Fax: (212) 305-5562
Direct visual inspection for cervical cancer screening
An analysis of factors influencing test performance
Article first published online: 15 MAR 2002
Copyright © 2002 American Cancer Society
Volume 94, Issue 6, pages 1699–1707, 15 March 2002
How to Cite
Denny, L., Kuhn, L., Pollack, A. and Wright, T. C. (2002), Direct visual inspection for cervical cancer screening. Cancer, 94: 1699–1707. doi: 10.1002/cncr.10381
- Issue published online: 15 MAR 2002
- Article first published online: 15 MAR 2002
- Manuscript Accepted: 31 OCT 2001
- Manuscript Revised: 15 OCT 2001
- Manuscript Received: 19 MAR 2001
- Bill and Melinda Gates Foundation
- AVSC International
- Cancer Association of South Africa
- Department of National Health, South Africa
- direct visual inspection;
- human papillomavirus;
- Papanicolaou smears;
- squamous intraepithelial lesions;
- cervical intraepithelial neoplasia
The authors evaluated direct visual inspection of the cervix after the application of 5% acetic acid (DVI) as a cervical cancer screening test for use in low-resource settings.
Four tests were used to screen 2754 previously unscreened women: DVI with and without ×4.5 magnification and differentiating between all lesions and well-circumscribed lesions; testing for high-risk types of human papillomavirus DNA using the Hybrid Capture II assay, cervical cytology, and Cervicography™. Women with positive results on any of the four screening tests were referred for colposcopy and histologic sampling. All women were tested for N. gonorrhea, C. trachomatis, T. vaginalis, and human immunodeficiency virus 1.
Histologically confirmed carcinoma was diagnosed in 21 women (0.8%), high-grade squamous intraepithelial lesions (SILs) were diagnosed in 96 women (3.5%), and low-grade SILs were diagnosed in 102 women (3.7%). The estimated sensitivity of DVI when performed without magnification for high-grade SILs was 70%, with an estimated specificity of 79%. Magnification did not significantly improve sensitivity for high-grade SILs (74% with magnification) but significantly reduced specificity (77%). Restricting the definition of a positive DVI test to a well-defined acetowhite lesion reduced sensitivity and significantly improved specificity. Infection with T. vaginalis, N. gonorrhea, and C. trachomatis did not alter sensitivity or specificity.
This study confirmed the utility of DVI as a primary screening test. Evaluation of the clinical effectiveness of screening with this low-cost and simple test in low-resource settings is necessary. Cancer 2002;94:1699–707. © 2002 American Cancer Society.
Most cases of cervical carcinoma can be prevented through screening. Although there are no randomized trials, case–control and cohort studies from a number of developed countries have shown that the incidence of and mortality from cervical carcinoma can be reduced with the implementation of mass, organized, cytologically based cervical cancer screening programs.1–7 Few if any developing countries, however, have been able to initiate and sustain effective cytologically based screening programs. Screening, when it has been implemented in developing countries, has tended to be sporadic or opportunistic, with little impact on the incidence of cervical carcinoma.8–13 The failure to establish screening programs in developing countries is an important reason why cervical carcinoma remains the most common cause of cancer death among women in these countries.
This has led to the search for alternative, technologically more appropriate methods of screening for cervical carcinoma and its precursors, which would overcome some of the barriers posed by cytologic screening. One such alternative is direct visual inspection of the cervix after the application of 5% acetic acid (DVI). Several studies have suggested that DVI has an ability nearly equivalent to that of cervical cytology to detect cervical carcinoma precursors.14–19 DVI offers several advantages for screening: It is a low-technology test with minimal infrastructural and equipment requirements, it provides an immediate on-site result, and nursing sisters and other allied health care workers can be trained to perform the test in basic primary care clinics.
One of the major concerns identified in evaluation studies of the performance of DVI is the large proportion of women classified as positive by the test, which results in low specificity and low positive predictive values. Other problems with DVI include different definitions used to evaluate the performance of DVI and the lack of reproducible methods of quality control. Thus, we sought to compare the performance of DVI with and without magnification and using different definitions of acetowhite lesions for a positive DVI test with the objective of identifying the characteristics of cervical lesions identified by DVI that would be predictive of true carcinoma precursors. In addition, we sought to investigate the influence of sexually transmitted infections, menopausal status, and other patient characteristics on test performance.
MATERIALS AND METHODS
From January 1998 to November 1999, 2754 women were recruited from a periurban settlement outside Cape Town, South Africa. Women age 35–65 years who had not been screened previously or treated for cervical carcinoma, who were not pregnant, and who had not undergone hysterectomy were eligible for the study. The study was approved by the Institutional Review Boards of Columbia University and the University of Cape Town.
A cross-sectional screening study was conducted in which women were screened with four independent tests for cervical carcinoma and its precursors. These tests included the DVI examination, a conventional Papanicolaou (Pap) smear, and testing for high-risk types of human papillomavirus (HPV) DNA using Hybrid Capture II™, and Cervicography™. The sample for the Pap smear was obtained first using an Accellon biosampler (MedScand, Hollywood, FL); an endocervical sample for HPV DNA testing was collected using a cone-brush that was placed in a Specimen Transport Medium (Digene Corporation, Gaithersburg, MD) collection tube. These samples were collected before acetic acid was applied for the DVI examination. A 35-mm photograph of the cervix was taken after reapplying 5% acetic acid to the cervix at the end of the examination (Cervigram™; National Testing Laboratories, St. Louis, MO).
HPV DNA was assayed at the University of Cape Town on a weekly basis using the second-generation Hybrid Capture II HPV DNA assay (Digene Corporation). The HPV DNA test was run according to the manufacturer's instructions using the microtiter plate-based format and probes for high oncogenic risk HPV types (i.e., types 16, 18, 31, 33, 35, 45, 51, 52, 56, 58, 59, and 68). HPV DNA determinations were read out as the ratio of relative light units (RLU) of HPV DNA in the sample to that of a positive control. All samples with RLU greater than the positive control (standard cut-off, > 1 pg/mL) were classified as positive. Pap smears were evaluated at the University of Cape Town Cytopathology Laboratory using the Bethesda System terminology, and cytologic diagnoses of low-grade SIL, high-grade SIL, or carcinoma were classified as positive. Cervigrams were evaluated at National Testing Laboratories and were reported using the company's standard terminology. Cervigram diagnoses of warrants colposcopy, low-grade SIL, high-grade SIL, or carcinoma were classified as positive.
Women were given the results of the DVI examination immediately. HPV DNA test results were available within 2–6 days after the initial examination, Pap smear results were available within 2 weeks, and Cervicography results were available within 8–12 weeks.
After meeting with a community health educator, written informed consent was obtained, and a brief oral questionnaire was administered. A nursing sister who had been trained in DVI and in obtaining Pap smears performed a screening examination on site. The nursing sister was trained to refer all women with acetowhite lesions as well as all women with cervical ulcers or cervical growths. Prior to performing DVI, 5% acetic acid was applied twice to the cervix over a 1-minute interval. The sister was trained to classify her findings on the DVI examination into five different categories (Table 1). The examination was performed in two steps: 1) application of acetic acid followed by examination of the cervix with the naked-eye (without magnification) followed by 2) a second application of acetic acid and examination with the use of magnification. A hand-held, ×4.5 magnification device with a green light (Aviscope; PATH, Seattle, WA) was used for magnification. The screening examination was designed to enforce a structured break between the inspection of the cervix with and without magnification. Findings from the first examination without magnification were recorded before and independent of the findings from second examination with magnification.
|Suspicious||Cervical ulcer or exophytic growth suspicious for cancer carcinoma|
|Definite lesion||Acetowhite lesion with well-circumscribed border|
|Nonconfluent scattered lesions||Focal, small, punctated areas of acetowhitening usually involving the transformation zone|
|Ill-defined lesions||Poorly circumscribed and faintly acetowhite|
|No lesion||No acetowhite lesion visible|
A nursing sister who had not performed any gynecologic or obstetric procedures in over 10 years was employed as the screening sister. Her initial training included attending colposcopy clinics, which were located at a tertiary teaching hospital in Cape Town, where she was trained under supervision to pass a speculum and to examine the lower genital tract. She was specifically trained to visualize the cervix, to identify the anatomic structures of the cervix (endocervix, ectocervix, and squamocolumnar junction), and to distinguish different types of white lesions that may be apparent after the application of 5% acetic acid. One-on-one tutorials were conducted in which normal anatomy, physiology, benign abnormalities of the cervix, and the pathogenesis of cervical carcinoma and its precursors were explained. This was followed by a 4-day intensive course that included hands-on clinical examinations as well as extensive review of photographs of normal and abnormal cervices. Although the training of the nursing sister took place over a 6-week period, the total number of hours spent amounted to an approximately 10-day training course. Feedback on the basis of colposcopic findings in terms of the appropriateness of DVI findings was provided intermittently over the course of the study.
Tests for Sexually Transmitted Diseases
Immediately after inserting the speculum, a Dacron swab was used to sample the vaginal fornices for the presence of Trichomonas vaginalis. The swab was placed in a pouch containing Diamond medium (In Pouch™; Biomed Diagnostics Inc., San Jose, CA), and the specimens were transported on a daily basis to the microbiology laboratory at the University of Cape Town for culture of T. vaginalis. The sample collected using the Accellon sampler was tested at Columbia University for N. gonorrhea and C. trachomatis DNA using the Hybrid Capture GC/CT DNA Assay followed by confirmatory GC specific and CT specific probes (Digene Corporation, Gaithersburg, MD). All women consented to provide a blood sample for anonymous, linked human immunodeficiency virus-1 (HIV-1) testing. All samples were screened for HIV-1 antibodies using the Abbott HIV kit on an Abbott AXSYM system (Abbott Laboratories, Chicago, IL). Positive results were confirmed with the Vironosticka HIV kit (Organon Teknika, Durham, NC). HIV results were linked to clinical data that were stripped of identifying information, as described previously.20
Women who had positive results from any one of the four screening tests were referred for on-site colposcopy and histologic sampling. All colposcopies were performed by the same colposcopist. Colposcopic lesions were graded using the Reid Colposcopic Index (RCI).21 Lesions that were classified as minor grade (RCI < 3) were biopsied, whereas lesions that were classified as significant (RCI ≥ 3) were excised electrosurgically using loop electrodes. An endocervical curettage was performed if no lesions were visible. Histologic samples were obtained from all clinically visible tumors.
All biopsies, loop excision specimens, and endocervical curettage specimens were evaluated at Columbia University, New York, and were masked to any clinical screening test information. The histologic results were reported using a two-tiered squamous intraepithelial lesion (SIL) terminology, with low-grade SIL equivalent to Grade 1 cervical intraepithelial neoplasia and high-grade SIL equivalent to Grade 2 and 3 cervical intraepithelial neoplasias combined.
Sensitivity, specificity, and positive and negative predictive values were estimated using standard formulae. True positive results were identified on the basis of histologic diagnoses of carcinoma, high-grade SIL, or low-grade SIL. True negative results were identified in patients who had no positive screening test results on any of the four screening tests used (DVI, cytology, HPV DNA testing, and Cervicography) or if one or more screening test was positive and had no evidence of carcinoma, high-grade SIL, or low-grade SIL on histologic sampling after colposcopy. Point estimates of sensitivity and specificity may be slightly inflated if some true positive results were missed by all four tests, but comparisons should be unaffected. Comparisons between the estimates of sensitivity and specificity of DVI using different criteria to define a screening test as positive and comparisons with cytology were tested with the McNemar test. Comparisons in estimated sensitivity and specificity by other patient characteristics were tested using chi-square tests.
Between January 1998 and November 1999, 2779 women presented for the study, and 2754 women met eligibility criteria and consented to enrollment. The sociodemographic profile of the study population is displayed in Table 2.
|No school||252 (9.2)|
|Some primary school||854 (31.0)|
|Some high school||1378 (50.0)|
|High school graduate||270 (9.8)|
|Shack on serviced site||1544 (56.1)|
|Shack on unserviced site||244 (8.9)|
|Employed (%)||1153 (41.9)|
|Married (%)||1250 (45.4)|
|Sexual partners last month (%)|
|Current contraceptive use (%)|
|Tubal ligation||654 (23.7)|
|Cigarette smoking||147 (5.3)|
|Median no. of pregnancies (range)||3 (0–15)|
|Median no. of live births (range)||3 (0–13)|
|Median age in yrs of first sexual intercourse (range)||17 (10–30)|
|Median no. of lifetime sexual partners (range)||2 (0–20)|
Of the 2754 women who enrolled in the study, 1207 women (44%) had a positive result on at least one of the four screening tests and were referred for colposcopy, of whom 1156 women (96%) underwent colposcopy. The other 51 women who were referred for colposcopy were lost to follow-up. Overall, two-thirds of the colposcopies (68%) were performed within 2 weeks of screening, and 91% were performed within 2 months of screening. Of the 2754 women who ere screened, histologically confirmed carcinoma was diagnosed in 21 women (0.8%), high-grade SIL was diagnosed in 96 women (3.5%), and low-grade SIL was diagnosed in 102 women (3.7%). There was no histologic evidence of disease in 932 women (33.8%). Another five women had inadequate histologic sampling, and no diagnosis could be made. The remaining 1547 women (56.2%) had no positive screening tests.
DVI after Application of Acetic Acid
DVI without magnification classified 683 women (25%) as positive, and DVI using ×4.5 magnification identified an additional 72 women, with 755 women (27%) classified as positive. Of the additional women who were identified as DVI positive with magnification, 20 of 72 women (28%) were classified with a definite lesion, and 42 of 72 women (58%) were classified with ill-defined lesions (Table 3). All women who were classified as positive without magnification also were classified as positive with magnification. Of the 755 lesions that were identified with magnification, 569 (75%) were classified as definite lesions, 47 (6%) were classified as nonconfluent scattered lesions, 131 (17%) were classified as ill-defined lesions, and 8 (1%) were classified as nonacetowhite exophytic growths suspicious for carcinoma. All but seven of the lesions that were identified without magnification were classified in the same category with and without magnification (Table 3).
|DVI classification without magnification||DVI classification with magnification|
|No lesion||Ill-defined lesion||Scattered||Definite lesion||Suspect carcinoma||Total|
Estimated Sensitivity and Specificity of DVI
The estimated sensitivity, specificity, and positive and negative predictive values of DVI examination based on four different criteria are shown in Table 4. The sensitivity for detection of high-grade SIL with and without magnification was 70% and 74%, respectively, when any lesion was classified as a positive result. When the definition of a positive DVI examination was restricted to those with definite lesions only, there was a statistically significant reduction in sensitivity for high-grade SIL both with and without magnification that was accompanied by an improved specificity for both categories (Table 4). The positive and negative predictive values were not affected appreciably by changes in the criteria for defining the DVI examination as positive (Table 4).
|Criteria for defining the DVI examination as positive||Sensitivity for carcinoma (n = 21 women)||Sensitivity for high-grade SIL (n = 96 women)||Sensitivity for low-grade SIL (n = 102 women)||Specificity (n = 2479 women)||Positive predictive value||Negative predictive value|
|Any lesion without magnification||85.7 (62.6–96.2)||69.8 (59.4–78.5)||57.8 (47.7–67.4)||79.3 (77.6–80.8)||12.9 (10.5–15.8)||96.3 (95.2–96.9)|
|Definite lesion without magnification||81.0 (57.4–93.7)||58.3 (47.8–68.2)b||48.0 (38.1–58.1)b||83.5 (82.0–84.9)b||13.7 (11.0–17.0)||95.5 (94.5–96.3)|
|Any lesions with magnification||85.7 (62.6–96.2)||74.0 (63.8–82.1)||59.8 (49.6–69.3)||76.6 (74.9–78.3)c||12.2 (10.0–14.9)||96.5 (95.6–97.2)|
|Definite lesion with magnification||81.0 (57.4–93.7)||58.3 (47.8–68.2)||50.0 (40.0–60.0)b||82.5 (81.0–84.0)b||13.1 (10.5–16.3)||95.6 (94.6–96.4)|
|Cytologyd||83.3 (57.7–95.6)||57.4 (46.8–67.4)||40.2 (30.8–50.4)||96.3 (95.5–97.0)||34.5 (28.0–41.6)||95.8 (94.9–96.5)|
Comparison with Cytology
The sensitivity of cytology for detection of high-grade SIL (57%) was significantly lower compared with the sensitivity of DVI with magnification (74%), but it did not differ significantly from the sensitivity of DVI without magnification or with a more stringent definition of a positive result (Table 4). If ASCUS was included in the definition of a positive cytology result, then the sensitivity of cytology to detect high-grade SIL increased to 64%, which was not significantly different from the sensitivity of DVI. Cytology had significantly better specificity compared with DVI across all conditions.
Analysis of DVI Successes and Failures
A cross tabulation of colposcopic findings, histology diagnosis, and DVI results is shown in Table 5. Of 25 women with negative DVI results and with histologically confirmed, high-grade SIL (i.e., false negative results or undercalls by DVI), 84.0% had an RCI ≥ 3, 12% had an RCI of 1–2, and 1 woman did not have a colposcopically evident lesion (Table 5). The patients with these false negative results had colposcopic findings similar to the colposcopic findings in patients with true positive results. Of 71 women with positive DVI results who had histologically confirmed, high-grade SIL (true positive results or correct calls by DVI), 64.8% had an RCI ≥ 3, 32.3% had an RCI of 1–2, and 1 woman had macroscopic carcinoma.
|Colposcopic findings||Diagnosis on colposcopically directed histologic sample (%)|
|Invasive carcinoma||High-grade SIL (CIN2, CIN3)||Low-grade SIL (CIN1)||Within normal limits|
|No lesion||0||0||1 (1.4)||1 (4.0)||1 (1.6)||2 (4.9)||123 (21.2)||95 (26.9)|
|1–2||0||0||23 (32.3)||3 (12.0)||35 (63.4)||26 (58.5)||408 (70.5)||234 (66.3)|
|3–8||2 (11.1)||3 (100)||46 (64.8)||21 (84.0)||25 (41.0)||13 (31.7)||46 (7.9)||24 (6.8)|
|Macroscopic carcinoma||16 (88.9)||0||1 (1.4)||0||0||0||2 (0.3)||0|
Of 579 women with positive DVI test results but with no histologic evidence of disease (i.e., false positive results or overcalls by DVI), 7.9% had an RCI ≥ 3, 70.5% had an RCI of 1–2 (suggestive of an insignificant lesion), 21.2% had no colposcopically evident lesion, and 2 women had colposcopic lesions suggestive of carcinoma (both of which were identified as tuberculosis on histologic review). The patients with these false positive results had colposcopic findings similar to the results seen in patients with true negative results (DVI negative with no evidence of disease) (Table 5).
Of the 3 malignancies that were missed by DVI, 1 of the 3 lesions was classified as a macroscopically visible tumor on colposcopy (established as Stage IIIb on later investigation), and the other 2 lesions were identified as lesions with RCIs of 3 and 8, respectively (both were diagnosed later as adenocarcinoma). All three patients with these lesions were treated with loop excision on their initial colposcopic examination.
Influence of Sexually Transmitted Infections on DVI Performance
The prevalence of sexually transmitted infections in the study population was high: 20% were positive for high-risk types of HPV DNA, 7% were HIV seropositive, 2% had endocervical infection with N. gonorrhea, 4% had endocervical infection with C. trachomatis, and 19% had culture-proven T. vaginalis infection. The prevalence of all of these infections declined significantly with increasing age (Table 6). Similarly, high-grade and low-grade SIL significantly declined with increasing age, but the prevalence of invasive carcinoma significantly increased.
|Age group prevalence (%)||Total no.a||Total %|
|35–39 yrs||40–49 yrs||50–65 yrs|
|Sexually transmitted infections|
|Any lesion with magnification||29.4||27.2||23.4b||2754||27.4|
There were no significant differences in the sensitivity and specificity of DVI by the presence or absence of N. gonorrhea, C. trachomatis, or T. vaginalis or by age, parity, contraceptive use, or across the duration of the study. The specificity of DVI was significantly lower among HIV seropositive women and showed a trend toward greater specificity among women age > 50 years. Stratifying the sample on reported premenopausal versus postmenopausal status rather than on age, a significantly greater specificity of DVI was observed among postmenopausal women.
This cross-sectional study of the performance characteristics of DVI as a primary screening test for cervical carcinoma and its precursors was performed among women of low socioeconomic status living in a low-resource environment. We confirmed the relatively high sensitivity of DVI for the detection of high-grade SIL and carcinoma, which we observed in our previous study in this population and which has been observed by others.14–19 In contrast to our prior study, which did not show any improvement of DVI performance with the use of a hand-held, ×2.5 magnifying device,18 the current study showed a marginal but nonsignificant improvement in sensitivity using a ×4.5, hand-held magnifying device, although this improvement was accompanied by a small, but significant reduction in specificity.
DVI using Different Definitions for an Acetowhite Lesion
In an attempt to improve the performance of DVI, the nursing sister was trained to describe acetowhite lesions as well-defined, ill-defined, or nonconfluent scattered lesions with and without magnification. Classifying acetowhite lesions as definite lesions without magnification was associated with a significant loss of sensitivity for high-grade SIL, from 70% for any acetowhite lesion to 58% for definite acetowhite lesions. This was accompanied by a significant improvement in specificity from 79% to 84%.
These findings have important implications for a screen-and-treat policy and for the standardization of training methods. The improved specificity using a stricter definition of a positive DVI test would reduce the number of women without cervical disease undergoing treatment. However, the reduction in sensitivity would result in the detection of fewer women with high-grade SIL and may reduce the impact on cervical carcinoma prevention in the screened population. A sensitivity of DVI for high-grade SIL of approximately 70% (with a positive test defined as any acetowhite lesion and without magnification) was equivalent to cytology in our study and compares favorably with the performance of cytology in many settings.22 The choice between increased sensitivity or increased specificity ultimately will be influenced by the safety and efficacy of screen-and-treat programs. If the outcome of screening and treating women on the basis of a positive DVI test were associated with a low complication rate, high acceptability among women, and a significant reduction in the prevalence of cervical carcinoma and its precursors, then a higher sensitivity at the expense of a lower specificity would be preferable in resource-poor settings.
Sankaranarayanan et al.15 trained paramedical personnel to screen 3000 Indian women using DVI (without magnification) and cytology. In that study, the screeners were trained to grade the acetowhite lesions as positive only if a distinct acetowhite area was noted (equivalent to a definite lesion in our study). If the acetowhitening was doubtful or faint, then DVI was considered negative. A positive DVI test, defined as distinct lesions, detected 90.1% of the histologically confirmed high-grade SIL compared with 58% in our study when a similar definition of a positive DVI test was used. Women with indistinct acetowhite lesions did not undergo colposcopy; therefore, the disease prevalence in this group is unknown. Using the definition of a distinct lesion to define a positive DVI test, 10% of the Indian women in the study by Sankarayanana et al. had positive test results. By comparison, using a definite lesion to define a positive DVI test in the current study led to classifying 21% of the women screened with positive test results. Larger studies and comparisons with different training methods need to be evaluated to establish whether the different results obtained by the two studies relate to different methods of training or to biologic factors related to the screened population.
Using any acetowhite lesion as the definition of a positive test without magnification would conceivably reduce the training and equipment requirements of the test, because less discrimination is required by the screener, and a magnifying device is not required. This becomes especially important in very low-resource settings, such as in rural clinics, where one nursing sister may be required to screen large numbers of women. The subjective nature of DVI suggests that the definition of a positive test should be made as simple as possible. Adequate training and supervision of personnel will be critical to implementation of DVI-based screening programs.
The performance of DVI was evaluated against the findings at colposcopy and after histologic sampling. Most of the patients with true high-grade lesions that were missed by DVI had significant lesions detectable on colposcopy. In these women, it appears that DVI simply failed to detect what were colposcopically unambiguous, high-grade lesions. There are a number of possible explanations for this. These include the subjective nature of the test, failure to apply sufficient acetic acid for an adequate period, poor visualization of the cervix, or screener fatigue. In addition, the DVI examination in this study was performed after the Pap smear was taken and the endocervix was sampled, which may have resulted in stripping of the cervical epithelium. In some instances, the colposcopic examination was delayed in women with clinically evident lower genital tract infections or severe atrophy so that these conditions could be treated, thus optimizing the conditions for the colposcopic examination compared with the DVI screening examination. Finally, colposcopy was performed only when it was known that a woman had one or more abnormal screening tests, which may have influenced the colposcopist's interpretation of the findings (i.e., a referral vs. a screening examination), although it should be noted that, overall, 44% of all women enrolled in this study underwent colposcopy.
Influence of Sexually Transmitted Infections on the Performance of DVI
We hypothesized when designing this study that low specificity of DVI may be attributed to the high rates of lower genital tract infection found in the study population. Interpreting the result of a DVI examination in a woman with an inflamed cervix is difficult, and it was our impression that this may have led the nurses conducting the screening to err on the side of an overcall rather than an undercall. This hypothesis was not supported by our data. The sensitivity and specificity of DVI for patients with high-grade SIL and carcinoma was not significantly different among women who were positive or negative for infection with T. vaginalis, N. gonorrhea, or C. trachomatis. These data suggest that pretreating women with antibiotics prior to the examination would not help the performance of DVI as a primary screening test.
The sensitivity of DVI for patients with high-grade SIL and carcinoma did not differ significantly by HIV status, but the specificity was lower among HIV positive women compared with HIV negative women. The higher false positive rate of DVI among HIV positive women may be related to the presence of other, undetermined infections or to a higher prevalence of high-grade lesions that may have been missed by colposcopically-directed biopsies in this group. Information on CD4 counts or clinical staging was not available for the HIV positive women.
This study confirmed the utility of DVI as a primary screening test in low-resource settings. Magnification with a ×4.5, hand-held device did not improve the detection of cervical carcinoma precursors and significantly decreased specificity. Applying stricter definitions to define a positive DVI test, i.e., defining a positive test as a definite lesion with a well-circumscribed border, significantly reduced sensitivity but increased specificity, reducing the number of women identified with positive test results. The presence of sexually transmitted infections did not significantly alter the performance of DVI. The only exception was a decrease in the specificity of DVI among HIV positive women. The sensitivity of DVI, however, was similar among HIV positive women and HIV negative women.
The ultimate utility of DVI as a primary screening test in low-resource settings will depend on the outcome of screen-and-treat screening programs. In countries where DVI is the only feasible method of screening due to the barriers imposed by poor infrastructure and limited resources, screening and immediately treating women with a positive DVI test offers an attractive alternative to the traditional methods of cytology, colposcopy, colposcopically directed biopsies, and treatment for women with histologically confirmed disease. The cost effectiveness of DVI screening compares favorably with other strategies.23 In detecting 70% of the cases of high-grade SIL in the screened population, the simplest definition of a positive DVI test without magnification is a feasible alternative to cytology in resource-poor settings. The safety, acceptability, and efficacy of screening and treating women on the basis of a positive DVI test currently is being evaluated in a prospective, randomized trial.
Data entry was provided by the Medical Research Council of South Africa. HPV DNA testing at University of Cape Town was performed by Patricia Papier and Anneli Visser; all cytology was read at the University of Cape Town Cytology Laboratory; and on-site histology was read by Dr. Helen Wainwright, University of Cape Town, Department of Anatomical Pathology.
- 1IARC Working Group on Cervical Cancer Screening. Summary chapter. In: HakamaM, MillerAB, DayNE, editors. Screening for cancer of the uterine cervix. Lyon: International Agency for Research on Cancer, 1986: 133–142.
- 2Trends in mortality from cervical cancer in the Nordic countries: association with organised screening programmes. Lancet. 1987; 30: 1247–1249., , .
- 3IARC Working Group on Evaluation of Cervical Cancer Screening Programmes. Screening for squamous cervical cancer: the duration of low risk after negative result of cervical cytology and its implication for screening policies. Br Med J. 1986; 293: 659–664.
- 4A screening programme for cervical cancer that worked. Cancer Surveys. 1988; 17(3): 403–416., .
- 5HakamaM, MillerAB, DayNE, editors. Screening for cancer of the uterine cervix. IARC scientific publication no. 76. Lyon: International Agency for Research on Cancer, 1986: 47–60.
- 7Screening for cervix cancer in developing countries. In: MillerAB, ChamberlainJ, DayNE, HakamaM, ProrokPC, editors. Cancer screening. Cambridge: Cambridge University Press, 1991: 184–198..
- 9Cancer in developing countries. Cancer Surveys. 1994; 20: 519–561..
- 21Colposcopy of cervical preinvasive neoplasia. In: SingerA, editor. Premalignant lesions of the lower genital tract. New York: Elsevier, 87–116..