Associations of Body Mass Index and Waist Circumference with the Risk of Head and Neck Cancer: A National Population-Based Study

Simple Summary A high body mass index is positively associated with multiple cancer types. However, there are controversies regarding such an association with head and neck cancer. This population-based study is the first report of association of body mass index and waist circumference with the risk of head and neck cancer. Low body mass index and low waist circumference were related to a risk of head and neck cancer. These findings highlight the importance of preventing and reversing body mass index to reduce head and neck cancer incidence. Abstract Background: We investigated the association between BMI and HNC subtype incidence in a cohort study of ten million people, adjusting for the effect of smoking and drinking. We also investigated the relationship between waist circumference (WC) and HNC subtype. Methods: All data used in this study originated from the Korean National Health Insurance Service database. We analysed subjects who had undergone health check-ups in 2009 and monitored subjects until 2018 (n = 10,585,852). Finally, 9,598,085 subjects were included after exclusions. We collected variables that could affect the risk of HNC. Cox proportional hazards regression analysis was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Results: The overall incidence of HNC was higher in the low BMI category (BMI < 18.5 according to WHO recommendations for Asian people) (HR: 1.322; 95% CI: 1.195–1.464) compared with the normal BMI category. Among the HNC cases, the incidence rates of laryngeal (HR: 1.3; 95% CI: 1.085–1.558), oral cavity (HR: 1.277; 95% CI: 1.011–1.611), and oropharyngeal (HR: 1.584; 95% CI: 1.25–2.008) cancers were higher in the low BMI category compared with the normal BMI category. No significant association was detected between low BMI and sinus cancer, salivary gland cancer, or nasopharyngeal cancer. The low WC category (<80 cm in men and <75 cm in women) was related to a risk of hypopharyngeal (HR: 1.268; 95% CI: 1.061–1.514) and laryngeal (HR: 1.118; 95% CI: 1.007–1.241) cancers. The HR for occurrence of HNC was high in underweight participants according to smoking status (1.219 for never smoker vs. 1.448 for ever smoker, p for interaction = 0.0015) and drinking status (1.193 for never drinker vs. 1.448 for ever drinker, p for interaction = 0.0044). Conclusions: Low BMI was associated with the risk of some types of HNC. The results of this study could assist etiological investigations and prevention strategies.


Introduction
Body mass index (BMI) is the most common indicator of general obesity, and waist circumference (WC) is an indicator of abdominal obesity. Previous studies have reported that an increase in BMI is positively associated with multiple cancer types, such as colon

Study Population
All data used in this study came from the Korean National Health Insurance Service (NHIS) database. The NHIS is the obligatory health insurance for about 97% of the Korean population except for Medicaid beneficiaries. The NHIS database includes patient demographics and the records of diagnosis, interventions, and prescriptions. The NHIS health check-up programs include anthropometric measurements, hearing and visual acuity checks, laboratory tests, past family, medical, and surgical history, and social history. Hospitals perform the health check-ups after being certified by the NHIS, which also regularly qualifies trained examiners. Past medical history, and health-related behaviours such as smoking, alcohol intake, and physical activity were collected using standardized self-reporting questionnaires. Therefore, the Korean NIHS data represent the entire Korean population without selection bias and have been used in many epidemiological studies.

Patient Selection and Cancer Ascertainment
We selected subjects who were >20 years and who had undergone a health check-up in 2009 (n = 10,585,852). We monitored subjects until 31 December 2018. Participants were censored if they (1) died, (2) immigrated and were no longer followed up by the NHIS, or (3) had new-onset HNC (primary outcome of the study). We excluded individuals with missing data (n = 746,403) and those with a history of another cancer before the health check-up (n = 153,456). We also applied a 1-year lag period to minimize detection bias. Subjects who diagnosed with cancer within one year of the health check-up were excluded to minimize the effect of cancer on BMI or WC, and subjects who had less than a year of follow-up data were also excluded (n = 87,908). Finally, 9,598,085 subjects were included in this study from baseline to the date of HNC diagnosis. Participants were diagnosed with HNC if they had admission records for HNC in their NHIS data from 2010 to 2018. Diagnoses were confirmed by the International Classification of Disease, Tenth Revision, Clinical Modification codes: C02, C03, C04, C05, and C06 for oral cavity cancer; C07 and C08 for salivary gland cancer; C11 for nasopharyngeal cancer; C01, C051, C099, and C103 Cancers 2022, 14, 3880 3 of 12 for oropharyngeal cancer; C12 and C13 for hypopharyngeal cancer; C10 for sinus cancer; and C32 for laryngeal cancer.

Definitions of BMI and WC
BMI was calculated as the weight divided by the square of the height (kg/m 2 ) and categorized using the WHO recommendation for Asian populations, (underweight ≤ 18.5, normal weight = 18.5-23, overweight = 23-25 kg/m 2 , preobesity = 25-30 kg/m 2 , obesity ≥ 30 kg/m 2 ) [12]. Abdominal obesity was defined as WC ≥ 90 cm in men and ≥85 cm in women, according to the definition of the Korean Society for the Study of Obesity. For analysis, WC was separated into levels of 5 cm increments: level 1, <80 cm in men and <75 cm in women; level 2, 80-85 cm in men and 75-80 cm in women; level 3, 85-90 cm in men and 80-85 cm in women; level 4, 90-95 cm in men and 85-90 cm in women; level 5, 95-100 cm in men and 90-95 cm in women; and level 6, ≥100 cm in men and ≥95 cm in women [13].
We categorized smoking status as non-smokers, ex-smokers, or current smokers. Alcohol drinking was categorized as non-drinking, mild (<30 g/day), and heavy (≥30 g/day), and regular exercise was defined as vigorous physical activity for at least 20 min/day. Income levels were divided by quartiles: Q1 (the lowest), Q2, Q3, and Q4 (the highest). Subjects with hypertension were defined as having at least one claim per year for the prescription of anti-hypertension medication under ICD-10 codes I10-13 and I15, or having a systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥90 mmHg without a claim for anti-hypertension medication under ICD-10 codes I10-13 and I15. Diabetes was defined as a fasting blood glucose level ≥ 126 mg/dL (≥7 mmol/L) or the presence of one or more claims per year for antihyperglycemic medications with ICD-10-CM code E10-14. Patients with a glomerular filtration rate < 60 mL/min/1.73 m 2 at the time of baseline evaluation were classified as the chronic kidney disease group.

Statistical Analysis
When death or cancer occurred within one year after the medical examination, subjects were excluded from the analysis to reduce bias. Cox proportional hazards regression analysis (follow-up duration is the primary time variable) was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between BMI and the risk of HNC. Model 1 was not adjusted; Model 2 was adjusted for age, sex, income, smoking, alcohol consumption, exercise, and chronic diseases that could affect BMI or WC, such as hypertension and diabetes. In Model 2, variables that could affect the incidence of HNC or influence of BMI and WC were adjusted for. Tests for multiplicative interactions were performed for BMI category and smoking and drinking. Statistical analyses were performed using SAS ver. 9.4 (SAS Institute, Cary, NC, USA). A p-value < 0.05 was considered significant.

Basic Characteristics
The general characteristics of the participants according to the BMI WHO recommendations for Asian populations are presented in Table 1. In total, 10,585,852 participants were recruited, and 987,767 participants were excluded. Finally, 9,598,085 participants were eligible for this study from 2009 to 2019. The median follow-up duration was 8.31 (8.12-8.57) years. Among them, 10,732 participants were newly diagnosed with HNC: 2972 with laryngeal, 2225 with oral cavity, 1814 with oropharyngeal, 929 with hypopharyngeal, 1101 with nasopharyngeal, 539 with sinus, and 1273 with salivary gland. The mean BMI was 17.57 ± 0.78 kg/m 2 in the underweight group, 21.13 ± 1.21 kg/m 2 in the normal weight group, 23.93 ± 0.57 kg/m 2 in the overweight group, 26.73 ± 1.29 kg/m 2 in the preobesity group, and 32 ± 6.89 kg/m 2 in the obesity group. Participants were more likely to be non-smokers than smokers or ex-smokers. However, the ratio of current smoker was highest in the obesity group compared with other BMI categories. Drinking was also distributed in the order of non-alcohol, mild, and heavy in all BMI categories, similar to smoking. The ratio of heavy drinkers was highest in obesity group compared with other BMI categories. WC and weight were higher with higher BMI categories. The correlation of BMI and WC was 0.705 (0.700 for male and 0.730 for female).

Figure 1.
Hazard ratios (HR) of head and neck cancer (HNC) and its subtypes according to body mass index (BMI). All HRs were adjusted for age, sex, income, smoking, alcohol consumption, regular exercise, diabetes, and hypertension.

Relationship between WC and HNC
The incidence rates of the HNC subtypes according to the WC categories are shown in Table 3. The overall incidence of HNC was higher for level I WC (HR: 1.104; 95% CI: 1.045-1.165) compared with the normal (level III) WC category. Among the HNCs, only the incidence rates of hypopharyngeal cancer (HR: 1.268; 95% CI: 1.061-1.514) and laryngeal cancer (HR: 1.118; 95% CI: 1.007-1.241) were higher in the level I WC category compared with the normal WC category.

Interactions of BMI with Smoking and Drinking on Occurrence of HNC
We examined the effect of smoking or drinking status on the risk of HNC according to BMI category. Table 4 provides HRs and 95% CIs among participants with ever smoker as compared with never smoker stratified by BMI. There was a positive interaction between BMI and smoking on occurrence of HNC (HR: 1.129; 95% CI: 1.024-1.451 for never smoker and underweight, HR: 1.335; 95% CI: 1.178-1.512 for ever smoker and underweight, p for interaction = 0.0015). The HR for occurrence of HNC was high in underweight participants according to never or ever drinker (1.193 vs. 1.448, p for interaction = 0.0044) ( Table 5).

Discussion
We observed an association between BMI and the incidence of HNC in this large cohort study. However, this association differed by cancer subtype. The incidence of laryngeal, oral cavity, and oropharyngeal cancer was higher in the low BMI category compared with the normal BMI category. Furthermore, the incidence of hypopharyngeal cancer was negatively correlated with higher BMI. These results were adjusted for all of the covariates listed in Table 1.
The results of our study are consistent with previous studies, as low BMI was associated with a higher risk of HNC. However, there were some differences in the HNC subtype risk analysis. The authors of the prospective NIH AARP cohort study showed an inverse association between HNC and BMI that was almost exclusively among current smokers (HR: 0.76) [14]. In a Dutch cohort study, oral cavity and oropharyngeal cancers had the strongest inverse association with BMI [6]. Our results confirm previous findings that oral, oropharyngeal, and laryngeal cancers are associated with a low BMI; however, the inverse association between BMI and oral cavity cancer failed to reach statistical significance (p = 0.0637). Our results suggest that the relationship between the risk of sinus, salivary gland, and nasopharyngeal cancers differed from that of other HNCs.
Drinking and smoking are known risk factors for HNC; however, the effects of drinking and smoking on BMI or weight differ among studies. In their prospective study, Gaudet et al. reported that moderate to heavy drinkers tend to have a lower BMI and a higher risk for HNC [7]. However, Traversy et al. suggested that heavy drinking is more consistently related to weight gain [15]. In this study heavy drinking was positively related to a high BMI (Table 1). A previous study proposed that smoking affects low BMI. Piirtola et al.
suggested that current smoking is associated with a lower BMI and that stopping smoking is associated with a higher BMI in their monozygotic twin study [16]. In the present study, the relationship between current smoking and BMI was inconsistent, as in previous studies, as the ratio of current smokers was higher with higher BMI. Thus, smoking status and alcohol consumption should be adjusted to analyse the relationship between the incidence of HNC and BMI. However, there was limited power in most previous studies to stratify smoking status or alcohol consumption, or both, due to the numbers of participants and the study durations.
There has been no exact explanation about the relationship between BMI and HNC risk [10]. Most studies have explained that smoking and drinking can be associated with being underweight or dysphagia, or that odynophagia could influence being underweight [5]. Some recent studies suggested that neuropeptides such as leptin were influenced by smoking and drinking and affect body weight by regulating satiety and food intake [17][18][19][20][21]. Gallina reported that leptin was associated with recurrence of malignancy in laryngeal cancer [21]. Here, we adjusted the results for the effects of smoking and drinking. Participants diagnosed within one year after determining their BMI were excluded to rule out the effect of cancer on BMI. Furthermore, most laryngeal cancer cases did not develop dysphagia. A possible explanation is that patients with a low BMI have low levels of vitamins or micronutrients such as vitamins A, C, and D, or folate [22][23][24][25][26]. These vitamins and micronutrients could act as antioxidants, which are important in the prevention of further injury or carcinogenesis during eating or phonation in the upper aerodigestive tract [27][28][29][30][31][32][33][34]. Nasopharyngeal cancer might have a different pattern from the other pharyngeal cancers in this study because it rarely occurs as a result of exposure to foreign substances [35].
Some studies have associated WC with a higher risk of HNC, but we observed a relative inverse but non-significant relationship between HNC and WC. Abdominal obesity, rather than general obesity, is considered to be one of the key features of metabolic syndrome [36]. Intra-abdominal fat has been hypothesised to be biologically different from fat in other areas with regard to tumour angiogenesis and cell proliferation [37]. These results were supported by a prospective cohort study of 3.5 million adults in Spain. In that study, the HR of laryngeal cancer differed according to BMI and WC, unlike other cancer types [38]. This difference may be due to the inconsistency between BMI and WC. In the present study, most of level 3 WC was comprised of preobesity or those that were overweight according to BMI category, and a normal BMI occupied 72.77% of level 1 WC. Therefore, WC was not a good reflection of a low BMI.
The strengths of our study are that the results were analysed in cohort data including more than 9.5 million participants over 10 years. Therefore, we investigated the association between BMI and HNC subtype with sufficient power after adjusting for smoking and drinking to minimize confounding bias. The NIHS database provided the health information for 97% of the Korean population. Therefore, although data represented only Koreans, selection bias was limited. Furthermore, the hospitals that performed the health check-ups were certified by the NHIS, which also qualifies trained examiners. These factors minimized the information bias. Another strength was excluding participants who were diagnosed with HNC within 1 year of the study to rule out the effects on diet or BMI of the patient. Finally, this study compared the effect of BMI and WC on the risk of the HNC subtype.
This study also had limitations. First, most of the participants were Korean, so it was difficult to assess the representativeness of races or ethnicity. Second, the effects of HPV on oropharyngeal cancer were not included. The causes and clinical manifestations of HPV-related oropharyngeal cancer are different from those of HPV-negative oropharyngeal cancer [39][40][41][42]. However, because the diagnosis in the Korea NHIS is based on the International Classification of Diseases, we could not distinguish between oropharynx cancer related to HPV. Third, the environmental and occupational exposure of the individuals could not be considered in this study. Although there are some case-control studies about the association between nutrients and HNC, this study did not contain detailed individual nutritional status or diet. Nutritional intake is one of the leading contributing factors to obe-sity. Future studies should include a more accurate and comprehensive dietary assessment to determine the role of diet and body composition in HNC development [25,43,44]. Finally, this study was designed as a 10 year retrospective cohort study. The lifestyle (including smoking or drinking status) of participants could be changed during the observation period; however, it is difficult to reflect these changes. Also, the causal relationship between changes in BMI, WC, and HNC risk is beyond the scope of this cohort study. Further, well-designed case-control studies are needed.

Conclusions
In this large cohort study, low BMI was associated with a higher risk of some types of HNC, confirming the results of previous studies. This study provides the novel finding that salivary gland, sinus, and nasopharyngeal cancers have different patterns from those of other HNCs in their relationships with BMI. Unlike BMI, WC was only related to the incidence of hypopharyngeal and laryngeal cancers.  Informed Consent Statement: Patient consent was waived for the retrospective data evaluation. Data Availability Statement: Data available on request due to data sharing restrictions. The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
All authors have no conflicts of interest to disclose.