Dataset Information

Single index methods for evaluation of marker-guided treatment rules based on multivariate marker panels.

ABSTRACT: Clinical practice may be enhanced by use of person-level information that could guide treatment choice and lead to better outcomes for both treated individuals and for the population. The scientific challenge is to identify and validate those factors that can reliably be used to target treatment, and to accurately quantify the expected treatment benefit as a function of candidate markers. Our proposal is to explicitly focus on smooth non-parametric evaluation of a canonical single index score that estimates the expected treatment benefit associated with patient characteristics. Our methods intentionally decouple the model used to generate the treatment benefit score from the methods that are adopted to evaluate the performance of the resulting single index score. We are motivated by the practical issue that model performance can not realistically be evaluated for every specific covariate value due to intrinsic sparseness. However, direct validation of a scalar treatment benefit score obtained through model-based dimension reduction is feasible, and we believe should be the focus of validation efforts. We also show that the canonical single index treatment benefit score can be used for selecting subsets of patients with enriched expected treatment response since patients can be easily ordered and grouped based on the scalar score. Our biomedical motivation comes from a recent randomized trial of steroid injections for low back pain where baseline clinical and imaging data are candidate measures for guiding therapeutic choice.

SUBMITTER: Skrivankova V

PROVIDER: S-EPMC5803489 | biostudies-literature | 2018 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Single index methods for evaluation of marker-guided treatment rules based on multivariate marker panels.

Skrivankova Veronika V Heagerty Patrick J PJ

Biometrics 20170807 2

Clinical practice may be enhanced by use of person-level information that could guide treatment choice and lead to better outcomes for both treated individuals and for the population. The scientific challenge is to identify and validate those factors that can reliably be used to target treatment, and to accurately quantify the expected treatment benefit as a function of candidate markers. Our proposal is to explicitly focus on smooth non-parametric evaluation of a canonical single index score th ...[more]

PMID: 28783868

Similar Datasets

Project description:BACKGROUND: Most women with a clinical presentation consistent with ovarian cancer have benign conditions. Therefore methods to distinguish women with ovarian cancer from those with benign conditions would be beneficial. We describe the development and preliminary evaluation of a serum-based multivariate assay for ovarian cancer. This hypothesis-driven study examined whether an informative pattern could be detected in stage I disease that persists through later stages. METHODOLOGY/PRINCIPAL FINDINGS: Sera, collected under uniform protocols from multiple institutions, representing 176 cases and 187 controls from women presenting for surgery were examined using high-throughput, multiplexed immunoassays. All stages and common subtypes of epithelial ovarian cancer, and the most common benign ovarian conditions were represented. A panel of 104 antigens, 44 autoimmune and 56 infectious disease markers were assayed and informative combinations identified. Using a training set of 91 stage I data sets, representing 61 individual samples, and an equivalent number of controls, an 11-analyte profile, composed of CA-125, CA 19-9, EGF-R, C-reactive protein, myoglobin, apolipoprotein A1, apolipoprotein CIII, MIP-1alpha, IL-6, IL-18 and tenascin C was identified and appears informative for all stages and common subtypes of ovarian cancer. Using a testing set of 245 samples, approximately twice the size of the model building set, the classifier had 91.3% sensitivity and 88.5% specificity. While these preliminary results are promising, further refinement and extensive validation of the classifier in a clinical trial is necessary to determine if the test has clinical value. CONCLUSIONS/SIGNIFICANCE: We describe a blood-based assay using 11 analytes that can distinguish women with ovarian cancer from those with benign conditions. Preliminary evaluation of the classifier suggests it has the potential to offer approximately 90% sensitivity and 90% specificity. While promising, the performance needs to be assessed in a blinded clinical validation study.

Dataset Information

Single index methods for evaluation of marker-guided treatment rules based on multivariate marker panels.

Publications

Single index methods for evaluation of marker-guided treatment rules based on multivariate marker panels.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets