Inuence of Sociodemographic Factors, and the Coverage and Offer of Health Services on Mortality Due to Oral and Oropharyngeal Cancer in Brazil: A 20-year Analysis

Background: To investigate the inuence of sociodemographic factors and variables related to oral health services in oral and oropharyngeal cancer mortality in Brazil. Results: The mortality rate was higher in men than in women; the Southeast and South regions had the highest rates, which increased with age. Regarding APC analysis, men aged 57 years or more and those born from the 1920s to 1955, presented the highest mortality rate, while women born between the 1920s and the 1930s had a higher rate ratio. Kaplan-Meier survival curves and Cox regression showed that black men living in the Midwest region had the lowest survival rate. Considering the correlations, the North and Northeast regions presented mortality rates inversely proportional to FPDC and NSTB, while the Southeast presented it only to FPDC. Conclusions: The sociodemographic variables analyzed exhibited an inuence on mortality and survival rates in relation to oral and oropharyngeal cancer. Regarding the oral health services, it was observed that preventive and diagnostic procedures are not being performed, which may be exacerbating the increase in the mortality rates observed.


Introduction
Oral and oropharyngeal cancers can affect the oral cavity (lip, tongue, and mouth [ICD-10: C00-06]), salivary glands (C07-08), and oropharynx (ICD-10: C09-C10). [1,2] Considering all types and sites of cancer cited, in 2018, an estimated 500 550 new cases were diagnosed worldwide, resulting in 250,565 deaths. [3] Approximately 370 000 new cases and 185 821 deaths have occurred in 2020, indicating that this neoplasm is a global public health problem. [4] In 2012, approximately 77% of deaths caused by oral cancer occurred in developing countries. [5] Among the continents, Latin America has the highest incidence rate, [6] wherein men from Cuba and Brazil had the highest age-adjusted incidence and mortality rates per 100 000 inhabitants. [7] Moreover, the mortality rate in Brazil has increased since the 1980s. [8] This disease has a multifactorial etiology, associated with both intrinsic and extrinsic factors. [9] Age and sex are considered risk factors [10]; therefore, oral cancer is more common in men than in women, and the majority patients are aged 50 years or older. [2,11] As extrinsic risk factors, chemical substances (primarily tobacco and alcohol) and biological agents (exposure to human papillomavirus [HPV] and immunosuppression) have been identi ed. [11] Therefore, it is important to understand whether there have been changes over the years, considering the effectiveness of public policies focused on risk factors.
Aspects such as access to health information, access and use of health services, exposure to carcinogenic factors, and inadequate nutrition interfere not only with early diagnosis, but also in estimating the neoplasm-associated survival, incidence, and mortality rates. [12][13][14] Therefore, as Brazil is a country of continental dimensions, with disparities in sociodemographic conditions and health indicators, [15] this study aimed to investigate the in uence of these sociodemographic factors and variables associated with the offer and coverage of oral health services in oral and oropharyngeal cancer mortality in Brazil.

Ethics
The database used in this research contains consolidated information, without identifying individuals, therefore, according to CNS Resolution No. 510, April 7, 2016, evaluation by the research ethics committee was not required. [16] Study design This study had a retrospective, longitudinal, observational, ecological, descriptive and inferential analytical study design.

Population
This study examined the secondary data of individuals who had died due to oral and oropharyngeal cancer in Brazil between 2000 and 2019. Oral and oropharyngeal cancer was de ned according to the 10th revision of the International Classi cation of Diseases (ICD-10) by codes C00 to C10. [1] Data base Sociodemographic and health indicators were analyzed, considering the offer and coverage of the oral health services network. The units of analysis investigated were the ve geographic regions of Brazil, according to the Brazilian Institute of Geography and Statistics (IBGE). The sociodemographic variables and those related to the provision and coverage of oral health services were obtained from the Department of Informatics of the Uni ed Health System (DATASUS), [17] which consolidates information from the National Registry of Health Establishments, the Outpatient Information System, and the Mortality Information System (SIM). The DATASUS is an open-access database, maintained by the Brazilian Ministry of Health. The population quantity used in the calculation of the rates was based on the projection made by IBGE. [18] To study the variables, data on the number of deaths due to oral cancer were collected using SIM.

Data extraction and variables
The number of deaths used to calculate the speci c mortality rate was obtained according to the place of occurrence, since in comparison with deaths by place of residence, no statistical difference was found (p ≥0.9999, one-way ANOVA with Sidak's post hoc test). The independent variables were geographic regions (North, Northeast, South, Southeast, and Midwest), sex (female and male), age (≤29 years, 30-39 years, 40-49 years, 50-59 years, 60-69 years, 70-79 years, and ≥80 years), and race/ethnicity according to skin color: white, yellow, brown, black, and indigenous. Unidenti ed data were excluded from analysis.
In order to analyze the correlation between oral cancer mortality rates and indicators associated with coverage and offer of oral health services, other independent variables were selected in primary care (PC) and specialized care (SC). The rst variable, oral health teams coverage (OHTC), represents the proportion of the population assisted by oral health teams in PC. [19] The second is related to the number of rst programmatic dental consultations (FPDC), which consists of the evaluation of general health conditions and dental clinical examination for diagnostic purposes.
Regarding SC, the chosen variables included the number of dental specialty centers (DSCs) where specialized dental services are offered, with emphasis on the diagnosis and detection of oral cancer, [20,21] as well as the number of soft tissue biopsies (NSTB) of the mouth performed for diagnostic purposes. All these data were transformed into speci c rates per 100,000 inhabitants for statistical analysis. However, as these health service variables are available only from the period between 2008−2018, the correlation was limited to this time interval.

Statistical analysis
Descriptive results were expressed as measures of central tendency and dispersion. To verify the normality of all variables, data were submitted to the D'Agostino and Pearson test. Then, to compare the mortality rates according to sex, age, and geographic region, one-way ANOVA (parametric) and Kruskal-Wallis (non-parametric) tests were used, with Tukey and Dunn's post-hoc tests, respectively. To analyze the mortality rate during the study period and to verify the correlation with the health services variables selected, Pearson's correlation coe cient (Pearson's r) was performed to analyze the parametric data, while Spearman's correlation (Spearman's r) was selected when the distribution was non-parametric. Pearson's and Spearman's r were classi ed as weak (r <0.33), moderate (r = 0.34−0.66), and strong (r >0.67).
The age-period-cohort (APC) analysis obtained by parameters estimated using the APC Web Tool (Biostatistics Branch, National Cancer Institute, Bethesda, MD, USA) [22] was used. The appropriate model was applied, which explains the identi cation problem to determine the variations in mortality due to the independent effects of age groups, death calendar periods, and birth cohorts. For all variables analyzed in this study, the following functions were estimated: net drift, all age deviations, all period deviations, all cohort deviations, all period rate ratios (RR), all cohort RR, and relation net drift with all local drifts. The Wald test, for evaluation of the asymptotic chi-square (χ2) distribution, was used to verify signi cant differences, with p < 0.05 considered statistically signi cant.
For the survival analysis, calculations were performed considering the life period from birth to the occurrence of the event corresponding to death. In addition, the log-rank, Breslow, and Tarone

Results
In the exploratory and descriptive analysis of the data, we identi ed 98,232 deaths due to oral and oropharyngeal cancer between 2000−2019, and males represented 78% of this total. The number of deaths increased over the years (2000-2019) in Brazil (Pearson r = 0.990; CI 95%−0.975; p<0.0001). The mortality rate was higher in men (4.04 ± 0.585/100 000 inhabitants) than in women (1.05 ± 0.208/100 000 inhabitants) (p<0.0001, independent t-test).
In relation to the female-speci c mortality rate between the regions, it was observed that the Northeast, Southeast, and South regions had the highest mortality rates ( Figure 1A). The Southeast and South regions also had the highest mortality rates among males ( Figure 1B).
Regarding mortality according to age group, the trend was the same in both females and males, with an increase in the mortality rate observed with increasing progressive age ( Figure 2).
The results obtained in the analysis of APC are shown for females ( Figure 3) and males ( Figure 4). Age, period, and cohort deviations were not statistically signi cant ( Figure 3A). All cohort RR, otherwise, was statistically signi cant, indicating that women born in the 1920-1930 cohort detained the largest RR ( Figure 3C). For the male population, all period and cohort deviations were not statistically signi cant ( Figure 4A). All age deviations and cohort RR, however, were statistically signi cant, as men aged 57 years or older had the highest RR ( Figure 4B), indicating that the mortality of these individuals increased with age. When looking at the cohort speci cally, men born between 1920 and 1955 presented an increase in RR ( Figure 4D).
From the survival curve analysis differences between sexes (5A), geographic regions (5B), and race (5C) could be observed. In relation to the sociodemographic variables (Table 1), the analysis of survival according to sex showed that men had a lower survival rate than women. Regarding the regions, the survival rate throughout the aging process is lower for the Southeast, South, and Midwest regions than for the North. The Northeast, however, had higher survival rates compared to the Southeast, South, and Midwest. In addition, the Midwest region showed a lower survival rate compared to the other regions. As for the variable race, White individuals had a higher survival rate than those who were Black and Brown. Individuals corresponding Black race, however, had a lower survival rate than the other races. Unlike Yellow, which has a higher rate in relation to the others, except for Indigenous, which did not present a statistically signi cant difference. Brown individuals also presented a lower rate than indigenous individuals.
Regarding the correlations performed, it was observed that OHTC presented a signi cant correlation in four regions (North, Northeast, South, and Midwest), being inversely proportional only in the south. Referring to the FPDC procedure, the correlation was signi cantly negative in the North, Northeast, and Southeast regions. Concerning the number of DSCs, the correlation was positive and signi cant in all regions investigated, as well as in Brazil, in contrast to the NSTB, which was statistically signi cant and negative only in the North and Northeast regions ( Table 2).

Discussion
Based on the results of this research, it can be observed that sociodemographic characteristics have an in uence on mortality due to oral and oropharyngeal cancer in the Brazilian population, also showing that this coe cient is higher in males than females, increases with age, and occurs more in the Southeast and South regions. In addition, the Black population, residents of the Midwest region, and males have lower survival rates. The in uence of the supply and coverage of health services was also observed, indicating that the prevention and diagnosis of oral and oropharyngeal cancer are not e cient, to the point of not decreasing the mortality rate.
Considering both sexes, the Southeast and South regions had the highest mortality rates, while the North region had the lowest. This difference had already been observed in other studies [8,13,23,24] and the implementation and/or improvement in noti cation systems, associated with a better organization of health services, and the unequal expansion of access to these services have been identi ed as possible causes. [6,8,25] In the survival analysis, the lowest rate was found in the Midwest region. It is necessary to point out that this is the rst research to compare survival rates between the geographic regions of Brazil, therefore, it should be considered that when analyzing and comparing these rates between countries, the differences observed are justi ed by the stage of the disease in diagnosis and the availability and quality of treatment. Furthermore, access to diagnosis and treatment can be di cult in some regions and localities, as well as for some speci c groups in Brazil, [13] which may justify the differences identi ed here.
Another nding is the annual average of deaths in women in the Northeast region, which was similar to the Southeast and South regions. Perea et al. [8] exhibited a trend of increased incidence and mortality from mouth and oropharynx cancer from 2002-2013; a phenomenon which was localized to this region. Although the highest coe cients are associated with males, more current studies show that this ratio between the sexes has been changing, with a downward trend in variation. [6,8,23] Therefore, there is a need to understand what may be associated with this speci c epidemiological pattern in the region.
Regarding age groups, the highest annual average mortality rates belonged to individuals over 50 years, for both sexes, a pattern found in a study conducted in Uruguay [26] and also in Brazil, from 2000-2013. [27] One study conducted by Rocha et al. [6] showed that the proportion of adults over 60 years in Brazilian municipalities was positively associated with the mortality rate from 2002-2012. The same pattern was observed when analyzing the trends in mortality from oral and oropharyngeal cancer in the capital of the Southeast region, where the highest mortality rate was observed among men over 60 years.
[28] When the sexes were analyzed separately, men over 80 years had the highest speci c rate (27.92/100 000 inhabitants), while in women, this rate was 16.08/100 000 inhabitants. These ndings show that the risk of dying from cancer increases with age, [23] and is in uenced by aging in the process of carcinogenesis and tumor growth, mainly due to immunosenescence, which leads to impairment of regenerative capacity and tissue repair. [29,30] The APC analysis identi ed the cohort effect on mortality due to oral and oropharyngeal cancer in both sexes, but an age effect was observed only in males. Perea et al. [31] conducted a similar analysis in the Brazilian population, with death data from 1983-2017; however, they identi ed that the risk of mortality increased from 40 years of age in men, which differs from this study, in which this increase was observed in men aged over 57 years. These same authors identi ed an age effect on mortality in the female population, with an increased risk of mortality from 55 years, [31] which also differs from the ndings reported here. However, as expected for a chronic disease, in both studies, mortality rates increased with age. It is worth remembering that both studies are pioneers in the analysis of APC in oral and oropharyngeal cancer in Brazil, which makes it impossible to compare with other analyses performed here.
Regarding the cohort effect, men born between 1920 and 1955 and women born between 1920 and 1930 presented a higher risk of death. These ndings could be justi ed by higher male exposure to risk factors, tobacco, and alcohol, which occur earlier and last for a long period of time. The female cohort seems to be more associated with the aging process, since the group of women with higher mortality rates were those aged 80 years or older. However, Perea, Antunes, and Peres [31] observed this effect in a cohort from 1958-1962, for both sexes, but we should point out that the cohort analysis was performed separately, according to the anatomical site. These authors provide as justi cation for their ndings the exposure of women to smoking in more recent decades, as well as the exposure to HPV, mainly due to its relationship with oropharyngeal cancer. [31] Regarding the sex differences found in this study, the survival rate in men was signi cantly lower than that in women, as had already been identi ed in other countries [4,32]. Shiboski, when investigating racial disparities in survival rates, also identi ed disparities between the sexes; white women exhibited higher survival than men, both black and white, while black women had lower survival than white men [33]. However, some studies conducted in the Southeast and Northeast regions' capitals did not identify differences in survival between the sexes. [12,34,35] Clinical factors such as site of injury, histological type, staging of the disease, type of treatment, time between diagnosis and initiation of treatment, nutritional status, and HPV exposure, also in uence the survival of individuals with oral cancer, [4,12,14,34,36,37] which may justify the difference found in this study compared to other ndings in the literature. Another possible explanation is that the regular use of health services, speci cally by women, [8] which can enable early diagnosis, encounters barriers such as wrong diagnoses and erroneous and/or late referrals, which lead to late diagnosis, with consequent worsening of prognosis and survival. [38] Regarding race, the survival analysis performed in this study indicated that black individuals had the lowest survival rates, corroborating studies conducted in the United States. [32,33] Differences in access and use of health services have been identi ed as possible causes. [33] A study in Brazil from 2000-2013 identi ed stability in the mortality rate for black men and a decrease for women of the same race. [27] These authors speculate that the expanding access to health services may be associated with this observation. [28] In this regard, public health policies need to be planned and implemented, considering the incentive to use these services by black individuals. Furthermore, in São Paulo from 2003-2009, mortality doubled in blacks, but remained unchanged among whites, [25] and may also be associated with greater tobacco and alcohol exposure of the low socioeconomic status racial and social stratum. [25] We should note that this study grouped browns and blacks, which may have in uenced the results. In addition, another study in Brazil investigated racial disparities in the use of public dental services and found that blacks and browns depend more on these services than do whites. [39] Delay in diagnosis is a major concern with regard to oral cancer and is responsible for at least 50% of the diagnoses made in advanced stages. [38,40] In Brazil, according to the National Oral Health Policy (PNSB), [21] the prevention and early detection of oral cancer can be attributed to oral health teams in PC, while DSC is responsible for diagnosis and referral for hospital treatment. [27] It is worth noting that the early identi cation of mouth lesions, the follow-up of suspected cases, and the guarantee of treatment recommended by the PNSB, depends on an articulated network of services. [21] However, studies indicate that oral health in the SUS still faces challenges with regard to the consolidation of SC, which prevents integration between the levels of care. [41] Therefore, we sought to correlate the mortality rate due to oral and oropharyngeal cancer, primarily with service coverage.
Mortality rates in the North, Northeast, and Midwest regions showed positive and signi cant correlations, both in relation to OHTC and the number of DSCs. In other words, the expansion of services accompanied the increase in the mortality rate. Considering that the network of oral health services still faces di culties in the integration of care [41], and that this interferes with diagnosis, referral, and treatment, it is assumed that the diagnosis of oral cancer is not being performed in the initial stages [6]. The South region was the only one to present a signi cant and inversely proportional correlation for OHTC, indicating the in uence of mortality by the opportunity to access health services for early diagnosis and treatment [27].
In the correlation between the FPDC and mortality rate, the North, Northeast, and Southeast regions presented signi cant and inversely proportional results; the higher the mortality, the lower the number of consultations performed. Considering that the FPDC is an opportune time for visual and tactile inspection of the oral cavity and identi cation of lesions with malignant potential [42], it is necessary to investigate the reasons associated with not performing this procedure in Brazilian regions, as it compromises early diagnosis and reduction in mortality [43].
The same reasoning can be applied when analyzing the NSTB performed, since the result was similar to the FPDC, being inversely proportional to the mortality rate in the North, Northeast, and national panorama regions. The identi cation of lesions on the oral mucosa should be prioritized at all points of health care [21], and the respective diagnoses should be established by biopsy [36,44]. However, the result obtained in the correlation indicates that these procedures are not being performed, resulting in increased mortality, since there is no diagnosis and, therefore, no treatment.
In order to plan, develop and also evaluate public health policies, especially those aimed at oral and oropharyngeal cancer, it is essential to understand the association between sociodemographic factors and mortality due to this neoplasm. However, as in other developing countries, it is necessary to recognize that there are some limitations on the quality of data collection in Brazil. A large amount of unknown or unreported data affects the reliability of the analysis carried out in studies with secondary databases, although this situation has been decreasing over the years. The ecological design of this research must also be considered, since there are no individual analyzes which could be required in a more accurate investigation, as occurs in control or cohort case studies.
Therefore, studies with other methodological approaches should be carried out to clarify the possible issues raised here.

Conclusion
In this study, we found that the mortality rate from oral and oropharyngeal cancer increased during the aging process, being more prevalent in men, who had a lower survival rate, regardless of the geographical region of Brazil. In addition, women and men born between 1920-1930 and 1920-1955, respectively, had a higher risk of death. Individuals of the Black race/skin color had the lowest survival rate. Finally, there was a lower survival rate in the Midwest and higher in northern Brazil. Regarding the oral health services indicators, it was observed that preventive and diagnostic procedures are not being performed, which may be associated with an increase in the mortality rate in Brazil.     Age-Period-Cohort (APC) analysis with Wald test (A), representation of the mortality rate by age (B), period (C) and cohort (D) in male.

Figure 5
Kaplan-Meier survival rates for individuals with oral cancer according to sex (A), geographic regions (B) and race (C).