| |
Global burden of cancers attributable to infections in 2008: a review and synthetic analysis: "HBV, HCV, HPV, and H pylori were together responsible for 1·9 million cases worldwide"
|
| |
| |
The Lancet Oncology June 2012
Download the PDF here
Summary
Background
Infections with certain viruses, bacteria, and parasites have been identified as strong risk factors for specific cancers. An update of their respective contribution to the global burden of cancer is warranted.
Methods
We considered infectious agents classified as carcinogenic to humans by the International Agency for Research on Cancer. We calculated their population attributable fraction worldwide and in eight geographical regions, using statistics on estimated cancer incidence in 2008. When associations were very strong, calculations were based on the prevalence of infection in cancer cases rather than in the general population. Estimates of infection prevalence and relative risk were extracted from published data.
Findings
Of the 12·7 million new cancer cases that occurred in 2008, the population attributable fraction (PAF) for infectious agents was 16·1%, meaning that around 2 million new cancer cases were attributable to infections. This fraction was higher in less developed countries (22·9%) than in more developed countries (7·4%), and varied from 3·3% in Australia and New Zealand to 32·7% in sub-Saharan Africa. Helicobacter pylori, hepatitis B and C viruses, and human papillomaviruses were responsible for 1·9 million cases, mainly gastric, liver, and cervix uteri cancers. In women, cervix uteri cancer accounted for about half of the infection-related burden of cancer; in men, liver and gastric cancers accounted for more than 80%. Around 30% of infection-attributable cases occur in people younger than 50 years.
Interpretation
Around 2 million cancer cases each year are caused by infectious agents. Application of existing public health methods for infection prevention, such as vaccination, safer injection practice, or antimicrobial treatments, could have a substantial effect on the future burden of cancer worldwide.
Funding
Fondation Innovations en Infectiologie (FINOVI) and the Bill & Melinda Gates Foundation (BMGF).
Introduction
Infection is recognised as a major cause of cancer worldwide. Prevention and treatment of infectious agents have already had a substantial effect on cancer prevention.1 A useful statistic to quantify this effect is the population attributable fraction (PAF), defined as the proportion of new cancer cases in a specific population that would have been prevented by a hypothetical intervention on a specific exposure. For infectious agents classified as carcinogenic to humans,2 we calculated the PAF worldwide and in eight regions, using GLOBOCAN statistics on estimated cancer incidence in 2008.3 Similar calculations have been done for cancer incidence data from 19904 and 2002.5 In this report, we substantially revised the methods to reduce uncertainties and biases resulting from lack of data on population-specific and age-specific infection prevalence. We also discuss a framework for calculating global attributable fractions that might be applied to other causes of cancer. Some physical or chemical carcinogens act synergistically with infectious agents to cause cancers; in these cases, the attributable fractions can add to more than 100%. We report the attributable fractions of infectious agents but do not report the contribution of any non-infectious cofactor.
Results
Table 2 shows the estimated number of cancer cases attributed to infection in 2008, in less developed and more developed regions. Of the estimated 12·7 million new cancers worldwide, around 2 million were attributable to infections, of which 1·6 million (80%) occurred in less developed regions. HBV, HCV, HPV, and H pylori were together responsible for 1·9 million cases worldwide. Figure 2 shows the contribution of these infectious agents to cancer burden in less developed and more developed regions.
Table 3 shows a breakdown by geographical region of the number of new cancer cases and the number attributable to infection. Overall, 16·1% of cancer cases in 2008 were attributable to infection. The proportion was higher in less developed countries (22·9%) than in more developed countries (7·4%). The attributable fraction varied greatly between regions, from 3·3% in Australia and New Zealand to 32·7% in sub-Saharan Africa.
Table 4 shows a more detailed breakdown of attributable cancer cases, according to sex, age group, and development status of the country. Cervical cancer accounted for half of the attributable cases in women. The burden of gastric cancer and liver cancer was much higher in men than in women. The total number of cases attributable to infection was much the same in men and women. This similarity between sexes was noted across age groups, apart from in individuals younger than 40 years, where women had a higher burden of infection-related cancer, on account of cervical cancer, in less developed and more developed regions (figure 3).
Discussion
The analysis described in this report and in the appendix shows that infection is an important contributor to the global cancer burden, with 16·1% of cancers diagnosed in 2008 being attributable to infections, although the contribution due to infection varies widely from region to region. The estimated burden of cancer in 2008 attributable to infections is an update of previous estimates for 20025 and 1990.4 Our estimates for 2008 are slightly lower than those for 2002, for global burden of cancer (16·1% vs 17·8%) and by development status (7·4% vs 7·7% for more developed regions; 22·9% vs 26·5% for less developed regions). Overall, the results are remarkably similar, in view of the change in methodology to incorporate retrospective calculation of PAF based on prevalence of infection in cancer cases. For the four main infections altogether, the relative contribution of HPV to cancer burden is similar in more developed and less developed areas. The contribution of H pylori is, however, proportionally larger in more developed countries, and that of HCV and HBV is larger in less developed countries.
Three recent studies have reported country-specific estimates of PAFs for infection-related cancer in the UK,18 South Korea,19 and China.20 PAF estimates for China and South Korea were 25·9% and 21·2% respectively, in line with our estimates of 26·1% for China and 22·5% for east Asia. The estimate for the UK was 3·1%, which is lower than our regional estimate of 7·0% for Europe. We found similar estimates of PAF for North America (4·0%) and Australia and New Zealand (3·3%), areas where infection prevalence is similar to the UK.
The main strengths of our approach are the use of the highest quality epidemiological evidence and incidence data available. The choice of infectious agents and cancer sites was taken from a review by an expert IARC working group.2 We chose only agents classified as carcinogenic to humans by this group, and only cancer sites with sufficient evidence of association with infection (see appendix). Including more cancer sitesÑeg EBV in gastric cancer, HBV in non-Hodgkin lymphoma, or HPV in oral cavityÑis a possible approach, but it would be more subjective and potentially misleading, since the strength of the published evidence is controversial. The high threshold of evidence we used might have prevented us from addressing cancers for which evidence of an infectious link is rapidly emerging. Nevertheless, we aimed to not exaggerate the importance of infections in cancer. Estimates of relative risks and infection prevalence were always derived from the same review, or from systematic or comprehensive reviews that we updated when necessary. Cancer incidence data were derived from GLOBOCAN 2008, or calculated using a consistent method when GLOBOCAN estimates were not available.
For most PAF calculations, we used estimates of infection prevalence based on case series rather than prevalence in the general population. This choice was made for two reasons. First, there are few large, high-quality, population-based surveys of infection prevalence that are representative of the general population. Such surveys tend to oversample young people or those at low risk for specific infections. Undertaking high-quality population-based surveys is a long and difficult process and has not been done at all in many countries, particularly in less developed regions. Cancer case series, which are available in less developed countries, are usually representative of the population served by the hospital. Patients with cancer need expert care, and the severity of the disease makes it very likely that these patients will seek appropriate treatment in specialised centres. Second, population-based surveys often use less sensitive or specific measurement methods than case series, because the best testing methods are often expensive, invasive, and less feasible on a large scale. For example, causation is difficult to determine from serology in cases of EBV-related or HPV-related cancers; more than 90% of the population is positive for EBV, and HPV serology is not site-specific. However, patients included in case series undergo many tests, often including direct detection of infectious agents and even gene expression in tumour tissue. Therefore, we are confident that calculating prevalence from case series increases the validity of PAF estimates.
The need to avoid the effect of time trends in infection prevalence when selecting the case series on which we based our calculations was also considered, but did not seem to be an issue, at least for the four main infections that drive most of the global PAF. Although improved living conditions have led to a steadily decreasing prevalence of H pylori infection in many populations, prevalence in gastric cancer cases from nested caseÑcontrol studies or other epidemiological studies seems to be very stable, around 90%, with no detectable secular trends. The case series we selected for estimating the prevalence of hepatocellular carcinoma attributable to HBV and HCV used only second-generation or third-generation ELISA for detection of HCV, with a total of 37 000 cases from 132 studies published from 1992 to 2009. The low sensitivity of first-generation anti-HCV ELISA has long been recognised and seems to differ between cases and controls.13 Most studies of HPV-related cancers have been done in the past 15 years using DNA detection techniques, and relevant studies of oropharyngeal cancers are even more recent.
Our approach has several limitations. First, some uncertainty in cancer incidence and infection prevalence is inherent in our estimations of PAF. Our attempt to obtain global estimates of infection prevalence by pooling local data sources inevitably requires extrapolation to countries with sparse data on cancer incidence or risk-factor prevalence. The main risk of this extrapolation is that estimates derived from a small amount of data are applied to a larger population that is substantially different, with possible amplification of bias. The weighting scheme we used to estimate regional infection prevalence might give undue weight to larger surveys done in smaller countries.
A second limitation is that strong assumptions were required for the calculations. For example, we assumed that relative risks for infection were constant across populations and sexesÑa common assumption in epidemiology. Generally, this assumption is not true when comparing populations or sexes with widely different baseline cancer incidence rates, since the multiplicative assumption behind relative risk estimation is only an approximation. Our PAF calculations were not strongly dependent on the constant relative risk assumption because we used the retrospective formula based on cancer cases, and the relative risks were uniformly large. A change in relative risk from 10 to 20, for example, makes only a 5% difference in the PAF estimate. For H pylori, HPV, HBV, and HCV, the order of magnitude of relative risk is generally constant worldwide when other known risk factors have been controlled for, so the assumption of a constant relative risk should not lead to substantial error.
A third limitation is the lack of high-quality epidemiological data for some of the cancer sites in this study (eg, EBV-related cancers, such as nasopharyngeal carcinoma), in areas of low cancer incidence. We based our estimates on the most recent and least subjective evidence, but the lack of data inevitably leads to uncertainty in the estimates.
Some of the assumptions used in our calculations were conservative. We restricted the effect of HPV in head and neck cancer to the oropharynx and base of the tongue, where the epidemiological and mechanistic evidence for a causal effect is strongest. HPV might be associated with other head and neck cancers, but this is impossible to quantify with current data. For H pylori, we based PAF calculations on a relative risk of 5·9; although this is the best estimate available, evidence from prospective studies and studies using western blot rather than ELISA suggests that it might be higher. Such studies yielded relative risks greater than 10,21Ñ25 which would increase the proportion of non-cardia gastric cancer attributable to H pylori from 75% to 90%. Likewise, we estimated that 75% of diffuse large B-cell lymphoma of gastric location is due to H pylori, but the attributable fraction could be nearer to 100%. The discovery of new associations between infections, particularly viruses, and cancer has been anticipated; however, studies have either disclosed associations with very rare cancers (eg, Merkel-cell carcinoma) or are yet to provide conclusive results (eg, for cutaneous HPV types and non-melanomatous skin cancer). Nevertheless, undiscovered associations could exist, which is another reason to conclude that our results probably underestimate the true burden of infection-associated cancers.
Attributable-risk calculations can be done for any environmental exposure, but are most useful, from a public health perspective, when relative risks are large and when interventions to reduce population exposure are feasible. Many infection-related cancers are preventable (panel 2), particularly those associated with HPV, H pylori, HBV, and HCV. Prophylactic vaccines have shown nearly 100% efficacy in preventing precancerous lesions of the cervix due to HPV types 16 and 18, among previously uninfected individuals. In Taiwan, the incidence of hepatocellular carcinoma in children and adolescents has been substantially reduced by a combination of immunoglobulin given at birth to prevent vertical transmission from mother to child at birth and childhood HBV vaccination.1 The current WHO recommendation is to vaccinate all infants against HBV as soon as possible after birth.26 Although no vaccine is available for HCV, iatrogenic transmission can be avoided with safer practices for injection and blood transfusion, and preference for oral drug delivery over injections where available. H pylori is a treatable infection, although the feasibility, effectiveness, and safety of large-scale eradication of H pylori infection in different age groups is not yet clear. Our finding that H pylori accounts for 46% of infection-associated cancers in more developed areas might reflect lower investment in research on prevention of gastric cancer compared with cervical and liver cancers. Such considerations underscore the difference between what is theoretically preventable, according to the assumptions of the PAF calculations, and what is preventable in practice. The importance of time must also be acknowledged; preventing infection-associated cancers in 2008 would have required intervention programmes many decades earlier.
Panel 2
Research in context
Systematic review
The relation between infectious agents and cancer was the subject of a comprehensive literature review done as part of the IARC Monographs programme. Global cancer incidence and mortality data were synthesised by the GLOBOCAN project to provide estimates of the global burden of cancer. We used these data sources to estimate the global burden of cancer due to infection, relying on existing systematic reviews of the literature to provide the quantitative inputs (relative risk and infection prevalence) required for calculation of attributable fractions. Previous global estimates of the proportion of cancers attributable to infection were done for 19904 and for 2002.5 Country-specific estimates have been provided for the UK,18 South Korea,19 and China.20
Interpretation
The present review extends previous findings by showing wide geographical variation in the fraction of cancers attributable to infection. It also underscores the importance of HPV, Helicobacter pylori, HBV, and HCV as cancer-related infectious agents. Since infections are an important and preventable cause of cancer worldwide, clinicians should support the implementation of available strategies for preventionÑie, vaccination against HBV and HPV, use of safe injection practices, and avoidance of parenteral treatment when oral treatment is available. Clinicians should also closely follow and, if possible, contribute to progress in areas where early detection of infection (eg, HPV) or treatment (eg, HCV and H pylori) could diminish cancer sequelae. Public health doctors and cancer-control specialists should appreciate the importance of infectious causes of cancer in different regions and age groups, particularly in low-income and middle-income populations. The 2011 UN high-level meeting on non-communicable diseases highlighted the growing global agenda for prevention and control of non-communicable diseases. Although cancer is considered a major non-communicable disease, a sizable proportion of its causation is infectious and simple non-communicable disease paradigms will not be sufficient.
In view of the high mortality rate of infection-associated cancers, the fraction of cancer deaths attributable to infections is probably higher than the 16·1% that our study generated. Although a full investigation of cancer death due to infection is beyond the scope of this report, we can estimate the mortality burden by applying the PAFs to the 7·5 million cancer deaths that occurred in 2008. These calculations suggest that 1·5 million cancer deaths were attributable to infectious agents, or roughly one in five deaths due to cancer worldwide.
|
|
| |
| |
|
|
|
