Prion disease is not one in a million

If there’s one number that everyone affected by prion disease has had drilled into them by reptition, it’s this: one in a million. For us, from the moment we first heard prion disease or Creutzfeldt-Jakob disease (CJD) raised as possible diagnoses for my mother-in-law, we were told this was a “one in a million” condition. It was comforting, in a way. One in a million seemed to provide hard facts to back our emotions — numerical validation for just how unlucky we felt. The disease was so terrible that I didn’t want to imagine more than one in a million people being affected by it. And if this disease was truly this rare, then it was understandable that so many neurologists had failed to diagnose or even suspect it, inevitable that we’d spent so many painful months on a diagnostic odyssey to different world-renowned hospitals and come up empty-handed.

But as Sonia and I left our old careers and re-trained as prion scientists to devote our lives to curing this disease, we came to understand that the “one in a million” figure, while not exactly false, is extremely misleading. It is a real statistic based on real data, but it doesn’t mean what many people think it means, and it gives rise to an incredible amount of confusion.

To understand why there is confusion, we need to define three different terms for different ways of measuring how common a disease is:

These metrics are all inter-related, of course, but they measure different things, and the numbers can turn out very different. Let’s take a look at how the numbers shake out in prion disease.

Let’s start with incidence. The U.S. National Prion Disease Pathology Surveillance Center in Cleveland publishes statistics on all cases referred with possible prion disease. In 2017, they saw 266 cases that turned out to be prion disease, out of a U.S. population of about 325 million. That’s 0.82 cases per million, or close to one in a million. As I said earlier, “one in a million” isn’t a myth, it is indeed a real statistic based on real data. But even the real data here come with a caveat. If you’ve lived through a loved one’s prion disease, you know that getting a diagnosis can be very challenging. One study tried to quantify the degree of underdiagnosis by looking across different countries [Klug 2013]. Some European countries have much more intensive surveillance systems than we do in the U.S., with their surveillance centers receiving up to ten times as many referrals from neurologists, per capita, as we do here. There are diminishing returns, of course — ten times the referrals does not mean ten times the cases — but those countries do see as many as 2 cases per million population per year. So even when speaking about incidence, the proper figure is probably more like 2 per million rather than 1 in a million.

Next, let’s consider prevalence. Prevalence is related to incidence, but it depends on how long people remain sick with the disease. Prion disease is very rapid: on the median, people only survive about 5 months from first symptom to death [Pocchiari 2004]. Thus, if there are 266 new cases in the U.S. in a year, but each only lives, say, half a year, that translates to 133 people sick at any given time. So the prevalence might be something like 0.5 per million, or, if you account for the underdiagnosis discussed above, perhaps 1 per million. (There is also a different concept called “genetic prevalence”, meaning the number of people with a genetic mutation that makes them likely to later develop prion disease — I will discuss this further below).

Finally, let’s consider lifetime risk. The question here is what is the probability that a random person — let’s say, from birth — will eventually develop prion disease. One very rough way to get at this is to start with incidence, and multiply by lifespan. If prion disease affects 2 people per million per year, and people live 80 years on average, then you have 80 chances to be one of those 2 in a million, so your chances of dying of prion disease are actually 2×80 = 160 per million. This is a crude calculation, though, which ignores the fact that people almost never get prion disease when they’re young — the age distribution in the population also matters. A slightly more sophisticated method would be to divide the incidence by the death rate, in order to ask what fraction of deaths are caused by prion disease. After all, we all die of something eventually, so lifetime risk of prion disease is just the risk that prion disease is that eventual cause. According to the U.N. Vital Statistics Report, Table 3, there were 2.7 million deaths in 2015 (the last year with complete data), meaning that about 0.85% of people (about 1 in 118) die every year. An incidence of 2 per million, divided by 0.0085 (or multiplied by 118) gives us a lifetime risk figure of about 235 per million (corresponding to 1 in every 4,255 people). This is higher than the 160 per million we came up with a moment ago, because immigration and births make the U.S. population skew young. In other words, a disproportionately small fraction of our population is in the age range where prion disease (or any other cause of death, on average) is likely to strike, and so a smaller number of observed cases corresponds to a higher lifetime risk. A different way to go about estimating lifetime risk is simply to review death certificates and ask what fraction of them list CJD or other forms of prion disease as the cause. Ryan Maddox at the CDC has been doing this for the U.S., and he reported in 2016 that about 1 in every 6,000 deaths is due to prion disease. This figure probably captures some cases that were diagnosed only postmortem or never got referred to the surveillance center, although there might still be some underdiagnosis at work here. Simon Mead has announced a figure of 1 in every 4,700 deaths in the U.K. Coming at it from several different angles and data sources, then, we converge on an answer that roughly 1 in every 5,000 people dies of prion disease, or in other words, the general population’s lifetime risk of prion disease is 1 in 5,000.

From three different ways of measuring how common a disease is, we can get three very different answers. 1 in 5,000 is very different from 1 in a million, but it’s not that either is wrong, they’re just measuring different things. The different statistics are appropriate for different purposes. Suppose you’re a scientist proposing a research study where you’ll ask newly diagnosed patients to volunteer to undergo some new brain scan to determine whether it’s diagnostically accurate, and you want to know how big a patient pool you can hope to draw from, then your relevant number probably really is 1 in a million — that’s the rate of new cases being diagnosed and reported to a centralized surveillance system. But if you ask a question like what are the odds that two people who knew each other should both at some point die of prion disease, then the relevant statistic is lifetime risk — 1 in 5,000. And if you mistakenly plug in the 1 in 1 million statistic in a place where the 1 in 5,000 statistic belongs, it’s easy to convince yourself that something is an outrageous coincidence, proof positive of something fishy going on, when in fact, everything is roughly as expected. Let’s look at a few examples.

What are the chances of a husband and wife both dying of prion disease? The probability of one specific person dying of prion disease is 1 in 5,000, so the probability of two specific people dying of prion disease is 1 in 5,000 squared, which is 1 in 25 million. There are apparently 60.8 million married couples in the U.S., so you’d expect that there will be 2 or 3 married couples alive today where both people eventually die of prion disease. It’s not common, but in a country this large, it will happen. The fact that one such married couple has been reported in the scientific literature [Brown 1998] is therefore not surprising.

What are the chances of two siblings, who do NOT have a genetic mutation in the prion protein gene, both dying of prion disease? We could do some fancy statistical modeling taking into account how many families have just 2 siblings, versus 3 or 4 or more siblings. Or, we could just start with a very rough ballpark estimate that, again, 1 in 5,000 squared, or 1 in every 25 million sibling pairs, will both die of sporadic prion disease. I’m aware of 3 sibling pairs where this has happened — one in the U.K. [Webb 2008], one in Switzerland [Frontzek 2015], and one in the U.S. (unpublished). There are surely at least 75 million sibling pairs where both have died in the past few decades across the U.S. and Europe, so again, there is no surprise in seeing 3 sibling pairs die of sporadic prion disease.

Here’s a slightly trickier one. What’s the chance of a person in a genetic prion disease family, who does not have the mutation themselves, dying of sporadic prion disease?

We are going to need an estimate of our denominator: how many people are out there in the world who are in genetic families but don’t have the mutation themselves. As an even more basic starting point, let’s consider how many people there are who do have mutations in the prion protein gene, PRNP. I mentioned earlier the concept of genetic prevalence — how many people carry a mutation that makes them likely to later develop a disease. As we’ve just established, about 1 in 5,000 people dies of prion disease, and about 10% of those have high-risk genetic mutations [Minikel 2016, Minikel 2018], so we can guess that the genetic prevalence of prion disease might be 1 in 50,000. A lot depends on how you measure — from birth vs. people alive today, which mutations you include, and so on, but even if it’s only 1 in 100,000, that’s still something like 3,000 people in the U.S., and maybe 10,000 across all the countries worldwide with prions surveillance systems in place. We can take this as sort of a loose upper bound, because this does include people who haven’t yet gotten sick, don’t know their genetic status, and maybe don’t even know that the disease runs in their family. To get a more conservative estimate, we could ask how many people have died of genetic prion disease and been reported to the biomedical establishment. Our largest case series captured 1,895 genetic cases across the U.S., Europe, Australia, and Japan [Minikel 2016] — but that’s only the cases that got diagnostic genetic testing, were reported to surveillance, occurred within a particular time window (about 1993 - 2014), and it doesn’t include some major genetic prion disease populations, such as those in Israel and Slovakia. The true number of cases could easily be twice that. So our estimate of people with mutations in known genetic prion disease families is somewhere in the 4,000 to 10,000 range worldwide.

Next, let’s ask how many people without mutations are in families with those people. Consider that for each person with a mutation, there is probably one spouse without a mutation, and one sibling without a mutation (since inheritance of the mutation is 50/50). Therefore, there might be on the order of 8,000 to 20,000 people in the developed world who don’t have mutations but are in a genetic prion disease family. And remember that 1 in 5,000 people dies of prion disease. So while it won’t be common, we might expect to hear of one or a few cases worldwide of a person without a mutation, but from a genetic prion disease family, dying of sporadic prion disease. It therefore does not seem surprising that two such cases are known in the biomedical literature [Capellari 2008, Areskeviciute 2018].

It’s important to keep the true lifetime risk of prion disease — about 1 in 5,000 — in perspective, because otherwise things that are perfectly within the bounds of expectation will appear too unlikely to be mere coincidence, and will invite some other explanation. If one assumes that prion disease is 1 in a million, then the chance of two spouses or two siblings both dying of prion disease is 1 in a million squared, which is 1 in a trillion, and there most definitely are not a trillion married couples or sibling pairs out there, so there must be some other reason that both people got sick. Different people are drawn to different explanations. Some people might speculate that prion disease was transmitted directly from one person to another, others might speculate that the two were both infected from some common source, others might propose a shared environmental risk factor, or a genetic risk factor outside of the prion protein gene, PRNP. Some of these explanations are more plausible than others — there has never been any evidence for casual transmission of prions between people, whereas there may well exist non-PRNP genetic risk factors. But none of these explanations are actually necessary when you appreciate that the number of these “coincidences” that people have observed is consistent with what we’d expect given how common prion disease is — 1 in 5,000, not 1 in 1 million.

To be fair, while the popular press has spun some outrageous stories about what might explain these coincidences, most of the researchers who have written about these cases are far more sophisticated. In practice, no one has done a calculation quite as simplistic as 1 in a million squared is 1 trillion. Instead, the papers cited above have looked at the incidence of cases per year reported in their country or region, integrated over some number of years spanning the deaths under discussion. But such calculations are too conservative, because they ignore how large the true denominator is — even if someone is talking about a married couple who died 5 years apart in Florida, it’s important to remember that they’d still probably be talking about if it was two best friends who died 9 years apart in Georgia. So a calculation based only on the exact circumstances observed doesn’t really capture the number of opportunities in the world for something “weird” to happen. In other words, the world is large enough that even unlikely things do happen, and they have to happen somewhere. If they happen to you, it seems too strange to be coincidence, but if you zoom out to the whole world, it’s not surprising at all.

This is important because when we falsely rule out random chance as an explanation for something, we may focus in on explanations that at best are a waste of time and at worst are harmful. Maybe a belief that something can’t just be chance motivates an expensive research investigation into genes or environmental causes that turns up nothing. But worse yet, the belief that something can’t just be chance can inspire wild conjecture and paranoia about prion transmission from person to person. Families affected by prion disease are already confronted by all manner of misconceptions and discrimination — we’ve heard stories of people being turned away by funeral homes or denied medical care because of people’s unfounded belief that prion disease is more readily transmissible than it is. False claims that co-occurrence of two cases in a family can’t be coincidence, and must therefore be a transmission event, feed into this paranoia. Thus, purely hypothetical explanations for chance events can end up doing harm that is all too real.

If you’re a researcher studying prion disease or a family affected by prion disease, people probably often ask you “oh, is that super rare?”. Next time this happens, give the best answer you can give: “it kills about 1 in 5,000 people.” Not super common, but not 1 in a million.