Question

A local clinical consultant colleague with a special interest in sarcoidosis (incidence 20 per 100,000) approaches you to help design a study to investigate his theory that sarcoidosis is caused by recent infection with a specific mycobacterium.

1. Describe an appropriate research design, and explain why that design is an appropriate choice to investigate this hypothesis.
2. Describe the general features of a confounding variable.

c. What techniques will you use to reduce confounding? Explain your answer with examples

Answer 1

Ans) a) Sarcoidosis is a rare condition. A study investigating aetiology would need to be an
observational study, and a case control study design would be most suitable.
b) A confounding variable is a variable, or exposure that is associated both with a
disease and with a causative agent that you are studying. A confounder should not be a variable that is „on the disease pathway‟ as that would be considered as an “effect modifier”. Confounding variables can increase the association between an exposure and an outcome, or decrease the association (i.e. they can be “positive” or “negative” confounders).
c) Confounding can be reduced by:
Design: e.g. matching – i.e. if sex is a confounder, every male case is matched with a male control; restriction – again if sex is a confounder, restricting the study to males only; not relevant in the above case [as it is an observational study] – but randomisation should eliminate confounding.
Analysis: stratified analysis – e.g. if sex is a confounder, analysing results for males separately to those with females. This is rarely performed now, as stratification reduces your sample size and power. Instead, potential
confounders are “adjusted for” using multivariate statistical methods in
observational studies.