Accuracy of Smartphone Images of the Cervix After Acetic Acid Application for Diagnosing Cervical Intraepithelial Neoplasia Grade 2 or Greater in Women With Positive Cervical Screening: A Systematic Review and Meta-Analysis

PURPOSE: Smartphones are used in cervical screening for visual inspection after acetic acid or Lugol's iodine (VIA/VILI) application to capture and share images to improve the sensitivity and interobserver variability of VIA/VILI. We undertook a systematic review and meta-analysis assessing the diagnostic accuracy of smartphone images of the cervix at the time of VIA/VILI (termed S-VIA) in the detection of precancerous lesions in women undergoing cervical screening. METHODS: This systematic review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Studies from January 1, 2010, to June 30, 2020, were assessed. MEDLINE/PubMed, Embase, CINAHL, Cochrane, and LILACS were searched. Cohort and cross-sectional studies were considered. S-VIA was compared with the reference standard of histopathology. We excluded studies where additional technology was added to the smartphone including artificial intelligence, enhanced visual assessment, and other algorithms to automatically diagnose precancerous lesions. The primary outcome was the accuracy of S-VIA for the diagnosis of cervical intraepithelial neoplasia grade 2 or greater (CIN 2+). Data were extracted, and we plotted the sensitivity, specificity, negative predictive value, and positive predictive value of S-VIA using forest plots. This study was prospectively registered with The International Prospective Register of Systematic Reviews:CRD42020204024. RESULTS: Six thousand three studies were screened, 71 full texts assessed, and eight studies met criteria for inclusion, with six included in the final meta-analysis. The sensitivity of S-VIA for the diagnosis of CIN 2+ was 74.56% (95% CI, 70.16 to 78.95; I(2) 61.30%), specificity was 61.75% (95% CI, 56.35 to 67.15; I(2) 95.00%), negative predictive value was 93.71% (95% CI, 92.81 to 94.61; I(2) 0%), and positive predictive value was 26.97% (95% CI, 24.13 to 29.81; I(2) 61.3%). CONCLUSION: Our results suggest that S-VIA has accuracy in the detection of CIN 2+ and may provide additional support to health care providers delivering care in low-resource settings.