COMMUNITY MEDICINE 5.

01
CRITICAL APPRAISAL
Dean M. Sosa & Assoc. | January 18, 2019
Transcriber group: 16B
INTRODUCTION WHAT IS EVIDENCE-BASED MEDICINE?
Alright fam! So this trans is a combination of the lecture Dean
Sosa gave plus a transcript of the two videos we were Evidence-based medicine is the integration of best evidence
supposed to watch. It’ll be divided into those three main parts. with clinical expertise and patient values. - Dave Sackett 2001
Good luck studying!

• PAGE 1 – DEAN SOSA LECTURE

• PAGE 2 - CRITICAL APPRAISAL OF THERAPY (VIDEO)
• PAGE 5 - CRITICAL APPRAISAL OF DIAGNOSIS (VIDEO)

PRACTICE OF MEDICINE BEFORE EBM

• Before Evidence-Based Medicine (EBM) thinking began to

impact on its structure, clinical practice relied on expert
advice — often driven by physiological reasoning and
individual clinicians’ experience.
• This emphasis resulted in significant gap between the
available evidence and the clinical practice.

HOW IMPORTANT IS EVIDENCE BASED MEDICINE PRACTICE?

• Applying the knowledge gained from large clinical trials to

patient care promotes consistency of treatment, and
optimal outcomes helps establish national standards of
patient care, and sets criteria to measure and reward Clinical Expertise > Research Evidence
performance-based medical practice.
• Implementation of EBM in the managed care setting
provides standards that have the potential to provide the
best medical care at the lowest cost.

AN EBM APPROACH TO EDUCATION

• 1990: Sackett’s “Just in Time” learning

• Evidence cart on ward rounds – 1995
• Looked up 2-3 questions per patient
• Took 15-90 seconds to find
• Change about 1/3 decisions
• Rounds took longer!
• *David Lawrence Sackett was an American-Canadian
medical doctor and a pioneer in evidence-based
medicine. He is known as one of the fathers of Evidence-
Based Medicine.
CRITICAL APPRAISAL

• Claims of effectiveness
• Claims of accuracy
• Claims on causation
• Claims on prognosis

1 of 10
 FOUR STEPS IN EBM o PICO
o P: population
1. Formulate answerable question o I: intervention
2. Track down the best evidence o C: comparison/comparator
3. Critically appraise the evidence o O: outcome
4. Individualise, based clinical expertise and patient • Research
concerns • Retrieving medical literature

FORMULATION OF DIAGNOSIS QUESTION

Is Positron Emission Tomography (Test/intervention) good test

for coronary disease (outcome)?

Is PET (test/intervention) a more sensitive and specific test

(outcome) in diagnosing coronary artery disease (patient) as
compared to coronary angiography (comparison)?
Its practice requires:
• Asking
• Acquiring
• Appraising
• Applying
• Assessing

Up next is the “Critical Appraisal of RCT for students —
therapeutic” video transcript. Feel free to watch the video
while reading this trans. It’s only 28 min and 54 seconds long.
Kaya mo yan!
EVIDENCE-BASED MEDICINE
• Defined as a systematic approach to the acquisition,
appraisal, and application of research evidence to the
decisions in health care
• Involves three skills such that evidence should be:
o acquired
o appraised
o applied
STEPS IN ACQUIRING EVIDENCE
Asking a focused clinical question
• Most important because if you ask the wrong question,
you will never find the right answer
• Perhaps the hardest skill to learn
• Should be well-stated in terms of the following variables:
o PEO
o P: patient population with a certain disease or
condition
o E: exposure (or treatments) to be administered
to these patients
▪ E1 or E – the experimental treatment
being evaluated (usually a new one)
▪ E2 or C – the control treatment to which
E1 is being compared to
o O: outcome (or conditions) that the treatment is
intended to prevent
• The highest level of evidence for testing effectiveness of a
treatment or a therapy is found in randomized controlled
double-blind studies and systematic reviews and meta-
analyses
• Randomized Controlled Trial
o A study in which participants are randomly
allocated to an experimental or comparison
group
o The experimental group gets an intervention
o The comparison group gets something different
(either no intervention, a placebo, or a different
intervention from that received by the
experimental group)
o Outcomes in each group are compared to
determine the effect of the intervention
o The purpose of randomization is to make the two
groups equal in which the only difference is the
intervention
o Results of RCT’s can infer causality which can
attribute differences in outcomes to the
differences in the treatment
STEPS IN APPRAISING ARTICLES
APPRAISING DIRECTNESS
• How well does the PEO or PICO in the study (or research
question) correspond with your own PEO or PICO (clinical
question)
• Example:
o Among children 1-5 years old age (P), how
effective is zinc supplementation (E1 or E)
compared to placebo (E2 or C) in preventing
acute diarrhea (O)?

2 of 10
 Research Question(s) Clinical Question(s)
P o Compare baseline characteristics between
E o Randomization is key
O o Helps to avoid performance bias
APPRAISING VALIDITY (INTERNAL VALIDITY)
• Done only after making sure that the clinical question
matches the research question
• Assess the risk of bias and decide if the results are
trustworthy
o Selection bias — systematic differences between
baseline characteristics of the groups that are
compared
o Performance bias — systematic differences
between groups in the care that is provided, or in
exposure to factors other than the interventions
of interest
o Detection bias — systematic differences
between groups in how outcomes are
determined
o Attrition bias — systematic differences between
groups in withdrawals from a study
QUESTIONS ANSWERED BY APPRAISING VALIDITY
• Q1: Were patients randomly assigned to treatment
groups?
o Randomization ensures that both groups are
comparable
o To check for randomization, look for the words
randomization, randomized, or randomly
assigned in the title, abstract or methodology
• Q2: Was allocation concealed?
o Allocation sequence — random sequence by
which patients are assigned to treatment groups
o Allocation concealment — measure taken to
ensure that the sequence is not altered
knowingly or unknowingly
▪ Placement of treatment assignment in
concealed envelopes
▪ Delegation of a third party to assign
treatment (for example, a pharmacist)
▪ Use of objective computers
o Used to avoid selection bias
▪ Groups should have similar baseline
characteristics
▪ Achieved through random sequence
generation and concealed allocation
▪ Bias is committed if a participant’s
decision to provide consent or a
recruiter’s decision to enroll a
participant is influenced by knowledge
of which group a patient would be in if
they participated (selection bias)
• Q3: Were baseline characteristics similar at the start of
the trial?
groups in terms of prognostic factors
• Q4: Were patients blinded to treatment assignments?
o Blinding helps in the objective assessment of the
treatment response
o Patients may have the tendency to report
symptoms more if they know they are receiving
placebo, which will make the active drug seem
better
o Tip is to look for identical preparations
• Q5: Were caregivers blinded to treatment assignments?
o Helps to avoid performance bias
o The goal is to provide equal treatments and
behaviour to the experimental and control
groups other than the intervention itself
• Q6: Were outcome assessors blinded to treatment
assignment?
o Helps to avoid detection bias
o The goal is to assess outcomes the same way for
both experimental and control groups
o Objective judgment of outcomes is necessary
• Q7: Were all patients analysed in the groups to which
they were originally randomized?
o Helps to avoid attrition bias
o Possible reasons why patients may not adhere to
treatment groups
▪ Forgetting how to take drugs properly
▪ Suffering from side effects of the drug
▪ Refusing to go on for no apparent
reason
o Intention To Treat (ITT)
▪ Once randomized, participants are
analysed in their assigned group
regardless of:
• Their status
• Whether they were lost to
follow-up
• Even if they never received
treatment, or
• Whether they crossed over to
the other treatment group
▪ Benefits:
• Groups stay equal
• The power of the study is
maintained
• An estimated of “real world”
effectiveness
• Q8: Was follow-up rate adequate?
o Adequacy of follow-up refers to minimization of
the number of patients who drop out of the
study, usually due to adverse events or
dissatisfaction
o Helps to avoid attrition bias
3 of 10
 o Study validity is threatened if the number of
participants lost to follow-up is high
o Performing sensitivity analyses would determine
significant dropout rate
o “5-and-20 rule of thumb”
▪ <5% loss to follow-up — little bias
▪ 5-20% loss — small bias
▪ >20% — serious threat to validity
CLASSIFICATION SCHEME FOR BIAS
- See table 1
IF NO SIGNIFICANT BIAS HAS BEEN COMMITTED
• Move on to the following steps:
o Appraising the results
o Assessing applicability
o Individualizing the results
APPRAISING RESULTS
There are two questions that need to be considered:
1. How large was the treatment effect?
2. How precise was the estimate of the treatment
effect?
Q1. How large was the treatment effect?
• There are two ways of expressing the results of article on
effectiveness depending upon the outcome of variable —
dichotomous or continuous variable
Ways of expressing effectiveness (see table 2)
• For dichotomous variable, it can be expressed as:
o Percentage/proportion (e.g. deaths per 100
patients)
o Rate (e.g. deaths per 100 patients per year)
• Comparing results between two groups with a
dichotomous variable can be expressed as:
o Proportion
▪ Relative Risk Reduction (RRR)
▪ Absolute Risk Reduction (ARR)
▪ Relative Risk (RR)
o Rate
▪ Hazard Ratio
• Continuous variables can be expressed as:
o Mean
▪ Mean difference can be used to
compare results between two groups
Measure of effectiveness for dichotomous outcome (see
table 3)
Q2. How precise was the estimate of the treatment effect?
• RRR, ARR and RR are examples of point estimates
• Confidence interval (CI) provides range of possible values
of the treatment effect.
o Usually reported at 95% level of confidence
▪ 95% CI means “we are 95% sure that
the true effect of the treatment lies
within this range”
• In a study on warfarin treatment, the reported RRR is
79% with a 95% confidence interval of 52% and 90%
(RRR: 79% (95% CI: 52%, 90%). What does that mean?
o That means that warfarin reduces risk of stroke
in patients with atrial fibrillation by 79% but the
actual RRR could range from 52-90%.
Interpreting Confidence Intervals (CI)
• When both ends of the CI are on the side of benefit, the
treatment is definitely beneficial or superior
o Ex. RR of 0.63 (95% CI: 0.53,0.73)
o It means that the confidence interval lies on the
side of benefit because it is less than 1, thus, the
treatment is beneficial or superior
• When both ends of the CI are on the side of harm, the
treatment is definitely harmful or inferior
o Ex. RR of 2.3 (95% CI: 1.5,3.1)
o The 95% confidence interval is between 1.5-3.1
then treatment is definitely harmful.
• When one end reflects important benefit and the other
end reflect important harm, the study is inconclusive.
o Ex. RR of 0.98 (95% CI: 0.3,2.5)
o For example, we now present you the results of a
relative risk, remember that you interpret a
relative risk of <1 as beneficial treatment and >1
as harmful. Now for the relative risk of 0.98, it is
seemingly beneficial. However, when you take a
look at the 95% confidence interval, the lower end
of 0.3 is beneficial while the higher end of 2.5 is
harm, that means that the study is inconclusive
• When one end reflects unimportant benefit and the
other end reflects unimportant harm, then for all intents
and purposes the two treatments being compared are
the same or equivalent.
o Ex. RR of 0.98 (95% CI: 0.95,1.2)
o Again, our example is the relative risk, 0.98 < 1
that means treatment is beneficial but when you
take a look at the 95% confidence interval, the
lower end is 0.95 < 1 while the higher end is 1.02
which is a little bit more than 1. This reflects
unimportant benefit for the lower confidence
interval and unimportant harm for the higher
confidence interval which means that the two
treatments being compared are the same or
equivalent.
4 of 10
 ASSESSING APPLICABILITY
• After evaluating validity and analyzing the results of the
trial, the next step is to decide if the results can be applied
to our own patients.
• In making decisions in applicability is best to use a
measure of effectiveness that is:
o Stable
o Constant across different population.
• Now we can be guided using SCRAPS
o S – Sex: consider physiologic, hormonal or
biochemical differences between sex that may
affect results of interventions
o C – Comorbidities: consider coexistent
conditions
▪ Are there any coexistent conditions in
the study that are not present in your
patient that can affect the results of the
intervention?
o R – Race: consider racial differences
o A – Age: consider age differences
o P – Pathology: consider actual disease under
study
o S – Socioeconomic factor: usually studies are
conducted in an ideal condition which is difficult
to apply in everyday life
INDIVIDUALIZING RESULTS
• It is done when biologic and socioeconomic factors will
not compromise effectiveness
• Studies report the effectiveness of a treatment in a
population as a whole, the benefits, cost and risks varied
greatly from patient to patient
• Variability affects the ARR to a greater extent than RRR
due to patient’s baseline risk
• Number needed to TREAT to avoid ONE unfavourable
outcome
• 1/risk difference (ARR) = NNT (better) description for
clinical significance)
• Example: ARR is 10%
• NNT = 1/0.10 = 10 (treat 10 patients to avoid ONE
unfavourable outcome
Whew! That’s the end of that! Up next is the “Critical Appraisal
of Diagnosis” video transcript. Feel free to watch the video and
read along with the trans. It’s only 49 min and 55 seconds.
Note that there is some overlap with the other video and this
video. Anyway, enjoy!
INTRODUCTION
• Reasons to run diagnostic tests
o Diagnose presence or absence of condition
o Plan an intervention (surgery)
o Monitor response to therapy
o Estimate risk of future evens
o Determine prognosis with unknown disease
• General objective: to apply the various rules of evidence
in deciding if the results of an article on a diagnostic test
are valid
• Specific objectives: to discuss the rational for each
appraisal step pertaining to the validity of claims on
accuracy
• To learn the definitions of, calculations for and
interpretation of sensitivity, specificity, positive predictive
value, negative predictive value, likelihood ratio, pre-test
probability, post-test probability, pre-test odds, post-test
odds, diagnostic and therapeutic threshold
STEPS IN APPRAISING ARTICLES ON DIAGNOSIS
• Appraising directness
• Appraising validity
• Appraising the results
• Assessing applicability
• Individualizing the results
ASKING A FOCUSED CLINICAL QUESTION ON THERAPY (PEO)
• P = the patient population on whom the test might be
done
• E = the exposure (in this case, the test to be performed,
referred to as the index test)
• O = the outcome (the condition that the test is supposed
to diagnose
ASKING A FOCUSED CLINICAL QUESTION ON DIAGNOSIS
Example: Among post-menopausal women presenting with a
breast mass (P), how accurate is fine needle aspiration biopsy
(E) in establishing the presence or absence of breast cancer
(O)?
APPRAISING DIRECTNESS
• Evaluate how well the PEO in the study (research
question) corresponds with your own PEO (clinical
question)
Research Clinical
Questions Questions
P
E
O
• Main question: Does the study provide a direct enough
answer in terms of patients, examination or test used, and
disease/outcome diagnosed?
• Evaluating directness: Does the study provide a direct
enough answer to your clinical question in terms of
patients (P), examination or test (E) used and disease or
outcome (O) being diagnosed?
5 of 10
• Why is it important? - Many times, the patient you are
managing, the test you are about to use, or the outcome
you are trying to prevent is not exactly the same as those
studied by the authors of a paper. In this situation, you
need to decide if you can use the study results at all. The
decision requires some expertise on the disease under
question.
• What to look for: Seek the opinion of an expert (this might
be you), or your colleagues.
APPRAISING VALIDITY
• To make sure the odds are not stacked for or against a
particular diagnostic test (bias), there are four criteria to
assess the validity of studies on diagnostic accuracy.
• They are phrased as yes or no questions. (the more yes,
the better the study)
• QUESTION 1: WAS THE REFERENCE STANDARD AN
ACCEPTABLE ONE?
o Appraising validity: Was the reference standard
an acceptable one?
o Why is it important: A reference standard that is
accepted in the scientific community as the
definition of disease is the yardstick with which
the performance of the test is measured
o What to look for: Look for standard diagnostic
test(s) used to ascertain presence of disease in
the study population.
• QUESTION 2: WAS “DEFINITION” OF THE INDEX TEST
AND THE REFERENCE STANDARD INDEPENDENT?
o Appraising validity: Was “definition” of the index
test and the reference standard independent?
o Why is it important: If any of the criteria for the
index test are part of the criteria for the
reference standard, we tend to observe a falsely
high level of agreement between the two. This
will result in an overestimate of the accuracy of
the index test.
o What to look for: Look for the criteria used to
interpret the index test and the reference
standard and make sure they do not overlap
• QUESTION 3: WAS “PERFORMANCE” OF THE INDEX TEST
AND THE REFERENCE STANDARD INDEPENDENT?
o Appraising validity: Was “performance” of the
index test and the reference standard
independent?
o Why is it important: If in a study setting the result
of the index test becomes the indication for the
reference standard, the agreement between the
index test and the reference standard will
increase making the index test look better than it
actually is.
o What to look for: Check whether measures were
taken to ensure that the result of the index test
was not an indication for doing the reference
standard.
• QUESTION 4: WAS “INTERPRETATION” OF THE INDEX
TEST AND THE REFERENCE STANDARD INDEPENDENT?
o Appraising validity: Was “interpretation” of the
index test and the reference standard
independent?
o Why is it important: If in a study setting,
knowledge of the results of an earlier index test
may influence interpretation of the reference
standard, making the index test seem better than
it really is.
o What to look for: Look for attempts to make sure
interpretation of one test did not affect
interpretation of the other.
APPRAISING THE RESULTS
• QUESTION 1: WHAT WERE THE LIKELIHOOD RATIOS OF
THE VARIOUS TEST RESULTS?
o Appraising results: What were the likelihood
ratios of the various test results?
o Why is it important: Likelihood ratios allow us to
convert pre-test to post-test probability, even if
tests have more than just 2 results (positive or
negative).
o What to look for: If LR’s [likelihood ratios] are not
reported, 2 x n [two by n] tables need to be
reconstructed, and calculations need to be made.
WHAT ARE THE RESULTS?
• Are the likelihood ratios for the test results presented or
data necessary for their calculation included?
• Validity of a Clinical Test: Sensitivity (Sn) and Specificity
(Sp)
• Sn: proportion of people with disease who have a positive
test; positivity in disease
• Sp: proportion of people free of disease who have a
negative test; negativity in health
SENSITIVITY (Sn) AND SPECIFICITY (Sp)
Reference Standard Total
Result
Test Result Present Absent
A B
Positive A+B
Negative C D C+D
Total A+C B+D A+B+C+D
Sn = A/A+C
Sp = D/B+D
Accuracy = A+D/A+B+C+D
6 of 10
SENSITIVITY (Sn) AND SPECIFICITY(Sp) (+) LR = (A/A+C) / (B/B+D)
(-) LR = (C/A+C) / (D/B+D)
Reference Standard Total *It’s not important to remember the formulas for calculating
Result likelihood ratio so long as it is understood that likelihood ratio
is a measure of how much the likelihood of a disease changes
Test Result Present Absent given a test result
a=TP b=FP
• LRs indicate by how much a given diagnostic test result will
Positive a+b
raise or lower the pretest probability of the target
Negative c=FN d=TN c+d disorder
Total a+c b+d a+b+c+d • LR = 1: post-test probability is exactly the same as the
pretest probability
• LR > 1: increase the probability that the disease is present
Sn = TP/TP + FN (better at ruling IN disease — confirmatory tests)
Sp = TN/TN + FP • LR < 1: decrease the probability of the disease (better at
TP – True Positive, FP – False Positive
ruling OUT disease — screening tests)
TN – True Negative, FN – False Negative
• *there are rarely tests with BOTH strongly positive and
Utility of a Clinical Test: Predictive Values (PV)
strongly negative likelihood ratios
(+) PV: Proportion of those with a positive test who actually ASSESSING APPLICABILITY
have the disease as measured by the gold standard
• QUESTION 1: ARE THERE BIOLOGIC ISSUES THAT MAY
(-) PV: Proportion of those with a negative test who actually AFFECT ACCURACY OF THE TEST?
are free of the disease as measured by the gold standard o Assessing applicability: Are there biologic issues
PREDICTIVE VALUES that may affect accuracy of the test? (Consider
the influence of sex, comorbidity, race, age, and
Reference Standard Total pathology)
Result o Why is it important: There should be adequate
consideration of variables that may affect test
Test Result Present Absent
performance and utilization in our population of
A B
interest
Positive A+B
o What to look for: Look for characteristics of the
Negative C D C+D study population and assess if the authors
Total A+C B+D A+B+C+D attempted to isolate variables or made statistical
adjustments
(+) PV = A/A+B = Post-test Probability of disease • QUESTION 2: ARE TEHRE SOCIO-ECONOMIC ISSUES THAT
(-) PV = D/C+D = Post-test Probability of no disease MAY AFFECT ACCURACY OF THE TEST?
Prevalence = A+C/A+B+C+D = Pretest Probability of disease o Assessing applicability: Are there socioeconomic
Likelihood Ratio issues that may affect accuracy of the test?
• Expresses the odds that a given level of a diagnostic test o Why is it important: Social, cultural, and
result would be expected in a patient with the target
economic contexts that may potentially affect
disorder or disease
diagnostic test performance have to be taken
• The probability of a test result among patients with the
disease, divided by the probability of that test result into account particularly for interventions that
among patients without the disease. require socio0cultural sensitivity (e.g.
questionnaires and/or provider compliance)
o What to look for: Check if the tests done were
Reference Standard Total
Result locally validated, you may need to do validation
studies in your locale. Look for attempts to
Test Result Present Absent compute costs.
A B
Positive A+B • SCRAPS
Negative C D C+D o S-Sex – consider physiologic, hormonal, or
Total A+C B+D A+B+C+D biochemical differences between sex that may
affect results of the test
o C-comorbidities – consider co-existent
conditions

7 of 10
 o R-race – consider racial differences • Therapeutic threshold: probability of disease above which
o A-age – consider age difference we are reasonably sure that we should stop testing and
o P-pathology: consider actual disease under study just get on with treatment
o S-socioeconomic factor — ideal condition • Testing range: between the two thresholds, the clinician
difficult to apply in everyday life is uncertain, so further tests are required
INDIVIDUALIZING THE RESULTS
SUMMARY
• Determine the impact that the test might have on your
patient’s probability of having the disease Articles on diagnostic tests are sometimes considered
• Convert the probability of the disease before the test (pre- difficult reading. By dissecting them one step at a time, reading
test probability) to the probability of the disease after the them becomes simple and straightforward.
test (post-test probability) First, appraise directness. Is the question asked
• Set the Diagnostic and Therapeutic Threshold important to you? Is the test something you are considering?
Is it available? Is it a feasible option? If not, maybe you should
PROBABILITY OF DISEASE GIVEN A TEST RESULT be looking for another article. If it does address an important
question for you, the next step is to appraise validity. Is the
Step How to do it reference standard acceptable? Were the two tests
independent in terms of definition, performance, and
1.Estimate the pre- Based on your history and PE, your interpretation? The answers to these questions will tell you
test probability as a clinical experience will give you a good whether the results are credible. If they’re not, why even
percentage estimate of the probability of disease. bother appraising the article? These two steps can actually
If the case is difficult, ask an expert save you from a lot of futile reading. If you decide to proceed
and interpret the results, look for likelihood ratios, or
2.Convert pre-test Probability information/data that might lead you to compute them
probability to odds Odds = -------------------- yourself. Before actually using these numbers yourself,
100-probability however, decide whether the information is applicable to your
setting: biologic, and socioeconomic factors should be
(Probability expressed as %) considered.
3.Multiply pre-test If everything is in order, directness, validity, results,
odds by the LR of the Post-test odds = pre-test odds x LR and applicability, you now have the information necessary to
test result to get the decide whether the test is something you’d like to request or
post-test odds (Probability expressed as %) if you are using them, interpret the results when you arrive.
Don’t become too engrossed with numbers. Sometimes it’s
4.Convert post-test Odds enough just to understand that likelihood ratios are a measure
odds back to post- Probability = ------------ x 100 of changes in disease probability. Test results with likelihood
test probability in per 1 + Odds ratios greater than 1 increase the disease probability. The
cent higher the likelihood ratio, the closer it can bring you into
(Probability expressed as %) confirming the disease. Test results with likelihood ratio less
than one decrease the disease probability. The lower the
likelihood ratio, the closer it can bring you to ruling out the
disease. Test results with likelihood ratio very close to one
have very little impact on your clinical or pretest estimates of
SETTING THE DIAGNOSTIC AND THERAPEUTIC THRESHOLD
disease probability.
When you have arrived at your post-test probability of
disease, you can make a clinical decision based on your
threshold of management and your patient’s preferences.
These are usually three choices: 1) stop testing and get on with
the treatment of the probable disease 2) stop testing and
reassure the patient the disease probability is low 3) do more
tests before you decide
*sorry bout the speaker, that’s there the entire video lol*
• Diagnostic threshold: the probability of disease below
which we are reasonably sure that we should stop testing
and just reassure the patient

8 of 10
 COMMUNITY MEDICINE 5.01
CRITICAL APPRAISAL
Dean M. Sosa & Assoc. | January 18, 2019
Transcriber group: 16B

TABLE 1: Classification Scheme for Bias

Type of bias Description Relevant domain in the “risk of

bias” tool
Selection Systematic differences between baseline Sequence generation
characteristics of the groups that are compared
Allocation concealment
Performance Systematic differences between groups in the Blinding of participants and
care that is provided, or in exposure to factors personnel
other than the interventions of interest
Addressing other potential
threats to validity
Detection Systematic differences between groups in how Blinding of outcome assessors
outcomes are determined
Addressing other potential
threats to validity
Attrition Systematic difference between groups in Incomplete outcome data
withdrawals from a study

TABLE 2: Ways of Expressing Effectiveness

Summary of results Comparison of results between two

Outcome
within each group groups
Dichotomous (e.g. lived or Proportions (e.g. deaths Relative risk reduction, absolute risk
died, blood pressure per 100 patients) reduction, relative risk
controlled or not)
Rate (e.g. deaths per 100 Hazard ratio = rate in treatment/rate in
patients per year control
Continuous (e.g. blood Mean (e.g. mean blood Mean difference = mean in control —
pressure in mmHg, quality pressure) mean in treatment
of life in a scale of 1-10)

1 of 10
 TABLE 3: Measure of Effectiveness for Dichotomous Outcome

Weight analogy Measure of effectiveness Interpretation

I lost 20 kg Weight difference I lost 2% of my risk Risk difference ARR (%)
(absolute weight = Wc – Wt Absolute Risk = Rc – Rt
>0%: Tx beneficial
reduction) Reduction (ARR) 0%: no effect of Tx
<0%: Tx harmful

I lost 25% of my Weight difference I lost 25% of my risk Risk difference / RRR (%)
weight (relative Original = (Wc-Wt) Relative Risk Reduction Original Risk
>0%: Tx beneficial
risk reduction) / Wc (RRR) 0%: no effect of Tx
(Rc – Rt) / Rc <0%: Tx harmful

I now weigh 75% New weight / My risk is now 0.75 of New risk / RR (decimal)
of what I weigh original weight what it was Original risk
<1: Tx beneficial
(relative risk) Relative Risk (RR) 1: no effect of Tx
Rt / Rc >1: Tx harmful

*Xc = Control; Xt = Treatment

Example
An RCT showed death rates of 4% in the treatment group and 5% in the placebo group. What would be
the RRR, ARR and RR?

RRR = (Rc – Rt) / Rc

(5 – 4) / 5 = 0.2
2. x 100% = 20% (answers should be in percentage/rate for RRR and ARR)

ARR = Rc – Rt
5% - 4% = 1%

RR = Rt / Rc
4 / 5 = 0.8

Interpretation
RRR: 20% > 0% =>Treatment is beneficial (it means that there is a 20% reduction of the risk of death in the
treatment group)
ARR: 1% > 0% =>Treatment is beneficial
RR: 0.8 < 1 => Treatment is beneficial

--END—
“Love is like an hourglass, with the heart filling up as the brain empties. – Jules Renard, French author”

Happy February fam! #ONE2021

REFERENCES:
• Recordings
• PPT photos
• Lecture notes
• Video lectures

2 of 10