Background: Participants' protocol adherence may influence assessments of the effectiveness of new female-controlled methods for sexually transmitted infection prevention.
Methods: In 2005 we conducted a randomized pilot study among female sex workers (FSWs) in Madagascar in preparation for sexually transmitted infection prevention trial of diaphragms and a vaginal microbicide. Participants (n = 192) were randomized into 4 arms: diaphragm plus microbicide (Acidform), diaphragm plus placebo gel hydroxyethyl cellulose (HEC), Acidform alone, or HEC alone. FSWs were seen weekly for 4 weeks. Using multivariable regression with generalized estimating equations, we assessed predictors of adherent product use during all sex acts in the last week. We collapsed the gel-diaphragm arms together and the gel-only arms together for this analysis.
Results: Between 43% and 67% of gel-diaphragm users (varying by visit) reported using study products during all sex acts in the last week, compared with 20% to 45% of gel-only users. Adherence increased with follow-up [visit 4 vs. visit 1 risk ratio (RR) for gel-diaphragm users: 1.55, P <0.01; for gel-only users, RR: 1.58, P = 0.01]. Gel-diaphragm users whose casual partners were never aware of products (RR: 2.02, P = 0.03) and who had experienced partner violence after requesting condom use (RR: 1.45, P <0.01) were more adherent. Gel-only users reporting lower sexual frequency (1–9 weekly acts vs. ≥19 acts, RR: 1.98, P <0.01) and no sex with primary partners in the past week (RR: 1.54, P = 0.02) were more adherent.
Conclusions: Gel-diaphragm users had better adherence than gel-only users, and predictors of adherence differed between groups. Addressing modifiable factors during counseling sessions may improve adherence.
A 4-week randomized pilot trial of female sex workers in Madagascar assessed demographic, behavioral, product- and trial-related factors associated with adherent use of candidate vaginal microbicide and diaphragm during sex.
*Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina; †UNC-MAD, Antananarivo, Madagascar; and ‡United States Centers for Disease Control and Prevention (CDC), Atlanta, Georgia
This study was funded by the United States Centers for Disease Control and Prevention through an Inter-Agency Agreement with the United States Agency for International Development and CONRAD. Abigail Norris Turner was supported by the Joseph E. Pogue Pre-Doctoral Fellowship at the University of North Carolina at Chapel Hill.
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
The Mad STI Prevention Group: Ny Lovaniaina Rabenja, Béatrice Fenozandry, Louisette Ralimamonjy, Sandra Charlie Razafindravoavy, Rakotondramanana Lalanirina Rotsihanitriniaina, Jennifer Legardy-Williams, Noro Ravelomanana, Justin Ranjalahy Rasolofomanana, Maurizio Maculuso, D'Nyce Williams.
Correspondence: Abigail Norris Turner, PhD, Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, McGavran-Greenberg Hall, CB 7435, Chapel Hill, NC 27599-7435. E-mail: firstname.lastname@example.org.
Received for publication July 8, 2008, and accepted October 3, 2008.
Female-controlled methods for prevention of hiv and sexually transmitted infections (STIs) are those that do not rely (or rely less) on cooperation from male partners. New female-controlled methods, including vaginal microbicides that could kill or deactivate sexually transmitted pathogens, are currently under development. Cervical barriers such as diaphragms, which were originally developed for contraception, are also being reconsidered for their ability to prevent disease.1
As with male condoms, effective protection using female-controlled methods is dependent on adherence. Recent research (largely short-term pilot studies) has found that adherence in microbicide and diaphragm studies was associated with the acceptability of product characteristics and instructions for use,2,3,4,5 understanding of study concepts,3,6 education,7 and partner approval.3–5,8–10 Knowledge about factors associated with consistent product use can be incorporated into counseling messages to potentially improve adherence.
We conducted a 4-week, randomized pilot study in preparation for an upcoming large-scale randomized controlled trial that will examine the effectiveness of the diaphragm with candidate vaginal microbicide for prevention of gonococcal and chlamydial infection among female sex workers (FSWs) in Madagascar. A primary objective of this pilot was to identify factors associated with adherent study product use to aid in the development of appropriate protocols and counseling messages for the upcoming trial.
Materials and Methods
Study methods are described in detail elsewhere.11 The pilot trial took place in 2005 in 4 cities in Madagascar: Antananarivo, Antsiranana, Mahajanga, and Toamasina.
Recruitment and Screening
Recruitment occurred though a peer-outreach program.12,13 Peer counselors approached women in community venues to invite them to the study clinic for formal screening. FSWs attending the same clinics for routine care were also invited to screen.
Eligibility criteria included having at least 4 different sex partners in the past month, using condoms in less than 100% of sex acts in the previous 2 weeks, age 15 to 55 years, not pregnant or planning pregnancy in the next 2 months, with no allergy to latex, and no physical abnormality making diaphragm use impossible.
At screening, women underwent pregnancy and urinary tract infection testing. Pregnant women and those with urinary tract infections were discontinued from screening.
Eligible women (n = 192) were randomized to 4 groups (n = 48 in each arm): (1) vaginal microbicide Acidform (TOPCAD, Chicago, IL) used with a latex All-flex Arcing Spring Diaphragm (Ortho Ortho-McNeil Pharmaceutical, Inc., Titusville, NJ); (2) inert control hydroxy ethylcellulose (HEC) placebo gel (ReProtect LLC, Baltimore, MD) used with a latex diaphragm; (3) HEC alone; and (4) Acidform alone. Randomization was stratified by site. Treatment assignments were partially masked (assignment of diaphragms was open, but neither staff nor participants was aware of gel assignments).
Protocols for Study Product Use
Women in the gel-diaphragm arms were instructed to use the diaphragm continuously, removing it once daily for cleaning. After cleaning, women used a study-provided applicator to dispense their assigned gel into the dome of the diaphragm, then immediately reinserted the diaphragm with gel. Participants in gel-only arms were instructed to insert their assigned gel intravaginally before each sex act, again using a study applicator.
All women were instructed to use male condoms with every act. Women returned all unused experimental products at the end of the study.
Enrollment and Follow-Up
At enrollment, participants received face-to-face interviews and counseling on use of condoms, diaphragms and/or gel, as appropriate for each study arm. All women received a pelvic exam and those randomized to use the diaphragm were fitted. Participants were given as many condoms as they wanted.
Follow-up visits occurred once weekly for 4 weeks. At each visit, women were interviewed about their sexual behavior and study product used. They underwent pelvic examination and received resupplies of assigned study products and condoms.
We compensated women $3.50 to 6.00 USD, depending on the visit, for their travel and time spent at the clinic.
All analyses were conducted using SAS (SAS Institute, Version 9.1.3, Cary, NC).
For 2 reasons, for this analysis we collapsed the Acidform-diaphragm and the HEC-diaphragm randomization arms into a single group (“gel-diaphragm users”) and the HEC-alone and Acidform-alone arms into 1 group (“gel-only users”). First, given the nature of the products and instructions for use, we hypothesized that adherence was more likely to vary by arm type than by individual study arm. Second, because of the small sample size, we wanted to increase our power to detect associations between various predictors and adherent product use. Because adherence could also vary by individual arm, we controlled for randomization arm in all models.
Adherence was defined dichotomously: self-reported use of assigned products (gel with diaphragm for the 2 gel-diaphragm arms, and gel for the 2 gel-only arms) during 100% of vaginal sex acts with all partners (husband, boyfriend(s), or casual partners) in the previous week, versus use of products at less than 100% of acts. We did not account for condoms in our adherence definition because condom use has been assessed elsewhere (Pettifor AE et al, unpublished data).11 Person-time during which women reported no sex was excluded from this analysis.
We used regression models with generalized estimating equations14,15 to assess the 1 week “risk” of adherent product use for women according to several predictors. We accounted for clustering resulting from repeated measurements on individual women and we specified the exchangeable working correlation matrix for all generalized estimating equations models.
Because of small sample size, we used Poisson regression with robust variance estimation16,17 rather than log-binomial regression18,19 to compute parameter estimates. We considered associations between adherent product use and several domains of participant characteristics: trial factors, demographic factors, reproductive factors, sexual behavior, product characteristics, and control/power issues.
Because we could achieve satisfactory model convergence with a limited number of predictor variables,20 and to reduce the probability of missing important confounding effects apparent only in the presence of other factors,21 we divided predictor variables by domain into 3 preliminary multivariable models (similar to Ingram et al.22): (1) trial and demographic factors; (2) reproductive and sexual behavior; and (3) product-related and control/power factors. Each preliminary model contained all variables specific to that domain.
We then conducted a manual backward elimination procedure with each domain-specific preliminary model to achieve the most parsimonious model, dropping variables one by one until all remaining predictors had Wald P values less than α = 0.25.23
With all variables surviving the backward elimination process of the domain-specific models, we constructed a single combined model. We then conducted another backward elimination procedure, removing variables one at a time until all remaining predictors had Wald test P values less than α = 0.05.
We conducted the preliminary and full model procedures separately for gel-diaphragm and gel-only users. The final models predicting adherent product use contained all variables that were significant predictors following the backward elimination procedure for either arm type.
Because of additional a priori concerns about confounding, we included randomization arm, age, visit, and sexual frequency in the last week in all domain-specific preliminary models, and we retained these variables in the final multivariable models.
All women gave written informed consent for screening and again for enrollment. The pilot study received ethical approval from the Comité d'Ethique auprès du Ministère de la Santé in Madagascar and the Institutional Review Boards at the University of North Carolina at Chapel Hill and the Centers for Disease Control and Prevention, United States.
Nearly all participants (n = 188, 98%) completed 4 weeks of follow-up. This analysis includes 93 gel-diaphragm users who contributed 364 person weeks (PWs) of follow-up time and 96 gel-only users who contributed 381 PWs. Eight PWs were excluded because women reported no sex during that time interval or because of incomplete data on product use.
By design, women were evenly distributed at enrollment by randomization arm and site (Table 1). Participants had a median age of 29 years, with a median age at first sex of 17 years and a median level of schooling of 6 years. Over half (61% in gel-diaphragm arms; 55% in gel-only arms) did not have a primary partner (husband or boyfriend), and a substantial minority (43% in gel-diaphragm arms; 35% in gel-only arms) shared their bedrooms with at least 2 additional people. Women in gel-diaphragm arms reported a median of 8 different partners and 10 acts in a typical week, whereas those in the gel-only arms had a median of 6 different partners and 9 acts in a typical week. Condom use with primary partners was less common than with casual partners: of those with a primary partner, 8% in the gel-diaphragm arms and 14% in the gel-only arms reported always or almost always using condoms with the primary partner, whereas 42% in the gel-diaphragm arms and 53% in the gel-only arms reported always or almost always using condoms with casual partners (Table 1).
Adherent Product Use Over Follow-Up
At the first follow-up visit, 43% of women in gel-diaphragm arms used study products during all acts in the previous week; this value increased to 49% at the second follow-up visit, 63% at the third follow-up visit, and 67% at the final follow-up visit. Overall, women in gel-only arms had lower compliance: 28% of women in gel-only arms at the first follow-up visit, 20% at the second visit, 31% at the third visit, and 45% at the final visit reported adherent use of gel during all acts with all partners in the last week.
Preliminary Multivariable Models: Domain-Specific Associations With Adherent Product Use
Measures of effect in Table 2 are adjusted for all variables in that domain as well as randomization arm, participant age, study visit, and number of acts in the previous week (Table 2).
Women's adherent use of diaphragm with assigned gel increased with each additional follow-up visit (Model 1, Table 2). Participants reporting previous use of female condoms for HIV prevention [RR: 1.42, 95% confidence interval (CI): 1.12–1.80], not having sex with a husband or boyfriend in the previous week (RR: 1.35, 95% CI: 1.00–1.81) (Model 2, Table 2), who did not conceal product use from casual partners (RR: 4.42, 95% CI: 1.23–15.91), whose partners were never aware of product use (RR: 2.00, 95% CI: 1.03–3.88), and who had experienced past violence from a casual partner after a suggestion of condom use (RR: 1.43, 95% CI: 1.14–1.80) (Model 3, Table 2) had significantly higher adherence with diaphragm and gel use during sex in the past week. In addition, reporting that study products were somewhat or very easy to conceal from casual partners (RR: 3.17, 95% CI: 0.92–10.90) and having no problems with study products in the past week (RR: 1.48, 95% CI: 0.96–2.28) (Model 3, Table 2) suggested higher adherence, though these associations were not statistically significant.
Women randomized to use their diaphragm with Acidform had significantly lower adherence than those using the diaphragm with HEC (RR: 0.68, 95% CI: 0.51–0.91) (Model 1, Table 2), and women reporting a past unwanted pregnancy also had lower adherence (RR: 0.73, 95% CI: 0.57–0.95) (Model 2, Table 2).
Adherence in the gel-only arms also improved significantly by the final follow-up visit (visit 4 vs. visit 1; RR: 1.66, 95% CI: 1.16–2.38) (Model 1, Table 2), but other significant predictors of adherence for the gel-only arms differed from factors identified for the gel-diaphragm arms. Women from Tamatave (RR: 1.63, 95% CI: 1.00–2.65) (Model 1, Table 2) and those reporting fewer weekly acts (1–9 acts vs. 19 or more acts; RR: 2.01, 95% CI: 1.25–3.23) (Model 2, Table 2) had higher compliance, as did women who reported no sex with a husband or boyfriend in the past week (RR: 1.73, 95% CI: 1.17–2.56) (Model 2, Table 2).
We observed a suggestion of higher adherence among women randomized to use Acidform-only versus those using HEC-only (RR: 1.35, 95% CI: 0.94–1.92), and among older women (≥35 years) compared to the youngest participants (16–24 years) (RR: 1.62, 95% CI: 0.94–2.81) (Model 1, Table 2), though these associations were not statistically significant.
Adherence was not associated in either group with education, site, socioeconomic factors, prior participation in an STI or pregnancy prevention study, or past use of birth control.
Final Multivariable Models
After final multivariable adjustment, the strongest significant predictor of consistent gel and diaphragm use in the previous week was reporting not concealing product use from casual partners (RR: 4.75, 95% CI: 1.33–16.96); women who reported that casual partners were never aware of product use (RR: 2.02, 95% CI: 1.07–3.80) were also more adherent (Table 3). Increased experience with study products (visit 4 vs. visit 1, RR: 1.55, 95% CI: 1.24–1.95) and a past experience of violence after a request for condom use (RR: 1.45, 95% CI: 1.17–1.80) remained significantly associated with increased adherence among gel-diaphragm users following final adjustment. Finding products easy to conceal from casual partners (RR: 3.32, 95% CI: 0.98–11.23) and reporting no sex with a husband or boyfriend in the past week (RR: 1.27, 95% CI: 0.98–1.65) were suggestive of higher adherence.
Users of Acidform with the diaphragm had significantly lower adherence than women randomized to use HEC with the diaphragm (RR: 0.77, 95% CI: 0.60–0.98), as did women who reported a past unwanted pregnancy (RR: 0.73, 95% CI: 0.57–0.93) (Table 3).
In gel-only arms, after final adjustment, higher adherence was also associated with more product experience (visit 4 vs. visit 1, RR: 1.58, 95% CI: 1.10–2.28). Women reporting fewer weekly acts (1–9 vs. 19 or more acts, RR: 1.98, 95% CI: 1.29–3.03) and no sex with a husband or boyfriend in the previous week (RR: 1.54, 95% CI: 1.07–2.22) were also more likely to report gel use during all acts in the previous week (Table 3). Two other factors were suggestive of increased adherence after final adjustment: randomization to Acidform rather than HEC (RR: 1.34, 95% CI: 0.94–1.91) and older age (≥36 years vs. 16–24 years; RR: 1.55, 95% CI: 0.97–2.48).
In this short pilot study of FSWs in Madagascar, adherent use of study products was substantially higher among women randomized to use gel with diaphragm than among women using gel-only. The gap in adherence by arm type was sustained during follow-up, though the proportion of women reporting adherent product use during all acts generally increased over time for both gel-diaphragm and gel-only users.
Reasons for the disparity in adherence by arm type may be related to the recommended protocols for use. Gel-diaphragm use was not coitally-dependent; rather, women applied gel into the dome of the diaphragm once daily, inserted it, and wore it continuously, whereas gel-only users were asked to apply gel intravaginally before every act. The increased volume of gel present in the vagina for gel-only users may have been less acceptable or may have resulted in fear of conflict with partners, discouraging consistent use. (We reported previously that partners of gel-only users had mentioned excessive lubrication and wetness in the previous week at 30% of follow-up visits).11 Of note, ease of covert product use was not a significant predictor of adherence among gel-only users.
Participant and product characteristics associated with adherent use differed for women randomized to gel-diaphragm and gel-only arms. After adjustment, gel-diaphragm users who reported that their casual partners were never aware of product use were significantly more likely to use their products compliantly, suggesting that women who were comfortable with covert use were more likely to use their products. Reinforcing this finding, gel-diaphragm users who reported that covert use was easy also had increased adherence, though this association was of borderline statistical significance. Interestingly, those reporting not concealing products from casual partners (reported at only 7 visits) also had significantly higher adherence. Though making inferences from such a small subgroup is difficult, this suggests that women who can engage in open communication with partners may also have higher levels of compliance. In light of these results, counseling sessions for the upcoming trial will include more extensive discussion and training related to covert use at baseline and during booster counseling at follow-up visits. Women will also be told in counseling sessions that although many women are able to use the products successfully during sex without their partner's knowledge, others choose to tell their partners about the products and are also able to use the products successfully. The finding that adherence increased over follow-up further suggests that more time dedicated to training in product use may improve participant compliance.
For gel-only users, associations between several predictors and adherent product use trended in the same direction as for gel-diaphragm users, but nearly all associations failed to achieve statistical significance. Longer duration of follow-up remained a significant predictor of adherent use of gel in the past week, and 2 other factors—reporting fewer acts and not having sex with a husband or boyfriend in the past week – were also significantly associated with use of gel during all acts. The relationship between number of acts and increased adherence is likely due in part to our definition of adherence. Women had to use their assigned products in 100% of acts to be considered adherent; the higher their number of acts, the more difficult it became to clear this threshold. For this reason, it is important that researchers adjust for the number of acts when measuring adherence. On the other hand, higher sexual frequency may also be 1 component of overall risky sexual behavior, and individuals with 1 risky behavior often have others. Not having sex with a husband or boyfriend in the previous week was also associated with higher adherence for both arm types (significantly for gel-only users, nonsignificantly for gel-diaphragm users). This fits with our observation that women seemed to use their products less compliantly with primary partners than with casual partners.
Randomization assignment was somewhat associated with adherence for both arm types, but in opposite directions. Among gel-diaphragm users, those randomized to use Acidform had lower compliance than those using HEC, whereas among the gel-only groups, women randomized to use Acidform had somewhat higher compliance than those using HEC. Interpretation of the differences in adherence by randomization arm is challenging because of the study's small sample size and short follow-up period; we unfortunately did not have sufficient power to explore adherence differences by individual randomization arm, though we did control for study arm in all multivariable analyses. One hypothesis is that continuous use led to greater discomfort with Acidform, because the diaphragm is placed over the cervix, holding the Acidform gel directly onto sensitive cervical tissue, whereas gel-only users were exposed to intravaginal Acidform use, with gel dispersed over a larger surface area but in lower concentration throughout the vagina. However, participant's self-reports in the trial do not bear this out (reports of discomfort were approximately equal among users of Acidform with the diaphragm and Acidform alone).11 Whatever the reason, variation in adherence between arms in the planned trial would have substantial implications for intent-to-treat analyses, because differences in STI rates between randomization arms could be due to the effect of the experimental products or simply because of differences in adherence.
Limitations and Strengths
Two intentional design aspects of this small pilot study—small sample size and short follow-up time—were nevertheless limitations. For example, as indicated earlier, we did not have adequate power to examine adherence by individual randomization arm; we similarly lacked the ability to explore adherence by partner type. Other predictors of adherence might have emerged with a larger sample. Our multivariate analyses attempted to minimize this sample size limitation by initially dividing variables into separate domains and assessing the individual and joint impact of each variable within these domains. This approach, however, could fail to detect confounding relationships across domains because some factors were dropped at the domain level and not taken forward to the combined model. The pattern of increased adherence with increased follow-up time may not persist past the short follow-up of this pilot. An additional limitation is that sexual behavior data, including measures used to calculate adherence, were self-reported and may suffer from recall or social desirability biases.
These limitations are offset by several strengths. First, no previous study has explored factors associated with adherence for a combination microbicide-diaphragm prevention package. Second, participants were randomized, reducing the likelihood of baseline imbalance in the distribution of characteristics that may be associated with adherence. Third, our analysis methods allowed proper computation of associations given repeated measurements on participants. Fourth, because FSW behavior has been documented to vary by partner type,24 we collected predictor data separately for casual and regular partners. Finally, the FSWs in this study are a highly relevant population group: they are at extremely high STI risk, yet have reduced ability to negotiate condom use.
The goal of this analysis was to identify factors that could be used to maximize adherent use of experimental study products in an upcoming effectiveness trial. Some variables that we hypothesized might be associated with product use were not related to adherence, including education, socioeconomic status, and vaginal hygiene practices. However, several factors were associated with increased adherence and will be integrated into counseling scripts and trial protocols.
At least 16 candidate vaginal microbicides are currently in various stages of human testing. A recent clinical trial among women in Zimbabwe and South Africa indicated that a prevention package of the diaphragm, lubricant gel, and condoms was no more effective than condoms alone at preventing HIV or STI acquisition,25,26 and differential adherence between the randomization arms may have masked a treatment effect.25 Finding an effective female-controlled method is clearly an important research priority, and exploring which characteristics make women choose to use it may be just as important.
1. Moench TR, Chipato T, Padian NS. Preventing disease by protecting the cervix: The unexplored promise of internal vaginal barrier devices. AIDS 2001; 15:1595–1602.
2. Hardy E, Hebeling EM, de Sousa MH. Devices for the administration of a vaginal microbicide: Use difficulties, adherence to use and preferred device. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract PC23.
3. Wandiembe SP, Ssali A, Basajja V, et al. Potential barriers to adherence to product use and cohort retention in microbicides efficacy trials. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract OB9.
4. Kasse MJ, Wandiembe S, Ssali A, et al. Acceptability and adherence to microbicide gel use instructions among couples participating in a pilot study for a phase III microbicide trial in rural SW Uganda. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract PC28.
5. Phillip JL, Morar N, Ramjee G. Assessment of informed consent in the Microbicide Development Programme Feasibility Study, Durban, South Africa. Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract PC62.
6. Mirembe BG, Mirembe F, Nakabiito C, et al. Strategies for recruitment, retention and adherence to study procedures for microbicide Phase III trial – Uganda. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract PB21.
7. Dladla-Qwabe N, Ntumba N, Govinden R, et al. Women's perception to being randomised to a gel or no gel arm: Implications for adherence to product intervention. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract OC13.
8. Magagula DP, Khoza L, Morar NS, et al. Role of men in ensuring adherence and compliance to microbicide gel use. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract OC17.
9. Toriesen K, Chanza H, Masoo W, et al. Male involvement may improve retention and gel adherence among women participating in a microbicide trial: Lilongwe, Malawi, HPTN 035. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract OC18.
10. Walaza S, Moyes J, Nanoo A, et al. Does participant disclosure of microbicide gel use to partner affect gel use? A pilot study in preparation for a randomized placebo controlled trial of a vaginal microbicide. In: Microbicides; 24–26 April 2006; Cape Town, South Africa. Abstract PC85.
11. Behets FM, Norris Turner A, Van Damme K, et al. Vaginal microbicide and diaphragm use for infection prevention research: A randomized acceptability and feasibility study among high-risk women in Madagascar. Sex Transm Dis 2008; 35:818–826.
12. Behets F, Norris Turner A, Van Damme K, et al. Acceptability and feasibility of continuous diaphragm use among sex workers in Madagascar. Sex Transm Infect 2005; 81:472–476.
13. Feldblum P, Hatzell T, Van Damme K, et al.
Results of a randomized trial of male condom promotion among Madagascar sex workers. Sex Transm Infect 2005; 81:166–172.
14. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13–22.
15. Zeger SL, Liang KY, Albert PS. Models for longitudinal data: A generalized estimating equation approach. Biometrika 1988; 44:1049–1060.
16. McNutt LA, Wu C, Xue X, et al. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 2003; 157:940–943.
17. Zou G. A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004; 159:702–706.
18. Greenland S. Interpretation and choice of effect measures in epidemiologic analyses. Am J Epidemiol 1987; 125:761–768.
19. Wacholder S. Binomial regression in GLIM: Estimating risk ratios and risk differences. Am J Epidemiol 1986; 123:174–184.
20. Harrell FJ. Predicting Outcomes: Applied Survival Analysis and Logistic Regression. Charlottesville, VA: University of Virginia, 2000.
21. Sun GW, Shook TL, Kay GL. Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis. J Clin Epidemiol 1996; 49:907–916.
22. Ingram DM, Miller WC, Schoenbach VJ, et al. Risk assessment for gonococcal and chlamydial infections in young children undergoing evaluation for sexual abuse. Pediatrics 2001; 107:73–79.
23. Hosmer DW Jr, Lemeshow S. Applied Logistic Regression, 2nd ed. New York, NY: Wiley, 2000.
24. Kayembe PK, Mapatano MA, Busangu AF, et al. Determinants of consistent condom use among female commercial sex workers in the democratic republic of Congo: Implications for interventions. Sex Transm Infect 2008; 84:202–206.
25. Padian NS, van der Straten A, Ramjee G, et al. Diaphragm and lubricant gel for prevention of HIV acquisition in southern African women: A randomised controlled trial. Lancet 2007; 370:251–261.
© Copyright 2009 American Sexually Transmitted Diseases Association
26. De Bruyn G, Shiboski S, Ramjee G, et al. The diaphragm and lubricant gel for prevention of cervical sexually transmitted infections: Results of a randomized controlled trial. In: International Society for Sex Transm Dis Research (ISSTDR) annual meeting; 29 July–1 August 2007; Seattle. Abstract 0–5.