Adherence to Diaphragm Use for Infection Prevention: A Prospective Study of Female Sex Workers in Kenya

Objective. To assess adherence to and acceptability of the diaphragm among 140 female sex workers in Kenya in a 6-month prospective study. Methods. At baseline and bimonthly visits, participants were interviewed on diaphragm knowledge, attitude, and practices. We used principal component analysis and logistic regression to identify predictors of consistent use. Results. At 50% of 386 bimonthly visits, women reported consistently using a diaphragm with all partners during the preceding 2 weeks. Consistent use was significantly higher at the 6-month than the 2-month visit. Women reported less covert use with “helping” (regular sex partners to whom she could go for help or support) than with “other” partners. Perceptions that diaphragms are easier to use than condoms and that their lack of coital interruption is important were associated with consistent diaphragm use with both partner types. Partner support of diaphragm use is correlated with consistent use with “helping” partners only while higher parity, consistent condom use, and perceived lack of need of condoms as a benefit of diaphragms were associated with consistent use with “other” partners. Conclusions. Diaphragm acceptance among female sex workers in Nairobi was high. Future studies should distinguish between partner types when evaluating diaphragm adherence.


Introduction
Cervical barriers such as diaphragms may protect women against the acquisition of sexually transmitted infections (STIs). Diaphragms could prevent N. gonorrhoeae and C. trachomatis from reaching the target columnar epithelial cells in the cervix [1] and protect against HIV acquisition by reducing the risk of the virus reaching HIV receptor-rich cells, which are present in larger numbers in the cervix than in the vagina [2][3][4][5][6]. In addition, diaphragms may reduce HIV susceptibility indirectly by protecting against other STIs [7,8] and could be an effective mechanism with which to deliver microbicide in the genital tract if a microbicide effective against HIV were identified [9]. Although diaphragms have not been shown to be effective against HIV acquisition in a randomized controlled trial [10], observational data suggest that consistent diaphragm use could protect against STIs and their long-term sequelae [11].
Even if diaphragms were proven to be efficacious for STI prevention and safe for use in diverse populations, they would still have to be acceptable to potential users in order to be used consistently. Although diaphragm 2 Infectious Diseases in Obstetrics and Gynecology use has been shown to have a high rate of acceptability among women, many factors could influence adherence to diaphragm usage instructions, including the attributes of specific types of diaphragms (e.g., their reusability), how they are used (e.g., when they need to be inserted and removed, whether they are used with a microbicide gel, and whether users are instructed to refrain from concomitant douching or intravaginal cleansing), and partner knowledge of their use [12][13][14][15][16][17][18][19][20][21]. Although diaphragms are "female-controlled" devices, partners' attitudes toward their use could affect women's adherence to consistent usage. Covert diaphragm use, especially with regular partners, can be difficult to maintain long-term, and unintended discovery of diaphragm use could cause conflict between women and partners who object to the use of diaphragms [22,23]. We conducted a 6-month prospective study to measure the prevalence of consistent diaphragm use and to identify determinants of consistent use among female sex workers in Kibera in Nairobi, Kenya.

Study Procedures.
The study was conducted from 2004 to 2005 among female sex workers (i.e., women who reported trading sex for money or gifts in the prior 2 weeks), 18-57 years of age, in the Kibera settlement in Nairobi, Kenya who had previously participated in a randomized trial of monthly chemical prophylaxis for STI control and prevention [24]. Findings on the safety of diaphragm use and changes in condom use in this population will be reported separately. Women were ineligible for the study if they reported consistent condom use during the previous 2 months or if they had a current pregnancy, gynecological abnormalities precluding diaphragm use, latex allergies, a partner with latex allergies, or a history of toxic shock syndrome.
At baseline, participants underwent a pelvic examination and were fitted for the silicone Milex Wide Seal diaphragm (Milex Inc., Chicago, IL). After being counseled on diaphragm use, women practiced inserting and removing their diaphragm under a clinician's observation. They were instructed to apply 1-2 teaspoons of K-Y Jelly (Johnson and Johnson; New Brunswick, New Jersey) to the cup and rim of the diaphragm for lubrication before inserting it and to keep the diaphragm in their vagina for at least 6 hours after coitus (without internal vaginal cleansing) but not for more than 24 hours. Women were given two diaphragms, a supply of K-Y Jelly, and male condoms.
At follow-up visits scheduled 1 week and 2, 4, 6 months after the baseline visit, women were counseled on safer sex practices and instructed to have their partners use a male condom during every coital act. They were counseled on the limited contraceptive effectiveness of diaphragms when used without spermicide and the unknown effectiveness of the diaphragm against STI acquisition. At all visits, they were tested for pregnancy and urinary tract infection and were treated syndromically for STIs and reproductive tract infections. The pelvic examination was repeated at the 6month visit.
Women were interviewed at the baseline visit and at the 2-, 4-, and 6-month follow-up visits to collect demographic data and data on their sexual behaviors; adherence to diaphragm use; and knowledge, attitude, and practices regarding diaphragms (as well as their perceptions of their partners' attitudes about diaphragm use). The questionnaires distinguished between diaphragm use with a "helping" partner and with all "other" partners. These partner types were established based on earlier formative research with the target population. A "helping" partner was defined as a regular sex partner to whom a woman could go for help or support if needed. If women had multiple partners in the past two weeks, they were asked to consider their main "helping" partner and their most recent "other" partner while answering the questionnaires. At all follow-up visits, participants completed a self-administered questionnaire about the frequency of their diaphragm, gel, and condom use for all acts with partners in the previous 2 weeks.

Statistical
Analysis. SAS 9.1.3 (SAS Institute, Cary, NC) was used for data analysis. Consistent use was defined with a categorical variable that measured diaphragm use (with or without K-Y Jelly) for all coital acts in the previous 2 weeks and analyzed separately for "helping" and "other" partners. The analyses were restricted to follow-up visits in which women reported at least one coital act in the preceding 2week period with the relevant partner type. The adjusted Wald method 24 was used to calculate 95% confidence intervals for estimates of the percentage of women who used a diaphragm consistently, and chi-squared tests were used to assess differences between the percentages of women who reported consistent diaphragm use and covert diaphragm use at their 2-month visit and the percentages who did so at their 6-month visit.
In separate principal component analyses, we assessed the extent to which 25 and 26 factors were related to diaphragm use with "helping" partners and "other" partners, respectively. The association of each factor with diaphragm use was based on participants' responses to statements about the diaphragm, which originally were measured with a three-item Likert scale; however, we recoded the responses for our analysis as "agree" versus "neutral or disagree." Principal component analysis is a method used to collapse a large number of partially correlated variables into fewer uncorrelated factors, each representing the combination of two or more of the initial variables [25]. This decreases the risk for type I errors (by reducing the number of statistical tests performed) as well as the risk for type II errors. Use of such a summary risk factor instead of multiple correlated items tends to produce stronger and more precise measures of association when a true association with the outcome of interest exists [26].
The derived variables then can be used in other analyses; for example, they can be regressed against the outcome of interest. We used the principal axis method to extract the components and the varimax method to produce orthogonal (uncorrelated) components. We based the decision to retain components on eigenvalues greater than 1 and the location of the break on the scree test. Factors that load for a given analysis are those that were given substantial weight when the principal component was constructed. We considered a factor to load for a component if the factor loading was at least .40 for the given component and less than .40 for the other components in the rotated factor pattern. The SAS procedure used for this analysis automatically standardized the component scores to unit variance. We retained two components in our subsequent analysis of factors related to diaphragm use with "helping" partners and with "other" partners. Tables 2 and 4 show the factors that loaded on each of the components. The communality estimate expresses the percentage of variance in the original variables that is accounted for by the retained principal components.
To identify determinants of consistent diaphragm use, we used logistic regression with generalized estimating equations, with the exchangeable working correlation matrix specified to adjust for correlation of standard errors between multiple records from one participant. After fitting individual models to perform bivariable analyses, we then fit a full model with all variables for the multivariable analysis. Using manual backward elimination, we removed factors that did not predict consistent diaphragm use based on an alpha of 0.05. Time-independent factors that were assessed as predictors of consistent diaphragm use were measured at baseline; these factors included age (≤27, 28-34, ≥35 years), marital status (never married and not cohabiting versus cohabiting, divorced, or widowed), highest educational level completed (<8 versus 9-12 years), ethnicity (Kikuyu versus other), weekly income (≤9 versus >9 USD), parity (0-1 versus ≥2 children), perceived importance of preventing pregnancy (not at all versus moderately or a lot), worry about pregnancy (not at all versus moderately or a lot), and worry about HIV (not at all versus moderately or a lot). Timedependent factors describing behaviors in the preceding 2 weeks were measured at the bimonthly visits, these included number of sex partners (1-5 versus ≥6), number of new sex partners (0, 1-2, ≥3), coital frequency with all partners (0-5, 6-15, ≥16 acts), any condom use with "helping" or "other" partners (yes versus no), a variable for the bimonthly (2, 4, or 6-month) visit, the components constructed in the principal component analysis; and the variables that did not load on the components created in the principal component analysis.
Only women who gave written, informed consent participated. Ethical review committees at the University of Nairobi, the University of Washington, the University of California, San Francisco, and the US Centers for Disease Control and Prevention approved the research.

Study Population.
Of the 180 women who were screened, 140 met the eligibility criteria and were enrolled in the study. Most women completed the 2-month (n = 134), 4-month (n = 130), and 6-month visits (n = 126). Overall, 138 (99%) of women enrolled in the study completed at least one of the bimonthly visits, and 121 (86%) completed all three bimonthly visits. Study attrition was due to pregnancy (n = 8), moving (n = 3), and loss to follow-up (n = 3). Analyses of diaphragm use with "helping" partners were based on data collected during 313 bimonthly visits by 121 women who reported coitus with a "helping" partner during at least one bimonthly visit. Analyses of diaphragm use with "other" partners were based on data collected during 362 bimonthly visits by 135 women who reported at least one act with any "other" type of partner during at least one bimonthly visit. Analyses of diaphragm use with all partners were based on data collected during 386 bimonthly visits by 138 women who reported at least one act with either partner type during at least one bimonthly visit. Participants' median age was 30 years (range, 18-55); 73% were cohabiting, divorced, or widowed; 81% had less than 8 years of education (Table 1). Most participants (63%) reported having previously used condoms for contraception. Only one woman reported having ever used a diaphragm before.

Adherence and Covert
Use. Women reported that they used a diaphragm with all partners during the preceding 2 weeks at 50% of the bimonthly visits, with all "helping" partners at 59% of the visits, and with all "other" partners at 64% of the visits. Among the subset reporting at least one act with the relevant partner type during the bimonthly visits, 41%, 45%, and 33% of women reported consistent diaphragm use with "helping," "other," or all partners, respectively. The percentage of women who reported consistent diaphragm use increased over the course of the study (Figure 1). The proportion of women who reported consistent diaphragm use was significantly higher at the 6-month visit than the 2-month visit for coital acts with "helping" partners (P = .04) and with "other" partners (P = .04) but not for coital acts with all partners (P = .09).
Results of analyses limited to visits by women who reported at least some diaphragm use with the relevant partner type showed that women reported always using the diaphragm without their "helping" partners' knowledge at 55% of the bimonthly visits and without their "other" partners' knowledge at 76% of the visits. The percentage of women who reported consistent covert use of the diaphragm during the preceding 2 weeks did not vary significantly  between the 2-and 6-month visits either with "helping" partners (58% and 51%, resp.; P = .56) or with "other" partners (77% and 73%, resp.; P = .82).

Diaphragm Use with "
Helping" Partners. Principal component analysis yielded two components associated with the use of a diaphragm with a "helping" partner ( Results from the unadjusted analysis showed that five variables were associated with reporting consistent diaphragm use with a "helping" partner in the prior 2 weeks: the time of the follow-up visit (6-month versus 2month visit), perceived partner support of diaphragm use, attitudes toward diaphragm attributes, the perception that using a diaphragm with gel use is at least as easy as using a condom, and the perceived importance of diaphragm use not interrupting sex (Table 3). In the multivariable analysis, however, the associations remained significant for only three variables: perceived partner support of diaphragm use (adjusted odds ratio [OR], 1.4; 95% confidence interval [CI], 1.1-1.7), the perception that using a diaphragm with gel use is at least as easy as using a condom (adjusted OR, 2.0; 95% CI, 1.2-3.1), and the perceived importance of diaphragm use not interrupting sex (adjusted OR, 2.8; 95% CI, 1.1-7.1).

Diaphragm Use with "Other" Partners.
Results from the principal component analysis showed two components significantly associated with diaphragm use with "other" partners ( Table 4). The first was a composite of factors concerning women's perceptions of whether their "other" partners supported their diaphragm use, including whether diaphragm use would cause their "other" partners to become mad, angry, argue with her, refuse to have sex, or hit or beat her. The second component was based on the same four factors as the second component in the analysis of diaphragm use with "helping" partners.
Nine variables were related to consistent diaphragm use with "other" partners in the previous 2 weeks ( Table 5). Four of these variables, however, were not associated with consistent diaphragm use in the adjusted model: study follow-up visit (6-month versus 2-month), number of recent sex partners, number of coital acts with all partners in the previous 2 weeks, and the second component. Only the following five variables were associated with consistent diaphragm use with "other" partners in the multivariable model: having ≥2 children (adjusted OR, 2.1; 95% CI, 1.1-4.0), using condoms consistently (adjusted OR, 2.1; 95% CI, 1.2-3.7), perceiving diaphragm use to be easier than condom use (adjusted OR, 2.3; 95% CI, 1.4-3.8), perceiving diaphragm use not interrupting sex as being important (adjusted OR, 2.2; 95% CI, 1.0-5.0), and perceiving the lack of need for condoms as a benefit of diaphragm + gel use (adjusted OR, 2.3; 95% CI, 1.3-4.2).

Reasons for Not
Using a Diaphragm. The most common explanation that women gave for not using a diaphragm was that they used condoms instead; 75% of the women who reported inconsistent diaphragm use during the previous 2 weeks cited this reason one or more times (Table 6). This explanation was provided more often to explain not using a diaphragm with "other" partners (82%) than with "helping" partners (38%). Women often cited "trust" and "knowing partners well enough not to use it" as reasons for not using diaphragms with "helping" partners but rarely given cited either as a reason for not using a diaphragm with "other" partners. Another frequently cited reason for not using a diaphragm was that sex was unexpected or that women did not have the diaphragm with them (cited by 30% as a reason with "helping" partners and by 53% as reasons with "other" partners). In contrast, negative attributes of the diaphragm (e.g., participant discomfort) were rarely reported as reasons for not using it.

Discussion
A substantial proportion of participants reported consistent diaphragm use in the prior 2 weeks with "helping" and "other" partners (59% and 64% of follow-up visits, resp.). Diaphragm use with each partner type increased over the course of the study. With one exception [12], results from studies of diaphragm use for infection control have shown little change over time in the percentage of women who reported consistent use during study visits [15,17,19,21]. The increase that we found in the percentage of women who used diaphragms consistently might be the product of a "learning curve," whereby women learned to use the device more consistently or became more comfortable with its use over time. Because of the low attrition rate in the study, the increase in the rate of consistent diaphragm use is unlikely to be attributable completely to early study discontinuation by women who did not like the diaphragm.
The perceptions that diaphragms are easier to use than condoms and that lack of interruption to sex is important were associated with consistent diaphragm use with both partner types. The association of other factors with consistent diaphragm use, however, differed by partner type. Women's perception of the degree to which their "helping"  partner would support diaphragm use (measured by the principal component analysis) was associated with recent consistent diaphragm use with this partner but not with "other" partners. This difference by partner type could be related to a higher rate of awareness of diaphragm use among "helping" than among "other" partners. Women might have felt a responsibility to inform "helping" partners about their diaphragm use or have judged covert use to be difficult to sustain long term with this partner. They also might have been more concerned about the consequences of a "helping" partner discovering their covert diaphragm use. On the other hand, condom-related factors (reporting consistent condom use and perceiving the lack of need for condoms as a benefit of diaphragm use) were associated with consistent diaphragm use with "other" partners but not with "helping" partners. Overall, our results suggest that future studies of the effect of diaphragm promotion on condom use should account for partner type.
The study's main limitation was its reliance on selfreported data, which has potential for reporting bias. For example, we cannot eliminate the possibility that the apparent increase in diaphragm use over the course of the study resulted from a change in the validity of self-reporting. Women might have overreported consistent diaphragm use more often as the study progressed if they developed closer rapport with the study staff and, thus, had a stronger motivation to "please" staff by reporting adherence to their instructions. Similarly, the association between consistent condom use and consistent diaphragm use with "other" partners could be spurious if the perceived social desirability of using both products caused some women to overreport their use of them. Another limitation is that because the study population consisted of female sex workers in Kibera, Nairobi who did not consistently use condoms and who were willing to try the diaphragm, study findings might not be generalizable to other populations. Finally, just as intention to use study products does not necessarily predict use [27,28], the level of adherence to consistent diaphragm use (as well as factors related to adherence) among participants in this study might be different in nonstudy settings or be impossible to sustain for periods longer than 6 months.
Although study participants were advised repeatedly of the unknown effectiveness of the diaphragm against STI acquisition and, when used for this purpose, against pregnancy, qualitative data from the exit interview (data not reported here) suggest that women were reluctant to return the diaphragm at the end of the study because they believed that it protected them from acquiring STIs. This finding has been observed elsewhere [14] and highlights the need for better methods for ensuring participant comprehension of key points during research participation. The quantitative measures of participant concern about HIV acquisition (as well as concern about pregnancy and the perceived importance of preventing pregnancy) were not associated with consistent diaphragm use.
Reasons cited for not using the diaphragm with "other" partners included having unexpected sex or not having the diaphragm available. Continuous diaphragm use (except during daily removal for cleaning) could address this barrier. By making its use independent of coitus, continuous use Infectious Diseases in Obstetrics and Gynecology of the diaphragm could increase the proportion of sex acts protected by the device and thus improve its effectiveness [29]. However, little research has been conducted on continuous diaphragm use [30], and the safety and effectiveness of such use would need to be proven in future studies before continuous use of diaphragms could be promoted.
Other research has demonstrated that relationship factors influence diaphragm use [31]. To our knowledge, though, this was the first prospective study to evaluate adherence to diaphragm use by partner type. We used only two classifications for partner type; a more refined understanding of partners might have yielded different results. Nevertheless, the differences we found in determinants of consistent diaphragm use during sex with "helping" partners and during sex with "other" partners suggest that future research should also distinguish between different types of partners when evaluating adherence to diaphragm use. Tailoring interventions and counseling messages based on women's types of partners could perhaps help improve efforts to promote consistent diaphragm use.