![[ Visit Client Website ]](images/banner.gif)
PURPOSE:
Outcomes in RA can be improved by application of the principal of tight control. Tight control entails rapid escalation of therapy to achieve maximal suppression of disease activity. The process is guided by frequent disease activity measurements. Current disease activity monitoring tools are based on clinical assessments, which can be subjective, and laboratory tests with poor predictive power. We are developing an objective quantitative laboratory test for disease activity using an algorithmic combination of blood-based biomarkers.
METHODS:
Clinical and laboratory measures were assessed for 424 RA patients collected cross-sectionally. Genome-scale expression profiling of whole blood RNA was assessed on Illumina microarrays. Univariate associations to DAS28-CRP3 were assessed through a linear model. Significant univariate terms identified were carried into a multi-stage multivariate screening process. Step 1 - All significant terms were randomly pooled into groups of ~100 markers and subjected to a forward stepwise multiple linear regression (FS-MLR) with a p to enter of 0.05. Step 2 - Terms significant in stage 1 were combined into a new FS-MLR. Step 3 - The model was bootstrap cross-validated with out-of-bag samples. Step 4 - The cross-validated model was tested on additional naïve samples not used to derive the model.
RESULTS:
ESR and CRP have low correlation to DAS28-CRP3 in this cross sectional cohort (R = 0.19 and 0.25 respectively). We identified 1,144 genes correlated with DAS28 (10% False Discovery Rate) in a set of 90 RA samples. The most parsimonious multivariate algorithm correlated to DAS28 included 9 markers (R = 0.80). The algorithms predictive power was validated in an independent set of 24 patient samples (R = 0.64). Correlation to DAS28 was markedly higher than ESR and CRP.
CONCLUSIONS:
Current laboratory tests of RA disease activity, ESR and CRP, are poorly associated with disease activity. Multivariate modeling identified a set of gene expression markers with relatively high correlation to disease activity. This marker set validated in an independent cohort with higher predictive value than ESR and CRP. These results provide a proof of concept. An ideal disease activity test would correlate highly with disease activity changes over time, quantify therapy response, and provide an accurate and subclinical measure of disease remission. Additional studies are ongoing in larger patient cohorts followed prospectively to improve the potential of this approach.
Outcomes in RA can be improved by application of the principal of tight control. Tight control entails rapid escalation of therapy to achieve maximal suppression of disease activity. The process is guided by frequent disease activity measurements. Current disease activity monitoring tools are based on clinical assessments, which can be subjective, and laboratory tests with poor predictive power. We are developing an objective quantitative laboratory test for disease activity using an algorithmic combination of blood-based biomarkers.
METHODS:
Clinical and laboratory measures were assessed for 424 RA patients collected cross-sectionally. Genome-scale expression profiling of whole blood RNA was assessed on Illumina microarrays. Univariate associations to DAS28-CRP3 were assessed through a linear model. Significant univariate terms identified were carried into a multi-stage multivariate screening process. Step 1 - All significant terms were randomly pooled into groups of ~100 markers and subjected to a forward stepwise multiple linear regression (FS-MLR) with a p to enter of 0.05. Step 2 - Terms significant in stage 1 were combined into a new FS-MLR. Step 3 - The model was bootstrap cross-validated with out-of-bag samples. Step 4 - The cross-validated model was tested on additional naïve samples not used to derive the model.
RESULTS:
ESR and CRP have low correlation to DAS28-CRP3 in this cross sectional cohort (R = 0.19 and 0.25 respectively). We identified 1,144 genes correlated with DAS28 (10% False Discovery Rate) in a set of 90 RA samples. The most parsimonious multivariate algorithm correlated to DAS28 included 9 markers (R = 0.80). The algorithms predictive power was validated in an independent set of 24 patient samples (R = 0.64). Correlation to DAS28 was markedly higher than ESR and CRP.
CONCLUSIONS:
Current laboratory tests of RA disease activity, ESR and CRP, are poorly associated with disease activity. Multivariate modeling identified a set of gene expression markers with relatively high correlation to disease activity. This marker set validated in an independent cohort with higher predictive value than ESR and CRP. These results provide a proof of concept. An ideal disease activity test would correlate highly with disease activity changes over time, quantify therapy response, and provide an accurate and subclinical measure of disease remission. Additional studies are ongoing in larger patient cohorts followed prospectively to improve the potential of this approach.
M. Turner, Riley Genomics, 5; N. Knowlton, Riley Genomics, 5; M.B. Frank, Riley Genomics, 5; C. Carson, None; A. Kumar, None; E. Arthur, None; R. Hynd, None; L. Willis, None; K. Welk, None; J. Osban, None; M. Darragh, None; L. Gonzales, None; T. Lockwood, None; T. Woodward, None; J. Ogar, None; P. Floyd, Riley Genomics Inc., 3; A. Martinez, Riley Genomics Inc., 3; C. Ligon, Riley Genomics Inc., 3; P. Smith, None; C. Sutton, Riley Genomics Inc., 3; P. Pathipvanich, None; Y. Jiang, None; M. Nguyen, None; B. Sliger, Riley Genomics Inc., 3; M. Centola, Riley Genomics, 1; Riley Genomics, 2; Riley Genomics, 5.