Celiac Food Roulette, Explained
May 16, 2016
Feeding me is a little like feeding a very confused slightly-alien who didn’t remember to bring important things like a tricorder or a sonic screwdriver, so can’t actually tell whether something is edible or not. I therefore take a lot of risks living on this planet, but haven’t rightly stopped to calculate (even in a back of the console sort of way) what size those risks are.
Bounding the desired risk
The effects of mistaking something for food vary in strength and duration, and depend on whether I’ve made other mistakes recently. But for a back of the envelope perspective, lets presume a single mistake causes approximately a week of significant illness. Then, if I want to be ill half the time, I should make the average risk of meals such that about one every two weeks goes wrong. Given about 5 meals or snacks in a day, that’s 1 in 70.
But noone wants to be sick half the time. Suppose I accept an average meal risk of 1 in 1000. That equates to an average of 12 days per year ill, in addition to the sundry colds, flus and seasonal allergies that already take their toll. Little enough for the effects that take months rather than weeks to dissipate to be under control, but doesn’t sound impossible.
So given that range, where a 1 in 70 average meal risk is dismal but a 1 in 1000 may be acceptable, what is the actual risk of food? Can such an acceptable bound be achieved, and if so then what is required to do it?
Bounding the achievable risk
Some foods will instantly create an effective 100% risk for the meal in which they appear. Anything labelled “contains wheat,” or with ingredients like “wheat flour” or “barley malt” or “rye” anything. These are the problems almost anyone with a passing familiarity with celiac will recognize. In theory I could blow an entire risk allowance on one bagel, but in practice the effects of a large dose are so phenominally worse than the cross-contamination and minor-ingredient effects I based the estimated one week duration on that it would be a horrible mistake.
Next up are those labeled “shared equipment” with wheat processing. I’ve seen cases where this was ok, but only extremely rarely. I’d call it a 99% risk. Then there are oats. Oats are a real problem. They don’t have to be labeled as contains wheat, but they are almost always cross-contaminated unless strictly tested. Where by almost always, I loosely assign a 99% risk to any oat product not strictly and persistently tested for gluten contamination.
Then comes “shared facility,” which is far more variable. It can mean anything from running right next to wheat things with flour in the air to an entirely different building on the site with good controls between. At a rough guess I’ll call these a 60% risk, although the variance is huge. Once in a while you’ll see “shared facility but [description of seperation procedures]”; I’m not counting those as shared, really. It tends to be consistent within a product and brand though.
Restaurants with GF menus fall somewhere between shared facility and shared equipment, with tremenduous variation but a tendency toward the high risk end. Because so many products are involved, as opposed to a steady few in manufacturing, risks are less consistent between meals and higher overall. I would put the average risk without further prior knowledge around 90%. Non-dedicated home kitchens tend to be less well seperated or cleaned, verging back up toward 99% risk.
Then come the subtle but significant risks…
Soy sauce: Well labeled on prepackaged food but a major source of oblivious failures in restaurants. Soy sauce contains wheat; tamari does not.
Malt flavoring: Can be made from corn, is usually made from barley. I’ve seen arguments about this one, apparently there’s a move away from barley so it may have dropped to around 80% risk.
Misreading the label: There are a lot of little details to go through, and sometimes one just gets sloppy or tired. Or eye strain; those fonts are tiny. I’d say I pick up something that looks ‘fine’ at first read, then later discover wheat or shared facility listed on it, about 5% of the time. For risk bounding purposes, I’ll run with that number, even though sometimes I catch it before consumption (but not often).
Natural flavors: Can contain malt flavoring. This is very rare, but does happen on occasion. I have no good data as to how often so I’m ballparking it at 1% risk.
Undeclared allergen, cross-contamination, etc.: Problems that can cause recalls… eventually. In the mean while, they cause pain. Not frequent but existent issue, call it 1% risk. More like 10% for oats.
Stuff I forgot: Maybe 3% risk. I forget things a lot and by definition I don’t remember what they are. But there are sneaky ingredients. Like starches. And .. I forget.
For simplicity, I’m going to lump all factors below malt flavoring together as about a 5% risk attached to any product not strictly labeled as gluten free. Due to FDA regs and good testing, I’m going to loosely and laughably claim labeled products are safe. Things can and do still go wrong, but this is a back of the envelope calculation.
All of these risks are without prior knowledge. With prepackaged products, consistency in production means prior experience is an excellent guide, so a personal database of known-safe-enough products can be built up over time. This is absolutely the key to cheap GF eating, since well tested and labeled products tend to be expensive, although that’s been getting better. Things can and do still go wrong, of course, so some bound would have to be put on that if this weren’t the back of an envelope.
Intersection of bounds
Suppose all untried foods are chosen as well as possible from foods not strictly labeled GF. By the above, I’ve given this a 5% risk. One untried food a day doesn’t sound unreasonable while still figuring things out… Until you note this equates to a 1 in 100 average risk, which isn’t much better than the 1 in 70 that was marked unacceptable. The desired risk threshold of 1 in 1000 can be met with one untried food per ten days, presuming all previously tried foods that remain in the diet are safe. Since mistakes usually have to be made twice for certainty, and not all low grade ones are caught quickly, this is generous.
A more numerically pleasant and realistic bound would be one untried not-GF-labeled food per two weeks. But only prepackaged foods, with full ingredients lists, that are not from dubious facilities. There is no room in such a scheme for food at a party, shared kitchens, friends trying to cook, restaurants, or indulging in any other risky bad ideas.
Suppose among these risky bad ideas one makes value judgements that give ‘risky’ meals, including those with the 5% risk scenario considered above, an overall average 50% risk. So some new restaurants, occasional friends kitchens, but mostly trying out new foods as safely as possible. Then the acceptable rate of risk taking drops by a factor of ten, to one every twenty weeks. Or just over two attempts at food discovery outside GF labeling taken in a year. (Three if you didn’t make the two weeks estimation above.)
There is a balance, I think. But it involves getting sick more often than is desirable, doing a lot of smiling stiffly at parties while turning things down, and gritting of the teeth when it’s summer new product season because advertising really does work and so many things look tempting.
For myself, I think my takeaway will be to try for no more than one careful-but-risky whim a month. And penny pinch and plan in order to take GF labeled well tested foods with me more often. Because I am dearly tired of being ‘careful’ yet sick half the time due to curiousity. One new restaurant a year and one new food a month is a ridiculously tight bound, even though it sounds pretty accurate to me. Thankfully, cooking from scratch can bound this in a far less psychologically stressful manner. But that’s a different calculation entirely.
Addendum: Tiny risks add up
Suppose, rather than being perfectly safe, prior good foods and GF labeled foods have an associated risk of only 0.05% (1 in 2000) per meal (which may contain multiple foods, so each must have a lower risk than that). Then, right there, the entire risk allowance is cut in half. With frequent risks, little bits matter a lot.