1

 Critically Appraising
the Evidence:
 Statistics for Therapy

2

 Clinical Statistics Calculator (Excel)
 Statistics for:
 Therapy
 Control Event Rate (CER) & Experimental Event Rate (EER)
 Number Needed to Treat (NNT)
 Absolute Risk Reduction (ARR)
 Relative Risk (RR)
 Odds
 Odds Ratio (OR)
 Practice Exercises

3

 If available, find the best evidence in secondary sources where analysis
has already occurred.
 If not preassessed, use critical appraisal worksheets to help you
through the process.

4

 Understanding the Limitations of the Author’s Analyses and
Interpretations of the Data
 Assessing Internal Validity
 Assessing External Validity
 Identifying Potential Confounding Variables

5

 View movie as:
 QuickTime (.mov)
 Flash (.swf)
 Doubleclick on video for fullscreen mode.

6


7

 Experimental Event Rate (EER)
 The proportion of patients (in the intervention) who experienced the
target disorder with the treatment
 Control Event Rate (CER)
 The proportion of patients (in the comparison group) who experienced
the target disorder without the treatment

8

 Experimental Event Rate (EER)
 Control Event Rate (CER)

9

 The estimated number of patients needed to be treated for every patient
benefiting from the treatment beyond baseline/control expectation
 So small numbers indicate greater effectiveness

10

 Essentially we want to know how many patients we must treat before we
can expect one successful treatment beyond what we would normally expect
without treatment.
 First consider, what we already know:
 We know both the proportions of how many patients still had the target
disorder that were treated (EER) and that were not treated (CER).

11

 With these two proportions we can find the proportion of successful
treatment beyond the baseline expectation (control).
 From there, we can determine how many we would need to treat to expect
one success.
 Now let’s start from the baseline: Find the proportion of
untreated patients with the target disorder (CER).

12

 Now we want to know how much better the treatment did than no treatment
at all. So we find the proportion of the treated patients that had the
target disorder (EER)
 Now we find their difference (CEREER), which we call the Absolute Risk
Reduction (ARR).

13

 That is if 10% of the untreated patients had the disease and 3% of the
treated patients had the disease, then the treatment helped 7% (=10%3%)
of the treated patients (which if left without treatment would still
likely have the disease). So 0.07 would be the ARR.
 Hence ARR is the probability that treatment will reduce the risk of a
given patient beyond baseline expectation.

14

 Now we know that the treatment reduced the risk of a proportion of
patients, which we call Absolute Risk Reduction and that ARR = CER
– EER.
 In other words, ARR is the number of successful treatments beyond
baseline expectation divided by the number of treated patients.
 Now we’re close. NNT is the number of treated patients per
successful treatments beyond baseline expectation. Can you see the
relationship yet?

15

 In this situation, we should look for a way to invert our number.
 We know that we can divide any number by 1 without altering it. So ARR =
ARR/1. So think of ARR as a fraction.
 Now think of the number of successful treatments beyond baseline
expectation as the unit on top of the fraction (i.e. ARR) and the number
of treated patients as the unit on the bottom.

16

 Now if we take the inverse (i.e. flip our fraction over) and calculate
1/ARR, we get a number where the units are flipped. Thereby, this number
has the number of treated patients as the unit on top of the fraction
(i.e. ARR) and the number of successful treatments beyond baseline
expectation as the unit on the bottom. This is exactly how we defined
NNT.

17

 NNT = 1/ARR = 1/CER – EER
 Often rounded up to nearest whole number

18


19


20

 The number of treated/exposed patients with the target outcome for every
patients in the control with the target outcome
 (Also used in therapy articles)
 RR = EER / CER

= (a/(a+b)) / (c/(c+d))

21


22

 The number of times the target outcome occurred in patients exposed to
the risk for each time the target outcome occurred in patients not
exposed to the risk.

23

 OR = (a/b) / (c/d)
 =
a*d / b*c
 A measure of association
 When large, there is greater association

24

 Critical Appraisal Practice Exercises

25

 EBM Glossary
 Critical Appraisal Practice Exercises

26

