The primary purpose of this study was to provide an acceptable budget for an oral cancer screening program based on decreased medical costs and increases QALYs gained.
We have shown that yearly oral cancer screening via visual inspection and manual palpation in the community-based setting for high-risk American males over 40 years age would be considered cost-effective with a budget of $3363 per screened person over the 40 year cycle. This figure is derived from both savings in costs of management ($258) and increase in quality-adjusted life years ($3105) of the Screen group.
Sensitivity analysis illustrates the impact of key variables on the value of a screening program. First, the level of penetrance into this high-risk group via screening greatly affects the effectiveness of the program. Put simply, the greater the number of mouths that can be examined the greater the opportunity to catch disease in its early stages. In addition, compliance with biopsy bears a tremendous impact on the cost-effectiveness of a screening program. This represents potential missed opportunities of early cancer treatment. Finally, by varying the incidence of precancer in this model, we were able to simulate both a general population cohort and an ultra high-risk cohort. The findings demonstrate the per capita dollar value of screening each of these cohorts.
We tested our model with a validation step to ensure real-world applicability. Direct values from the literature were not available for either validation reference value. Prevalence of leukoplakia was estimated by assuming a relative risk of 2.5 for our high-risk population. Leukoplakia prevalence within the model at mid-cycle appeared to closely follow this estimated reference value of leukoplakia prevalence. The second validation criterion, death rates from oral cancer was derived from SEER data for oral cavity & pharynx cancer death rates from 1993–2002 in Detroit males aged 50–74. According to a 2009 American Cancer Society report, 30% of deaths from oral cavity and oropharynx result from oropharynx primary lesions.34
Removing this fraction, our model validation shows a model value of 0.0402% and a new reference value range from 0.0165%–0.0412%. While the model value remains within the range, it remains at the upper extreme suggesting that the model may have mildly overestimated death rates in this population.
We sought to develop a conservative model of the burden and costs of oral cancer to strengthen our findings. The utility of the cohort in the healthy state was chosen to be 0.84. We incorporated a disutility of 0.16 in the healthy population to bias against an intervention which decreases a relative disutility such as screening. We also defined high-risk as regular, recent (within last year) users of tobacco or alcohol. The relative risk for development of premalignant lesions increases from 1.9 for past tobacco and current light drinkers (<15g/day) to 10.2 for current tobacco users and heavy drinkers. We selected a low relative risk of roughly 2.5 which represents past tobacco users and moderate alcohol users.16
We did not model the potential reduction in prevalence in the screening arm as a result of patient education.6
It has been shown that cessation of tobacco can lead to a reduction or disappearance of lesions in 44-80% of subjects17,35
With respect to diagnosis, treatment and workup costs, only direct costs were modeled in our analysis. Head and neck cancer, particularly advanced disease, bears enormous indirect costs including lost wages and the time of caregivers.36
We also did not model indirect costs of biopsy including anxiety for biopsies as a result of false positive oral examinations. With specificity of oral exam near 99%, this represents a relatively small proportion of the population. Costs of complications were not included in this analysis. Recent literature demonstrates that 17% of costs for chemotherapy for advanced head and neck carcinoma are ascribed to complication-related treatment.37
We intentionally did not include costs of the program in our analysis. We felt that the modeling of program costs would require major assumptions and potentially undermine the credibility of our study. Our primary objective was to provide financial parameters to direct further research for screening program development.
Our analysis has several limitations. In regards to uncertainty about the natural history of oral cancer, we chose one year as our cycle length and effectively applied a 3-year minimum cycle to oral cancer. In our model, the lesion starts as normal tissue, progresses to leukoplakia at 0.8% per year and then to Stage I/II cancer at 0.8% per year and to Stage III/IV at 50% per year. It is likely that aggressive variants of squamous cell carcinoma exhibit aggressive behavior with a shorter time course. Similarly, we categorized all lesions as oral tongue for purposes of standardization. Floor of mouth and buccal mucosa lesions present different treatment challenges which were not entertained in this study.
We attempted to utilize data from the United States for oral cancer but encountered difficulty in obtaining specific data. For example, we used data from national SEER reports which combine oral and oropharyngeal carcinoma into a single category. As a result, 5-year survival rates used in this study represented a composite of oral and oropharyngeal carcinoma survival data. While the National Cancer Database (NCDB) separately records oral cancer data, it is not a population-based registry and does not provide incidence and prevalence figures. Hence, we were not able to use the NCDB oral cancer figures in our study. Similarly, incidence and utility data were sparse. Incidence data in a large prospective study for premalignant lesions in the United States was not available. While India has produced much literature on this topic given the high prevalence of disease, we used incidence data from a large Japanese study to more closely resemble the risk factors of males in the United States.19
Data regarding utilities for oral precancer and cancer states emanates from a small British study in which lesions were classified as premalignant, small (<2cm) and large (>2cm). While Stage II lesions are >2cm, we applied utilities for small (<2cm) to Stage I and II given similar treatment options.
All cancers in this model were presumed to be oral tongue for purposes of standardization. Given that this high-risk population is at risk for head and neck squamous cell carcinomas including other regions of oral cavity, oropharynx, and larynx, consideration should be given to include direct and indirect examination of these areas during screening. While not explicitly included in this model, screening for multiple sites during the head and neck examination could increase the overall cost-effectiveness of screening.
The United Kingdom National Health Service’s (NHS) Health Technology Assessment Programme published a report in 2006 on the cost-effectiveness of screening for oral cancer in primary care. Similar to our analysis, a Markov analysis was undertaken to identify subpopulations benefiting from such an intervention. The results demonstrated that opportunistic high-risk screening, particularly men, by general dental practitioners to be a ‘practical proposition’ with an incremental cost-effectiveness ratio below our presumed $75,000 per QALY threshold. While this report examined a national health system and an office-based screening setting, the identification of high-risk males as unique opportunities bears mention.
While the design and execution of a community-based screening program is beyond the scope of this paper as previously stated, the ‘next steps’ warrant some discussion. For a particular cohort of 40-year old men, the budget for a screening program amounts to roughly $84 ($3363 divided by 40 year cycle) per individual per year. Further research needs to clarify the feasibility of such a screening endeavor. Factors determining feasibility include: number of high-risk males in a community, costs of training and employment of health workers, costs of space, ability to perform biopsy during screening, costs of biopsy materials and pathologist reading fees, tobacco and alcohol prevention education materials. There appears to be a strong benefit to develop such a program in affiliation with a university assuming government funds are not available. The costs for training and hiring health workers can be substantially reduced through the use of medical and dental students. Trained health workers have proven nearly as accurate as practicing generalists and dentists in the identification of suspicious lesions.28
Though the universities may assist in defraying costs of personnel and other resources, recruitment of the desired patient population remains the limiting factor. As previous studies have shown, free, office-based head and neck cancer screenings attract low-risk women as opposed to high-risk men. The cohort examined in this model is exclusively composed of men with recent, regular use of alcohol and/or tobacco. This high-risk cohort often does not seek regular medical care and access, therefore, to this population remains paramount. Significant capital investment will be required to seek out, incentivize and educate high-risk individuals to participate in an oral cancer screening.
With regards to community-based screening, one North American study has reported remarkable success with screenings. Poh et al9
reported a 98% participation in screenings held at a free dental clinic in an impoverished, high-risk area in downtown Vancouver. Incidentally, 2 out of 200 patients screened had biopsy-positive oral cancer and 8 patients had biopsy-proved precancer. Despite the small sample size, the number of subjects afflicted with oral cancer illustrates the extraordinary burden of disease within this community. The authors attributed much of the success of their program to their community-based approach. Participants were found to be reluctant to move outside very small boundaries; proximity to the screening site played a critical role in their participation.