Vaginal yeast infection testing: GYNI microscopy study

In one specialist clinic, the investigational GYNI system had 84.06% sensitivity (58/69) and 65.85% specificity (81/123) for Candida albicans against the study's composite reference standard. The estimate is limited by the small, single-site sample and an imperfect reference standard; a positive result did not show that Candida caused symptoms, and the study did not assess patient outcomes.

The study evaluated a point-of-care microscopy and pH system across five vaginal conditions, including infectious and noninfectious alternatives. Its results show test-specific performance against the study's reference rules; they do not establish that organism detection alone identifies the cause of symptoms or that using the device improves care.

Lev-Sagie and colleagues conducted a prospective, cross-sectional accuracy study at one vulvovaginal-disorders clinic in Jerusalem, Israel, from December 2020 through October 2022. The study recruited 226 women aged 18 or older who reported discharge, odor, itching, burning, pain, or dryness. GYNI results were interpretable for 192 samples; 34/226 (15.0%) samples were excluded from the accuracy analyses. Some participants had more than one condition: 22/192 (11.5%) samples had coinfections, including 12 with both bacterial vaginosis and C. albicans.

The GYNI™ system used one vaginal swab with a cytobrush and cartridge, saline and 10% KOH microscopy, pH measurement, symptom inputs, and a multi-label image classifier. It returned suggested condition(s), pH, and a microscopy heatmap; the study says results were available within five minutes of scan initiation and were blinded to the clinician. The test was intended to support, not replace, history-taking and examination.

Condition-specific composite reference standards used specialist wet-mount microscopy plus laboratory tests where available: CHROMagar culture for Candida, multiplex PCR for Trichomonas vaginalis, Amsel criteria for bacterial vaginosis, and wet mount with clinical criteria for vaginal atrophy, aerobic vaginitis/desquamative inflammatory vaginitis (AV/DIV), and cytolytic vaginosis. For Candida, either microscopy or culture could make the reference standard positive. These criteria are imperfect comparators, and a positive culture does not by itself show that Candida caused symptoms.

Among the 192 included samples, GYNI accuracy for C. albicans was 139/192 (72.40%). Sensitivity was 58/69 (84.06%; 95% CI 73.67–90.86%) and specificity was 81/123 (65.85%; 95% CI 57.11–73.64%). Positive predictive value was 58/100 (58.00%; 95% CI 48.21–67.20%), and negative predictive value was 81/92 (88.04%; 95% CI 79.85–93.19%). These estimates describe agreement with the study's composite reference standard, not proof that a positive Candida result explained symptoms.

For bacterial vaginosis, accuracy was 163/192 (84.90%), sensitivity 30/33 (90.91%; 95% CI 76.43–96.86%), and specificity 133/159 (83.65%; 95% CI 77.12–88.59%). For vaginal atrophy, the corresponding values were 175/192 (91.15%), 22/26 (84.62%; 95% CI 66.47–93.85%), and 153/166 (92.17%; 95% CI 87.06–95.37%). For cytolytic vaginosis, they were 189/192 (98.44%), 12/14 (85.71%; 95% CI 60.06–95.99%), and 177/178 (99.44%; 95% CI 96.89–99.90%). For AV/DIV, they were 181/192 (94.27%), 22/25 (88.00%; 95% CI 70.04–95.83%), and 159/167 (95.21%; 95% CI 90.83–97.55%). The smaller condition groups make those estimates less precise.

The paper did not report accuracy estimates for non-albicans Candida or T. vaginalis because case numbers were too low. In a separate analysis of the specialist's assessment against Candida culture—not the GYNI test—C. albicans accuracy was 189/203 (93.10%), sensitivity 72/83 (86.75%), specificity 117/120 (97.50%), and agreement κ=0.855 (95% CI 0.782–0.927).

These results apply to interpretable samples from symptomatic adults examined at one specialist clinic and compared with the study's particular reference definitions. They do not establish performance in other clinics, populations, or workflows, or show that detected Candida caused a person's symptoms. The study also does not establish patient benefit or how the system affects treatment decisions.

The study was small and single-site: 192 of 226 recruited samples were analyzed, and condition-positive counts ranged from 14 to 69. The 34 excluded samples lacked interpretable device results; the authors described a camera-stabilization issue for the first 30 and several further issues, but did not account for all exclusions. For Candida, a positive culture or wet mount could make the composite reference positive, so its definition did not establish that Candida caused each participant's symptoms. The same specialist examined participants, contributed the reference microscopy, and labeled training images; the authors also noted that Amsel criteria limited the bacterial-vaginosis comparison. Non-albicans Candida and T. vaginalis accuracy was not reported, and there was no follow-up or assessment of treatment, patient, or safety outcomes.

This study is contextual. It adds single-clinic estimates for an investigational point-of-care system across several causes of vaginal symptoms, but its composite comparators, limited sample, and lack of patient outcomes do not replace guideline evidence or change the current conclusions about symptom overlap, testing, or Candida colonization.

No practical change. The study does not support self-diagnosis or treatment from this test result alone, and it does not change current guidance on clinical evaluation and appropriate testing.

The article contains no funding statement. Gyntools Ltd supplied the GYNI scanner and collection tools, paid institutional ethics committee service and insurance fees, and provided raw data for the study team's independent analysis, a sample heatmap, and device images. The first author held shares in GynTools Ltd and an unpaid advisory role. The other authors declared no competing financial or nonfinancial interests.

Ahinoam Lev-Sagie, Doris Strauss, and Avraham Ben Chetrit. “Diagnostic performance of an automated microscopy and pH test for diagnosis of vaginitis.” npj Digital Medicine. 2023;6:66. Published 13 April 2023. Read the original article (DOI: 10.1038/s41746-023-00815-w; PMID: 37055473).