Shapiro, Samuel*; Hoffman, Margaret*; Constant, Deborah*; Rosenberg, Lynn†; Carrara, Henri‡; Allan, Bruce Rider*; Marais, Dianne Jean*; Passmore, Jo-Ann Shelley*; Williamson, Anna-Lise¶∥
In 2003 we reported findings from a case-control study among women below the age of 60 years in which we evaluated the risk of invasive cancer of the cervix in relation to the use of hormonal contraceptives.1 There was no evidence to suggest that contraceptive hormones increase the risk, or that they potentiate the carcinogenicty of high-risk human papillomavirus (HPV) infection. A history of ever having received a Papanicolaou (Pap) smear was recorded as a potentially confounding variable. In the course of analyzing the data we found an unexpectedly large reduction in the risk of cervical cancer among Papanicolau test recipients2: among the control women whose median age was 44 years, relative to never having had a Papanicolau test, the odds ratio estimates for lifetime histories of 1, 2, and ≥3 smears were 0.4, 0.2, and 0.2, respectively (test for trend, P = 0.0003). We hypothesized that Pap smears may protect against cervical cancer, not only by the identification of high-risk dysplastic lesions, but possibly by some additional mechanism.
We next turned our attention to the determinants of HPV infection among the control women.3 The odds ratios for the lifetime receipt of 1, 2, and ≥3 Pap smears were 0.7, 0.5, and 0.5, respectively (test for trend, P = 0.04). We hypothesized that the minor trauma involved in taking a Papanicolau test may induce partial host immunity against HPV infection.
Prompted by our findings concerning HPV, we then evaluated the effect of Pap smears on 2 other sexually transmitted viral infections (STIs) [herpes simplex virus 2 (HSV2), and human immunodeficiency virus (HIV) 1 and 2]. Here we present findings for all 3 STIs (HSV2, HPV, and HIV).
The study was approved by the Institutional Review Boards of the University of Cape Town and Boston University. Written informed consent was obtained from all participants. After data collection was completed, all identifying information was deleted, and the files were rendered anonymous.
Details of the methods have been described,1–3 and relevant aspects are mentioned here. The Western Cape Provincial Administration of South Africa operates hospitals and community health centers that provide medical services to patients who do not have health insurance. The present study was confined to colored and black women who used those services; we had to exclude white women because they seldom used the services.
In the Provincial services, Pap smears are routinely taken using Aylesbury spatulas. Until 2000, Papanicolau test screening was performed on an opportunistic basis, when women attended family planning clinics or antenatal services. Since 2000, in conformity with international recommendations for developing countries,4 women using the public health services have been offered 3 free Pap smears, commencing after the age of 30 years, and thereafter at about 10-year intervals.5
Virtually all uninsured women with cervical cancer are seen at the community health centers, and referred to gynecologic oncology clinics located in 2 tertiary care hospitals. Incident cases of stage 1B or more advanced cervical cancer were prospectively identified at the clinics from January 1998 to December 2001. To be eligible for inclusion the women had to be 18–59 years of age, and to have lived within 150 km of Cape Town during the preceding 6 months. A standard questionnaire (including questions about lifetime histories of Pap smears) was administered by trained nurse interviewers. There were 526 cases, of whom 2 refused to participate. The median age was 44 years.
The controls were frequency-matched to the cases in an approximate ratio of 3:1 for decade of age, ethnic group, and area of residence. They were recruited when they attended Provincial Administration hospitals or community health centers for conditions (such as trauma or respiratory infections) judged to be unrelated to contraceptive use or cervical cancer risk. The same questionnaire was administered. In addition, cervical scrapings were taken to test for HPV infection, and at the same time Pap smears were also taken.
There were 1654 potential controls, of whom 107 (6%) refused to participate—38 (2%) because they had recently undergone a Papanicolau test. Six potential control women were found to have stage 1A cervical cancer, and were excluded. Among the remaining 1541 women, the median age was 44 years.
As part of the study protocol, and with a view to testing future hypotheses, we obtained and stored blood samples at −70°C. The present analysis was carried out among 1540 control women for whom samples were available; the total numbers of women whose sera were assayed for each test varied depending on the amount of serum available.
Aliquots of sera were tested for HSV2 by enzyme-linked immunosorbent serologic assay (ELISA) screening, and for HIV by ELISA AXSYM screening (Abbott) followed by confirmatory ELISA (Vironastica) and latex agglutination (Capillus) if necessary. Cervical scrapings were tested for high-risk HPV infection (types 16/18/31/33/35/39/45/51/52/56/58/59/68) using the Hybrid Capture Test (Digene Corporation, Gaithersburg, MD).
Lifetime histories of Pap smears received before recruitment into the study were compared among those who tested positive and negative for HSV2, HPV, and HIV. Unconditional multiple logistic regression was used to adjust for confounding. The initial exploratory models included the following factors: ever-receipt of Pap smears, lifetime number of Pap smears received, date of last receipt of a Papanicolau test, age, ethnic group, lifetime number of sexual partners, age at first sexual intercourse, education (none or primary school/more advanced), parity, area of residence (urban/rural), injectible contraceptive use (ever/never), oral contraceptive use (ever/never), cigarette smoking (ever/never), and alcohol consumption (ever/never). Odds ratio point estimates and their 95% confidence intervals (CIs) for the receipt of Pap smears were compared when individual factors were removed or included in the models, and factors that did not change the estimates were dropped from the final models (see below).
Among the 1540 control women, lifetime histories of 0, 1, 2, 3, and ≥4 Pap smears were given by 27%, 31%, 17%, 10%, and 15%, respectively. Among 1345 women who were tested for HSV2, 884 (66%) were positive. For HPV, 255/1522 (17%) tested positive, and for HIV, 78/1365 (6%). Table 1 gives prevalence rates of infection with HSV2, HPV, and HIV according to ever- and never-receipt of Pap smears within strata of various individual characteristics. In all strata the infection rates were lower among women who had received Pap smears.
Table 2 gives crude and adjusted odds ratio estimates for each of the infection outcomes according to ever-receipt of Pap smears, and according to the lifetime number received; Table 3 gives odds ratio estimates according to elapsed time since the last Papanicolau test. For HSV2 and HIV infection, the crude odds ratios were lower than the adjusted estimates, and for both outcomes the differences between crude and adjusted results were principally accounted for by ethnic group, age, and lifetime number of sexual partners. For HPV infection the crude and adjusted estimates were similar. For all 3 outcomes, the area of residence, parity, age at first sexual intercourse, contraceptive use, current smoking, and alcohol use were virtually without effect on the odds ratio estimates, and those factors were dropped from the final models. Below, only the adjusted estimates are considered further.
For the most common outcome, HSV2 infection, the adjusted odds ratio of infection for ever relative to never having had a Papanicolau test was 0.7 (95% CI = 0.5–0.9) (Table 2). For the lifetime number of Pap smears the odds ratios declined from 0.9 for 1 Papanicolau test to 0.5 for 4 or more (test for trend confined to Papanicolau test recipients, P = 0.02). For time since last Papanicolau test (Table 3), the odds ratios declined from 0.8 for at least 10 years before to 0.4 for <1 year before (test for trend, P = 0.002) (Table 3). Among women who had only a single Papanicolau test, the odds ratio for infection for women whose Papanicolau test was <10 years previously (120 women) was 0.5 (0.3–0.8).
For HPV infection, the second-most common outcome, the odds ratio for infection associated with ever having a Papanicolau test was 0.5 (95% CI = 0.3–0.7) (Table 2). For lifetime number, the odds ratio declined from 0.7 among women with 1 Papanicolau test to 0.5 among women with at least 3 Pap smears (test for trend, P = 0.04). For last Papanicolau test 5–9 and <5 years previously, the odds ratios of HPV infection were 0.4 (0.3–0.7) and 0.5 (0.4–0.8), respectively (Table 3). Among women who had only a single Papanicolau test, the odds ratio for last receipt <10 years previously (39 women) was 0.5 (0.3–0.8).
For HIV infection, the least common outcome, the odds ratio associated with most recent Papanicolau test within the last 5 years was 0.4 (0.2–0.9) (Table 3). Among women who had only a single Papanicolau test, the odds ratio of HIV infection for last receipt within the last 10 years (15 women) was 0.5 (0.2–1.2).
There were 1337 women who were tested for all 3 infections. Among them, prevalence rates of infection with a single virus were as follows. For HSV2, the prevalence was 58% (465/799) for women who had ever had a Papanicolau test and 74% (195/264) for women never having a Papanicolau test [crude OR = 0.5; adjusted OR = 0.8 (0.6–1.1)]. For HPV, the prevalence was 9% (35/369) for ever and 13% (10/79) for never having had a Papanicolau test [crude OR 0.7; adjusted odds ratio, 0.7 (0.3–1.7)]. For HIV there were insufficient data (7 infected women).
We have previously hypothesized that the minor trauma involved in taking a Papanicolau test may trigger a host immune response against HPV infection,3 with resultant reduction in the risk of cervical cancer. Based on the present findings we now generalize the hypothesis, and suggest that Pap smears may also provoke partial host immunity against HSV2, and, possibly, against HIV as well.
For what was by far the most common outcome, HSV2 infection, ever-receipt of a Papanicolau test was associated with an estimated 30% reduction in the risk, and there was a significant dose-response gradient; among women who had received 3 or more smears, the risk reduction was about 50%. The reduced risk also appeared to be most marked when the last smear had been received within the last 5 years—suggesting that any resistance to HSV2 infection conferred by Pap smears may be relatively short-lived, and perhaps be due to a localized cellular immune response.
HPV infection was relatively common, and a similar pattern was evident; again there was a significant dose-response gradient, with an estimated 2-fold reduction in the risk among women who received 2 or more Pap smears, and even the receipt of a single smear was associated with a 30% reduction. However, the pattern for HPV was a little different in that the reduced risk following the receipt of a Papanicolau test appeared to persist for longer (up to 9 years) than was the case for HSV2 infection (up to 4 years). HPV infection differs from HSV2 and HIV in that most women spontaneously clear their infections within a few years.6 It is possible that Pap smears may both confer partial immunity against infection with HPV and also accelerate clearance among women who are already infected.
HIV infection was the least common outcome. The findings were not robust, and with scanty data there was an estimated 60% reduction in the risk among women who had their most recent Papanicolau test within the last 5 years—again raising the possibility of a short-term risk reduction. However, that estimate was based on only 9 infected cases.
The overall pattern observed in this study was one of apparent short-term reduction in the risk of viral STIs among women who received Pap smears, evident even among women who had received only a single Papanicolau test. For 2 of the 3 outcomes for which there were sufficient data (HSV2 and HPV), the findings were not materially changed when women coinfected with more than one virus were excluded, although with more limited numbers the results were unstable.
Despite the consistency of the overall findings, they must be regarded as tentative and interpreted with caution. This study has certain limitations, the most serious of which is the possibility of residual confounding. For HSV2 and HPV, the systematically weaker odds ratios after adjustment support that possibility. A general limitation of studies that adjust for sexual activity is that they can adjust only for what is reported, and underreporting may be greatest among women who have had the greatest number of partners. In addition, we had no information on the sexual behavior of the male partners, and we cannot exclude the possibility that women who had Pap smears selected partners who were at lower risk for STIs. Confounding could also have occurred if there were unmeasured or incompletely measured differences in health behaviors (eg, drug use), or in socioeconomic status. It is also possible that women who seek Papanicolau test screening might make other health choices that put them at reduced risk of STIs (eg, use of condoms).
Finally, the immunologic data were cross-sectional, and we had no information on the time of onset of the infections. To overcome that limitation, longitudinal studies would be needed in which data are prospectively collected.
Information bias cannot explain the findings, because no hypothesis existed when the data were collected, and samples were tested without knowledge of Papanicolau test history. Selection bias was also unlikely; the recruitment rate was 94%, only 2% of the women refused to participate specifically because they had recently undergone a Papanicolau test, and control conditions such as trauma or respiratory infection were independent of ever having had a Papanicolau test.
We are not aware of other epidemiologic studies that have reported possible protective effects of Pap smears against viral STIs. However, there is limited evidence to suggest that trauma to the cervix may be associated with protection. One study7 has reported a decrease in the risk of HPV infection (adjusted for age and number of sexual partners) associated with an increasing frequency of sexual intercourse and an increasing number of pregnancies. In addition, it has been proposed (based mainly on studies among prostitutes, and partly on anecdotal evidence) that frequent sexual intercourse in some women may block the transmission of HPV and HIV.8 It has also been suggested that the protective effect may be lost when prostitutes quit their occupations and stop having frequent intercourse.9
With regard to possible mechanisms, there is only limited information on what factors may initiate host immune responses to viral STIs. However, because HPV infections commonly regress,6 it is clear that an effective immune response to at least one viral infection does occur. If minor trauma to the cervix can trigger such a response, one question that arises is whether the trauma associated with the use of an Aylesbury spatula is sufficiently traumatic. To examine that possibility, in a separate follow-up study we compared various markers of inflammation in cervico-vaginal lavage specimens obtained from healthy women at baseline, and again 2 weeks later; Pap smears were administered at baseline, using Aylesbury spatulas.10 At 2 weeks, inflammatory cytokines associated with cell-mediated immunity (interleukin-12 and tumor necrosis factor-α) and T-cell regulation (interleukin-10) were elevated.
Not all the evidence is consistent: although the acquisition of viral STIs appears in some studies to be inversely associated with frequency of intercourse,7–9 other studies have demonstrated that cervicovaginal inflammation places women a higher risk of acquiring HIV infection.11–13 Proposed mechanisms have been the recruitment of highly activated and Chemokine receptor CCR5-expressing target cells, and epithelial disruption.
Some indirect evidence may also be relevant. It has been suggested that cervical biopsy (which is more traumatic than a Papanicolau test) may bring about regression of cervical intraepithelial neoplasia 1, or inhibit its progression to cervical intraepithelial neoplasia 3, by eliciting an immune or inflammatory reaction.6,14 However, in recent prospective data, among women whose Pap smears were diagnosed as atypical squamous cells of undetermined significance and who also tested positive for HPV, cervical biopsy did not reduce the subsequent risk of carcinoma-in situ.15
Our findings raise the possibility that Pap smears may provoke a host immune response, and hence confer partial protection against viral STIs. The data also suggest that repeated Pap smears may confer added protection, and that any protective effect of a single smear may be relatively short-lived. We present these data as a hypothesis. The findings need to be independently confirmed in longitudinal studies that provide precise information on time sequences. The possibility of confounding, particularly residual confounding due to underreporting of sexual activity or other relevant health behaviors, can confidently be excluded only in a randomized trial.
Data collection and the collection of specimens were carried out by Saint P. Gribble, Lungiswa Makayi, and Beverley Arendse. The study administrator was Eleanor Marks. We thank the Western Cape Provincial Administration for access to the oncology clinics and community health centers. Special thanks are due to Zena Stein, Columbia University, based on our earlier HPV findings, she suggested that we should check whether Pap smears may also protect against HIV, and that suggestion led to the present analysis.
1. Shapiro S, Rosenberg L, Hoffman M, et al. Risk of invasive cancer of the cervix in relation to the use of injectable progestogen contraceptives and combined estrogen/progestogen contraceptives. Cancer Causes Control. 2003;14:485–495.
2. Hoffman M, Cooper D, Carrara H, et al. Limited Pap screening associated with reduced risk of cervical cancer in South Africa. Int J Epidemiol. 2003;32:573–577.
3. Shapiro S, Carrara H, Allan BR, et al. Hypothesis: the act of taking a Pap smear reduces the prevalence of human papillomavirus infection: a potential impact on the risk of cervical cancer. Cancer Causes Control. 2003;14:953–957.
4. International Agency for Research on Cancer (IARC). Working group on cervical cancer screening. In: Hakama M, Miller AB, Day NE, eds. Screening for Cancer of the of the Uterine Cervix. Lyon: IARC; 1986.
5. National Department of Health. National Guidelines for Cervical Cancer Screening Programme. Pretoria, South Africa: NDOH; 2000.
6. Ostor AG. Natural history of cervical intraepithelial neoplasia: a critical review. Int J Gynecol Pathol. 1993;12:186–192.
7. Herero R, Schiffman M, Hildesheim A, et al. Epidemiology of type-specific HPV infection in Guanacaste. Presented at: Proceedings of the 20th International Papillomavirus Conference; October 4–9, 2002; Institut Pasteur, Paris. Abstract 0070.
8. Laurence J. Opinion paper. Repetitive and consistent cervicovaginal exposure to certain viral pathogens appears to protect against their sexual acquisition in some women: potential mechanisms. J Reprod Immunol. 2003;58:79–91.
9. Kaul R, Kimani J, Dong T, et al. Late seroconversion in HIV “resistant” Nairobi prostitutes is associated with a preceding decrease in HIV exposure. Presented at: 7th Conference Retroviruses and Opportunistic Infections; January 30 to February 2, 2000; San Francisco. Abstract 489.
10. Passmore J-AS, Morroni C, Shapiro S, et al. Papanicolaou smears and inflammatory cytokine responses. J Inflamm. 2007;4:8
11. Lawn SD, Subbarao S, Wright TC, et al. Correlation between human immunodeficiency type 1 RNA levels in the female genital tract and immune activation associated with ulceration of the cervix. J Infect Dis. 2000;181:1950–1956.
12. Wright TC, Subbarao S, Ellerbrook TV, et al. Human immunodeficiency virus 1 expression in the female genital tract in association with cervical inflammation and ulceration. Am J Obstet Gynecol. 2001;184:279–285.
13. Hladik F, Lentz G, Delpit E, et al. Coexpression of CCR5 and IL2 in human genital but not blood T cells. Implications for the ontogeny of the CCR5+ Th1 phenotype. J Immunol. 1999;163:2306–2313.
14. Koss LG, Stewart FW, Foote FW, et al. Some histological aspects of behavior of epidermoid carcinoma in situ and related lesions of the uterine cervix. Cancer. 1963;16:1160–1211.
15. Cox JT, Schiffman M, Solomon D. Prospective follow-up suggests similar risk of subsequent cervical intraepithelial neoplasia grade 2 or 3 among women with cervical intraepithelial neoplasia grade 1 or negative colposcopy or directed biopsy. Am J Obstet Gynecol. 2003;188:1406–1412.
© 2007 Lippincott Williams & Wilkins, Inc.