The study was approved by the Committee on Human Subjects at the University of California, San Francisco. All participants provided written informed consent; because minors can consent to contraceptive services in California and because the study posed minimal risk, parental consent for participation for minors was waived. The study was conducted from September 2005 to July 2008 at four Planned Parenthood clinics in the Northern California: Vallejo, Richmond, East Oakland, and Hayward. We enrolled consecutive women who presented for reproductive health care at the study clinics and selected to initiate a hormonal contraceptive after they received standard clinical care from clinic providers. Women were eligible for inclusion in the study if they were aged 15 to 24, not married, could read English or Spanish, were not pregnant (self-report) or were not desiring pregnancy within the next year, and were able to provide written informed consent and comply with study procedures. Women had to have not used the contraceptive method selected in the past (“initiators”); however, they may have used another hormonal contraceptive.
Research staff collected data from enrolled participants via self-administered electronic questionnaires at baseline and at 3, 6, and 12 months after baseline. Questionnaires were pilot-tested to ensure that the instrument was at an appropriate reading level and understood by respondents. The questionnaires, which were available in English and Spanish, were accessed by web link on laptop computers and were administered either at the clinic site or at a location convenient for the participant or by telephone. Study participants received $30 for completing the baseline and 12-month follow-up questionnaires and $20 for completing the 3- and 6-month follow-up questionnaires.
Sociodemographic and reproductive health measures collected at baseline were from previously validated questionnaires and included income, race and ethnicity, education, and employment status.17-20
Reproductive health measures included information on sexual partners, hormonal contraceptive use, and history of pregnancy.
The selection of other baseline study measures was guided by the theory of reasoned action, which posits that the best predictor of a given behavior is intention to engage in the behavior.21
Intention is influenced by two factors: 1) the individual’s attitudes toward taking the action, which reflects one’s beliefs and values about consequences of engaging in the behavior; and 2) their view of social expectations related to the behavior. Participants were asked the question, “how sure are you that you will use the baseline method selected for 1 year?” to gauge intent to use contraception (answer choices: very sure, sure, somewhat sure, and not at all sure).6
To assess pregnancy desire, participants were asked when they wanted to have a child (answer choices: never, in the next year, in the next 1–2 years, in more than 2 years, and do not know) and how they would feel if they got pregnant in the next 3 months (answer choices: very upset, somewhat upset, I would not care, somewhat pleased, very pleased, and do not know).
To understand consequences related to method use, we explored perceptions of side effects of combined hormonal birth control methods using questions developed from information obtained in focus groups conducted in the formative stages of the study.22,23
Study participants were asked to rank how much they agreed or disagreed with 13 negative statements about attributes of combined hormonal contraceptive methods, such as “it is dirty to put the ring in the vagina.” A Likert scale was used (strongly agree, agree, neither agree nor disagree, disagree, strongly disagree) and a negative attributes scale adding scores from the 13 items was developed, with a higher score indicating more negative perceptions of combined hormonal methods. The score had fair internal consistency (Cronbach α
=.62). To assess perceptions of social expectations, participants also were asked questions about how their partners and peers felt about their use of various hormonal contraceptive methods and if they thought their parents would be upset if they knew they used birth control.
The follow up questionnaires focused on the main study outcomes: contraceptive use and pregnancies over the 12-month follow-up period. Participants were asked if they started the method selected, if they were using it currently, and whether they used the method continuously (without breaks). If participants discontinued the method, then they were asked how long they used it, the main reason for discontinuation, subsequent methods used (switching), and birth control method used during last sexual encounter. When use of another method was reported, use of any hormonal method or an intrauterine device was considered use of an effective method. Information on pregnancies was collected via self-report at each follow-up, urine pregnancy tests were performed at 6 and 12 months during in-person follow-up visits, and medical record chart review was performed at the study sites at the end of the 12-month follow-up period.
Using data on contraceptive continuation from the 1995 National Survey of Family Growth and clinical trials of the patch and the ring, anticipated continuation rates were 50% for pill users and 60% for patch, ring, and depot medroxyprogesterone acetate. It was estimated that a sample size of 375 pill, patch, ring, and depot medroxyprogesterone acetate initiators would allow us to detect a significant difference in continuation rates at 12 months with a one-tailed test. Calculations were based on survival analysis with comparison of two groups with α
of .05, β
of 80%, and 20% attrition rate at 1 year.11-13,24
To allow for adequate comparisons, participants initiating any of the four methods were enrolled until a sample of 400 women initiating pills was reached. Enrollment of pill initiators was then discontinued while continuing to enroll women initiating the other three methods.
Women who completed at least one follow-up are included in the analyses; for those lost to follow-up during the study period, data are included up until the point when they were lost to follow-up (ie, censored). To compare women with no follow-up data to women contributing study data, we completed an attrition analysis. We examined characteristics of the study population by contraceptive method selected at baseline comparing the patch, ring, and depot medroxyprogesterone acetate initiators to pill initiators using χ2 statistics for categorical variables and F tests from one-way analysis of variance for continuous variables. We plotted 12-month contraceptive continuation by method with Kaplan-Meier estimates of the probability of discontinuation. Life table analysis was used to estimate survival rates for contraceptive continuation. The log-rank test was used to compare survival rates for different methods. Comparisons of main reasons reported for contraceptive discontinuation by method were performed using χ2 analysis.
To assess factors associated with method discontinuation, we estimated Cox proportional hazards models with contraceptive method and covariates from the baseline data, including sociodemographic characteristics, reproductive history, pregnancy desire, and attitudes toward contraception.25
For the survival analysis, we estimated time to the event (contraceptive discontinuation or first pregnancy). Each woman contributed observation time to the analysis for the period in which she participated. Individuals were censored when lost to follow-up, when the event occurred, or when they exited the study. Unadjusted and adjusted hazards ratios and 95% confidence intervals were estimated. Likelihood ratio tests were used to estimate the final simplest model. Missing values are not included in the analyses. To check the assumptions of the proportional hazards models, we estimated the Schoenfeld residuals.
Contraceptive use was categorized into one of three patterns over the study year: (1) continuation of the method selected at baseline (with or without breaks); (2) discontinuation of the baseline method followed by switching to another effective method; and (3) discontinuation of the baseline method without subsequent use of another effective method. Time to first pregnancy was estimated using survival analysis, and Cox proportional hazards were estimated to compare pregnancy hazards by baseline method selected and contraceptive use pattern. All analyses were conducted using STATA 11 (Stata Corporation, College Station, TX).