2.7 - 9: Confounding, matching and standardisation

Question 1

Define confounders:

Answer 1

• A third variable associated with the exposure of interest
• Independently associated with the risk of disease (should not lie on casual pathway between the two)
• Explains partially or fully the relationship between E and D

Question 2

Describe some possible consequences of confounding:

Answer 2

• Creation of an apparent relationship (spurious) between E and D
• Masking a true relationship between E and D
• Causing an overestimate or underestimate of true effect

Question 3

Methods to control confounding:

Answer 3

Design based:

• Randomisation; promoting balance over potential confounders
• Restriction; limiting participation of those similar in relation to the confounder
• Matching; selecting controls to be similar to cases in terms of C

Analysis based:

• Stratification; estimating pooled estimate of association measure; adjust for C effect
• Standardisation; controlling C using external population to adjust
• Multivariate analysis; regression models including C variable in model

Question 4

How is stratification carried out?

Answer 4

• Allowing association between exposure and outcome to be examined within strata of confounding variable
• Mantel-Haenszel method most widely used
• Tabulate each of the k
• Use Mantel-Haeszel estimate of common odds ratio; producing weighted average

Question 5

Types of matching:

Answer 5

• Group: constant ratio of cases and controls within broad strata
• Individual: Matching controls to each case (e.g. 1:1)
• The rationale in a matched case-control is to eliminate confounders by design

Question 6

Describe some advantages and disadvantages of matching:

Answer 6

+

• Control confounders by elimination
• Gain in efficiency
• Avoid/minimise selection bias

-

• More complicated study design
• Not possible to study the effect of matching variables on the outcome of interest
• Overmatching
• Note: Risk, RD, RR cannot be estimated in matched case-control studies

Question 7

Types of standardisation:

Answer 7

• Direct: the disease rates in the population of interest are applied to the ‘standard’ population
• Indirect: the disease rates in the standard population are applied to the population of interest
• -> Both involve calculating expected E numbers of events (e.g. deaths) and comparing them to the observed number of events O
• Commonly based on age or age/gender strata

Question 8

Direct age standardised event ‘rate’ per 1,000:

Answer 8

• See notes

Question 9

SMR: What is it, how is it calculated?

Answer 9

• Standardised mortality ratio
• See notes page for formula
• Express as a %
• Represents ratio of two indirectly standardised rates