Temporal associations of human papillomavirus infection with cervical cytologic abnormalities

Temporal associations of human papillomavirus infection with cervical cytologic abnormalities Attila T. Lorincz, PhD,. Mark H. Schiffman, MD: William J. Jaffurs, MD,c John Marlow, MD,d Allison P. Quinn," and Gary F. Temple, MD" Gaithersburg and Bethesda, Maryland, and Washington, D.C. The relationship between infection with different human papillomavirus types and cervical intraepithelial neoplasia was studied in a group of 398 women seen in a private gynecology practice in Washington, D.C. Each woman was assessed for human papillomavirus infection by Southern blot hybridization analysis of cervical cells obtained by swab. The human papillomavirus results were correlated with the results of Papanicolaou smears taken the same day and with data abstracted from medical records regarding past cervical disease. Subjects with normal cytologic findings at the time of human papillomavirus testing were followed up for an average of 2 to 3 years with additional Papanicolaou smears. At the time of human papillomavirus testing, 58% (19 / 33) of women with cervical intraepithelial neoplasia had detectable human papillomavirus deoxyribonucleic acid in contrast to 10% (28 / 289) of women with normal cytologic findings (p < 0.001). This association persisted after statistical adjustment for age and current use of oral contraceptives, a factor that appeared to increase the detection of human papillomavirus. Among women with no current cytologic evidence of neoplasia, human papillomavirus detection was more likely in those with a history of past genital neoplasia (p = 0.05). In the follow-up study, 15% (3 of 20) of cytologically normal women who were human papillomavirus-positive at baseline subsequently exhibited cervical cells suggestive of cervical intraepithelial neoplasia compared with only 5% (9 of 195) of human papillomavirus-negative women. However, this difference reflected recurrent and not incident neoplasia. (AM J OasTET GVNECOL 1990;162:645-51 .)

Key words: Human papillomavirus, cervical intraepithelial neoplasia, temporal associations Human papillomaviruses (HPV) are the etiologic agents of human warts and have been associated with a wide variety of benign and malignant epithelial lesions.I-4 Of the 60 types of HPV that are known, approximately 15 types are commonly found in the anogenital mucosa. One or more of these HPV types is found in the majority of cases of cervical intraepithelial neoplasia. HPV deoxyribonucleic acid (DNA) is also found in normal genital epithelia adjacent to HPV-containing lesions, as well as in women without any evidence of disease. 5 -7 It has been speculated that lesions that contain HPV arise from normal tissue containing HPV, and that HPV infection plays a central role in the genesis of premalignant and malignant lesions. Laboratory evidence to support this causal h ypothesis is strong and growing, but adequate epidemiologic support is not yet From the Department of Molecular Diagnostics, Bethesda Research Laboratories, Dzvision of Life Technologies, Inc., Gaithersburg: the Environmental Epzdemiology Branch, National Cancer Instztute, Bethesda/ and the Departments of Pathology' and Continumg Medical Education," Columbia Hospital for Women, Washmgton, D.C. Recezved for publicatIOn June 1, 1989,-accepted November 21, 1989_ R epnnt requests: Attila Lorincz, PhD, Department of Molecular Diagnostics, B ethesda R esearch Laboratories, DiVISion of Life Technologus, Inc., P.O. Box 6009, GaIthersburg, MD 20877. 611118438

available. Most cross-sectional population studies that associate HPV infection and cervical neoplasia have failed to consider potential confou nding factors that need to be considered before the association of HPV infection and cervical neoplasia is interpreted as causal. Prospective epidemiologic data with regard to the risk of HPV infection in cytologically normal women are lacking. This investigation was undertaken to clarify the relationship of HPV infection with cervical disease, taking into account possible confounding variables and focusing on temporal relationships between infection and disease. In particular, we sought to determine whether the presence of HPV DNA in women with normal cytologic findings is a short-term predictor of subsequent abnormal Papanicolaou smears.

Material and methods The study population included women of upper socioeconomic status who attend a private gynecology practice in Washington, D.C. that specializes in colposcopy and laser surgery (J. M.). The majority of the patients had been referred to the practice for one of the following reasons: abnormal Papanicolaou smears or genital condylomas, exposure to diethylstilbestrol in utero, or infertility. The remainder were seen for routine gynecologic care. Women were enrolled in the

645

646 Lorincz et al.

study between December 1984 and July 1985 when an initial Papanicolaou smear and HPV test were performed. Samples of exfoliated cervical cells for human papillomavirus testing were obtained from the endocervical canal and the transformation zone after Papanicolaou smear with a cotton-tipped swab and were placed in a tube containing 1 ml of TE buffer (10 mmol/L Tris hydrochloride and 1 mmol/L ethylenediaminetetraacetic acid, pH 8) plus 0.5% sodium dodecyl sulfate. The cells on the swab were dislodged by vigorous shaking for 15 seconds and the swab was discarded. Tubes with exfoliated cervical cells were refrigerated at 4°C for 1 to 4 days before transport to the HPV analysis laboratory (A. T. L, A. Q), then either processed immediately or frozen at - 70° C. HPV DNA was detected and typed by low-stringency and high-stringency Southern blotting procedures, described in detail elsewhere. s Briefly, cell samples were digested with proteinase K plus dithiothreitol, and DNA was extracted with phenol-chloroformisoamylalcohol and precipitated with ethanol. The DNA pellets were resuspended in TE buffer and aliquots were digested with Pst I. Samples were electrophoresed in 1% agarose gels and then blotted onto nitrocellulose. A photograph of the ethidium bromidestained gel and the use of known DNA standards allowed the inclusion of only samples with more than 2 /Lg of total DNA per lane. Of the original study population, about 5% of samples were found by ethidium bromide staining to be inadequate for HPV testing. Results are presented only for the 398 women with adequate HPV results. Southern blot filters were hybridized with DNA probes of HPV types 6, 11, and 16 prepared by nick translation (3 x lOs dpm/ /Lg) and washed under lowstringency conditions (- 35° C) to identify all HPV types including currently unknown types. To identify specific HPV types, nylon filters (Hybond-N; Amersham, Arlington Heights, Ill.) were prepared that contained immobilized DNA of all the HPV -positive samples. These filters were reprobed sequentially at high stringency with individual HPV DNAs of types 6, 11, 16, 18,31,33,35,42,43,44,45,51,52, and 56. All analyses for HPV DNA were performed in a masked fashion with no knowledge of the cytologic diagnoses. Moreover, to prevent falsely positive results, all HPV DNA probes were electrophoretically purified and samples were reprobed with pBR322 DNA to identify the presence of bacterial plasmid sequences. 6 All pBR-positive samples were regarded as negative for HPV unless they also had clearly identifiable HPV -reactive sequences. For the data analysis, the HPV typing data were combined into four biologically distinct groups, according

March 1990 Am J Obstet Gynecol

to their degree of association with cancer in previous investigations. HPV types 6, 11, 42, 43, and 44 were categorized as the "type 6 group." They are principally found in condylomas and low-grade intraepithelial neoplasia, and only rarely in cancer. 1·3,9 HPV types 31, 33, 35, 39, 45, 51, 52, and 56 were combined as the "type 31" group. They are associated with a wide variety of lesions including condylomas, intraepithelial neoplasia, and 20% to 25% of cancersJ.4 (Lorincz AT. Unpublished data). HPV -16 and HPV -18, the types most strongly associated with invasive disease, are found in 60% to 70% of cervical, vaginal, and vulvar cancers and in less severe lesions. I -4 Type 18 was not found in any samples from the study population; thus this group was presented in the results as "type 16." Although conceptually useful, the distinction between these three groups is not absolute. A fourth group, "uncharacterized HPV types," may include both new HPV types and HPV types for which probes were not used in this study. The study population was also categorized on the basis of results of the enrollment Papanicolaou smear (W. J. J.), masked to HPV results. Some diagnostic categories were combined to provide larger numbers: The "normal" group included individuals with either normal cervical cells or benign inflammatory cells, the "reactive atypia" class comprised patients with cells suggestive of metaplastic, glandular, or undefined atypia, and the "cervical intraepithelial neoplasia" grou p of patients had changes suggestive of flat condyloma or definite cervical intraepithelial neoplasia. The inclusion of subjects with condylomatous changes in the cervical intraepithelial neoplasia category follows the recent recommendations of the Bethesda conference on cervical cytologic nomenclature. lO Women who had cytologically normal findings and received no treatment at enrollment were followed up with repeat Papanicolaou smears as they returned for routine clinical care. Study follow-up ended on Sept. 30, 1988. During the course of follow-up, patients who were found to have abnormal Papanicolaou' smears were monitored by repeated Papanicolaou smear or colposcopy and had biopsy if suspicious areas were detected on the cervix after the application of 5% acetic acid. Histopathologically confirmed lesions were treated with either cryosurgery or the carbon dioxide laser. The medical record of each study participant was abstracted, masked to HPV results, to obtain a complete history of cervical pathologic conditions and treatments both before the enrollment date (past medical history) and during follow-up. Also abstracted were details regarding possible confounding variables that could affect the association of HPV infection and cervical neoplasia, including oral contraceptive use, day of the menstrual cycle, or pregnancy at the time of HPV testing,

Volume 162 Number 3

exposure to DES in utero, and age. Abstracts were completed by a trained medical abstractor who used a standardized form developed for this study. Ten percent of the abstracts were repeated (M. H. S.) and confirmed for accuracy. The measures of association used in the data analysis were Fisher's exact test (two-tail) for 2 x 2 tables and X' statistic for larger tables. II In the case of ordered categories (e.g., increasing age), the 1 df Mantel X' for trend was used. I. To permit statistical control of potential confounding variables, the cross-sectional data from enrollment were also analyzed with logistic regression models (BMDP LR) to provide estimates of the crude and adjusted relative risks associating cytologic abnormality with HPV infection. 13 A relative risk of 10.0 for HPV infection, for example, would indicate that cervical neoplasia occurs 10 times as often among infected women than among noninfected women. Ninety-five percent confidence intervals were also computed; if the interval excludes 1.0, this corresponds roughly to statistical significance at the level of p = 0.05. In the prospective analysis in which time of followup was a consideration, Cox proportional hazards modeling (BMDP 2L) was used. 13 For each subject in the prospective analysis, the date of enrollment was taken as the start of follow-up. For women with new abnormalities during follow-up, the exit date was considered the earliest date that a subsequent Papanicolaou smear or biopsy result showed cervical intraepithelial neoplasia. For women with no known new cervical neoplasia during follow-up, the censor date was taken as the date of the last normal Papanicolaou smear before Sept. 30, 1988. This multipart investigation can be viewed therefore as three discrete temporal analyses. First, the enrollment Papanicolaou smears and HPV tests were compared in a cross-sectional analysis, with consideration of potentially confounding variables assessed in the medical abstract. Then the issue of past disease that could affect current HPV results was considered by comparison of HPV results at enrollment with relevant details of past medical history from the patient records. Finally, the prospective aspect of the study was considered, associating the enrollment HPV results with subsequent clinical course. Results

The median age of the 398 women in the study population was 34 years (mean, 35; range, 15 to 79 years). Thirty-eight women in the group (10%) had previous hysterectomies according to their medical records; these women were excluded from analyses unless indicated. As shown in Table I, the prevalence of HPV at en-

Human papillomavirus and cervical neoplasia 647

rollment was strongly associated with cytologic diagnosis (p < 0.001). Ten percent of 289 women with cytologically normal cervices had HPV DNA of any type. The prevalence of HPV DNA in cervical cells of women with reactive atypia resembled the results for normal women. In contrast, 58% of the 33 women with cervical intraepithelial neoplasia had HPV DNA. HPV type-specific results yielded sparse data in many comparisons; thus statistical significance testing was not performed. A few apparent trends are noteworthy. Human papillomaviruses of the type 6 group were detected in similarly low percentages (:s3%) of samples from all diagnostic categories. In contrast, the prevalence of HPV-16 and the type 31 group was increased in women with cervical intraepithelial neoplasia, thus largely explaining the overall association of HPV positivity with severity of lesion. Infection of the same cervical sample with more than one HPV type was relatively uncommon. Of the 52 women infected with HPV DNA, only four (8%) had two HPV types, none had three HPV types, and preferential HPV type combinations in the samples were not apparent. We examined whether selected demographic and behavioral characteristics of the patients correlated with the diagnostic categories and with the presence of HPV DNA, and thus might confound the association of HPV infection with cervical intraepithelial neoplasia. Factors considered included current pregnancy, current oral contraceptive use, diethylstilbestrol exposure in utero, day of the menstrual cycle, and age divided into three categories «31, 31 to 40, >40 years). Women who were using oral contraceptives were significantly more likely to have cervical intraepithelial neoplasia (p = 0.03). Currently pregnant women were also slightly more likely to have cervical intraepithelial neoplasia (p = 0.10), as were older women (p for trend = 0.30). In contrast, women with a history of diethylstilbestrol exposure in utero were less likely to have cervical intraepithelial neoplasia (p = 0.24). Day of the menstrual cycle was not associated with cytologic diagnosis. With regard to HPV DNA detection, current oral contraceptive use was strongly associated with HPV positivity (p < 0.001). This relationship, shown in Table II, was most apparent among subjects with cytologically normal smears and those that showed reactive atypia. All type-specific HPV groups showed the effect of oral contraceptive use except the type 6 group, for which very small numbers prevented a meaningful comparison. Current pregnancy was also related to increased HPV detection (p = 0.08), but there were only 12 pregnant women in the study group; this association was mainly a result of three women with cervical intraepi-

648

Lorincz at al.

March 1990 Am J Obstet Gynecol

Table I. Prevalence of HPV DNA in exfoliated cervical cells from 360 women, by cytologic diagnosis and HPV type* HPV DNA resultt HPV-6 group

Any HPV:f: CytolOgiC dtagnosis

No.

Normal (n = 289) Reactive atypia (n = 38) Cervical intraepithelial neoplasia (N = 33)

28 5 19

I

%

No.

10 13 58

6 I

o

I

HPV-3J group

%

No.

2 3

7

o

3 8

I

Uncharacterized types

HPV-16

%

No.

2 8 24

o

8

8

I

%

No.

3

7 I 3

o

24

I

% 2

3 9

* Excludes 38 women with hystereclOmies. The HPV-6 group consists of types 6, 11,42,43, and 44 ; the HPV-31 group consists of types 31, 33, 35, 45, 51, 52, and 56. tHPV results include four double infections: one woman with types 33 and 44 classified as type 33, one woman with types 31 and 52 classified as type 31, and two women with types 33 and 52 classified as type 33. :j:Association of HPV detection (any type) with cytologic diagnosis is statistically significant (p < 0.001) .

Table II. HPV DNA detection in current users and nonusers of oral contraceptives by cytologic diagnosis* HPV DNA result Cytologtc diagnosis

Normal (n = 222) Current oral contraceptive user Nonuser Reactive atypia (n = 28) Current oral contraceptive user Nonuser CIN (n = 27) Current oral contraceptive user Nonuser

n

HPV detected (%)

pt

42 l80

26 6

<0.001

5 23

60 9

0.03

II

73 56

0.45

16

* Excludes 38 women with hysterectomies and an additional 83 women for whom current oral contraceptive use could not be ascertained by chart review. tp value for HPV detection among oral contraceptive current users versus nonusers, by cytologic diagnosis.

thelial neoplasia who had HPV type 16 DNA detected. A decrease in HPV detection was observed in the 10 1 women exposed to diethylstilbestrol compared with those not exposed (p = 0.03). This difference was most pronounced among women with Papanicolaou smears that showed reactive atypia. Day of the menstrual cycle was not significantly associated with HPV detection. HPV detection did appear initially to decrease with age (p for trend = 0.05), but this trend could be entirely explained as a correlate of oral contraceptive use, which declined strongly with age. When the women with cervical intraepithelial neoplasia were compared with those without neoplasia (reactive atypias grouped with normal smears), the overall estimated relative risk of neoplasia associated with HPV positivity was 12.1 (95 % confidence interval, 5.6 to 26.3). Statistical adjustment for the effects of oral contraceptive use, diethylstilbestrol history, or pregnancy by logistic regression modeling did not materially affect

this estimate, which remained above 10.0 in all analyses. Moreover, the strong association of oral contraceptive use with cervical intraepithelial neoplasia, which was apparent in the univariate analysis, disappeared when the HPV data were considered. Among women whose enrollment cytologic findings were normal, those with histories of genital neoplasia were more likely than those without such a history to have HPV detected (Table Ill). Similar effects were seen in women with abnormal cytologic findings at enrollment but were statistically nonsignificant, perhaps reflecting the small numbers of patients. Histories were categorized as positive for neoplasia if there was mention in the records of previous Papanicolaou smears or biopsy results that indicated cervical neoplasia (approximately 70% of the positive histories) , or if there was mention of cervical therapies generally prescribed for neoplasia, such as cryosurgery, laser surgery, or cone biopsy (another 20%). Approximately 10% of the

Volume 162 Number 3

pOSItive histories were a result of genital condylomas or carcinomas that did not involve the cervix; exclusion of these few women from the analysis did not change the overall results. The finding of increased HPV detection in women with histories of genital neoplasia was not restricted to anyone of the type-specific HPV groups. HPV results were examined separately for the 38 women with hysterectomies who had been excluded from the main analysis. Only 5% of these women had cytologic evidence of intraepithelial neoplasia in their vaginal specimens, less than the 9% with cervical intraepithelial neoplasia among the women with intact cervices. In contrast, the prevalence of HPV detection among the women with hysterectomies was found to be 21 %, slightly higher than the prevalence in women with intact cervices (14%). When women were stratified into two groups on the basis of history of cervical neoplasia, HPV was still found to be as prevalent in the women with hysterectomies as in the rest of the study group. To determine how often HPV-positive women who had initially cytologically normal findings would exhibit cytologic abnormalities on follow-up, prospective data were obtained on 215 (66%) of the 324 women in the study population with cytologically normal Papanicolaou smears at enrollment, including 30 of the 38 women with hysterectomies. The subjects for whom follow-up slides were available were older than the remainder of the study population and more likely to have a history of diethyk;tilbestrol exposure. There were no significant differences between the groups in oral contraceptive use or pregnancy. The women in the prospective study were classified according to history of genital neoplasia, and as positive or negative for HPV DNA (Table IV). Twelve women had Papanicolaou smears during follow-up that were suggestive of neoplasia. Although women with HPV detected at enrollment had a higher rate of apparent progression to disease (3 of 20, 15%) than did HPVnegative women (9 of 195, 5%), this comparison may be misleading because it ignored the influence of past medical history. In fact, 10 of the 12 apparent progressions occurred in women with a history of genital neoplasia. Therefore the analysis was limited to the subgroup of 123 women with normal cytologic findings at enrollment but who had no history of neoplasia. Within this group, 110 women were HPV-negative at enrollment; they were followed up for an average of 896.2 days, with a mean of 3.5 follow-up Papanicolaou smears. Thirteen women had HPV detected at enrollment; they were followed up for an average of 844.4 days and had a mean of 3.2 follow-up Papanicolaou smears. Thus the number of follow-up Papanicolaou smears and the length of follow-up did not differ

Human papillomavirus and cervical neoplasia 649

Table III. HPV DNA detection in 360 women according to history of genital neoplasia, by current cytologic diagnosis* Current cytolof51-c diagnosis

Normal (n = 289) Reactive atypia (n = 38) Cervical intraepithetial neoplasia (n = 33)

History of neoplasiat

HPV DNA detectIOn (%)

p:f:

No (n = 147) Yes (n = 142) No (n = 11)

6 13 0

0.05

Yes (n = 27) No (n = 6)

19 33

Yes (n = 27)

63

0.30 0.36

*Excludes 38 women with hysterectomies. t History includes Papanicolaou smear or cervical biopsy of cervical neoplasia; previous cryosurgery, laser surgery, or cone biopsy; or history of genital condylomas. +P value for HPV detection according to medical history, by cytologic diagnosis.

greatly by HPV status at enrollment. However, 23% of the HPV -positive women had recurrent disease during follow-up, compared with only 6% of HPV -negative women. The estimated relative risk of recurrence for HPV-positive women when the varying durations of follow-up for individual subjects were considered was 3.7 (95% confidence interval, 1.0 to 14.6). It should be emphasized that the women in whom recurrences were noted often had extensive histories of cervical disease and treatments. As the most extreme example, one of the three HPV-positive women was a 72-year-old woman with a previous history of cervical intraepithelial neoplasia and a hysterectomy, who developed vaginal intraepithelial neoplasia that was confirmed by biopsy. All other recurrences involved cervical neoplasia. Comment

The results show that HPV DNA is detectable by Southern blot hybridization in approximately 60% of women with cervical intraepithelial neoplasia, compared with only about 10% of women with normal cytologic findings or reactive atypia. The data also support previous findings that suggest that cervical intraepithelial neoplasia and cancers have a more restricted set of HPV types than benign cervical lesions and normal cervical epithelia. I . 4 A variety of the approximately 15 types of mucosotropic HPV were found in the samples from women with normal cytologic findings and reactive atypia. In contrast, women with cervical intraepithelial neoplasia more often were found to be infected with HPV type 16 and and the type 31 group.

650

Lorincz at al.

March 1990

Am J Obstet Gynecol

Table IV. Subsequent Papanicolaou smears showing neoplasia among 215 women with normal cytologic findings at the start of follow-up by enrollment HPV DNA results and history of genit~l neoplasia Subsequent neoplasia History of neoplasia*

No Yes

Enrollment HPV DNA result

No. of subjects

No HPV detected HPV detected No HPV detected HPV detected

85 7

110 13

Mean follow-up (days)

692.4 791.0 896.2 844.4

No.

2

o 7 3

1

%

2

o 6

23t

* History includes Papanicolaou smear or cervical biopsy result showing cervical neoplasia; previous cryosurgery, laser surgery, cone biopsy, or hysterectomy; or history of genital condylomas. tReiative risk of recurrent neoplasia for women with HPV detected at enrollment compared with women with no HPV detected, taking follow-up time into account, is 3.7 (95% confidence interval, 1.0 to 14.6).

Infection of the same cervical specimen with multiple HPV types was infrequent in our study population, consistent with earlier results from our laboratory. 1, 2 In contrast, other investigations have found multiple infections to be more common." 4 In addition, HPV type 18 was not detected in this study, unlike other reports that are based on pathologic findings of similar severity.' Therefore the true frequencies of type 18 infection and of multiple infections are unclear and should be the subject of future interlaboratory comparisons, similar to those that have already demonstrated interlaboratory variability in the detection and typing of HPV DNA in clinical specimens. ' • The hormonal factors assessed in our investigation appeared to influence HPV test results. Most strikingly, current oral contraceptive use was associated both with HPV detection and with cervical neoplasia, but the association of oral contraceptive use with neoplasia could be explained by the HPV detection data. This finding merits further attention because oral contraceptive use has been postulated to increase the risk of invasive cervical cancer. 15 As an alternative interpretation, oral contraceptive use might represent a proxy for a risk factor that we could not assess in our investigation, such as recent sexual activity. Current pregnancy also was associated with increased HPV detection in our study, whereas HPV detection was decreased in women with a history of diethylstilbestrol exposure in utero. These findings could have resulted from the inclusion in the study population of a large number of women with conditions, such as abnormal Papanicolaou smears or condylomas, that are related to HPV detection. The possibility of bias in our cross-sectional results emphasizes the need to find less selected groups of women for future studies of HPV infection, especially if viral prevalence or the determinants of infection are to be determined. The effect of age on HPV prevalence has been studied elsewhere with conflicting results. 'S ,17 Consistent with another recent study,18 there was no apparent

trend in HPV detection with age in our data when the correlates of age, especially oral contraceptive use, were considered. It seems that age cannot be viewed as a simple variable but rather must be studied in combination with age-related covariates, including contraceptive method, current pregnancy, and possibly even current sexual habits, a topic that was not assessed in this investigation. The strong association of cervical neoplasia and HPV detection persisted after other risk factors for which we had data were considered. We believe that the association of cervical neoplasia with HPV detection is real, and that future cross-sectional studies should focus on the relationship of HPV infection to other established risk factors for cervical neoplasia, particularly aspects of sexual behavior. With regard to temporal associations, women in this investigation with a medical history of genital neoplasia were more likely to have HPV detected than those without such a history. This observation suggests that investigators who attempt to estimate the true prevalence of HPV infection among normal women cannot rely only on current cytologic or histologic diagnosis but rather must consider past medical history. The prevalence of HPV is likely to be lower in the general population than the estimates reported from clinic-based series. It cannot be determined from our data whether the finding of HPV in cytologically normal women with a history of neoplasia represents latent, persistent infection in normal epithelium or the failure of the enrollment Papanicolaou smear to detect subtle, ongoing foci of disease. The finding of high prevalence of HPV infection in vaginal cuff scrapes from women with hysterectomies may suggest that asymptomatic viral carriage in the vagina is relatively common, even in the absence of past cervical disease. A major purpose of this investigation was to determine whether HPV results could predict subsequent clinical course in women with normal cytologic findings at enrollment. Despite over 2 years of average follow-

Volume 162 Number 3

up, only two women without a history of genital neoplasia exhibited new disease, and thus we were not able to study the etiologic relationship of HPV infection to the subsequent development of neoplasia. To address this key question will require very large studies with prolonged follow-up; at least two such investigations in the United States and Great Britain are currently underway. Investigators must carefully distinguish between new and recurrent disease because recurrent disease is likely to be much more common in most clinicbased prospective series. In our data, HPV infection appeared to predict short-term recurrence, although this was based on a small number of patients. If our results with regard to recurrence are confirmed by larger series, HPV testing may have a role in the clinical surveillance of women already diagnosed and treated for cervical neoplasia. REFERENCES 1. Lorincz AT, Temple GF, Kurman RJ, Jenson AB, Lancaster WD. Oncogenic association of specific human papillomavirus types with cervical neoplasia. JNCI 1987;79: 671-7. 2. Reid R, Greenberg M, Jenson AB, et al. Sexual~y t~an~­ mitted papillomaviral infections. I. The .anat?mlc dlstr~­ bution and pathologic grade of neoplastIc leSIons assocIated with different viral types. AM J OBSTET GYNECOL 1987; 156:212-22. 3. Fuchs PG, Girardi F, Pfister H. Human papillomavirus DNA in normal, metaplastic, preneoplastic and neoplastic epithelia of the cervix uteri. Int J Cancer 1988;41 :41-5. 4. Bergeron C, Ferenczy A, Shah KV, Naghashfar Z. Multicentric human papillomavirus infections of the female genital tract: correlatio? of viral types with abno~mal mitotic figures, colposcoplc presentatIon, and locatIon. Obstet Gynecol 1987;69:736-42. 5. Ferenczy A, Mitao M, Nagai N, Silverstein SJ, Crum CPo Latent papillomavirus and recurring genital warts. N Engl J Med 1985;313:784-8.

Human papillomavirus and cervical neoplasia 651

6. Lorincz AT, Temple GF, Patterson JA, Jenson AB, Kurman RJ, Lancaster WD. Correlation of cellular atypia and human papilloma virus deoxyribonucleic acid sequences in exfoliated cells of the uterine cervix. Obstet Gynecol 1986;68:508-12. 7. Toon PG, ArrandJR, Wilson LP, Sharp DS. Human papillomavirus infection of the uterine cervix of women without cytological signs of neoplasia. Br Med J 1986; 293: 1261-4. 8. Lorincz AT, Lancaster WD, Temple GF. Cloning and characterization of the DNA of a new human papillomavirus from a woman with dysplasia of the uterine cervix. J Virol 1986;58:255-29. 9. Lorincz AT, Quinn AP, Goldsborough MD, Sc?midt BJ, Temple GF. Cloning and partial DNA sequencmg of two new human papillomavirus types associated with condylomas and low-grade cervical neoplasia. J Virol 1989;63: 2829-34. 10. National Cancer Institute Workshop. The 1988 Bethesda system for reporting cervical/vaginal cytologic diagnoses. JAMA 1989;262:931-4. . 11. Zar JH. Biostatistical analysis. Englewood Chffs, New Jersey: Prentice-Hall, 1974. 12. Mantel N. Chi-square tests with one degree of freedom; extensions of the Mantel-Haenszel procedure. J Am Stat Assoc 1963;58:690-70. 13. Rothman K. Modern epidemiology. Boston: Little Brown, 1986. 14. Brandsma J, Burk R, Lancaster W, Pfister H, Schiffman MH. Interlaboratory variation as an explanation for varying prevalence estimates of human papillomavirus infection. Int J Cancer 1989;43:260-2. 15. Brinton LA, Higgins GR, Lehma? HF, ~t al. .Long te.rm use of oral contraceptives and rIsk of mvaslve cervIcal cancer. Int J Cancer 1986;38:339-44. 16. de Villiers E-M, Schneider A, Miklaw H, et al. Human papilloma virus infections in women with and without abnormal cervical cytology. Lancet 1987; 1: 703-6. 17. Meanwell CA, Blackledge G, Cox MF, Maitland NJ. HPV 16 DNA in normal and malignant cervical epithelium: implications for the aetiology and behaviour of cervical neoplasia. Lancet 1987;1:703-7. . 18. Reeves WC, Brinton LA, Garcia M, et al. Human papt!lomavirus infection and cervical cancer in Latin America. N EnglJ Med 1989;320:1437-41.