EBM and EB Guidelines

1
EBM and E-B Guidelines

• EBM integrates evidence, expertise, and the
  unique biology and values of individual patients.
  
• Local EB Provision ought to integrate evidence,
  expertise, and the unique biology and values of
  the local scene.

2
EBM and E-B Guidelines

• The best evidence comes from systematic reviews
  (such as Cochrane) and/or E-B journals of 2º
  publication
• Much more likely (than personal search and
  critical appraisal) to be true
• Saves the clinicians precious (scarce!) time
• Avoids error and duplication of effort

3
EBM and E-B Guidelines

• But NO systematic review can (or should try to)
  identify the 4 Bs
• Burden
• Barriers
• Behaviours
• Balance
• They can ONLY be determined at the local (or even
  patient) level

4
1. Burden

• The burden of illness, disability, and untimely
  death that would occur if the evidence were NOT
  applied
• the consequences of doing nothing

5
2. Barriers

• Patient-values preferences
• Geography
• Economics
• Administration/Organisation
• Tradition
• Expert opinion

6
3. Behaviours

• The behaviours required from providers and
  patients if the evidence is applied.
• All that guidelines can do is specify the former!

7
4. Balance

• The opportunity cost of applying this guideline
  rather than some other one.

8
Killer Bs

• Burden too small to warrant action.
• Barriers ultimately down to patients values.
• Behaviours may not be achievable.
• Balance may favour another guideline over this
  one.

9
Two monumental wastes of time and energy

• First, national/international evidence-summarising
  groups prescribing how patients everywhere
  should be treated.
• Their expertise predicting the health
  consequences if you do treat.
• Their ignorance the local Bs, and whether
  killer Bs are operating.

10
Two monumental wastes of time and energy

• Second, local groups attempting to systematically
  review the evidence.
• Their expertise identifying the local Bs and
  eliminating the killer Bs
• Their ignorance searching for all relevant
  evidence Chinese performing tests for
  heterogeneity.

11
Applying a study result to my patient

• Never interested in generalising
• Am interested in a special form of extrapolation
  particularising

12
Extrapolating (particularising) to my individual
patient

• First and foremost Is my patient so different
  from those in the trial that its results can make
  no contribution to my treatment decision?
• if no contribution, I restart my search
• if it could help, I need to integrate the
  evidence with my clinical expertise and my
  patients unique biology and values...

13
To add Clinical Expertise and Patients Biology
Values

• What is my patients RISK ?
• of the event the treatment strives to prevent?
• of the side-effect of treatment?
• What is my pts RESPONSIVENESS?
• What is the treatments FEASIBILITY in my
  practice/setting?
• What are my patients VALUES ?

14
To add Clinical Expertise and Patients Biology
Values

• I begin by considering Risk and Responsiveness
  for the event I hope to prevent with the
  treatment
• The report gives me (or I can calculate) an
  Absolute Risk Reduction ARR for the average
  patient in the trial.
• ARR probability that Rx will help the average
  patient.

15
For example, Warfarin in nonvalvular atrial
fibrillation

• After 1.8 years of follow-up in an RCT
• Control Event Rate (placebo) 4.3
• Exper. Event Rate (warfarin) 0.9
• so, for the average patient in the trial, the
  probability of being helped, or Absolute Risk
  Reduction (CER - EER) 3.4 ACPJC
  199311842

16
How can I adjust that ARR for my pts Risk and
Responsiveness?

• Could try to do this in absolute terms
• my Patients Expected Event Rate PEER
• and multiply that by the RRR
• and factor in my Patients expected
  responsiveness
• Clinicians are not very accurate at estimating
  absolute Risk and Responsiveness

17
How can I adjust that ARR for my pts Risk and
Responsiveness?

• Clinicians are pretty good at estimating their
  patients relative Risk and Responsiveness
• So, I express them as decimal fractions
• frisk (if at three times the risk, frisk 3)
• fresp (if only half as responsive e.g., low
  compliance, fresp 0.5)

18
How can I adjust that ARR for my pts Risk and
Responsiveness?

• probability that Rx will help my patient ARR x
  frisk x fresp
• If ARR is 3.4
• and I judge that their frisk is 3
• and that their fresp is 0.5
• then the probability that warfarin will help my
  patient 3.4 x 3 x 0.5 5.1

19
Must also consider the probability that I will do
harm

• In the case of warfarin serious bleeding
  (requiring transfusion) from the g-i tract, or
  into the urine, soft tissues or oropharynx.
• Absolute Risk Increase 3 at 1 yr, so ARI
  estimated to be 5 in 1.8 years
  ACPJC 199412052

20
and adjust the probability of harm for my patient

• Again, can express my clinical judgement in
  relative terms fharm
• Given my patients age, I judge their fharm to
  be doubled 2
• then the probability that Rx will harm my patient
  ARI x fharm 5 x 2 10

21
Can now begin to estimate the Likelihood of Help
vs. Harm

• Probability of help ARR (embolus) x frisk x
  fresp 5.1
• Probability of harm ARI (haemorrhage) x
  fharm 10
• My patients Likelihood of Being Helped vs.
  Harmed LHH is (5.1 to 10) or 2 to 1
  against warfarin!
• or is it ?

22
The LHH has to include my patients values

• I need to take into account my patients views
  (preferences, utilities) about the relative
  severity
• of the bleed I might cause
• to the embolus I hope to prevent
• Expressed in relative terms s
• if the bleed is half as bad as the embolus, then
  s 0.5

23
On in-patient services in Oxford and Toronto

• When Dr. Sharon Straus has described a typical
  embolic stroke (with its residual disability) and
  typical moderate bleed (brief hospitalisation and
  transfusion but no permanent disability)
• for most of her patients, a bleed is only 1/5th
  as bad as a stroke
• so the s is 0.2

24
So the LHH becomes

• ARR for embolus x frisk x fresp vs. ARI
  for bleed x f-harm x s
• 3.4 x 3 x 0.5 5.1 vs. 5 x 2 x 0.2
  2
• LHH 5.1 to 2 or 2.5 to1
• (I am more than twice as likely to help than harm
  my patient if they accept my offer of Rx)

25
We can work out the LHH for most patients lt6
minutes

• To be feasible on our service has to be
  do-able in 3 minutes.

26
Reactions from our patients

• All are grateful that their values/opinions are
  being sought
• 1/3 want to see the calculations, perhaps change
  their value for s, and make up their own minds.
• 1/3 adopt the LHH as presented.
• 1/3 say Whatever you tell me, doctor!