Human Papillomavirus Infection and Time to Progression and Regression of Cervical Intraepithelial Neoplasia
Journal of the National Cancer Institute
Human Papillomavirus Infection and Time to Progression and Regression of Cervical Intraepithelial Neoplasia
Nicolas F. Schlecht 0 1
Robert W. Platt 0 1
Eliane Duarte-Franco 0 1
Maria C. Costa 0 1
Joa˜ o P. Sobrinho 0 1
Jose´ C. M. Prado 0 1
Alex Ferenczy 0 1
Thomas E. Rohan 0 1
Luisa L. Villa 0 1
Eduardo L. Franco 0 1
0 DOI: 10.1093/jnci/djg037 Journal of the National Cancer Institute , Vol. 95, No. 17, © Oxford University Press 2003, all rights reserved
1 Affiliations of authors: N. F. Schlecht, E. L. Franco (Departments of Oncology and Epidemiology and Biostatistics), R. W. Platt (Departments of Epidemiology and Biostatistics and Pediatrics), E. Duarte-Franco (Department of Oncology), McGill University , Montreal, Quebec , Canada; M. C. Costa, J. P. Sobrinho, J. C. M. Prado, L. L. Villa , Ludwig Institute for Cancer Research , Sa ̃o Paulo , Brazil; A. Ferenczy, Department of Pathology, McGill University; T. E. Rohan, Department of Epidemiology and Population Health, Albert Einstein College of Medicine , New York, NY. McGill University , 546 Pine Ave. West, Montreal, Quebec , Canada H2W 1S6
Background: Little is known about the duration of precancerous cervical lesions in relation to human papillomavirus (HPV) infection. We estimated rates of progression and regression and sojourn times of cervical squamous intraepithelial lesions (SILs) according to HPV status. Methods: We used data from a longitudinal study of HPV infection and cervical neoplasia in Sa˜o Paulo, Brazil. Cervical specimens were taken from 2404 women for Pap cytology and polymerase chain reaction-based HPV testing every 4-6 months over a period of 8 years. We used actuarial and non-actuarial analyses to measure time to and rates of lesion progression and regression according to status and type of HPV infection. Results: During follow-up, 118 low-grade SIL (LSIL), 24 high-grade SIL (HSIL), and 173 atypical squamous cells of undetermined significance (ASCUS) events were detected. Mean time to progression from ASCUS to LSIL or worse and from LSIL to HSIL or worse was shorter in women with oncogenic HPV types than in women with no HPV infection (mean times for ASCUS progression were 67.0 and 88.0 months, respectively, in women with oncogenic HPV and no HPV, difference = 21.0 months, 95% confidence interval [CI] = 11.3 to 30.7 months; mean times for LSIL progression were 73.3 and 83.5 months, respectively, difference = 10.2 months, 95% CI = -0.15 to 20.6 months). Half of the LSILs regressed to normal or ASCUS within 6 months. Mean times for regression from ASCUS to normal, from LSIL to ASCUS or normal, and from HSIL/cervical intraepithelial neoplasia 2 to ASCUS or normal were longer for women with oncogenic HPV types (16.8 months, 95% CI = 7.5 to 26.2 months; 13.8 months, 95% CI = 8.8 to 18.7 months; and 17.1 months, 95% CI = 4.1 to 30.1 months, respectively) than for women with non-oncogenic HPV types (7.7 months, 95% CI = 5.2 to 10.2 months; 7.8 months, 95% CI = 5.3 to 10.2 months; 8.9 months, 95% CI = 3.3 to 14.6 months) or for women with no HPV infection (7.6 months, 95% CI = 6.9 to 8.4 months; 7.6 months, 95% CI = 6.4 to 8.7 months; and 7.0 months, 95% CI = 5.0 to 8.9 months, respectively). Conclusion: Precursor lesions of the cervix persist longer and progress more quickly in women with oncogenic HPV infections than in women with non-oncogenic infections or without HPV. Testing cervical lesions for oncogenic HPVs may help identify those that are likely to progress rapidly. [J Natl Cancer Inst 2003;95:1336-43]
cervical cancer (
). This sequence forms the premise on which
cytologic screening for cervical cancer is based and corresponds
to an underlying multistep carcinogenic process in the
development of cervical intraepithelial neoplasia (CIN) (
squamous intraepithelial lesions (LSILs) may progress to
highgrade SILs (HSILs) and invasive cervical cancer or may regress
to a normal state (
). However, few studies of cervical neoplasia
have evaluated lesion recurrence (
) or disease progression (3)
over time. Likewise, lesion progression or regression has not
been evaluated in relation to the presence of human
papillomaviruses (HPV), the main etiologic agents in the initiation of
cervical neoplasia (
). The use of a biomarker that can predict
the rate of progression or regression and the duration of the
preinvasive stages of cervical cancer could represent an
attractive means for targeting screening or chemoprevention.
Beginning in 1993, we initiated a cohort study involving
repeated measurements of HPV infection and cervical cytology
in women attending a comprehensive maternal and child health
program that serves low-income families living in
neighborhoods located in the northern sector of the city of Sa˜o Paulo,
). In this population, early precursor lesions are
generally not treated, which enabled us to evaluate prospectively the
occurrence of SIL events at regular intervals over time. In
particular, we sought to measure the frequency and rates of
progression and regression, as well as the durations (i.e., sojourn
time) of early cervical precursor lesions according to their HPV
SUBJECTS AND METHODS
Two study nurses approached 4990 women from daily lists of
outpatients in the family medicine, gynecology, and family
planning clinics at the Vila Nova Cachoeirinha municipal hospital in
Sa˜o Paulo, Brazil, for an interview. Women who were
tially eligible were presented with a detailed overview of the
study and invited to participate. Recruitment began in November
1993 and continued until March 1997.
Women were eligible to participate if they 1) were between
18 and 60 years of age; 2) were permanent residents of Sa˜o
Paulo (city), Brazil; 3) were not currently pregnant and had no
intention of becoming pregnant during the next 12 months; 4)
had an intact uterus and no current referral for hysterectomy; 5)
reported no use of vaginal medication in the previous 2 days; and
6) had not received treatment for cervical disease in the previous
6 months. In addition to these criteria, women were considered
eligible only if they expressed willingness to comply with all
scheduled return visits, at least for the initial 2 years.
Subjects entered the study only after giving signed informed
consent. The study protocol was approved by institutional
ethical and research review boards of the participating institutions in
Canada (McGill University) and Brazil (Ludwig Institute for
Cancer Research). A detailed description of the design and
methods of the study has been published previously (
participants were seen every 4 months in the first year (at 0, 4,
8, and 12 months) and every 6 months thereafter. Delays in
returning for a given follow-up appointment were allowed; the
visit numbering sequence was maintained, even when subjects
returned for their follow-up visit after the scheduled date, with
the information and specimens collected being assigned to the
originally scheduled follow-up visit. As a result, the same
number of scheduled visits was retained, precluding missing interval
visits. However, all time-to-event analyses were based on the
actual time of the visits. Cervical specimens were taken for
conventional Pap smear and HPV testing at every visit. The
study nurses also performed a detailed interview at enrollment to
collect information on sociodemographic factors, reproductive
health, sexual activity, and smoking status. Information on
sexual activity and reproductive health was also collected at each
return visit during the first 12 months and annually thereafter.
Cervical Cell Specimens
An Accelon biosampler (Medscand, Hollywood, FL) was
used to collect ecto- and endocervical samples. After cells were
smeared on a glass slide and fixed for Pap cytology, the sampler
containing the residual exfoliated cells was immersed in a tube
containing Tris–EDTA buffer (pH 7.4) (
). The exfoliated cell
samples were sent to the Ludwig Institute for Cancer Research
in Sa˜o Paulo for storage and HPV testing. The Pap smears were
shipped to McGill University for coding and classification by an
expert cytopathologist (A. Ferenczy) who was blinded to
previous cytology outcomes and to HPV results for the same and
previous samples. The cytopathology reports were based on the
1991 Bethesda system for cytologic diagnoses (
). Referral to
colposcopy was made by the Study Management Center at
McGill University on the basis of either local or review cytology
reports or the cervicography examination, which was performed
once every 2 years for all women.
The progression and regression states were classified by
lesion severity (LSIL or HSIL) and were subcategorized within
these lesional grades. LSILs were separated into those with
koilocytosis (LSIL/HPV) and those without koilocytosis but with
squamous abnormalities (LSIL/SQ). Similarly, HSILs were
separated into those indicative of CIN2 (HSIL/CIN2) and those
indicative of CIN3 (HSIL/CIN3).
Cervical specimens were tested for the presence of HPV
DNA by using the MY09/11 polymerase chain reaction (PCR)
). The amplified products were typed by
hybridization with individual oligonucleotide probes specific for 27
genital HPV types (10). Amplified products that hybridized with
the generic probe but with none of the type-specific probes were
further tested by restriction fragment length polymorphism
) to extend the range of identifiable HPV types. To
verify the specificity of the hybridizations, we included more
than 30 type-specific positive controls in each HPV test run. To
check the integrity of the host DNA extracted from the
specimens, assays also included an additional set of primers to
amplify the -globin gene (
). HPV types were separated into two
groups by level of oncogenicity. Oncogenic types included
HPV16, -18, -31, -33, -35, -39, -45, -51, -52, -56, -58, -59, and
-68; non-oncogenic types included 6/11, -26, -32, -34, -40, -42,
-44, -53, -54, -55, -57, -62, -64, -66, -67, -69, -70, -71, -72, -73,
-81, -82, -83, -84, CP6108, and other unknown types. All HPV
assays were done on coded specimens, with no identification
that could link specimens from the same woman. Appropriate
precautions were taken to reduce the possibility of specimen
For the analyses reported here, follow-up began in November
1993 and continued through mid-April 2002. Subjects with
atypical squamous cells of undetermined significance (ASCUS),
LSIL, or HSIL on cytology were included in each risk set at the
time of their first detected result. Depending on the index lesion
of interest, prevalent cases of equivalent grade of abnormality at
enrollment were excluded from the risk set at baseline. Subjects
who were biopsied were censored at the time of their biopsy if
no transition event had occurred before the biopsy date to reduce
the potential for interference by the biopsy procedure on
estimates of time to regression. Women who dropped out of the
study were censored at their last visit date. In time-to-event
analyses, time to event was measured from the date of the index
visit (i.e., the first instance of an abnormal cytologic result) to
the date of the visit at which transition to a more or less severe
cytologic category (for estimates of progression or regression,
respectively) was first detected or, for censored subjects, to the
last recorded return visit date. Such data represent
intervalcensored data because the exact dates of HPV infection and SIL
incidence are not known. The time to regression from HSIL or
LSIL was defined as the time from the index visit until the first
follow-up visit at which a subject presented with LSIL, ASCUS,
or a normal Pap smear (depending on the regression endpoint of
interest), whether or not a worse cytologic event had been
detected during intervening visits. Times to progression or
regression were evaluated for the overall group at risk by age and
ethnic group and for groups stratified by HPV status at the visit
when the index lesion was first detected or at the visit closest to
the index event at which a valid HPV result was obtained.
Oncogenic status of HPV in the index specimen was coded
according to the following hierarchic categories: 1) no HPV detected,
2) only non-oncogenic types detected in the index specimen, and
3) any oncogenic HPV type detected.
The cumulative probability of a lesion’s remaining in the
same stage or progressing to the next stage was estimated by
actuarial analysis using Kaplan–Meier curves (
) as a function
of the length of follow-up, stratified by HPV status at the index
visit. The life table method was also used to estimate the
proportion of women who remained positive for a precursor lesion
during follow-up without progressing to a higher stage,
according to HPV status at the index visit (
). Ninety-five percent
confidence intervals (CIs) for the actuarial estimates were
calculated using the standard error of the cumulative probability at
the end of a particular interval in which an event occurred (
or, for binomial proportions, via exact estimation. Statistical
comparisons (two-sided tests) of lesion sojourn times according
to HPV status were performed by using the log-rank test.
Progression and regression density estimates were also calculated
by dividing the number of incident events (of progression or
regression, respectively) by the length of follow-up in
personmonths of subjects at risk (i.e., those with the index lesion
event). Density estimates were likewise stratified by HPV status
of the index lesion. Density rates and mean durations were
compared by pooled variance z tests and t tests, respectively.
Statistical analyses were performed by using SPSS version 11.0
(SPSS, Chicago, IL) and PEPI version 4.0× (Sagebrush Press,
Salt Lake City, UT).
For non-actuarial estimates of mean lesion duration, we used
a standard formula based on the epidemiologic tenet that, within
a stationary population and in the absence of migration, the
prevalence proportion (P) is a function of the incidence rate (I)
and of the mean duration (D) of the condition. Therefore, the
average duration can be estimated with the general formula D
P/[I × (1 – P)], where P is calculated as a weighted average of
the point prevalence over time for each lesion grade. This
formula holds provided that the point prevalence within each
stratum is less than 10% (
), a condition that was met for all of the
individual lesion grades analyzed in this study. The
nonactuarial estimates were stratified by the cumulative HPV status
over the first year of follow-up, taking into account test results
from the first four visits.
Of the 4990 women initially identified, 3589 women initially
met the eligibility criteria and were invited to participate in the
study. Between November 1993 and March 1997, 2528 women
were enrolled in the study, a response rate of 70.4%. A further
66 women who were found not to fit the eligibility criteria
were excluded after enrollment. Fifty-one women (2.1%)
presented with prevalent lesions (LSIL or HSIL) at enrollment,
and seven women had unsatisfactory baseline cytology results.
Thus, 2404 women remained for analysis. Among these
women, 131 (5.4%) developed incident SIL events: 118
developed LSIL and 24 developed HSIL over the period of
followup, although these were not mutually exclusive events. Among
women without any abnormal smears (i.e., excluding those
with ASCUS or worse) at enrollment (N 2361), there
were 173 incident reports of ASCUS within the 8-year
followup period. The prevalence of oncogenic and non-oncogenic
HPV types was similar across visits and ranged from 7.8% to
9.2% for oncogenic types and from 6.2% to 7.6% for
nononcogenic types. With respect to HPV prevalence by level of
incident abnormality detected by cytology, 32% of ASCUS
smears were HPV-positive (31 oncogenic types and 24
nononcogenic types), whereas 61% of LSIL smears were
HPVpositive (45 oncogenic types and 28 non-oncogenic types), as
were 71% of HSIL smears (15 oncogenic types and two
The total follow-up time for subjects with no lesions at
enrollment was 128 129 person-months, for a mean follow-up
duration of 53.3 months per subject. The average interval between
visits was 4.8 months for the first year and 6.8 months for
subsequent years. Although visits were scheduled according to
the study design to occur every 6 months after the first year, the
actual intervals between visits in the second and subsequent
years ranged from 2.9 months to 81.3 months.
The majority of ASCUS (147/173) and LSIL (104/118)
events detected by cytology regressed to a lower grade during
follow-up (Table 1). Actuarial analyses showed that half of these
abnormalities regressed within 6 months of first detection. The
overall mean time to regression of ASCUS abnormalities was
shorter than that of LSILs. The mean time to regression from
LSIL to ASCUS or normal was generally longer for lesions with
oncogenic HPV types (13.8 months) than for lesions with
nononcogenic HPV types (7.8 months) (difference 6.0 months,
95% CI –0.7 to 12.7 months) or for HPV-negative lesions
(7.6 months) (difference 6.2 months, 95% CI 1.0 to 11.4
months, respectively). Median durations, however (i.e., 6.1, 5.3,
and 6.0 months for LSILs with oncogenic HPV types, with
non-oncogenic HPV types, and without HPV, respectively),
differed less than the respective means. Similar differences in
regression times within levels of HPV status were observed for the
subcategories of LSIL.
To evaluate mean sojourn times with respect to cumulative
HPV status during the first year of follow-up, we also calculated
mean lesion duration based on a non-actuarial formula (see the
“Subjects and Methods” section) in which subjects were
stratified by cumulative HPV status during the first year of follow-up.
Given the particular relevance of HPV16 in cervical cancer
etiology, subjects who tested positive for that type in any of the
visits were evaluated separately from those with other oncogenic
HPV types. The mean ASCUS duration in women who were
HPV-negative at all four visits was 7.9 months, which was lower
than that in women with only non-oncogenic HPV types (10.5
months), in women with any oncogenic HPV type (excluding
HPV16 at any visit [15.4 months]), and in women positive for
HPV16 at any visit (13.4 months). The equivalent mean LSIL
durations were 8.9, 10.3, 12.2, and 13.4 months for HPV-negative
women and women with non-oncogenic HPV types, oncogenic
HPV types, and HPV16, respectively. The equivalent mean HSIL
durations were 7.6, 5.7, 15.6, and 57.0 months, respectively.
With respect to progression to a higher preinvasive lesion
grade, we observed the reverse of the pattern described above for
regression by HPV infection status (Table 2). That is, lesions in
subjects with no HPV detected in the index specimen took
longer to progress than those with oncogenic HPV. For example,
mean time to progression from ASCUS to LSIL or worse and
from LSIL to HSIL or worse was shorter for women with
oncogenic HPV types than for women with no HPV infection (for
ASCUS progression, means 67.0 and 88.0 months,
respectively; difference 21.0 months [95% CI 11.3 to 30.7
months]; for LSIL progression, means 73.3 and 83.5 months,
respectively; difference 10.2 months [95% CI –0.15 to
20.6 months]) or in women with non-oncogenic HPV types, with
respect to LSIL- to HSIL progression. Among women with
persistent LSIL (n 24, data not shown), 20% progressed to HSIL
or cervical cancer during follow-up.
*ASCUS atypical squamous cells of undetermined significance; LSIL/HPV low-grade squamous intra-epithelial lesion (LSIL) with koilocytotic atypia induced by a productive
HPV infection; LSIL/SQ LSIL showing squamous effects equivalent to cervical intra-epithelial neoplasia grade 1; HSIL/CIN2 high-grade squamous intra-epithelial lesion (HSIL)
with moderate dysplasia equivalent to cervical intra-epithelial neoplasia grade 2. ND not determined; — not estimable because of insufficient follow-up time for the group or because
all regression events had already occurred at a previous time point.
†Index lesion is defined as the first detected event of the stated cytologic abnormality.
‡Numbers of lesions regressed/total number of index lesions. Number of HPV stratum-specific samples may not add up to the overall number if valid HPV results were unavailable for
§Regression density rate is equal to the number of incident events per 100 person-months.
Estimates from actuarial analysis using the Kaplan–Meier method.
¶Proportion of subjects remaining with lesion at end of interval period (6, 12, 18, or 24 months) derived by life table analysis.
Any oncogenic HPV type in index sample.
**No HPV16-positive specimens were identified.
To evaluate whether lesions progressed more quickly with
age, we estimated rates of progression separately for younger
and older women. On average, women aged 31–65 years
progressed to HSIL from an incident LSIL more rapidly (mean time
to progression 77.9 months) than women aged 16–30 years
(mean time to progression 88.4 months, difference 10.5
months [95% CI 1.5 to 19.5 months]). By contrast, the mean
time to progression from ASCUS to HSIL was shorter in
younger women (81.8 months) than in older women (90.4
months, difference 8.6 months [95% CI 4.0 to 13.2
months]). After stratification by HPV status, older women with
oncogenic HPV types, including HPV16, had a higher
cumulative risk of progression to HSIL than younger women, regardless
of baseline abnormality (Fig. 1), although the differences were
not statistically significant. Mean times to progression from
LSIL to HSIL or cancer for women with oncogenic HPV
infections were 68.4 months in women aged 31–65 years and 75.6
months in women aged 16–30 years (difference 7.2 months,
95% CI –9.8 to 24.2 months). Mean times to progression
from ASCUS to HSIL or cancer for women with oncogenic HPV
infections were 73.4 months and 80.4 months in older and
younger women, respectively (difference 7.0 months, 95% CI
–10.2 to 24.2 months).
We also evaluated rates of progression and regression for
subjects with respect to ethnic origin (data not shown). The
regression density rates of incident ASCUS, LSIL, and HSIL to
normal for whites (n 1542) and nonwhites (n 856) were
not statistically significantly different (differences in lesion
regression rates 0.61 regressed abnormalities per 100
personmonths [95% CI –10.0 to 11.2]; 2.4 per 100 person-months
[95% CI –9.8 to 14.6]; and 4.7 per 100 person-months [95%
CI –66.7 to 76.1], respectively). Similarly, we saw little
difference in rates of progression to HSIL or worse between ethnic
Given the potential for biopsy interventions to interfere with
the natural history of HSIL, we estimated the time to regression
for HSILs for women based on whether their biopsies were
performed before or after a regression event to ASCUS or
Index lesion and
HPV status in
ASCUS to LSIL or worse
*ASCUS atypical squamous cells of undetermined significance; LSIL/SQ low-grade squamous intra-epithelial lesion (LSIL) showing squamous effects
equivalent to cervical intra-epithelial neoplasia grade 1; LSIL/HPV LSIL with koilocytotic atypia induced by a productive HPV infection; HSIL/CIN3
high-grade squamous intra-epithelial lesion (HSIL) with severe dysplasia equivalent to cervical intra-epithelial neoplasia grade 3.
†Index lesion is defined as the first detected event of the stated cytologic abnormality.
‡Number of lesions progressed/total number of index lesions. Number of HPV stratum-specific samples may not add up to the overall number if valid HPV results
were unavailable for some samples.
§Progression density rate is equal to the number of incident events per 100 person-months.
Estimates from actuarial analysis using the Kaplan–Meier method. Median time to progression was not estimated as less than 50% of the index lesions progressed.
Mean times were restricted to the longest follow-up time, regardless of event status.
¶Proportion of subjects progressing to worse abnormality at end of interval period (6, 12, or 18 months) derived by life table analysis.
Any oncogenic HPV types in baseline smear.
**All baseline (index) LSIL events were oncogenic HPV-positive.
mal was detected by cytology during study follow-up (data not
shown). As the study progressed, recommendations for biopsy
became more aggressive. Biopsies were done at the discretion of
the attending colposcopists if lesional tissue could be visualized
under magnification. Of the 24 women with incident HSIL
included in the analyses of regression, 23 had a biopsy, at an
average of 27.9 months after the first cytologic detection of
HSIL. Some lesions appeared to regress as determined by
cytology before the biopsy was performed. We therefore evaluated
rates of regression separately among women whose biopsy was
taken before regression was determined by cytology and among
women whose biopsy was taken after (including the subject who
did not have a biopsy). Irrespective of biopsy status, the mean
time to regression of all incident HSILs was 12.8 months (95%
CI 7.6 to 18.0 months). Although the number of HSILs with
(n 12) and without (n 12) possible biopsy interference was
small, it appeared that women who had a biopsy before a
regression event (to ASCUS or normal) was detected by cytology
retained their lesions longer (mean time to regression 16.9
months) than those whose biopsy was performed after regression
(mean 8.3 months; difference 8.6 months, 95% CI –1.0
to 18.1 months). The mean duration of HSIL persistence after
having the biopsy (calculated by subtracting time to biopsy from
time to regression) was 5.6 months.
To date, few studies have investigated rates of progression
and regression of preinvasive cervical lesions in relation to
cervical cancer risk factors (
). The available evidence on
the natural history of HPV and cervical neoplasia suggests that
a relationship between the likelihood of precursor lesions
persisting and progressing to cancer is dependent on the
characteristics of the HPV infection (
). In the current
longitudinal study, we found that precursor lesions of the cervix detected
by cytology persisted longer and were more likely to progress in
women with oncogenic HPV infections than in women with
non-oncogenic HPV infections or in uninfected women. Using
repeated screening by cytology, we were able to evaluate the
progression and regression of cervical lesions over time in a
more systematic fashion than that in previous studies, which
involved registry data and screening programs based on passive
Our study does have several limitations. First, given that our
outcome ascertainment was based on cytologic analysis, one
potential limitation is misclassification of lesion outcome
history, even though the cytologic assessments were carefully
conducted in a reference laboratory following a strict quality-control
protocol. We opted for an intensive, expert cytologic review of
all subjects in the study every 4–6 months and referred all
instances of HSILs for colposcopy. This approach reduced the
likelihood of unnecessary biopsies, which can interfere with the
natural history of early lesions (
). Nevertheless, the
occurrence of false-negative Pap tests could have resulted in
underestimates (or shorter estimates) of regression time and in either
overestimates or underestimates of progression time, depending
on whether these test results occurred at lesion outset or during
the sojourn period. In a prospective study of histologically
diagnosed carcinoma in situ (CIS), McIndoe et al. (
) found that
54% of the cases had normal cytology during screening after a
punch biopsy, with only one case of six developing invasive
carcinoma within 4 years. We therefore censored subjects at the
time of their biopsy, anticipating that the procedure could
influence the rates of disease determined by cytology.
Studies using both histology and cytology to follow the
natural history of cervical neoplasia have shown no effect of limited
sampling by punch biopsy on the short-term course of dysplasia
). In a review of 27 studies of CIN, Mitchell et al. (
observed similar probabilities of regression, persistence, and
progression based on biopsy evaluation and cytology. We found
that HSIL persisted for an average of 5.6 months following a
biopsy. Although persistent lesions are more likely to be
biopsied before regression than lesions of short duration, we found
no evidence that biopsy affected the persistence of HSILs, at
least in the short term. Although more aggressive standards for
biopsy were adopted after a number of years into the study by
the local colposcopists (all women referred for colposcopy are
currently obtaining biopsies following our recommendations),
we cannot exclude the possibility that the biopsy procedures
performed earlier in the study were preferentially done for
lesions that appeared more severe on colposcopy. Such lesions
may have taken longer to regress, regardless of whether a biopsy
had been done.
Another possible limitation was misclassification of HPV
status. Misclassification was unlikely because we used a highly
sensitive PCR-based testing method. The estimates of HPV
positivity by cytologic category in the present study are not
substantially different from other published estimates for ASCUS and
LSIL patients (
), which reflects positively on the quality of the
cervical specimens taken, as well as on the sensitivity of the
PCR-based method we used for HPV DNA detection. In this
analysis, we chose to use the HPV DNA test results obtained
from the same specimen that was used for Pap cytology. This
decision was based on the assumption that finding HPV in
incident lesions is a proxy for prior HPV infection states that led
to the lesion. Cross-sectional assessment of HPV and lesion
status precludes the determination of directionality in the
association between infection and lesion development. However, this
approach was chosen to generate data similar to those obtained
in screening and on which triage decisions are usually based.
We used two methods to calculate the mean duration or
sojourn time of incident cytologic lesions: an actuarial method and
a non-actuarial formula—the prevalence–incidence relation.
Estimates of mean duration based on actuarial probability
estimates indicated that lesions with oncogenic HPV infections
persisted longer than lesions with non-oncogenic HPV infections or
than lesions without HPV infection. In general, the average
duration estimates based on the non-actuarial formula were similar
to those obtained with the actuarial method. However, the mean
duration of HSIL estimated by the non-actuarial formula was
longer than the equivalent estimate by actuarial analysis for
subjects with HPV16. The non-actuarial prevalence–incidence
relation method is appropriate for estimating average duration of
incident conditions, such as cytologic abnormalities consistent
with ASCUS, LSIL, or HSIL that rarely exceed a prevalence of
10% in most clinical settings (
). However, the non-actuarial
formula does not account for censored data (i.e., incomplete
observations due to lesions that have not cleared at the study
closing date or losses to follow-up). By contrast, actuarial
estimates of time to progression were restricted to the longest
follow-up time available. As a result, mean times were
underestimated when the largest observed analysis time occurred for a
censored subject. This effect was greater for strata with smaller
numbers of observations. Therefore, we did not separate out the
subjects with HPV16 in the actuarial analyses. This difference in
estimates from the actuarial and non-actuarial analyses
underscores the fact that, due to the heterogeneity of follow-up times
and censoring, no single statistical summary measure is
appropriate to capture the average progression or regression times of
lesions in a repeated-measurements cohort study. Whenever
appropriate, therefore, we reported all three summary statistics
(non-actuarial means, actuarial means, and actuarial medians) to
provide a more complete picture of transit times.
It is conceivable that using shorter cytologic screening
intervals could have resulted in more precise estimates of the
duration of lesion sojourn times. Indeed, the reduction in
followup frequency to every 6 months after the first year (when return
visits were scheduled at 4 months) may have increased the
observed sojourn time. However, this effect would have been
equal across HPV groups because all cytology and HPV
evaluations were carried out blindly with respect to previous results
for the same subject, making comparisons valid on a relative
Several reviews and meta-analyses have attempted to
summarize rates of progression and regression along the continuum
of cervical neoplastic changes. Östor (
) reported decreasing
probabilities of regression for different CIN grades of increasing
severity detected by histology. Mitchell et al. (
probabilities of regression, persistence, and progression of all grades
of CIN combined to any higher grade lesion of 34%, 41%, and
25%, respectively. Melnikow et al. (
) calculated the following
weighted average rates of progression to HSIL at 24 months
according to baseline cytologic abnormality: 7.1%, 20.8%, and
23.4% for ASCUS, LSIL, and HSIL persistence, respectively.
Conversely, average rates of regression to a normal Pap smear
were 68.2% for ASCUS, 47.4% for LSIL, and 35.0% for HSIL.
Using mild dysplasia as the referent category, Holowaty et al. (
found relative risks of CIS of 8.1 (95% CI 6.2 to 10.7) for
women with moderate dysplasia and 22.7 (95% CI 16.0 to
32.1) for those with severe dysplasia within a 2-year period.
They also observed that rates of progression were higher during
the first 2 years following a positive Pap smear than in
subsequent years. This increase was interpreted to be the result of
underrating the original smear. Their recommendation,
therefore, was to repeat smears within 6 months of the first positive
smear rather than 1 year later. We propose a less aggressive
recommendation because one-half to one-third of all LSIL and
HSIL/CIN2 lesions in our study regressed to ASCUS or normal
within 6 months. Repeat screening with a shorter delay would
therefore detect a substantial proportion of lesions that would
regress spontaneously. Furthermore, given a mean time to
progression for LSILs of 85.7 months, most repeat cytology
screenings before 1 year would not indicate whether a lesion is likely
Of the reviews indicated above, few evaluated progression or
regression by risk stratifiers. Holowaty et al. (
) examined the
influence of parity, age, oral contraceptive use, and number of
positive smears on the relative risk of progression of LSILs.
They found no relationship between these factors and the
relative risk of progression. Similarly, in our study, the mean time to
progression was not statistically significantly different between
age groups, although we found that, among women with
highrisk oncogenic HPV types, progression rates were higher in
women 31 years and older than in younger women. These
findings support the World Health Organization’s recommendation
for focusing screening on older women at risk for cervical cancer
). We also observed shorter progression transit times from
ASCUS to LSIL and worse in women with non-oncogenic
(lowrisk) and oncogenic (high-risk) HPV types compared with
women who had HPV-negative abnormalities. This observation
provides further evidence of the importance of colposcopic
evaluation of women with HPV-positive ASCUS smears,
particularly those with high-risk oncogenic HPV types (23).
In conclusion, using screening tests for oncogenic HPVs may
help identify those lesions that are likely to progress quickly to
more advanced stages. HPV testing of women with abnormal
Pap smears may therefore help identify women who might
benefit from colposcopic evaluation and, if appropriate,
chemopreventive treatment. The ability to identify subjects whose lesions
are likely to take longer to progress could also be cost-saving by
increasing the follow-up intervals and reducing the morbidity
which may result from potentially unnecessary invasive
diagnostic and therapeutic procedures.
(1) Meijer CJ , van den Brule AJ , Snijders PJ , Helmerhorst T , Kenemans P , Walboomers JM . Detection of human papillomavirus in cervical scrapes by the polymerase chain reaction in relation to cytology: possible implications for cervical cancer screening . IARC Sci Publ 1992 ; 119 : 271 - 81 .
(2) Mitchell MF , Hittelman WN , Hong WK , Lotan R , Schottenfeld D. The natural history of cervical intraepithelial neoplasia: an argument for intermediate endpoint biomarkers . Cancer Epidemiol Biomarkers Prev 1994 ; 3 : 619 - 26 .
(3) Holowaty P , Miller AB , Rohan T , To T. Natural history of dysplasia of the uterine cervix . J Natl Cancer Inst 1999 ; 91 : 252 - 8 .
(4) Ho GY , Burk RD , Klein S , Kadish AS , Chang CJ , Palan P , et al. Persistent genital human papillomavirus infection as a risk factor for persistent cervical dysplasia . J Natl Cancer Inst 1995 ; 87 : 1365 - 71 .
(5) Woodman CB , Collins S , Winter H , Bailey A , Ellis J , Prior P , et al. Natural history of cervical human papillomavirus infection in young women: a longitudinal cohort study . Lancet 2001 ; 357 : 1831 - 6 .
(6) Bosch FX , Lorincz A , Munoz N , Meijer CJ , Shah KV . The causal relation between human papillomavirus and cervical cancer . J Clin Pathol 2002 ; 55 : 244 - 65 .
(7) Franco E , Villa L , Rohan T , Ferenczy A , Petzl-Erler M , Matlashewski G . Design and methods of the Ludwig-McGill longitudinal study of the natural history of human papillomavirus infection and cervical neoplasia in Brazil. Ludwig-McGill Study Group . Rev Panam Salud Publica 1999 ; 6 : 223 - 33 .
(8) Herbst AL . The Bethesda system for cervical/vaginal cytologic diagnoses . Clin Obstet Gynecol 1992 ; 35 : 22 - 7 .
(9) Bauer HM , Ting Y , Greer CE , Chambers JC , Tashiro CJ , Chimera J , et al. Genital human papillomavirus infection in female university students as determined by a PCR-based method . JAMA 1991 ; 265 : 472 - 7 .
(10) Hildesheim A , Schiffman MH , Gravitt PE , Glass AG , Greer CE , Zhang T , et al. Persistence of type-specific human papillomavirus infection among cytologically normal women . J Infect Dis 1994 ; 169 : 235 - 40 .
(11) Bernard HU , Chan SY , Manos MM , Ong CK , Villa LL , Delius H , et al. Identification and assessment of known and novel human papillomaviruses by polymerase chain reaction amplification, restriction fragment length polymorphisms, nucleotide sequence, and phylogenetic algorithms . J Infect Dis 1994 ; 170 : 1077 - 85 .
(12) Kaplan EL , Meier P . Nonparametric estimation for incomplete observations . J Amer Stat Assoc 1958 ; 53 : 457 - 81 .
(13) Cox DR . Regression models and life tables . J Royal Stat Soc 1972 ; 34 : 187 - 220 .
(14) Peto R , Pike MC , Armitage P , Breslow NE , Cox DR , Howard SV , et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. I. Introduction and design . Br J Cancer 1976 ; 34 : 585 - 612 .
(15) Freeman J , Hutchison GB . Prevalence, incidence and duration . Am J Epidemiol 1980 ; 112 : 707 - 23 .
(16) McIndoe WA , McLean MR , Jones RW , Mullins PR . The invasive potential of carcinoma in situ of the cervix . Obstet Gynecol 1984 ; 64 : 451 - 8 .
(17) Ostor AG . Natural history of cervical intraepithelial neoplasia: a critical review . Int J Gynecol Pathol 1993 ; 12 : 186 - 92 .
(18) Melnikow J , Nuovo J , Willan AR , Chan BK , Howell LP . Natural history of cervical squamous intraepithelial lesions: a meta-analysis . Obstet Gynecol 1998 ; 92 : 727 - 35 .
(19) Londesborough P , Ho L , Terry G , Cuzick J , Wheeler C , Singer A. Human papillomavirus genotype as a predictor of persistence and development of high-grade lesions in women with minor cervical abnormalities . Int J Cancer 1996 ; 69 : 364 - 8 .
(20) Schlecht NF , Kulaga S , Robitaille J , Ferreira S , Santos M , Miyamura RA , et al. Persistent human papillomavirus infection as a predictor of cervical intraepithelial neoplasia . JAMA 2001 ; 286 : 3106 - 14 .
(21) Richart RM . Influence of diagnostic and therapeutic procedures on the distribution of cervical intraepithelial neoplasia . Cancer 1966 ; 19 : 1635 - 8 .
(22) Youkeles L , Forsythe AB , Stern E . Evaluation of Papanicolaou smear and effect of sample biopsy in follow-up of cervical dysplasia . Cancer Res 1976 ; 36 : 2080 - 4 .
(23) Solomon D , Schiffman M , Tarone R . Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial . J Natl Cancer Inst 2001 ; 93 : 293 - 9 .
(24) Miller AB , Nazeer S , Fonn S , Brandup-Lukanow A , Rehman R , Cronje H , et al. Report on consensus conference on cervical cancer screening and management. Int J Cancer 2000 ; 86 : 440 - 7 .
(25) Lawson HW , Lee NC , Thames SF , Henson R , Miller DS . Cervical cancer screening among low-income women: results of a national screening program, 1991 - 1995 . Obstet Gynecol 1998 ; 92 : 745 - 52 .
Supported by an intramural grant from the Ludwig Institute for Cancer Research, by Public Health Service grant CA70269 (to E. L. Franco , L. L. Villa , A.
Ferenczy , and T. E. Rohan ) from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services, and by grant MA13647 (to E. L. Franco and A. Ferenczy ) from the Canadian Institutes of Health Research . N. F. Schlecht is a recipient of a pre-doctoral scholarship , R.
We thank Maria L. Baggio and Lenice Galan for management of the patients and specimen collection, Silvaneide Ferreira and Romulo Mayamura for HPV testing, and Juliette Robitaille for cytology reviews . Manuscript received January 21 , 2003 ; revised June 18, 2003 ; accepted July 3 , 2003 .