| BACKGROUND:Irritable bowel syndrome(IBS)is a functional bowel disease and is defined by its symptoms of recurrent abdominal pain and changes in bowel habits.There appears to be no evidence which confirms that there are specific histological or biochemical abnormalities which can be used to confirm the symptoms reported by people with IBS.There are four types of IBS:IBS with predominant diarrhoea(IBS-D),IBS with predominant constipation(IBS-C),IBS with mixed bowel habits(IBS-M),and IBS unclassified.IBS is similar to "syndrome of TongXie"recorded in Chinese medicine classical literatures.Tongxie Yaofang,a classical prescription for "syndrome of TongXie",is considered to be the main clinical prescription for IBS-D.Existing systematic reviews(SRs),including preliminary studies by our research team,have shown that traditional Chinese medicine(TCM)is effective for IBS.In recent years,sufficient evidence.from a large number of randomized controlled trials(RCTs)on TCM for IBS has been published in both domestic and international journals and demonstrated that IBS is an ideal disease for building an evaluation model of efficacy.Complex networks(CNs)and Bayesian networks(BNs)are two of the main techniques used for data mining and analysis.CNs can be used to describe and mine the structure and the internal implicit connections of complex systems in the form of network diagrams.This method has spread to many disciplines including:business,communications,economics,marketing,medicine,public health,political science,psychology and sociology.BNs express the joint probability distribution among variables in the form of a graph.This technique is widely adopted in medical,industrial,financial,computer,military,ecological,agricultural and animal husbandry fields.In the TCM field,CNs and BNs are primarily applied to determine patterns of drug compatibility and develop diagnostics research,respectively.As far as we know,there have been no studies to explore the relationship between the elements:participants(P),interventions(I),comparisons(C)and outcomes(O)which affect the clinical efficacy of TCM.PURPOSE:Taking IBS as an example,this project aims to explore in depth the relationship between P,I,C and O in RCTs of TCM for IBS using CNs and BNs techniques,in order to explore theoretical models and data analysis path suitable for the evaluation of TCM efficacy.METHODS:(Study 1:Establishment and preliminary application of a PICO efficacy evaluation model for irritable bowel syndrome based on complex network techniques)An electronic database search was systematically conducted from MEDLINE,Embase,CENTRAL,CNKI,Wanfang,VIP,and Sinomed(due November 2019).RCTs on TCM for IBS met the inclusion criteria were screened and obtained.Participants of included studies were IBS patients diagnosed by Kuris,Manning or Rome criteria.Interventions in the experimental groups were Chinese medicines,including single Chinese medicine,a mixture of several Chinese medicines,Chinese proprietary medicine,or Chinese herbal extracts.There were no detailed restrictions on types of drugs,dosage form,means of delivery,dosage,and course of treatment.Interventions in control groups could be no treatment,placebo,routine western medicine,or symptomatic treatment.Outcomes included overall symptoms improvement of IBS(total effective rate and overall symptom score),quality of life,relapse rates,improvement of specific symptoms(abdominal pain,bloating,diarrhea or constipation)and improvement of psychological states(depression and anxiety).Two of our team members extracted data independently.If disagreement existed after the consensus meeting,the author of this paper would arbitrate any differences in opinions from team members.This study defined the nodes in the IBS-PICO-Complex Network(IBS-PICO-CN)as P,I/C and O.P consisted of age(P1),course of the disease(P2),setting(P3),and TCM syndrome differentiation type(P4);I/C was made up of Chinese medicine(Z),a combination of Chinese and western medicine(Z+X),or western medicine/placebo(X);O contained 10 outcomes,for example total efficiency rate(O).Data processing was conducted by following steps:unifying units of continuous variables,converting score data into improvement rates,and performing discrete assignment based on the experience of clinical experts and methodological experts.The Numpy package of Python software(v3.6)was used to establish 0-1 matrixes between nodes in CN based on the "co-occurrence principle".We also quantified the measures for importance of the network nodes,for example closeness centrality.The singular value decomposition method was adopted simultaneously to calculate the correlation value of two nodes.The correlation value of the two nodes were then added to obtain the cumulative correlation value of the multi-node combination,as the basis of identifying the most and least effective options.(Study 2:Establishment and preliminary application of a PICO efficacy evaluation model for irritable bowel syndrome based on Bayesian network techniques)In this study,the nodes defined in the IBS PICO Bayesian network(IBS-PICO-BN)were the same as those in Study 1.Following the data from Study 1,a more detailed discrete assignment was performed.According to the experience and knowledge of experts in gastroenterology and methodology,the IBS-PICO-BN structural model was constructed,and the parameter learning of IBS-PICO-BN was conducted based on the data.Furthermore,based on "Bayes’ theorem",Bayesian inference was performed to calculate the conditional probability distribution of the multi-node combination options using the R software(v3.6.3)bnlearn package.The most and least effective options were selected according to the size of the conditional probability.(Study 3:Herbal medicine for irritable bowel syndrome:an overview of systematic reviews and Bayesian network meta-analysis)Databases searched in this study was consistent with those used in Study 1.Searches were conducted up to February 2020.SRs of herbal medicine for IBS were screened and analyzed for the overview.The original studies embedded in the SRs included in this study were RCTs or randomized crossover trials with a washout period.The participants were IBS patients diagnosed by Rome,Manning,or Kuris criteria.Interventions in experimental groups were herbal medicines(single herbs,extracts,herbal mixtures,or Chinese proprietary medicines)alone or in combination with routine western medicines with oral delivery of drugs.There were no restrictions on dosage form.The control group interventions were routine western medicines.Co-interventions were allowed as long as all arms of the randomized allocation received the same co-intervention.The outcomes of each SR reported included at least one of the following:(1)Global symptoms or specific symptom relief,measured by score or number of improved cases;(2)Quality of life;(3)Relapse rate;(4)Psychological states.Two members conducted independent screening and data extraction from the literature.A measurement tool 2 of systematic reviews assessment(AMSTAR2)and preferred reporting items for systematic reviews and meta-analyses(PRISMA)were used to evaluate the methodological quality and reporting quality of SRs.The Cochrane risk of bias(ROB)tool was used to evaluate the risk of bias of RCTs included.Pairwise meta-analysis and Bayesian network meta-analysis(NMA)was used to quantitatively synthesis and pool the relative risk(RR),mean difference(MD)and their 95%confidence interval(CI).The ranking of the drugs was obtained by calculating the area under the cumulative ranking curves(SUCRA).RESULTS:(Study 1)The rate of unreported data of other outcomes were more than 50%among all of the included studies.However,the rate of unreported data of total effective rate was only 4.91-5.74%.Therefore,only the total effective rate was included by the O elements of IBS-PICO-CN model constructed in this study.Based on the 623 RCTs(518 for IBS-D and 105 for IBS-C)of TCM for IBS,the IBS-PICO-CN model was constructed by CNs method.After combining with the singular value decomposition,a total of 122 potential most(total effective rate over 90%)and least effective(total effective rate less than 75%)PICO combination options were qualitatively excavated.This result contributed information to inform clinical practice.These options suggested:(1)The most effective options for IBS-D were the options involving Chinese medicine alone or combination of Chinese and western medicine.It meant that the following treatment options(or the combinations of PIC)might make the outcome of total effective rate exceed 90%:① for outpatients who were 30-40 years old with liver stagnation and spleen deficiency,or spleen deficiency and dampness accumulation,patients were treated with the prescriptions aimed to soothe the liver and invigorate the spleen,or strengthen the spleen and eliminate dampness;② for outpatients who were over 40 years old and had a disease course of 4-6 years,with liver stagnation and spleen deficiency or spleen deficiency and dampness accumulation,patients were treated with prescriptions to soothe the liver and invigorate the spleen,or strengthen the spleen and eliminate dampness;③ for inpatients who were 30-40 years old and had a disease course of not less than 4 years,with liver stagnation and spleen deficiency or spleen deficiency and dampness accumulation,patients were treated with the prescriptions to soothe the liver and invigorate the spleen,or strengthen the spleen and eliminate dampness;④ for outpatients who were 30-50 years old and had a disease course of less than 4 years,with liver stagnation and spleen deficiency,patients were treated with the prescriptions to soothe the liver and invigorate the spleen,combined with antidiarrheal agents and probiotics,or antispasmodic agents and probiotics;⑤ for outpatients who were over 50 years old and had a disease course of less than 4 years,with liver stagnation and spleen deficiency,patients were treated with prescriptions to soothe the liver and invigorate the spleen,combined with antidiarrheal agents,or antispasmodic agents and probiotics.(2)The least effective options for IBS-D were the options involving western medicine alone.Since plenty of options were obtained,the options involving liver stagnation and spleen deficiency pattern and spleen deficiency and dampness accumulation pattern were used as examples.The following treatment options(or the combinations of PIC)might generate an outcome with a total effective rate of less than 75%.For patients with liver stagnation and spleen deficiency:① age over 30 years old,treated with antispasmodics alone;② age 30-50 years old,course of disease in 2-6 years,treated with probiotics alone;③ age 30-40 years old,course of disease in 4-6 years,treated antidiarrheal agents alone or combined antispasmodic and probiotics;④ not less than 50 years old,with a disease course of 4-6 years,treated with probiotics alone.For patients with spleen deficiency and dampness accumulation:① age 30-40 years old,with a disease course of 2-6 years,treated with antidiarrheal agents or probiotics or antispasmodics alone,or combined antispasmodics and probiotics;② age 40-50 years old,course of disease in 2-4 years,treated with antidiarrheal or probiotics or antispasmodics,or combined antispasmodics and probiotics;③ age 40-50 years old,course of disease in 4-6 Years,treated with probiotics or antispasmodics alone.(3)The most effective options for IBS-C were the options involving Chinese medicine alone or combination of Chinese and western medicine.It meant that the following treatment options(or the combinations of PIC)might make the outcome of total effective rate exceed 90%:① for outpatients who were 30-40 years old and had a disease course of 2-4 years,with weakness of spleen and stomach,or liver depression and qi stagnation,these patients were treated with the prescriptions to invigorate the spleen and replenish qi,or soothe the liver and regulate qi;② for outpatients who were 40-50 years old with weakness of spleen and stomach,these patients were treated with the prescriptions which had the principles of invigorating spleen and replenishing qi;③ for outpatients who were 40-50 years old and had a disease course of 2-4 years,with liver depression and qi stagnation,these patients were treated with the prescriptions to soothe the liver and regulate qi;④ for outpatients who were 30-40 years old and had a disease course of 2-4 years,with liver depression and qi stagnation,the prescriptions used were to soothe the liver and regulate qi combined with a gastrointestinal motility agent;⑤for outpatients who were 40-50 years old and had a disease course of more than 2 years,with liver depression and qi stagnation,these patients were treated with the prescriptions which soothed the liver and regulated qi in combination with a gastrointestinal motility agent or antispasmodic agent.(4)The least effective options for IB S-C were the options involving western medicine alone.It meant that the following treatment options(or the combinations of PIC)might make the outcome of total effective rate less than 75%:① for outpatients who were 30-50 years old with liver depression and qi stagnation and treated with gastrointestinal motility agent alone,of which were 40-50 years old and had a disease course of 4-6 years and treated with laxatives alone;②for outpatients who were 30-50 years old and had a disease course of 2-4 years,with liver depression and qi stagnation,and treated with the laxatives or antispasmodic agent alone;③ for outpatients who were 30-50 years old and had a disease course of 2-4 years,with weakness of spleen and stomach,and treated with the laxatives or antispasmodic agent or gastrointestinal motility agent alone;④ for outpatients who were 40-50 years old and had a disease course of 2-6 years,with liver stagnation and spleen deficiency,and treated with the laxatives or gastrointestinal motility agent alone;⑤ for outpatients who were 40-50 years old and had a disease course of 4-6 years,with fluid insufficiency and heat binding,and treated with the laxatives or gastrointestinal motility agent alone.(Study 2)Similarly as with Study 1,only the total effective rate was included in IBS-PICO-BN models constructed in this study.In this study,the IBS-PICO-BN basic model was constructed consisting of 6 nodes(setting,age,course of disease,TCM syndrome differentiation pattern,intervention,total effective rate)and 5 directed edges(setting→intervention,age→intervention,course of disease→intervention,TCM syndrome differentiation pattern→intervention,intervention→total effective rate).The IBS-PICO-BN complex model was constructed consisting of 6 nodes(same as the basic model)and 8 directed edges(based on the basic model,another 3 directed edges were added:age→course of disease,age→TCM syndrome differentiation pattern,course of disease→TCM syndrome differentiation pattern).Since parameter learning and BNs inference were carried out,110 potential most(total effective rate over 90%)and least effective(total effective rate less than 50%)PICO combination treatment options were found.These options could provide a reference for clinical practice.These options suggested:(1)Among the most effective options for IBS-D,there were 78 options with a total effective rate of 90-95%.No matter which model was used to excavate these options,all the options identified involved either Chinese medicine alone or a combination of Chinese and western medicine.Neither western medicine alone nor placebo involved in the options.Comparing the results determined from the two BN models,68.75%(33/48)of the options were the same.These options were all distributed in the range of conditional probability greater than 0.7,suggesting that the options had better robustness and might have greater assurance of the reliability of efficacy prediction.There were 10 options with a total effective rate over 95%,all of which were options with Chinese combined with western medicines,but no options with only Chinese medicine or western medicine alone.It indicated that the Chinese combined with western medicine,especially combined with antispasmodics,might have a probability of exceeding 0.5 to make the total effective rate greater than 95%.There were some differences between the options from the basic model and the complex model.This finding showed that the relationship between age,course of disease,and TCM syndrome differentiation pattern might have a greater impact on the options involving liver depression and spleen deficiency pattern and weakness of spleen and stomach pattern.(2)The least effective options for IBS-D were the options only involving placebo.Of these,none involved either Chinese medicine alone or western medicine alone or the combination of Chinese and western medicine.(3)Among the most effective options for IBS-C,there were 15 options with a total effective rate of 90-95%.No matter which model was used to excavate these options,all the options identified involved Chinese medicine alone.Comparing results determined from the two BN models,75%(6/8)of the options were the same.These options had only a few different probability distributions in the results of the two models,suggesting these options have better robustness.Due to the limited number of RCTs included in IBS-C,the least effective options for IBS-C have not been found.(Study 3)A total of 63 SRs and 189 RCTs were included in Study 3.In general,the quality of methodology of included studies were very low.This could be explained by one or more of the following reasons:not providing a list of excluded studies,not providing a registered protocol in advance,incomplete searching strategies or strategies that were not comprehensive,no discussion about publication bias,and no analysis of the influence on results from the bias of individual studies.The overall quality of reporting was modest.The major problems were lack of registered protocols and non-standard reporting of the process of inclusion and exclusion.Regarding the evaluation of ROB of included RCTs,the risk of bias was relatively high in the blinding of participants.The risk of bias from the concealment of random sequence,the blinding of outcome assessment,and other biases were unclear.The methodological problems generally existed in the included SRs,for example,quantitatively synthesis was still performed when I2>80%(44/63),the control medicines contained Chinese medicine components(9/63).This situation led to a low reliability of the synthesis results.Therefore,based on overviewing the current evidence from the SRs,the RCTs that met the inclusion criteria embedded in included SRs were further screened.Head-to-head comparisons of meta-analysis and Bayesian NMA were conducted for quantitative synthesis.The NMA results showed that in the comparison of "Chinese medicine vs Western medicine"for IBS-D,the soothing liver,invigorating spleen and eliminating dampness prescriptions(95.60%)might be the best choice,followed by invigorating the spleen and replenishing qi prescriptions(92.33%),antidiarrheal agents combined with probiotics and anti-anxiety and depression agents(90.13%),invigorating spleen and eliminating dampness prescriptions(85.52%),soothing liver and invigorating spleen prescriptions(81.47%),probiotics(74.22%),warming and toxifying the spleen and kidney prescriptions(70.82%),invigorate the spleen,clearing heat and resolving dampness prescriptions(62.68%),soothing liver,invigorating spleen,and reconcile cold and heat prescriptions(62.04%),antispasmodics combined with anti-anxiety and depression agents(49.04%),antispasmodics combined with probiotics(47.47%),antidiarrheal agents combined with probiotics(45.78%),antidiarrheal agents combined with anti-anxiety and depression agents(38.60%),antidiarrheal agents combined with calcium channel blockers(38.48%),symptomatic treatments(36.82%),antidiarrheal agents(34.68%),antispasmodics combined with antidiarrheal agents(33.46%),antispasmodics(30.40%),5HT3 receptor antagonists(29.30%),anti-anxiety and depression agents combined with calcium channel blockers and symptomatic treatments(16.20%),probiotics combined with anti-anxiety and depression agents(14.77%),gastrointestinal motility agents(14.45%),antidiarrheal agents combined with anti-anxiety and depression agents and calcium channel blockers(14.30%),which might be the worst choice.In the comparison of "Chinese medicine vs Western medicine" for IBS-C,those prescriptions that regulated qi and moistened the intestines and those that soothed the liver and regulated qi might be the best choice,followed by nourishing yin and moistening the intestines prescriptions(63.71%),soothing liver and invigorating spleen(48.40%),antispasmodics(43.49%),gastrointestinal motility agents combined with anti-anxiety and depression agents(37.86%),gastrointestinal motility agents(26.05%),laxatives(14.30%),which might be the worst choice.A pairwise meta-analysis of IBS-D showed that(1)Herbal medicine versus Placebo:Sishen Wan improved the total effective rate(RR 2.06,95%CI 1.71-2.47,2 studies,n=322)and reduced the relapse rate(RR 0.27,95%CI 0.17-0.44,2 studies,n=200)were better than placebo;the Tongxiening,a Chinese patent medicine,was better than placebo for improving abdominal pain(RR 1.99,95%CI 1.62-2.44,2 studies,n=499).(2)Herbal medicine versus Western medicine:modified Tongxie Yaofang could significantly reduce the total score of IBS-BSS(MD-43.97,95%CI-56.67 to-31.28,4 studies,n=361),abdominal pain score(MD-11.53,95%CI-16.21 to-6.85,2 studies,n=207),abdominal distension score(MD-5.15,95%CI-8.49 to-1.82,3 studies,n=267)and relapse rate(RR 0.40,95%CI 0.26-0.63,4 studies,n=190).(3)Herbal medicine combined with western medicine versus Western medicine:modified Tongxie Yaofang combined with probiotics were better than probiotics alone in improving the total effective rate(RR 1.29,95%CI 1.12-1.49,3 studies,n=208).Modified Shenling Baizhu San combined with probiotics exhibited no difference compared with probiotics alone in improving the total effective rate(RR 1.31,95%CI 0.99-1.73,3 studies,n=174,I2=61%).Invigorating spleen and replenishing qi prescriptions combined with probiotics might be better than probiotics alone in improving the total effective rate(RR 1.26,95%CI 1.10-1.45,2 studies,n=186).Warming and tonifying the spleen and kidney prescriptions combined with probiotics might be better than probiotics alone in improving the total effective rate(RR 1.32,95%CI 1.10-1.58,2 studies,n=174).Soothing the liver and invigorating the spleen prescriptions combined with antispasmodics might be better than antispasmodics alone in improving the total effective rate(RR 1.18,95%CI 1.09-1.27,6 studies,n=450),of which,modified Shenling Baizhu San combined with trimebutine be better than trimebutine alone in improving the total effective rate(RR 1.19,95%CI 1.09-1.31,2 studies,n=262).Given the limited number of included studies,and the difference in scoring criteria,the data of RCTs for IBS-C and IBS-M were not quantitatively synthesized,but only described qualitatively.CONCLUSIONS:Two theoretical models of efficacy evaluation,IBS-PICO-CN and IBS-PICO-BN,were established through CN method and BN methods in this project.Using the two models with data from 623 RCTs of TCM for IBS,122 potential most effective options and 110 potential least effective options were found.It provided a potential reference for clinical practice.An overview of 63 SRs and 189 RCTs of herbal medicines for IBS of Bayesian NMA suggests that the prescriptions which focused on soothing the liver,strengthening the spleen and eliminating dampness,and the prescriptions to regulate qi,and moisten intestines might be the best choices for patients with IBS-D and IBS-C,respectively.IMPLICATIONS:In this paper,two models were built and exploited to determine a series of treatment options.These could be used as hypotheses for future clinical empirical research.The options excavated from the IBS-PICO-CN model could provide a direction for the retrieval of evidence in the development of evidence-based clinical guidelines and enhance the evidence for subgroup analysis to a certain extent.The options excavated from the IB S-PICO-BN models with their conditional probabilities,provided a reference point for evaluating the quality of evidence used in guideline development.The method of constructing IBS-PICO-BN basic models would be useful for basic model-building of BN efficacy evaluation for other internal diseases.Then according to the specific conditions of different diseases,corresponding complex models of BN efficacy evaluation could be constructed.IBS-PICO-BN models are expected to assist the development of TCM intelligent diagnosis and treatment systems.The models provided a new inference algorithm based on evidence-based efficacy evaluation;also IBS-PICO-BN complex model based on clinical expert experience might assist the construction of a knowledge base for an intelligent diagnosis and treatment system. |
PDF Full Text Request