Comments on: Limitations of HPV DNA Testing in Screening of Cervical Adenocarcinomas

Dear Editor, We read with interest the article on human papillomavirus (HPV) DNA testing for cervical cancer in Brazil, authored by Zeferino et al (Brazilian Association for the Lower Genital Tract Pathology and Colposcopy (ABPTGIC, in the Portuguese acronym), as well as the subsequent letter to the editor, by Dr de Alcantara Segura, and the authors’ reply.1,2 We agree that detection of adenocarcinomas represents a significant challenge for cervical cancer screening programs. In the United States, squamous cancer rates have seen a decline since the introduction of the Pap smear and molecular testing, but adenocarcinoma rates have remained unchanged (►Fig. 1). Moreover, Australia, which arguably has a best in class vaccination and screening program, is reporting a steady increase in adenocarcinoma rates.3 As Dr. de Alcantara Segura points out, cytology is an insensitive method for screening adenocarcinomas. This has also been shown to be the case in the US, where Katki et al4 reported that 63% of adenocarcinomas showed normal cytology results. In addition, screening for adenocarcinomaprecursors byHPV is clearly less sensitive than squamous cancer precursors, as evidenced by the lack of progress in reducing the adenocarcinoma incidence since the introduction of molecular testing in the 2000s (►Fig. 1). A review of the recently published literature has identified a number of potential causes for theapparent lackofHPVdetectionwithincervical adenocarcinoma, including the following: (i) Historical use of the Hybrid Capture 2 (HC2) assay and an associated higher rate of HPV negative cervical adenocarcinoma cases reported5,6; (ii) low viral load of HPVand detection only with nested polymerase chain reaction (PCR) methods7; (iii) L1 vs E6/E7 detection (HPV integration and L1 deletion)8,9; (iv) non-HR HPV genotypes associated with some adenocarcinomas10; and (v) rare histological subtypes thateitherhavelowviral loadsorareHPVnegative.11,12 There are some cervical adenocarcinoma tissues that appear to be non-HPV derived. Clear cell adenocarcinoma can arise from prenatal diethylstilbestrol exposure that are not believed to be associated with high-risk HPV infections. Likewise, gastric-type adenocarcinomas are considered to be either HPV-negative, or at least, have very low levels ofHPVpresent.13 Inparticular, target regionselectionhasasignificant impact on an HPV DNA test’s ability to detect cancers. Themajority of assays target the L1 regionof thevirus, and thegene isprone to deletion during virus integration as disease progresses.14–16 Viral integrationcanalso occurearly in thediseaseprocess and is detected in routine liquid-based cytology specimens of precancer patients.14,17 While it has long been reported that E6/E7 target regions are more sensitive than L1, some have argued that this is due to technical difficulties with isolating DNA from fixed tissue (formalin-fixed, paraffin embedded [FFPE]) samples.18 The use of L1-target region assays, and the older detectionmethods in the cross-sectional study by de Sanjose et al19 may have contributed to the quoted poor detection rate (62%) of adenocarcinomas. More recent studies using more robust methods and shorter target sequences estimate the rate of HPV-negative cancers at approximately 10% using L1-based PCR assays.20,21While an improvement, it still represents an unacceptably high false-negative rate. Assays that target the E6/E7 genes rarely undergo deletion because of the requirement for E6/E7 gene products in oncogenesis. Thus, assays that target this region of the virus should be more sensitive in detecting cancers. Recent publications confirm that this is in fact the case. Molijn et al8 reported a 7% increase in adenocarcinoma detection using type-specific E6 PCR versus a very sensitive L1-based SPF10-PCR assay. Lagheden et al21 also reported the results of a comprehensive analysis of HPV type detection using tissue blocks from the Swedish cancer registry collected from 2002 to 2011. The authorsweresuccessful inobtaining69%ofall cell blocks (from 22 different biobanks), and samples were extracted using standardized procedures with hematoxylin & eosin (H&E) confirmatory staining of the first and last sections of each block for re-review. A total of 37 different HPV types were detected using L1 Modified General Primers (MGP) and a Laurence M Vaughan's ORCID is https://orcid.org/0000-00032976-3215.

Dear Editor, We read with interest the article on human papillomavirus (HPV) DNA testing for cervical cancer in Brazil, authored by Zeferino et al (Brazilian Association for the Lower Genital Tract Pathology and Colposcopy (ABPTGIC, in the Portuguese acronym), as well as the subsequent letter to the editor, by Dr de Alcantara Segura, and the authors' reply. 1,2 We agree that detection of adenocarcinomas represents a significant challenge for cervical cancer screening programs. In the United States, squamous cancer rates have seen a decline since the introduction of the Pap smear and molecular testing, but adenocarcinoma rates have remained unchanged (►Fig. 1). Moreover, Australia, which arguably has a best in class vaccination and screening program, is reporting a steady increase in adenocarcinoma rates. 3 As Dr. de Alcantara Segura points out, cytology is an insensitive method for screening adenocarcinomas. This has also been shown to be the case in the US, where Katki et al 4 reported that 63% of adenocarcinomas showed normal cytology results. In addition, screening for adenocarcinoma precursors by HPV is clearly less sensitive than squamous cancer precursors, as evidenced by the lack of progress in reducing the adenocarcinoma incidence since the introduction of molecular testing in the 2000s (►Fig. 1). A review of the recently published literature has identified a number of potential causes for the apparent lack of HPV detection within cervical adenocarcinoma, including the following: (i) Historical use of the Hybrid Capture 2 (HC2) assay and an associated higher rate of HPV negative cervical adenocarcinoma cases reported 5,6 ; (ii) low viral load of HPV and detection only with nested polymerase chain reaction (PCR) methods 7 ; (iii) L1 vs E6/E7 detection (HPV integration and L1 deletion) 8,9 ; (iv) non-HR HPV genotypes associated with some adenocarcinomas 10 ; and (v) rare histological subtypes that either have low viral loads or are HPVnegative. 11,12 There are some cervical adenocarcinoma tissues that appear to be non-HPV derived. Clear cell adenocarcinoma can arise from prenatal diethylstilbestrol exposure that are not believed to be associated with high-risk HPV infections. Likewise, gastric-type adenocarcinomas are considered to be either HPV-negative, or at least, have very low levels of HPV present. 13 In particular, target region selection has a significant impact on an HPV DNA test's ability to detect cancers. The majority of assays target the L1 region of the virus, and the gene is prone to deletion during virus integration as disease progresses. [14][15][16] Viral integration can also occur early in the disease process and is detected in routine liquid-based cytology specimens of precancer patients. 14,17 While it has long been reported that E6/E7 target regions are more sensitive than L1, some have argued that this is due to technical difficulties with isolating DNA from fixed tissue (formalin-fixed, paraffin embedded [FFPE]) samples. 18 The use of L1-target region assays, and the older detection methods in the cross-sectional study by de Sanjose et al 19 may have contributed to the quoted poor detection rate (62%) of adenocarcinomas. More recent studies using more robust methods and shorter target sequences estimate the rate of HPV-negative cancers at approximately 10% using L1-based PCR assays. 20,21 While an improvement, it still represents an unacceptably high false-negative rate. Assays that target the E6/E7 genes rarely undergo deletion because of the requirement for E6/E7 gene products in oncogenesis. Thus, assays that target this region of the virus should be more sensitive in detecting cancers. Recent publications confirm that this is in fact the case. Molijn et al 8 reported a 7% increase in adenocarcinoma detection using type-specific E6 PCR versus a very sensitive L1-based SPF10-PCR assay. Lagheden et al 21 also reported the results of a comprehensive analysis of HPV type detection using tissue blocks from the Swedish cancer registry collected from 2002 to 2011. The authors were successful in obtaining 69% of all cell blocks (from 22 different biobanks), and samples were extracted using standardized procedures with hematoxylin & eosin (H&E) confirmatory staining of the first and last sections of each block for re-review. A total of 37 different HPV types were detected using L1 Modified General Primers (MGP) and a Luminex bead readout. Of the 2,932 retrieved blocks, 2,850 (97%) gave valid PCR results (beta globin internal cellularity control detected). A total of 527 (18%) samples were found to be HPV-negative but approximately 8% of these were negative for invasive cancer on reviewing the histology, and were thus excluded from the analysis. The DNAs from the 483 remaining L1-negative cases were re-tested using E6/E7 type specific primers, and 89 additional cancer cases were detected. This corresponds to an 18% (89/483) improvement or a 3.2% (89/ 2806) gain in overall HPV-negative cancer detection using just the HPV16/18 target regions (65% of the cancer cases in the study cohort). If one assumes an equal detection rate for all 14 high-risk types, this would result in an approximately 4.9% (136/2806) improvement in overall cancer detection. Thus, based on this large cancer cohort (representing approximately 1.57 million screened women), 22 an estimated 28% (136/483) increase in HPV-negative cancer detection might be obtained using E6/E7 versus L1 primer sets.
While it is not possible to detect all cancer using either cytology and or HPV testing, there is little doubt that HPV testing is the most sensitive tool. In addition to the Ronco et al 23 study cited by Zeferino and colleagues, Schiffman et al 24 performed a comprehensive analysis in a cohort of over 1.2 million women with screening histories preceding cancers (n ¼ 623) and precancers (n ¼ 5369). They found that HPV testing identified more women subsequently diagnosed with cancer (P < 0.001) and precancer (P < 0.001), except for within 12 months when cytology showed improved cancer detection (P ¼ 0.10). However, this represented, at most, 5 cases per million women per year, and the authors questioned the utility of continued co-testing versus primary HPV screening -"Excessive screening in an attempt to prevent every case could have minimal cancer prevention benefits while increasing the harms of screening".
The decision on which screening paradigm to adopt is often driven by both the science as well as regional or country-specific parameters, including infrastructure, societal risk tolerance, and funding/reimbursement. However, it is important that clinicians and decision makers carefully weigh the safety and effectiveness of different approaches so as to maximize the impact of any screening program. Adenocarcinomas continue to represent a significant screening challenge, but it is hoped that with advancements in screening tools and the impact of vaccination, this performance gap will soon begin to close.
Finally, the issue with HPV-negative cervical adenocarcinoma poses both a challenge for cervical cancer screening programs and an important scientific and clinical question about the etiology of cervical glandular lesions. In our opinion, this situation necessitates a "call to arms" within the scientific and clinical community, in which a systematic investigation of cervical adenocarcinoma cases using deep nucleic acid sequencing analysis is warranted to answer the question about HPV association with all cervical adenocarcinomas. It is likely that these studies will lead to next generation cervical cancer screening technologies and improved adenocarcinoma detection.

Authors' Reply
Dear Editor, We are grateful for the insightful comments made by the authors and the well-structured review of literature, which indicates a new step on the debate of cervical cancer screening. The detection of adenocarcinomas represents a challenge for cervical cancer screening programs, and the performance of HPV tests on preventing them is debatable.
Even considering these issues, in regions where the incidence of cervical cancer is middle or high, the robust evidence about the DNA-HPV test performance supports the recommendations for implementing this technology instead of cytology, as long as accompanied by improvements on the programmatic strategy to ensure access and high coverage for the target population. It is clear that the benefits observed in the first stage of the HPV-DNA test implementation will be more evident in the control of squamous neoplasms. In the second stage, it is possible that the limitations pointed out in the detection of glandular neoplasia will be more evident. This debate is fairly current in countries with low rates of cervical cancer, which have already had controlled incidence rates.
In middle-income countries such as Brazil, where the incidence of squamous cell carcinoma is high, the impact of the implementation of HPV test-based screening surpasses the doubtful HPV test performance on detecting cervical adenocarcinoma. By now, it may only indicate a reduction in expectations regarding the impact to be observed with the new technique.
In the near future, new screening technologies might overcome these limitations. In addition, ongoing vaccination policies may contribute to reduce the impact of this issue. A more detailed analysis of the clinical and biomolecular characteristics of endocervical adenocarcinomas is quite opportune. Furthermore, a better understanding of such characteristics could provide new tools for screening test and strategies to improve adenocarcinoma detection.