Cervical Cancer Screening in Cameroon: Interobserver Agreement on the Interpretation of Digital Cervicography Results

Transcription

1 ORIGINAL ARTICLE Cervical Cancer Screening in Cameroon: Interobserver Agreement on the Interpretation of Digital Cervicography Results Simon Manga, BSN, MSc, 1,2 Groesbeck Parham, MD, 3 Nkoum Benjamin, PhD, 2 Kathleen Nulah, MSc, 1 Lisa Kennedy Sheldon, PhD, 4 Edith Welty, MD, 1 Javier Gordon Ogembo, PhD, 5 Leslie Bradford, MD, 5 Zacharie Sando, MD, MPH, 6 Ray Shields, MD, 7 and Thomas Welty, MD, MPH 1 Objective: The World Health Organization recommends visual inspection with acetic acid (VIA) for cervical cancer screening in resourcelimited settings. In Cameroon, we use digital cervicography (DC) to capture images of the cervix after VIA. This study evaluated interobserver agreement of DC results, compared DC with histopathologic results, and examined interobserver agreement among screening methods. Method: Three observers, blinded to each other's interpretations, evaluated 540 DC photographs as follows: (1) negative/positive for acetowhite lesions or cancer and (2) assigned a presumptive diagnosis of histopathologic lesion grade in the 91 cases that had a histopathologic diagnosis. Observer A was the actual screening nurse; B, a reproductive health nurse; C, a gynecologic oncologist; and D, the histopathologic diagnosis. We compared inter-rater agreement of DC impressions among observers A, B, and C, and with D, with Cohen kappas. Results: For interpretations of DC, (negative/positive) strengths of agreement of paired observers were the following: A/B, moderate [K, 0.54; 95% confidence interval (CI), ], A/C, fair (K, 0.37; 95% CI, ), and B/C, moderate (K, 0.45; 95% CI, ). For presumptive pathologic grading, strengths of agreement for weighted Ks were as follows: A/B, moderate (K, 0.42; 95% CI, ); A/C, fair (K, 0.33; 95% CI, ); B/C, fair (K, 0.54; 95% CI, ); A/D, moderate (K, 0.59; 95% CI, ); B/D, moderate (K, 0.58; 95% CI, ); and C/D, moderate (K, 0.50; 95% CI, ). Conclusions: Interobserver agreement of DC interpretations was mostly moderate among the 3 observers, between them and histopathology, and comparable to that of other visual-based screening methods, i.e., VIA, cytology, or colposcopy. Key Words: cervical cancer, screening, acetic acid, digital cervicography, interobserver agreement (J Lower Gen Tract Dis 2015;19: 00 00) Cervical cancer is the second leading cause of cancer death in women in Cameroon, with age-standardized incidence and 1 Cameroon Baptist Convention Health Services, Bamenda, Cameroon; 2 Catholic University of Central Africa, Yaoundé, Cameroon; 3 University of North Carolina, NC, USA; 4 University of Massachusetts, Boston, MA, USA; 5 Catholic University of Central Africa, Yaoundé, Cameroon; 6 Yaoundé Gyneco-Obstetric and Pediatric Hospital, Yaoundé, Cameroon; and 7 The Hill Group, Bellingham,WA, USA Correspondence to: Thomas Welty, Cameroon Baptist Convention Health Services, 939 Flynn Lane, McCall, ID, US. thomaswelty@gmail.com Disclosure of sources of Funding: This project was supported with grant from University of North Carolina AIDS International Training and Research Program (Grant No. 2D43TW001039). IRB status: The Cameroon Baptist Convention Health Services and University of North Carolina institutional review boards reviewed and approved the protocol for this study. Copyright 2015 American Society for Colposcopy and Cervical Pathology ISSN: DOI: /LGT mortality rates per 100,000 women of 30.0 and 17.5, respectively. 1 Recent World Health Organization (WHO) guidelines for screen-and-treat programs for low- and middle-income countries (LMIC) recommend visual inspection with acetic acid (VIA), high-risk human papillomavirus (hrhpv) DNA testing, or cytologic examination for primary screening, and stress the importance of immediate treatment when possible, to assure that women with precancers are not lost to follow-up. 2,3 Visual inspection with acetic acid, the least costly option, is affected by visual acuity, light source, and training and produces no permanent image. The carehpv test (Qiagen, Gaithersburg, MD), designed for LMIC, is most sensitive but is expensive and requires 90-specimen batched laboratory testing. 4,5 Cytologic examination requires extensive health care infrastructure and multiple patient visits for evaluation and treatment, and to overcome low cytologic sensitivity. 6,7 Positive cytology and hrhpv tests commonly require triage with colposcopy to determine the true nature of the underlying lesion. Colposcopes are expensive, not portable, and easily damaged by electrical surges common in LMICs. There is a critical need for a point-of-care screening method that is affordable, sustainable, produces a permanent record, can be performed accurately by various levels of health care workers, and linked to immediate treatment. Visual inspection with acetic acid enhanced by DC enables various levels of health care providers to rapidly ascend the learning curve for diagnosing precancers and cancers through frequent reviews of cervical photographs (cervicographs) by peers and experts, produces a permanent image matched to each woman's history and physical examination, and allows same-day treatment or further diagnostic procedures by defining the limits of lesions, thus meeting the aforementioned criteria. The Cameroon Baptist Convention Health Services (CBCHS), a faith-based organization with a network of 81 health facilities in 6 of Cameroon's 10 regions ( org), has a nurse-led Women's Health Program (WHP) that has used DC to screen 46,000 women for cervical cancer screening since The WHP uses digital cameras to project 30 magnified real-time images onto a commercial 13-inch television (TV) screen that can be seen by both the woman and the provider. We found TVs give a much clearer image than projecting onto a computer screen, are inexpensive, and require fewer connecting devices. The cervicographs taken are linked to each woman's medical record in the WHP database. Cervical abnormalities visible on the TVare used to describe the findings to each woman screened, thus empowering her to participate in treatment decisions. Cervicographs are archived for follow-up comparison, staff training, consultation, and quality improvement. Women's Health Program nurses offer same-day cryotherapy, punch biopsy, referral for loop electrosurgical excision procedure (LEEP), or other indicated evaluations. This study evaluated interobserver Journal of Lower Genital Tract Disease Volume 19, Number 4, October

2 Manga et al. Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 agreement among 3 health professionals in interpreting cervicographs and compares their presumptive pathologic lesion grades with available histopathologic results. METHODS After a 2007 seminar in Cameroon on cervical anatomy and pathophysiology, DC, cryotherapy, and LEEP by the Zambian Cervical Cancer Prevention Program, 8 the WHP developed cervical cancer policies and standard operating procedures based on the Zambian model but adapted to Cameroon-specific needs and capacity. We used Olympus SP-510 Ultra Zoom digital cameras with an Olympus MCON mm macro extension lens (Olympus Corporation of Americas, New Orleans, LA) to take 3 cervicographs of each cervix: the first, immediately after insertion of the vaginal speculum; the second, 2 minutes after application of a 3% to 5% acetic acid solution; and the third, immediately after application of Lugol iodine. The Olympus camera that we used is no longer manufactured, so we have switched to the Canon SX50 HS camera with a Canon 58-mm close-up lens 500D (Canon, USA; Melville, NY), which works equally well. We used a Black Diamond 200 lumen icon headlamp (Black Diamond Equipment, Ltd, Salt Lake City, UT) to illuminate the cervix. This study evaluated only DC after the application of acetic acid. In , WHP screened a total of 14,376 women. Of these, 8,271 (57.5%) were DC negative and 1,459 (10.2%) were DC positive. For the remaining 4,646 (32.3%), Observer A, who comprised 10 original screening nurses, had interpreted the DC examination as either inadequate or uncertain. Observer A had been trained to interpret results as inadequate if the ectocervix was completely negative for acetowhite lesions, but the entire transformation zone could not be fully visualized (usually because the squamocolumnar junction was beyond sight in the endocervical canal). We designed this study retrospectively to assess interobserver agreement of cervicographs that were clearly diagnosed as positive or negative. We accepted Observer A's interpretation of uncertain because of the high rate of false-positive acetowhite lesions in women with cervicitis, which would have confounded the results of our study. Therefore, cervicographs initially evaluated as inadequate or uncertain were excluded from the study. All cervicographs were de-identified and only the WHP identification number was used to match them with available histopathologic results. In this retrospective cross-sectional study, archived cervicographs that were previously taken during cervical cancer screening at 5 stationary WHP clinics and multiple mobile outreach clinics and reported as positive or negative were included in the study. Positive cervicographs are those that observer A felt had acetowhite lesions or were suspicious for cancer. Negative cervicographs represent the normal cervixes that had no acetowhite areas. Outreach clinics were conducted in either a donated US Army ambulance converted to a gynecologic suite or in established village clinics. The DC interpretations by the WHP nurses who originally screened the patients and took the cervicographs were considered observer A. Only cervicographs read by observer A as negative or positive were included. Of these 10 nurses, 4 held bachelor's degrees in nursing (BNS); 4 were stateregistered nurses; one was a registered nurse-midwife; and one was a state-registered reproductive health nurse. Observer B (SM, an CBCHS WHP Supervisor, BNS, M.Sc.), who had received an additional 2 months of training from colleagues in Zambia on DC and LEEP, interpreted the same archived cervicographs. Observer B trained and supervised the other 10 WHP nurses. The third observer, C (GP, a gynecologic oncologist who directs the Zambian program), also interpreted the same cervicographs. In addition to reading the cervicographs as negative, positive, or suspicious for cancer, observers A, B, and C also gave their presumptive pathologic grading [negative, cervical intraepithelial neoplasia (CIN) low-grade, CIN high-grade, or suspicious for cancer]. For the 91 cases that had tissue specimens sent for histopathologic examination, we compared their presumptive pathologic grading to the histopathologic reports. Before the study, observer A had performed cervical punch biopsies on a subset of cervixes that appeared suspicious for cancer or had lesions that were difficult to diagnose. Appropriately trained WHP nurses performed LEEPs on a referral basis on eligible lesions at CBCHS hospitals, where surgical backup was available and excised tissue was sent for histopathologic diagnosis. These tissue specimens FIGURE 1. Kappas and 95% confidence intervals for interobserver agreement in negative/positive cervicograph interpretation between paired observers A, B, and C. A C has significantly lower agreement than A B American Society for Colposcopy and Cervical Pathology

3 Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 Interobserver Agreement on Cervicographs were sent for histopathologic diagnosis to the Gyneco-Obstetric and Pediatric Hospital in Yaoundé, the capital city. The presumptive pathologic grading of cervicographs by observers A, B, and C were compared to the histopathologic reports (observer D). Of note, the pathology request forms that had accompanied the specimens to the pathologist included observer A's presumptive pathologic grading and, thus, observer D was not blinded to observer A's interpretations. To measure interobserver agreement for cervicographic interpretations (negative, positive, or suspicious for cancer), an on line calculator ( was used to calculate Cohen kappa (K) statistic with 95% confidence intervals [CIs]. The K statistic adjusts the percent agreement for the degree of agreement expected by chance alone. 9 To compare DC-based presumptive pathologic grading between paired observers (A B, A C, and B C), and between each of these observers' readings and observer D's histopathologic diagnoses, a weighted K was calculated to reflect the degree of discrepancy in the lesion grading when 2 observers' readings are compared. 9 Both the unweighted and the weighted K values were interpreted as follows: K < 0.20, poor; , fair; , moderate; , substantial; and , almost perfect. 10 To assure adequate power, we used STATA (College Station, TX) to calculate sample size at a CI of 95% and a precision of 0.05, yielding our sample size of 554 cervicographs. Of the 554 cervicographs, 266 were previously interpreted by observer A as positive and 288 as negative. To allow equal numbers of cervicographs from each of the 5 sites, we increased the sample size to 555, or 111 cervicographs per site. We used a random number table to select cervicographs from each site and divided them into approximately half previously interpreted as negative and half, as positive. We excluded 15 of these 555 cervicographs because of poor quality and used 540 for the study. From among the positives, purposive sampling was used to select 91 cervicographs for which histopathologic results were available. All cervicographs were saved in random distribution in an electronic file by the WHP data manager (K.N.) for blinded review by observers B and C. Both observers B and C then reported their interpretations independently to the data manager, who then developed tables of all reported interpretations by observers A, B, and C, as well as the histopathology report diagnoses (observer D) on the subset for which histopathologic reports were available. The CBCHS and University of North Carolina institutional review boards reviewed and approved the protocol for this study. We obtained written consents during screening from all women whose cervicographs were used in the study during cervical cancer screening in the clinics, including consent to use their cervicographs in scientific publications with no identifiers. RESULTS For negative/positive DC interpretations, strengths of agreement between paired observers were as follows: A B, moderate (K, 0.54; 95% CI, ); A C, fair (K, 0.37; 95% CI, ); B C, moderate (K, 0.45; 95% CI, ). Table 1 shows a comparison of interpretations of cervicographs by observers A, B, and C. A C had significantly lower agreement than A B. None of the other differences in kappas (K) were statistically significant. The agreement between pairs was moderate except for A C, which was fair. For presumptive pathologic grading, strengths of agreement for unweighted/weighted K were as follows: A B, moderate (K, 0.42/0.42; 95% CI, ); A C, fair (K, 0.31/0.33; 95% CI, ); B C, fair (K, 0.39/0.54; 95% CI, ); A D, moderate (K, 0.53/0.59; 95% CI, ); B D, moderate (K, 0.50/0.58; 95% CI, ); and C D, moderate (K, 0.40/0.50; 95% CI, ). Comparisons of presumptive pathologic grading by observers A, B, and C with histopathologic diagnosis (observer D) and the K coefficients are provided in Table 2. There were no statistically significant differences in the interobserver agreements among these comparisons. All of the weighted K coefficeints indicated moderate interobserver agreement between pairs except A C, which was fair agreement. Figure 1 graphically displays the Ks with 95% CI for interobserver correlations between paired observers A, B, and Cs' negative/positive interpretations. A-B agreement was significantly higher than A C for negative/positive interpretations. Figure 2 compares the weighted Ks with 95% CI for observers A, B, and Cs' presumptive pathologic grading. The 95% CIs overlap, indicating that the differences are not significant. Figure 3 compares observers A, B and Cs' presumptive pathologic grading with the histopathologic diagnoses (observer D). The 95% CIs overlap, indicating that the differences are not significant. Figure 4 shows 3 cervicographs each from 9 women, observers A, B, and Cs' interpretations, and in cases that had specimens sent to pathology, observers A, B, and Cs' presumptive pathologic grading and the histopathologic diagnosis (observer D). The visual criteria used to assign presumptive pathologic grading are shown at the bottom of Figure 4. DISCUSSION All interobserver agreements in this study had moderate strength, except for the A-C comparisons, which were fair. The 10 WHP nurses who screened the women (observer A) were less experienced than observers B and C, suggesting that experience may increase interobserver agreement. However, observer A had the advantage of real-time evaluation of the cervixes, whereas observers B and C had only cervicographs to review. If histopathology is considered the criterion standard, then observer A's TABLE 1. Comparisons of Positive Versus Negative VIA-DC Photograph Interpretations, Kappa Values, and CIs Between Observers A and B, A and C, and B and C A B Observer B Observer A Positive Negative Total Pos Neg Total: A C Observer C Observer A Pos Neg Total Pos Neg Total: B C Observer C Observer B Pos Neg Total Pos Neg Total: A and B A and C B and C Kappa % CI % agreement 77.4% 68.6% 75.3% Strength of agreement Moderate Fair Moderate 2015 American Society for Colposcopy and Cervical Pathology 3

4 Manga et al. Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 TABLE 2. Comparisons of Presumptive Pathologic Grades Between Observers A, B, C and Histopathologic Diagnosis (Observer D), Kappa Values, and CIs A B Observer B A-B Observer A Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 58.7% Total Strength of agreement Moderate A C Observer C A-C Observer A Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 46.8% Total Strength of agreement Fair A D Observer D A-D Observer A Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 67.5% Total Strength of agreement Moderate B C Observer C B-C Observer B Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 55.7% Total Strength of agreement Moderate B D Observer D B-D Observer B Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 60.0% Total: Strength of agreement Moderate C D Observer D C-D Observer C Grade* Total Unweighted Kappa % CI Weighted Kappa % CI % agreement 53.9% Total Strength of agreement Moderate *0, normal; 1, CIN low-grade; 2, CIN high-grade; 3, suspicious for cancer. interpretations had slightly better agreement than those of observers B and C, although this was not statistically significant. One limitation of this study was that observer D was not blind to observer A's presumptive pathologic grading, which was recorded on the pathology request forms. However, since there was no statistically significant difference between observers A, B, or Cs' presumptive pathologic grading and D s histopathologic diagnoses, it is unlikely that this influenced the outcomes of the study. The Cameroon government has conducted intermittent cervical cancer screening campaigns using VIA and/or cytologic examination but lacks adequate funding and health care infrastructure for effective, large-scale, population-based screening. A cost-effective screen-and-treat strategy for scale-up is essential to expand cervical cancer prevention to most Cameroonian women, especially in poor rural areas. Thus, visual inspection methods are the most promising screening tool. Four studies performed in Cameroon have found varying sensitivities for VIA and cytologic examination The range for VIA sensitivity was from 25% to 70%, and that for cytology was from 47% to 90%. In comparison, 2 meta-analyses in other LMIC showed VIA sensitivities of 77% and 80% 15,16 compared American Society for Colposcopy and Cervical Pathology

5 Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 Interobserver Agreement on Cervicographs FIGURE 2. Comparison of weighted Kappa values and 95% confidence intervals for presumptive pathologic grading between observers A, B, and C. to cytology, which was 57%. 15 In human immunodeficiency virus (HIV)-positive women, test sensitivity increased from 65% with VIA alone to 75% with DC, 17 suggesting that DC improves sensitivity, although part of this difference may be related to nurse versus physician observers. A study of test performance of carehpv, VIA, and cytology to detect CIN2+ among 16,961 women in India, Nicaragua, and Uganda showed the following sensitivities: cervical carehpv, 81.5%; vaginal carehpv, 69.6% to 71.3%; VIA, 21.9% 73.6%; and cytology, 40.7% to 73.7%. 5 A study of HIV-positive women in Botswana comparing interpretations of DC by 4 trained nurse-midwives and one gynecologist using 63 Samsung cell phone cervicographs demonstrated substantial interobserver agreement (K, 0.71). 18 Two other studies of VIA interpretations showed fair to moderate agreement (Ks, ) and sensitivities of 84% to 89%. 19,20 Cervical cytology (Pap) interobserver agreement, which was poorest for atypical squamous cells of undetermined significance (ASC-US), improved from 55.1% to 82.3% when cytology categories were reduced from the Bethesda classification to 2 (negative versus positive). 21 Among HIV-infected women in Zambia, DC was at least as good as cytology for detecting highgrade precancers. 22 In this study, DC was compared to cytology at a CIN 2+ cutoff, and cytology sensitivity and specificity were 61% and 58%, respectively, compared to DC (84% and 58%, respectively). Colposcopic interobserver agreement studies based on reviewing digitized colposcopic photographs of acetic acid stained cervixes (similar to reviewing DC photographs) showed a K of 0.57 (moderate agreement). 23 Another triage study showed poor agreement, 24 and, in the ASC-US or low-grade squamous intraepithelial lesion (ASC-US/LSIL), Ks for various Reid index parameters ranged from 0.11 to 0.26 (poor to fair). 25 Histopathologic interpretation of tissue specimens is usually regarded as the criterion standard, but interobserver agreement among pathologists is only minimally better than that of VIA, DC, or colposcopy. K values among pathologists improved from 0.56 to 0.71 with greater standardization of classification. 26,27 The K for interpretation of hematoxylin and eosin stain alone (0.57) improved to 0.75 with addition of P16 staining. 28 In another study, the weighted K for interpretation of cervical biopsies was 0.36 but was better for higher-grade lesions. 26,29 In our study, the K values for VIA-DC interpretation were mostly in the moderate range of 0.43 to 0.59, with only 2 values in the fair range of 0.33 to 0.37, and thus are comparable to those of cytologic examination, colposcopy, and even histologic examination with hematoxylin and eosin alone. Digital cervicography has several advantages over nonenhanced visual screening methods: (1) It obtains a permanent image for secondary review and comparison of cervicographs over time; (2) it has the potential to increase sensitivity in comparison to VIA alone because it magnifies the image before taking the cervicograph and enables zooming-in on suspicious areas; (3) it empowers screened women to view their cervix with realtime 30 images on a TV screen and receive education enabling them to participate in informed decision-making; (4) it enables new staff to more rapidly acquire diagnostic skills and improve their quality of care by discussion of image interpretation through cervicograph review; (5) it provides a cervicograph database for inclusion with patients' medical records, expert consultation, and FIGURE 3. Comparison of weighted Kappa values and 95% confidence intervals for presumptive pathologic grading of observers A, B, and C and the pathologist's (Observer D's) histopathologic diagnoses American Society for Colposcopy and Cervical Pathology 5

6 Manga et al. Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 FIGURE 4. Comparison of observers A, B, and Cs' interpretations and their presumptive pathologic grading and the histopathologic diagnosis (observer D) for cervicographs from nine women. program evaluation; and (6) it provides immediate results so that women can receive same-day treatment, thus reducing loss to follow-up. These advantages have enabled the Zambian Cervical Cancer Prevention Program to screen more than 200,000 women,30 and our program has now screened more than 46,000 women, demonstrating the feasibility of scaling up DC to have a public health impact in sub-saharan Africa. All treatment decisions require adequate visual evaluation of the cervix, including lesion size, location, and severity. It is critical that DC providers become skilled in taking excellent cervicographs of fully exposed cervixes including the entire transformation zone without blurring. Using a camera, rather than a colposcope, makes it more difficult to manipulate the cervix with a cotton-tipped applicator or endocervical speculum to gain full view of a squamocolumnar junction beyond sight in the endocervical canal. Because of the high rate of inadequate interpretations, the WHP has retrained the nurse clinicians to perform additional manipulations to fully expose the squamocolumnar junction such as using a larger speculum, opening the speculum wider, lifting the anterior lip with a cotton tip applicator or using an endocervical speculum. Since cervicitis creates false-positive acetowhite lesions, we treat it according to the WHO/US Centers for Disease Control and Prevention treatment guidelines and ask women to return for DC after completing treatment. We therefore opted to exclude cervicographs interpreted by observer A as inadequate or uncertain because we felt they would confound valid interobserver comparisons. In retrospect, exclusion of the cervicographs that observer A diagnosed as inadequate or uncertain is possibly 6 a weakness of this study, since significant pathology may have been missed. Although DC could miss atypical glandular lesions confined to the endocervix, it is unlikely that we missed many squamous lesions. We plan future studies to investigate the use of HPV testing to manage women whose DC is inadequate or uncertain. CONCLUSION Our study's levels of agreement in DC interpretation among 3 health care professionals with differing levels of training and between these professionals and histopathologic interpretations are comparable to those in studies of VIA, cytology, and colposcopy. Digital cervicography has several important advantages over other screening methods such as same-day treatment of precancer, low cost, and permanent cervicographic records that can be used for quality improvement and follow-up care. With limited external private funding and low fees, the WHP has now screened more than 46,000 women. Therefore, policymakers should consider DC as an option for primary cervical cancer screening in LMIC, and its role as a secondary test after hrhpv testing warrants further investigation. ACKNOWLEDGMENTS The authors thank Adaora A. Adimora, MD, MPH, Professor of Medicine, Division of Infectious Diseases, University of North Carolina; and Julius Atashili, MD, MPH, PhD of the University of Buea, Cameroon for their invaluable support of the project. The dedication and hard work of the WHP nurses and 2015 American Society for Colposcopy and Cervical Pathology

7 Journal of Lower Genital Tract Disease Volume 19, Number 4, October 2015 Interobserver Agreement on Cervicographs the support of Professor Pius Tih, Director of CBCHS made this study possible. REFERENCES 1. Ferlay J, Soerjomataram I, Ervik M. GLOBOCAN 2012: Cancer Incidence and Mortality Worldwide: IARC Cancer Base No. 10 [Internet], International Agency for Research on Cancer, WHO. Guidelines for screening and treatment of precancerous lesions for cervical cancer prevention: supplemental material: GRADE evidence-to-recommendation tables and evidence profiles for each recommendation. Geneva: WHO; 2013 Contract No.: WHO/RHR/ WHO. Comprehensive Cervical Cancer Control: A Guide to Essential Practice. 2nd ed. Australia: 2014 Contract No.: Qiao YL, Sellors JW, Eder PS, et al. A new HPV-DNA test for cervical-cancer screening in developing regions: a cross-sectional study of clinical accuracy in rural China. Lancet Oncol 2008;9: Jeronimo J, Bansil P, Lim J, et al. A multicountry evaluation of carehpv testing, visual inspection with acetic acid, and Papanicolaou testing for the detection of cervical cancer. Int J Gynecol Cancer 2014;24: Denny L, Quinn M, Sankaranarayanan R. Screening for cervical cancer in developing countries. Vaccine 2006;24:S Denny LA, Sankaranarayanan R, De Vuyst H, et al. Recommendations for cervical cancer prevention in sub-saharan Africa. Vaccine 2013;31 (suppl 5):F Mwanahamuntu MH, Sahasrabuddhe VV, Pfaendler KS, et al. Implementation of 'see-and-treat' cervical cancer prevention services linked to HIV care in Zambia. AIDS 2009;23:N Fleiss JL. Statistical Methods for Rates and Proportions. NewYork:John Wiley & Sons Inc; 1981: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33: Bigoni J, Gundar M, Tebeu PM, et al. Cervical cancer screening in sub-saharan Africa: a randomized trial of VIAversus cytology for triage of HPV-positive women. Int J Cancer 2014;137: Tebeu PM, Fokom-Domgue J, Crofts V, et al. Effectiveness of a two-stage strategy with HPV testing followed by visual inspection with acetic acid for cervical cancer screening in a low-income setting. Int J Cancer 2015;136:E Untiet S, Vassilakos P, McCarey C, et al. HPV self-sampling as primary screening test in sub-saharan Africa: implication for a triaging strategy. Int J Cancer 2014;135: Doh AS, Nkele NN, Achu P, et al. Visual inspection with acetic acid and cytology as screening methods for cervical lesions in Cameroon. Int J Gynaecol Obstet 2005;89: Sauvaget C, Fayette JM, Muwonge R, et al. Accuracy of visual inspection with acetic acid for cervical cancer screening. Int J Gynaecol Obstet 2011;113: Arbyn M, Sankaranarayanan R, Muwonge R, et al. Pooled analysis of the accuracy of five cervical cancer screening tests assessed in eleven studies in Africa and India. Int J Cancer 2008;123: Firnhaber C, Mao L, Levin S, et al. Evaluation of a cervicography-based program to ensure quality of visual inspection of the cervix in HIV-infected women in Johannesburg, South Africa. J Low Genit Tract Dis 2015;19: Quinley KE, Gormley RH, Ratcliffe SJ, et al. Use of mobile telemedicine for cervical cancer screening. J Telemed Telecare 2011;17: Hillmann Ede C, Dos Reis R, Monego H, et al. Cervical digital photography for screening of uterine cervix cancer and its precursor lesions in developing countries. Arch Gynecol Obstet 2013;288: Sherigar B, Dalal A, Durdi G, et al. Cervical cancer screening by visual inspection with acetic acid interobserver variability between nurse and physician. Asian Pac J Cancer Prev 2010;11: Sherman ME, Dasgupta A, Schiffman M, et al. The Bethesda interobserver reproducibility study (BIRST). Cancer Cytopathol 2007;111: Bateman AC, Parham GP, Sahasrabuddhe VV, et al. Clinical performance of digital cervicography and cytology for cervical cancer screening in HIV-infected women in Lusaka, Zambia. J Acquir Immune Defic Syndr 2014;67: Sellors JW, Jeronimo J, Sankaranarayanan R, et al. Assessment of the cervix after acetic acid wash: inter-rater agreement using photographs. Obstet Gynecol 2002;99: Jeronimo J, Massad LS, Castle PE, et al. Interobserver agreement in the evaluation of digitized cervical images. Obstet Gynecol 2007;110: Massad LS, Jeronimo J, Schiffman M. Interobserver agreement in the assessment of components of colposcopic grading. Obstet Gynecol 2008; 111: de Vet HC, Knipschild PG, Schouten HJ, et al. Sources of interobserver variation in histopathological grading of cervical dysplasia. JClin Epidemiol 1992;45: De Vet HC, Koudstaal J, Kwee W-S, et al. Efforts to improve interobserver agreement in histopathological grading. J Clin Epidemiol 1995;48: Bergeron C, Ordi J, Schmidt D, et al. Conjunctive p16ink4a testing significantly increases accuracy in diagnosing high-grade cervical intraepithelial neoplasia. Am J Clin Pathol 2010;133: Basu P, Kamal M, Ray C, et al. Interobserver agreement in the reporting of cervical biopsy specimens obtained from women screened by visual inspection with acetic acid and hybrid capture 2. IntJGynecolPathol2013; 32: Parham GP, Mwanahamuntu MH, Kapambwe S, et al. Population-level scale-up of cervical cancer prevention services in a low-resource setting: development, implementation, and evaluation of the cervical cancer prevention program in Zambia. PLoS One 2015;10:e American Society for Colposcopy and Cervical Pathology 7