Statistics Show Strengths and Weaknesses of Genetics-Common Disease Studies

CAMBRIDGE, Mass. (February 12, 2003) — When the $100M HapMap project was announced late last year, it stoked a decades-long debate surrounding the “common-disease, common variant” hypothesis.

Can hunting for links between common genetic variants—single letter variations in DNA sequence—and common diseases help reveal why some individuals are more susceptible to common diseases like diabetes and hypertension than others? Or, as some scientists argue, does this strategy of finding candidate disease genes waste resources because seemingly similar studies vary widely in their outcomes?

Thus far, researchers have used gut feelings rather than hard numbers to drive their arguments, both pro and con.

Now, a Whitehead Genomics team led by Joel Hirschhorn answers the question, using statistics rather than impressions. Their work, published recently in Nature Genetics, concludes that the strategy, called genetic association studies, has merit and should also be enhanced by designing the studies better, making them large enough, and reporting negative as well as positive findings.

Genetic association studies typically compare the DNA of individuals with the disease with healthy controls to see whether a genetic variant is more prevalent in affected individuals than in the general population. Studies of genetic contributions to complex diseases are hard to design because they follow a pattern different from the classic, rare, Mendelian disorders where a single genetic variant can account for the disease.

Most common disorders are complex, and in these cases, many genetic variants contribute, but each merely increases the risk of disease. An individual’s chance of being affected by the disease is determined by a combination of these genetic variants and environmental factors, such as diet. As a result, the effect of a single variant on the disease may be subtle, influencing factors such as risk, age of onset, or disease severity.

This makes designing good case-control studies challenging. Because of the modest effects of individual variants, studies might find no significant difference between the cases and controls.

Negative results may go underreported because such outcomes, on their own, hold little interest. People tend to stash away such results; it’s called “the desk drawer phenomenon”, says Hirschhorn, who is also an Assistant Professor of Genetics at Children’s Hospital and Harvard Medical School. This biased underreporting of negative outcomes falsely inflates the strength of variant-disease association.

Critics argue that common variants raise the disease risk by such a small amount that any signals are too subtle and “association” studies would fail to detect them. They also argue that the difficulty in reproducing association studies means that positive findings that have been reported represent statistical noise rather than true signals. But the Hirschhorn team, including first author and student Kirk Lohmueller, rigorously analyzed 25 different gene-disease associations, each of which had been the subject of multiple published reports.

They showed that although most of the associations were in fact likely noise, some were in fact real signals, even when the desk drawer phenomenon is taken into account. This suggests that there are many common variants in the human genome with modest but real effects on common disease risk, and that studies using large samples will convincingly identify such variants (seemingly weak signals can be spotted, and enhanced if the sample size were bigger).

“Modest signals might sometimes be interesting even when they don’t end up being strongly predictive of disease. They might help explain the mechanism of disease and provide drug targets,” says Hirschhorn. “For example, one associations—between PPARG and type 2 diabetes—involves the protein that is targeted by thiazolidinediones, which are drugs used to treat diabetes. If the thiazolidinediones hadn’t already been discovered to be a treatment for type 2 diabetes, this association would have pointed in that direction.”

The study’s main conclusion is that a fraction of already-reported associations between common variants and common disease are likely correct. This finding is encouraging, since it suggests that many more such associations could be found in the future, once more common variants are tested for a role in disease.

This possibility provides further impetus for the HapMap project, which will determine the patterns of common variation throughout the genome. As a result of this project, testing common variants for association with disease will be both easier and more comprehensive. A Whitehead Genome Center team, led by David Altshuler, Mark Daly, and Stacey Gabriel, is a major participant in the HapMap project, which is a public-private effort to build the next generation map of the human genome.

Called a “haplotype map,” this effort is expected to make it easier, faster, and perhaps cheaper to find genes that predispose us to common diseases such as diabetes and cancer.

“We’ve only scratched the surface so far in accumulating common genetic variants. We expect the HapMap and large association studies should identify lots of other variants that are real genetic risk factors for disease,” says Hirschhorn. The report has already caught the attention of genetics researchers. For example, biostatistician Tim Rebbeck (University of Pennsylvania) said that he had alerted several colleagues about the paper. The work, he noted, is a great teaching tool because it quantifies the strengths and weakness of association method for understanding the genetic bases of common diseases.

Citation

Lohmueller, K. E., Pearce, C. L., Pike, M., Lander, E. S., & Hirschhorn, J. N. (2003). Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease. Nature genetics, 33(2), 177-182.