Evaluation of the Cancer Transition Theory in the US, Select European Nations, and Japan by Investigating Mortality of Infectious- and Noninfectious-Related Cancers, 1950-2018

Key Points Question As nations develop, do they experience a systematic pattern in cancer trends by type, distinguishing between infectious-related and noninfectious-related cancers? Findings This cross-sectional study of 6 countries’ cancer mortality data from 1950 to 2018 found that a crossover in trends between the 2 main types of cancers (infectious-related and noninfectious-related) took place around 1990 in Japan and in the mid-1950s in Norway. For the other countries in the study, the trends in the 2 types of cancers do not intersect as they do for Japan and Norway, but those other nations still exhibit a cancer transition with declining rates of infectious-related cancers and rates of noninfectious-related cancers initially increasing, before eventually declining. Meaning These findings support the theory that cancer transitions are occurring in the US, select European nations, and Japan.


Introduction
Cancer is one of the leading causes of death worldwide, and because of marked declines in cardiovascular disease in several countries, cancer has become the leading cause of death in many high-income populations. [1][2][3] Although the important role of infection in causing certain cancer types is established and quantified (2.2 million cancer cases were attributable to infection in 2018), 4-6 the view of cancer as a prime example of a "degenerative and man-made disease" 7 remains. 8,9 Indeed, most global reports, including those of the World Health Organization (WHO), categorize cancer as a key noncommunicable disease. 10 The article by Omran 7 on the epidemiological transition has influenced a number of academic fields. 11, 12 In his article, Omran 7 postulated that in the third and last stage of the transition, noncommunicable and chronic ailments such as heart disease and cancer replace acute, infectious diseases as prime killers. In this age of degenerative and lifestyle-related diseases, infectious diseases experience a progressive decline but do not disappear entirely. 7 Omran's theory, 7 however, is overly simplified because it overlooks the fact that infection is often an important cause of cancer. 13 For instance, we now know that the bacterium Helicobacter pylori (H pylori), hepatitis B and C viruses, and human papillomavirus are important causes of cancers of the stomach, liver, and cervix, respectively. The number of new cancer cases in 2012 attributable to infectious agents was about 79% for stomach cancer, 73% for liver cancer, and 100% for cervical cancer. 13 Gersten and Wilmoth 14 first introduced and developed the concept of the cancer transition, which they meant to be analogous to Omran's epidemiological transition. The authors analyzed cancer trends in Japan from 1951 until 1997 and found that cancers associated with infectious causes were declining, whereas those associated with noninfectious causes were increasing. There has thus been a playing out of the epidemiological transition within the broad and complex set of diseases that define cancer.
Work investigating and attempting to extend the cancer transition theory has been limited, perhaps numbering no more than a few articles. However, Knaul and colleagues 15 have seriously engaged with the theory and presented cross-sectional data from GLOBOCAN (2008) 16 suggesting that cancers associated with infectious causes become less prominent as many nations move along the gradient from less to more developed. A similar finding has been reported by Fidler and colleagues 17 in a more recent article using a different data set. The aim of the present study is to contribute to the nascent literature on the cancer transition theory, and in so doing, examine the relevance of the epidemiological transition theory to understand current mortality trends. This article focuses on evidence of cancer transition in the US, select European nations, and Japan, differentiating between cancers associated with an infectious cause and other cancers.

Methods
Because all of the statistical information used in this cross-sectional study is derived from publicly available, deidentified, aggregated death records, the University of California, Berkeley institutional review board has determined that this study is exempt from review and does not require informed The International Classification of Diseases (ICD), implemented by all study countries during the period under consideration, has gone through numerous revisions, from the sixth in 1950 to the tenth currently. These ICD revisions introduce some disruptions in the data series as illustrated by trends in the death rates from liver cancers in both Japan and France. However, as documented in the literature, cancer as a disease is much better defined than many others and these disruptions are small enough that a general picture of the overall trends emerges very clearly. 32,33 Similarly, differences in diagnostic and coding practices do not appear to be large enough across countries to bias international comparisons in cancer mortality trends. 34, 35 We define the cancer transition as the process through which a country's population moves from a situation where rates of infectiousrelated cancers are greater than those of noninfectious-related cancers, to the converse situation where rates of noninfectious-related cancers are higher.

Statistical Analysis
The 95% CIs for the age-standardized death rates were calculated using the Keyfitz formula. 36 Statistical analysis was performed using RStudio statistical software version 1.3.959 (Integrated Development for R, PBC) from April 2020 to February 2021.

Results
Trends in the age-standardized death rates from infectious-related cancers and noninfectiousrelated cancers in the 6 study countries are presented on Figure 1. Trends in the specific cancers making up each of the 2 categories (infectious-related and noninfectious-related cancers) are presented in Figure 2. noninfectious-related cancers). A comparison between the epidemiological and cancer transition theories is presented schematically in the Table. In all countries studied here, mortality from infectious-related cancers has declined steadily throughout the period, except for Japan (Figure 1). At the beginning of the study period, the age-standardized death rates from cancers associated with infectious causes ranged from 50.4 deaths per 100 000 population (95% CI, 49.9-50.8 deaths per 100 000 population) in the US to

Infectious-related cancers Noninfectious-related cancers
Error bars denote 95% CIs. 159.0 deaths per 100 000 population (95% CI, 157.6-160.4 deaths per 100 000 population) in Japan, a ratio of 1 to 3.2. Note, however, that the series starts in 1959 in the US, but in 1950 for Japan.

JAMA Network Open | Global Health
In Japan, the only country where mortality from these cancers increased, the peak was reached at 179.0 deaths per 100 000 population (95% CI, 177.6-180.5 deaths per 100 000 population) in 1956.
At the end of the period (2016-2018, depending on the country), the rates for infectious-related cancers ranged from approximately 20 deaths per 100 000 population (21.2 deaths per 100 000  where it contributed <50%) and was associated with the overall decline of mortality from this type of cancer ( Figure 2; eTable 1, eTable 2, eTable 3, eTable 4, eTable 5, and eTable 6 in the Supplement).
Stomach cancer's share was reduced to 20% to 30% (46% in Japan, where the rate has been much higher than elsewhere) by the end of the 2010s. The death rate from cervical cancer also exhibited a decline in all study countries throughout the period. Mortality from liver cancer did not exhibit any systematic trend across the 6 countries, with international variations in both levels and direction.
At the beginning of the study period, all countries exhibited an increasing trend in the noninfectious-related cancers. Every country but Sweden experienced a peak around 1990, with Sweden's peak occurring in 1974. In all but the US and England and Wales, mortality from noninfectious-related cancers were at their lowest at the beginning of the study period. Initial rates Japan and Norway are the only 2 countries where trends in mortality from the 2 groups of cancers intersect. Norway's intersection in 1955 occurred earlier than Japan's, which was in 1988.

Discussion
In this study, we used temporal cancer-specific mortality data to assess whether the US, England and Wales, France, Norway, Sweden, and Japan have experienced comparable trends in cancer by type, albeit on a different schedule. Overall, there is compelling evidence supporting the cancer transition theory.
Of the more than 100 cancers responsible for the total cancer burden, the 9 cancers we examined account for more than two-thirds of all cancer deaths worldwide in 2018. 1 In investigating the extent of the transition from infectious-related to noninfectious-related cancers, we have identified specific time points of convergence in Japan and Norway, with both countries exhibiting a crossover in trends from the 2 types of cancers during the study period. Visually, one way Japan's trends are striking is that they intersect later in the period and most closely resemble an X pattern ( Figure 1A).
Despite their considerable geographic, political, and cultural diversity, all 6 countries investigated here have experienced aspects of the transition sequence and appear to have been going through a systematic pattern of change in cancer mortality that can be summarized as follows: (1) initially, infectious-related cancer mortality rates are high and noninfectious-related cancer mortality are low; (2) then, infectious-related cancer mortality starts declining, while noninfectiousrelated cancer mortality increases; and (3) subsequently, infectious-related cancer mortality reaches a low level, while noninfectious-related cancer mortality starts declining after reaching its peak.
Of primary importance in our study is the grouping and analysis of infectious-related vs noninfectious-related cancers to identify signs of cancer transition. A caveat to such an analysis is the extent to which the overall trends mask some of the underlying diversity in specific cancers. Still, despite concealing heterogeneity across countries, grouping cancers by more or less infectiousoriented causes helps to reveal a concise picture of the cancer transition.
To conceptualize the cancer transition that occurs as nations develop, we apply elements of Omran's theory of the epidemiological transition. 7 Omran proposed 3 stages in the epidemiological transition, the last being a move to the age of degenerative and lifestyle-related diseases where illnesses such as stroke and cancer rise in prominence, while infectious diseases like malaria, tuberculosis, and smallpox become less common. We consider Omran's understanding of cancer at the time as a noncommunicable or noninfectious-related disease, the classic view of cancer's causes.
In the last stage explicated by Omran, infectious diseases are increasingly brought under control, but not entirely. Historically, acute and infectious diseases tended to strike at younger ages, and when these diseases were brought under better control, people lived to increasingly older ages when noncommunicable, chronic diseases tended to strike.  viruses are causes of liver cancer. 13 One major difference between Omran's infectious diseases and the diseases associated with infection identified in our study is that the former are typically acute in nature whereas the latter are chronic and tend to develop over extended periods of time. As a consequence, the change in the mix of cancer mortality, with a shift from infectious causes to lifestyle-associated factors such as smoking, has not been temporally nested within the epidemiological transition. The cancer transition does not precisely mirror the epidemiological transition, but it is similar in the sense that the leading causes of mortality have been shifting from those associated with infections to those associated with other factors, prominent among which are individual behaviors such as smoking, drinking, diet, and physical activity.

JAMA Network Open | Global Health
With regard to H pylori, more than half of the world's population in 2015 was infected with the bacterium, 45 but only a small fraction of those infected will go on to develop stomach cancer. 13 The prevalence of H pylori is declining in most countries of the world 46,47 not through any specific medical measures, but as a result of general improvements in sanitation, housing conditions, access to clean water, food availability and freshness, and reduced salt intake. 14,48,49 The human papillomavirus is transmitted sexually, with the majority of sexually active individuals of both sexes acquiring it at some point during their lifetime. 50 Prior to the vaccine becoming available in 2006, 51,52 screening and treatment were the only means to reduce cervical cancer incidence and mortality. 53 Research is continuing into creating more affordable vaccines, which will especially help less developed countries where current levels of vaccination are a fraction of what they are in more developed countries. 52,54 Reductions of cervical cancer will only be evident decades after vaccination, but a surrogate for cervical cancer-high-grade cervical intraepithelial neoplasia-has been substantially reduced in populations with high vaccine coverage. 55 The last cancer associated with infectious causes that we investigated is liver cancer, and persistent hepatitis B and C viruses together account for approximately 90% of hepatocellular carcinoma, the most common form of liver cancer. 56 A vaccine for hepatitis B has been available since the early 1980s, and was progressively administered to infants as part of national childhood immunization programs in 90 countries in the late 1980s and 1990s. Hepatitis B decreases have been observed in most countries. 57,58 Currently there is not a vaccine for hepatitis C, but there is antiviral treatment, which cures approximately 90% of those infected with the virus. 59 To summarize, there are reasons to expect that mortality from cancers associated with infectious causes will decline even further in the future in high-to low-income countries.
As for the combined noninfectious-related cancers, we found a plateauing or decline in 5 of the 6 study countries starting between 1985 and 2000. The pattern of risk factors over the last few decades in these countries for the individual cancers are mixed. A decline in total physical activity has been documented in Japan 60,61 and the US. 62 Although there does not appear to be countryspecific data for the other nations we examine, one study found that in high-income Western countries, between 2001 and 2016, physical inactivity increased by 5%. 63 Another important trend is that Japan and France seem to have largely resisted a degradation of their diets, 64,65 but not the US 66 and England and Wales. 67 There is not enough literature to determine the extent to which food consumption has changed in Norway and Sweden.
The mean body mass index (BMI) in 2014 for Japanese men and women was in the normal range, but for the other countries, mean body mass indexes are in the overweight category for at least one sex, if not both. 68 Alcohol consumption from 1990 to 2016 nearly leveled in the US and Japan, increased slightly in England and Wales, Norway, and Sweden, and declined markedly in France. 69 In the countries we examine, smoking prevalence has declined, which has resulted in, or which is expected to result in, lower lung cancer rates with a 30-to 40-year lag. 70 Sweden has had the greatest decrease in annualized rate of change of smoking prevalence between 1990 and 2015,

JAMA Network Open | Global Health
whereas British men and French women have made the least amount of progress. 71 Importantly, the ability to both detect and treat some cancers has improved dramatically, such as for cancers of the breast 72 and colorectum. 73 Prostate-specific antigen screening remains controversial, 74,75 but treatment for prostate cancer has become more effective. 76 When esophageal cancer is detected early, evolving therapies have improved the cure rate. 77 For pancreatic cancer, mortality rates have decreased little despite advances in imaging and surgery. 78 Considering these somewhat conflicting changes in the underlying factors associated with noninfectious causes of cancer, it is difficult to predict how fast mortality will decline in the future and how low a level it will ultimately reach. However, we found no reason to expect a reversal of their favorable trends in the study countries.
Our study's findings support and extend the hypothesis of Gersten and Wilmoth 14 that the cancer transition can be considered analogous to the epidemiological transition. Put another way, one can say that the cancer transition parallels the epidemiological transition in that there is an epidemiological transition also within the complex system of cancers linked to specific causes. This linkage between the epidemiological and cancer transition theories is presented schematically in the Table. A main implication of our analysis, then, is that although Omran's theory 7 does not apply to part of the third stage of cancer transitions, in which noninfectious-related cancers have peaked and have, in some settings, begun to decline, cancer remains bound in epidemiological transition theory.
Indeed, the theory continues to remain a useful framework for understanding fundamental mortality patterns for cancer, a leading cause of death worldwide.

Limitations
This study had some limitations. Our analysis was constrained by the lack of comparable cause-of-

Conclusions
We examined cancer mortality trends in the US, select European nations, and Japan, and found substantial support for the cancer transition theory. This theory claims that as nations develop, rates of infectious-related cancers decline, whereas rates of noninfectious-related cancers increase. These trends parallel, and show the continued relevance of, the originally formulated epidemiological transition theory, which has had considerable influence on a number of academic fields.