We have searched our topic of interest via electronic databases (PubMed, Embase and the Cochrane Library) in August 2016 by utilizing Medical Subject Heading terms which are listed as follows: “autism”, “autism spectrum disorders”, “diet”, “gluten-free”, “casein-free”, “gluten”, “casein”, “GFCF” and their combinations. Afterwards, references of the selected articles and systematic reviews were searched by two authors in order to eliminate the risk of overlooking (Study selection process can be viewed from Figure 1).

This category includes 9 different systematic reviews listed as follows: Millward et al, 2004, Christison and Ivany, 2006, Millward et al, 2008, Mulloy et al, 2010, Mulloy et al, 2011, Buie 2013, Hurwitz et al, 2013, Marı´-Bauset et al, 2014 and Elder et al 2015 (Table 4).

Systematic review conducted by Christison and Ivany in 2006 concludes that there is not sufficient data to support GFCF diet in ASD treatment by analyzing six uncontrolled clinical trials and a single-blind randomized controlled trial. We believe that the review conducted by Christison and Ivany is inadequate to reach a valid conclusion due to the absence of control group in most of the studies 23.

Hurwitz et al, 2013 reached the conclusion that elimination of gluten and casein from diet of an ASD patient would not cause significant improvement in ASD symptoms by analyzing 5 study (same conclusion with Christison and Ivany, 2006) 39. Moreover, Hurwitz et al, 2013 states that 3 out of 5 study demonstrate that GFCF diet is not beneficial for ASD patients, two of which were reliable studies since they were blinded studies comparing participants’ behavior and found consistent outcomes. On the other hand, other 2 studies claim that GFCF diet has positive impacts on ASD patients. However, these two studies do not reveal reliable results because of not having placebo control (high risk of bias)39.

Millward et al, 2004 states that the results about cognitive skills, linguistic ability and motor ability are unclear, whereas, there is an important beneficial effect in autistic traits associated with GFCF diet. Millward et al, 2004 touches upon the fact that there are wide confidence intervals in the outcomes since their review is a small-scale study20. On the other hand, Millward et al, 2008 concludes that there is not statistically significant difference in any of the traits between the control group and treatment group 27. Similarly, Marı´-Bauset et al, 2014, Mulloy et al, 2010 and Mulloy et al, 2011 argue that there is not sufficient data to support opioid-excess theory and, thus GFCF diet in treatment of ASD 313341.There is a consensus on studies’ about this topic being weak and limited . Moreover, all of these studies indicate that GFCF diet can be resorted by medical professionals if the patient experiences dramatic behavioral improvement associated with GFCF diet or if the patient has a confirmed allergy towards gluten and/or casein 2027313341.

Even though Buie 2013 reaches the same conclusion as all other systematic reviews, Buie 2013 emphasizes the possibility that inconsistent and insufficient data regarding to GFCF diet might be related to genetic differences and possible gene expression differences between individuals suffering from ASD 34.The emphasis of Buie 2013 about the individualistic differences and individual-based treatment strategies should not be overlooked.

This category includes six studies all of which are well-conducted randomized controlled trials aim to reach out a conclusion about the efficiency of GFCF diet in ASD treatment (Table 4).

Sponheim, 1991 has both conducted double-blind gluten/placebo studies (3 patients) and open study (4 patients), thus become a fundamental study in this category. Sponheim, 1991 argues that GF diet is not beneficial for ASD patients, also in some instances it becomes a negative factor that causes further social isolation of patient 13. Elder et al, 2006 conducted randomized, double-blind study with 15 patient (age 2-16) and argues that there is no connection between GFCF diet and ASD symptoms despite the presence of improvement in some patients (statistically insignificant). One must consider the fact that results of Elder et al, 2006 are preliminary and based on a short-term treatment, thus not complete 21.

Hyman et al, 2016 is a double-blind, placebo-controlled study conducted with 14 children (age 3-5) which demonstrates no statistically significant benefit associated with GFCF diet treatment for patients suffering from ASD (Similar results were established by Hyman et al, 2010 in a randomized, double-blind study) 3043. Depending of the size of study group (72 patients) and longer treatment period (24 months) Whiteley et al, 2010 is an attention-grabbing study. It promotes the same conclusion reached by other studies in this group, though, Whiteley et al, 2010 does not provide conclusive evidence due to lack of placebo control 32. Even though these studies utilize distinct assessment methods to evaluate patiens’ conditions they reveal supportive outcomes 21263043.

Despite the fact that two studies in this category, Knivsberg et al, 2002 and Johnson et al, 2011, are designed differently we believe that they are equal in terms of trustworthiness and “Level of Evidence” (Table 4).

Positive correlation between GFCF diets and ASD treatment has been demonstrated for the first time in this systematic review by Knivsberg et al, 2002. Even though the study group is not large (20 participants) use of two-tailed Mann–Whitney U-test, p-test and mean value analysis reduces the risk of bias. The comparison between control group and diet group indicates significant improvement in the diet group even though small improvements are observed also in the control group.These statistically insignificant improvements in the control group might be related to the fluctuations in autism symptoms as argued by the authors 19.On the other hand, Johnson et al, 2011, is a non-randomized, single blinded study which also rejects the relation between gluten and casein free diet and autistic behaviors 35.

Lucarelli et al, 1995 is a double blind study which examines the effect of gluten-free and casein-free diet on patients individually. Lucarelli et al, 1995 concludes that significant improvement is associated with GF and CF diets. Additionally, Lucarelli et al, 1995 indicates that casein challenge is related to the worsening of autistic symptoms which becomes a supportive evidence to the interest of this study 14.

Similar results are established by Knivsberg et al, 1990 and Knivsberg et al, 1995 both of which utilize urinary peptide level analysis and standard assessment questionaries 1215. Also, Cade et al, 2000 demonstrates that schizophrenia or autism is associated with exorphins formed due to incomplete digestion of gluten and casein in the intestine by immunoglobulin analysis as claimed by opioid-excess theory 18.

On the other hand, controversial findings are established by Whiteley et al, 1999 which argues that GFCF is associated with behavioral and social improvements whereas no significant decrease in specific urinary compounds are observed expected by opioid-excess theory. Therefore, outcomes of this study point out the possibility that possible success of eliminating gluten and casein from diet might be related to a different reason, rather than opioid-excess theory 16.

Patel and Curtis, 2007 is a qualitative study performed with a small study group comprised of 10 participants. This study indicates significant improvement in social interaction, concentration, writing, language, and behavior in all of the participants and decrease in urinary heavy metal levels (lead) in every participant 25. Nazni et al, 2008 investigates the efficiency of gluten-free diet and casein-free diet separately and concludes that they are both beneficial for autistic patients (Higher improvements are observed in gluten-free diet compared to casein-free diet). Other study in this category conducted by Mageshwari and Minitha, 2006 provide supportive evidence to the conclusion reached by Patel and Curtis, 2007 and Nazni et al, 2008 and demonstrates improvements in behavioral symptoms of autism 2428.

This category consists of five case reports and two studies based on parental reports. Even though Irvin, 2006 does not suggest any association between ASD and gluten-free diet, Knivsberg et al, 1999, Hsu et al, 2009 and Genius, 2010 claim otherwise 17222938 Moreover, Herbert and Buckley, 2013 shows improvements in autistic symptoms, language skills and auditory sensitivity. 40 These studies will not be further discussed since they are case reports which may include high levels of bias associated with small study groups and unreliable assessment methods.

After completing the discussion about the efficiency of GFCF diet for patients suffering from ASD one should consider adverse effects associated with the implementation of GFCF diet. Despite the fact that GFCF diet is a relatively new treatment method for ASD numerous studies exist about its’ safety, as well as its’ efficiency. Although 2 studies, Cornish et al, 2002 and Johnson et al, 2011, report no significant side-effects associated with GFCF diet in any kind of macro or micronutrient, most of the studies report lower bone density and calcium intake 354445464748. In addition to lower calcium intake and bone density Neumayer et al, 2012 also reports lower vitamin D and lactose intake from gut 49.

Another aspect of gluten free diet, effects on microbiome in human gut, is investigated in Bonder et al, 2016. According to their 4 weeks of study gluten-free diet results in moderate but significant changes in human gut microbiome, especially for 8 bacteria species. For instance, remarkable decrease in Veillonellaceae, known for lactate fermentation, and Ruminococcus bromii, known for starch degradation is reported 50 Nevertheless, there is an obvious need to conduct more comprehensive research about the safety of GFCF diet since none of the previous studies were conducted with large study groups.