Abstract. cytology and biopsies, patient recall, treatment, quality management, and overall high cost. To overcome

Transcription

1 Int J Gynecol Cancer 2006, 16, Optical coherence tomography as a diagnostic aid to visual inspection and colposcopy for preinvasive and invasive cancer of the uterine cervix P.F. ESCOBAR*y, L. ROJAS-ESPAILLAT*y, S. TISCIz, C. ENERSONy, J. BRAINARD, J. SMITHz, N.J. TRESSERk, F.I. FELDCHTEINk, L.B. ROJAS{ & J.L. BELINSON* *Department of Gynecology and Obstetrics, Section of Gynecologic Oncology, The Cleveland Clinic Foundation, Cleveland, Ohio; ypreventive Oncology International, Morelia, Mexico; zuniversity of North Carolina, Chapel Hill, North Carolina; Department of Pathology, Section of Gynecologic Pathology, The Cleveland Clinic Foundation, Cleveland, Ohio; khospital Maternidad Nuestra Senora de la Altagracia, Santo Domingo, Dominican Republic; and {Imalux Corporation, Cleveland, Ohio Abstract. Escobar PF, Rojas-Espaillat L, Tisci S, Enerson C, Brainard J, Smith J, Tresser NJ, Feldchtein FI, Rojas LB, Belinson JL. Optical coherence tomography as a diagnostic aid to visual inspection and colposcopy for preinvasive and invasive cancer of the uterine cervix. Int J Gynecol Cancer 2006;16: The purpose of this study was to determine the sensitivity and specificity of optical coherence tomography (OCT) under two well-defined clinical settings. First, as an aid to cervical cancer screening, using visual inspection with acetic acid (VIA) in low-resource settings, and the second, as an adjunct to the traditional management of abnormal cervical cytology with colposcopy and biopsy. Patients referred for colposcopy with atypical squamous cells of undetermined significance were accrued for the study. Each subject underwent VIA and colposcopy. OCT was performed in all VIA- and colposcopy-positive areas and at the squamocolumnar junction in all four quadrants. The sensitivity of VIA for cervical intraepithelial neoplasia 2 was 76% (95% CI 58 88). When OCT was applied to VIA as a secondary screen, the specificity improved from 34% (95% CI 27 41) to 61% (95% CI 60 74). With liberal diagnostic criteria for the majority of the colposcopy examinations, OCT showed an even greater relative improvement in specificity. OCT proved to be a fair diagnostic modality (receiver operating characteristic curve 0.73) adjunctive to VIA and colposcopy. On the basis of the above findings, we believe that this technology could potentially show greatest utility in the management of cervical dysplasia in low-resource settings where a single episode of care is most desirable KEYWORDS: cervical cancer, optical coherence tomography. The incidence of cervical cancer varies widely between and within regions throughout the world (1,2).Thisis attributable in large part to the variable access to cytologic screening programs to detect and treat preinvasive disease of the cervix (3). The World Health Organization estimates that 288,000 deaths per year Address correspondence and reprint requests to: Jerome L. Belinson, MD, Department of Gynecology and Obstetrics, Section Gynecologic Oncology, The Cleveland Clinic Foundation, 9500 Euclid Avenue, Desk A-81, Cleveland, OH 44118, USA. belinsj@ ccf.org doi: /j x are caused by cervical cancer, making it the second most common cause of female cancer deaths worldwide. According to the World Health Organization, approximately 510,000 cervical cancer cases are reported each year. Asia, Latin America, and Africa account for 76% of all cases (4). The traditional cytology-based screening programs that have been successful in reducing deaths from cervical cancer by 70% require a complex infrastructure that is difficult to maintain in most developing countries (5). This includes obtaining and processing the cytology and biopsies, patient recall, treatment, quality management, and overall high cost. To overcome # 2006, Copyright the Authors Journal compilation # 2006, IGCS

2 1816 P.F. Escobar et al. these obstacles, various biophotonic methods have been investigated that allow rapid and noninvasive evaluation of the cervix. One of the methods that has potential for developing world application is optical coherence tomography (OCT). OCT is an infrared light imaging technology, based on low-coherence interferometry. It is a noninvasive method that produces a real-time, high spatial resolution microstructural image of living tissues (6 8). OCT s efficacy in detecting early neoplasia in various types of mucosa has been demonstrated both in animal models and in humans (9,10). Few reports on OCT in the gynecological setting have been published to date. A review of early results on the use of OCT in gynecology can be found in Shakhova et al. (11). OCT imaging of neoplastic and normal gynecological tissues was studied in vitro by Pitris et al. (12), and later in vivo by Escobar et al. (13) who reported that normal ectocervical tissue could be reproducibly recognized. This initial experience with OCT suggested that the use of OCT as an optical biopsy in cervical cancer was feasible. It was noted that a colposcopist should be able to rapidly examine the cervix and identify immature metaplasia, a common cause of a false-positive colposcopy. It follows that using OCT to select biopsy sites could markedly reduce the number, and consequently, the cost and morbidity of biopsies. The current study, a continuation of our previous study, was designed to determine the sensitivity and specificity of OCT under two well-defined clinical settings. First, as an aid to cervical cancer screening using unmagnified visual inspection with acetic acid (VIA) in low-resource settings, to accurately characterize normal squamous epithelium of the cervix and cervical intraepithelial neoplasia (CIN). This could allow point of service diagnosis and treatment without biopsy confirmation. The second setting would be applying OCT to traditional cervical cancer diagnostic programs along with colposcopy and biopsy. b) patients had to sign an informed consent document stating that they understood the investigational nature of the proposed study. Exclusion criteria were: a) if subject was a prisoner; b) was pregnant; c) had a prior hysterectomy; or d) had prior treatment for preinvasive or invasive cervical cancer. When the patients arrived for their examination, the nature of the clinical study was explained in detail and informed consent was obtained for all participants. Demographic and epidemiologic information was obtained from each participant. The clinical protocol took place in one examination room where each woman underwent all tests and procedures that were performed by gynecological oncologists or gynecological oncology fellows. The protocol for each patient included five clearly defined steps (Fig. 1): Step 1 Unmagnified VIA. Acetic acid (5%) was applied to the cervix and after 1 min VIA was performed. The visual inspection diagnosis was recorded by quadrant. Quadrants defined as abnormal had dense white lesions with sharp borders located in the transformation zone. A diagnosis of abnormal/cancer was made if there was a visible friable mass with an irregular surface or an ulcer. Step 2 A digital photograph was taken of the cervix following visual inspection. Materials and methods The human subject review boards of both The Cleveland Clinic Foundation and The Hospital Maternidad Nuestra Senora de la Altagracia approved the study. Patients who were diagnosed with an abnormal Papanicolaou test (Pap smear atypical squamous cells of undetermined significance) or with suspicious lesions of the uterine cervix were eligible for the study. To be qualified for entry into this study, patients had to meet all the following criteria: a) 18 years of age and older; Figure 1. Study protocal flowchart.

3 Sensitivity and specificity of OCT under two well-defined clinical settings 1817 Step 3 Colposcopy. Acetic acid (5%) was reapplied to the cervix, colposcopy was performed, and observations were recorded by quadrant. If a quadrant was abnormal, the data reflected the most severe abnormality within that quadrant. Step 4 OCT. The OCT device (Imalux Corporation, Cleveland, OH), operating at 980 nm center wavelength and having lateral resolution 25 lm and in-depth resolution of 11 lm in tissue, was used to take readings from 2, 4, 8, and 10 o clock at the squamous columnar junction (SCJ) as well as any additional abnormal areas found by VIA, colposcopy, or both VIA and colposcopy. Step 5 Biopsies. Biopsies were taken in each quadrant and appropriately labeled from all VIA, colposcopic, and VIA- and colposcopic-positive areas. In addition, a biopsy was taken from each quadrant at 2, 4, 8, or 10 o clock at the SCJ. Thus, more than one biopsy per quadrant was taken depending on the VIA and colposcopic impressions. All biopsies were performed with a bronchoscopy biopsy instrument that has a 2 mm jaws and is virtually painless for most patients. An endocervical curettage was also done on every patient. The histologic slides of all biopsy and endocervical curettage specimens were read by author J.B. A team of gynecological pathologists at The Cleveland Clinic served as consultants for problem cases. All women found to have high-grade precancer or true invasive cancer were cared for according to standard treatment protocols at each clinical site (The Cleveland Clinic Foundation and Hospital Maternidad Nuestra Senora de la Altagracia). Patient information, pathology, and OCT images were then de-identified and coded to preclude identification by investigators. Three investigators (P.F.E., L.R.E., and J.L.B.) interpreted the coded OCT images after all patient examinations had been completed. They each diagnosed the images independently using the following scale: OCT images were graded as normal if a well-organized, simple two-layer structure was seen with a sharp interface between the surface (squamous) epithelium and the underlying stromal layer (dense connective tissue), abnormal if the tissue was unstructured with no apparent interface present, and indeterminate if irregularities on the images suggested artifacts or physiologic alterations (swelling of the epithelium, edema of the stromal layer, or inflammation) and did not meet diagnostic criteria for normal or abnormal. In order to provide information as to the application of OCT for limited and varying clinical circumstances, the images were read three separate times for each patient. First, the images were read knowing only the patient s age. Second, they were read with knowledge of the VIA impression by quadrant and with access to a nonmagnified digital image of the cervix. Finally, the OCT images were interpreted with full knowledge of the referral Pap results, the colposcopic diagnosis by quadrant, and viewing a magnified digital photograph of the cervix. Statistical design and analysis The aim of this study was to assess the diagnostic capabilities of OCT alone, and in combination to VIA and colposcopy, in preinvasive and invasive lesions of the uterine cervix. Sample size calculation assumed that with 200 patients the study will be powered (80%) to detect clinically meaningful differences in accuracy between the two procedures. Assuming a 5% exclusion rate, a total of 210 patients were recommended for this study. The University of North Carolina at Chapel Hill served as the central statistical and data management center. The performance characteristics of the diagnostic strategies were evaluated by calculating the sensitivity and specificity according to the standard definition; these results are presented with their corresponding 95% confidence intervals (CI). Since all eligible women received the reference standard (cervical biopsy), the above estimates were calculated directly in our study. Those women with CIN 2 or greater were considered as positive cases for the reporting of the above estimates. Receiver operating characteristic curves (ROC) were also generated for the three sets of OCT readings. The difference between the area under the curves was tested to detect if there was a difference in the diagnostic capability of the strategies (Fig. 2). The data were also analyzed using generalized estimating equations. All data management and statistical analysis were performed with STATA 8 (StataCorp LP, College Station, TX). Results Two hundred twenty data forms reflected subject enrollment. Two hundred ten women were examined at the Hospital Maternidad in the Dominican Republic and 10 at The Cleveland Clinic Foundation. Eight subjects were eliminated. One due to age,18 years, two with heavy bleeding (one cancer and one marked atrophy), one with normal Pap, two with recent cone biopsies, one with a blank form (number inadvertently skipped), and one unknown resulting in 212 evaluable patients.

4 1818 P.F. Escobar et al. Sensitivity Specificity oct_extra_1 ROC area: oct_extra_2 ROC area: oct_extra_3 ROC area: Figure 2. Receiver operating characteristic curve of optical coherence tomography (OCT), visual inspection with acetic acid (VIA), and colposcopy for lesions distant from the SCJ. The mean age of the study group was years (range 18 80). The mean number of pregnancies was 3.77 (range 0 17). One hundred eighty-nine women (89%) were premenopausal, and 23 (11%) were postmenopausal. Eighty-two percent of the women were married or cohabitating. Forty-two percent of the women had a symptomatic vaginal discharge at the time of the examination. Forty-eight of the patients (23%) presented with atypical squamous cells of undetermined significance), 142 (67%) with low-grade squamous intraepithelial lesion, and 22 (10%) had high-grade squamous intraepithelial lesion on Pap smear. Pathology on cervical biopsies taken revealed 179 women (84%) with greater than or equal to CIN 1, while thirty-three women (16%) had greater than or equal to CIN 2. A total of 1215 OCT images were matched to the final histologic diagnosis. Table 1 displays the sensitivity and specificity of OCT alone and in combination with VIA and colposcopy. The data are interpreted as a patient true disease status (all lesions) for lesions at the SCJ and for lesions distant from the SCJ. Overall, the sensitivity and specificity of blinded OCT (no knowledge of the Pap smear, VIA or colposcopy results, or the appearance of the digital image) for the diagnosis of CIN 2 was 56% (95% CI 39 72) and 59% (95% CI 52 66), respectively. The per person sensitivity and specificity of VIA for CIN 2 was 76% (95% CI 58, 88) and 34% (95% CI 27 41), respectively. The sensitivity and specificity of VIA 1 OCT for CIN 2 was 53% (95% CI 30 64) and 61% (95% CI 62 75), respectively. The sensitivity and specificity of colposcopy for CIN 2 was 39% (95% CI 24 57) and 71% (95% CI 64 77), respectively. The sensitivity and specificity of colposcopy 1 OCT for CIN 2 was 46% (95% CI 30 64) and 69% (95% CI 62 75), respectively (Table 2). If we analyze each OCT reading with the corresponding matched biopsy as an independent event using generalized estimating equations, the sensitivity and specificity for OCT as an aid to VIA and colposcopy are 33% (95% CI 14 60) and 91% (95% CI 86 94), respectively. ROC curves were also generated to access the quality of OCT as a diagnostic test for lesions at the SCJ and lesions distant from the SCJ. OCT demonstrated to be a fair diagnostic tool (ROC 0.73) adjunctive to VIA and colposcopy in accurately identifying lesions distant from the SCJ (Fig. 2). Discussion The objective of this clinical study was to determine the sensitivity and the specificity of this technology under two well-defined clinical settings. First, as an adjunct to cervical cancer screening using VIA in lowresource settings, and second, as an additional technique applied to traditional cervical cancer screening programs along with colposcopy and biopsy. We envision that in a low-resource setting, OCT directed by VIA could improve the specificity of the screening examination in detecting CIN 2 allowing for a single episode of care that includes both diagnosis and treatment without histologic confirmation. This point of care model for diagnosis and treatment

5 Sensitivity and specificity of OCT under two well-defined clinical settings 1819 Table 1. Characteristic Demographic characteristics Distribution Age (years) 35 (18 80) Reproductive age, n (%) Premenopausal 189 (89) Postmenopausal 23 (11) Pregnancy history 3.77 (0 17) Lifetime sexual partners, n (%) (48) 2 78 (37) 3 23 (11) 4 9 (4) Contraceptive history, n (%) Surgical 82 (39) Hormonal 18 (9) Barrier 8 (4) IUD 10 (5) Lactation 1 (,1) Other 4 (2) None 88 (42) Vaginal discharge, n (%) No 122 (57.55) Yes 89 (41.98) Unknown 1 (0.47) Referral diagnosis, n (%) ASCUS 48 (23) LSIL 142 (67) HSIL 22 (10) Biopsy results, n (%) CIN (84) CIN 2 33 (16) IUD, Intrauterine device; ASCUS, atypical squamous cells of undetermined significance; LSIL, low-grade squamous intraepithelial lesion; HSIL, high-grade squamous intraepithelial lesion; CIN, cervical intraepithelial neoplasia. of preinvasive disease would be ideal in areas with a limited health care infrastructure, limited human and financial resources, and with limited ability to have a patient recall system for follow-up care. In the developed world (moderate to high-resource setting), OCT could theoretically perform an optical biopsy. The optical biopsy would be taken at areas determined to be suspicious by colposcopy allowing for the selection of key areas for biopsy. This could reduce the number of unnecessary biopsies taken, while increasing the number of histologically significant biopsies (true positives). A typical application would be areas of extensive squamous metaplasia (both mature and immature) or determining the margins for a loop electro-excision procedure in a young patient with a wide transformation zone. In order to accomplish our dual objectives, a complex study design was developed. The Hospital Maternidad Nuestra Senora de la Altagracia in the Dominican Republic was selected as the major clinical research study site for several reasons. The cervical dysplasia clinic at this hospital is the only clinic within the Dominican Republic s public health system that evaluates abnormal cervical cytology. The clinic is well established and evaluates approximately 3500 patients annually. It is equipped with an examination room and a standard colposcope and the fact the clinic is open 5 days per week allowed efficient patient accrual. The potential for bias to be introduced by the clinicians recording their VIA, colposcopic, and OCT impressions in one examination room was great. We did not allow our clinicians to view the OCT images at the time of the examination. This was intended to properly blind clinical investigators from patient information, pathology, and most importantly VIA and colposcopic findings. Once a VIA and colposcopic impression was given by the clinical investigators and recorded, an OCT device operator handed the OCT probe device to the clinician, who in turn applied the probe to the appropriate areas of the cervix without observing any of the OCT images. The images were then recorded, matched, and interpreted remotely after all patients had been examined. As noted in the methods, all of the images were first read knowing only the patient s age. Second, they were read knowing the VIA impression by quadrant, and viewing a nonmagnified digital image of the cervix. Finally, the OCT images were interpreted with full knowledge of the referral Pap, the colposcopic diagnosis by quadrant, and viewing a magnified digital photograph of the cervix. There was fair to moderate agreement between the gynecological oncologists assessing the OCT images (j ¼ ). Test reliability depends, in large part, on the method of measuring the presence or absence of early disease (14). OCT technology is rater-dependent, and even though all clinical investigators in this study were similarly trained before the interpretation of the study images, the technology is still based on a visual interpretation and therefore subject to greater variability than an objective modality. OCT proved to be a fair diagnostic modality (ROC 0.73) adjunctive to VIA and colposcopy in accurately identifying lesions distant from the SCJ. An understanding of the histology of the cervix is critical to the use of OCT as it relates to the interpretation of the images and pathologic correlation. The cervix is covered by both columnar and stratified nonkeratinizing squamous epithelia. The SCJ is the most important colposcopic landmark, as this is where the most severe lesions are frequently located (15). Abnormal areas in the SCJ that lie close to or within the endocervical canal are difficult to evaluate for

6 1820 P.F. Escobar et al. Figure 3. Structured OCT image of normal uterine ectocervix (on left); unstructured OCT image of ectocervix (on right). several reasons. First, columnar epithelium near the transformation zone is comprised a single layer of mucin-secreting cells that architecturally lies in complex longitudinal folds and invaginations as it approaches the endocervical canal. On OCT, this has an unstructured appearance. Second, the disposable OCT probe sheath contains a membrane configured to be firmly applied over the area of interest to position the OCT probe perpendicular to the tissue surface. OCT readings obtained in areas that lie close to or within the endocervical canal potentially contained angled images, disrupting the optical architecture and influencing the sensitivity and specificity of the modality. Our success with images lying on the ectocervix, remote from the SCJ, could have been predicted. The ectocervix normally contains a layered structure consisting of squamous epithelium overlying dense connective tissue. On OCT, this is seen as a layered image with a clear distinction between the epithelium and the connective tissue. Thus, an unstructured area that is remote to the SCJ should not be confused with the normal endocervix and could be correctly interpreted as abnormal (Fig. 3). On the basis of the above findings, we believe that in order to truly measure the efficacy of this modality, future clinical studies should include real-time readings with OCT. This would theoretically reduce the false positives caused by improper positioning at the SCJ. Furthermore, by reviewing the OCT images during the gynecological examination, the clinician will be able to fully integrate the patient history (including Pap smear results), the visual examination before and after acetic acid, and the colposcopy impression. The OCT optical biopsy impression could then be compared to the results of histologic biopsies taken at the precise area that corresponds to the OCT image. A Figure 4. OCT image of chronic inflammation. Figure 5. OCT image of carcinoma in situ.

7 Sensitivity and specificity of OCT under two well-defined clinical settings 1821 Figure 6. OCT image of cervical intraepithelial neoplasia III. study with OCT impressions completed in real-time is currently being conducted, therefore, modeling the use of the OCT as a single episode of care imaging modality. Sensitivity and specificity estimates for VIA based on several published cross-sectional studies have varied from 29% to 95%, and 49% to 91%, respectively (16 19). In this study, the sensitivity and specificity of VIA alone was 76% and 34%. OCT as an adjunct to VIA for the detection of CIN 2 was 53% (95% CI 34 96) and 61% (95% CI 60 74), respectively, nearly doubling the specificity of VIA alone. In both cases, CIN 1 was considered as negative, because it is not a reproducible diagnosis and greater than 80% regress or do not progress, therefore it does not require any immediate treatment according to the standard of care (15). The variability among VIA studies can be explained, in part, by a lack of standardized test definitions, the background skills and experience of the clinicians, the underlying prevalence of disease, verification bias, and information bias (14). The clinician doing the VIA and colposcopy in the Hospital Maternidad was taught the VIA technique for the study. The VIA data supports the known learning curve for visual techniques in that sensitivity rapidly approaches the norm for the technique, but as the examiner gains experience it is reflected in improved specificity. We believe the data clearly supports the use of OCT in lowresource settings especially considering real-time use that we believe will further improve specificity (Figs 4,5,6). In summary, a technology such as OCT that is capable of performing real-time, cross-sectional microtissue imaging of the uterine cervix could be an influential adjunct to VIA and colposcopy. In particular, we believe this technology could potentially show greatest utility in the management of cervical dysplasia in low-resource settings where a single episode of care is most desirable. The algorithm would be primary screening with VIA and then OCT as the secondary screen for VIA positives. Appropriate treatment or referral in one clinical visit would then be based on the OCT findings. Recently, Goldie et al. (20) employed computer-based models to assess the cost-effectiveness of a variety of cervical cancer screening strategies in India, Kenya, Peru, South Africa, and Thailand. Interestingly, the most cost-effective strategies were those that required the fewest visits, and incorporated visual inspection of the cervix with acetic acid or DNA testing for human papillomavirus resulting in improved follow-up testing and treatment. Table 2. Sensitivity and specificity of OCT imaging alone and OCT as an adjunct to VIA and colposcopy for the detection of CIN 2 OCT indeterminate OCT abnormal Sensitivity CI Specificity CI Sensitivity CI Specificity CI Overall OCT 94 (78 98) 11 (07 16) 56 (39 72) 59 (52 66) VIA 1 OCT 94 (78 98) 16 (12 23) 53 (36 70) 61 (54 68) VIA 1 colpo 1 OCT 91 (74 97) 18 (13 24) 46 (30 64) 69 (62 75) SCJ images OCT 93 (76 98) 14 (09 19) 55 (37 72) 64 (57 71) VIA 1 OCT 93 (76 98) 20 (15 26) 52 (34 96) 67 (60 74) VIA 1 colpo 1 OCT 93 (76 98) 22 (17 29) 41 (25 60) 72 (65 79) Non-SCJ images OCT 83 (51 96) 43 (35 51) 50 (23 77) 79 (71 84) VIA 1 OCT 75 (43 92) 52 (44 60) 42 (18 70) 82 (75 87) VIA 1 colpo 1 OCT a 83 (51 96) 50 (42 58) 50 (23 77) 85 (79 90) Colpo, colposcopy. a ¼ ; a ROC ¼ ( ).

8 1822 P.F. Escobar et al. References 1 Sawaya GF, Brown AD, Washington AE, Garber AM. Clinical practice. Current approaches to cervical-cancer screening. N Engl J Med 2001;344: Pisani P, Parkin DM, Bray F, Ferlay J. Estimates of the worldwide mortality from 25 cancers in Int J Cancer 1999;83: DeMay RM. Common problems in Papanicolaou smear interpretation. Arch Pathol Lab Med 1997;121: World Health Organization. State of the art of new vaccines: research & development. Geneva: World Health Organization, Belinson J, Qiao YL, Pretorius R et al. Shanxi Province Cervical Cancer Screening Study: a cross-sectional comparative trial of multiple techniques to detect cervical neoplasia. Gynecol Oncol 2001;83: Huang D, Swanson EA, Lin CP et al. Optical coherence tomography. Science 1991;254: Brezinski ME, Tearney GJ, Bouma BE et al. Optical biopsy with optical coherence tomography. Ann N Y Acad Sci 1998;838: Fujimoto JG, Pitris C, Boppart SA, Brezinski ME. Optical coherence tomography: an emerging technology for biomedical imaging and optical biopsy. Neoplasia 2000;2: Pan YT, Lavelle JP, Bastacky SI et al. Detection of tumorigenesis in rat bladders with optical coherence tomography. Med Phys 2001; 28: Pitris C, Fujimoto JG, Brezinski ME. Imaging neoplasia. In: Bouma BE, Tearney GJ, eds. Handbook of optical coherence tomography. Basel, NY: Marcel Dekker, Inc., 2002: Shakhova NM, Feldchtein FI, Sergeev AM. Applications of optical coherence tomography in gynecology. In: Bouma BE, Tearney GJ, eds. Handbook of optical coherence tomography. Basel, NY: Marcel Dekker, Inc., 2002: Pitris C, Goodman A, Boppart SA, Libus JJ, Fujimoto JG, Brezinski ME. High-resolution imaging of gynecologic neoplasms using optical coherence tomography. Obstet Gynecol 1999;93: Escobar PF, Belinson JL, White A et al. Diagnostic efficacy of optical coherence tomography in the management of preinvasive and invasive cancer of uterine cervix and vulva. Int J Gynecol Cancer 2004; 14: Denny L, Kuhn L, Pollack A, Wright TC Jr. Direct visual inspection for cervical cancer screening: an analysis of factors influencing test performance. Cancer 2002;94: Ostor AG. Natural history of cervical intraepithelial neoplasia: a critical review. Int J Gynecol Pathol 1993;12: Slawson DC, Bennett JH, Herman JM. Are Papanicolaou smears enough? Acetic acid washes of the cervix as adjunctive therapy: a HARNET study. Harrisburg Area Research Network. J Fam Pract 1992;35: Belinson JL, Pretorius RG, Zhang WH, Wu LY, Qiao YL, Elson P. Cervical cancer screening by simple visual inspection after acetic acid. Obstet Gynecol 2001;98: Sankaranarayanan R, Basu P, Wesley RS et al. Accuracy of visual screening for cervical neoplasia: results from an IARC multicentre study in India and Africa. Int J Cancer 2004;110: Sankaranarayanan R, Chatterji R, Shastri SS et al. Accuracy of human papillomavirus testing in primary screening of cervical neoplasia: results from a multicenter study in India. Int J Cancer 2004;112: Goldie SJ, Gaffikin L, Goldhaber-Fiebert JD et al. Alliance for Cervical Cancer Prevention Cost Working Group. Cost-effectiveness of cervical-cancer screening in five developing countries. N Engl J Med 2005;353: Accepted for publication March 31, 2006