Analysis Of Clinical Factors Affecting Outcome Of IVF And Establishment Of A Model For Predicting Live Birth Rates

[Background]China developped assisted reproductive technology(ART) since1980s.After20years develepment,China has absorbed the advanced experience from the west countries and also developped technology that suited to our chinese people.Nowadays,we can make the women stay at a perfect stage by controlling skillfully on the every step of ART. In spite of these, ART can not reach a high success rates as all of us expect.ART is a kind of technology that requires a high degree of laboratory environment and doctor’s proficiency,so it also spends a lot of money and energy of the infertile couples.Because every infertile couples have their own characteristic,they will face different outcomes with the same therapeutic milieu.Since now, many foreeign investigators have focused on the factors affecting IVF outcomes.Many indices were used to establish the outcomes of patients going on IVF which could give a predicted value before the real treatment.China has no predictive models for IVFoutcomes.Our research focus on the Chinese people,analyse the index known before the real treatment that can affect the outcome of IVF,try to establish a simple model to help predict the outcomes. Part Ⅰ[Objective]By retrospectively analyzing the clinical data of patients under treatment of IVF in Nanfang hospital from2003to2009, this study aims to understand how basic clinic factors affect the outcomes of IVF,and tries to establish a predict model to assess the patients before real treatment.[Methods]This study retrospectively analyzed the4834IVF/ICSI/FET cycles of Nanfang Hospital from2003to2009. Exclusion criteria:(1)Treatment cycles that were not IVF, such as IUI, GIFT;(2) the use of oocyte/embryo/sperm donation;(3)cycle cancellation;(4)lack of follow-up.We collected the basic clinic datas of patients which could be knew before the treatment.We established the databases using spss13.0.Chi-square test(or Linear-by-Linear Association test) was used to compare the live birth rates among the different group of different indices.Univariate logistic analyse and multivariate analyse were used to observe the relationship between the predictor candidates and live birth rates.P<0.05meant there were statistic difference.The predictive ability of the model was assessed by determining the discrimination,using the area under the curve of receiver operator characteristics(AUROC),and its calibration.Calibration referred to the level of correspondence between the calculated live birth chances and the observed proportion of live births. We used the scatter diagram to observe calibration.[Results]1、The average age of the overall patients was32.29±4.45,the average duration was5.20±3.41,The top three cause were tubal,combination known causes and male.The pregnacy rate was37.28%.Live birth rate was29.85%.The ectopic pregnancy rate was3.83%.The abortion rate was16.09%。2、Univariate analyse showed that the female age,the cause of infertility,the kind of ART,the history of previous IVF,the number of cycles,duration of infertility,previous pregnancies,previous livebirths,kind of infertility would affect the IVF outcome.3、Multivariate analyse showed that the female age,the cause of infertility,the kind of ART, the history of previous IVF would affect the IVF outcome.4、The factors affect live birth rates4.1Female ageThe highest livebirth rates were in the age-group25-30years with live birth rate of37.9%. Younger women in the age-group under25years had a little lower live birth rates of37.8%.There was a sharp decline in older women.In the age-group of older than40years,the live birth rates were only11.9%.There were29cycles in the group of older than45years with no success.The live birth rates declined with the age grew(Linear-by-Linear Association test was106.532, P<0.001),this trend was significantly.4.2Cause of infertilityThe highest live birth rates were in the cause-group unknown which was39.4%.The group of male cause was a little lower with36.3%.The combination known causes was the lowest group whose live birth rate was only26.3%.There were statistical differences among different groups(χ2=23.495, P=0.001).4.3Treatment typeOur research classfied the treatment to three types:IVF-ET、ICSI-ET、FET(IVF受精)、FET(ICSI受精)The highest live birth rates were in the type-group ICSI-ET which was34.5%, The lowest live birth rates were in the type-groupFET(ICSI受精)which was15.4%. There were statistical differences among different groups(χ2=107.916, P<0.001).4.4History of treatmentWomen who had at least one previous IVF treatment had lower rate of live birth than women who had no history of treatment.(36.7%VS20.3%, P<0.001).5、Establishment of a model for predicting live birth rates and its evaluationWe established regression equation in multivariate logistic analyse. The regression equation had statistically sense with its-2Log likelihood296.089 (df=14,P<0.001).The area under ROC was0.653,the standard error was0.008, P<0.001,95%CI(0.636,0.669).Choosing P=0.3064as the cut off level.the sensitivity of the model was64.0%,and the specificity of the model was60.1%.The Hosmer-Lemeshow test value was5.098,自由度=8, P=0.747.It meant that there were no significant difference between the predicted frequency and the observed frequency. Consequently,we thought the model was rather good.We compared the observed to predicted frequency of successful live birth by tenths of the distribution of the linear prediction model by scatter diagrams,the scatter dots trended to be the angular bisector of the axis.The model can provide a basis for counselling and informing couples of their prognosis in terms of low,moderate,or high odds of success.[Conclusion]1、Fmale age,the cause of infertility,the kind of ART,the history of previous IVF could be used to predict the outcome of IVF before embryo transplantation treatment.2. The discrimination of the models were poor,but the calibration of the models were excellent.These models coule be used to assess patients before treatment and calculate their likely prognosis in terms of low or high odds of success.Part Ⅱ[Objective]On the basis of part Ⅰ,we add some clinic indices to make full use of the datas,and want to impove the predictive ability of the model.[Methods]This study retrospectively analyzed the3014IVF/ICSI cycles of Nanfang Hospital from2003to2009. Exclusion criteria:(1) Treatment cycles that were not IVF, such as IUI, GIFT;(2) treatment cycles that were not fresh embryo transfer;(3) the use of oocyte/embryo/sperm donation;(4) cycle cancellation;(5) lack of follow-up. We collected the basic clinic datas of patients which could be knew before the treatment.We established the databases using spss13.0.Chi-square test (or Linear-by-Linear Association test)was used to compare the live birth rates among the different group of different indices.Univariate logistic analyse and multivariate analyse were used to observe the relationship between the predictor candidates and live birth rates.P<0.05meant there were statistic difference.The predictive ability of the model was assessed by determining the discrimination,using the area under the curve of receiver operator characteristics(AUROC),and its calibration.Calibration referred to the level of correspondence between the calculated live birth chances and the observed proportion of live births.We used the scatter diagram to observe calibration.[Results]1、Univariate analyse showed that the female age,the cause of infertility, duration of infertility,basal antral follicle count, the number of cycles, previous pregnancies, kind of infertility,previous livebirths, the history of previous IVF,basal FSH level,basal FSH/LH would affect the IVF outcome.2、The screening of affected factors,Establishment of a model for predicting live birth rates and its evaluationMultivariate analyse showed that the female age, basal antral follicle count,the history of previous IVF would affect the IVF outcome. The regression equation had statistically sense with its-2Log likelihood189.384(df=8,P<0.001).The area under ROC was0.642,the standard error was0.010, P<0.001,95%CI (0.622,0.663).Choosing P=0.3412as the cut off level,the sensitivity of the model was68.2%,and the specificity of the model was53.5%.The Hosmer-Lemeshow test value was2.261, P=0.944.It meant that there were no significant difference between the predicted frequency and the observed frequency. Consequently,we thought the model was rather good.We compared the observed to predicted frequency of successful live birth by tenths of the distribution of the linear prediction model by scatter diagrams,the scatter dots trended to be the angular bisector of the axis.The model can provide a basis for counselling and informing couples of their prognosis in terms of low,moderate,or high odds of success.[Conclusion] 1. Female age,basal antral follicle count, the history of previous IVF could be used to predict the outcome of IVF before fresh embryo transplantation treatment.2. The discrimination of the models were poor,but the calibration of the models were excellent.These models could be used to assess patients before treatment and calculate their likely prognosis in terms of low or high odds of success.3. The fresh embryo transfer cycles model was better because of more predictor candidates.[Summary]Our study retrospectively analyzed the IVF/ICSI/FET cycles of Nanfang Hospital from2003to2009.We analyzed the factors affecting outcome of IVF and established models for predicting live birth rates.For all embryo transfer cycles, Fmale age,the cause of infertility,the kind of ART,the history of previous IVF could be used to predict the outcome of IVF before embryo transplantation treatment.For fresh embryo transfer cycles, female age,basal antral follicle counts, the history of previous IVF could be used to predict the outcome of IVF.The discrimination of the models were poor,but the calibration of the models were excellent.These models could be used to assess patients before treatment and calculate their likely prognosis in terms of low or high odds of success which could help the patients to make realistical dicisions.