Estimation of Cancer Burden Attributable to Infection in Asia

Background Some infectious agents have been shown to be human carcinogens. The current study focused on estimation of cancer burden attributable to infection in different regions of Asia. Methods By systematically reviewing previous studies of the infection prevalence data of 13 countries in Asia and relative risks of specific cancers, we calculated the population attributable fraction of carcinogenic infections. Using data from GLOBOCAN 2012, the overall country-specific and gender-specific number of new cancer cases and deaths resulting from infection were estimated. Results Across 13 principal Asian countries, the average prevalence and range was 6.6% (0.5% in Japanese women to 15.0% in Vietnamese men) for hepatitis B virus (HBV), 2.6% (0.3% in Iran to 5.1% in Saudi Arabia) for hepatitis C virus (HCV), 7.9% (2.8% in Pakistan to 17.7% in China) for human papillomavirus (HPV), and 61.8% (12.8% in Indonesia to 91.7% in Bangladesh) for Helicobacter pylori (HP). The estimated total number of cancer cases and deaths caused by infection in these 13 countries were 1 212 026 (19.6% of all new cancer cases) and 908 549 (22.0% of all deaths from cancer). The fractions of cancer incidence attributable to infection were 19.7% and 19.5% in men and women, respectively. The percentages of cancer deaths attributable to infection were 21.9% and 22.1% in men and women, respectively. Among the main infectious agents, HP was responsible for 31.5% of infection-related cancer cases and 32.8% of infection-related cancer deaths, followed by HBV (28.6% of new cases and 23.8% of deaths), HPV (22.0% of new cases and 27.3% of deaths), and HCV (12.2% of new cases and 10.6% of deaths). Conclusions Approximately one quarter of all cancer cases and deaths were infection-associated in Asia, which could be effectively prevented if appropriate long-term controls of infectious agents were applied.


INTRODUCTION
Chronic infection with potentially carcinogenic agents is recognized as a major risk factor of human cancer 1,2 and was estimated to be responsible for around 2 million new cancer cases worldwide in 2008. The overall population attributable fraction (PAF) for infectious agents was 16.1% worldwide, 3 indicating that 16.1% of new cancer cases could be prevented by elimination of such infectious agents. This fraction is higher in developing countries (22.9%) than in developed countries (7.4%). Asia, which contains over 60% of the world's population, plays an increasingly important role in international economics and trade. As a part of global analysis, the burden of infection-associated cancer in Asia has been estimated. 3,4 While the diversity of environment, culture, and economics in different countries contributes to the variety of infectious agents and their prevalence in the general population, the effects of those factors have not been revealed in previous studies.
The aim of the current study is to assess the cancer burden attributable to infection in 13 Asian countries based on review of nation-specific prevalence of carcinogenic infectious agents and comparison of risk estimates between exposure and infection-associated cancer.

Geographical areas
As residential environment, social-economic status, habits, and culture vary in different places among different groups of people, we chose countries and regions with the largest population sizes to obtain reliable and stable estimates, while also taking population representativeness into consideration. With the population data of almost all Asian countries and regions available in GLOBOCAN 2012, we chose a cut-off point of 30 million. As a consequence, 13 countries were included and divided into four areas: East Asia (Korea, China, and Japan), Southeast Asia (Indonesia, the Philippines, Thailand, and Vietnam), Middle South Asia (Bangladesh, India, Iran, and Pakistan), and West Asia (Turkey and Saudi-Arabia), following the geographical definition of the 2012 revision of the United Nations' World Population Prospects. 5 Although Iraq was on the list of 13 countries, few studies of infection prevalence in Iraq were available. Subsequently, we substituted Saudi Arabia, a smaller West Asian country with a population of 27 million, for Iraq because more information on infection prevalence was available for Saudi Arabia.
Definition of exposure Infectious agents with sufficient evidence for their carcinogenicity were defined in an IARC monograph series. 2 The infectious agents and related cancers used in the current study are listed in Table 1. Other carcinogenetic infectious agents, such as Schistosoma haematobium and human T-cell lymphotropic virus type 1 (HTLV-1), were not included in the current analysis because the prevalence of the infectious agents in Asia or the data of cancer incidence and/or mortality were not available in published studies or were only available in low quality studies (eg, studies with small sample sizes or poor study designs).

Prevalence of infectious agents and relative risk of cancers and infectious agents
Searches were conducted in Pubmed, Google Scholar, and Chinese National Knowledge Infrastructure (CNKI) for all published studies in Chinese and English. The keywords of prevalence-oriented searching were "prevalence", the names of infectious agents, and the country names. Prevalence data published between 1997 and 1999 were collected, fixing the exposure time of infectious factors around 1997. The intervals from exposure to the infectious agents to diagnosis of related cancer were all assumed to be 15 years in the current study. 6 If the data from 1997-1999 for some countries were not available, we expanded our search to data published between 1995 and 2000. We included case-control studies with large sample sizes only if population-based cross-sectional studies were not available. As a consequence, prevalence data is based on the general population without adjustment for age. In the search for relative risks (RRs), we included keywords of "meta-analysis", "cohort study", "case-control study", and the names of infectious agents and relevant cancers. The priority for data selection was (1) meta-analysis or pooled analysis in Asia; (2) large-scale case-control study in an Asian country; and (3) multinational meta-analysis or meta-analysis from non-Asian countries. For instance, we included (2) only if (1) was not available. The biomarkers of infectious agents were human papillomavirus (HPV) DNA, HPV L1 antibody, serum-antibody against Helicobacter pylori (HP), anti-HIV antibody, anti-hepatitis C virus (HCV) antibody, and hepatitis B surface antigen (HBsAg), which are considered to be relatively sensitive and accurate measurements of prevalence in the general population. RRs in these studies are calculated by comparing the probability of developing cancer in an exposed population to that in a non-exposed population (cohort study) or estimated with odds ratios (ORs) in casecontrol studies. We prioritized results from the latest published studies over older studies because the time effect and newly discovered confounding biases are more likely to be taken into consideration in more recent studies. We assumed that RRs were constant in different countries and adopted generally accepted RRs, which were abstracted from international studies or meta-analysis of worldwide prevalences. Similarly, if the sex-stratified prevalence was not accessible, it was assumed that both sexes were exposed equally to infectious factors, and sex-stratified prevalence was replaced by overall prevalence. The prevalence among patients was adopted when the general prevalence was not available. To account for the differences between general population and patient prevalences, Formula B (which we discuss in the Methods section), was used in PAF calculation. As 85%-95% of the general population had positive serology tests (IgG) for Epstein-Barr virus (EBV) infection when young, 7,8 it was not necessary to match the prevalence of EBV in calculation of cancer burden. As a result, the PAFs of EBV-related cancers, which were calculated from caseseries prevalence, were applied. 3

Incidence and mortality data of infection-related cancer
The cancer-specific sex-stratified incidence and mortality in each country were obtained from GLOBOCAN 2012 and a cancer register project (cancer incidence of Japan). 9 The numbers of new cases of cancer and cancer-related deaths that were attributable to infection were estimated by multiplying the overall numbers of new cases of cancer and cancer-related deaths with the corresponding PAF. The data of anal cancer, penile cancer, vulvar cancer, and vaginal cancer were included in the category of others and unspecified cancers in the GLOBOCAN project. Therefore, we referred to Cancer Incidence in Five Continents Vol. X (CI5X), 10 which recorded cancer-sorted incidence data worldwide from 2003 to 2007, to estimate the proportions of these cancers. For countries which were not listed in the CI5X (Bangladesh, Indonesia, Vietnam, Iran, and Saudi Arabia), we assumed geographic similarity of cancer incidence patterns in the same area of Asia. For example, Bangladesh is located in Middle South Asia, so the cancer incidence data was assumed to be consistent with that of Pakistan.
Specific data on incidence and mortality related to noncardia and cardia gastric cancer, hepatocellular carcinoma, and cholangiocarcinoma in Korea was obtained in an original article 11 because the national registry was of high quality with broader categories than those of GLOBOCAN 2012.
It was assumed that Burkitt's lymphoma accounted for 2.6% of the total non-Hodgkin's lymphoma (NHL) cases in East/Southeast Asia (however, the proportion in Japan is assumed to be <0.1%) and 14.8% in West Asia. In developing areas of Asia, noncardia gastric cancer constituted 80.0% and 87.0% of gastric cancers in men and women, respectively, while in developed areas, the respective proportions were 80.0% and 88.0%. 4 We assumed the proportions of hepatocellular carcinoma and cholangiocarcinoma in liver cancer to be 80% and 15%, respectively. The assumed incidence of cholangiocarcinoma was higher than the world average due to a markedly high prevalence of liver flukes in the study areas. 12 By applying these findings to Levin's formula AF ¼ 13 (where P is the prevalence of infectious agent in general population, Pc is the prevalence among patients, and RR is the RR of cancer among infectious agents exposed population), PAFs were calculated. Combining the data of cancer incidence and mortality in each country from GLOBOCAN 2012, the overall and specific numbers of new cancer cases and cancerrelated mortality caused by infection were estimated. We hypothesized that infection had no effect on survival. Table 2 shows the data source of RRs applied in the current study, the biomarkers of each infectious agent, the study design, and the target population of each study. The RRs of human papillomavirus (HPV)-related anal cancer, vulvar cancer, vaginal cancer, and penile cancer were replaced with corresponding generally accepted PAFs because no data for RRs were available. 14 Oncogenic HPV is the etiological agent of cervical cancer, while EBV and Kaposi's sarcomaassociated herpes virus/human immunodeficiency virus (KSHV/HIV-1) play vital roles in nasopharynx cancer (100.0% of nasopharynx cancers were assumed to be EBV-related in medium-and high-risk areas, and 90.0% elsewhere 4 ) and Kaposi's sarcoma (the RRs were 97.5 in men and 202.7 in women, respectively). 15 Therefore, the PAFs of these cancers were assumed to be 100. Information about the prevalence of etiologic agents in the general population of the study countries and the RRs of corresponding cancers, as well as gender-specific prevalence of infection by carcinogenic infectious agents, are shown in Table 3. Table 4 specifies PAFs of infectious agents and cancer sites in different countries among men and women. Table 5 shows total estimated new cases and deaths caused by infection.

Overview
Among six principle infectious agents, the most new cancer cases were attributed to HP (31.5%), followed by HBV (28.6%), and HPV ( Table 6 illustrates the new cancer cases and deaths due to HPV infection.

Epstein-Barr virus
The PAF of nasopharyngeal carcinoma caused by EBV was 90 in most Asian regions. 4 In Southern China, an area with a high nasopharyngeal carcinoma burden, the fraction was

Hepatitis B virus and hepatitis C virus
Asia has a heavy disease burden from hepatitis B virus (HBV) and hepatitis C virus (HCV), both of which contribute to hepatocellular carcinoma and cholangiocarcinoma. Moreover, HCV is related to NHL. Table 7 lists the numbers of new cases and deaths from cancers resulting from HCV and HBV infection in the target countries.

Helicobacter pylori
The cases and deaths of gastric cancer caused by HP are shown in Table 8.

DISCUSSION
The present study is the first to assess the burden of potentially oncogenic infections and their related cancers in Asia. Based

7.7% of cases in developed countries).
Our results were somewhat different from previous nation-specific PAF calculations 11,85 because most of the RRs in our study were summarized from international studies that were generalizable to different Asian countries; for example, we chose the RR of 18.8 for HBV, which was taken from a study with a large sample size, 21 rather than the RRs in Shin's study 11 (24.45 in men and 33.73 in women), which were taken from a metaanalysis of Korean data and only represented the Korean population. Moreover, we estimated the number of cancers based on GLOBOCAN2012, while previous studies used Korean data from 2007. This earlier data found lower prevalence than those of GLOBOCAN2012, and the proportion of infection-related cancers in all cancers was consequently lower. However, our results are consistent with Parkin's estimates for developing countries because the majority of countries in our study are developing countries (except for Japan and Korea). Specifically, 20.5% of new cancer cases were related to infection in the 11 developing Asian countries, and 14.5% of new cancer cases were related to infection in Korea and Japan. Because the background prevalence of infectious agents in Korea and Japan differed from other developed areas (mainly Europe and North America), the proportion of infection-related cancers in developed countries was doubled in our study. Another study in 2008 3 revealed that 16.1% of new cancer cases were attributed to infection worldwide, and the proportion was higher in less developed countries (22.9%) than in more developed countries (7.4%). While both Parkin's and Martel's studies focused on global infection-related cancer burden, our study is designed in a region-specific way, in which we analyzed the data from 13 Asian countries to illustrate the infection background of cancer cases and deaths in Asia in 2012. The findings of our study provide comparable data for different Asian countries, which enable us to gain insight into the regional status of infectious agents that are linked to cancer incidence. The potential burden of disease (BOD) described in the Asian region could be significantly ameliorated with intervention measures, such as prophylactic vaccinations for HBV and HPV. In addition, this description of the BOD offers a baseline reference for governments seeking to address the public health issues of infectious disease and cancer. Combining the outcomes of the current study with the data of medical expenses and living expenditures of corresponding countries, the economic burden caused by infection could be calculated. These economic burden calculations can illustrate to policymakers the importance of taking immediate action on infection control for these agents, which may improve the living standard of the public, reduce unnecessary loss of lives, and reduce financial burden.
The gender-specific patterns of incidence and mortality caused by different infectious agents (shown in Figure) are of practical consequence when designing preventive interventions and treatment for targeted population. For men, HP and HBV are the main causative factors of gastric cancer and liver cancer, respectively, together comprising over three quarters of infection-related cancer cases and deaths. HBV infection contributes to twice the liver cancer burden in men as in women, and much work remains to reduce the high prevalence of HBV in Southeast/East Asia. In women, HPV and HP were associated with heavy burden of cervical and gastric cancers, respectively. These findings indicate that more attention should be paid to prevention of HPV in women, and both prevention and treatment of HP play vital roles in reducing cancer burden in Asia for both genders.

HIV
Although infection with HIV alone will not cause many common HIV-associated cancers, it has been recognized as a major contributing factor to some specific cancers because cancer incidence rises significantly with combined infection of HIV and strong cancer-causing infectious agents. The carcinogenic mechanisms of HIV are immune deficiency and suppression, which lead to the failure of immunological surveillance of tumor cells and susceptibility to carcinogenic factors. The contribution of HIV to the development of many cancers is complicated; for example, HIV can increase the risk of anal cancer, which is caused by HPV. To avoid overcounting for these co-infections, we only estimated cases and deaths of Kaposi's carcinoma and NHL, in which HIV plays a main role, though in comparison to European or North American countries, the prevalence of Kaposi's carcinoma in Asia is quite low. 15 The numbers of cancer cases and deaths related to HIV infection might be overestimated because we considered the estimated prevalence of HIV in low-prevalence areas as 0.1%, while UNAIDS 38 reported prevalence below 0.1% in the general population. The current study estimates that HIV contributed to 9018 new cases and 5721 deaths from cancer in the study area. However, the rates of HIV infection started to rise after the low-prevalence stage of the early 1990s, increasing from 0.1% to 1.3% in 2009. 83 The pattern of HIV spread in Asia has gradually changed, from sporadic new cases in the general population and epidemic in high-risk groups to a low-level epidemic in the general population. As a consequence, the numbers of HIV-associate cancer cases and deaths is expected to increase in a few years, especially in regions such as Thailand and India. The HIV prevalence of these areas is much higher than in the rest of the surveyed Asian countries, which is expected to cause health and social problems, including short lifespan and poverty. Fortunately, the utilization of antiretroviral therapy and behavioral interventions in Asian areas have had initial success in reducing costs of HIV control. 83 However, much work is still required to control the spread of HIV in developing countries.

HP
There has been controversy over the direct evidence linking HP infection and gastric cancer in Asia, 78 where there is some doubt that the high infection rate of HP (eg, in Japan and Bangladesh 58 ) is a predictor of the high prevalence of gastric cancer. At an individual level, HP has been associated with the occurrence of gastric cancer, but some Asian countries have a low prevalence of gastric cancer despite a high HP prevalence. Alcohol and cigarette consumption in these areas may cause many of these gastric cancers. However, a Japanese study 40 revealed that both groups testing strongly positive and weakly positive for HP antibody showed significantly higher incidence of gastric cancer compared to those testing negative. We adopted a combined analysis, which illustrated that the OR for non-cardia gastric cancer and HP was 5.9 (95% CI, 3.4-10.3) compared with those testing negative for HP, 18 while the relationship was less significant between cardia gastric cancer and HP infection (RR1.6; 95% CI, 1.0-2.5). 19 Consequently, we estimated that approximately 436 501 new cancer cases and 324 042 deaths were caused by HP in 2012. The use of regular and rational combinations of antibiotics is the optimal secondary prevention strategy against HP infection, but such use of antibiotics is the primary prevention method for gastric cancer. Using regular gastroscope screening tests, early lesions on the gastric epithelium can be detected before they progress to cancer. 86 HPV In our study, new cancers in the oral cavity, oropharynx, cervix, vulva, and vagina that were associated with high-risk types of HPV totaled 278 403 cases, ranking HPV the highest among the principle infectious cancer-causing agents in women. HPV was also estimated to be responsible for 162 762 cancer deaths in Asia, making HPV the secondranked infectious factor leading to cancer deaths. To estimate the burden of HPV-related vulvar, vaginal, penile, and anal cancers, we applied a geographic similarity method in cases where country-level data was not available in CI5X. Namely, neighboring countries' data were used to substitute for the corresponding incidence or mortality of the country with missing data. Admittedly, this method has some limitations, because the incidence and mortality of cancers vary by country. However, it is likely to be a more accurate way of estimating the sub-category cancer burden with the present database than by assuming the same incidence and mortality across Asia without considering geographic factors.
Since the currently licensed prophylactic HPV vaccines (Gardasil for type 6, 11, 16, and 18 and Cervarix for type 16 and 18) marketed in North America, Europe, Australia, 87 and some parts of Asia became available, the expectation is that HPV-related cancers maybe largely prevented if these vaccines are adopted with high coverage. Both vaccines have been shown to be effective, immunogenic, and safe. 88 However, the vaccines have not been approved to market in most areas of Asia. Where they are approved, such as Hong Kong and Korea, they are high-priced and less likely to be listed in social medical insurance, therefore largely only available to those who can afford them. Where governments have endorsed and paid for vaccines, such as in Australia, 87 reductions in rates of HPV-related infections 89 and high-grade lesions (ie, lesions coded as cervical intraepithelial neoplasia of grade 2 or worse or adenocarcinoma in situ) 90 have already been seen. The vaccines are currently accessible through school-based government-funded programs only in a few regions, such as Malaysia, Bhutan, and Japan. The predicted outcome is that the HPV-related cancer burden will decline if the primary prevention measure of HPV vaccines is applied in the general population by governments. In addition, screening with cytology tests, visual inspection, or HPV-DNA detection are practical secondary prevention methods that can be employed against cervical cancer in low-resource countries and areas.
HCV and HBV HCV and HBV were highly epidemic up to the early 1990s in many areas of Asia, resulting in a relatively high incidence of liver cancer decades later. However, the late 1990s saw the introduction of the HBV vaccine among all newborns, particularly babies of women who were chronically infected with HBV. Widespread HBV vaccination has been markedly effective in reducing the HBV epidemic in areas where good coverage has occurred. A Taiwanese study showed that universal HBV vaccination for newborns provided long-term protection for up to 20 years, which enabled prevention of the infection before adulthood 91 ; meanwhile, obligatory screening of blood and organ donations has significantly curbed HCV transmission. Together, HBV and HCV caused 431 086 new liver cancer cases and 406 779 deaths in 2012, representing 77.61% of liver cancer cases and 76.6% of deaths. In addition, liver flukes, such as Clonorchis sinensis and Opisthorchis viverrini, were also causal factors of liver cancer (responsible for 1701 new cholangiocarcinoma cases in our estimation). C. sinensis is still an endemic parasite in areas of river basins in East Asia, while O. viverrini remains epidemic in Thailand. 4,92 In this case, improving public awareness of the risk of the consumption of raw or insufficiently cooked fish, management of fecal sewage, and the supervision of intermediate hosts in epidemic areas are of great importance in controlling fluke infection and preventing related liver cancers.
Inevitably, our study has some limitations. First, not all prevalence data of infectious agents in the target countries were available. For instance, the prevalence of EBV in most countries and the prevalence of HPV in Bangladesh were not readily available from the databases to which we had access. We used overall PAFs for the cancers caused by EBV in all target countries and PAFs of HPV in Bangladesh. Moreover, the RRs we adopted are not country-specific, except for some RRs taken from independent studies on infection-related cancer burden in Korea 10 and China. 85 Numbers of cancer deaths in Japan were taken from the data of Monitoring Cancer Incidence in Japan (MCIJ) Project. The mortality data reported in MCIJ were lower than those in GLOBOCAN 2012, and we recognized them as more reliable and valid due to the wider regions that the program covered and the high-quality systematic registration method the MCIJ Project employed. According to the proportion of cancer deaths, some cancer categories in the MCIJ report were split into the corresponding sub-categories, as in GLOBOCAN, to facilitate estimation. As we were focusing on estimated numbers of infection-related cancer cases and deaths rather than comparing cancer incidence and mortality among different countries, we did not do age or sex standardization when using two data sources. Age-standardized data were available in both the MCIJ and GLOBOCAN. Old research was used to extrapolate the prevalence within a defined period, which affected the results because older technology for detection of infectious agents was limited. In addition, the geographic similarity assumption was utilized to tackle the inaccessible proportions of other and unspecified cancers (eg, penile, vulvar, and anal cancer) of the countries not included in CI5X. The data of CI5X came from a different time period (1993)(1994)(1995)(1996)(1997), and the rest of data were from 1997 to 1999. Some of the RRs were extracted from non-Asian studies if the data could not be obtained in reliable Asian studies. As a consequence, the estimations in our study ignored some geographic variation among these countries. There might be some systematic differences between larger and smaller countries in Asia. In addition, the sample sizes of some prevalence studies of infectious agents in some countries were small or contained severe bias because the subjects were blood donors or hospital-based populations rather than general populations, which makes it hard to extract valid prevalence. For example, the majority of studies of HBV and HCV prevalence in Japan were based on the registry system of blood donation, 39,40 resulting in limited representativeness. As liver flukes are endemic in a limited area in Asia and accurate data were not available, rates of cholangiocarcinoma and urinary bladder cancer caused by liver flukes were not included in our study. The exclusions of cholangiocarcinoma and urinary bladder cancers in calculating the prevalence of cancers and deaths may have caused a slight underestimation of the deaths and new cases attributable to this infection.
In conclusion, infectious factors play a major role in the etiology and progression of various cancers, which contributes to about one quarter of all cancer cases and cancer-related deaths. Infection by these agents not only results in the loss of life, but also imposes heavy economic burden on families and societies, both directly and indirectly. Adopting long-term measures (including primary and secondary prevention) to prevent infection from the principal cancer-causing agents is an efficient way to reduce rates of infection-related cancers. Tools are available to prevent many of these diseases and should be utilized.