The incidence trends of oral cancers worldwide from 1988 to 2012 and the prediction up to 2030

This paper aims to analyze the time trend of OCs incidence in 43 countries (1988–2012) and predict the incidence trend of OCs (2012–2030).


| INTRODUCTION
Head and neck squamous cell carcinoma (HNSCC) ranks as the sixth most common malignant tumor worldwide. In 2020, HNSCC caused more than 800 000 new cancer cases and 450 000 deaths. 1 Oral cancer (OC) is one of the main types of HNSCC. Over the past 30 years, the incidence of OCs has remained roughly the same. 2 However, due to differences in geographical environment, lifestyles and customs, and the degree of economy, the incidences vary significantly across different countries, with the highest incidence of OCs in South Asia. In addition, Asia has a relatively high OC burden due to its higher population density. 3,4 The occurrence of OCs is closely related to factors such as smoking, alcohol consumption, smokeless tobacco chewing, and poor oral hygiene. 5 Tobacco and alcohol are commonly recognized as risk factors for OCs. In Western countries, approximately 75% of OCs patients have a history of smoking and drinking. 6 Numerous studies have shown that quitting smoking and drinking can significantly reduce the risk of OCs. 7,8 Almost 600 million individuals chew areca nut worldwide, especially in South Asia, Southeast Asia, and the Pacific islands. The number and duration of chewing areca nut and the chewing duration are both significant risk factors for OCs. 9 Smoking, drinking, and chewing smokeless tobacco can also lead to poor oral hygiene, increasing the risk of OCs. 10 Regarding human papilloma virus (HPV) infection, unlike the high correlation of HPV (30.8%) with oropharyngeal cancers, the correlation between HPV and OCs is only 2.2%. 11 OCs have high incidence and recurrence rates. 12 Since the tongue and other organs in the oral cavity have rich lymphatic networks, the lymphatic metastasis rate of OCs is also high. 13 Most cases are in advanced stages at the time of diagnosis, which seriously affect the patients' postoperative survival due to large coloboma on the head and face after resection. 14 Meanwhile, the heterogeneous nature of OC, as a solid tumor, makes it difficult to have effective nonsurgical therapies such as targeted therapy. 15 In addition, the low correlation of HPV increases the difficulty of oral cancer vaccine development. 16 Therefore, OCs are likely to be a more serious social burden, and predicting the incidence trend of OCs is important in cancer management. Because few studies have predicted the future incidence of OCs worldwide. This study investigated the trend of OC incidence to predict the future incidence of OCs in various countries and regions worldwide and provide guidance and reference for social and medical preparations in advance.

| Data resource
The database on the temporal trend for Cancer Incidence in Five Continents (CI5plus) was developed and maintained by the International Association of Cancer Registries, a professional association dedicated to promoting the goals and activities of global cancer registries. 17 The CI5plus database deals with the number of new cancer cases classified by year, age group, gender, and cancer site, and the corresponding risk populations in the 124 selected populations (denominator) from 108 cancer registries published in the CI5 monographs, providing a database for exploring the morbidity and incidence trend of OCs worldwide. The International Agency for Research on Cancer has established strict data quality criteria based on comparability, integrity, accuracy, and timelines. The recorded and registered data that fulfill the standards would be included in CI5. Therefore, the quality of data sources must be regularly evaluated. 18 During 1998-2012, the CI5 register covered 14% of the world's population, with coverage rates of 2% for Africa, 6% for Asia, 8% for Central and South America, 42% for Europe, 78% for Oceania, and 95% for North America. Thereinto, the coverage rates of Europe, Oceania, and America were significantly higher than those of other regions. The cancer site dictionary contains 170 diagnostic units, which comprise cancer sites at the third digit level of the International Classification of Diseases, 10th revision (ICD-10) categories, and, for some 13 cancer sites, by histological subtype. Different countries are regrouped according to their geographic locations, including Africa, America (Caribbean, Central America, South America, and North America), Asia (East Asia, Southeast Asia, South Asia, and West Asia), Europe (Eastern Europe, Northern Europe, Southern Europe, and Western Europe), and Oceania.

| Definition of OCs
We focused on the malignant tumors originating from oral mucosal epithelium; therefore, we excluded adenocarcinomas of parotid glands, tonsils, and so forth. Meanwhile, we excluded oropharyngeal malignant tumors with a higher P16 (HPV) positivity rate due to different therapeutic methods and prognoses.
Based on the classification of head and neck tumors by the National Comprehensive Cancer Network (NCCN), 19 we determined the specific sites of OCs and extracted the data on the lip (C00), tongue (C01-C02), and mouth (C03-C06) from the CI5plus database according to ICD-10. We aggregated the data on OC cases at the country level if data from more than one registry were available in one country. As the availability of registered data was different across countries, we selected the period from 1988 to 2012 to ensure that most of the studied countries had data points within this period. It is worth noting that in the NCCN OC, the site of tongue cancer has been defined as the anterior two-thirds of the tongue; however, since most tongue cancers occur in this area and are treated with similar methods, all the data on the tongue were included in this study.

| Statistical analysis
We calculated the age-standardized incidence rate (ASR) of OC per 100 000 people for all age groups and cut-off age groups (40-59 and ≥60 years of age) using the world's standard population from the World Health Organization (WHO). The calculation is based on the number of annual new cases with a specific gender in a specific country of the 5-year age groups (0-4 to 85+ years). The human development index (HDI), a useful classifier of cancer globalization, involves the key dimensions of human development, such as long and healthy life, extensive knowledge, and decent living standards. 20 The selected countries are divided into two categories, 21 high-income countries and 22 middle-and low-income countries, with the median HDI (0.88) as the cut-off value.
We used Joinpoint regression 21 to quantify the overall incidence trends, which fitted a series of connected straight lines to the incidence to test the statistically significant changes in the trends. Then, the regression line was fitted to the natural logarithm of the rate over calendar years to calculate the annual percent change (APC) for each identified trend: where β i represents the slope of the cycle. We also estimated the average APC (AAPC) under the assumption that only one phase was involved in the entire study period. 22 The Bayesian age-period-cohort model was used to predict the incidence of OC by 2030. 23 Bayesian methods are less dependent on parametric hypotheses and can solve the problems that age-period-cohort models frequently face concerning identification due to the invalid parametric constraint hypotheses. In addition, correlated Bayesian ageperiod-cohort models with a priori random walk smoothing have been developed. 24 It is hypothesized that the a priori random walk 1 (RW1) presents a constant trend over time scales, while the a priori random walk 2 (RW2) exhibits a linear time trend. Studies have shown that models with a priori RW2 give more reliable predictions than those with a priori RW1. 25 Considering that one of the main goals of this study was to predict the incidence of OC by 2030, a Bayesian age-period-cohort model with a priori RW2 was adopted. The Joinpoint software version 4.9.1.0 (Surveillance Research and Applications Branch, National Cancer Institute, Bethesda, MD) was used for analyses. Other analyses and data visualizations were performed using R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria).

| Observed trends
The data in this study covered 108 cancer registries in 43 countries. Figure 1 showed the worldwide distribution of the data in this study and the number of cancer registries in each selected country. Figure 2 presented the temporal trends and regional distribution of OC incidence in the selected countries from 1988 to 2012. Among the 43 countries selected, eight countries had an ASR >6/100 000, 11 had an ASR <3/100 000, and 24 had an ASR between 3 and 6 in 1988. The countries with ASRs >6/100 000 in 1988 were considered the key countries for OC occurrence and ranked from high to low ASR as follows: India, Spain, France, Slovakia, Australia, Switzerland, Philippines, and Germany, among which India, Spain, and France were three countries with the highest ASR (ASR >9/100 000), and the ASRs for the rest of the key countries were <8/100 000. In 2012, the ASRs in these countries showed a declining trend, most of which fell below 6, and that of the Philippines dropped to below 3/100 000 (1.5/100 000). India and Australia were the exceptions, where the incidence declined but still remained at a high level. In addition, for the countries with ASRs in the 3-6 range, the ASRs exhibited no significant change, among which Martinique, Columbia, Malta, Israel, and Bahrain were the exceptions, and those with ASRs <3/100 000 still maintained a low incidence. We arranged the countries by regional division. It can be seen from the heat map that the morbidity of OC was mainly concentrated in South Asia, Central Europe, and Oceania, while East Asia, the Middle East, and South America had a low risk of OCs. Table 1 shows the ASRs of OC per 100 000 between 1988 and 2012 for both genders. Among them, India had the highest ASR (13.7/100 000, 11.83/100 000) during this period. In terms of regions, South Asia and Oceania had the highest ASR in 1988 (9.24/100 000) and 2012 (6.74/100 000). From 1988 to 2012, the ASRs of most countries showed a declining trend, with the Philippines exhibiting the most significant decreasing tendency (from 6.36/100 000 in 1988 to 1.52/100 000 in 2012). Among the selected 43 countries, only 14 countries showed an upward trend of ASRs from 1988 to 2012, that is, Estonia, the Czech Republic, Netherlands, UK, Italy, Demark, Norway, Brazil, Ecuador, China, Japan, Cyprus, Kuwait, and Turkey, among which Kuwait presented the most significant increasing amplitude (from 1.58/100 000 in 1988 to 2.99/100 000 in 2012). In addition, gender-specific ASR trend charts were presented for different countries in Figure 3, and were classified by the national development index. It was found that the ASR of OC in males was much higher than in females in most countries. However, from 1988 to 2012, the ASR of OC in males decreased significantly in most countries, while it    Table 3 shows the classification of all countries by HDI. The ASR of low-and middle-income countries decreased more than that of high-income countries, with a percentage change of À48.62% versus À15.78% in ASR. In addition, the median decline amplitude in ASRs in highincome countries (median ASR of 4.65/100 000 in 1988, median ASR of 4.41/100 000 in 2012, with a percentage change of À5.16% in median ASR) was also lower than that in low-and middle-income countries (median ASR of 4.63/100 000 in 1988, median ASR of 2.68/100 000 in 2012, with a percentage change of À42.16% in median ASR). It is worth noting that India, as the country with the highest incidence, still exhibited a lower declining rate in ASR of OC (À13.66%) compared to the overall higher declining rate in low-and middle-income countries. Note: Age-standardized incidence rates, using the world standard population (per 100 000 person-years). Abbreviation: ASR, age-standardized incidence rates.  3.5 | Trends for age groups Figure 4 shows the incidence trend of OCs in the young (aged 40-59) and elderly (aged ≥60) groups in high-income and low-and middle-income countries. In all the selected countries, the ASR for the elderly group was higher than that in the young group, and in most of the selected countries, the young group exhibited a changing trend similar to the elderly group. There were 21 countries where both the incidence rates in young and elderly groups decreased, among which the Philippines exhibited the most significant decrease amplitude (young group: À70.25%; elderly group: À77.89%). On the other hand, in nine countries, the incidence rates in both groups increased, among which Kuwait exhibited the most significant increasing amplitude (young group: 115.74%; elderly group: 66.86%). Regarding regions, only East Asia exhibited an increasing trend in both age groups. Concerning age groups, the ASRs of the young group in 15 countries increased, with the most significant increase in Ecuador (207.25%), while that in 28 countries decreased, among which Bahrain exhibited the most significant decrease amplitude (À72.12%). Regarding the elderly group, 16 countries displayed increased ASRs, among which Estonia exhibited the most significant increase amplitude (60.16%), and 26 countries showed a downward trend, with the most significant decline in the Philippines (À77.89%). The detailed data are presented in Table 4. Figure 5 shows the variation trends of the OC sites in selected regions. From a global perspective, the overall incidence of OCs in the mouth and lips showed a downward trend, while that at the tongue showed an upward trend. In Europe, except in Western Europe, the incidence of OC in the mouth, tongue, and lips has decreased in descending order in recent years. In Asia, the incidence of oral cancers in the mouth in South Asia showed a rapidly declining trend, and that of the tongue and lips maintained a relatively stable low level in East Asia and the Middle East. However, unlike the extremely low incidence of OCs at the lips in East Asia, there was a period of a substantial decline in the Middle East. In addition, in North America, OCs in the tongue had the highest incidence, and Oceania exhibited an exceptionally high incidence of lip cancers and the lowest incidence in the mouth. Figure 6 shows the incidence of OCs from 1988 to 2012 in different countries and the predicted incidence from 2012 to 2030 according to the Bayesian age-period-cohort model. Southern Asia, Western Europe, Eastern Asia, and Oceania were the four regions with a rising incidence of OCs. Among them, the ASR of OCs in Southern Asia and Western Europe would reach about 10/100 000. In contrast, the rising ASR of OCs in Oceania was relatively mild, reaching about 8/100 000 in 2030. The ASR of OCs in Eastern Asia was at a lower level for a long time; however, according to the prediction, the ASR of OCs in Eastern Asia will reach approximately 5/100 000 in 2030. The incidence of OCs in other regions exhibited stable or slowly declining trends, and the ASR of OC in 2030 will be less than 5/100 000. In Europe, the UK, and the Czech Republic were predicted to be with the fastest increase in the ASR of OCs, reaching 15/100 000 and 10/100 000 in 2030, respectively. The ASR of OC in Austria, Switzerland, Norway, and Poland also exhibited a rising trend. Italy and Croatia were the two countries with the most significant declining trend. According to our prediction, India will have the highest ASR of OCs worldwide, reaching about 23/100 000 in 2030.

| Trends of prediction
Thailand and Japan are the other two countries in Asia where the ASR of OC showed a rising trend. The predicted ASR of OCs in the remaining countries of Asia was low (below 5/100 000). The incidence of OCs in Oceania and America showed no significant change in the prediction, and that in Australia will remain high in 2030.

| Main findings
This study comprehensively profiled the morbidity and trend of OCs worldwide and conducted predictions by the Bayesian model. It should be noted that some countries are predicted to remain a higher incidence and a rising trend, although there has been a general improvement in the control of tobacco and alcohol consumption, the popularization of HPV vaccination, and social hygiene and medical insurance levels. 26 From 1988 to 2012, South Asia and Oceania were always the F I G U R E 4 Temporal change in age-standardized incidence rates of OC between 1988 and 2012 divided into upper part of high-income countries and lower part of middle-and low-income countries. regions with a high incidence of OCs. Meanwhile, the predicted incidence of OCs in Central and South Asia and Oceania by 2030 will still maintain a higher level. Another country deserving attention is Japan, where the incidence of OC maintained a gradually rising trend from 1988 to 2030. Concerning Europe, countries deserving attention are the United Kingdom, the Czech Republic, and Austria. The incidence of OCs in the remaining countries in Europe and the entire America exhibited a downward trend or remained at a low level.
The following inspirations were achieved by analyzing the incidence of OCs from different dimensions: First, the incidence of OCs showed apparent discrepancy in different regions, genders, and different age groups. Second, the incidence of OCs at different sites exhibits obvious regional characteristics. Third, the incidence of OCs worldwide presented an overall trend of decreasing and an increasing trend in local regions. Additionally, compared with HDI, regional characteristics are a factor with a greater correlation to the incidence of OCs. Therefore, studying the incidence of OCs in different regions is more conducive to preventing and treating OCs.

| Discussion of findings
Our studies predicted that South Asia would bear the greatest burden of OCs, which was similar to other studies. 26,27 Since smokeless tobacco involves chewing and maintaining tobacco in the mouth, long-term contact with the oral mucosa results in a higher correlation with OCs than tobacco smoking and alcohol. 28 South Asia is the region with the largest consumption of smokeless tobacco all over the world, and the consumption of smokeless tobacco may destroy the normal flora in the oral cavity, which is the main reason for the high incidence of OCs. 29,30 In Bayesian prediction model, India has the highest OCs burden. Other studies have also suggested this result. 31 Considering that the Indian government has made some progress in controlling tobacco and improving sanitation. There are two possible reasons for the increase in oral cancer in India. 32 First, Tobacco use among young adolescents remains an important public health challenge worldwide. Studies show that tobacco use among young people in South Asia is still on the rise. 33,34 Worryingly, adolescents in India do not have a clear understanding of the harms associated with areca nut and believe that areca nut is different from tobacco and has less impact on health. 35 Secondly, compared with the higher consumption of tobacco and alcohol in males than in females, there is no gender difference in the consumption of smokeless tobacco in regions with a tradition of smokeless tobacco consumption, which contribute to the higher incidence of OCs in females in India than other regions. 36 Therefore, education about OCs is as important as policies.
Eastern Asia was one of the regions with a low incidence of OCs from 1988 to 2012. However, according to our prediction, the OC incidence in Japan shows a gradually increasing trend by 2030, consistent with previous studies. 37 Likewise, smoking and alcohol drinking are also independent risk factors for the morbidity of OC in Japan. 38,39 In addition, it is worth noting that heated tobacco products are more popular among young and wealthy people in Japan. 40 Thus, controlling such new tobacco products may be the focus for Japan to control the morbidity of OCs in the future.
Oceania has always maintained a high level of OC incidence. Due to the geographical proximity of Oceania and South Asia, developed countries like Australia have more immigrants from South Asia, and studies have shown that their knowledge about OC is at a relatively low level. 41,42 In addition, the HPV infection rate of indigenous people in Oceania far exceeds the international average levels, which is also one of the reasons for the high incidence of OCs. 43 However, the most relevant causes may be the high rates of smoking and drinking in Oceania, and research has shown that 15 525 (13%) cancer cases in 1 year in Australia could be attributed to smoking, of which 1973 (59%) cases belonged to OCs and pharyngeal carcinomas. 44 This phenomenon is particularly remarkable in northern Australia. 45 With changes in public health policies in countries such as Australia, risk factors like tobacco and alcohol consumption can be further controlled, which may prohibit the increase of OC incidence in the future. 46 In the present study, the incidence of OCs in the United Kingdom also exhibited an increasing trend, consistent with previous studies. 47,48 We explored the potential reasons for this phenomenon. First, the United Kingdom has many immigrants from South Asian countries such as India, Pakistan, and Bangladesh. The incidence of OCs was higher among South Asian immigrants due to lifestyle habits and economic levels. 49 Second, studies have shown that in the UK, in addition to the factor of Asian immigration, deprivation, and heavy drinking are closely related to the incidence and mortality of OCs. 50 Among them, deprivation indicates a lower educational level and insufficient recognition of OCs. 51 Concerning the analysis of other dimensions, although the incidence of OCs reflected a clear gender distinction, the incidence of OCs in males showed a significant declining trend. In addition to the higher consumption of tobacco and alcohol by males, the reason for gender differences also included the males' poorer oral hygiene habits than the females'. 52 Therefore, the inhibitory effect on OC in males is apparent with the development of medical care and public health, the popularization of health concepts, and the decreased consumption of tobacco and alcohol. Our study demonstrated that the incidence of OCs was higher in the elderly population than in the relatively younger population, but the change trends were basically the same in different countries. Regarding OCs at different sites, compared with other regions, we noticed that Oceania and North America exhibited the highest incidence of OC at the lip and tongue, respectively. Lip and tongue cancers were the first cancers that were reported to be associated with smoking, and the high incidence in North America and Oceania might be related to the higher smoking rates in these two regions. 26,53 4.3 | Strength, limitations, and future directions Compared with previous epidemiologic studies on OCs, this study used a robust Bayesian age-period-cohort prediction model to make a longer-term prediction of the incidence of OCs. 54 Similar approaches have been adopted and proved feasible in predicting other cancers. 55,56 The data results presented in this study could be used to predict the burden of OCs and provide guidance for medical hygiene in various regions. However, this study still has certain limitations. First, this study conducted a secondary data analysis and was limited to the data source. Most selected countries were in North America and Europe, and the coverage of high-quality data on Asian, African, and South American countries was limited. Second, The CI5plus data is only updated to 2012, however, because of the comprehensiveness of its data, it is difficult to find a replacement for the CI5plus data. Third, the Bayesian age-period-cohort prediction model could only make predictions based on the provided underlying age distribution trend for the incidence but could not accurately predict the impact of some public concerns, such as the COVID-19 pandemic or the impact of policy changes on OCs. Therefore, the results of this study can only be used as a reference for epidemiological research rather than as the basis for determining future incidence rates. As a result, it is necessary to establish an incidence diagnosis and prediction model associated with the incidence-related factors and the public health system of each region to better analyze the incidence of OCs and reduce bias.

| CONCLUSION
There are significant regional differences in the incidence of OCs It should be noted that countries such as India, Thailand, the United Kingdom, Austria, the Czech Republic, and Japan showed a rising trend. Meanwhile, the incidence in Oceania still remain at a higher level. Considering the well-known risk factors of OCs such as smokeless tobacco and alcohol, further actions and policies can be made to reduce their consumption, and knowledge related to oral hygiene and OCs should be further popularized worldwide, especially in countries exhibiting a rising incidence of OCs.

AUTHOR CONTRIBUTIONS
Dong-Ning Lu and Wan-Chen Zhang was a major contributor to writing the manuscript. Yi-Ning Zhang, Cheng-Ying Shao, and Yan-Ze Lin were responsible for the acquisition of the presented data and were involved in the revision process of the manuscript. Chuan-Ming Zheng was responsible for designing the manuscript. Ming-Hua Ge analyzed and interpreted the data. Jia-Jie Xu and Ming-Hua Ge participated in the extensive revision process of the manuscript. All authors have read and approved the final manuscript.