Recent and Projected Trends in Oral Tongue Cancer in the United States: A Demographic Shift in Case Burden as Early Onset Increases Among Females Subside

Abstract Background Oral tongue cancer (OTC) incidence has increased rapidly among young (< 50 years) non-Hispanic White (NHW) individuals in the United States (U.S.) over the last two decades; however, it is unknown if age-associated trajectories have persisted. Furthermore, incidence trends for all 50 U.S. states and the District of Columbia have never been investigated. Materials and methods Using U.S. Cancer Statistics data, we investigated incidence trends from 2001–2019, overall and according to age, sex, race/ethnicity, and state of residence. We used age-period-cohort analysis to explore temporal patterns among birth cohorts and to project future trends and case counts. Results OTC incidence increased across all age, sex, and racial/ethnic groups, with marked increases observed among the NHWs (2.9%/year; 95%CI, 2.2%-3.7%). Incidence among NHWs increased in most U.S. states, particularly in the Southeast. Increases were significantly greater among NHW females compared to males (3.6%/year vs 2.6%/year; P  = 0.022). Increases among females aged 50–59 years were most notable and significantly outpaced increases among younger females (4.8%/year [95% CI, 4.1%-5.4%] vs. 3.3%/year [95% CI, 2.7%-3.8%]; P  < .001). While both NHW male and female birth cohorts from 1925 to 1980 saw sustained increases, rates stabilized among females born after 1980. Should trends continue, the burden of new OTC cases among NHWs in the U.S. is projected to shift to older individuals (33.1% versus 49.3% aged ≥ 70) and females (86% case increase versus 62% among males). Conclusion The period of rapidly increasing OTC incidence among younger NHW females in the U.S. is tempering and giving way to greater increases among older females, suggesting that a birth cohort effect may have influenced previously observed trends. Recent increases among NHWs aged ≥ 50 of both sexes have matched or outpaced younger age groups. Continuing increases among older individuals, particularly females, will lead to a shift in the OTC patient profile over time.


INTRODUCTION
Over the past 25 years, head and neck cancer incidence has been declining in the United States (U.S.) with decreasing prevalence of tobacco use. 1 Two head and neck cancer subsites-oropharyngeal cancer squamous cell carcinoma (OPSCC) and oral tongue cancer (OTC)-have been the exception.3][4] The rise in OPSCC incidence has been attributed to human papillomavirus infection (HPV). 2 A corresponding cause for OTC has yet to be identi ed. 4,5e rising incidence of OTC in the U.S. was rst recognized to be occurring among young non-smokers. 6bsequent studies over the past two decades con rmed a rapid rise in OTC incidence among young non-Hispanic White (NHW) individuals, particularly among young females without the traditional risk factors of heavy tobacco or alcohol use. 3,7][10][11][12] This has led many in the eld to consider early onset OTC (diagnosed at < 50 years) as a distinct clinical entity, with research primarily focused on understanding the differences between early versus typical onset (diagnosed at ≥ 50 years) OTC. 3,4,8,12Previous U.S. studies quantifying OTC incidence trends in the U.S. used the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) Program data, which captures at most 48% of the U.S. population. 3,4,8,13,14The relative rarity of OTC combined with the limited size of the SEER Program database has restricted the ability to study more granular patterns by age and geographic region, potentially masking important information about the geographic distribution of cases and evolving age trends.With respect to OPSCC, recent studies have reported that the increasing incidence among younger individuals has moderated and the case burden is now shifting to older age groups. 15It remains unknown if OTC is also following a similar pattern.
The objective of this study was to provide a comprehensive analysis of recent temporal and geographic trends in OTC incidence in the U.S. and to project future case burden.To accomplish this, we used the 2001-2019 U.S. Cancer Statistics (USCS) Public Use Database, which covers 98% of the U.S. population, allowing us to consider trends in narrower age groups across both sexes and provide the rst state-level analysis of OTC incidence.

Data Sources
We analyzed incident cancer cases from the USCS Public Use Database for the years 2001 through 2019. 16This database contains submissions from the Center for Disease Control and Prevention's (CDC) National Program of Cancer Registries (NPCR) and the NCI SEER program, covering approximately 98% of cancer cases in all 50 states and the District of Columbia during this time.

Study Population
We identi ed oral tongue cancer cases using topography codes C02.0-02.3 from the International Classi cation of Diseases for Oncology version-3 (ICD-O-3). 17We excluded cases classi ed as C02.8 (Overlapping lesion of tongue) and C02.9 (Malignant neoplasm: Tongue, unspeci ed) to avoid possible misclassi cation of oropharyngeal base of tongue tumors as oral tongue tumors.Staging was applied according to SEER Summary Stage 2000.We included only malignant tumors and of all histological subtypes.This study did not use personal identifying information from the USCS Public Use Database or involve interaction with human subjects; and thus, was deemed exempt from institutional review by the University of Kentucky Institutional Review Board.
We analyzed patient demographics using the following variables: age at time of diagnosis, sex, race/ethnicity, and state of residence at time of diagnosis.Due to limited case numbers, available racial/ethnic categories were reduced to Hispanic, non-Hispanic Black (NHB), non-Hispanic White (NHW), and other non-Hispanic races (Other NH).For some analyses we further categorized state of residence into Southeast, Southwest, Northeast, Midwest, Rocky Mountain, Paci c, and Non-contiguous regions.

Statistical Analysis
Incidence rates were age-adjusted to the 2000 U.S. standard population and given per 100,000 persons using SEER*Stat version 8.4.0.Rates were suppressed for any subpopulation and year/year range containing fewer than 16 cases.We estimated OTC incidence rates by sex, age, race/ethnicity, stage at diagnosis, and region of residence at diagnosis.We used Joinpoint software version 4.9.1.0to calculate trends in OTC annual incidence rates and to estimate annual percentage changes (APC) and average APCs (AAPC). 18Joinpoint regression ts piecewise log-linear trends to the rates using a permutation test and provided estimates for APC of segments and AAPC as a weighted average of the APCs.To determine whether the AAPC trends were different from zero, a t test was conducted for zero joinpoints, and a z test was conducted for one or more joinpoints.Statistical signi cance was assessed at a level of P < 0.05, and all hypotheses were two-sided.For the state-level analysis, we estimated OTC incidence rates for 2001-2005 and 2015-2019 by sex, and state of residence at diagnosis.Temporal changes were quanti ed using incidence rate ratios (IRR).
We performed age-period-cohort analysis to estimate OTC trends by age, period, and birth cohort.We accessed case counts and population estimates for the 5-year ranges of 2005-2009, 2010-2014, and 2015-2019 from the USCS Public Use Database in SEER*Stat.The use of 13 age groupings and three calendar periods resulted in 15 partially overlapping birth cohorts (age-period-cohort analysis was restricted to ages 20 and older to ensure non-zero case counts in all age groups used).Age-period-cohort models were built using the Age Period Cohort Analysis web tool created by the NCI. 19To project the future burden of OTC, we used the same case counts and population estimates as above along with population projections for the 5-year ranges of 2020-2024, 2025-2029, and 2030-2034 from the U.S.
Census Bureau via CDC WONDER. 20Case projection estimates were calculated using the package (version 1.1) with Poisson link function in the R statistical programming language (version 4.0.5). 21,22Sensitivity analysis was performed on case projections by using the same model on data from 2001-2015 to project the average annual number of new cases for 2016-2020.This was compared to the average annual number of new cases observed in 2016-2019 (Additional File 1: Figure S1).For males, the model t to the 2001-2015 data predicted the average annual observed case count to within 4% for all but one age group, and all age groups within 8%.For females, the n or dpred n or dpred model predicted the average annual observed case count to within 5% for all but one age group, and all age groups within 10%.The 'Under 30 years' age group was omitted from sensitivity analysis because case counts used for projections were known underestimates due to suppression.

Patient Characteristics
There were 58,661 new cases of OTC identi ed between the years of 2001 and 2019.Men (57.6%), non-Hispanic Whites (83.7%), those aged 60 years or older (58.0%), and individuals with localized stage disease at diagnosis (62.7%) comprised the majority of cases (Additional File 2: Table S1).

OTC Incidence by Sex, Age, and Race/Ethnicity
While rates across all combinations of sex and race/ethnicity generally increased with age, the most notable increases were observed among NHWs (Fig. 1).Between 2001 and 2019, OTC incidence among NHW persons increased 2.9% per year (95% CI, 2.2-3.7%;Table 1).OTC incidence rate ratios increased among all age groups 30 years and older for both NHW males and females between 2001-2005 and 2015-2019 (Additional File 3: Table S2).OTC incidence remained largely the same among other racial/ethnic groups over the same period.Given the presence of substantial OTC incidence rate increases across NHW sexes and age groups compared to other races/ethnicities, all subsequent analyses were restricted to NHWs.
Relative declines in OTC incidence were observed in only four states (California, Maine, Nevada, Hawaii) for males and two states (New Mexico and Maine) for females.These ndings are consistent with APC estimates on regions with available data (Additional File 7: Table S5).The most marked increases occurred in the Southeast where OTC incidence rose by 4.3% per year among males and by 5.5% per year among females.

Cohort-speci c Trends in OTC Incidence among NHWs
Age-period-cohort analysis revealed increasing trends in OTC incidence among both NHW male and female birth cohorts between 1925 and 1995 (Fig. 4; Additional File 11: Figure S5 and Additional File 12: S6).The most recent birth cohort considered (individuals born between 1995 and 1999) had IRRs of 6.5 (95% CI, 3.9 to 10.7) for males and 5.

DISCUSSION
Our study is the rst comprehensive analysis of nationwide OTC patterns using the 2001-2019 USCS Public Use Database, which covers nearly 100% of the U.S. population in all 50 states and the District of Columbia.Using this nationwide data, we are the rst to show that the most substantial changes in OTC incidence rates have been concentrated in the Southeast region of U.S., particularly among Kentucky, Maryland, Tennessee and North Carolina.We are also the rst to show a substantial shift in incidence rate increases and case burden from early onset OTC among NHW females to NHW females aged 50-59 years.We show through analysis of birth cohorts that rate increases have attenuated in NHW females born since 1980.Incidence rates among NHW males continued previously observed increases, with a signi cant long-term trend of increase occurring among males aged 50 and older, but no substantial shift in overall case burden demographics.Altogether, this suggests a recent change in the yet to be identi ed risk factors underlying the previously documented rise in early onset cases among NHW females and calls for reevaluating the early-versus-typical onset paradigm that has dominated recent OTC literature.
Using observed cases strati ed by age and year of diagnosis, and accounting for the estimated distribution of the U.S. population over the next decade, we project continued and substantial increases in the total burden of new OTC incidence among both NHW males and females in future years.Should these trends continue over the next decade, new oral tongue cancer cases in the U.S. will increasingly be diagnosed among older individuals and the burden of disease will be shared more evenly between male and female patients than currently observed.
This study is consistent with and adds meaningful updates to prior research that found alarming increases in OTC incidence among younger NHW females in the U.S. In particular, Patel et al. showed rapid OTC incidence increases among 18-44 year old NHW females and stable rates among females > 44 years using data from 1975-2007. 3Tota et al. followed by nding rapid increases among NHW females < 50 years and the beginnings of increases in females ≥ 50 years using data from 1973-2012. 4In both studies, the younger age group was de ned by individuals born after 1962.The current analysis adds an additional 7 years of cancer registry data (2001-2019), with some of the "younger females" identi ed by Patel et al and Tota et al now falling into the 50-59 years old age group.We report NHW females aged 50-59 as having experienced the largest increases in age-speci c incidence-signi cantly outpacing increases among females less than aged 50.This, combined with the ndings from our ageperiod-cohort analysis, suggests that a cohort effect may explain the previously observed increases in early onset OTC among females and argues for a shift away from the early versus typical onset dichotomy that has dominated OTC research over the last 20 years.
This study also shows similarities to patterns observed among oropharyngeal cancer in the U.S., both in terms of the geographic distribution of cases and the recently reported moderation of increasing OPSCC incidence rates among younger individuals. 15,23Currently, the factors driving OTC incidence increases remain unknown.Similarities with OPSCC incidence suggest there may be shared risk factors for developing these two subtypes of head and neck cancer.HPV infection as a common etiology is unlikely given prior studies demonstrating a lack of HPV detection in OTC tumors. 5,24,257][28][29] Our ndings, which provide the rst data on the geographic distribution of OTC incidence increases as well as the rst evidence of a possible birth cohort effect, may help focus future research efforts towards identifying etiologic factors.
The results of this study have several meaningful implications for clinical practice.The pro le of those at risk for OTC continues to evolve, with our ndings indicating a need for heightened awareness and screening of older NHW persons and persons residing in the Southeast U.S. Surgical resection is the mainstay of treatment for OTC, which may be limited in older individuals, leading to a need for novel therapeutic approaches in older populations.1][32] It is worth considering whether these characteristics will follow the same shift in age patterns as we report with incidence.It's also worth noting that some studies have shown decreased relative and overall survival among OTC patients aged 50 or older. 11,33,34Lastly, although we have shown a leveling off in risk increases among younger NHW females, the overall rate of OTC among both NHW males and females remains much higher than historically observed and will continue to present at greater frequencies across patients of all ages.
Our study had several limitations.First, individual-level risk factors, such as HPV, smoking, obesity, and genetic conditions, are not routinely captured in cancer registries.Therefore, the impact of these and other risk factors on increasing incidence cannot be directly measured, leaving us to speculate on potential etiologies.Second, the possibility that HPV-related base of tongue cancers may have been misclassi ed as cancers of the oral tongue could lead to an overestimate of OTC incidence.To mitigate this risk, we excluded overlapping lesions of the tongue and malignancies not otherwise speci ed.The fact that observed increases in our study occurred in both NHW males and females while HPV-related base of tongue cancer has primarily been increasing in NHW males provides further evidence against the impact of possible misclassi cation.Third, limited case numbers among birth cohorts from 1990-2000 may result in unstable age-period-cohort estimates among younger individuals under 30 years old during the timeframe of this study.Fourth, the results of future case projections should be interpreted with caution, given limited observed case data among individuals under 20 years of age.Despite these limitations, the greatest strength of our study is the use of nationwide, high-quality, population-based registries that allowed us to perform analyses on the most comprehensive and contemporary OTC incidence data available in the U.S.

CONCLUSIONS
This study shows an epidemiologic shift in recent patterns of OTC incidence in the U.S. Though rates of OTC are still broadly increasing across the population, the alarming rise of early onset cases among young NHW females appears to be giving way to rapid increases among females aged 50 years and older -indicating that previously observed increases seen among younger individuals could be attributable to a birth cohort effect.Given recent trends and the current age distribution of the U.S. population, the population of new OTC cases will likely become older and more evenly split between males and females over the next decade, which has important implications for future research as well as diagnosis and treatment.

Figure 1 Age
Figure 1

Figure 2 State
Figure 2

Figure 3 Trends
Figure 3

Figure 5 Age
Figure 5

Table 1
Oral tongue cancer incidence (2001-2019) among non-Hispanic White persons: National Program of Cancer Registries and Surveillance, Epidemiology, and End Results Datasets Rates were calculated as the number of cases per 100,000 person-years and age-adjusted to the 2000 US standard population; a Cases include malignant tumors that matched selection criteria b to future population estimates (Additional File 15: TableS9).Overall, females are projected to experience a larger increase in cases than males (86% increase versus 62% increase, respectively).While males were diagnosed with 35.9% more cases of OTC than females in 2015-2019, the above shifts combined with changing population age demographics project to shrink this discrepancy by almost half (to 18.4%) in 2030-2034.
7 (95% CI, 3.4 to 9.7) for females when compared to individuals of the same sex in the 1940 reference birth cohort.Both NHW males and females experienced IRR increases in successive birth cohorts from 1925 to 1955 (IRR Male 4.2% [95% CI, 4.0-4.5%],IRRFemale4.4% [95% CI, 4.1-4.7%]).Male IRR increases tempered slightly starting with the 1955 birth cohort, though signi cant increases per successive 5-year birth cohort continued through the end of the observed period.Increases in successive female birth cohort IRRs maintained the same trajectory from 1925 until the 1980 birth cohort, at which point they stabilized across remaining birth cohorts.This mirrors our results from the earlier age-speci c trends, which showed no signi cant annual incidence rate increase among females under 30 years of age between 2001 and 2019.Females 80 + years: 3.6 to 6.0 per 100,000) (Figs.5A and 5C; Additional File 14: TableS8).This would result in the case burden skewing more heavily towards individuals aged 70 years and older (47% of total male cases in 2030-2034 versus 30% in 2015-2019, and 54% of total female cases in 2030-2034 versus 37% in 2015-2019).To ensure the shift in case burden was not simply the result of changing age demographics, we compared our projections to projected cases where 2015-2019 ASIRs remain xed and are applied