Friday, 2 March 2012

But it's only a survey study of the GFCF diet and autism...

Treelight @ Paul Whiteley
I briefly touched upon the recent study published by Pennesi and Cousino Klein* examining the efficacy of a gluten- and casein-free (GFCF) diet for children with autism a few days back. Having had some contact with the authors and a chance to look at the full-text of the paper, I feel I can post some more details about how the study was done and just what can be inferred from the results obtained.

I freely admit that there is a bit of a conflict of interest here on my part because it's an area that I've done some work on. This probably also accounts for the number of 'my research says this' statements that litter this post for which I will apologise now.

Had the results reported in this most recent paper been less kind to this kind of dietary intervention, I dare say my scientific mettle might have been tested as it was when another trial was reported on a few years back, although as far as I know not published in peer-review format yet. Suffice to say that I am slightly more relieved that years of my life have not been debunked with this latest offering; at least on this occasion.

Anyway, back to the recent GFCF study and some details:

  • The aim of the study was to examine subgroups within the autism spectrum and various dietary implementation factors related to the GFCF diet.
  • A 90-item online questionnaire on various aspects around the use of a GFCF diet was developed and administered via SurveyMonkey, a popular website used quite a bit in research and marketing circles. 
  • Completed responses collected over 5 months in 2008 (N=387) were analysed. 
  • Two groups of parental responses were initially analysed: (a) those who removed all foods containing gluten and casein from the diet of their child with autism (n=223) and (b) those who only removed some gluten- and/or casein-containing foods from their child's diet (n=70). When comparing whole dieters vs. partial dieters, significantly more improvements were noted for whole dieters based on measures of core and peripheral autism related behaviours (echolalia, repetitive behaviours, etc), social behaviours (eye contact, social response, pointing, language production) and physiological symptoms (bowel functions, seizures, etc).
  • Groupings based on reported dietary errors (N=271), ranging from absolutely no breaks at all to more frequent diet-breakers and everything in between both in and outside of the parental home, suggested that breaking the diet now and again was not necessarily the be-all-and-end-all when it came to outcome. So fairly consistently those who broke the diet one or two times a year were actually reported to show more improvements than those who were strictest with the diet.
  • The length of time diet was in place also suggested some interesting trends; those who followed diet for more than 2 years showed the greatest number of improvements when it came to autism-related and physiological symptoms. Having said that, there was also a bit of spike in the 6-12 month duration group across the survey item groups.
  • Probably one of the most important findings from this paper is in relation to experiences of children with autism and comorbid gastrointestinal (GI) or allergy-related symptoms in terms of responses across the survey groups. These children seemed to do consistently better than those without GI or allergy symptoms. This trend also continued when looking at those diagnosed with a food allergy or suspected to have a food sensitivity.

Based on these results, a few things in particular caught my eye.

The suggestion that those with autism comorbid to bowel issues did better on diet is interesting if not exactly new. I think back to the Harland Winter trial on the GFCF diet which still needs to report its findings and how this might further inform any relationship (or not). Given also the various studies pointing towards diet and bowel issues in autism such as the rates of lactose intolerance and the carbs angle, one can perhaps see where there might be some overlap. The allergy side of things has also been covered in previous research and it will be interesting to see how the 'defining the gluten spectrum' game-changer paper might also feed into any future work in this area.

The other interesting detail relates to the time on diet - effectiveness data. I can remember my first study in this area which seemed to confirm that dietary intervention of this type is not a 'quick-fix', short-term thing. One of the important things we found in our last autism dietary research foray was some indication of a plateau effect from such dietary intervention. That is, after a period of time, diet didn't exactly 'stop working' but rather results suggested that positive changes to outcome slowed down compared with earlier testing occasions. Aside the possibility of a faltering placebo effect(!), I never really had a good explanation for this, aside from either some kind of homeostatic mechanism related to gut bacteria or gut permeability kicking in - our bodies love balance and equilibrium - or that a GFCF diet might affect more peripheral areas of autism (e.g. inattention, hyperactivity) rather than core areas. This last point implies that what you see [simplistically] is a kind of peeling away of the accompanying issues to reveal the true 'autistic core' presentation over the months of dietary intervention. What the new study adds to that issue is to pinpoint where that plateau might occur (after about a year) but at the same time also show a sort of 'second wind' whether due to diet or perhaps other factors (maturation, other interventions, etc).

If I was in real scientific rottweiler mode, it would be very easy to savage this study based on lots of different confounders and potential forms of bias being present. One could easily ask questions like 'how do you know respondents were parents of children with autism?' 'how can you confirm diagnosis?' and 'do concepts like constipation and diarrhoea mean the same to everyone?'. And the answers to these questions would be 'we don't', 'we can't' and 'not necessarily'. Don't even start down the path of bias via self-selection, retrospective recall et al. Having said that, and with specific reference to the reporting of bowel issues in autism, the recent paper by Gorrindo and colleagues** hinted that parents "were sensitive to the existence, although not necessarily the nature, of GID [gastrointestinal dysfunction]". I'm sure than many parents are indeed over-joyed to hear that they do officially now know the difference between normal bowel habits and problem bowel habits in their own children!

The point however is that this study knew what is was and what limitations it had. It didn't claim to be the gold standard randomised, double-blind, placebo-controlled trial of a GFCF diet despite the fact that quite a few people have pointed out that it wasn't.

OK maybe I would take some issue with one line in the conclusion section: "These findings provide additional support for the use of this diet in treatment of some children with ASDs..". They do but they don't insofar as the methodological restraints on a survey methodology. The continuation of the sentence is however important: ".. and substantiate the importance of further investigations into the various, nuanced biological and diet implementation factors that interact to optimize and attenuate the diet’s effectiveness for treating these children".

I think Dr Martha Herbert summarised it all quite well: “This study supports the need for taking diet impacts in autism seriously and pursuing them, and learning not only whether they work – because they won’t work for everybody – but also how they work when they do” according to comments on this site. I can't really say more than that.

* Pennesi C. & Cousino Klein L. Effectiveness of the gluten-free, casein-free diet for children diagnosed with autism spectrum disorder: Based on parental report. Nutritional Neuroscience. February 2012.
DOI: 10.1179/1476830512Y.0000000003

** Gorrindo P. et al. Gastrointestinal Dysfunction in Autism: Parental Report, Clinical Evaluation, and Associated Factors. Autism Research. January 2012.
DOI: 10.1002/aur.237


  1. It is long past time that the GFCF diet gets a more serious look. It isn't always needed and doesn't always work, but when it does it can work some wonders. It is a shame that one of the truly effective treatments for autism lingers in the backwoods of alternative medicine.

    I only wish that there were a properly designed RCT that documented the effectiveness's. But before that could happen there would need to be a better way of determining which children with autism are going to benefit and which aren't.

    Which, along those same lines, is exactly what the forthcoming trial of CM-AT did. That study only included children with autism who showed specific biological markers in their stool samples. I have a feeling that that study, when released, is going to be a milestone in autism research.

    Now if only someone would do the same with the GFCF diet....

    One other thing, the diet's "plateau effect" that you mention isn't that hard to explain. I don't think that "autism" is some monolithic entity that only has one method of action in the body. Rather it is something of a self-reinforcing disruption that impacts multiple parts of the body via multiple pathways. So the behaviors that we call "autism" are the combined effect of multiple systems in the body being out of balance at the same time. Some of these effects are going to impact the same functional area, making that one area worse than any single pathway could.

    Under a model like this, "fixing" or correcting one pathway doesn't mean that you automatically fix everything else. Nor does it mean that the "fix" can restore the damage that was done by the pathway being out of balance in the first place.

    For example, if a child is having problems absorbing nutrients because of GI issues and the GFCF corrects the absorption problem, that does not mean that the damage caused by the nutritional deficiencies during key developmental periods would be reversed. Nor does it mean that the developmental milestones that were never reached are automatically mastered.

    The other thing to consider is what would happen if the diet were removed. It may look like a plateau has been reached, but if you remove the GFCF diet you might well see a regression. I know from personal experience that my twin daughters, after years on the GFCF diet, are still sensitive to casein and somewhat sensitive to gluten. When they have a dietary infraction we see all sorts of interesting behaviors and they get to spend some extra quality time in the bathroom.

  2. Many thanks for the comment MJ.

    I agree with your sentiments and suggestion following Dr Fallon's example (which by the way is a very good example of best research practice).

    In many respects you may have also provided a good starting point for any future trial of the GFCF diet: concentrate on those participants who do see a regression when diet is not in place.