Building Multi-Marker Algorithms for Disease Prediction–-The Role of Correlations among Markers

Abstract

A widely held viewpoint in the field of predictive biomarkers for disease holds that no single marker can provide high enough discrimination and that a panel of markers, combined in some type of algorithm, will be needed. Motivated by a recent study where 27 additional markers for ovarian cancer, many of which had good predictive value alone, failed to substantially increase the predictive ability of the primary marker of CA125, we explore the effect of additional markers on the area under the ROC curve (AUC). We develop a statistical model based on the multivariate normal distribution and linear algorithms and use it to explore how the magnitude and direction of statistical correlation among the markers (in diseased and in non-diseased) is critical in determining the added predictive value of additional markers. We show mathematically and empirically that if the additional marker(s) is negatively correlated with the primary marker, then it will always be able to provide increased AUC when combined with the primary marker (as compared to that obtained with the primary marker alone), even if it has little predictive ability on its own. In contrast, if the additional marker(s) is positively correlated with the primary marker, then it is unlikely to substantially increase the AUC when combined with the primary marker, even when it has good predictive ability on its own. Thus, univariate analyses alone may not be the best approach in choosing which markers to combine in a predictive panel of markers; patterns of statistical correlation should be considered in ranking top-performing biomarkers.

A 4-gene panel as a marker at chromosome 8q in Asian gastric cancer patients

“…The type of associations between genes/proteins was illustrated as the networks legend indicated. markers, combined in some type of algorithm, will be necessary in order to produce the requisite level of predictive ability [10]. As was seen from Fig.…”

“…In addition, these biomarkers serve as critical tools in understanding carcinogenesis at a molecular level [17]. The fact that most cancers, including GC, are quite heterogeneous, is what has, in part, led many to the conclusion that linear combinations of multiple markers will be needed to identify a high percentage of the cases [10].…”

“…Though some serum tumor markers have been discovered, including carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA19-9) [7], and CA72-4 [8], the sensitivity of single biomarker in tumor diagnosis is low (usually less than 40%) with high "false-positives" [9]. A widely held viewpoint in the field of predictive biomarkers for disease is that no single marker can provide high enough discrimination between diseased cases and nondiseased and that the use of multiple markers, combined in some type of algorithm, will be necessary to produce the high specificity and sensitivity of predictive ability [10], such as a combination of miR-375, miR-424 and miR-92a with an accuracy of 94% in discriminating colorectal carcinomas from adenomas [11].…”

Serum proteins reflecting inflammation, injury and repair as biomarkers of disease activity in ANCA-associated vasculitis

“…Multiple markers appeared to distinguish between active AAV and remission better than did ESR or CRP, but, more importantly, provided complementary information to ESR and CRP based on low correlation coefficients. 38 Markers that distinguish between clinically obvious disease states are the most promising ones for further studying the more important and challenging goals of biomarker development. These goals have not been enumerated for vasculitis; we propose that they include staging of current internal organ involvement, distinguishing active vasculitis from infections or other conditions, distinguishing mildly active vasculitis from remission, predicting permanent and/or progressive damage to organ systems, predicting response to treatment or risk of relapse, and assessing efficacy of treatment early in its course.…”

ecancermedicalscience

“…Cross-tabulation analysis was carried out and the significance (χ 2 ) and likelihood ratio (LR) were calculated from the chosen cutoff values. The ROC analysis was first conducted on individual markers and then in combination to explore the potential that a panel of markers can provide improved performance [15]. …”