Table 2.

The diagnostic efficiency of models in differentiating NSCLC cases and controls

Training setTest set
AUC (95% CI)SN (%)SP (%)PPV (%)NPV (%)Positive LRNegative LRAUC (95% CI)SN (%)SP (%)PPV (%)NPV (%)Positive LRNegative LR
ADC vs. control
 IDH1a0.858 (0.823–0.889)56.095.091.469.511.190.460.810 (0.771–0.844)56.889.183.268.55.220.48
 CEA0.600 (0.553–0.644)38.882.968.258.92.270.740.568 (0.521–0.631)36.682.065.957.62.030.77
 CA1250.704 (0.661–0.745)14.199.294.155.016.920.870.679 (0.635–0.722)14.599.697.155.134.740.86
 Cyfra21-10.648 (0.603–0.691)30.887.570.057.22.470.790.556 (0.509–0.601)28.283.361.555.01.680.86
 IDH1 + CEA + CA125 + Cyfra21-10.890 (0.858–0.917)63.495.092.373.312.690.380.838 (0.802–0.870)57.790.084.569.15.750.47
SCC vs. control
 IDH1a0.778 (0.739–0.815)42.095.089.661.68.410.610.740 (0.698–0.778)43.489.180.360.73.990.63
 CEA0.534 (0.488–0.579)15.182.947.448.90.881.020.519 (0.474–0.564)16.082.047.648.90.891.02
 CA1250.747 (0.706–0.785)12.299.293.752.514.690.880.752 (0.711–0.790)13.999.697.153.133.300.86
 Cyfra21-10.846 (0.811–0.877)64.987.584.170.95.190.400.833 (0.797–0.865)66.482.979.870.73.870.41
 IDH1 + CA125 + Cyfra21-10.914 (0.886–0.938)63.395.092.871.712.650.390.893 (0.862–0.919)66.492.590.072.98.820.36
NSCLC vs. control
 IDH1a0.817 (0.786–0.845)48.795.095.048.59.750.540.773 (0.741–0.804)49.989.190.047.44.590.56
 CEA0.565 (0.528–0.602)26.582.975.336.41.550.890.543 (0.505–0.580)25.982.073.936.01.440.90
 CA1250.726 (0.692–0.759)13.499.296.936.816.020.870.717 (0.682–0.750)14.299.698.537.134.000.86
 Cyfra21-10.751 (0.717–0.782)48.587.588.446.43.880.590.699 (0.664–0.733)48.082.984.644.72.800.63
 IDH1 + CA125 + Cyfra21-10.896 (0.871–0.918)60.495.096.054.912.080.420.853 (0.825–0.879)58.491.693.252.86.980.45
  • Abbreviations: LR, likelihood ratio; NPV, negative predictive value; PPV, positive predictive value.

  • aThe diagnostic cut-off value was 2.19 U/L.