Introduction

Human papillomavirus (HPV) is the most common sexually transmitted infection (STI) in the United States. Persistent cervical infection with a high-risk type of HPV is necessary for the development of cervical cancer. Although HPV vaccination is the only known intervention to prevent HPV-associated precancerous lesions, it has been hypothesized that intrauterine device (IUD) use may alter the natural history of HPV infections, preempting development of precancerous lesions of the cervix and cervical cancer. In fact, a 2011 meta-analysis of 26 studies from around the world demonstrated a statistically significant decrease in cervical cancer associated with IUD use; women who reported ever using an IUD had approximately half the likelihood of being diagnosed with cervical cancer compared with never users after adjusting for lifetime number of screening pap tests. The interpretation of this study is limited since it was cross-sectional and it did not correct for many possible confounders including some sexual and health behaviors. For instance, more intensive screening, diagnosis, and treatment among IUD users compared with nonusers or differential sexual behavior may still have resulted in differing exposure to HPV between IUD users and nonusers.

However, on a biological level, it is plausible that IUDs may protect against cervical cancer through their association with inflammation in the genital tract, which may result in rapid clearance of infections and/or protection against initial HPV acquisition. Some studies have shown that cytokine levels are elevated in the setting of persistent HPV infections suggesting that recruitment of immune mediators is necessary for clearance of the virus. Conversely, other studies have shown that chronic inflammation in the setting of HPV promotes development of high-risk premalignant lesions. The effect of inflammation in the genital tract on HPV is complex and there is a paucity of data on the specific effect of IUDs on cervical HPV infection.

The goal of our study was to evaluate the association between IUD use and high-risk HPV acquisition and clearance in a large prospective cohort.

Materials and Methods

This is an analysis of the San Francisco Natural History of HPV Cohort described elsewhere in detail. Briefly, from October 2000 through October 2006, 676 women were enrolled from 2 family planning clinics in San Francisco, CA. Women and girls ages 13-22 years with <5 years of sexual experience were included. Those who were immunosuppressed or had a history of ablation or excision of the cervix were excluded. Participants were followed up until study close in June 2014. Each visit included a speculum examination with collection of cervicovaginal lavages for high-risk HPV testing, and an interviewer-administered questionnaire to obtain information on demographics and sexual behaviors. Samples were obtained yearly, or sooner if a study subject was symptomatic, to test for Chlamydia trachomatis and Neisseria gonorrhoeae. HPV testing was done by linear array (Roche Molecular Diagnostics, Alameda, CA) and a novel Luminex-based approach, PGMY-LX, which was shown to have a sensitivity and specificity comparable to linear array in this population and is described elsewhere in detail. Sexual and substance use behavior and contraceptive use, including IUD use, were assessed by interview-administered questionnaire. IUD use was verified at the time of speculum examination but the type of IUD strings observed were not recorded.

Incident high-risk HPV infection was defined as the first visit where high-risk HPV DNA was detected following at least 2 preceding visits in which high-risk HPV DNA was not detected. Multiple episodes of acquisition were not allowed for this analysis. Time to high-risk HPV acquisition was defined as the time from first visit with no high-risk HPV DNA detection to first high-risk HPV detection.

HPV clearance was defined as having 2 visits negative for all high-risk HPV types following a visit with a positive high-risk HPV type. Time to high-risk HPV clearance was defined as the time from first high-risk HPV DNA detection to first of 2 negative visits. Prevalent infections detected at baseline were included in the clearance analysis but not the acquisition analysis.

Initial analyses focused on descriptive comparisons of baseline characteristics between women classified as ever users and nonusers of IUDs (or women who did not use an IUD ever during study follow-up). Significance testing was based on the χ ² /Fisher exact test for categorical and the t test/Wilcoxon rank sum test for continuous characteristics.

We considered 2 complementary approaches to investigating the association between IUD use and HPV outcomes: The first was based on a new user design. A new user design identifies all of the participants in a population who start a course of treatment and study follow-up begins at the same time as initiation of the treatment. The goal of this approach is to mimic trial assignment. The rationale is to attempt to balance possibly time-varying confounding factors between users and nonusers, similar to what might be achieved in a randomized trial. Results from observational data analyzed using a new users design were shown to be similar to data from randomized trials. We matched each IUD user to 5 women who were not using IUDs based on user’s age and date of IUD initiation. IUD initiation date was estimated within 3 months. IUD users contribute data beginning at the time they initiated their IUD (or on entry into the study if they had an IUD at baseline).

The outcomes were clearance for those who were HPV positive and acquisition for those who were HPV negative following initiation of the IUD (or start date for matched control). Between-group comparisons of cumulative outcome risks were based on Kaplan-Meier estimates and log-rank tests.

Because the new user approach uses only a subset of available data, and does not allow explicit control for time variation in IUD use and other variables potentially related to HPV outcomes, we also analyzed data from the entire cohort using a Cox proportional hazards model, including time-varying indicators of IUD use, and adjusting for fixed and time-dependent predictor variables including age at baseline, age at first intercourse, STI during study follow-up, HPV vaccination status, condom use since last visit, new partner since the last visit, combined hormonal contraceptive use (pill, patch, and ring), smoking since last visit, and pregnancy.

Statistical analysis was performed using software (SAS, Version 9.4; SAS Institute Inc, Cary, NC).

Participants gave written consent according to guidelines approved by the Committee for Human Research, University of California, San Francisco ( ClinicalTrials.gov NCT01366742 ).

Materials and Methods

This is an analysis of the San Francisco Natural History of HPV Cohort described elsewhere in detail. Briefly, from October 2000 through October 2006, 676 women were enrolled from 2 family planning clinics in San Francisco, CA. Women and girls ages 13-22 years with <5 years of sexual experience were included. Those who were immunosuppressed or had a history of ablation or excision of the cervix were excluded. Participants were followed up until study close in June 2014. Each visit included a speculum examination with collection of cervicovaginal lavages for high-risk HPV testing, and an interviewer-administered questionnaire to obtain information on demographics and sexual behaviors. Samples were obtained yearly, or sooner if a study subject was symptomatic, to test for Chlamydia trachomatis and Neisseria gonorrhoeae. HPV testing was done by linear array (Roche Molecular Diagnostics, Alameda, CA) and a novel Luminex-based approach, PGMY-LX, which was shown to have a sensitivity and specificity comparable to linear array in this population and is described elsewhere in detail. Sexual and substance use behavior and contraceptive use, including IUD use, were assessed by interview-administered questionnaire. IUD use was verified at the time of speculum examination but the type of IUD strings observed were not recorded.

Incident high-risk HPV infection was defined as the first visit where high-risk HPV DNA was detected following at least 2 preceding visits in which high-risk HPV DNA was not detected. Multiple episodes of acquisition were not allowed for this analysis. Time to high-risk HPV acquisition was defined as the time from first visit with no high-risk HPV DNA detection to first high-risk HPV detection.

HPV clearance was defined as having 2 visits negative for all high-risk HPV types following a visit with a positive high-risk HPV type. Time to high-risk HPV clearance was defined as the time from first high-risk HPV DNA detection to first of 2 negative visits. Prevalent infections detected at baseline were included in the clearance analysis but not the acquisition analysis.

Initial analyses focused on descriptive comparisons of baseline characteristics between women classified as ever users and nonusers of IUDs (or women who did not use an IUD ever during study follow-up). Significance testing was based on the χ ² /Fisher exact test for categorical and the t test/Wilcoxon rank sum test for continuous characteristics.

We considered 2 complementary approaches to investigating the association between IUD use and HPV outcomes: The first was based on a new user design. A new user design identifies all of the participants in a population who start a course of treatment and study follow-up begins at the same time as initiation of the treatment. The goal of this approach is to mimic trial assignment. The rationale is to attempt to balance possibly time-varying confounding factors between users and nonusers, similar to what might be achieved in a randomized trial. Results from observational data analyzed using a new users design were shown to be similar to data from randomized trials. We matched each IUD user to 5 women who were not using IUDs based on user’s age and date of IUD initiation. IUD initiation date was estimated within 3 months. IUD users contribute data beginning at the time they initiated their IUD (or on entry into the study if they had an IUD at baseline).

The outcomes were clearance for those who were HPV positive and acquisition for those who were HPV negative following initiation of the IUD (or start date for matched control). Between-group comparisons of cumulative outcome risks were based on Kaplan-Meier estimates and log-rank tests.

Because the new user approach uses only a subset of available data, and does not allow explicit control for time variation in IUD use and other variables potentially related to HPV outcomes, we also analyzed data from the entire cohort using a Cox proportional hazards model, including time-varying indicators of IUD use, and adjusting for fixed and time-dependent predictor variables including age at baseline, age at first intercourse, STI during study follow-up, HPV vaccination status, condom use since last visit, new partner since the last visit, combined hormonal contraceptive use (pill, patch, and ring), smoking since last visit, and pregnancy.

Statistical analysis was performed using software (SAS, Version 9.4; SAS Institute Inc, Cary, NC).

Participants gave written consent according to guidelines approved by the Committee for Human Research, University of California, San Francisco ( ClinicalTrials.gov NCT01366742 ).