A short introduction to epidemiology Chapter 10: Interpretation Neil Pearce Centre for Public Health Research Massey University, Wellington, New Zealand
Chapter 10 Interpretation Appraisal of a single study Appraisal of all of the available evidence
Interpretation of Evidence From Epidemiological Studies Populations do not randomize themselves by exposure status They do not always respond to requests to participate in epidemiological studies They may supply incomplete exposure information They cannot be asked about unknown risk factors It is not possible to do perfect studies, and we have to make decisions based on imperfect information
Summary of Study Design Issues Reduce random error by making the study as large as possible and through appropriate study design Minimize selection bias by having a good response rate (and selecting controls appropriately in a case- control study) Ensure that information bias is non-differential and keep it as small as possible Minimize confounding in the study design and control for it in the analysis
Appraisal of a Single Study: Random Error What is the magnitude and precision of the effect estimate? Are the study findings consistent with those of previous studies?
Cohort Studies of Shipyard Welding and Lung Cancer
Appraisal of a Single Study: Systematic Error What are the likely strengths and directions of possible biases?
Selection Bias Selection bias is any bias arising from the way that study participants are selected (or select themselves) from the source population If selection bias cannot be avoided or controlled, then it may still be possible to assess its likely strength and direction
Healthy Worker Effect in a Longitudinal Study of FEV 1 and Exposure to Granite Dust
Information Bias May occur when there is misclassification of exposure or disease If misclassification of exposure (or disease) is unrelated to disease (or exposure) then the misclassification is non-differential If misclassification of exposure (or disease) is related to disease (or exposure) then the misclassification is differential
Information Bias Is information bias likely to be differential or non-differential? If it is non-differential, then a positive findings unlikely to be explained by misclassification, but a negative finding may be a “false negative”
Confounding Occurs when the exposed and non-exposed groups in the source population are not comparable, because of inherent differences in background disease risk If there is the potential for uncontrolled confounding, then it is important to attempt to assess its likely strength and direction
Assessment of Possible Confounding by Smoking in a Study of Lung Cancer and Occupation
Appraisal of a Single Study The two most common criticisms of epidemiological studies are: the possibility of uncontrolled confounding; misclassification of exposure or disease (information bias) Uncontrolled confounding is often weaker than might be expected Non-differential information bias will usually produce false negative findings
Chapter 10 Interpretation Appraisal of a single study Appraisal of all of the available evidence
Appraisal of All of the Available Evidence: Criteria for Assessing Causality (Bradford-Hill) Criteria based on epidemiological evidence Temporality Specificity Consistency Strength of association Dose-response
Meta-Analysis: Benefits Meta-analysis may reduce the possibility of false negative results because of small numbers in specific studies It may enable the effect of exposure to be estimated with greater precision
Cohort Studies of Shipyard Welding and Lung Cancer
Meta-Analysis: Limitations Strikingly different results can be obtained depending on which studies are selected Meta-analysis reduces random error but does not necessarily reduce systematic error, and may even increase it Meta-analysis therefore involves the same issues as in a report on a single study, and both quantitative and narrative elements are required
Meta-Analysis: Assessment of Possible Biases An advantage of meta -analyses is that possible biases can be addressed using actual data rather than hypothetical examples For example, if smoking information is not available in all studies, the extent of confounding by smoking can be assessed in those studies in which smoking information is available Similarly, the possibility of information bias can be assessed by contrasting particular studies
Case-Control Studies of Phenoxy Herbicides and STS
Case-Control Studies of Phenoxy Herbicides and NHL
New Zealand Case-Control Study of Phenoxy Herbicides and NHL
Appraisal of All of the Available Evidence: Criteria for Assessing Causality (Bradford-Hill) Criteria based on comparing epidemiological evidence with evidence from other sources Plausibility Coherence
Biological Plausibility Many major epidemiological findings (e.g. on occupational carcinogens) were not biologically plausible at the time they were first discovered In many instances it has taken many years in the laboratory to ascertain the mechanism involved in established epidemiological findings Biological implausibility should not, by itself, be used to dismiss epidemiological findings
Interpretation of Evidence From Epidemiological Studies The most common criticisms of epidemiological findings are There may be uncontrolled confounding Information on exposure and/or disease is not perfect The findings lack biological plausibility
Interpretation of Evidence From Epidemiological Studies None of these considerations are sufficient in themselves to dismiss the findings of an epidemiological study Assessment of epidemiological findings should be based on all of the available evidence It is important to assess the likely strength and direction of possible biases
A short introduction to epidemiology Chapter 10: Interpretation Neil Pearce Centre for Public Health Research Massey University, Wellington, New Zealand