Copyright © 1995 by Steven J. Milloy. All rights reserved. First edition. Published by the Cato Institute, 1000 Massachusetts Avenue, N.W., Washington, D.C. 20002. Library of Congress Catalog Number: 95-72177. International Standard Book Number: 0-9647463-2-8.
Chapter 9
Cooking Up Biological Plausibility
Earlier we said biological plausibility is an essential element of the epidemiology equation. If your statistical association doesn't make some sort of biological sense, you're in trouble. For our purposes, biological plausibility can be developed in two ways: either from the linear nonthreshold model we discussed earlier or from toxicology experiments. We'll briefly discuss biological plausibility from the linear nonthreshold model and then spend most of this chapter on the technique of toxicology for risk assessment.
Instant biological plausibility
Biological plausibility from the linear nonthreshold model is simple. Whatever happens at high doses also happens at low doses. Cancer from too many medical X-rays is biologically plausible because of the studies of the A-bomb survivors. Forget that this is completely contradicted by those same studies. Skin cancer from low levels of arsenic is biologically plausible because high levels of arsenic are associated with skin cancer. Forget arsenic is actually an essential nutrient in trace amounts.
That any level of exposure to mercury is dangerous is biologically plausible because we know that high levels of exposure can lead to mercury poisoning. Never mind that mercury was used for more 350 years as an effective treatment for syphilis before it was replaced in the 20th Century by an arsenic compound.
Lung cancer from radon in the home is biologically plausible because of the observations from underground uranium miners. Don't let small details (such as areas with higher levels of radon are associated with lower lung cancer rates) bother you. Lung cancer from ETS is biologically plausible because smokers have a higher risk of lung cancer. Never mind that passive smokers are inhaling a chemically different type of smoke in a physically different manner. High exposures, low exposures, different exposures. They're all the same...
Baked biological plausibility
Toxicology is the study of poisons. It's not a traditional public health discipline mainly because poisons, as commonly thought of, are not a traditional public health issue. But so much for tradition. Today, toxicology is a multipurpose weapon in the public health arsenal.
Toxicology can be used to create a risk from scratch. For example, as you are reading this, the U.S. Government's National Toxicology Program is "investigating" whether alcohol causes cancer. There are only a few inconclusive epidemiologic studies for this association. So the government may have to "jump start" the research to go anywhere at all.
Toxicology can be used as a life support system to keep a risk alive. In the 1970s, toxicologists first reported dioxin is an extremely potent carcinogen in laboratory animals. However, epidemiology studies have since failed to associate dioxin with cancer in humans to any convincing degree. Nevertheless, the mere existence of the dioxin toxicology has fueled unending research (worth well over $1 billion) whether dioxin causes cancer and other health problems in humans.
Lastly, let's talk about toxicology's use for biological plausibility purposes. But there's one caveat: Even if you have the best toxicology in the world, few will believe you unless you've got human data that "proves" your point. That's why, for example, those who have been pushing the dioxin risk are still running in place after almost 20 years. The dioxin epidemiology is really disappointing, particularly considering how promising the dioxin toxicology was.
Most of the general public can distinguish — or are willing to distinguish — the difference between what happens to rats in "rigged" experiments and humans in the real world. So the dioxin toxicology itself hasn't been able to carry the day. Nevertheless, proponents of dioxin have obviously done quite well for themselves.
Being a public health-type, you probably don't know the first thing about toxicology, or certainly not enough to design and run your own bioassays (that's what this type of toxicology experiment is called). Unless you can find "canned" toxicology (studies that have already been conducted and published in the literature), you'll have to team up with a skilled toxicologist.
By skilled, I don't mean a toxicologist that is necessarily a "real" scientist — they're not inclined to engage in this sort of silliness...I mean, activity. You want someone who understands your need to find risk and who knows all the tricks to maximize your chances. You need someone who can pick the right animals, give them the right doses in the right way and who knows what to look for when the animals are autopsied.
Picking the right species
Since we don't typically study poisons directly in human subjects, we use rats, mice and other types laboratory animals. What species of animals should you use? You're probably going to need at least 300 animals (trust me). Generally speaking, rats and mice tend to be the most popular for bioassays because they're a lot cheaper, take up a lot less space, require a lot less care and are a lot less endearing than some other animals.
But remember: cutting corners may cost you in the long run. That is, the least expensive animals may also be the least sensitive to what you are exposing them. Not all animals react the same way to chemicals. What's toxic to a dog may have no effect — or a reduced effect — on a rat. So, instead of saving money by getting rats, you could actually waste your money by picking an animal that doesn't respond the way you want.
So which species do you pick? Unfortunately, there is no easy way to predict which species will give you the best results. It's basically bioassay roulette. If cost is not a showstopper, you could run your bioassay in several different species such as rats, dogs and rabbits. You then base your results on the species that is the most sensitive in the bioassay. The advantages of this are several.
First, at least you'll know you haven't picked the least sensitive species on which to base your results. Second, if the results of all the bioassays show all the different species get whatever you're looking for (bravo!), you have a self-validating series of interspecies experiments. On the other hand, if results between species groups are inconsistent (the rats get cancer but the dogs and rabbits don't), you may have to explain away the inconsistency. Although you probably could finesse this issue in your writeup, it's probably easier to ignore the data you don't want. My guess is you wouldn't be the first.
So remember, the most important thing is to find the proper species of test animal. This is the species most likely to help you "prove" your risk. Penny-wise is pound-foolish, as they say.
Maximizing the dose
Now that you've picked the right species of animal, how much do you expose them to? Simple: As much as they can stand. If your toxicologist were to expose the animals to the amount of your substance that humans ordinarily come into contact with, you would likely need, and I mean this literally, millions of animals... and millions of cages, tons of food and an army of lab assistants. This would probably blow your budget, even with one of those generous federal grants.
Such large numbers are needed because, even if the risk you're looking at really exists, it is likely to be so small it can only be detected by looking at a very large population. But, as you can see, large populations are impractical. Fortunately, there's a way around this dilemma; it's called the "maximum tolerated dose" or MTD for short.
The MTD is the highest amount of your substance that can be administered to the animals without killing them or making them noticeably ill just from being poisoned. But wait a minute, you say... Isn't that what we're trying to do anyway? Poison them? Well.. yes, but only in the long run. You don't want to just out-and-out poison them to death right away. Your results will be meaningless, and your competition...peers, I mean...will just laugh at you.
It's a well-established principle of toxicology that the dose makes the poison. In other words, everything is a poison; it's just a matter of dose. Chocolate ice cream can be a poison in high enough amounts. In trace amounts, arsenic is an essential nutrient. So you don't want to be accused of just "poisoning" your animals. The MTD basically maximizes the probability that a bioassay with only 50 animals will turn up something useful to you without simply poisoning the animals.
How high are MTD doses? MTDs can be anywhere from 10,000 times to 100,000 times what humans may be exposed to. But this has no relationship to reality, you say? Perhaps but remember, we're trying to manufacture risk and make you a star. Who's got time for common sense?
What to do with rats that just won't cooperate
Sometimes the logical or most obvious way to expose the laboratory animals to your substance won't do the trick. For example, if you want to look at dandruff shampoo as a cancer risk (and some have), you might consider applying the shampoo (or its key ingredient) to the skin of your laboratory animals.
Or, let's say you are interested in fiberglass as a cancer risk when inhaled. You might consider making your laboratory animals breathe air that contains fiberglass fibers. Now, suppose that in both these cases, the logical and intuitive routes of exposure produce nothing useful in the bioassays. What do you do?
Don't panic. Don't ever panic. There's almost always some way to fix things and get your intended results. If you can't go in through the front door, try the back door. If skin applications of dandruff shampoo don't work, trying feeding. When inhaling fiberglass doesn't work, try injecting it into the lungs. If feeding doesn't work, try inhalation. There are a number of pathways into an animal. One of them should work, even if by mere chance! Don't worry about the relevance of the pathway. It's like horseshoes and hand grenades — close is plenty good enough.
What to look for
If you're interested in a cancer risk, when your toxicologist is looking at your animal's body parts, she's going to be looking for cancerous tumors. The more tumors the better. However, it may be that your animals have developed noncancerous (benign) tumors as well as cancerous ones. If your toxicologist only counts the cancerous tumors, there may not be enough tumors for you to identify a statistically significant risk.
So you tell your toxicologist to count both the cancerous and benign tumors. You see, in toxicology for public health, we simply pretend that benign tumors are really just like cancerous tumors. (Does this type of fantasy sound familiar? Remember the linear nonthreshold model?) It's like basing your batting average on how many times you hit the ball, fair or foul, not just on your base hits. A grounder or pop fly is as good as double. In this case, every tumor, cancerous or benign, is a home run!
Wrapping it up
Once your toxicologist counts the tumors, she'll tell whether there are significantly more tumors in animals exposed to your substance than in the control group... almost exactly what we do in epidemiology studies. If so, you've got a positive association between your substance and cancer (in animals at least). This is just what you were looking for. This is good enough for biological plausibility.
If there's no positive association, you've got three options. First, you can go back to the drawing board and pick a new target. This will cost time and money. Second, you can rerun the bioassay, maybe trying a different species or a different exposure pathway. This will also cost time and money. Third, you can ignore the results of your bioassay and take your chances on the epidemiology alone. If you pick Option No. 3, your epidemiology must be extra-convincing to overcome your failure to produce on the toxicology side.
Click here for Chapter 10: Peer "Review"
Click here to return to the home page
Copyright © 1997 Steven J. Milloy. All rights reserved. Site developed and hosted by WestLake Consulting.