Prevention of spontaneous preterm birth remains one of the most important challenges in modern maternity care. Whilst an association between ascending infection and preterm birth is undisputed, research focussed on finding effective preventive treatments has been disappointing[1]. To date, most treatment trials (e.g. for bacterial vaginosis, Ureaplasma urealyticum, Chlamydia trachomatis, trichomoniasis) have found little effect on the rate of preterm birth[2]. In contrast, in a randomized controlled trial of early antenatal screening (15-19 weeks) and treatment for asymptomatic bacterial vaginosis, candidiasis and/or trichomoniasis in early pregnancy, Kiss and colleagues reported a 46% reduction in the spontaneous preterm birth rate[3]. Post-hoc subgroup analyses of this trial suggest the benefit was primarily among those women who were treated for asymptomatic candidiasis.

The results of observational studies of the association between candidiasis and preterm birth have been mixed. Two cohort studies conducted in high-risk obstetric populations in the United States in the 1980 s found no significant association between preterm birth and moderate to heavy growth of Candida albicans or other Candida species at 22-30 weeks gestation[4, 5], In contrast, retrospective analyses of the prevalence of preterm birth in population-based data from Hungary found that vaginal clotrimazole treatment of candidiasis during pregnancy was associated with a significantly higher mean gestational age, resulting in a 34-64% reduction in the prevalence of preterm birth[6–8]. A similar reduction in preterm birth (49%) was observed in retrospective study of Latina women in New York who were treated with intravaginal azoles for Candida vaginitis[9]

Pregnant women have a two-fold increase in the prevalence of vaginal colonization by Candida species compared with non-pregnant women[10]. This association is influenced by increased levels of circulating oestrogens and deposition of glycogen and other substrates in the vagina during pregnancy[10]. Trials of treatment of candidiasis in pregnancy have been limited to women with symptomatic candidiasis (thrush) and the outcome limited to successful eradication of Candida colonization and amelioration of symptoms, not pregnancy outcomes[11]. Women with symptomatic candidiasis cannot ethically be randomized to a 'no treatment' trial arm, whereas currently, asymptomatic Candida colonization is not considered to require treatment. Consequently, we undertook a pilot study to determine the feasibility of conducting a large randomized controlled trial to answer the clinical question 'In pregnant women with asymptomatic candidiasis does treatment with clotrimazole reduce the incidence of preterm birth?' In addition to assessing procedural feasibility, the aim of the pilot study was to determine the rate of asymptomatic vaginal candidiasis <20 weeks gestation and at 24-28 weeks gestation in an unselected maternity population.

This open-label pilot study was conducted in a single Australian tertiary obstetric hospital with recruitment between May 2008 and December 2009. Eligible women included those with a singleton pregnancy of 12 to 19 weeks gestation, aged 18 years or over, no symptoms of vaginal candidiasis at the time of recruitment and no known sensitivity to clotrimazole. Women were recruited by research midwives during a routine antenatal clinic visit. All women provided written informed consent prior to entry into the study. Participants were advised at study entry to seek medical attention from their local doctor or other antenatal care provider if symptoms of thrush developed at any stage in their pregnancy.

Baseline data (age, gravidity, parity and previous preterm birth) were collected at the time of obtaining informed consent and women were instructed on the self-collection of a vaginal swab for isolation of Candida species. Such self-collection is commonly used for screening in pregnancy[12]. All women with Candida isolated on the initial swab and a random sample of women without candidiasis were asked to provide a second vaginal swab at 24-28 weeks.

All vaginal swabs were received within 24 hours of collection and inoculated onto full plates of chromogenic agar (CHROMagar™Candida, CHROMagar, Paris, France) to isolate, provisionally speciate and semi-quantitatively enumerate Candida species (scant, moderate, or heavy). Cultures were incubated for 72 hours at 35 degrees Celsius in 5% CO₂. Identification to species level was based on colony morphology and color.

We have demonstrated the feasibility of performing a randomized controlled trial of clotrimazole for pregnant women with asymptomatic candidiasis. The participation rate (64%) was acceptable, women were not inconvenienced by participation in the study, laboratory testing and medication dispensing were problem free, and the follow-up rate was 99%. Although there were no statistically significant differences in outcomes between treatment groups, the findings were in the direction of a treatment effect.

We chose a PROBE design to overcome some specific disadvantages associated with a double-blind, placebo-controlled trial for answering our clinical question[13, 14]. First, for our hypothesis, a placebo arm would not represent current clinical practice (no screening and no treatment). Second we were concerned that knowledge of Candida colonization might change participants behavior such that they would seek active therapy (clotrimazole vaginal preparations are available over the counter in Australia). And finally, as a vaginally administered placebo will necessarily contain an alcohol preservative, it may be biologically active and have an independent effect on vaginal flora.

Findings from this pilot study support the hypothesized role of Candida in the causal pathway to preterm birth. First, there was a higher spontaneous preterm birth rate in women with untreated asymptomatic candidiasis compared to those without candidiasis (6.25% versus 2.99%, RR = 2.2 95%CI 0.5-8.7) consistent with Candida colonization as a risk factor for preterm birth[3, 15, 16]. Second, there was a tendency towards a reduction in preterm birth for those women treated with clotrimazole, consistent with that reported by Kiss[3], Although the randomized trial design minimises bias by balancing both known and unknown prognostic factors in the assignment of treatments[17], the wide confidence intervals around the estimates are also consistent with a null finding or one of increased risk.

This pilot study established the rate of asymptomatic vaginal colonization with Candida early in pregnancy (20%) and at 24-28 weeks (76% untreated and 49% treated, P < 0.01). Our observed colonization rate of 20% is consistent with the few published studies. Kiss reported a 14% colonization rate at 15-19 weeks gestation in asymptomatic women with singleton pregnancies[3]. A multicentre prospective US study of an ethnically diverse population enrolled at 23-26 weeks gestation reported that 10% of women had moderate-heavy colonization with Candida and among a subset from a single centre the total Candida colonization rate was 22% (58% light and 42% moderate-heavy colonization) with 87% C albicans and 94% of women asymptomatic[4]. Two other single centre US studies reported candidiasis rates at 22-30 weeks gestation of 14-38% in diverse populations of pregnant women[5, 18].

The post-treatment colonization rate of 48% in our pilot study was higher than most, but not all (median 22%, range 0-73%) treatment trials of symptomatic (culture positive) vaginal candidiasis (thrush)[11]. The latter trials varied widely in gestational age at enrolment (14-36 weeks gestation), dose and type of imidazole (1%-2%, clotrimazole, miconazole & econazole), duration of imidazole therapy (1-14 days) and the duration between treatment and follow-up [11]. Importantly in the pilot study, the 10-week average duration between the enrolment and 24-28 week culture was notably longer than the thrush treatment trials (1-5 weeks)[11]. We cannot determine whether treatment led to eradication followed by recolonization or simply a reduction in the level of Candida growth in individual participants. A larger trial could determine whether the observed trend to a reduced level of moderate-heavy Candida colonization (55% to 42% light colonisation) among treated women occurred by chance or was a true effect.

We considered carefully the timing of the intervention and chose to screen for and treat candidiasis in the early/mid part of the second trimester for a number of reasons. First, early intervention is consistent with the protocol in the Kiss trial[3] in which screening and subsequent treatment of candidiasis occurred between 15 weeks and 19 weeks gestation. Second, accumulating evidence suggests that the role of infection in preterm birth is a chronic process.⁴ Organisms detected in the uterus before membrane rupture are typically of low virulence, probably accounting for both the chronicity of intrauterine infections and the frequent absence of overt clinical signs of infection.⁴ And third, although the trials of treatment of bacterial vaginosis overall demonstrate no benefit in the reduction of preterm birth, a positive treatment effect is seen in those trials that initiated treatment early in the second trimester[2]. The current Cochrane Review on bacterial vaginosis concludes that: "treatment before 20 weeks' gestation may reduce the risk of preterm birth. This needs to be further verified by future trials." If early pregnancy is the period of vulnerability to establishment of inflammatory responses to low virulence organisms that will increase the risk of preterm birth, then candidiasis late in pregnancy may be unimportant.

The safety of any proposed intervention in pregnancy is of paramount importance. Clotrimazole is classified as a Category A drug, which has been used by a large number of pregnant women without any proven increase in the frequency of malformations or other direct or indirect harmful effects on the fetus having been observed[19, 20]. Local application of clotrimazole vaginal pessaries or cream is generally well tolerated although skin reactions (burning, stinging, or redness) can occur occasionally. We chose a 6-day treatment (rather than shorter course) because this is supported by the Cochrane Systematic Review of treatment for Candida eradication in pregnancy[11] and was the regimen used in the Kiss trial[3].

We believe a large, adequately powered, randomized trial of clotrimazole to prevent preterm birth in women with asymptomatic candidiasis is both feasible and warranted. This pilot study has informed the development and funding application for such a trial[21]. Should this intervention prove effective, a screening protocol could be readily incorporated into routine antenatal care and treatment is available and widely accepted. If it can be demonstrated that treatment of asymptomatic candidiasis reduces preterm births this will change current practice internationally and will directly impact the management of every pregnant woman.