18: Cross-Tabulated Counts

1
Chapter 18Part C Matched Pairs
2
18.6 Matched Pairs

• Binary data
• Match on extraneous factors that can confound
  results
• 11 match

3
Case-Control Example

• Case-control study on colon polyp re-occurrence
• Cases controls matched on age, clinic, sex and
  other potentially confounding factors
• Let exposure low fruit/veg consumption
• Ascertain exposure in case-control pairs
• 45 pairs case exposed and control not exposed
• 24 pairs control was exposed and case not exposed

Case exposed Case non-exposed
Control exposed ? 24
Control non-exposed 45 ?
4
Matched-Pairs Odds Ratio
Case E Case E-
Control E a b
Control E- c d

Interpret like any other odds ratio i.e, as a
relative risk
5
Colon polyp recurrence / low fruit illustrative
data
Case E Case E-
Cntl E ? 24
Cntl E- 45 ?
88 higher risk w/ low fruit/veg consumption
6
Confidence Interval for OR
Matched Pairs Matched Pairs Matched Pairs
Case E Case E-
Control E a b
Control E- c d
7
95 CI for OR Colon polyps illustrative example
Interpret like any other CI
8
WinPEPI gt PairEtc gt A.
Input
WinPEPI produces these superior confidence limits
9
Cohort Matched-PairsExposed and non-exposed
matched pairs

• Smoking and non-smoking twins (exposure
  smoking)
• Disease earlier death
• 17 twin-pairs in which smoking twin died first
• 5 twin-pairs in which non-smoking twin died first

Smoker D Smoker D-
Non-smoker D 0 5
Non-smoker D- 17 0
10
Hypothesis Test

• A. H0 OR 1 (no association)
• B. McNemars z (or equivalent chi-square)
• C. P-value

11
Twins Mortality Example
Conclude Good evidence against H0 (or other
reasonable P-value interpretation)