Analysis of Gluten Free Casein Free Diet, Article Critique Example

Abstract

This paper evaluates the methods section presented in the article by Elder J., Shankar M., Shuster J., Therique D., Burns S., and Sherrin L. on the Gluten-Free, Casein-Free Diet in Autism.

Evaluation of Methods

Sample

Demographic characteristics:The authors had potential biases in the demographics for their study.  To start, the authors do not include some demographic information regarding the sampled individuals. They do not report the location or cultural background of any of the individuals. In addition, the authors used only 3 girls in this study, versus 12 boys. It is possible that gender can play an important role in diet and autism and future studies by these authors should include a fair gender analysis, if possible, to generalize the results.   The individuals are also different ages with a mean age of 7.32, with the youngest age being 2 and the oldest age being 16. The use of this wide range of age can results in a data error.  The author does not state if there has been previous research on the differences in autism in age groups of children, especially children undergoing the Gluten-Free and Casein-Free diet.  Furthermore, the individuals included in the study were mostly Caucasian, with only one Asian individual; therefore, it is not representative of the entire population.  The authors do indicate, however, that each individual that was included in the study was based on a diagnosis of an autistic disorder according to DSM IV criteria and according to the score above the cut-off on each symptom domain of the Autism Diagnostic Interview. The authors do not address each individual that was included in the study in regard to their scores according to the DSM IV criteria or the Autism Diagnostic Interview. Therefore, again, it is possible there is a bias and that the sample is not representative of the entire autistic spectrum of diagnosed individuals.

Sample size: Elder et al., (2006) did not indicate how they determined the sample size they used during the study. The sample size was only 15, which can cause statistical analysis to have a low power. Therefore, the small sample size in this study could cause sampling error. Elder et al., (2006) used children ranging from age 2-16 with a mean age of 7.32 years. This sample size is limited for a couple of reasons. For one, the sample size is small. Only 15 children represented a population of autistic children that participate in the Gluten- free  Casein-free Diet.  For instance, the sample size in the study described in Elder et al., (2006) conducted by Cade et al., 2000, is 270 individuals. The study hypothesized that schizophrenia is associated with the body’s absorption of “exorphins” contained in gluten and casein.  Elder et al., (2006) also includes in the paper a study related to GFCF diet in autism where antibodies were measured in 50 children diagnosed with autism. In addition, Elder et al., (2006) proceeds to include Arnold et al., (2003) in which included 26 children with autism on a regular diet, 10 on a gluten and casein free diet and 26 children with developmental delays to serve as controls.  In sum, the authors analyzes these previous studies and indicates that the studies did provide information on GFCF dietary effects, but also indicates that the information is limited due to a small sample size. In addition, Elder et al., (2006) states that there needs additional clinical trials due to the previous smallsampling size studies; however, their study only includes 15 children, an even smaller sample size than the previous studies indicated.

Sampling techniques for the sample: The authors used two instructed to sample and describe the individuals in the study. The techniques were CARS and the Autism Diagnostic Interview-Revised (ADI-R).  The authors used the Childhood Autistic Rating Scale (CARS) to assess the autistic features in each individual. The CARS is specifically designed for autism and therefore, it is an appropriate method for this study. In addition, the authors do specify the exact profiling for what is being assessed, such as the individual relationship with others, imitation, emotional expression, body use, peculiarities in object use; resistance to change, visual, auditory, and tactile responsiveness, anxiety, verbal and nonverbal communication, activity level, and intellectual ability. The ADI-R technique for sampling was used to assess autism in the each individual by interviewing the caregivers for each subject.  The authors do indicate that the ADI-R is used for differential diagnosis of pervasive developmental disorders; therefore the appropriate for use in an autistic study is unknown. The ADI-R does, however, cover areas consistent with autistic disorders, such as social interaction, communication and repetitive behaviors and stereotyped patterns as mentioned by the authors, possibly validating the appropriateness to use the technique in the study.

Use of volunteers:The study does not indicate the use of any volunteers. The authors state that for the CARS portion of the sampling, an evaluation team completed the assessment on a scale using 15 items to which they reached aconsensus on the assessment for each individual. The authors do not state how many people were on the team and whether or not they were professionals who work with autistic children. In addition, the authors indicate that the first author in the paper, Elder, was the individual who trained on the ADI-R and assessed the caregivers for the study. This is indicating that only one person was responsible for assessing each caregiver. In one instance, this can show consistency with results; however, this can also lead to a bias opinion on the evaluation or assessment for those individuals.

Procedures

Type of research design: The authors conducted a randomized double blind repeated measures crossover study.  The authors performed techniques to evaluate each potential child. After assessment and parental consent the authors randomly assigned children to either the Gluten-Free Casein-Free Diet or a placebo diet. The authors made sure that the children, parents and the investigative team except for the data manager and dietician were unknown to the diet.  The diet for each child participant was monitored under the direction of the University of Florida’s GCRC Bionutrition staff.Each individual was provided meals and snacks from GCRC’s metabolic kitchen for 12 weeks.In addition, the authors indicate the food was based on the child’s food preferences.The authors do not indicate if the difference in food between individuals could alter results.In addition, the authors do not indicate the quantity of food difference in older subjects versus younger subjects. One other issue with the study was that the parents were in control of feeding the children and were dependent for giving the information to the researchers. Therefore,the researchers had to rely on the parent’s observations.  In addition, the researchers had to rely on whether or not the parents were recording accurate information on the food intake.

The authors used different instruments to measure dependent variables. In addition to using the CARS to screen individuals, the authors used it to make a baseline assessment and to compare the value to the control and experimental dietary conditions. The authors also took urine samples to test for peptide levels at specific days during the trial period. The samples were each analyzed for gluten and casein levels. The only problem in this part of the design is that some individuals were older children. There is no indication if levels vary with age.  The study also incorporated an Ecological Communication Orientation (ECO) Language Sampling Summary which recorded child behavior and In-Home Observation. These two protocols were previously used in studies conducted by the first author Elder. Elder (1995); Elder & Goodman, 1996; Elder, et al., 2003). A research assistant for the study was trained in videotaping techniques as well as unobtrusiveness during home visits. It does not indicate if the same research assistant was used for each individual or during each home visit. In addition, the authors indicate that trained coders were used to assess the behavior of both the parent or caregiver and child.  Again, it does not indicate if there were different coders for each subject. The authors do indicate that the coders did not know the status of the individuals in regard to the type of diet the subject was receiving during the trial (placebo vs. GFCF).

In order to analyze the results, the authors conducted a two-sided, two sample T-test. The authors do not do a good job of describing the methods for the two-sided, two sample T-test. The readershould be able to read the methods and be able to repeat the experiment; however the methods for analysis are confusing. For instance, the authors state, “Irrespective of the treatment order assignment, the dependent variable was the value in period 2 minus the value in period 1”. This is confusing to the reader. The authors then proceed to state that the two-sample T-test has an advantage over the one-sample T-test; however, if the reader doesn’t understand how they conducted the two-sample T-test, it is hard to understand the advantages.

Control group: The exact procedure or methods for the type of food for the placebo subjects was not clearly indicated or described by the authors.

Consistency of design with purpose and hypothesis: The design of the experiment itself is consistent with the purpose of the study. The goal of the researchers was to evaluate the effects of gluten-free, casein-free diet on the subject’s autistic symptoms using urinary peptide levels. In addition, the researchers wanted to evaluate the role of parent’s behavior and placebo effects in regard to the gluten-free, casein-free diet. The researchers did accomplish these goals in that they assessed each individual child’s behavior, as well as parental behavior and formulated a consensus on the child’s autism. In addition, the researchers were able to collect urinary samples to check the peptide levels on each individual subject throughout the study.

Clarity of defining independent variables:The dependent variables in the study were clearly defined, as stated in the Methods section. The authors have an entire section dedicated to measuring the dependent variables, “Instruments and Procedure for Measuring Dependent Variables”. The independent variables or I.V., however, are not clearly defined.  The reader then needs to come to his or her own conclusion on which variable in the experiment is independent. As scientists we know that the independent variable is part of the experiment that is manipulated, which would lead us to think that the dietary intake of food was the I.V.

Validity of Instrument: Elder et al., (2006) did provide information about the validity of the instruments used in their study.  The authors specifically state for CARS that “the instrument’s validity has been assessed as good under various conditions. It’s interrater reliability has been well documented and the CARS manual reports a test-retest reliability of .88 on total CARS scored based on 91 cases over a 12 month-period”.  In addition, in regard to the ADI-R test, the authors indicate that the interrater reliability scored .62-.89. This assures the reader that the techniques have been previously used and applied to these types of studies and therefore has validity.

Threats to external validity: Threats to external validity included compromised sample size or population and gender for the study. The results cannot be generalized with a degree of confidence. The instruments, dependent variables and independent variables, however, did show reliability for this study.Elder et al., (2006) states that reliability and validity of the ECO Scales was established through previous research with an interrater score at .93.  Furthermore, the authors indicate that The In-Home Observation analysis from the coders had an interrater score at .82 to 1.0.

The study could have also been exposed to ecological validity. The In-Home Observation study may have been a concern in this study because the parents were interviewed. It is possible that the parents may have consciously or unconsciously misrepresented the truth. As it is hard to avoid this type of results, they did have a control group which increases the validity of the experiment.

Threats to internal validity:Threats to internal validity include variations in the independent variable. The independent variable is assumed to be the dietary intake of the subjects. There were variations in the independent variable as the authors indicate that the types of food given to the subjects varied depending on the preference of the subjects. Therefore, each subject did not receive the same exact meal or snack. In addition, the authors do not indicate whether or not the change in the preference affected the type of nutrition each subject was receiving.

Instrumentation

Description of Instrument and types of measures: Elder et al., (2006) adequately described the two main instruments used in the study to evaluate and assess children for autism. The authors provide the name of each scale and the validity of each instrument. In addition, Elder et al., (2006) describes his use of the instruments in previous studies. The authors use four inventories to assess the dependent variables in order to find compatible subjects for their study: Childhood Autism Rating Scale (CARS), Urinary Peptide Levels (UPL), and Ecological Communication Orientation Language Sample Summary (ECO). Responses to each scale were concluded in consensus with a team of researchers. The inventories can provide the researchers with a great deal of information; however, the subjects may not provide accurate information resulting in data error and the results of the study can be interpreted with some invalidity.

Standardization of instrument and appropriateness for sample: The scales used by the authors in this study and the degree to which the scales were standardized were not indicated.  Furthermore, the authors do not compare the demographics to a larger population. This can compromise the study, as well of the results. In addition, the study subjects may have been a biased sample because they were mostly Caucasian with only 3 girls, the rest males.

Appropriateness of sampling techniques: The authors do not exactly state where the recruitment of subjects came from. In addition, although the type of research design was random, the selection of the subjects were not random because they needed to be assessed before chosen to participate in the study; therefore, the study may not have been as representative as it could have been.

Instrument Reliability: The authors do report reliability on the instruments they used. For instance, the ECO Scale reliability is reported as an interrater agreement of 91.2, stability of .28-.86 and an internal consistency using split halves method at .93. The reliability is also reported for CARS and ADI-R as .88 and .89, respectively. These measures are adequate as the methods have been conducted in previous studies and documented in primary literature. The authors used these methods for those specific reasons.

References

Arnold, G. L., Hyman, S. L., Mooney, R., & Kirby, R. S. (2003). Plasma amino acids profiles in children with autism: Potential risk for nutritional deficiencies. Journal of Autism and Developmental Disorders, 33(4), 449-454.

Cade, R., Privette, M., Fregley, M., Rowland, N., Sun, Z., Zele, V. et al. (1999). Autism and schizophrenia: Intestinal disorders, Nutritional Neuroscience, 3, 57-72.

Elder, J.H., Shankar, M., Shuster, J., Theriaque, D., Burns, S., Sherrin, L. 2006.  The Gluten-Free, Casein-Free Diet in Autism: Results of APremliminary Double Blind Clinical Trial. Journal of Autism and Developmental Disorders. 36(3):413-420.