During the past 50 years, the mortality rate from cervical cancer has declined by more than 70% in the United States,1 after the widespread availability of Papanicolaou smears. Approximately 50 million Papanicolaou smears are done in the United States annually,2 and most of the smear results are normal. Although it is clear that women who have never or infrequently had Papanicolaou smears should be screened, the best strategy for screening recently screened women, especially those with documented normal smear results, is not known.
Published guidelines suggest that women with a history of normal smears might be screened less often than annually, but it remains unclear how often smears should be done.3 Other guidelines suggest that women can be screened up to every 3 years.4 Given the large number of smears done each year in the United States, the differences between annual screening and screening every 3 years in terms of absolute numbers of annual smears is profound. Determining the optimal screening interval after a normal smear is important because unnecessary screening results in false-positive smears that often lead to needless additional tests and patient concern.
Our purpose was to determine the incidence of cervical cytologic abnormalities after normal smears, focusing our analyses on how those rates vary by age and time since last normal smear. On the basis of our results, we discuss implications for current clinical practice and make recommendations for future studies.
Materials and Methods
We analyzed data from the National Breast and Cervical Cancer Early Detection Program, an initiative sponsored by the Centers for Disease Control and Prevention (CDC) that has provided mammograms and cervical smears to low-income, uninsured women throughout the United States since 1991. Data collection methods were described previously.5–7 The CDC established minimum data elements to be collected from each woman who receives screening services, including demographic characteristics, screening results, diagnostic procedures, and histologic outcomes. Although data collection forms vary among screening locations, local program officials standardize data categories before submitting information to the CDC.
Our study population comprised women who had a first program Papanicolaou smear reported as normal between 1991 and 1998 and at least one other subsequent smear within the next 36 months. We hypothesized that second smears done within 9 months of the first one might have been part of a diagnostic workup, or a follow-up to an abnormal smear before program enrollment, not for screening. Therefore, we excluded women (n = 2926, 2.1%) who had smears within 9 months of the first one from the analysis, after determining that those smears were significantly more likely to be abnormal (ie, atypical squamous cells of undetermined significance [ASCUS] or worse) than smears done 9–12 months after the last normal smear (P = .001). We were most interested in screening for squamous cell lesions because epidemiologic trend data suggested that widespread screening has not been effective in reducing incidence of invasive adenocarcinoma.8,9 Therefore, we also excluded women with second smears that showed glandular cell abnormalities (n = 178, 0.1%).
Each woman's age was calculated based on the birth date she reported at enrollment. Age at the time of the normal smear was used as her age throughout the study period. On the basis of results of analyses that showed where abrupt changes in incidences of smear abnormalities occurred, we grouped women into four age groups (under 30, 30–49, 50–64, and over 65 years old). Race/ethnicity was determined by an algorithm used by the CDC for reporting program statistics. If a woman was identified as Hispanic, she was Hispanic regardless of any other designations. If she was white, black, Asian-Pacific Islander, or American Indian/Alaska Native, we classified her as such. We classified women who met none of those criteria as other.
We constructed screening categories on the basis of the interval in months between the first and second smears. We chose intervals simulating clinical testing intervals, 9–12 months, 13–24 months, and 25–36 months from normal smears. We also evaluated time since normal smear in months as a continuous variable to assess trends in incidence rates of cytologic abnormalities.
Cervical smears were read at regional laboratories, and results were reported using Bethesda System categories of normal; infection, inflammation, or reactive changes; ASCUS; low-grade squamous intraepithelial lesion (SIL); high-grade SIL; suggestive of squamous cell carcinoma, and other. Most smear results in the other category were glandular abnormalities (eg, glandular atypia, atypical endocervical glands). When possible, we reclassified smear results reported as other on the basis of textual information.
We calculated incidence rates of smear interpretations by dividing the number of women with each smear result by the number of women rescreened within each age and time category. Smears read as high-grade SIL and suggestive of squamous cell cancer were combined into the category high-grade SIL or worse because those are smears for which no controversy exists as to their clinical importance. The number of women needed to screen was computed by taking the inverse of the incidence rates of high-grade SIL or worse in each age category. That number refers to the women who needed to be screened in each age group to find each cytologic abnormality. When appropriate, we age-adjusted incidence rates using the age distribution of the 1996 National Breast and Cervical Cancer Early Detection Program population. To evaluate the effect of more time between screening smears on the incidence of high-grade SIL or worse, we regressed the age-adjusted incidence of smears categorized as high-grade SIL or worse by time from the normal smear in months. Differences in rates were compared using χ2 tests and regression analysis techniques, and an alpha of .05 was considered significant.
Among 620,063 women tested over 7 years, 128,805 had initial smears that were read as normal and had second cervical smears 9–36 months later. The average age of the rescreened cohort was 48.9 years (range 12–96 years). The mean time to having the second smear was 15.7 months (range 9–36 months). Most second smears (94.1%) were interpreted as either normal or infection, inflammation, or reactive changes. The group was racially and ethnically diverse; over 40% of the women were not white. Demographic information and results of the second smear for all women are detailed in Table 1.
The crude incidence rates of new smear abnormalities within 3 years of normal smears by age are shown in Table 2. The highest incidence of cytologic abnormalities was among women less than 30 years old. As age advanced, the incidence of ASCUS, low-grade SIL, and high-grade SIL or worse decreased markedly (P < .001 for each category). Over the 3 years after normal smear results, the incidence of new smears interpreted as high-grade SIL or suggestive of squamous cell cancer (high-grade SIL or worse) was 66 of 10,000 for women under 30 years old, 22 of 10,000 for those 30–49, 15 of 10,000 for those 50–64, and 10/10,000 for those over 65 years. The number needed to screen in the 3 years after normal smears to find each case of high-grade SIL or worse increased substantially, with 152 women under 30 years old required to find each case of high-grade SIL or worse compared with 455 women aged 30–49 years, 667 women aged 50–64 years, and 1000 women over 65 years.
We calculated age-adjusted incidence rates of cytologic abnormalities per 10,000 women screened at 9–12 months, 13–24 months, and 25–36 months from normal smears (Table 3). The incidence of ASCUS, the most common cytologic abnormality, did not change significantly in time from normal smears (P = .36). The incidence of smears interpreted as low-grade SIL increased as time since normal smear increased (P = .01). The differences in incidence of high-grade SIL or worse after a normal smear for various intervals were not statistically significant (P = .42). No additional women with smears suggestive of squamous cell cancer were identified per year (1.4 of 10,000 at 9–12 months compared with 0.8 of 10,000 at both 13–24 and 25–36 months).
When we evaluated time from normal smear as a continuous variable (9–36 months), we found no significant association between the age-adjusted incidence of high-grade SIL or worse and time within that interval (P = .09, data not shown). To evaluate a nonlinear association, we fit the data to a quadratic polynomial function and found no significant relationship (P = .33).
Previous analyses of this population found that about half of smears categorized as high-grade SIL or worse were associated with high-grade cervical histology,7 so the average number of women that need to be screened to find each case of high-grade histology associated with high-grade cytology is at least two times greater than we estimated.
Women screened 1, 2, and 3 years after normal smears had the same risk of developing high-grade SIL or worse. The stable rate could be caused by the low incidence of high-grade SIL or worse within 3 years of normal smears, the low false-negative rate of the first normal smear, or other unmeasured confounders. Previous prospective cohort studies showed that the incidence of invasive cancer was low within 3 years of normal smears.10 Case-control studies concluded that annual, biennial,11,12 and triennial12,13 screening conferred equal protection from developing invasive squamous cell cancer. Our study complements those data by showing the lack of dependence of incident high-grade SIL or worse on time within such a short interval.
Within 3 years of normal cytology results, the incidence of ASCUS was stable and the incidence of low-grade SIL increased. Low-grade abnormal smears comprise over 95% of cytologic abnormalities after normal smears, and their treatment is controversial. Women who are screened too frequently might have worse health outcomes if many low-grade smears of no clinical importance are identified for each additional case of high-grade found. Low-grade abnormalities often prompt further tests and procedures that cause unfounded patient anxiety.14,15 We report only cytologic outcomes and not underlying histologic outcomes, and although some low-grade cytologic abnormalities lead to diagnoses of high-grade cervical histology, most do not.
Our study has several strengths and limitations. The CDC program data base was designed to be used for descriptive statistics, not hypothesis testing, so data from individual clinical sites might vary in quality. However, the relative difference between comparison groups should be preserved. The program uses data from many different laboratories and clinical settings, so results might be representative of actual screening settings, rather than those approximated in controlled trials. The population served by the CDC program is of relatively low income, so generalizability to other groups might not be appropriate. Low socioeconomic status has been shown to be an independent predictor of cervical dysplasia and cancer, so our findings might overestimate their incidence in lower-risk populations. Women who avail themselves of frequent screening, specifically 9–12 months after normal smears, might be low-risk women concerned about their health or high-risk women with histories of abnormal smears who have been told to get annual smears. Women who wait longer to get cervical smears might have been told to do so by their providers because they are at low risk. We do not know the other risk factors for cervical cancer in those women, and we do not know whether those risk factors affected our results. Women who wait longer to get their second smears might be at higher risk of dysplasia because of cofactors associated with nonadherence. However, if that were true, we would have seen a higher rate of high-grade SIL or worse in women's second smears at longer intervals. Smears that showed glandular cell abnormalities might also lead to diagnoses of underlying high-grade squamous lesions. Our exclusion of them likely did not affect our major conclusion. In supplemental analyses, those abnormalities were uncommon (157 of 128,962; 0.12%), and incidence rates were not associated with increasing time since smear (P = .788).
In the United States, many practitioners still do annual cervical smears, and ACOG states that annual cervical cytologic screening for most women is “prudent and warranted.”16 Annual screening has been standard practice for many years, and practitioners have been concerned about changing a screening system that has been highly effective in controlling cervical cancer.4 Studies supporting such a change have been based on observations in populations outside the United States, leading many practitioners to question the generalizability of results to their clinical practices. Many practitioners are concerned that the main motivation for screening less often is to save money at the expense of patient care. The issue is complicated by a medicolegal environment in the United States that places high value on not missing any women with important cervical disease regardless of the costs and risks of screening in terms of false-positive testing.
Practitioners and policymakers need the best available information with which they can make decisions about screening strategies for large populations of women that optimize the benefits of screening and minimize the harm. The true benefits of identifying additional cases of high-grade histology in terms of decreasing morbidity and mortality from cervical cancer need to be weighed against the risk of false-positive testing. Compared with issues of cost, enumerating risks of overscreening in terms of unnecessary interventions provides a more compelling reason for screening less frequently. Optimal screening strategies for women with recent normal cytology results, however, should be devised on the basis of comprehensive modeling studies that incorporate risks and benefits of repetitive screening. Effective methods of informing women about risks and benefits should also be determined to facilitate optimal patient-provider decision making.
1. Cannistra SA, Niloff JM. Cancer of the uterine cervix. N Engl J Med 1996;334:1030–8.
2. Richart RM. Screening. The next century. Cancer 1995;76:1919–27.
3. Shingleton HM, Patrick RL, Johnston WW, Smith RA. The current status of the Papanicolaou smear. Can Cancer J Clin 1995;45:305–20.
4. U.S. Preventive Services Task Force. Guide to clinical preventive services: An assessment of the effectiveness of 169 interventions. Report of the U.S. Preventive Services Task Force. Baltimore, MD: Williams & Wilkins, 1996.
5. May DS, Lee NC, Nadel MR, Henson RM, Miller DS. The National Breast and Cervical Cancer Early Detection Program: Report on the first 4 years of mammography provided to medically underserved women. AJR Am J Roentgenol 1998;170:97–104.
6. Henson R, Wyatt SW, Lee NC. The National Breast and Cervical Cancer Early Detection Program: A comprehensive public health response to two major health issues for women. J Public Health Manage 1996;2:36–47.
7. Lawson HW, Lee NC, Thames SF, Henson R, Miller DS. Cervical cancer screening among low-income women: Results of a national screening program, 1991–1995. Obstet Gynecol 1998;92:745–52.
8. Platz CE, Benda JA. Female genital tract cancer. Cancer 1995;75:270–94.
9. Nieminen P, Kallio M, Hakama M. The effect of mass screening on incidence and mortality of squamous and adenocarcinoma of cervix uteri. Obstet Gynecol 1995;85:1017–21.
10. Lynge E, Poll P. Risk of cervical cancer following negative smears in Maribo County, Denmark, 1966–1982. IARC Sci Publ 1986;1986:69–86.
11. Shy K, Chu J, Mandelson M, Greer B, Figge D. Papanicolaou smear screening interval and risk of cervical cancer. Obstet Gynecol 1989;74:838–43.
12. Sato S, Makino H, Yajima A, Fukao A. Cervical cancer screening in Japan. A case-control study. Acta Cytol 1997;41:1103–6.
13. Clarke EA, Hilditch S, Anderson TW. Optimal frequency of screening for cervical cancer: A Toronto case-control study. IARC Sci Publ 1986;1986:125–31.
14. Jones MH, Singer A, Jenkins D. The mildly abnormal cervical smear: Patient anxiety and choice of management. J R Soc Med 1996;89:257–60.
15. Bell S, Porter M, Kitchener H, Fraser C, Fisher P, Mann E. Psychological response to cervical screening. Prev Med 1995;24:610–6.
16. American College of Obstetricians and Gynecologists. Routine cancer screening. ACOG committee opinion no. 185. Washington, DC: American College of Obstetricians and Gynecologists, 1997.