HPV and lung cancer: A systematic review and meta‐analysis

Abstract Background Lung cancer has emerged as a global public health problem and is the most common cause of cancer deaths by absolute cases globally. Besides tobacco, smoke infectious diseases such as human papillomavirus (HPV) might be involved in the pathogenesis of lung cancer. However, data are inconsistent due to differences in study design and HPV detection methods. Aim A systematic meta‐analysis was performed to examine the presence of HPV‐infection with lung cancer. Methods and Results All studies in all languages were considered for the search concepts “lung cancer” and “HPV” if data specific to HPV prevalence in lung cancer tissue were given. This included Journal articles as well as abstracts and conference reports. As detection method, only HPV PCR results from fresh frozen and paraffin‐embedded tissue were included. Five bibliographic databases and three registers of clinical trials including MEDLINE, Embase, Cochrane Library, and ClinicalTrials.gov were searched through February 2020. A total 4298 publications were identified, and 78 publications were selected, resulting in 9385 included lung cancer patients. A meta‐analysis of 15 case‐control studies with n = 2504 patients showed a weighted overall prevalence difference of 22% (95% CI: 12%‐33%; P < .001) and a weighted overall 4.7‐fold (95% CI: 2.7‐8.4; P < .001) increase of HPV prevalence in lung cancer patients compared to controls. Overall, HPV prevalence amounted to 13.5% being highest in Asia (16.6%), followed by America (12.8%), and Europe (7.0%). A higher HPV prevalence was found in squamous cell carcinoma (17.9%) compared to adenocarcinoma (P < .01) with significant differences in geographic patterns. HPV genotypes 16 and 18 were the most prevalent high‐risk genotypes identified. Conclusion In conclusion, our review provides convincing evidence that HPV infection increases the risk of developing lung cancer.


| INTRODUCTION
Lung cancer is estimated to be the leading cause of cancer-related mortality worldwide, with 2.1 million new lung cancer cases and 1.8 million predicted deaths worldwide in 2018. 1 Although smoking by far has been identified as the most important risk factor in lung cancer, other interactions with environmental and/or genetic risk factors as well as infectious diseases have been identified to contribute to the pathogenesis of lung cancer as well.
Viral infections, such as human papillomavirus (HPV) infections have been reported to be an important risk factor of cervical cancer if genotypes with a high oncogenic risk are found. Since the first identification of human papillomavirus, more than 200 different subtypes have been identified They are classified into high-risk HPV types (16, 18, 31, 33, 39, 45, 51, 52, and 58) and low-risk HPV types (6, 11, 42, 43, and 44). 2 In some other publications, a differentiation between high-, intermediate-, and low-risk HPV types can be found. 3 Although HPV infection has been identified as a potential contributor to the pathogenesis in lung cancer in certain populations, such as never smokers, its role still remains controversial. Numerous tests, such as nucleic acid amplification, HPV DNA-based in situ hybridization, immunohistochemistry, and cytology are available for HPV-testing and screening. 4,5 The current study focused on the prevalence of HPV infections in lung cancer patients in which HPV detection was performed by means of PCR from fresh frozen and/or paraffin-embedded tissue to first minimize differences in HPV prevalence due to methodological bias and second to rely on the method with the highest sensitivity to detect HPV positivity, which has been proven to have the highest sensitivity in earlier studies. 4,5 We conducted and report here a systematic review on the issue above.

| METHODS
The methods of the systematic review and meta-analysis were specified in advance and published in a protocol registered with PROSPERO. Reporting of this meta-analysis was done according to the recommendation of Stroup et al for reporting observational studies. 6

| Evidence search and meta-analysis
The digital databases Embase (via Ovid, 1974-present) were searched for completed studies. All searches were last updated on February 6, 2020. We deviated from the protocol; in that, we did not search the German Clinical Trials Register due to its search interface giving erroneous results. An initial, sensitive search strategy for the concepts "lung cancer" AND "HPV" was developed for Embase by a medical librarian in cooperation with subject matter experts and then adapted to the other databases. Controlled terms from the databases' thesauri and a broad range of synonyms were used. No limits such as for study type, publication type, publication date, or language were applied. Search strategies that allow for reproducing the searches are documented in Appendix 1. Database searches were carried out by a medical librarian. The reference lists of included studies and of relevant systematic reviews were screened for additional studies. Records from the database searches were imported into Endnote software for deduplication. Screening by title and abstract and subsequent full-text assessment were done in Covidence. Titles and abstracts of the publications were analyzed by three independent reviewers (F.K., J.K., and C.S.) for relevance and matching inclusion criteria. Analysis of the publications was done according to prespecified inclusion and exclusion criteria.
All studies reporting HPV prevalence in primary lung cancer cases in adults were included. Case reports were excluded. As detection method, only PCR from fresh frozen and/or paraffinembedded tissue were included. All types of tissue sampling method were included. HPV detection in archival tumor tissue was included as well. Only studies that provide data specific to HPV prevalence in lung cancer tissue were included. No exclusions were made based on language. Journal articles as well as abstracts and conference reports were included if they met the inclusion criteria. Journal articles that reported about not only cases of HPV detection in primary lung cancer but, for example, in head and neck cancer as well, were included but only the data of the primary lung cancer group were extracted.

| Statistical analysis
The total number of cases, as well as the number of positive and negative HPV detections, was collected from the selected records, and HPV prevalences were calculated by means of the extracted patient data. The Chi-squared-test of independence was used to analyze whether prevalence rates differ between continents. Furthermore, a meta-analysis was performed on a small subset of case-control studies regarding HPV prevalence. Prevalence difference (PD) and prevalence ratio (PR) both accompanied with the corresponding 95% confidence intervals were estimated for each study. To estimate PR in studies with no HPV positive cases, 0.5 was added to each cell of the 2 × 2 table as usually recommended. Random-effect models were used to determine the weighted averages of PD and PR while allowing for heterogeneity of effects. The Q-statistic as a measurement for between-study heterogeneity and I 2 -statistic for quantification of the proportion of total variation due to heterogeneity were calculated.
Analyses were performed using R version 4.0.3 (The R Foundation for Statistical Computing), the meta-analysis by using the metafor package. For all comparisons, a P value <.05 was considered as statistically significant.

| Evidence Search
The database searches were last updated on February 6, 2020 and yielded a total of 4525 records. Following deduplication, 3135 publications were evaluated on relevance for the research question. A total of 2754 of the titles and abstracts did not relate to the current research and were excluded. In summary, 381 publications were entered into the full text review. Full texts of three possibly relevant publications could not be obtained despite some efforts and therefore were not available 7-9 for further analyses. The remaining 378 fulltexts were assessed for eligibility. After applying the inclusion and exclusion criteria, 78 publications were included in this systematic review. Reasons for exclusion were as follows: No PCR data were reported (n = 80). HPV detection method was not detailed (n = 2).
Duplication of the data (n = 22). Case reports (n = 9). Corrections and/ or comments on screened publications (n = 15). Systematic reviews and meta-analysis (n = 29). Overview articles (n = 29). HPV detection was not done in lung biopsies (n = 32). HPV prevalence analyzed in cancers other than lung cancer or on metastasis (n = 6). Missing data on HPV prevalence (n = 40). Same patients in separate publications (n = 7). Same information in different languages (n = 4). Abstract published in a different journal than the full text (n = 12). HPV prevalence in lung cancer in special patient groups, for example, patients after lung transplantation, immunocompromised patients, butchers, and respiratory papillomatosis (n = 7). Unfinished studies (n = 4). No data on sampling method were provided (n = 2). This review process was performed according to the PRISMA statement. Figure 1 depicts the flow of citations reviewed for the meta-analysis.
A total of 15 publications were case-control studies, in which normal lung tissue was used as a control (see Table 1).
The studies were stratified according to the geographical region The other publications were published in Chinese (n = 3), French (n = 1), and German (n = 1). In order to get information on as many cases as possible not only journal articles but every type of available The rate of smokers was 64.6% and ranged from 0% to 100%.

| Meta-analysis of 15 case-control studies
A total of 1750 lung cancer cases and 754 controls were analyzed, which were derived from 15 case-control studies ( Table 1). One of them is from America, 10 are from Asia, and four from Europe. The overall HPV prevalence was detected to be 31.3% (548/1750) in the lung cancer group and 5.5% (42/754) in the control group (P < .001). Figure 2 shows the HPV prevalence derived from case-control studies as well as divided by different continents. Comparing HPV prevalence of patients with lung cancer and controls in a meta-analysis, using the 15 casecontrol studies with a total of 2504 patients, a higher prevalence could be found for the lung cancer patients for prevalence difference (PD = 0.22; 95%-CI, 0.12-0.33; P < .001) as well as prevalence ratio (PR = 4.7; 95% CI, 2.7-8.4; P < .001). A forest plot summarizing the data and the effect estimates is shown in Figure 3. Due to the large confidence intervals of the PRs, only PDs are presented graphically.
According to the Q-statistic, a significant difference in between-study heterogeneity could be identified

| HPV prevalence
Of all included patients with lung cancer (n = 9385), HPV was detected to be positive in 1268 cases. The overall HPV prevalence was

Study
Carpagna (2011) Cheng (2004) Cheng (2001) Eberlein (1992) Fan (2015) Galvan (2012) Gatta (2012) Li (1995) Lu (2016) Nadji (2007) Robinson (2016) Wang (2008) Wang (2010) Yu (2015) Zhang ( F I G U R E 4 Overall HPV, HPV 16, and HPV 18 prevalence in all analyzed lung cancer cases and between analyzed continents. The highest HPV prevalence was detected in Asia followed by The Americas and Europe. Overall and on all three continents the prevalence of HPV 16 was significantly higher than for HPV 18. The highest HPV 16 prevalence was detected in The Americas followed by Asia and Europe. The highest HPV 18 prevalence was found in Asia followed by The Americas and finally Europe F I G U R E 5 HPV prevalence in SCC vs AC. There was no statistically significant difference between the HPV prevalence in SCC and AC in the studies from America (P = .78). Statistically significant differences were found in studies from Asia (P < .01) and Europe (P < .01). On a global observation HPV prevalence in SCC was significantly higher (P < .01) when compared to AC

| Histology and HPV prevalence
Only the information on primary squamous cell carcinoma (SCC) and primary adeno carcinoma (AC) of the lung was collected. In the remaining cases, it was neither one of them or the histological subtype was not detailed. There were 2750 cases of SCC and 2887 cases of AC. In total, 29.3% of the included cases were squamous cell carcinomas and 30.8% were adenocarcinomas.
The overall HPV prevalence in adenocarcinomas (n = 265) was calculated to be 9.2%. In contrast, the highest HPV prevalence in AC was calculated in the Americas (11.1%), followed by Asia (10.4%), and Europe (6.0%).
When the HPV prevalences of SCC and AC are compared, the difference is statistically highly significant (P < .01), which is due to a significantly higher HPV prevalence in SCC (P < .01) in Asia, whereas no differences in prevalence were found in The Americas and Europe based on histological subtypes of lung cancer. Figure 5 shows the calculated HPV prevalences.

| DISCUSSION
Growing evidence supports the association between HPV-infection and lung cancer but the relationship is still debatable. The aim of derived from cohort studies or nested case-control studies are given.
In addition, cofounders of possible importance such as smoking status, gender, age, immunosuppressive co-medications, oncogenic driver mutations, and estrogenic signaling pathways have not been taken into considerations, which limit the value of the results reported. Furthermore, not all HPV subtypes were assessed due to missing specification in many studies, and no transcriptional activity of the HPV genotypes found was included in the meta-analysis. Since only PCR was included as HPV detection method but this not being the only way to detect HPV, which can potentially bias the study's results further.
In conclusion, our systematic review provides evidence that HPV infection might increase the risk of developing lung cancer. Whereby relevant regional differences with respect to prevalence and histological subtypes were found with a predominance of squamous cell carci-

ETHICAL STATEMENT
Not applicable.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available in