Article Text
Abstract
Objectives The purpose of the study was to obtain a better understanding of the relative importance of personal factors, male partner factors and situational factors, in determining condom breakage in a population of female sex workers (FSWs) in Bangalore.
Methods The authors conducted a cross-sectional study that included a face-to-face interview and condom application test, with 291 randomly selected FSWs in Bangalore, India, in early 2011.
Results Ninety-seven per cent of respondents noted condom use at last sex; 34% reported a condom breakage in the last month. Combining individual, situational and partner aspects of condom breakage into one logistic regression model and also controlling for client load, the authors found that partner and situational factors were dominant since the only significant predictors of condom breakage included being a paying client (adjusted odds ratio 4.61, 95% CI 1.20 to 17.58, p=0.025), the condom being too small for the penis (adjusted odds ratio 2.29, 95% CI 0.97 to 5.40, p=0.056) or too big for the penis (adjusted odds ratio 4.29, 95% CI 1.43 to 12.80, p=0.009) and rough sex (adjusted odds ratio 6.39 CI 3.55 to 11.52, p<0.001).
Conclusions Condom use among Bangalore FSWs is now very high. However, condom breakage is still a not uncommon event and puts women and their clients at unnecessary risk of infection. It may be difficult to eliminate the problem completely, but every effort should be made to discuss with sex workers the findings of this survey that point to possible personal markers of risk seen in the univariate analysis and to highlight the importance of avoiding rough sex and of ensuring the condom fits the client.
- Condoms
- female sex workers
- India
- condom breakage
- HIV
- sexual health
- AIDS
- epidemiology (general)
- epidemiology
- sexual behaviour
- STD control
- statistics
- adolescent
Statistics from Altmetric.com
- Condoms
- female sex workers
- India
- condom breakage
- HIV
- sexual health
- AIDS
- epidemiology (general)
- epidemiology
- sexual behaviour
- STD control
- statistics
- adolescent
Introduction
Condoms are an important tool in the fight against sexually transmitted infections, if properly used.1–4 There have been many reports of factors associated with condom breakage, although few are from less developed countries or in commercial sex settings, and few studies have examined breakage in a comprehensive fashion. Poor technique, incorrect handling, lack of experience and other personal characteristics appear to be important determinants of condom breakage.5–11 Also important are the fit of the condom,6 12–14 the duration of the sex act,15 the roughness of the sexual episode,5 16 the use of drugs and alcohol17 and erectile dysfunction.18 Lack of lubrication or use of oil-based lubrication have been also found to be associated with condom breakage.19 20 There is conflicting evidence of whether double condoms reduce or facilitate breakage.21 Among female sex workers (FSWs), women who have recently experienced sexual violence appear to be at greater risk for condom breakage.17 22
A recent study in four districts of Karnataka found that condom breakage in the previous month among FSWs was associated with young age (under 20 years), being divorced/separated/widowed, frequent alcohol use, entertaining clients in lodges/rented rooms or brothels (rather than at home or in public places), anal sex, inconsistent condom use and having never seen a condom demonstration.23 However, this study had no data on actual breakage events. The purpose of the current study was to obtain a better understanding of the relative importance of personal factors that might be markers of risk, and situational factors (the sex act itself and male partner factors), in determining condom breakage in a population of FSWs in Bangalore.
Methodology
We conducted a cross-sectional study with FSWs in Bangalore city in early 2011. Bangalore was selected as the study area because it has over 20 000 FSWs, with a documented HIV prevalence in 2008 of 8.3%.24 The sex worker ‘hotspots’ (main solicitation locations) have been mapped in these areas, and from these data and using a randomised time location cluster design, we sampled 65 clusters in approximately half the city where a sex worker collectively called Swati Mahila Sanga manages an HIV interventions programme, using a standard methodology.25 In each cluster, we attempted to recruit five women, for a total sample size of 325 women (based on an expected condom breakage rate of 10%±3.5%, in the past month and ∼90% response rate, for a final sample of 300).
As the first part of a two-phase study (the second phase being a prospective telephone survey, not reported here), we conducted quantitative face-to-face interviews that included questions on all the parameters shown in figure 1 and parameters that have been found to be associated with breakage in other studies. Parameters were grouped into either the respondents' personal characteristics, which might be seen as markers for breakage, or the circumstances of the specific sex acts discussed (including aspects of the male partner). The respondents were asked details of the last time that they used a condom that did not break and the last time (in the previous month only) when they used a condom that did break. Five supervisors and nine experienced investigators with various language skills were trained for 6 days in how to conduct the interviews. The interviews were conducted in early 2011 in a private confidential setting. The respondents were then given a penis model and condom and asked to demonstrate condom application. We developed a scoring system out of 10, based on five elements often noted in training materials (checking lubrication and expiry date, opening the packet carefully, isolating the teat properly, pinching the teat and unrolling fully onto the penis), and agreed the criteria with the observers for scoring each on a 3-point scale (0–2).
All data were entered into CSPro (US Census Bureau, Washington DC, USA) and then analysed using SPSS V.18 (IBM Corporation, Armonk, NY, USA) and STATA V.10 (STATA Corporation, TX, USA). Sample cluster weights were added to the data to correct for different selection probabilities and non-response in each primary sampling unit.
In total, 324 eligible women were approached to participate, of which 25 refused. After a tip-off, and subsequent questioning, six women admitted to the team that they had come along with friends and pretended to be sex workers in order to get a phone that they had heard was to be provided in the second phase of the study: we then excluded them from the analysis. Thus, a total of 293 women participated (90.4% response rate). Two of the women reported never using a condom, so our analysis focuses on 291 women.
We first conducted univariate analyses on breakage and the respondents' individual characteristics (291 women); then we compared the situational/partner variables during the last episode of condom use when the condom did not break (all 291 respondents) with the situational/partner variables during the last episode of condom use in the last 1 month when it did break (99), a total of 390 episodes. In the logistic regression models, we included only the variables that had a p value <0.1 in univariate analysis and controlled for weekly client load. We fitted three different models: in the first model, we used a logistic regression analysis that included only the characteristics of the respondent (n=291 women). The second model included only the situational/specific partner variables (n=390 condom use episodes) and used random effects intercept multi-level logistic regression with a random intercept only for the ‘individual woman’ level. The third model used multi-level logistic regression and included both women's and condom use episode characteristics; the sex worker characteristics were regarded as level 2 and episodes of condom use as level 1 for the 390 episodes of condom use since we had two episodes of condom use nested within the 99 sex workers who experienced breakage.
The study was approved by the Institutional Ethical Review Board of St John's Medical College and Hospital, Bangalore, India. All respondents gave witnessed verbal consent. Lists of names/telephone numbers were kept securely and destroyed at the end of the study. Participants were reimbursed for travel costs and lunch only in this phase of the study.
Results
Individual characteristics
Respondent socio-demographic and sex work profile
Respondent ages ranged from 18 to 49 (mean 29.5 years). Just under half were illiterate, and more than half were married/cohabiting, with 39% divorced, widowed or separated. Sexual debut was at a mean age of 16.9 years, and sex work began at a mean age of 24.8 years. Client volume varied, from 1 to 12/day (mean 2.9), 1 to 7 days/week (mean 3.7) and from 1 to 48 clients/week (mean 10.5) (table 1).
Risk behaviours and programme exposure
Sixty-four per cent of respondents reported they never drank alcohol, although one-quarter did so at least once a week. Overall, 70% of respondents reported being asked for anal sex sometime (mean times 3.5/month), whereas only 18% reported they had ever had anal sex. Almost one-fifth said they had been beaten or forced into sex in the previous year; almost 9% had been arrested at some time. Almost all the respondents had seen a condom demonstration at some time, three-quarters in the previous 3 months (table 1).
Condom use
All but two of the respondents had used a condom at some time, and of these, 97.0% reported using one at last sex with a client. The majority of respondents noted that they had at some time experienced condom breakage (61%), and 99 women (34%) said this had happened accidentally at least once within the last month, with 12% experiencing a breakage at last use. Of those who had experienced a break in the previous month, one-quarter experienced it once, 40% experienced it twice and the rest more than twice. Approximately 10% of women reported having a breakage weekly and 16% at least monthly. However, almost one in five said that breakage was a rare event and 40% reported it had never happened to them (table 1).
Condom application test
In total, 286 women agreed to participate in the condom application test and 17 (5.9%) broke the condom while trying to apply it on the model. Twenty per cent gained full marks for checking the expiry date and integrity of the packet, 53% for opening the packet properly, 48% for holding the condom properly, 20% for pinching the teat correctly and 42% for unrolling the condom properly onto the penis model. Seventeen per cent of respondents unrolled the condom completely before applying it. Of the total possible score of 10, 49% scored between 1 and 5 and 51% scored between 6 and 10 points.
Situation and partner characteristics
Table 2 shows the details of the last condom sex (partner and circumstances). Most sex was with a paying client and was almost all vaginal sex. One-third of the sex acts was described as rough sex, with 5% also being accompanied by threats. The men were generally reported to be sexually experienced and familiar with condom use, but the condom was reportedly unrolled before placing it on the penis 13% of the time. The male client put the condom on himself 61% of the time and provided the condom in just over one-third of cases. Approximately 20% of sex acts involved double condom use. The respondents assessed the size of the condom vis-à-vis penis size. About 83% were thought to be a good fit, though 10% were deemed too small and 6% too big. About half the men were reported to have had overly ‘hard’ erections.
Univariate analyses of variables associated with breakage
We examined the association between individual FSW characteristics listed in figure 1, with experiencing at least one breakage in the 1 month before the survey (table 3). In the univariate analysis, younger age, being illiterate, entertaining clients in lodges or rented rooms, having sexual debut and first commercial sex at a young age, using alcohol regularly and having a history of forced sex and arrest were significantly associated with reporting condom breakage. Having a history of breakage was also associated with breakage in the last month, for example, 91% of those who said they routinely had breaks once a month and had a break in the index month (p<0.001) (data not shown). However, this was excluded from the regression analysis as we felt that respondents would be biased towards answering positively to this question if they had had a break in the index month. Other characteristics, including the condom application test score, were not associated with breakage.
The respondents were asked to describe the circumstances around the last sex act with a condom when it did not break (n=291 episodes) and the last time that they had sex with a condom in the last month when it did break (n=99 episodes), a total of 390 episodes. Table 4 shows the partner and situational factors on univariate analysis significantly associated with having a break in the previous month. Important factors associated with breakage were the sex partner being a paying client, his use of alcohol, sex lasting longer than usual, the man having a very hard erection, the sex being furtive, the sex being rough and the condom being a poor fit. Other situational and partner factors were not significantly associated with breakage.
Regression models
In the first multivariable logistic regression model exploring the sex workers' individual characteristics, only regular use of alcohol, starting sex work before the age of 21, having <15 clients/week and having a history of forced sex remained significantly associated with condom breakage (table 5). In model 2 multi-level regression examining situational and partner variables, factors that remained significantly associated with breakage were the partner being a paying client, rough sex, sex lasting longer than usual, furtive sex and the condom being a poor fit. In model 3, combining individual, situational and partner aspects of condom breakage into one multi-level logistic regression model, we found that partner and situational factors were dominant since the only significant predictors of condom breakage included being a paying client, the condom being too small or too big and rough sex (table 5).
Discussion
Condoms are the mainstay of HIV prevention efforts among this population in Bangalore and their use is reportedly high.26 We wanted to understand why condoms sometimes break and thus jeopardise women's and client's safety. This multi-level study is the first one of its kind in this type of population: by combining personal markers of risk and actual circumstances around breakage, we were able to observe the overwhelming importance of the partners and the sex act itself. The key determinants were if the partner was a paying client, rough sex and condom fit. Many of the factors found to contribute to condom breakage in other studies were not found in this study, probably because we used a multi-level analysis that reduced the importance of personal markers and also because we were able to obtain many more situational parameters than other studies.
This study had certain limitations. First, the data were retrospective, with multi-partner women asked to recall unusual events that happened up to 1 month before; therefore, details might not always have been accurate. Recall might also have been tainted by their own, possibly erroneous perceptions of causes of breakage or by community rumours, for example, about poor quality government condoms. Our analysis was limited by having only one or two observations per woman. It is also possible that respondents associated the interviewers with the local HIV prevention project, thus exaggerated condom use, although rates are consistent with other studies.26
Our study found that rough sex was a key determinant of breakage, as found in the few other studies that have included it as a measure.5 22 Violence is common and well documented in these situations.27–29 Rough sex might not always be foreseen (except if the man has been drinking or if he appears violent); women reported that since they started agreeing to sex over the phone, they were sometimes surprised when then several men appeared and that then the sex was often rough. Women need to know how to extricate themselves from such a situation before sex or if they find themselves in a difficult situation and how to slow the client down.
The women in our study noted a problem with condom fit that sometimes led to breakage, an issue found in other studies.6 12–14 Recently, the BBC reported on a small study of penis size in Indian men that concluded that most condoms in India were too big for Indian men (http://news.bbc.co.uk/2/hi/6161691.stm). Another study in Kerala found large variations in erect penis sizes: penis length ranged from 10.5 to 17 cm and circumference from 9 to 13.5 cm.30 Our own measurement of the government Nirodh used by the majority of sex workers in most sex acts showed a non-stretched circumference of 10 cm and length of 20 cm, which would apparently be adequate for most men. However, it may be that the men with very short penises, in the 10–12 cm range, cannot completely unroll the condom and might have trouble keeping it on. On the other hand, men with large circumference penises may have problems stretching the condom rim enough to properly apply it, especially if in the dark or in a hurry, leading to rough handling and breakage. More quantitative and qualitative research is needed on this issue and on the need to have free condoms in different sizes.
It is difficult to know how to translate these findings into preventative actions, as avoidance of the men who appear to ‘cause’ the problem, either by having difficulties with condom fit or by engaging in rough sex, is not straightforward. One suggestion for now might be to ensure that sex workers understand the risks of breakage, learn skills that can slow men down and have condoms that can fit larger men available at all times for when they encounter extreme tumescence. More longitudinal prospective research is needed to validate these results and to obtain better estimates of the rates of breakage. In addition, modelling these results to estimate the risk of HIV infection attributable to condom breakage would be useful in terms of evaluating the level of effort required to address the problem. More qualitative research is also needed to help understand which men engage in rough and protracted sex and why and how sex workers can manage prolonged erections, minimise such encounters or reduce the harm therein.
In conclusion, condom use among Bangalore sex workers is now very high, with 97% reporting use at last sex with a client, and there is evidence that rates of HIV and sexually transmitted infections are declining26 as a result of their diligence in practising safer sex. However, condom breakage is still not an uncommon event and puts women and their clients at unnecessary risk of infection. It may be difficult to eliminate the problem completely, but every effort should be made to discuss these findings that point to possible personal markers of risk seen in the univariate analysis and that highlight the importance of avoiding rough sex and of ensuring the condom fits the client. The sex worker organisations must seek solutions to diminish these risks to the extent possible.
Key messages
It is important to understand the relative importance of personal factors, partners and situational factors, in determining condom breakage among female sex workers.
Condom breakage is still a not uncommon event and puts women and their clients at unnecessary risk of infection.
We found that partner and situational factors were dominant: being a paying client, the condom being too small or too big for the penis and rough sex.
More qualitative research is needed to understand condom fit and aspects of rough sex and how sex workers can manage this.
Acknowledgments
We would like to thank the study interviewers, staff of Swati Mahila Sangha and all the women who participated.
References
Footnotes
Funding This research was funded by the Bill & Melinda Gates Foundation. The views expressed herein are those of the authors and do not necessarily reflect the official policy or position of the Bill & Melinda Gates Foundation, grant number 33978.
Competing interests None.
Patient consent Verbal witnessed consent was used. This is a normal procedure for sex work research in India as many subjects are illiterate.
Ethics approval Institutional Ethical Review Board of the St John's Medical College and Hospital, Bangalore, India.
Provenance and peer review Not commissioned; externally peer reviewed.