Potter, Joseph E. PhD; McKinnon, Sarah PhD; Hopkins, Kristine PhD; Amastae, Jon PhD; Shedlin, Michele G. PhD; Powers, Daniel A. PhD; Grossman, Daniel MD
Almost half of women who initiate use of oral contraceptives (OCPs) discontinue the method during the first year of use.1 Among the reasons given for OCP discontinuation are the side effects experienced by pill users, as well as poor compliance with the regimen of taking one pill each day.2–4 Access issues also have been found to contribute to OCP discontinuation. A study of contraceptive use patterns in the United States found that 10% of gaps in OCP use were attributable to difficulty paying for a method and lack of time for medical visits.5 Additionally, a study of Medicaid claims data from California showed that women given 13 pill packs had greater OCP continuation and fewer gaps in coverage than women who received three or fewer packs.6 Most recently, a clinical trial randomizing participants to receive a 3- or 7-month supply of OCPs showed significantly higher continuation rates for women receiving more pill packs (51% compared with 35%).7
Another way to reduce access barriers would be to make OCPs available over the counter (OTC).8 Two surveys of U.S. women found that between 40% and 60% of nonusers said they would be more likely to use OCPs if they were available in a pharmacy without a prescription.9,10
The objective of this study is to estimate how both of these measures of access, the number of pill packs dispensed and providing OTC access to the pill, are associated with OCP continuation. To simulate the latter measure, we take advantage of the natural experiment along the United States–Mexico border, where U.S. residents can access OCPs without a prescription in Mexican pharmacies for approximately U.S. $5 per cycle.11 By comparing OCP continuation between users who obtained pills in U.S. clinics and those who obtained them OTC in Mexican pharmacies, we gain insight into what might be expected with regard to continuation if OCPs were made more easily accessible in the United States.
MATERIALS AND METHODS
The Border Contraceptive Access Study was conducted in El Paso, Texas, from 2006 through 2008. El Paso, Texas, with a population of 800,000 people, is located on the United States–Mexico border directly across from Ciudad Juárez, Mexico. The majority of the El Paso population is Hispanic/Latino and many residents regularly cross the border into Mexico for health services because of lower costs and convenience as well as social ties, cultural familiarity, and perceived quality of care.12–15
Using convenience sampling, we recruited current OCP users aged 18 to 44 stratified into two groups: 1) El Paso residents who use OCPs and obtained their last pack at a family planning clinic in El Paso (target n=500) and 2) El Paso residents who use OCPs and obtained their last pack OTC at a pharmacy in Mexico (target n=500). There were no inclusion criteria other than age, source of last pack, and residency in El Paso. Respondents were interviewed up to four times at 3-month intervals. Further information on recruitment strategies and the interview process is available in an earlier article.11 For completing the baseline interview and each of the follow-up interviews, participants were offered a small compensation (up to $75 total). This study was approved by the Institutional Review Boards at the University of Texas at Austin and University of Texas at El Paso.
The initial (or baseline) questionnaire collected information on the participant's background (age, marital status, parity) as well as the month and year that a participant obtained her last pill pack(s), the number of packs acquired at that time, where she got her packs, experience of side effects, the duration of her current OCP segment, and how long she planned on using OCPs as her method of contraception. In all subsequent interviews, a woman was asked about changes in her contraceptive practice during the previous 3 months, the source and number of pill packs obtained if she resupplied since the previous interview, and, again, how long she planned to use OCPs. The second and third interviews were administered via telephone, whereas the final (fourth) interview was conducted in person.
In total, we recruited 1,046 OCP users who completed baseline interviews, 514 women who obtained OCPs from Mexican pharmacies and 532 women who obtained OCPs from El Paso family planning clinics. Overall, 90% of participants completed the study, with only 105 women lost to follow-up. Of the 965 women who completed the second interview and thus were eligible for this analysis, information for 24 was not used because of missing or incomplete data.
Sociodemographic characteristics, side effects of OCPs, and planned OCP use were calculated for clinic and OTC users separately. The statistical significance of differences between these groups was determined using χ2 tests of the homogeneity of proportions. Next, following Westhoff et al,16 we defined OCP discontinuation as having not taken any active OCP for more than 7 days and probed in each interview for any discontinuation. If the woman reported stopping, then we asked for the date of discontinuation (month and year) and the reason for stopping. In addition to 13 precoded reasons for discontinuation, we probed with an open-ended question about why the respondent stopped using OCPs. We recoded the reason for stopping into four categories: 1) stopped to become pregnant or because no longer needed contraception; 2) stopped to switch to another method; 3) became pregnant while using OCP; and 4) stopped because of side effects or other problems with OCPs.
We separated exposure to the risk of discontinuation into separate segments corresponding to the intervals between interviews (approximately 90 days) for a maximum of three segments. The length of time a woman planned to continue using OCPs was updated at each successive interview. For women who had resupplied between interviews, we further divided their use segments on the date they reported last obtaining pill packs and updated source of OCPs and number of packs accordingly.
We calculated Kaplan-Meier estimates of continuation for clinic and cross-border OTC users based on the elapsed time since the baseline interview. These quantities are estimated for all reasons for discontinuation, as well as for selected method-related reasons (3 and 4 above: method failure, side effects, and other reasons). We focus on these two types of discontinuation because they are especially likely to lead to an unwanted pregnancy, unlike stopping to switch to another method or to become pregnant.1 Log-rank tests were used to assess equality of survivor functions.
We estimated three Cox proportional hazards models to test for differences in the risk of discontinuation for the selected method-related reasons between women who obtained their OCPs in El Paso clinics and those who obtained their OCPs OTC in Mexico. In the first, the source of contraception is the only predictor of discontinuation. In the second, we also adjust for differences between women in the two arms of the study identified in a previous analysis (age, country of birth, country where education was completed, receipt of government assistance, U.S. health insurance, and border-crossing frequency11), as well as three additional variables assessed in the baseline interview that could be expected to have a direct effect on a woman's motivation to continue using OCP: duration of use, experience of side effects attributable to the pill, and how long the woman reported she planned to use OCPs.
In a further effort to assess the effect of access and convenience on OCP continuation, we estimated a third model that divided the El Paso clinic users according to the number of pill packs they received the last time they resupplied. Users were separated into those receiving one to five packs and those receiving six or more packs, with source and number of packs updated on the basis of the information collected at successive interviews. If having to visit the clinic to resupply pill packs constituted a significant barrier to use, then we would expect to find higher rates of discontinuation among women provided fewer packs. However, because family planning clinics in El Paso were operating under severe budget constraints at the time of the study, clinic personnel were wary of giving out large numbers of packs to clients they thought might soon discontinue and waste the OCPs. Thus, the number of packs dispensed was likely based on the clinic personnel's prognosis of continuation and therefore was subject to indication bias.17
We attempt to diminish any confounding resulting from indication by adjusting for characteristics found to be associated with the number of pill packs dispensed to clinic users as well as all the other predictors included in the second model. For all Cox proportional hazards models, estimates of the effects of risk factors were obtained using the Efron method for handling tied event times.18 Additionally, we performed conventional diagnostic tests for nonproportional effects, outliers, and influential observations to check the robustness of our findings.19–20
The target sample size for the study, 500 clinic users and 500 cross-border OTC users, was chosen with multiple analyses in mind because we planned to analyze knowledge of how to use the pill, use of preventive screening services, and prevalence of contraindications, in addition to continuation. With a projected cumulative loss to follow-up of 19%, the target sample size would allow us to test for a 10-percentage-point one-sided difference between proportions (from 0.7 to 0.6) in the two cohorts of the study with power of 0.9 and an alpha level of .05. All analyses were performed using Stata 10.0.21
Social and demographic characteristics for the OCP users eligible for this analysis are shown in Table 1 according to their source of OCPs at the baseline interview. In comparison with El Paso clinic users, cross-border OTC users were more likely to be older, have higher parity, be born in Mexico, and have completed their last year of schooling in Mexico. They were less likely to have U.S. health insurance and, not surprisingly, crossed the border more frequently. The cross-border OTC users also had been using OCPs for longer and were less likely to report experiencing side effects at the baseline interview. However, the answers women gave to the question regarding how long they expected to use OCPs were distributed quite evenly between the two cohorts. Among the cross-border OTC users, the proportion reporting that they only planned to use OCPs for 3 months or less was higher, whereas the proportion reporting that they planned to use OCPs for 2 years or more was lower compared with El Paso clinic users.
Of the 941 OCP users, 216 discontinued use during the approximately 9-month period of observation (Table 2). Among those who discontinued use, 15% did so because of method failure, 19% stopped to become pregnant, and nearly 60% stopped because of side effects or other reasons. Only 14 women reported stopping OCPs to switch to another method of contraception. However, a larger proportion of participants switched source of OCPs during follow-up: 44 cross-border OTC users (as measured at the baseline interview) switched to another source for OCPs (20 became clinic users and 24 switched to another source such as a private doctor's office; data not shown). Among women who were initially El Paso clinic users, 40 became cross-border OTC users and 27 switched to another source.
Life table estimates of continuation at 9 months after the initial interview were nearly 5% higher for cross-border OTC users than for El Paso clinic users (Fig 1). This difference was consistent for discontinuation for all reasons and discontinuation attributable to method-related reasons. The log-rank test for equality of survivor functions was rejected at P<.01 in both pairs of life tables.
In the first Cox proportional hazard model with no covariates, the risk of discontinuation attributable to method-related reasons is significantly higher for El Paso clinic users compared with cross-border OTC users (hazard ratio 1.48, 95% confidence interval 1.07–2.04). After adjusting for covariates in the second model, the estimated risk ratio increased slightly (hazard ratio 1.58, 95% confidence interval 1.11–2.26).
The final Cox model of discontinuation for method-related reasons separated clinic users based on the number of pill packs they were given the last time they resupplied (Table 3). We found that approximately 35% of clinic users received six or more packs, and that clinic users who were both married and had children were especially likely to have been supplied with six or more packs (data not shown). Thus, this model adjusts for length of planned use, duration of use at baseline, side effects, and predictors of source, and also for parity and marital status. Relative to cross-border OTC users, the adjusted rate of discontinuation is nearly identical for clinic users who received six or more packs but 80% higher (hazard ratio 1.80, 95% confidence interval 1.22–2.65) for clinic users supplied with fewer than six packs. In this model, as we expected, length of planned use reported at baseline was strongly associated with the hazard of method-related discontinuation, but experience of side effects was not a significant predictor of this hazard. None of the other variables included in the model was significantly associated with discontinuation. Tests revealed no departure from the proportional hazards assumption in any of the models and, in particular no evidence of a nonproportional effect for the main variables of interest, source, and number of packs (P>.67).
We observed lower rates of OCP discontinuation for cross-border OTC users than for El Paso clinic users both in life table analyses and hazard models that adjusted for user characteristics and contraceptive goals and experience. Only when the comparison was restricted to clinic users who were dispensed six or more pill packs when they last resupplied was continuation among clinic users comparable to that of the cross-border OTC users. Beyond source of contraception, the only other significant predictor of discontinuation was a woman's report of how long she intended to use OCPs, suggesting that planned duration of use is an important piece of information clinicians should consider.
The location of the study in El Paso, Texas, afforded us the opportunity to compare the experience of users who accessed the pill OTC in a setting in which this option had been available for a considerable length of time and was, in addition to family planning clinics, a regular source of OCPs for low-income women, especially those without private health insurance. However, a limitation of this natural experiment, in addition to convenience sampling, is that the increase in access provided by moving OCPs over the counter in the United States would be greater than that now available to El Paso residents who have to cross the border to get to a pharmacy where a prescription is not required. On the other hand, it is possible that once OCPs were available without a prescription in the United States, the price might be set higher than that prevailing in Mexican pharmacies. Also, whereas detailed baseline and follow-up questionnaires permitted us to capture extensive data on the number of pill packs dispensed to clinic users, how long a woman planned on using the pill, whether she had previously experienced side effects, and the month and year of discontinuation, responses to these questions may have been subject to recall failure and social desirability bias.22 Still, any such bias would likely affect both groups equally.
Cross-border OTC users and clinic users differed in a number of characteristics, some of which could be expected to be associated with OCP continuation. Duration of use, age, marital status, insurance, and parity are among the variables that have been found to be associated with OCP discontinuation in previous studies.1,23–25 However, our concern that the women who elected to get their OCPs in Mexico might have been more motivated to continue using the pill than El Paso clinic users was alleviated by the similarity between the two groups of women in their reports of how long they planned to use the pill. On this basis, clinic users would be expected to have lower rates of discontinuation than cross-border OTC users. One difference between the women that could explain higher discontinuation among clinic users is the greater proportion reporting that they had experienced side effects. Nonetheless, after adjusting for prospective measures of both planned use and side effects, the hazard ratios still showed significantly lower discontinuation for method-related reasons for cross-border OTC users.
Although we cannot rule out the possibility that the lower discontinuation rate of clinic users receiving six or more packs was attributable to indication bias, this bias seems unlikely to have played an important role. In theory, clinicians could have access to additional information regarding a woman's motivation, knowledge of how to use OCPs correctly, and previous experience with this method,17 but we did not find any significant association between women's reports of how long they planned to use the pill and the number of packs dispensed to clinic users, suggesting that clinicians either do not ask women about their plans or are unwilling to trust what women tell them about planned duration of use. Moreover, the size of the effect is consistent with the findings from a recent clinical trial.7 It is also consistent with the findings from a large observational study using California claims data,6 but our observational study includes a number of controls that were not available in the California study.
These results provide evidence suggesting that removing the prescription requirement for OCPs, in addition to making it easier for women to initiate OCP use, would not have an adverse effect on continuation and might well improve it. Moreover, either by way of dispensing a larger number of pill packs or by way of OTC provision, access appears to have a greater effect on continuation than side effects, consistent with other prospective studies.26 The large effect of the number of pill packs dispensed to clinic users in the final model suggests that the need for repeated clinic visits for resupply is an impediment to continuation. Also, whereas estimated continuation rates for cross-border OTC users and clinic users receiving six or more packs were similar, this should not be taken to imply that increasing the number of pill packs is a substitute for changing the prescription status of OCPs. Both would likely expand access, but only moving OCPs over the counter would leave the need for clinic or doctor visits up to women and provide the user with the possibility of sending someone else to pick up her OCPs for her.11
Given the parallel findings from an RCT and a large observational study,6,7 increasing the number of packs dispensed at clinics and pharmacies would clearly be an important step to increase the convenience of OCP use. Although this would not require Food and Drug Administration approval, it would require changes on the part of providers and insurances. For example, some state Medicaid programs limit birth control to 30- or 90-day supplies,27 as do many private insurances. The only potential down-side of providing more pill packs is product wastage if a significant number of packs are not used.
Switching an OCP product OTC would require U.S. Food and Drug Administration approval and would involve considerations of OTC use beyond continuation. Another analysis of the Border Contraceptive Access Study found that contraindications to combined oral contraceptives were more common among cross-border OTC users than clinic users,28 and a future analysis will report on the use of preventive screening services in the two cohorts. As part of an OTC switch application, the Food and Drug Administration would require an actual use study, which would provide additional evidence concerning the effect of OTC access on OCP effectiveness that would complement our results from this natural experiment drawn from the border setting of El Paso, Texas.
1. Vaughan B, Trussell J, Kost K, Singh S, Jones R. Discontinuation and resumption of contraceptive use: results from the 2002 National Survey of Family Growth. Contraception 2008;78:271–83.
2. Potter L. Oral contraceptive compliance and its role in the effectiveness of the method. In: Cramer J, Spilker B, editors. Patient compliance in medical practice and clinical trials. New York (NY): Raven Press; 1991. p. 195–205.
3. Rosenberg MJ, Burnhill MS, Waugh MS, Grimes DA, Hillard PJ. Compliance and oral contraceptives: a review. Contraception 1995;52:137–41.
4. Rosenberg MJ, Waugh MS. Oral contraceptive discontinuation: a prospective evaluation of frequency and reasons. Am J Obstet Gynecol 1998;179:577–82.
5. Frost JJ, Singh S, Finer LB. US women's one-year contraceptive use patterns, 2004. Perspect Sex Reprod Health 2007;39:48–55.
6. Foster DG, Parvataneni R, de Bocanegra HT, Lewis C, Bradsberry M, Darney P. Number of oral contraceptive pill packages dispensed, method continuation, and costs. Obstet Gynecol 2006;10:1107–14.
7. O'Connell K, Roca C, Westhoff CL. The impact of pack supply on oral contraceptive continuation: a randomized controlled trial. Contraception 2010;82:185–6.
8. Trussell J, Stewart F, Potts M, Guest F, Ellertson C. Should oral contraceptives be available without prescription? Am J Public Health 1993;83:1094–9.
9. Landau SC, Tapias MP, McGhee BT. Birth control within reach: a national survey on women's attitudes toward and interest in pharmacy access to hormonal contraception. Contraception 2006;74:463–70.
10. Grossman D, Fernández L, Hopkins K, Amastae J, Potter JE. Perceptions of the safety of oral contraceptives among a predominately Latina population in Texas. Contraception 2010;81:254–60.
11. Potter JE, White K, Hopkins K, Amastae J, Grossman D. Clinic versus over-the-counter access to oral contraception: choices women make in El Paso, Texas. Am J Public Health 2010;100:1130–6.
12. Bastida E, Brown HS, Pagan JA. Persistent disparities in the use of health care along the US-Mexico border: an ecological perspective. Am J Public Health 2008;98:1987–95.
13. Amastae J, Fernandez L. Transborder use of medical services among Mexican American students in a U.S. border university. J Borderland Studies 2006;21:77–87.
14. Fernandez L, Howard C, Amastae J. Education, race/ethnicity and out-migration from a border city. Pop Res Policy Rev 2007;26:103–24.
15. Potter JE, Moore A, Byrd T. Cross-border procurement of contraception: estimates from a postpartum survey in El Paso, Texas. Contraception 2003;68:281–7.
16. Westhoff C, Heartwell S, Edwards S, Zieman M, Cushman L, Robilotto C, et al. Initiation of oral contraceptives using a quick start compared with a conventional start: a randomized controlled trial. Obstet Gynecol 2007;109:1270–6.
17. Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias—Effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007;297:278–85.
18. Efron B. The efficiency of Cox's likelihood for censored data. J Am Stat Assoc 1977;72:557–65.
19. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrika 1982;69:239–41.
20. Therneau TM, Grambsch PM. Modeling survival data: extending the Cox model. New York (NY): Springer; 2000.
21. StataCorp. Stata Statistical Software Release 10. College Station, TX: StataCorp LP; 2007.
22. Stuart GS, Grimes DA. Social desirability bias in family planning studies: a neglected problem. Contraception 2009;80:108–12.
23. Frost JJ, Darroch JE. Factors associated with contraceptive choice and inconsistent method use, United States, 2004. Perspect Sex Reprod Health 2008;40:94–104.
24. Frost JJ, Singh S, Finer LB. Factors associated with contraceptive use and nonuse, United States, 2004. Perspect Sex ReprodHealth 2007;39:90–9.
25. Trussell J, Vaughan B. Contraceptive failure, method-related discontinuation and resumption of use: results from the 1995 National Survey of Family Growth. Fam Plann Perspect 1999;31:64–72, 93.
26. Westhoff CL, Heartwell S, Edwards S, Zieman M, Stuart G, Cwiak C, et al. Oral contraceptive discontinuation: do side effects matter? Am J Obstet Gynecol 2007;196:6.
27. Crowley JS, Ashner D, Elam L. State Medicaid: outpatient prescription drug policies: findings from a national survey, 2005 update. Washington (DC): The Kaiser Family Foundation; 2005.
28. Grossman D, White K, Hopkins K, Amastae J, Shedlin M, Potter JE. Contraindications to Combined Oral Contraceptives Among Over-the-Counter Compared With Prescription Users. Obstet Gynecol 2011;117:558–65.