Variants of human papillomavirus (HPV) type 31 have been shown to be related both to risk of cervical lesions and racial composition of a population. It is largely undetermined whether variants differ in their likelihood of persistence. Study subjects were women who participated in the ASCUS-LSIL Triage Study and who had a newly detected HPV31 infection during a two-year follow-up with six-month intervals. HPV31 isolates were characterized by sequencing and assigned to one of three variant lineages. Loss of the newly detected HPV31 infection was detected in 76 (47.5%) of the 160 women (32/67 with A variants, 16/27 with B variants and 28/66 with C variants). The adjusted hazard ratio associating loss of the infection was 1.2 (95% CI, 0.7–2.1) for women with A variants and 2.1 (95% CI, 1.2–3.5) for women with B variants when compared with those with C variants. Infections with A and C variants were detected in 50 and 41 Caucasian women and in 15 and 23 African-American women, respectively. The likelihood of clearance of the infection was significantly lower in African-American women with C variants than in African-American women with A variants (p = 0.05). There was no difference in the likelihood of clearance between A and C variants among Caucasian women. Our data indicated that infections with B variants were more likely to resolve than those with C variants. The difference in clearance of A vs. C variants in African-Americans, but not in Caucasians, suggests a possibility of the race-related influence in retaining the variant-specific infection.
Human papillomavirus 31 (HPV31) is one of the oncogenic types most closely related to HPV16.1 Studies of HPV31 DNA intratypic variations have demonstrated the presence of a variety of variants in all populations examined.2–9 As suggested by an analysis of the whole HPV31 genome,10 these variants could be phylogenetically classified into one of three lineages (i.e., group A, B and C). Results from our recent study indicated that such a genotypic classification reflects the clinical relevance of the variants, as women with A or B, compared with C, variants were at a significantly higher risk of cervical intraepithelial neoplasia grade 2 or 3 (CIN2/3).11 Given the importance of persistent infection in the development of cervical cancer and its precursor lesion,12–21 it would be interesting to see whether these variants differ in their likelihood of persistence.
Similar to the race-associated distribution of HPV16 and HPV18 variants, the distribution of HPV31 variants is related to the racial composition of a population, with A variants being predominant among Caucasian women and C variants being predominant among African-American women.21 The race-associated distribution of the variants may result from long-term sexual mixing patterns in the population or genetic influences of the host which preferentially predispose women to establish and/or retain infection with particular variants. As reported previously,22 HPV16 and 18 variants persisted longer in the host in which they have co-evolved, i.e., European variants persisted longer in Caucasian women and African variants persisted longer in African-American women. It is unknown, but interesting to see, whether the same is true for HPV31 variants.
To address our research interests, we compared the likelihood of loss of the variant-specific infection, overall and by race, among women with a newly detected HPV31 infection. Knowing the relationship between HPV31 variants and their likelihood of clearance may further refine our understanding of HPV-related pathogenesis.
Material and Methods
Study subjects and specimens
Study subjects were women who participated in the Atypical Squamous Cells of Undetermined Significance (ASC-US) and Low-Grade Squamous Intraepithelial Lesion (LSIL) Triage Study (ALTS). A detailed description of the ALTS design and study population has been reported elsewhere.23 Briefly, between January 1997 and December 1998, 5,060 women with a Pap of ASC-US or LSIL in the previous six months were enrolled and randomly assigned into one of three trial arms. These arms differed only in criteria for referral to colposcopy and colposcopically directed biopsy of visible lesions at enrolment with a referral being given to all women in the immediate colposcopy arm, those with oncogenic HPV types in the HPV triage arm and those with a cytologic diagnosis of high-grade SIL (HSIL) in the conservative management arm. All participants underwent an entry procedure at enrollment including interview, Pap smear and testing for HPV DNA and then returned every six months over two years for cervical cytology and HPV DNA testing. During follow-up, women were re-referred for colposcopy and biopsy if they had a cytologic diagnosis of HSIL. Women were scheduled for colposcopy as part of their exit visit at month 24. Women with CIN2/3 were treated with a loop electrosurgical excision procedure (LEEP). The ALTS protocol was approved by the institutional review boards at the National Cancer Institute and at each of the four clinical centers involved in the trial.
Specimens from ALTS participants who were without HPV31 infection at enrollment but became positive during follow-up (i.e., with a newly detected HPV31 infection) were included in the present study. As reported previously,11 the first positive sample from all HPV31-positive women and a random set of the last positive sample were tested for variants; the identical variant was found in each of the paired positive samples. Data on HPV typing results, clinical diagnoses and characteristics of 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.
Characterization of HPV31 variants
This was a second analysis of behavior of HPV31 variants among ALTS participants. A detailed protocol for characterization of HPV31 variants and lineage classification was reported elsewhere.11 Briefly, sequence variations of the 1,000-bp DNA fragments from nucleotide position 7,810 to 897 were determined by polymerase chain reaction (PCR)-based direct DNA sequencing. A viral isolate was defined as a distinct variant if as compared with the prototype24 and other isolates, there was one or more nucleotide alterations in the region analyzed. These variants were then phylogenetically classified as A, B or C variant lineages. As recommended recently,10 the prototype was included in the A lineage.
In assessment of the likelihood of variant-specific persistence, we used loss of the infection (or clearance) as the endpoint, defined as without PCR-detectable HPV31 DNA following the onset of HPV31 infection. Kaplan-Meier analyses were performed to estimate the mean and median time to loss of the variant-specific infection. A log-rank test was used to assess the significance of differences in the overall likelihood of clearance between the variants. The time variable was defined as the onset date of the infection until the date of clearance. Those women who completed the follow-up without clearing their infection or were lost to follow-up were censored at their last available positive visit. As HPV status was assessed only at visits that were scheduled every six months, the precise onset time was unknown. We assumed that the event of the infection occurred at one third of the time interval prior to the first HPV31-positive visit; the event of clearance occurred at one third of the time interval subsequent to the latest positive visit (before the negative detection). Results remained similar when the event was assumed to occur at the mid-point between the visits. For simplicity, these results were not presented. Sixteen women had a single missing visit (i.e., no HPV testing results) prior to the first positive detection. The onset of the infection for these women was calculated using the first positive and the preceding negative date. For women with a single missing visit prior to the date on which HPV31 DNA became undetectable (n = 6), the time of clearance was calculated using the latest positive and the following negative date. Women with a single missing visit between HPV31-positive visits (n = 5) were considered to be consecutively positive. The distribution of the variants between women with and without a missing visit was comparable (data not shown).
Cox proportional hazard regression analysis (37) was used to assess hazard ratios (HR) and their 95% confidence intervals (CI) for the association between HPV31 variants and the likelihood of loss of the infection. To identify appropriate covariates for adjustment, we evaluated risk factors for persistence of newly detected HPV31 infections by univariate analyses. Potential risk factors included age at enrollment, study arm, sexual behavior, use of hormonal contraceptives, number of Pap tests per year in the past five years, current smoking status and co-infection with non-HPV31 oncogenic types. Variables with p < 0.2 in univariate analyses were entered into multivariate models as covariates. We used a backward stepwise regression to construct the final models, with p < 0.20 as the criterion for entering and removing variables. Covariates included in the final model were study arm, number of Pap tests per year in the past five years (<1 vs. ≥1) and co-infection with non-HPV31 oncogenic types (i.e., HPV16/18/33/35/39/45/51/52/56/58/59/68). Three women had an HPV31-negative test between two positive visits. The intervening negativity was treated as being positive in an additional analysis. Twenty-three women had a diagnosis of CIN2/3 at visits where HPV31 DNA was detected. Parallel analyses were performed by censoring women with CIN2/3 at the time of initial diagnosis or treating CIN2/3 cases as being positive through the last available visit. The assumption of proportional hazard was examined with Schoenfeld residuals, showing no significant changes in hazard ratios with increasing follow-up time.
To determine whether persistent infection was a risk factor for CIN2/3, we compared the risk of CIN2/3 at the first positive visit vs. that at subsequent positive visits. In this analysis, the time of the first positive detection was rescaled as month 0, the next visit as month 6 and the visit after the next as month 12. CIN2/3 diagnosed at the time concurrent to the first HPV31-positive detection was considered to be related to a short positive duration. Those without CIN2/3 at the first positive visit who continued to be HPV31-positive at the next visit were at risk for CIN2/3 that was related to two consecutive positive tests (a surrogate of persistent infection) and so on for those with three consecutive positive tests. Logistic regression analysis (34) was used to estimate risk of CIN2/3 by the timing of HPV31 positivity. In ALTS, CIN2/3 cases were more likely to be diagnosed at exit than during follow-up because of different criteria for referral to colposcopy. Thus, odds ratios (OR) were adjusted for visits of the first HPV31-positive detection in addition to lineage of the variants and co-infection with non-HPV31 oncogenic types. Considering that a simple adjustment for visits of the first positive detection may not adequately control for referral-related confounding effects, we further performed Fisher's exact mid-p test to compare proportions of CIN2/3 at exit between women who continued to be positive for a pre-existing infection and those who had a newly detected infection at exit.
The mean age at enrollment by the variant lineage and the mean length of follow-up since the first positive detection (without taking censoring into account) were compared using one-way ANOVA or Student's t-test when appropriate. We used chi-square test to compare proportions of A and C variants between Caucasian and African-American women and characteristics of women who were included in the study vs. those who were excluded. All tests of significance were at the 5% two-sided significance level.
Of the 205 ALTS participants having a newly detected HPV31 infection during follow-up, 37 were excluded because their first positive samples were unavailable (n = 2) or tested negative by PCR-based DNA sequencing (n = 35). Considering the increased uncertainty on the onset time of the infection, we additionally excluded eight women who had two consecutive missing visits prior to the first positive detection. This left 160 women in the analysis. Compared to women included in the study, those excluded were less likely to be current hormonal contraceptives users (31% vs. 50%, p = 0.03) or report having ≥6 lifetime male sex partners (36% vs. 51%, p = 0.09); but no appreciable differences were observed with respect to study arm, age, race, current smoking status, number of Pap tests per year in the past five years and co-infection with non-HPV31 oncogenic types at the time of the first positive detection (data not shown).
Of the 160 women included, 67 (41.9%), 27 (16.9%) and 66 (41.2%) were positive for HPV31 A, B and C variants, respectively. The mean age (±SD) at enrolment was 24.5 (±7.8) years for women with A variants, 22.7 (±4.6) years for women with B variants and 25.3 (±7.1) years for women with C variants (p = 0.29). Infection with A, B and C variants was detected in 22 (41.5%), 8 (15.1%) and 23 (43.5%) women in the immediate colposcopy arm, 17 (39.5%), 8 (18.6%) and 18 (41.9%) women in the HPV triage arm and 28 (43.8%), 11 (17.2%) and 25 (39.1%) women in the conservative management arm, respectively. Forty-three (64.2%) of 67 women with A variants, 15 (55.6%) of 27 women with B variants and 33 (50.0%) of 66 women with C variants reported having <1 Pap test per year in the last five years. The mean length (±SD) of follow-up since the onset of the infection was 11.5 (±7.3) months for women with A variants, 11.4 (±7.6) months for women with B variants and 11.5 (±6.3) months for women with C variants (p = 0.99). Twenty-one women did not return for follow-up at month 24 (7 with A variants, 3 with B variants and 11 with C variants). There was no difference in proportions of completing the scheduled exit visit for women with HPV31 A, B or C variants (p = 0.54).
Loss of the newly detected HPV31 infection by variant lineage
During follow-up, loss of the newly detected HPV31 infection was observed in 76 (47.5%) of the 160 women (32/67 with A variants, 16/27 with B variants and 28/66 with C variants). Table 1 shows the mean and median time to clearance from the onset of the infection, with infections with B variants displaying a rapid clearance. After adjusting for study arm, number of Pap tests per year in the past five years and a time-dependent covariate of co-infection with non-HPV31 oncogenic types, the HR associating loss of the infection was 1.2 (95% CI, 0.7–2.1) for women with A variants and 2.1 (95% CI, 1.2–3.5) for women with B variants when compared with those with C variants (Table 2).
Table 1. Mean and median time to HPV31 DNA clearance from the onset of the infection
Table 2. Hazard ratios (HR) for the association between HPV31 variants and loss of the infection among women with a newly detected infection
Three women had an HPV31-negative test result between two positive visits. An analysis that treated the intervening negative test as being positive yielded an adjusted HR of 1.5 (95% CI, 0.9–2.5) for infections with A variants and 2.0 (95% CI, 1.2–3.5) for infections with B variants when compared with those with C variants. Twenty-three women had a diagnosis of CIN2/3 at the HPV31-positive visits. When these women were censored at the time of CIN2/3 diagnosis, the adjusted HR associating with loss of the infection was 1.4 (95% CI, 0.8–2.6) for women with A variants and 2.0 (95% CI, 1.1–3.6) for women with B variants when compared with those with C variants. An analysis that arbitrarily treated CIN2/3 cases as being positive through the last available visit did not appreciably alter the estimate of the association (HR = 1.3, 95% CI, 0.8–2.4 for infections with A vs. C variants; HR = 2.0, 95% CI, 1.1–3.5 for infections with B vs. C variants). The association remained similar when the analysis was restricted to women without a diagnosis of CIN2/3 or to women with at least one follow-up visit subsequent to the first HPV31-positive detection (data not shown).
We previously reported an increased risk of CIN2/3 associated with HPV31 A or B, compared with C, variants. However, data from this study showed a more rapid clearance of infections with A or B variants than those with C variants. This brings about a desire to examine risk of CIN2/3 by the length of HPV31 positive duration within this dataset.
Development of CIN2/3 by the timing of HPV31 positivity
HPV31 infection was initially detected at the first, second, third and fourth follow-up visits in 51, 45, 32 and 32 women, respectively. As shown in Table 3, 9 (5.6%) of the 160 women had a diagnosis of CIN2/3 at the time concurrent to the first positive detection (6/67 with A variants, 1/27 with B variants and 2/66 with C variants). Of the 76 women without a diagnosis of CIN2/3 at the first positive visit who continued to be positive at the next visit, 10 (13.2%) developed CIN2/3 at that visit (7/28 with A variants, 2/11 with B variants and 1/37 with C variants). Four (14.8%) of the 27 women with three consecutive positive visits had a diagnosis of CIN2/3 12 months after the first positive detection (2/10 with A variants, 1/3 with B variants and 1/14 with C variants). Coinfection with non-HPV31 oncogenic types was detected in 96 (60.0%) of the 160 women at the first positive visit, 33 (43.4%) of the 76 women at the second of two consecutive positive visits and 14 (51.9%) of the 27 women at the third of three consecutive positive visits.
Table 3. Odds ratios (OR) for associating risk of CIN2/3 with the timing of HPV31 positivity
Compared with the risk of CIN2/3 at the first positive visit, the likelihood of having a diagnosis of CIN2/3 at subsequent positive visits was significantly higher (OR = 4.3, 95% CI, 1.4–13.0 for those with two consecutive positive visits and OR = 6.5, 95% CI, 1.5–29.2 for those with three consecutive visits, adjusted for HPV31 variant lineage, co-infection with non-HPV31 oncogenic types and visits of the first HPV31-positive detection). HPV31 DNA was detected at exit in 70 women, including 38 with a pre-existing infection (i.e., persisting for ≥2 visits) and 32 with a newly detected infection. Women who continued to be positive for the pre-existing infection were more likely to have an exit diagnosis of CIN2/3 than those with the newly detected infection at exit, although the difference was not statistically significant (8/38 vs. 2/30, mid-Pexact = 0.09).
Variant-specific clearance of the newly detected HPV31 infection by racial group
Results from our previous analysis (Hulbert A, DeFilippis V, Koutsky LA, et al. unpublished data) demonstrated a race-associated distribution of HPV31 variants, i.e., A variants being predominant among Caucasian women and C variants being predominant among African-American women. A question raised is whether the persistence of the variants differs by race of study subjects.
Due to the limited number of infections with B variants (n = 27) and only four women belonging to a non-Caucasian and non-African-American race category, the analysis of the race-associated clearance of HPV31 variants was confined to Caucasian and African-American women who were positive for either A or C variants. Infections with A and C variants were detected in 50 (54.9%) and 41 (45.1%) Caucasian women and in 15 (39.5%) and 23 (60.5%) African-American women, respectively (p = 0.11). The mean length (±SD) of follow-up since the onset of the infection was 11.9 (±7.2) months for Caucasian women with A variants and 11.1 (±6.8) months for those with C variants (p = 0.58); it was 9.7 (±7.6) months for African-American women with A variants and 12.7 (±5.2) months for those with C variants (p = 0.20).
The overall likelihood of clearance of the infection was significantly lower for African-American women with C variants than for African-American women with A variants (Fig. 1a, log rank test, p = 0.05). Among African-American women, the mean time to loss of the infection from the onset of the infection was 12.5 months (95% CI, 9.5–15.5) for those with C variants and 7.6 months (95% CI, 4.5–10.7) for those with A variants. The analyses yielded similar results when treating the intervening negative test as being positive, censoring women with CIN2/3 at the time of diagnosis or treating CIN2/3 cases as being positive through the last available visit (data not shown). Among Caucasian women, there was no evidence of difference in clearance between infections with A and C variants (Fig. 1b, log rank test, p = 0.85).
In this two-year prospective study of women with a newly detected HPV31 infection, we found that infections with B variants were more rapidly to resolve than those with C variants. The difference could not be explained by factors that correlated to persistence of HPV31 infection such as study arm, number of Pap tests per year in the past five years and co-infection with non-HPV31 oncogenic types. Because women included in this study were those without a detectable HPV31 infection at enrollment in ALTS, a prevalent infection-related bias was minimized.
One concern is that HPV31 infection may resolve as a consequence of therapeutic procedure for cervical lesion.25, 26 In ALTS, women with a histologic diagnosis of CIN2/3 were treated with LEEP. CIN2/3 cases might have had a prolonged positive duration had the lesion not been excised. Also, the influence of the treatment-induced elimination of the virus would be more pronounced for women with A or B variants than for those with C variants because of the association between CIN2/3 and HPV31 variants. To address whether this could explain our findings, we reassessed the risk association by censoring the follow-up time at the date of CIN2/3 diagnosis, arbitrarily treating CIN2/3 cases as being positive through the last available visit or restricting the analysis to women without a diagnosis of CIN2/3. We also noted that some women had a single negative test between HPV31 positive visits. Considering that the intervening negativity may result from a natural fluctuation of viral DNA or variability in the sampling procedure, it was treated as being positive in the additional analysis. The consistency of the results obtained from various analytic strategies suggests that our findings are robust.
The published reports regarding the variant-specific persistence for a given type of HPV are rare; most of them have focused on HPV16 and HPV18 variants.27–31 Our results agree with a recent report from a case-control study nested in the population-based cohort,31 showing that infections with HPV31 A or B variants were less likely to persist than infections with C variants, although the difference was not statistically significant (p = 0.22). In that study, HPV31 A and B variants were grouped together without a distinction between the prevalent and incident infection. To the best of our knowledge, the present report is the first to address the variant-specific clearance among women with a newly detected HPV31 infection.
Given a previous report regarding the increased risk of CIN2/3 associated with HPV31 A or B, compared with C, variants,11 we would have expected a reverse direction of loss of the variant-specific infection because it is generally thought that risk of CIN2/3 increases with increasing duration of persistence. In view of the apparent conflicting results between the variant-associated risk of CIN2/3 and likelihood of viral DNA clearance, a question raised is whether a persistent infection still plays a role in risk of CIN2/3 in this setting. As noted, CIN2/3 was more likely to be diagnosed at exit than during follow-up because of different criteria for referral to colposcopy implemented in ALTS. Also, most new infections were initially detected during follow-up. We tried to mitigate the effects of the referral criteria on outcome ascertainment by adjusting for visits of the first HPV31-positive detection. To further minimize this distorting effect, we compared the rates of CIN2/3 detected at exit only (assuming under the same referral criteria) between women who had a newly detected HPV31 infection at exit and those who continued to be positive for a pre-existing infection. Consistent with previous reports,12–20 women with a persistent HPV31 infection were at a significantly increased risk of CIN2/3. The underlying mechanisms for the conflicting association (i.e., infections with C, compared with A or B, variants were less likely to resolve but conferred lower risk of CIN2/3) are not clear. Obviously, the association of CIN2/3 with HPV31 variants was not mediated through a prolonged positive duration, although persistent infection was a risk factor for CIN2/3. It is possible that HPV31 B variants as compared with C variants may be more aggressive, thereby leading to a rapid progression from the infection to the development of CIN2/3. Analogue to this is the scenario of high- vs. low-risk HPV types: infections with some low-risk types, although conferring lower risk of cervical lesion, could persist longer than those with some high-risk types (except for HPV16).
An interesting finding of this study is the race-associated clearance of HPV31 variants which enriches our understanding of a role of the viral-host interaction in defining persistent infection. In agreement with the previous report of the race-associated distribution and persistence of HPV16 and HPV18 variants,22 infections with HPV31 C, compared with A, variants were less likely to resolve among African-American women. Although no opposite trend was observed among Caucasian women, the difference seen among African-American women disappeared among Caucasian women. It is possible that among African-American women, infections with C variants, the predominant ones, were more likely to be established and retained than those with A variants because of the adeptness between the host and virus during a long-term co-evolution. The reason for a lack of the opposite trend among Caucasian women is not clear. One possibility is that the race-related growth advantage of HPV31 A over C variants, if present, might be compromised by effects of some unknown factors in this group of women. Alternatively, infections with C variants tended to persist longer than those with A variants but such a difference was attenuated among Caucasian women by the host factors that favor retaining the A variants. Nevertheless, the difference in clearance between infections with A and C variants seen among African-American women but not among Caucasian women suggests a potential role of the race-related genetic factors in retaining or clearing the variant-specific infection.
Several limitations of the study should be addressed. The results of this study were based on a relatively small number of HPV31 infections, particularly for analyses of race-associated clearance of variant-specific infection. Confirmation of these findings by large-scale studies is warranted. Secondly, ALTS participants had a cytologic diagnosis of ASC-US or LSIL within six months prior to enrollment. Thus, a so-called “new” infection could be either newly acquired (a true incident infection) or reactivated from a pre-existing latent infection. It is currently unknown whether these two kinds of new infections differ by variants and/or in lengths of persistence. Thirdly, we noticed that 21 women exited the trial before the month 24 visit, and 27 women had a single missing visit right before the first positive test, prior to the negative reversion, or between the positive visits. Biases could have been introduced if loss to follow-up had been differentially related to HPV31 variants. However, the mean length of follow-up since the first positive detection was comparable among women with A, B or C variants, and so was the proportion of completing the exit visit. Also, the distribution of HPV31 variants was comparable between women with and without a missing visit. Finally, ALTS participants were followed every six months for only two years. We were unable to count infections that occurred and disappeared within the interval. This may lead to an overestimate of the length of positive duration. On the other hand, as in all cohort studies of the newly detected HPV infection, the reported mean and median duration might be underestimated because of right-censoring of infections that did not clear by the end of follow-up. Presently, however, there are no data to suggest that these features would be differentially related to HPV31 variants.
In summary, our data indicated that in a population with Caucasian and African-American women as the majority, infections with HPV31 B variants were more likely to resolve than those with C variants. The difference in the likelihood of clearance between A and C variants seen among African-American women but not among Caucasian women suggests a possibility of the race-related influence in retaining the variant-specific infection. Given the etiologic importance of persistent infection, further studies to examine mechanisms involving variant-specific clearance, particularly the race-associated disparities, are of great interest.
This study was part of the project ancillary to the ALTS clinical trial but does not represent the ALTS Group. The authors would like to thank the ALTS Group for providing cervical samples and ALTS data for this study. Dr. Mark Schiffman is collaborating with Roche Molecular Systems Inc. (Alameda, CA), which supplies free testing of cervical specimens for his research. They do not control any aspect of resultant data. Other authors have no commercial or other associations that might pose a conflict of interest.