Evaluating vaginitis

Abstract

Vaginal complaints are extremely common, accounting for approximately 10 million office visits each year.1 Current research has important limitations. Vaginal symptoms and signs inaccurately predict the etiologic agent. Microbial identification from vaginal secretions does not prove causation. About 30% of women with vaginal complaints remain undiagnosed despite extensive testing.2 In this issue of the Journal of General Internal Medicine, Carr et al.3 present a cost-effectiveness analysis which concludes that a “shotgun” approach to undiagnosed vaginitis costs less than sequential testing. This study underscores the challenge physicians have in choosing the best (most complete and least expensive) strategy for testing and treating these complaints. The “shotgun” approach assessed by Carr et al. included: pH testing; yeast culture; DNA probes for Gonorrhea and Chlamydia; and Gram's stain and Trichomonas culture when the vaginal pH is greater than 4.9. Adhering to this approach would save $8 to $76 and lessen symptoms by up to 1.3 days. Because rigorous cost-effectiveness analyses are difficult to perform with the results dependent on correct assumptions,4 we are concerned that several assumptions used in this analysis are oversimplified, limiting the applicability of the study's results. This strategy may work for undifferentiated patients, but general internal medicine patients are differentiated. Previous studies have documented that causes of vaginitis vary depending on a woman's menopausal status and age. Candidiasis, bacterial vaginosis, and trichomoniasis represent the most common causes of vaginitis in premenopausal women.2 In contrast, up to 40% of postmenopausal women have vaginal symptoms because of atrophic vaginitis, which can mimic the vaginal discharge of infectious etiologies.5 The prevalence of Gonorrhea and Chlamydia are highest in women aged 15 to 19 years, and African-American women in this age group have rates eight times higher than white women.6 We can further risk stratify patients with certain comorbidities (diabetes mellitus and immunosuppression), recent antibiotic use, or steroid use. In these patients, the pretest probability of Candida as the etiologic agent, with targeted testing and potential empiric therapy, would likely dominate strategies incorporating DNA probes for Gonorrhea and Chlamydia. Likewise, if a patient already self-medicated with an over-the-counter antifungal therapy without symptom improvement, the pretest probability for Candida would decrease in favor of other etiologic microbes.7 We also question the major utility assumptions in this analysis. When considering adverse outcomes, the authors assumed that treatment side effects and vaginitis symptoms would cause similar decreases in quality of life. Yet, fluconazole is safe and well tolerated with a low (2.8%) incidence of significant adverse side effects or laboratory abnormalities leading to discontinuation of the drug.8 We wonder whether women would consider gastrointestinal symptoms from fluconazole as equal to symptoms of vaginitis, especially as vaginitis can cause extreme anxiety and severe social and sexual dysfunction in addition to the physical discomfort.9 The authors did a very nice job of considering further evaluation costs with follow-up office visits and possible referral to specialists. However, nurse practitioners often evaluate acute vaginal complaints, and phone triage of vaginitis is common.10 In the age of physician extenders, phone triage, and e-medicine, this analysis may overestimate the cost of traditional follow-up office visits. This would be difficult to account for in any model but may alter the results of the analysis. This cost-effectiveness analysis—demonstrating that a “shotgun” approach to undiagnosed vaginitis is less expensive than sequential testing—does a great service in helping us understand a very complicated problem. Further refinements of the model should address differentiated patients, incorporating age, sexual risk factors, and comorbidities. Symptom severity and medication side effects are likely not equivalent for all women, thus limiting generalizability. Finally, in our changing world of health care access, prior trials of self-medication and physician extenders will add more uncertainty. This common problem deserves further study and we applaud the authors for advancing the work in managing this complex symptom.