Abstract
We examined the concurrence of multiple human papillomavirus (HPV) infections in 47,617 women who underwent cervical screening in New Mexico between December 2007 and April 2009 using the LINEAR ARRAY HPV Genotyping Test (Roche Diagnostics, Indianapolis, Indiana), which detects 37 different types of HPV. Our primary goal was to examine the distributions of multiple HPV types with a special interest in negative interactions, which could signal the possibility of type replacement associated with a common niche if some HPV types were prevented by vaccination. Multiple infections were found to be more common than expected under independence, but this could largely be accounted for by a woman-specific latent heterogeneity parameter which was found to be dependent on age and cytological grade. While multiple infections were more common in young women and in those with abnormal cytology, greater heterogeneity was seen in older women and in those with normal cytology, possibly reflecting greater variability in exposure due to current or past HPV exposure or due to heterogeneity in related HPV reactivation or in immune responses to HPV infection or persistence. A negative interaction was found between HPV 16 and several other HPV types for women with abnormal cytology but not for those with normal cytology, suggesting that type replacement in women vaccinated against HPV 16 is unlikely to be an issue for the general population.
Keywords: cervical screening, human papillomavirus, multiple infections, population heterogeneity
Editor's note:An invited commentary on this article appears on page 1076.
Infection with more than 1 type of human papillomavirus (HPV) is commonly observed in women who are infected with HPV (1–7). Furthermore, the number of HPV types is often observed to significantly exceed expectation under the assumption that infection with different types is independent at the population level. It has been suggested that the excess risk of multiple infections can be explained by a woman-level random effect attributable to unobserved covariates such as current and past HPV exposure and immune response to these viruses (5, 6). Under such a model, infections would be independent given a woman's unobserved random effect. Little is known about the epidemiology or determinants of multiple infections. In particular, do they vary according to age, existence and grade of concurrent cytological abnormality, or geographical region? It is also unclear as to whether there is a potential synergism or antagonism between specific types, or whether multiple infections occur between specific types in a manner that would be predicted by their individual prevalences. This is highly relevant for assessing the impact of HPV vaccines, as any negative interaction of a specific type with either HPV 16 or HPV 18 would suggest that type replacement might occur if types 16 and 18 were eliminated by vaccination. Such an interaction between HPV 16 and HPV 6/11 was seen in a retrospective serological study by Luostarinen et al. (8), and further assessment of potential HPV type interactions using HPV DNA measures in cervical samples is highly desirable.
Here, in a large cohort of New Mexico women undergoing routine cytological screening, we explored the age-specific prevalence of single and multiple HPV infections according to HPV type, the extent to which multiple HPV infections exceeded expectation under independence, and whether the excess could be explained by a women-level random effect. To carry out this analysis, we used a single-parameter frailty model developed recently by Cuzick and Yang (9). The single parameter of the frailty model captures the heterogeneity or population spread in susceptibility to multiple infections. Other investigators have briefly considered pairwise infections or infections of higher-order multiplicities based on HPV genotypes detected, but they did not develop a model for them (1, 2, 5, 7). We carried out this analysis in 3 age groups (≤30 years, 31–49 years, and ≥50 years) and 5 cytological groups (normal, atypical squamous cells of unknown significance (ASC-US), atypical squamous cells—cannot exclude HSIL (ASC-H), atypical glandular cells of undetermined significance (AGUS), low-grade squamous intraepithelial lesion (LSIL), and high-grade squamous intraepithelial lesion (HSIL)). We also examined interactions of HPV 16 and HPV 18 with other HPV types, interactions between low-risk and high-risk HPV types, and interactions between HPV types within the α9 and α7 species.
METHODS
Population
The data set consisted of a stratified sample of women who underwent cytological screening at clinical practices served by 7 major laboratories in the state of New Mexico from December 2007 to April 2009. These laboratories processed 79% of all cervical cytological specimens in the state. A summary of the sampling scheme is given in Figure 1. Specimens for HPV genotyping were randomly selected from each of these 7 laboratories within 4 strata defined by age (≤30 years vs. >30 years) and by whether the results of cytological screening were normal or abnormal, with the sampling fractions varying across strata. A total of 47,617 women were included in these analyses. The sample sizes across the 3 age groups (≤30 years, 31–49 years, and ≥50 years) and the 5 cytology groups, assigned using the Bethesda System (10)—normal, ASC-US, ASC-H, AGUS, LSIL, and HSIL—are shown in Table 1. Women aged 30 years or younger with normal cytology constituted the largest sample group, while women aged 50 years or older with normal cytology constituted the largest population group. More details on the study population can be found in Wheeler et al. (11).
Figure 1.
Selection criteria used in a study of concurrence of multiple human papillomavirus (HPV) infections in 47,617 women who underwent cervical screening, New Mexico, 2007–2009.
Table 1.
Sample Sizes Used in a Study of Concurrence of Multiple Human Papillomavirus Infections Among 47,617 Women Who Underwent Cervical Screening, by Age and Cytological Outcome, New Mexico, 2007–2009
| Age Group, years | Cytology Group |
|||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Normal |
ASC-US |
AGUS |
LSIL |
ASC-H |
HSIL |
Abnormal |
All |
|||||||||
| No. | Row % | No. | Row % | No. | Row % | No. | Row % | No. | Row % | No. | Row % | No. | Row % | No. | Row % | |
| ≤30 | 23,070 | 75.9 | 4,079 | 13.4 | 102 | 0.3 | 2,510 | 8.3 | 336 | 1.1 | 308 | 1.0 | 7,335 | 24.1 | 30,405 | 100 |
| 31–49 | 5,647 | 56.3 | 3,014 | 30.0 | 260 | 2.6 | 748 | 7.5 | 219 | 2.2 | 143 | 1.4 | 4,384 | 43.7 | 10,031 | 100 |
| ≥50 | 4,897 | 68.2 | 1,653 | 23.0 | 186 | 2.6 | 282 | 3.9 | 97 | 1.4 | 66 | 0.9 | 2,284 | 31.8 | 7,181 | 100 |
| All ages | 33,614 | 70.6 | 8,746 | 18.4 | 548 | 1.2 | 3,540 | 7.4 | 652 | 1.4 | 517 | 1.1 | 14,003 | 29.4 | 47,617 | 100 |
Abbreviations: AGUS, atypical glandular cells of undetermined significance; ASC-H, atypical squamous cells—cannot exclude HSIL; ASC-US, atypical squamous cells of unknown significance; HSIL, high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion.
Liquid-based cytological specimens (SurePath (Becton, Dickinson and Company, Franklin Close, New Jersey) or ThinPrep (Hologic, Inc., Bedford, Massachusetts)) were tested for 37 different types of HPV using the LINEAR ARRAY HPV Genotyping Test (Roche Diagnostics, Indianapolis, Indiana), hereafter called the HPV LA. For technical reasons, we omitted HPV types 52 and IS39 from the original 37 HPV types in the analyses; the former is inferred indirectly from the presence of a combination probe in the absence of HPV types 33, 35, and 58, and the latter is a subtype of HPV 82 and was therefore counted as HPV 82 in our analyses. Thus, the data for these analyses consisted of the binary outcome for each of the 35 HPV types as well as age and cytological group for each woman.
For each age and cytology subgroup, we examined the prevalence of single and multiple HPV infections with the 35 HPV types. We also examined the independence of different HPV types and the extent to which the interactions could be explained by a frailty or random effect which was not type-specific but woman-specific and possibly related to differences in sexual exposure to HPV genotypes overall (past or recent) or differences in the ability to clear latent or newly acquired HPV infections. We also examined concurrences between HPV types belonging to the α9 species (types 16, 31, 33, 35, 58, and 67), types belonging to the α7 species (types 18, 39, 45, 59, 68, and 70), high-risk types (types 16, 18, 31, 33, 35, 39, 45, 51, 56, 58, and 59), and low-risk types (types 6, 11, 40, 42, 54, 61, 72, and 81).
HPV genotype testing
The HPV LA is a qualitative test for 37 HPV genotypes incorporating selective polymerase chain reaction amplification with biotinylated PGMY 09/11 L1 region consensus primers and colorimetric detection of amplified products bound to immobilized HPV genotype-related oligonucleotide probes on a LINEAR ARRAY HPV genotyping strip. PGMY-based HPV genotyping with the HPV LA and a prototype line blot assay have been previously described in detail (11). Following vigorous mixing of the original liquid cytological specimens, 500-µL aliquots of SurePath (Becton, Dickinson and Company) or ThinPrep (Hologic, Inc.) were transferred to 12-mm × 75-mm polypropylene tubes, and DNA was purified using a Cobas X421 robot (Roche Molecular Systems, Inc., Pleasanton, California). The robot performed proteinase K digestion and inactivation with the final DNA eluate (150 µL) delivered into a 96-well QIAamp plate (Qiagen, Inc., Valencia, California). Fifty microliters (50 µL) of purified DNA was transferred to a tube with 50 µL of HPV LA master mix, and the mixture was amplified by polymerase chain reaction using the 96-well gold-plated GeneAmp PCR System 9700 from Applied Biosystems (Life Technologies, Grand Island, New York) as specified by the manufacturer. Controls for contamination and assay sensitivity were included in each 96-well assay.
Using the Roche HPV LA detection kit, hybridizations were automated using Tecan ProfiBlot-48 robots (Tecan Austria GmbH, Grödig, Austria). The HPV LA detects 37 types of HPV, 35 of which were analyzed here. Two independent readers interpreted the presence of HPV genotypes using a reference template provided by the manufacturer. Any discrepancies were identified by means of a custom computer application and were adjudicated by a third review.
Statistical analysis
We investigated the excess risk of multiple HPV type-specific infections across the 3 age groups (≤30 years, 31–49 years, and ≥50 years) and the 5 cytology groups (normal, ASC-US, ASC-H, AGUS, LSIL, and HSIL). The excess risk of k-fold concurrences is expressed as the ratio of the sum of observed concurrences (O) to the sum of expected (under independence) concurrences (E) for all type combinations of order k (O/E). Note that (k + 1)-fold concurrences would be counted as k + 1 separate k-fold concurrences. For low marginal prevalences, these excess risks can also be interpreted as odds ratios of a first-order Taylor approximation of a logistic regression model. Since these ratios are specific for the order of the concurrence, we considered a frailty model in which a multiplicative woman-specific random effect is used to unify these excess risks across all orders. Specifically, we fitted a 1-parameter multiplicative γ-frailty model (with unit mean) to all orders of the observed concurrences. More details on the model are given elsewhere by Cuzick and Yang (9).
The single parameter (σ) of the frailty model can be viewed as a heterogeneity parameter which summarizes the excess risks observed in concurrences of all orders. Thus, σ is a measure of the population spread in susceptibility to multiple infections. In particular, if σ is evaluated solely from pairwise concurrences, it is simply the O/E ratio minus 1. In practice, σ can be estimated using a minimum variance approach by combining the estimated σ values from the second-, third-, and fourth-order concurrences (9) and the covariance matrix of these 3 estimates. We used this parameter to examine the variability in excess risk across the different age and cytology groups, as well as for important subgroups of the 35 HPV types—high-risk types, low-risk types, and types belonging to the α9 and α7 species. We also evaluated σ for concurrences between classes of HPV types, including concurrences between high-risk and low-risk HPV types (high-risk/low-risk), between HPV 16 and all other types (16/other), and between HPV 18 and all other types (18/other). For high-risk/low-risk, 16/other, and 18/other, σ was evaluated using only pairwise HPV infections. Estimates were unavailable for groups that exhibited a significantly reduced risk of being infected with multiple types (denoted by σ < 0) and groups with insufficient data. Tests for trend (1 degree of freedom) across age and cytology groups were also conducted. Assessment of goodness of fit via standard techniques (details given by Cuzick and Yang (9)) was not helpful because there were many cells with small numbers of events, so we used an index of dispersion D to summarize the overdispersion of a fitted model, defined as the ratio of the sum of the squared difference between observed and expected cell counts (for fitted σ) to the sum of expected cell counts.
RESULTS
The prevalence of the 36 HPV types (including HPV 52) is given in Table 2 for each of the age and cytology subgroups, ordered by their overall prevalence. HPV 62 was the most common type for women aged 31 years or older and for women with normal cytology, whereas HPV 16 was the most common type in younger women (age ≤30 years) and women with abnormal cytology (most notably the cytology categories LSIL, ASC-H, and HSIL). All HPV types were more common in women with LSIL than in those with normal cytology. However, only the following high-risk HPV types were more than twice as common in women with HSIL as in women with ASC-US: types 16, 18, 31, 33, 35, and 58. HPV 51, another high-risk type, was the second most common type in women with HSIL and almost met this criterion. We also noted an excess risk of HPV 82 in women with HSIL versus women with ASC-US, but this is not conventionally considered a high-risk type.
Table 2.
Marginal Prevalences (%) of 36 Types of Human Papillomavirus Among 47,617 Women Who Underwent Cervical Screening, by Age and Cytological Outcome (Adjusted for Sampling Fractions), New Mexico, 2007–2009
| HPV Type | Age Group, years |
Total (n = 333,275) | Cytology Group |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ≤30 (n = 107,313)a | 31–49 (n = 122,292) | ≥50 (n = 103,670) | Normal (n = 310,349) | ASCUS (n = 14,253) | AGUS (n = 873) | LSIL (n = 5,882) | ASC-H (n = 1,067) | HSIL (n = 851) | Abnormal (n = 22,926) | ||
| 16 | 7.1 | 2.1 | 1.3 | 3.5 | 2.6 | 10.5 | 8 | 19.3 | 30 | 46.8 | 14.9 |
| 62 | 4.1 | 2.5 | 2.3 | 3.0 | 2.7 | 6.3 | 2.2 | 9.1 | 5.6 | 5.8 | 6.8 |
| 53 | 4.8 | 2.2 | 1.6 | 2.9 | 2.4 | 7.4 | 3.7 | 17.6 | 7.1 | 6.5 | 9.8 |
| 84 | 4.2 | 1.9 | 1.3 | 2.4 | 2.2 | 5.2 | 1.3 | 9.7 | 4.8 | 3.5 | 6.1 |
| 54 | 3.7 | 2.0 | 1.4 | 2.3 | 2.1 | 5.2 | 2.6 | 8.1 | 4.6 | 5.2 | 5.8 |
| 89 | 4.3 | 1.6 | 1.1 | 2.3 | 2.0 | 5.8 | 2.1 | 10.1 | 5.8 | 5.7 | 6.8 |
| 51 | 4.9 | 1.2 | 0.9 | 2.3 | 1.7 | 7.4 | 3.5 | 18.7 | 11.3 | 13.9 | 10.6 |
| 66 | 4.3 | 1.5 | 1.2 | 2.3 | 1.8 | 6.3 | 1.5 | 18.7 | 6.4 | 6.7 | 9.3 |
| 61 | 2.9 | 1.7 | 1.8 | 2.1 | 1.9 | 4.4 | 2.6 | 7.1 | 5.6 | 5.1 | 5.1 |
| 39 | 4.4 | 1.4 | 0.5 | 2.1 | 1.7 | 6.7 | 2.2 | 12.7 | 7.2 | 7.0 | 8.1 |
| 59 | 4.2 | 1.2 | 0.9 | 2.1 | 1.8 | 6.1 | 1.1 | 9.6 | 5.8 | 5.5 | 6.8 |
| 52b | 3.7 | 1.4 | 0.7 | 1.9 | 1.6 | 5.5 | 2.8 | 7.3 | 8.9 | 6.3 | 6.0 |
| 31 | 3.5 | 1.3 | 0.6 | 1.8 | 1.4 | 5.8 | 3.2 | 8.9 | 12.8 | 12.5 | 7.1 |
| 42 | 3.2 | 0.7 | 0.7 | 1.5 | 1.2 | 5.7 | 1.1 | 8.8 | 3.9 | 2.3 | 6.1 |
| 83 | 1.5 | 1.3 | 1.3 | 1.3 | 1.2 | 2.7 | 1.5 | 3.6 | 2.4 | 2.3 | 2.8 |
| 56 | 2.5 | 0.8 | 0.5 | 1.2 | 0.9 | 4.3 | 1.5 | 13.1 | 4.8 | 4.7 | 6.5 |
| 18 | 2.4 | 0.7 | 0.6 | 1.2 | 0.9 | 4.0 | 4.7 | 6.8 | 6.7 | 10.2 | 5.1 |
| 58 | 2.5 | 0.7 | 0.4 | 1.2 | 0.9 | 4.0 | 1.5 | 7.6 | 6.7 | 10.3 | 5.2 |
| 6 | 2.6 | 0.6 | 0.4 | 1.2 | 0.9 | 3.1 | 1.1 | 8.0 | 5.3 | 3.3 | 4.4 |
| 81 | 1.4 | 1.0 | 0.9 | 1.1 | 1.0 | 2.7 | 1.5 | 4.1 | 2.0 | 1.6 | 2.9 |
| 45 | 1.9 | 0.7 | 0.6 | 1.1 | 0.9 | 3.0 | 2.4 | 5.1 | 4.5 | 2.9 | 3.6 |
| 70 | 1.0 | 1.2 | 0.8 | 1.0 | 0.9 | 2.0 | 1.8 | 3.3 | 2.0 | 2.3 | 2.4 |
| 55 | 1.5 | 0.9 | 0.6 | 1.0 | 0.9 | 2.4 | 1.3 | 3.6 | 2.0 | 2.2 | 2.6 |
| 73 | 1.8 | 0.6 | 0.5 | 0.9 | 0.8 | 2.5 | 0.7 | 5.7 | 3.9 | 3.9 | 3.4 |
| 35 | 1.5 | 0.7 | 0.4 | 0.9 | 0.7 | 2.9 | 0.9 | 5.6 | 5.2 | 7.0 | 3.8 |
| 68 | 1.3 | 0.6 | 0.5 | 0.8 | 0.7 | 2.2 | 1.1 | 3.9 | 3.4 | 1.0 | 2.6 |
| 82c | 1.5 | 0.4 | 0.3 | 0.7 | 0.5 | 2.1 | 1.1 | 4.2 | 3.0 | 5.7 | 2.8 |
| 67 | 1.6 | 0.2 | 0.1 | 0.6 | 0.5 | 2.4 | 0.0 | 5.1 | 3.6 | 2.4 | 3.0 |
| 72 | 0.5 | 0.6 | 0.5 | 0.6 | 0.5 | 1.0 | 0.4 | 1.3 | 0.9 | 0.6 | 1.0 |
| 40 | 1.2 | 0.3 | 0.2 | 0.5 | 0.4 | 1.7 | 0.8 | 3.1 | 0.9 | 2.0 | 2.0 |
| 33 | 0.9 | 0.3 | 0.2 | 0.5 | 0.4 | 1.6 | 1.1 | 2.9 | 3.8 | 4.7 | 2.1 |
| 71 | 0.2 | 0.2 | 0.4 | 0.3 | 0.2 | 0.5 | 0.4 | 0.5 | 0.1 | 0.2 | 0.4 |
| 11 | 0.4 | 0.1 | 0.1 | 0.2 | 0.1 | 0.6 | 0.2 | 1.0 | 0.3 | 0.4 | 0.7 |
| 26 | 0.2 | 0.0 | 0.0 | 0.1 | 0.1 | 0.3 | 0.2 | 0.8 | 0.3 | 0.6 | 0.5 |
| 69 | 0.0 | 0.0 | 0.1 | 0.0 | 0.0 | 0.2 | 0.0 | 0.2 | 0.5 | 0.0 | 0.2 |
| 64 | 0.1 | 0.0 | 0.0 | 0.0 | 0.0 | 0.2 | 0.2 | 0.4 | 0.6 | 0.0 | 0.2 |
Abbreviations: AGUS, atypical glandular cells of undetermined significance; ASC-H, atypical squamous cells—cannot exclude HSIL; ASC-US, atypical squamous cells of unknown significance; HPV, human papillomavirus; HSIL, high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion.
a Estimated number of women in the population who had at least 1 cytological screen during the time period of the sample.
b Not included in subsequent analysis.
c Includes subtype 82_IS39.
Table 3 shows the proportions of women with single HPV infections, multiple HPV infections, and any HPV infection for the different age and cytology subgroups. Data are given for all HPV types and separately for high-risk and low-risk HPV types. Overall, rates of multiple HPV infection were higher in women with abnormal cytology and decreased with age. The LSIL cytology group had the highest multiple infection rates at all ages, although there was little difference across the LSIL, ASC-H, and HSIL groups for high-risk multiple HPV infections in younger women. Multiple infection rates for ASC-US and AGUS were generally intermediate between those for the normal cytological outcome and the other abnormal cytology groups. Both single and multiple low-risk HPV types were always more common in women with LSIL and least common in women with normal cytology. The multiple infection rates and proportions for high- and low-risk HPV types separately were generally lower than those for all 35 HPV types, reflecting infections with both low- and high-risk HPV types.
Table 3.
Prevalences (%) of Single-Type, Multiple-Type, and All Human Papillomavirus Infections Among 47,617 Women Who Underwent Cervical Screening, According to Age and Cytological Outcome, New Mexico, 2007–2009
| Age Group and Type of HPV Infection |
Cytology Group |
|||||||
|---|---|---|---|---|---|---|---|---|
| Normal | ASC-US | AGUS | LSIL | ASC-H | HSIL | Abnormal | All | |
| ≤30 years | ||||||||
| Single | 19.81 | 25.79 | 21.57 | 26.69 | 28.27 | 36.69 | 26.61 | 20.60 |
| Multiple | 16.36 | 49.55 | 35.29 | 69.60 | 60.42 | 61.36 | 57.21 | 21.10 |
| Any | 36.17 | 75.34 | 56.86 | 96.29 | 88.69 | 98.05 | 83.82 | 41.70 |
| HR typesa | ||||||||
| Single | 15.48 | 34.98 | 29.41 | 43.71 | 45.83 | 60.06 | 39.44 | 18.27 |
| Multiple | 5.00 | 21.87 | 17.65 | 32.35 | 34.82 | 33.77 | 26.49 | 7.49 |
| Any | 20.48 | 56.85 | 47.06 | 76.06 | 80.65 | 93.83 | 65.93 | 25.76 |
| LR typesa | ||||||||
| Single | 10.23 | 21.75 | 15.69 | 27.85 | 20.83 | 18.18 | 23.56 | 11.78 |
| Multiple | 1.85 | 6.01 | 2.94 | 7.81 | 5.36 | 3.90 | 6.46 | 2.38 |
| Any | 12.08 | 27.75 | 18.63 | 35.66 | 26.19 | 22.08 | 30.02 | 14.16 |
| 31–49 years | ||||||||
| Single | 14.26 | 21.6 | 21.92 | 37.03 | 39.73 | 55.24 | 26.25 | 14.93 |
| Multiple | 5.88 | 19.64 | 11.54 | 50.94 | 27.40 | 40.56 | 25.57 | 6.99 |
| Any | 20.13 | 41.24 | 33.46 | 87.97 | 67.12 | 95.80 | 51.82 | 21.92 |
| HR typesb | ||||||||
| Single | 7.46 | 18.55 | 18.08 | 42.25 | 44.75 | 70.63 | 25.57 | 8.47 |
| Multiple | 0.76 | 5.44 | 3.46 | 16.98 | 10.96 | 19.58 | 8.03 | 1.17 |
| Any | 8.22 | 23.99 | 21.54 | 59.22 | 55.71 | 90.21 | 33.60 | 9.64 |
| LR typesc | ||||||||
| Single | 6.02 | 12.67 | 7.31 | 24.20 | 14.16 | 9.79 | 14.3 | 6.49 |
| Multiple | 0.69 | 2.49 | 1.15 | 6.68 | 2.74 | 4.20 | 3.19 | 0.83 |
| Any | 6.71 | 15.16 | 8.46 | 30.88 | 16.89 | 13.99 | 17.50 | 7.32 |
| ≥50 years | ||||||||
| Single | 11.25 | 19.72 | 18.28 | 37.23 | 23.71 | 51.52 | 22.85 | 11.65 |
| Multiple | 4.68 | 15.85 | 6.45 | 50.35 | 23.71 | 24.24 | 19.92 | 5.20 |
| Any | 15.93 | 35.57 | 24.73 | 87.59 | 47.42 | 75.76 | 42.78 | 16.86 |
| HR typesb | ||||||||
| Single | 4.80 | 13.97 | 11.83 | 34.75 | 29.90 | 59.09 | 18.35 | 5.27 |
| Multiple | 0.65 | 2.12 | 1.08 | 15.96 | 5.15 | 7.58 | 4.03 | 0.77 |
| Any | 5.45 | 16.09 | 12.90 | 50.71 | 35.05 | 66.67 | 22.37 | 6.04 |
| LR typesc | ||||||||
| Single | 4.98 | 13.19 | 5.91 | 31.91 | 7.22 | 9.09 | 14.54 | 5.31 |
| Multiple | 0.55 | 2.54 | 1.08 | 10.64 | 3.09 | 3.03 | 3.46 | 0.65 |
| Any | 5.53 | 15.73 | 6.99 | 42.55 | 10.31 | 12.12 | 17.99 | 5.96 |
| All ages | ||||||||
| Single | 14.98 | 23.30 | 20.64 | 29.55 | 31.31 | 43.38 | 25.92 | 15.74 |
| Multiple | 8.69 | 33.50 | 14.56 | 64.43 | 44.55 | 51.37 | 41.89 | 10.98 |
| Any | 23.68 | 56.80 | 35.19 | 93.98 | 75.86 | 94.75 | 67.81 | 26.71 |
| HR typesb | ||||||||
| Single | 9.05 | 25.71 | 18.24 | 42.74 | 43.21 | 62.73 | 31.99 | 10.63 |
| Multiple | 2.02 | 12.83 | 5.48 | 28.05 | 22.91 | 26.84 | 17.44 | 3.08 |
| Any | 11.07 | 38.54 | 23.72 | 70.79 | 66.12 | 89.57 | 49.43 | 13.71 |
| LR typesc | ||||||||
| Single | 6.97 | 17.18 | 8.51 | 27.43 | 16.74 | 14.87 | 19.37 | 7.83 |
| Multiple | 1.00 | 4.21 | 1.48 | 7.80 | 4.19 | 3.87 | 5.01 | 1.28 |
| Any | 7.97 | 21.39 | 9.99 | 35.22 | 20.93 | 18.74 | 24.39 | 9.10 |
Abbreviations: AGUS, atypical glandular cells of undetermined significance; ASC-H, atypical squamous cells—cannot exclude HSIL; ASC-US, atypical squamous cells of unknown significance; HPV, human papillomavirus; HR, high-risk; HSIL, high-grade squamous intraepithelial lesion; LR, low-risk; LSIL, low-grade squamous intraepithelial lesion.
a Results are shown for all HR and LR types separately.
b HPV types 16, 18, 31, 33, 35, 39, 45, 51, 56, 58, and 59.
c HPV types 6, 11, 40, 42, 54, 61, 72, and 81.
The observed-versus-expected (under independence) counts for the 595 pairwise concurrences are plotted in Figure 2 for the normal and abnormal cytology subgroups. The average observed:expected (O/E) ratios are illustrated as slopes of the fitted lines. The slopes were lower in women with abnormal cytology and increased with age (data not shown). Similar trends were seen for the higher-order concurrences (e.g., triple infections).
Figure 2.
Observed numbers of pairwise human papillomavirus infections versus expected numbers (under independence) among 47,617 women who underwent cervical screening, New Mexico, 2007–2009. Solid lines correspond to the fitted excess risks, and dashed lines correspond to no interaction. Outlying pairs are noted with an X.
Table 4 shows that the heterogeneity parameter for all 35 HPV types was smallest in women aged 30 years or younger and increased with age (P = 0.02). It was also significantly lower in women with abnormal cytology, especially for the LSIL and HSIL groups. Heterogeneity in infection rates for different subsets of HPV types is shown in Table 5 for different cytology groups. It is clear that for each subgroup of HPV types, the greatest heterogeneity is seen in women with normal cytology. This is followed by the ASC-US and AGUS groups and the ASC-H group. LSIL and HSIL generally show the least amount of heterogeneity across all subgroups of HPV infections. Of note is the lower heterogeneity for concurrences between HPV 16 and other HPV types than for the other subgroups, including all α9 types. Further details on the heterogeneity parameter, broken down by HPV subgroup, age, and cytology outcome, are shown in Web Table 1 (available at http://aje.oxfordjournals.org/).
Table 4.
Estimates of the Heterogeneity Parameter σ (SD) Among 47,617 Women Who Underwent Cervical Screening, by Age and Cytological Outcome, New Mexico, 2007–2009
| Age Group, years |
Cytology Group |
P Valuea | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Normal | ASC-US | AGUS | LSIL | ASC-H | HSIL | Abnormal | All | ||
| ≤30 | 1.64 (0.02) | 0.41 (0.02) | 0.57 (0.10) | 0.14 (0.01) | 0.37 (0.06) | 0.17 (0.06) | 0.29 (0.01) | 1.40 (0.01) | 0.001 |
| 31–49 | 2.32 (0.03) | 1.28 (0.07) | 1.18 (0.25) | 0.22 (0.03) | 0.68 (0.11) | 0.24 (0.07) | 0.89 (0.04) | 2.45 (0.03) | 0.001 |
| ≥50 | 3.17 (0.06) | 2.04 (0.18) | 0.63 (0.24) | 0.20 (0.03) | 1.49 (0.35) | 0.07 (0.07) | 1.49 (0.10) | 3.42 (0.07) | 0.004 |
| All ages | 2.63 (0.02) | 0.82 (0.02) | 1.28 (0.14) | 0.17 (0.01) | 0.58 (0.07) | 0.21 (0.05) | 0.56 (0.01) | 2.50 (0.02) | 0.0005 |
Abbreviations: AGUS, atypical glandular cells of undetermined significance; ASC-H, atypical squamous cells—cannot exclude HSIL; ASC-US, atypical squamous cells of unknown significance; HSIL, high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion; SD, standard deviation.
a P for the difference in σ between normal and abnormal cytology groups.
Table 5.
Estimates of the Heterogeneity Parameter σ (SD) Among 47,617 Women Who Underwent Cervical Screening, by Cytological Outcome, New Mexico, 2007–2009a
| HPV Types | Cytology Group |
P Valueb | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Normal | ASC-US | AGUS | LSIL | ASC-H | HSIL | Abnormal | All | ||
| 35 types | 2.63 (0.02) | 0.82 (0.02) | 1.28 (0.14) | 0.17 (0.01) | 0.58 (0.07) | 0.21 (0.05) | 0.56 (0.01) | 2.50 (0.02) | 0.0005 |
| HR typesc | 3.07 (0.04) | 0.70 (0.03) | 0.65 (0.17) | 0.06 (0.02) | 0.23 (0.07) | σ < 0 | 0.40 (0.02) | 2.79 (0.03) | 0.001 |
| LR typesd | 2.60 (0.06) | 1.00 (0.08) | 3.25 (0.71) | 0.31 (0.04) | 0.63 (0.19) | 0.48 (0.22) | 0.78 (0.04) | 2.48 (0.05) | 0.002 |
| HR types/LR typese | 2.31 (0.05) | 0.75 (0.05) | 0.80 (0.38) | 0.08 (0.03) | 0.55 (0.14) | 0.28 (0.13) | 0.49 (0.03) | 2.26 (0.04) | 0.001 |
| α9 speciesf | 3.04 (0.08) | 0.41 (0.05) | 1.18 (0.53) | 0.12 (0.04) | 0.12 (0.09) | σ < 0 | 0.28 (0.03) | 2.63 (0.06) | 0.004 |
| 16/otherg | 2.27 (0.06) | 0.65 (0.06) | 0.84 (0.38) | 0.02 (0.04) | 0.01 (0.08) | σ < 0 | 0.30 (0.03) | 2.19 (0.05) | 0.003 |
| α7 speciesh | 2.01 (0.07) | 0.93 (0.07) | 0.33 (0.44) | 0.30 (0.06) | 0.76 (0.17) | σ < 0 | 0.68 (0.05) | 2.24 (0.06) | 0.003 |
| 18/otheri | 3.00 (0.11) | 0.92 (0.11) | 0.55 (0.41) | 0.08 (0.06) | 0.28 (0.22) | σ < 0 | 0.50 (0.06) | 2.83 (0.08) | 0.004 |
Abbreviations: AGUS, atypical glandular cells of undetermined significance; ASC-H, atypical squamous cells—cannot exclude HSIL; ASC-US, atypical squamous cells of unknown significance; HPV, human papillomavirus; HSIL, high-grade squamous intraepithelial lesion; LSIL, low-grade squamous intraepithelial lesion; SD, standard deviation.
a When multiple HPV types occurred less often than would be expected by chance, we denote this by σ < 0.
b P for the difference in σ between normal and abnormal cytology groups.
c HPV types 16, 18, 31, 33, 35, 39, 45, 51, 56, 58, and 59.
d HPV types 6, 11, 40, 42, 54, 61, 72, and 81.
e HR types versus LR types.
f HPV types 16, 31, 33, 35, 58, and 67.
g HPV 16 and all other types.
h HPV types 18, 39, 45, 59, 68, and 70.
i HPV type 18 and all other types.
To explore this further, we also examined specific HPV genotype pairs that exhibited strong departures from the model. Figure 2 and Web Table 2 show positive interactions with P < 10−4 and negative interactions with P < 10−3, which could be associated with type replacement. The strongest deviation was a positive interaction between types 56 and 66, found across all groups, mostly notably in younger women with normal cytology. Another consistent positive interaction was observed between HPV 51 and HPV 82. Cross-reaction between the probes used for detection of these HPV types (56/66 and 51/82) by the HPV LA assay has been suggested previously (12). Overdispersion among women with normal cytology was reduced from D = 2.0 to D = 1.5 with the exclusion of these 2 outliers.
Among the negative interactions in women with abnormal cytology, the strongest negative interactions (P < 10−4) were between HPV 16 and other HPV types (types 35, 51, 53, 56, and 70). Three pairs among the remaining 9 negative interactions also included HPV 16. No negative interactions between HPV 16 and other types were seen in women with normal cytology. A single significant negative interaction between HPV 72 and HPV 84 was found in women with normal cytology, but this observation may have been unreliable because of the rarity of these HPV types. We found no evidence of interaction between HPV 6 and HPV 16 (P > 0.05 for all cases) or between HPV 11 and HPV 16 (P > 0.03 for all cases) in either the normal or the abnormal cytology group or in any age group. Although the overdispersion is still statistically significant, the removal of the 2 excess event interactions and the 34 pairwise infections between HPV 16 and other HPV types reduced model overdispersion from 4.7 to 2.4 among women with abnormal cytology. Overdispersion was small across the 3 age groups given normal or abnormal cytology (D < 1.8), with the exception of younger women (≤30 years) with abnormal cytology (D = 3.9).
DISCUSSION
We investigated the extent to which multiple HPV concurrences of 35 HPV types varied among 47,617 women from New Mexico, according to age and cytology outcome. To summarize these data, we used a model in which the probability of infection with multiple HPV types is determined by age, cytology outcome, the type-specific prevalences, and an unobserved woman-specific random effect. This random effect is not HPV type-specific and is summarized by a single heterogeneity parameter. We used this heterogeneity parameter to explain the variability across women in the risk of having multiple HPV infections. Note that the frailty model used here was developed to capture excess risk for multiple infections of all multiplicities (summarized using the parameter σ), although we have shown illustrations for double infections only (Figure 2). The model accounts for an increased risk on an individual level (but not on a type-specific level), as opposed to a reduced risk of being infected with additional type(s). This approach yields greater statistical power to observe a negative interaction, by allowing for the expected excess risk to be positive overall.
Significant heterogeneity in the risk of having HPV multiple infections was observed among women of all ages and in all cytology groups. While multiple HPV types were most common in young women with low-grade cytology, the largest heterogeneity was found in older women with normal cytology. Although we are not aware of any specific study or survey, this is most easily explained by the assumption that sexual exposure is more uniform in younger women and more variable in older women, particularly as related to past and current HPV exposure. Another possible explanation for the variability in multiplicity is that there is greater variability in the risk that an HPV infection will become persistent in older women, possibly because of differences in the degree to which the immune system is capable of clearing infections in older women. Most of the infections found in older women are likely to be persistent (13, 14), due to their likelihood of being related to earlier exposure at a young age, when sexual activity is generally higher (3). If true, this could also be associated with variation in the reappearance of quiescent infection as women age, which has been seen in some populations (15–17).
More detailed information on HPV exposure, such as number of (recent) sexual partners, is not available in such a large population study with routinely collected data. Vaccination was very uncommon in this population (18) but will be important in future cohorts. In a future study, we intend to examine the relationship of multiple infections with age, abnormal cytological grade, and histological type in greater detail.
We also evaluated the heterogeneity parameter for important subsets of the 35 HPV types—high-risk, low-risk, α9 species, and α7 species. In particular, lower heterogeneity was observed for concurrences within the high-risk types, within types belonging to the α9 species, and between HPV 16 and other types than for all HPV types. This was most clearly seen in the abnormal cytology groups. One explanation is that the abnormal tissue is more often infected with only a single HPV type (19) and that this could dominate the viral load in samples so that infections with a lower copy number might not be readily detected.
When looking for pairs not fitting the model, we found positive interactions between HPV 56 and HPV 66 and between HPV 51 and HPV 82, which are phylogenetically closely related—an observation resulting from sequence similarities and cross-reactive probes. Of greater interest were the negative interactions of HPV 16 with 8 other types, including the high-risk types 35, 51, 56, and 58. This raises the possibility of type replacement after vaccination against HPV 16, whereby the removal of HPV 16 by vaccination might more easily enable other high-risk HPV infections to become persistent. However, this was seen only in women with abnormal cytology (including the low-grade cytological groups ASC-US and LSIL), despite adequate statistical power in those with normal cytology (>99%). It is possible that when abnormal cytology is present, the greater transforming ability of HPV 16 results in much higher genome copy numbers than for other concurrent types. The relatively higher abundance of HPV 16 compared with other types present could result in underdetection of low-copy HPV coinfections, especially when amplification primers are shared by HPV 16 and the other coinfecting HPV types. This could also explain the finding of σ < 0 seen in the HSIL group, which suggests that our model may be inadequate when these other mechanisms are involved. Of note was the lack of negative interaction of HPV 16 with the low-risk types HPV 6 and HPV 11, which has been seen in 1 previous large study using serological analysis (8). The lack of negative interactions in women with normal cytology, which is more representative of what may be observed for the general population, suggests that type replacement is less likely to occur after vaccination.
In summary, with a few exceptions of positive interactions due to technical issues, we felt that the model fitted the data well with minimal type-specific interactions, although interactions of HVP 16 with other types were seen in women with abnormal cytology.
Supplementary Material
ACKNOWLEDGMENTS
Author affiliations: Wolfson Institute, Queen Mary University of London, London, United Kingdom (Zihua Yang, Jack Cuzick); and Department of Pathology, University of New Mexico Health Sciences Center, Albuquerque, New Mexico (William C. Hunt, Cosette M. Wheeler).
This work was funded by a grant to C.M.W. from the National Institute of Allergy and Infectious Diseases (grant U19AI084081) and a program grant to J.C. from Cancer Research UK (grant A-10404).
Reagents for the LINEAR ARRAY HPV Genotyping Test and equipment for automating the HPV genotyping assays were provided by Roche Molecular Systems, Inc. (Pleasanton, California).
The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Z.Y. and W.C.H. take responsibility for the integrity of the data and the accuracy of the data analysis.
C.M.W. has received funding for HPV vaccine studies through her institution, the University of New Mexico; has had travel costs covered by GlaxoSmithKline (Brentford, United Kingdom) and Merck and Company, Inc. (Whitehouse Station, New Jersey); and has received equipment and reagents for HPV genotyping from Roche Molecular Systems, Inc. J.C. reports advisory board involvement, study grants, and honoraria from Qiagen, Inc. (Valencia, California); Hologic Gen-Probe (San Diego, California); Roche Molecular Systems; Becton, Dickinson and Company (Franklin Close, New Jersey); Genera Biosystems Ltd. (Scoresby, Victoria, Australia); OncoHealth Corporation (Mountain View, California); and Abbott Laboratories (Abbott Park, Illinois).
REFERENCES
- 1.Kong X, Gray RH, Moulton LH, et al. A modeling framework for the analysis of HPV incidence and persistence: a semi-parametric approach for clustered binary longitudinal data analysis. Stat Med. 2010;29(28):2880–2889. doi: 10.1002/sim.4062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Maucort-Boulch D, Plummer M, Castle PE, et al. Predictors of human papillomavirus persistence among women with equivocal or mildly abnormal cytology. Int J Cancer. 2010;126(3):684–691. doi: 10.1002/ijc.24752. [DOI] [PubMed] [Google Scholar]
- 3.Plummer M, Schiffman M, Castle PE, et al. A 2-year prospective study of human papillomavirus persistence among women with a cytological diagnosis of atypical squamous cells of undetermined significance or low-grade squamous intraepithelial lesion. J Infect Dis. 2007;195(11):1582–1589. doi: 10.1086/516784. [DOI] [PubMed] [Google Scholar]
- 4.Merikukka M, Kaasila M, Namujju PB, et al. Differences in incidence and co-occurrence of vaccine and nonvaccine human papillomavirus types in Finnish population before human papillomavirus mass vaccination suggest competitive advantage for HPV33. Int J Cancer. 2011;128(5):1114–1119. doi: 10.1002/ijc.25675. [DOI] [PubMed] [Google Scholar]
- 5.Trottier H, Mahmud S, Prado JCM, et al. Type-specific duration of human papillomavirus infection: implications for human papillomavirus screening and vaccination. J Infect Dis. 2008;197(10):1436–1447. doi: 10.1086/587698. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Vaccarella S, Franceschi S, Snijders PJ, et al. Concurrent infection with multiple human papillomavirus types: pooled analysis of the IARC HPV Prevalence Surveys. Cancer Epidemiol Biomarkers Prev. 2010;19(2):503–510. doi: 10.1158/1055-9965.EPI-09-0983. [DOI] [PubMed] [Google Scholar]
- 7.Chaturvedi AK, Katki HA, Hildesheim A, et al. Human papillomavirus infection with multiple types: pattern of coinfection and risk of cervical disease. J Infect Dis. 2011;203(7):910–920. doi: 10.1093/infdis/jiq139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Luostarinen T, af Geijersstam V, Bjørge T, et al. No excess risk of cervical carcinoma among women seropositive for both HPV16 and HPV6/11. Int J Cancer. 1999;80(6):818–822. doi: 10.1002/(sici)1097-0215(19990315)80:6<818::aid-ijc4>3.0.co;2-t. [DOI] [PubMed] [Google Scholar]
- 9.Cuzick J, Yang Z. A frailty model for interaction between multiple events. J Multivar Anal. 2013;122:133–147. doi: 10.1016/j.jmva.2013.07.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Solomon D, Davey D, Kurman R, et al. The 2001 Bethesda System: terminology for reporting results of cervical cytology. JAMA. 2002;287(16):2114–2119. doi: 10.1001/jama.287.16.2114. [DOI] [PubMed] [Google Scholar]
- 11.Wheeler CM, Hunt WC, Cuzick J, et al. A population-based study of human papillomavirus genotype prevalence in the United States: baseline measures prior to mass human papillomavirus vaccination. Int J Cancer. 2013;132(1):198–207. doi: 10.1002/ijc.27608. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wentzensen N, Schiffman M, Dunn ST, et al. Multiple human papillomavirus genotype infections in cervical cancer progression in the Study to Understand Cervical Cancer Early Endpoints and Determinants (SUCCEED) Int J Cancer. 2009;125(9):2151–2158. doi: 10.1002/ijc.24528. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Meijer CJ, Snijders PJ, Castle PE. Clinical utility of HPV genotyping. Gynecol Oncol. 2006;103(1):12–17. doi: 10.1016/j.ygyno.2006.07.031. [DOI] [PubMed] [Google Scholar]
- 14.Castle PE, Schiffman M, Herrero R, et al. A prospective study of age trends in cervical human papillomavirus acquisition and persistence in Guanacaste, Costa Rica. J Infect Dis. 2005;191(11):1808–1816. doi: 10.1086/428779. [DOI] [PubMed] [Google Scholar]
- 15.González P, Hildesheim A, Rodríguez AC, et al. Behavioral/lifestyle and immunologic factors associated with HPV infection among women older than 45 years. Cancer Epidemiol Biomarkers Prev. 2010;19(12):3044–3054. doi: 10.1158/1055-9965.EPI-10-0645. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rodríguez A-C, Schiffman M, Herrero R, et al. Low risk of type-specific carcinogenic HPV re-appearance with subsequent cervical intraepithelial neoplasia grade 2/3. Int J Cancer. 2012;131(8):1874–1881. doi: 10.1002/ijc.27418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Rositch AF, Burke AE, Viscidi RP, et al. Contributions of recent and past sexual partnerships on incident human papillomavirus detection: acquisition and reactivation in older women. Cancer Res. 2012;72(23):6183–6190. doi: 10.1158/0008-5472.CAN-12-2635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cuzick J, Myers O, Hunt WC, et al. A population-based evaluation of cervical screening in the United States: 2008–2011. Cancer Epidemiol Biomarkers Prev. 2014;23(5):765–773. doi: 10.1158/1055-9965.EPI-13-0973. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Quint W, Jenkins D, Molijn A, et al. One virus, one lesion—individual components of CIN lesions contain a specific HPV type. J Pathol. 2012;227(1):62–71. doi: 10.1002/path.3970. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.