoE ε4 in Peruvians is higher than we have observed in non-Hispanic white populations. Given the high admixture of AI ancestry in the Peruvian population, it suggests that the AI genetic ancestry local to the ApoE gene is contributing to a strong risk for AD in ε4 carriers. Our data also support the findings of an interaction between the genetic risk allele ApoE ε4 and the ancestral backgrounds located around the genomic region of ApoE gene.

Linkage of Alzheimer disease families with Puerto Rican ancestry identifies a chromosome 9 locus

The genetic admixture of Caribbean Hispanics provides an opportunity to discover novel genetic factors in Alzheimer disease (AD). We sought to identify genetic variants for AD through a family-based design using the Puerto Rican (PR) Alzheimer Disease Initiative (PRADI). Whole-genome sequencing (WGS) and parametric linkage analysis were performed for 100 individuals from 23 multiplex PRADI families. Variants were prioritized by minor allele frequency (<0.01), functional potential [combined annotation dependent depletion score (CADD) >10], and co-segregation with AD. Variants were further ranked using an independent PR case-control WGS dataset (PR10/66). A genome-wide significant linkage peak was found in 9p21 with a heterogeneity logarithm of the odds score (HLOD) >5.1, which overlaps with an AD linkage region from two published independent studies. The region harbors C9orf72, but no expanded repeats were observed in the families. Seven variants prioritized by the PRADI families also displayed evidence for association in the PR10/66 (p < 0.05), including a missense variant in UNC13B. Our study demonstrated the importance of family-based design and WGS in genetic study of AD.

Lower Levels of Education Are Associated with Cognitive Impairment in the Old Order Amish

BACKGROUND

Lower education has been reported to be associated with dementia. However, many studies have been done in settings where 12 years of formal education is the standard. Formal schooling in the Old Order Amish communities (OOA) ends at 8th grade which, along with their genetic homogeneity, makes it an interesting population to study the effect of education on cognitive impairment.

OBJECTIVE

The objective of this study was to examine the association of education with cognitive function in individuals from the OOA. We hypothesized that small differences in educational attainment at lower levels of formal education were associated with risk for cognitive impairment.

METHODS

Data of 2,426 individuals from the OOA aged 54-99 were analyzed. The Modified Mini-Mental State Examination (3MS-R) was used to classify participants as CI or normal. Individuals were classified into three education categories: <8, 8, and >8 years of education. To measure the association of education with cognitive status, a logistic regression model was performed adding age and sex as covariates.

RESULTS

Our results showed that individuals who attained lowest levels of education (<8 and 8) had a higher probability of becoming cognitvely impaired compared with people attending >8 years (OR = 2.96 and 1.85).

CONCLUSION

Even within a setting of low levels of formal education, small differences in educational attainment can still be associated with the risk of cognitive impairment. Given the homogeneity of the OOA, these results are less likely to be biased by differences in socioeconomic backgrounds.