Epidemiologic Features of Vulvovaginal Candidiasis among Reproductive-Age Women in India

Background. Vulvovaginal candidiasis is characterized by curd-like vaginal discharge and itching, and is associated with considerable health and economic costs. Materials and Methods. We examined the incidence, prevalence, and risk factors for vulvovaginal candidiasis among a cohort of 898 women in south India. Participants completed three study visits over six months, comprised of a structured interview and a pelvic examination. Results. The positive predictive values for diagnosis of vulvovaginal candidiasis using individual signs or symptoms were low (<19%). We did not find strong evidence for associations between sociodemographic characteristics and the prevalence of vulvovaginal candidiasis. Women clinically diagnosed with bacterial vaginosis had a higher prevalence of vulvovaginal candidiasis (Prevalence 12%, 95% CI 8.2, 15.8) compared to women assessed to be negative for bacterial vaginosis (Prevalence 6.5%, 95% 5.3, 7.6); however, differences in the prevalence of vulvovaginal candidiasis were not observed by the presence or absence of laboratory-confirmed bacterial vaginosis. Conclusions. For correct diagnosis of vulvovaginal candidiasis, laboratory confirmation of infection with Candida is necessary as well as assessment of whether the discharge has been caused by bacterial vaginosis. Studies are needed of women infected with Candida yeast species to determine the risk factors for yeast's overgrowth.


Introduction
Vulvovaginal candidiasis is caused by overgrowth of Candida yeast species in the vagina and is characterized by curd-like vaginal discharge, itching, and erythema [1]. Vulvovaginal candidiasis has been associated with considerable direct and indirect economic costs [2], enhanced susceptibility to HIV infection [3], and is being investigated for a potential relationship with preterm birth [4]. Treatment of vulvovaginal candidiasis is warranted when a woman presenting with a complaint of symptoms consistent with vulvovaginal candidiasis also has laboratory confirmation of the presence of Candida from a vaginal specimen. Short-course azole-based treatment regimens are considered effective and safe [5] and are accessible and affordable in most settings.
Much of the epidemiologic literature concerning vulvovaginal candidiasis reports on studies in which women were queried on their self-reported history of vulvovaginal candidiasis [6], but without laboratory-confirmation of infection by Candida. Other studies, in which investigators only measure the presence of Candida infection of the vagina [1], are not able to identify women with symptomatic vulvovaginal candidiasis disease; this latter study design is frequently employed for studies conducted in lowincome settings. Few studies have diagnosed vulvovaginal candidiasis through laboratory confirmation of infection in symptomatic women, and few studies have measured the incidence of confirmed cases of vulvovaginal candidiasis.
The lack of representative data on the epidemiologic features of laboratory-confirmed vulvovaginal candidiasis has been evident throughout the time in which vulvovaginal candidiasis has evolved from being considered a "nuisance infection" to a clinically relevant condition [7,8]. In India, only two studies have been conducted in which laboratoryconfirmed vulvovaginal candidiasis was diagnosed in a community-based sample. Bang [9,10]. However, neither study assessed the incidence of or identified risk factors for vulvovaginal candidiasis.
As reduction of HIV transmission and of adverse birth outcomes remain public policy priorities in India [11], and studies have shown gynecological morbidity is extremely common [12][13][14][15], additional investigation of the epidemiologic features of vulvovaginal candidiasis is warranted.

Materials and Methods
We examined the incidence, prevalence, and potential risk factors for vulvovaginal candidiasis among a cohort of women originally recruited for a study to examine the relationship of lower genital tract infections and incident Herpes simplex virus type 2 infection [16]. The recruitment and baseline descriptive features of the cohort of 898 sexually active, nonpregnant women between 16 and 30 years of age from communities around Mysore, India has been previously described [17,18]. Briefly, between 2005 and 2006, women in the cohort completed three study visits (at baseline and at three and six months), comprised of a structured interview, a clinical examination, and collection of cervicovaginal specimens for laboratory testing. Trained interviewers collected sociodemographic and behavioral information, as well as reports of symptoms associated with gynecologic morbidity.

Clinical and Laboratory
Methods. Study physicians performed a pelvic examination of each participant and recorded signs of vaginal abnormalities. During the examination, swabs of the posterior fornix of the vagina and blood specimens were collected, and vaginal pH measured. Consistent with the US Centers for Disease Control and Prevention (CDC) 2006 diagnostic criteria [19] established prior to the baseline visit [20], vaginal Candida infection was assessed from culture (InTray Colorex Yeast, BioMed Diagnostics, White City, OR, USA) though speciation of positive samples was not recorded. Participants with positive culture for Candida were diagnosed as having vulvovaginal candidiasis if they reported vaginal itching or discharge and had vaginal erythema or discharge observed on clinical examination. As part of the Nugent-criteria scoring for laboratory diagnosis of bacterial vaginosis [21], the average number of Lactobacillus-like cells (morphotypes) detected over several visual fields in light microscopy was enumerated, and given a score corresponding to averages of 0, <1, 1-4, 5-29, and ≥30. Laboratory investigations were completed at Holdsworth Memorial Hospital and Vikram Hospital in Mysore, India. Clinical and laboratory diagnostic criteria for the gynecological conditions examined in this cohort are detailed in Table 1. Women were treated according to CDC guidelines [19].

Statistical Methods.
Our outcome measure was laboratory-confirmed diagnosis of vulvovaginal candidiasis. First, we describe prevalence and incidence of vulvovaginal candidiasis using frequencies and percentages. We calculated the visit-specific prevalence of vulvovaginal candidiasis and assessed whether this prevalence changed over the course of the study. Next, we tabulated the number of women's clinic visits with various vaginal symptoms reported or vaginal signs observed. Among observations in which these vaginal signs or symptoms were recorded, we calculated the proportion in which a diagnosis of vulvovaginal candidiasis was made. Finally, we examined the relationship between the prevalence of vulvovaginal candidiasis and sociodemographic characteristics and possible risk factors. Categorical variables are the same as used in previous analyses; cut-points of previously continuous measures were defined to create balanced strata of participants at the baseline visit [16,17]. We used a separate univariable regression model for each characteristic and risk factor. While counts (e.g., the number of diagnoses of vulvovaginal candidiasis) are reported as observed, percentages, prevalences, trends in prevalence, 95% confidence intervals, and P values were estimated using generalized estimating equations (GEE) regression models. GEE allows for parameter estimation when observations are correlated, in this case, due to multiple observations per participant over the course of the study. The GEE models were specified with binary family, identity link, and exchangeable correlation. The GEE models were bootstrapped with 500 repetitions to estimate robust standard errors.
Statistical analysis was conducted using Stata 11.2 (Stata Corp, College Station, TX, USA). The study was approved by the Committee for the Protection of Human Subjects at the University of California, Berkeley, and the Asha Kiran Institutional Review Board of Mysore, India; all participating women provided written informed consent.

Results
The 898 women participating in the study contributed a total of 2551 study visits, with 800 women (89%) attending all three study visits and another 53 women (6%) attending two visits. The median age of the women at the baseline visit was 26 years (Interquartile range [IQR] 24-29 years), and participants had been with their current sex partner for a median of nine years (IQR 6-12 years). Over one-quarter of the women had no formal education (27%). Nearly one in three women identified as Muslim (29%). Few women were nulliparous (15%), and very few women reported using oral contraception, condoms, or an intrauterine device at any The prevalence of vulvovaginal candidiasis declined over the three study visits from 77/893 (9%) at baseline to 65/840 at three months (8%) and 38/795 (5%) at six months (P value for trend < 0.001). The incidence of vulvovaginal candidiasis was modest: Of the 1487 baseline or three month study visits in which a woman was not diagnosed with vulvovaginal candidiasis, 72 (5%) were positive for vulvovaginal candidiasis at the next study visit. Repeat diagnoses of vulvovaginal candidiasis were common. Of the 137 baseline or three-month study visits in which a woman was diagnosed with vulvovaginal candidiasis, 30 (28%) were again diagnosed with vulvovaginal candidiasis at the next study visit.

Cross-Sectional Analysis of Vulvovaginal Candidiasis.
We did not find strong evidence for associations between sociodemographic characteristics and the diagnosis of vulvovaginal candidiasis. The prevalence of vulvovaginal candidiasis among women who first had sex before 15 years of age (Prevalence 4.4%, 95% CI 2.6, 6.3) appeared lower than for women who first had sex between 15 and 18 years (Prevalence 7.5%, 95% CI 6.1, 9.0) or over 19 years of age (Prevalence 8.2%, 95% CI 5.5, 11.0). There was a large difference in the prevalence of vulvovaginal candidiasis

Discussion
We examined the incidence, prevalence, and potential risk factors for vulvovaginal candidiasis among a cohort of reproductive-age women in Mysore, India. Among this cohort, we found evidence that a presumptive diagnosis of vulvovaginal candidiasis based only on presence of signs or symptoms, in absence of laboratory confirmation, would be mostly incorrect. Consistent with previous research [24], we could not identify behavioral risk factors for vulvovaginal candidiasis, which provides impetus for additional investigation into intrinsic factors such as the composition of vaginal flora, the presence or absence of genetic factors, and the features of the host and local immune response. Vaginal discharge, itching, and erythema, while quite common, were insufficient to diagnose vulvovaginal candidiasis in the absence of laboratory confirmation. Had syndromic diagnosis been used to diagnose vulvovaginal candidiasis in this cohort, the positive predictive values would have been very low (15-41%). Our results are consistent with other studies detailing the overtreatment that results from the use of syndromic diagnosis based on vaginal discharge to diagnose vaginal conditions [25][26][27][28]. Previous findings also demonstrate that a minority of women with vaginal discharge have vulvovaginal candidiasis [25,[29][30][31]. Thus, the diagnosis of vulvovaginal candidiasis based solely on signs or symptoms leads to overestimation of the prevalence of vulvovaginal candidiasis and its overtreatment, while leaving the actual cause of the vaginal symptoms untreated. This finding of misdiagnosis based on symptoms is also relevant for women who self-diagnose vulvovaginal candidiasis.
The prevalence of laboratory-confirmed vulvovaginal candidiasis we observed is consistent with the results of two other community-based studies in India [9,10]. Given that reproductive tract conditions account for nearly half of the days of illness experienced among women in this region of India [32], it is critical to understand the incidence and prevalence of individual conditions; to our knowledge, this is the first study from India to describe the incidence of and possible risk factors for vulvovaginal candidiasis.
The study visit-specific point prevalence of vulvovaginal candidiasis in this cohort decreased from 9% to 5%, and 72% of treated women were negative for vulvovaginal candidiasis at their next study visit, indicating successful provision of treatment. Only 20% of those infected with Candida were diagnosed as having vulvovaginal candidiasis, much lower than the 53% found in a community-based study in Tamil Nadu, India [9]. We were not able to determine whether the 28% of women with a diagnosis of vulvovaginal candidiasis on two consecutive visits were cases in which, despite treatment, vulvovaginal candidiasis had cleared and then recurred. More likely, the repeat diagnoses at consecutive visits represent instances in which the vulvovaginal candidiasis was caused by Candida species not susceptible to fluconazole treatment. Previous research in India has found a high proportion of women infected by non-albicans Candida species [29,33], which are more resistant to treatment with azoles [34,35].
Of the sociodemographic and behavioral characteristics we examined, only age at initiation of sexual activity appeared to be associated with the prevalence of vulvovaginal candidiasis, such that those with later initiation of sexual activity had a higher prevalence of vulvovaginal candidiasis. As the number of years women had been with their sex partners was not associated with vulvovaginal candidiasis, these two sociodemographic results appear discrepant and warrant additional investigation.
We found a positive association between having clinically diagnosed bacterial vaginosis and vulvovaginal candidiasis. As both diagnoses include vaginal discharge as a component of their respective diagnostic criteria, it is very likely there Infectious Diseases in Obstetrics and Gynecology 5  Prevalences, 95% confidence intervals and P values calculated using generalized estimating equations with binary family, identity link, and exchangeable correlation. * 0.05 ≤ P < 0.10. * * P < 0.05. * * * A socioeconomic index was calculated using the first factor from a principle components analysis of household consumer goods, toilet type, financial instruments, and stove type, and then recoded into a binary score of low and high socioeconomic status [23].
is misclassification between vulvovaginal candidiasis and clinically defined bacterial vaginosis. For example, women infected with Candida may have discharge caused by bacterial vaginosis and could thus be misdiagnosed with vulvovaginal candidiasis [36]. Our findings emphasize the problems inherent in making diagnoses of vaginal conditions based on clinical examination alone [34,37,38]. We found some evidence that the prevalence of vulvovaginal candidiasis varied with the presence of Lactobacillus morphotypes. The evidence for a relationship between the prevalence of vulvovaginal candidiasis and the presence of Lactobacillus in the vagina is conflicting, including studies in which the H 2 O 2 -production status of Lactobacillus was considered [24,[39][40][41][42][43][44]. Recently, a prospective cohort study of female sex workers in Kenya found the presence of Lactobacillus, regardless of H 2 O 2 -production status, was positively associated with prevalent vulvovaginal candidiasis (adjusted odds ratio (aOR) 2.3, 95% CI 0.8, 6.4), a relationship that was strengthened after restricting the analysis to women without a diagnosis of bacterial vaginosis (aOR 3.8, 95% CI 1.3, 10.8) [39].
The loss of vaginal Lactobacilli is the hypothesized mediator for the relationship between the receipt of antibiotics and the risk of vulvovaginal candidiasis [1,5,45,46]. The mediation hypothesis also underpins the long-standing interest in use of probiotic interventions to reduce the risk of developing vulvovaginal candidiasis [1,47,48]; the results here do not provide strong support for this hypothesis.
Strengths of this study include a large effective sample size derived from the use of participants' repeated observations, which allows for measurements of prevalence and incidence. Additionally, other studies of vulvovaginal candidiasis in India use samples of symptomatic women recruited from clinics or used syndromic diagnosis and as a result were not able to estimate the community-level prevalence of vulvovaginal candidiasis. Our study is one of the few to examine the prevalence of vulvovaginal candidiasis across a range in the number of Lactobacillus morphotypes detected in the vagina; examination of a dose-response provides better evidence, if any, of a biological relationship.
There are also important limitations of our study to consider. First, the women in this cohort were recruited by nonrandom sampling; unmeasured sampling bias can limit the generalizability of these results. Second, because of the cross-sectional nature of our analysis, we cannot make causal interpretations for the variations in the prevalence of vulvovaginal candidiasis observed here. Third, we were not able to speciate the Candida organisms detected. The associations between sociodemographic characteristics and potential risk factors and the prevalence of vulvovaginal candidiasis may differ by the Candida species which infect women, which are known to vary considerably by geographical location [1]. 7 Fourth, we could not verify whether participants were selfmedicating between visits with antibiotics or antifungals, which would influence the incidence and prevalence measurements of vulvovaginal candidiasis. Further, we could not immediately followup with women treated for vulvovaginal candidiasis, and so could not verify whether treatment was successful. Finally, given the limited duration of the study, we could not identify a subset of women with recurrent vulvovaginal candidiasis, an important condition with epidemiologic features distinct from acute vulvovaginal candidiasis [49].

Conclusions
We found that syndromic diagnosis will result in substantial overdiagnosis and overtreatment of vulvovaginal candidiasis-negative women. For correct diagnosis of vulvovaginal candidiasis laboratory confirmation of vaginal infection with Candida is necessary as is a mean of assessing whether the discharge has been caused by bacterial vaginosis. Absent accurate means of diagnosing vulvovaginal candidiasis, women remain at risk for vulvovaginal candidiasisassociated negative birth outcomes and acquisition of sexually transmitted infections. Follow-up studies are needed of women infected with Candida yeast species to determine the risk factors for the yeast's overgrowth, as it appears that the examination of behavioral risk factors does not appear to be a fruitful avenue for further inquiry.