Cluster analysis for syndromes of real-world coronary heart disease with angina pectoris

Abstract

Syndromes of coronary heart disease with angina pectoris were analyzed to provide guidance for clinical practice and to improve accuracy of traditional Chinese medicine (TCM) diagnoses and efficacy of TCM treatment. A total of 860 cases with coronary heart disease with angina pectoris were selected from TCM Clinical Research Information Sharing System for TCM clinics and research. Syndromes were automatically extracted with the cluster method and were analyzed to provide objective evidence for clinical studies. Final syndrome classifications were recognized and confirmed by clinical experts. Popular syndromes included Qi and blood deficiency, blood stasis and obstruction collaterals, liver depression and spleen deficiency, and Qi stagnation and blood stasis. Syndromes Qi and blood deficiency and blood stasis and obstruction collaterals accounted for 28.61% of total syndromes, whereas liver depression and spleen deficiency and Qi stagnation and blood stasis accounted for 26.44%. The main syndrome elements comprised Qi deficiency, blood deficiency, blood stasis, and Qi stagnation.

Keywords

syndrome differentiation / real-world study / coronary heart disease with angina pectoris / cluster analysis

Introduction

Coronary heart disease with angina pectoris became a significant factor affecting health and quality of life, and its mortality rate increases annually. Thus, a better plan must be urgently developed to improve efficacy of treatment and quality of life. Traditional Chinese medicine (TCM) possesses significant advantages in treatment of coronary heart disease with angina pectoris [¹–¹²]. In TCM clinical practice, syndrome differentiation serves as the core of treatment for coronary heart disease with angina pectoris. Correct syndrome recognition will improve efficacy of treatment for this disease. Therefore, determining syndrome features of coronary heart disease with angina pectoris from effective clinical cases bears significance. Related studies focused on syndrome classification based on data mining methods. Various classification techniques, e.g., naïve Bayes, support vector machine, k nearest neighbor, and latent structure models, were applied in quantitative syndrome analysis [¹³–²⁰]. Better results were obtained for several important diseases, e.g., hepatitis, viral hepatitis, and diabetes. By introducing these data mining models, researchers can mine from some effective cases objective syndrome feature distribution for diseases. Such feature distribution can be used to characterize the underlying relationship between a syndrome and its specific symptoms. However, few studies explored the underlying syndrome features for coronary heart disease with angina pectoris. Li et al. focused on inquiry diagnosis model to determine the effective inquiry style for coronary heart disease [⁴]. Chen et al. focused on phenotypic networks to explore biological basis for coronary heart disease with angina pectoris [⁷]. Zhou et al. analyzed syndrome features of coronary heart disease with angina pectoris based on the cluster method [¹–³]. However, their research data originated from collection of clinical records. That is, pre-designed clinical manifestations were used to collect symptom information. Compared with real-world clinical data for coronary heart disease with angina pectoris, collection of clinical records requires additional cost for researchers and hence further burdens clinical doctors. Therefore, efficiency of research can be improved by finding consistent evidence between real-world and theoretical data for coronary heart disease with angina pectoris. In this study, we selected 860 cases of coronary heart disease with angina pectoris from TCM Clinical Research Information Sharing System (hereinafter “Sharing System”) for TCM clinics and research [²¹]. Sharing System was structured for clinical practice and research and is extensively applied in domestic hospitals because of its high efficiency. In this study, objective syndrome classifications included those first automatically extracted from 860 cases of coronary heart disease with angina pectoris using the cluster method. Cluster algorithm is the most robust method and is available for many data styles in data mining methods. Second, clinical experts on coronary heart disease with angina pectoris discussed and determined the final syndrome classification and optimized syndrome features. Final syndrome results thoroughly reflect syndrome differentiation experience of patients with coronary heart disease with angina pectoris. Simultaneously, better syndrome features provide objective diagnostic evidence for clinical doctors for improvement of accuracy of syndrome diagnosis and efficacy of clinical treatment for coronary heart disease with angina pectoris.

This paper consists of the following sections. Materials and Methods introduce case selection, data pre-processing, and analysis method. Results discuss evaluation of performance of syndrome classification on selected cases. Conclusions and recommendations for future work are also discussed.

Materials and methods

Case selection

A total of 860 selected patients of coronary heart disease with angina pectoris were scheduled to receive treatment in the Guanganmen Hospital of Traditional Chinese Medicine from January 2012 to December 2013. These patients were effectively treated to reduce frequency of chest pain. Clinical information of these patients were collected by the Sharing System for TCM clinics and research. Different from solely observed symptoms and treatment plan in randomized controlled trials (RCT), treatment of these patients is based on syndrome differentiation of clinical doctors. Symptoms were collected and recorded based on clinical manifestation of patients. A total of 860 patients provided their clinical information, e.g., population, symptoms, syndromes, and treatment. Each patient possesses records for at least two treatments.

Data pre-processing

Extraction, transformation, and loading (ETL) software was used to pre-process clinical manifestations, e.g., population, symptom, tongue, pulse and treatment, of 860 patients with coronary heart disease with angina pectoris. Fig. 1 shows the interface of ETL.

ETL software includes some functional interfaces, e.g., database restore, operation database, detailed database, data transformation, data cleaning, data load, and semantic mergence. In this study, data pre-processing steps for coronary heart disease with angina pectoris are as follows: combining similar clinical information; deleting errors or completed clinical information; unifying different types of clinical information; and quantifying clinical information by 0 and 1. Noisy records were removed by filtering.

General materials

Among the 860 selected patients, 389 were males, and 471 were females, with an average age of 60.86±8.16 years; all were diagnosed with coronary heart disease with angina pectoris with an average disease course of 16.44±5.46 months. Finally, 115 clinical manifestations, i.e., 86 symptoms, 18 tongues, and 11 pulses, were extracted and generated to explore underlying syndromes of coronary heart disease with angina pectoris after data pre-processing.

Analysis method

All data were analyzed using SAS (Version 9.3, SAS Institute, Chicago, IL, USA) software. PROC VARCLUS in SAS was used to obtain clusters of symptoms, tongues, and pulses. Cluster results were used to analyze and explain syndrome features of coronary heart disease with angina pectoris. Variables of clinical manifestations were selected and contained in one cluster at P<0.05, which will be considered significant for all analyses.

Results

Symptom indicators

Table 1 presents the top 20 clinical symptoms of coronary heart disease with angina pectoris. “Lack of strength” was the most dominant symptom in 860 included patients.

Clustering results

Symptoms were clustered to six groups to characterize syndrome features. Table 2 shows relevance of six clusters, and Table 3 shows clustering results on 115 symptoms.

Syndrome distribution

Table 4 shows clustering results on symptoms and accuracy of syndrome differentiation.

Discussion

Cluster analysis can be used to determine underlying syndrome differentiation without prior knowledge of syndrome classification. Results objectively present the underlying relationship among TCM clinical symptoms. In this study, clustering results were obtained from 860 patients of coronary heart disease with angina pectoris; these patients were treated effectively by syndrome differentiation.

First, we observed that each cluster features high inter-heterogeneity with each other, as indicated in Table 2. The best heterogeneity was observed in cluster 5 (liver depression and spleen deficiency; and Qi stagnation and blood stasis). Average relevance coefficient of clusters 5 and 2 reached 0.0483 and 0.0944, respectively, which were smaller than those of other clusters. However, the highest relevance coefficient was noted in cluster 6, i.e., Yin deficiency and depleted fluid, blood stasis, and obstruction collaterals. Inter-relevance coefficients of all clusters were small, that is, all clusters featured good inter-heterogeneity.

Second, the best intra-consistency was detected in clusters 4 (Qi and blood deficiency, blood stasis, and obstruction collaterals) and 5 (liver depression and spleen deficiency; and Qi stagnation and blood stasis) (Tables 2 and 3). The numbers of symptoms in 246 and 227 patients of coronary heart disease with angina pectoris reached 17 and 10, respectively. For clusters 4 and 5, patient-to-symptom ratios totaled 14.47% and 22.70%, respectively. All clusters showed high patient-to-symptom ratios, that is, all clusters presented good intra-consistency.

According to Table 4, Qi and blood deficiency, blood stasis, and obstruction collaterals are the main syndrome types for patients of coronary heart disease with angina pectoris. This result is consistent with diagnosis information of “chest pain” or “pectoral pain with stuffiness” category in TCM.

In summary, syndrome features were similar to results in references [¹–³], which were based on collection of clinical records. Among 860 included patients, Qi deficiency, blood deficiency, blood stasis, and Qi stagnation are the main syndrome elements for coronary heart disease with angina pectoris. Calculated results agree with those in actual clinical practice. Angina pectoris belongs to “chest pain” or “pectoral pain with stuffiness” categories in TCM. Syndromes also include stagnation of the heart blood, phlegm, and Yin-cold and deficiencies of the heart-Yin and kidney-Yin, Yang and Yin, and Yang-Qi in pectoral pain with stuffiness. Some of the most important pathogeneses include deficiency and stagnation of Qi and blood. Results presented the main therapy rule of activating blood and promoting Qi. This finding conforms to the rule of “relieving the secondary in an urgent case and removing the primary in a chronic case.”

Conclusions

We conclude that similar syndrome features can be captured from real-world clinical data of coronary heart disease with angina pectoris. Preciseness of results were verified according to TCM theory, and they provide objective evidence for TCM clinical practice and studies.

References

[1] Zhou J, Ji W, Lu H, Tang M, Li J, Nie Y. Study of the pattern characteristics and diversity of stable and unstable angina pectoris based on cluster analysis. World J Inter Tradit West Med (Shi Jie Zhong Xi Yi Jie He Za Zhi) 2012; 7(1): 43–45 (in Chinese)

[2] Zhou JX, Tang M, Li J. Analysis of Chinese syndrome features and combination laws of 2029 patients with coronary heart disease angina. Chin J Integr Tradit West Med (Zhongguo Zhong Xi Yi Jie He Za Zhi) 2011; 31(6): 753–755 (in Chinese)

[3] Zhou JX, Yu GY, Nie YY, Tang M, Jie L. Chinese medicine syndrome characteristic based on clustering analysis of 2029 cases of angina pectoris of coronary heart disease. Chin J Inform Tradit Chin Med (Zhongguo Zhong Yi Yao Xin Xi Za Zhi) 2011; 18(8): 19–21 (in Chinese)

[4] Li DT, Li J, Wang J, Zhang MZ, Wei LB, Jiang RX. Study of TCM syndromes diagnostic criterion of coronary heart disease. J Shandong Univ Tradit Chin Med (Shandong Zhong Yi Yao Da Xue Xue Bao) 2011; 35(2): 117–120 (in Chinese)

[5] Nelson LC, Ryan LJ. The treatment of cardiovascular disease with acupuncture. Med Acup 2014; 26(6): 333–340

[6] Liu W, Xiong X, Yang X, Chu F, Liu H.The effect of Chinese herbal medicine gualouxiebaibanxia decoction for the treatment of angina pectoris: a systematic review. Evid Based Complement Alternat Med 2016; 2016: 8565907

[7] Chen J, Lu P, Zuo X, Shi W, Zhao H, Luo L, Yi J, Zheng C, Yang Y, Wang W. Clinical data mining of phenotypic network in angina pectoris of coronary heart disease. Evid Based Complement Alternat Med 2012; 2012: 546230

[8] Huang J, Tang X, Ye F, He J, Kong X. Clinical therapeutic effects of aspirin in combination with Fufang Danshen Diwan, a traditional Chinese medicine formula on coronary heart disease: a systematic review and meta-analysis. Cell Physiol Biochem 2016; 39(5): 1955–1963

[9] Li P, Ma H. Efficacy of Qingxin Huatan therapy in regulating cardiac autonomic nerve function in patients with chronic stable angina: study protocol for a randomized controlled trial. Clin Trials Degener Dis 2016; 1(2): 51–57

[10] Li S, Xu H. Integrative western and Chinese medicine on coronary heart disease: where is the orientation. Evid Based Complement Alternat Med 2013; 2013: 459264

[11] Yang T, Hu K. Study on distribution rules of TCM four diagnostic information based on complex network in patients with CHD angina pectoris. International Conference on Bioinformatics and Biomedicine, December 15–18, 2016. Shenzhen, China. 1385–1388

[12] Hao PP, Jiang F, Chen YG, Yang J, Zhang K, Zhang MX, Zhang C, Zhao YX, Zhang Y. Traditional Chinese medication for cardiovascular disease. Nat Rev Cardiol 2015; 12(2): 115–122

[13] Zhao Y, He L, Xie Q, Li G, Liu B, Wang J, Zhang X, Zhang X, Luo L, Li K, Jing X. A novel classification method for syndrome differentiation of patients with AIDS. Evid Based Complement Alternat Med 2015; 2015: 936290

[14] Zhao YF, He LY, Liu BY, Li J, Li FY, Huo RL, Jing XH. Syndrome classification based on manifold ranking for viral hepatitis. Chin J Integr Med 2014; 20(5): 394–399

[15] Zhao Y, Xie Q, He L, Liu B, Li K, Zhang X, Bai W. Comparison analysis of data mining models applied to clinical research in traditional Chinese medicine. J Tradit Chin Med 2014; 34(5): 627–634

[16] Gu Z, Qi X, Zhai X, Lang Q, Lu J, Ma C, Liu L, Yue X. Study on TCM syndrome differentiation of primary liver cancer based on the analysis of latent structural model. Evid Based Complement Alternat Med 2014; 2014: 761565

[17] Li G, Yan S, You M, Sun S, Ou A. Intelligent ZHENG classification of hypertension depending on ML-kNN and feature fusion. Evid Based Complement Alternat Med 2012; 2012: 837245

[18] Soni J, Ansari U, Sharma D, Soni S. Predictive data mining for medical diagnosis: an overview of heart disease prediction. Int J Comput Appl 2014; 8(17): 43–48

[19] Wu D, Lin J, Wang P, Wang Y, Lu F. Analysis on composition principles of prescriptions for erectile dysfunction by using traditional Chinese medicine inheritance system. Chin Tradit Patent Med (Zhong Cheng Yao) 2016; 38(4): 755–759 (in Chinese)

[20] Li Y, Zheng G, Liu L. Treatment rules of Sinomenium acutum by text mining. World Chin Med (Shi Jie Zhong Yi Yao) 2015; 10(6): 823–827 (in Chinese)

[21] Liu B, Zhou X, Zhang R, Wang Y, Xie Q, Guo Y, Zhang X, Zhou X, He L, Zhang L, Song G, Zhang Y, Zhang H, Li B, Zhao S. Basic requirements of TCM Electronic Medical Record System in Medical Treatment and Clinical Scientific Research Information Sharing System. Chin Digital Med (Zhongguo Shu Zi Yi Xue) 2012; 7(10): 57–60 (in Chinese)

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag GmbH Germany