Analysis of GFCF Diet in Autism, Article Critique Example

Abstract

This paper reviews the results as well as the discussion presented in, The Gluten-Free, Casein-Free Diet in Autism, an article by Elder et al (2006).

Analysis of Article on Gluten-Free, Casein-Free Diet in Autism

Elder et al (2006) investigated the extent to which gluten-free and casein-free (GFCF) diet treated autism problem. The authors used “randomized, double blind repeated measures crossover design”. They used a sample of fifteen children with the ages of 2 to 16 years who suffered from autism spectrum disorder. The sample was kept on the specific diet for twelve weeks through which data was collected on urinary peptide levels and autistic symptoms. The analyzed data did not show any statistically considerable findings even though the parents of the children under study reported notable improvement on their children. Since the expected results were not found, through the research conducted, room for more study is left with a direction on how to conduct the research given.

The study was arranged in a superb order. They started by introducing the topic in which Autism was defined followed by the reason driving the study. It was proved that the problem was increasing and thus needed an urgent solution before it blew beyond proportion. Literature review in which earlier attempted trials on autism diet reports were reviewed. Purpose statement and study goals were defined. The method of study was shown in which the children were selected through sampling methods in which written consent was provided by the children parents. The below parts of this critic paper evaluates the presentation of the results by the authors as well as the discussion they gave for their study.

Evaluation of Results

Descriptive Statistics

There were two instruments of description namely Childhood Autism Rating Scale (CARS), and the Autism Diagnostic Interview-Revised (ADI-R). All the fifteen items in CARS instrument were evaluated on a seven-likert scale. This instrument was tested and proved for validity and reliability as it had been used by many earlier researchers. The (‘f. coefficient of this instrument has been set at 0.94. All the children were observed for the set time using CARS and the scale filled.

Through ADI-R, a semi-structured interview was done on the mothers and caregivers of the children in the sample. ADI-R includes three main areas namely “Impairment in reciprocal social interaction (SI), communication (CO), and repetitive behaviors and stereotyped patterns (RB)” (p. 416). Dependent variables were measured using Childhood Autism Rating Scale (CARS), Urinary Peptide Levels (UPL), Ecological Communication Orientation (ECO) Language Sampling Summary, and In-Home Observation.

The children, participants, were screened using CARS. Baseline assessment of each participant was also done and then compared with the control dietary experiments. Using UPL, urine samples were taken on 21^st, 42^nd, 53^rd, and 84^th days and tested for gliadorphin, casomorphin, gluten peptides, and urinary casein levels and them compared with the normal levels of <0.95. ECO Language Sampling Summary was used to record a child’s behavior. In home-observation, they videotaped the behaviors of the children as they interacted with their caretakers. The researchers ensured that they used methods, which have been used by earlier researchers and tested for validity and reliability.

Standard deviations and frequencies of dependent variables were recorded in form of a table (p. 419). The analyzed data did not show any significant changes even after monitoring children as they fed on the specific diet for the specified period. The fact that the research was a pilot study meant results were not reliable. The authors failed to represent descriptive data on all combinations involved in the scenarios making it impossible to reveal statistical analysis. Most of the analysis was based on the feelings and attitudes of the caregivers as they argued that they saw improvement on the children after feeding on the diet. The parents gave positive comments on their children because they knew that they were under GFCF diet. Regrettably, the tables were complex and with incomplete data making it impossible to describe the sample adequately.

Statistical Assumptions

The greatest assumption in the study conducted by Elder et al (2006) was “children with autism are at high risk for amino acid deficiencies and may benefit from a structured diet” (p. 415). They assumed that their sample had high amino acid deficiencies than the control group in order to ensure comparison. Improvement was to be determined by the children being compared with children who did not suffer from the autism disorder. As a result, the two classes of individuals, the sample and the control group, had to have different characteristics. This has been seen as an assumption by other researchers who warrant further investigation on this issue.

The fact that the children under study stayed at their homes meant that an assumption had to be made that their caretakers fed them strictly on GFCF diet. This cannot be established because some parents might have failed to adhere to the diet only and the results could not be as expected. The researchers had to assume that the children were fed strictly on the diet they advised while in reality, they might have been fed something else by their siblings. Since “participants were provided all meals and snacks from the GCRC’s Metabolic Kitchen for 12 weeks” (p. 417), this was a good base of their study by assuming that the participants could not feed on anything else because they had all they needed in their diet. Elder et al (2006) did not show prove that their data was free of homogeneity, independence, and linearity of variance adding to their assumptions.

Inferential Statistics

As mentioned earlier, Elder et al (2006) used a number of statistical methods to analyze the data they collected into a form that is easily understood by the people who used the data. The authors started by identifying the sample of children to use in which the participants had to be children between the ages of 2 and 16 years and suffering from autism disorder. Demographic variable of the participants were identified and compared with dependent variables. The authors used covariance analysis appropriately.

Elder et al (2006) first applied multivariate covariance analysis simultaneously on the dependent variables that were being studied. The study conducted by Elder et al (2006) was mainly a pilot study. A two-sided T-test was constructed in which the dependent variable value was realized by taking period 2 value less period 1 value. Further, the mean obtained from AB ordering was compared to that of BA ordering. The main dependent measures were identified as CARS and ECO. A 2 sample T-Test was seen to be advantageous over 1-sample T-Test in that it ignores order of treatments. Further, a 1-sample T-Test is limited in larger variances.

The authors had incomplete tables on major variables and this forced them use a missing random model for week 12 and 15 subjects whose data was not provided. The realized significant results were analyzed using CARS. However, since some figures lacked in the drawn tables, it was difficult to follow the discussion the authors gave. This called for an inclusion of other figures to make the tables complete thus clarifying the results.  Nevertheless, the employed statistics were appropriate and adequate in answering the authors’ hypotheses.

Evaluation of Discussion

Restatement of Purpose and Study Goals

Elder et al (2006) need to restate their purpose of the study in that what they need to achieve is overstated bearing in mind that their sample size is too small. They state, “The purpose of our overall research program is to provide critical feedback loops for families of children with ASD including the most current scientifically sound information and incorporating family observations into treatment planning and development” (p. 415). This means that they are out to come up with better results compared to the results presented by earlier researchers on the same topic. However, this is difficult to achieve since earlier researchers used a bigger sample size and yet came with unreliable results. Since there is no cure on autism disorder so far, researchers in this field should use a very sample size and long time to report reliable results.

It has been noted that even though, “Recent studies provide interesting information regarding hypothesized GFCF dietary effects on physiology, behavior and cognition, they are limited by small sample sizes” (p. 415). Elder et al (2006) should not repeat this mistake. Otherwise, they should have changed their purpose of the study to something lesser that could easily be achieved with their small sample size. Their study goals are also too high to achieve with such a small sample size. Thus means that Elder et al (2006) need to restate both their study purpose and study goals.

Discussion of Implications

Elder et al (2006) made sure that their findings were related to earlier research on the same field in order to make their results’ presentation clear and without confusion. They once mentioned that their results were more reliable compared to those of prior researchers implying that they were doing comparisons. They needed to have a better research by relating their results to their set study purpose and goals to accomplish. Mentioning that their study, “Clearly, was difficult to conduct but undoubtedly viewed as important by the participating families”, shows that they were not prepared despite the fact that they are the ones who set the goals to achieve. The authors mention, “Irrespective of the treatment order assignment, the dependent variable was the value in period 2 minus the value in period 1. The mean for the AB ordering was then compared to the mean for the BA ordering” (p. 418). This makes the results unclear since one is left to wonder what the authors meant by the terms period 1 and 2 as well as AB and BA ordering. They managed to summarize their results but made them less informative. No explanations are linked to any hypothesis making it more difficult to clarify the results.

Even though no relationships were done on the research findings and the study purposes, relating them to earlier research made the results reliable and realistic. All used methods of data analysis were proven reliable since they had been tested by earlier researchers for usability and applicability. Before explaining how they used CARS as one of their descriptive methods, Elder et al (2006) started by proving its reliability as they say, “The instrument’s validity has been assessed as good under various conditions” (p. 416) by Schopler et al (1980) and Schopler et al (1986) to having a reliability level of 0.88. ADI-R has also been proven by Rutter and Couteur (1994) to having an “interrater reliability from .62-.89”. Giving this proves before using the methods as their main descriptive methods shows reliability of the results even before considering their data collection procedures.

Even with the limitations associated with their research, Elder et al (2006) gave imprecise but acceptable discussion that gives a good direction for future research. They proved that with follow-ups of the participants to ensure that they fed only what was prescribed, increasing the sample size, and increasing study time was enough to come with more logical results. This thus makes it easier for researchers who will be interested in the same topic since they will know where to start.

Alternative Explanations

The authors did not give alternative explanations on their discussion. The considered children vary in a number of ways. This makes the explanation illogical since with difference in severity of the autism disorders in the participants involved, it is impossible to come up with informed research. The findings were limited to the variability of the characteristics of the participants since it was not possible to come up with sound results when in reality, the treatment offered had to work differently on different participants.

Identification of Potential Limitations

Elder et al (2006) knew clearly that their research faced significant limitations. To start with, they mentioned, “The sample size was small and heterogeneous, thus, possibly contributing to a Type 2 error” (p. 419). This limitation was the greatest bearing in mind that they related their research to earlier research and yet used a small sample compared to that of their references. Since they knew that they experienced this limitation they had advise to future researchers on the same topic that they need to “include either larger, more homogeneous samples, or in-depth individual study using rigorous intrasubject, single subject experimental measures with replication across subjects” (p. 419). This was a surety that they faced other limitations as well including heterogeneity of the participants. This was in terms of “age, severity of autism, and cognitive abilities” making it “difficult to draw meaningful conclusions about the group as a whole” (p. 418-419). Additionally, the authors faced other limitations including the fact that “soy products may affect urinary pep tides and thus introduce confounds”, unreliability of CARS instrument despite being tested for applicability, and “parental placebo effects related to the GFCF diet effectiveness”.

It was noted that the participants might have sneaked other meals, which are not prescribed making the results unreliable.  Data was meant to be drawn from participants who were feeding on GFCF diet only. Therefore, feeding on diet, which has not been prescribed, meant that the results could be as expected and this the results were not valid. Elder et al (2006) concluded in their research that GFCF diet did not have any relationship with decreasing autism disorders. However, bearing in mind that the participants might have been fed on other diets leaves a room for different results in the future.

Elder et al (2006) proposed, “Future research should include a combination of direct observational methods as well as several additional instruments suited to repeated measures designs that have well-established psychometrics” (p. 419). This is a proof that their methods did not give room for repeated measures making it impossible to analyze their data as expected. Considering the number of limitations and their strength, Elder et al (2006) leaves some questions to be answered on the reliability of the analyzed results.

Identification of Future Directions

Elder et al (2006) knew that their research was not successful but only opened a way for further research. They identified all possible limitations faced and gave advice on what ought to have been done. To start with, they argued, “Future research should include a combination of direct observational methods as well as several additional instruments suited to repeated measures designs that have well-established psychometrics (p. 419) implying that there were other better methods that could be used and help come up with the better results. They made sure that they identified any limitations they faced making it possible for future research since researchers should not repeat same mistakes.

Further research should be done with participants living in distant areas. They need to be monitored by the researchers. Further, sample size should be increased, select participants with similar characteristics and prolong research period. Since a solution to autism disorder has not been identified yet, further research needs to be done in which the possibility of the diet raising the levels of urinary peptide, which were measured in the study. All these need clarification, which can only be done through future research. The fact that their data lacked clarity calls for future research in which clarification is done on the analysis made.

Reference

Elder, J. H., Shankar, M., Shuster, J., Theriaque, D., Burns, S. & Sherrinl, L. (2006). “The Gluten-Free, Casein-Free Diet In Autism: Results of a Preliminary Double Blind Clinical Trial.” Journal of Autism and Developmental Disorders, 36(3): 413-420.