Relative diagnostic performance and feasibility of filter paper for storing self-collected vaginal samples for human papillomavirus testing in Brazil: A cross-sectional study

OBJECTIVE: To evaluate the agreement and relative diagnostic performance of any-type high-risk human papillomavirus detection in self-collected vaginal samples stored on cellulose filter paper compared with clinician-collected cervical samples stored in SurePath™, using the Cobas 4800 human papillomavirus test. Acceptability and feasibility were evaluated as secondary outcomes. METHODS: This cross-sectional study was conducted across six Barretos Cancer Hospital Prevention Units in Brazil. Participants provided paired samples: a self-collected vaginal sample stored on cellulose filter paper and a clinician-collected cervical sample stored in SurePath™. Human papillomavirus testing was performed using the Cobas 4800 human papillomavirus test. Primary outcomes included overall agreement, kappa coefficient, relative sensitivity, relative specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve. Acceptability was assessed via questionnaire, and feasibility was evaluated descriptively. RESULTS: Among 1,290 paired sample sets submitted, 158 (12.2%) had an invalid result in at least one method, leaving 1,132 valid pairs for diagnostic performance analyses. Among valid pairs, high-risk human papillomavirus positivity was 37.4% (423/1,132) in self-collected filter-paper samples and 44.5% (504/1,132) in clinician-collected SurePath™ samples. Overall agreement was 90.9% (1,029/1,132) with a kappa of 0.813 (p < 0.001). Positive predictive value and negative predictive value were 97.4% and 87.0%, respectively. Relative sensitivity was 81.8% (95% confidence interval, 78.1 to 85.0%), relative specificity was 98.3% (95% confidence interval, 96.9 to 99.1%), and area under operating characteristic curve was 0.90 (95% confidence interval, 0.88 to 0.92). Self-sampling was highly acceptable: 99% (1,280/1,290) rated it easy, and 75% (971/1,290) preferred self-collection. CONCLUSIONS: Self-collected vaginal samples stored on cellulose filter paper showed strong agreement and high relative specificity compared with clinician-collected cervical samples in SurePath™. However, lower high-risk human papillomavirus positivity, reduced relative sensitivity, and higher invalid-result frequencies indicate that protocol optimization and clinical validation against cervical disease endpoints are required before implementation.