WHEN USED CORRECTLY and consistently, male latex condoms are the most effective method to prevent sexually transmitted diseases (STDs), including human immunodeficiency virus (HIV), among sexually active couples.1 However, for pregnancy prevention other more effective methods are available.2 This trade-off leads to a dilemma of how to counsel women who are at risk for both pregnancy and STDs. One approach is to counsel women to use one method for contraception and a second method for STD prevention. Unfortunately, some women who are at risk for STDs either cannot or will not consistently negotiate the use of a male condom.3,4 Given these realities, and the evidence that nonoxynol-9 (N-9) can provide some protection against gonorrhea and chlamydia,5,6 the New York State Department of Health AIDS Institute has advocated hierarchical counseling and the use of N-9 as a backup to condoms7,8 despite disagreement among researchers.9–11 We evaluated whether offering a choice of barrier methods can increase overall barrier method use without decreasing condom use in women using oral contraceptives (OCs) for contraception.
We enrolled 167 sexually active OC users who were at least 18 years of age and who attended a publicly funded family planning clinic in Texas. Eligible women had to meet Catania's definition type B of being at risk for HIV (two or more partners in past 5 years, transfusion recipient, injection drug user, hemophiliac or risky sexual partner).12 Although no consensus exists on which criteria best predict risk in different settings,13,14 we chose a definition that had been proposed in the literature at the time of protocol development.
All participants received standardized hierarchical counseling about ways to protect against STDs from one of three female counselors or a female nurse (Table 1, first row). They were then randomized into two groups (Condom or Choice group). Women in the Condom group (N = 81) received three dozen condoms and were asked to use them during each act of intercourse during the next 2 months. Women in the Choice group (N = 86) received three dozen condoms and three dozen N-9 film along with further counseling (Table 1, second row). If they ran out, they were encouraged to return to the clinic for more free supplies. Counseling with clients at this clinic typically lasts about 45 minutes to 1 hour. Women who were enrolled in this study received about 1 to 2 hours of counseling and study directions, including detailed instructions on how to use the study products. The sessions with the Choice group tended to last slightly longer.
Participants kept daily coital diaries to record each act of intercourse and barrier method use. During each of the three 2-month follow-up visits, diaries were reviewed by study staff and transcribed onto data forms. Participants received $20 per completed follow-up visit. The study protocol and consent forms were approved by Family Health International's Protection of Human Subjects Committee.
Two outcome variables were calculated from coital diary data: (1) percentage of coital acts per woman protected by at least one barrier method and (2) percentage of coital acts per woman protected by condoms. The means of the two outcomes for each study group were compared using a repeated measures analysis of covariance (ANCOVA). Group by visit interaction, group effect at each visit, as well as overall group effect were assessed. For our final model, we chose covariates that were either (1) associated with the outcome per criteria outlined by Senn15 or (2) found to be imbalanced in the women returning for follow-up. Correlations among the chosen variables were calculated using a Spearman rho statistic to assess potential problems with collinearity, and no problems were found.
Most loss to follow-up occurred between enrollment and the first follow-up visit; the cumulative loss for all three intervals was 40% in the Condom group and 57% in the Choice group. These rates in loss to follow-up were large and seemingly dissimilar (P = 0.22). Thus, in an exploratory analysis, a three-state transitional model16 was used to evaluate the potential bias of the differential loss to follow-up. For each observed time period, participants were classified into one of three potential outcomes: (1) high compliance; (2) low compliance; and (3) loss to follow-up. The cutpoint for high/low compliance was ≥ than 50% of acts protected. Under a first order Markov assumption, the current time period compliance would only depend on the previous compliance. To evaluate the intervention effect adjusting for differential loss to follow-up and potential confounders, an ordinary multilogistic regression was used that included previous outcome as a predictor in the model. A detailed comparison of the transitional model to the marginal model is presented elsewhere.17
Randomization appeared to have worked fairly well with a few notable exceptions (Table 2). In the enrolled population (Condom group N = 81; Choice group N = 86), only two background characteristics were initially imbalanced (no condom use in the past 6 months and STD in the past 5 years). However, the differential loss to follow-up caused further imbalance among three background characteristics (STD in the past 5 years, OC user for more than 1 year, and married or in union).
The Choice group protected a higher mean percentage of acts with a barrier method throughout the three periods (month 1 to 2: 29% vs. 22%; month 3 to 4: 33% vs. 21%; and month 5 to 6: 35% vs. 19% (Fig 1). Standard deviations (SD) ranged from 37% to 41% for the Choice group and 32% to 34% for the Condom group. Three covariates were individually associated with the outcome (no condom use in last 6 months, education, and currently has an STD). We included these three, along with three other covariates that we considered to be meaningfully imbalanced in our final model (married or in union, used OC for more than 1 year, and had an STD in the past 5 years). After adjusting for these six covariates, overall barrier method use was significantly higher in the Choice group (adjusted p value = .012) (Table 3).
Mean percentage of acts protected with condoms was also higher for the Choice group owing to the combined use of condoms and N-9 film. This difference became more evident the longer the study progressed (month 1 to 2: 23% vs. 22%; month 3 to 4: 30% vs. 21%; and month 5 to 6: 33% vs. 19%, adjusted P = 0.036). SD ranged from 34% to 41% in the Choice group and 32% to 34% in the Condom group.
Owing to the skewness of the percentage of acts protected, we repeated the analysis with a square root transformation of the outcome variables (mean acts protected by any barrier method and mean acts protected by a condom). Results remained virtually unchanged.
The exploratory analysis with the transitional multilogistic regression approach showed that the differential loss to follow-up was a result of the Choice group participants with low barrier compliance in a previous interval not returning for the subsequent follow-up visit. Adjusting for this and the same covariates used in the repeated measures analysis, mean percentage of acts protected with any barrier method remains significantly higher (adjusted p value = 0.013) in the Choice group (Table 4). Mean percentage of acts protected by condoms was significantly higher as well (adjusted p value = 0.046).
In this convenience sample of family planning clients, providing choice of barrier methods improved overall compliance. We found that providing OC users at risk of STDs with N-9 film in addition to condoms, supported by appropriate hierarchical counseling, increased overall barrier method use without decreasing condom use. Unfortunately we had a relatively large loss to follow-up; this may be partially explained by a sizable proportion of the participants being students who left the area after the academic year. Because this large and differential loss to follow-up could influence the outcome measures, this study should be given the evidential weight of an observational cohort study rather than a successful randomized controlled trial.18
What influence could our differential loss to follow-up have had on our results? Two covariates were imbalanced from the beginning; no condom use in past 6 months was higher in the Choice group (56% vs. 44%) and STD in the past 5 years was higher in the Condom group (30% vs. 20%). It is plausible that the imbalance of these two covariates would make the Choice group less likely to use barrier methods during the study because they have not used condoms in the past and may consider themselves at lesser risk for STD. The differential loss to follow-up further accentuated the difference in one of the covariates (STD in the past 5 years: 31% vs. 6%) and made two further covariates imbalanced; the women remaining in the Condom group were more likely to be married or in union (36% vs. 27%) and more likely to have used OCs for more than one year (70% vs. 58%). It could be argued that more stable couples who used OCs for a long time are less likely to use barrier methods, thus this imbalance would downwardly bias the percentage of acts protected in the Condom group. Finally, it is plausible that women lost to follow-up would have lower compliance after discontinuing the study, thus making the Choice group appear to have higher barrier compliance than the Condom group.
Controlling for all collected potential confounders in a repeated measures ANCOVA, the Choice group had a statistically significant higher level of overall barrier method use. In addition, a transitional multilogistic regression approach produced similar results, providing some assurance that the measured effect of increased barrier method use among the Choice group is real. However, we cannot rule out that an imbalance in unmeasured covariates could have contributed to these results.
These findings can be compared with data from a three-arm, randomized study of Colombian sex workers.19 In that study, the group counseled to use spermicides as a backup when condoms could not be used (first intervention group) increased overall barrier method use slightly when compared with the group counseled only to use condoms (96.9% vs. 94.7%). However, the group significantly decreased condom use (78.1% vs. 94.5%). Their second intervention group received counseling similar to our intervention group. This counseling emphasized the use of both condoms and spermicides together and resulted in increased overall barrier method use (97.3% vs. 94.7%) with only a slight reduction in condom use (92.3% vs. 94.5%) in the Colombian study.
These two studies suggest that when clients are counseled to use two barrier methods concurrently, the concern that women will replace the proven effective condom with the potentially less effective spermicide can be minimized. Moreover, a choice of methods increased overall barrier use in these two studies. Additional barrier methods, including the female condom, need to be promoted in the method mix,20 and counseling needs to emphasize the relative effectiveness of the various methods.
In the end, the decision of what method(s) to use should be made by informed consumers. The provider's responsibility is to ensure that the client is fully informed about all the available options. Our data are encouraging because when given choices, women practice increasingly safer behaviors.
1. Cates W Jr, Stone KM. Family planning, sexually transmitted diseases and contraceptive choice: A literature update-part I. Fam Plann Perspect 1992; 24:75-84.
2. Hatcher RA, Trussell J, Stewart F, et al. Contraceptive Technology, 16th rev. ed. New York: Irvington Publishers, 1994.
3. Santelli JS, Davis M, Celentano DD, Crump AD, Burwell LG. Combined use of condoms with other contraceptive methods among inner-city Baltimore women. Fam Plann Perspect 1995; 27:74-78.
4. Weisman CS, Plichta S, Nathanson CA, et al. Consistency of condom use for disease prevention among adolescent users of oral contraceptives. Fam Plann Perspect 1991; 23:71-74.
5. Niruthisard S, Roddy RE, Chutivongse. Use of nonoxynol-9 and reduction in rate of gonococcal and chlamydial cervical infections. Lancet 1992; 339:1371-1375.
6. Louv WC, Austin H, Alexander WJ, Stagno S, Cheeks J. A clinical trial of nonoxynol-9 for preventing gonococcal and chlamydial infections. J Infect Dis 1988; 158:518-523.
7. New York State Department of Health AIDS Institute. Women and HIV Prevention: Methods of Personal Protection, a Policy Statement. Focus on AIDS in New York State, July 1992.
8. Cleary J, Winters S. Female condom: Efficacy, acceptability and relationship to the women's hierarchy of risk reduction. Division of HIV Prevention. New York State Department of Health AIDS Institute, January 1994.
9. Stein ZA. Editorial: The double bind in science policy and the protection of women from HIV infection. Am J Pub Health 1992; 82:1471-1472.
10. Rosenberg MJ, Gollub EL. Commentary: Methods women can use that may prevent sexually transmitted disease, including HIV. Am J Pub Health. 1992; 1473-1478.
11. Cates W Jr, Stewart FH, Trussell J. Commentary: The quest for women's prophylactic methods-hopes vs science. Am J Pub Health 1992; 82:1479-1482.
12. Catania J, Coates TJ, Stall R, et al. Prevalence of AIDS-related risk factors and condom use in the United States. Science 1992; 258:1101-1106.
13. Cates W Jr. A risk-assessment tool for integrated reproductive health services. Int Fam Plann Perspect 1997; 23:36-38.
14. Thomas T, Choudhri S, Kariuki C, Moses S. Identifying cervical infection among pregnant women in Nairobi, Kenya: Limitations of risk assessment and symptom-based approaches. Genitourin Med 1996; 72:334-338.
15. Senn S. Testing for baseline balance in clinical trials. Stat Med 1994; 13:1715-1726.
16. Diggle PJ, Liang KY, Zeger SL. Analysis of Longitudinal Data. New York: Oxford University Press, 1994.
17. Chen PL, Wong E, Dominik R, Steiner MJ. A transitional model for assessing barrier methods compliance adjusting for nonrandom loss to follow-up, presented at the Joint Statistical Meeting of the American Statistical Association, Anaheim, CA, 1997.
18. U.S. Preventive Services Task Force. Guide to clinical preventive services, 2nd ed. Baltimore: Williams & Wilkins, 1996:862.
19. Farr G, Castro LAA, Disantostefano R, Claassen E, Olguin F. Use of a spermicide and impact of prophylactic condom use among sex workers in Santa Fe de Bogota, Colombia. Sex Transm Dis 1996; 23:206-212.
20. Gollub EL, Stein Z, El-Sadr W. Short-term acceptability of the female condom among staff and patients at a New York City hospital. Fam Plann Perspect 1995; 27:155-158.
21. Novello AC, Peterson HB, Arrowsmith-Lowe JT, From the Surgeon General, US Public Service. Condom use for prevention of sexual transmission of HIV infection. JAMA 1993; 269:2840.