Clinical Risk Calculator for MS: How Close Are We?
Decoding MS risk factors is less like fitting together a jigsaw puzzle and more like balancing a Jenga tower, with layer upon layer of complex interactions and unpredictable outcomes if something changes
When we need it, we have the Framingham risk assessment tool to determine our risk for having a heart attack in the next 10 years. Enter a few bits of information about yourself—age, sex, cholesterol and blood pressure numbers, blood pressure meds—and voila!—you learn that the 10-year risk for having a heart attack for a person with your numbers is less than 1%. Phew.
What we don’t have, however, is a way to check our risk over any time frame of developing multiple sclerosis (MS), even when we have relatives with the condition. With all the advances in genetic analysis and improved understanding of environmental risk factors, an algorithm for determining such risk might seem like a near-future reality. It’s not. In fact, according to several researchers working on exactly that, we’re just at the beginning of the long and winding road to a clinically useful risk-assessment calculator for a disease that affects as many as 450,000 people in the United States and more than 2 million worldwide.
Why can’t the marriage of genetics and environment achieve fruition sooner? As they say on Facebook about relationships, it’s complicated.
“We’re working on that, but to be honest we’re really not there yet because … there are environmental risk factors and genetic risk factors that have been identified, but we actually don’t understand how they act together,” said Emmanuelle Waubant, M.D., Ph.D., a professor of neurology at the University of California, San Francisco (UCSF), in an interview with MSDF. Because different research groups have historically focused either solely on genetics or solely on environment, studies have rarely addressed the overall risk with all of these factors in combination, she said. “So that’s why there’s no way we’re in a situation to come up with risk.”
Philip De Jager, M.D., an associate professor of neurology at Brigham and Women’s Hospital in Boston, agreed. “We’re still very close to the beginning of the whole process,” he told MSDF. The genetics aspect is a fairly straightforward part of the equation in the sense that genetics is both easy to measure and reliable, he noted. But the drawback of the risk gene variants identified so far, De Jager said, is that they don’t explain much of the risk. “There’s a portion of the risk that’s related to genetics I think that we can estimate very well, but it’s just a small portion of the overall risk.”
According to Waubant, the strongest genetic risk factors identified to date could confer a two- to threefold increase in MS risk. That might sound like a lot. However, she also pointed out that as much as 70% of the white population carries such a risk variant, but that the “vast, vast majority” will never develop MS.
That means a lot of people are walking around with tiny gene sequence changes that studies have correlated with MS, but for most of those folks, MS will never pop up on their diagnostic radars. This overlap in variants between affected and unaffected populations implies that screening at the general population level for MS-related variants to identify people at high risk might not be effective, according to Cathryn Lewis, Ph.D., a statistical geneticist and professor at King’s College London. “There may be more targeted groups where genetic screening is useful,” she told MSDF, “for example, in people that have a strong family history of MS.”
De Jager’s group is engaged in research with just such a targeted group, the Genes and Environment in Multiple Sclerosis (GEMS) research study. “Of course, we’ve known for decades that family members are at higher risk than the general population, although the absolute risk is still pretty low at 2 to 3%,” he said. “So what we really need to do are enormous studies to find those going through the transition” from non-MS to MS. His team is following a cohort consisting of about 2,500 people with first-degree relatives who have MS, tracking their “unique collection of genetic variants and life history experience” in an effort to develop and test an algorithm in a group of people with a preexisting higher risk.
As De Jager noted, because of the increased risk for people with a first-degree relative with MS, researchers have long known about the genetic component. Pinning it down to use it clinically to identify anyone at risk has proved more elusive. Results of a recent paper by Di Santo et al., published in PLOS ONE, illustrate these difficulties. In spite of working with 110 known MS-related risk variants, these authors found that with this suite of genetic changes, 5% of the population would be identified as being at increased risk of MS. That adds up to about a fourfold increased risk compared to your average Joe or Joanne with no family history of the disorder. According to Lewis, “Although having a fourfold increased risk of disease may sound high, since the prevalence of MS in the population is only about 1 in 1,000, even these people with the highest inherited risk of MS only have about a 4 in 1,000 chance of developing MS.”
So what’s the point of genetics if they’ve revealed so little so far and can explain such a tiny percentage of the risk? Pierre-Antoine Gourraud, Ph.D., M.P.H., a professor of neurology at UCSF, said it’s because genes are so beautifully easy to pin down. “The reason why it’s easier to do genetics is that it’s very easy to measure,” he noted in an interview with MSDF. “We know how to measure genetic variation, and this is very stable because when you have the genes you have them for your entire life. Almost anything else is more difficult to measure because it’s varied.”
It turns out that as beautifully measurable and well-behaved as gene variants might be, even the broader genetic picture is messy. We might be inclined to think of MS risk factors as a puzzle with pieces that will interlock if only we can find them and figure out how they fit. But they’re much more like a Jenga tower, with layer upon layer of complex interactions and unpredictable outcomes if something changes, ranging from minor instability to total collapse.
“There are a couple of hundred variants that are going to be important,” De Jager said, “but … we don’t know whether it’s the total number of variants you have or the variants in a certain pathway.” That question is an important one, he said, because having 50 variants distributed across a lot of disconnected pathways might not be as bad as having 50 variants in the same pathway all pointing to the same disease outcome.
And there’s more. Changes that don’t lie within or near a gene can still influence the gene, such as enhancer sequences that sit thousands of molecular steps away from the gene they affect. And, of course, no gene is an island: They all act under environmental influences. Environmental factors blur the risk picture so much, we’re staring at MS risk as rendered by Monet.
Overall, we don’t understand much about the interactions between genetics and environment. “We have some examples of that but nothing well validated,” De Jager said. “[We] just have a few hints. With MS, it’s still emerging.”
In terms of the weight of environmental risk factors for MS, what has emerged is not much more compelling than the gene variants. Waubant points to smoking. “Smoking is associated with an increase of up to 70% in risk,” she said. “Not even a twofold factor. Epstein-Barr virus, about a twofold increase. Vitamin D about a twofold increase.” Overall, she said, the contribution of any of these factors to total risk is “pretty modest,” and, echoing De Jager, she said, “It’s not been clearly studied how all of these risk factors in combination may increase overall risk or not.”
One of the clinical uses of the Framingham risk calculator is that some of its factors are modifiable. Cholesterol can be modified. Blood pressure can be modified. Whether a person smokes can be modified. These possibilities point to clinical utility not only in terms of identifying risk but also in terms of doing something about it. What would such a calculator for MS risk achieve?
According to Waubant, from a public health standpoint, some environmental risk factors might make good targets. “If low vitamin D is confirmed to be associated with a doubling of risk of MS,” she said, “then maybe pediatricians should be careful in terms of how much vitamin D supplementation children are receiving.” But not all environmental factors are modifiable. An Epstein-Barr virus infection, once you’ve had it, is not something you can modify, and it might even be something you don’t necessarily want to prevent. Waubant noted that it may be that Epstein-Barr, given its high rate of infection, is a virus that plays an important role in shaping the immune system. At any rate, vaccination against it isn’t currently possible and, she said, “might not even be safe. We don’t know.”
Predicting risk for a condition isn’t the only application of a risk calculator. Another possible use is to ascertain which disease type a person most likely has and what course it might follow. With MS, these issues aren’t simply matters of semantics. Progression is variable, and the forms of MS—primary progressive MS, relapsing-remitting MS, secondary progressive MS—are also variable. Response to therapy also varies from person to person. Researchers hold out hope that genetic and environmental risk factors could someday interact in a clinical calculator for disease phenotype, prognosis, and therapeutic targeting, in addition to predicting disease risk.
But if the contribution of gene variants to susceptibility is a slippery entity to pin down, the phenotype and disease progression genetics are like algorithmic quicksilver. “In a way,” said UCSF’s Gourraud, “it’s quite striking to see how much we know about the genetics of susceptibility and how little we have about the genetics of severity or the genetics of MS course.”
Environmental factors might be more promising for progression predictions. Examples of environmental factors that might lend themselves to such identification are vitamin D and smoking. Both have been associated with the risk of having a worse disease course following MS diagnosis, Waubant said. Targeting these factors in risk assessment “becomes a little more manageable,” she said, “because you’re not talking about trying to decrease exposure in millions and millions of people [who] have a 0.1% chance in their lifetime. Now you’re talking about modifying the rate of progression in patients who already have the disease.”
Ah, but there’s yet another rub: the matter of timing. It’s one thing to identify a risk factor but something else to figure out when the exposure to that factor matters. As Waubant said, “Genes, of course, you’re born with them, but these environmental factors? It may be that exposure to cigarette smoking is more important before age 10 or it doesn’t matter. We don’t know.” Another issue is how much of a dose of the factor is relevant. “For things like vitamin D level or smoking,” Waubant said, “there’s also graduation for how much you’re getting or not getting and for how long or at which time of your life.”
If you feel like tugging at your hair in frustration—with both hands—that’s understandable. With all of the risk factors by all of the timing issues by all of the possible interactions by all of the dose-of-exposure conundrums, is it even tenable to develop an algorithm for disease risk, phenotype identification, or progression prediction at all?
In spite of the apparently unpromising outlook for an effective calculator for MS risk prediction, De Jager thinks it will happen. “We need to do a lot more, but our [upcoming] paper and several papers with the UCSF group, for example, are building these models,” he said. “Eventually, something from this will come into clinical practice.” People tend to underestimate how long it takes to move from lab to clinic with tools like this. “In some cases, it will be fast,” he said, “but even from our first paper back in 2009, it was clear that the genetics themselves weren’t going to be enough.”
Gourraud is also optimistic, if that’s the right word for “cautious but hopeful.” “Eventually, you may be able to reach the goal to help make decisions about treatments,” he said. “Maybe help us in understanding why some treatments are working very well in some groups of patients and not in others.”
Actually, even now, most of these findings apply to only one group, period, and that group is, to quote Waubant, “white people.” Lewis said that even genome-wide association studies could be missing some genetic variants, noting that those that have been captured are most of the common variations “at least in European populations.” Ancestry is thus another consideration in developing an appropriate risk algorithm, according to Waubant. “These risk factors may be very different if we’re talking about a population of white individuals or African-American individuals,” she noted. “Ancestry is kind of important because lots of the genetic work has been done in cohorts of mostly white people, and so we still have to confirm that some of these genetic factors are also risk factors for people of different ancestry.” Some groups have begun to focus on nonwhite populations in recent years.
Lewis is betting on something far more specific than genome-wide association studies and the most frequent risk variants when it comes to predicting personal health. “The genetics of MS has come an amazing distance over the last few years but still has a long way to go before the findings can be translated to benefit patients,” she said. “I suspect the success will come from whole genome sequencing, being able to identify anywhere in the genome where there is an increase in risk. Ultimately we’re going to be walking around with our genome sequence on a USB stick in our pocket. It’s going to be valuable not just for MS but for all sorts of chronic diseases.”
Key open questions
- In an algorithm that includes genetic risk factors, what is the utility for the patient?
- What is the role of epigenetics, if any, in this complex risk picture?
- Will we someday all be walking around with our genome sequences on a USB stick?
Disclosures and sources of funding
Pierre-Antoine Gourraud is a member of the MS Discovery Forum Scientific Advisory Board and is the founder of Methodomics (2008) and MIRA Medicine (2013). Philip De Jager is a member of the Scientific Advisory Board of the Accelerated Cure Project, the nonprofit publisher of MSDF. None of the other sources quoted in this article had any disclosures to make.