Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

Erythrocyte Superoxide Dismutase, Glutathione Peroxidase, and Catalase

Activities and Risk of Coronary Heart Disease in Generally Healthy
Women: A Prospective Study
By: Shuman Yang, Majken K. Jensen, Eric B. Rimm, Walter Willett, & Tianying Wu
American Journal of Epidemiology, Volume 180 (9), pp. 901-908, 2014.
INTRODUCTION:
Review of literature--Is this adequate so that you can understand why the study was
done?

The authors Yang et al. examined the literature surrounding the question of the
predictive ability of levels of serum antioxidant enzymes superoxide dismutase
(SOD), glutathione peroxidase (GPx), and catalase (CAT) and coronary heart
disease (CHD) risk. They found that a meta-analysis had been conducted that
consisted of 42 case-control and three cohort studies suggested that there was an
inverse relationship between the activity of SOD, GPx, or CAT and risk of CHD.
However, in the majority of the studies included in the meta-analysis, the blood
samples used to measure the activity of the antioxidant enzymes had been taken
post-non-fatal, acute, coronary events, and such studies had small sample sizes.
Only one existing prospective study had examined the association between an
antioxidant enzyme (GPx) and the risk of myocardial infarction (MI) in a
generally healthy population. Thus, the authors determined that there was a need
for a prospective study with a large sample size of generally healthy subjects to
determine if there was an association between serum levels of SOD, GPx, and
CAT and incidence of CHD. The authors determined that the effect of fasting on
the aforementioned antioxidant enzymes needed to be assessed, give that there has
been evidence showing that fasting blood samples improves the ability of some
biochemical measurements in predicting risk of CHD.

Rationale--Is the reasoning behind the study objectives or hypotheses clearly described?
Given the background information, what is the next logical step in elucidating the
problem?

The authors clearly stated that their study assessing the predictive ability of SOD,
GPx, and CAT was needed, given the state of current research. They also showed
that while fasting has been proven to influence the hepatic synthesis of
antioxidant enzymes, this association has not been examined in human subjects.
In addition, the effect of fasting on the ability of such enzymes to predict CHD
has not been examined. The authors showed that the next logical step in
elucidating the predictive ability of SOD, GPx, and CAT on future CHD was to
determine if such association was influenced by the duration of fasting prior to
blood sample collection.

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

Hypotheses or Objectives--Are either hypotheses or objectives clearly stated? These

should be specific enough to lead directly to the methodology.

The authors clearly stated their hypotheses that by using a prospective study
design the authors were going to examine the association of the activity of SOD,
GPx, and CAT and CHD events in the future, and if any such association was
influenced by the fasting duration prior to the collection of the blood samples.
With these hypotheses and with the information laid out in the literature review
the authors indicate the need for a large sample-size prospective study of
generally healthy individuals. For this purpose, they chose the Nurses Health
Study (NHS), a large-scale, prospective study with 121,701 enrolled free-living,
generally healthy subjects.

STUDY DESIGN AND METHODS

Design: Answer as many of the following as possible:
General approach: descriptive or hypothesis testing?

The general approach of this study was hypothesis testing.

Level of measurement: individual or ecological?

This study had an individual level of measurement.

Specific design: cohort, historical cohort, cross-sectional, case-control, experimental,

quasi-experimental, other.

The specific design of this study was case-control.

Population at Risk: What is the population from which the study sample has been
selected? This needs to be specified to determine to which population (if any) the results
may be generalized. This is sometimes called external validity.

The population that this study sample was taken was from the NHS, a study of
121,701 women aged 30-55 years, employed as nurses in 1976 in the United
States. The study used participants from the NHS 43-70 years of age who were
free of cardiovascular disease at the time, and who had provided blood samples.
The population to which these results can be generalized is United States females
of 43-70 years of age, with nursing-level education and employment, and includes
smokers and non-smokers, and females of varying BMIs, physical activity levels,
degree of alcohol intake, pre- or post-menopausal status, family history of MI,
personal histories of presence or absence of diabetes and hypertension, use of
hormone replacement therapies, blood triglyceride and cholesterol statuses, and
levels of C-reactive protein. Men, women of les than 42 years of age, and
individuals with a presence of cardiovascular disease were not included in this
study, and thus the results are not able to be generalized to these sub-groups of the
greater United States population.

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

Sampling Method and Sample Size: How was the study sample selected from a defined
population at risk? What sampling method was used? What was the resulting sample
size?

Out of the 121,701 women enrolled in the NHS in 1976, 32,826 provided fasting
(at least 8 hours post feeding) blood samples to the researchers between 1989 and
1991. The participants had to be free of cardiovascular diseases and cancers at the
time of the blood draw. These 32,826 participants were then followed from 1989
to 2004 (for a median of eight years and for a maximum of 15), and upon followup, 474 new cases of CHD were identified. CHD included non-fatal MI, and fatal
CHD. Non-fatal MI was defined in accordance with the WHO criteria of: MI
symptoms plus either elevated cardiac enzyme levels, or diagnostic
electrocardiographic changes. Fatal CHD was determined if autopsy or hospital
records confirmed CHD as cause of death, if death certificate indicated CHD as
cause of death, if underlying CHD was the most plausible cause of death, or if
there was prior evidence of CHD available. Physicians, who were blinded as to
the baseline risk factors of the participants, conducted all documentation. After
the 474 cases had been identified, the researchers used risk set sampling at the
time of the diagnosis of the cases to randomly select controls in a ratio of 2:1. The
factors used were within two years of age, smoking status, within two months of
blood collection, and fasting status (< 8 hours post feeding or 8 hours post
feeding). The final analysis had a sample size of 1,093 subjects, consisting of 365
cases and 728 controls, representing 290 of the original matched case-control sets.

Selection and Exclusion Criteria for Study Group: What criteria were used in selecting
the sample from the population at risk? After initial selection, what criteria were used to
exclude individuals? What possible selection bias might have occurred?

Out of the 32,826 participants of the NHS from whom blood samples were drawn,
all with existing CHD and cancers were excluded. After the initial 474 cases were
identified as described above, the researchers randomly selected controls in a ratio
of 2:1, matching factors of age, time of blood draw, smoking status, and fasting
status. Out of these case-control sets, only 290 were used. There is no mention of
how the 184 case-control sets were excluded. This leads to an unknown level of
selection bias, as the criteria used to choose the 290 case-control pairs used for
analysis was not given.

Dependant variable: What was/were the dependant variable(s) according to the study
objectives or hypothesis, and how were they measured?

The dependent variables in this study were incidence of future CHD events as
determined by fatal CHD or non-fatal MI in the subjects, and the relative activity
of the antioxidant enzymes SOD, GPx, and CAT based on levels of fasting prior
to blood draw. The CHD events were measured as documented by physicians (in
the non-fatal MI cases), and by hospital and autopsy records, and death

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

certificates. The antioxidant enzyme activities were measured individually, using

SOD assay kit number 706002, GPx assay kit number 703102, with the
erythrocyte lysate measured at different wavelengths, and CAT activity was
measured using method first described by Beers et al. (later modified by Aebi), in
which the erythrocyte solution was treated with hydrogen peroxide, and
wavelength was subsequently measured.
Independent Variable: What were the key independent variables, and how were they
measured?

The independent variables were the level of antioxidant activity, measured as

described above, and the amount of time the subjects had fasted prior to blood
draw, as reported by the subjects themselves.

Potential Confounding or Effect-Modifying Variable: Make a list of these, including

those that have authors have considered plus other you think may be important. Group
them in 3 categories:
- Controlled by study design (by matching, randomization, etc.)

Age, smoking status, month of blood collection, and fasting status (length of time
before the blood draw) were controlled for by matching.
- Controlled by analysis

BMI, physical activity levels, levels of alcohol intake, pre- or post-menopausal

status, family history of MI, personal history of diabetes, personal history of
hypertension, use of hormone replacement therapy, serum cholesterol, serum
triglycerides, and levels of C-reactive protein were all controlled for by analysis.
- Not controlled

Race/ethnicity and prescription pharmaceuticals were not controlled for, nor was
time of onset of CHD.

Information Bias Controlled by Study Design: Did the study design try to prevent or
minimize information bias, e.g. through blinded measurement?

Information bias was controlled for by study design, because the physicians doing
the measurements of CHD in the study subjects were blinded to the baseline risk
factors that the study participants reported.

RESULTS
Missing Data: Evaluate the results carefully for completion rate, drop-outs, compliance,
or other types of missing data. How have the authors addressed these issues? Can you
identify any problems that the authors have not noted?

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

The completion rate for this study was the 1,093 participants included in the final
analysis. Since this was a nested study in the NHS, there was no mention of
dropouts. Compliance was not addressed in the discussion of the study, and an
error in compliance could have occurred if the participants were not truthful about
the time of their last meal prior to blood draw. There does not appear to be other
forms of missing data.

Major Findings: Briefly summarize the major results of the study in terms of magnitude
of effects, risk estimates, etc.

There was a null association found between the levels of antioxidant enzyme
activity and future CHD events. There was a significant association between SOD
activity and CHD in participants who had fasted for fewer than 12 hours prior to
blood draw (RR 1.25, 95% CI: 1.02-1.52), however this association was not seen
in those participants who had fasted for greater than 12 hours prior to blood draw.
This is a very small increase in risk, and there is not much that can be elucidated
from it, as overall the markers of SOD, GPx, and CAT were not associated with
increase in risk for CHD.

Control for Confounding: How were variable from your lists above examined in the
analysis, e.g. stratification, adjustment, or other multivariate methods? Which were
identified as important confounders?

Conditional and unconditional logistic regression models were used to examine

the association between SOD, GPx, and CAT and increased future incidence of
CHD. Covariates in the unconditional models included the matching factors of
age, prior or current smoking status, time and month of blood collection, in
addition to the factors assessed to be confounders, including BMI, levels of
physical activity, alcohol intake, pre- or post-menopausal status, family history of
MI, personal histories of presence or absence of diabetes and hypertension, use of
hormone replacement therapies, blood triglyceride and cholesterol statuses, and
levels of C-reactive protein. The conditional models included all of the factors
motioned above, but excluded the matching factors.

To determine the association between length of time fasting prior to blood draw
and antioxidant enzyme activity, the data was stratified into six groups of varying
lengths of fasting time, and used unconditional logistic regression models to
examine the association of interest. The researchers treated the enzyme activity as
continuous variables and the length of time spent fasting as categorical variables
to determine the final cut-point of length of time spent fasting, using the Wald P
value for assessing significance for the interaction. The fasting strata with similar
results were then collapsed, and the association of RR was assessed.

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

Statistical Analysis: Were the methods of statistical analysis appropriate and clearly
described? Did the authors consider statistical significance, power, and/or confidence
intervals? Were these used appropriately?

The statistical methods used in the analysis were appropriate and clearly
described. The authors appropriately noted the confidence interval and the
statistical significance of every analysis performed.

DISCUSSIONS AND CONCLUSIONS

Internal Validity: Did you draw the same conclusions from the results as the authors?
What alternative orders of explanation (e.g. systematic or sampling error) were
considered by the authors and what others can you identify? Are these problems serious
enough to jeopardize the internal validity of the study?

The internal validity of this study appears to be adequate; there are not any
apparent threats to the internal validity.

Given the statistical analysis presented, I drew the same conclusions as the
authors. There was a null association between the antioxidant enzyme activity
level and future incidents of CHD, and this was clearly laid out in the tables
presented. The authors proposed that this null association was seen because there
is no predictive ability of the activity of the antioxidant enzymes and future CHD,
even if the activity of such enzymes changes post-cardiovascular event. Given the
size of the study and the single blood draw, I believe that the authors correctly
interpreted the data they found.

There was an increase in the RR of CHD in the subjects who had fasted for less
than 12 hours prior to blood draw, and this association was also apparent, and was
statistically significant. The authors identified two possible hypotheses for this
result, the one being that the increase in SOD and CHD risk in those fasting less
than 12 hours may indicate an activation of the bodys antioxidant defense
system, which resembles a pre-diabetic state, indicating perhaps that those
with CHD in the future experience greater post-prandial oxidative stress than
healthier subjects. The second hypothesis is that SOD (the only enzyme that
showed the positive association) is an antioxidant enzyme that is seen before GPx
or CAT, and this may explain why there was not a similar association seen with
these two enzymes. It is possible that the diet of those predisposed to CHD is such
that within 12 hours of eating they experience higher levels of SOD, and thus a
fasting blood draw may be an indicator of CHD risk in the future, however it is
not clear why this association exists, and thus needs to be researched further
before any sort of diagnostic conclusion could be drawn from this discovery.

External Validity: If you consider the internal validity of the study was adequate, it is
appropriate to determine whether the results have external validity, i.e. can they be

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

generalized or extrapolated, and if so, to what population? Refer back to what you said
about Population at Risk when you answer this.

These results have external validity, however the authors point out that the
participants of the NHS had lower than national average levels of CHD, and were
generally healthier than the national average. This means that the participants of
this study may not be representative of the population in general, and this could
limit the external validity. The results of the NHS can be generalized to those who
fit the characteristics of the participants, but this is a small sector of the
population at large, consisting of females of 43-70 years of age with the
educational and occupational level of that of a nurse, and are employed.

Consistency with Other Studies: Do the authors attempt to integrate the results with
those of previous studies? What findings are inconsistent?

The authors of the study relate their findings to the general body of research, and
indicate that their findings are consistent with one prospective study, but are
inconsistent with two others, with regards to the null association between GPx
and CHD. The authors pointed out that no other prospective study has
investigated the association of CAT and CHD, and thus it is the first of its kind.
The null association found between CAT and CHD is in disagreement with some
case control studies that found an inverse relationship between CAT and CHD.
This difference may be due to the fact that the blood collected in this study was
taken prior to the development of CHD, whereas all other CAT-CHD studies had
blood collected after the diagnosis of CHD.

Theoretical Importance of Results: Do the results add anything to existing knowledge

about the problem under study? How might these results affect public health?

These results do not have much potential to affect public health, due to the fact
that this study is the first large-scale, prospective, antioxidant enzyme CHD risk
study performed. In this way this is an important study in that it takes the next
step in understanding the way that SOD, GPx, and CAT interact with CHD. This
study raises questions as to why these enzymes may not have predictive value
prior to CHD events, and yet there is an association seen once CHD develops,
which should be explored further. As no public health changes can be made from
this research, more research is needed.

Practical Importance of Results: As a result of this study, should individuals or agencies

do anything differently from before?

There are no practical changes that can be made in light of these results, however
agencies such as the American Heart Association, the National Cancer Institute,
the National Institute for Environmental Health Services, and the National
Institutes of Health would be wise to continue funding such research so that a
clearer picture of how native antioxidants interact with the development of CHD.

Michelle Garvie Systematic Assessment of an Epidemiological Study

April 5, 2015

Further Study: What directions for further research do the authors suggest? Can you add
anything to this?

More research into the specifics bridging the gap between the existing body of
research showing that there is an association between SOD, GPx, and CAT in the
presence of CHD, but the lack of evidence of the predictive value of the enzymes.
More research should be done to understand the relationship between the higher
SOD levels and higher risk of CHD in the individuals who had fasted for less than
12 hours before blood sampling.