Confounding and effect modification Manish Chaudhary BPH(IOM, TU), MPH(BPKIHS)
Confounding and effect modification Confounding refers to the effect of an extraneous variable that entirely or partially explains the apparent association between the study exposure and the disease. Confounding is a distortion in the estimated measure of effect due to the mixing of the effect of the study factor with the effect of other risk factor(s). If we do the analysis by ignoring the potential confounding factors, we might get an obscure conclusion on the association between factors.
AB C Criteria for confounders It is a risk factor of the study disease (but it is not the consequence) It associates with exposure under study (but not with the consequence of such exposure). It is about of interest of current study ( i.e. an extraneous variable) In the absence of exposure it indendently able to cause disease (outcome)
Example of confounding Assume a case- control study of association between stomach cancer and smoking Smoker Non-smoker Cases 170 (a) 80 (b) Control 80 (c) 170 (d) 250 OR= 4.5 The risk of getting stomach cancer among smoker is 4.5 times higher than non smoker
Example of confounding Stratified by drinking habit(alcohol) Drinking(-)Drinking(+) Smoker Non- smoker Cases 2050 Control OR= 2.0 Smoker Non- smoker Cases Control The risk of getting stomach cancer among smoker is 2 times higher than non smoker in both drinker and non drinker OR= 2.0
Example of confounding First criteria of confounder, drinking habit relates to stomach cancer Drinking + Drinking - Cases Control OR= 16.0 Drinking habit is a risk factor of stomach cancer
Second step is to determine whether drinking habit relates to smoking – 70/250 (28%) of non smoker drink – (72%) of smoker drink Smoking and drinking are related. Final third criteria is main interest of the study is stomach cancer and smoking. Drinking habit is out of interest of ie extraneous factor. All three criteria of confounder are true.
Effect modifier If the degree of association between the exposure of interest and study out come varies according to third variable, the third variable is then called effect modifier.
Example Incidence of MI in relation to smoking in age group AgeSmokin g MINon MItotalRR < Ps= Ps= 0.5 Total
In this example the effect of smoking is not constant over age strata Risk of MI is 1.4 times in younger age while risk is 3.6 times in older age group. Overall risk of MI is 2.1 times But there is no relationship between smoking and age.(same in both age group)
Look at another table AgeMINon MITotalRR < Age is related to MI. Thus age is effect modifier in this example
Control of confounding In research design During data analysis phase Three methods to control confounding during the design phase of the study: – randomization – restriction – matching
Error of measurement 1. Instruments poor calibration or lack of sensitivity 2. Observer's variation – Intra- observer variations: Semi skilled observers are often inconsistent in diagnosis of the same specimen presented to him blindly on different occasions. – Inter - observer variation: Several observers do not always agree on the diagnosis of the same specimen. 3. Observer's lack of skill or experience to use the apparatus or to give interpretation of diagnosis 4. Patient's lack of cooperation 5. Patients are not measured in the same manner, under the same condition or atmosphere
Summary