Roddy RE, Zekeng L, Ryan KA, Tamoufé U, Tweedy KG. Effect of Nonoxynol-9 Gel on Urogenital Gonorrhea and Chlamydial InfectionA Randomized Controlled Trial. JAMA. 2002;287(9):1117-1122. doi:10.1001/jama.287.9.1117
Author Affiliations: Family Health International, Durham, NC (Mr Roddy and Mss Ryan and Tweedy); Laboratoire de Santé Hygiène Mobile, Ministry of Public Health (Dr Zekeng), and Care and Health Program (Mr Tamoufé), Yaoundé, Cameroon.
Context Nonoxynol-9 has been suggested as a vaginal microbicide to protect against
common sexually transmitted infections.
Objective To compare nonoxynol-9 gel and condom use (gel group) vs condom use
alone (condom group) for the prevention of male-to-female transmission of
urogenital gonococcal and chlamydial infection.
Design and Setting Randomized controlled trial conducted at 10 community clinics and 10
pharmacies in Yaoundé, Cameroon, between October 1998 and September
2000, with 6 months of follow-up.
Participants High-risk population of 1251 women (excluding sex workers) being treated
for or who had symptoms of sexually transmitted infections. Three were excluded
from the gel group (0.5%) and 7 from the condom group (1%) because of no follow-up
Interventions Nonoxynol-9 gel (100 mg) and condoms or condoms only.
Main Outcome Measure A positive test result for gonococcal or chlamydial infection by the
ligase chain reaction assay; secondary outcome measure was a positive test
result for human immunodeficiency virus (HIV).
Results The rate ratio (RR) for new urogenital infections was 1.2 for the gel
group vs condom group (95% confidence interval [CI], 0.9-1.6; P = .21). The gel group had 116 diagnosed gonococcal infections, chlamydial
infections, or both for a rate of 43.6 per 100 person-years, and the condom
group had 100 infections for a rate of 36.6 per 100 person-years. The RR for
gonococcal infection in the gel group vs the condom group was 1.5 (95% CI,
1.0-2.3) and for chlamydial infection was 1.0 (95% CI, 0.7-1.4). There were
5 new cases of HIV infections in the gel group and 4 in the condom group.
Three women in each group became pregnant during the study.
Conclusion Nonoxynol-9 gel did not protect against urogenital gonococcal or chlamydial
Sexually transmitted infections (STIs) have important repercussions
on reproductive health and are a major cause of morbidity and mortality worldwide.
Both Neisseria gonorrhoeae and Chlamydia trachomatis are common STIs in sub-Saharan Africa.1 Primary STI prevention strategies include abstinence,
mutually exclusive sexual relationships with an STI-negative partner, and
the use of male or female condoms. All of these strategies require male consent
Nonoxynol-9 was chosen for study because it is the most commonly used
spermicide worldwide and is 1 of only 2 spermicides marketed in the United
States. Although data are limited, the other major spermicides available outside
the United States do not appear to have a more favorable safety profile, little
is known about their in vivo effectiveness with regard to STI prevention,
and no data are available regarding clinical activity against human immunodeficiency
Nonoxynol-9 is a nonionic detergent that has been used as a spermicide
for more than 50 years. It can inactivate many sexually transmitted pathogens
in vitro, including N gonorrhoeae, C trachomatis, Haemophilus ducreyi, Treponema pallidum, Trichomonas vaginalis, and herpes simplex virus.2- 4
Both in vitro and animal data support the hypothesis that intravaginal application
of nonoxynol-9 may help protect women from HIV and other STIs.5- 8
Five randomized placebo-controlled trials of nonoxynol-9 for STI prevention
have compared 4 products: 2 gels,9,10
a sponge,11 and a film.12,13
The incidence of STI was reduced anywhere from 0% to 40% in these studies
(rate ratios [RRs], 1.0-0.6); however, the 95% confidence intervals (CIs)
for the RRs included 1, indicating possibly no protective effect. One study
using the film did not find any effect of nonoxynol-9 on the rate of disease
incidence for HIV, N gonorrhoeae, or C trachomatis.13 One study found a marginally
significant increase in the rate for gonorrhea, but the authors did not adjust
for the multiple comparisons made in the analysis.10
A meta-analysis of these studies would be difficult to interpret because different
formulations have been assessed and the sponge study used a different measure
We studied Conceptrol Contraceptive Gel with 4% nonoxynol-9 (Advanced
Care Products, Raritan, NJ). Its moderate dose, small (3.5 mL total) amount
of product inserted vaginally, and single-use applicators make it an attractive
product. A gel theoretically might be more readily available for protection
than a film. The gel contains a lower dose of nonoxynol-9 than a sponge and
may cause less epithelial irritation. A criticism of many of the previous
nonoxynol-9 studies is that they were conducted among sex workers who had
intercourse many times daily and with multiple partners. We chose a high-risk
population but excluded sex workers.
Although not widely used, spermicides are available in Cameroon. Pharmacies
carry a variety of nonoxynol-9 products, including gels, suppositories, and
foaming tablets. The Cameroon National Association of Family Planning and
Well-being also distributes nonoxynol-9 spermicides to clients.
Condoms are the standard of care for STI/HIV prophylaxis. A placebo
gel without preservatives and without some antibacterial effect in vitro could
not be manufactured. This study was designed to compare nonoxynol-9 gel and
condom use with condom use alone for the prevention of male-to-female transmission
of genitourinary gonorrhea and chlamydial infection.
The National Ethics Review Committee, Ministry of Public Health, Yaoundé
(Cameroon), and the Protection of Human Subjects Committee, Family Health
International, Durham, NC, approved the research. Each participant was required
to give written consent twice: once to be screened for the study and again
The target population was sexually active women who were not sex workers
but were at risk for gonorrhea and chlamydial infection, as indicated by having
symptoms of STIs immediately before contact with our study referral staff.
The participants were at least 18 years of age, had at least 1 coital act
weekly, were willing to learn the results of the HIV tests, and were willing
not to self-medicate with antibiotics during study participation. The participants
could not have a history of an adverse reaction to products containing latex
or nonoxynol-9, be using a spermicide, be pregnant, or have a seropositive
HIV test result.
We stationed referral staff at 10 clinics and 10 pharmacies. Women receiving
STI treatment at these sites were asked to speak to referral staff. Many clinics
do not conduct laboratory tests for STIs but treat according to symptoms.
Many women do not visit clinics but visit only the pharmacies to get treatment
for symptoms, which made the pharmacies a good site for contacting potential
participants. The staff explained the general purpose of the study and the
eligibility requirements and referred appropriate women to the study clinic
for screening after they had completed their treatment.
At the clinic, the staff gave another brief study explanation, and the
woman entered the consent process for screening. Each woman was given a detailed
study explanation by face-to-face counseling and the counselor's reading the
consent form to the woman and addressing her questions, which generally took
45 minutes to 1 hour to complete. Each woman received HIV pretest and condom
counseling. Participants received a gynecological examination and had specimens
collected for ligase chain reaction testing for N gonorrhoeae and C trachomatis from the uterine cervix
(Abbott LCx Probe System, Abbott Laboratories, Abbott Park, Ill) and a saline
and potassium hydroxide wet mount for bacterial vaginosis and T vaginalis. They also provided a blood sample for HIV testing (Enzygnost
Anti-HIV 1/2 Plus, Aventis Bhering, King of Prussia, Pa; Multispot HIV 1/HIV
2, Bio-Rad Laboratories, Hercules, Calif; and Novopath HIV-1 Immunoblot Kit,
Bio-Rad Laboratories). Two independent laboratories examined the LCx N gonorrhoeae and C trachomatis
tests for possible inhibition by nonoxynol-9. There were no false-positive
or false-negative results. All women who had a curable STI were treated at
no charge and invited to return for rescreening 2 weeks after they had completed
treatment. Women who did not have N gonorrhoeae, C trachomatis, or HIV infection were permitted to enter
the study enrollment consent process within 30 days of screening.
At enrollment, potential participants received posttest counseling,
were tested for pregnancy with a urine test, were informed further about the
nonoxynol-9 gel study, gave a second consent, provided an oral mucosal transudate
specimen (OraSure, Epitope Inc, Beaverton, Ore) for HIV testing (Wellcozyme
HIV 1 and 2 GACELISA [gamma antibody capture enzyme-linked immunosorbent assay],
Murex, Oxford, England), and gave a urine specimen for testing for N gonorrhoeae and C trachomatis by ligase
chain reaction. After the enrollment interview, the counselor received the
randomization assignment from the clinic coordinator, told participants their
assigned group, and provided condom counseling to both groups and gel use
instructions to the gel group. Participants chose 1 of 10 pharmacies for their
follow-up visits. All participants were asked to use condoms as often as possible,
and those in the gel group were asked to apply 1 applicator of gel up to 1
hour before each coital act. Free condoms and gel were provided to participants.
A statistician who was not otherwise involved with the study developed
the allocation sequence by using a random-number generator and randomly varied
permuted blocks of 4, 8, and 12. Group assignments were concealed in sequentially
numbered, sealed, opaque envelopes. Trained staff opened the envelopes only
after participants were properly enrolled. The clinic coordinator maintained
the envelopes in a secure office, and they were unavailable to the study counselors
Masking participants or clinic staff to treatment assignment was impossible.
However, assignments were not known by the laboratory staff and were revealed
to Family Health International study investigators and analysts only after
all data were entered in the study database and final-analysis programs had
been verified using dummy treatment codes.
Participants made monthly follow-up visits for up to 6 months at their
chosen pharmacy. Pharmacies were chosen as follow-up sites because they were
located closer to the study participants, making follow-up visits easier,
provided a secure place for storage of study materials, and were capable of
providing a discreet area for study procedures. The staff interviewed participants
about sexual behavior and product use and collected a urine specimen for N gonorrhoeae and C trachomatis
testing by ligase chain reaction. At the 3- and 6-month visits, participants
gave an oral mucosal transudate specimen for HIV testing. Participants visited
the study clinic when they experienced adverse events or when STIs were detected.
If a participant required a gynecological examination, a cervical swab was
taken for N gonorrhoeae and C trachomatis testing.
Adverse events were asked about at the monthly follow-up visits. If
an adverse event was reported, the participant was referred to the study clinic
for examination. If the adverse event was gynecological, the study clinicians
conducted a standardized pelvic examination that was identical to the examination
Gel and condom use was assessed by interview. We asked the number of
vaginal sex acts the woman had had in the 7 days before the interview, and
among those acts, the number of times gels, condoms, and condoms and gel together
were used. From these data, we were able to calculate how many coital acts
were with gel or condom use alone and how many were without gel or condom
Our primary end point was the occurrence of a positive ligase chain
reaction test (urine or cervical swab) for either N gonorrhoeae or C trachomatis after the subject had had
a negative test result for these organisms. Our secondary end point was the
occurrence of HIV (defined as 2 different positive enzyme-linked immunosorbent
assay [ELISA] results or a positive ELISA and a positive Western blot, after
a negative HIV test result). We collected a cervical swab whenever a pelvic
examination occurred (screening and gynecological adverse events) and collected
urine samples at the enrollment visit and normal follow-up visits. This approach
allowed us to reduce the number of pelvic examinations, reduce the inconvenience
to the study participants, decentralize the follow-up process, and reduce
loss to follow-up.
This trial was designed to have at least 90% power to reject a null
hypothesis of no difference in infection rates (at the .05 significance level),
assuming the gel was 50% effective in preventing infection with N gonorrhoeae or C trachomatis. Nearly 90
infections (total in both groups) would be needed to achieve the desired power
according to a 2-sided log-rank test. Achieving this power required enrolling
approximately 1200 women for 6 months of follow-up each, assuming an incidence
of infection in the condom group of 20 cases for every 100 person-years and
a loss to follow-up of 10% at 6 months.
A preplanned interim analysis was conducted on data collected through
the first 53 study infections. We intended to stop the trial if a significant
treatment effect was observed, an independent data and safety monitoring board
decided on ethical grounds to discontinue the study, or Family Health International
decided to stop the study on administrative grounds. Dummy treatment indicators
were used to mask data analysts, study investigators, and data and safety
monitoring board members to group assignments. The Lan-DeMets spending function
with O'Brien-Fleming boundaries was used to account for multiple looks at
the data, resulting in a nominal significance level of .00006 (corresponding
to a z score of 4.01) for a test of the primary outcome14 (EaSt Software, Cytel Software Corp, Cambridge, Mass).
The interim log-rank test of no difference in infection rates was not
significant at the required level, and the data and safety monitoring board
recommended that the study continue. For the final analysis, the primary outcome
used a nominal significance level of .04997 to account for this interim look.
Differences in infection rates between the 2 treatment groups were assessed
with 2-sided log-rank tests on an intent-to-treat basis. The event time was
defined as the midpoint of the following: the first follow-up visit at which
an incident infection was detected and the time of the previous visit at which
the participant was determined to be free from infection. Women who either
did not test positive or first tested positive more than 7 months after enrollment
were considered censored at either 7 months or the time of the last visit
at which they were free from infection, whichever was first. Cumulative infection
probabilities were based on the Kaplan-Meier method and Greenwood formula
for SEs. Crude incidence rates, RRs, and 95% CIs were based on the ratio of
the number of new cases to the total length of treatment exposure in the interval.
All calculations were based on time to first infection.
Baseline factors were identified as potential confounders before analysis
of the data: age, marital status, whether the participant was living with
a partner, number of children, number of coital acts in the 7 days before
enrollment, number of sex partners in the 30 days before enrollment, whether
condoms were used during the last coital act, gonorrhea or chlamydia at screening,
history of spermicide use, history of contraceptive use (oral contraceptives,
injectables, or intrauterine device), history of douching, history of STI
before the one that made the participant eligible for the study, and whether
previous condom use caused irritation. Cox proportional hazards regression
models were used to evaluate the effect of baseline factors on time to first
infection as well as potential interaction effects with the study gel.
Screening for the study began in October 1998, and the first participant
was enrolled in November 1998. Enrollment continued until June 2000, and the
study was stopped in September 2000. Figure
1 shows the participant flow from first contact with study referral
staff until completion of the study.
We excluded 3 participants in the gel group and 7 in the condom-only
group from the effectiveness analysis because they had no follow-up data.
Only HIV-seronegative women were to be enrolled in the study; however, 4 women
who were HIV seropositive at either screening or enrollment were enrolled
and later discontinued from the study. These women are included in the intent-to-treat
analysis. We had no reports of women in the condom-only group receiving or
using nonoxynol-9 gel during the study.
Participants in the 2 groups were similar in all clinical and demographic
respects (Table 1). The majority
of the women were single and had high educational attainment. Additionally,
the majority of participants had had sex with 1 or 2 partners in the 30 days
before enrollment, and most averaged 3 coital acts weekly. Women in both groups
(2.5%; 17 in the gel and condom group and 14 in the condom-only group) had
a confirmed positive serology test result for syphilis.
From baseline, no changes in the reported number of coital acts weekly,
the number of partners, or the number of new partners occurred during the
study. Reported condom use during follow-up was 81.4% of coital acts in the
gel group and 86.9% for the condom group. Condoms reportedly broke or slipped
off about 4.5% of the time in both groups. Among the gel group, gel use was
reported for 76.3% of the acts, both gel and condoms were used for 63.0%,
and only gel was used for 13%. No gel or condoms were used 5.6% of the time
in the gel group.
The rate of new urogenital infections was slightly higher in the gel
group (RR, 1.2; 95% CI, 0.9-1.6; P = .21) (Table 2). There were 15 swab events and
101 urine events in the gel group. There were 7 swab events and 93 urine events
in the condom group. Adjusting for significant covariate effects and stratifying
on history of STI before the episode that made the women eligible for enrollment,
the estimated Cox proportional hazards model RR for gel use vs condom use
alone was no different from the crude estimate. The N gonorrhoeae RR was 1.5 (95% CI, 1.0-2.3), and chlamydial infection RR was 1.0
(95% CI, 0.7-1.4) for gel vs condom group. Only 5 HIV infections were detected
in the gel group and 4 in the condom group, making a meaningful comparison
impossible because of small numbers and low statistical power.
A total of 214 women in the gel group and 215 in the condom group had
at least 1 adverse event. There were 292 examinations in the gel group and
296 in the condom group for adverse events. The type and frequency of adverse
events were similar in both groups. Reported symptoms were also similar in
both groups. Three women in each group became pregnant during the study follow-up.
No differences occurred between the groups for bacterial vaginosis, yeast
infections, trichomoniasis, pelvic inflammatory disease, or genital ulcers
that were identified at clinic visits for adverse event.
Self-reported condom use was frequent but inconsistent in both randomized
groups. Reported gel use was less consistent than condom use in the gel group.
The gel group reported more coital acts with no or incomplete condom protection
(23.2% vs 17.6% for the condom-only group). However, the gel group had fewer
coital acts with no gel or condom use (5.6% vs 13.1%). These numbers suggest
that the gel group substituted gel use for condom use. A similar pattern of
behavior was reported among sex workers in Colombia.15
Two laboratories assessed whether the gel inhibited the ligase chain
reaction tests for N gonorrhoeae and C trachomatis and found no problems. If the nonoxynol-9 gel had inhibited
our assay, we would have expected detection of fewer infections in the gel
group. We do not believe the nonoxynol-9 gel interfered with the testing,
since there were more infections in the gel group.
The excessive number of gonorrheal infections among the gel group may
have several explanations. First, gonorrhea may have a higher transmissibility
than chlamydia, estimated at 50% vs 20%,16
which would make inconsistent condom use less forgiving for gonorrheal than
for chlamydial infection. Second, nonoxynol-9 may make the epithelial cells
more receptive to bacterial attachment by organisms with fimbriae, such as Escherichia coli, or pili, such as N
gonorrhoeae.17 Finally, at screening,
slightly more participants were infected with gonorrhea in the gel group than
the condom group (5% vs 3%); therefore, the gel group may have had more exposure
A major concern when we designed this study was that loss to follow-up
for the condom-only group would be much higher than for the gel group. We
feared that women who did not get the study product would be upset or discouraged
and not return for follow-up. However, differential loss to follow-up did
not occur, and total loss to follow-up was quite low, thus not affecting the
The findings of this study are consistent with the findings from an
earlier study conducted in Cameroon among sex workers using the nonoxynol-9
film.13 The film study used a 70-mg product,
and the findings may have resulted from the smaller amount of nonoxynol-9
and possible problems with the nonoxynol-9 distribution from the film. In
this study, we sought to address both these potential problems by using a
higher dose of nonoxynol-9 and a formulation that was perhaps more readily
available. Initially, the findings of this study appear to be at odds with
those from another 100-mg gel study conducted in Alabama among a non–sex-worker
population.9 That study found a statistically
significant decrease in infections, but only at the 90% level, not the 95%
level. It is difficult to compare this study with the nonoxynol-9 sponge study
conducted among sex workers.11 The sponge contains
both barrier and chemical characteristics that are impossible to separate,
and the investigators chose to use the number of infections over the number
of examinations (because the number of examinations was different between
the 2 experimental groups) instead of the usual number of events over time.
This nonoxynol-9 gel, although safe to use, did not protect against
gonorrheal and chlamydial infection.