Sage Journals: Discover world-class research

Abstract

Background. A barrier to early-stage oral cavity cancer detection is the lack of a defined population and screening regimen satisfying risk–benefit considerations. Methods. We constructed a microsimulation model, Simulation of Cancers of the Oral cavity and Risk Exposures (SCORE), that incorporates risk profiles defined by smoking and alcohol exposure. SCORE simulates the development and progression of oral potentially malignant disorders (OPMD) representing benign, dysplastic, or malignant lesions in the US population starting at age 40 y. OPMD high-risk characteristics of malignant transformation informed a biopsy decision rule. SCORE was calibrated to national cancer registry data. We compared life expectancy in those aged 40 to 60 y with OPMDs, cancer incidence, and cancer-specific deaths across screening strategies with and without the biopsy decision rule, assuming screening every 3 y starting at age 50 y. Results. In US men, all screening strategies reduced cancer incidence and cancer-specific mortality by at least 26% and 20% compared with no screening. Whether with or without a biopsy decision rule, life expectancy among those aged 40 to 60 y with OPMDs was 36.37 ± 0.01 life-years, a gain of 0.03 life-years. However, the use of the biopsy rule improved diagnostic efficiency with 8 biopsies per treatable diagnosis. Screening with or without the biopsy decision rule in high-risk men demonstrated comparable benefit, reducing cancer-specific deaths by 27% and incidence by 20% compared with no screening. Meanwhile, in the non-high-risk subpopulation, applying the biopsy rule avoided the harms of excess procedures, reducing lifetime biopsies by 38% versus biopsy of all OPMDs while preserving reductions in cancer burden. Conclusions. SCORE enables virtual trials of various screening regimens and target populations. Given the time and cost of clinical trials, SCORE may facilitate the evaluation of new technologies and clinical recommendations.

Highlights

A new oral cancer simulation model with risk factors including degrees of smoking and alcohol exposure, oral lesion features, and sex incorporates more accurate and precise representation of patient risk categories.

We evaluated screening strategies for oral potentially malignant disorders with or without risk-stratified biopsy referral in both the general population and subpopulations defined by degrees of smoking and alcohol exposure.

Men with a high degree of both smoking and alcohol exposure exhibited a significant reduction in cancer-specific deaths and cancer incidence from screening programs for oral potentially malignant disorders.

Screening with risk-stratified biopsy, using a surgical treatment threshold of moderate dysplasia or worse, yielded the greatest efficiency in term of biopsies needed to detect 1 treatable case.

Keywords

oral potentially malignant disorders oral cavity cancer oral dysplasia OPMD microsimulation decision analysis

The incidence of and mortality from oral cavity and pharyngeal cancers has not improved over recent decades, and the prognosis remains poor despite advances in cancer treatment.^1,2 Oral cavity cancers constitute two-thirds of all oral cavity and pharyngeal cancers, and more than 90% of these are oral squamous cell carcinoma (OSCC).^1,3–5 The US Cancer Statistics Public Use Database reported a steady rise in the burden of OSCC cases, with 10,616 in 2001 to 13,856 in 2017 among all ages.⁶ The age-adjusted incidence of OSCC also steadily increased from 2005 to 2019, with an annual percentage change of 1.07%, based on the data from the National Program of Cancer Registries and Surveillance, Epidemiology, and End Results (SEER) Program.⁷ For OSCC, survival is highly dependent on the cancer stage at diagnosis; 5-y survival for cancer stage I and II combined is 74% compared with 38% in cancer stage III and IV combined.^8–11 Thus, the detection of early-stage OSCC or dysplasia, which are at risk of malignant transformation, represent a major opportunity for better patient mortality and morbidity. Visual–tactile examination entails clinicians taking a look in the oral cavity using their hands to manually reposition and aims to discover oral potentially malignant disorders (OPMDs), specifically leukoplakia and erythroplakia, which may represent both premalignant or malignant lesions but also include some benign lesions with overlapping appearance. Even when performed by trained experts,¹² inaccurate clinical assessment can lead to false positives and excess referrals for biopsy, the gold standard for diagnosis, or false negatives and nonreferral. These limitations represent major barriers to the design and implementation of potential screening programs.

The American Dental Association has recommended visual–tactile oral examination for all adults as part of initial, routine, and emergency dental visits.¹³ Despite the recommendation, visual–tactile examination is inconsistently performed,^14–16 and only 29% of US adults have reported ever receiving an oral cancer examination.¹⁶ Fear of excess unnecessary referrals and patient burden contribute to such reluctance on the part of dental providers, given the rarity of oral cancer in the general population.^14,15 Screening programs in some countries such as India¹⁷ and Taiwan¹⁸ are based on the visual–tactile examination of high-risk individuals (i.e., those with risk factors such as tobacco, areca nut, or heavy alcohol use),¹⁹ but in the United States, questions remain about the specific population that would likely experience favorable benefits compared with harms and the potential screening schedule associated with optimal clinical utility. Since trials for less common cancers are limited by cost and duration, disease simulation modeling is applied to understand whether test and population characteristics exist that support screening trials. Such models estimate the comparative effectiveness of test-and-treat decisions in terms of reduction of late-stage disease burden, life expectancy (LE), and also overall diagnostic efficiency.^20,21

Risk stratification based on risk exposures and OPMD features could be a key determinant of precision oral cavity screening. Men have a 28% increased odds of oral lesions²² and an 87% higher risk of developing OSCC compared with women,^22,23 and heavy alcohol consumption and high smoking intensity are consistently associated with increased odds of developing OSCC.^24–26 The combination of alcohol consumption and tobacco smoking also produces a synergistic effect.²⁴ In addition, patients with OPMDs with clinical features such as nonhomogeneous texture, mixed red/white color, size exceeding 200 mm², and high-risk anatomic site carry an increased risk of malignant transformation (MT).^27–29

Our purpose was to develop a new disease simulation model for the development and progression of OPMDs and oral cavity cancers in US men and women at 40 y of age. The model incorporates specific risks for evaluating more precise screening and management pathways. We evaluated the effectiveness of potential screening, beginning at age 50 y, by visual–tactile examination conducted during a routine dental visit, in both the overall population (all risk groups) and subpopulations ranging from very low to high risk according to degrees of smoking and alcohol exposure, using the medical sector perspective.³⁰

Methods

Simulation of Cancers of the Oral cavity and Risk Exposures (SCORE) is a discrete time, state transition microsimulation model that expanded upon a predecessor model³¹ with an inclusion of behavioral risk factors and OPMD features. The model simulated a closed hypothetical cohort of 40-y-old US men and women using a lifetime horizon. Modeling from age 40 y allows for the accumulation and progression of OPMDs prior to screening initiation at age 50 y, capturing the influence of behavioral risk exposures on lesion development and progression that may affect screening outcomes. The individual-level characteristics of the cohort include smoking and alcohol exposure, presence/absence of OPMD, and various OPMD features. This article provides a description of SCORE, including model structure, model inputs, and associated assumptions, and is provided in accordance with the model reporting guidelines (ISPOR-SMDM).^30,32–35

Natural History Model Overview

SCORE models the natural history of cancer development and progression in patients presenting both with and without OPMDs using a monthly cycle. The model also includes benign disorders with visual characteristics that overlap those with dysplastic OPMDs and cancers. The model simulated patients with the same age and sex and assigned individual risk behaviors and OPMD features that inform differential risks of disease development and progression. For patients presenting with OPMDs, SCORE assigned OPMD features including phenotypes, appearance, anatomic sites, and initial disease severity by randomly sampling from the categorical distributions informed by oral cancer screening studies (Table 1). The spectrum of disease severity (disease states) included benign disorders such as reactive/frictional keratoses; dysplastic OPMDs with grades ranging from mild, moderate, and severe dysplasia or carcinoma in situ; and OSCC with stages ranging from stage I to IV. Disease transitions follow a Markov model with nonhomogeneous individual transition matrix based on individual-level characteristics and OPMD features, resulting in probabilistic model outcomes. SCORE was implemented in Python 3.9.

Table 1

Base-Case Parameters for the US Population

Variable	Value (Range)	Distribution	Source
US population characteristics
Proportion of oral lesion in US adults at the age of 40 y	0.05 (0.045−0.17)		Villa and Gohel,³⁶ Laronde et al.³⁷
Prevalence of OPMD in US adults at the age of 40 y (%)	0.31 (0.11–0.90)		BCE calibratedVilla and Gohel,³⁶ Mello et al.³⁸
Risk profile distribution by sex			Centers for Disease Control and Prevention (CDC) and National Center for Health Statistics (NCHS)³⁹
Percentage of smoking and alcohol exposure in male
Very low-risk subpopulation: Never-smoker and 0–6 drinks/wk	63%
Low-risk subpopulation: Smoker and 0–6 drinks/wk	26%
Intermediate subpopulation: Smoker and 7–20 drinks/wk	8%
High-risk subpopulation: Smoker and ≥21 drinks/wk	3%
Percentage of smoking and alcohol exposure in female
Very-low-risk subpopulation: Never-smoker and 0–6 drinks/wk	75%
Low-risk subpopulation: Smoker and 0–6 drinks/wk	21%
Intermediate subpopulation: Smoker and 7–20 drinks/wk	2%
High-risk subpopulation: Smoker and ≥21 drinks/wk	2%
Oral lesion characteristic: phenotype, subtype, and anatomic sites
Proportion of oral lesion by phenotype
Leukoplakia (OL)	0.900		Expert opinion
Erythroplakia (OE)	0.061		Assume^a
Lichen planus (OLP)	0.039		Assume^a
Proportion of OL by subtype			Schepman et al.,⁴⁰ Kierce et al.⁴¹
Homogenous leukoplakia (HOL)	0.67 (0.60–0.74)
Nonhomogenous leukoplakia (NHOL)	0.33 (0.26–0.34)
Proportion of HOL by anatomic site			Ho et al,.⁴² Holmstrup et al.,⁴³ Scheifele et al.⁴⁴
Tongue	0.07
Floor of mouth	0.25
Others	0.68
Proportion of NHOL by anatomic site			Ho et al.,⁴² Holmstrup et al.,⁴³ Scheifele et al.⁴⁴
Tongue	0.15
Floor of mouth	0.2
Others	0.65
Proportion of OE by anatomic site			Shafer and Waldron,⁴⁵ Lumerman et al.⁴⁶
Tongue	0.0
Floor of mouth	0.0
Others	1.0
Proportion of OLP by anatomic site			Laniosz et al.,⁴⁷ Bardellini et al.⁴⁸
Tongue	0.31
Floor of mouth	0.01
Others	0.68
Initial state of precancers and cancers by lesion type
Initial proportion of benign and dysplasia grades in HOL			Schepman et al.,⁴⁰ Holmstrup et al.,⁴³ Chaturvedi et al.,⁴⁹ Shearston et al.⁵⁰
Benign	0.87
Mild dysplasia	0.08
Moderate dysplasia	0.03
Severe dysplasia	0.02
Initial proportion of benign and dysplasia grades in NHOL			Schepman et al.,⁴⁰ Holmstrup et al.⁴³
Benign	0.45
Mild dysplasia	0.25
Moderate dysplasia	0.09
Severe dysplasia	0.21
Initial proportion of benign and dysplasia grades in OE			Lumerman et al.⁴⁶
Benign	0
Mild dysplasia	0.32
Moderate dysplasia	0.24
Severe dysplasia	0.44
Initial proportion of benign and dysplasia grades in OLP	Same as HOL		Assume
Initial proportion of malignancy (any lesion types)	0.05 (0.02–0.075)
Initial proportion of cancer stages in malignant lesions (any lesion types)			National Cancer Institute⁵¹
Stage I	0.42
Stage II	0.16
Stage III	0.13
Stage IV	0.29
Natural history
OPMD development
Annual probability of developing OPMD^c	0.0005 (0.0003–0.0019)	Normal ( $\bar{x}$ = 0.00055, s = 0.0006)	Calibrated^b
Age multipliers applied to probability of developing OPMD
Age 40–44 y	0.46 (0.34–1.58)	Weibull (α = 3.9, β = 1.09)	Calibrated^b
Age 45–49 y	0.78 (0.43–0.91)	Weibull (α = 8.6, β = 0.75)	Calibrated^b
Age 50–54 y	1.21 (0.32–1.71)	Weibull (α = 3.8, β = 1.15)	Calibrated^b
Age 55–59 y	0.77 (0.38–1.83)	Weibull (α = 3.8, β = 1.15)	Calibrated^b
Age 60–64 y	1.48 (0.4–2.05)	Weibull (α = 4.9, β = 1.44)	Calibrated^b
Age 65–69 y	2.16 (0.78–2.49)	Weibull (α = 6.6, β = 1.87)	Calibrated^b
Age 70–74 y	1.8 (0.88–2.37)	Weibull (α = 6.6, β = 1.87)	Calibrated^b
Age 75–79 y	1.68 (0.76–2.49)	Weibull (α = 6.6, β = 1.87)	Calibrated^b
Age 80–84 y	3.48 (1.9–4.69)	Weibull (α = 8.1, β = 3.74)	Calibrated^b
Age 85+ y	4.21 (1.64–4.6)	Weibull (α = 8.1, β = 3.74)	Calibrated^b
Risk ratio of OPMD development by smoking status (Ref = never-smoker)
Smoker	5.00 (0.84–5.55)		Shulman et al.,²² Villa and Gohel³⁶
Proportion of benign and mild dysplasia grade among newly developed OPMD of type HOL/OLP	Ratio of benign to mild dysplasia is similar to that from initial proportion of benign and mild dysplasia grades in HOL		Assume
Benign	0.92
Mild dysplasia	0.08
Proportion of benign and mild dysplasia grades among newly developed OPMD of type NHOL	Ratio of benign to mild dysplasia is similar to that initial proportion of benign and mild dysplasia grades in NHOL		Assume
Benign	0.64
Mild dysplasia	0.36
Proportion of benign and mild dysplasia grades among newly developed OPMD of type OE	Ratio of benign to mild dysplasia is similar to that initial proportion of benign and mild dysplasia grades in OE		Assume
Benign	0.00
Mild dysplasia	1.00
Progression among dysplasia grades
Annual probability of regressing from mild dysplasia to benign	0.0042 (0.0018–0.007)		Silverman et al.⁵²
Annual probability of progressing from mild to moderate dysplasia^c	0.0006 (0.0001–0.0021)	Normal ( $\bar{x}$ = 0.0005, s = 0.0006)	Calibrated^b
Annual probability of progressing from moderate to severe dysplasia^c	0.001 (0.0002–0.0037)	Normal ( $\bar{x}$ = 0.001, s = 0.0011)	Calibrated^b
Age multipliers applied to probability of progression from less severe dysplasia grade to more severe dysplasia grade
Age 40–44 y	0.41 (0.12–0.55)	Weibull (α = 3.4, β = 0.36)	Calibrated^b
Age 45–49 y	0.52 (0.18–0.81)	Weibull (α = 3.6, β = 0.53)	Calibrated^b
Age 50–54 y	0.94 (0.31–1.76)	Weibull (α = 2.5, β = 0.94)	Calibrated^b
Age 55–59 y	1.13 (0.54–1.99)	Weibull (α = 3.8, β = 1.33)	Calibrated^b
Age 60–64 y	1.53 (0.99–2.13)	Weibull (α = 6.4, β = 1.74)	Calibrated^b
Age 65–69 y	1.9 (1.18–2.31)	Weibull (α = 7.1, β = 1.91)	Calibrated^b
Age 70–74 y	1.88 (0.93–2.51)	Weibull (α = 8.1, β = 2.15)	Calibrated^b
Age 75–79 y	2.51 (1.32–2.87)	Weibull (α = 8.4, β = 2.39)	Calibrated^b
Age 80–84 y	2.59 (1.79–3.28)	Weibull (α = 10, β = 2.79)	Calibrated^b
Age 85+ y	3.75 (2.13–4.7)	Weibull (α = 7, β = 3.81)	Calibrated^b
Hazard ratio of dysplasia progression of dysplastic leukoplakia (Ref = never-smoker)
Smoker	0.86 (0.22–0.86)		Rock et al.⁵³
Risk ratio of dysplasia progression of dysplastic OPMDs excluding leukoplakia by risk profile(Ref = very-low-risk subpopulation^d)			BCE calibrated^b
Low-risk subpopulation^d	1.75 (0.72–2.25)	Weibull (α = 10, β = 1.8)	Jaber et al.,⁵⁴ Morse et al.⁵⁵
Intermediate subpopulation^d	1.80 (0.66–2.31)	Weibull (α = 10, β = 1.9)	Jaber et al.,⁵⁴ Morse et al.⁵⁵
High-risk subpopulation^d	2.85 (2.07–4.15)	Weibull (α = 10, β = 3)	Jaber et al.,⁵⁴ Morse et al.⁵⁵
Progression from dysplasia to malignancy
Annual probability of progressing from mild dysplasia to malignancy^c	0.0005 (0.0001–0.0007)	Normal ( $\bar{x}$ = 0.0003, s = 0.0001)	Calibrated^b
Annual probability of progressing from moderate dysplasia to malignancy^c	0.0004 (0–0.0008)	Normal ( $\bar{x}$ = 0.0004, s = 0.0002)	Calibrated^b
Annual probability of progressing from severe dysplasia to malignancy^c	0.0038 (0.0031–0.0044)	Normal ( $\bar{x}$ = 0.0036, s = 0.0003)	Calibrated^b
Age multipliers applied to probability of progression from dysplasia to malignancy	Same as age multipliers for progression among dysplasia grades		Calibrated^b
Risk ratio of oral cavity occurrence by risk profile (Ref = very-low-risk subpopulation^d)			BCE calibrated^b
Low-risk subpopulation^d	1.35 (0.7–4.5)	Weibull (α = 10, β = 2.9)	Schwartz et al.^24,56
Intermediate subpopulation^d	2.20 (1.9–5.7)	Weibull (α = 10, β = 3.2)	Schwartz et al.^24,56
High-risk subpopulation^d	5.08 (1.9–5.7)	Weibull (α = 10, β = 5.3)	Schwartz et al.^24,56
Risk ratio of malignancy by lesion site (Ref = other anatomic sites)
Tongue or floor of mouth	2.70 (1.31–5.5)		Evren et al.⁵⁷
Risk ratio of malignancy by lesion phenotype (Ref = OL/HOL)
OE	1.73 (1.34–3.56)		Lumerman et al.,⁴⁶ Silverman et al.,⁵² Iocca et al.,⁵⁸ Banoczy⁵⁹
OLP	0.18 (0.15–2.44)		Iocca et al.,⁵⁸ Warnakulasuriya et al.⁶⁰
NHOL	2.28 (1.27–4.28)		Rock et al.⁵³
Progression among cancer stages
Annual probability of progressing from cancer stage I to stage II^c	0.12 (0.08–0.17)	Beta (α =24, β = 181)	Calibrated^b
Annual probability of progressing from cancer stage II to stage III^c	0.28 (0.23–0.31)	Beta (α = 175, β = 485)	Calibrated^b
Annual probability of progressing from cancer stage III to metastasis^c	0.31 (0.28–0.36)	Beta (α = 232, β = 500)	Calibrated^b
Age multipliers applied to probability of progression from less severe cancer stage to more severe cancer stage	Same as age multipliers for progression among dysplasia grades		Calibrated^b
Progression to death
Annual mortality rate of cancer-specific death	0.172 (0.127–0.344)		Mucke et al.⁶¹
Probability of surgical mortality for dysplastic OPMD	0.000001		Wiemer et al.⁶²
Probability of surgical mortality for oral cavity cancer	0.0001		Bhattacharyya and Fried⁶³
Risk ratio of background mortality (Ref = very-low-risk subpopulation^d)			Qin et al.,⁶⁴ Xi et al.⁶⁵
Low-risk subpopulation^d	2.25
Intermediate subpopulation^d	2.29
High-risk subpopulation^d	2.38
Screening characteristics
Sensitivity of visual exam for differentiating benign lesions from carcinoma	0.71 (0.60–0.85)		Downer et al.^66–68
Specificity of visual exam for differentiating benign lesions from carcinoma	0.97 (0.93–0.98)		Downer et al.⁶⁶
Sensitivity of visual exam for differentiating benign lesions from mild or moderate dysplasia	0.50 (0.25–0.75)		Epstein et al.⁶⁹
Specificity of visual exam for differentiating benign lesions from mild or moderate dysplasia	0.97 (0.93–0.98)		Downer et al.⁶⁶
Sensitivity of visual exam for differentiating homogeneous white OPMDs from other types of OPMDs	1.0 (0.50–1.0)		Assume
Specificity of visual exam for differentiating homogeneous white OPMDs from other types of OPMDs	1.0 (0.75–1.0)		Assume
Sensitivity of scalpel biopsy for differentiating benign lesions from dysplasia or carcinoma	0.98 (0.90–1.0)		Giunta et al.⁷⁰
Specificity of scalpel biopsy for differentiating benign lesions from dysplasia or carcinoma	1.0 (0.90–1.0)		Giunta et al.⁷⁰

BCE, base-case estimate; OPMD, oral potentially malignant disorder.

The initial proportion of erythroplakia and lichen planus in total was assumed to be 0.1. Using the ratio of prevalence of erythroplakia to lichen planus 0.17:0.11,^38,71 the initial proportions of erythroplakia and lichen planus were 0.06 and 0.04, respectively.

The value-calibrated parameters were derived from the best-fitted parameter set with the smallest sum of chi-square. Ranges and distributions of the calibrated parameters were derived from multiple best-fitted parameter sets that fulfilled criteria of outcomes with 95% confidence intervals (CIs) with at least 50% overlap of the 95% CIs of the targets, respectively. Each specified distribution was truncated at defined upper and lower bounds (parameter ranges) to ensure positivity.

Transition probability used in the model depends on various individual-level patient characteristic and OPMD features and is calculated by the following formula: $p' = p \times age multiplier \times Π_{i = 1}^{n} R R_{i}$ where $p$ is annual probability and $R R_{i}$ is risk ratio of factor $i$ , which is a risk factor listed in the same subcategories of transition. For example, the probability of progressing from severe dysplasia to malignancy depends on age, patient risk profile, lesion site, and lesion phenotypes.

The subpopulations were categorized into 4 categories based on degrees of smoking and alcohol use. The very-low-risk subpopulation represented nonsmokers with low degrees of alcohol use. Low-risk, intermediate-risk, and high-risk subpopulations represented smokers with low, intermediate, and high degrees of alcohol use, respectively.

Dysplastic OPMDs progressed by increasing in grades of dysplasia or to OSCC stages based on the conditional transition probability of the current disease state (Figure 1). Mild dysplasia was susceptible to either regressing to a normal state or to progression in dysplasia grade.⁵² During each monthly cycle, all patients were subject to background mortality, where we applied the US life table and risk ratios of background mortality to account for higher mortality rates in those with higher degrees of exposure to smoking and alcohol.^64,65 Only patients with metastasis were additionally subject to oral cancer death. We assumed the transition probabilities in adults aged ≥40 y increased with age following 1) increasing observational oral cavity cancer incidence by age from SEER⁵¹ and 2) the association between older ages and MT rates based on Cox proportional hazard analysis using data on patients with OPMDs from a large retrospective study.⁶⁰

Figure 1

State transition diagram for the natural history model of benign disorders, dysplastic oral potentially malignant disorders (OPMDs), and oral squamous cell carcinoma (OSCC).

Patient Risk Factors and OPMD Features

As smoking and alcohol exposure are well-established risk factors associated with the development of OPMD³⁶ and MT,^24,56 we constructed 4 patient risk profiles capturing the two factors. The joint prevalence of smoking and alcohol use across degrees of exposure by sex were informed by our assessment of the National Health and Nutrition Examination Survey (NHANES) longitudinal series 2009–2018³⁹; for example, 63% of men were nonsmokers, with the lowest degree of alcohol exposure (very low-risk subpopulation), while the remaining 27% were smokers from low to high degrees of alcohol exposure (subpopulations of low, intermediate, and high risk) (Table 1). Smokers in the three risk profiles were defined as long-time smokers (lifetime smoking duration ≥21 y) with both low smoking intensity (1–20 cigarettes per day) and high smoking intensity (≥21 cigarettes per day). Rates of OPMD formation and progression in dysplasia or cancer differed depending on risk exposures. Compared with nonsmokers, smokers were more likely to develop dysplastic OPMDs and those with higher degrees of alcohol exposure were associated with a higher risk of OSCC occurrence.^24,54–56 However, leukoplakia, the most prevalent type of OPMD, exhibited higher rates of progression from dysplasia to OSCC when developing in the absence of smoking and alcohol use, as previously reported.⁵³

We modeled the development of 3 common phenotypes of OPMDs that are defined by the main features of color and texture: leukoplakia (further delineated as homogeneous and nonhomogeneous leukoplakia), lichen planus, and erythroplakia.^45,72–74 Evidence has shown that nonwhite color or nonhomogeneous texture are among high-risk features for MT.⁷⁵ For example, erythroplakia (i.e., red-colored OPMDs) exhibits the highest rate of MT among the 3 common phenotypes.⁵⁸ A meta-analysis reported the overall rates of MT of 33.1%, 9.5%, and 1.4% in erythroplakia, leukoplakia, and lichen planus, respectively. Within the 2 subtypes of leukoplakia, the nonhomogeneous subtype is associated with a 2.3 times higher rate of MT.⁵³ In addition to color and texture, certain anatomic sites also portend greater risk; OPMDs on the tongue or floor of mouth exhibit 2.7 times higher risk of MT than those located at other sites do.⁵⁷ Accordingly, we incorporated OPMD phenotypes associated with different rates of MT into the model, with a representative spectrum of histology and anatomic location (Table 1).

Model Calibration

To ensure the model reflects trends in oral cavity cancer, we calibrated SCORE using SEER data from 2004 to 2010. We used more recent SEER data (2011–2017) for subsequent validation, described in the next section. To represent current population risk behaviors, we incorporated smoking and alcohol use data from NHANES 2009–2018.

For the overall population comprising 4 subpopulations of very-low-risk to high-risk individuals (Table 1), we calibrated SCORE to several targets: 1) age-specific oral cavity cancer incidence, 2) OSCC stage distribution at the age of diagnosis of 50 y, and 3) cancer-specific mortality risk derived from SEER 17 data based on oral cancer cavity cases diagnosed in 2004 to 2010.⁵¹ Thirty-six clinical parameters relevant to natural history were simultaneously calibrated, including the prevalence of OPMD for adults at age 40 y, the probability of OPMD development, the probability of progression in dysplasia grades, and probabilities of progression of dysplasia to malignancy and to successive cancer stages (Table 1). Differential rates of OSCC development by patient risk profile were generated using a risk ratio specific to each profile compared with never-smokers. To search for the best-fitted parameters, the sum of chi-square was used as the goodness-of-fit measure.^76,77 To generate the model outcomes, we simulated the overall population of 2 million with 500 repetition iterations accounting for first-order uncertainty (Supplementary Table S1). SCORE was calibrated for US men and women separately using a heuristic search consisting of 3 steps: 1) eliminating infeasible parameter spaces, 2) reducing parameter spaces using Bayesian optimization with an objective function of minimizing the sum of the chi-square between the model outcome and targets,^78,79 and 3) iteratively performing Latin-hypercube sampling on the reduced parameter spaces. Correlation coefficients were analyzed to assess the independence assumption of parameters. We define the best-fitted parameter sets as those that fulfilled 2 criteria: 1) the mean of model outcomes must fall within the 95% confidence interval (CI) of the calibration target, and 2) the 95% CI of the model outcomes must overlap by at least 50% with the 95% CIs of the targets. We then fit distributions across all the best-fitted parameters. The parameter set with the smallest sum of chi-square is defined as the base-case parameter set. We derive the ranges of calibrated parameters from the best-fitted parameter sets.

Model Validation

We validated SCORE against age-specific oral cavity cancer incidence, stage distribution, and cancer-specific mortality derived from SEER data from 2011 to 2017.⁵¹ The predefined validation criterion is overlapping 95% CIs between the model outcomes and targets for incidence and mortality and standard error (SE) ≤0.5% for stage distribution, respectively. We additionally validated our model for the overall population of men by comparing SCORE’s LE in 50-y-old men to the SEER Oral Cancer Survival Calculator (OCSC).^80,81 This calculator provides 3 submodels to estimate LE without cancer depending on patients’ age and health conditions and uses a competing risk model to estimate the probabilities of death from oral cancer and death from other causes using National Health Interview Survey.⁸² To validate, we assessed whether model LE fell within the ranges generated using the SEER OCSC.

Application: Evaluation of Representative Screening Strategies

To provide an example of the application with SCORE, we evaluated 4 hypothetical screening strategies against no screening. Thus, 5 strategies were compared: 1) No Screening, 2) Biopsy Referral for all OPMDs with Moderate Treatment Threshold: Screening with biopsy referral for all OPMDs with surgical treatment threshold of moderate dysplasia or worse, 3) Biopsy Referral for all OPMDs with Severe Treatment Threshold: Screening with biopsy referral for all OPMDs with surgical treatment threshold of severe dysplasia or worse, 4) Risk-stratified Referral for Biopsy with Moderate Treatment Threshold: Screening with decision to refer for biopsy based on a positive smoking history or for OPMDs that are nonhomogeneous, nonwhite, or at a high-risk site⁷⁵ with a surgical treatment threshold of moderate dysplasia or worse, and 5) Risk-stratified Referral for Biopsy with Severe Treatment Threshold: same biopsy referral criteria as 4) but with surgical treatment threshold of severe dysplasia or worse (Figure 2 and Supplementary Table S2). With no screening, only patients with cancer presented to care, underwent biopsy, and then received treatment according to OSCC stage at diagnosis. For the screening strategies, screening was assumed to initiate at the age of 50 y, occurred every 3 y in the base-case analysis, and ended at the age of 75 y. We assumed patients have 100% adherence to dentist visits for screening and received the visual–tactile examinations during their visits. When an initial risk stratification was included in the visual–tactile examination, only OPMDs with the concerning features were referred for biopsy. Positive biopsies were treated with surgical excision while lesser disease states underwent surveillance. Two definitions of a positive biopsy, namely, moderate and severe treatment thresholds, were considered in the 4 management strategies. When moderate dysplasia was the threshold for positivity, surgical excision was performed for moderate or severe dysplasia or localized carcinoma. Meanwhile, surveillance was undertaken for nondysplastic OPMDs and mild dysplasia. When the threshold was severe dysplasia, severe dysplasia and carcinoma were then treated with surgery. The surveillance duration was 5 y and occurred every 6 mo during the first year and every year thereafter.

Figure 2

Screening schematics for the 5 strategies.

In the base case, we assumed that the sensitivity and specificity of visually differentiating homogeneous white OPMDs from other types of OPMDs (red or nonhomogeneous white) was 100% in the risk-stratified referral for biopsy strategies. In the sensitivity analysis, the sensitivity and specificity of visually differentiating homogeneous white OPMDs from other OPMDs varied from 100% to 71% and 97%, respectively, such that the performance was similar to distinguishing between benign and malignant lesions on the standard visual tactile examination.⁶⁶ Patients were subject to surgical mortality rates according to the literature.^62,63 For each strategy, we simulated each subpopulation of 2 million patients and 500 repetition iterations with the starting age of 40 y using a lifetime horizon and all-cause mortality from the 2011 US life table.⁸³ The outcomes of the overall population were derived from the weighted average of subpopulation outcomes using a distribution of 4 subpopulations—very-low, low, intermediate, and high risks, defined by degrees of smoking and alcohol exposure (Table 1)—as corresponding weights. Using the medical sector perspective,³⁰ we evaluated the primary outcomes of LE in patients who developed OPMDs from age 40 to 60 y and the intermediate outcomes of cancer incidence, cancer-specific deaths, and the number of biopsies accumulated over lifetime. The evaluation was conducted for the overall cohort of men and women separately in the base-case analysis. To identify efficient strategies, we compared the tradeoff of benefit gain and harm of screening strategies, both against no screening. Benefit gains were calculated as a reduction in the cumulative cancer incidence and cancer-specific deaths. Harm was calculated as an increase in the number of biopsies. The strategies were sorted by the order of benefits, and any strategies that were strongly dominated by others were ruled out (less benefit and more harm). We calculated the incremental benefit–harm ratio for each strategy by comparing it with the next most effective strategy. The strategy was efficient if it provided higher benefit with less harm as compared with alternative strategies.^84,85 The identified efficient strategies were presented as a frontier plot, where the dominated strategies were under the frontier.

For the scenario analyses, we evaluated 2 scenarios: 1) a longer screening interval of 10 y and 2) infrequent dental visits in which an adherence to a dental visit was reduced from 100% to 50%.^86,87 Sensitivity analysis was performed by varying major parameter values (1-way deterministic analysis) to assess the impact on model results. Major parameters included the prevalence of OPMD at the age of 40 y, clinical parameters relevant to natural history, and screening characteristic parameters (Table 1).

Results

Model Calibration

SCORE was well-calibrated to all 3 targets from SEER. Among the 100 best-fitted parameter sets, we found weak correlations among the 36 calibrated parameters, supporting the validity of independence assumption. Specifically, of 630 calibrated parameter pairs representing all unique combinations of 2 parameters selected from 36 calibrated parameters (i.e., 36 choose 2, without repetition; $C_{2}^{36} = 630$ ), 96% of the pairs had correlation coefficients below ±0.2 and 70% had coefficients below ±0.1, and only 2 pairs had correlation coefficients exceeding 0.3 (0.31 and 0.36). The ranges of the sum of chi-squares were 1.55 to 10.69 (standard error [SE]: 0.23) and 1.72 to 7.06 (SE: 0.13) for men and women, respectively (Figure 3). Based on the best-fitted parameters (Table 1), SCORE matched cancer incidences in both men and women, slightly overestimating the incidence in men aged 50 to 60 y and women aged 60 to 65 y (Figure 3A and C). Simultaneously, SCORE calibration was closely matched to stage at diagnosis with SE ≤ 0.3% and SE ≤ 0.4% across all cancer stages in men (Figure 3B) and women (Figure 3D), respectively. There was no significant difference in the 5-y cancer-specific survival for patients with stage IV cancer at age 50 y, which was 41.6% (95% CI: 41.3%–41.9%) in SCORE, compared with the observed 41.4% (95% CI: 36.4%–46.3%) in SEER.

Figure 3

Model calibration results for the overall population comparing Simulation of Cancers of the Oral cavity and Risk Exposures (SCORE) model outcomes against Surveillance, Epidemiology, and End Results (SEER) 2004–2010.

Model Validation

In the overall population, SCORE provided well-matched outcomes to observations from SEER 2011–2017 with the sum of chi-squares ranging from 2.97 to 13.03 (SE: 0.35) (Supplementary Figure S1). The 95% CIs of cancer incidence rates from SCORE overlapped with those from SEER 2011–2017 in all age groups with a slightly overestimated incidence in age 55 to 60 y and age 75 to 85+ y (Supplementary Figure S1A). Compared with SEER, SCORE estimated a closely matched distribution of stages at diagnosis with SE ≤ 0.42% across all stages (Supplementary Figure S1B). In addition, SEER 2011–2017 reported the 5-y cancer-specific survival rate at stage IV cancer of 38.4% (95% CI: 33.3%–43.6%), while SCORE simulated the 5-y rate of 41.0% (95% CI: 33.1%–48.8%). In addition, for 50-y-old men, SCORE estimated an LE without cancer of 28.5 y (95% CI: 26.8–30.3) and 19.8 y (95% CI: 17.8–22.2) in nonsmokers and smokers, respectively. The estimated LE fell within SEER OCSC reported ranges of 23 to 34 y for nonsmokers and 17 to 30 y for current smokers.

Application: Evaluation of Representative Screening Strategies

All screening strategies yielded similar LE in patients who developed OPMDs from age 40 to 60 y of 36.37±0.01 life-years (+0.03 y compared with no screening) in the general population of 50-y-old men and 30.97 ± 0.01 life-years (+0.02 y compared with no screening) in the high-risk subpopulation (Table 2). Screening, whether with or without risk-stratified biopsy, and paired with a treatment threshold of moderate dysplasia, provided the highest reduction in cumulative oral cancer incidence of 30%. All screening strategies yielded a similar reduction in cancer-specific death of at least 20% compared with no screening. Compared with screening with biopsy referral for all OPMDs, screening with risk-stratified biopsy referral yielded 30% fewer biopsies (1,300–1,314 v. 1,877–1,977 biopsies per 100,000 patients), leading to 8 to 12 biopsies needed to diagnose 1 treatable condition in persons receiving first biopsies and with minimal difference in LE in patients aged 40 to 60 y who developed OPMDs, cancer incidence, or cancer-specific death.

Table 2

Comparison of Primary and Intermediate Outcomes of Screening and No Screening Strategies for the Base Case of Screening Starting at Age 50 y in US Men with a 3-y Screening Interval

Strategy	LE in People Aged 40–60 y with OPMDs (y)	Number per 100,000 Persons				Biopsies Needed to Detect 1 Treatable Case^a
Strategy	LE in People Aged 40–60 y with OPMDs (y)	Lifetime Incidence	Cancer-Specific Death	Cumulative Number of Biopsies	Precancer and Cancer Diagnosed	Biopsies Needed to Detect 1 Treatable Case^a
Overall population
No screening	36.342(36.264–36.414)	91.5(90.0–93.0)	23.5(22.8–24.2)	106(104–107)	56(55–58)^a	NA
Screening with biopsy referral for all OPMDs with surgical treatment threshold of moderate dysplasia or worse	36.380(36.317–36.448)	64.3(63.0–65.6)	17.4(16.7–18.1)	1,877(1,870–1,884)	170(168–173)^b	11.0(10.9–11.1)
Screening with biopsy referral for all OPMDs with surgical treatment threshold of severe dysplasia or worse	36.369(36.292–36.437)	67.1(65.8–68.5)	17.8(17.1–18.5)	1,977(1,969–1,983)	118(117–120)^c	16.7(16.5–16.9)
Screening with risk-stratified biopsy with surgical treatment threshold of moderate dysplasia or worse	36.373(36.309–36.444)	64.7(63.2–66.1)	18.5(17.8–19.2)	1,314(1,308–1,319)	164(162–166)^b	8.0(7.9–8.1)
Screening with risk-stratified biopsy with surgical treatment threshold of severe dysplasia or worse	36.371(36.300–36.438)	67.7(66.3–69.2)	18.7(18.0–19.3)	1,300(1,294–1,306)	113(112–115)^c	11.5(11.3–11.6)
High-risk subpopulation
No screening	30.950(30.929–30.973)	186.9(185.0–189.0)	42.3(41.5–43.3)	219(216–221)	126(125–128)^b	NA
Screening with biopsy referral for all OPMDs with surgical treatment threshold of moderate dysplasia or worse	30.972(30.950–30.992)	129.5(127.8–131.2)	33.2(32.4–34.0)	3,666(3,657–3,674)	376(373–378)^c	9.8(9.7–9.8)
Screening with biopsy referral for all OPMDs with surgical treatment threshold of severe dysplasia or worse	30.963(30.940–30.988)	136.6(134.7–138.2)	34.0(33.1–34.8)	3,824(3,816–3,832)	244(242–247)^d	15.7(15.5–15.8)
Screening with risk-stratified biopsy with surgical treatment threshold of moderate dysplasia or worse	30.974(30.950–30.997)	129.7(128.0–131.3)	33.1(32.2–33.9)	3,666(3,658–3,673)	376(373–378)^c	9.8(9.7–9.8)
Screening with risk-stratified biopsy with surgical treatment threshold of severe dysplasia or worse	30.973(30.952–30.993)	136.8(134.9–138.5)	33.5(32.7–34.3)	3,625(3,617–3,632)	239(236–241)^c	15.2(15.0–15.3)

LE, life expectancy; OPMDs, oral potentially malignant disorders. All outcomes are reported as mean (95% interpercentile range).

The numerator is the number of patients with OPMDs receiving first biopsies, and the denominator is the number of treatable cases from first biopsies.

Treatable cases include only early-stage cancers consisting of cancer stage I and II.

Treatable cases include precancers consisting of moderate and severe dysplasia and early-stage cancers.

Treatable cases include precancers of severe dysplasia only and early-stage cancers.

For the first scenario analysis, we evaluated an increase in screening intervals to 10 y from the initial assumption of 3 y. This further reduced the biopsy burden by approximately 244 to 409 biopsies per 100,000 individuals without a substantial change in LE in patients aged 40 to 60 y who developed OPMDs or overall cancer incidence but with a reduction in early-stage cancer detection by 10% to 13% compared with 3-y screening intervals (depending on the chosen treatment threshold) (Supplementary Table S3). In the second scenario analysis, we evaluated screening strategies with 3-y screening intervals with a decrease in adherence to a dental visit to 50% from the initial assumption of 100%. Compared with the base-case analysis, there was no substantial change in LE in patients aged 40 to 60 y who developed OPMDs or overall cancer incidence (Supplementary Table S4). However, there was a reduction in early-stage cancer detection by 7% to 9% in the general population and by 7% to 10% in the high-risk population, respectively (depending on the chosen treatment threshold). For all scenarios, there were minimal differences in the overall LE across all 5 strategies (Supplementary Table S6).

In women, similar to men, screening with risk-stratified biopsy strategies yielded a significantly lower biopsy burden (960–961 v. 1,402–1,468 biopsies per 100,000 patients) with similar LE (−0.01 y), cumulative oral cancer incidence, and cancer-specific death when compared with screening with biopsy referral for all OPMDs (Supplementary Table S5). In addition, screening with risk-stratified biopsy strategy and a treatment threshold of moderate dysplasia led to reductions in the cumulative number of detected treatable cases of 3% to 6% in women and 3% to 4% in men, compared with biopsy referral for all OPMDs, depending on treatment threshold.

Furthermore, we investigated the outcomes by patient risk profiles. Considering the incidence reduction and biopsy burden in very-low-risk men, both screening with and without risk-stratified biopsy and paired with a treatment threshold of moderate dysplasia were on the frontier (Figure 4). Specifically, screening with a biopsy referral for all OPMDs yielded an additional reduction in the incidence of 0.27 per 100,000 patients with an increment in biopsies of 308 per 100,000 patients compared with risk-stratified biopsy (Figure 4A). Similar results were observed in low-risk men. In intermediate-risk and high-risk men, screening with risk-stratified biopsy with a treatment threshold of moderate dysplasia yielded similar benefits and harms to screening with a biopsy referral for all OPMDs at the same treatment threshold. Screening with biopsy referral for all OPMDs with a moderate treatment threshold was on the frontier, providing an additional incidence reduction of only 0.15 per 100,000 patients, compared with screening with risk-stratified biopsy at the same treatment threshold (Figure 4A). However, when comparing cancer-specific death averted to biopsy burden, screening with risk-stratified biopsy with a treatment threshold of moderate dysplasia was on the frontier, with only an additional 0.12 cancer-specific deaths averted per 100,000 patients (Figure 4B). Screening with and without a risk-stratified biopsy with a treatment threshold of moderate dysplasia yielded the highest biopsy efficiency of 7 to 13 biopsies needed to detect 1 treatable case across all 4 subpopulations (Supplementary Figure S2).

Figure 4

Comparison between the following benefit gained and burden of screening strategies compared with no screening.

Sensitivity Analysis

In one-way sensitivity analysis, each major parameter was varied across a plausible range. The OSCC incidence and disease-specific death were affected mostly by changes in the probability of developing an OPMD, the probability of progressing from mild dysplasia to malignancy, the probability of progressing from moderate to severe dysplasia (Supplementary Figures S3 and S4). The comparative favorability of strategies remained stable.

Discussion

We developed, calibrated, and validated a new microsimulation model (SCORE) of the natural history of OPMD development and progression of OPMD and OSCC to enable virtual trials in US men and women aged 40 y and their subpopulations. SCORE differs from other existing models with the incorporation of major behavioral risk factors and OPMD features that are associated with disease development and progression and allowed for various phenotypes of OPMD development in a cohort of patients under simplified assumptions. As such, SCORE can be used to bridge the gap in population screening evaluation for the United States and evaluating various OPMD screening and management strategies that use behavioral risk factors and/or OPMD features. Accordingly, we evaluated the potential benefits and harms of alternate screening strategies.

Our results showed that screening with risk-stratified biopsy led to a minimal improvement in primary and intermediate outcomes in very-low-risk persons. However, both screening and no screening strategies led to a similar overall LE in the general population and subpopulations, primarily due to the low prevalence of OPMDs among US men and women and the high proportion of benign disorders among OPMDs. Screening strategies, whether with or without risk-stratified biopsy, yielded comparably significant improvement in cancer-specific death and cancer incidence in high-risk persons. Compared with screening strategies with biopsy referral for all OPMDs, screening strategies with risk-stratified biopsy could greatly reduce biopsy burden in patients with low exposure to smoking and alcohol use while maintaining similar levels of reduction in cancer incidence and cancer-specific death and increase in biopsy efficiency. This finding underscores the importance of the use of risk stratification of OPMDs in management decisions. Furthermore, risk stratification models that improve upon simple decision rules based on low versus non-low-risk OPMD features may further improve the benefits and harms considerations for screening. In addition, SCORE demonstrated that population screening for OPMDs best improved cancer detection when risk-stratified biopsy tempered the procedure burden and enabled diagnostic yield comparable with established cancer screening programs such as cervical cancer screening.⁸⁸ We believe that SCORE may be useful for guiding health policy and identify high-priority clinical trials in precision oral cancer screening.

A prior decision analytics study by Dedhia et al.⁸⁹ examined a yearly community-based screening program in 40-y-old US high-risk men who regularly use tobacco and/or alcohol over a time horizon of 40 y and found that screening is more effective and less costly compared with no screening. Our model emphasizes additional risk factors that are associated with malignant transformation: degrees of smoking and alcohol exposure ranging from very low to high, dysplasia grade spectrum in “precancer,” and OPMD features of color, texture, and sites. Specifically, SCORE modeled various dysplasia grades exhibiting within precancer, while Dedhia et al. modeled only one “precancer” state. However, both models considered the potential regression of dysplastic OPMD to a healthy state. SCORE also assumed older ages were associated with higher probabilities of disease progression, whereas Dedhia et al. assumed constant probabilities of progression. SCORE uniquely simulates US patients with and without OPMD development and incorporates both patient risk factors and OPMD features that were influential to OPMD development and MT.

Our study had several limitations. Model assumptions of disease development and progression were simplified using a multiplicative effect for behavioral risk factors and OPMD features observed to elevate the risk of malignancy. The existing literature reported a greater than multiplicative effect when only behavioral risk factors were considered in association with risk of malignancy,⁹⁰ although confounding and selection bias may also be reflected. Some of the model parameters including the initial proportions of OPMD anatomic sites and grades by each OPMD phenotype were informed by US survey data on oral lesions or described OPMD characteristics in patients with OPMDs, and misclassification errors are possible.^{40,42–44,46–49} As OPMDs were not common in younger patients, there is a limited literature on screening studies for younger age groups. Although mortality risks should be adjusted for major comorbidities and smoking and alcohol exposures, we include only an increase in all-cause death due to smoking and alcohol due to the lack of comorbidity-adjusted life tables that accounted for smoking and alcohol exposure.

In conclusion, SCORE is a new microsimulation tool for the virtual trial for specific risk groups who may undergo screening and management for OPMDs. We applied SCORE to the US population and key risk-stratified subpopulations, and our results point to specific future research needs evaluating screening strategies augmented by rule-based biopsy referral using risk factors and specifically the possibility that more advanced testing technologies could further increase the efficiency in biopsy. By enabling analyses that assess the downstream impact of risk factors, OPMD features, screening regimen, and other test performance, SCORE will be available to address additional policy and practice questions for improving oral cancer control.

Supplemental Material

sj-docx-1-mpp-10.1177_23814683251353226 – Supplemental material for Simulation Modeling of Oral Cancer Development with Risk Stratification: How Potential Screening Programs Can Be Evaluated

Supplemental material, sj-docx-1-mpp-10.1177_23814683251353226 for Simulation Modeling of Oral Cancer Development with Risk Stratification: How Potential Screening Programs Can Be Evaluated by Mutita Siriruchatanon, Emily R. Brooks, Alexander R. Kerr, Denise M. Laronde, Miriam P. Rosin and Stella K. Kang in MDM Policy & Practice

Footnotes

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr. Kang reports royalties from Wolters Kluwer for unrelated work. The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support for this study was provided entirely by a grant from the National Institutes of Health/National Institute of Dental and Craniofacial Research (R01DE030169, principal investigator: Kang). The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report. The data generated in this study are not publicly available due to protections of patient privacy and the scope of informed consent, but summaries of the data may be available upon reasonable request from the corresponding author

Ethical Considerations

This study is based entirely on simulated data generated using the authors’ simulation model. Model inputs were derived from previously published studies, all of which are appropriately referenced. The study does not involve any human participants, identifiable personal data, or real-world experimental interventions. Therefore, ethical approval from an institutional review board was not required. As the data were fully simulated, there are no ethical concerns related to human or animal research.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

ORCID iD

Mutita Siriruchatanon

Data Availability

The data generated in this study are not publicly available due to protections of patient privacy and the scope of informed consent, but summaries of the data may be available upon reasonable request from the corresponding author.

References

Siegel

Miller

Fuchs

Jemal

. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33.

Siegel

Naishadham

Jemal

. Cancer statistics, 2013. CA Cancer J Clin. 2013;63(1):11–30.

Sloan

Gale

Hunter

, et al. Malignant surface epithelial tumours: squamous cell carcinoma. In: El-Naggar

Chan

JKC

Grandis

, et al. eds. WHO Classification of Tumours of the Head and Neck. 4th ed. Lyon (France): IARC Press; 2017:81–90.

Choi

Myers

. Molecular pathogenesis of oral squamous cell carcinoma: implications for therapy. J Dent Res. 2008;87(1):14–32.

Chamoli

Gosavi

Shirwadkar

, et al. Overview of oral cavity squamous cell carcinoma: risk factors, mechanisms, and diagnostics. Oral Oncol. 2021;121:105451.

Stepan

Mazul

Larson

, et al. Changing epidemiology of oral cavity cancer in the United States. Otolaryngol Head Neck Surg. 2023;168(4):761–8.

Salma

Tara

Zaid

A-Q

, et al. Trends in incidence of oral cavity squamous cell carcinoma in the United States 2001–2019. Surg Oncol Insight. 2024;1(2):100055.

Garzino-Demo

Zavattero

Franco

, et al. Parameters and outcomes in 525 patients operated on for oral squamous cell carcinoma. J Craniomaxillofac Surg. 2016;44(9):1414–21.

van Dijk

Brands

Geurts

Merkx

Roodenburg

. Trends in oral cavity cancer incidence, mortality, survival and treatment in the Netherlands. Int J Cancer. 2016;139(3):574–83.

10.

Listl

Jansen

Stenzinger

, et al. Survival of patients with oral cavity cancer in Germany. PLoS One. 2013;8(1):e53415.

11.

Luryi

Chen

Mehra

Roman

Sosa

Judson

. Treatment factors associated with survival in early-stage oral cavity cancer: analysis of 6830 cases from the National Cancer Data Base. JAMA Otolaryngol Head Neck Surg. 2015;141(7):593–8.

12.

Gourin

Kaboli

Blume

Nance

Koch

. Characteristics of participants in a free oral, head and neck cancer screening program. Laryngoscope. 2009;119(4):679–82.

13.

Lingen

Abt

Agrawal

, et al. Evidence-based clinical practice guideline for the evaluation of potentially malignant disorders in the oral cavity: a report of the American Dental Association. J Am Dent Assoc. 2017;148(10):712–27.e10.

14.

Applebaum

Ruhlen

Kronenberg

Hayes

Peters

. Oral cancer knowledge, attitudes and practices: a survey of dentists and primary care physicians in Massachusetts. J Am Dent Assoc. 2009;140(4):461–7.

15.

Barao

DMH

Essex

Lazar

Rowe

. Detection of early-stage oral cancer lesions: a survey of california dental hygienists. J Dent Hyg. 2016;90(6):346–53.

16.

Moyer

, U.S. Preventive Services Task Force. Screening for oral cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med. 2014;160(1):55–60.

17.

National Health Mission, Ministry of Health & Social Welfare, India. Operational Guidelines: Prevention, Screening and Control of Common Non-communicable Diseases: Hypertension, Diabetes and Common Cancers (Oral, Breast, Cervix). New Delhi: Government of India; 2016.

18.

Chuang

Chen

, et al. Population-based screening program for reducing oral cancer mortality in 2,334,299 Taiwanese cigarette smokers and/or betel quid chewers. Cancer. 2017;123(9):1597–609.

19.

Bouvard

Nethan

Singh

, et al. IARC perspective on oral cancer prevention. N Engl J Med. 2022;387(21):1999–2005.

20.

Çağlayan

Terawaki

Chen

Rai

Ayer

Flowers

. Microsimulation modeling in oncology. JCO Clin Cancer Inform. 2018;2:1–11.

21.

Rutter

Zaslavsky

Feuer

. Dynamic microsimulation models for health outcomes: a review. Med Decis Making. 2011;31(1):10–8.

22.

Shulman

Beach

Rivera-Hidalgo

. The prevalence of oral mucosal lesions in US adults: data from the Third National Health and Nutrition Examination Survey, 1988–1994. J Am Dent Assoc. 2004;135(9):1279–86.

23.

Weatherspoon

Chattopadhyay

Boroumand

Garcia

. Oral cavity and oropharyngeal cancer incidence trends and disparities in the United States: 2000–2010. Cancer Epidemiol. 2015;39(4):497–504.

24.

Schwartz

Doody

Fitzgibbons

Ricks

Porter

Chen

. Oral squamous cell cancer risk in relation to alcohol consumption and alcohol dehydrogenase-3 genotypes. Cancer Epidemiol Biomarkers Prev. 2001;10(11):1137–44.

25.

Barasch

Morse

Krutchkoff

Eisenberg

. Smoking, gender, and age as risk factors for site-specific intraoral squamous cell carcinoma. A case-series analysis. Cancer. 1994;73(3):509–13.

26.

Mayne

Morse

Winn

. Cancers of the oral cavity and pharynx. In: Schottenfeld

Fraumeni

, eds. Cancer Epidemiology and Prevention. 3rd ed. Oxford (UK): Oxford University Press; 2006. p 674–96.

27.

Aguirre-Urizar

Lafuente-Ibáñez de Mendoza

Warnakulasuriya

. Malignant transformation of oral leukoplakia: systematic review and meta-analysis of the last 5 years. Oral Dis. 2021;27(8):1881–95.

28.

Giuliani

Troiano

Cordaro

, et al. Rate of malignant transformation of oral lichen planus: a systematic review. Oral Dis. 2019;25(3):693–709.

29.

Reichart

Philipsen

. Oral erythroplakia—a review. Oral Oncol. 2005;41(6):551–61.

30.

Roberts

Russell

Paltiel

, et al. Conceptualizing a model: a report of the ISPOR-SMDM modeling good research practices task force-2. Med Decis Making. 2012;32(5):678–89.

31.

Kang

Mali

Braithwaite

Kerr

McDevitt

. Point-of-care characterization and risk-based management of oral lesions in primary dental clinics: a simulation model. PLoS One. 2020;15(12):e0244446.

32.

Caro

Briggs

Siebert

Kuntz

, ISPOR-SMDM Modeling Good Research Practices Task Force. Modeling good research practices—overview: a report of the ISPOR-SMDM modeling good research practices task force-1. Med Decis Making. 2012;32(5):667–77.

33.

Siebert

Alagoz

Bayoumi

, et al. State-transition modeling: a report of the ISPOR-SMDM modeling good research practices task force—3. Value Health. 2012;15(6):812–20.

34.

Briggs

Weinstein

Fenwick

, et al. Model parameter estimation and uncertainty analysis: a report of the ISPOR-SMDM modeling good research practices task force working group-6. Med Decis Making. 2012;32(5):722–32.

35.

Eddy

Hollingworth

Caro

, et al. Model transparency and validation: a report of the ISPOR-SMDM modeling good research practices task force-7. Med Decis Making. 2012;32(5):733–43.

36.

Villa

Gohel

. Oral potentially malignant disorders in a large dental population. J Appl Oral Sci. 2014;22(6):473–6.

37.

Laronde

Williams

Hislop

, et al. Decision making on detection and triage of oral mucosa lesions in community dental practices: screening decisions and referral. Community Dent Oral Epidemiol. 2014;42(4):375–84.

38.

Mello

Miguel

AFP

Dutra

, et al. Prevalence of oral potentially malignant disorders: a systematic review and meta-analysis. J Oral Pathol Med. 2018;47(7):633–40.

39.

Centers for Disease Control and Prevention (CDC), National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Data. Hyattsville (MD): US Department of Health and Human Services, Centers for Disease Control and Prevention; 2009–2018. Available from: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx.

40.

Schepman

van der Meij

Smeele

van der Waal

. Malignant transformation of oral leukoplakia: a follow-up study of a hospital-based population of 166 patients with oral leukoplakia from The Netherlands. Oral Oncol. 1998;34(4):270–5.

41.

Kierce

Shi

Klieb

Blanas

Magalhaes

. Identification of specific clinical risk factors associated with the malignant transformation of oral epithelial dysplasia. Head Neck. 2021;43(11):3552–61.

42.

P-S

Chen

P-L

Warnakulasuriya

Shieh

T-Y

Chen

Y-K

Huang

. Malignant transformation of oral potentially malignant disorders in males: a retrospective cohort study. BMC Cancer. 2009;9(1):1–7.

43.

Holmstrup

Vedtofte

Reibel

Stoltze

. Long-term treatment outcome of oral premalignant lesions. Oral Oncol. 2006;42(5):461–74.

44.

Scheifele

Reichart

Dietrich

. Low prevalence of oral leukoplakia in a representative sample of the US population. Oral Oncol. 2003;39(6):619–25.

45.

Shafer

Waldron

. Erythroplakia of the oral cavity. Cancer. 1975;36(3):1021–8.

46.

Lumerman

Freedman

Kerpel

. Oral epithelial dysplasia and the development of invasive squamous-cell carcinoma. Oral Surg Oral Med Oral Pathol Oral Radiol Endod. 1995;79(3):321–9.

47.

Laniosz

Torgerson

Ramos-Rodriguez

, et al. Incidence of squamous cell carcinoma in oral lichen planus: a 25-year population-based study. Int J Dermatol. 2019;58(3):296–301.

48.

Bardellini

Amadori

Flocchini

Bonadeo

Majorana

. Clinicopathological features and malignant transformation of oral lichen planus: a 12-years retrospective study. Acta Odontol Scand. 2013;71(3–4):834–40.

49.

Chaturvedi

Udaltsova

Engels

, et al. Oral leukoplakia and risk of progression to oral cancer: a population-based cohort study. J Natl Cancer Inst. 2020;112(10):1047–54.

50.

Shearston

Fateh

Tai

Hove

Farah

. Malignant transformation rate of oral leukoplakia in an Australian population. J Oral Pathol Med. 2019;48(7):530–7.

51.

National Cancer Institute. Surveillance, Epidemiology, and End Results (SEER) - SEER research data, 17 registries, Nov 2022 Sub (2000–2020). 2023. Available from: www.seer.cancer.gov.

52.

Silverman

Gorsky

Lozada

. Oral leukoplakia and malignant transformation—a follow-up-study of 257 patients. Cancer. 1984;53(3):563–8.

53.

Rock

Rosin

Zhang

Chan

Shariati

Laronde

. Characterization of epithelial oral dysplasia in non-smokers: first steps towards precision medicine. Oral Oncol. 2018;78:119–25.

54.

Jaber

Porter

Gilthorpe

Bedi

Scully

. Risk factors for oral epithelial dysplasia—the role of smoking and alcohol. Oral Oncol. 1999;35(2):151–6.

55.

Morse

Katz

Pendrys

, et al. Smoking and drinking in relation to oral epithelial dysplasia. Cancer Epidemiol Biomarkers Prev. 1996;5(10):769–77.

56.

Schwartz

Daling

Doody

, et al. Oral cancer risk in relation to sexual history and evidence of human papillomavirus infection. J Natl Cancer Inst. 1998;90(21):1626–36.

57.

Evren

Brouns

Wils

, et al. Annual malignant transformation rate of oral leukoplakia remains consistent: a long-term follow-up study. Oral Oncol. 2020;110:105014.

58.

Iocca

Sollecito

Alawi

, et al. Potentially malignant disorders of the oral cavity and oral dysplasia: a systematic review and meta-analysis of malignant transformation rate by subtype. Head Neck. 2020;42(3):539–55.

59.

Banoczy

. Follow-up studies in oral leukoplakia. J Maxillofac Surg. 1977;5(1):69–75.

60.

Warnakulasuriya

Kovacevic

Madden

, et al. Factors predicting malignant transformation in oral potentially malignant disorders among patients accrued over a 10-year period in South East England. J Oral Pathol Med. 2011;40(9):677–83.

61.

Mucke

Wagenpfeil

Kesting

Holzle

Wolff

. Recurrence interval affects survival after local relapse of oral cancer. Oral Oncol. 2009;45(8):687–91.

62.

Wiemer

Nathan

Heggestad

, et al. Safety of outpatient procedural sedation administered by oral and maxillofacial surgeons: the mayo clinic experience in 17,634 sedations (2004 to 2019). J Oral Maxillofac Surg. 2021;79(5):990–9.

63.

Bhattacharyya

Fried

. Benchmarks for mortality, morbidity, and length of stay for head and neck surgical procedures. Arch Otolaryngol Head Neck Surg. 2001;127(2):127–32.

64.

Qin

Magnussen

Steffen

Zhao

. Light cigarette smoking increases risk of all-cause and cause-specific mortality: findings from the NHIS cohort study. Int J Environ Res Public Health. 2020;17(14):5122.

65.

Veeranki

Zhao

Yan

. Relationship of alcohol consumption to all-cause, cardiovascular, and cancer-related mortality in U.S. adults. J Am Coll Cardiol. 2017;70(8):913–22.

66.

Downer

Moles

Palmer

Speight

. A systematic review of test performance in screening for oral cancer and precancer. Oral Oncol. 2004;40(3):264–73.

67.

Downer

Evans

Hughes Hallet

Jullien

Speight

Zakrzewska

. Evaluation of screening for oral cancer and precancer in a company headquarters. Community Dent Oral Epidemiol. 1995;23(2):84–8.

68.

Ikeda

Downer

Ishii

Fukano

Nagao

Inoue

. Annual screening for oral cancer and precancer by invitation to 60-year-old residents of a city in Japan. Community Dent Health. 1995;12(3):133–7.

69.

Epstein

Guneri

Boyacioglu

Abt

. The limitations of the clinical oral examination in detecting dysplastic oral lesions and oral squamous cell carcinoma. J Am Dent Assoc. 2012;143(12):1332–42.

70.

Giunta

Meyer

Shklar

. The accuracy of the oral biopsy in the diagnosis of cancer. Oral Surg Oral Med Oral Pathol. 1969;28(4):552–6.

71.

Tang

Zheng

, et al. Global prevalence and incidence estimates of oral lichen planus: a systematic review and meta-analysis. JAMA Dermatol. 2020;156(2):172–81.

72.

Warnakulasuriya

Ariyawardana

. Malignant transformation of oral leukoplakia: a systematic review of observational studies. J Oral Pathol Med. 2016;45(3):155–66.

73.

Warnakulasuriya

. Oral potentially malignant disorders: a comprehensive review on clinical aspects and management. Oral Oncol. 2020;102:104550.

74.

Warnakulasuriya

Kujan

Aguirre-Urizar

, et al. Oral potentially malignant disorders: a consensus report from an international seminar on nomenclature and classification, convened by the WHO Collaborating Centre for Oral Cancer. Oral Dis. 2021;27(8):1862–80.

75.

Speight

Khurram

Kujan

. Oral potentially malignant disorders: risk of progression to malignancy. Oral Surg Oral Med Oral Pathol Oral Radiol. 2018;125(6):612–27.

76.

Karnon

Vanni

. Calibrating models in economic evaluation: a comparison of alternative measures of goodness of fit, parameter search strategies and convergence criteria. Pharmacoeconomics. 2011;29(1):51–62.

77.

van der Steen

van Rosmalen

Kroep

, et al. Calibrating parameters for microsimulation disease models: a review and comparison of different goodness-of-fit criteria. Med Decis Making. 2016;36(5):652–65.

78.

Frazier

. Bayesian optimization. In: Recent Advances in Optimization and Modeling of Contemporary Problems. INFORMS; 2018. p 255–78.

79.

Frazier

. A tutorial on Bayesian optimization. 2018. arXiv preprint arXiv:180702811.

80.

Davies

Hankey

Wang

, et al. A new personalized oral cancer survival calculator to estimate risk of death from both oral cancer and other causes. JAMA Otolaryngol Head Neck Surg. 2023;149(11):993–1000.

81.

Davies

Hankey

Wang

, et al. Key points for clinicians about the SEER oral cancer survival calculator. JAMA Otolaryngol Head Neck Surg. 2023;149(11):1042–6.

82.

Cho

Wang

Yabroff

, et al. Estimating life expectancy adjusted by self-rated health status in the United States: national health interview survey linked to the mortality. BMC Public Health. 2022;22(1):141.

83.

Arias

. United States life tables, 2011. Natl Vital Stat Rep. 2015;64(11):1–63.

84.

Gold

. Cost-Effectiveness in Health and Medicine. Oxford (UK): Oxford University Press; 1996.

85.

Vilaprinyo

Forne

Carles

, et al. Cost-effectiveness and harm-benefit analyses of risk-based screening strategies for breast cancer. PLoS One. 2014;9(2):e86858.

86.

Ferreira

Rodrigues

Moreira

Ribeiro

Longatto-Filho

. Breast cancer screening adherence rates and barriers of implementation in ethnic, cultural and religious minorities: a systematic review. Mol Clin Oncol. 2021;15(1):139.

87.

Lake

Shusted

Juon

, et al. Black patients referred to a lung cancer screening program experience lower rates of screening and longer time to follow-up. BMC Cancer. 2020;20(1):561.

88.

Wentzensen

Walker

Gold

, et al. Multiple biopsies and detection of cervical cancer precursors at colposcopy. J Clin Oncol. 2015;33(1):83–9.

89.

Dedhia

Smith

Johnson

Roberts

. The cost-effectiveness of community-based screening for oral cancer in high-risk males in the United States: a Markov decision analysis approach. Laryngoscope. 2011;121(5):952–60.

90.

Hashibe

Brennan

Chuang

, et al. Interaction between tobacco and alcohol use and the risk of head and neck cancer: pooled analysis in the International Head and Neck Cancer Epidemiology Consortium. Cancer Epidemiol Biomarkers Prev. 2009;18(2):541–50.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.39 MB