Endgames Statistical Question

Confounding in case-control studies II

BMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c5290 (Published 29 September 2010) Cite this as: BMJ 2010;341:c5290
  1. Philip Sedgwick, senior lecturer in medical statistics
  1. 1Section of Medical and Healthcare Education, St George’s, University of London, Tooting, London
  1. p.sedgwick{at}sgul.ac.uk

Researchers conducted a nested case-control study to investigate whether breast feeding protected infants against pneumonia.1 All children born in urban areas of Southern Brazil in 1993 were included in the cohort. Cases were a total of 152 infants who had been born in 1993, admitted to hospital when aged 28­364 days and diagnosed with pneumonia. Controls were a total of 2391 infants selected from the cohort. Cases and control were not matched.

The mothers of all cases and controls were interviewed about the infant’s past diet, with information collected on type of milk consumed, use of fluid supplements, plus use of solid and semisolid supplements. Each food type was investigated as a possible risk factor for the development of pneumonia. The results are presented in the table, which shows the odds ratios of pneumonia according to type of food given.

View this table:

Odds ratios for developing pneumonia according to type of food given

Which of the following, if any, are true?

  • a) The effects of age and sex as potential confounders could be removed if cases and controls were matched for these variables.

  • b) For each risk factor, the size of the difference between the unadjusted and adjusted odds ratios permits the extent of confounding to be determined. …

View Full Text

Log in

Log in through your institution


* For online subscription