Table 4 The capacity to predict 2 and 6 months treatment outcomes by 11 biomarkers (BM) which changed consistently on ATT was analyzed applying logistic regression, leave-10-out cross-validation while adjusting for micronutrient supplementation.

From: BLR1 and FCGR1A transcripts in peripheral blood associate with the extent of intrathoracic tuberculosis in children and predict treatment outcome

BM entered in the model

 

Prediction by BM measured at baseline

Prediction by BM measured after 2 months

Prediction by the change in BM from baseline to 2 months

Outcome at 2 months

Outcome at 6 months

Outcome at 6 months

Outcome at 6 months

Int. est. a

Cross-validation

Int. est. a

Cross-validation

Int. est. a

Cross-validation

Int. est. a

Cross-validation

single biomarkers

BCL2

0,614

0,568

0,762

0,75

0,651

0,587

0,759

0,648

BLR1

0,636

0,614

0,738

0,702

0,619

0,587

0,704

0,63

CASP8

0,614

0,58

0,75

0,762

0,635

0,619

0,704

0,685

CD3E

0,636

0,568

0,726

0,714

0,651

0,587

0,741

0,667

CD4

0,602

0,545

0,714

0,679

0,603

0,556

0,685

0,63

CD19

0,625

0,557

0,762

0,714

0,651

0,587

0,667

0,667

FCGR1A

0,625

0,591

0,738

0,702

0,651

0,635

0,685

0,537

FPR1

0,625

0,58

0,75

0,714

0,651

0,603

0,722

0,667

IL7R

0,625

0,557

0,75

0,75

0,651

0,587

0,741

0,667

TGFBR2

0,614

0,591

0,75

0,75

0,651

0,571

0,704

0,574

MMP9

0,58

0,557

0,738

0,702

0,698

0,683

0,778

0,759

signatures

FCGR1A & BLR1

0,648

0,568

0,738

0,714

0,667

0,603

0,685

0,537

11 markers *

0,727

0,523

0,762

0,679

0,73

0,619

0,759

0,593

10 markers **

0,705

0,523

0,786

0,643

0,746

0,651

0,759

0,593

  1. a Internal estimate of accuracy. * Cross-validation of the combination of all 11 biomarkers that consistently changed with treatment (BCL2, BLR1, CASP8, CD19, CD3E, CD4, FCGR1A, FPR1, IL7R, MMP9, and TGFBR2). ** Cross-validation of the combination of the same markers as above, but excluding CD4.
  2. Biomarkers measured at the start of ATT (baseline), at 2 months, and the corresponding change in score were assessed. Prediction accuracy in the cross-validation estimates are bolded.