Marker research
Methodological work
The Stratos initiative considers evaluation of prediction models in Topic Group 6.
Overview papers
A framework for traditional and modern performance measures, Epidemiology 2010
Incremental value in testicular cancer case study, Eur J Clin Invest 2012
Incremental value in cardiovascular disease (Framingham data), Rev Esp Cardiol 2011
Incremental value in cardiovascular disease (Framingham data), in Dutch, Ned Tijdschr Geneeskd 2012
NRI
With Michael Pencina:
LTTE on calculation of NRI for survival outcomes, Ann Intern Med. 2010
Extensions of NRI (prospective formulation for survival and competing risk data, weighting with costs), Stat Med 2011
Relations between NRI and 3 utility-based measures (ΔNB), ΔRU), wNRI) with ovarian cancer case study, Med Decis Making 2013
With Stuart Baker:
Theory and estimation of the test tradeoff, Int J Biostat 2012
With Kirsten Mühlenbruch:
Impact of risk cut-offs and number of risk categories, Eur J Epidemiol 2013
With Maarten Leening and others:
Review of reporting issues for NRI in top medical journals
Increase in c
With Peter Austin:
C in relation to variance and odds ratio of a continuous predictor, BMC Med Res Methodol. 2012
Increase in c by prevalence and odds ratio of the marker, c of the reference model, Stat Med 2013
In progress: increase in c for survival outcomes
Early HTA
Early health technology assessment of novel biomarkers (Postmus, Stat Med 2012)
Landmark paper on HTA perspective for markers: (Hlatky et al, Circulation 2009)
Perspectives on NRI by others
Relation between increase in c and NRI (Mihaescu, AJE 2010)
Relation between effect size and increase in c, NRI and IDI (Pencina, Stat Med 2012)
More examples of effect size and increase in c, NRI and IDI (Pencina, AJE 2012)
IDI and NRI are not to be trusted with miscalibrated models (Hilden, Stat Med 2013)
NRI may not be helpful in evaluating prediction models (Vickers&Pepe, Ann Int Med 2014)
Applications
NRI of transient elastography, Liver Int 2013
Coronary heart disease risk prediction, Eur J Prev Cardiol 2012
Coronary artery calcification (CAC) and heart failure risk, JACC Cardiovasc Imaging 2012
Small vessel disease on brain MRI and stroke risk, JNNP 2012
Prediction of presence of coronary artery disease, BMJ 2012
Arguments have been put forward that applications should move away from NRI to differences in decision-analytic performance measures
In progress: measures for improvement
Following on [[http://www.ncbi.nlm.nih.gov/pubmed/23439103| (Decision making between coronary artery bypass surgery and percutaneous coronary intervention for individual patients: SYNTAX score II, Lancet 2013), research is ongoing on quantifying improvement in performance.
One proposal was to consider a net reclassification risk graph; illustrated with Framingham data and other examples .
Another idea is to consider cost-effectiveness, i.e. individualized risk thresholds depending on long term mortality risks.
