Diagnosing Bacterial Vaginosis with a Novel, Clinically-Actionable Molecular Diagnostic Tool

Abstract

15Bacterial vaginosis is a common condition among women of reproductive age and is 16 associated with potentially serious side-effects, including an increased risk of preterm birth. 17Recent advancements in microbiome sequencing technologies have produced novel insights into 18 the complicated mechanisms underlying bacterial vaginosis and have given rise to new methods 19 of diagnosis. Here we report on the validation of a quantitative, molecular diagnostic algorithm 20 based on the relative abundances of ten potentially pathogenic bacteria and four commensal 21Lactobacillus species in research subjects (n = 172) classified as symptomatic (n = 149) or 22 asymptomatic (n = 23). We observe a clear and reinforcing pattern among patients diagnosed by 23 the algorithm that is consistent with the current understanding of biological dynamics and 24 dysregulation of the vaginal microbiome during infection. Using this enhanced assessment of 25 the underlying biology of infection, we demonstrate improved diagnostic sensitivity (93%) and 26 specificity (90%) relative to current diagnostic tools. Our algorithm also appears to provide 27 enhanced diagnostic capabilities in ambiguous classes of patients for whom diagnosis and 28 medical decision-making is complicated, including asymptomatic patients and those deemed 29 "intermediate" by Nugent scoring. Ultimately, we establish CLS2.0q as a quantitative, sensitive, 30 specific, accurate, robust, and flexible algorithm for the clinical diagnosis of bacterial vaginosis -31 importantly, one that is also ideal for the differential diagnosis of non-BV infections with 32 clinically similar presentations. 33 34

Molecular Diagnosis of Vaginitis: Comparing Quantitative PCR and Microbiome Profiling Approaches to Current Microscopy Scoring

“…In conclusion, our study confirmed that molecular methods are more accurate and objective for BV diagnosis than microscopy, as qPCR was able to provide higher resolution with species-level data. Additionally, the BV interpretive algorithm was able to resolve the BV status for samples designated altered flora by Nugent (24). The comprehensive qPCR assay included possibilities for alternative diagnoses, revealing a higher prevalence of AV pathogens than expected.…”

“…The approach involves a linear transformation that yields a set of weighted species-specific values which are then combined to produce a score for each patient. In previous evaluations, it has shown an overall sensitivity of 93% and a specificity of 90% compared to gold-standard testing (24).…”

Bacterial Vaginosis: Current Diagnostic Avenues and Future Opportunities

“…Although Amsel's criteria and the Nugent scoring system are considered as the “gold standard” for BV diagnosis, problems still exist because of the “interobserver variability” and the fact that the “intermediate vaginal microbiome” do not necessarily indicate disease progression to BV or vice versa (van de Wijgert et al, 2014 ). The advances in machine learning and its application in other fields have been followed by attempts to apply computer algorithms in BV diagnosis (Baker et al, 2014 ; Beck and Foster, 2014 , 2015 ; Carter et al, 2014 ; Song et al, 2017 ; Jarvis et al, 2018 ).…”

“…Computer algorithms could potentially have a wide range of applications that may help clinicians and researchers to search for models that identify features relevant to BV diagnosis, to assess relative bacterial abundance data (qPCR) to diagnose BV, or to analyze bacterial morphotypes on microscope images for more accurate Nugent scoring results (Beck and Foster, 2014 , 2015 ; Carter et al, 2014 ; Song et al, 2017 ; Jarvis et al, 2018 ). One of the first attempts to apply machine learning algorithms in BV diagnosis was done by Beck and Foster ( 2014 ), where the authors first grouped the correlations in microbial relative abundance data from studies by Ravel et al ( 2011 ) and Srinivasan et al ( 2012 ) and built different classification models (based on Amsel's criteria or Nugent scoring) using three different types of machine learning algorithms [“genetic programming” (GP), “logistic regression” (LR) and “random forest” (RF)].…”

“…Computer algorithms can also be incorporated into a molecular diagnostic tool to assess the relative abundance data of BV-associated bacteria for diagnosing BV. In a recent study conducted by Jarvis et al ( 2018 ), a novel qPCR-based diagnostic tool (that detects ten BV-associated bacterial species and four Lactobacillus species) with a diagnostic machine learning algorithm (CLS2.0q) was developed. Evaluation of this molecular diagnostic tool with 172 women (149 symptomatic and 23 asymptomatic) resulted in a sensitivity of 93% and a specificity of 90% when compared with Nugent scoring (Jarvis et al, 2018 ).…”