Table 4.

Diagnostic accuracy of targeted plasma metabolomics in CFS

SexClassifiersAUROC*95% CIrdCV accuracyPermutation P value2 × 2§ accuracy2 × 2 sensitivity2 × 2 specificity
Males8-analyte example0.940.84–1.00.840.0010.900.910.89
1-analyte example#0.710.50–0.880.620.0090.720.730.72
Females13-analyte example||0.960.87–1.00.900.0010.930.910.95
1-analyte example#0.680.42–0.860.580.0090.680.700.67
  • n = 18 control males and 22 CFS males, and n = 21 control females and 23 CFS females.

  • * AUROC, area under the receiver operator curve reflects the overall accuracy of diagnosis using these analytes.

  • rdCV, repeated random subsample (2/3 in, 1/3 out) double cross-validation.

  • Permutation P values represent the probability that the RF classification of cases and controls using the specified analytes could be obtained by chance.

  • § Values calculated by standard 2 × 2 contingency table analysis.

  • 8:analytes in males, phosphatidyl choline PC(16:0/16:0), glucosylceramide GC(18:1/16:0), 1-P5C, FAD, pyroglutamic acid (also known as 5-oxoproline), HICA, l-serine, and lathosterol.

  • # 1-analyte: phosphatidyl choline PC(16:0/16:0).

  • || 13-analytes in females: THC(18:1/24:0), phosphatidyl choline PC(16:0/16:0), hydroxyproline, ceramide(d18:1/22:2), lathosterol, adenosine, phosphatidylinositol PI(16:0/16:0), FAD, 2-octenoylcarnitine, phosphatidyl choline plasmalogen PC(22:6/P18:0), phosphatidyl choline PC(18:1/22:6), 1-P5C, and CDCA.