Volume 26, Number 2, June 2000
Two Approaches to Managing Vaginal Discharge Lead To Overtreatment, Missed Infections and Wasted Funds
Generally accepted clinical strategies for diagnosing reproductive tract infections may be relatively inaccurate in areas where the prevalence of infections is low, resulting in misdiagnosis and overtreatment. A study conducted in 1997 at five health centers in Matlab, Bangladesh, to evaluate the efficacy of two approaches used to identify infected women without having to perform expensive laboratory tests found that both approaches overestimated the prevalence of reproductive tract infection, classified uninfected women as possibly infected and missed most of the few women who had serious genital infections.1 As a result, anywhere from 36% to 87% of recurrent funds expended to treat women on the basis of these approaches were wasted.
Because the laboratory tests needed to detect reproductive tract infections are expensive and may be difficult to perform under conditions in developing countries, syndromic management is often used to identify and treat women having a reproductive tract infection. This approach assumes that by asking a symptomatic client a few questions about her risk factors, a clinician can judge with a good degree of certainty whether she has an infection that needs treatment. Once women without infections are eliminated, presumptive treatment of the remainder is expected to eradicate most infections.
Researchers in Bangladesh set out to evaluate the accuracy of two algorithms for syndromic management of women complaining of abnormal vaginal discharge. Over a five-month period during 1997, all women in Matlab who visited a reproductive health center and complained of vaginal discharge or some related problem (such as genital itching, lower abdominal pain or pain during sex) were asked if they would participate in the study. All 465 symptomatic women agreed; they averaged 30.2 years of age, nearly all (97%) were married and most (94%) reported vaginal discharge.
In one approach, which was based on recommendations from the World Health Organization (WHO), health workers conducted a risk assessment for each client, noting the woman's personal characteristics and asking her about her sexual history and whether her partner had symptoms of an infection. If the woman's partner was symptomatic or if the woman herself had several risky characteristics, she was then treated for a broad range of potential cervical infections; if she did not meet these criteria, she was treated only for bacterial vaginosis and yeast infection.
The second approach involved visual inspection of the vagina: For each woman complaining of vaginal discharge, the health worker carried out a speculum examination, looking for visual signs of yeast, bacterial vaginosis, trichomonas, gonorrhea or chlamydia. Clients with signs of any of these infections were treated appropriately; those with no signs of specific infections were not treated. Health workers then took swabs of cells from the cervix and vagina, blood samples and urine samples, and used standard laboratory tests to diagnose infections in all participating women.
Among the 418 women for whom results were available, laboratory tests indicated that 32% had at least one endogenous infection (i.e., bacterial vaginosis or yeast infection). The proportions with specific infections ranged from 19% with bacterial vaginosis and 13% with candida infection to 1% or fewer with gonorrhea or chlamydia. These proportions were mostly unchanged when the sample was restricted to the 320 women for whom complete laboratory results on all infections were available.
A comparison of results from the two diagnostic approaches revealed that the WHO algorithm identified all women with bacterial vaginosis, yeast infection or trichomonas (i.e., its sensitivity was 100%). However, the algorithm also incorrectly identified all uninfected women as needing treatment (for a specificity of 0%). As a result, the positive predictive value of the WHO algorithm (i.e., the likelihood that it identified only those who actually had a disease) was low: 19% for bacterial vaginosis, 12% for yeast infection and 2% each for trichomonas and for the cervical infections gonorrhea and chlamydia.
The speculum-based algorithm was not as effective as the WHO algorithm at identifying women with an infection, pinpointing one-third of clients with bacterial vaginosis (32%) or trichomonas (33%) and nearly three-fifths of those with a yeast infection (59%). Visual inspection was much more specific than the WHO algorithm, however, correctly identifying anywhere from 80% (for trichomonas) to 97% (cervical infections) of those who were not infected. Still, because this method missed the majority of infected women, its positive predictive value was only somewhat better than that of the WHO algorithm--29% for bacterial vaginosis, 28% for yeast infection and 3% for trichomonas. Moreover, it was very unsuccessful with the most serious infections: Only one of the three women with gonorrhea or chlamydia was identified using this approach, but eight uninfected women were incorrectly assessed as having a cervical infection.
The WHO algorithm proved to be relatively expensive: Because it resulted in all women being treated, the total treatment cost per woman (US$1.22) and cost per true case (US$3.61) were somewhat greater for this approach than the respective costs of the speculum-based algorithm would have been (US$0.38 and US$2.75). The cost for every true cervical infection treated was $130 for the WHO algorithm; no cost could be calculated for the speculum-based algorithm, as no woman was correctly treated.
Moreover, with the WHO algorithm, 87% of the costs related to treatment and staff time went to unnecessary treatment, compared with just 36% of costs associated with visual inspection. On the other hand, while the latter approach wasted less money, it failed to identify infection in 78% of the women infected with trichomonas, all of the women with cervical infections and 58% of those with endogenous infections.
The researchers conclude that neither of the approaches they examined was adequate for detecting sexually transmitted infections in a low-prevalence situation. Both resulted in overdiagnosis of infection: This would represent a problem not only because of the unnecessary expense of treating uninfected women, but also because of the potential social consequences for women mistakenly told that they had a reproductive tract infection--particularly a sexually transmitted disease such as gonorrhea.
They recommend that in situations where the prevalence of sexually transmitted infections and endogenous vaginal infections is relatively low, efforts must be made to improve the accuracy of algorithms and diagnostic tests. In particular, alternative explanations for vaginal discharge need to be identified. Overtreatment will not be eliminated simply by improving existing algorithms, they add, if uninfected women continue to appear at health facilities with concerns about vaginal discharge.
The author of an accompanying commentary observes that undernourished and anemic women in south Asia often come to health facilities complaining of vaginal discharge (along with other symptoms, such as backache and dizziness).2 Traditional understandings of physiology may lead women to interpret natural genital secretions as representing a loss of the body's vital essence. Such women may have "a culturally shaped illness," she argues, one that "is not necessarily associated with reproductive-tract disease. Rather, the illness reflects a general state of being unwell, linked to overwork, poor nutrition, and social stress."
The researchers conclude that in situations such as that in Matlab, where the overall prevalence of sexually transmitted diseases is low, it may not be efficient to locate infection control programs at providers such as family planning clinics, since few clients there will prove to be infected in the first place. What could prove more cost-effective, they suggest, is targeting populations that are at elevated risk of sexually transmitted infection, especially sex workers and their clients. Indeed, they add, algorithms for syndromic management may "work better in men with clinically observable discharge than in women," and situating efforts at sites providing family planning services effectively excludes men from being examined.--M. Klitsch
1. Hawkes S et al., Reproductive-tract infections in women in low-income, low-prevalence situations: assessment of syndromic management in Matlab, Bangladesh, Lancet, 1999, 354(9192):1776-1781.
2. Trollope-Kumar K, Symptoms of reproductive-tract infection--not all that they seem to be, Lancet, 1999, 354(9192):1745-1746.