Surveys of condom use in women based on self-report may be seriously unreliable, to the extent of overestimating true use by 100%, the 2010 International Microbicides Conference heard yesterday.
This finding comes from a South African survey, but if it applies broadly it may have significant effects on trial design and mathematical modelling of the effect of microbicides.
Mitzi Gafos of the Africa Centre for Health and Population Studies discussed results from a longitudinal survey of 1177 women at the Africa Centre for Health and Population Studies (ACHPS), a rural centre north of Durban. The study was part of the MDP301 trial of the PRO2000 microbicide, which failed to demonstrate efficacy last year.
Local HIV prevalence is extremely high at 22%, peaking at over 50% in women aged 25 to 29.
South African surveys of condom use report that there has been a considerable increase in condom use in the last decade, with the proportion of women reporting condom use at last sex increasing from 27% in 2002 to 68% in 2008.
ACHPS reported similar figures in its year-long study, with an increase of condom use at last sex from 58% to 68% over the year.
However, these reported increases in condom use have not been accompanied by any decline in HIV incidence, except possibly in teenagers, though the methodology of the national survey that found evidence for this decline has been criticised. HIV incidence in women attending ACHPS has stayed at 4.4% a year, with no decline at all since 2003.
Over the longitudinal study, the researchers noticed that women very rarely reported that they ‘sometimes’ used condoms. Forty-six per cent, averaged over the trial, reported ‘always’ using condoms, 48% ‘never’ and only 6% ‘sometimes’. This did not square with numerous other condom surveys that suggest that inconsistent use is very common.
Four times during the study, at weeks 4, 24, 40 and 52, the researchers asked more detailed questions about sexual behaviour and asked whether the women had had sex in the previous week; and if so, when, what kind of sex (oral, vaginal or anal), and whether a condom was used.
This was designed to produce a more accurate minimum estimate of inconsistent, ‘sometimes’ condom use so that, if a woman was given the opportunity to report, for instance, having sex twice during the week but only using a condom once, this might produce a more accurate account of inconsistent use.
However, these surveys taken individually mirrored the monthly ones: the proportion reporting that they used a condom on every occasion when they had sex in the last week was 52%, 64%, 65% and 66% at the four time points and the proportion reporting inconsistent use 6%, 3%, 4% and 3%.
However, if reports from the four time points were compared, the researchers found that different women reported inconsistent use each time – to the extent that the proportion of women reporting inconsistent use at least once during the study was 36%, the proportion reporting 100% use was 41% and the proportion reporting that they never used condoms 21%.
When these data was combined with more qualitative data including a sample of in-depth qualitative interviews, sex diaries kept by the women, and the fact that a ‘biomarker event’ – HIV or STI diagnosis or pregnancy – occurred in 24% of the subset of women who were taking part in the PRO2000 study – researchers estimated that the true level of consistent condom use was in fact 25% – half of that seen in self-report. Gafos commented that given that the absence of a biomarker was not proof of condom use, the 25% may be a maximum.
She commented that although some of the over-reporting was due to social desirability bias – not wanting to risk the disapproval of the researchers – qualitative interviews showed that women tried to report their condom use as accurately as they could.
Sometimes they simply could not remember what had happened during the last week, and sometimes there was evidence that they genuinely thought men had worn a condom (and were astonished by events like pregnancy) when there was anecdotal evidence that men sometimes used stratagems to deceive women such as putting on a condom that they had perforated.
The main problem however was that while women might see themselves as consistent condom users, and that this self-image might cause selective recall, life circumstances conspired to change condom usage patterns suddenly and unpredictably.
Women did not so much adopt high-risk behaviours as enter high-risk situations, said Gafos, such as the start or ending of a relationship, unaccustomed alcohol use, partners returning home or emotionally stressful life events.
“The data we have in a province like KwaZulu-Natal consistently overestimates condom use. Cross-sectional data, even in visits that are not that far apart, do not capture true use,” commented Gafos.
She added: “Some of the answers will come from qualitative work, to find out what changes for a woman and why she finds she can, then can’t, then can use a condom.”
Reliable measures of condom usage are important when estimating the likely impact of microbicides: other presentations at the conference showed that mathematical models will underestimate the impact of microbicides if condom use is thought to be higher than it is.
Gafos M et al. How many women really achieve consistent condom use over the course of a year? Evidence from rural KwaZulu-Natal. 2010 International Microbicides Conference, Pittsburgh, abstract 193, 2010.