Comparison of Diagnostic Accuracy between Self-Collected and Physician-Collected Samples for HPV DNA Detection in Women

Abstract

Background
Cancer screening is critical for early detection and longer survival. While physician-collected samples are routine, self-sampling is more convenient, private, and accessible. This study compares the diagnostic performance and agreement of self- and physician-collected samples for cervical HPV detection.
Methods
A hospital-based cross-sectional study was conducted among 1,000 participants at BP Koirala Memorial Cancer Hospital, Bharatpur, Chitwan, Nepal, from July to December 2024. Ethical approval was obtained from the Nepal Health Research Council (Ref. No. 53_2024), and informed consent was taken from all participants. Both self-sampling and physician-collected cervical samples were obtained. Demographic data were gathered via structured questionnaires. The physician-collected sample served as the gold standard. Diagnostic performance metrics, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy were calculated for self-sampling. Agreement between methods was assessed using Cohen’s Kappa coefficient, and associations were tested with the Chi-square test.
Results
Self-sampling exhibited high concordance with physician-collected sampling, with a Cohen’s Kappa of 0.885, indicating almost perfect agreement. The Chi-square test (χ² = 786.068, p < 0.001) confirmed a significant association between methods. The diagnostic performance of self-sampling was 95.29% sensitivity, 98.36% specificity, 84.38% positive predictive value (PPV), 99.56% negative predictive value (NPV), and 98.10% accuracy.
Conclusions
Self-sampling demonstrates excellent diagnostic accuracy and strong agreement with physician-collected samples for detecting cervical HPV. These results indicate that self-sampling is a reliable, accessible, and private alternative for cancer screening. Adoption of self-sampling could strengthen screening programs, especially in underserved regions. Further research is needed to assess scalability and long-term impact on early cancer detection.