Accuracy of detection of high‐grade cervical intraepithelial neoplasia using electrical impedance spectroscopy with colposcopy

Objective To determine if electrical impedance spectroscopy (EIS) improves the diagnostic accuracy of colposcopy when used as an adjunct. Design Prospective, comparative, multi‐centre clinical study. Setting Three colposcopy clinics: two in England and one in Ireland. Population Women referred with abnormal cytology. Methods In phase 1, EIS was assessed against colposcopic impression and histopathology of the biopsies taken. In phase 2, a probability index and cut‐off value for the detection of high‐grade cervical intraepithelial neoplasia (HG–CIN, i.e. grade CIN2+) was derived to indicate sites for biopsy. EIS data collection and analyses were performed in real time and blinded to the clinician. The phase‐2 data were analysed using different cut‐off values to assess performance of EIS as an adjunct. Main outcome measure Histologically confirmed HG–CIN (CIN2+). Results A total of 474 women were recruited: 214 were eligible for analysis in phase 1, and 215 were eligible in phase 2. The average age was 33.2 years (median age 30.3 years, range 20–64 years) and 48.5% (208/429) had high‐grade cytology. Using the cut‐off from phase 1 the accuracy of colposcopic impression to detect HG–CIN when using EIS as an adjunct at the time of examination improved the positive predictive value (PPV) from 78.1% (95% CI 67.5–86.4) to 91.5%. Specificity was also increased from 83.5% (95% CI 75.2–89.9) to 95.4%, but sensitivity was significantly reduced from 73.6% (95% CI 63.0–82.5) to 62.1%, and the negative predictive value (NPV) was unchanged. The positive likelihood ratio for colposcopic impression alone was 4.46. This increased to 13.5 when EIS was used as an adjunct. The overall accuracy of colposcopy when used with EIS as an adjunct was assessed by varying the cut‐off applied to a combined test index. Using a cut‐off set to give the same sensitivity as colposcopy in phase 2, EIS increased the PPV to detect HG–CIN from 53.5% (95% CI 45.0–61.8) to 67%, and specificity increased from 38.5% (95% CI 29.4–48.3) to 65.1%. NPV was not significantly increased. Alternatively, applying a cut‐off to give the same specificity as colposcopy alone increased EIS sensitivity from 88.5% (95% CI 79.9–94.4) to 96.6%, and NPV from 80.8% (95% CI 67.5–90.4) to 93.3%. PPV was not significantly increased. The receiver operator characteristic (ROC) to detect HG–CIN had an area under the curve (AUC) of 0.887 (95% CI 0.840–0.934). Conclusions EIS used as an adjunct to colposcopy improves colposcopi