Articles were included in the review if they included participants who underwent cervicovaginal self-sampling for HPV DNA testing; measured the accuracy, concordance, and acceptability of cervicovaginal self-sampling and clinician sampling for HPV; focused on Asian patients; were conducted in Asian countries and were in English. The included studies were randomized controlled trials, prospective cohort studies, cross-sectional studies, comparative studies, and other non-randomized controlled trials. Studies that did not use vaginal or cervical specimens for examination were excluded from the review. Studies that focused on non-Asian populations, or did not report relevant outcomes related to the accuracy of self-sampling, concordance with clinician-collected samples, or women’s acceptance of self-sampling, were excluded.

The PubMed, Cochrane Library, Cumulative Index to Nursing and Allied Health Library (CINHAL), and Web of Science databases were searched for studies reported from the establishment of the database to 31 October 2022. A final update of the search was completed before the final extraction and synthesis of the results on 23 February 2023. The reference lists of the included articles were also screened to identify publications that met the eligibility criteria. Database-specific Boolean operators (AND, OR, NOT) and truncation symbols (* and “ “) were used.

The following search terms were used to identify eligible studies:

1. Cervical dysplasia OR cervical intraepithelial neoplasia OR cervix neoplasms OR papillomavirus OR papillomavirus, human OR human papillomavirus OR papillomavirus, infections

AND

2. Self-collected OR self-test OR self-obtained OR self-sampling

AND

3. Asia OR Asian OR Afghanistan OR Armenia OR Azerbaijan OR Bahrain OR Bangladesh OR Bhutan OR Brunei OR Cambodia OR China OR Cyprus OR Georgia OR India OR Indonesia OR Iran OR Iraq OR Israel OR Japan OR Jordan OR Kazakhstan OR Korea, North OR Korea, South OR Kuwait OR Kyrgyzstan OR Laos OR Lebanon OR Malaysia OR Maldives OR Mongolia OR Myanmar OR Nepal OR Oman OR Pakistan OR Palestine OR Philippines OR Qatar OR Saudi Arabia OR Singapore OR Sri Lanka OR Syria OR Tajikistan OR Thailand OR Timor-Leste OR Turkmenistan OR Turkey OR United Arab Emirates OR Uzbekistan OR Vietnam OR Yemen.

The more detailed search strategies of each database were shown in Supplementary Table 2.

Descriptive data were extracted independently by two authors, and a third reviewer was consulted to resolve any differences in data collection. The citation, objectives, location, population characteristics, description of the type of HPV screening, description of any additional intervention components, study design, sample size, numerical outcomes, results, and limitations were extracted from each included study.

After finalizing the data extraction, two authors reviewed the data and the full texts to accurately classify HPV self-sampling.

The reported data regarding screening accuracy, participation, attendance, response, and compliance were combined to determine the cervical cancer screening outcomes. Conventional cytology (Pap smears), VIA, or colposcopy data were also gathered. When more than one control group was reported, the intervention group was compared to the least intensive sampling strategy group.

Two independent reviewers evaluated the risk of bias for all included studies by using the Quality Assessment Tool for Diagnostic Accuracy Studies-2 (QUADAS-2).

Heterogeneity was assessed using Cochran’s Q test and the I² statistic. Begg’s rank correlation test was performed to further assess publication bias. A funnel plot was used to visualize publication bias.

Human papillomavirus self-sampling was defined as the process in which women insert a self-sampler into their vagina to collect isolated cells. In contrast, HPV clinician sampling involved clinicians or healthcare workers inserting a vaginal speculum into the woman’s vagina to obtain a cervical smear using a sampler.

The diagnostic test sensitivity and specificity were based on colposcopy-confirmed cases of high-grade squamous intraepithelial lesion (HSIL), previously called cervical intraepithelial neoplasia 2+ (CIN2+) or CIN3+, and detection of cervical cancer and HPV infection. The sensitivity was defined as the number of identified cases of HSIL and cervical cancer (positive for both HPV and colposcopy) divided by the total number of colposcopy-confirmed cases. Specificity was defined as the number of cases without HSIL or cervical cancer (negative on both HPV and colposcopy) divided by the total number of colposcopy-negative cases. The HPV detection rate was defined as the HPV-positive cases divided by the total number of women enrolled. Agreement was defined as the concordance between self-sampled HPV tests and clinician-sampled HPV tests (the percentage of agreement with positive test results and the percentage of agreement with negative test results). Acceptability was defined as the percentage of women willing to participate in the HPV test and their preference between HPV self-sampling and clinician sampling.

A total of 573 articles were retrieved, comprising 124 studies from PubMed, 215 from Web of Science, 26 from the Cochrane Library, and 208 from CINAHL, including 135 duplicate titles. Therefore, 438 articles were screened against the eligibility criteria. Following the exclusion of 241 articles based on their titles and abstracts, the full texts of 195 articles were read, and 67 studies were ultimately included in the systematic review (Figure 1).

Most of the included studies were cross-sectional studies (n = 62). The remaining studies were randomized controlled trials (n = 1), prospective cohort studies (n = 1), prospective population-based studies (n = 1), and prospective randomized crossover studies (n = 2) (Supplementary Table 3).

The patient populations of the included studies were women in China (28 studies) (Belinson et al., 2001; Belinson et al., 2003; Belinson et al., 2010; Belinson et al., 2012; Chang et al., 2002; Chen S. et al., 2014; Chen W. et al., 2014; Chen K. et al., 2016; Chen Q. et al., 2016; Chou et al., 2016; Du et al., 2021; Goldstein et al., 2020; Guan et al., 2012; Guan et al., 2013; He and He, 2020; Li et al., 2022; Ngu et al., 2022; Qiao et al., 2008; Qin et al., 2016; Tisci et al., 2003; Twu et al., 2011; Wang et al., 2014; Wang et al., 2017; Wong et al., 2016; Wong et al., 2018; Wong et al., 2020; Zhang et al., 2020; Zhao et al., 2013), Thailand (nine studies) (Gottschlich et al., 2019; Kittisiam et al., 2016; Nilyanimit, 2014; Nutthachote et al., 2019; Oranratanaphan et al., 2014; Phoolcharoen et al., 2018a; Phoolcharoen et al., 2018b; Ploysawang et al., 2023; Trope et al., 2013), Japan (seven studies) (Aiko et al., 2017; Hanley et al., 2016; Onuma et al., 2020; Ozawa et al., 2023; Satake et al., 2020; Terada et al., 2022; Yoshida et al., 2011), Malaysia (seven studies) (Abdullah et al., 2018; Ahmad et al., 2021; Khoo et al., 2021; Latiff et al., 2015a; Latiff et al., 2015b; Ma’som et al., 2016; Tan et al., 2021), India (six studies) (Anand et al., 2022; Asthana and Labani, 2015; Bhatla et al., 2009; Kuriakose et al., 2020; Madhivanan et al., 2021; Sowjanya et al., 2009), Korea (three studies) (Cho et al., 2019; Seo et al., 2006; Shin et al., 2019), Nepal (two studies) (Johnson et al., 2014; Shrestha et al., 2021), Singapore (Lim et al., 2022), Mongolia (Tsedenbal et al., 2022), Cambodia (Thay et al., 2019), Vietnam (Hanh, 2006), and Brunei (Chaw et al., 2022).

A total of 19 studies evaluated the sensitivity and specificity of clinician-collected and self-collected HPV testing for diagnosing CIN. A total of 35 studies reported the detection rates of HPV using both self-sampling and clinician sampling methods. A total of 29 studies examined concordance between clinician-collected and self-collected HPV testing. A total of 33 studies assessed women’s acceptance and preference rates for HPV self-sampling.

A total of 13 HPV detection methods are discussed in this review, including seven WHO-approved testing methods: HC2 (Qiagen, Germantown, MD, United States), careHPV (Qiagen, Gaithersburg, MD, United States), AmpFire (Atila BioSystems, United States), SeqHPV (BGI Shenzhen, Shenzhen, China), Cervista (Hologic, Marlborough, MA, United States), matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF, BGI Shenzhen, Shenzhen, China), and Cobas HPV test (Roche Molecular Systems, Inc., United States). Additionally, six other methods are introduced, including HPVDNA Chip™ (Biomedlab Co., Seoul, South Korea), PGMY PCR (Roche Molecular Systems, Inc., United States), Easy-Chip HPV Blot (King Car Yuanshan Research Institute, Taiwan, China), RealTime High Risk HPV assay (Abbott Molecular Inc., Abbott Park, IL), Anyplex II HPV kit assay (Seegene, Seoul, South Korea), and Linear Array HPV Genotyping test (Roche Diagnostics, United Kingdom). HC2, Cobas HPV test, and Cervista have received FDA/CE-IVD approval. HC2 test detects the presence of 13 HR-HPV types using full genome probes complementary to HPV DNA, specific antibodies, signal amplification, and chemiluminescent detection (Belinson et al., 2001). HPVDNA Chip uses HPV and β-globin primers to amplify the target HPV DNA through PCR under specific conditions, and the amplification products are labeled with Cy5-dUTP, which could contain 22 type-specific probes (15 for the high-risk group and seven the low-risk group) (Seo et al., 2006). PGMY PCR uses the PGMY09/11 L1 consensus primer system for PCR amplification and a reverse line blot detection strip that individually identifies 22 high-risk types (Bhatla et al., 2009). The careHPV assay, adapted from the HC2 assay, is a qualitative test for HR-HPV detection, targeting 14 HR-HPV types through hybridization of HR-HPV DNA with a cocktail of RNA probes and chemiluminescence signal amplification (Qiao et al., 2008). Easy-Chip HPV Blot contains 39 type-specific probes that are immobilized on a 14.4 mm × 9.6 mm nylon membrane, which is used for reverse-blot hybridization and detects HPV DNA in a single assay (Twu et al., 2011). Cervista is a signal-amplification method for the qualitative detection of 14 HR-HPV types (Belinson et al., 2012). MALDI-TOF is a mass spectrometry method that uses a multiplex primary PCR also for the same 14 HR-HPV types detected by Cervista (Belinson et al., 2012). Cobas HPV test is a real-time PCR assay that detects 14 HPV types, with HPV16 and HPV18 detected individually and the other 12 HPV types detected as a pooled group (Chen Q. et al., 2016b Terada et al., 2022). The AmpFire method is a nucleic acid amplification technique for qualitative detection of HR-HPV, using HR-HPV-specific primers and fluorescent probes to amplify the viral genomic DNA (including the E6/E7 region) under isothermal conditions. This method does not require DNA extraction or purification and can directly detect HPV from lysed clinical samples in one step (Zhang et al., 2020). The SeqHPV assay is a high-throughput HPV genotyping method based on multiplex PCR and next-generation sequencing, capable of detecting 14 HR-HPV types (Du et al., 2021). The Abbott m2000rt automatic biochemical analyzer was used for real-time fluorescence quantitative PCR detection. The detection boundary value of cycle threshold (CT) was 32.0, and the internal quality control target boundary value of CT was 35.0. Abbott HR-HPV assay could detect 14 HR HPV types (Abdullah et al., 2018; Aiko et al., 2017; Chen W. et al., 2014; Chen K. et al., 2016; Chou et al., 2016; Hariprasad et al., 2023; Li et al., 2022; Nutthachote et al., 2019; Page et al., 2021; Phoolcharoen et al., 2018a; Satake et al., 2020; Tan et al., 2021; Wong et al., 2020; Yoshida et al., 2011) simultaneously, and specifically identifies HPV16 and HPV18 (Qin et al., 2016). Anyplex^{^TM} II HPV 28 real-time PCR test simultaneously detects 19 HR-HPV and 9 low-risk HPV types, using dual priming oligonucleotides and a melting curve analysis method of tagging oligonucleotide cleavage and extension (Cho et al., 2019). Linear Array HPV Genotyping test (Roche Diagnostics, United Kingdom) combines consensus PCR and reverse-hybridization amplification products to detect 36 genital HPV genotypes. Because it has been clearly defined and validated in research and clinical applications, it is often considered the reference method for genital HPV genotyping (Yoshida et al., 2011).

All the studies included in this systematic review were assessed for risk of bias (Figure 2). The Cohen’s kappa value between two independent reviewers was 0.839. Most of the studies included in the analysis were cross-sectional and did not employ random patient selection or allocation. As a result, the risk of bias in several domains was found to be high or unclear. Specifically, 21 studies were assessed as having a high risk of bias in the “Patient Selection” domain (Aiko et al., 2017; Chen Q. et al., 2016; Goldstein et al., 2020; Gottschlich et al., 2019; Hanh, 2006; Hanley et al., 2016; Johnson et al., 2014; Kuriakose et al., 2020; Madhivanan et al., 2021; Nilyanimit, 2014; Oranratanaphan et al., 2014; Ozawa et al., 2023; Phoolcharoen et al., 2018a; Qin et al., 2016; Seo et al., 2006; Shrestha et al., 2021; Thay et al., 2019; Twu et al., 2011; Wang et al., 2014; Wong et al., 2018; Yoshida et al., 2011). This high risk was attributed to the non-random selection of participants, which could introduce selection bias and limit the generalizability of the findings. One study was considered to have a high risk of bias in the “Index Testing” domain (Kittisiam et al., 2016), due to the use of a non-standardized or poorly validated diagnostic test, which could affect the accuracy of the results. In the “Reference Standard” domain, only one study was deemed to have an unclear risk of bias (Shrestha et al., 2021), due to a lack of detailed information regarding the reference standard used. In the “Flow and Timing” domain, four studies were assessed as having an unclear risk of bias (Hanh, 2006; He and He, 2020; Ngu et al., 2022; Shrestha et al., 2021), which was due to incomplete reporting of participant flow or unclear timing of tests, potentially leading to attrition or measurement bias. Notably, all studies were judged to have a low risk of diagnostic bias, as the diagnostic criteria were predefined prior to the availability of results, ensuring the objectivity of the assessment.

The Cochran’s Q statistic was highly significant (Q = 1.8 × 10⁶, P = 0.000), indicating substantial heterogeneity among the studies. Additionally, the I² statistic was calculated to be 100%, suggesting that nearly all of the variability in effect sizes across studies could be attributed to differences between studies rather than random error. Begg’s test yielded a significant p-value (P < 0.05), further suggesting the presence of potential publication bias. This finding implies that smaller studies or studies with non-significant results may be underrepresented or unpublished, which could have influenced the overall effect size observed in the meta-analysis. The funnel plot (Figure 3) exhibited signs of asymmetry, suggesting the presence of publication bias. Specifically, there appears to be an over-representation of studies with larger effect sizes, while smaller studies with negative or null results may be underrepresented.

Two studies found that the accuracy of HPV self-sampling is comparable to that of physician sampling (Aiko et al., 2017; Belinson et al., 2012). In the study of Belinson et al. (2012) when using the MALDI-TOF mass spectrometry system for HPV detection, the sensitivity of self-sampling for identifying CIN 3+ was equivalent to that of clinician sampling. However, when utilizing Cervista, the sensitivity for detecting CIN 3+ in self-collected specimens was only 70.9%, compared to 95.0% for clinician-collected samples (Belinson et al., 2012). In the study of Onuma et al. (2020) the sensitivity of HPV self-sampling and clinician sampling for the detection of CIN 2+ were both 100% (Aiko et al., 2017). In studies of Du et al. (2021), Zhang et al. (2020), the sensitivity for detecting CIN 2+ was higher in self-sampling than in clinician sampling. While in the other three studies of Belinson et al. (2001) the sensitivity for detecting CIN 3+ was 81% and 98% in self- and clinician-collected samples. In the remaining 12 studies, the sensitivity of HPV self-sampling was slightly lower than physician sampling, with values ranging from 59.4% to 87.5%, while the specificity of HPV self-sampling was identical to clinician sampling (Table 1).

In the detection of HPV, 17 studies reported the detection rates were higher in clinician sampling (Anand et al., 2022; Asthana and Labani, 2015; Belinson et al., 2001; Chang et al., 2002; Chen K. et al., 2016; Kuriakose et al., 2020; Latiff et al., 2015a; Madhivanan et al., 2021; Ma’som et al., 2016; Satake et al., 2020; Seo et al., 2006; Singh et al., 2023; Terada et al., 2022; Thay et al., 2019; Twu et al., 2011; Wang et al., 2014; Wang et al., 2017), while in 15 studies this rate was higher in self-collected samples (Belinson et al., 2003; Cho et al., 2019; Du et al., 2021; Latiff et al., 2015a; Nutthachote et al., 2019; Qin et al., 2016; Satake et al., 2020; Seo et al., 2006; Terada et al., 2022; Thay et al., 2019; Wong et al., 2016; Wong et al., 2018; Yoshida et al., 2011; Zhang et al., 2020; Zhao et al., 2013). Two studies evaluated that detection rates of both sampling methods were the same (Anand et al., 2022; Nilyanimit, 2014). In 13 studies, the difference in detection rates was not more than 1% (Aiko et al., 2017; Asthana and Labani, 2015; Belinson et al., 2001; Bhatla et al., 2009; Chang et al., 2002; Chen W. et al., 2014; Hanh, 2006; Kuriakose et al., 2020; Lim et al., 2022; Satake et al., 2020; Wang et al., 2017; Zhang et al., 2020; Zhao et al., 2013; Table 2). In the studies by Wong and Yoshida et al., multiple types of HPV infections were found to occur more frequently with self-sampling compared to clinician sampling (Wong et al., 2016; Yoshida et al., 2011).

A total of 29 studies reported an agreement between HPV self-sampling and clinician sampling. A total of 24 reported a high or nearly perfect agreement between self-sampling and clinician sampling for the detection of HPV DNA. Specifically, 18 studies demonstrated an agreement exceeding 90% (Anand et al., 2022; Bhatla et al., 2009; Chen K. et al., 2016; Chen Q. et al., 2016; Du et al., 2021; Johnson et al., 2014; Kuriakose et al., 2020; Latiff et al., 2015a; Madhivanan et al., 2021; Ngu et al., 2022; Nilyanimit, 2014; Nutthachote et al., 2019; Satake et al., 2020; Seo et al., 2006; Sowjanya et al., 2009; Tsedenbal et al., 2022; Wang et al., 2014; Wong et al., 2016). Three studies assessed the agreement in both collecting methods samples using two assays for the detection of HPV (Chen W. et al., 2014; Du et al., 2021; Sowjanya et al., 2009), and one study evaluated the concordance of both sampling methods in three HPV testing assays (Cho et al., 2019). Some new HPV assays such as SeqHPV and careHPV showed higher agreement in self- and clinician-collected samples. In studies of Chen W. et al. (2014), Du et al. (2021) when the same sample was tested using different detection methods, the consistency of clinician-sampled samples was higher than that of self-sampled samples.

However, three studies have reported poor agreement between self- and clinician sampling results for the detection of HPV. Twu et al. (2011) reported low agreement between vaginal and cervical specimens using the EasyChip HPV Blot (k = 0.37) (Table 3).

A total of 29 studies have assessed women’s overall acceptance of HPV self-sampling. The lowest reported acceptance was 40.3% (95% CI: 38.49%–42.11%) (Trope et al., 2013), while the highest reached 100% (Anand et al., 2022; Ploysawang et al., 2023). In 27 of the 29 studies, acceptance exceeded 60% (Abdullah et al., 2018; Ahmad et al., 2021; Aiko et al., 2017; Anand et al., 2022; Chen S. et al., 2014; Cho et al., 2019; Chou et al., 2016; Goldstein et al., 2020; Gottschlich et al., 2019; Guan et al., 2012; Hanley et al., 2016; Khoo et al., 2021; Kittisiam et al., 2016; Li et al., 2022; Ma’som et al., 2016; Ngu et al., 2022; Oranratanaphan et al., 2014; Phoolcharoen et al., 2018b; Ploysawang et al., 2023; Shrestha et al., 2021; Singh et al., 2023; Sowjanya et al., 2009; Thay et al., 2019; Tisci et al., 2003; Trope et al., 2013; Wong et al., 2016; Wong et al., 2020). A total of 13 studies indicated that women preferred self-sampling over clinician sampling (Goldstein et al., 2020; Gottschlich et al., 2019; Hanh, 2006; Khoo et al., 2021; Li et al., 2022; Lim et al., 2022; Madhivanan et al., 2021; Ploysawang et al., 2023; Shin et al., 2019; Shrestha et al., 2021; Trope et al., 2013; Wong et al., 2016; Wong et al., 2018), however, three studies found a preference for clinician sampling instead (Aiko et al., 2017; Ngu et al., 2022; Tsedenbal et al., 2022; Table 4).

When asked about their preferred location for self-sampling, four studies found that participants preferred to perform the test at the clinic rather than at home (Belinson et al., 2001; Chen K. et al., 2016; Kittisiam et al., 2016; Zhang et al., 2020). In contrast, three studies reported a preference for sampling at home (Onuma et al., 2020; Seo et al., 2006; Tan et al., 2021).

This review is not without limitations. The characteristics of the participants enrolled in the primary studies differed, as did the sample sizes. In addition, the methods for HPV testing and sampling devices were not described in several studies. The diagnostic accuracy of the HPV tests was not uniform across the studies, and the intervals between self-sampling and clinician sampling were also inconsistent. Another limitation of this review is that only studies conducted in East Asia, Southeast Asia, and South Asia were included. Finally, gray literature and conference abstracts were not included in this review, and the exclusion of non-English articles may have limited the comprehensiveness of the analysis. These exclusions could introduce potential bias, as studies published in languages other than English or in gray literature might have different characteristics or findings compared to those published in peer-reviewed journals. Consequently, the findings of this review may not fully represent the entire body of literature, and future research should consider including non-English studies and gray literature to provide a more comprehensive understanding of the topic.