Reaching the Limits of Genome-wide Significance in Alzheimer Disease: Title and subTitle BreakBack to the Environment

In this issue of JAMA, Seshadri et al¹ present a 3-stage approach to genome-wide association studies (GWAS) involving more than 35 000 individuals to identify novel genes for late-onset Alzheimer disease (AD). As has been the case for an increasing number of common, complex diseases, results for a few new loci reached genome-wide significance and other previously reported associations were replicated. The authors also addressed the contributions of these genes to disease risk prediction and conclude, perhaps not surprisingly, that the loci did not significantly improve AD risk prediction. Nevertheless, Seshadri et al point out that the results implicate biological pathways that may provide important targets for interventions.

Seshadri et al have provided an exemplary demonstration of the process by which information from multiple sources, often including multiple consortia and replication samples, potentiates finding reliable, significant results. However, the value of continued attempts to find genetic effects of diminishing importance remains uncertain. Important questions are whether these small effect sizes have any value in understanding disease pathogenesis and what truly are the clinical implications of this line of study?

Alzheimer disease is one of the most heritable common, complex disorders, with a heritability of 60% to 80%,² and is one of the few diseases for which a single susceptibility gene, apolipoprotein E (APOE), gives rise to a substantial risk. Many early linkage studies, often of early-onset cases, helped researchers considerably in understanding disease mechanisms; eg, the amyloid cascade in AD.³ It is therefore gratifying to see that the loci now being identified may provide insights into other mechanisms of AD pathogenesis. Seshadri et al point out that the loci for which they find significant associations appear to be in pathways that have a plausible connection to dementia because the genes in those loci are expressed in the brain and play a role in neuronal differentiation or function. Nevertheless, considerable work will be necessary to understand the complex nexus by which these genes contribute to pathogenesis. Even if the mechanism is well delineated, it is unlikely that this insight will lead to new targets for intervention in the foreseeable future.

Has the success of GWAS for identifying associations with P<10⁻⁸ perhaps clouded researchers' perspective and seduced clinicians into expecting similar success in risk prediction? There is no reason to be surprised or disappointed. Finding statistically significant associations that contribute essentially nothing to improved risk prediction only verifies the notion that AD is a polygenic disorder; ie, that potentially tens of thousands of risk alleles, each with a small effect, are important for liability to disease. What is remarkable is that the single gene APOE is as important as it is for this complex disorder.

In contrast to many disorders that may be considered to be the extreme of a normal distribution of a physiological parameter (such as hypertension is to blood pressure), AD is not simply the extreme of a distribution of memory performance. Rather, the diagnosis is based on evidence from a number of criteria—essentially, the equivalent of a cluster analysis. Might it be that AD is not just polygenic but is the result of risk alleles in a cluster of genes (most often including APOE) where some constellations of risk alleles are important in some individuals while other combinations are important in other individuals? Or might differing combinations of risk alleles and environmental triggers (manifested as gene-environment interactions) be thwarting the ability to predict risk? Currently, massive computational limitations prevent exploring cluster-based approaches in a genome-wide framework. However, there may be some potential in applying cluster- or pathway-based analyses to genes that have been identified in GWAS.

Clearly, researchers need to pay much more attention to environmental risk and protective factors. The AD concordance rate for monozygotic twins is at most 61%,² and heritability is lower with increasing age of AD onset,⁴ suggesting increasing importance of the environment with increasing age. Environmental risk factors may accumulate with age or act as triggers of a disease process in a susceptible brain. Very large sample sizes, on the order of those in GWAS consortia such as those Seshadri et al¹ report, are necessary for detecting significant gene-environment interactions. Possibly more could be gained by focusing efforts in these consortia on incorporating information on environmental risk and protective factors in further collaborative efforts than in further pursuit of gene identification or replication. Many cohorts that have contributed information to GWAS have at least some information on selected risk factors other than age, sex, and APOE genotype. The next challenge is to take a step beyond gene identification and move into consideration of genetic risk in the context of environmental risk and protective factors.

What does this newly found genetic knowledge tell clinicians? It is a fresh reminder that family history is very important, even for late-onset disease that was once thought to be sporadic. But the world is facing an escalation in AD prevalence now that life expectancy is well past 75 years in most developed countries. Greater portions of the adult population will recognize signs of failing memory and cognitive impairment among their parents who are now among the oldest old. Lessons from increasing numbers of epidemiological studies with prospective information indicate that changes in midlife behavior, particularly those that are also conducive to cardiovascular health, can reduce risk of dementia or at least postpone onset.⁵ Findings such as those reported by Seshadri et al¹ reinforce the futility of using individual genetic risk profiling for AD beyond collecting information on age, sex, family history, and APOE status. The challenge for the clinician today is to ensure that individuals in midlife engage in the well-established, personally advantageous preventive behaviors already associated with benefit.