Why are casecontrol studies deceitful

1
Why are case-control studies deceitful ?

• Jørn Wetterslev M.D., Ph.D.
• Copenhagen Trial Unit
• Centre for Clinical Intervention Research
• Rigshospitalet

2
Presentation

• Definition
• Example from anaesthesia
• Hypothetical advantages
• Selection of cases and controls
• Selection bias
• Seductive power

3
Case-control study design
4
Case-control study in anaesthesiology

• Impact of Anesthesia Management Characteristics
  on
• Severe Morbidity and Mortality
• M. Sesmu Arbous, Anneke E. E. Meursing, Jack W.
  van Kleef,
• Jaap J. de Lange, Huub H. A. J. M. Spoormans,
  Paul Touw,
• Frans M. Werner, Diederick E. Grobbee.
• Anesthesiology 2005 102257 68

5
Advantages ?

• Sample size sparing ?
• Time sparing ?
• Powerful ?
• Ethical as no experiments are performed ?

6
Dutch case-control study of anaesthesa-related
morbidity and mortality1995-1997

• Does anaesthesia staffing resources and quality
  in management of patients influence patient
  outcome of deaths and coma during the first 24
  postoperative hours ?

7
Design

• Nested case-control study (prospective) with
  retrospectively matched controls
• Within a prospectively defined cohort of 869,483
  patients for anaesthesia
• Detecting the cases and the controls along the
  road of time, but retrospectively matched

8
Patients cases controls

• Cases were patients who either remained
  comatose or died within 24 hrs of undergoing
  anaesthesia
• Controls were matched patients who neither
  remained comatose nor died within 24 hours of
  undergoing anaesthesia

9
Selection of cases in article Anesthesiology 2005

• Immediately after a death or coma, the
  Procedure Questionnaire and an anonymous
  anaesthesia and recovery form were submitted to
  the study centre

10
Selection of controls

• As soon as a case was reported, a
  control was drawn from a randomly selected
  hospital using tables of random numbers. The
  control was matched for sex and age within 5
  yrs of the case. The correspondent ensured
  that the controls were randomly drawn
  from the operating schedule. Instructions
  were given on how to proceed in case no one, or
  more than one person, fulfilled the matching
  criteria.

11
Patients cases controls

• The cohort 869,483 patients
• 807 cases
• 883 controls

12
Association of risk factors pre-, intra-, and
postoperative

• Checks and use of equipment
• Presence of physician
• Pain medication

13
Results Preoperative anaesthesia management
risk factors

• Equipment check, performed with a checklist and
  protocol, was associated with a decreased
  risk of perioperative morbidity and
  mortality as opposed to no check
• odds ratio 0.64 (0.43 to 0.95)
• Documentation of the check was similarly as-
  sociated with a decreased risk
• odds ratio 0.61 (0.40 to 0.92)

14
Results Intraoperative anaesthesia management
risk factors

• Direct (intercom) availability of the
  anesthesiologist during maintenance compared
  with indirect (by means of telephone, beeper,
  or walkie-talkie) availability was associated
  with a significantly lower risk
• odds ratio 0.46 (0.31 to 0.66)

15
Results Postoperative anaesthesia management
risk factors

• Postoperative administration of opiates as
  opposed to no pain medication was associated
  with a decreased risk of coma or death
• odds ratio 0.16 (0.11 to 0.25)
• local anesthetics
• odds ratio 0.06 (0.01 to 0.40)
• combination of opiates and local
  anaesthetics
• odds ratio 0.32 (0.14 to 0.75)

16
Selection of cases revisited

• Immediately after a death or coma, the
  Procedure Questionnaire and an anonymous
  anesthesia and recovery form were submitted to
  the study center
• The investigator is responsible for the tracking
  of cases within 24 hours of anaesthesia ! Can you
  depend on that ?

17
Selection of controls revisitedin article
Anesthesiology 2005

• As soon as a case was reported, a
  control was drawn from a randomly selected
  hospital using computer-generated tables of
  random numbers. The control was matched for
  sex and was aged within 5 yr of the case. The
  request for a control patient was made to
  the correspondent of the selected hospital by
  telephone and by letter

18
Selection of controls revisitedin design article
Anaesthesia 1998

• .As soon as a case was reported, two controls
  will be drawn from two randomly selected
  hospitals. The correspondent in these two
  hospitals ensured that the controls will be
  randomly drawn from the operating schedule,
  Instructions were given if no-one or more than
  one fulfilled the matching criteria..

19
Selection of cases and controls

• The representativeness of the cases in the total
  cohort of 869,483 patients is plausible only if
  all cases are tracked !!
• The controls should be representative of the
  total cohort, especially with respect to the
  determining, or anaesthesia related, risk
  factors.

20
Exposure to the risk factors in controls in
the total cohort

pcontrols frequency of exposure to a risk
factor among the controls pundiseased
frequency of exposure in the total cohort,
without cases If pcontrols ? pundiseased
bias is introduced into the case- control
study
21
Exposure to the risk factors in controls in
the total cohort

• If pcontrols lt pundiseased
• the influence of the risk factor certainly will
  be overestimated
• Chapter 6, Woodward M. Epidemiology, Study
  Design And data Analysis, Texts in Statistical
  Science, 1999 Chapman Hall/CRC.

22
Disadvantages ?

• Sample size sparing ? Possibly, but why ?
• Time sparing ? 10 years of investigation ?
• Powerful ? Artificial universe !?!
• Ethical as no experiments are performed ? No !
• As conclusions may be misleading

23
Conclusions

• The advantages of case-control studies may be
• Hypothetical
• Selection bias is imminent and prominent
• Seductive power

24
Thank you !