How to critically appraise a harm study to detect bias

18
Critically Appraising a Harm Study- Part 1 Terry Shaneyfelt, MD, MPH Associate Professor of Medicine UAB School of Medicine ebmteacher.com

description

Annotated slides that guide you through how to critically appraise a harm study (either cohort or case-control study) for various biases or design flaws.Please download these slides and view them in PowerPoint so you can view the annotations describing each slide.

Transcript of How to critically appraise a harm study to detect bias

Page 1: How to critically appraise a harm study to detect bias

Critically Appraising a Harm Study- Part 1

Terry Shaneyfelt, MD, MPH

Associate Professor of Medicine

UAB School of Medicine

ebmteacher.com

Page 2: How to critically appraise a harm study to detect bias

Appraising harm studies: 3 steps

1. Are the results valid?

2. What are the results?

3. Will they help me

care for my patients?

Page 3: How to critically appraise a harm study to detect bias

1. In a cohort study were there clearly defined groups of patients similar in all important ways other than the exposure of interest?

2. In a case-control study did the cases and controls have a similar chance of being exposed in the past to the factor of interest?

3. Were exposures and outcomes measured the same way in both groups?

4. Was follow-up sufficiently long and complete?

Are The Results Valid?

Page 4: How to critically appraise a harm study to detect bias
Page 5: How to critically appraise a harm study to detect bias

• Patients in each cohort should be similar for prognostic factors that are known to be assocwith the outcome

• Look at Table 1

• If imbalances are seen….

1. Are they important?

2. Are they dealt with in some way?

3. What else is missing?

1) In a cohort study were the groups similar at the start of the study except for exposure?

Page 7: How to critically appraise a harm study to detect bias

• 3 questions:

1. Are the differences important?– Yes!

2. Were they adjusted for?– Yes

3. Residual confounding? – Likely

1) Were the NHS cohorts

similar?

Page 8: How to critically appraise a harm study to detect bias

• Both cases and control should have equal opportunity of being exposed

• Controls:• Should meet same inclusion/exclusion criteria

• Should be at risk for developing the outcome (could become a case)

• What about this study?– Is HRT associated with uterine cancer?

– Cases: women with uterine cancer

– Controls: men without uterine cancer

In a case-control study did the cases and controls have a similar chance of being exposed in the

past to the factor of interest?

Page 9: How to critically appraise a harm study to detect bias

Please watch Part 2 of Critically Appraising a Therapy Study

Page 10: How to critically appraise a harm study to detect bias

Critically Appraising a Harm Study- Part 2

Terry Shaneyfelt, MD, MPH

Associate Professor of Medicine

UAB School of Medicine

ebmteacher.com

Page 11: How to critically appraise a harm study to detect bias

1. In a cohort study were there clearly defined groups of patients similar in all important ways other than the exposure of interest?

2. In a case-control study did the cases and controls have a similar chance of being exposed in the past to the factor of interest?

3. Were exposures and outcomes measured the same way in both groups?

4. Was follow-up sufficiently long and complete?

Are The Results Valid?

Page 12: How to critically appraise a harm study to detect bias
Page 13: How to critically appraise a harm study to detect bias

• Investigators report a 3-fold increase in risk of melanoma in individuals working with radioactive materials

– What if physicians, concerned about a possible risk, searched more diligently in those exposed to radiation than those who weren't?

• Researchers should:

– Be blinded to exposure status (cohort studies) and to outcome status and hypothesis (case-control studies)

– Use same methods in both groups (standardized)

3) Were exposures and outcomes measured the same way in both groups?

Page 14: How to critically appraise a harm study to detect bias

Sources of Exposure Information

• Pre-existing Records• Inexpensive, available,

unbiased

• May be incomplete

• Interviews & Questionnaires (Patients, MDs, Family)

• Complete data

• Bias (e.g. recall bias)

• Proxy Measures (e.g. Job title, Proximity)

• Inaccurate

• Direct Measurement

• Current, future

Page 15: How to critically appraise a harm study to detect bias

3) Were exposures and outcomes

measured the same way in both groups?

Page 16: How to critically appraise a harm study to detect bias

• Follow-up needs to be long enough for participants to develop the outcome(s) of interest

• Ideally no patients are lost to follow-up

• Loss to f/u can introduce bias if – Those who are lost have different outcomes

– Differences in f/u between exposed and unexposed cohorts

4) Was follow-up sufficiently long and complete?

Page 17: How to critically appraise a harm study to detect bias

4) Was follow-up sufficiently long and complete?

Page 18: How to critically appraise a harm study to detect bias

• Moderate risk for bias (residual confounding)- in hindsight!

• Options:

– Read the results and realize the findings might not be accurate

– Find another study

Assessment of the NHS