2. Definition of EBP
Evidence-based practice includes which of the following:
A. Best available evidence
B. Individual clinical experience
C. Patient values
D. All of the above
3. Definition of EBP
Evidence-based practice includes which of the following:
A. Best available evidence
B. Individual clinical experience
C. Patient values
D. All of the above
4. Five steps of EBP
What is the first step in applying EBM concepts to
answer a clinical question?
A. Ask the question
B. Acquire the evidence
C. Appraise the evidence
D. Apply the evidence
C. Assess the whole process
5. Five steps of EBP
What is the first step in applying EBM concepts to
answer a clinical question?
A. Ask the question
B. Acquire the evidence
C. Appraise the evidence
D. Apply the evidence
C. Assess the whole process
The five As
6. Types of clinical studies
What is the best design to study the prevalence of a
disease?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
7. Types of clinical studies
What is the best design to study the prevalence of a
disease?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
8. Types of clinical studies
What is the best design to study the incidence of a
disease?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
9. Types of clinical studies
What is the best design to study the incidence of a
disease?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
10. Types of clinical studies
What is the best design to study the efficacy of an
intervention?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
11. Types of clinical studies
What is the best design to study the efficacy of an
intervention?
A- Cross-sectional study
B- Case-control study
C- Cohort study
D- RCT
12. Types of clinical studies
Which is the best design to study the etiology of
a disease?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. RCT
13. Types of clinical studies
Which is the best design to study the etiology of
a disease?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. RCT
14. Case-control & cohort study etiology of a disease
Cohort study better than case-control study (less bias)
15. Types of clinical studies
Which is the best trial design to study the prognosis of
a disease?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. RCT
16. Types of clinical studies
Which is the best trial design to study the prognosis of
a disease?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. RCT
17. Types of clinical studies
• Case report/case series
• Ecological study
• Cross-sectional study
• Case control study
• Cohort study
• Randomized clinical trial
Primary research
• Systematic review
• Meta-analysis
Secondary research
18. Question type & study design
Study DesignQuestion
Intervention RCT
Incidence & prognosis Cohort study
Prevalence Cross-sectional study
Etiology & risk factors Cohort or case-control
Diagnosis Comparison w gold standard
In each case, SR of all available studies better than individual study
23. The cohort study is the gold-standard of
analytical epidemiology
24. Diagnostic study
Which is the best trial design to study the accuracy of a
diagnostic test?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. Comparison of diagnostic test with gold standard test
25. Diagnostic study
Which is the best trial design to study the accuracy of a
diagnostic test?
A. Cross-sectional study
B. Case-control study
C. Cohort study
D. Comparison of diagnostic test with gold standard test
26. Structure for a study of diagnostic test
Suspected target
condition
Guyatt G et all. Users’ guides to medical literature: manual for EBP.
McGraw-Hill, New York, USA, 2nd edition, 2008.
Accuracy of diagnostic test compared to gold standard
Gold standard test
Positive
Negative
Diagnostic test
Positive
Negative
27. Accuracy of a diagnostic test
• Dichotomous test (only 2 results)
Sensibility (Sn) & Specificity (Sp)
Positive Predictive Value (PPV)
Negative Predictive Value (NPV)
Likelihood Ratios + & – (LR)
Diagnostic Odds Ratio (OR)
• Multilevel test (> 2 results)
Receiver Operating Characteristic (ROC)
Newman TB & Kohn MA. Evidence-based diagnosis.
Cambridge University Press, Cambridge, UK, 1st edition, 2009.
with 95%CI
28. Accuracy of a diagnostic study
A study of diagnostic accuracy of arterial blood gas for diagnosis
of pulmonary embolus (PE) included 212 patients with suspected
PE, 49 of whom were subsequently determined to have PE.
Of the 49 patients with PE, 41 had abnormal alveolar-arterial
oxygen gradient (A-a)DO2.
Of the 163 patients without PE, 118 had abnormal (A-a)DO2.
What is the sensitivity, specificity, PPV, NPV, LR positive and
negative of (A-a) DO2 for the diagnosis of PE?
29. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive
Negative
Column totals
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
30. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive
Negative
Column totals 49
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
31. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41
Negative
Column totals 49
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
32. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41
Negative 8
Column totals 49
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
33. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41
Negative 8
Column totals 49 163
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
34. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118
Negative 8
Column totals 49 163
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
35. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118
Negative 8 45
Column totals 49 163
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
36. Arterial blood gas for diagnosis of pulmonary embolus
Construction of 2 X 2 table
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
49 patients had PE, 41 of them had abnormal (A-a) DO2
163 patients without PE, 118 of them had abnormal (A-a) DO2
37. Arterial blood gas for diagnosis of pulmonary embolus
Sensibility
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
=
Denominator = Column totals
Sensibility
41
49
= 0.84
38. Arterial blood gas for diagnosis of pulmonary embolus
Specificity
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
=
Denominator = Column totals
Specificity
45
163
= 0.28
39. Arterial blood gas for diagnosis of pulmonary embolus
Positive predictive value (PPV)
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
=
Denominator = Row totals
PPV
41
159
= 0.26
40. Arterial blood gas for diagnosis of pulmonary embolus
Negative predictive value (NPV)
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
Denominator = Row totals
NPV
45
53
= = 0.85
41. Arterial blood gas for diagnosis of pulmonary embolus
Prevalence
Gold standard test
CECT Row totals
Disease present Disease absent
Diagnostic test
Arterial blood gas
Positive 41 118 159
Negative 8 45 53
Column totals 49 163 212
Prevalence = 0.23=
49
212
42. LR for a positive test
LR + = 0.84 / (1 – 0.28 ) = 1.17
LR + = Sensitivity / (1 – Specificity)
Corresponds to clinically “ruling in disease”
Probability that the patient has true positive,
rather than false positive test
43. LR for a negative test
LR – = (1 – 0.84 ) / 0.28 = 0.57
LR – = (1 – Sensitivity ) / specificity
Corresponds to clinically “ruling out disease”
Probability that the patient has true negative
rather than false negative test
44. Diagnostic study
The ideal diagnostic test has a:
A. High sensitivity and high specificity
B. High sensitivity and low specificity
C. Low sensitivity and high specificity
D. Low sensitivity and low specificity
45. Diagnostic study
The ideal diagnostic test has a:
A. High sensitivity and high specificity
B. High sensitivity and low specificity
C. Low sensitivity and high specificity
D. Low sensitivity and low specificity
46. ? ?
In a diagnostic test if:
Pretest probability: 30%
Likelihood ratio positive (LR+): 9
What is the post-test probability?
?
?
Computing post-test probability
49. In a diagnostic test if:
Pretest probability: 30%
Likelihood ratio negative (LR– ): 0.1
What is the post-test probability?
Fagan nomogram
Pre-test
probability
Likelihood
ratio
Post-test
probability
Computing post-test probability
51. Randomized controlled trials
A recent RCT found that 29% of diabetics with coronary
heart disease treated with pravastatin suffered a recurrent
coronary event during 5 years of follow-up, while 37% of
the placebo group suffered recurrent coronary events.
• What is the relative risk for recurrent events?
• What is the absolute risk reduction for recurrent events?
• What is the number needed to treat to prevent one
recurrent event?
52. RCT/ Relative risk reduction
A recent RCT found that 29% of diabetics with coronary
heart disease treated with pravastatin suffered a recurrent
coronary event during 5 years of follow-up, while 37% of
the placebo group suffered recurrent coronary events.
Relative risk reduction (RRR)
Risk in treatment group / risk in control group
(0.37 – 0.29) / 0.37 = 22%
53. RCT/ Absolute risk reduction
A recent RCT found that 29% of diabetics with coronary
heart disease treated with pravastatin suffered a recurrent
coronary event during 5 years of follow-up, while 37% of
the placebo group suffered recurrent coronary events.
Absolute risk reduction (ARR)
Risk in control group – Risk in treatment group
37% – 29% = 8%
54. RCT/ Number needed to treat
A recent RCT found that 29% of diabetics with coronary
heart disease treated with pravastatin suffered a recurrent
coronary event during 5 years of follow-up, while 37% of
the placebo group suffered recurrent coronary events.
Number needed to treat (NNT)
1 / Absolute risk reduction
1 / 0.08 = 12.5
55. Systematic review & meta-analysis
All the followings regarding systematic review and
meta-analysis are true EXCEPT:
A. a systematic review may include a meta-analysis
B. a systematic review may not include a meta-analysis
B. a meta-analysis may not include a systematic review
D. a systematic review should always include a meta-analysis
56. Systematic review & meta-analysis
All the followings regarding systematic review and
meta-analysis are true EXCEPT:
A. a systematic review may include a meta-analysis
B. a systematic review may not include a meta-analysis
B. a meta-analysis may not include a systematic review
D. a systematic review should always include a meta-analysis
57. Systematic review & meta-analysis
Systematic reviews
(SR)
Meta-analyses
(MA)
MA may, or may not, include a SR
Egger M et all. Systematic reviews in health care: Meta-analysis in context.
BMJ Publishing Group, London, 2nd edition, 2001.
58. Systematic review & meta-analysis
When data are combined from smaller studies into
a larger sample size, which can then be statistically
evaluated in a more robust fashion than the smaller
samples, the following term is applied:
A. Prospective study
B. Case-control study
C. Cohort study
D. Double-blind clinical trial
E. Meta-analysis
59. Systematic review & meta-analysis
When data are combined from smaller studies into
a larger sample size, which can then be statistically
evaluated in a more robust fashion than the smaller
samples, the following term is applied:
A. Prospective study
B. Case-control study
C. Cohort study
D. Double-blind clinical trial
E. Meta-analysis
60. Forest plot in meta-analysis
The diamond of a meta-analysis of RCTs reveals these results:
Odds ratio: 1.75 – 95% confidence interval: 1.69 , 1.79
• Are the results statistically significant?
A- Yes B- No C- cannot tell
• Are the results precise?
A- Yes B- No C- cannot tell
• Is there a publication bias?
A- Yes B- No C- cannot tell
61. Forest plot in meta-analysis
The diamond of a meta-analysis of RCTs reveals these results:
Odds ratio: 1.75 – 95% confidence interval: 1.69 , 1.79
• Are the results statistically significant?
A- Yes B- No C- cannot tell
• Are the results precise?
A- Yes B- No C- cannot tell
• Is there a publication bias?
A- Yes B- No C- cannot tell
62. Clinical scenario -1
• You have a 60-year-old patient with acute biliary
pancreatitis and non-infected pancreatic necrosis
on CECT. You wonder if prophylactic antibiotics
prevents infection of non-infected pancreatic necrosis
& decreases mortality.
• You identify a meta-analysis of RCTs evaluating the
effect of prophylactic antibiotics to prevent infected
necrosis and decrease mortality.
Bai Yu & al. Am J Gastroenterol 2008 ; 103 : 104 – 110.
63. Antibiotic prophylaxis & pancreatic necrosis
Bai Y et al. Am J Gastroenterol 2008 ; 103 : 104 – 110.
Forest plot
Is the result statistically significant?
Is the result precise?
Is there heterogeneity?
64. • Diamond cross horizontal line Not statistically significant
• Wide confidence interval Low precision
• I2 = 23.2% Low level of heterogeneity
Antibiotic prophylaxis & pancreatic necrosis
OR: 0.81 – 95% CI: 0.54-1.22 – I2: 23.2%
Bai Y et al. Am J Gastroenterol 2008 ; 103 : 104 – 110.
65. Bai Y et al. Am J Gastroenterol 2008 ; 103 : 104 – 110.
Antibiotic prophylaxis & pancreatic necrosis
Funnel plot/Publication bias
Is there a publication bias?
66. Is there a publication bias?
Funnel plot did not show significant asymmetry,
but the number of included studies is low (7 studies)
67. Clinical scenario -2
• You are consulted regarding the peri-operative management
of a 66- year-old man undergoing hip replacement. He is a
smoker and has a history of type 2 diabetes & hypertension.
Because he has multiple cardiovascular risk factors, you
consider using perioperative β-blockers to reduce the risk of
postoperative cardiovascular complications and death.
• You identify a recently published systematic review and meta-
analysis evaluating the effect of perioperative β-blockers on
nonfatal myocardial infarction, stroke and death.
Bouri S et al. Heart. 2014;100(6):456-464
68. Death in patients receiving perioperative β-blockers
Bouri S et al. Heart. 2014;100(6):456-464
69. Is the result statistically significant?
Is the result precise?
Is there heterogeneity?
70. Is the result statistically significant?
Diamond doesn’t cross line of no effect: statistically significant
Diamond cross line of no effect: not statistically significant
Look at the diamond with the point estimate & 95% CI
71. The diamond
Perera R, Heneghan C, Badenoch D. Statistics Toolkit.
Blackwell Publishing Ltd, Oxford, 1st edition, 2008.
Shows combined point estimate (RR or OR)
with 95% CI
72. Death in patients receiving perioperative β-blockers
Bouri S et al. Heart. 2014;100(6):456-464
73. Is the result precise?
Look at the diamond with the point estimate & 95% CI
Narrow confidence interval: High precision
Wide confidence interval: Low precision
74. Death in patients receiving perioperative β-blockers
Bouri S et al. Heart. 2014;100(6):456-464
75. Statistical significance & CI
(a) Statistically significant, low precision
(b) Statistically significant, high precision
(c) Not statistically significant, low precision
(d) Not statistically significant, high precision
Glasziou P et al. Evidence based practice workbook. Blackwell, 2nd edition, 2007.
76. Influence of sample size on CI precision
Width of CI (precision of the estimate)
decreases with increasing sample size
Peat JK, et al. Health science research. Allen & Unwin, Australia, 1st ed, 2001.
77. Is there heterogeneity?
Visual evidence of heterogeneity in the forest plot
Simon SD. Statistical evidence in medical trials: What do the data really tell us?
Oxford University Press, Oxford, 1st edition, 2006
Qualitative assessment
Quantitative assessment
I-squared ˂ 25% Low heterogeneity
I-squared 25 – 50% Moderate level of heterogeneity
I-squared ˃ 50% High level of heterogeneity
78. Heterogeneity
Do the pieces fit together?
Simon SD. Statistical evidence in medical trials: What do the data really tell us?
Oxford University Press, Oxford, 1st edition, 2006
80. Death in patients receiving perioperative β-blockers
Bouri S et al. Heart. 2014;100(6):456-464
81. • Not statistically significant
• Low precision
• Moderate level of heterogeneity
Death in patients receiving perioperative β-blockers
Bouri S et al. Heart. 2014;100(6):456-464
OR: 0.94 – 95% CI: 0.63-1.40 – I2: 30%
82. Death in patients receiving perioperative β-blockers
Funnel plot/Publication bias
Bouri S et al. Heart. 2014;100(6):456-464
Is there a publication bias?
83. Is there a publication bias?
“Funnel plot did not show significant asymmetry,
but this cannot definitively exclude publication bias”
Number of included studies is low (˂ 30 studies)
86. Clinical scenario -3
• HEV seroprevalence (anti-HEV IgG) in patients on
maintenance hemodialysis (HD) ranges from 0% to 44%.
Chronic hepatits E is encountered in immunocompromized
patients. You wonder if it is worthy to test your HD patients
for HEV markers especially for those who are planned to
receive renal transplantation.
• You identify a recently published systematic review and
meta-analysis evaluating the seroprevalence of HEV in
HD patients.
Haffar et al. Aliment Pharmacol Ther 2017;46:790-799.
87. Haffar et al. Aliment Pharmacol Ther 2017;46:790-799.
88. Is the result statistically significant?
Is the result precise?
Is there heterogeneity?
89. • Statistically significant
• Low precision
• High level of heterogeneity
HEV seroprevalence in hemodialysis patients
OR: 2.47 – 95% CI: 1.79-3.40 – I2: 75.2%
Haffar et al. Aliment Pharmacol Ther 2017;46:790-799.
98. HEV seroprevalence in hemodialysis patients
Funnel plot/Publication bias
Haffar et al. Aliment Pharmacol Ther 2017;46:790-799.
Egger test: p = 0.83
Is there a publication bias?
99. Is there a publication bias?
Funnel plot: no publication bias
Egger test (p = 0.83): no publication bias
100. Publication bias (Funnel plot)
Funnel plot is used to detect publication bias in trials
included in a meta-analysis.
A. True B. False
Publication bias will result in asymmetry of funnel plot.
A. True B. False
The p value of Egger test is 0.932 which indicates that
publication bias exists in trials included in the meta-analysis.
A. True B. False
Sedgwick P et al. How to read a funnel plot. BMJ 2013;346:f1342.
101. Publication bias (Funnel plot)
Funnel plot is used to detect publication bias in trials
included in a meta-analysis.
A. True B. False
Publication bias will result in asymmetry of funnel plot.
A. True B. False
The p value of Egger test is 0.932 which indicates that
publication bias exists in trials included in the meta-analysis.
A. True B. False
Sedgwick P et al. How to read a funnel plot. BMJ 2013;346:f1342.
102. • We have non-invasive means of early detection
of HEV infection with sensitive and specific assays
• Benefits of such testing clearly outweigh its risks
• Cost-effectiveness of such approach remains to be elucidated
HEV seroprevalence in hemodialysis patients
Conclusion
Haffar et al. Aliment Pharmacol Ther 2017;46:790-799.
104. Critical appraisal of a systematic review -1
Murad MH et al. JAMA. 2014;312(2):171-179
Address sensible clinical question Focused question
Search for relevant studies exhaustive PubMed, Embase,…
Selection/assessment of studies reproducible κ agreement
Results ready for clinical application yes
Address confidence in estimates of effect Assess risk of bias
Address heterogeneity
Evaluate credibility of the methods of SR
If no, the second judgment will not be possible
Questions Answers
105. Medline
Embase
Cochrane Trials Registry
Comparing PubMed, Embase & Cochrane
Overlap of 2 databases: 34%
PubMed: better coverage of US journals
EMBASE: better coverage of European journals
Smith BJ et al. Med J Aust 1992 ; 157 : 603 – 11.
106. Interpretation of different values of kappa
Kappa from Greek letter κ
Value of kappa Strength of agreement
0 – 0.20 Poor
0.21– 0.40 Fair
0.41– 0.60 Moderate
0.61– 0.80 Good
0.81–1.00 Very good
Perera R, Heneghan C & Badenoch D. Statistics toolkit.
Blackwell Publishing & BMJ Books, Oxford, 1st edition, 2008.
kappa score of 0.6 indicates good agreement
107. Critical appraisal of systematic review -2
Murad MH et al. JAMA. 2014;312(2):171-179
Risk of bias across studies Cochrane tool of bias
Results consistent in studies (heterogeneity) I-squared & p value
Precision of the results Wideness of 95% CI
Results apply to my patient (indirectness) Age, sicker,….
Concern about reporting bias Funnel plot, Egger test
Rate the confidence in the effect estimates
Questions Answers
Reasons to increase the confidence rating
“Large treatment effect over a short time”
112. Directness of evidence
Four types Examples
Populations Older, sicker or more co-morbidity
Oseltamivir for prophylaxis of avian flu
Interventions Surgery undertaken by subspecialists in referral
centers vs general surgeons in the community
Outcomes Surrogate outcome
Digestive side effects of NSAIDs
Indirect comparison No head-to-head comparisons between several
drugs of the same class (eg: biphosphonates)
Guyat GH et al. Journal of Clinical Epidemiology 2011;64:1303-1310.
113. Serious GI events:
perforation -bleeding
Clinical ulcers
Endoscopic ulcers
Relative severity
GI Symptoms
Relative frequency
NSAID-related GI side effects