Accuracy of visual screening for cervical neoplasia: Results from an IARC multicentre study in India and Africa
Version of Record online: 24 MAR 2004
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 110, Issue 6, pages 907–913, 20 July 2004
How to Cite
Sankaranarayanan, R., Basu, P., Wesley, R. S., Mahe, C., Keita, N., Mbalawa, C. C. G., Sharma, R., Dolo, A., Shastri, S. S., Nacoulma, M., Nayama, M., Somanathan, T., Lucas, E., Muwonge, R., Frappart, L., Parkin, D. M. and for the IARC Multicentre Study Group on Cervical Cancer Early Detection (2004), Accuracy of visual screening for cervical neoplasia: Results from an IARC multicentre study in India and Africa. Int. J. Cancer, 110: 907–913. doi: 10.1002/ijc.20190
- Issue online: 26 MAY 2004
- Version of Record online: 24 MAR 2004
- Manuscript Accepted: 27 JAN 2004
- Manuscript Revised: 21 JAN 2004
- Manuscript Received: 22 OCT 2003
- Bill & Melinda Gates Foundation
- visual inspection;
- acetic acid;
- Lugol's iodine;
- cervical neoplasia;
Visual inspection-based screening tests, such as visual inspection with 4% acetic acid (VIA) and with Lugol's iodine (VILI), have been proposed as alternatives to cytology in mass screening programs. To date, there is only limited information on the accuracy of these tests in detecting High-grade Squamous Intraepithelial Lesions (HSIL). Eleven cross-sectional studies involving 56,939 women aged 25–65 years were conducted in Burkina Faso, Congo, Guinea, India, Mali and Niger to evaluate the accuracy of VIA and VILI performed by health workers. A common protocol and questionnaire was used. For final diagnosis, all women were investigated with colposcopy and biopsies were taken when necessary. Data from the studies were pooled to calculate sensitivity, specificity and predictive values of the tests for the detection of HSIL. Of the screened women, 16.1% and 16.4% were positive on examination using, respectively, VIA and VILI; 1,063 were diagnosed with HSIL. The pooled sensitivity, specificity, positive and negative predictive values for VIA were 76.8% (95% CI: 74.2–79.4%), 85.5% (95% CI: 85.2–85.8%), 9.4% (95% CI:8.8–10.8%) and 99.5% (95% CI:99.4–99.6%), respectively. The values were 91.7% (95% CI: 89.7–93.4%), 85.4% (95% CI: 85.1–85.7%), 10.9% (95% CI: 10.2–11.6%) and 99.8% (95% CI:99.7–99.9%), respectively for VILI. The range of sensitivity and specificity for VIA was 56.1–93.9% and 74.2–93.8%, respectively, between studies and were 76.0–97.0 % and 73.0–91.3% for VILI. VILI had a significantly higher sensitivity than VIA in detecting HSIL, but specificity was similar. VILI appears to be a more accurate visual test for use in screening and treatment programs in low-resource settings. © 2004 Wiley-Liss, Inc.
Cancer of the cervix uteri is the most common cancer among women in the developing world, and 60% of all cases occur in the countries of sub-Saharan Africa, South Asia and Latin America.1 Cervical cytology screening programs in the developed countries of Europe and North America have been followed by substantial reduction in disease burden. However, screening programs do not exist in most developing countries, and, in those few countries where they do, in only a few instances has a significant reduction in incidence and mortality from cervical cancer been observed.2, 3 The success of prevention programs based on cytology screening depends upon availability of adequate technical personnel, good quality cytology smears, adequate laboratory services with internal and external quality control, a good organisation to ensure high coverage of the target population with screening and diagnosis, treatment and follow-up of screen-positive women. In many developing countries, technical, manpower and financial resources are inadequate to provide the necessary infrastructure. The difficulties in introducing cervical cytology screening in low-resource settings have prompted the evaluation of simple and inexpensive noncytological methods of detecting precursor lesions, such as visual inspection of the cervix after application of 3–5% acetic acid (VIA) or after the application of Lugol's iodine (VILI).
The International Agency for Research on Cancer (IARC) conducted a number of cross-sectional studies in India and Africa, in collaboration with the national institutions, to estimate and compare the sensitivity and specificity of VIA and VILI to detect high-grade cervical intraepithelial neoplasia (HSIL) and to study the accuracy of these screening tests by subgroups of study centres, region, type of study population, age group, education, marriage, number of pregnancies, menopausal status and visibility of squamocolumnar junction (SCJ).
MATERIAL AND METHODS
During the years 1999 to 2003, 11 cross-sectional studies of cervical cancer screening were conducted in 9 centres in Africa and India. Similar protocols were used in all centres, which were approved by the institutional review boards and ethics committees of IARC and the collaborating national institutions. The study design is shown in Figure 1. All women participating in the study were tested sequentially with index screening tests VIA and VILI and the reference investigation, namely, colposcopy with or without biopsy; biopsies were taken in women with abnormal or suspicious findings on colposcopy. All the testing was done independently and blinded, so results of previous tests were unknown. True positive disease was defined as HSIL and the reference standard for final diagnosis was histology, or colposcopy if histology was not available or inconclusive.
The study participants were apparently healthy, ambulant, asymptomatic women who were aged 25–65 years with an intact uterus and with no past history of cervical neoplasia recruited from 9 locations in 6 countries: Bamako (Mali), Brazzaville (Congo), Conakry (Guinea), Jaipur (India), Kolkata (India), Mumbai (India), Niamey (Niger), Ouagadougou (Burkina Faso) and Thiruvananthapuram (India). Women who had inconclusive colposcopy with no biopsy were excluded from the study. Consecutive series of participants defined by selection criteria were opportunistically recruited from the resident populations in the cities and surrounding districts, except in 2 studies (Kolkata B and Thiruvananthapuram B) where the participants were from defined geographical locations (community-based). Women were encouraged to attend open access screening clinics by publicity and information delivered through person to person communications, group meetings organised by voluntary associations, printed media, radio (Congo, Guinea, Mali) and television messages (Mali).
Women attended open access screening clinics conducted on scheduled days. A female health worker explained the purpose of the study and the screening tests and interventions to participants attending the opportunistic clinics, and leaflets in the local language on cervical cancer prevention and early detection were distributed. A printed consent form was read out to the study participants and their signature or left thumb impression was obtained. The health worker then administered a structured questionnaire to collect information on sociodemographic and reproductive variables. After the interview, screening tests and diagnostic investigations were carried out.
VIA and VILI were performed by trained female health workers; overall 51 health workers were involved in testing. The health workers had a variety of different educational qualifications: auxiliary nurse midwives (n=33) (Burkina Faso, Congo, Guinea, Mali and Niger), registered nurses (n=2) (Thiruvananthapuram), cytotechnicians (n=2) (Thiruvananthapuram), university graduates in science and arts subjects (n=9) (Jaipur and Kolkata) or high-school graduates (n=5) (Mumbai). All were trained in the performing and reporting of VIA and VILI during a 5-day intensive course, using a training manual prepared by IARC.4 The training included lectures, discussions, review of photographs of normal and abnormal cervix, as well as practical experience observing examinations of 150–200 volunteer subjects. All received instructions in treatment of lesions by cryotherapy. Refresher courses (1–2 days) were given for all health workers during the study period.
In the screening clinic, each participant was tested with VIA, followed by colposcopy and VILI; the woman was placed in a modified lithotomy position on an examination couch with leg rests, knee crutches or stirrups. The health worker exposed the cervix by inserting an unlubricated bivalve vaginal speculum, examined the cervix with the aid of a halogen focus lamp and identified the SCJ. Four percent acetic acid was then applied to the cervix using a cotton swab and excess mucus was cleared. VIA findings were visible 1 min after application of the acetic acid and the test results were then documented. Categorisation of VIA test results is presented in Table I. Number of test providers in each study site is given in Table II.
|VIA findings were categorised as negative when any of the following findings were observed:|
|• No acetowhite lesions, or faint, ill-defined, bluish white or doubtful lesions|
|• Acetowhitening on cervical polyps or on nabothian cysts|
|• Dot- or streak-like acetowhitening on the cervix|
|• White line-like prominent squamocolumnar junction (SCJ) after application of acetic acid|
|• Angular or geographic acetowhite lesions far away from the SCJ or the external os, if SCJ was not visible (satellite lesions)|
|VIA was categorized as positive when any of the following were observed:|
|• Well-defined, opaque, acetowhite lesions touching the SCJ or the external os, if SCJ was not visible|
|• A large circumferential acetowhite lesion surrounding the external os|
|• Pre-existing wart or leukoplakia turning intensely white after application of acetic acid|
|• Ulceroproliferative growth turning densely acetowhite after application of acetic acid|
|VILI was reported as negative when any of the following were observed:|
|• Normal cervix where squamous epithelium turned mahogany brown or black and, when the SCJ was visible, the columnar epithelium beyond it did not change color|
|• Scattered or diffuse ill-defined or patchy, non- or partial-iodine uptake areas in the transformation zone or all over the cervix, not restricted to the transformation zone|
|• Thin, yellow, noniodine uptake areas with angular, or digitating margins, resembling geographical areas, distant from the SCJ|
|VILI was categorized as positive when any of the following were observed|
|• Well-defined dense, thick, bright, mustard- or saffron-yellow, iodine non-uptake areas in the transformation zone, touching the SCJ|
|• Circumferential yellow non-iodine uptake area around the external os|
|• Ulceroproliferative growth turning yellow after the application of iodine solution.|
|Study site||Number of test providers2||Number of participants||> 40 years old (%)||VIA positivity (%)||VILI positivity (%)||HSIL (%)|
Following VIA, all women were examined by colposcopy, with the colposcopist being unaware of the VIA test findings. Colposcopy was carried out by medical staff (27 gynaecologists and 7 non-gynaecologists). They received intensive training in a 15-day course using a manual prepared by IARC.5 Colposcopic assessment of the cervix in our study followed the classic technique of assessing abnormal areas after application of acetic acid followed by Lugol's iodine, which also allowed assessment of Reid's score for evaluation of colposcopic abnormalities.5 Biopsies were directed only after colposcopic assessment with Lugol's iodine. The colposcopic assessment entered in the study proforma was based on findings of colposcopy after application of both acetic acid and iodine. One-day refresher sessions took place every 4–6 months during the study. All were trained in treatment methods, such as cryotherapy and loop electrosurgical excision procedure (LEEP).
When the colposcopic examination after applying acetic acid was completed, a second health worker (blinded to the findings on VIA and colposcopy) applied Lugol's iodine to the cervix and, after naked eye examination, recorded the test result. Categorisation of VILI test results is presented in Table I. All the testing was done independently and blinded, so results of previous tests were unknown.
Following VILI, the same doctor repeated colposcopy, blinded to the findings of VILI, to assess the cervix again following iodine application; then punch biopsies were obtained from colposcopically assessed abnormal areas on the cervix. Pictorial atlases of VIA, VILI and colposcopy were used for consultation during test procedures. Biopsy specimens were fixed in 10% formalin and were processed and reported by the pathology laboratories of the collaborating national institutions. Histopathologists were provided with a brief description of the colposcopic findings. The pathologists (n=17) and laboratory technicians (n=14) responsible for processing and reporting on biopsy material received special refresher courses prior to the commencement of screening. Internal and external quality control measures were introduced in the pathology laboratories. A sample of histopathology slides was reviewed in a reference laboratory for inter-observer agreement.
The final reference diagnosis was based on histopathology findings or, if no biopsy was taken or inconclusive, on colposcopy. For neoplasia, histopathology (if tissue samples are appropriately taken) is considered to identify the true disease most accurately. For cervical neoplasia, colposcopy and directed biopsy are commonly used and are widely accepted reference standard (“gold standard”) investigations.6, 7, 8 However, when histology is not available, or is not conclusive, colposcopic diagnosis may be accepted as the true disease.7, 8
HSIL were considered as true positive disease to calculate sensitivity, specificity and predictive values of the screening tests in order to obtain conservative estimates of accuracy and to avoid spectrum-bias. Data entry took place in each study site using standard computer software. Checks for consistency and analysis were carried out in IARC using STATA 7.0. Sensitivity, specificity and predictive values and their 95% confidence intervals were calculated using 2 × 2 tables and standard formulae.9
Our study was carried out during 1998 to 2003; the recruitment was initiated in March 1998 and completed in March 2003. A total of 56,939 women were recruited, of whom 12,768 were recruited from 2 community surveys in defined geographical regions; 54,981 women included in the final analysis after excluding 1,958 women (3.4%) who had inconclusive colposcopy and no biopsy. Most of the excluded women were postmenopausal (52.5% vs. 9.8% of the included women) and did not have a visible squamocolumnar junction (SCJ) (78.8% vs. 11.6% of the included women). Number of included participants in each study is given in Table II. The distribution of age group, education, marital status, number of pregnancies, menopausal status and visibility of SCJ among the included women is given in Table III. All the included women (n=54,981) were tested with VIA and had colposcopy. Only 49,080 included women who received VILI as participants in Kolkata A study did not receive VILI. 9,590 (17.4%) women included in the study received biopsy and the proportion of women who received biopsy varied from 8.9% to 37.2% between the study centres.
|Subgroup||Number||Detection of HSIL (%)||VIA||VILI|
|Positivity (%)||Sensitivity (%)||Specificity (%)||Positivity (%)||Sensitivity (%)||Specificity (%)|
|Value||95% CI||Value||95% CI||Value||95% CI||Value||95% CI|
|Number of pregnancies|
VIA was positive in 8,848 (16.1%) of 54,981 tested women and VILI was positive in 8,060 (16.4%) of 49,080 tested women (Fig. 1); the range of test positivity was 6.6% to 27.4% between the study centres for VIA and 9.3% to 28.7% for VILI (Table II). Test positivity by region (India or Africa), type of study population, visibility of SCJ, age, menstrual status, education, marital status and pregnancies are shown in Table III. The detection rate of HSIL by study centres is given in Table II and by participant characteristics in Table III.
A cross-tabulation of the results of the index screening tests by the results of the reference standard is given in Table IV. Among women tested with VIA, the final diagnosis was normal in 50,280 (91.4%) women, low-grade squamous intraepithelial lesion (LSIL) in 3,369 (6.1%), HSIL in 1,063 (1.9%) and invasive cancer in 269 women (0.5%). Among women tested with VILI, the final diagnosis was normal in 44,892 (91.5%) women, LSIL in 3,019 (6.2%), HSIL in 938 (1.9%) and invasive cancer in 231 women (0.4%). Of the 938 HSIL detected among 49,080 women tested with both VIA and VILI, both tests detected 732 (78.0%), VIA detected 15 lesions (1.6%) that had been missed by VILI and VILI detected 129 lesions (13.8%) that had been missed by VIA. Both tests missed 62 (6.6%) lesions; 59.1% of the LSIL, 80.9% of the HSIL and 94.1% of the cancer cases were confirmed by biopsy. No adverse events such as bleeding and allergic reactions occurred after administration of screening tests and the reference standard.
When a random sample of 736 histopathology slides from study sites (classified into 6 diagnostic categories: 251 normal, 88 atypia/HPV changes, 178 LSIL, 135 HSIL, 66 invasive cancer and 18 inconclusive cases) were reviewed in our reference pathology laboratory in Lyon, the agreement rate between the study pathologists and the reference pathologists was 70.8% (kappa: 0.62). The agreement rates differed by study sites: it varied from 45.7% to 94.3% between the study centres; the kappa value varied between 0.27 to 0.92.
Estimates of sensitivity, specificity and predictive values of VIA
The final disease status of women tested by VIA is given in Table IV. The sensitivity of VIA in detecting HSIL was 76.8% (95% CI: 74.2–79.4%) and the specificity was 85.5% (95% CI: 85.2–85.8%) (Table V). The positive and negative predictive values were 9.4% (95% CI: 8.8–10.8%) and 99.5% (95% CI: 99.4–99.6%), respectively. The sensitivity varied from 56.1% to 93.9% among the study centres; specificity ranged from 74.2% to 93.8% (Table V, Fig. 2). The sensitivity, specificity and their 95% confidence limits by sub-groups of participant characteristics (age group, type of study population, type of test providers, visibility of SCJ and menopausal status) are given in Table III. The sensitivity of VIA in detecting HSIL and invasive cancer was 79.3%(95% CI: 76.9–81.4%) and the specificity was 85.5%(95% CI: 85.1–85.8%); the positive predictive value was 11.8%(95% CI: 11.1–12.5%) and the negative predictive value was 99.4%.(95% CI: 99.3–99.5%).
|Value||95% CI||Value||95% CI||Value||95% CI||Value||95% CI|
Estimates of sensitivity, specificity and predictive values of VILI
The final disease status of women tested by VILI is given in Table IV. The sensitivity of VILI in detecting HSIL was 91.7% (95% CI: 89.7–93.4%) and the specificity was 85.4% (95% CI: 85.1–85.7%) (Table V). The positive and negative predictive values were 10.9% (95% CI: 10.2–11.6%) and 99.8% (95% CI: 99.7–99.9%), respectively. The sensitivity varied from 76.0% to 97.0% among the study centres; specificity ranged from 73.0% to 91.3% (Table V, Fig. 2). The sensitivity, specificity and their 95% confidence limits by subgroups of participant characteristics are given in Table III. The sensitivity of VILI in detecting HSIL and invasive cancer was 92.2% (95% CI: 90.5–93.7%) and the specificity was 85.4%(95% CI: 85.0–85.7%); the positive predictive value was 13.3%(95% CI: 12.5–14.0%) and the negative predictive value was 99.8% (95% CI: 99.7–99.8%).
The suitability of a screening test relates not only to its simplicity and safety but also to its accuracy, as measured by sensitivity and specificity.10 These can be evaluated in cross-sectional studies, provided that a suitable reference test is carried out, to distinguish the subjects who are truly positive from those who are negative. A study will suffer from “verification bias” if the reference investigation is restricted to test positive individuals only or additionally to a sample of test negative persons.9 Such bias is known to inflate sensitivity estimates and may be avoided if all individuals receive the reference investigation, irrespective of the test results.11
The characteristics of VIA as a screening test have been investigated in several cross-sectional studies. VILI is similar to the Schiller iodine test advocated in the 1930s.12 Although iodine solution has been widely used by colposcopists to assess and to delineate lesions before directing biopsy and treatment, it has not been used as a screening test for many years. Our experience in using Lugol's iodine during colposcopy, and the observation that health workers found it easier to detect the colour patterns produced by iodine staining, rather than the whitening produced by acetic acid, led us to re-evaluate VILI as a screening test.
The results reported here represent the largest experience to date on the test qualities of VIA and VILI with minimal verification bias. A common study protocol and uniform training of the workers in testing and diagnosis allowed pooled data analysis and comparison of results between centres. The large sample size permitted the evaluation of test characteristics by sub-groups of variables of interest in programme settings.
VIA and VILI were carried out by different test providers independently and blinded; no consultations between the health workers and the colposcopists were permitted, and assessments were carried out blind to the findings of other tests. These measures ensured the independent assessment of the 2 tests. VILI was performed following the application of acetic acid for colposcopy for logistical reasons since the epithelial staining following iodine remains for a long time (up to 30–45 min), and introducing VILI early in the sequence of tests would have greatly prolonged the time needed to examine each participant, as the women need to wait for an additional hour for the iodine stain to disappear before acetic acid can be applied for colposcopy.
Our study also reflected the heterogeneous service delivery conditions that exist in real programme settings, such as a large number of test providers with different educational backgrounds, large number of colposcopists with different lengths of experience and large number of pathologists and the varying levels of development of health services. While the study locations in India have moderately developed health care services, the African study sites are in countries with some of the least developed health care systems in the world. Considerable attention was paid to ensure good quality colposcopy and histopathology reporting in our studies. Prior to our study, no cervical cancer early detection service had existed in any of the study centres in the African countries included in our study. The project provided the opportunity to train a core group of service providers in all the countries included in the study and to improve histopathology facilities and reporting, particularly in the African study sites.
Previous cross-sectional studies7, 13, 14, 15, 16, 17, 18, 19, 20 have reported on the accuracy of VIA, but all except 2 studies7, 13, 14 suffered from verification bias. Since these studies were conducted in a period when interest in VIA was evolving, different definitions were used for categorisation of VIA results. True positive disease in these studies included both HSIL and invasive cancer. Sensitivity estimates may be inflated and biased when cancer is included in the true positive disease, since cancer is easier to detect by the naked eye; it is also possible that patients with advanced cancer might have been recruited into the studies. In spite of these potential drawbacks, the above studies established the value of VIA as a promising alternative to cytology.
If the true positive disease was defined as HSIL and invasive cancer (as in published studies of VIA), the sensitivity, specificity, positive and negative predictive values of VIA in our study were 79.3% (95% CI: 76.9–81.4%), 85.5% (95% CI: 85.1–85.8%), 11.8% (95% CI: 11.1–12.5%) and 99.4% (95% CI: 99.3–99.5%), respectively; these corresponding values for VILI in our study were 92.2% (95% CI: 90.5–93.7%), 85.4% (95% CI: 85.0–85.7%), 13.3% (95% CI: 12.5–14.0%) and 99.8% (95% CI: 99.7–99.8%), respectively. The sensitivity of VIA for HSIL and cancer in our study was similar to those reported from Harare, Zimbabwe7 (76.7%, 95% CI: 70.3–82.3%) and Shanxi, China13, 14 (71%, 95% CI:60–80%), but the overall specificity in our study was substantially higher (64.1%, 95% CI: 61.9–66.2% in Zimbabwe and 74%, 95% CI: 72–76% in China). The higher specificity seems to be due to a stringent definition of VIA categories and to the rigorous training of the health workers in our study.
There are no published results for VILI in recent decades with which to compare our results, but VILI had significantly higher sensitivity and specificity than those reported for VIA in the Zimbabwe7 and Chinese13, 14 studies.
The sensitivity of both VIA and VILI in our study is higher than that reported for cytology in Harare, Zimbabwe7 and in recent meta-analyses of Pap smear screening.21, 22 In a meta-analysis of 62 studies of cytology conducted between 1984 and 1992, the mean sensitivity was found to be 58% (range 11–99%) and mean specificity 68% (range 14–97%).21 In another meta-analysis involving 94 studies, similar values were observed.22 The sensitivity and specificity of VILI in our study is similar to that reported for HPV testing by HC II method in Shanxi, China13 and in other countries23, 24, 25, 26, 27 and higher than that reported from Harare, Zimbabwe.28 Currently, the cost of HPV tests are prohibitively high for use in developing countries.
There are several advantages to VIA and VILI as screening tests in low-resource settings, i.e., they are simple, inexpensive tests and do not require a sophisticated laboratory infrastructure. The visual test providers can be trained in much shorter training periods (5–10 days) as compared to the training of cytotechnicians (12–24 months). The immediate availability of the test result permits diagnostic procedures (colposcopy with or without biopsy) and treatment to be performed at the time of the screening visit. This avoids the inevitable loss to follow-up that occurs when women must be recalled following positive cytology or HPV tests. Our results clearly demonstrate that both VIA and VILI can identify the majority of cases of HSIL, although the variation in sensitivity between the study centres illustrates the provider-dependent nature of both tests. The high negative predictive values of both tests means that women who are test negative can be reassured with confidence that they are disease free. Since the sensitivity of visual tests (particularly VILI) is high, repeated testing at short intervals as with other tests with low sensitivity, such as cytology, can be avoided. It should be emphasised that high-levels of sensitivity is critically important to ensure success of prevention programs in low-resource settings, where once in a life time or less-frequently repeated (e.g., at 10-year intervals), low-intensity screening seems to be the most feasible prevention approach.
However, the specificity of VIA and VILI are still low, compared to that of good quality conventional cytology. Some 15% of women will be false positives, due to low specificity (around 85%), and will require diagnostic workup (e.g., colposcopy/biopsy), or may receive treatment unnecessarily, if screen-positive women are treated without diagnostic triage in a single-visit “screen and treat” approach. It remains to be seen if specificity can be improved without substantial loss in sensitivity, by standardising reporting categories and training strategies. Almost all information on the test performance of visual tests come from clinical research settings and there is currently limited information on how these tests will perform when introduced for wide-spread routine use in real-life settings. Since visual tests are essentially subjective tests, there is some concern on their reproducibility, particularly in routine practice. A fair degree of agreement (agreement rate 64.5%, kappa value 0.38) was observed between the master trainer and test providers in our study using 36 cervical photographs after acetic acid application; the agreement rates varied 52.8 to 80.2% (range in kappa values: 0.15 to 0.65) among the study centres. In one recently completed study using photographs of acetic acid impregnated cervix, a moderate to substantial degree of agreement was observed among expert trainers in different study settings.29 Quality assurance of visual screening in field conditions also poses a major challenge. Close monitoring of test positivity and disease detection rates as well as periodic retraining are essential to maintain good standards of visual testing under field conditions.
The reference investigation in our study consisted of histopathology or, if no biopsy was taken or inconclusive, colposcopy. This means that a colposcopic diagnosis of HSIL, in the absence of a biopsy, was also considered as true positive disease: 80.9% of HSIL in our study was confirmed by histology and the rest by colposcopy only. If we excluded the HSIL cases confirmed by colposcopy only from the study and recalculated the screening test characteristics, the results were similar to those obtained using the above reference standard for all practical purposes: the sensitivity of VIA was 74.9% (95% CI: 71.8–77.8%) and that of VILI was 91.1% (95% CI: 88.8–93.0%); the specificity of VIA was 85.9% (95% CI: 85.6–86.2%) and that of VILI was 86.2% (95% CI: 85.9–86.5%). The positive predictive values of VIA and VILI were 8.1% and 9.7% respectively; the negative predictive values were 99.5% and 99.8% respectively. These findings validate the choice of reference standards used in our study.
There is yet no universally accepted uniform definition of test results for VIA and VILI. It is an urgent priority to establish standard definitions and approaches for these tests to avoid the considerable variation in the way they are applied and interpreted. VILI is clearly more sensitive than VIA (p<0.001), a finding that is consistent between the study sites and according to different participant characteristics in our study. The specificity of the 2 tests is similar (p=0.16). No allergic reaction to iodine has been noted along the study. The statistically significant difference observed between the sensitivity estimates of VIA and VILI in our study could be attributed to the large sample size of this pooled analysis, but the consistency of sensitivity and the consistent direction of difference within each study site (Fig. 2) means that the difference in sensitivity is truly clinically and programmatically important. Thus VILI appears to be a more accurate and clinically useful visual screening test than VIA.
The balance of costs vs. effectiveness of visual screening has been addressed in model-based cost effectiveness studies.30, 31 These suggest that visual inspection is likely to be a more cost-effective method of screening than either cytology or HPV testing, given a wide range of assumptions about natural history and costs of different procedures. Eventually, the results of on-going randomised intervention trials of visual screening32, 33, 34 will provide real data on relative costs and effectiveness, so that the place of visual screening in routine programmes in low-resource settings can be assessed based on the best possible evidence.
We thank the following colleagues for their technical support in training our project staff and investigators in the study sites: Histopathology: Dr. P. Couty, Hotel Dieu, Lyon, France; Colposcopy, cryotherapy, LEEP: Dr. J.W. Sellors, Program for Appropriate Technology in Health (PATH), Seattle, WA; Dr. S.S. Sundar, John Radcliffe Hospital, Oxford, UK; Dr. J. Liaras, Dr. M. Marien, Dr. P. Mathevet, Hôpital Edouard Herriot, Lyon, France; Dr. R. Ananth, Chennai, India; Dr. S. Ratnakarn, Faculty of Medicine, University of Khon Kaen, Khon Kaen, Thailand and Dr. S. Koonsaeng, National Cancer Institute, Bangkok, Thailand. The authors are thankful to Mrs. E. Bayle for her assistance in preparing this article.
- 4A practical manual on visual screening for cervical neoplasia: IARC Technical Publication No. 41. Lyon: International Agency for Research on Cancer, IARC Press, 2003., .
- 5Colpscopy and treatment of cervical intraepithelial neoplasia: a beginners' manual: IARC non-serial publication. Lyon: IARC Press, 2003., .
- 9Santos SilvaI., ed. Cancer epidemiology, principles and methods. Lyon: International Agency for Research on Cancer, 1999.
- 10Principles and Practice of Screening for Disease. Public Health Papers No. 34. Geneva: World Health Organization, 1968., .
- 12Early diagnosis of carcinoma of the cervix. Surg Gynecol Obstet 1933; 66: 210–20..
- 32Early detection of cervical cancer by visual inspection methods: A summary of completed and on-going studies in India. Salud Publica Mex 2003; 45; S274–82., , , , , , , , , , .