Last month, results from a widely publicized study¹ seemed to spell doom for millions of individuals in the US and around the world. Authors Fortea et al. reported that having two copies of the APOE-ε4 gene variant, long known to be associated with increased risk of Alzheimer’s disease (AD), instead represented a form of genetically determined AD – meaning that, for the roughly 2% of the population carrying this genotype, development of the disease is not only more likely, it is nearly guaranteed.
The grim news has naturally been met with alarm and despair, but is the hopelessness justified? Or is this yet another case of bold claims and flashy headlines without sufficient scientific validation?
A brief refresher on AD genetics and the APOE gene
Though we have previously discussed APOE gene variants and AD risk in detail on The Drive, a few overview points are worth reiterating in order to provide context for Fortea et al.’s study.
Genetics have long been known to affect AD risk, but different genes vary in their relative impact. For certain key genes (most notably, PSEN1, PSEN2, and APP), mutations are virtually deterministic – conferring a nearly 100% chance of developing AD, typically at a relatively young age. However, such highly penetrant mutations are rare, accounting for only ~1% of all AD cases combined.
More often, genetics plays a subtler role, with more common genetic variants imparting modest increases or decreases in risk. Of these non-deterministic genes, APOE is recognized as having the strongest association with AD risk. Three different variants of this gene (known as “alleles”) exist in the human population – ε2, ε3, and ε4 – and of these, the ε4 allele is associated with the highest risk of developing AD. However, each of us has two copies of the APOE gene (making for six possible combinations of alleles), and while having one copy of the ε4 allele is associated with a moderate elevation in risk relative to other genotypes (about 2 to 3x), having two copies raises risk far more dramatically (about 8 to 10x). Yet even among these APOE-ε4/ε4 individuals, development of Alzheimer’s disease is not regarded as genetically determined – in other words, it is far from inevitable, with prior work showing that risk associated with this genotype varies based on several factors including biological sex, ethnicity, environment, and lifestyle. At least, such was the mindset prior to Fortea et al.’s recent publication.
About the study
The investigators set out to assess whether an APOE-ε4/ε4 genotype might indeed constitute a highly penetrant, genetically determined form of Alzheimer’s disease. They utilized postmortem data from 3,297 brain donors from the National Alzheimer’s Coordinating Center (NACC), as well as clinical data – including data on AD biomarkers and on diagnoses of mild cognitive impairment, dementia, and AD – from 10,039 living individuals across five large multicenter cohorts. The researchers used this information to evaluate the prevalence and progression of AD among APOE-ε4/ε4 individuals, principally comparing against APOE-ε3/ε3 individuals (the most common APOE genotype, accounting for an estimated 50-55% of the overall population).
Results indicated that the penetrance of the APOE-ε4/ε4 genotype may be close to 100%, far greater than previously thought. The authors reported that “nearly all” postmortem APOE-ε4/ε4 brains demonstrated AD-related neuropathological changes classified as either “high” or “intermediate.” In living subjects, data on abnormal amyloid and tau levels in cerebrospinal fluid indicated that levels of both biomarkers were consistently higher among ε4/ε4 individuals than among ε3/ε3s starting around age 55. Further, APOE-ε4/ε4 individuals were found to reach diagnostic milestones 7–10 years earlier than APOE-ε3/ε3 individuals, with ε4/ε4s experiencing early AD symptoms at an average age of 65.6, mild cognitive impairment at age 71.8, dementia at age 73.6, and death at age 77.2.
Together, these results suggest that APOE-ε4/ε4 is highly penetrant for AD – to the point that it might be regarded as deterministic rather than as a mere risk factor – and that it is associated with earlier disease onset than cases of AD occuring in the context of other genotypes.
What do these results add to existing knowledge?
As explained earlier, the APOE-ε4 allele was already known to confer an elevated risk of AD in a dose-dependent manner, with the APOE-ε4/ε4 genotype associated with far greater risk than any other APOE genotype. Previous research has also already established that disease onset occurs at a younger age, on average, among APOE-ε4/ε4 individuals than among those with other genotypes.² So what new information does Fortea et al.’s study provide?
The novel finding of this study is that the APOE-ε4/ε4 genotype demonstrates near-complete penetrance for AD and should thus be considered to be a genetically determined form of the disease rather than simply a risk factor. Indeed, if we take these results at face value, AD would be far more heritable than previously thought, and AD caused by the APOE-ε4/ε4 genotype would represent not only the most common form of genetically determined AD, but one of the most common Mendelian genetic diseases in general.
If we take these results at face value.
There’s more to AD than biomarkers
A strength of this study is its reliance on multiple readouts; postmortem neuropathology assessments, AD biomarker data from living subjects, and data on cognitive symptoms were all consistent in pointing toward the authors’ conclusion that APOE-ε4/ε4 is deterministic for AD. But importantly, these lines of evidence were all independent. Neuropathology data from a given brain might suggest the presence of AD, but that does not guarantee that the patient in question had ever exhibited clinical symptoms or had an AD diagnosis. Likewise, we have no indication that higher levels of AD biomarkers among APOE-ε4/ε4 individuals corresponded to greater cognitive deficits among the same individuals.
AD is a clinico-pathological diagnosis, meaning that it requires both positive results from biomarker tests and the presence of cognitive symptoms. While the investigators provide data on age of onset for cognitive symptoms and AD diagnoses, they do not share information on the proportion of individuals who met these clinical milestones. In other words, their conclusions regarding prevalence derive from neuropathology and biomarker data alone, which are notoriously imperfect correlates of functional cognitive decline.
A case study in sampling bias
Further, even if we do assume alignment across the various lines of evidence, consistency in results should not be confused for accuracy in results, as systematic errors can influence multiple data streams in similar ways. In the case of the Fortea et al. study, systematic error took the form of heavy sampling bias in the datasets used for their various analyses.
Imagine walking into an AD support group meeting, taking a quick poll, and finding that everyone present has been diagnosed with AD. Would you then conclude that 100% of the world’s population has AD? Of course not. And yet, when a similar process and logic was applied by Fortea et al., the news was published in Nature Medicine and broadcast across mainstream media.
Postmortem data were obtained from the NACC, while clinical data from were obtained from five separate cohorts: Alfa plus (Alzheimer’s and Families cohort), the A4 clinical trial, OASIS (Open Access Series of Imaging Studies), the Alzheimer’s Disease Neuroimaging Initiative, and the Wisconsin Registry for Alzheimer’s Prevention. All six of these data sources were aimed at studying dementia or AD specifically, and many targeted participants at higher than average AD risk independent of APOE genotype (e.g., those with evidence of amyloid plaques or cognitive impairment at baseline). Thus, we can safely assume that each of these cohorts included a much higher proportion of individuals with AD or at high risk of AD than would be found in the overall population, and while such cohorts have utility for studying biomarkers, progression, treatments, and countless other aspects of the disease, extrapolating their data to determine prevalence of AD in the general population is as meaningless as our hypothetical AD support group analysis.
Indeed, the NACC specifically states on its website that their data “are not suited for an analysis of dementia incidence or prevalence” because their sampling (which was based on volunteers and referrals) cannot be assumed to represent the broader population. In a cohort intended for Alzheimer’s research, who do you expect is most likely to be among those volunteers and referrals?
For those who need further convincing, we need only look at Fortea et al.’s results from APOE-ε3/ε3 individuals. The authors report that 50% of ε3/ε3 postmortem brains had high or intermediate levels of AD-related neuropathology changes, and 54% of ε3/ε3s in the pooled clinical cohorts had been diagnosed with AD dementia. If we extrapolate these data to the general population, we would need to conclude that the vast majority of humans on the planet have at least a 1-in-2 chance of AD, which very clearly is not the case!
Frankly, the fact that this enormous flaw was overlooked by the journal, the popular press, and the study authors themselves is nothing short of shameful and irresponsible. The result of this oversight has been undue emotional stress for those who have (or care about someone who has) an APOE-ε4/ε4 genotype – and perhaps, an unjustified loss of faith in risk mitigation strategies (discussed in detail in AMA #46) for which we have prior evidence of efficacy.
No need for hopelessness
This study appears to paint a bleak picture in which Alzheimer’s disease is an unavoidable fate for individuals with two copies of the APOE-ε4 allele, but fortunately, closer inspection reveals that this hopelessness is unfounded. In actuality, we can blame biased sample selection across all included cohorts for vastly inflating the apparent risk of AD associated with this genotype – and indeed, with any genotype – well beyond what would be seen in a general population.
It’s worth noting, however, that this sampling bias does not necessarily invalidate the within-cohort comparisons between those with APOE-ε4/ε4 genotypes versus those with APOE-ε3/ε3 – i.e., the results showing earlier disease development among the former and a general pattern of higher prevalence (though we shouldn’t put any stock in the exact ratio of risk between genotypes as reported here). Our critique of Fortea et al.’s findings with regard to overall AD prevalence among ε4/ε4s thus should not be interpreted as any indication that this genotype isn’t a significant risk factor for the disease – but again, we already had ample evidence that this was the case.
Yet “elevated risk” is a far cry from “guarantee” – leaving plenty of room for each of us to influence our own fate through other risk-mitigation strategies such as maintaining metabolic and cardiovascular health. Previous work has shown such strategies to be effective irrespective of APOE genotype, further demonstrating that AD is far from a sealed fate for APOE-ε4/ε4s. So to the millions who carry this high-risk genotype, there’s no reason to lose hope, but there’s every reason to take action toward cognitive and overall health.
For a list of all previous weekly emails, click here.
References
1. Fortea J, Pegueroles J, Alcolea D, et al. APOE4 homozygozity represents a distinct genetic form of Alzheimer’s disease. Nat Med. 2024;30(5):1284-1291. doi:10.1038/s41591-024-02931-w
2. Locke PA, Conneally PM, Tanzi RE, Gusella JF, Haines JL. Apolipoprotein E4 allele and Alzheimer disease: examination of allelic association and effect on age at onset in both early- and late-onset cases. Genet Epidemiol. 1995;12(1):83-92. doi:10.1002/gepi.1370120108
