Table E1. ROC Analysis of the ADC Values for Discriminating between Lymphoma and SCC, SCC and NPC , and Lymphoma and NPC with the Whole-Node and Single-Section Techniques Using a Cutoff Point for Optimizing Sensitivity and Specificity with Equal Weighting

Comparison and Technique	AUC (95% CI)	Optimal Threshold*	Sensitivity (95% CI)	Specificity (95% CI)	PPV	NPV
Lymphoma vs SCC
Whole node	0.98 (0.93�1.00)	0.893 ^a	0.94 (0.73�1.00)	1.00 (0.63�1.00)	1.00 (0.80�1.00)	0.89 (0.52�1.00)
Single-section node	0.99 (0.97�1.00)	0.824 ^a	0.94 (0.73�1.00)	1.00 (0.63�1.00)	1.00 (0.80�1.00)	0.89 (0.52�1.00)
Whole node	0.98 (0.93�1.00)	0.893 ^b	1.00 (0.63�1.00)	0.94 (0.73�1.00)	0.89 (0.52�1.00)	1.00 (0.80�1.00)
Single-section node	0.99 (0.97�1.00)	0.824 ^b	1.00 (0.63�1.00)	0.94 (0.73�1.00)	0.89 (0.52�1.00)	1.00 (0.80�1.00)
SCC vs NPC
Whole node	0.94 (0.86�1.00)	0.945 ^c	0.89 (0.65�0.99)	0.88 (0.64�0.99)	0.89 (0.65�0.99)	0.88 (0.64�0.99)
Single-section node	0.88 (0.77�0.99)	0.894 ^c	0.83 (0.59�0.96)	0.76 (0.50�0.93)	0.79 (0.54�0.94)	0.81 (0.54�0.96)
Whole node	0.94 (0.86�1.00)	0.945 ^d	0.88 (0.64�0.99)	0.89 (0.65�0.99)	0.88 (0.64�0.99)	0.89 (0.65�0.99)
Single-section node	0.88 (0.77�0.99)	0.894 ^d	0.76 (0.50�0.93)	0.83 (0.59�0.96)	0.81 (0.54�0.96)	0.79 (0.54�0.94)
Lymphoma vs NPC
Whole node	0.74 (0.54�0.95)	0.841 ^e	0.53 (0.28�0.77)	0.88 (0.47�1.00)	0.90 (0.56�1.00)	0.47 (0.21�0.73)
Single-section node	0.81 (0.63�0.99)	0.694 ^e	0.76 (0.50�0.93)	0.75 (0.35�0.97)	0.87 (0.60�0.98)	0.60 (0.26�0.88)
Whole node	0.74 (0.54�0.95)	0.841 ^f	0.88 (0.47�1.00)	0.53 (0.28�0.77)	0.47 (0.21�0.73)	0.90 (0.56�1.00)
Single-section node	0.81 (0.63�0.99)	0.694 ^f	0.75 (0.35�0.97)	0.76 (0.50�0.93)	0.60 (0.26�0.88)	0.87 (0.60�0.98)

Note.� AUC = area under curve, CI = confidence interval, NPV = negative predictive value, PPV = positive predictive value.

* Optimal in the sense of maximizing sensitivity and specificity with equal weighting.

^a Values larger than the threshold indicate stronger evidence for SCC.

^b Values smaller than the threshold indicate stronger evidence for lymphoma.

^c Values larger than the threshold indicate stronger evidence for SCC.

^d Values smaller than the threshold indicate stronger evidence for NPC.

^e Values larger than the threshold indicate stronger evidence for NPC.

^f Values smaller than the threshold indicate stronger evidence for being lymphoma.

Table E2. ROC Analysis of the ADC Values for Discriminating between Lymphoma and SCC, SCC and NPC, and Lymphoma and NPC with the Single-Section Technique Using a Cutoff Point Chosen to Maximize Sensitivity at a Specificity of 100%

Comparison	AUC (95% CI)	Highest Specificity Cutoff*	Sensitivity (95% CI)	Specificity (95% CI)	PPV	NPV
Lymphoma vs SCC
Predicting SCC	0.99 (0.97�1.00)	0.824 ^a	0.94 (0.73�1.00)	1.00 (0.63�1.00)	1.00 (0.80�1.00)	0.89 (0.52�1.00)
Predicting lymphoma	0.99 (0.97�1.00)	0.767 ^b	0.88 (0.47�1.00)	1.00 (0.81�1.00)	1.00 (0.59�1.00)	0.95 (0.74�1.00)
SCC vs NPC
Predicting SCC	0.88 (0.77�0.99)	1.093 ^a	0.39 (0.17�0.64)	1.00 (0.80�1.00)	1.00 (0.59�1.00)	0.61 (0.41�0.79)
Predicting NPC	0.88 (0.77�0.99)	0.764 ^c	0.47 (0.23�0.72)	1.00 (0.81�1.00)	1.00 (0.63�1.00)	0.74 (0.46�0.83)
Lymphoma vs NPC
Predicting NPC	0.81 (0.63�0.99)	0.788 ^d	0.53 (0.28�0.77)	1.00 (0.63�1.00)	1.00 (0.67�1.00)	0.50 (0.25�0.75)
Predicting lymphoma	0.81 (0.63�0.99)	No suitable cutoff	�	�	�	�

Note.� AUC = area under curve, CI = confidence interval, NPV = negative predictive value, PPV = positive predictive value.

*The cutoff points were chosen to produce 100% specificity and the highest possible sensitivity.

^a Values larger than the threshold indicate stronger evidence for SCC.

^b Values smaller than the threshold indicate stronger evidence for lymphoma.

^c Values smaller than the threshold indicate stronger evidence for NPC.

^d Values larger than the threshold indicate stronger evidence for NPC.

Malignant Cervical Lymphadenopathy: Diagnostic Accuracy of Diffusion-weighted MR Imaging