Predictive Modeling Of Relapse Risk In Vulvovaginal Candidiasis Based On The Changes Of Vaginal Microbiota

Background:Vulvovaginal candidiasis(VVC)is a common lower genital tract infectious disease in women of childbearing age,which is closely related to vaginal flora disorder.After VVC treatment,it is still easy to relapse or even 4 or more times in a year to develop recurrent vulvovaginal candidiasis(RVVC).The clinical symptoms and signs caused by repeated attacks of RVVC seriously affect women's physical and mental health,marital harmony and social stability.However,there is no effective treatment at present.Recent studies have confirmed that the vaginal flora of VVC patients is diversified and RVVC is relatively single.After antifungal treatment,the changes of vaginal flora structure of VVC patients are more significant than that of RVVC,and the latter flora structure is similar in the attack and intermittent stages of the disease.It has also been suggested that the vaginal flora of VVC patients who recurred for the first time after treatment was significantly different from that of healthy women,and the flora structure tended to be stable with the increase of recurrence times.The purpose of this study was to detect and analyze the vaginal flora status and clinical information of VVC patients after the first treatment,and to construct a prediction model of VVC recurrence risk by using machine learning algorithm,in order to explore and clarify the related factors of VVC recurrence.It lays a theoretical foundation for early intervention in the future and prevention of its progress for RVVC.Objectives:The difference of vaginal flora between relapsed and non-recurrent patients with VWC within 6 months was compared.Based on the composition of vaginal flora in VVC patients,the clinical information indexes were integrated,and the prediction model of recurrence risk of VVC was constructed.Methods:Chapter 1:VVC was first diagnosed in the gynecological clinic of the first people's Hospital of Yunnan Province from August 2017 to January 2018.The vaginal secretions of all the patients in the group were collected at the time of revisit after the first treatment.According to the follow-up within half a year after treatment,the patients were divided into two groups:recurrence group(Yes)and non-recurrence group(No).After the samples were processed,the amplified products were sequenced by Illumina high-throughput sequencing technique,and then the species richness and composition of vaginal flora between the two groups were compared by BIPES biological information analysis method.The diversity index was statistically processed by SPSS20.0 software and Wilcoxon Signed Ranks test was used.P<0.05,the difference was statistically significant.At the same time,we used LEfSe(LDA=2)online statistical analysis tool to find the difference between recurrence group and non-recurrence group.Chapter 2:Using the grouping method and sequencing information of the patients in the first chapter,the relevant clinical information of the patients in the past one month was recorded at the time of the first visit to the hospital.It included age,vaginal washing habits,contraceptive methods,sexual intercourse methods,use of antibiotics,hormone use and VVC score at the first attack.At the same time,the random forest prediction model of machine learning method was used to analyze the relationship between bacteria,clinical information and recurrence outcome,and the related variables were screened out.Then the verification set is substituted into the model,and the OOB(Out-of-bag)error rate is minimized and the contribution of each variable is predicted.Finally,the effectiveness of the model is evaluated by obtaining the area under the curve(Area Under ther Curve,AUC)of the prediction model and the working characteristics of the subjects(Receivert Operating Characteristic curve,ROC).Result:Chapter 1:(1)After the first treatment of VVC patients,Lactobacillus was the dominant genus in the vaginal flora,and there were Streptococcaceae,Pseudomonadaceae,Gardnerella,Prevotella,Atopobium,Finegoldia and other bacteria in the vaginal flora of some patients.Compared with the non-recurrent group,the recurrent group had an increasing trend in Lactobacillus and Prevotell,but a decreasing trend in Streptococcaceae,Pseudomonadaceae,Gardnerella and Atopobium.(2)Compared with the non-recurrent group,the flora richness of the recurrent group was higher than that of the non-recurrent group,and the flora structure of the recurrent group was more stable than that of the non-recurrent group.(3)LEfSe analysis was used to find the different genera between the two groups.It was found that the richness of Enterobacteriales,Vibrionales,Enterobacteriaceae and Phenylobacterium increased in the recurrence group,but in the non-recurrence group,the abundance of Coriobacteriales,porphyromonadaceae,Coriobacteriaceae,Porphyromonas,Coriobacteriia was increased.Chapter 2:The prediction model was established to screen out seven related variables related to the risk of recurrence of VVC in clinical information and flora information.The inclusion of variable validation set into the model makes the OOB error rate reach a minimum of 28.06%.At the same time,according to the calculated contribution value,the vaginal washing habit was the most important related factor.Porphyromonas,Atopobium and the use of antibiotics was negatively correlated with the risk of recurrence of VVC.Chryseobacterium,vaginal washing habits,age and hormone drugs use were positively correlated.In addition,the area under the curve of the prediction model and the working characteristics of the subjects is 0.959,and the 95%confidence interval is(0.929<0.989).The prediction efficiency of the model is located in the confidence interval.Conclusion:There were differences in the richness and composition of vaginal flora after the initial treatment of VVC,and it was related to its future recurrence.Combined with vaginal flora information and clinical information,a risk model for predicting the recurrence of VVC can be successfully established,and the prediction effect is good.The model suggests that Porphyromonas,Atopobium and the use of antibiotics is the protective factor for the risk of VVC recurrence,while Chryseobacterium,vaginal washing habits,age and hormone drugs use are the risk factors for recurrence.Among them,vaginal washing habit is the most important factor for the recurrence of VVC.