Research Article

Noninferiority and Equivalence Evaluation of Clinical Performance among Computed Radiography, Film, and Digitized Film for Telemammography Services

Table 6

Equivalence tests for accuracy, PPV, sensitivity, specificity, and AUC by paired devices ().

VariableCompared devices Mean difference () SE⁢(1-2α)% confidence interval for equivalence testing
() device() deviceLowerUpper

AccuracyLUMIXICR 0.0100.0142 −0.0130.034−6.31<0.001Ha
LUMIXCR −0.0140.0133 −0.0360.008−6.46<0.001Ha
LUMIXFILM −0.0470.0152 −0.072−0.022−3.49<0.001Ha
ICRCR −0.0240.0138 −0.047−0.002−5.46<0.001Ha
ICRFILM −0.0570.0162 −0.084−0.031−2.650.004Ha
FILMCR 0.0330.01640.0060.060−4.11<0.001Ha

PPVLUMIXICR −0.0030.0052 −0.0110.006−18.69<0.001Ha
LUMIXCR −0.0070.0049 −0.0150.001−18.79<0.001Ha
LUMIXFILM −0.0090.0050 −0.017−0.001−18.28<0.001Ha
ICRCR −0.0050.0047 −0.0120.003−20.32<0.001Ha
ICRFILM −0.0070.0040 −0.0130.000−23.08<0.001Ha
FILMCR 0.0020.0027 −0.0020.007−35.76<0.001Ha

SensitivityLUMIXICR 0.0350.02030.0010.068−3.21<0.001Ha
LUMIXCR −0.0170.0219 −0.0530.019−3.78<0.001Ha
LUMIXFILM −0.0650.0246−0.105 −0.025−1.430.08Ho
ICRCR −0.0520.0210 −0.087 −0.017−2.280.01Ha
ICRFILM−0.1000.0256−0.142 −0.057 −0.020.49Ho
FILMCR 0.0480.0316 −0.0040.100−1.660.05Ha

SpecificityLUMIXICR −0.0120.0192 −0.0430.020−4.60<0.001Ha
LUMIXCR −0.0120.0158 −0.0380.014−5.59<0.001Ha
LUMIXFILM −0.0310.0182 −0.061 −0.001−3.79<0.001Ha
ICRCR 0.0000.0173 −0.0280.028−5.79<0.001Ha
ICRFILM −0.0190.0182 −0.0490.011−4.42<0.001Ha
FILMCR 0.0190.0124 −0.0010.040−6.50<0.001Ha

AUCLUMIXICR 0.0090.0646 −0.0980.115−1.4120.08Ho
LUMIXCR −0.0160.0350 −0.0730.042−2.4080.008Ha
LUMIXFILM −0.0260.0412 −0.0930.042−1.8070.04Ha
ICRCR −0.0240.0660−0.1330.084−1.1450.13Ho
ICRFILM −0.0340.0508−0.1180.049−1.2920.10Ho
FILMCR 0.0100.0463 −0.0660.086−1.9450.03Ha

Each comparison was calculated from 980 observations (70 cases, seven radiologists, and two devices). PPV: positive predictive value, ROC: receiver operating characteristic, AUC: area under ROC curve, SE: standard error of the mean, Ho: null hypothesis, Ha: alternative hypothesis for testing equivalence, α: significance of the test (0.05), δ: difference of the means allowed to achieve equivalence, : test for difference of compared devices, that is, = (Difference ()δ)/SE, : retained hypothesis equivalence at δ level (“Ha” indicates equivalence achieved and “Ho” indicates failing to reject the null hypothes).
Ho: difference ()δ = 0.
Ha: difference ()δ < 0.