The Role of Ethnicity in Alzheimer’s Disease: Findings From The C-PATH Online Data Repository

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia and accounts for 60–80% of all dementia cases [1]. It is estimated that approximately 35 million people lived with dementia worldwide in 2010, and this number is expected to be doubled by 2030 and then tripled by 2050 [2]. With emerging evidence of an increasing trend of AD around the world, treatment and preventative strategies must be developed, especially for groups at particularly high risk of AD.

Advancing age [3] and genetic mutations [3] are recognized risk factors. The role of ethnicity [4], family history of AD [3, 5], previous head injury [6], and the development of depression before the onset of dementia [7] are less conclusive but are considered to have a strong link to the development of AD. However, an increasing rate of AD in older populations among ethnic minorities is projected [8, 9].

Studies investigating AD by ethnicity have observed that African Americans and Cuban Hispanics only had greater risks of AD compared with Caucasians [10]. The reasons behind these ethnic differences remain unclear. Some authors believe such ethnic differences might be explained by genetic factors [12], but no known genetic factors can solely explain this distinction. Hypotheses to explain these ethnic contrasts in AD include disparities in the prevalence of co-morbidities and socioeconomic status. High blood pressure or type II diabetes are more common among older Hispanics and African Americans [12]. Furthermore, these diseases might trigger cerebrovascular dysfunction, which possibly leads to AD pathology [12]. Reduced socioeconomic status and low education achievement among minorities could also play a role [13]. The mechanisms behind the relationship between low socioeconomic status, education, and the increased risk of AD remain unclear.

Interest in early onset AD (EOAD) has increased [14, 15]. Some studies reveal that EOAD incidence is estimated to be 4–12 per 100,000 population under the age 65 years [16, 17]. The true burden of EOAD remains unclear due to major variations in study design and methodology [18]. Nevertheless, EOAD cases cause more burden to society and families due to loss of relationships and productivity in the early stages of life. Investigations in early onset dementia have shown that multiple etiologies result in early onset dementia [19]; that cerebrovascular risk factors might be important [20] and might have a more aggressive natural history [21]; and that ApoE ɛ2 might be more represented than ApoE ɛ4 [22]. There might also be an impact of cognitive reserve in EOAD and neuropsychological differences [23–26]. To our knowledge, there is no published research on the impact of ethnicity on EOAD.

This study aims to understand the possible role of ethnicity in EOAD. With valuable information from over 6,000 AD subjects who participated in 24 clinical trials (control or placebo arm) and shared their information to the C-Path Online Data Repository (CODR), this investigation explored the demographics, co-morbidities, family history, ApoE genotypes, and cognitive performances by several ethnic groups: Caucasian, Asian, African American, Hispanic, and others. The results might help to identify risk factors of EOAD, especially in minority groups.

METHODS

The CODR is composed of the placebo arm of 24 clinical trials into AD from 9 major pharmaceutical companies; the trials vary in length from 3 months to 2 years; the data is de-identified and at the time of analysis (January 2014) contained information on 6,500 subjects. There were no differences in the recruitment strategies between the different trials from which patients were incorporated into the database; there was no bias to early stages of dementia in the ethnic groups investigated. Subjects had given informed consent for entry into the clinical trials according to good clinical research practice.

The CODR is available for Critical Path Institute consortia members and qualified researchers to upload and work on valuable scientific data. Information relevant to biomarkers of drug toxicity, neurodegenerative diseases, medical history, concomitant medications, ApoE genotype, and patient-reported outcomes and cognitive assessments are stored and regularly maintained by the Coalition Against Major Diseases (CAMD) Data Working Group. In conjunction with the Clinical Data Interchange Students Consortium (CDISC), standards for collecting, sorting, and interchanging AD clinical trial data was developed. The clinical trial data has been turned into a new format, and Ephibian, an information technology solutions company, have been remapping it into a workable database. It is open to CAMD members as well as external qualified researchers with approval.

AD was diagnosed using NINCDS ADRA criteria according to McKhan [27] and these criteria were applied to people under the age of 65 years.

Information on other biomarkers of AD— such as imaging, biofluids, expanded genetics beyond ApoE status, exact name of test drug candidates from sponsor companies, and background therapies per individual cases— is, however, not available. Information collected in each trial is not necessarily consistent but factors that are of interest, including demographics (age, gender, ethnicity/race), self-reported medical history (co-morbidities and family history of AD), medications, ApoE genotypes, and cognitive performance at each visit (Mini-Mental Status Examination (MMSE) for all trials and Alzheimer’s Disease Assessment Scale-cognitive subscale (ADAS-Cog) for half of the trials).

The number of subjects in trials ranged from 57 to 719. Most subjects were from the United States (56.8%), followed by Canada (5.5%) and Germany (3.7%). There are two ethnic variables in the CODR database— ethnicity (Hispanic or Latino/not Hispanic or Latino) and race (Caucasian, Asian, Black or African American, and others). Information from these two variables was used to categorize all 6,500 subjects into following races/ethnicities: Caucasians (89.0%), Asians (4.7%), African Americans (2.7%), Hispanics (2.4%), and Native Alaskans, Native Americans, and Hawaiians (1.1%). The designation of race and ethnicity were determined by self-report and not by genetic AD mixture or skin pigmentation.

RESULTS

Table 1 presents the distribution of demographics and risk factors by different ethnicities. Of the 6,500 subjects, females accounted for 56.2% and the average age was 73.7 (SD = 8.6) years. People who had AD onset before age 65 accounted for 15.7% of the total participants. For subjects that reported a family history of AD (n = 2,523), nearly two-thirds had at least one first degree relative with AD. Approximately two-thirds of subjects in each ethnic group, apart from Native Alaskans, Americans, or Hawaiians, had at least one first degree relative with AD, making familial genetic EOAD unlikely; a finding supported by our research which suggests that most EOAD are sporadic, i.e., we define EOAD as individuals under the age of 65 years with the diagnosis of AD [5]. ApoE genotyping was available in 2,382 subjects; 56% with an ApoE ɛ4 allele and 8% of subjects with an ApoE ɛ2 allele.

The chi-square tests revealed that the distributions of age, gender, EOAD status, co-morbidities, family history of AD, and ApoE genotypes were significantly different by ethnicity. Compared with Caucasians, a greater proportion of ethnic minorities were females. The mean age of African Americans, Native Alaskans, Americans, and Hawaiians were younger than other ethnicities. Greater proportions of hypertension, high cholesterol, or type II diabetes were found among Asians, which was the lowest ethnic group reporting anxiety or depression. In contrast, Hispanics had lower proportions of hypertension, high cholesterol, and type II diabetes, which is in contrast with other studies in the literature and might relate to the selection of patients for the CODR cohort. Compared with Caucasians, Asians and Hispanics had higher numbers of at least one first degree relative with a history of AD, whereas African Americans and others had lower proportions. Nearly two-thirds (64%) of African Americans were found to have ApoE ɛ4 allele compared to 58% of Caucasians. The proportions with ApoE ɛ2 allele among African Americans, Native Alaskans, Americans, or Hawaiians was almost double that of Caucasians.

Table 2 shows the difference in distribution of ApoE genotype by ethnicity. Most subjects were found to have a combination of ApoE ɛ4 and ApoE ɛ3 (41%) or two ApoE ɛ3 alleles. A combination of two ApoE ɛ4 alleles accounted for 13% , followed by the combination of ApoE ɛ3 and ApoE ɛ2 and then ApoE ɛ4 and ApoE ɛ2, totaling 8% of subjects. The only individual homozygous for the ApoE ɛ2 allele was Caucasian. While a high proportion of Asians and Hispanics had two ApoE ɛ3 alleles, the most common combination for Caucasians and African Americans was ApoE ɛ3 and ApoE ɛ4.

The ethnic differences in ADAS-Cog and MMSE at the first 7 visits over 2 years are presented in Table 3. The trend of MMSE and ADAS-Cog were similar: While Caucasians, African Americans, and Hispanicsshowed cognitive deterioration; Asians and other minorities remain the same or improved slightly compared with baseline. The mixed model showed that the MMSE was significantly interacting with ethnicity and time (p = 0.0008), suggesting the deterioration rate in MMSE was affected by ethnicity. A similar trend was found with the ADAS-Cog, although this interaction was not statistically significant (p = 0.06) (Table 4).

The results of logistic regression revealed that African Americans, Native Americans, Alaskans, and Hawaiians had increased odds of developing EOAD compared with Caucasians (Table 5). The increased odds remained after controlling for gender, co-morbidities, revealing that African Americans had 60% higher odds (95% CI: 1.1–2.4) and Native Americans, Alaskans, and Hawaiians had double the odds (95% CI: 1.2–3.5) of developing AD before age 65 (Fig. 1).

DISCUSSION

By sharing information on 6,500 subjects with AD, CODR provides a unique opportunity to answer many unknown factors of AD. With a large sample size and a comprehensive collection of risk factors of AD, this study found significant ethnic differences in demographics, co-morbidities, family history of AD, ApoE genotypes, cognitive functions, and their deterioration rate. A greater proportion of ApoE ɛ2 allele and increased odds of EOAD were also found among African Americans, Native Americans, Alaskans, and Hawaiians. This may help understand the ethnic disparities in AD and EOAD.

CODR is the first database that collects data from multiple clinical trials around the world. Information collected is ready to be shared by pharmaceutical companies and made available for qualified researchers. The ultimate goal is to help accelerate new research on brain diseases, such as dementia. The large sample size can provide power to detect significant differences and robust results, which is often the limitation of studies in this area. There is also the potential to examine time effects due to a proportion of subjects with multiple trial visits.

This study utilized CODR data to better understand the ethnic difference in AD and EOAD, which has been prioritized in the AD research agenda [28]. The results of this study may shed light on other dementia research that is limited by sample size. Nevertheless, the interpretations of results from CODR should be cautious, as there are limitations in the CODR database. For example, the terminology between trials may not be consistent, despite standards that have been developed. Also, not every contributor supplied the key data elements, such as ApoE genotypes or family history of dementia. Furthermore, patients in clinical trials represent an accessible population for research but might not be representative of the AD community. The nature of the trials is unknown, therefore constraints might have been placed on age and cognitive scores. Other impediments include the fact that most trials require subjects to have dependable informants. Moreover, certain comorbidities, medications, behavioral complications, and severe dementia might have caused some patients to be excluded; these considerations placing caveats on our findings. The CAMD group continues working on such issues to ensure the data quality and completeness as well as improving the easy use of data. More detail can be found on the CAMD website [29].

The majority of participants were females, which reflects a recent WHO report that estimated that prevalence of AD among females is 19–29% higher than males [30]. The age of the AD cohort varies with separate studies; however, the CODR cohort tends to be younger with a higher proportion of EOAD compared with other studies [4, 32]. All participants in this study are grouped into Caucasian, Asian, African American, Hispanic, and other minorities (Native Americans, Alaskans, and Hawaiians), which are similar to other studies that explored ethnic differences [4, 11]. Although Hispanics and African Americans together only accounted for 5% of the CODR cohort, there were over 150 subjects in each ethnicity, which make comparisons statistically robust.

Most ethnic differentials found in this study are consistent with previous research. The prevalence rates of type II diabetes and hypertension were higher in African Americans and other minorities, which is similar to previous findings, showing that the hypertension and diabetes might promote the development of AD [33], particularly in African Americans and/or Hispanics [32, 35]. Nevertheless, the role of hypertension and type II diabetes in the pathway of dementia/AD remains uncertain [10]. The Hispanics in the CODR seem to have lower prevalence rates of hypertension, hypercholesterolemia, and type II diabetes compared to their counterparts. The reduced prevalence of cerebrovascular risk factors within Hispanics is probably explained by differences between Hispanic populations in vascular risk factors.[36] It is likely that the Hispanic participants in the CODR were from a relatively healthy cohort and recruited for a particular trial. All Hispanic participants had a first degree relative with AD providing some support for this hypothesis. Asian participants, on the other hand, had the highest prevalence rates of hypertension, high cholesterol, and type II diabetes but the lowest rate of anxiety or depression. These ethnic differences in co-morbidity rates were adjusted for in the logistic regression to further examine the odds of EOAD by ethnicity.

Despite the well-known association between ApoE ɛ4 presence and the risk of AD, this relationship is less evident among minorities [11]. In this study, the presence of ApoE ɛ4 and ApoE ɛ2 was significantly different by ethnicities: While African Americans had the highest proportion of ApoE ɛ4, African Americans, Native Americans, Alaskans, and Hawaiians all had higher proportions of subjects with ApoE ɛ2. As all subjects in CODR had AD, it is not possible to further examine how ApoE genotypes affect the risk of AD.

This study was able to examine the odds of EOAD by ethnicity, adjusting for gender and co-morbidities of patients, with a reasonable sample size for eachethnic group. The results revealed that, compared with Caucasians, African Americans, Native Americans, Alaskans, and Hawaiians had an increased odds ratio of developing AD before age 65. This finding might be explained by the greater proportions of African Americans, Native Americans, Alaskans, and Hawaiians with ApoE ɛ2, which is consistent with the results from previous research which suggests a possible link between ApoE ɛ2 and EOAD [37]. Although some studies reported that ApoE ɛ2 may have some protective effects on late onset AD [38, 39], interaction between ApoE ɛ2 and ethnicity may mask the true effect of ApoE ɛ2. We suggest that future studies probing the ApoE effect in AD should consider the effects of ethnicity.

This is the first study, to our knowledge, reporting cognitive assessments with time in different ethnic groups. The results of cognitive function assessments suggested that, not only did Asians have better cognitive function over time compared with other ethnicities, but also the rate of deterioration in cognition for Asians was slower. This might relate to the fact they had the lowest baseline result, and improved the most with repeat measurements. As there is no previous data comparing cognitive function assessments across several ethnicities, some ethnic differences have been seen [40]. African Americans with AD had significantly lower visual-spatial score at baseline compared with Caucasians; however, the decline in cognition overtime was not significantly affected by ethnicity [40]. Other studies noticed that educational achievement and other socioeconomic status played a significant role in cognitive deterioration in AD [13, 41]. As education and other socioeconomic status surrogates are not found in the CODR, the results are likely to be confounded. It is recommended that in future the CODR collects information on education and other socioeconomic status variables from participatingtrials.

There are limitations to this study. Quality and consistency of data collected from several trials may be problematic but can be overcome with detailed guidelines and standardized procedures to minimize errors. Incomplete information on key factors, such as ApoE genotype is not collected/reported by all trials, may be another weakness. It is estimated that 20–30% of patients diagnosed as AD and enrolled in pharmaceutical industry trials might be amyloid negative using PET amyloid imaging, suggesting that they might not have AD [42]; notwithstanding recent neuropathological studies which indicate that 25% of people diagnosed with mild to moderate AD, with a MMSE 16–26, have sparse neuritic amyloid plaques on neuropathological examination, with a higher percentage in those who are APOE ɛ4 non-carriers [43].

These findings suggest a significant proportion of individuals diagnosed with mild to moderate AD will have low levels of amyloid. Therefore, additional markers such as tau PET imaging might be necessary, in addition to amyloid imaging, for patient selection for research and treatment. Amyloid and other imaging is not currently available in the CODR database. In addition, amyloid scanning might only be useful for trials targeting the amyloid protein, with limitations as a diagnostic tool in AD. We believe the large sample size should somewhat neutralize these difficulties. Similar methodology to pool data to augment sample size and power has been employed in other studies [4]. Education and socioeconomic status surrogates, regarded as key risk factors for AD [44], are not found in the database. Reporting such information should be encouraged. Finally, as subjects were recruited from different trials and from all over the world, the differences in environment/tools/ protocols may have an impact on the results. The MMSE and the ADAS-Cog were used as tools to assess cognitive function internationally, but ethnicity and language might influence the cultural validity of our findings [45]. Therefore, caution should be used when interpreting our results. The patients used in these trials were diagnosed using NINCDS-ADRA criteria for AD and we accept the limitations of the diagnosis given cultural, linguistic, educational, and other biases.

It should be acknowledged that study subjects were recruited from clinical trials and thus the findings may not be generalizable to other populations. It has been documented that such patients (i) are usually more educated; (ii) have had other treatments; (iii) are more likely to have a family history of dementia; and (iv) are more likely to participate in therapeutic studies early in their natural history. Interpretation of findings of this study should be cautious as these inclinations might introduce biases into our results; however, a great effort was made to adjust as many risk factors as we can in the statistical modelling. The CODR is encouraged to collect more details from subjects in order to justify the findings better. Considering the unequal sample size per ethnic group may be a challenge when comparing difference across ethnicities, we used Caucasian as the reference group whenever possible which provides a robust estimate for comparison with other ethnicities. Meanwhile we propose further research into ethnicity and dementia be considered.