Meta-Analysis of Variations in Association between APOE ɛ4 and Alzheimer’s Disease and Related Dementias Across Hispanic Regions of Origin

INTRODUCTION

As a leading cause of morbidity and mortality, 12.7 million adults in America are projected to be affected by Alzheimer’s disease and related dementias (ADRD) 2050 [1].

Apolipoprotein E (APOE) ɛ4 is the strongest known genetic risk factor for ADRD, but has primarily been examined for populations with European ancestry [1]. Emerging research has shown racial and ethnic variations in the magnitude of association between the APOE ɛ4 allele and ADRD risk [1–3]. Studies researching this association among Hispanic groups within and outside of the United States (US) have produced inconsistent results. For example, among a Caribbean Hispanic population in New York City, there is a weaker association between APOE ɛ4 presence and ADRD risk with sporadic Alzheimer’s disease (AD), compared to a stronger association with late onset cases with a strong family history of AD [4]. Additionally, APOE ɛ4 has a lower prevalence and weaker association with ADRD risk among Mexican Hispanic populations as compared to non-Hispanic White populations [5, 6].

These differences in the associations between APOE ɛ4 and ADRD have been hypothesized to be related to the heterogeneity of Hispanic genetic ancestry [5, 6]. Genetic ancestry of Hispanic groups varies among Amerindian, European, and African depending on country of origin [5]. As differences in the association between APOE ɛ4 and ADRD have been established depending on African or Amerindian ancestry, there is likely a difference in the associations between APOE ɛ4 and ADRD based on regions of origin across Hispanic populations [5, 7]. However, there remain inconsistencies, as one recent study found African admixture had no effect on the association between APOE ɛ4 and dementia when comparing Non-Hispanic White, African American, Caribbean Hispanic, and Hispanic American groups [8].

Studies have also suggested an association between social and environmental lifestyle factors and ADRD risk [9, 10]. The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) [11], showed an effect of lifestyle interventions on cognitive function of Finnish older adults at risk of dementia, leading to an adapted global approach led by the Alzheimer’s Association (WW-FINGERS) [12]. Of existing studies that address the relationship between APOE ɛ4 and ADRD among Hispanic populations, most focus on specific populations in South America. There are limited studies in this area on other Hispanic groups such as Mexican Hispanic groups. Hispanic populations are heterogenous in their geographic locations and genetic ancestry, and the population of older Hispanic individuals continues to grow rapidly [13]. Thus, a comprehensive assessment of the associations between APOE ɛ4 and ADRD by Hispanic subgroup is needed.

To address this gap, this meta-analysis aimed to evaluate the variation in the association between APOE ɛ4 and ADRD risk amongst Hispanic populations by region of origin. We also explored the association between APOE ɛ4 and mild cognitive impairment (MCI), as MCI can be viewed as an early stage of the disease continuum for ADRD [1]. A better understanding of the association between APOE ɛ4 and ADRD in different Hispanic subgroups will help to improve understanding of the underlying pathology and ultimately to reduce disparities in ADRD.

METHODS

This meta-analysis is registered in PROSPERO (CRD42021221598) and was conducted according to PRISMA 2020 guidelines [14]. A comprehensive search strategy consisted of search terms related to Hispanic and Latin American origin, ADRD, cognitive decline, and APOE (Supplementary Table 1). PubMed, Embase, Scopus, and PsycInfo were chosen as the primary databases for this meta-analysis and reference lists of the included studies were hand-searched to identify additional eligible studies. Studies were also manually searched for from publications of well-known population-based ADRD studies such as the Dementia 10/66 [15]. The final search was completed on February 15, 2022 to capture any studies that were published before that date.

The following inclusion criteria was used to identify eligible studies. Studies should: 1) focus on Hispanic and Latino populations of any origin, i.e., Spanish, Caribbean Hispanic, Latin American, Mexican, Cuban, Puerto Rican, self-identified Hispanic; 2) address the presence of APOE ɛ4 allele as a genetic risk factor with a focus on dementia, ADRD, cognitive function, and/or MCI as the study outcomes; 3) specify the regional ethnic origin of the Hispanic individuals; 4) be designed as population-based studies; 5) involve an adult population (18+); 6) be published in English, and 7) be dataset based original studies.

Studies were excluded if the study: 1) did not involve Hispanic ethnic groups; 2) did not specify the ethnic region of Hispanic origin (i.e., Cuban Hispanic, Mexican Hispanic, etc.); 3) did not include data on APOE ɛ4; 4) did not specify allele/genotype frequency for each ethnic subgroup; 5) study participants were recruited through familial relation or analyses were conducted at the genomic level rather than at an individual level; 6) outcomes did not include dementia, ADRD, cognitive function, and/or MCI; 7) non-dataset based original studies (e.g., editorial, trial design paper, case reviews), systematic reviews/meta-analysis; 8) thesis and other non-peer reviewed studies; and 9) non-English study.

A dual review approach was implemented throughout the screening and data extraction [16]. Three reviewers independently screened articles and resolved any resulting conflicts. When multiple articles reported data from the same parent study/dataset, only one study with the largest sample size, the most specific breakdown of APOE and a focus on ADRD outcomes was included. The following information was extracted independently from each included study: study design, country setting, region of origin of study population, sample size, age, education, sex, frequency of APOE genotypes, frequency of population with at least one copy of APOE ɛ4, study outcomes, and outcome measures. Two reviewers independently conducted risk of bias assessments using the Newcastle-Ottawa scale presented in Table 1 [17]. Discrepancies in scores were resolved between the two reviewers.

Meta-analysis was conducted in Stata 16 SE. Odd Ratios (ORs) and 95% Confidence Interval (CI) were extracted and/or calculated for developing ADRD or MCI respectively for individuals with the APOE ɛ4 allele (ɛ4+) versus individuals without the APOE ɛ4 allele (ɛ4–) by Hispanic ethnicity region. Adjusted ORs were extracted if available and meta-analysis was conducted to calculate the pooled adjusted risk for developing ADRD. The pooled adjusted estimate for developing MCI was unable to be obtained due to small sample size. The DerSimonian-Laird random effect model was used due to high heterogeneity across studies for each region [18–20]. Q value and I² were used to assess the heterogeneity among the included studies.

RESULTS

There were 2,248 articles identified in the initial database search and one study was identified from reference lists. Of the 2,249 articles, 308 were screened for full-text eligibility. A total of 20 articles were included in this meta-analysis (Fig. 1). Among 20 selected studies, a total of 16,178 participants were included, with 6,259 cases of ADRD, 829 cases of MCI, and 9,090 cognitively normal controls.

OPEN IN VIEWER

Fig. 1

PRISMA Flow Diagram of eligible article selection from initial database searches in PubMed, Embase, Scopus, and PsycInfo and reference lists.

Six Hispanic ethnic regions were represented: South American (n = 8), Mexican Hispanic (n = 6), Cuban Hispanic (n = 4), Caribbean Hispanic (Puerto Rico and the Dominican Republic, n = 2), Spanish (n = 2), and Central American (Panama and “Central American”, n = 2). One study included data on Caribbean, South American, Central American, and Cuban regions [6]. All other studies focused on one region of origin. Within the South American region, four studies focused on a Colombian population [21–24], there was one study focused on each of the Peruvian [25], Venezuelan [26], and Argentinian [27] populations, and one study did not specify a country of focus [6]. Five studies were conducted with Hispanic groups living in the US [5, 28–30], four in Colombia [21–24], three in Mexico [31–33], two each in Cuba and Spain [34–37], and one each in Argentina, Panama, Peru, and Venezuela [25–27, 38].

Ten of the included studies are based on data from one or more large parent studies that have produced multiple publications. The represented parent studies include: Hispanic Community Health Study/Study of Latinos (HCHS/SOL) [6], Study on Aging and Dementia in Mexico (SADEM) [33], Argentinian Alzheimer’s Disease Neuroimaging Initiative (ADNI) [27], Playa Dementia and Alzheimer Study (EDAP) [35], Project FRONTIER (Facing Rural Obstacles to Health Now Through Intervention, Education & Research) [28], the Texas Alzheimer’s Research & Care Consortium (TARCC) [28], The Maracaibo Aging Study [26], Panama Aging Research Initiative (PARI) [38], 10/66 Cuba [36], the Washington Heights and Inwood Columbia Aging Project (WHICAP) [30], Northern Manhattan Study (NOMAS) [30], and Prevalence Survey of Dementia in the Mexico City Elderly Population (ESEC) [31].

Tables 2A–2C present the study characteristics and demographic information for each study by ADRD, MCI, and cognitively normal outcomes, respectively. Sample sizes varied and ranged from 18 to 2,063 among cognitively normal groups, 24 to 148 among MCI groups, and ranged from 28 to 2,451 among the groups with ADRD. Eighteen out of 20 studies reported age data. The pooled mean age was 67.72 for cognitively normal cases, 67.91 for individuals with MCI, and 76.46 for individuals with ADRD. Sixteen out of 20 studies reported data on sex. Education was reported in 13 out of the 20 studies. Three studies reported education as a categorical variable [6, 36] and 10 studies reported education in years of education. Among these 10 studies, the pooled average years of education was 9.45 for MCI (n = 7), 7.64 for cognitively normal groups (n = 7), and 5.93 for ADRD (n = 8). Seven studies did not report data on education [5, 37].

Most studies used the National Institute for Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria to assess ADRD outcomes (n = 11) [21, 38]. Other criteria used included the Diagnostic and Statistical Manual of Mental Disorders (DMS-IV-R) criteria for dementia (n = 7) [22, 36], National Institute on Aging-Alzheimer’s Association (n = 2) [27, 28], the National Institute of Neurological Disorders and Stroke Association Internationale Pour la Recherche et Enseignement en Neurosciences (NINDS-AIREN) (n = 1) [33], 10/66 dementia diagnosis algorithm (n = 1) [36], and unspecified “standardized criteria” (n = 1) [5].

To classify MCI, four studies used criteria established by Peterson and Morris [27, 34] and one study used National Institute on Aging-Alzheimer’s Association criteria [6].

The reports on frequency/proportion of APOE genotypes (ɛ2/ɛ2, ɛ2/ɛ3, ɛ2/ɛ4, ɛ3/ɛ3, ɛ3/ɛ4, ɛ4/ɛ4) and alleles (ɛ2, ɛ3, ɛ4) varied. Only six studies reported data on the frequency and/or proportion of all possible genotypes and alleles [5, 36]. Five studies reported on all genotypic frequencies and no allelic frequencies [23, 38], and two studies reported on some genotypes or some alleles but not all [25, 38], and seven studies only reported on ɛ4 data [6, 37]. As the data on the frequency of APOE ɛ4 carriers (ɛ4+) and APOE ɛ4 non-carriers (ɛ4–) were available or could be calculated for each study, this meta-analysis focuses on assessing the relationship between the presence of APOE ɛ4 and the risks of developing MCI and/or ADRD. Frequencies of ɛ4–and ɛ4+ individuals for ADRD and MCI outcomes and controls are reported by region of origin in Table 3.

APOE ɛ4 and ADRD risk

Based on unadjusted estimates, there was a statistically significant association with APOE ɛ4 presence and risk of developing ADRD for Caribbean Hispanic (OR: 1.88, 95% CI: 1.66–2.14), Central American (OR: 5.08, 95% CI: 2.01–12.87), Cuban (OR: 3.69, 95% CI: 1.86–7.33), and South American (OR: 3.51, 95% CI: 2.26–5.45) groups (Supplementary Figure 1). The association between APOE ɛ4 and ADRD was not statistically significant for the Mexican group. There was only one included study in each of the Caribbean Hispanic [30], Central American [38], and Spanish groups [37], thus the presented estimate of the APOE ɛ4/ADRD association for these groups is not robust.

Twelve studies reported adjusted OR values with common adjustments being age, sex, and education [5, 38]. Results from pooled estimates based on adjusted ORs showed that there was a statistically significant association between APOE ɛ4 and ADRD for the South American (OR: 4.61, 95% CI: 2.74–7.75) group (Fig. 2). There was not a statistically significant association between APOE ɛ4 and ADRD for the Mexican group. Adjusted estimates were not presented for the Central American, Cuban Hispanic, Caribbean Hispanic, and Spanish groups due to a lack of studies with adjusted data.

OPEN IN VIEWER

Fig. 2

Forest plot of adjusted analysis of ADRD association with APOE ɛ4. APOE ɛ4+, individuals with the APOE ɛ4 allele; APOE ɛ4–, individuals without the APOE ɛ4 allele.

Heterogeneity from adjusted analysis was not calculated for Central American and Cuban Hispanic groups, since these groups were represented by one study each. Heterogeneity was moderate for the South American group (I² = 47.72%) and substantial for the Mexican Hispanic group (I² = 67.66%).

In meta-regression for unadjusted analysis, higher mean age and more females included did not significantly increase the effect among studies focused on South American, Mexican, or Cuban Hispanic groups. For the adjusted analysis, higher mean age and more females included did not significantly increase the effect among studies focused on South American or Mexican Hispanic groups.

DISCUSSION

This meta-analysis evaluated the variation of the association between APOE ɛ4 presence and ADRD risk among Hispanic groups from six different ethnic regions of origin. There was a significant association between APOE ɛ4 presence and ADRD risk for South American groups based on adjusted analysis. In unadjusted analysis, APOE ɛ4 was significantly associated with ADRD risk for Caribbean Hispanic, Central American, Cuban, and South American groups. There was not a statistically significant association between APOE ɛ4 and ADRD risk for the Mexican group in both adjusted and unadjusted analyses. Only the Central American group showed a significant association between APOE ɛ4 presence and MCI risk.

The variation in the association of APOE ɛ4 and ADRD risk has been supported by studies evaluating the influence of genetic ancestry on the association of APOE ɛ4 with ADRD as well as MCI [6, 40].

In this meta-analysis, the South American countries included were Colombia, Argentina, Venezuela, and Peru. APOE ɛ4 presence and ADRD risk varied between among included studies of different South American countries, as well as studies within the same countries. This contrast may be attributed to differences in the ethnic makeup of the South American group, a very large and diverse region. One study focused on Caucasian Mestizos in Columbia found a strong association between APOE ɛ4 and ADRD risk [24]. Contrastingly, a different Colombian based study found no association in APOE ɛ4 presence and ADRD risk in men, but a significant association for women [21]. Similarly, one study conducted in Venezuela only found a significant association between APOE ɛ4 and AD risk for women [26]. Therefore, future studies are needed that cover a wide array of countries in South America to further reveal the association between APOE ɛ4 and AD. In this meta-analysis, while the heterogeneity of Hispanic populations is acknowledged by assessing the association between APOE ɛ4 and ADRD by region, the heterogeneity among studies within each region remains high. Heterogeneity indicated by high I² values points to large variation in sample characteristics within regions, not only across regions. This variation in the strength of APOE ɛ4 as a genetic risk factor for ADRD, suggests that in certain ethnic subgroups, there may be additional important ADRD risk factors other than APOE ɛ4, such as environmental exposures, stressful life events, nativity status, and immigration history, or how well a population is assimilated in their country.

We were not able to further test the impact of nativity for two reasons: 1) limited number of studies were conducted in the US and 2) and among those few US based studies, lack of comprehensive data to differentiate first versus second (or more) generation of immigration. Future research should consider comparing the associations in three groups (e.g., Mexicans, Mexican immigrants in the US, and Mexican Americans). Assessing outcomes by nativity status and immigration history are important areas for future research.

Although some studies adjusted for potential other risk factors, such as education, sex, and age, limited studies adjusted for lifestyle and comorbidities [23, 33]. The diagnoses of ADRD and MCI have evolved over time with the advances in disease biomarkers. Thus, more recent diagnoses of either condition are much more robust than diagnoses based on older criteria [41]. Because the contemporary diagnosis standards rest on biological evidence of disease, less diagnostic variability is observed in more recent studies. This consistency would improve comparison across different cultures and allow greater precision in inferences around the role of different regions of origin in the relationship of genetic factors to ADRD risk. Additionally, ADRD risk among Hispanic populations residing in the US may be different than Hispanics residing in their home countries due to changes in environment and lifestyle [42]. Future studies are needed to assess the possibility of environmental and lifestyle risk factors, by focusing on one group with similar environmental exposures while accounting for other risk factors as has been done in studies of ADRD risk in other sociocultural groups [43].

The variation in APOE ɛ4 and ADRD risk may also be due to additional genetic or biochemical factors associated with ADRD in Hispanic subgroups. Genetic factors other than APOE have been found to be associated with ADRD for other ethnic groups. ATP-binding cassette transporter protein (ABCA7) was found to double the risk of AD among African American individuals [44]. Mutations in genes coding for presenilin 1 (PSEN1), presenilin 2 (PSEN2), and amyloid precursor protein (APP) have been found to be risk factors for early-onset familial ADRD (EOAD) [45]. It is possible that among the regions of origin with weaker association between APOE ɛ4 and ADRD, there is a stronger association of ADRD with these genetic factors and/or higher prevalence of EOAD as compared to late-onset ADRD, which shows a higher association with APOE ɛ4 [45].

Genetic ancestry may play a role in the association between APOE ɛ4 presence and ADRD risk among Hispanic populations [6]. It has been found that Hispanic groups with higher proportions of European ancestry have a stronger association between APOE ɛ4 and ADRD as compared to those with African or Amerindian ancestry [6]. Additionally, Amerindian ancestry may have a protective effect against the effect of APOE ɛ4 on ADRD development [6]. Still, a recent study comparing Non-Hispanic White, African American, Caribbean Hispanic (Cuba, Puerto Rico, Dominican Republic, Venezuela), and Hispanic American individuals found ancestry had no effect on association between APOE ɛ4 and ADRD risk [8]. In this meta-analysis, only three of the four studies focused on Cuban Hispanic groups included information on genetic ancestry. Samples from the three studies had similar genetic ancestry makeup despite moderate heterogeneity across studies. Therefore, more research is needed to further assess the role of genetic ancestry in the association between APOE ɛ4 and ADRD risk.

There is a need for increased studies assessing APOE ɛ4 prevalence and ADRD association among Central American and Spanish populations as well as a need for greater representation of heterogenous Hispanic groups amongst studies of ADRD risk factors. Only one included study focused on ADRD within a Spanish group and one study focused on a Central American group in Panama. There may be a need for more studies conducted in Central American countries other than Panama. There were limited studies focused on Central American and Cuban Hispanic groups with more rigorous design that allowed us to account for factors associated with ADRD risk. Study quality was poorest among studies focusing on Spanish populations. There were only two included Spanish studies, both with a score of 5 out of 9. This may indicate a need for more higher-quality studies focused on Spanish populations.

Three non-English studies were excluded from this meta-analysis. Two of the non-English studies were based on Mexican populations [46, 47]. The third study was the Spanish translation of an English study that is included in this meta-analysis [27]. Thus, it is unlikely that this meta-analysis is missing significant data from studies published in Spanish. Future studies should consider incorporating data from non-English articles if applicable.

Another future research direction is a more comprehensive investigation of the association of APOE ɛ4 with MCI using data from studies that apply the new biological definitions of MCI due to AD [41]. The diagnosis of MCI is unstable when based solely on the clinical features of the disease, with many diagnosed with MCI reverting to cognitively normal [1]. Knowing the association between APOE ɛ4 and MCI due to AD may help identify individuals with a higher risk of progressing to dementia and identify factors that might mitigate disease progression. Among the 20 included studies, only seven assessed the association between APOE ɛ4 and MCI. Therefore, additional studies with diverse Hispanic groups are needed to investigate the association of APOE ɛ4 and/or other risk factors with MCI.

The APOE ɛ4 allele is significantly associated with increased ADRD risk for the South American group but not for other Hispanic subgroups. This variation suggests that there are environmental factors and/or genetic factors outside of APOE ɛ4 that contribute to the risk of ADRD among diverse Hispanic groups. A better understanding of ethnic variations in both genetic and environmental ADRD risk factors may help to reduce the globally disproportionate burden of ADRD among Hispanic populations.