Evidence Based Medicine

Front 1

best study for prognosis

Back 1

cohort

Front 2

best study for aetiology

Back 2

cohort

Front 3

best study for intervention

Back 3

RCT

Front 4

best study for diagnosis

Back 4

diagnostic test evaluation

Front 5

Levels of Prevention - definition

Back 5

•
Primary prevention - protecting healthy people from developing a disease or experiencing an injury in the first place
•
Secondary prevention - the goal is to halt or slow the progress of disease (if possible) in its earliest stages
•
Tertiary prevention – goals include preventing further physical deterioration and maximising quality of life

Front 6

Why we can’t we infer reliable estimates of effect from case reports?

Back 6

1. random error
2. systematic error:
- publication bias (more for +ve rather than -ve results)

Front 7

The role of case reports

Back 7

- hypothesis generation
- Awareness raising of rare and unexpected complications of interventions
› Convincing when the natural history is uniform and poor and when the
intervention is completely and universally effective
- The “parachute” effect
- Dialysis for symptomatic end-stage
kidney disease

Front 8

Why where the HRT for menopause observational studies intially so wrong?

Back 8

- may have been selection bias?
- or they where more health conscious (selection bias) people more likely to get better

Front 9

Role of observational studies

Back 9

› Hypothesis generating for treatment effects
› Essential for evaluating prognosis and therefore who to treat
› Often the only evidence we have (simple, feasible)

Front 10

First trial

Back 10

James Lind
1747 6 treatments for scurvy
>150yr lag before citrus was used to treat scurvy

Front 11

Formulating a well built question

Back 11

- eg for BP lowering agents we want the outcome to be CVR and not BP lowering because all BP lowering drugs lower BP
PICO

Front 12

A checklist for reporting
(Cochrane risk of bias tool)

Back 12

› Were methods for sequence generation reported?
› Was there adequate allocation concealment?
› Was there complete outcome data?
› Was there intention to treat analysis?
› Was there blinding of participants, investigators, outcomes
assessors, data analysts?
› Was there selective outcome reporting?
› Other sources of bias?
- Cointerventions
- Commercial
› Yes, No, Unclear

Front 13

Intention to treat

Back 13

this is when they are given the medication and not checked up to see if they were compliant or not

› Analysis by intention to treat preserves randomization.

Front 14

Bias in treatment assignment in controlled clinical trials

Back 14

Front 15

Adequate allocation concealment

Back 15

› Adequate
- central randomisation
- numbered/coded bottles/containers
- drugs prepared by the pharmacy
- serially numbered, opaque, sealed envelopes
› Inadequate
- alternation, date of birth

Front 16

Should we include?

Back 16

› A patient who ceased medication after only 1% of
medication?
› A patient who died before starting randomised treatment?
› A patient whose baseline lab results were in error and was
ineligible?
› A patient who developed cancer shortly after starting
treatment?

Front 17

Commercial bias

Back 17

“The sponsor .. has the right to postpone the
publication, for a period not to exceed 18 months..
the sponsor can discontinue the study.. if the aim of
the trial has become no longer of interest..a detailed
publication policy will be developed..the steering
committee agrees to include the modifications
requested by the sponsor”

Front 18

What should the comparator be?

Back 18

› Evidence-based standard of care
- Not placebo unless “nothing” is evidence-based standard of care
› How can clinicians and policy makers evaluate an intervention against the
clinically appropriate comparator when the intervention has only be
compared against second-best or placebo?

Front 19

Checklist of thinks to report in a RCT

Back 19

Front 20

how to read and critique a RCT

Back 20

› Read efficiently – prioritise the abstract, methods, tables+figures
› Comparator: Is it the best available?
› Outcome: Is it patient relevant or validated surrogate?
› What is the risk of bias?
- Randomisation sequence described and valid?
- Was there adequate allocation concealment?
- Was there blinding x4?
- Was there complete outcome data?
- Was there intention to treat analysis?
- Was there selective outcome reporting?
- Was there other sources of bias?
- Co-interventions?
- Funding?
82

Front 21

Confidence intervals

Back 21

› Sample size determines precision not validity
- a large trial may be biased
› Confidence intervals
- if the same study were repeated 100 times then the results of 95 of these 100
studies would lie somewhere within the 95%CI
- you can be 95% confident that the true result lies within the 95%CI

Front 22

P-values
e.g P-value = 0.04

Back 22

› p=0.04 means “If the null hypothesis were true (ie no
difference between two groups) then similar data or more extreme to that observed would be obtained 4% of the time by chance alone”
› Does not protect against bias

Front 23

P-values from
confidence intervals

Back 23

› If the confidence interval includes the “no effect” value
= not statistically significant
› If the confidence interval does NOT include the “no effect” value
= statistically significant

Front 24

practice question

Back 24

@ th end of the lecture

Front 25

ARR

Back 25

absolute risk reduction= risk with no Tx - Rsk with tx

Front 26

NNT

Back 26

1/ARR