|Home | About | Journals | Submit | Contact Us | Français|
Lung cancer is the leading cause of cancer mortality worldwide. A possible carcinogenic role of human papillomavirus (HPV) has been investigated for >20 years and has major clinical and public health implications. We performed a meta-analysis to assess the prevalence of HPV16 and HPV18 in primary lung cancers (2435 subjects from 37 published studies). The overall HPV prevalence ranged from 0.0 to 78.3% with large heterogeneity across geographic regions and histological tissue types. A higher proportion, 50% (7/14), of the European studies reported low or no HPV prevalence (0–10%) compared with the Asian studies, 22% (4/18). When the analysis was limited to HPV16 and HPV18 prevalence, a higher prevalence in Asia (HPV16=11.6% and HPV18=8.8%) than in Europe (HPV16=3.5% and HPV18=3.6%) was observed. Studies using HPV-specific primers resulted in higher prevalence rates than consensus HPV primers (HPV16: Asia=13% and Europe=6%; HPV18: Asia=13% and Europe=5%). Further studies are needed to elucidate the role of HPV in lung carcinogenesis with careful thought given to study design and laboratory detection methods for a more accurate assessment of HPV status in lung tumors.
Lung cancer is the leading cause of cancer-related mortality worldwide. Although cigarette smoking is the most important risk factor responsible for 90% of lung cancer cases, <20% of cigarette smokers develop lung cancer. Other factors including occupational or environmental exposure to radon and asbestos, certain metals, air pollution, coal smoke, hormones, genetic susceptibility and infection with oncogenic types of human papillomavirus (HPV) have been implicated in lung carcinogenesis (1,2). While the role of HPV in the causation of cervical carcinoma is well established, it has been estimated that 15–20% of all human cancers could be related to oncogenic HPVs, and HPV has been implicated as a risk factor for a subset of head and neck squamous cell carcinomas (3–5). The role of HPV as a cocarcinogen in the pathogenesis of lung cancer has been investigated for >20 years through both clinical and laboratory research.
There appears to be a wide variation in the prevalence of HPV infection in lung cancer around the globe. A review by Syrjänen reported HPV DNA in 21.7% of the 2468 lung cancers analyzed; the HPV status varied from 0 to 100% and that HPV16 was the most common type detected (6). Similarly, a more recent systematic review of HPV in lung cancer was performed in 2007 (7) and reported that the mean incidence of HPV was 24.5%. The article reported considerable heterogeneity between studies. However, the quality of studies and reasons for the heterogeneity was not evaluated. Because of the major clinical impact of lung cancer worldwide and the vast clinical and public health implications of a possible role of HPV in bronchial carcinogenesis, we have performed a meta-analysis on an updated literature review. For the first time, the quality of these studies was evaluated in order to explore the reasons for the heterogeneity observed between studies and we have assessed the prevalence of HPV16 and HPV18 in primary lung cancers.
A comprehensive search of the MEDLINE database using the PUBMED search engine was carried out to identify studies that assessed the presence of HPV infection in human subjects with primary lung cancers published up to 24 January 2009. The keywords used for the search were (HPV OR human papillomavirus) AND (lung OR bronchogenic) AND (cancer OR carcinoma).
Studies that were eligible to be included in the meta-analysis included those that tested for presence of HPV in the lung tissue of patients diagnosed by histopathology to have primary lung cancer. In studies that also assessed HPV in non-lung primary cancers or lung metastases of non-lung primary cancers, only the primary lung cancer data were abstracted and used for the meta-analysis. Among studies that assessed other molecular markers (besides HPV DNA) in lung cancer, only relevant information pertaining to HPV was abstracted. Published English language and foreign language articles, of any study design, were included for this meta-analysis.
In review articles, studies performed on animal models and/human cell lines, case reports and studies related to lung involvement in recurrent respiratory papillomatosis were excluded from the meta-analysis. Studies done primarily on benign lesions of the lung as well as studies on patients with preexisting conditions such as cervical intraepithelial neoplasia and human immunodeficiency virus infection were also excluded from the meta-analysis.
The bibliographic search yielded 382 articles. After a manual review, 55 articles assessing HPV status in primary human lung cancers were identified. A manual review of the references of relevant publications was also performed to identify additional studies. A review of the bibliography of two review articles yielded one additional article for a total of 56 publications. From this list, nine articles were excluded as they contained overlapping data or the same data were published in English and a foreign language. Among the articles that had overlapping data, the article with the larger number of subjects and/or the most recent article and/or the one that contained more complete information was used. For one article (8), the data were combined with another, published by the same author/group (9) and counted as a single study. For one article (10), only the data from the Niigata region of Japan were used since the data for the Okinawa region of Japan published in this study were overlapping with the study by Miyagi et al. (9). One article reported data for HPV16 and HPV18 (11), whereas data for HPV6 and HPV11 were reported in a subset of the same population in another article (12). Both articles were counted as a single study.
Polymerase chain reaction (PCR) is the gold standard for HPV detection. For the studies to be more comparable, only those that utilized PCR as the primary detection method were included in this analysis. Therefore, 10 additional studies were excluded because the predominant test method was in situ hybridization (13–18), immunohistochemistry (19,20), Hybrid Capture II or Southern blot (21,22).
Overall a total of 37 articles (35 studies) were included in the meta-analysis. There were three Chinese and one Japanese language articles with English language abstracts. For one study, the corresponding author was contacted to obtain more detailed information (23) and the data provided through personal correspondence were used for the meta-analysis.
For each study, the data extracted included author and year of publication, country, sample size, sample type, all the techniques used to detect HPV, type of HPV types detected and the histological types of lung carcinoma. The data related to bronchioloalveolar carcinoma (BAC) were abstracted as a separate category as BAC has a unique biology compared with other subtypes of non-small cell lung carcinoma and a unique clinical and radiological presentation with a different response to systemic treatment compared with conventional lung adenocarcinoma (24). Cancer stage, gender, age and smoking status of the subjects were also abstracted whenever available and the type of consensus primer or the type of initial set of primers used was also recorded. Studies where HPV DNA was identified in the subjects, but could not be genotyped either because of use of limited number of type-specific primers or an unknown type of HPV, the HPV infection was classified as HPV X.
Data relating to HPV prevalence in specific histological subtypes of lung cancer as well as data related to the prevalence of specific HPV types were abstracted from only a subset for articles since all did not provide this detailed information.
A limited number of studies reported HPV status in lung cancer cases as well as controls. Furthermore, the type of controls included in each study varied between tissue from cancer-free controls and healthy non-cancer lung tissue from cancer cases. Therefore, odds ratios could not be calculated. The proportion of primary lung cancers that tested positive for HPV was calculated and reported. The Q statistics were calculated to test for heterogeneity among the studies included in this analysis (25) and were considered statistically significant when the P-value was <0.10. The I2 statistic was also used to assess the extent of between-study heterogeneity (26,27). I2 values that are 50% or higher indicate large between-study heterogeneity, whereas values of 25–50% indicate moderate between-study heterogeneity. The between-study heterogeneity was explored by conducting subgroup analyses based on stratification by various study and patient characteristics. An influence analysis was also performed by removing one study at a time to identify individual studies that might contribute to the heterogeneity findings. The calculations for the confidence intervals (CIs) for I2 statistics and two-sample proportion tests were performed using STATA SE (version 10) software (StataCorp LP, College Station, TX). Meta-estimates were calculated when heterogeneity was not observed and publication bias was assessed using the Comprehensive Meta Analysis Version 2 software (28).
Of the 37 data sets (35 articles) included in this analysis (Table I), 18 studies were carried out in Asian countries (8–12,23,29–42), 14 in European countries (43–56) and 1 in North America and two in South America. Overall, there were 2435 primary lung cancer cases included, 1378 Asian patients, 918 Europeans and 139 North and South Americans. The number of cases included in each study varied from 8 to 313 patients, with an average of 70 cases per study.
The prevalence of HPV infection ranged from 0.0 to 78.3% with considerable variation for both Asia and Europe. Among the Asian studies, the prevalence varied in China from 11.8 to 55.0% and in Japan from 0.0 to 78.3%. Only one study each was conducted in Korea, Iran and India; the HPV prevalence was 45.5, 25.6 and 5.0%, respectively. Among the European studies, the HPV prevalence ranged from 0.0 to 69.2%; the prevalence in the only study conducted in North America was 5.9% and was lower than that of the two studies conducted in South America (27.8 and 29.0%) (Table I). A higher proportion 50% (7/14) of the European studies reported low or no HPV prevalence (0–10%) compared with the Asian studies 22% (4/18).
The type of lung specimens used to test HPV status were primarily either paraffin-embedded or fresh-frozen tissue. There was significant variability with respect to the type of HPV primers used (summarized in Table I), each of which amplified a different size of HPV target sequence. For all studies combined, the prevalence of HPV was 20.7% and was higher than what was observed in the 10 excluded studies that utilized non-PCR methods for HPV detection, where the prevalence was 11.9% (data not shown). Overall, there was large heterogeneity observed between the studies included in this analysis (PCR: Q-test P-value <0.0001, I2=91%, 95% CI=88–93%) with evidence of publication bias (Egger's test P-value <0.0001).
Similarly, large heterogeneity was observed between studies in Asia and Europe (Asia: Q-test P-value <0.001, I2=89%, 95% CI thinsp;=85–93%; Europe: Q-test P-value <0.001, I2=89%, 95% CI=83–93%) and there was evidence of publication bias for these studies (Asia: Egger's test P-value=0.029; Europe: Egger's test P-value=0.028). For the South American studies, no heterogeneity was observed between the two studies (Q-test P-value=0.897) and for this reason, we were unable to assess publication bias. When the three American studies were combined, we observed large heterogeneity (Q-test P-value=0.05, I2=67%, 95% CI=0–90%) with no evidence of publication bias (Egger's test P-value=0.217).
The majority of studies included all histological subtypes of lung carcinoma, whereas others restricted their study to one or more specific histological subtypes. The histology-specific HPV prevalence was collected by all but four studies (34,41–43). Thirty studies provided HPV data for squamous cell carcinomas, 22 for adenocarcinoma, 5 for adenosquamous carcinoma, 15 for large cell carcinoma, 6 for BAC and 13 for small cell carcinoma. For squamous cell carcinomas, the HPV prevalence ranged from 0.0 to 48.1%, whereas the prevalence ranged from 0.0 to 44.4% for adenocarcinoma, 0.0 to 100.0% for adenosquamous carcinoma, 0.0 to 75.0% for large cell carcinoma, 0.0 to 50.0% for BAC and 0.0 to 100.0% for small cell carcinoma.
There were 16 studies in Asia and 14 studies in Europe that reported data for HPV16 in lung tumors. Among them, different HPV primers were used to detect HPV16. Overall, 14 studies used HPV16-specific primers, 13 used consensus HPV primers and for three studies, the type of PCR primers used was not reported (15,16,18). The meta-estimate for overall HPV16 prevalence in Asia and Europe was 7.1% (95% CI=4.7–10.6%) and large heterogeneity between the studies was observed (Q-test P-value <0.0001; I2=79%, 95% CI=70–85). For the studies that used type-specific HPV primers, all but two used E6 and/or E7 type-specific PCR primers, whereas one study used p16-1 and p16-2R primers (40) and the other used E1 primers (55). For the studies that used consensus HPV primers, all but two used L1 consensus HPV primers and two studies used consensus primers that amplified an E6–E7 HPV sequence (38,48).
For the Asian studies only, the meta-estimate for HPV16 was 11.4% (95% CI=7.2–17.5%) and there was large heterogeneity between studies (Q-test P-value <0.0001; I2=78.0%, 95% CI=65.0–86.0). A sensitivity analysis was performed and two studies appeared to influence the overall meta-estimate (11,12,35) and were excluded from all future analyses of HPV16 prevalence in Asia. The resulting overall meta-estimate of HPV16 in Asia was 11.6% (95% CI=9.5–14.2) with modest heterogeneity (Q-test P-value=0.073, I2=38%, 95% CI=0–67) (Figure 1A). When the studies were stratified according to the type of HPV primers used, the meta-estimate for studies that used HPV16 type-specific primers was 13.4% (95% CI=10.5–16.8); for consensus HPV primers, it was 8.6% (95% CI=5.1–14.0), whereas for other PCR-based studies, the meta-estimate was 6.4% (95% CI=3.0–13.1). There was modest heterogeneity between the studies that used consensus HPV primers (Q-test P-value=0.092, I2=58%, 95% CI=0–88), whereas no heterogeneity was observed for studies using HPV16 type-specific primers (Q-test P-value=0.250, I2=23%, 95% CI=0–64) and other PCR methods (Q-test P-value=0.410, I2=0%, 95% CI=0–90).
The meta-estimate for HPV16 in Europe was 3.8% (95% CI=1.9–7.4) and large heterogeneity was observed between studies (Q-test P-value <0.0001, I2=68%, 95% CI=44–82). Sensitivity analysis identified two studies that appeared to influence the meta-estimates (35,45) and, therefore, were subsequently excluded from further analyses. The overall HPV16 meta-estimate was 3.5% (95% CI=2.3–5.3) with no heterogeneity (Q-test P-value=0.208, I2=24%, 95% CI=0–61) (Figure 1B). The meta-estimate for the studies that utilized HPV16-specific primers was 5.6% (95% CI=3.3–9.4) and no heterogeneity was observed (Q-test P-value=0.240, I2=29%, 95% CI=0–74). For studies that used consensus HPV primers, the meta-estimate was 1.6% (95% CI=0.8–3.2) and there was no heterogeneity between these studies (Q-test P-value=0.944, I2=0%, 95% CI=0–68). Publication bias was present for the studies conducted in both Asia and Europe (Asia: Egger's test P-value=0.015, Europe: Egger's test P-value=0.007).
The same studies (16 Asian and 14 European) that reported data for HPV16 also reported data for HPV18 prevalence in lung tumors. Overall, the HPV18 prevalence was 5.6% (95% CI=3.5–9.0%). Similarly, there was large heterogeneity for all these studies combined (Q-test P-value <0.0001, I2=81%, 95% CI=74–87).
For the Asian studies, the meta-estimate for HPV18 was 9.7% (95% CI=5.6–16.3) and there was large heterogeneity (Q-test P-value <0.0001, I2=84%, 95% CI=75–89). Sensitivity analysis identified three studies that appeared to influence the meta-estimate (11,12,26) and these studies were excluded from future analyses of HPV18 prevalence. The resulting meta-estimate of HPV18 prevalence in Asia was 8.8% (95% CI=6.0–12.8), with modest heterogeneity (Q-test P-value=0.029, I2=46%, 95% CI=0–71) (Figure 2A). For studies that used HPV18-specific PCR primers, the meta-estimate was 13.0% (95% CI=10.2–16.4) and no heterogeneity was observed (Q-test P-value=0.213, I2=27%, 95% CI=0–67). For studies that used consensus HPV primers and those that used other PCR primers, the meta-estimates were 7.3% (95% CI=4.3–12.0) and 1.8% (95% CI=0.4–6.9), respectively. No heterogeneity was observed in these two subgroups (consensus HPV primers: Q-test P-value=0.273, I2=23%, 95% CI=0–92; other PCR studies: Q-test P-value=0.779, I2=0%, 95% CI=0–90).
The meta-estimate for HPV18 in Europe was 2.8% (95% CI=1.4–5.6%) and large heterogeneity was observed between these studies as well (Q-test P-value=0.007, I2=55%, 95% CI=18–75). After performing a sensitivity analysis, a single study appeared to influence the meta-estimate and was excluded from further HPV18 analyses (45). The resulting meta-estimate of HPV18 prevalence in Europe was 3.6% (95% CI=2.3–5.7) and heterogeneity was no longer observed (Q-test P-value=0.144, I2=30%, 95% CI=0–64) (Figure 2B). The meta-estimate for studies that used HPV18-specific PCR primers was 5.1% (95% CI=2.9–9.0) and there was no between-study heterogeneity (Q-test P-value=0.173, I2=37%, 95% CI=0–77). For the studies that used consensus HPV primers, the meta-estimate was 1.9% (95% CI=0.8–4.1) and no heterogeneity was observed (Q-test P-value=0.476, I2=0%, 95% CI=0–68). Publication bias was observed for both Asia and Europe (Egger's test P-value 0.002 for Asia and P < 0.0001 for Europe).
There were three studies conducted in the Americas, one in North America (57) and two in South America (58,59). All three studies used consensus HPV primers for HPV detection. HPV16 prevalence was reported in the South American studies only and the overall meta-estimate was 17.2% (95% CI=11.1–25.7). There was no heterogeneity between these studies (Q-test P-value=0.652 and I2=0%); publication bias could not be assessed since there were only two studies. All three studies reported HPV18 prevalence and the meta-estimate was 5.1% (95% CI=2.4–10.3). There was neither heterogeneity between the studies (Q-test P-value= 0.933, I2=0%,95% CI=0–90) nor publication bias (Egger's test P-value =0.097).
A recent systematic review of HPV infection in lung cancer has been performed (7). The study summarized the rates of HPV in lung cancer according to geographic region and reported that there was considerable heterogeneity between the studies. In the present study, we included an updated review of the literature and for the first time, a meta-analysis to formally assess the prevalence of HPV16 and HPV18 in primary lung cancers across geographical regions while exploring the reasons for the variation in HPV prevalence. Furthermore, the present analysis only includes studies that utilized PCR-based methods in order to reduce the heterogeneity between studies that might be attributed to differences in sensitivity of the various HPV detection methods. Consistent with the previous systematic review, a wide variation in the overall prevalence of HPV-positive lung tumors was still observed (0.0–78.3%) with large heterogeneity between studies. Stratification of the studies according to geographic region and histological subtypes did not resolve the heterogeneity between the studies until subset analyses of HPV16 and HPV18 prevalence was performed.
The majority of studies used paraffin-embedded tissue for the determination of HPV status with variation in the types of HPV primers used. The sensitivity of HPV detection varies due to differences in amplification efficiency between different types of HPV primers, especially when utilizing paraffin-embedded tissue samples. Therefore, the observed variability in HPV prevalence between studies was not surprising. Significant DNA degradation is known to occur with paraffin-embedded tissue. Therefore, it may be challenging to perform PCR amplification of long DNA fragments [i.e. >300 bp (60)]. Some of the studies included in the present analysis used consensus HPV primers that target the amplification of a 450 bp segment in the L1 gene and might prove to be less sensitive for detecting HPV sequences in the lung tumor tissues. In contrast, other studies used HPV type-specific primers that are usually designed to amplify shorter sequences of HPV DNA (i.e. <200 bp) and might be more sensitive for detecting HPV DNA sequences. Therefore, it is possible that the detection rate for HPV using HPV type-specific PCR primers may be higher than other PCR methods. This premise is supported by the findings in our study. We performed stratified analyses of HPV16 and HPV18 prevalence according to the type of HPV primers (consensus versus type specific). For the studies in both Asia and Europe, the meta-estimates of HPV16 was higher for the subgroups that used HPV type-specific primers (Asia=13.4% and Europe=5.6%) compared with those studies that used consensus HPV primers (Asia=8.6% and Europe=1.6%). Similar observations were made for HPV18. We were unable to compare these data with the studies conducted in North and South America since all these studies used consensus primers for HPV detection. In general, the meta-estimates for HPV16 and HPV18 were higher for Asia than in Europe except for the HPV16 studies that used consensus HPV primers. The highest HPV16 prevalence in this case was reported in South America (17.2%), rather than Asia (8.6%) and Europe (1.6%) although these findings should be viewed with caution due to the fact that there were only two studies in South America compared with four in Asia and nine in Europe.
Two different publications from Taiwan were published by the same group reported HPV prevalence of 51 and 75%, respectively (11,12,41). Both studies used consensus primers (MY09/MY11) for HPV detection; however, slight differences in the method were reported. Cheng et al. performed a second-round PCR using type-specific primers for HPV16, HPV18, HPV6 and HPV11, whereas Lin et al. used in the second-round PCR, type-specific primers for HPV16 and HPV18 only. This methodological difference may be a possible explanation for the difference in prevalence since Lin et al. may have missed detection of other HPV types. Our findings suggest that methodological issues may contribute to the observed variation in HPV prevalence between studies. Furthermore, the potential for contamination, especially in the earliest studies where PCR assays were still very prone to contamination, may also be a contributing factor. Therefore, higher quality studies are needed in order for more accurate assessments of the prevalence of HPV in lung tumors.
It is possible that the variation of HPV prevalence might be due to differences in environmental factors including sexual behavior, smoking as well as other ethnic and cultural differences that influence HPV infection between various geographical regions (8,18,23). We and others have shown that women with a first diagnosis of cervical cancer (an HPV-related cancer) have an increased risk of developing second primary lung tumors (61,62). The burden of cervical cancer varies by geographic region and according to the World Health Organization, the highest age-standardized incidence rates for this disease are observed in South America (28.6/100000), followed by Asia (15.4/100000), Europe (11.9/100000) and North America (7.7/100000) (63). These observed differences may be primarily attributed to differences in screening practices rather than HPV infection rates. Nevertheless, studies of the geographic distribution of HPV prevalence in lung tumors and a possible relationship with cervical cancer are needed. Studies have also suggested that there may be an inverse relationship between smoking prevalence and HPV infection among persons with tumors arising in the oropharynx (64–66). Therefore, differences in smoking habits in geographic regions may also contribute to the variability of HPV prevalence in lung tumors; this hypothesis still needs to be addressed by ad hoc studies.
HPV16 and HPV18 are the most prevalent HPV types detected in invasive cervical cancers worldwide (67), and HPV16 is the most common type detected in head and neck squamous cell carcinomas worldwide (68). A few studies support the carcinogenic role of HPV in lung cancer by detecting the virus in the tumor cells but not in the adjacent normal epithelium (9,10,35). While HPV DNA detection might infer a possible involvement of the virus in lung tumors, DNA status alone does not prove a carcinogenic role in lung tumors. The pathogenicity of HPV relies on the expression of the HPV oncoproteins E6 and E7 and studies of cervical cancers have confirmed that these oncogenes are always expressed (69). A limited number of studies have presented evidence that the HPV E6 and E7 oncogenes are expressed in lung carcinoma tissues (37,44), but the etiologic link between HPV and lung cancer has not been firmly established. Therefore, further studies to explore the relationship between HPV infection and lung cancer are warranted. In addition to the Hill criteria for establishing causality (plausibility, strength of association, consistency, specificity, temporality, biological gradient, coherence, experimental evidence and analogy), the criteria described by Gillison et al. (70) are recommended for determining a causal relationship between HPV and lung cancer. These include evaluations of the presence or absence of integrated HPV genomes, expression of the viral oncogenes (E6/E7), association of HPV lung tumors with sexual behavior and increased incidence in populations that are immunosuppressed (such as human immunodeficiency virus-positive and transplant patients). It is important to evaluate metastases from the lung in order to further investigate whether HPV may be pathogenically related. However, to our knowledge, these studies have not been performed but are needed to further strengthen the pathogenic role of virus in lung tumors.
Although initial studies pointed to the role of HPV infection in squamous cell carcinoma of the lung (15), subsequent studies have detected HPV in other histological types of lung carcinoma. The true prevalence of HPV in these tissue types could not be determined due to the large heterogeneity between studies. Sexual behaviors, along with individual susceptibility to HPV infection, and viral smoking interaction may contribute to the observed differences in HPV prevalence. Another limitation in this study was that we were also unable to determine the prevalence of HPV according to smoking status and gender because this information was scanty in the studies that were included in this analysis. Therefore, stratified analyses of HPV infection by smoking status, gender and histology although not possible with the available data, is certainly warranted in future studies.
The main limitation of the present study is publication bias. While publication bias evaluates whether small studies give significant different results than larger studies, the lack of enough studies and the presence of large and significant between-study heterogeneity may also contribute to this finding (63,71). This limitation emphasizes the need for caution in interpreting the summary estimates and also indicates the need for additional high-quality studies to elucidate the role of HPV in lung carcinogenesis.
In summary, based on our findings, heterogeneity in our overall estimate of the prevalence of HPV in primary lung cancers may be attributed to differences in prevalence with geographical regions, methodological differences, including methods used to detect HPV, PCR protocols, sensitivity and specificity of these methods, as well as host factors. Our findings show that the prevalence of HPV16 and HPV18 suggests that HPV16 and HPV18 may be associated with lung tumors, especially in Asia rather than Europe, however, further studies are needed to investigate the role of the virus in this disease.
University of Pittsburgh Specialized Program of Research Excellence in Lung Cancer (P50 CA 090440).
Conflict of Interest Statement: None declared.