Predicting left ventricular systolic dysfunctionBMJ 2000; 321 doi: https://doi.org/10.1136/bmj.321.7265.895 (Published 07 October 2000) Cite this as: BMJ 2000;321:895
Combining test results gives best predictive effect
- Frank Dobbs (firstname.lastname@example.org), senior lecturer
- Department of Primary Health Care and General Practice, University of Plymouth, Plymouth PL6 8BX
- Ninewells Hospital and Medical School, Dundee DD1 9SY
- University of Birmingham, Birmingham B15 2TH
- Hightown Surgery, Banbury OX16 9DB
- Horton Hospital, Oxford Radcliffe Hospitals NHS Trust, Banbury OX16 9AL
EDITOR—Landray et al chose a cut-off level for brain natriuretic peptide of 17.9 pg/ml.1 They said that other levels (10 or 76 pg/ml) did not improve the predictive characteristics. However, this level correctly predicted left ventricular systolic dysfuntion in only 64 of 126 patients. A level of 76 pg/ml correctly predicted 111 cases out of 126 (calculated from the sensitivity and specificity figures) and is therefore clearly superior. This proportion is similar to that found by Smith et al for a cut-off level of 26.7 pmol/l.2 Both these levels are about two to three times the median level of the normal population.
The other screening tests assessed by Landray et al are likely to be independent predictors, as they are proxies for different aspects of heart failure and have been shown by Nielsen et al, using multiple logistic regression, to be statistically independent.3 It would therefore be possible to combine the predictive effect of these tests using bayesian analysis. The B-score method 4 5 can …
Log in using your username and password
Log in through your institution
Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
Sign up for a free trial