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

Abstract

Objective. The aim of this study was to evaluate and compare the clinical performance of different alternatives to implement low-cost screening telemammography. We compared computed radiography, film printed images, and digitized films produced with a specialized film digitizer and a digital camera. Material and Methods. The ethics committee of our institution approved this study. We assessed the equivalence of the clinical performance of observers for cancer detection. The factorial design included 70 screening patients, four technological alternatives, and cases interpreted by seven radiologists, for a total of 1,960 observations. The variables evaluated were the positive predictive value (PPV), accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curves (AUC). Result. The mean values for the observed variables were as follows: accuracy ranged from 0.77 to 0.82, the PPV ranged from 0.67 to 0.68, sensitivity ranged from 0.64 to 0.74, specificity ranged from 0.87 to 0.90, and the AUC ranged from 0.87 to 0.90. At a difference of 0.1 to claim equivalence, all alternatives were equivalent for all variables. Conclusion. Our findings suggest that telemammography screening programs may be provided to underserved populations at a low cost, using a film digitizer or a digital camera.

Breast cancer early detection: A phased approach to implementation

“…Task‐sharing and task‐shifting approaches have been used successfully in different settings to surmount the workforce shortage for breast cancer early detection, including training and centralized services. Additional support can be supplied through digital or e‐health and telemedicine 40,43‐47 …”

Identifying an accurate self‐reported screening tool for alcohol use disorder: evidence from a Swiss, male population‐based assessment

“…There are no (statistical or clinical) standards for defining the margin of equivalence. A previous study in another field reported that it was difficult to obtain significant results with a margin of less than 10% , so we used this value in the study. Our findings might be used as a starting‐points for defining the margin of equivalence in other studies.…”

“…As described in the study protocol, we used several proportion non‐inferiority tests to ensure a sufficient sample size in both groups (with and without AUD) . We set the following parameters: alpha = 5%, power = 80% and margin of equivalence = 10% . We tested sensitivity values between 20 and 80% for self‐reported AUD and we used sensitivity = 50% because it was the worst scenario in term of sample size.…”

Diagnostic Accuracy of the French Version of the Adult Attention Deficit / Hyperactivity Disorder Self-Report Screening Scale for DSM-5 (ASRS-5)

“…We used a non-inferiority test to define how many participants were needed to obtain a similar AUC as in the first validation study (Ustun et al 2017). We used alpha=5%, power=80%, margin of equivalence=10% (Salazar et al 2016), AUC=0.94 (Ustun et al 2017), and allocation ratio=1. A total of n=98 (49 participants in each group) was needed.…”