Today's post is about a piece of writing that I don't really recommend: A report from the French Food Safety Agency called
This report is a splendid example of irrational "scientific" thinking: Intelligent people following what they consider good rules, and ending up with a completely ridiculous result. If you take the logic of the of this report seriously, it would be wrong of health care personell to give mouth-to-mouth first aid to victims of drowning today:
a) The method carries with it several risk factors, like blowing air into the patient's stomach, which can cause the patient to vomit and get gastric acid into his airways. The method also carries with it the risk of over-inflation and lung damage, particularly if the patient is a child with low lung capacity. Heart compressions, which are often advocated along with the blowing, carry with them a severe risk of fractured ribs.
b) No double-blind study has yet been done, where drowned patients have been randomly been assigned to the treatment group and placebo group, and where enough care has been taken to keep observers in the dark about what patient is getting mouth-to-mouth treatment and who's been given placebo.
b) No double-blind study has yet been done, where drowned patients have been randomly been assigned to the treatment group and placebo group, and where enough care has been taken to keep observers in the dark about what patient is getting mouth-to-mouth treatment and who's been given placebo.
Before I get back to the drowning victims, I'd like to say something about seven distinct types of irrationality that we may be dealing with here.
1) First and foremost, the authors of this report seem to be suffering from a severe case of irrational loss aversion. This factor was first convincingly demonstrated by Amos Tversky and Daniel Kahneman. Several studies have shown that the motivating force of a potential loss (like the social awkwardness from being on a diet) is much stronger than a corresponding gain (like recovering from autism). This is sometimes also called the pseudocertainty effect – the tendency to make risk-averse choices if the expected outcome is positive, but make risk-seeking choices to avoid negative outcomes.
* On the one hand, they are taking very seriously a small risk for a modestly negative outcome: "No data are available on growth or nutritional status of autistic children subjected to a gluten-free, casein-free diet. Therefore, it is impossible to contend that such a diet has no harmful effect in the short, medium or long term".
* On the other hand, they are completely disregarding the possibility of a positive outcome.
* On the other hand, they are completely disregarding the possibility of a positive outcome.
2) Secondly, it seems safe to guess is that the loss aversion in this case is being reinforced by omission bias – the tendency to judge harmful actions as worse, or less moral, than equally harmful omissions (inactions).
* The alternative they are advising against, requires practical action.
* The alternative that they don't mind, consists of inaction.
* The alternative that they don't mind, consists of inaction.
3) Thirdly, the way the authors justify their position makes me suspect that they also suffer from the observer expectancy effect – they may have manipulated unconsciously the criteria for what they consider valid, in order to find the result they expected.
* On the one hand, they advise that studies should be assigned a credibility of zero, unless they fulfill certain formal criteria: A control group (autistic children without dietary intervention), random allocation of treatment or placebo, and a double blind protocol. These criteria happen to exclude all evidence for an effect from diet on autism.
* On the other hand, this last study is assigned high credibility, in spite of the fact that the study design is such that it didn't actually test the hypothesis that most parents and supporters believe in. It's as if they'd finally gotten around to doing a double-blind placebo-controlled study of mouth-to-mouth resuscitation, ... and were giving the patients 2 inhalations each, without clearing their airways of water first.
* On the other hand, this last study is assigned high credibility, in spite of the fact that the study design is such that it didn't actually test the hypothesis that most parents and supporters believe in. It's as if they'd finally gotten around to doing a double-blind placebo-controlled study of mouth-to-mouth resuscitation, ... and were giving the patients 2 inhalations each, without clearing their airways of water first.
4) Fourth, they also seem to have been at risk of falling for the Ambiguity effect - the tendency to avoid options for which missing information makes the probability seem "unknown." I see this as a verision of the Zero-risk bias - the preference for reducing a small risk to zero over a reduction in a larger risk).
* In autism, the total amount of uncertainty will be reduced if nothing is done (although in favour of a truly awful outcome).
5) Fifth, the authors must also have been at risk of Hyperbolic discounting - the risk of having a stronger preference for or against something, the closer to the present the cost or payoff are in time or space.
* In autism, the cost of treatment (financial and in the form of hassle) is certain and immediate, while the benefit is uncertain and several months (years) into the future.
6) Sixth, the authors have also been at risk for the Bandwagon effect – the tendency to do or believe things because many other people do or believe the same. This is related to "groupthink", "herd behavior" and the Semmelweis reflex – the tendency to reject new evidence that contradicts an established paradigm and the Availability cascade - a self-reinforcing process in which a collective belief gains more and more plausibility through its increasing repetition in public discourse (or "repeat something long enough and it will become true").
* No medical authorities have yet had the courage to be the first to make a rational cost-risk-benefit analysis of what this treatment has to offer autistic children.
7) Lastly, it's also possible that the authors (or the authority figures that have created today's medical paradigm in this area) are suffering from Confirmation bias – the tendency to search for or interpret information in a way that confirms one's preconceptions. This factor will be particularly strong if the preconceptions in question have the power cause severe and irreparable damage.
* In autism, it can be postulated that the medical authorities will be less and less disposed towards changing their policy, the more children that have suffered irreparable brain damage as a result of those policies.
There is probably a strong synergistic interaction between all these kinds of irrationality.
Now, back to the drowning victims. Do I hear voices saying that it's unethical not to treat them with whatever means we have at hand, as long as there's a chance that we might save them?` Thank you very much. That's exactly my point. Why shouldn't the same rule be applied to autistic children? Untreated autism leaves the patient to suffer for a lifetime (irreversible brain damage), along with his parents (burned-out), his siblings (neglected), and the rest of society (stuck with the bill when the patient grows up).
We still have a chance of saving at least some of these children. Why are these people insisting that we don't even try?
- o 0 o -
Before you can say you have a rational solution to the autism-and-diet question, you need to do the math involved in a structured risk analysis. I recommend that everyone who is interested in this issue, start by setting up a risk analysis matrix around at least five risk factors:
1) The patient 's physical growth is retarded because of a well monitored GFCF diet,
2) The patient develops irreversible brain damage, from untreated autism
3) The patient's social life becomes more complicated because of a GFCF diet,
4) The patient's social life becomes more complicated because of autism
5) The patient's parents incur some expense because of the GFCF diet,
2) The patient develops irreversible brain damage, from untreated autism
3) The patient's social life becomes more complicated because of a GFCF diet,
4) The patient's social life becomes more complicated because of autism
5) The patient's parents incur some expense because of the GFCF diet,
They should then assign a probability of 1-100% to each of these factors, under the following alternatives:
a) The patient does not try the diet.
b) The patient tries the diet, has no effect from it, and discontinues it after 6 months.
c) The patient tries the diet, experiences an average *) reduction in symptoms, and continues.
d) The patient tries the diet and recovers completely.
*) The average outcome of dietary experiments is still unknown. Even if it were known, it would be impossible to extrapolate from that figure to what was going to happen to one specific patient. For that patient, the full range of outcomes must still be reckoned with, from nothing to a full recovery. The actual treatment decision must therefore always be based on possibilities, rather than probabilities. Even so, I've attempted to assign a value to this column, based on my best assessment of the total amount of evidence available.
b) The patient tries the diet, has no effect from it, and discontinues it after 6 months.
c) The patient tries the diet, experiences an average *) reduction in symptoms, and continues.
d) The patient tries the diet and recovers completely.
*) The average outcome of dietary experiments is still unknown. Even if it were known, it would be impossible to extrapolate from that figure to what was going to happen to one specific patient. For that patient, the full range of outcomes must still be reckoned with, from nothing to a full recovery. The actual treatment decision must therefore always be based on possibilities, rather than probabilities. Even so, I've attempted to assign a value to this column, based on my best assessment of the total amount of evidence available.
Next, they should assigned a numerical value (for example 1-100) for the importance of each of the problems above, relative to the worst possible outcome in the most serious of the categories, to make them comparable.
Lastly they should multiply probability and importance. The resulting numbers will tell something about how important you consider it is to guard against each of these factors.
My conclusion at the bottom of my personal risk analysis matrix, was that a lifelong diet needed to produce a recovery rate of around 4%, or a reduction in severity of the autistic outcome of the same size order, in order to outweigh the costs. If we take into consideration that patiens who don't benefit from the diet, can discontinue it after 6 months, the figure falls to under 1%.
This is all based on my personal value system, i.e. how I (for example) prioritize a 100% certain loss of social convenience (being able to eat anything in birthday parties) versus a somewhat lower chance of irreversible brain damage.
I'm longing for the day when the Government's medical researchers starts doing this kind of math. When they come clean about their priorities. When they start looking at ALL the available evidence, including the lab reports that show the opoid peptides right there, physically present in the blood and urine ... just because it isn't part of a randomized double-blind placebo controlled crossover study.
And I'm longing for the day when the same people start paying attention to study designs, and start laughing - along with me - of studies that are so badly designed that they can't prove anything - even though they fulfill all those sensible criteria, with randomization and crossover and all.
:-J
