Reduction in Human Papillomavirus Vaccine Type Prevalence Among Young Women Screened for Cervical Cancer in an Integrated US Healthcare Delivery System in 2007 and 2012–2013

Abstract

Human papillomavirus (HPV) vaccine was first introduced in the United States in 2006 for girls and women, and in 2009 for boys and men. HPV vaccine is recommended for routine vaccination of girls and boys aged 11 or 12 years [1]. Vaccination is also recommended for women through age 26 years and men through age 21 years who were not vaccinated at the routine age [1]. There are now 3 HPV vaccines available; the bivalent HPV vaccine (targeting HPV types 16 and 18), the quadrivalent HPV vaccine (targeting HPV types 6, 11, 16, and 18) and the 9-valent HPV vaccine (targeting HPV types 6, 11, 16, 18, 31, 33, 45, 52, and 58) [2]. HPV vaccines have high efficacy for prevention of HPV-16/18–associated precancers [1].

Demonstrating population effectiveness in prevention of HPV-associated cancers will support the vaccination program; however, because infection occurs most often in young adults and cancer develops years later, the impact of vaccines on cancers may take a decade or more to observe. Early vaccine impact, such as reduction in HPV vaccine-type prevalence (VT HPV), is important for vaccine programming and policy [2]. A few studies in settings with different HPV vaccine coverage have already demonstrated reductions in VT HPV [3]. In a nationally representative population from the United States, there was a 56% reduction in VT HPV among 14–19-year-olds in 2007–2010 compared with the prevaccine era; no significant reductions were noted in other age groups [4]. US clinic-based studies in high-risk teens and young adults have also found significant reductions in VT HPV prevalence [5, 6]. Few studies have focused on vaccine impact in women aged 20–29 years, but assessing HPV vaccine impact in young women is important for monitoring long-term protection and vaccine impact.

In this study, we compared HPV prevalence in young women screened for cervical cancer at Kaiser Permanente Northwest at age 20–29 years in 2007, before widespread vaccine introduction, and in 2012–2013, 6 years after national vaccine introduction. During this time frame, all HPV vaccine delivered in this system was quadrivalent HPV vaccine. We also evaluated predictors of VT HPV and any HPV in the vaccine era.

METHODS

Data collection methods are described in detail in a previous assessment of prevaccine era HPV prevalence in 2 integrated healthcare delivery systems [7]. We used existing samples collected from screened women for this assessment. Residual SurePath (TriPath Imaging) liquid cytology cervical specimens from women aged 20–29 years undergoing routine cervical cancer screening at Kaiser Permanente Northwest were retained consecutively until target numbers were obtained in each age group (20–24 and 25–29 years). Sample size calculations indicated that 2000 specimens in each of the 2 age groups would be needed to detect a 40% reduction in HPV-16 and/or HPV-18 with 80% power.

Specimens were collected from May through December 2007 from unvaccinated women and from June through August 2012 and May through August 2013 from vaccinated and unvaccinated women. Retrieved specimens were stored for <37 days at ambient temperature and shipped to the Centers for Disease Control and Prevention (CDC). In 2007, data on age group and cervical cytology results were obtained and linked to each specimen. In 2012 and 2013, additional data on vaccination (dates, doses), demographics, and human immunodeficiency virus (HIV), chlamydia, and pregnancy testing were retained and linked to each specimen. Data on patient characteristics, such as race and ethnicity, were obtained from administrative and electronic health records. Family poverty was calculated based on geocoded census tract information from household income. Information about HPV vaccination was extracted from the electronic medical record. This study was approved by the institutional review board at Kaiser Permanente Northwest.

Laboratory Methods

SurePath specimens (approximately 3 mL) were stored at 4°C when received at the CDC, until testing. Those received in 2007 were processed and extracted using ABI Prism 6100 Nucleic Acid Prep Station and NucPrep reagents (Applied Biosystems), as described elsewhere [7]. In 2012, the DNA extraction method was changed to an automated magnetic bead system, Chemagic MSM1 (Chemagen), after verifying that DNA yields were the same or better than with the 2007 method. After 1 mL of the specimen was centrifuged for 5 minutes at 11 000g, medium was removed from the pellet. Cells were resuspended in 200 µL of nuclease-free water, 110 µL of lysis buffer EL1, and 20 µL of proteinase K from the Pathogen NA Extension Kit (Perkin Elmer Chemagen Technology). After 1-hour incubation at 70°C, the lysed emulsion was mixed with 340 µL of lysis buffer EL2 and incubated for another 15 minutes. The lysate was processed with a Chemagic MSM1 programmed for SurePath specimens using buffers from the Viral NA/gDNA kit (Perkin Elmer). The purified DNA was eluted in a final volume of 100 µL. Four water blanks were processed with each set of 96 samples to control for cross-contamination.

HPV genotypes were determined using the Research Use Only Linear Array (LA) HPV Genotyping Test (Roche Molecular Diagnostics) and HPV-52 quantitative polymerase chain reaction, as described elsewhere [8]. Specimens that were HPV negative and failed to amplify the cellular positive control (β-globin) were considered inadequate and excluded from the study.

We compared prevalence for any HPV, VT HPV (6, 11, 16, and/or 18), high-risk (HR) HPV, low-risk (LR) HPV, HR VT HPV (16 and/or 18), LR VT HPV (6, and/or 11), and HR non-VT HPV in the prevaccine and vaccine era. Any HPV included ≥1 of 37 HPV types. The HR non-VT included 12 clinically relevant types: types 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68 [9]. We also evaluated the 5 additional HPV types targeted by the 9-valent HPV vaccine (types 31, 33, 45, 52, and 58) and 3 types with some evidence of cross-protection in some vaccine efficacy studies (types 31, 33, and 45) [10].

Statistical Methods

To compare the HPV prevalence between 2007 and 2012–2013, we calculated prevalence ratios (PRs) and 95% confidence intervals (CIs) assuming binomial distribution, and we produced the corresponding P values using χ² tests. To identify the characteristics associated with VT HPV and any HPV in the vaccine era, we performed bivariate analysis of characteristics and calculated associated unadjusted PRs, their 95% CIs, and P values assuming binomial distribution. Factors associated with VT HPV and any HPV in bivariate analysis were included in 2 separate multivariable log-binomial regression models to obtain the adjusted PRs (aPRs). Only vaccine doses in which the vaccination was administered ≥30 days before specimen collection were included as a vaccine dose.

We created a variable for recent sexually transmitted infection and pregnancy testing (categorical variable) by using electronic laboratory data for Chlamydia trachomatis and HIV testing in the year preceding specimen collection, and pregnancy testing in the 6 months preceding specimen collection. Interactions between factors were evaluated, and an interaction between number of doses and age at HPV vaccine initiation was observed. Because of this interaction and because there were few persons with only 1 or only 2 doses, the multivariable analysis of VT HPV excluded female subjects with 1 or 2 vaccine doses. All other analysis included all study subjects. The analyses were conducted using SAS software, version 9.3.

RESULTS

Overall, 4138 adequate specimens were available in 2007 (prevaccine era), 2059 in 2012, and 2112 in 2013 (vaccine era). The number of inadequate samples was <0.5% in both time periods (0.3% in 2007 and 0.2% in 2012–2013).There were no significant differences in the demographic characteristics of women between 2012 and 2013 so the data for vaccine-era years were combined.

Characteristics of women were similar in 2007 and 2012–2013 by age group and cytologic findings (data not shown). About half the study population was in each age category (20–24 and 25–29 years) in both study periods. Most Papanicolaou test results were normal in 2007 and in 2012–2013 (92.1% and 92.7%, respectively). Few specimens (n = 19) were missing cytology results.

The prevalence of VT HPV decreased from 10.6% in 2007 to 6.2% in 2012–2013 (PR, 0.6; (95% CI, .5–.7) (Table 1). There were also significant differences for both HR VT HPV (HPV types 16 and/or 18) and LR VT HPV (HPV types 6 and/or 11). Prevalence of any HPV type, any HR HPV type, any LR HPV type, and HR non-VT HPV was significantly higher in the vaccine era than in the prevaccine era (Table 1). No differences in prevalence were observed for HPV types 31, 33, 45, 52, or 58 or types 31, 33, or 45.

In the vaccine era, 31.9% of women in our study had initiated HPV vaccination; 6.5% received only 1 dose, 4.8% received only 2 doses, and 20.7% received all 3 doses (Table 2). Only 1.8% of women received their first dose at <15 years of age; most received their first dose at age 15–18 years (51.2%). Most women were white (71%) and had unknown Hispanic ethnicity (76.9%). In the past year, 14.1% had been tested for chlamydia and 13.8% for HIV; 15.9% had had a pregnancy test in the past 6 months.

In bivariate analysis among all vaccine-era subjects, vaccine initiation was significantly associated with lower VT HPV, compared with no vaccination, as was age <19 years at vaccination (Table 3). In multivariable analysis excluding vaccinated women with only 1 or 2 vaccine doses, factors associated with VT HPV were receipt of the first dose of the 3-dose series before age 19 years compared with no vaccination (aPR, 0.1; 95% CI, .1–.3), and having had a chlamydia, HIV, or pregnancy test recently compared with no test (aPR 1.4; 95% CI, 1.1–1.8) (Table 4). Factors significantly associated with any HPV in the vaccine era in both bivariate and multivariable analyses of all women were vaccine initiation at age ≥19 years compared with no vaccination (aPR, 1.2; 95% CI, 1.1–1.4), age group 20–24 years compared with 25–29 years (aPR, 1.3; 95% CI, 1.1–1.4), and any chlamydia, HIV, or pregnancy test in the past 6 months to 1 year compared with no test (aPR, 1.2; 95% CI, 1.1–1.3) (data not shown).

DISCUSSION

Our study demonstrated a 40% reduction in VT HPV prevalence in the vaccine era compared with the prevaccine era in women aged 20–29 years screened for cervical cancer in the northwest United States. In our population, 32% of women from the vaccine era had initiated the vaccination series; two-thirds of them received the complete 3-dose series. This is similar to what was demonstrated in national surveys, with 36.9% of women aged 19–26 years reporting receipt of ≥1 dose of HPV vaccine in 2013 [11]. In the vaccine era, HPV vaccine initiation before 19 years of age was associated with decreased VT HPV prevalence.

A number of studies in the United States and globally have demonstrated reduction in VT HPV prevalence since HPV vaccine introduction [3]; these settings may use different vaccines and target age groups. A representative survey of the US population demonstrated a 56% reduction in VT HPV in girls and women aged 14–19 years; however, this study did not find reductions in older groups, probably because it was conducted early during vaccine implementation [4]. Two clinic-based studies of high-risk teens and young adults in the United States also found reductions in VT HPV prevalence [5, 6]. One included girls and young women aged 13–26 years from 2 primary care clinics in the United States; in this study, VT HPV decreased from 31.7% to 13.4% [5]. Another study of young women from an urban setting in Indiana found that VT HPV decreased from 24% to 5.1% [6].

Studies outside the United States have also demonstrated decreases; a study from the United Kingdom that used specimens from the national chlamydia screening program found a difference in vaccine era and prevaccine era HR VT HPV among girls and women aged 16–20 years but not in those aged 22–24 years [12]. In Australia, where high vaccine coverage was achieved in the routine target age group soon after the beginning of the vaccine program in 2006, and with moderately high coverage also in the catch-up population, a repeat cross-sectional study comparing the prevaccine with the vaccine era found a significant decrease in VT HPV prevalence from 29% to 7% in 18–24-year-old women [13]. Also, although not seen in the comparison between prevaccine and vaccine eras, within the vaccine era, vaccinated women had a lower prevalence of some non-VT HPV types (types 31, 33, and 45) [13].

In addition to reductions in VT HPV prevalence, our study also found increases in any HPV, any HR HPV, any LR HPV, and any HR non-VT prevalence between the prevaccine and vaccine eras. We did not find any differences in non-VT HPV types 31, 33, 45, 52, and 58. Although our study was conducted in the same healthcare institution in 2007 and 2012–2013, and our sampling methods were the same (continuous sampling of 20–29-year-olds until sample sizes were attained in each age group), various factors may have contributed to these changes over time. There were some differences in laboratory processing of specimens in 2012–2013, including a more consistent interval from collection to freezing of specimens, and changes in the extraction methods to improve sensitivity of HPV detection in the laboratory. Differences in sexual behavior in the population may also have accounted for changes in HPV prevalence between 2007 and 2012–2013. Data from a national survey showed an increase in number of lifetime and past-year sex partners in women in this age group between these time periods (CDC unpublished data). As noted, a proxy for sexual behavior, testing for HIV, chlamydia, or pregnancy, was an independent predictor of both VT HPV and any HPV in the vaccine era. Even with these increases in any-HPV prevalence, there were significant reductions in VT HPV prevalence.

Our assessment has several limitations. For one, our population was women screened for cervical cancer and although cervical cancer screening adherence is high within Kaiser Permanente Northwest, this population may not be representative of all women enrolled in the integrated healthcare delivery system. Furthermore, cervical cancer screening recommendations changed during the study period; screening is currently recommended starting at age 21 years, but our prevaccine- and vaccine-era age groups included 20-year-olds. However, when we removed 20-year-olds from the vaccine era, the VT HPV prevalence estimates did not significantly change. We also did not collect sexual behavior information and therefore were unable to assess this in the analysis; however, we did assess a proxy of sexual behavior in the vaccine era by using HIV, chlamydia, and pregnancy testing. Also our sample size was insufficient to evaluate HPV type prevalence by vaccine dose, but we intend to evaluate this after more years of data collection. Finally, vaccine receipt was determined from the electronic medical record, and vaccination status may not always have been accurate. Although we found a reduction in VT HPV prevalence among women aged 20–29 years, this was an early assessment because the vaccine was available for girls and women only since 2006, and fewer women aged 20–29 years would have been vaccinated. Our assessment of VT HPV will continue in this population and vaccination uptake will likely increase, providing a better opportunity to assess vaccine impact.

In summary, in this early assessment of an insured population of 20–29-year-old women in the northwest United States, we found reductions in VT HPV prevalence indicating HPV vaccine impact. An ongoing assessment in this population will be important to evaluate reductions in VT HPV by vaccine dose. The reduction in VT HPV was notable in those vaccinated before 19 years of age, which adds to the body of evidence supporting vaccination in early adolescence before HPV exposure.

Notes

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC).

Financial support. The project was supported by funds from the Division of STD Prevention, CDC.

Potential conflicts of interest. All authors: No potential conflicts of interest.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References

Author notes