- Email: [email protected]
Risk factors for rapid-onset cervical cancer
Risk factors for rapid-onset cervical cancer Allan Hildesheim, PhD,a Olympia Hadjimichael, MPH,b Peter E. Schwartz, MD,b Cosette M. Wheeler, PhD,c Willard Barnes, MD,d David M. Lowell, MD,b Jerry Willett, MD,d and Mark Schiffman, MDa Bethesda, Maryland, New Haven, Connecticut, Albuquerque, New Mexico, and Washington, D.C. OBJECTIVES: The current study was designed to elucidate risk factors associated with the development of cervical cancer during the course of routine Papanicolaou smear screening (rapid-onset cervical cancer). STUDY DESIGN: Four hundred eighty-three women diagnosed with invasive cervical cancer, representing 73% of all such tumors diagnosed in Connecticut between 1985 and 1990, were studied. Papanicolaou smear screening and risk factor information was obtained by questionnaire and physician record review. Results from human papillomavirus deoxyribonucleic acid testing by polymerase chain reaction of tumor samples were available for 278 study participants. Prediagnostic Papanicolaou smear slides were reviewed for 67% of cases with a screening history. Screening history information, slide review, and questionnaire data were used to classify women as having rapid-onset cervical cancer (n = 43), possible rapid-onset cervical cancer (n = 111), or normal-onset cervical cancer (n = 329). RESULTS: Compared with normal-onset cases, rapid-onset cases tended to be younger (P = .001) and were more likely to be white (P = .002), diagnosed with adenocarcinomas or adenosquamous carcinomas (P = .001), and diagnosed with early-stage disease (P = .001). Cases diagnosed as possible rapid-onset disease tended to have a profile that was intermediate to that observed for rapid-onset and normal-onset cases. Human papillomavirus deoxyribonucleic acid was detected in 75.2% of cases tested. Compared with women who tested positive for human papillomavirus type 16 or other, those positive for human papillomavirus type 18 had a relative risk for rapid-onset disease of 1.6 (95% confidence interval 0.52-4.9). No significant association was observed between type 18 and possible rapid-onset disease when possible rapid-onset cases were compared with women diagnosed with normal-onset cervical cancer (relative risk 0.67, 95% confidence interval 0.29-1.6). Oral contraceptive use, cigarette smoking, number of pregnancies, and a maternal history of cervical cancer were not significantly associated with rapid-onset disease. CONCLUSIONS: Results from this study suggest that the risk factors associated with the development of rapidonset cervical cancer are similar to those for normal-onset disease. (Am J Obstet Gynecol 1999;180:571-7.)
Key words: Rapid-onset cervical cancer, human papillomaviruses, epidemiology
Cervical cancer typically develops over many years after initial infection with the human papillomavirus (HPV), persistence and progression of this infection to high-grade intraepithelial neoplasia, and eventual invasion. The incidence of HPV infection characteristically peaks soon after initiation of sexual activity, usually in the late teens and early twenties, and is closely followed by a smaller peak in the incidence of HPV-induced, cytologically evident low-grade cervical lesions.1 The majority of HPV infections and associated low-grade lesions regress without treatment, but in a minority of women they persist and progress to high-grade lesions.2-4 The peak inci-
From the National Cancer Institute,a Yale University School of Medicine,b the University of New Mexico School of Medicine,c and Georgetown University.d Received for publication July 9, 1998; revised September 2, 1998; accepted November 16, 1998. Reprint requests: Allan Hildesheim, PhD, Interdisciplinary Studies Section, Environmental Epidemiology Branch, DCEG, National Cancer Institute, 6130 Executive Blvd, Room 443, Rockville, MD 20852. 6/1/95843
dence of high-grade lesions occurs in the late twenties and early thirties, a median of 5 to 10 years after the peak occurrence of low-grade dysplasia.1 Invasion occurs on average a decade or more later, as the result of as-yet-unknown final oncogenic events. The majority of invasive cervical tumors are diagnosed after age 45 years. The long preclinical phase of cervical neoplasia and the existence of observable precursor lesions have made cervical cancer screening by Papanicolaou smears possible. Papanicolaou smear screening has proved to be highly successful in countries where widespread screening has been implemented for several decades. This is demonstrated by the drastic reduction of cervical cancer incidence and mortality in countries where successful screening programs have been established.5 In spite of the success of Papanicolaou smear programs in reducing the occurrence of invasive cervical cancer, cervical cancer will develop in a small proportion of women during the course of routine Papanicolaou smear screening, suggesting a potential for a much-faster disease progression than is the norm.6 The notion of 571
572 Hildesheim et al
“rapid-onset cervical cancer” was suggested by Bain and Crocker6 who noted that as many as 25% of invasive cervical cancer cases could be classified as rapid onset on the basis of an interval of less than 3 years between a normal cytologic screening and the development of invasive disease. Others have reported a similar phenomenon.7-9 It is now understood that rapid-onset cervical cancer as a clinical entity results at least in part from inadequate screening (inadequate specimen collection, Papanicolaou smear interpretation, or follow-up).10 However, it is probable that in a subgroup of women defined clinically as having rapid-onset disease the disease does in fact arise quickly. Indirect evidence for truly rapid-onset disease is provided by studies that have indicated that high-grade cervical lesions, the immediate precursor of invasive cervical cancer, can arise within short time periods without evidence of preceding low-grade disease.11 In addition, the fact that cervical cancer is sometimes diagnosed among women in their late teens and early twenties suggests that the phenomenon of rapid-onset disease, although uncommon, exists. It is important to determine whether the etiology of these rapidly progressing cases is different from that of the more commonly observed normal-onset cases. One possibility is that true rapid-onset cases are associated with particularly oncogenic HPV types. It is known that, of the more than 30 HPV types that infect the cervix, a small subgroup accounts for the majority of cervical cancer diagnosed worldwide (HPV types 16, 18, 31, 33, and 45). Among HPV types known to be associated with cancer development, heterogeneity in the transforming ability has been suggested in laboratory and clinical studies.12, 13 Most notable are the data that suggest that HPV-18 has a higher transforming ability than HPV-16.12 It is thus plausible that HPV-18 may be strongly linked to the development of rapid-onset disease.13 In addition to HPV, persons in whom rapid-onset cervical cancer develops may differ from those in whom normal-onset disease develops with respect to other factors associated with cervical cancer, such as oral contraceptive use, cigarette smoking, and reproductive behavior. The current study was designed to examine whether the distribution of HPV types and other cervical cancer risk factors is different among rapid-onset cervical cancer cases compared with the more common normal-onset disease. To examine this issue, we investigated 483 women diagnosed with cervical cancer in Connecticut over a 5year period; these women were classified as having been diagnosed with either rapid- or normal-onset cervical disease, and their risk factor profile was compared. Methods The current study is part of a population-based investigation undertaken in Connecticut to examine patterns of cervical cancer screening and ways in which screening could be improved to further reduce cervical cancer incidence.10, 14
March 1999 Am J Obstet Gynecol
After institutional review board approval, all histologically confirmed incident cervical cancer cases were sought by a rapid ascertainment system set up through the Connecticut tumor registry. During the period from March 1, 1985, through February 23, 1990, 664 cervical cancer cases were diagnosed among Connecticut residents. Information on Papanicolaou smear screening history and on exposure to known or suspected cervical cancer risk factors was obtained for 483 of the 664 eligible women (73%) by a structured telephone interview and review of physician records. Nonresponse was accounted for by physician refusal (12%), patient refusal (6%), and inability to locate subject or next of kin (9%). An attempt was made to retrieve and review cytology slides and reports from the last 3 Papanicolaou tests performed on study participants before the diagnosis of cancer. Because of the nature of laws regulating the length of storage of Papanicolaou smears by cytology laboratories, we were most successful at obtaining slides performed in the 2 to 3 years preceding diagnosis. Thus we were successful at retrieving slides for 79% of cases who reported at least 1 Papanicolaou smear in the 3 years preceding diagnosis and 67% of all cases who reported ever having had a Papanicolaou smear. All slides were reviewed by a single expert pathologist (D.M.L.) who was masked as to the original diagnosis. In cases where a discrepancy was noted between the original diagnosis and that of the study pathologist, a second study pathologist (M.J.M.) was asked to review the slides and a consensus diagnosis was reached. On the basis of reported Papanicolaou smear screening history, slide review, and questionnaire data, women were classified as having “rapid-onset cervical cancer” if the following criteria were met: (1) at least 1 normal Papanicolaou smear was taken within 5 years of cancer diagnosis and review of the slide confirmed normalcy and adequacy of the sample (including presence of endocervical cells) or (2) the diagnosis of cancer was made within 10 years of having initiated sexual activity. In addition, women were classified as having “possible-rapid onset cervical cancer” if the following criterion was met: At least 1 normal Papanicolaou smear was taken within 5 years of diagnosis but review (to confirm normalcy and adequacy) was not possible or incomplete. Forty-three women met the criteria for rapid-onset disease (31 on the basis of screening history, 8 on the basis of the interval since initiation of sexual activity, and 4 on the basis of both criteria) and an additional 111 women were classified as having possible rapid-onset disease on the basis of the criterion listed. The remaining 329 patients were classified as having “normal-onset cervical cancer.” Among the 329 patients classified as having normal-onset cervical cancer, 137 reported never having been screened and 110 were screened more than 5 years preceding the diagnosis of cancer. The remaining 82 women had a Papanicolaou smear within 5 years of diagnosis, but
Hildesheim et al 573
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
the smear was either incorrectly read as normal (n = 30) or the abnormal reading was inadequately followed up (n = 52), leading to the development of cancer. We also defined rapid-onset disease with a 3-year instead of a 5-year cutoff level. Analyses with this alternative definition yielded similar results, so results with the 5-year cutoff value are presented. To assess the presence and type of HPV in tumors from participants in our study, we attempted to retrieve histologic blocks from all study participants. We were successful in this endeavor for 379 (78%) of the 483 participants. From each available block, one 5-µm hematoxylineosin–stained slide followed by two 10-µm curled sections were prepared. Stained slides were reviewed (J.W.) to confirm the presence of invasive tumor material, and in cases where more than 1 block was available per woman the most representative block or blocks were selected for HPV deoxyribonucleic acid (DNA) testing. In selecting specimens for HPV DNA testing, blocks with extensive necrotic tissue or blood were avoided. Among the 379 women for whom blocks were retrieved, reviewed, and tested, 93% had a single block selected for HPV testing, 6% had 2 blocks selected for testing, and 1% had 3 blocks selected for testing. One 10-µm section was used for HPV DNA testing. On arrival at the laboratory, samples were deparaffinized and DNA was extracted as previously described.15 In brief, samples were centrifuged at 14,000g for 30 seconds, deparaffinized in 1 mL of octane, extracted in 1 mL of absolute ethanol, and resuspended in 50 to 100 µL of digestion buffer (50 mmol/L TRIS [trishydroxymethylaminomethane], pH 8.5, 1 mmol/L EDTA [ethylenediaminetetra-acetic acid], 400 µg/mL proteinase K, 4 mmol/L magnesium chloride, and 1% laureth-12 detergent). Resultant crude digests were subjected to heat inactivation at 95°C for 10 minutes. Specimens were amplified for human β-globin (using the GH20 and PC04 primer pair, which targets a 268-bp product) and with the 4 HPV primer systems described below. Two HPV primer sets that target the L1 (HPV-16 L1)16 and URR (HPV-16 URR)17 regions of HPV-16 were used, as was 1 primer set targeting the L1 region of HPV-18 (HPV-18 L1)16 and a degenerate primer system (WD72/76) that targets the HPV E6 gene of a broad spectrum of HPV types.18 Amplification was carried out under the following conditions: Five microliters of crude digest was added to 100 µL of master mix containing 2.5 mmol/L Mg++ and a standard buffer system. The reactions were amplified in a Perkin-Elmer 9600 for 40 cycles. The following amplification profile was used for all primer pairs: 95°C for 1 minute, 55°C for 1 minute, and 72°C for 2 minutes, followed by a 5-minute extension at 72°C. Detection of polymerase chain reaction amplicons was carried out as follows: For the HPV-16 and HPV-18 type-specific primers both ethidium bromide gel analysis and hybridization assays were performed as previously described.16 The WD72/76 reactions were evaluated by hybridization with a battery of HPV type-specific probes pre-
viously described.18 HPV-16 URR amplification products were evaluated by staining 2% agarose gels with ethidium bromide after electrophoresis.17 A woman was considered positive for HPV-18 if the tumor tissue tested positive for HPV-18 with either the HPV-18 L1 or the WD72/76 primer sets. Similarly, a woman was classified as positive for types other than HPV-18 if the tumor tissue tested negative for HPV-18 and positive for HPV-16 or other HPV types with either the HPV-16 L1 primer set, the HPV-16 URR primer set, or the WD72/76 primer set. Finally, women were classified as HPV DNA negative if the tumor tissue tested negative to all 4 HPV primer sets used but tested positive for β-globin amplification. One hundred one women (27% of those tested) whose samples tested negative for all HPV primer sets and to the β-globin primers were considered inadequate and excluded. Our finding that 27% of samples were inadequate for HPV testing and that inadequate samples tended to be older (P < .001) confirms the well-known difficulty of testing for multiple HPV types out of aging fixed tissue.15 Contingency table analyses were performed to compare the distribution of various factors among women classified as having rapid-onset, possible rapid-onset, and normalonset disease. Factors examined included age, race, histologic features, stage at diagnosis, HPV, reproductive history, oral contraceptive use, smoking, and a maternal history of cervical cancer. Pearson’s χ2 test was used to evaluate whether differences between groups were statistically significant. Logistic regression analyses were also performed comparing rapid-onset and possible rapid-onset cases against women diagnosed with normal-onset cervical cancer. Because HPV is known to be present in approximately 90% of cervical cancer lesions, HPV-16/other–positive women were considered as the referent group in analyses that examined the association between HPV and rapid-onset disease. The statistical significance of risk estimates was determined by the 95% confidence interval. A confidence interval that excludes 1.0 can be interpreted as being statistically significant. The significance of trends was tested for by including the categoric variable of interest as continuous in the logistic model. Results Analysis was performed on the 329 women classified as having normal-onset cervical cancer, 111 women classified as having possible rapid-onset cancer, and 43 women with rapid-onset disease. Women diagnosed with rapidonset tumors were found to differ significantly from those diagnosed with normal-onset disease with respect to all demographic and clinicohistologic factors examined (Table I). Women with rapid-onset disease were younger than those with normal-onset disease; 41.9% of rapid-onset cases were younger than 35 years old compared with 8.8% of normal-onset cases. Although this finding is partly the result of the definition of rapid-onset
574 Hildesheim et al
March 1999 Am J Obstet Gynecol
Table I. Demographic and clinicohistologic features of cervical cancer cases by rapid-onset status
Factor Age at diagnosis <35 y 35-44 y 45-54 y 55-64 y ≥65 y Race† White Other Histologic diagnosis Squamous cell Adenocarcinoma or adenosquamous carcinoma Stage at diagnosis‡ I II III/IV
Normal-onset cases (n = 329)
Possible rapidonset cases (n = 111)
Rapid-onset cases (n = 43)
29 (8.8%) 74 (22.5%) 63 (19.2%) 65 (19.8%) 98 (29.8%)
23 (20.7%) 43 (38.7%) 19 (17.1%) 17 (15.3%) 9 (8.1%)
18 (41.9%) 12 (27.9%) 8 (18.6%) 4 (9.3%) 1 (2.3%)
249 (78.6%) 68 (21.4%)
94 (87.9%) 13 (12.1%)
42 (97.7%) 1 (2.3%)
282 (85.7%) 47 (14.3%)
78 (70.3%) 33 (29.7%)
25 (58.1%) 18 (41.9%)
160 (54.4%) 84 (28.6%) 50 (17.0%)
68 (67.4%) 27 (26.7%) 6 (5.9%)
29 (82.9%) 2 (5.7%) 4 (11.4%)
Statistical significance* P = .001
P = .002 P = .001
P = .001
*P value from Pearson χ2 test comparing distribution among 3 case groups. †Sixteen participants with missing information on race excluded. ‡Fifty-three participants with missing information on stage excluded.
disease used (see Methods), analysis that excluded those women classified as having rapid-onset disease because of a short latency between onset of sexual activity and diagnosis of cancer still suggested a younger age distribution of rapid-onset cases relative to normal-onset cases (P = .001). In addition, a real difference in the age distribution of rapid-onset and normal-onset cases is suggested by the fact that 20.7% of cases classified as having possible rapid-onset cancer were younger than 35 years old compared with 8.8% of normal-onset cases. Rapid-onset cases were more often white than normal-onset cases were. Only 2.3% of rapid-onset cases were not white compared with 21.4% of normal-onset cases and 12.1% of possible rapid-onset cases. Histologic diagnoses also varied among groups. Close to half of rapid-onset cases (41.9%) were diagnosed as having tumors with a glandular component compared with only 14.3% of normalonset cases. Women classified as having possible rapidonset disease had an intermediate proportion of cases diagnosed with a glandular tumor (29.7%). Finally, women diagnosed with rapid-onset disease were more likely to be diagnosed at an early stage (82.9% stage I compared with 54.4% and 67.4% for the normal-onset and possible rapid-onset groups, respectively). We next examined the association between rapid-onset disease and other nonviral factors. Overall findings are presented in Table II. Analyses stratified on age (<40 years and ≥40 years), Papanicolaou smear screening history (positive history of at least 1 smear in the 5 years preceding diagnosis), race (white), and histologic diagnosis (squamous carcinomas and adenocarcinoma or adenosquamous carcinomas) were also performed. Although women diagnosed with rapid-onset cancer were more likely than normal-
onset cases to report long-term oral contraceptive use (defined as use for >3 years)—51.2% versus 26.7%—this difference was largely the result of the fact that rapid-onset cases were young and therefore more likely to have used oral contraceptives. After age adjustment, long-term oral contraceptive users were found to have a nonsignificant 1.6-fold increased risk of rapid-onset disease compared with never users (95% confidence interval 0.61-4.4). In addition, when analysis was restricted to those women who reported a screening history in the 5 years preceding diagnosis, no effect of oral contraceptive use was observed (relative risk 0.99, 95% confidence interval 0.32-3.1 for long-term users relative to nonusers). An effect of longterm oral contraceptive use was observed for possible rapidonset cases (relative risk 2.7 for long-term users relative to nonusers), but this effect too was greatly reduced when analyses were restricted to women with a history of screening in the preceding 5 years (relative risk 1.5, 95% confidence interval 0.67-3.6 for long-term users relative to nonusers). Interestingly, the oral contraceptive use effect was greater among cases diagnosed with adenocarcinomas or adenosquamous carcinomas than among women diagnosed with squamous tumors. Among the group of women diagnosed with adenocarcinoma or adenosquamous carcinoma, long-term oral contraceptive users had a relative risk of rapid-onset disease of 3.4 compared with nonusers (95% confidence interval 0.61-19). The comparable estimate for cases diagnosed with squamous tumors was 1.2 (95% confidence interval 0.32-4.7). When cigarette smoking was examined, no clear pattern of association was observed. Women who reported smoking for ≤20 years were found to be at no increased risk of rapid-onset disease (relative risk 1.0, 95% confi-
Hildesheim et al 575
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
Table II. Distribution and risk of rapid-onset cervical disease associated with oral contraceptive use, cigarette smoking, number of pregnancies, and family history of cancer
Factor Oral contraceptive use† Never ≤3 y >3 y P trend Cigarette smoking‡ Never ≤20 y >20 y P trend No. of pregnancies§ 0-1 2-3 ≥4 P trend History of maternal cervical cancer No/don’t know Yes
Normal-onset cases Possible (n = 329) rapid-onset cases (No.) (n = 111) (No.).
Relative risk*
95% Confidence interval
Rapid-onset cases (n = 43) (No.)
Relative risk*
95% Confidence interval
137 (54.6%) 47 (18.7%) 67 (26.7%)
24 (23.3%) 32 (31.1%) 47 (45.6%)
1.0 2.8 2.7 P = .008
1.4-5.6 1.4-5.4
9 (20.9%) 12 (27.9%) 22 (51.2%)
1.0 1.4 1.6 P = .34
0.50-4.0 0.61-4.4
114 (37.0%) 67 (21.8%) 127 (41.2%)
37 (34.6%) 39 (36.4%) 31 (29.0%)
1.0 1.1 0.84 P = .65
0.62-2.0 0.49-1.5
18 (42.8%) 17 (40.5%) 7 (16.7%)
1.0 1.0 0.56 P = .31
0.48-2.2 0.21-1.5
63 (20.4%) 130 (42.1%) 116 (37.5%)
25 (23.6%) 50 (47.2%) 31 (29.2%)
1.0 0.96 0.85 P = .60
0.53-1.7 0.45-1.6
17 (42.5%) 15 (37.5%) 8 (20.0%)
1.0 0.54 0.49 P = .14
0.24-1.2 0.18-1.3
319 (97.0%) 10 (3.0%)
108 (97.3%) 3 (2.7%)
1.0 0.70
0.18-2.7
40 (93.0%) 3 (7.0%)
1.0 1.7
0.41-6.8
*Age-adjusted relative risk estimates calculated comparing possible rapid-onset and rapid-onset case groups with normal-onset case group. †Eighty-six participants with missing information on oral contraceptive use excluded. ‡Twenty-six participants with missing information on smoking excluded. §Twenty-eight participants with missing information on number of pregnancies excluded.
dence interval 0.48-2.2) and those who reported smoking for >20 years had a reduced relative risk of 0.56 (95% confidence interval 0.21-1.5) compared with never smokers. Similarly, in analyses comparing possible rapid-onset cases against normal-onset cases, the relative risk estimates were 1.1 (95% confidence interval 0.62-2.0) and 0.84 (95% confidence interval 0.49-1.5) for ≤20 years of smoking and >20 years of smoking, respectively. Number of pregnancies was not significantly associated with the development of rapid-onset cervical cancer. Compared with women with none or a single pregnancy, those who reported 2 to 3 and ≥4 pregnancies had ageadjusted relative risk and 95% confidence interval estimates of 0.54 (0.24-1.2) and 0.49 (0.18-1.3, respectively (P trend .14). The equivalent estimates when possible rapid-onset cases were compared with normal-onset cases were 0.96 (0.53-1.7) and 0.85 (0.45-1.6) (P trend .60). Similar results were obtained when women with no pregnancies were classified as the baseline category against which women who reported 1, 2-3, and ≥4 pregnancies were compared. A nonsignificant 70% increase in the risk of development of rapid-onset disease was observed among cases with a positive maternal history of cervical cancer (relative risk 1.7, 95% confidence interval 0.41-6.8). This effect was reduced to 50% in subsequent analyses restricted to women with a positive history of screening in the 5 years preceding diagnosis (relative risk 1.5, 95% confidence interval 0.29-8.0). All 3 women diagnosed with rapid-onset
cancer who reported a positive family history were younger than 40 years old. In analyses restricted to young women (<40 years old), those who reported a maternal history of cervical cancer were at a 7.1-fold increased risk of rapid-onset disease (95% confidence interval 1.0-49). A maternal history of cervical cancer was not associated with increased risk of possible rapid-onset disease. Results of analyses that examined the association between HPV and rapid-onset disease are presented in Table III. HPV DNA was detected in tumor samples from 209 (75.2%) of the 278 women for whom HPV data were available. Among those positive for HPV, 45 (21.5%) were found to be positive for HPV-18. The remaining 164 (78.5%) positives were found to have HPV types other than HPV-18. All but 1 of the subjects positive for HPV types other than HPV-18 were found to be positive for HPV-16. Compared with women positive for HPV-16 and other types, those positive for HPV-18 had a relative risk of rapid-onset cervical cancer of 1.6 (95% confidence interval 0.52-4.9) and a relative risk of possible rapid-onset disease of 0.67 (95% confidence interval 0.29-1.6). Comment Consistent with previous reports, we observed that women classified as having rapid-onset cervical cancer were more likely than normal-onset cases to be diagnosed with glandular tumors (adenocarcinomas and adenosquamous carcinomas).9 Although glandular tumors have been reported to account for 10% to 20% of cervical cancer cases,
576 Hildesheim et al
March 1999 Am J Obstet Gynecol
Table III. Distribution and risk of rapid-onset cervical disease associated with HPV DNA
Factor HPV HPV-16/other HPV-18 Negative
Normal-onset Possible cases (n = 187) rapid-onset (No.) cases (n = 66) (No.)
105 (56.2%) 30 (16.0%) 52 (27.8%)
45 (68.2%) 9 (13.6%) 12 (18.2%)
Relative risk*
1.0 0.67 0.62
95% Confidence interval
Rapid-onset cases (n = 25) (No.)
Relative risk*
95% Confidence interval
0.29-1.6 0.30-1.3
14 (56.0%) 6 (24.0%) 5 (20.0%)
1.0 1.6 0.96
0.52-4.9 0.31-3.0
*Age-adjusted relative risk estimates calculated comparing possible rapid-onset and rapid-onset case groups with normal-onset case group.
among our group of rapid-onset cases nearly 42% of cases were diagnosed with a carcinoma containing a glandular component.19 The suggestion that glandular tumors of the cervix are more aggressive than their squamous counterparts has been supported by some studies that observed a poorer prognosis of cases diagnosed with adenocarcinomas compared with squamous cervical tumors,20-22 whereas other studies have observed no differences in survival between these 2 groups.23, 24 Given that a large proportion of glandular tumors in our study were classified as rapid-onset disease, our study supports the notion that glandular tumors are more aggressive than the more commonly diagnosed squamous tumors. However, our observation that 83% of rapid-onset tumors were diagnosed at an early stage does not support the notion that rapid-onset disease is more aggressive than normal-onset tumors. The possibility exists that the excess of glandular tumors among our rapidonset cases is the result of false-negative Papanicolaou smear screening tests. Glandular tumors of the cervix are more likely than their squamous counterparts to arise in the endocervical canal and as such are more prone to being missed by routine Papanicolaou smear screening. Although we attempted to reduce misclassification by classifying as rapid onset only those women for whom review of previous normal Papanicolaou smear slides confirmed both the normalcy and adequacy of the smear (including presence of an endocervical component), prevalent abnormalities resulting from suboptimal sampling of the endocervical canal cannot be ruled out. Our results agree with previous observations that rapidonset cervical cancer occurs at a younger age than its normal-onset counterpart.9 Our age effect is partly the result of the criteria used to define rapid-onset disease (ie, invasive cancers diagnosed within 10 years of initiation of sexual activity were considered to be rapid onset). However, analyses that excluded women classified as having rapid-onset disease as a result of a short interval between onset of sexual activity and diagnosis of cancer still suggested a younger age distribution for rapid-onset relative to normal-onset cases. Our study is the first to systematically examine the possibility that rapid-onset cervical cancer is associated with particularly oncogenic HPV types, such as HPV-18. Results from our study do not suggest a strong association be-
tween HPV type and rapid-onset disease, although a nonsignificant 1.6-fold increase in risk of rapid-onset disease was observed. Because in our study type-specific primers were used to detect HPV-16 and HPV-18 and other types were detected with use of a single generic primer set (WD 72/76), it could be argued that type-specific associations with rapid-onset disease for HPV types other than HPV-16 and HPV-18 were missed. The fact that a larger proportion of normal-onset than rapid-onset cases were HPV negative (Table III), however, argues against this possibility. Another possibility is that rapid-onset disease is associated with specific HPV variants. Our study was not designed to examine whether variants of HPV-16 or HPV-18 are associated with risk of rapid-onset cervical cancer. Should continuing studies demonstrate functional differences in the biologic activity of different HPV variants, examination of the association of these variants with rapidonset disease would be warranted. Among the other exogenous factors examined, none were found to be significantly associated with rapid-onset cervical cancer. The apparent excess of oral contraceptive use among rapid-onset cases was accounted for in large part by differences in age and Papanicolaou smear screening habits. It is noteworthy, however, that among women in our study with adenocarcinomas or adenosquamous carcinomas long-term oral contraceptive users were at a 3.4fold increased risk of rapid-onset disease, a finding that is consistent with previous reports that have found that oral contraceptive use is strongly linked to the development of cervical tumors with a glandular component.25 Although a small proportion of cases reported a mother with a history of cervical cancer, our observation that women diagnosed with rapid-onset disease were more likely to report such a history and that the effect of a positive family history was particularly striking among young women is worth mention. Focused studies of these early-onset cervical cancers might provide clues as to host factors (including host immune handling of HPV) involved in the pathogenesis of this disease. In summary, our results confirm that women in whom rapid-onset cervical cancer develops are more likely than women diagnosed with normal-onset disease to have a tumor containing a glandular component. Also, our find-
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
ings indicate that rapid-onset cervical cancer is more likely than its normal-onset counterpart to develop among women younger than 35 years old. Examination of HPV and other factors, including oral contraceptive use, cigarette smoking, number of pregnancies, and a family history of cervical cancer, suggests that none of these factors are strongly linked to the development of rapid-onset disease. REFERENCES
1. Schiffman MH. Recent progress in defining the epidemiology of human papillomavirus infection and cervical neoplasia. J Natl Cancer Inst 1992;84:394-8. 2. Hildesheim A, Schiffman MH, Gravitt P, Glass AG, Greer C, Zhang T, et al. Persistence of type-specific human papillomavirus infection among cytologically normal women in Portland, Oregon. J Infect Dis 1994;169:235-40. 3. Nasiell K, Roger V, Nasiell M. Behavior of mild cervical dysplasia during long-term follow-up. Obstet Gynecol 1986;67:665-9. 4. Ho GYF, Bierman R, Beardsley L, Chang CJ, Burk RD. Natural history of cervicovaginal papillomavirus infection in young women. N Engl J Med 1998;338:423-8. 5. Pinten J, Adami H, Bergstrom R, Dillner J, Friberg L, Gustafsson F. Strategies for global control of cervical cancer. Int J Cancer 1995;60:1-26. 6. Bain RW, Crocker DW. Rapid onset of cevical cancer in an upper socioeconomic group. Am J Obstet Gynecol 1983;146:366-71. 7. Figge DC, Bennington JL, Schweid AI. Cervical cancer after initial negative and atypical vaginal cytology. Am J Obstet Gynecol 1970;108:422-8. 8. Dunn JE, Crocker DW, Rube IF, Erickson CC, Coleman SA. Cervical cancer occurence in Memphis and Shelby County, Tennessee, during 25 years of its cervical cytology screening program. Am J Obstet Gynecol 1984;150:861-4. 9. Benoit AG, Krepart GV, Lotocki RJ. Results of prior cytologic screening in patients with a diagnosis of stage I carcinoma of the cervix. Am J Obstet Gynecol 1984;148:690-4. 10. Janerich DT, Hadjimichael O, Schwartz PE, Lowell DM, Meigs JW, Merino MJ, et al. The screening histories of women with invasive cervical cancer, Connecticut. Am J Public Health 1995;85:791-4. 11. Koutsky LA, Holmes KK, Critchlow CW, Stevens CE, Paavonen J, Beckmann AM, et al. A cohort study of the risk of cervical intraepithelial neoplasia grade 2 or 3 in relation to papillomavirus infection. N Engl J Med 1992;327:1272-8.
Hildesheim et al 577
12. Barnes W, Woodworth C, Waggoner S, Stoler M, Jenson AB, Delgado G, et al. Rapid dysplastic transformation of human genital cells by human papillomavirus type 18. Gynecol Oncol 1990;38:343-6. 13. Kurman RJ, Schiffman MH, Lancaster WD, Reid R, Jenson AB, Temple GF, et al. Analysis of individual human papillomavirus types in cervical neoplasia: a possible role for type 18 in rapid progression. Am J Obstet Gynecol 1988;159:293-6. 14. Hadjimichael O, Janerich DT, Lowell DM, Meigs JW, Merino MJ, Schwartz PE. Histologic and clinical characteristics associated with rapidly progressive invasive cervical cancer: a preliminary report from the Yale cancer control research unit. Yale J Biol Med 1989;62:345-50. 15. Greer CE, Wheeler CM, Manos MM. PCR amplification from paraffin-embedded tissues: sample preparation and the effects of fixation. In: Dieffenbach CW, Dveksler GS, editors. PCR primer: a laboratory manual. New York: Cold Spring Harbor Laboratory Press; 1995. p 99-112. 16. Van den Brule AJC, Meijer CJLM, Bakels V, Kenemans P, Walboomers JMM. Rapid detection of human papillomavirus in cervical scrapes by combined general primer-mediated type-specific polymerase chain reaction. J Clin Microbiol 1990;28:273943. 17. Tidy JA, Parry GCN, Ward P, Coleman DV, Peto J, Malcolm ADB, et al. High rate of human papillomavirus type 16 infection in cytologically normal cervices. Lancet 1989;1:434. 18. Resnick RM, Cornelissen MTE, Wright DK, Eichinger GH, Fox HS, ter Schegget J, et al. Detection and typing of human papillomavirus in archival cervical cancer specimens by DNA amplification with consensus primers. J Natl Cancer Inst 1990;82:1477-84. 19. Kjaer SK, Brinton LA. Adenocarcinomas of the uterine cervix: the epidemiology of an increasing problem. Epidemiol Rev 1993;15:486-98. 20. Brand E, Berek JS, Hacker NF. Controversies in the management of cervical adenocarcinoma. Obstet Gynecol 1988;71:261-9. 21. Eide TJ. Cancer of the uterine cervix in Norway by histologic type, 1970-84. J Natl Cancer Inst 1987;79:199-205. 22. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet Gynecol 1991;77:912-7. 23. Platz CE, Benda JA. Female genital tract cancer. Cancer 1995;75:270-94. 24. Anton-Culver H, Bloss JD, Bringman D, Lee-Feldstein A, DiSaia P, Manetta A. Comparison of adenocarcinoma and squamous cell carcinoma of the uterine cervix: a population-based epidemiologic study. Am J Obstet Gynecol 1992;166:1507-14. 25. Ursin G, Peters RK, Henderson BE, d’Ablaing G III, Monroe KR, Pike MC. Oral contraceptive use and adenocarcinoma of cervix. Lancet 1994;344:1390-4.
Key words: Rapid-onset cervical cancer, human papillomaviruses, epidemiology
Cervical cancer typically develops over many years after initial infection with the human papillomavirus (HPV), persistence and progression of this infection to high-grade intraepithelial neoplasia, and eventual invasion. The incidence of HPV infection characteristically peaks soon after initiation of sexual activity, usually in the late teens and early twenties, and is closely followed by a smaller peak in the incidence of HPV-induced, cytologically evident low-grade cervical lesions.1 The majority of HPV infections and associated low-grade lesions regress without treatment, but in a minority of women they persist and progress to high-grade lesions.2-4 The peak inci-
From the National Cancer Institute,a Yale University School of Medicine,b the University of New Mexico School of Medicine,c and Georgetown University.d Received for publication July 9, 1998; revised September 2, 1998; accepted November 16, 1998. Reprint requests: Allan Hildesheim, PhD, Interdisciplinary Studies Section, Environmental Epidemiology Branch, DCEG, National Cancer Institute, 6130 Executive Blvd, Room 443, Rockville, MD 20852. 6/1/95843
dence of high-grade lesions occurs in the late twenties and early thirties, a median of 5 to 10 years after the peak occurrence of low-grade dysplasia.1 Invasion occurs on average a decade or more later, as the result of as-yet-unknown final oncogenic events. The majority of invasive cervical tumors are diagnosed after age 45 years. The long preclinical phase of cervical neoplasia and the existence of observable precursor lesions have made cervical cancer screening by Papanicolaou smears possible. Papanicolaou smear screening has proved to be highly successful in countries where widespread screening has been implemented for several decades. This is demonstrated by the drastic reduction of cervical cancer incidence and mortality in countries where successful screening programs have been established.5 In spite of the success of Papanicolaou smear programs in reducing the occurrence of invasive cervical cancer, cervical cancer will develop in a small proportion of women during the course of routine Papanicolaou smear screening, suggesting a potential for a much-faster disease progression than is the norm.6 The notion of 571
572 Hildesheim et al
“rapid-onset cervical cancer” was suggested by Bain and Crocker6 who noted that as many as 25% of invasive cervical cancer cases could be classified as rapid onset on the basis of an interval of less than 3 years between a normal cytologic screening and the development of invasive disease. Others have reported a similar phenomenon.7-9 It is now understood that rapid-onset cervical cancer as a clinical entity results at least in part from inadequate screening (inadequate specimen collection, Papanicolaou smear interpretation, or follow-up).10 However, it is probable that in a subgroup of women defined clinically as having rapid-onset disease the disease does in fact arise quickly. Indirect evidence for truly rapid-onset disease is provided by studies that have indicated that high-grade cervical lesions, the immediate precursor of invasive cervical cancer, can arise within short time periods without evidence of preceding low-grade disease.11 In addition, the fact that cervical cancer is sometimes diagnosed among women in their late teens and early twenties suggests that the phenomenon of rapid-onset disease, although uncommon, exists. It is important to determine whether the etiology of these rapidly progressing cases is different from that of the more commonly observed normal-onset cases. One possibility is that true rapid-onset cases are associated with particularly oncogenic HPV types. It is known that, of the more than 30 HPV types that infect the cervix, a small subgroup accounts for the majority of cervical cancer diagnosed worldwide (HPV types 16, 18, 31, 33, and 45). Among HPV types known to be associated with cancer development, heterogeneity in the transforming ability has been suggested in laboratory and clinical studies.12, 13 Most notable are the data that suggest that HPV-18 has a higher transforming ability than HPV-16.12 It is thus plausible that HPV-18 may be strongly linked to the development of rapid-onset disease.13 In addition to HPV, persons in whom rapid-onset cervical cancer develops may differ from those in whom normal-onset disease develops with respect to other factors associated with cervical cancer, such as oral contraceptive use, cigarette smoking, and reproductive behavior. The current study was designed to examine whether the distribution of HPV types and other cervical cancer risk factors is different among rapid-onset cervical cancer cases compared with the more common normal-onset disease. To examine this issue, we investigated 483 women diagnosed with cervical cancer in Connecticut over a 5year period; these women were classified as having been diagnosed with either rapid- or normal-onset cervical disease, and their risk factor profile was compared. Methods The current study is part of a population-based investigation undertaken in Connecticut to examine patterns of cervical cancer screening and ways in which screening could be improved to further reduce cervical cancer incidence.10, 14
March 1999 Am J Obstet Gynecol
After institutional review board approval, all histologically confirmed incident cervical cancer cases were sought by a rapid ascertainment system set up through the Connecticut tumor registry. During the period from March 1, 1985, through February 23, 1990, 664 cervical cancer cases were diagnosed among Connecticut residents. Information on Papanicolaou smear screening history and on exposure to known or suspected cervical cancer risk factors was obtained for 483 of the 664 eligible women (73%) by a structured telephone interview and review of physician records. Nonresponse was accounted for by physician refusal (12%), patient refusal (6%), and inability to locate subject or next of kin (9%). An attempt was made to retrieve and review cytology slides and reports from the last 3 Papanicolaou tests performed on study participants before the diagnosis of cancer. Because of the nature of laws regulating the length of storage of Papanicolaou smears by cytology laboratories, we were most successful at obtaining slides performed in the 2 to 3 years preceding diagnosis. Thus we were successful at retrieving slides for 79% of cases who reported at least 1 Papanicolaou smear in the 3 years preceding diagnosis and 67% of all cases who reported ever having had a Papanicolaou smear. All slides were reviewed by a single expert pathologist (D.M.L.) who was masked as to the original diagnosis. In cases where a discrepancy was noted between the original diagnosis and that of the study pathologist, a second study pathologist (M.J.M.) was asked to review the slides and a consensus diagnosis was reached. On the basis of reported Papanicolaou smear screening history, slide review, and questionnaire data, women were classified as having “rapid-onset cervical cancer” if the following criteria were met: (1) at least 1 normal Papanicolaou smear was taken within 5 years of cancer diagnosis and review of the slide confirmed normalcy and adequacy of the sample (including presence of endocervical cells) or (2) the diagnosis of cancer was made within 10 years of having initiated sexual activity. In addition, women were classified as having “possible-rapid onset cervical cancer” if the following criterion was met: At least 1 normal Papanicolaou smear was taken within 5 years of diagnosis but review (to confirm normalcy and adequacy) was not possible or incomplete. Forty-three women met the criteria for rapid-onset disease (31 on the basis of screening history, 8 on the basis of the interval since initiation of sexual activity, and 4 on the basis of both criteria) and an additional 111 women were classified as having possible rapid-onset disease on the basis of the criterion listed. The remaining 329 patients were classified as having “normal-onset cervical cancer.” Among the 329 patients classified as having normal-onset cervical cancer, 137 reported never having been screened and 110 were screened more than 5 years preceding the diagnosis of cancer. The remaining 82 women had a Papanicolaou smear within 5 years of diagnosis, but
Hildesheim et al 573
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
the smear was either incorrectly read as normal (n = 30) or the abnormal reading was inadequately followed up (n = 52), leading to the development of cancer. We also defined rapid-onset disease with a 3-year instead of a 5-year cutoff level. Analyses with this alternative definition yielded similar results, so results with the 5-year cutoff value are presented. To assess the presence and type of HPV in tumors from participants in our study, we attempted to retrieve histologic blocks from all study participants. We were successful in this endeavor for 379 (78%) of the 483 participants. From each available block, one 5-µm hematoxylineosin–stained slide followed by two 10-µm curled sections were prepared. Stained slides were reviewed (J.W.) to confirm the presence of invasive tumor material, and in cases where more than 1 block was available per woman the most representative block or blocks were selected for HPV deoxyribonucleic acid (DNA) testing. In selecting specimens for HPV DNA testing, blocks with extensive necrotic tissue or blood were avoided. Among the 379 women for whom blocks were retrieved, reviewed, and tested, 93% had a single block selected for HPV testing, 6% had 2 blocks selected for testing, and 1% had 3 blocks selected for testing. One 10-µm section was used for HPV DNA testing. On arrival at the laboratory, samples were deparaffinized and DNA was extracted as previously described.15 In brief, samples were centrifuged at 14,000g for 30 seconds, deparaffinized in 1 mL of octane, extracted in 1 mL of absolute ethanol, and resuspended in 50 to 100 µL of digestion buffer (50 mmol/L TRIS [trishydroxymethylaminomethane], pH 8.5, 1 mmol/L EDTA [ethylenediaminetetra-acetic acid], 400 µg/mL proteinase K, 4 mmol/L magnesium chloride, and 1% laureth-12 detergent). Resultant crude digests were subjected to heat inactivation at 95°C for 10 minutes. Specimens were amplified for human β-globin (using the GH20 and PC04 primer pair, which targets a 268-bp product) and with the 4 HPV primer systems described below. Two HPV primer sets that target the L1 (HPV-16 L1)16 and URR (HPV-16 URR)17 regions of HPV-16 were used, as was 1 primer set targeting the L1 region of HPV-18 (HPV-18 L1)16 and a degenerate primer system (WD72/76) that targets the HPV E6 gene of a broad spectrum of HPV types.18 Amplification was carried out under the following conditions: Five microliters of crude digest was added to 100 µL of master mix containing 2.5 mmol/L Mg++ and a standard buffer system. The reactions were amplified in a Perkin-Elmer 9600 for 40 cycles. The following amplification profile was used for all primer pairs: 95°C for 1 minute, 55°C for 1 minute, and 72°C for 2 minutes, followed by a 5-minute extension at 72°C. Detection of polymerase chain reaction amplicons was carried out as follows: For the HPV-16 and HPV-18 type-specific primers both ethidium bromide gel analysis and hybridization assays were performed as previously described.16 The WD72/76 reactions were evaluated by hybridization with a battery of HPV type-specific probes pre-
viously described.18 HPV-16 URR amplification products were evaluated by staining 2% agarose gels with ethidium bromide after electrophoresis.17 A woman was considered positive for HPV-18 if the tumor tissue tested positive for HPV-18 with either the HPV-18 L1 or the WD72/76 primer sets. Similarly, a woman was classified as positive for types other than HPV-18 if the tumor tissue tested negative for HPV-18 and positive for HPV-16 or other HPV types with either the HPV-16 L1 primer set, the HPV-16 URR primer set, or the WD72/76 primer set. Finally, women were classified as HPV DNA negative if the tumor tissue tested negative to all 4 HPV primer sets used but tested positive for β-globin amplification. One hundred one women (27% of those tested) whose samples tested negative for all HPV primer sets and to the β-globin primers were considered inadequate and excluded. Our finding that 27% of samples were inadequate for HPV testing and that inadequate samples tended to be older (P < .001) confirms the well-known difficulty of testing for multiple HPV types out of aging fixed tissue.15 Contingency table analyses were performed to compare the distribution of various factors among women classified as having rapid-onset, possible rapid-onset, and normalonset disease. Factors examined included age, race, histologic features, stage at diagnosis, HPV, reproductive history, oral contraceptive use, smoking, and a maternal history of cervical cancer. Pearson’s χ2 test was used to evaluate whether differences between groups were statistically significant. Logistic regression analyses were also performed comparing rapid-onset and possible rapid-onset cases against women diagnosed with normal-onset cervical cancer. Because HPV is known to be present in approximately 90% of cervical cancer lesions, HPV-16/other–positive women were considered as the referent group in analyses that examined the association between HPV and rapid-onset disease. The statistical significance of risk estimates was determined by the 95% confidence interval. A confidence interval that excludes 1.0 can be interpreted as being statistically significant. The significance of trends was tested for by including the categoric variable of interest as continuous in the logistic model. Results Analysis was performed on the 329 women classified as having normal-onset cervical cancer, 111 women classified as having possible rapid-onset cancer, and 43 women with rapid-onset disease. Women diagnosed with rapidonset tumors were found to differ significantly from those diagnosed with normal-onset disease with respect to all demographic and clinicohistologic factors examined (Table I). Women with rapid-onset disease were younger than those with normal-onset disease; 41.9% of rapid-onset cases were younger than 35 years old compared with 8.8% of normal-onset cases. Although this finding is partly the result of the definition of rapid-onset
574 Hildesheim et al
March 1999 Am J Obstet Gynecol
Table I. Demographic and clinicohistologic features of cervical cancer cases by rapid-onset status
Factor Age at diagnosis <35 y 35-44 y 45-54 y 55-64 y ≥65 y Race† White Other Histologic diagnosis Squamous cell Adenocarcinoma or adenosquamous carcinoma Stage at diagnosis‡ I II III/IV
Normal-onset cases (n = 329)
Possible rapidonset cases (n = 111)
Rapid-onset cases (n = 43)
29 (8.8%) 74 (22.5%) 63 (19.2%) 65 (19.8%) 98 (29.8%)
23 (20.7%) 43 (38.7%) 19 (17.1%) 17 (15.3%) 9 (8.1%)
18 (41.9%) 12 (27.9%) 8 (18.6%) 4 (9.3%) 1 (2.3%)
249 (78.6%) 68 (21.4%)
94 (87.9%) 13 (12.1%)
42 (97.7%) 1 (2.3%)
282 (85.7%) 47 (14.3%)
78 (70.3%) 33 (29.7%)
25 (58.1%) 18 (41.9%)
160 (54.4%) 84 (28.6%) 50 (17.0%)
68 (67.4%) 27 (26.7%) 6 (5.9%)
29 (82.9%) 2 (5.7%) 4 (11.4%)
Statistical significance* P = .001
P = .002 P = .001
P = .001
*P value from Pearson χ2 test comparing distribution among 3 case groups. †Sixteen participants with missing information on race excluded. ‡Fifty-three participants with missing information on stage excluded.
disease used (see Methods), analysis that excluded those women classified as having rapid-onset disease because of a short latency between onset of sexual activity and diagnosis of cancer still suggested a younger age distribution of rapid-onset cases relative to normal-onset cases (P = .001). In addition, a real difference in the age distribution of rapid-onset and normal-onset cases is suggested by the fact that 20.7% of cases classified as having possible rapid-onset cancer were younger than 35 years old compared with 8.8% of normal-onset cases. Rapid-onset cases were more often white than normal-onset cases were. Only 2.3% of rapid-onset cases were not white compared with 21.4% of normal-onset cases and 12.1% of possible rapid-onset cases. Histologic diagnoses also varied among groups. Close to half of rapid-onset cases (41.9%) were diagnosed as having tumors with a glandular component compared with only 14.3% of normalonset cases. Women classified as having possible rapidonset disease had an intermediate proportion of cases diagnosed with a glandular tumor (29.7%). Finally, women diagnosed with rapid-onset disease were more likely to be diagnosed at an early stage (82.9% stage I compared with 54.4% and 67.4% for the normal-onset and possible rapid-onset groups, respectively). We next examined the association between rapid-onset disease and other nonviral factors. Overall findings are presented in Table II. Analyses stratified on age (<40 years and ≥40 years), Papanicolaou smear screening history (positive history of at least 1 smear in the 5 years preceding diagnosis), race (white), and histologic diagnosis (squamous carcinomas and adenocarcinoma or adenosquamous carcinomas) were also performed. Although women diagnosed with rapid-onset cancer were more likely than normal-
onset cases to report long-term oral contraceptive use (defined as use for >3 years)—51.2% versus 26.7%—this difference was largely the result of the fact that rapid-onset cases were young and therefore more likely to have used oral contraceptives. After age adjustment, long-term oral contraceptive users were found to have a nonsignificant 1.6-fold increased risk of rapid-onset disease compared with never users (95% confidence interval 0.61-4.4). In addition, when analysis was restricted to those women who reported a screening history in the 5 years preceding diagnosis, no effect of oral contraceptive use was observed (relative risk 0.99, 95% confidence interval 0.32-3.1 for long-term users relative to nonusers). An effect of longterm oral contraceptive use was observed for possible rapidonset cases (relative risk 2.7 for long-term users relative to nonusers), but this effect too was greatly reduced when analyses were restricted to women with a history of screening in the preceding 5 years (relative risk 1.5, 95% confidence interval 0.67-3.6 for long-term users relative to nonusers). Interestingly, the oral contraceptive use effect was greater among cases diagnosed with adenocarcinomas or adenosquamous carcinomas than among women diagnosed with squamous tumors. Among the group of women diagnosed with adenocarcinoma or adenosquamous carcinoma, long-term oral contraceptive users had a relative risk of rapid-onset disease of 3.4 compared with nonusers (95% confidence interval 0.61-19). The comparable estimate for cases diagnosed with squamous tumors was 1.2 (95% confidence interval 0.32-4.7). When cigarette smoking was examined, no clear pattern of association was observed. Women who reported smoking for ≤20 years were found to be at no increased risk of rapid-onset disease (relative risk 1.0, 95% confi-
Hildesheim et al 575
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
Table II. Distribution and risk of rapid-onset cervical disease associated with oral contraceptive use, cigarette smoking, number of pregnancies, and family history of cancer
Factor Oral contraceptive use† Never ≤3 y >3 y P trend Cigarette smoking‡ Never ≤20 y >20 y P trend No. of pregnancies§ 0-1 2-3 ≥4 P trend History of maternal cervical cancer No/don’t know Yes
Normal-onset cases Possible (n = 329) rapid-onset cases (No.) (n = 111) (No.).
Relative risk*
95% Confidence interval
Rapid-onset cases (n = 43) (No.)
Relative risk*
95% Confidence interval
137 (54.6%) 47 (18.7%) 67 (26.7%)
24 (23.3%) 32 (31.1%) 47 (45.6%)
1.0 2.8 2.7 P = .008
1.4-5.6 1.4-5.4
9 (20.9%) 12 (27.9%) 22 (51.2%)
1.0 1.4 1.6 P = .34
0.50-4.0 0.61-4.4
114 (37.0%) 67 (21.8%) 127 (41.2%)
37 (34.6%) 39 (36.4%) 31 (29.0%)
1.0 1.1 0.84 P = .65
0.62-2.0 0.49-1.5
18 (42.8%) 17 (40.5%) 7 (16.7%)
1.0 1.0 0.56 P = .31
0.48-2.2 0.21-1.5
63 (20.4%) 130 (42.1%) 116 (37.5%)
25 (23.6%) 50 (47.2%) 31 (29.2%)
1.0 0.96 0.85 P = .60
0.53-1.7 0.45-1.6
17 (42.5%) 15 (37.5%) 8 (20.0%)
1.0 0.54 0.49 P = .14
0.24-1.2 0.18-1.3
319 (97.0%) 10 (3.0%)
108 (97.3%) 3 (2.7%)
1.0 0.70
0.18-2.7
40 (93.0%) 3 (7.0%)
1.0 1.7
0.41-6.8
*Age-adjusted relative risk estimates calculated comparing possible rapid-onset and rapid-onset case groups with normal-onset case group. †Eighty-six participants with missing information on oral contraceptive use excluded. ‡Twenty-six participants with missing information on smoking excluded. §Twenty-eight participants with missing information on number of pregnancies excluded.
dence interval 0.48-2.2) and those who reported smoking for >20 years had a reduced relative risk of 0.56 (95% confidence interval 0.21-1.5) compared with never smokers. Similarly, in analyses comparing possible rapid-onset cases against normal-onset cases, the relative risk estimates were 1.1 (95% confidence interval 0.62-2.0) and 0.84 (95% confidence interval 0.49-1.5) for ≤20 years of smoking and >20 years of smoking, respectively. Number of pregnancies was not significantly associated with the development of rapid-onset cervical cancer. Compared with women with none or a single pregnancy, those who reported 2 to 3 and ≥4 pregnancies had ageadjusted relative risk and 95% confidence interval estimates of 0.54 (0.24-1.2) and 0.49 (0.18-1.3, respectively (P trend .14). The equivalent estimates when possible rapid-onset cases were compared with normal-onset cases were 0.96 (0.53-1.7) and 0.85 (0.45-1.6) (P trend .60). Similar results were obtained when women with no pregnancies were classified as the baseline category against which women who reported 1, 2-3, and ≥4 pregnancies were compared. A nonsignificant 70% increase in the risk of development of rapid-onset disease was observed among cases with a positive maternal history of cervical cancer (relative risk 1.7, 95% confidence interval 0.41-6.8). This effect was reduced to 50% in subsequent analyses restricted to women with a positive history of screening in the 5 years preceding diagnosis (relative risk 1.5, 95% confidence interval 0.29-8.0). All 3 women diagnosed with rapid-onset
cancer who reported a positive family history were younger than 40 years old. In analyses restricted to young women (<40 years old), those who reported a maternal history of cervical cancer were at a 7.1-fold increased risk of rapid-onset disease (95% confidence interval 1.0-49). A maternal history of cervical cancer was not associated with increased risk of possible rapid-onset disease. Results of analyses that examined the association between HPV and rapid-onset disease are presented in Table III. HPV DNA was detected in tumor samples from 209 (75.2%) of the 278 women for whom HPV data were available. Among those positive for HPV, 45 (21.5%) were found to be positive for HPV-18. The remaining 164 (78.5%) positives were found to have HPV types other than HPV-18. All but 1 of the subjects positive for HPV types other than HPV-18 were found to be positive for HPV-16. Compared with women positive for HPV-16 and other types, those positive for HPV-18 had a relative risk of rapid-onset cervical cancer of 1.6 (95% confidence interval 0.52-4.9) and a relative risk of possible rapid-onset disease of 0.67 (95% confidence interval 0.29-1.6). Comment Consistent with previous reports, we observed that women classified as having rapid-onset cervical cancer were more likely than normal-onset cases to be diagnosed with glandular tumors (adenocarcinomas and adenosquamous carcinomas).9 Although glandular tumors have been reported to account for 10% to 20% of cervical cancer cases,
576 Hildesheim et al
March 1999 Am J Obstet Gynecol
Table III. Distribution and risk of rapid-onset cervical disease associated with HPV DNA
Factor HPV HPV-16/other HPV-18 Negative
Normal-onset Possible cases (n = 187) rapid-onset (No.) cases (n = 66) (No.)
105 (56.2%) 30 (16.0%) 52 (27.8%)
45 (68.2%) 9 (13.6%) 12 (18.2%)
Relative risk*
1.0 0.67 0.62
95% Confidence interval
Rapid-onset cases (n = 25) (No.)
Relative risk*
95% Confidence interval
0.29-1.6 0.30-1.3
14 (56.0%) 6 (24.0%) 5 (20.0%)
1.0 1.6 0.96
0.52-4.9 0.31-3.0
*Age-adjusted relative risk estimates calculated comparing possible rapid-onset and rapid-onset case groups with normal-onset case group.
among our group of rapid-onset cases nearly 42% of cases were diagnosed with a carcinoma containing a glandular component.19 The suggestion that glandular tumors of the cervix are more aggressive than their squamous counterparts has been supported by some studies that observed a poorer prognosis of cases diagnosed with adenocarcinomas compared with squamous cervical tumors,20-22 whereas other studies have observed no differences in survival between these 2 groups.23, 24 Given that a large proportion of glandular tumors in our study were classified as rapid-onset disease, our study supports the notion that glandular tumors are more aggressive than the more commonly diagnosed squamous tumors. However, our observation that 83% of rapid-onset tumors were diagnosed at an early stage does not support the notion that rapid-onset disease is more aggressive than normal-onset tumors. The possibility exists that the excess of glandular tumors among our rapidonset cases is the result of false-negative Papanicolaou smear screening tests. Glandular tumors of the cervix are more likely than their squamous counterparts to arise in the endocervical canal and as such are more prone to being missed by routine Papanicolaou smear screening. Although we attempted to reduce misclassification by classifying as rapid onset only those women for whom review of previous normal Papanicolaou smear slides confirmed both the normalcy and adequacy of the smear (including presence of an endocervical component), prevalent abnormalities resulting from suboptimal sampling of the endocervical canal cannot be ruled out. Our results agree with previous observations that rapidonset cervical cancer occurs at a younger age than its normal-onset counterpart.9 Our age effect is partly the result of the criteria used to define rapid-onset disease (ie, invasive cancers diagnosed within 10 years of initiation of sexual activity were considered to be rapid onset). However, analyses that excluded women classified as having rapid-onset disease as a result of a short interval between onset of sexual activity and diagnosis of cancer still suggested a younger age distribution for rapid-onset relative to normal-onset cases. Our study is the first to systematically examine the possibility that rapid-onset cervical cancer is associated with particularly oncogenic HPV types, such as HPV-18. Results from our study do not suggest a strong association be-
tween HPV type and rapid-onset disease, although a nonsignificant 1.6-fold increase in risk of rapid-onset disease was observed. Because in our study type-specific primers were used to detect HPV-16 and HPV-18 and other types were detected with use of a single generic primer set (WD 72/76), it could be argued that type-specific associations with rapid-onset disease for HPV types other than HPV-16 and HPV-18 were missed. The fact that a larger proportion of normal-onset than rapid-onset cases were HPV negative (Table III), however, argues against this possibility. Another possibility is that rapid-onset disease is associated with specific HPV variants. Our study was not designed to examine whether variants of HPV-16 or HPV-18 are associated with risk of rapid-onset cervical cancer. Should continuing studies demonstrate functional differences in the biologic activity of different HPV variants, examination of the association of these variants with rapidonset disease would be warranted. Among the other exogenous factors examined, none were found to be significantly associated with rapid-onset cervical cancer. The apparent excess of oral contraceptive use among rapid-onset cases was accounted for in large part by differences in age and Papanicolaou smear screening habits. It is noteworthy, however, that among women in our study with adenocarcinomas or adenosquamous carcinomas long-term oral contraceptive users were at a 3.4fold increased risk of rapid-onset disease, a finding that is consistent with previous reports that have found that oral contraceptive use is strongly linked to the development of cervical tumors with a glandular component.25 Although a small proportion of cases reported a mother with a history of cervical cancer, our observation that women diagnosed with rapid-onset disease were more likely to report such a history and that the effect of a positive family history was particularly striking among young women is worth mention. Focused studies of these early-onset cervical cancers might provide clues as to host factors (including host immune handling of HPV) involved in the pathogenesis of this disease. In summary, our results confirm that women in whom rapid-onset cervical cancer develops are more likely than women diagnosed with normal-onset disease to have a tumor containing a glandular component. Also, our find-
Volume 180, Number 3, Part 1 Am J Obstet Gynecol
ings indicate that rapid-onset cervical cancer is more likely than its normal-onset counterpart to develop among women younger than 35 years old. Examination of HPV and other factors, including oral contraceptive use, cigarette smoking, number of pregnancies, and a family history of cervical cancer, suggests that none of these factors are strongly linked to the development of rapid-onset disease. REFERENCES
1. Schiffman MH. Recent progress in defining the epidemiology of human papillomavirus infection and cervical neoplasia. J Natl Cancer Inst 1992;84:394-8. 2. Hildesheim A, Schiffman MH, Gravitt P, Glass AG, Greer C, Zhang T, et al. Persistence of type-specific human papillomavirus infection among cytologically normal women in Portland, Oregon. J Infect Dis 1994;169:235-40. 3. Nasiell K, Roger V, Nasiell M. Behavior of mild cervical dysplasia during long-term follow-up. Obstet Gynecol 1986;67:665-9. 4. Ho GYF, Bierman R, Beardsley L, Chang CJ, Burk RD. Natural history of cervicovaginal papillomavirus infection in young women. N Engl J Med 1998;338:423-8. 5. Pinten J, Adami H, Bergstrom R, Dillner J, Friberg L, Gustafsson F. Strategies for global control of cervical cancer. Int J Cancer 1995;60:1-26. 6. Bain RW, Crocker DW. Rapid onset of cevical cancer in an upper socioeconomic group. Am J Obstet Gynecol 1983;146:366-71. 7. Figge DC, Bennington JL, Schweid AI. Cervical cancer after initial negative and atypical vaginal cytology. Am J Obstet Gynecol 1970;108:422-8. 8. Dunn JE, Crocker DW, Rube IF, Erickson CC, Coleman SA. Cervical cancer occurence in Memphis and Shelby County, Tennessee, during 25 years of its cervical cytology screening program. Am J Obstet Gynecol 1984;150:861-4. 9. Benoit AG, Krepart GV, Lotocki RJ. Results of prior cytologic screening in patients with a diagnosis of stage I carcinoma of the cervix. Am J Obstet Gynecol 1984;148:690-4. 10. Janerich DT, Hadjimichael O, Schwartz PE, Lowell DM, Meigs JW, Merino MJ, et al. The screening histories of women with invasive cervical cancer, Connecticut. Am J Public Health 1995;85:791-4. 11. Koutsky LA, Holmes KK, Critchlow CW, Stevens CE, Paavonen J, Beckmann AM, et al. A cohort study of the risk of cervical intraepithelial neoplasia grade 2 or 3 in relation to papillomavirus infection. N Engl J Med 1992;327:1272-8.
Hildesheim et al 577
12. Barnes W, Woodworth C, Waggoner S, Stoler M, Jenson AB, Delgado G, et al. Rapid dysplastic transformation of human genital cells by human papillomavirus type 18. Gynecol Oncol 1990;38:343-6. 13. Kurman RJ, Schiffman MH, Lancaster WD, Reid R, Jenson AB, Temple GF, et al. Analysis of individual human papillomavirus types in cervical neoplasia: a possible role for type 18 in rapid progression. Am J Obstet Gynecol 1988;159:293-6. 14. Hadjimichael O, Janerich DT, Lowell DM, Meigs JW, Merino MJ, Schwartz PE. Histologic and clinical characteristics associated with rapidly progressive invasive cervical cancer: a preliminary report from the Yale cancer control research unit. Yale J Biol Med 1989;62:345-50. 15. Greer CE, Wheeler CM, Manos MM. PCR amplification from paraffin-embedded tissues: sample preparation and the effects of fixation. In: Dieffenbach CW, Dveksler GS, editors. PCR primer: a laboratory manual. New York: Cold Spring Harbor Laboratory Press; 1995. p 99-112. 16. Van den Brule AJC, Meijer CJLM, Bakels V, Kenemans P, Walboomers JMM. Rapid detection of human papillomavirus in cervical scrapes by combined general primer-mediated type-specific polymerase chain reaction. J Clin Microbiol 1990;28:273943. 17. Tidy JA, Parry GCN, Ward P, Coleman DV, Peto J, Malcolm ADB, et al. High rate of human papillomavirus type 16 infection in cytologically normal cervices. Lancet 1989;1:434. 18. Resnick RM, Cornelissen MTE, Wright DK, Eichinger GH, Fox HS, ter Schegget J, et al. Detection and typing of human papillomavirus in archival cervical cancer specimens by DNA amplification with consensus primers. J Natl Cancer Inst 1990;82:1477-84. 19. Kjaer SK, Brinton LA. Adenocarcinomas of the uterine cervix: the epidemiology of an increasing problem. Epidemiol Rev 1993;15:486-98. 20. Brand E, Berek JS, Hacker NF. Controversies in the management of cervical adenocarcinoma. Obstet Gynecol 1988;71:261-9. 21. Eide TJ. Cancer of the uterine cervix in Norway by histologic type, 1970-84. J Natl Cancer Inst 1987;79:199-205. 22. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet Gynecol 1991;77:912-7. 23. Platz CE, Benda JA. Female genital tract cancer. Cancer 1995;75:270-94. 24. Anton-Culver H, Bloss JD, Bringman D, Lee-Feldstein A, DiSaia P, Manetta A. Comparison of adenocarcinoma and squamous cell carcinoma of the uterine cervix: a population-based epidemiologic study. Am J Obstet Gynecol 1992;166:1507-14. 25. Ursin G, Peters RK, Henderson BE, d’Ablaing G III, Monroe KR, Pike MC. Oral contraceptive use and adenocarcinoma of cervix. Lancet 1994;344:1390-4.