Use of relatively expensive tests for cervical cancer only after abnormalities have been identified through lower-cost procedures fails to identify some women with disease but allows many more women in resource-poor areas to be screened. Moreover, according to a study conducted among women living in a poor community outside of Cape Town, South Africa, women who receive human papilloma-virus (HPV) testing, cervicography or a Pap smear only if visual inspection of the cervix identifies cervical disease are less likely to undergo unnecessary treatment than are women who are initially screened using one of those tests.1
In January 1996, flyers and radio advertisements were used to recruit and enroll 1,423 women aged 35-65 from a settlement outside of Cape Town. Partic- ipants were referred to an outpatient clinic, where they received a gynecologic examination from a health worker specially trained in screening techniques. Each examination consisted of the same sequence of screening methods: a Pap smear, an HPV deoxyribonucleic acid (DNA) test,* direct visual examination of the cervix and cervicography (a photograph of the cer-vix). All women were asked to return to the clinic in 2-6 days to obtain their results.
Women with a positive visual exam or a highly positive HPV DNA assay (levels at least 10 times the control value) underwent colposcopy when they returned to the clinic. All low-grade lesions were biopsied; high-grade lesions were treated by loop electrosurgical excision. Endocervical curettage was performed if no lesion was evident.
Results from Pap smears and cervicography were not available for several weeks. Any woman whose results from those tests indicated low- or high-grade lesions or cervical cancer but who had not had a colposcopy as a result of HPV testing or direct visual inspection was recalled for colposcopy and any necessary treatment.
Among women found to have high-grade lesions or cancer according to cervical biopsy, endocervical curettage or loop excision, the researchers calculated the proportion who had positive results on a given test; they also calculated the proportion of women with negative pathological findings or negative results on all four screening tests who had positive results on a given test. The results of the individual screening tests were then used to estimate values for two-stage screening procedures in which women who had positive results from direct visual inspection would receive a Pap smear, an HPV DNA test or cervicography.
A total of 1,335 women (94%) had complete screening data; their median age was 39, and 16% were aged 50 or older. Thirty-two percent of the women screened were identified by one of the four screening tests as having a cervical abnormality; of these, 84% (363) subsequently had a colposcopy. After adjusting for loss to follow-up, the researchers found that 5% of all women screened had low-grade lesions, 3% had high-grade lesions and fewer than 1% were diagnosed with cancer.
Direct visual inspection of the cervix and Pap smears each identified 30 of the 37 women confirmed through cervical biopsy as having either cancer or a high-grade lesion. These screening procedures also produced positive findings for, respectively, 228 and 73 women who were later determined to be disease free (a false-positive result). Highly positive HPV DNA test results correctly identified 18 women with cancer or high-grade lesions and produced 46 false-positive results. Use of the standard HPV DNA testing criterion (any level higher than the positive control value) found 27 of the women with disease, but produced positive results for 166 disease-free women. Cervicography accurately identified 26 women with cancer or high-grade lesions and gave positive results for 116 disease-free women.
Two-stage screening using direct visual inspection of the cervix in combination with either a Pap smear or cervicography would have accurately identified 23 of the 37 women who had cancer or high-grade lesions, while direct cervical inspection followed by HPV DNA testing would have correctly identified 21 women as requiring treatment. These two-stage screenings would have resulted in 28, 52 and 78 false-positive findings, respectively.
After adjustment for loss to follow-up, the proportion of women with cervical disease identified by individual methods ranged from 71% for cervicography to 82% for Pap smears. The HPV DNA test using the standard cutoff level identified 72% of such women; use of the higher cutoff level identified only 47%. With the exception of screening using the more restrictive HPV DNA test, all two-stage procedures identified a lower percentage of women needing treatment than did any of the individual methods: Providing a Pap smear or cervicography only after positive results from visual inspection of the cervix identified 58% of women with cancer or high-grade lesions, while using HPV DNA testing (based on the higher cutoff level) after visual inspection identified 51%.
The proportion of disease-free women with positive results on individual screening methods ranged from 6% for HPV DNA testing using the higher cutoff level to 19% for direct visual inspection. Overall, the two-stage screening procedures were less likely than tests used in isolation to result in false-positive results: The proportion of disease-free women with positive results on two-stage procedures was 2% for direct cervical observation combined with Pap smears, 4% for visual inspection of the cervix combined with HPV DNA testing and 6% for visual inspection combined with cervicography.
For every 1,000 women screened, direct visual inspection and Pap smears--the two most sensitive single tests--would accurately identify 24 women and 26 women, respectively, in need of treatment. A two-stage procedure entailing direct visual inspection followed by a Pap test or cervicography would correctly identify 18 women per 1,000 screened, while following cervical inspection with an HPV DNA test would correctly identify 16 women needing treatment.
If all women whose initial screening with any of the four individual techniques produced abnormal results were referred for colposcopy, 206 women per 1,000 screened would undergo the procedure, 182 of them unnecessarily. Based on Pap smear results following direct visual inspection, 40 women per 1,000 would have a colposcopy, 22 of them unnecessarily; those figures would be 60 and 41 for direct visual inspection followed by HPV DNA screening, and 80 and 62 for visual inspection followed by cervicography.
The authors estimated that if resources limited use of Pap smears to 1,000 women, reliance on that test alone in a population similar to the one in their study would correctly identify 26 cases of cervical disease among every 1,000 women screened. In comparison, reliance on a two-stage procedure would mean that 4,854 women could obtain initial screening using direct visual inspection, of whom 1,000 would have abnormal results. These women would receive Pap smears, which would identify 89 as having high-grade lesions or cervical cancer. Thus, initial use of direct visual inspection would allow screening of almost five times as many women, and the same number of Pap smears would identify more than three times as many in need of treatment.
The investigators acknowledge that use of a two-stage screening procedure rather than traditional one-stage screening would result in a smaller proportion of women with cervical disease being identified. They argue, however, that in communities with limited resources, the use of a relatively expensive screening procedure only when abnormalities have been identified by a relatively inexpensive method affords a greater number of women access to cervical cancer screening. Moreover, by reducing substantially the number of women who unnecessarily undergo treatment, two-stage screening limits women's exposure to adverse health outcomes such as cervical narrowing, hemorrhage, infection and risk of infection with or shedding of HIV.--K. Mahler
1. Denny L et al., Two-stage cervical cancer screening: an alternative for resource-poor settings, American Journal of Obstetrics and Gynecology, 2000, 183(2):383-388.