3Years Follow-up Outcome Study Of Traditional Chinese Medicine Intervention In Ischemic Stroke And The Construction Of Risk Model

1BackgroundLarge sample, long time follow-up helps to study the long-term prognosis of disease, while mortality and recurrence are the clinical events which are the greatest impact on ischemic stroke and the patients most want to avoid, and are the end outcome evaluation index. At present, ischemic stroke study time is not long, usually less than one year, and the sample size is limited, taking the long-term outcome of death and recurrence as the evaluation index is unusual in the field of traditional Chinese medicine(TCM) follow-up. In addition, early comprehensive rehabilitation program intervention study for ischemic stroke patients is less, and the influence on the long-term prognosis remains unclear.On the other hand, risk prediction model could guide the early warning. Research at home and abroad almost paid attention to modern risk factors predicting the prognosis of the disease, lack of the TCM items, which was difficult to meet the clinical needs of TCM clinical practice. Influencing factors screening is the most important step in the construction of risk model. Cox single-factor and multi-factor analysis is the general method, while it unable to automatically handle missing data in the survival data, analysis of numerous covariate data will lost some information.2ObjectiveThe paper has three aims. In the first place, we want to observe natural outcome of the ischemic stroke patients treated by early rehabilitation program intervention after discharge, and study whether early TCM intervention program can influence ischemic stoke death and recurrence in3years. In the second place, this study analyses the death and recurrence influencing factors of ischemic stroke patients, and construct the corresponding risk model based on modern risk factors and TCM syndrome, in order to provide reference proposal for second prevention. In the third place, we hope to evaluate two influencing factors screening method used in the survival data, so as to provide methodological guidance for model building in the future.3Content3.1Observe the all-cause mortality, mortality related to ischemic stroke, mortality directly related to ischemic stroke, mortality indirectly related to ischemic stroke, and recurrence of ischemic stroke, and compare survival and recurrence of the long-term prognosis in patients with early intervention of TCM comprehensive rehabilitation plan and modern rehabilitation through3-year follow up. In addition, death and recurrence number needed to treat(NNT) are assessed.3.2Build a3-year prognosis risk model of ischemic stroke by screening modern risk factor and TCM syndrome which are influencing factors of death and recurrence between the two groups. Cox single-factor and multi-factor analysis and random survival forests(RSF) are the two screening methods used by us.3.3Evaluate the two screening methods of influencing factors according to the3-year forecasting performance of the risk model judged by area under the receiver operating characteristic curve(ROC).4Method4.1Cases criterion1059patients with acute ischemic stroke were enrolled in12clinical units throughout the country in2008June27to June3,2010. The TCM and Western medicine(WM) diagnosis, inclusion, exclusion, exit, withdrawal criteria respectively was established.4.2Research Design4.2.1Design TypeA prospective, multicenter, pragmatic randomized controlled trial method was adopted.4.2.2Random AllocationClinical research central stochastic system, in accordance with the ratio of2:1,1059patients were randomly divided into705cases of Chinese medicine comprehensive rehabilitation program group and354cases of modern rehabilitation programs group, achieving the concealment of random allocation program.4.2.3Treatment ProgramsTCM group was given decoction of Chinese medicine, traditional Chinese medicine injections, acupuncture and massage therapy, treatment for10-21days. WM group was given the modern rehabilitation, including the setting of the limbs in good position, passive movement of the joint, limb function training, the rehabilitation process of suppressing spasms, the course of10-20days. Basic medical treatment to follow the specific methods of "prevention and treatment guidelines of China cerebrovascular disease" developed in2007.4.2.4Evaluation IndexDeath and recurrent events were followed up for3years.4.2.5Influencing factors collected indexInfluencing factors collected index include:general information, basic condition, TCM syndrome, primary comorbidities, whether to insist on taking drugs after discharge,27factors, etc.4.3Statistical analysis methodsWe generally described the characteristics of the different groups, death and without death crowd, recurrence and without recurrence crowd by SPSS16.0software. Using imput.rsf, survival, Random Survival Forest, survivalROC software package of R2.15.3software for data analysis.4.3.1Outcome ResearchDescribe the the all-cause mortality, mortality related to ischemic stroke, mortality directly related to ischemic stroke, mortality indirectly related to ischemic stroke, and recurrence of ischemic stroke occurred of different groups in3years by the method of survival analysis, comparing the differences in proportion of death and recurrence of different groups by the log-rank test.4.3.2Influence factors analysis and the establishment of risk modelsWith27survey factors as the covariate, we screen the influencing factors of prognosis of different groups respectively using the Cox single-factor and multi-factor method, RSF method, establish multivariate Cox risk model based on the screening influencing factors to predict the risk of death and recurrence in3years.4.3.3Risk model long-term predict performance evaluationThe area under ROC curve to evaluate the long-term forecast performance of the risk model, compare the ability of screening influence factors using Cox single-factor and multi-factor approach and RSF method.5Result5.1Completion of follow-up860cases(including80cases death) has completed the3-year follow-up work in1059cases by the end of December2012, the follow-up rate was81.2%. The average follow-up time of TCM group and WM group is3.22and3.18years respectively, and the longest follow-up time is4.37and4.35years respectively.5.2The general descriptionThe proportion of TCM group patients after discharge insisting on taking the Chinese medicine is higher than WM medicine group, there is statistically significant(P<0.05), while other items shows no significance between the groups(P>0.05).Take Whether to death as a grouping variable, the results shows that age, national institute of health stroke scale(NIHSS) score, the proportion of tanshimengshen syndrome and relapse, systolic blood pressure, diastolic blood pressure, low density lipoprotein cholesterol(LDL-C) values on admission are higher in the death group(P<0.05). However, fugl-meyer assessment(FMA) score, barthel index(BI) score, high density lipoprotein cholesterol(HDL-C) values on admission are lower in the death group(P<0.05).Take Whether to relapse as a grouping variable, the results shows that LDL-C values on admission and the proportion of combined with dyslipidemia are higher in the relapse group(P<0.05). However, HDL-C values on admission are lower in the relapse group(P<0.05).5.3Outcome Research 5.3.1Mortality analysis5.3.1.1All-cause mortalityThrough3-year follow up,80cases died.49cases died in the TCM group(N=705), risk proportion of death is6.95%.31cases died in the WM group(N=354), risk proportion of death is8.76%. The average survival time of TCM group and WM group is2.38and2.32years respectively, no statistical significance is found between the groups through log-rank test(P>0.05).5.3.1.2mortality related to ischemic strokeThrough3-year follow up, there are28cases died because of ischemic stroke.15cases died in the TCM group(N=705), risk proportion of death is2.13%.13cases died in the WM group(N=354), risk proportion of death is3.67%, no statistical significance is found between the groups through log-rank test(P>0.05).5.3.1.3mortality directly related to ischemic strokeThrough3-year follow up, there are15cases died directly related to ischemic stroke.9cases died in the TCM group(N=705), risk proportion of death is1.28%.6cases died in the WM group(N=354), risk proportion of death is1.69%, no statistical significance is found between the groups through log-rank test(P>0.05).5.3.1.4mortality indirectly related to ischemic strokeThrough3-year follow up, there are13cases died indirectly related to ischemic stroke.6cases died in the TCM group(N=705), risk proportion of death is0.85%.7cases died in the WM group(N=354), risk proportion of death is1.98%, no statistical significance is found between the groups through log-rank test(P>0.05).5.3.2Recurrence analysisThrough3-year follow up,76cases relapsed.47cases relapsed in the TCM group(N=705), risk proportion of death is6.67%.29cases relapsed in the WM group(N=354), risk proportion of death is8.19%. The average free-relapse time of Chinese medicine group and WM group is2.28and2.22years respectively, no statistical significance is between the groups found through log-rank test(P>0.05).5.3.3Death NNT evaluationCompared with WM goup, there was1death case short out of every55patients in TCM group. NNT95%CI is (-59,18). As a result of the95%CI include0, so the comparison between the two rehabilitation programs reducing the risk of death shows no statistical significance.5.3.4Recurrence NNT evaluationCompared with WM goup, there was1recurrence case short out of every66patients in TCM group. NNT95%CI is (-53,20). As a result of the95%CI include0, so the comparison between the two rehabilitation programs reducing the risk of recurrence shows no statistical significance.5.4Screening for influencing factors of death and relapse and risk model establishment5.4.1Variable screen and model establishment based on Cox single-factor and multi-factor analysis5.4.1.1Death within three years risk prediction modelTake survival time as outcome variable, influencing factors of death within three years are as below:age(X1,P=0.000), FMA score(X2, P=0.000), tanshimengshen syndrome(X3, P=0.000), whether to relapse(X4, P=0.000), Systolic pressure(X5, P=0.000), LDL(X6, P=0.014), HDL(X7,P=0.002), whether the patients continue to take antithrombotic drugs after discharge (Xg, P=0.001), whether the patients continue to receive a basic curative treatment after discharge(X9,P=0.003). Then build the risk model: h(t|X)=h0(t)exp(0.064*X1-0.022*X2+1.874*X3+1.590*X4+0.026*X5+0.329*X6-0.586*X7-0.952*X8-0.935*X9)5.4.1.2Relapse within three years risk prediction model ofTake no relapse time as outcome variable, influencing factors of relapse are:BI score(X1, P=0.032), fasting plasma glucose level(X2, P=0.037), combined with dyslipidemia(X3, P=0.003). Then build the risk model as below: h(t|X)=h0(t)exp(-0.008*X1+0.072*X2+0.842*X3)5.4.2Variable screen and model establishment based on RSF5.4.2.1Death within three years risk prediction modelTake survival time as outcome variable, influencing factors for death within three years are as below:age(X1,P=0.000), BI score(X2, P=0.000), tanshimengshen syndrome(X3, P=0.000), whether to relapse(X4, P=0.000), Systolic pressure(X5, P=0.000), LDL(X6, P=0.003), HDL(X7, P=0.000), combined with dyslipidemia(X8, P=0.002), whether the patients continue to take antithrombotic drugs after discharge (X9, P=0.000), whether the patients continue to receive a basic curative treatment after discharge(Xio, P=0.000). Then build the risk model as follows: h(t|X)=h0(t)exp(0.066*X1-0.023*X2+1.439*X3+1.112*X4+0.026*X5+0.356*X6-0.637*X7+0.804*X8-1.236*X9-1.437*X10)5.4.2.2Relapse within three years risk prediction modelTake no relapse time as outcome variable, influencing factors of relapse including age(Xi, P=0.076), NIHSS score(Xi, P=0.051), BI score(X3, P=0.037), tanshimengshen syndrome(X4, P=0.062), fasting plasma glucose level(X5, P=0.038), combined with dyslipidemia(X6, P=0.000). Then build the risk model: h(t|X)=ho(t)exp(0.018*X1+0.011*X2-0.009*X3+0.475*X4+0.068*X5+0.817*X6)5.5Death and relapse risk model evaluation5.5.1Death risk model evaluationThe Cox death risk prediction model built based on screening influencing factors via RSF method(AUC=0.795) had better performance in3-year predication than Cox single-factor and multiple factors analysis(AUC=0.782).5.5.2Relapse risk model evaluationThe Cox relapse prediction model built based on screening influencing factors via RSF method(AUC=0.621) performed better in3-year predication than Cox single-factor and multiple factors analysis(AUC=0.605).6Conclusion6.1After3-year period follow-up, it inllustrates that for TCM group both all-cause motality, mortality related to ischemic stroke, mortality directly related to ischemic stroke, mortality indirectly related to ischemic stroke, and recurrence hazard are lower than WM group(P>0.05). Indeed, survival time and relapse-free time in TCM group are much higher than WM group, which indicates a better developing trend for TCM group. Compared with WM goup, there was1death case short out of every55patients in TCM group. Compared with WM goup, there was1recurrence case short out of every66patients in TCM group. However, considering that this study is influened by sample size, there is no obvious significance between the groups.6.2Take survival time as an endpoint variable, and research method is random survival forests (RSF). Consequentially, impact factors in3year are:age, BI score, tanshimengshen syndrome, whether to recurrence, systolic blood pressure, LDL-C, HDL-C, combined with dyslipidemia, insistency of antithrombotic drugs intake after discharged, and insistency of basic treatment drugs.6.3Set relapse-free time as endpoint variable(Research method:RSF). Influencing factors on3-year recurrence indicate:age, NIHSS score, BI score, tanshimengshen syndrome, fasting blood-glucose, combined abnormal lipid metabolism.6.4This study has builded ischemic stroke3-year death and recurrence prediction model and prognosis index. Cox proportional hazard model is established by screening infulencing factors under RSF. In comparison to Cox single-factor and multi-factor screening method, the model3-year prediction performance based on RSF shows a better vision.7Innovation7.1Taking death and recurrence as assessment indicators, it analyses the outcome effects of early intervention of TCM rehabilitation therpy by a3-year long-trem follow up study in the first time.7.2Screening important morden risk factors or TCM syndrome of ischemic stroke prognosis among a mass of concomitant variables. Besides, it takes into account missing value of survival data.7.3Characteristic TCM content is integrated in risk assessment model of ischemic stroke prognosis, which increases TCM clinical significance.