Human Papillomavirus (HPV) Type 16 and Type 18 DNA Loads at Baseline and Persistence of Type-Speciﬁc Infection during a 2-Year Follow-Up

Background. Studies of viral load–associated persistence of human papillomavirus (HPV) infection are rare, with inconsistent results reported. Methods. The study were for HPV 16 and HPV type 18 (HPV-18), respectively, the time of into in the ASCUS-LSIL (Atypical Squamous Cells of Undetermined Signiﬁcance–Low-Grade Squamous Intraepithelial Lesion) Triage Study returned 1 or more times for HPV testing during a biannual 2-year follow-up. The numbers of HPV-16 and HPV-18 copies per nanogram of cellular DNA at baseline were measured by use of real-time polymerase chain reaction. Results. Women with, compared with women without, persistent infection at month 6 of follow-up had a higher viral load at enrollment ( , for HPV-16; for HPV-18). The association of each 1-log 10 increase P ! .001 P p in viral load with persistence of HPV-16 or HPV-18 during the ﬁrst 6 months of the study was statistically signiﬁcant among women with multiple HPV types at enrollment (for HPV-16: odds ratio [OR], 1.53 [95% conﬁdence interval {CI}, 1.29–1.82];

performed in Costa Rica [5], ∼25% of women with single prevalent HPV infections had concurrent cervical cytologic abnormalities. Given that persistent infection with oncogenic HPV types is a prerequisite for the development of cervical cancer [6][7][8][9][10][11], knowing the determinants that are associated with viral persistence may help to improve programs for cervical cancer control. Because the HPV DNA load reflects the productivity of viral replication, whether the level of viral DNA is able to predict the likelihood of persistence deserves consideration.
The association of a higher HPV DNA load with persistence of the infection has been reported in some studies [12][13][14][15][16][17] but not in others [18,19]. These studies are, however, often limited by a small sample size and by the use of measurements obtained at only 2 points in time to define viral persistence. A few studies were limited by the use of a semiquantitative approach to measure the DNA of a group of HPV types. It is unclear whether association of the HPV DNA load with persistence of infection differs with increases in the length of time from when the initial positive test result was obtained (ie, it is unclear how persistence is defined).
In the present study, we evaluated the association between the HPV-16 and HPV-18 DNA loads at baseline and the typespecific persistence of infection during a 2-year follow-up of women who participated in the Atypical Squamous Cells of Undetermined Significance (ASCUS)-Low-Grade Squamous Intraepithelial Lesion (LSIL) Triage Study (ALTS).

Study subjects.
Study subjects were women enrolled in ALTS, a large-scale randomized clinical trial designed to compare strategies for the management of women with a referral Papanicolaou test result indicating the presence of ASCUS or LSIL. A detailed description of the design and study population of the ALTS has been reported elsewhere [20,21]. In brief, between January 1997 and December 1998, a total of 5060 women who received a Papanicolaou test result of ASCUS or LSIL in the previous 6 months were enrolled in the study and randomly assigned to 1 of 3 trial arms. All participants underwent an entry procedure at enrollment that included completing an interview, providing a Papanicolaou smear sample for cytologic evaluation, and undergoing testing for the presence of HPV DNA. Three trial arms differed in their criteria for referral for colposcopy and biopsy at enrollment. Regardless of the study arms to which they were assigned, these women returned at 6-month intervals over 2 years for cervical cytologic evaluation and HPV testing. During follow-up, women were re-referred for colposcopy and biopsy if cytologic evidence of high-grade squamous intraepithelial lesion (HSIL) was detected. At the time of exit from the study, all women with persistent low-grade lesions or HPV-positive ASCUS were also referred for colposcopy. Women with biopsy-confirmed cervical intraepithelial neoplasia grade 2 or higher (уCIN2) received appropriate treatment immediately. The ALTS protocol was approved by the institutional review boards at the National Cancer Institute and at each of the 4 clinical centers involved in the trial.
ALTS participants were eligible to participate in the present study if they had HPV-16 and/or HPV-18 DNA detected in their enrollment cervical samples by polymerase chain reaction (PCR)-based reverse-line blot assay [22]. PCR amplicons were subjected to reverse-line blot hybridization for the detection of 27 types (HPV types 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51-59, 66, 68, 73, and 82-84) [23]. During the trial, the typing capacity of the reverse-line blot was expanded from detection of 27 types to 38 types, by inclusion of an additional 11 types (HPV types 61, 62, 64, 67, 69-72, 81, 85, and 91). Of 1071 eligible women (759 of whom were HPV-16 positive, 258 of whom were HPV-18 positive, and 54 of whom were positive for both types), 19 (11 of whom were HPV-16 positive and 8 of whom were HPV-18 positive) were excluded because of a lack of samples for viral quantification. We also excluded 76 women, including 1 woman whose enrollment sample was positive for HPV-18 but negative for cellular DNA and 75 women (61 of whom were HPV-16 positive and 14 of whom were HPV-18 positive) who did not have any follow-up visits. Therefore, a total of 741 women with HPV-16 infection at baseline and 289 with HPV-18 infection at baseline were included in the analyses. The HPV-16 or HPV-18 DNA load at baseline was similar in women with or without a follow-up visit (data not shown). Data on HPV typing, cervical cytologic and histologic findings, and characteristics of the study subjects were obtained from the ALTS database. The protocol for this study was approved by the institutional review board at the University of Washington.
Quantification of HPV-16 and HPV-18 DNA loads. The number of HPV-16 and HPV-18 E7 DNA copies and the amount of cellular DNA in cervical swab samples at enrollment were measured by multiplex real-time PCR, as described elsewhere [24,25]. In brief, the assay was performed in a reaction volume of 25 mL with the use of the TaqMan Universal PCR Master Mix Kit (Applied Biosystems). Amplification was done using the Applied Biosystems 7900 HT. Two logarithmic-phase 5-point standard curves were implemented in each set of realtime PCR assays, one for HPV-16 or HPV-18 DNA and the other for cellular DNA. Each sample was assayed in triplicate. The viral load was normalized to the input amount of cellular DNA and was expressed as the number of E7 copies per nanogram of cellular DNA.
E7 DNA was undetectable in 58 HPV-16-positive samples and 21 HPV-18-positive samples by means of real-time PCR. Considering that the negative result might be caused by a tiny amount of viral DNA, a value of 1 viral copy per nanogram of cellular DNA was assigned to each sample. Similar results were obtained when these samples were excluded from the analysis (data not shown).
Statistical analyses. The normalized HPV-16 and HPV-18 DNA loads were log 10 transformed. The mean value of the triplicate measurements was used for analyses. A linear regression model [26] was used to compare the log 10transformed HPV-16 or HPV-18 DNA loads at enrollment for women whose results remained positive and those whose results became negative at each follow-up visit while adjusting for cervical cytologic findings at enrollment, current smoking status, and coinfection with other HPV types. Coinfection was defined based on the testing results of 27 HPV types for 52% of the enrollment samples and of 38 HPV types for 48% of the enrollment samples. In ALTS, the events of becoming viral DNA negative were ascertained at 6-month intervals. A woman was eligible for analysis of HPV status at a given visit if she had type-specific HPV DNA detected at all scheduled previous visits. For example, a woman with HPV-16 infection at baseline was eligible for analysis of HPV-16 status at the month 6 visit; if she continued to be HPV-16 positive at month 6, she would be eligible for the analysis of HPV-16 status at month 12. Although this analysis is straightforward, it excluded all visits after an initial missing visit.
By use of a logistic regression model, a second analysis was performed to evaluate odds ratios (ORs) and 95% confidence intervals (CIs) for the association of viral load at baseline (to be fitted as a continuous covariate) with the risk of having a positive result at various follow-up visits. The ORs were adjusted for coinfection with other types, current smoking status, and cervical cytologic findings at enrollment. The significance of the interaction between viral load and time (ie, the followup visit) was assessed using a likelihood ratio test. For illustrative purposes, the probability of remaining type-specific positive at various follow-up visits was plotted for the 25th, 50th, or 75th percentile of the HPV-16 (or HPV-18) DNA load at baseline. We also examined coinfection with other types, current smoking status, or cervical cytologic findings at enrollment, to determine whether they modified the effect of viral load at baseline in this model.
To determine whether estimates of persistence by viral load would be distorted by the presence of уCIN2, we performed parallel analyses in which women with уCIN2 were censored at the time of initial diagnosis. The results remained similar; for simplicity, these results were not presented.
We used Student's t test to compare viral load by age at enrollment, race, current use of hormonal contraceptives, lifetime number of sex partners, current smoking status, coinfection with other types, and HPV variant. Differences in viral load according to the number of visits followed and the cervical cytologic findings at baseline were tested by 1-way analysis of variance. Of women infected with multiple HPV types at enrollment, the proportions of coinfection with the same species type between women with and without detectable HPV-16 or HPV-18 at month 6 were compared using the test. The 2 x coinfection types were phylogenetically classified as non-HPV-16 a-9 species (ie, HPV-16 related, including HPV31, 33, 35, 52, 58, and 67) for analysis of HPV-16 and as non-HPV-18 a-7 species (ie, HPV-18 related, including HPV39, 45, 59, 68, 70, and 85) for analysis of HPV-18. All statistical tests were at the 5% 2-sided significance level.

RESULTS
The present study included 741 women with HPV-16 and 289 women with HPV-18 infection at enrollment. Forty-one of the women were positive for both HPV-16 and HPV-18. The mean (‫ע‬SD) value of the log 10 -transformed numbers of HPV-16 E7 copies per nanogram of cellular DNA was 2.82 ‫ע‬ 1. 33 ). As shown in Table 1, within the study P p .39 population, the HPV-16 DNA load at baseline differed significantly according to age, race, current smoking status, coinfection with other HPV types, and cervical cytologic findings at enrollment. The HPV-18 DNA load at baseline differed significantly according to current smoking status, coinfection with other HPV types, and cervical cytologic findings at enrollment. Table 2 shows the differences in the viral load at enrollment between women with and without persistence of the infection at consecutive visits. Women who remained type-specific positive, compared with those who became type-specific negative at month 6, had a significantly higher baseline HPV-16 DNA load ( ) or baseline HPV-18 DNA load ( ). P ! .001 P p .01 Among those who remained HPV-16-or HPV-18-positive at month 6, there was no appreciable difference in viral load at baseline according to type-specific positivity at the subsequent follow-up visit at month 12. Similarly, subsequent persistence of infections that were still present at month 12 or month 18 was not predicted by viral load at enrollment. We next assessed the association between a 1-log 10 increase in viral load at baseline and type-specific positivity at various follow-up visits. The interaction between the order of followup visits and the viral load at baseline was statistically significant (by likelihood ratio test, for women with HPV-16 and P p .05 for women with HPV-18); it was included in the lo-P p .04 gistic model. With adjustment for current smoking status, coinfection with other types, and cervical cytologic findings at enrollment, the ORs for the association of each 1-log 10 increase in the viral load at baseline with HPV-16 positivity noted at months 6, 12, 18, and 24 were 1. 35   excluded (data not shown). We also tested the model by treating the intercurrent visit when findings were negative as a visit when findings were positive; the results remained similar (data not shown). Figure 1 illustrates the probabilities of remaining type-specific positive at various follow-up visits according to viral load at baseline. The 25th, 50th, and 75th percentiles of the HPV-16 or HPV-18 DNA load were chosen to denote low, medium, and high viral loads, respectively. Regardless of the viral load at enrollment, the probabilities of type-specific positivity decreased dramatically over time. Consistent with the estimates of risk association, probabilities of type-specific positivity at month 6 differed substantially according to the viral load at baseline, but the differences diminished quickly as the duration of follow-up increased, and they disappeared at the month 18 and 24 visits.
Given a significant association between the higher viral load at enrollment and type-specific positivity at month 6, we further examined whether the association was modified by viral loadassociated factors. As shown in Table 3, in contrast to a null association of viral load at baseline with type-specific positivity at month 6 among women who had monotype infection at Table 2 ). The magnitude of the P p .34 risk association was not meaningfully different according to current smoking status or cervical cytologic findings at enrollment. Among women with multiple HPV types at enrollment, coinfection with HPV-16-related types from the a-9 species was detected in 76 (39.4%) of 193 women with and 116 (45.1%) of 257 women without detectable HPV-16 DNA at month 6 ( ). Baseline coinfection with HPV-18-related P p .22 types from the a-7 species was detected in 22 (28.9%) of 76 women with and 40 (36.7%) of 109 women without detectable HPV-18 DNA at month 6 ( ). P p .27

DISCUSSION
In this 2-year longitudinal study of women with HPV-16 and/ or HPV-18 infection at enrollment, we found that the HPV-16 or HPV-18 DNA load at baseline was significantly higher among women with, compared with women without, a persistent infection during the first 6-month follow-up. However, for women who remained positive for viral DNA at visits occurring at month 6, 12, or 18, the probabilities of infection persisting for another 6 months did not differ appreciably by HPV-16 or HPV-18 DNA load at enrollment. This finding suggests that the higher viral load of the prevalent infection predicted greater short-but not long-term persistence.
The loss of strength in the association over time cannot be explained by loss to follow-up or the length of follow-up, because the viral load was equivalent both between women with and without follow-up and according to the number of followup visits. Ascertainment bias was not an issue, because viral load measurement was performed without information on viral persistence. A substantial number of ALTS participants received a therapeutic procedure for уCIN2. In view of the association between viral load and the risk of уCIN2 [27][28][29] and that between уCIN2 treatment and viral clearance [30,31], a diagnosis of уCIN2 might be an intermediate step in the pathway from viral load to persistence. Therefore, we performed parallel analyses by censoring уCIN2 cases at the time of initial diagnosis rather than by adjusting for a diagnosis of уCIN2. The comparable results between the analyses with and without censoring suggested that it is unlikely that the loss of strength in the association over time was biased by уCIN2 treatment.
Studies of the association between HPV DNA load and persistence of infection are rare, with inconsistent findings reported [12][13][14][15][16][17][18][19]. None of the previously published studies, except for one by van Duin et al [15], examined risk association over time. In that study, the cervical samples that were assayed were a mixture of samples obtained at study entry and during follow-up that were divided into groups according to time (ie, 3-12 months, 13-24 months, and 25-59 months) before the end of follow-up. Viral clearance was defined as no HPV-16 DNA detected in the last follow-up sample or absence of HPV-16 DNA in у2 consecutively obtained samples. van Duin and colleagues found a significant association between viral clearance and a lower median HPV-16 DNA load in the group of samples obtained 3-12 months before the end of the study but not in the other 2 groups of samples. Although the study by Duin et al [15] was limited by a small sample size and a lack of normalization of individual viral loads, its findings somewhat concur with our results.
The mechanism for the association of a higher HPV-16 or HPV-18 DNA load at baseline with short-but not long-term persistence is not clear. One possibility is that a subset of prevalent infections was close to the end of the natural course of HPV infection. The viral load of these infections would be likely to decrease with time, thereby reducing the average viral load for women without a positive finding at month 6. Alternatively, infections with low viral loads might, in some cases, denote contamination by sexual partners or other viral states destined to produce short-term viral DNA persistence.
However, it is difficult to determine how to explain our finding that the association of a higher viral load with shortterm persistence of infection was seen among women with coinfection with other HPV types but not among those with monotype infection, especially for HPV-16. It is possible that some coinfection-related factors play a role in the association of viral load with short-term persistence. As suggested by reduced rates of HPV clearance in individuals with HIV infection and/or a low CD4 cell count [32][33][34][35][36], the cellular immune response of the host is thought to be critical to clearance of the infection. In a recent study [37], we observed statistically significant associations between a lower HPV-16 DNA load and coinfection with HPV-16-related types and between a lower HPV-18 DNA load and coinfection with HPV-18-related types. Given that closely related HPV types may share some conserved T cell epitopes to elicit cross-reactive immune responses [38,39], the coinfection-associated reduction in viral load was presumed to be related to cross-reactivity of the cellular immune response. Accordingly, the cellular immunity induced by other HPV types may help with clearance of the infection by targeting the infected cells and/or limiting the virus for viral DNA replication or production of viral progeny.
Given the aforementioned interpretation, one may expect rapid clearance in women with coinfection with other HPV types, compared with those without such coinfection. However, previous studies of the natural history of HPV infection demonstrated either no [40] or a positive [41,42] association between coinfection and persistence. Trottier et al [42] explained the positive association by saying that women with multiple types might have had a poor immune response to the virus that consequently permitted persistence of the infection. Interestingly, in our study, among women with multiple HPV types, we observed a slightly lower proportion of coinfection with the same species types in women with, compared with women without, detectable type-specific positivity at month 6. This difference, although not statistically significant, somewhat corroborates a previous report indicating that the extent of the cellular immune response to HPV-16 L1 viruslike particle vaccine correlated with the phylogenetic distance of HPV types [43]. Thus, the results of the present study extend the previous interpretation and raise the intriguing possibility that the association of viral load with short-term persistence among women with coinfection may be in part attributable to differences in cross-reactivity of the cellular immune response induced by the same species type versus that induced by other species types.
It should be pointed out that a lack of the association between the HPV-16 or HPV-18 DNA load and short-term persistence of infection among women with monotype infections appeared to be dissimilar to a recent report from the ALTS participants who had single infections [44]. They found that persistence was associated with a higher viral load. In that study, however, the exposure was semiquantitative measurement of the viral load by use of Hybrid Capture (Digene Diagnostic), and the end point was the average type-specific persistence of a group Excluded were 5 HPV-16-positive women whose enrollment samples were inadequate for cytologic diagnosis. Because of the limited number of women with HSIL who remained HPV-18 positive at the month 6 follow-up visit, women with HSIL were grouped with those with LSIL for analysis of risk association.
of high-risk types. Although the estimated viral load determined by real-time PCR versus Hybrid Capture may be comparable for samples with a single HPV type, the range of types summarized as an outcome in that study extended to 14 types that included HPV-16 and HPV-18. This may in part explain the discrepancy in the results of the 2 studies. Several study limitations should be addressed. First, women included in the present study were those who were found to be positive for HPV-16 and/or HPV-18 at enrollment. Because we did not observe the time of acquisition of infection, we were unable to define the time from initial infection to regression. By selecting participants with prevalent HPV infection, we might have biased our study population toward women with a prolonged duration of positive findings. However, the influence of this possible bias is likely to be nondifferential. Second, we were unable to assess the effects of change in viral load on the persistence of infection, because, in this study, only the viral load at baseline was measured. Although this does not affect the validity of the findings, data on type-specific positivity subsequent to the viral load at each of follow-up visits would further provide a dynamic view of the risk association. Third, in ALTS, HPV positivity was detected by interval. The disappearance and occurrence within the interval remain undetermined. Such a misclassification may lead to overestimation of the length of viral persistence. It is unknown whether this would be differentially related to viral load. Last, because of the detection limit of PCR, whether the undetectable HPV DNA is the result of clearance of the infection or a latent infection in basal cells cannot be determined.
In summary, our data indicated that the persistence of typespecific infection in the first 6 months after enrollment was significantly associated with the HPV-16 or HPV-18 DNA load at baseline. Among women who continued to be positive for type-specific HPV DNA at month 6, 12, or 18 visits, persistence for another 6 months was unrelated to the viral load at baseline. The association between a higher viral load and short-term persistence of infection among women with multiple HPV types but not among those with monotype infections suggests a potential role of coinfection-related cross-reactivity of cellular immune response in clearance of infection.