Core Components of the Metabolic Syndrome in Nonalcohlic Fatty Liver Disease
Industrial-Project-Report
1. DIETARY MODIFICATIONS OF
AGE PRODUCTS IN
NON ALCOHOLIC FATTY LIVER
DISEASE
Abstract
Demonstrating the methodology of evaluating whether dietary
restriction of Advanced Glycation Products (AGE’s) can
significantly manage Non-Alcoholic Fatty Liver Disease (NAFLD)
based on simulated data.
Evangelos Matselis (3508326)
Pulasthi Gunasekara (3492351)
2. 1
ACKNOWLEDGEMENT
This project consumed vast amount of work, research and dedication. Still, implementation would
not have been possible if we did not have a support of many individuals and organizations.
Therefore we would like to extend our sincere gratitude to all of them.
First of all we are grateful to Ms. Kaye Marion, Dr. Stelios Georgiou and Dr. James Baglin from RMIT
University for providing necessary support and guidance concerning project implementation.
Without their superior knowledge and experience, the Project would lack in quality of outcomes, and
thus their support has been essential.
We are also thankful to Dr. Chris Leung and Laurence Alexander Jacuzzi from Austin Hospital for
technical support in the implementation and for providing necessary information regarding the
project.
Nevertheless, we express our gratitude toward our families for their kind co-operation and
encouragement which help us in completion of this project.
4. 3
INTRODUCTION
Nonalcoholic fatty liver disease (NAFLD) is one of the most common causes of chronic liver disease.
It encompasses a spectrum of conditions associated with lipid deposition in hepatocytes. It ranges
from steatosis (simple fatty liver) to nonalcoholic steatohepatitis (NASH–fatty changes with
inflammation and hepatocellular injury or fibrosis), to advanced fibrosis and cirrhosis. Studies
suggest that although simple fatty liver is a benign condition, NASH can progress to fibrosis and lead
to end-stage liver disease. The disease is mostly silent and is often discovered through incidentally
elevated liver enzyme levels. It is strongly associated with obesity and insulin resistance and is
currently considered by many as the hepatic component of the metabolic syndrome (Carey et al.,
2013).
Advanced glycation end-products (AGE) are known as substances that can be factors in aging and in
the development or worsening of many degenerative diseases such as diabetes, atherosclerosis,
chronic renal failure, alcoholic fatty liver disease etc. They form in foods following processing and
cooking at high temperatures and are generated spontaneously in cells; however, under conditions of
hyperglycemia and lipid peroxidation, their levels are higher than usual, which contribute to the
development of diseases such as the nonalcoholic fatty liver disease (NAFLD). Currently, the
recommended therapy for NAFLD consists of lifestyle modification, specifically dieting, weight
reduction and increased physical activity (Thoma et al., 2012).
As the dietary modifications are considered as one of the therapies for controlling NAFLD, it is
believed that Low-AGE’s diet leads to improvement of patients’ health while High-AGE’s diet is
expected to aggravate it. However, it is yet controversial whether the dietary restriction modifications
provide significant improvement of the serological markers of NAFLD - diagnosed patients.
PROJECT OBJECTIVE
This study will demonstrate the methodology of evaluating whether the dietary restriction of
advanced glycation products (AGE’s) can significantly manage non-alcoholic fatty liver disease
(NAFLD) based on simulated data.
DATA COLLECTION
Study participants are required to participate in this study for a total of nine weeks. Each participant
undergoes a process that consists of four different stages.
During the initial stage, patients are educated about the study and the process. After the initial stage,
Low-AGE’s diet period, wash out period and the High- AGE’s diet period follows, running for two
weeks each. Screening tests such as serum, urine and blood test, height, weight, BMI measurements
etc. are carried out before after each diet period.
5. 4
METHODOLOGY
Sample size Determination
Determining the appropriate sample size is of extreme importance, especially in the clinical area, as
data collection is often a costly procedure and the results and conclusions have to be as valid as
possible. According to the client’s specifications, a clinically significant result should have a standard
deviation of paired differences of 6, a difference in means of 2.7 and a statistical power of 0.8. Based
on these specifications, a power test was carried out and resulted in a sample size of 41 patients, to
achieve a probability of 80 percent that the study will detect a treatment difference at a two-sided
significance level of 0.05 (See Appendix C, Figure 1).
Data simulation
The data provided by the client were not enough to meet the sample size of 41 determined
previously. The dataset consisted of 10 patients with only 5 of them meeting the inclusion criteria
(For inclusion criteria see Appendix A). An analysis of such a small sample size could result in weak
or even not valid results and conclusions. Also, analyzing a sample of size greater than 30 is different
from analyzing a sample of size 4. Thus, it was decided to demonstrate the methodology of
evaluating whether the dietary restriction of AGE’s can significantly manage NAFLD based on
simulated data.
Simulating the data may not be the most important part of this study, as such a procedure will not be
required when the data will be collected. However, analysing and simulating appropriate and
realistic data proved to be a very time-consuming, as well as a very interesting process.
As mentioned above, each patient undergoes a set of tests before and after each diet period. There is a
large set of variables to be measured during these tests. In order for the simulated data to be as
realistic as possible, all measurement results for all the variables in all different treatments for each
patient should be associated. Apart from that, correlation should also exist patient-wise for each
variable and treatment. To achieve that, multivariate normal distributions were simulated. The
multivariate normal distribution is used to describe, at least approximately, any set of (possibly)
correlated real-valued random variables, each of which clusters around a mean value. In simple
terms, instead of each treatment column being individually simulated using their mean and standard
deviation, which would result in no association between the treatment results, all treatment columns
for all variables were simulated together, using a vector of the means of each treatment and their
covariance matrix. However, it has to be mentioned that in order to simulate multivariate normal
distributions, non-normality of some of the data had to be ignored.
As stated earlier, the original dataset consisted of 10 patients, with only five of them meeting the
inclusion criteria, as the rest were diabetics. The two-sample t - tests that were carried out between
the diabetics and the non-diabetics did not provide evidence of any significant difference in the
means between these two groups for all the variables and all the treatments except for glucose and
HbA1c (See Appendix B: Table 1). Thus, by ignoring the mean difference in the two variables
6. 5
mentioned before, the means that were used for the data simulation were calculated from all 10
patients, regardless if they met the inclusion criteria or not, in order to have a better estimate of the
means.
As the simulation was based on the covariance matrix the simulated data appeared to be similar to
the original data (See Appendix B: Table 2). It has to be declared that the “Lipase” marker was not
included in the simulation and the analysis as it contained incomplete data. Furthermore some of the
markers were not simulated as they are functions of other markers.
Analysis
To achieve a sample of size 41, 36 out of the 1000 simulated data were randomly selected along with
the results of the five patients that met the inclusion criteria. Before proceeding with the analysis, the
following assumptions had to be met,
Assumption 1: Dependent variable should be continuous
Assumption 2: The same subjects should be present in all stages
Assumption 3: The distribution of the differences between the two related groups should be
approximately normally distributed.
Assumptions 1 and 2 were obviously met, as our markers were measured at the interval or ratio level
and all patients were present in all stages. Regarding assumption 3, since the sample size was greater
than 30, the Central Limit Theorem (CLT) was invoked and it could be assumed that the differences
were approximately normal. As CLT claims, even for non – normal data the distribution of the
sample means is approximately normal, no matter what the distribution of the original data is, as
long as the sample size is larger than 30 and all samples are of the same size.
Since all assumptions were met, the analysis was carried out and paired-sample t-tests were
conducted for each variable to compare measurements between the following stages:
Results before and after Low AGE’s diet period (Pre-Low AGE’s vs Post Low AGE’s)
Results before and after High AGE’s diet period (Pre-High AGE’s vs Post High AGE’s)
EXPECTATIONS
The clients’ expectations for this research can be summarised as follows:
Anthropometric markers
No significant change after Low AGE’s Diet period
No significant change after High AGE’s Diet period
Metabolic and Other Markers
No significant change or significant decrease after Low AGE’s Diet period
Significant increase after High AGE’s Diet period
7. 6
RESULTS
Prior focusing on the results it is worth drawing the attention on the assigned research question
which was whether the dietary restriction modifications provide significant improvement of the
serological markers of NAFLD - diagnosed patients.
The significance of the mean differences of each serological markers for the two diet types were
tested and recorded (See Appendix B: Table 3 & Appendix C: Figure 2).
The p-values of the tests represent the probabilities of mistakenly rejecting the null hypothesis. For a
test to show evidence of significant mean difference, the p-value must be less than 0.05.
The hypothesis for the paired sample t-tests was the following:
𝝁 𝜟 = 𝟎
𝝁 𝜟 ≠ 𝟎
∗ 𝝁 𝜟: 𝒑𝒐𝒑𝒖𝒍𝒂𝒕𝒊𝒐𝒏 𝒎𝒆𝒂𝒏 𝒐𝒇 𝒕𝒉𝒆 𝒅𝒊𝒇𝒇𝒆𝒓𝒆𝒏𝒄𝒆𝒔
The results can be summarized as below under each serological marker category
Anthropometric Markers
Out of 4 anthropometric markers only the “Waist” behaved as the clients’ expectations.
Metabolic Markers
Out of 6 metabolic markers none of the markers behaved as expected.
Other Markers
5 out of 16 markers behaved as expected.
“Platelet”, “NAFLD Fibrosis Score”, “Bili.Tot.”, “Bili.Unconj.”, “GGT”
CONCLUSION
Although most of the markers showed a signifcant change in their means after at least one diet type,
only 6 out of 26 markers tested, showed a significant improvement in patients health according to
clients’ expectations, whereas the rest did not perform as expected.
Hence based on the above and to the analysts’ understanding of the industry, it can be concluded
that dietary restriction modifications do not seem to have significantly improved the NAFLD -
diagnosed patients’ health.
8. 7
However it is worth mentioning that the above conclusion may alter based on the importance levels
of the above six serologcal markers to the patient’s health which is yet to be decided by the client.
SUGGESTIONS
Should the client believe that there is an associaltion between some of the markers, then a test of
bivariate associations (Pearson’s product moment correlation) between all the markers for each stage
could be conducted, as a change in one marker may cause an impact in the behavior of another.
Furthermore, as in the current study the patients are just assigned diet schedules for the two diet
periods, providing them with pre-prepared diets produced by the researchers could result in less
deviation from the appropriate diets and more reliable conclusions.
Finally, more detailed analysis could be undertaken regarding the significance level (alpha value),
which was set at 0.05 by the client. This means that 5 percent of the time, a true null hypothesis will
be rejected. Testing multiple hypothesis over the same data, increases the probability of falsely
rejecting at least one of the null hypothesis, which could result in misleading conclusions. It is highly
recommended that an adjustment of the alpha value, such as the Bonferroni Correction, will be
considered in future studies. The Bonferroni Correction preserves the overall significance level, by
dividing the significance level with the number of comparisons, but it can become extremely
conservative if the number of tests considered increases, which means that rejecting any individual
null hypothesis could become more difficult than it should be.
9. 8
REFERENCES
1. Carey, E., Wieckowska, A. and Carey, W. (2013). Nonalcoholic Fatty Liver Disease. [online]
Clevelandclinicmeded.com. Available at:
http://www.clevelandclinicmeded.com/medicalpubs/diseasemanagement/hepatology/nonalco
holic-fatty-liver-disease/ [Accessed 29 May 2016].
2. Santos JC, e. (2013). Development of nonalcoholic hepatopathy: contributions of oxidative stress and
advanced glycation end products. - PubMed - NCBI. [online] Ncbi.nlm.nih.gov. Available at:
http://www.ncbi.nlm.nih.gov/pubmed/24084729 [Accessed 29 May 2016].
3. Mouzaki M, e. (2013). Intestinal microbiota in patients with nonalcoholic fatty liver disease. - PubMed -
NCBI. [online] Ncbi.nlm.nih.gov. Available at: http://www.ncbi.nlm.nih.gov/pubmed/23401313
[Accessed 29 May 2016].
4. Semba, R., Gebauer, S., Baer, D., Sun, K., Turner, R., Silber, H., Talegawkar, S., Ferrucci, L. and
Novotny, J. (2014). Dietary Intake of Advanced Glycation End Products Did Not Affect
Endothelial Function and Inflammation in Healthy Adults in a Randomized Controlled Trial.
Journal of Nutrition, 144(7), pp.1037-1042.
10. 9
APPENDICES
A)
Inclusion Criteria for the Research
Males and Females aged between 18 and 70
Imaging of Biopsy-diagnosed
Non-diabetic
B)
Table 1: Two sample t-test for means between diabetics and non-diabetics
Table 2: Comparison of original and simulated data for fat
11. 10
Table 3: Paired t –test results for mean difference between diets
C)
Figure 1: Power test for sample size determination
