EPI-820 Evidence-Based Medicine (EBM)

1
EPI-820 Evidence-Based Medicine (EBM)

• LECTURE 1 INTRODUCTION
• Mat Reeves BVSc, MS, PhD, Dip ACVS
• Department of Epidemiology
• Michigan State University

2
Objectives

• 1. Understand principles of EBM, what it is and
  what it isnt.
• 2. Define HSR, and understand components of
  clinical epidemiology
• 3. Understand distinction between disease and
  illness and their consequences
• 4. Understand the different definitions of normal
  and abnormal
• 5. Understand clinical application of RR, RD and
  NNT

3
I. What is Evidence Based Medicine?

• Evidence-based medicine (EBM) is the
  conscientious, explicit and judicious use of the
  current best evidence in making decisions about
  the care of individual patients (Sackett 1996).

4
IMPORTANT CONCEPTS

• synthesis of individual clinical expertise and
  external evidence from systematic research.
• stresses expertise in information gathering,
  synthesis and incorporation
• de-emphasizes memorization.
• rebellious disregard for authoritarian expert
  opinion.
• relies heavily on medical informatics technology
  (e.g., WEB based searches of MEDLINE)

5
EBM aims to improve patient care by

• more effective and efficient use of diagnostic
  tests
• better use of individual patient preferences
• better use of prognostic markers
• better use of effective but safe therapeutic/
  rehabilitative/preventive treatments.

6
5 steps for using EBM in clinical practice

• 1. Convert need for clinical information into
  answerable questions.
• 2. Track down the information using
  library- research techniques .
• 3. Critically appraise evidence - validity and
  clinical applicability.
• 4. Make a decision!
• 5. Evaluate and continually monitor clinical
  performance.

7
What are Health Service Research and Outcome
Research?

• Health Service Research (HSR) is the integration
  of epidemiologic, sociologic, economic and other
  analytical sciences to the study of health
  services. Focus is on the needs, demands, supply,
  use, and outcomes of health care delivery.

8
HSR involves evaluation of health services

• structure - resources, facilities and manpower.
• process - where, how and by whom is health care
  provided.
• output - amount, nature and cost of health
  services
• outcome - what are the measurable benefits of
  health care?
• Outcomes Research the study of the outcomes of
  interventions ( Clinical epidemiology).

9
II. What is clinical epidemiology?

• Clinical epidemiology is the methodological
  foundation for EBM
• quantitative approach (or "science") applied to
  the "art" of clinical practice.
• provides foundation for thorough, efficient
  clinical practice
• addresses questions at issue during
  clinician-patient interactions

10
Issue Question?

• Normal/abnormal? Is the patient sick?
• What abnormalities are associated with
  disease?
• Diagnosis How accurate are diagnostic tests?
• Risk factors What factors are associated with
  disease risk (or poor outcome)?
• Prognosis What is the likely outcome?
• Treatment How does treatment change course?
• Prevention How does early detection improve
  outcome? Can we prevent disease?
• Cause What factors result in the disease?
  What is the underlying pathogenesis?

11
Clinical epidemiology helps physicians to

• be thorough yet efficient
• characterize uncertainty
• select and interpret diagnostic test information
• choose optimal therapeutic intervention (
  clinical decision analysis).

12
Disease versus illness

• Disease / Illness / Sickness?

• Pneumonia
• Yellow Fever
• Hypertension
• Sleeping sickness
• Diabetes
• Femur fracture
• Obesity
• Depression
• Chronic Fatigue syndrome

• Near sightedness
• Hypothyroidism
• Osteoporosis
• Chronic back pain
• Atrial fibrillation
• Strep pneumonia infection
• Multiple sclerosis
• Chronic Hepatitis B infection

13
Components and implications of health - disease,
illness and sickness

• What is disease?
• No accepted definition
• Susser (1979)
• a physiological or psychological (homeostatic)
  disturbance

14
What is disease?

• Determinants
• Access to medical care - since it is usually
  diagnosed (determined) by medical professionals
• Case definition (medical)
• Understanding
• Variable - excellent to very poor
• Focus
• Doctors (centered on patho-physiology)
• Medical Profession Approach
• Cure through intervention
• Measurement
• Disease-free survival

15
What is illness?

• Susser (1979)
• The subjective state of the person who feels
  aware of not being well
• Patients expression of ill-health symptoms
  such as pain, nausea, depression, tiredness etc
  (individual subjective expression of symptoms)
• Interaction of underlying disease/problem and the
  patients reaction to it. (N.B. disease may or
  may not be present!)

16
What is illness?

• Determinants
• Learned behaviour (Hayes, 1978)
• Part socially determined (influenced by culture,
  ethnicity, age, gender)
• Understanding
• Often poor (e.g., backache, alcoholism,
  complaining well)
• Or good but no cure (e.g., diabetes, arthritis)
• Focus
• Patient
• Medical Profession Approach
• Tertiary prevention palliation, improve
  functional capacity, decr. complications
• Measurement
• QALY, DALY

17
What is sickness?

• Susser (1979)
• The state of social dysfunction i.e., the role
  that the individual assumes when ill
• Social role that person assumes when ill

18
What is sickness?

• Determinants
• Heavily influenced by the setting or predicament
  (socially determined)
• Influenced by environment (e.g., peace vs war),
  culture, time period, ethnicity, age, gender
• Understanding
• Usually poor
• Focus
• Patient/society
• Medical Profession Approach
• Tertiary prevention?
• Measurement
• QALY, DALY

19
IV. Normality and abnormality

• Last (1995)
• a) within the usual range of variation in a given
  population or group
• b) in good health or indicative or predictive of
  good health (normal indicates a low probability
  of disease)
• c) pertaining to the normal, Gaussian
  distribution (a range of values i.e., 2 SD)

20
Clinical use of normal

• Normal often means typical of the general
  population
• e.g., Chol 200mg/dl in 50 yr US male
• Normal also means not requiring further follow-up
  or intervention.
• requires judgement and experience

21
Laboratory use of normal

• Normal is usually defined in terms of a reference
  range 2 SD
• Limitations
• 1. Assumes data are normally (Gaussian)
  distributed but most clinical data arent!
• 2. Choice of 2 SD is arbitrary - why not choose
  90 or 99?
• 3. No general relationship between the degree of
  statistical unusualness and clinical disease
  e.g., anemia and clinical signs, and serum
  cholesterol and AMI

22
Laboratory use of normal

• Limitations Contd
• 4. The reference range depends on the reference
  population used.
• a) may include diseased persons.
• b) not the population that undergoes tests (they
  are healthy!).
• c) age/sex/race composition of population and
  subject selection process needs to be considered.
• 5. The reference range may or may not predict
  abnormality accurately. What is the Se and Sp of
  the reference range?
• 6. The exact reference limits are unstable (
  based on a small number of individuals in the
  tails of the distribution)

23
V. Risks and rates (practical applications)

• A. Cumulative Incidence rate
• CIR Num. of newly disease indv. for a specific
  time period
• Total number of population-at-risk for same time
  period
• ranges from 0 to 1 (its a proportion!)
• must be accompanied by a specified time period
• average risk

24
Risks and rates

• B. Incidence Density Rate
• IDR Number of newly disease individuals
  
• Sum of time periods for all disease-free
  indv.-at-risk
• denominator is "person-time" or "population time"
• ranges from 0 to infinity
• measure of the instantaneous force or speed
• dimension reciprocal of time i.e., time-1

25
Clinical Interpretation of Risks, Rates and
Relative Risks (RR)

• Clinical uses of risk and rates are often
  confused
• Both provide meaningful statements regarding
  commonness
• BUT in most clinical environments
• numerator is easy to obtain
• denominator is very difficult to estimate!!

26
Relative Risk (RR)

• Measure of the magnitude of association
• Typical clinical interpretation (RCT)
• what is the relative probability of the event in
  the treatment group compared to the control
  group
• not a very useful measure of the impact of
  treatment or risk factor intervention (need RD
  and/or PAF)
• Note
• RR is not reciprocal in nature (unlike the OR)
• most cases of disease do not have the risk factor
  - unless the RR is large (gt5) and the prevalence
  high (gt25), and
• the majority of people with the risk factor will
  not have disease (Rose, 1982)

27
Relative Risk Reduction (RRR)

• RRR 1 RR
• If RR 0.75, then RRR 25
• the proportionate risk difference or by how
  much in relative terms is the event rate
  decreased

28
The Risk Difference (RD) (or attributable risk)

• More clinically useful measure of the
  relationship between a risk factor and disease
• RD Risk (exposed) - Risk (unexposed)
• Table 1.1. Ten Year Risk of Coronary Death, Men
  45-65, Whitehall Study
• Serum Chol. Risk RR RD
• Quintile I 2.9
• 1.9 5.4 - 2.9 2.5
• Quintile V 5.4

29
Risk Difference (RD)

• RD the excess risk that a patient faces, or
• what is the absolute difference in event rates
  between the treatment and control groups
• Men in Quintile 5 have a 2.5 increased risk of
  death from CHD cf men in Quintile 1 over
  10-years.
• RD can be negative (implying protective effect).
• RD is much more meaningful/useful when facing
  important clinical decisions

30
Table 1.2. Comparison of RR and RD a
hypothetical example of Hormone Replacement
Therapy, Women 60-69

• Baseline 10-yr 10-yr RD
• Mortality Risk RR (HRT) (HRT - None)
• CHD
• 10 0.65 6.5 - 10 -3.5
• Breast CA
• 2 1.25 2.5 - 2 0.5
• Net Benefit - 3.0

31
Table 1.3. Two Year Risk of Sudden Death
Associated with QT Prolongation Relative to
History of AMI (Algra et al, 1991)

• Hx AMI No Hx AMI
• No QT prolongation 3.5 1.1
• QT prolongation 7.0 2.5
• RR (QT vs No) 2.0 2.4
• RD (QT vs No) 3.5 1.4

32
Risk Difference (RD)

• Clinical impact of QT prolongation is much
  greater in men with a prior MI - a fact only
  appreciated by the RD.
• A RD may be large enough to be of major clinical
  importance even when RR is only modest.
• RR is often misinterpreted by clinicians and
  patients when comparing benefits and risks (see
  Naylor, 1993 Forrow 1992 Malenka 1993 Bucher
  1994).
• In contrast to RR, the RD depends on background
  disease incidence, and can vary markedly from one
  population to another. So, you cannot translate
  RD calculated in one population to another!
• RD (like the CIR) is time dependent - the value
  will increase over time

33
The Number Needed To Treat (NNT)

• Very useful clinical measure that conveys similar
  information as the RD (Laupacis et al., 1988).
• NNT 1 / RD
• Example
• Among patients with prior MI, the RD between
  patients with and without QT prolongation is
  3.5. So,
• Approximately 29 patients with QT prolongation
  would have to undergo treatment to prevent one
  sudden death
• (i.e., NNT 1/0.035 28.6).

34
The Number Needed To Treat (NNT)

• The NNT is usually calculated for interventions
  that are known to do good i.e., where the RR is lt
  1.0.
• The NNT directly summarizes the effort needed to
  gain potential clinical benefit How many
  patients do I need to treat to prevent one
  event?
• High NNT bad, Low NNT good
• NNT depends on the efficacy of the intervention
  ( RR) and the underlying baseline risk
• Number needed to harm and number needed to screen.

35
Table 1.4. Effect of Base-line Risk and Relative
Risk of intervention of NNT
Base-line Risk () Relative Risk of Intervention Relative Risk of Intervention Relative Risk of Intervention Relative Risk of Intervention
0.5 0.75 0.80 0.90
60 3 7 8 11
30 7 13 17 33
10 20 40 50 100
5 40 80 100 200
1 200 400 500 1000
0.1 2000 4000 5000 10000
36
Clinical Use of the OR (vs RR)(See Sinclair 1994)

• OR are commonly reported in RCTs and
  meta-analyses, despite the fact that their
  clinical use and interpretation is difficult
• OR lacks intuitive clinical appeal
• Mistaken identify problem (its not a RR!)
• OR deviates substantially from RR when
• treatment effects are large and/or
• when baseline risks are high (Fig 2)
• Influence of changing baseline risks influences
• Interpretation of sub-group findings
• Predicted effect in other clinical populations