Diagnostic performance of DNA probe-based and PCR-based molecular vaginitis testing

ABSTRACT Vaginitis is usually diagnosed empirically, microscopically, via cultures, or by molecular testing for the detection of bacterial vaginosis (BV), vulvovaginal candidiasis (VVC), or Trichomonas vaginalis (TV). The DNA probe–based technique detects BV by identifying Gardnerella vaginalis, VVC by identifying Candida spp., while real-time PCR-based detection methods identify BV by algorithmic analysis of the absence or presence of known vaginal flora. We examined 8,878 total orders placed for DNA probe–based identification (ID) and 10,464 total orders placed for molecular panel ID. We found that PCR-based BV test positivity reduced from 30% to 23% compared with the population tested with DNA probe–based testing. We also found that PCR-based testing VVC positivity increased from 6.3% and 11.6% when compared with DNA probe–based testing. Bayesian generalized linear analysis estimated a lower mean proportion of positive tests for BV in PCR-based molecular panels than DNA probe testing suggesting an under-call of BV. The same models estimated a higher mean proportion of positive tests for molecular vaginal panels than DNA probe testing suggesting an increased detection of candidal vaginitis. In addition, the mean (SD) age for patients with Candida albicans was 40.5 (40.0–41.1) years. Patients with Candida glabrata (now N. glabrata) were 5.2–8.1 (mean 6.7) years older than patients with Candida albicans. Our retrospective data analysis found that BD Max MVP’s ability to discriminate between vaginal candidiasis versus other yeast will help to implement CDC (Centers for Disease Control and Prevention)-recommended treatment options. We also believe that providers’ inattention to non-albicans treatment could be an issue nationwide. IMPORTANCE Using retrospective data from U.S. Food and Drug Administration-approved/cleared molecular vaginal panels, molecular methods were found to have higher detection for Candida vaginitis and lower detection for bacterial vaginitis when compared to probe-based methods. In addition, the differentiation of Candida and non-Candida yeast has not reached the physician community as we observed noncompliance in recommended therapy. Furthermore, the pros and cons of migrating to molecular testing from conventional microscopy for identifying bacterial vaginitis and fungal vaginitis have been examined and reported in this paper. Interestingly, the mean (SD) age for patients with Candida albicans was 40.5 (40.0–41.1) years. Patients with N. glabrata were 5.2–8.1 (mean 6.7) years older than patients with Candida albicans.

and vaginal discharge sample analysis (Amsel criteria) (2,3).Yeast vaginal infection is considered the second most common cause of vaginitis whereas VVC caused by Candida albicans is one of the main sources of the condition (4).Healthcare providers usually diagnose VVC by examining a small sample of vaginal discharge under the microscope to see 2-to 4-μm-diameter of yeast cells or requesting laboratory tests.Positive fungal culture of a vaginal sample does not provide a definitive diagnosis as some women can have naturally colonized Candida in their vagina without causing any clinical symptoms (5).Although the Nugent score, Amsel criteria, and microscopic smears have been used for initial vaginosis/vaginitis diagnosis, pathogenesis and clinical presentation are more a complex process than direct correlation on microscopy or culture-based diagnosis.Many of the causative bacterial organisms like Atopobium vaginae and Prevotella cannot be cultured by conventional methods.Molecular diagnosis of bacterial vaginosis became more popular based on accurate causative organism identification.Additionally, some Lactobacillus spp.like L. crispatus and L. jensenii are important to maintain healthy bacterial flora, while Lactobacillus iners are shown to grow in high numbers in bacterial vaginosis.Therefore, conventional culture-based methods are not feasible to isolate and quantify such type of speciation.
However, molecular-based methods could predict these complex bacterial popu lation dynamics.Therefore, commercial molecular platforms for identifying vaginitis became popular in laboratory medicine (6).Based on the College of American Pathol ogist (CAP) proficiency testing (PT) program's participant summary on molecular vaginal panels in 2022, BD Max MVP (Becton Dickinson, Franklin Lakes, USA) and Aptima platforms (Hologic, Marlborough, USA) are the main commercial U.S. Food and Drug Administration (FDA)-approved/cleared platforms.In 2018, our lab migrated from BD Affirm DNA (Becton Dickinson) probe-based vaginitis testing to BD Max MVP based on its improved accuracy compared with BD Affirm for the detection of BV and Candida spp., and no difference for the detection of TV was observed between the two tests (7).In addition, BD Max MVP differentiates the primary source of fungal infection in Candida spp.versus the secondary source of yeast vaginitis in non-Candida/yeast spp.mainly responsible for yeast vaginitis/vaginosis.Candida spp.identified are C. albicans, C. tropicalis, C. parapsilosis, and C. dubliniensis, while non-candida yeast identified by this platform is the more difficult to treat Nakaseomyces (formerly Candida glabrata) and Pichia kudriavzevii (formerly Candida krusei) (7).More than 90% of Candida vaginitis are reported to be a C. albicans infection, and Nakaseomyces is responsible for the second most common vaginal yeast infection (8).Therefore, positive BD Max MVP Candida spp.can be comfortably inferred as C. albicans vaginitis.
The main aim of this study was to investigate if BD Max MVP has a better diagnostic ability compared to our previous BD Affirm DNA probe method.We also examined N. glabrata vaginitis/vaginosis for therapeutic appropriateness and compliance of our providers with Centers for Diseases Control and Prevention (CDC) guidance for treat ment.

MATERIALS AND METHODS
The study was a retrospective analysis performed in a single institution (Central Texas Veterans Health Care System, Temple, TX) from September 2015 to January 2023 conducted as a process improvement analysis.We examined 8,878 total orders placed for DNA probe-based identification (ID) (from September 2015 to September 2018 and 1,0464 total orders placed for molecular panel ID from September 2018 to January 2023).Deidentified data were used for this process improvement project, and institutional review board approval is not required.Institutional research chair approved publishing these data.

Testing platforms
The BD Affirm VPIII Microbial Identification Test (Becton Dickinson) is a DNA probe test intended for use in the detection and identification of Candida species (no differen tiation), Gardnerella vaginalis, and Trichomonas vaginalis nucleic acid in vaginal fluid specimens from patients with symptoms of vaginitis/vaginosis.The test is based on the principles of nucleic acid hybridization.In nucleic acid hybridization tests, comple mentary nucleic acid strands align to form specific, double-stranded complexes called hybrids.The test uses two distinct single-stranded nucleic acid probes for each organism, a capture probe, and a color development probe, which are complementary to unique genetic sequences of the target organisms.The capture probes are immobilized on a bead embedded in a Probe Analysis Card (PAC), which contains a separate bead for each target organism.The color development probes are contained in a multi-well Reagent Cassette (RC).The final step is reading the results of color development on each of the target organism beads and controls.
The BD MAX MVP (Becton Dickinson) performed on the BD MAX MVP System is an automated qualitative in vitro diagnostic test for the direct detection of DNA targets from bacteria associated with bacterial vaginosis (qualitative results reported based on the detection and quantitation of targeted organism markers), Candida species associated with VVC, and Trichomonas vaginalis from vaginal swabs in patients who are symptomatic for vaginitis/vaginosis.The test utilizes real-time PCR for the amplification of specific DNA targets and utilizes fluorogenic targetspecific hybridization probes to detect and differentiate DNA from: (i) BV through algorithmic analysis of lactobacilli (L.

Statistical methods
Bayesian binomial generalized linear models (GLM) with the number of positives as the outcome and the total number of tests as the denominator (trials) with the predictor as the device were used to estimate the mean proportion of positives for the two devices and the estimated mean difference in proportion between them, with statistical uncertainty quantified by 95% uncertainty intervals.The difference in age distribution between those with C. albicans and those with Nakaseomyces was estimated using a Bayesian GLM with skew-normal distribution, as the distributions of ages were strongly right-skewed.
Central Texas Veterans Health Care System (CTVHCS) has a unique patient population (women veterans over 18 years old, residing in central Texas), with a very high number of patient data points, and we found that two tests had similar TV positivity rates in agreement with the previous study (7).Both groups had similar negative percentages.Based on those facts, we assumed that the two groups of samples have identical chances of detection of vaginitis.Even though this comparison has limitations noted in the Discussion section, this is the best model we can use for such a large number of patient data sets for quality improvement assessment.

No difference in the detection of TV between the two testing platforms
Both populations have approximately a 2% positivity rate for TV (Fig. 1).This confirms a previous finding that there was 100% concordance between BD AffirmTV and BD Max MVP-TV verification and validation (9).In addition, the percentage of samples negative for TV was 61% from both populations (Fig. 1A and B).

BD Max MVP reported a lower BV percentage in a large sample size com pared with BD Affirm
We found that BD max MVP BV positivity reduced from 30% to 23% compared with the population tested with BD Affirm (Fig. 1B).There were 2,664 positives out of 8,878 total tests for BD Affirm and 2,453 positives out of 10,464 tests for BD Max MVP for the time mentioned above (Fig. 1A).The estimated mean proportion of positive tests for BD Affirm was 0.300 (0.290-0.309) and for BD Max MVP was 0.234 (0.226-0.243).The estimated mean difference in the proportion of positive tests was 0.066 (0.053-0.078).Thus, the proportion of tests that were positive for the BD Affirm device was 0.066 (0.053-0.078) higher than for BD Max MVP (Fig. 2A).

BD Max MVP reported a higher vaginal candidiasis percentage
In contrast to the results for BV, for vaginal candidiasis, BD Max MVP comparatively reported more vaginal candidiasis than BD Affirm.For vaginal candidiasis, there were 563 positives out of 8,878 total tests for BD Affirm and 1,427 positives out of 10,464 total tests for BD Max MVP.BD Affirm reported approximately 6% of total Candida spp.from 88,878 patients tested, while BD Max MVP platform reported approximately 13.5% positivity from 10,464 patients of combined Candida spp.and other two yeasts (Fig. 1).Using the same kind of model we used for BV, for Candida: the estimated mean proportion of positive tests for BD Affirm was 0.063 (0.058-0.069) and for the BD Max MVP it was 0.136 (0.130-0.143).The estimated mean difference in the proportion of positive tests was 0.073 (0.065-0.081), for proportion of the BD Max MVP minus proportion of BD Affirm.Thus, the proportion of tests that were positive for the BD Max MVP was 0.073 (0.065-0.081) higher than for BD Affirm (Fig. 2B).

BD Max MVP showed a difference in mean age for C. albicans versus N. glabrata
When we plotted distribution of Candida infection based on age (Fig. 3), we found that it had right-skewed distribution.Using a skew-normal distribution for modeling the age, the mean age for patients with C. albicans was 40.5 (40-41.1)years, while patients with C. glabrata showed a mean of 6.7 (5.2-8.1) years older with the mean age of 47.2 (45.7-48.6).
We investigated if our hospital providers made use of this new platform's species differentiation capability by reviewing 190 cases of N. glabrata vaginitis/vaginosis cases for therapeutic appropriateness and compliance with CDC guidance for treatment.Unfortunately, 114 out of 190 times, N. glabrata vaginitis was treated with fluconazole.In addition, seven out of 14 Pichia kudriavzevii (formerly C. krusei) cases were treated with fluconazole, while three patients were correctly treated with clotrimazole cream, and four patients were not treated based on clinicians' decisions.

DISCUSSION
We found that the newer BD Max MVP platform had reduced positivity as compared to BD Affirm for BV.In contrast to the results for BV, for vaginal candidiasis, BD Max MVP comparatively reported more vaginal candidiasis than BD Affirm.
BD Affirm predicts BV solely based on the presence of Gardnerella vaginalis, whereas BD max MVP utilizes algorithmic analysis of Lactobacillus and other bacteria involved in BV, which the manufacturer claim makes more accurate predictions based on their data analysis (6).Therefore, more accurate calling may better reflect the actual incidence of lower percentage of BV by BD Max MVP.In addition to our data, previous publications examined the diagnostic performance of two of these molecular assays in real-time comparison.In a verification study, using a total of 215 symptomatic women, they examined the sensitivity and specificity of the BD Max MVP compared with the BD Affirm platform (7).They found that the sensitivity and specificity of BD Max MVP for BV was 96.2% and 96.1%, respectively, compared to 96.2% and 81.6% for BD Affirm.The  The difference in specificity between the two assays for BV is due to the algorithmic detection of a combination of markers for BD Max MVP.The BD Max MVP Bacterial Vaginitis algorithm is predicated on Lactobacillus spp.for normal flora, G. vaginalis, A. vaginae, Megasphaera-1, and BVAB-2 as indicators of BV.In contrast, BD Affirm detects G. vaginalis alone which can be part of the normal vaginal flora (7).This could be an explanation for the increased detection of BV by BD Affirm as suggested by our data and previously published reports that BD Max MVP is s superior test (9).Initial validations were performed comparing microscopy-based gold standard Nugent's score and Amsel's criteria (6).Very recently, FDA cleared another PCR-based MVP test that introduces a BV flora-specific algorithm for BV combined with the detection of Candida species (Candida group and Candida glabrata/krusei) and Trichomonas vaginalis, from a single sample on Cepheid's GeneXpert systems (10).This will pave the way for multiple options for future vaginitis testing on molecular platforms.However, BV cannot completely be predicted based on the selected list of bacterial presence as collective Lactobacillus spp.and few other bacteria cannot directly correlate with pathogenicity and virulence.Therefore, whole-genome sequencing (WGS) and metagenomic analysis of vaginal samples in comparison with positive MVP would further solidify the accuracy of BV calling.In the meantime, the improved specificity rate of BV in MVP immensely helps in antibiotic stewardship.
We found that Candida and yeast vaginitis were detected more frequently by the BD Max MVP compared with the BD Affirm.A recent study reported that MVP significantly outperformed Affirm for the detection of Candida spp.MVP differentiates N. glabrata and Pichia kudriavzevii (formerly Candida krusei), which cause VVC, but are clinically indistinguishable from C. albicans.Speciation of Candida for VVC may be important as 50% of N. glabrata have decreased sensitivity to fluconazole, and Pichia kudriavzevii (formerly Candida krusei) is intrinsically resistant to the antifungal agent (7).However, this ultra-sensitive detection may be a double-sided sword.Despite the increased detection, it definitely helps clinicians to decide the correct antifungal regimen.However, as we observed that most of the providers resort to traditional fluconazole treatment, we are not sure of a strong awareness among providers of this new testing capability.In addition, approximately 20% of healthy women carry Candida in the normal vaginal flora.This assay is recommended to be used in symptomatic cases (11).Therefore, very sensitive testing for a targeted Candida spp.may be the reason for increased detection of Candida vaginitis leading to unnecessary antifungal therapy.As we mentioned for BV, a similar WGS analysis of VVC may address this issue.
Molecular vaginal panel can differentiate Candida albicans/Candida spp.from Nakaseomyces (formerly Candida glabrata) and Pichia kudriavzevii (formerly C. krusei).For C. albicans vaginitis, the most common treatment choices are antifungal medicine applied inside the vagina or a single dose of fluconazole taken by mouth (12).But fluconazole is not the drug of choice for the second most common vaginal yeast infection-causing agents N. glabrata and P. kudriavzevii.A longer duration of therapy (7-14 days) with a nonfluconazole azole regimen (oral or topical) is recommended by CDC (12).Therefore, molecular vaginal panels have more advantage in deciding proper therapeutic regimens as compared to DNA probe testing which only reports Candida spp.including non-albicans yeast as well making therapeutic selection challenging.
It is intriguing to observe an age-dependent yeast infection between C. albicans and N. glabrata.Our raw data and the models show that C. albicans vaginitis occurs in relatively younger women compared to N. glabrata vaginitis.This could be due to hormonal changes, immunity, and or virulence of the organism.This observation could be a new frontier to investigate further.
Limitations: We did not have a side-by-side comparison, and statistical analysis was based on the strong assumption that the two groups of samples have identical chances of detection and that the only difference is the testing method.Based on geographical location, very high numbers of patient data points compared to a normal verification exercise, and unique patients' population to CTVHCS, we assumed that the two groups of samples have identical chances of detection and that the only difference is the device.Our assumption for further modeling is justified as negative populations of both testing platforms were approximately 61% (Fig. 1A and B) In addition, both populations have approximately a 2% positive rate for TV (Fig. 1).This justifies our population compari son assumptions as it has been reported previously that there was 100% concordance between BD AffirmTV and BD Max MVP-TV verification and validation (7).Even though this comparison has limitations, this is the best model we can use for such many patient data sets for quality improvement assessment.
In conclusion, we found that BD Max MVP platform had lower detection rates for BV and high detection rates for Candida and yeast vaginitis with similar detection rates for TV.The ability to differentiate the Candida species gives clinicians the ability to choose appropriate therapy if treatment is indicated.But conversely, it can be responsible for overtreatment by clinicians.Hence, these molecular tests should be used to confirm clinical suspicion rather than as a screening test.

FIG 1
FIG 1 Summary of bacteria and yeast vaginitis cases reported by CTVHCS from September 2018 to January 2023.Red bars represent data gathered from BD Affirm, while the blue bars represent data gathered by BD Max MVP platform.

FIG 2
FIG 2 Mean proportion of positive testing: bacterial vaginitis (A) and Candida/ yeast vaginitis VVC (B) between BD Max MVP and BD Affirm.
sensitivity and specificity of BD Max MVP for Candida spp. was 98.4% and 95.4%, respectively, compared to 69.4% and 100% for BD Affirm.BD Max MVP and BD Affirm both showed 100% concordance for the detection of TV.The authors claimed that BD Max MVP has improved the accuracy of BV and VVC reporting.It is important to detect TV in the differential diagnosis of vaginitis since patient symptoms often overlap.Our finding of the same rate of TV positivity using BD Max MVP and BD Affirm in two large cohorts solidify this previous report.

FIG 3
FIG 3 Age distribution in a histogram representation for C. albicans and N.glabrata vaginitis.