Answers

b—In the above study, two groups of individuals were selected on the basis of their disease status: glioma (the cases) or no glioma (the controls). The aim of the case-control study was to identify potential risk factors for glioma. Cases and controls were compared for past exposure to proposed risk factors, including mobile phone use, to ascertain whether such factors were associated with a raised or reduced risk of glioma.

Cases and controls were matched for age and sex. Matching involves finding a control of similar age and same sex for each individual case. Matching by age is often to within five years of the age of the case. More than one control may be matched to each case to increase the precision of statistical analyses.

Matching by age and sex ensures that any difference between cases and controls cannot be a result of differences in the matched variables (answer b). The effects of age and sex as potential confounding variables have, therefore, been controlled for. Matching by age and sex means that adjustment for these potential confounders was made when designing the study, which is more efficient than making adjustments during statistical analyses. However, we cannot then investigate the matched variables as risk factors for glioma. Age and sex may be related to the development of glioma itself, but they were not of interest as potential risk factors in this study.

Answer a is false. Assessment bias is when information recorded by the researcher does not equate with the experiences of the study participant. This can occur if the researcher intentionally records information inaccurately or elicits information in a manner that favours one group over the other. Assessment bias will not have occurred in the above study, however, because a self report questionnaire and not an interview was used to collect information about potential risk factors. Matching will not affect the accuracy of information collected about potential risk factors.

Answer c is false. Allocation bias is the systematic difference in the characteristics of individuals assigned to groups, with particular reference to the allocation of patients to treatment groups in trials. In a case-control study, however, the cases and controls are selected on the basis of their disease status—no allocation of individuals occurs.

Answer d is false. Recall bias can occur when cases recall information about potential risk factors more accurately than do controls. For example, cases are more likely than controls to recall details about mobile phone usage because of their glioma. Recall bias will not be influenced by matching.

Cite this as: BMJ 2009;339:b4581