Being an Informed Consumer of Drug Research

1
Being an Informed Consumer of Drug Research

• Robert E. McGrath
• Fairleigh Dickinson University

2
Outline

• Obstacles to objective decision-making in
  pharmacotherapy
• Review of research terminology
• Accurately estimating drug effects
• Utility analysis

3
Industry Impact on Data Sources

• Pharmaceutical industry funds half of all CE on
  medication (Holmer, 2001). CE presenters tend to
  be more positive about the funders product than
  presenters without support (Bowman, 1986).
• Villanueva, Peiro, Librero, Pereiro (2003)
  44.1 of claims in pharmaceutical ads were not
  supported by the reference, most frequently
  because the ad recommended the drug for a patient
  group not treated in the study.
• 87 of practice guideline authors who responded
  admitted pharmaceutical industry funding
  (Choudhry, Stelfox, Detsky, 2002).
• Industry is even a major supporter of
  bioethicists (Elliott, 2004)

4
Implicit Information-Gathering(Avorn, Chen,
Hartley, 1982)

• Practicing physicians rated scientific sources
  much more important influences on prescribing
  than commercial sources.
• Also gauged knowledge in two cases where the
  message about medications from the scientific
  literature contradicted the commercial
  literature.
• The majority of doctors responded in a manner
  consistent with commercial literature.

5
The Principle of Least Effort(Haug, 1997)

• When seeking information about cutting-edge
  treatments, physicians tend to choose easily
  available information sources, even if it is of
  low quality, over higher-quality sources that
  require more effort.

6
Personal Misestimation of Treatment Effectiveness

• Cognitive Errors (Arkes, 1981)
• Covariance Misestimation
• Expectancies
• Logical Errors post hoc, ergo propter hoc
• Natural history of the disorder
• placebo effects

Improved Didnt Improve
Received Medication A B
Didnt Receive Medication C D
7
Becoming a Critical Consumer

• Being a critical consumer means critically
  evaluating research
• Lack of access to research data
• The Internet!

8
Statistical Terminology

• p The probability of your sample outcome if the
  null hypothesis is true. For two groups, the
  probability of this sample difference between
  group means if the difference is 0 in the
  populations. For a correlation, the probability
  of this sample correlation, if the correlation is
  0 in the population.
• a The p value at which you are willing to reject
  the null hypothesis that the population value
  0. The probability of rejecting the null
  hypothesis if the null hypothesis is true
  (incorrect rejection Type I error).
• The problem Population differences or
  correlations rarely equal 0.

9
Statistical Terminology (contd)

• Power (1 - ß) The probability of rejecting the
  null hypothesis if the null hypothesis is false
  (incorrect rejection). A function of
• a ?a, ?power
• Sample size ?sample size, ?power
• Effect size ?effect size, ?power
• Effect size The size of the difference or
  correlation in the population or sample.
• The larger the effect, the easier it is to reject
  the null hypothesis (greater power)
• Common measures
• d The difference between means divided by the
  standard deviation
• r The standard correlation coefficient

10
More Effect Sizes

• Odds ratio Odds of improvement in the treatment
  group divided by odds of improvement in control
  group (declining in popularity)
• Risk ratio Probability of improvement in the
  treatment group divided by probability of
  improvement in control group
• Number needed to treat The number of cases
  needed to be treated to have one more positive
  outcome. Smaller is better. E.g., NNT 4 means
  you will get 1 more positive outcome for meds
  than placebo for every 4 treated.

Improved Not Improved
Meds n11 n12
Placebo n21 n22
11
Examples

• Odds ratio
• Risk ratio
• NNTN

Improved Not Improved
Meds 20 80
Placebo 4 96
12
Methodological Terminology

• Last Observation Carried Forward (LOCF) An
  analysis in which participants last observation
  is used, even if they dropped out. All
  participants are included.
• Observed Cases (OC) An analysis restricted to
  participants who completed the entire protocol
• Evidence is poor that OC effects are larger
  (Breier Hamilton, 1999 Kirsch, Moore,
  Scoboria, Nicholls, 2002)
• LOCF significance tests are more powerful.
• Meta-analysis An integration of prior research
  findings across studies. Focus on size of effects
  rather than significance.

13
SchizophreniaAbilify (aripiprazole)

1. Google Abilify. Go to www.abilify.com.

14

1. Click on For Healthcare Professionals
2. Click on Efficacy
3. Click on Symptom Improvement

15
(No Transcript)
16

• Google PANSS
• Positive and Negative Syndrome Scale (PANSS)
• Kay, Fiszbein, Opler (1987)
• 30-item scale
• 16 general psychopathology symptom items
• 7 positive symptom items
• 7 negative symptom items
• completed by the physician
• Each item is scored on a 7-point severity scale
• A patient with schizophrenia entering a clinical
  trial typically scores 91.

17

• Positive Symptoms
• Negative Symptoms
• General Symptoms

Delusions Grandiosity
Conceptual disorganization Suspiciousness/persecution
Hallucinatory behavior Hostility
Excitement
Blunted affect Difficulty in abstract thinking
Emotional withdrawal Lack of spontaneity/flow of conversation
Poor rapport Stereotyped thinking
Passive/apathetic social withdrawal
Somatic concern Motor Retardation Disturbance of volition
Anxiety Uncooperativeness Poor impulse control
Guilt Feelings Unusual thought content Preoccupation
Tension Disorientation Active social avoidance
Mannerisms and posturing Poor attention
Depression Lack of judgment/insight
18

1. After 4 weeks, Abilify reduced PANSS score by 14
   (15 of 91)

1. Positive score only improved by 5 points

1. Negative score only improved by 3 points

1. About half of the effect had to do with general
   symptoms

19
Mean improvement in HAM-D score 2.4 (LOCF)-3.5
(OC) points Mean improvement in mood score .4
(OC) -.5 (LOCF) points Conclusion It doesnt
take much to get this guy golfing again!
20
Why Therapy is Better

• The utility (clinical significance) of an
  intervention is a function of three factors
• The size of the effect ?effect, ?utility
• The treatments value ?value, ?utility
• The costs or risks ?cost/risk, ?utility
• Interpreting effect sizes (Cohen, 1988)
• d small .20 medium .50 large .80
• r small .10 medium .30 large .80

21
Examples of Utility Analysis

• Physicians Aspirin Study r .034 (Rosenthal,
  1990)
• ECT (Carney et al., 2003)
• d .91 versus placebo mean Hamilton difference
  9 points
• d 1.01 versus meds mean difference 5 points

22
Comparing Meds to Therapy

• Greater risks must be offset by greater value
• Lasser, Allen, Woolhandler, Himmelstein, Wolfe,
  Bor (2002) Among drugs FDA approved 1975-1999,
  8.2 acquired an additional black box warning
  2.9 were withdrawn
• Kathol Henn (1982) Half of serious adult
  overdoses involved tricyclics (dated article)

23
Comparing Meds to Therapy (contd)

• Therapy can be at least as effective as meds
• Therapy equal to or better than meds for
  depression, even severe (Antonuccio, Danton,
  DeNelsky , 2004)
• Mean d for treating cognitive problems in
  schizophrenia with
• Meds .22 (Mishara Goldberg, 2004)
• Cognitive rehab .45 (Krabbendam Aleman, 2003)
• Increasing evidence total cost for therapy is
  cheaper for depression (Antonuccio et al., 2004)
  and anxiety disorders (Heuzenroeder et al., 2004)

24
Why Therapy is Better (contd)

• Comparison
• Effect size Therapy Meds
• Value Meds Therapy
• Risks Meds gt Therapy
• Cost Meds Therapy
• Therapy gt Meds

25
Being an Informed Consumer

• Be aware that information may be biased, even if
  it comes from trustworthy sources
• Monitor your own use of meds
• How many are on prescription?
• What are they taking?
• How many are taking multiple meds?
• How long are they maintained on meds?
• Outcomes?
• Do the results match your beliefs?