Professional Documents
Culture Documents
KWF 098
KWF 098
EPIDEMIOLOGY
Volume 156 Copyright © 2002 by The Johns Hopkins
Bloomberg School of Public Health
Number 8
Sponsored by the Society for Epidemiologic Research
October 15, 2002 Published by Oxford University Press
ORIGINAL CONTRIBUTIONS
Herpes Simplex Virus and Risk of Cervical Cancer: A Longitudinal, Nested Case-
Control Study in the Nordic Countries
Received for publication November 19, 2001; accepted for publication May 28, 2002.
Human papillomaviruses (HPVs) play the major role in cervical carcinogenesis. The authors reevaluated the role of
herpes simplex virus type 2 (HSV-2) in this multistage process by conducting a longitudinal, nested case-control study
using 1974–1993 data and comparing the results with those from a meta-analysis of studies. A Nordic cohort of 550,000
women was followed up for an average of 5 years, after which 178 cervical carcinoma cases and 527 controls were
identified. HSV-2; HPV-16, HPV-18, and HPV-33; and Chlamydia trachomatis antibodies were determined at baseline
by HSV-2 glycoprotein gG-2 and HPV virus-like-particle enzyme immunoassays and by using the
microimmunofluorescence method. The relative risk of cervical carcinoma was calculated by conditional logistic
regression. Longitudinal studies on HSV-2 and cervical neoplasia were identified through MEDLINE (National Library
of Medicine, Bethesda, Maryland), and weighted mean relative risks were calculated. Smoking (relative risk = 1.6, 95%
confidence interval (CI): 1.1, 2.3) and HPV-16/HPV-18/HPV-33 (relative risk = 2.9, 95% CI: 1.9, 4.3) were both
associated with cervical carcinoma. The smoking- and HPV-16/HPV-18/HPV-33–adjusted relative risks for HSV-2 were
1.0 (95% CI: 0.6, 1.7) and 0.7 (95% CI: 0.3, 1.6), respectively, for HPV seropositives. In the meta-analysis, the relative
risk for HSV-2 was 0.9 (95% CI: 0.6, 1.3). In both sets of data, HSV-2 did not play a role in cervical carcinogenesis.
Abbreviations: CI, confidence interval; ELISA, enzyme-linked immunosorbent assay; HPV, human papillomavirus; HSV-2,
herpes simplex virus type 2; RR, relative risk.
Since the end of the 1960s, herpes simplex virus type 2 study (1), together with inconsistent detection of HSV-2
(HSV-2) had been considered the major cause of invasive DNA and consistent identification of human papillomavirus
cervical carcinoma, but a longitudinal seroepidemiologic (HPV) DNA in cervical carcinoma (2), revised this paradigm
Reprint requests to Dr. Matti Lehtinen, University of Tampere, School of Public Health, POB 607, FI-33101 Tampere, Finland (e-mail: llmale@uta.fi).
in the early 1980s. However, assessing an infectious etiology samples to screen for congenital infections (11). The blood
of chronic disorders is difficult, and cohort studies that samples are drawn at all maternity clinics, and about 98
include complete follow-up are less vulnerable to different percent of all pregnant women donate samples to the Finnish
biases (3). Hence, most reliable results are provided by a Maternity Cohort bank. In 1993, this cohort had collected
combination of longitudinal design, a population-based 710,000 samples from 390,000 women; the samples were
setting, and state-of-the-science exposure assessment. stored at –25°C.
During the 1990s, considerable improvements took place The Janus Project was established in Norway in 1973 (12).
in the serologic diagnosis of herpes simplex and HPV infec- In 1991, the Janus Serum Bank had collected 424,000 serum
tions. HSV-2 antibodies can now be determined by a glyco- samples from 293,000 donors; again, the samples were
protein gG-2 enzyme-linked immunosorbent assay (ELISA) stored at –25°C. About 145,000 women were recruited
that suffers only minimally from cross-reactivity between during routine health examinations. During phase I (1974–
herpes simplex virus type 1 and HSV-2 because a majority of 1978) and phase II (1986–1991), the participation rates were
glycoprotein gG-2 is coded by a unique segment of the HSV- 85 percent and 75 percent, respectively.
2 genome (4–6). The HPV virus-like particle ELISA is a The Västerbotten Project was established in northern
highly type-specific and reproducible, albeit not very sensi- Sweden in 1986 (13). Each year, all residents of Väster-
TABLE 1. Characteristics of the Nordic cohort for a nested case-control study of the cervical carcinoma
risk associated with previous exposure to sexually transmitted microorganisms, 1974–1993
Am J Epidemiol 2002;156:687–692
Herpes Simplex Virus and Cervical Cancer Risk 689
For each case, three female controls who were cancer free Statistical analysis
at the time of the case’s diagnosis were selected randomly
Relative risks and their 95 percent confidence intervals
and were matched for age at serum sampling (±2 years),
were estimated for invasive cervical carcinoma and squa-
length of time that the serum sample was stored (±2 months),
mous cell carcinoma by using conditional logistic regression
and cohort area (Finland, Sweden, and subcohorts of the (17) with Stata computer software (version 5.0; Stata Corpo-
Janus Serum Bank in Norway). The earliest prediagnostic ration, College Station, Texas). Unconditional logistic
sample was chosen. If three controls could not be found, the regression analysis was applied for HPV-seropositive cases
matching criteria were widened: the age of six controls and controls, including the matching variables in the model.
differed by more than 4 years (maximum, 4.4 years) from The weighted mean relative risk was calculated by taking
that of the case; the maximum difference in storage time was a weighted average of the log relative risk (RR) from eligible
5 months. Samples from seven (1.3 percent) of 527 controls studies, the weight assigned to each log RR being propor-
could not be located. tional to the inverse of its variance (18). The variances were
derived from the published confidence intervals or from the
Laboratory methods pertinent frequency data by means of standard formulae. The
Am J Epidemiol 2002;156:687–692
690 Lehtinen et al.
TABLE 2. Relative risk of cervical carcinoma associated with previous exposure to herpes simplex virus type 2, overall and among
human papillomavirus* seropositives, the Nordic countries, 1974–1993
% %
No. No. RR† 95% CI† RR 95% CI RR 95% CI RR 95% CI RR 95% CI
positive positive
Invasive cervical
carcinoma 178 15 525‡ 12 1.3 0.8, 2.1 1.2 0.7, 2.0 1.3 0.8, 2.1 1.0 0.6, 1.7 1.0 0.6, 1.7
Squamous cell
carcinoma 148 16 437 12 1.4 0.9, 2.4 1.3 0.8, 2.3 1.4 0.8, 2.4 1.1 0.6, 2.0 1.1 0.6, 1.9
Invasive cervical
carcinoma* 64 17 95 24 0.7 0.3, 1.6 0.7 0.3, 1.6 0.7 0.3, 1.6
Squamous cell
carcinoma* 59 17 85 24 0.7 0.3, 1.7 0.7 0.3, 1.6 0.7 0.3, 1.7
0.4, 1.3) and 0.9 (95 percent CI: 0.5, 1.9), respectively, were DISCUSSION
obtained, indicating lack of any increased risk of cervical
neoplasia, irrespective of the method used to determine In the present study, previous HSV-2 infection was not
HSV-2 antibodies (table 3). When the glycoprotein gG-2 associated with any excess risk of subsequent development
of cervical carcinoma. Similarly, in a meta-analysis of
ELISA was used, the grand mean relative risk for all longitu-
comparable longitudinal seroepidemiologic studies that
dinal studies was 0.9 (95 percent CI: 0.6, 1.3).
yielded a remarkably narrow confidence interval, HSV-2
was not associated with any excess risk of cervical carci-
noma. Since the study by Choi et al. (19) was conducted, not
one of the longitudinal seroepidemiologic studies on HSV-2
and cervical neoplasia (1, 20–22) has found a significantly
increased relative risk, but all were underpowered to detect a
small excess risk. However, the upper 95 percent confidence
limit of our final meta-analysis, 1.3, indicates that if an
excess risk of cervical neoplasia were associated with HSV-
2, it would be very small.
Our results also indicate that the previous estimates of the
effect of HSV-2 on cervical neoplasia found in many cross-
sectional case-control studies (23–26) were biased. Possible
sources of bias include the cross-sectional design, inade-
quate power, misclassification, and confounding due to
uncontrolled risk factors, for example, HPV. Adjustment for
smoking and C. trachomatis, known risk factors for cervical
neoplasia/surrogates of risk-taking behavior (11, 27),
reduced our point estimates; further adjustment for HPV-16/
HPV-18/HPV-33 removed the excess risk from the point
estimate for women who tested positive for HSV-2 anti-
bodies. Moreover, HSV-2 antibodies were associated with
no excess risk for HPV-16/HPV-18/HPV-33–seropositive
women, also indicating that HSV-2 is not a cause of invasive
cervical carcinoma.
FIGURE 1. Human papillomavirus types 16, 18, and 33 and smok- With few exceptions, HSV glycoprotein gG-2 ELISAs are
ing-adjusted relative risks of invasive cervical carcinoma (ICC, n = now considered highly sensitive (>95 percent) and reproduc-
178) and squamous cell carcinoma of the uterine cervix (SCC, n =
150) associated with herpes simplex virus type 2 infection, by lag
ible (coefficient of variation, ≤5 percent) (6, 28). The test we
between serum sampling and cancer diagnosis in a Nordic cohort of used should not have biased the relative risk toward unit risk.
550,000 women followed up for an average of 5 years between 1974 Although our final relative risk estimates were consistent
and 1993. Relative risks for lag categories 1–12, 13–59, 60–119, and with all of the previous longitudinal study results based on
≥120 months: ICC (downward triangles)—2.2 (95% confidence inter- the early nonstandardized tests, our increased crude relative
val (CI): 0.5, 9.4), 0.5 (95% CI: 0.2, 1.1), 1.6 (95% CI: 0.5, 5.7), and
1.7 (95% CI: 0.4, 6.5), respectively; SCC (upward triangles)—2.2 risk estimates for invasive cervical carcinoma and squamous
(95% CI: 0.5, 9.5), 0.5 (95% CI: 0.2, 1.2), 1.6 (95% CI: 0.4, 5.5), and cell carcinoma suggest that the previous results were indeed
3.1 (95% CI: 0.7, 15), respectively. affected by a lack of test validity. Together with lower spec-
Am J Epidemiol 2002;156:687–692
Herpes Simplex Virus and Cervical Cancer Risk 691
TABLE 3. Meta-analysis of longitudinal seroepidemiologic studies of herpes simplex virus type 2 and
cervical neoplasia* using the II/I ratio† or glycoprotein gG-2 enzyme-linked immunosorbent assay to
determine previous herpes simplex virus type 2 infection with serum antibodies
Mean follow-
Method used and Person-years of Cases
up time OR‡ 95% CI‡ Study and year (reference no.)
country follow-up (no.) (no.)
(years)
II/I
Canada 88,000 56 1.9 2.1 0.3, 15 Choi et al., 1977 (19)
Czech Republic 33,000 33 1.6 0.7 0.2, 3.0 Vonka et al., 1984 (1)
United States 7,000 23 3.5 0.8 0.2, 3.3 Adam et al., 1985 (21)
Finland 181,000 32 5.0 1.0 0.3, 2.8 Lehtinen et al., 1992 (20)
* Carcinoma in situ (alone or included in cervical intraepithelial neoplasia) and invasive cervical carcinoma.
† A ratio of herpes simplex virus type 2 and herpes simplex virus type 1 antibody levels with a predefined cutoff
level of 0.85 (23).
‡ OR, odds ratio; CI, confidence interval.
§ The gC-2 assay was used.
¶ Combines all studies in which glycoprotein gG-2 enzyme-linked immunosorbent assay was used.
ificity and sensitivity of the II/I assays and the early glyco- both indistinguishable and readily detectable in the incident
protein gG-2 ELISAs, regression toward the mean probably invasive cervical carcinoma cases.
biased the previous results toward unit risk. In our longitu- Although our study did not associate a short time period
dinal study, we were able to assess a more accurate crude from serum sampling to cancer diagnosis with a significantly
relative risk; however, it proved to be confounded by HPV increased risk, the possibility that cervical neoplasia predis-
exposure, as indicated by both adjusting for HPV-16/HPV- poses to HSV-2 infection, that is, reverse causality, should
18/HPV-33 antibodies and restricting the analysis to HPV- not be neglected as an explanation for the association found
seropositive cases and controls only. Thus, the meta-analysis in the previous cross-sectional case-control studies. Longitu-
dinal design removes much of the possible bias due to anti-
of the previous longitudinal studies yielded a probably
body response to the cross-reactive tumor-specific antigens,
correct result for the wrong reasons or as a result of chance.
or amplification of the virus by the occult neoplasia, irre-
Why, then, did the cross-sectional studies find an associa- spective of the serologic test used (1, 20, 30). Although the
tion? Invasive cervical carcinoma patients produce an nonsignificant increase of HSV-2–associated risk of squa-
autoantibody response to an HSV-2–inducible tumor- mous cell carcinoma by increasing lag may deserve consid-
specific tissue polypeptide that is recognized by the II/I eration in even larger studies, we conclude that HSV-2 is not
assay (29). This finding also applies to the antibodies deter- likely to be causally associated with invasive cervical carci-
mined by HSV-2 nonstructural antigens (25, 26, 30, 31) noma or squamous cell carcinoma.
suggested to be found in invasive cervical carcinoma (32).
Cross-reactivity between viral infected cell protein (ICP) 8
and a homologous/functionally identical cellular protein,
proliferating cell nuclear antigen, which is abundant in inva- ACKNOWLEDGMENTS
sive cervical carcinoma, is one possible explanation (33–37). This study was supported by the Nordic Cancer Union and
It is plausible that ICP8 antibodies and proliferating cell the Academy of Finland. The Janus Serum Bank is owned by
nuclear antigen autoantibodies, as well as the HSV-2 anti- the Norwegian Cancer Society. The serum samples were
bodies and autoantibodies measured by the II/I assay, were provided following approval by the institutional review boards.
Am J Epidemiol 2002;156:687–692
692 Lehtinen et al.
The authors thank Prof. Vladimir Vonka for stimulating 17. Breslow NE, Day NE, eds. Statistical methods in cancer
discussions and Dr. Vera Abeler for histologic classification. research. Vol 1. The analysis of case-control studies. Lyon,
This is publication number 20 of the Nordic Biological France: International Agency for Research on Cancer, 1980.
Specimen Banks study group on Cancer Causes and Control (IARC scientific publication no. 32).
(NBSBCCC). 18. Der Simonian R, Laird M. Meta-analysis in controlled trials.
Control Clin Trials 1986;7:177–88.
19. Choi NW, Shettigara PT, Abu-Zeid HAH, et al. Herpesvirus
infection and cervical anaplasia—a seroepidemiological study.
Int J Cancer 1977;119:167–71.
REFERENCES 20. Lehtinen M, Hakama M, Aaran RK, et al. Herpes simplex virus
1. Vonka V, Kanka J, Jelinek I, et al. Prospective study on the type 2 infection and cervical cancer: a prospective study of 12
relationship between cervical neoplasia and herpes simplex years of follow-up in Finland. Cancer Causes Control 1992;3:
type-2 virus. Int J Cancer 1984;33:61–8. 333–8.
2. Durst M, Gissmann L, Ikenberg H, et al. A new papillomavirus 21. Adam E, Kaufman RH, Alder-Storthz K, et al. A prospective
DNA from cervical carcinoma and its prevalence in cancer study of association of herpes simplex virus and human papillo-
biopsy samples from different geographic regions. Proc Natl mavirus infection with cervical neoplasia in women exposed to
Am J Epidemiol 2002;156:687–692