ORIGINAL ARTICLE
Accuracy of Screening for Potentially Malignant Disorders of the Oral Mucosa by Dentists in Primary Care Luis Cláudio Sartoria/Paulo Frazãob Purpose: To examine the accuracy of a screening programme for potentially malignant disorders of the oral mucosa by visual inspection in primary health care. Materials and Methods: The study was based on secondary data from the Primary Care Information System maintained by seven units of family health in São Paulo City managed by a non-governmental agency. The reference population was composed of 15,072 residents 50 years old or more of both genders. The study population comprised 2,980 individuals. During screening in community settings, the oral mucosa was examined by trained dentists and distributed into two categories: (a) screen negative (b) screen positive. All participants underwent comprehensive clinical exams by a general dental practitioner supervised by a specialist. Individual records were grouped in a working dataset. Point and 95% confidence interval estimates were calculated regarding measures of sensitivity (Se), specificity (Sp) and positive and negative predictive values (PPV and NPV, respectively). Results: 18.0% of the population was considered screen positive. A total of 133 lesions (4.5%) were identified and 8 cases of oral cancer were confirmed, which corresponded to a prevalence rate of 27 cases in 10,000 people, a much higher rate than expected. The measures found were Se: 91.7% (85.3–95.6), Sp: 85.4% (84.1–86.7), PPV: 22.7% (19.3–26.5), NPV: 99.5% (99.2–99.8). The visual screen presented high accuracy. Conclusion: The test presented high sensibility and specificity values. From a public health point of view, the high accuracy levels showed the importance of oral health teams on family health strategy for more comprehensive primary care. Targeting risk groups and delegating the screening to community health agents may improve PPV and coverage. Key words: control, detection, oral neoplasms, screening accuracy Oral Health Prev Dent 2012; 10: 53-58
C
ancer ranks second among disease-caused deaths in Brazil. About 50% of the deaths from oral cancer occur in the first year after diagnosis in American and Caribbean regions. Another 10% to 20% die within five years of the diagnosis. According to the National Cancer Institute, the incidence in Brazil is nearly 14,120 new cases per year with over 3,000 deaths occurring annually. Oral cancer occupies fifth place among the highest incidence
a
Assistant Manager for Primary Health Care, Brazilian Ministry of Health, São Paulo, Brazil.
b
Professor, Public Health School, University of São Paulo, Brazil.
Correspondence: Paulo Frazão, Av Dr Arnaldo 715 – São Paulo, CEP 01246-905, Brazil. Tel: +55-11-3061-7957, Fax: +55-11-3061-7835. Email:
[email protected]
Vol 10, No 1, 2012
Submitted for publication: 17.02.11; accepted for publication: 31.05.11
cancers in males and the seventh place in females. In Brazil, the mortality rate for oral cancer remained at around 2 deaths per 100,000 men and 0.5 deaths per 100,000 women from 1979 through 2002 (Boing et al, 2009). Smokers and alcohol users who are male are most vulnerable. Oral cancer is lethal more often than it needs to be, because people tend to ignore symptoms, so it is typically caught in late stages. Doctors and dentists also bear great responsibility for the diagnostic delay. For these reasons, oral cancer is a public health problem in Brazil (Narvai and Frazão, 2008). However, some authors consider that the early detection through regular screening can be an effective control strategy for oral cancer (Downer et al, 1995; Jullien et al, 1995; Nagao and Warnakulasuriya, 2003), although evidence of
53
Sartori/Frazão
screening effectiveness from randomised controlled trials is limited (Kujan et al, 2006). If oral lesions are readily detectable, survival can be improved after treatment during early stages. Visual inspection with or without toluidine blue dye as an adjunct, oral self-examination and cytological analysis are the main investigative methods for screening oral pre-malignant and malignant lesions (Van der Waal, 2009). In a cluster-randomised controlled trial conducted in Kerala, India, an area with a very high prevalence of oral cancer, Sankaranarayanan et al (2005) found a significant reduction in oral cancer mortality over a 9-year period among high-risk individuals. Those authors concluded that visual inspection of the oral cavity can be a simple, acceptable and accurate screening test for oral neoplasia. Screening programmes can be conducted by several strategies, such as invitational or opportunistic (in regular visits), in primary care settings. The first can be carried out by medical and dental general personnel or by community health agents. Both can be directed to high-risk groups or to the population as a whole (Metha et al, 1986; Warnakulasuriya and Pindborg 1990; Downer et al, 1995; Ikeda et al, 1995; Sankaranarayanan, 1997). Different outcomes can be assessed in the screening strategy evaluation from the lesion diagnosis stage results up to morbidity, mortality and survival rates. According to the terminology proposed by Van der Waal (2009), oral pre-malignant and malignant lesions will be named potentially malignant disorders of the oral mucosa in this paper. The aim of this study was to determine the accuracy of screening for potentially malignant disorders of the oral mucosa by visual inspection carried out by general dental practitioners of primary health care in a population 50 or more years of age.
MATERIALS AND METHODS The study was based on secondary data from the Primary Care Information System maintained by seven units of family health in São Paulo City. Access to data was authorised by Casa de Saúde Santa Marcelina, a non-governmental organisation in charge of the management of the units on behalf of the Health Department of São Paulo City, responsible for conducting the municipal health system as a whole. The research protocol was approved by Ethics Committee of the Public Health School, University of São Paulo (No. 729/02).
54
A screening programme for oral cancer was carried out between 1999 and 2001 in all primary care settings managed by the non-governmental organisation. Although the incidence has been increasing in the younger age groups, the programme targeted individuals over 50 years old, as the age above which oral cancer has been more frequent. The screening programme was a component of a larger endeavor aimed to motivate people to seek primary health care for an oral consultation. The selected units were required to present an oral health team and administrative records on the screening programme for oral cancer covering all areas. The reference population was composed of 15,072 residents 50 years old or more of both genders registered in defined areas of study units. All those who attended the screening appointment and underwent clinical oral examination within 90 days were considered eligible for the study. The study population comprised 2,980 individuals. The data were collected by different dentists during screenings in community settings and clinical oral examinations at the dental office within family health units. Fifteen general dental practitioners received 16 h of intensive formal training on the visual inspection technique with the aid of a small wooden paddle. It included a 4-h theoretical phase in which the characteristics of normal oral mucosa and the main clinical aspects of abnormal oral mucosa were addressed. This was followed by two 4-h sessions of clinical training involving up to five general dental practitioners at a time. The first session consisted of exercises on visual inspection of the oral mucosa, and the second session comprised activities for collecting biopsies from the oral cavity. The individuals were examined in community settings under natural lighting in the screening programme. Test reproducibility for controlling intraand inter-examiner error was not feasible. The standardised set of operational diagnostic criteria for conducting screening programmes was quite simple in order to facilitate its application by community health agents. The observations were categorised into screen positive and screen negative: apparently healthy oral mucosa was considered screen negative; oral mucosa presenting a lesion that had persisted over 14 days – regardless of its appearance – was considered screen positive. During the three months subsequent to that step, all participants underwent comprehensive clinical examination in the dental office located at the primary care setting. The general dental practi-
Oral Health & Preventive Dentistry
Sartori/Frazão
tioner did not know the result of the screening test and could consult an oral diagnosis specialist for discussion of the cases. Lesions persisting over 14 days with no known cause but presenting clinical aspects such as (1) ulcers with indurated edges, (2) white lesions (leukoplakias), (3) red lesions, (4) granulated, verrucous or ulcerated vegetative lesions (papules or nodules) and (5) pigmented and keratogenous lesions were considered potentially malignant disorders of the oral mucosa. Biopsy samples were obtained from these individuals and one pathology laboratory performed all histological analyses. The reference examination for the final diagnosis of cancer was supported by a comprehensive clinical examination including histology. In addition, lesions persisting over 14 days and presenting clinical aspects such as hyperplasia from chronic trauma (ill-fitting dentures) but no response after elimination of the possible cause were also considered potentially malignant disorders. The gold standard was a comprehensive clinical examination by a general dental practitioner supported by a specialist and histological analysis when required. A working dataset containing information about observed conditions in screening and administrative records related to dental care was created. Three groups of variables were linked: (a) cadastral data – unit code, screening year, acronym of subject, date of birth, gender, area and family code; (b) screening positive or negative tests; (c) dental care data – lesion classification, biopsy requirement and diagnosis. A smaller number of individuals undergoing biopsy were excluded because histological findings had still not been recorded. Search procedures for possible false negative cases were adopted using the Hospital Information System. The hospital ad-
mission data with the diagnosis of oral cancer according to the International Statistical Classification of Diseases and Related Health Problems 10th Revision (codes C00 to C14) and the Hospital Information System for procedures related to diagnosis for malignant tumors (codes SIH 43000002, 72500000 and 85500712) as well as characteristics such as gender, age and household were scanned. A contingency table was elaborated to calculate sensibility and specificity measures and positive/ negative predictive values (Fletcher et al, 1996).
RESULTS The study population according to gender is presented in Table 1. Among 15,072 residents 50 years of age or above, 2,908 (19.8%) individuals were screened and 65.4% were women. One out of five residents accepted invitational screening and also underwent clinical oral examination. All addresses were checked to search for false negative cases. No resident admitted to hospitals with a diagnosis of oral cancer was found in study area. Five hundred thirty-seven participants (18.0%) were considered screen positive (Table 2). In the clinical oral examinations, 133 lesions were diagnosed (4.5%), of which 99 (74.4%) cases required laboratory exams. Out of 133 positive diagnoses, 22 were hyperplasia from chronic trauma (ill-fitting dentures), of which 15 required biopsies. Oral cancer was found in a total of 8 cases, comprising 0.27% or 27 per 10,000 residents. Among 1,032 male residents, 158 (15.3%) were considered screen positive. By clinical oral examination, 42 lesions (4.1%) were diagnosed and among these 29 (2.8%) were biopsied. Three cases of oral
Table 1 Number and percentage of referenced and screened residents of ≥50 years old according to gender Reference population
Screened residents
Gender
N
%
N
%
Male
6,648
44.1
1,032
34.6
Female
8,424
55.9
1,948
65.4
Total
15,072
100.0
2,980
100.0
Source: Primary Care Information System. Family Health Programme, Casa Saúde Santa Marcelina, Sao Paulo, Brazil, 2002
Vol 10, No 1, 2012
55
Sartori/Frazão Table 2 Number and percentage of study population according to screening test and clinical oral examination for potentially malignant disorders of the oral mucosa Test
Clinical oral examination diagnosis Positive
Negative
Total
N
%
N
%
N
%
Positive
122
22.7
415
77.3
537
100.0
Negative
11
0.5
2432
99.5
2443
100.0
TOTAL
133
4.5
2847
95.5
2980
100.0
Se = 91.7% (95% CI 85.3–95.6) Sp = 85.4% (95% CI 84.1–86.7)
PPV = 22.7% (95% CI 19.3–26.5) NPV = 99.5% (95% CI 99.2–99.8)
Sensibility (Se), specificity (Sp), positive and negative predictive values. Point and 95% confidence interval estimates. Family Health Programme, Casa Saúde Santa Marcelina, Sao Paulo, Brazil, 2002.
cancer were confirmed – a rate of 29 per 10,000 men. Very similar values were obtained for women: 379 (19.5%) were considered screen positive, 91 lesions (4.7%) were diagnosed by oral examination, 71 (3.6%) were biopsied and 5 cases were confirmed – a rate of 26 per 10,000 women. Out of 9 confirmed cases, seven (87.5%) were squamous cell carcinomas. The results thus did not differ significantly by gender. According Table 2, the majority of the screen negative results were confirmed as negative diagnoses (2,432 true negative cases among the 2,443 screened) after clinical oral examination. A positive diagnosis was confirmed in 122 (22.7%) cases among 537 screen positives (positive predictive value). Out of 133 cases with positive diagnoses, only 11 were not identified in the screening programme. High estimates of sensibility and specificity were found.
DISCUSSION Screening programme data, collected by general dental practitioners using visual inspection in seven primary health units, were assessed in this article. The participation of one in five residents may be considered high for a diagnostic accuracy study. The data from comprehensive clinical exams were provided by general dental practitioners that were blinded to the screening results. The examiners were supported by a specialist and histological reports when required. The records allowed producing an analysis similar to a prospective community trial on the screening accuracy for potentially malig-
56
nant disorders of the oral mucosa. All those that underwent clinical oral examination after screening were considered eligible. The most commonly used cancer screening programme, i.e. Pap smears for the cervical region, is also not supported by randomised trials but by consistent evidence from other types of study designs, the results of which show that lesions are detected in early stage with better clinical outcomes. The test presented high sensibility and specificity values similar to those mentioned in the literature (Warnakulasuriya and Pindborg 1990; Downer et al, 1995; Ikeda et al, 1995; Mathew et al, 1997) using dentists or community agents as screeners. The sole exception was the sensibility result (59%) of Metha et al (1986), in which the first examiner was a community agent. Unequal proportions of men and women participated in this study. An increase in the male accuracy values for a balanced situation could be expected. The positive predictive value (22.7%) found could be considered low from a clinical point of view, but acceptable if the measurements of rare events are taken into account, as pointed by Fletcher et al (1996). Factors related to classification criteria, duration of the lesions and accuracy of clinical oral examination (among others) may result in a higher number of false positive cases, explaining the values found here. The negative predictive value (99.5%) was similar to other research in which the screening was carried out by community agents (Metha et al 1986; Warnakulasuriya and Pindborg 1990; Mathew et al, 1997) and higher in comparison to that found by Ikeda et al (1995) using dentists as screeners. The
Oral Health & Preventive Dentistry
Sartori/Frazão
observed negative predictive value showed that a negative test almost completely excluded the possibility of an oral lesion diagnosis. A limitation of the data was a lack of information about test reproducibility, a limiting factor in other oral screening studies (Speight et al, 2006). As duplicate exams on other days were not feasible and the standardised criteria were quite simple, it is plausible to assume that the data were sound. There was no information about compliance rate specifically related to the invitation for the programme. This makes it impossible to present a patient flow chart and makes the comparison with other studies difficult, where compliance rates for the invitational screening and for the clinical oral examination were measured. In some studies, a significant number of screened-positive participants did not attend the examination to confirm the diagnosis. Out of 1,220 participants considered positive by community agents in the study by Warnakulasuriya and Pindborg (1990), 660 attended the examination. Among the 32 subjects considered positive in the study by Ikeda et al (1995), 25 were examined. The proportion of pre-malignant lesions found in 2,980 screened participants was 1.68% (50 cases). Metha et al (1986), using community agents as examiners, assessed tobacco users in an Indian district and observed similar values. Other studies found higher values (Warnakulasuriya and Pindborg 1990; Downer et al, 1995; Ikeda et al, 1995; Mathew et al, 1997). Higher-risk age groups could explain these differences. The study by Mathew et al (1997) took risk factors into account for selecting the population to be screened. With the exception of the studies carried out by Ikeda et al (1995) in Japan and Downer et al (1995) in England, all of the others referred to high-incidence regions for oral cancer (India and Sri Lanka). It is possible that the proportion found here (1.68%) could overestimate the number of potentially malignant disorders due to hyperplasia associated with ill-fitting dentures as a criterion of inclusion. However, as one of the first studies on a Brazilian population, it seemed better to adopt a conservative criterion. Moreover, out of 22 participants presenting hyperplasia from chronic trauma (ill-fitting dentures), 15 had no response after elimination of supposed cause and required biopsies. The majority of the data from the scientific literature about oral lesions are restricted to lesions determined in small samples, which does not permit a comprehensive view of the occurrence of these
Vol 10, No 1, 2012
lesions in representative samples of the population. Eight cases of oral cancer among those screened were confirmed by histological findings – 5 in females and 3 in males – with all of them diagnosed at stage I. Seven of them were squamous cell carcinomas, the most common kind. The gross incidence rate of oral cancer per 100,000 residents was 5.1 among women and 16.0 among men in São Paulo City in 1993. A decade later, the incidence values remained similar: 7.4 and 17.9, respectively, according to the National Cancer Institute. These numbers increase dramatically with age. The rate for the 35- to 64-year-old population was 7.0 and 34.0 per 100,000 residents for women and men, resp. For the population 65 or more years of age, recognised as at high risk for the disease, the rate was 36.8 per 100,000 residents for women and 97.3 per 100,000 among men in the same year (Antunes et al, 2001). The diagnosis of 8 cases in 2,980 screened subjects could be interpreted as a high value, but it is important to consider that this measure represents only a proportion related to the positive cases for oral cancer and are not valid as a population-based incidence or prevalence estimate. Although the positive predictive value can be low for rare events in high-specificity tests (Fletcher et al, 1996), it must be emphasised that the values found could be higher if the screening programme were directed at individuals, families and high-risk groups. Moreover, the need of training community health workers should be assumed. Studies in which screening was performed by this kind of worker presented high sensibility and specificity (Metha et al, 1986; Warnakulasuriya and Pindborg 1990; Mathew et al, 1997). In a visual-inspection randomised trial carried out by community agents, a non-significant reduction of 21% in deaths was observed among the group receiving the screening. Among tobacco and alcohol users, the difference was 34%, which was statistically significant (Sankaranarayanan et al, 2005). The training of community health agents who can assist people in oral self-examination is feasible as part of the Family Health Strategy, where many activities are supported by the health agents and are targeted at a catchment area population. This proposal could make increasing the coverage viable. It could also provide positive impulses for the cost-effectiveness of such activities, in addition to contributing to the prevention and control of the disease.
57
Sartori/Frazão
An extensive review of the available data from the United Kingdom concluded that while there was no strong direct evidence of benefit, high-risk opportunistic screening by general dental practitioners might be cost-effective (Speight et al, 2006). The simplicity of the procedure and the minimal risks involved compared to the potential benefits would be enough to justify it. Nevertheless, further studies must be encouraged to analyse the incidence of oral cancer in the elderly and treatment costs, as well as other aspects of screening programme evaluation such as the time interval between activities in the same population. About 234,000 community health agents worked in primary health care in Brazil in 2009. The expansion of primary health care under the principles of family health strategy is essential for improving the National Health System toward universal, equitable and comprehensive care. Indicators related to several health programmes have shown improvements (Macinko et al, 2007; Barreto and Aquino 2009). The findings of this study also have implications for continuing this expansion.
7.
8.
9.
10.
11.
12. 13. 14.
15.
CONCLUSION 16.
In conclusion, the test presented high sensibility and specificity values. From a public health point of view, the high accuracy levels emphasised the importance of oral health teams in family health strategy to facilitate more comprehensive primary care.
17.
18.
REFERENCES
dents of Tokoname City, Japan. Community Dent Health 1995;12:83–88. Jullien JA, Downer MC, Zakrzewska JM, Speight PM. Evaluation of a screening test for the early detection of oral cancer and precancer. Community Dent Health 1995;12:3–7. Kujan O, Glenny AM, Oliver RJ, Thakker N, Sloan P. Screening programmes for the early detection and prevention of oral cancer. Cochrane Database Syst Rev 2006;19;3:CD004150. DOI: 10.1002/14651858.CD004150.pub3 Macinko J, Souza MFM, Guanais FC, Simões CCS. Going to scale with community-based primary care: An analysis of the family health program and infant mortality in Brazil, 1999–2004. Soc Sci Med 2007;65:2070–2080. Mathew B, Sankaranarayanan R, Sunilkumar KB, Kuruvila B, Pisani P, Nair MK. Reproducibility and validity of oral visual inspection by trained health workers in the detection of oral precancer and cancer. Br J Cancer 1997;76:390–394. Metha FS, Gupta PC, Bhonsle RB, Murti PR, Daftary DK, Pindborg JJ. Detection of oral cancer using basic health workers in an area of high oral cancer incidence in India. Cancer Detect Prev 1986;9:219–225. Nagao T, Warnakulasuriya KAAS. Annual screening for oral cancer detection. Cancer Detect Prev 2003;27:333–337. Narvai PC, Frazão P. Saúde bucal no Brasil: muito além do céu da boca. Rio de Janeiro: Ed. Fiocruz; 2008. Sankaranarayanan R, Ramadas K, Thomas G, Muwonge R, Thara S, Mathew B, Rajan B. Effect of screening on oral cancer mortality in Kerala, India: a cluster-randomised controlled trial. Lancet 2005;365:1927–1933. Sankaranarayanan R. Health care auxiliaries in the detection and prevention of oral cancer. Oral Oncol 1997;33:149– 154. Speight PM, Palmer S, Moles DR, Downer MC, Smith DH, Henriksson M, Augustovski F. The cost-effectiveness of screening for oral cancer in primary care. Health Technol Assess 2006;10:1–144. Van der Waal I. Potentially malignant disorders of the oral and oropharyngeal mucosa; terminology, classification and present concepts of management. Oral Oncology 2009;45:317–323. Warnakulasuriya S, Pindborg JJ. Reliability of oral precancer screening by primary health care workers in Sri Lanka. Community Dent Health 1990;7:73–79.
1. Antunes JLF, Biazevic MGH, Araújo ME, Tomita NE, Chinellato LEM, Narvai PC. Trends and spatial distribution of oral cancer mortality in São Paulo, Brazil, 1980–1998. Oral Oncol 2001;37:345–350. 2. Barreto ML, Aquino R. Recent positive developments in the Brazilian Health System [invited Editor’s Choice]. Am J Public Health 2009;99:8. 3. Boing AF, Peres MA, Antunes JLF. Mortality from oral and pharyngeal cancer in Brazil: trends and regional patterns, 1979–2002. Pan Am J Public Health 2006;20:1–8. 4. Downer MC, Evans AW, Hughes HC, Julien JA, Speight PM, Zakrzewska JM. Evaluation of screening for oral cancer and precancer in a company headquarters. Community Dent Oral Epidemiol 1995;23:84–88. 5. Fletcher RH, Fletcher SW, Wagner EH. Epidemiologia clínica: elementos essenciais. 3. ed. Porto Alegre: Artmed, 1996. 6. Ikeda N, Downer MC, Ozowa Y, Inoue C, Mizuno T, Kawai T. Characteristics of participants and non-participants in annual mass screening for oral cancer in 60-year-old resi-
58
Oral Health & Preventive Dentistry
