One statistical test is sufficient for assessing new predictive markers
2011

Evaluating New Predictive Markers in Medicine

Sample size: 500 publication 10 minutes Evidence: high

Author Information

Author(s): Andrew J Vickers, Angel M Cronin, Colin B Begg

Primary Institution: Memorial Sloan-Kettering Cancer Center

Hypothesis

Does the addition of a new predictor improve the predictive accuracy of existing clinical markers?

Conclusion

Using both likelihood ratio and Wald tests is more effective for assessing new predictors than comparing areas under the ROC curve.

Supporting Evidence

  • The likelihood ratio and Wald tests showed superior power compared to AUC tests.
  • The study demonstrated that using both ROC and regression tests can lead to inconsistent conclusions.
  • The area test was found to be overly conservative in detecting incremental predictive accuracy.

Takeaway

When doctors want to know if a new test is better than old tests, they should use specific methods to check if it really helps, instead of just comparing scores.

Methodology

A simulation study was conducted using logistic regression to compare the effectiveness of new predictors against established ones.

Potential Biases

Potential bias due to the use of patient-specific predictors in ROC analysis.

Limitations

The study primarily focused on the statistical properties of tests without considering clinical applicability.

Participant Demographics

Not specified.

Statistical Information

P-Value

0.03

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/1471-2288-11-13

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication