There are a variety of TSE diseases in mammals, including scrapie (sheep), kuru (humans, Papua New Guinea, from eating human brains), Creutzfeldt-Jakob Disease (CJD), variant CJD, Chronic Wasting Disease (deer, elk), Feline Spongiform Encephalopathy, Bovine Spongiform Encephalopathy (BSE, discovered in 1986). TSEs are invariably fatal. They cause lesions and holes in the brain, leaving the brain looking sponge-like—hence "spongiform." Incubation period varies, but generally years in large mammals (like cows and us). But once symptoms appear, incapacitation and death follow relatively quickly.
TSE is transmissible (by definition). There's a genetic component to the susceptibility and the incubation time, and a possibility that infection can be passed from mother to offspring. (The "natural" CJD also runs in families, apparently due to genetics, not infection.) Whatever the causative agent is, it replicates in the "host." It is quite hard to "denature": it survives heat, radiation, etc. It alters the three-dimensional structure (folding) of a kind of protein, prion (pRP). The agent may be the same protein it affects, but abnormally folded. (Although Stanley Prusiner won the Nobel Prize in 1997 for work on prions, whether prions cause TSE remains controversial. In particular, it is possible that TSE is caused by a virus that has not yet been identified.) According to one theory, the natural form unfolds periodically; if the abnormal form is present, it guides the unfolded protein into the wrong configuration. This cascades.
The infectious agent is concentrated in the brain and some other tissues (CNS, distal ileum in cattle).
BSE is transmissible to cats, dogs, goats, deer, mice, pigs, monkeys. It is unusual for infectious agents to be able to cross "species barriers," which makes BSE more frightening.
The transmission routes for BSE include contaminated feed (by far the dominant route), maternal (the calf of an infected cow may become infected; there is some evidence that this occurs, but the evidence is weak), and through contact with infected animals (there is no evidence that this occurs, but it cannot be ruled out completely). Cattle are most susceptible to BSE before they are two years old. The incubation period for the disease is quite long: on the order of 5–7 years. BSE also occurs "in the wild" (sporadically) at some rate in cattle (types L and H BSE, thought to be genetic). Similarly, there are sporadic cases of TSE in humans (CJD) and other species (e.g., scrapie in sheep and chronic wasting disease in deer and elk).
For an excellent lecture by Prof. Roy Anderson on TSE and the BSE epidemic in the U.K., see http://vega.org.uk/video/programme/72. The following videos are news stories and a short lecture on prion diseases.
There was a BSE epidemic in the United Kingdom, starting around 1986. It was eventually controlled, primarily through a ban on feeding ruminants to ruminants. But more than 184,000 cases were diagnosed with BSE and over 4 million cattle were culled from the herd as a preventive measure. There were only 11 new cases in the U.K. in 2010. Banning the feeding of ruminants to ruminants essentially stops the transmission, although there are sporadic cases and some evidence that maternal transmission is possible.
A number of people contracted vCJD from eating infected cattle. vCJD is generally fatal within a year of showing symptoms.
BSE has been found in the U.S., but no infected cattle were found between 2006 and 2012, when an (asymptomatic) infected dairy cow was discovered at a processing plant.
In May 2003, a BSE-infected beef cow was discovered in Alberta, Canada. The U.S. Department of Agriculture banned imports of Canadian cattle and beef in response. In November 2003 the U.S. Department of Agriculture proposed allowing beef and cattle less than 30 months old to be imported, with special restrictions on how they are slaughtered and their intestines are disposed of (the distal ileum, part of the small intestine, can contain a concentration of the infectious agent). The U.S.D.A. cited Canada's 1997 ban on feeding mammals to ruminants as part of the justification: feed bans are very effective. (The U.S. considers it adequate to ban feeding ruminants to ruminants.) http://www.gpo.gov/fdsys/pkg/FR-2003-11-04/pdf/03-27611.pdf
The Ranchers-Cattlemen Action Legal Fund (R-CALF) sued the federal government to keep out Canadian cattle, citing public health concerns. An expert witness, Dr. Louis Anthony Cox, wrote declarations on their behalf, estimating that opening the border would result in about 9–11 infected cattle being introduced into the U.S. each year, that the chance of importing at least one is essentially 100%. He also argued that it was essentially certain that at least one person per year would become infected with vCJD as a result. Dr. Cox's first declaration in the matter is here.
I analyzed Dr. Cox's analysis (and portions of the Federal Register and the epidemiological literature and studies by various European government agencies) on behalf of the U.S.D.A. and came to rather different conclusions. My first declaration in the matter is here. Dr. Cox responded to my declaration and I to his; that exchange is not discussed here.
How did Dr. Cox arrive at 11 infected cattle per year?
According to Dr. Cox, at that time the Canadian government tested about \(2769 + 4000 + 1300 = 8069\) cattle and found 3 with BSE, a rate of 0.037%. The cattle tested were not a random sample, however: Canada targeted animals thought to have higher risk of BSE. In Europe, the incidence in "high-risk" was about 60 times higher than in the general herd. Dr. Cox estimated that the rate in the Canadian herd is about 6.25 per million \((3/8069)/60 \approx 6.2\). He said that if imports resumed, the U.S. would import about 1.7 million cattle per year. On that basis, he concludes that the U.S. should expect to import about \( 6.25 \times 1.7 = 10.6\) infected cattle per year.
Dr. Cox says there is a greater than 99% chance of importing at least one infected cow per year if the infection rate is even 3 per million in the Canadian herd (not as high as his original estimate of 6.25 per million). Where does this probability come from?
Imagine tossing a biased coin that has chance \(p\) of landing heads, \(n\) times, with all tosses independent of each other. The chance that the coin lands heads at least once is \(1 - (1-p)^n\). If Dr. Cox's estimates of the rate and of the number of cattle to be imported were right, if importing an infectious animal were like a coin toss, if the tosses were all independent, and if the chance the coin lands heads were the same for every animal, then the chance of importing at least one infectious animal would be \(1 - (1-0.000003)^{1,700,000} \approx 99.4\%\).
Using similar logic, Dr. Cox estimates that there would be about 72 servings of infected beef in the U.S. as a result of importing Canadian cattle. This number comes from multiplying 3.6 BSE-positives per million cattle times a 1% rate of failing to remove all the infective material times a billion pounds of beef imported times two servings per pound. He estimates that about 40% of those servings would be consumed by individuals "susceptible" to vCJD. (Both the 1% rate and the 40% rate appear to be entirely speculative.) From that he estimates that there would be about \(72 \times 0.4 = 28.8\) human infections of vCJD per year.
Moreover, he estimates that the chance of at least one such infection per year is essentially 100%. Where does that number come from? Again, if we view infections as if they resulted from independent tosses of a coin with a fixed chance 40% of heads, the chance of getting at least one head in 72 tosses would be \(100\% - 60\%^{72}\), which is essentially 100%. Alternatively, if we view infections as if they resulted from two billion independent tosses of a coin, one toss per half pound of beef: If the coin lands heads, the half pound is contaminated with infected SRM and is fed to a susceptible person. The chance the coin lands heads is the same for every half pound: \(0.0000036 \times 0.01 \times 0.4 = 0.000000144\%\).
Let's think about Dr. Cox's numbers and assumptions. Dr. Cox said that the rate of BSE in imported cattle would be about 1/60 of the rate in the high-risk group. That ignores the fact that only young cattle could be imported, and that BSE takes a long time to develop. The number should have been about 1/3000 of the rate in the herd as a whole, based on what was observed in the European Union (EU). (In the UK, only one in 10,000 BSE-positives were in cattle under 30 months old.) Dr. Cox did not take into account variations by age (including whether the animal was born after the feed ban), geography, etc. The approximation he used to find probabilities does not fit the Canadian data—nor the data for other countries, including the UK, France, and Switzerland.
Dr. Cox assumes that if imports resumed, the U.S. would bring in 1.7 million head of cattle per year. The National Cattlemen's Beef association estimated the number to be 900,000, and the U.S.D.A. estimated it to be 1.4 million.
Meat and other tissue—except the tonsils and distal ileum—are rarely infectious until at least 32 months after exposure. The regulations would require slaughter before the animal reached 30 months of age. Meat has never been found to be infectious.
All four infected Canadian cattle that had been discovered by the time of these proceedings were from a region 280 miles across; they were 6–8 years old at the time of diagnosis; and three of the four were exposed to meat and bone meal processed before the feed ban by the same renderer, shipped in a one-month period in 1997.
Dr. Cox ignored the rules that would govern imports when he assessed the risk. Most importantly, he ignored the restriction that cattle must be slaughtered before they are 30 months old. That restriction has a profound effect on the risk. Animals that young would have been born substantially after the feed ban (and Canada's feed ban was a good one). Hence, they pose little risk beyond the extremely low rate of "sporadic" (naturally occurring) cases. In the extremely rare event that such a young animal is infected, because the incubation time for the disease is so long, the amount of infectious agent would be low: infectivity grows very slowly at first. And even that small amount of infectious agent would be removed largely, if not completely, because it is concentrated in specific high-risk material (SRM: nervous tissue, distil ileum, tonsils, etc.). While it is impossible to quantify the risk without making unrealistic assumptions with no basis in the data, it is clear that the risk is very, very small. The U.S.D.A regulations are an example of "defense in depth." Nothing is ever certain, but the regulations provide failsafes upon failsafes.
Rates are not probabilities. If some event has a probability, then in repeated trials, the fraction of times it occurs tends to converge to the probability. But just because something has a rate does not mean it is the result of a random process. Dr. Cox invented a probability model to turn an estimated rate of BSE in the Canadian herd into an estimated probability that one or more infected animals will be imported. In his model, whether an animal is infected is like whether a coin lands heads, all animals share a single coin, and all tosses of that coin are independent. He uses essentially the same approach to estimate the chance that at least one person will be infected with vCJD as a result of importing Canadian cattle. But this model has no basis in fact, and actually contradicts the existing data for Canada, the U.K., France, and Switzerland: non-sporadic BSE cases cluster in space and time, because they are linked through contaminated food sources. Hence the tosses are not independent. And it is as if different animals have different coins, because their exposures differ—especially those born before a feed ban compared to those born after a feed ban. Whether infection or infectiousness of imported cattle can reasonably be modeled as a coin toss even for a single animal is arguable; I believe it cannot, because there is no identifiable source of randomness at play. Animals are not selected for import by sampling at random from the Canadian herd. Whether any particular animal that is to be imported has BSE or is infectious is an unknown fact, not a random event.
Dr. Cox argues that violations of his assumptions would only make matters worse. For instance, he says that ignoring heterogeneity—variations in the rate of infection across geography, age, time, etc., makes his result optimistic. To see that this assertion is incorrect, consider several scenarios with and without heterogeneity.
Scenario 1. There are four cows in Canada, two with BSE. Select two at random. \(P(\mbox{at least one has BSE}) = 3/4 \). This is an example of homogeneity.
Scenario 2. There are four cows, two in British Columbia, two in Alberta. One in each province has BSE. Select one from each province at random. \(P(\mbox{at least one has BSE}) = 3/4 \). This situation is still homogeneous: we have two strata, BC and Alberta, but the rates in the two strata are equal.
Scenario 3. There are four cows, two in British Columbia, two in Alberta. Two in Alberta have BSE; none in BC does. Select one from each province at random. \(P(\mbox{at least one has BSE}) = 1 \). Here we have heterogeneity: the rate of infection is higher in Alberta than in BC. In this case, Dr. Cox is right: heterogeneity would increase the chance of importing infected animals.
Scenario 4. There are four cows, two in British Columbia, two in Alberta. Two in Alberta have BSE; none in BC does. Select two (i.e., both) from BC. \(P(\mbox{at least one has BSE}) = 0 \). Here the heterogeneity is as before, but if we do not sample from the stratum known to be infected, heterogeneity decreases the chance of importing infected animals.
Scenario 5. There are four cows, two over 5 years old, two under 2 years old. The older two have BSE; the younger two do not. Select two (i.e., both) younger cows. \(P(\mbox{at least one has BSE}) = 0 \). This situation is closest to that of Canada: animals born substantially after the feed ban are very unlikely to be infected, and only those will be imported (because only those are under 30 months old, under the proposed regulations).
Dr. Cox estimated that if imports resumed, 1.7 million cattle would come in from Canada. The cattle industry and the federal government estimated values 20% to 50% lower. Dr. Cox estimated that the incidence of BSE in the Canadian herd was 6.25 per million. That estimate was based on a number of animals tested that was too small, and on a ratio of the rate of infection between high-risk animals and the general herd that was too small (compared with experience in the european epidemic). If those two errors are corrected, the estimated incidence in the herd falls to 3 per million. In the E.U., the rate of BSE-positive cattle under 30 months old was 2,800 to 10,000 times smaller than the rate for older cattle. Dr. Cox omitted this factor. If these factors are included, Dr. Cox's model implies that we should expect to import one BSE-positive animal every 7–110 centuries. However, as we have seen, Dr. Cox's model has serious problems; merely correcting its inputs does not make it right. I do not think that the range 7–110 centuries means very much. It is definitely not my estimate of the risk, just an illustration that Dr. Cox's model does not show that the risk is high, if the incorrect inputs to his model are corrected.
Part of Dr. Cox's argument is that the U.S.D.A. assessment of risk was inadequate because it didn't put a number on the risk. Dr. Cox put a number on the risk, a number that is large (more than a 99% risk of importing at least one infected cow per year, essential certainty of infecting at least one person with vCJD per year). Notwithstanding the fact that he plugged the wrong numbers into the formulae he was using, we have seen that the formulae depended on some rather stylized assumptions about how animals are infected, imported, slaughtered, and distributed to susceptible consumers. The idea that importing infectious cattle is like a large number of independent tosses of the same coin is central to his approach.
Some things are random because we make them so, for instance, randomized controlled experiments. In randomized experiments, we know things are random because the experimenter ensured that they are random. And we know that random samples are random, because we made them so.
Some things are random because that's how Nature produces them. Stirring soup before tasting it randomizes what ends up in the spoon. According to quantum mechanics, matter and energy are intrinsically random. Radioactive decay appears to be random.
But it is also common to analyze questions by making up a probability model for the data, according to which what we see is as if it results from coin tosses or drawing numbered slips of paper from a well-stirred bowl. Then, we must beware. Our ignorance or uncertainty about how things happen does not make those things random. Positing that they are random, with no evidence that they are, introduces extra information that can be misleading. To claim that something is like a coin toss is a very strong assumption: Many things have two possibilities, but are not random like coin tosses. A store can be open or closed; a light can be on or off; a mushroom can be poisonous or safe; a cow can have BSE or not. The fact that we might not know which of the two possibilities holds does not mean that it is determined as if by tossing a coin.
R-CALF won an injunction in trial court, but the 9th Circuit Court of Appeals overturned the injunction and imports of Canadian beef and cattle resumed. A number of BSE-infected animals have been found in North America—both the U.S. and Canada—but the number remains small and there has been no epidemic on either side of the border.
Source: Centers for Disease Control
The total number of infected cattle found in Canada through 2012 is 19, of which 2 were identified as H-Type BSE. The total in the U.S. through 2012 is 4, of which at least 3 were identified to be H-Type BSE. The total in the U.K. through 2010 is 184,500 in more than 35,000 herds. At the peak in early 1993, there were over 1,000 new cases per week in the U.K. https://www.cdc.gov/prions/bse/bse-north-america.html
There have been 227 cases of vCJD worldwide, 1996–2017, according to CDC. As of 2009, there had been 217:
Since variant CJD was first reported in 1996, a total of 217 patients with this disease from 11 countries have been identified. As of October 2009, variant CJD cases have been reported from the following countries: 170 from the United Kingdom, 25 from France, 5 from Spain, 4 from Ireland, 3 from the United States, 3 in the Netherlands, 2 in Portugal, 2 in Italy, and one each from Canada, Japan, and Saudi Arabia. Two of the three U.S. cases, two of the four cases from Ireland and the single cases from Canada and Japan were likely exposed to the BSE agent while residing in the United Kingdom. One of the 25 French cases may also have been infected in the United Kingdom. http://www.cdc.gov/ncidod/dvrd/vcjd/factsheet_nvcjd.htm
There was another case in Canada in 2011, apparently contracted elsewhere. As of this writing, the numerous cases of BSE and vCJD that Dr. Cox said were inevitable (100% probability of at least one infectious cow imported per year, and 100% probability of at least one vCJD infection as a result) have not materialized. BSE and vCJD have long incubation periods, but it has been rather more than ten years since imports resumed—somewhat longer than the incubation period in cattle, and comparable to the incubation period for vCJD in humans.
For comparison, the Centers for Disease Control estimate that about 3,000 people in the U.S. die annually from foodborne illnesses. http://www.cdc.gov/foodborneburden/index.html
The U.S. tests about 40,000 cattle per year for BSE. Suppose the tests were of randomly selected cattle (every animal equally likely to be selected—this is not true because many of the tests are targeted at "high risk" groups of cattle). Suppose all the test results are negative. What can we say about the prevalence of BSE in the U.S. herd? 40,000 is a tiny fraction of cattle—is it too few?
If the prevalence were 1%, what would the chance of at least one positive test result be?
\(P(\mbox{at least one positive}) = 100\% - P(\mbox{all negative}) = 100\% - (99\%)^{40,000} \approx 100\%. \)
Even if the prevalence of infected cattle were 0.0075% (75 per million), the chance of at least one positive test would be
\(100\% - (99.999925\%)^{40,000} \approx 95\%.\)
As of January 2013, there were estimated to be roughly 90 million head of cattle in the U.S.. A rate of 0.0075% would be 6,750 cattle in total. Do you think testing 40,000 head of cattle selected at random would be an adequate level of surveillance for BSE? To screen for BSE in the U.S. herd, would you select cattle at random, or would you target high-risk groups? What groups do you consider high risk? Which do you think matters more, the absolute number of infected cattle, or the percentage of cattle that are infected?
How would you decide how much surveillance is enough? Does the cost of surveillance matter? What other things might that money be spent on? Are there ways of spending the money that are likely to save more lives than BSE screening?
Source: various, including http://www.fda.gov/food/resourcesforyou/consumers/ucm103263.htm, http://www.fda.gov/downloads/Food/FoodSafety/FoodborneIllness/FoodborneIllnessFoodbornePathogensNaturalToxins/BadBugBook/UCM297627.pdf , http://www.cdc.gov/nczved/divisions/dfbmd/diseases/typhoid_fever/, http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0004460/.
Why did R-CALF litigate? Do you think it was primarily a question of public health and the health of the U.S. herd? What effect does opening the border to import Canadian cattle and beef have on the economics of beef and beef production, even if no BSE is imported?