A Smartphone-Based Approach for Triage of Human Papillomavirus-Positive Sub-Saharan African Women: A Prospective Study

• Main Authors: Urner, Esther, Delavy, Martine, Catarino, Rosa, Viviano, Manuela, Meyer-Hamme, Ulrike, Benski, Anne-Caroline, Jinoro, Jeromine, Heriniainasolo, Josea Lea, Undurraga, Manuela, De Vuyst, Hugo, Combescure, Christophe, Vassilakos, Pierre, Petignat, Patrick
• Format: Article
• Language: English
• Published: JMIR Publications 2017-05-01
• Series: JMIR mHealth and uHealth
• Online Access: http://mhealth.jmir.org/2017/5/e72/

BackgroundSub-Saharan African countries are marked by a high incidence of cervical cancer. Madagascar ranks 11th among the countries with the highest cervical cancer incidence worldwide. ObjectiveThe aim of the study was to evaluate the performances of digital smartphone-based visual inspection with acetic acid (D-VIA) and Lugol’s iodine (D-VILI) for diagnosing cervical precancer and cancer. MethodsHuman papillomavirus (HPV)-positive women recruited through a cervical screening campaign had D-VIA and D-VILI examinations with endocervical curettage (ECC) and cervical biopsy. Three images were captured for each woman (native, D-VIA, D-VILI) using a smartphone camera. The images were randomly coded and distributed on 2 online databases (Google Forms). The D-VIA form included native and D-VIA images, and the D-VILI form included native and D-VILI images. Pathological cases were defined as cervical intraepithelial neoplasia grade 2 or worse (CIN2+). Physicians rated the images as non-pathological or pathological. Using the ECC and cervical biopsy results as references, the sensitivity and specificity of D-VIA and D-VILI examinations for each and all physicians were calculated. ResultsAltogether, 15 clinicians assessed 240 images. Sensitivity was higher for the D-VIA interpretations (94.1%; 95% CI 81.6-98.3) than for the D-VILI interpretations (78.8%; 95% CI 54.1-92.1; P=.009). In contrast, the specificity was higher for the D-VILI interpretations (56.4%; 95% CI 38.3-72.9) than for the D-VIA interpretations (50.4%; 95% CI 35.9-64.8; P=.005). ConclusionSmartphone-based image for triage of HPV-positive women is more accurate for detecting CIN2+ lesions with D-VIA than D-VILI, although with a small loss of specificity.