It is time to improve the diagnostic workup of oropharyngeal cancer with circulating tumor HPV DNA: Systematic review and meta‐analysis

The possibility of detecting circulating tumor HPV DNA (ctHPVDNA) in plasma in patients with oropharyngeal cancer has been demonstrated in several reports. However, these data are from small cohorts and available tests for detection of ctHPVDNA are not fully validated. The aim is to evaluate sensitivity, specificity, and accuracy of ctHPVDNA by ddPCR to define its efficacy in the clinical setting for the diagnosis of HPV + OPSCC. A comprehensive search of three different databases: MEDLINE, Embase, and Cochrane Library databases. A total of 998 patients were evaluated from the 13 studies. OPSSC p16+ were 729, while controls p16− were 269. The meta‐analytic study estimated the diagnostic performance of ctHPVDNA as follows: pooled sensitivity and specificity of 0.90 (95% CI: 0.82–0.94) and 0.94 (95% CI: 0.85–0.98), respectively; positive and negative likelihood ratios of 12.6 (95% CI: 4.9–32.1) and 0.05 (95% CI: 0.02–0.13), respectively. ddPCR for ctHPVDNA has good accuracy, sensitivity, and specificity for diagnosis of HPV + OPSCC. ctHPVDNA kinetic represents a great reliable opportunity to improve diagnostic and therapeutic management of cancer patients and could open new perspectives for understanding tumor biology.

0.05 (95% CI: 0.02-0.13),respectively.ddPCR for ctHPVDNA has good accuracy, sensitivity, and specificity for diagnosis of HPV + OPSCC.ctHPVDNA kinetic represents a great reliable opportunity to improve diagnostic and therapeutic management of cancer patients and could open new perspectives for understanding tumor biology.
circulating tumor HPV-DNA, digital drop PRC, head and neck cancer, liquid biopsy, OPSCC, oropharyngeal squamous cell carcinoma

| INTRODUCTION
Human papillomaviruses (HPVs) are a family of small, non-enveloped DNA viruses that infect skin and mucosal epithelia.HPV genome consists of a double-stranded circular DNA genome approximately 8000 base pairs (bp) in length, enclosed in an icosahedral capsids of 72 pentamers. 1HPVs infect mitotically active basal keratinocytes in epidermis or mucosal epithelia, and their life cycle is closely related to cell differentiation within the squamous epithelium.][4] To date, 225 HPV genotypes have been fully characterized based on sequence information collected by the International HPV Reference Center. 5Thirteen of >200 HPVs are HR and are associated with human cancer, including ano-genital and oropharyngeal cancers. 6mong HR HPVs, HPV16 is the most common and best studied type of HPV and serves as a paradigm for all HPVs.Oropharyngeal squamous cell carcinoma (OPSCC) has one of the fastest growing incidences in high-income countries. 7,8In United Kingdom and United States, incidence of man HPV + OPSCC is exceeding that of woman cervical cancer. 9HPV positive and HPV negative OPSCC are getting higher in the last 10 years, and epidemiological data suggests that the HPV + OPSCC is increasing quicker. 10,11Worldwide, the proportion of HPV+ HNSCC was 33% in 2021, of these, 85%-96% were HPV-16 associated. 10National Comprehensive Cancer Network (NCCN) guidelines recommend use of tissue biopsy for detection of p16 IHC, a surrogate marker of HR HPV infection, for the diagnosis of HPV + OPSCC.Current present diagnostic procedures are invasive, costly and user dependent. 124][15][16][17] Moreover, p16 IHC on FNA-C specimens is hardly detectable and consensus between pathologists and institutions is lacking.Thus, often tissue biopsy in operation theater is required to prove the diagnosis of HPV + OPSCC, causing additional invasive procedure, diagnostic delays and cost increase. 18Despite cell free DNA (cfDNA) is a promising biomarker for diagnosis and monitoring of cancer treatment as well as for detection of residual disease and recurrences, [19][20][21] its use is limited by increasing challenges linked to sensitivity (detection of somatic cfDNA alterations against a normal DNA background) and specificity (cfDNA alterations are not intrinsically specific for a given type of cancer). 22,23][26] The possibility of detecting ctHPVDNA in plasma in patients with HPV + OPSCC has been demonstrated in several reports.However, this data is from small cohorts and study protocols and methodology differ between research groups.Meta-analyses can overcome these problems in studies of rare cancers such as HPV + OPSCC and are more reliable than single case-control studies.Thus, the aim of this meta-analysis is to evaluate sensitivity, specificity, and accuracy of ctHPVDNA by ddPCR to define its efficacy in the clinical setting for the diagnosis of HPV + OPSCC.

| MATERIALS AND METHODS
Systematic review and meta-analysis were conducted following the PRISMA guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analyses).

| Study eligibility criteria
The eligibility criteria for studies included in this meta-analysis were designed according to the Population Intervention Comparators Outcomes Study (PICO) framework.The population included patients with HPV + OPSCC.The intervention was the use of Liquid biopsy (LB) for ctHPVDNA detection by droplet digital-PCR (ddPCR) at first diagnosis.The comparator test was a negative result of ctHPVDNA detection at first diagnosis in HPV-negative HNSCC carcinoma or control group.The outcome was to establish the value of ctHPVDNA as diagnostic biomarker at first diagnosis.Peer-reviewed publications in English were included, with no restrictions to the publication year.

| Search strategy
Authors conducted a literature search on articles published until February 2023 using three different databases: MEDLINE, Embase, and Cochrane Library searching for studies examining the diagnostic performance of ctHPVDNA at the diagnosis in patients with HPV + OPSCC (Table S1).
The articles were surveyed applying the selection criteria on the title and abstract (phase 1) and then on the full text of those deemed appropriate after the first analysis (phase 2).In addition, a manual search was conducted for references from the selected studies.Duplicate abstracts were carefully removed.A flow chart illustrating all steps of the screening procedure is shown in Figure 1.

| Data extraction
A standardized electronic data collection form was used independently by two reviewers (FC, GD) to extract the data from each of the included studies such as the first

| Statistical analysis
The hierarchic summary ROC curve (HSROC) analysis was employed to test the diagnostic performance, based on a regression model that accounts for within-and betweenstudy variability.The random effects meta-analysis was carried out using the restricted maximum likelihood method.
The pooled sensitivity and specificity, the diagnostic odds ratio (DOR), positive and negative likelihood ratios were calculated.Results were reported with a 95% confidence interval (CI) for all the analyses.Moreover, care was taken to apply a minimal correction factor of 0.1 for "0" events, as it has been already shown to reduce the distortion of data for excessive correction. 27A subgroup meta-analysis was also executed dividing the studies in three groups according to the probe used, E6-E7 HPV-DNA, E6 HPV-DNA, E7 HPV-DNA.In the hypothesis that small samples (60 patients or less) could lead to biased results, another subgroup meta-analysis was conducted dividing the studies in <60 and >60 patients studies.All the analyses were performed using R software for statistical computing (R 2.10.1;"meta" and "mada" package).

| Risk of bias
The Quality Assessment of Diagnostic Accuracy Studies second edition (QUADAS-2) was applied to calculate the potential risk of bias and quality of included studies. 28he seven items of QUADAS-2 checklist were scored in all included articles.The risk of bias was rated high (H), low (L), or unclear (U) according to the QUADAS-2.

| Study selection
The preliminary search, according to the scheme defined, led to the identification of 311 articles.After the removal of duplicates, 194 articles were detected.

| Probes for ctHPVDNA detection by ddPCR
All studies used droplet digital PCR (ddPCR).Twelve of the studies extracted ctHPVDNA from plasma, and one extracted ctHPVDNA from serum (Table 1).Primers/ probes used by these studies were different, as showed in Table 1.In addition, analyses were performed to detect only E7 oncogene (mostly from HPV16) in 5 studies, only E6 in 4 and E6-E7 together in 4, respectively.

| ctHPVDNA meta-analysis for diagnostic accuracy in HPV + OPSCC
A total of 998 patients were evaluated from the 13 studies included in the meta-analysis.Seven hundred and twenty-nine patients had biopsy-proven HPV + OPSCC using p16 as a surrogate whereas 269 controls were p16 negative.The main clinical characteristics of the included studies are summarized in Table 1.

| Qualitative assessment
Quality assessment based on the QUADAS-2 is shown in Table S2, and the overall risk of bias was rated low.Included studies fulfilled the items "patient selection," "index test," "reference standard," and "flow and timing" of the risk of the bias section and all three items of the applicability concerns section ("patient selection," "index test," and "reference standard").Previous meta-analyses on the diagnostic accuracy of quantitative cfDNA testing demonstrated that this technology is better than conventional biomarkers for several cancer diagnoses including gastric cancer, 42 lung cancer, 43 thyroid cancer, 44 HN cancer, 45 and colon cancer. 46o the best our knowledge, this is the first metaanalysis exploring the accuracy of ctHPVDNA by ddPCR in patients at diagnosis of HPV + OPSCC.
This meta-analysis combined outcomes from 729 HPV + OPSCC patients and 269 p16À controls.Overall test performance demonstrated that ddPCR for ctHPVDNA is a sensitive, specific, and accurate test for diagnosis of HPV + OPSCC.The meta-analytic study estimated the diagnostic performance of ctHPVDNA as follows: pooled sensitivity and specificity of 0.90 (95% CI: 0.82-0.94)and 0.94 (95% CI: 0.85-0.98)(Figure 1), respectively.Good diagnostic tests have positive likelihood ratio (LR) LR+ >10 and their positive result has a significant contribution to the diagnosis.Good diagnostic tests have LRÀ <0,1. 47Our meta-analysis shows positive and negative LR of 12.6 (95% CI: 4.9-32.1)and 0.05 (95% CI: 0.02-0.13),respectively.The diagnostic odds ratio (DOR) gives a general estimate of diagnostic accuracy. 48A value above 200 is generally accepted as a good diagnostic test, from our analysis a DOR of 353 was calculated.The area under the curve averages diagnostic accuracy of the ctHPVDNA.A perfect diagnostic test has an AUC of 1 whereas a nondiscriminating test has an area 0.5. 49Generally, we can say that AUC between 0.9 and 1 is excellent, while a value between 0.8 and 0.89 is very good.Our meta-analysis shows AUC of 0.81, which corresponds to a "very good" diagnostic test.All together the results that come out from the meta-analysis demonstrated that may should consider ctHPVDNA test applicable in clinical practice.
Subgroup analysis was performed for primers/probes for E6 versus E6/E7 and versus E7.In addition, another subgroup analysis was performed for cohort size: small cohort with less than 60 patients vs large cohort with more than 60 patients.However, statistical regression data showed that these subgroup differences were not statistically significant.
Taken together, these results indicated that study design did not significantly affect diagnostic accuracy.
Liquid biopsy is the sampling and analysis of nonsolid biological tissue, such as urine, saliva, cerebrospinal fluid, and blood.To date, blood sampling is the most widely used type of liquid biopsy. 50In liquid biopsy, circulating viral DNA is easier to be identified than genomic DNA.Viral DNA has greater stability than mutated tumor DNA and this stability makes virus-related tumors unique among solid tumors. 51,52The study of circulating DNA kinetics represents a great opportunity to improve the diagnostic and therapeutic protocol of cancer patients and opens new perspectives for understanding tumor biology. 53For non-keratinizing or undifferentiated histology of nasopharynx, NCCN guidelines suggest considering testing for EBV in tumor and blood.EBV DNA burden in serum or plasma can be quantified using polymerase chain reaction (PCR) targeting EBV DNA genomic sequences.5][56] Thus, it is reasonable that ctHPVDNA could have the same diagnostic/ prognostic efficacy as ctEBVDNA. 57,58The opportunity of detecting ctHPVDNA in plasma in patients with HPV + OPSCC has prompted several research groups to focus their studies on this promising biomarker.Studies that used ddPCR in diagnosis of HPV + OPSCC were included in this meta-analysis.The analysis highlighted the accuracy of the method, but at the same time the lack of standardization and validation of the test.The choice to include only studies that used ddPCR derives from the evidence that it is the most appropriate technology in clinical setting. 59,60Real-time quantitative PCR (q-PCR) is the simplest and cheapest technology to detect HPV DNA.However, sensitivity of qPCR is limited by background fluorescent dye and need of a standard curve to quantify HPV copy number.These limitations are more consisting with a low HPV copy number because the signal-to-noise ratio is not large enough to distinguish one or a few copies of HPV DNA from false positives.Instead, Next Generation Sequencing (NGS) and ddPCR were equally more sensitive than q-PCR in detecting HPV DNA in plasma.In addition, ddPCR is more cost effective than NGS when quantification of few targets is desired. 35n clinical setting, the use of ctHPVDNA during the diagnostic workup could improve accuracy, reduce costs and shorten diagnostic time. 61In a prospective study, Siravegna et al. 38 compared standard clinical approach versus imaging/physical exam findings associated to ctHPVDNA liquid biopsy.Although in limited number of patients (70 cases and 70 controls) the study showed that the diagnostic accuracy of this noninvasive approach was significantly better than standard of care (Youden index 0.937 vs. 0.707, P 1/4 0.0006), costs were 36%-38% lower, and the median diagnostic interval was 26 days shorter.
Despite meta-analysis indicated that study design and primers/probes did not affect the diagnostic accuracy, the ctHPVDNA protocol needs to be standardized.It is necessary to optimize the path from blood sampling during the first visit to the molecular analysis in laboratory.
ctHPVDNA is an extremely interesting biomarker because, in addition to modifying the diagnostic workup, it could also modify the therapeutic workup.
An attractive application of ctHPVDNA is target follow up.Monitoring kinetic of ctHPVDNA could achieve the goal of fewer follow-up visits. 62owever, HPV + OPSCCs are rare and demonstrating the utility of ctHPVDNA from small case-control studies is difficult.It is not easy for a single center to obtain significant data on rare pathologies.
For this reason, it is desirable to promote large multi-center studies that promote a unique protocol to standardize the times of sampling and the methods of extraction and analysis of ctHPVDNA.

| Limitations
It must be pointed out that this meta-analysis has some limitations.Low number of studies and low sample sizes limit the generalizability of results.Despite meta-analysis indicated that study design and primers/probes did not affect the diagnostic accuracy, heterogeneity may be related to other factors such as patients age, tumor site, TNM stage and differences in experimental protocols (i.e., time of blood sample processing), which could not be analyzed in the current study because of loss of data or unrecognizable details.Finally, only four of the included articles used an HPV negative HNSCC as control.

| Conclusion
In conclusion, this meta-analysis demonstrated that ddPCR for ctHPVDNA has good accuracy, sensitivity and specificity for the diagnosis of HPV + OPSCC.ctHPVDNA represents a great reliable opportunity to improve diagnostic and therapeutic management of cancer patients, and could open new perspectives for understanding tumor biology.The ctHPVDNA protocol needs to be standardized.It is necessary to optimize the path from blood sampling during the first visit to the molecular analysis in laboratory.In next years, studies on larger and detailed patients' cohorts and continued improvements in assay methodology and technology could allow ctHPVDNA in routine clinical use.

1
Diagnostic accuracy of ctHPVDNA displayed by forest plots estimating (A) sensitivity, (B) specify at the diagnosis in patients with HPV + OPSCC (confidence interval [CI] in brackets).[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 1 Main features of the included studies.
Diagnostic accuracy of ctHPVDNA displayed by forest plots estimating positive likelihood ratio (PLR), and negative likelihood ratio (NLR) at the diagnosis in patients with HPV + OPSCC (confidence interval [CI] in brackets).[Color figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 3 Diagnostic accuracy of ctHPVDNA displayed by forest plots estimating diagnostic odds ratio (DOR) at the diagnosis in patients with HPV + OPSCC (confidence interval [CI] in brackets).[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 4 Diagnostic accuracy of ctHPVDNA displayed by forest plots estimating (A) sensitivity, (B) specify at the diagnosis in patients with HPV + OPSCC (confidence interval [CI] in brackets) among subgroups (> or <60 patients).[Color figure can be viewed at wileyonlinelibrary.com] Sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV).'s name, year of publication, study design, country, number of patients, cancer site, HPV status of cancer, number of pretreatment blood tests, HPV status in blood and method for the detection of viral DNA.The extracted outcomes about the diagnostic accuracy of ctHPVDNA as a detection test for disease progression in patients affected by HPV + OPSCC were the number of true positives, false positives, true negatives, and false negatives.
T A B L E 2 author