Evidence Based

Ophthalmology
Noel D. Atienza, MD, MSc
1. What is Evidence-based Medicine?

Evidence-Based Medicine is a systematic

approach to the acquisition, appraisal and
application of research evidence to guide
healthcare decisions.
Components of the EBM definition

A systematic approach: EBM allows practitioners to

assess new technology in an
efficient and logical manner,
…acquisition: the skill to conduct an efficient
literature search

…appraisal: the skill to criticize and decide

if results are credible or not

…and application of research evidence:

The skill to utilize the
information in the care
of specific populations

…to guide decisions Evidence, plus professional

expertise and
patients‘preferences.
2. What are the components of the EBM
Cycle?

Clinical Question
Systematic Search for “Best Evidence”
Critical Appraisal of Published Articles
Data Integration
Medical Heuristics
3. Elements of a well-built Clinical Question

Patient or Problem
Intervention or Exposure
Comparison Group
Outcome
Methodology
4. What are some important steps in
conducting an efficient literature search?
Step 1: Phrase the clinical question precisely,
(P,I,C,O, and M.)
Step 2: Rank the Concepts according to order of
importance
Step 3: Expand the post important concept to account
for variations in terminology
Step 4: Intersect the search yields (A AND B, etc..)
Step 5: Examine yield for hits and misses and modify
search if necessary
1. Which of the following is
classified as an analytical study
design?

A. Cross-sectional
B. Case Series
C. Case-Control Study
D. Ecological survey
2. In this type of study, the subjects are
recruited based on the presence or absence
of a disease / outcome

A. Case-Control study
B. Cohort Study
C. Randomized Controlled Trial
D. Case Series
3. In this type of study, the subjects are
recruited based on the presence or
absence of risk factors. They are followed
up for the development an outcome.

A. Case Control study

B. Cohort study
C. Randomized Controlled Trial
D. Case Series
4. This type of clinical study on human is
also considered as an experimental form
of research.

A. Case Control Study

B. Cohort Study
C. Randomized Controlled Trial
D. Meta-analysis
5. Randomization, blinding and intention-
to-treat analysis are hallmarks of a well
designed _____.

A. Case Control Study

B. Cohort Study
C. Randomized Controlled Trial
D. Meta-analysis
 Study Designs

Observational / Descriptive:
Case report, Case Series
Cross Sectional Studies

Observational / Analytical:
Case-Control Studies
Cohort studies

Experimental: Randomized Controlled Trials

Integrative: Syst Review / Meta-Analysis

 Cross-Sectional Study Design

Exposure status and Disease status are measured

at one point in time

Useful for chronic illnesses (gradual onset, long

duration)

Prevalence studies
Less costly than cohort studies
 Case-Control Studies

Subjects recruited based on their disease status

(+) disease or (–) disease

Retrospective study to ascertain exposure status

Useful for illnesses with a long latency period,

rare illnesses, or for evaluation of a wide range
of potential etiologic exposures (HARM studies)
 Cohort Studies

Subjects recruited based on their exposure status.

Prospective study
to determine the occurrence of an outcome.
to describe the incidence of certain outcomes
over time;
to analyze associations between risk factors and
the outcomes
Can establish a temporal sequence between
predictor and outcome variables.
 Retrospective Cohort Study:

Sample of subjects with the outcome of interest.

Similar to the prospective cohort study except that

baseline measurements, follow-up, and
outcomes all happened in the past

Only possible if there is adequate data on the risk

factors and outcome
 Clinical Trials, RCT
The direct comparison of two or more treatment
groups; uses an experimental design

Subjects randomly assigned to a treatment or to a

comparator group

Subjects and Researchers “blinded” to the actual

exposure

Outcomes are measures and compared using

statistical analysis to determine significant
effects
6. The validity of this type of study is affected
by selection bias, since the subjects are
already known to have the outcome of
interest.

A. Case Control Study

B. Cohort study
C. Randomized Controlled Trial
D. Meta-analysis
7. This type of analytical study best
demonstrates the temporal relationship
between a risk factor and the outcome.

A. Case Control study

B. Cohort Study
C. Randomized Controlled Trial
D. Meta-analysis
8. This form of study is a type of Integrative
Research that pools together data from
previous studies in order to increase power.

A. Case Control study

B. Cohort Study
C. Randomized Controlled Trial
D. Meta-analysis
9. This form of study design may demonstrate
the highest level of evidence in clinical
research.

A. Case Control study

B. Cohort Study
C. Randomized Controlled Trial
D. Meta-analysis of RCTs
 Point Estimates for Clinical Studies

Clinical Trials
RR, ARR, NNT
Harm studies
Odds Ratio (OR), Hazards Ratio (HR)
Cohort Studies
RR, Survival Analysis
Meta-analysis
Pooled RR, MD
 Interval Estimates

Example: 95% Confidence Interval

Estimate of accuracy and reliability of

statistical estimates. (Narrow, wide)

Can estimate statistical significance of

the point estimates
10. The Odds Ratio (OR) for a Harm Study
would tend to show a harmful effect if the
value is

A. Above Zero (0)

B. Less than 1.0 à Beneficial
C. Greater than 1.0 à Harmful
D. Less than zero
11. The Odds Ratio (OR) and its 95%
Confidence Interval for a Cohort Study /
Prognosis Study shows a significant harmful
effect with this value

A. 0.85 (0.55 – 1.21)

B. 2.5 (1.25 – 4.22) à Sig. Harm
C. 1.8 (0.85 – 5.45)
12. If the Odds Ratio and 95% C.I. for a
particular Harm Study shows a
significant harmful effect, we would
also expect the p-value of that statistic
to be
A. p < 0.05
B. p > 0.05
13. In Randomized Controlled Trials, the
Number-Needed-to Treat (NNT) is the
reciprocal of the

A. RR (Risk Ratio)
B. RRR (Risk Ratio Reduction)
C. ARR (Absolute Risk Reduction)
14. In RCTs, the baseline characteristics of
subjects recruited into the treatment
group and the placebo group would tend to
show no statistical differences due to

A. Double Blinding
B. Randomization / Random Allocation
C. Intention to treat Analysis
D. Placebo effect
15. This maneuver aims to limit the effect of
observer bias on the part of the outcome
assessor.

A. Double Blinding
B. Randomization
C. Intention to Treat Analysis
D. Placebo effect
Case Study 1:
To evaluate the effects of lasers on Diabetic
Retinopathy, the study subjects for the treatment and
control groups must have similar Hbaic, BP, Lipid Profile,
and Disease Severity.
(Note: Extrapolating treatment results with varying
disease severity could produce biased results. )

How will the study design limit the effects of these

confounding variables?

Random allocation, double blinding (both the patient and

the physicians / outcome assessors) – reduces the
likelihood of selection and measurement bias
Case Study 2:

Treatment effects of Trabeculectomy for glaucoma may

differ depending on the severity of disease
(for example, early versus advanced OAG)

How can surgical treatment trials be done to avoid errors

in the outcomes of treatment?

In surgical treatments, blinding may not be feasible. But

the surgical technique can be standardized as well as the
postoperative care.

Statisticians will need to statistically “control” for

differences in predictive factors affecting the outcomes of
treatment. Example: stratified analysis based on severity of
disease – reduces the likelihood of selection bias.
16. For RCTs, this phenomenon explains the
beneficial effects even among subjects who
received an inert substance instead of the
trial drug.

A. Double Blinding
B. Randomization
C. Intention to treat Analysis
D. Placebo effect
17. In the analysis of RCT data, the subjects who
dropped out or were lost to follow up must be
included in the analysis.
This procedure is called _________.

A. Analysis of Variance
B. Control of Bias
C. Intention to Treat Analysis (ITT)
D. Control of Confounding
Intention to Treat Analysis: Studies should account for
subjects who did not complete the trial. The reasons for
loss to follow-up may differ between the 2 groups.

Ø Participants in the intervention group may be more likely

to drop out than the control group if they experience
ocular adverse effects from the drug such as stinging or
burning.
Ø Subjecting the stats to a worse-case scenario could
result in a different result. If the results remain the same,
then the conclusions as “robust.”
18. A major source of bias in Systematic
Reviews that would tend to favor a beneficial
effect

A. Publication bias
B. Heterogeneity
C. Information Bias
D. Selection Bias
19. If a researcher wants to prove the beneficial
effects of antioxidants on the progression of
Diabetic Retinopathy, the best study design
would be
A. Case Series
B. Case Control Study
C. Cohort Study
D. Randomized Controlled Trial
20. If a researcher wants to prove that
preservatives contained in eye drops result in
more adverse effects, the appropriate study
design would be a ___________.
A. Case Series
B. Case Control Study
C. Cohort Study
D. Randomized Controlled Trial
Information Bias (or Measurement bias or Investigator
bias) – occurs when the outcome assessments are
performed by researchers who are knowledgeable of the
treatment allocation. The investigator may tend to
underestimate or overestimate the effects of the
intervention – resulting in errors in the study results.

Follow-up duration: must be adequate and determined

for all participants. Glaucoma may progress over long
periods, and therefore trials that assess visual field loss
would require a longer follow-up such as 5 years.
20. If a researcher wants to prove that
preservatives contained in eye drops result in
more adverse effects, the appropriate study
design would be a ___________.
A. Case Series
B. Case Control Study
C. Cohort Study
D. Randomized Controlled Trial
 Integrative Studies

Systematic Review:

Meta-Analysis: Statistical combination of

results from several studies
Meta-analysis: Forest Plots

Significant beneficial effect: pooled RR less than 1.0

Significant heterogeneity of studies
 Meta-analysis: Forest Plots

No significant beneficial effect (95% CI crosses 1.0)

 Diagnostic Studies

Cross-sectional design
May have blinding, randomization
Validity Criteria:
Diagnostic uncertainty
Masked interpretation
Reference (Gold) Standard
 Diagnostic Studies

Point Estimates:
Sensitivity
Specificity
Predictive values
Likelihood Ratios
Pre and Post-test Probabilities
Area under the ROC Curve