Uncertainty in baseline risk can reduce confidence in estimates of effect

Abstract text
Background: Because relative effects are more consistent across risk groups than absolute effects, systematic reviews focus on relative effects. Clinical decision-making, however, requires estimates of absolute effect, typically from applying estimates of relative risk to baseline risk (control group risk). The Cochrane approach to rating confidence in estimates of effect in Summary of Findings (SoF) tables has focused on uncertainty in relative effects.

Objectives:To demonstrate how limitations in evidence regarding baseline risk decrease confidence in estimates of absolute effect.

Methods:We searched the 9th iteration of the American College of Chest Physicians Antithrombotic Guidelines for recommendations in which limitations in evidence regarding baseline risk decreased confidence in estimated absolute treatment effects.

Results:We found examples in four categories of reasons for rating down confidence in estimates.

Risk of bias and indirectness: In estimating baseline risk of stroke in patients with atrial fibrillation, authors noted that data from a large administrative database suggested lower rates of non-fatal thromboembolism in untreated patients than did results from RCTs. They chose to use estimates from the RCTs because of concerns that the observational evidence systematically underestimated stroke risk. Nevertheless, concerns about unrepresentativeness of the RCT populations remain.

Imprecision: A guideline from AT9 addressing thromboprophylaxis for women undergoing assisted reproduction who develop ovarian hyperstimulation syndrome found that of 49 patients with the condition, two developed deep venous thrombosis: the 95% confidence interval around the 4.1% point estimate has limits of 1.1% to 13.7% resulting in estimates of 4 to 110 events prevented per 1000 treated.

Inconsistency: Authors addressing thromboprophylaxis in surgical patients estimated a risk of 2.1% in patients undergoing craniotomy from 8 studies that showed rates from 0% to 10%.

Conclusions:Systematic reviews should consider uncertainty in baseline risk as a reason for rating down confidence in estimates of absolute effect in SoFs.

Guyatt G1, Spencer F1, Iorio A1, Vandvik P2, You J1
1 McMaster University, Canada
2 Knowledge Centre for the Health Services, Norway
Presenting author and contact person
Presenting author: 
Gordon Guyatt
Contact person Affiliation Country
Gordon Guyatt (Contact this person) McMaster University, Faculty of Health Sciences Canada
Date and Location
Oral session C14O1
Wednesday 3 October 2012 - 11:00 - 11:20