Low positive predictive value of the oral brush biopsy in detecting dysplastic oral lesions

Authors

  • Vinodh Bhoopathi BDS, MPH,

    1. Division of Dental Public Health, Boston University School of Dental Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Sadru Kabani BDS, MS,

    1. Oral and Maxillofacial Pathology Department, Boston University School of Dental Medicine, Boston, Massachusetts
    Search for more papers by this author
  • Ana Karina Mascarenhas BDS, MPH, DrPh

    Corresponding author
    1. Division of Dental Public Health, Boston University School of Dental Medicine, Boston, Massachusetts
    • Boston University School of Dental Medicine, Division of Dental Public Health, 715 Albany Street 560, Boston, MA 02117===

    Search for more papers by this author
    • Fax: (617) 638-6381


Abstract

BACKGROUND:

The authors evaluated the effectiveness of the oral brush biopsy technique as a diagnostic tool in detecting dysplastic oral lesions.

METHODS:

In this cross-sectional study, pathologic reports (n = 152) from the scalpel biopsies (tissue samples) in patients who previously tested either “positive” (n = 3) or “atypical” (n = 149) for dysplasia by brush biopsy (OralCDx) were evaluated. Information on the age and sex of the patient, the site of the lesion, the brush biopsy results, and the histopathologic diagnosis of the scalpel biopsy was collected. The positive predictive values (PPVs) for “abnormal,” “atypical,” and “positive” brush biopsies were determined.

RESULTS:

Overall, the PPV of an abnormal brush biopsy was only 7.9% (95% confidence interval [CI], 4.2%-13.4%), and the PPV of an “atypical” brush biopsy was 7.4% (95% CI, 3.7%-12.8%). Of the 3 positive brush biopsies, only 1 was identified as dysplastic. The proportion of false-positive biopsy results was as high as 92.1% (95% CI, 86.6%-95.9%).

CONCLUSIONS:

The OralCDx technique overestimated dysplastic lesions and produced a high number of false-positive results. Cancer 2009. © 2009 American Cancer Society.

Oral cancer is the most frequent malignancy of the head and neck region1 and is a significant public health problem all over the world: Its prevalence is high in poor countries, and its survival is low in developed countries. Almost 90% of all oral cancers are oral squamous cell carcinomas (OSCC).2 In the United States alone, it is expected that 30,990 newly diagnosed cases of oral pharyngeal cancer would be diagnosed at the end of 2006, and approximately 7430 patients were expected to die of oral cancer.3 According to data from the Surveillance, Epidemiology, and End Results Program, the age-adjusted incidence rate of oral cancer, including both the oral cavity and the pharynx, between 1999 and 2003 was 10.5 cases per 100,000 population.4 The low 5-year cancer survival rates for oral cancer compared with other major cancers5 is attributed to late diagnosis and, thus, late treatment of the disease.6-8

Several adjunct diagnostic tests have been developed to improve the clinician's diagnosis of OSCC, such as toluidine staining, chemiluminescence, autofluoroscence, and oral brush biopsy. The OralCDx oral brush biopsy (OralCDx Laboratories, Inc., Suffern, NY) is 1 such diagnostic tool that, over the past several years, has been promoted vigorously as an effective test for detecting dysplastic lesions. It was designed to screen innocuous appearing oral epithelial abnormalities for dysplasia or cancer.9 However, there is controversy in the literature: Some researchers believe in the effectiveness of this technique and support its use as a diagnostic tool,9-11 whereas others have reservations.12-15 Therefore, the objective of the current study was to evaluate the effectiveness of the OralCDx brush biopsy by measuring the positive predictive value (PPV) of the test. We focused on lesions that were diagnosed as “abnormal” (atypical or positive) by the test.

MATERIALS AND METHODS

Data for this study were obtained from the Oral and Maxillofacial Pathology Department at the Boston University School of Dental Medicine. The cases (tissue samples) were referred by oral surgeons to the department for histopathologic evaluation. These cases were scalpel biopsies of lesions that tested “atypical” or “positive” by oral brush biopsy. Consecutive cases that were referred from January 2000 to June 2006 were acquired, resulting in a total of 152 cases. This cross-sectional study was approved by the Institutional Review Board at the Boston University Medical Campus.

Information, such as the age and sex of the patient, the site of the lesion, brush biopsy results, and histopathologic diagnosis of scalpel biopsy, was retrieved from pathologic reports of the scalpel biopsy results. The results of an oral brush biopsy are categorized by OralCDx Laboratories as “negative” when there is no evidence of precancerous or cancerous activity observed in the specimen submitted, “atypical” if the sample contains abnormal cells and may indicate a precancerous spot, and “positive” if there are dysplastic cells present that indicate an oral precancerous or cancerous lesion.16

Descriptive analyses were performed using means and proportions. The overall PPV and associated exact 95% confidence intervals (95% CIs) for an “abnormal” brush biopsy (positive or atypical) was calculated. The PPVs for an “atypical” OralCDx brush biopsy and a “positive” OralCDx brush biopsy also were calculated. In addition, the PPV for an atypical OralCDx brush biopsy was calculated stratified by age, sex, and site of lesion. Odds ratios and 95% CIs were calculated.

RESULTS

The mean age (±standard deviation) of the study population (n = 152) was 55 ± 14.4 years. About 54% of the study population was aged <55 years. The study sample comprised of more men (53%) than women. The brush biopsy results for the 152 cases included 3 positive biopsies and 149 atypical biopsies.

The histopathologic results of the scalpel biopsies reported in Table 1 were used to determine the true-positive rate, the false-positive rate, and the PPV. Of 152 cases, 12 were truly dysplastic (1 positive case and 11 atypical cases), suggesting a small number of true positives. Because only 12 of 152 brush-biopsied lesions were truly dysplastic, the probability that an abnormal brush biopsy truly would detect a dysplastic lesion (PPV) was 7.9% (95% CI, 4.2%-13.4%). Table 1 also reports the scalpel biopsy diagnoses for the nondysplastic lesions, the majority of which were reactive/inflammatory lesions (80.3%).

Table 1. Histopathologic Results of Scalpel Biopsy (N=152)
Scalpel Biopsy ResultsNo. of Patients%
Dysplastic lesions127.9
 Premalignant95.9
 Malignant32
Nondysplastic lesions14092.1
 Reactive9864.5
 Inflammatory2415.8
 Immune-mediated117.2
 Developmental42.6
 Pigmented21.3
 Normal10.6

Table 2 reports the lesions by site and the distribution of lesions in high-risk sites. Almost 60% (n = 91) of the lesions that were diagnosed as abnormal by the brush biopsy were from areas that have a low risk of cancer. The remaining abnormal lesions (40%) were identified in high-risk areas, such as the soft palate, the junction of the hard and soft palates, the ventral part of tongue, the lateral part of tongue, and the floor of the mouth.

Table 2. Characteristics and Distribution of Lesion Site (N=152)
Lesion siteNo. of Patients%
Type of lesion site  
 Low-risk area9159.9
 High-risk area6140.1
Distribution of high-risk lesion sites  
 Lateral tongue2717.8
 Floor of mouth127.9
 Ventral tongue117.2
 Soft palate74.6
 Junction of soft and hard palate42.6

The overall PPV of an “atypical” brush biopsy was 7.4% (95% CI, 3.7%-12.8%), because only 11 of 149 brush-biopsied lesions truly were dysplastic. The PPV of an atypical brush biopsy stratified by age, sex, and lesion site is presented in Table 3. The PPV was higher among patients aged ≥55 years (7.8%; 95% CI, 2.5%-17.2%) and among women (PPV, 8.6%; 95% CI, 3.2%-17.7%) compared with their counterparts. The probability that an atypical brush biopsy truly would detect a dysplastic lesion in a high-risk area was 18.3% (95% CI, 9.5%-30.4%). There were no dysplastic lesions identified in the low-risk areas (PPV, 0%).

Table 3. Positive Predictive Value (PPV) of Atypical Brush Biopsy Stratified by Age, Gender, and Site of Lesion
Parameter(N=149)True PositivePPV, %95% CI
  1. CI indicates confidence interval.

Age, y    
 ≥556457.82.5-17.2
 <557456.82.2-5.1
Sex    
 Women7068.63.2-17.7
 Men7956.32-14.1
Lesion site    
 High-risk site601118.3%9.5-30.4
 Low-risk site8900

Of the 3 positive brush biopsy cases, only 1 truly was positive for dysplastic changes; therefore, the PPV was 33.3% (95% CI, 0.08%-90.6%), which was much higher than the PPV of an atypical brush biopsy (7.4%). Although the difference was not statistically significant, a “positive” brush biopsy was 6.3 times more likely to detect a dysplastic lesion than an “atypical” brush biopsy.

DISCUSSION

The oral brush biopsy has been on the market for >5 years, a large number of dental professionals use it, and some dental schools teach its use. There is an ongoing debate in the literature regarding its effectiveness as a diagnostic tool. In a recent systematic review, it was concluded that this technique is neither beneficial nor harmful.17

The overall PPV for an abnormal brush biopsy in our study (both “positive” cases and “atypical” cases) was 7.9%. A PPV of 7.9% for a test to screen lesions that easily are visible to the naked eye raises the question of its true effectiveness. The 95% CI ranged from 4.2% to 13.4% and could be interpreted as the worst-case (4.2%) and best-case (13.4%) scenarios. A low PPV (7.9%) suggests a very high number of false-positive results. False-positive reports unnecessarily evoke anxiety and fear among those patients who truly do not have the disease. The management of a patient with a positive or atypical brush biopsy is referral for a scalpel biopsy, which can lead to further patient stress and anxiety and also can result in complications related to the scalpel biopsy and, ultimately, higher healthcare costs.

OralCDx Laboratories has reported that 30% to 44% of all brush-biopsied “atypical” lesions would be dysplastic.16 However, the PPV for an atypical brush biopsy in our study was only 7.4%, which is substantially lower than that reported above by OralCDx Laboratories. When an “atypical” brush biopsy truly detects dysplasia in only 1 of 13 patients, as suggested by our study results, it would seem legitimate to question the value of its use and to recommend the re-evaluation of this diagnostic test. In addition, the costs of the procedure need to be taken into account. Currently, there are costs associated with doing the brush biopsy procedure and laboratory costs for processing the brush biopsy. When an “atypical” or “positive” result occurs, there are further costs of the scalpel biopsy and costs associated with histopathologic evaluation. With the concern of rising healthcare costs, the cost-effectiveness of the OralCDx procedure should be evaluated.

In the current study, of 3 “positive” brush biopsy cases, only 1 lesion truly was dysplastic, suggesting a PPV of 33.3%. It is noteworthy that only 3 cases of a “positive” brush biopsy were reported (2%) from the data collected between January 2000 and June 2006. A PPV of just 33.3% for a positive brush biopsy that is supposed to detect dysplastic characteristics clearly, according to the criteria16 (changes indicating definitive precancerous or cancerous characteristics), is unsettling and, again, is in sharp contrast to the data presented by OralCDx Laboratories. A useful comparison is the PPV of Papanicolaou smears, which reportedly is 3.4% for mild-to-moderate dysplasia, 90.6% for a diagnosis of severe dysplasia/carcinoma in situ, 94.5% for a diagnosis of carcinoma in situ or microinvasive carcinoma, and 95.5% for an initial diagnosis of invasive carcinoma.18 Conceptually, the PPV of brush biopsies also could be compared with the PPV of other screening tests that have low PPV, such as that for ovarian cancer screening, in which the lowest acceptable PPV for a screening test is 10%,19 which still is higher than the PPV in the current study. It also should be noted again that ovarian cancer lesions are not visible to the naked eye compared with oral dysplastic lesions, for which the OralCDx brush biopsy is used as a screening tool.

In this study, all dysplastic oral lesions (100%) were located in high-risk areas. Because the PPV of an atypical brush biopsy in a high-risk area was higher at 18.3%, and because there were no dysplastic lesions identified in a low-risk area,20 the value of using this technique in a low-risk area is unclear. The low PPV for a high-risk site suggests that there is a high chance that atypical brush biopsy also overestimated dysplastic lesions even in high-risk areas, again producing a high number of false-positive results. Because scalpel biopsy is the only recommended procedure for diagnosing a lesion in a high- or low-risk area, and because brush biopsy overestimates dysplastic lesions even in high-risk areas, its use for correctly detecting a dysplastic lesion even in a high-risk area, again, is very unclear. Therefore, we recommend that, rather than undergoing a brush biopsy, the patient should be referred to the appropriate dental or medical professionals for further evaluation and possibly a scalpel biopsy. This also would reduce the time between diagnosis and definitive treatment.12

Limitations of our study include the sample size, referral bias, and that we were able to calculate only the PPVs, because our study participants included only those who had brush biopsies that tested either “atypical” or “positive” and not those who had negative results. Thus, we were unable to calculate the sensitivity, specificity, and negative predictive value for an abnormal brush biopsy technique, calculations that would have enhanced this study's results. However, our data represent effectiveness in a “real world” setting compared with the studies reported previously.9, 10 In addition, we recommend that an independent, multisite study that includes both the “positive” and “negative” lesions from OralCDx brush biopsies is needed to determine the sensitivity, specificity, PPV, and negative predictive values, along with a cost-effectiveness analysis, before any type of universal recommendations can be made regarding the use of brush biopsies.

In conclusion, the current results indicated that the overall PPV for an abnormal brush biopsy was 7.9%. The PPV for an atypical brush biopsy technique was only 7.4% compared with 33% to 40% of the PPV that the OralCDx Laboratories claim. The PPV for a positive brush biopsy was only 33.3%. The effectiveness of the OralCDx brush biopsy was questionable in high-risk sites, let alone low-risk sites. Overall, the OralCDx technique overestimated dysplastic lesions and produced a high number of false-positive results.

Conflict of Interest Disclosures

The authors made no disclosures.

Ancillary