This article has Open Peer Review reports available.
Participant characteristics associated with withdrawal from a large randomized trial of spermicide effectiveness
© Raymond et al; licensee BioMed Central Ltd. 2004
Received: 06 July 2004
Accepted: 01 October 2004
Published: 01 October 2004
In most recent large efficacy trials of barrier contraceptive methods, a high proportion of participants withdrew before the intended end of follow-up. The objective of this analysis was to explore characteristics of participants who failed to complete seven months of planned participation in a trial of spermicide efficacy.
Trial participants were expected to use the assigned spermicide for contraception for 7 months or until pregnancy occurred. In bivariable and multivariable analyses, we assessed the associations between failure to complete the trial and 17 pre-specified baseline characteristics. In addition, among women who participated for at least 6 weeks, we evaluated the relationships between failure to complete, various features of their first 6 weeks of experience with the spermicide, and characteristics of the study centers and population.
Of the 1514 participants in this analysis, 635 (42%) failed to complete the study for reasons other than pregnancy. Women were significantly less likely to complete if they were younger or unmarried, had intercourse at least 8 times per month, or were enrolled at a university center or at a center that enrolled fewer than 4 participants per month. Noncompliance with study procedures in the first 6 weeks was also associated with subsequent early withdrawal, but dissatisfaction with the spermicide was not. However, many participants without these risk factors withdrew early.
Failure to complete is a major problem in barrier method trials that seriously compromises the interpretation of results. Targeting retention efforts at women at high risk for early withdrawal is not likely to address the problem sufficiently.
Retention of participants has been a consistent problem in clinical studies of barrier contraceptive methods. For example, in six large studies of condoms, diaphragms, and spermicides conducted in the past decade, more than 30% of the participants failed for reasons other than pregnancy to complete the intended six months or six menstrual cycles of follow-up [1–6] Such high dropout rates seriously compromise the interpretation of trial results.
Issues regarding the design of barrier method studies have become increasingly important to researchers and public health scientists since the onset of the HIV epidemic because of the urgent need for methods to prevent this disease and other sexually transmitted infections. Numerous new barrier contraceptive methods and microbicides are currently in various stages of development and testing. Devising effective approaches to maximize retention in these studies will be critical.
In this analysis, we used data from a large, recently completed randomized trial of the efficacy and safety of five spermicide products to determine whether we could identify specific subgroups of participants who were at particular risk for failure to complete the trial. Our goal was to provide information that might assist in the development of targeted approaches to improve follow-up in future trials.
The primary purpose of this randomized trial was to estimate and compare the probability of pregnancy during six months of typical use of five nonoxynol-9 spermicide products. Safety, acceptability, and product use were additional specified outcomes. The trial was conducted at 14 sites in the United States between June 1998 and August 2002. The study was approved by the institutional review boards at each site and at Family Health International. All participants signed written informed consent forms before enrollment.
A full description of the trial procedures has been published previously . In brief, the study enrolled 1536 healthy, sexually active women aged 18–40 years who had no history suggestive of subfecundity, who were at low risk for sexually transmitted infections, and who stated that they were willing to rely on a spermicide as their only contraceptive method for 7 months and to accept a moderate risk of pregnancy. At the enrollment visit, each volunteer had an interview, pelvic examination, Pap smear, wet prep, and urine pregnancy test. After eligibility was established, she completed a self administered questionnaire that included a question about strength of desire to avoid pregnancy. Each eligible participant was randomly assigned to one of the five study spermicide groups. She was given a supply of her assigned spermicide and a diary on which to record relevant information daily throughout the study. Some participants at two centers were enrolled into a substudy to evaluate colposcopic effects of the spermicides. Participants were encouraged but not required to inform their partners about the study except at one center, where the Institutional Review Board required signed consent of the partner.
Follow-up visits were scheduled at 4, 17, and 30 weeks after admission. Each participant was also asked to return to the study site if she wished to discontinue use of the spermicide. At each visit, the participant was interviewed, and a urine pregnancy test was done. At the 4-week and final visits, she completed a seven-page acceptability questionnaire. A pelvic examination was performed at the final visit and at other visits as indicated. Colposcopy substudy participants had a vaginal colposcopy at each follow-up visit. Each participant was asked to do a pregnancy test at home 2, 10, and 23 weeks after admission and to telephone the site with the result. If a participant missed a scheduled contact, study procedures required that staff make at least four attempts to contact her by at least two different modalities (telephone, mail, etc.) If they could not contact her directly, staff were to try to reach her through an alternate contact person identified by the participant at admission. Compensation for completion of all scheduled visits in the primary study ranged from $120 to $400 at the 14 study sites; at most sites, the amount was divided evenly among the separate visits.
In this analysis, we included all randomized participants except for 22 who were discovered to have been pregnant at admission and who therefore contributed no data to the primary analysis. We classified each of the remaining 1514 participants as having completed the study if she considered the spermicide to be her primary contraceptive method for at least 183 days after randomization, or she became pregnant before she stopped relying on it. Otherwise, she was classified as having failed to complete the study. We assigned each participant's last day in the analysis as the earliest of the following dates: the estimated date of fertilization of a pregnancy; the date she was last known to have been relying primarily on the assigned spermicide for contraception; the latest date her pregnancy status could be reliably determined; and 183 days after randomization. These rules were the same as those used in the prior primary pregnancy analyses .
We assessed the associations between failure to complete and 17 baseline factors of interest, which were prespecified before the analysis. Among the subset of participants who were in the analysis for at least 6 weeks, we examined the associations between final status category and various factors that characterized their experience during the first 6 weeks in the study. Factors were categorized in part to ensure substantial numbers of participants in each level. Hypotheses about the effects of factors on completion status were tested using chi square tests, Fisher's exact tests, Mantel Haenszel tests. Parameters estimated by multivariable logistic regressions were tested using Wald tests. We included factors in regression models if they were associated with the outcome (alpha<0.10) in bivariable analyses. None of the included factors were highly correlated. In both bivariable and multivariable analyses, a p-value of <0.05 was considered to indicate a significant association.
Reasons for early withdrawal
Participants N = 635
Discontinued by site
Site closure/staff error
Participant decision, reason provided*
Unwillingness to continue study visits
Objections from partner†
Desire to change contraceptive method†
Separation from partner
Side effects or other medical events†
Cessation of sexual activity
Dissatisfaction with spermicide†
Distrusted contraceptive efficacy†
Desire for pregnancy
Mistaken suspicion of pregnancy
Participant decision, reason unknown
Protocol compliance by final status category
Final status category
Did not complete study
Number of participants
Median days in analysis per participant
Mean % of expected follow-up visits completed*
Mean % of expected diary days recorded*
Mean % of expected pregnancy tests completed*
Association between baseline factors and failure to complete study
Did not complete study
Relative Risk (95% confidence interval)
1.29 (1.15 – 1.45)
> 25 years
single not living with partner
1.21 (1.07 – 1.36)
married or living with partner
1.14 (1.0 1 – 1.28)
Baseline coital frequency
≥8 acts per month
1.21 (1.07 – 1.37)
≤7 acts per month
Geographic region of US
1.28 (1.12 – 1.46)
0.93 (0.80 – 1.09)
1.25 (1.08 – 1.44)
Recruitment rate at study site
≤4 per month
1.23 (1.09 – 1.38)
> 4 per month
1.31 (1.16 – 1.47)
Participation in colposcopy study
1.26 (0.99, 1.61)†
Association between early study experience and failure to complete study*
Experience during first 6 weeks
Did not complete study
Relative Risk (95% confidence interval)
Completed at least one follow-up visit
0.68 (0.53 – 0.87)
Completed at least one pregnancy test within first 4 weeks
0.73 (0.41 – 1.31)
Provided diary information for each day
0.71 (0.54 – 0.94)
Used spermicide at every coital act
0.70 (0.57 – 0.87)
Of those who completed follow-up visit
Reported medical complaints
1.17 (0.92 – 1.50)
Disliked spermicide somewhat or a lot
1.05 (0.63 – 1.76)
Distrusted contraceptive efficacy
1.15 (0.87 – 1.52)
Disliked timing of application
1.01 (0.79 – 1.29)
Complained about messiness
0.88 (0.68 – 1.13)
Had problems with insertion
1.06 (0.83 – 1.35)
Reported that partner disliked spermicide
1.18 (0.90 – 1.56)
Happy with spermicide†
0.82 (0.64 – 1.05)
Discussion and conclusions
In analyzing data from longitudinal studies, researchers commonly assume that the experience of participants who withdraw early, had they stayed in the study, would have been similar to the experience of those who completed. However, this assumption is generally impossible to confirm and is often implausible. If the assumption is false, the study findings may substantially misrepresent the likelihood of the outcome in the study population. If the degree of misrepresentation is not consistent across study groups, comparisons could be seriously biased. Indeed, some expert epidemiologists have suggested that a trial with losses of greater than 20% of the participants "would be unlikely to successfully withstand challenges to its validity" [8, 9].
Our study, like other recent barrier contraceptive method studies, did not even approach this standard: 42% of our enrolled participants did not complete the trial. Furthermore, the participants who failed to complete were different in key ways from those who did – they reported significantly more frequent coitus at baseline, and they also were more likely to be younger, unmarried, and poorly compliant with study procedures and method use in the first few weeks after admission. All of these characteristics were associated to some extent with elevated risk of pregnancy in our population , which suggests that our high withdrawal rate indeed may have distorted our findings: the pregnancy probabilities that we reported may be underestimates.
Clearly, increased attention to preventing this problem in future studies is imperative. In performing this analysis, our intention was to explore the potential impact of focusing retention efforts on participants with characteristics that are associated with failure to complete. However, although we did find some factors that were significantly associated with early withdrawal, none was highly predictive; that is, many participants without these factors failed to complete the study, and many with these factors did complete. Therefore, applying special efforts only to the high-risk participants would not likely have been sufficient to raise completion rates to desirable levels. In future trials, aggressive follow-up measures should be instituted universally. Such efforts might include assigning individual "case-workers" to participants, using novel means for communicating with the participants, such as pagers, conducting visits at participants' homes or at other locations convenient for them, providing specific reimbursement for expenses such as travel, parking, and child care, or providing extra incentives for completing follow-up. Researchers should be mindful, however, that one downside to some of these approaches is that they might influence participants' use of the study product or other behaviors related to the study outcome, which is detrimental if the goal of the trial is to estimate effectiveness during "typical use" of the product.
In our study, participants who enrolled at study centers where enrollment was slow were at increased risk of failure to complete the study. The reason for this association is unclear. Factors at these centers that hindered enrollment also may have adversely affected participants' interest in remaining in the study. Alternatively, in responding to pressure to hasten recruitment, these centers may have enrolled women who were not good candidates for study completion. This latter possibility emphasizes the need to maintain a careful balance between recruitment and retention goals: rapid recruitment of participants who then drop out of the study is not beneficial to the study as a whole.
The amount of reimbursement promised to our participants was strongly associated with final completion status in the bivariable analysis, but this effect was not significant when adjusted for other factors in our multivariable model. Numerous prior studies have shown that modest monetary incentives (e.g., $20 or less) increase response rates to surveys or short follow-up studies [10, 11] Some data also suggest that the value of the incentive matters, although possibly with diminishing returns as the value increases [12, 13]. However, the effect of higher levels of compensation in longer trials such as ours has not been rigorously studied. The possibilities that large financial incentives could be coercive, weaken generalizability, or encourage bogus participation are important concerns .
We were surprised that several of the factors that we expected would be associated with early withdrawal did not show significant associations in this analysis. When we began this analysis, we presumed that one reason for both slow enrollment and poor follow-up rates in barrier method trials is the relatively poor efficacy of these products: women may consider them to be temporary or backup methods and thus may be unwilling to use them as their sole or primary contraceptive for the 6–12 month duration of these studies. However, in our study, participants who strongly wished to avoid pregnancy or who had completed their desired family size were not more likely than others to drop out, nor were participants who expressed concerns about contraceptive efficacy early in the trial. Furthermore, neither early medical problems nor other complaints about the spermicides were predictive of withdrawal. These findings differ from that of a previous randomized trial of spermicides conducted mostly in developing countries. In that trial, participants who initially liked the assigned product very much were more likely than others subsequently to complete the study and to use the product for a longer period of time after admission .
In one respect, the poor retention rate in our study and in other barrier method trials is a result of the design of these studies, which typically call for censoring data (and in most barrier method trials, terminating active follow-up) when participants stop relying on the assigned contraceptive method. This design prohibits a true intent-to-treat analysis and is consequently a potential source of bias. Clearly, retention would be higher if the trials were designed at the outset to follow all subjects for the full intended duration of follow-up, even if they switched contraceptive methods. However, data from participants who are not using the method under study are not necessarily relevant to the efficacy and safety of the method. For the results of these trials to be meaningful, as many subjects as possible must not only complete follow-up but also continue to use the method during the full follow-up period. In our study, almost all the women who gave a reason for withdrawing early either cited problems with the spermicide or indicated that they wished to switch to another method after leaving the study. Our results are consistent with the findings of the 1995 National Survey of Family Growth, which showed that more than 47% of US spermicide users stopped relying on the method within the first 6 months of use . These findings are discouraging: they suggest that even if the retention in the study could be improved by aggressive follow-up techniques, the likelihood of significant extension of method use is low.
Our results suggest that to reduce bias potentially introduced by a large proportion of participants failing to complete the study, future barrier contraceptive method researchers should consider approaches in addition to those directly aimed at tracking and retaining individual participants. For example, both to reduce the burden on participants and to help the study staff maintain focus on follow-up, limiting data collection to critical variables may be appropriate. Complete collection of key data is clearly preferable to inadequate collection of less important data. Reducing the planned duration of follow-up would also certainly reduce withdrawals; although a larger sample size would be needed to provide the desired levels of precision and power, this disadvantage might be overcome if the shorter study were more attractive to potential participants. Given the large proportion of women who stop using the method earlier than 6 months, it is not clear that 6-month pregnancy probabilities are clinically needed anyway. Adding a run-in period to the trial before randomization might be helpful in excluding participants likely to drop out very early after admission, although such an addition might deter enrollment of other women as well, which is also a problem in these trials. Finally, innovative study designs to measure product efficacy should be evaluated. The design proposed by Steiner et al., which compares the one-month pregnancy probability in a relatively small number of women using a contraceptive method to the probability in women using a placebo, offers an alternative to the traditional 6–12 month trial . It showed some promise in a pilot study and is currently being further tested in a study of a new candidate spermicide.
Support for this study was provided by Family Health International (FHI) with funds from the National Institute of Child Health and Human Development (NICHD) contract number N01-HD-7-3271. The views expressed in this article do not necessarily reflect those of FHI or NICHD.
- Frezieres RG, Walsh TL, Nelson AL, Clark VA, Coulson AH: Evaluation of the efficacy of a polyurethane condom: results from a randomized, controlled clinical trial. Fam Plann Perspect. 1999, 31: 81-7.View ArticlePubMedGoogle Scholar
- Raymond E, Dominik R, Spermicide Trial Group: Contraceptive effectiveness of two spermicides: a randomized trial. Obstet Gynecol. 1999, 93: 896-903. 10.1016/S0029-7844(99)00002-2.PubMedGoogle Scholar
- Mauck C, Glover LH, Miller E, Allen S, Archer DF, Blumenthal P, Rosenzweig A, Dominik R, Sturgen K, Cooper J, Fingerhut F, Peacock L, Gabelnick HL: Lea's Shield: a study of the safety and efficacy of a new vaginal barrier contraceptive used with and without spermicide. Contraception. 1996, 53: 329-35. 10.1016/0010-7824(96)00081-9.View ArticlePubMedGoogle Scholar
- Mauck C, Callahan M, Weiner DH, Dominik R: A comparative study of the safety and efficacy of FemCap, a new vaginal barrier contraceptive, and the Ortho All-Flex diaphragm. The FemCap Investigators' Group. Contraception. 1999, 60: 71-80. 10.1016/S0010-7824(99)00068-2.View ArticlePubMedGoogle Scholar
- Steiner MJ, Dominik R, Rountree RW, Nanda K, Dorflinger LJ: Contraceptive effectiveness of a polyurethane condom and a latex condom: a randomized controlled trial. Obstet Gynecol. 2003, 101: 539-47. 10.1016/S0029-7844(02)02732-1.PubMedGoogle Scholar
- Walsh TL, Frezieres RG, Peacock K, Nelson AL, Clark VA, Bernstein L: Evaluation of the efficacy of a nonlatex condom: results from a randomized, controlled clinical trial. Perspect Sex Reprod Health. 2003, 35: 79-86. 10.1363/3513003.PubMedGoogle Scholar
- Raymond EG, Chen PL, Luoto J, Rountree RW, Barnhart KT, Bradley L, Creinin MD, Heine MW, Poindexter A, Wan L, Martens M, Schenken R, Nicholas CF, Blackwell R, Archer DF, Holmes M: Contraceptive effectiveness and safety of five nonoxynol-9 spermicides: a randomized trial. Obstet Gynecol. 2004, 103: 430-9.View ArticlePubMedGoogle Scholar
- Schulz KF, Grimes DA: Sample size slippages in randomised trials: exclusions and the lost and wayward. Lancet. 2002, 359: 781-5. 10.1016/S0140-6736(02)07882-0.View ArticlePubMedGoogle Scholar
- Sackett DL, Straus SE, Richardson WS, Rosenberg W, Haynes RB: Evidence Based Medicine. 2000, Edinburgh: Churchill Livingstone, 2Google Scholar
- Kissinger P, Kopicko JJ, Myers L, Wustrack S, Elkins W, Farley TA, Martin D: The effect of modest monetary incentives on follow-up rates in sexually transmitted disease studies. Int J STD AIDS. 2000, 11: 27-30. 10.1258/0956462001914869.View ArticlePubMedGoogle Scholar
- Halpern SD, Asch DA: Commentary: Improving response rates to mailed surveys: what do we learn from randomized controlled trials?. Int J Epidemiol. 2003, 32: 637-8. 10.1093/ije/dyg184.View ArticlePubMedGoogle Scholar
- Asch DA, Christakis NA, Ubel PA: Conducting physician mail surveys on a limited budget. A randomized trial comparing $2 bill versus $5 bill incentives. Med Care. 1998, 36: 95-9. 10.1097/00005650-199801000-00011.View ArticlePubMedGoogle Scholar
- Halpern SD, Ubel PA, Berlin JA, Asch DA: Randomized trial of 5 dollars versus 10 dollars monetary incentives, envelope size, and candy to increase physician response rates to mailed questionnaires. Med Care. 2002, 40: 834-9. 10.1097/00005650-200209000-00012.View ArticlePubMedGoogle Scholar
- Patten SB, Li FX, Cook T, Hilsden RJ, Sutherland LR: Irritable bowel syndrome: are incentives useful for improving survey response rates?. J Clin Epidemiol. 2003, 56: 256-61. 10.1016/S0895-4356(02)00586-3.View ArticlePubMedGoogle Scholar
- Raymond E, Dominik R, Díaz S, Ledesma L, Alvarado A, Bassol S, Fernández V, Morales E, Turkson S, Carlos G: Contraceptive effectiveness of two spermicides: a randomized trial. Obstet Gynecol. 1999, 93: 896-903. 10.1016/S0029-7844(99)00002-2.PubMedGoogle Scholar
- 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 (93): 64-72.View ArticlePubMedGoogle Scholar
- Steiner MJ, Hertz-Picciotto I, Schulz KF, Sangi-Haghpeykar H, Earle BB, Trussell J: Measuring true contraceptive efficacy. A randomized approach – condom vs. spermicide vs. no method. Contraception. 1998, 58: 375-8. 10.1016/S0010-7824(98)00124-3.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2288/4/23/prepub
This article is published under license to BioMed Central Ltd. This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.