Ethics, informed consent and statistics Paul S. Mueller, MD

1
Ethics, informed consent and statistics

• Paul S. Mueller, MD, MPH, FACP
• Division of General Internal Medicine
• Mayo Clinic Rochester

2
Questions we will cover today

• What are the elements of informed consent?
• Do people perceive risk similarly? If not, why
  not?
• My test is positive (negative). What does
  that mean?
• The treatment prevents (cures, etc.) a disease by
  50. Is it a good treatment?

3
"There are lies, damned lies and statistics."
Mark Twain
4
Informed consent

• Underlying ethical principle respect for patient
  autonomy
• Elements of informed consent
• Information
• Patient voluntarily agrees with plan
• Patient has decision-making capacity

5
AMA code on informed consent2000-2001 (8.08)
The patients right of self-decision can be
effectively exercised only if the patient
possesses enough information to enable
intelligent choices. The patient should make his
or her own determinations on treatment. The
physicians obligation is to present the medical
facts accurately
6
Informed consent legal aspects

• Based upon negligence principles
• State law governs malpractice
• Differing state standards shaped by case law
• Professional practice standard customary for
  other clinicians to do
• Reasonable person standard what a reasonable
  person needs to know (most states)

7
Information what would a reasonable patient want
to know?

• Nature of the intervention
• Benefits of intervention
• Risks
• Alternatives (and their benefits and risks)

8
Risk

• What is risk?
• Websters (1999) The chance of injury, damage,
  or loss dangerous chance hazard
• Understanding risk is complex
• Objective quantitative patients and clinicians
  have limited comprehension of the quantitative
  aspects of risk
• Subjective how important is it to the patient?

9
(No Transcript)
10
Risk Perception
High
Area of minimal concern
Area of most concern
likelihood
Area of least concern
Area of moderate concern
Low
Low
High
impact
11
Situations during which risk is commonly
discussed with patients

• Diagnostic tests
• If test X is positive, the chance of disease Y
  is
• Treatments
• The chance disease Y is cured by treatment X
  is...
• Treatment X reduces the risk (or recurrence) of
  disease Y by

12
If the test is positive, do I have the disease?
13
Dark circles have disease open circles no
disease Pink/diagonal lines positive test blue
negative test
Bad test for these people!
PPV probability of disease if test is positive
14
The chance a patient with a positive test has the
disease is

• True positives ? (true positives false
  positives)
• Here 24 ? (24 14) 63
• This is known as positive predictive value what
  we usually want to know!

• Similar concept negative predictive value
  (probability patient doesnt have disease if test
  is negative)
• Here 56 ? (56 6) 90

15
Bad test for these people!
Dark circles have disease open circles no
disease Pink/diagonal lines positive test blue
negative test
Sensitivity probability of positive test if
disease is present
16
The chance a test will be positive if the patient
has the disease is

• True positives ? (true positives false
  negatives)
• Here 24 ? (24 6) 80
• This is known as sensitivity
• Measures the effectiveness of a test

• Similar concept specificity (probability test is
  negative if the patient doesnt have the disease)
• Here 56 ? (56 14) 80

17
Dark circles have disease open circles no
disease Pink/diagonal lines positive test blue
negative test
A perfect test has 100 predictive value,
sensitivity and specificity. No such test exists.
18
Real example screening mammography

• For women with no FH of breast cancer
• Sensitivity 70-88
• Specificity 89-91
• Positive predictive value 1-6
• Sensitivity and specificity increase with age
• Mammography is not a perfect screening test for
  breast cancer

Ann Intern Med 2000133855-863
19
The bottom line

• Unfortunately, no test is perfect
• When a test is positive, the chance the patient
  has the disease is almost never 100
• Not all patients with positive tests have the
  disease the test is intended to detect

20
When a test is negative, and the disease is still
suspected, what do clinicians usually do next?
21
How do I know if the treatment that my doctor
suggests is good?
22
Hierarchy of evidence
23
A randomized controlled trial
Good outcome 7/10 70
Treatment new
Bad 3/10 30
Good 4/10 40
Treatment old
20
Bad 6/10 60
24
Worth noting

• Almost every trial demonstrates
• Some people get better on their own without
  treatment
• Not all people who are treated get better

25
By how much is treatment new better than
treatment old?

• Absolute risk reduction risk old treatment -
  risk new
• Here 60 - 30 30
• The risk of the bad outcome on treatment new is
  30 less than on treatment old.

• Relative risk (risk old treatment - risk new) ?
  risk old
• Here (60 - 30) ? 60 .50
• On treatment new, the chance of the bad outcome
  is 50 the risk of being on treatment old.

26
Expressed as graph
Absolute risk reduction 30

27
What is the problem with relative risk?
For each scenario, relative risk is the same even
though the absolute risk reduction is markedly
different!
28
For each scenario, relative risk reduction is
50, but the absolute risk reduction is much
different
ARR 30
Percentage
ARR 3
ARR 0.3
29
(No Transcript)
30
Accessed 6/14/2004
31
Physicians Health StudyNew Engl J Med
1989321129-135

• N 22,071
• 11,037 received aspirin (ASA) and 11,034 placebo
• Incidence (risk) of MI in ASA group was
  255/100,000 per year or 0.26 per year

• Incidence (risk) in placebo group was 440/100,000
  or 0.44 per year
• Absolute risk reduction of MI with ASA 0.18 per
  year
• Relative risk reduction 44

32
Accessed 6/14/2004
Cites Lancet 19943441383-1389
33
4S TrialLancet 19943441383-1389
N 4,444
Relative risk reduction (8.5 - 5.0) ? 8.5
0.42 or 42
Absolute risk reduction 8.5 - 5.0 3.5
34
Warfarin anticoagulation reduces the risk of
stroke in patients with atrial fibrillation (AF).
Should all patients with AF be anticoagulated?
Thousands of patients screened for these trials
were never enrolled because of co-morbid diseases.
35
Accessed 6/14/2004
Risk of stroke 85/8102 (1.05) placebo versus
127/8506 (1.49) EP for a hazard ratio of 1.41
or a 41 increased risk of stroke
Cites JAMA 2002288321-333
36
Be careful when contemplating risk

• What are the characteristics of the patients
  enrolled in the study?
• Which risk? Absolute or relative?
• Physicians, the media, medical journals, industry
  and lawyers often talk in terms of relative risk
• However, absolute risk reduction is often more
  relevant to patients

37
Number needed to treat (NNT) a better way of
communicating risk?

• NNT number of persons needed to treat to
  prevent (cure, etc.) one case
• NNT 1/absolute risk reduction
• Physicians Health Study NNT 1/0.0018 556

• Physicians Health Study would have to treat 556
  patients with ASA to prevent one MI
• 4S trial NNT 1/0.035 29 would have to treat
  29 patients to prevent one coronary death

38
(No Transcript)
39
Conclusions

• Medical statistics can be are complex and can be
  confusing
• No test is perfect
• Results of trials are often presented in terms of
  relative risk, which may be irrelevant to
  patients
• Effective communication of risk is essential for
  informed decision-making

40
Acknowledgements

• Amit K. Ghosh, MD, FACP
• BMJ 9/27/2003 a large portion of this issue is
  devoted to communicating risk to patients