Data Mining-Based Analysis of Chinese Medicinal Herb Formulae in Chronic Kidney Disease Treatment

Abstract

Background: Traditional Chinese medicine (TCM) has long been used to treat chronic kidney disease (CKD) in Asia. Its effectiveness and safety for CKD treatment have been confirmed in documented studies. However, the prescription rule of formulae for Chinese medicinal herbs is complicated and remains uncharacterized. Thus, we used data mining technology to evaluate the treatment principle and coprescription pattern of these formulae in CKD TCM treatment.

Methods: Data on patients with CKD were obtained from the outpatient system of a TCM hospital. We established a Chinese herb knowledge base based on the Chinese Pharmacopoeia and the Chinese Materia Medica. Then, following extraction of prescription information, we deweighted and standardized each prescribed herb according to the knowledge base to establish a database of CKD treatment formulae. We analyzed the frequency with which individual herbs were prescribed, as well as their properties, tastes, meridian tropisms, and categories. Then, we evaluated coprescription patterns and assessed medication rules by performing association rule learning, cluster analysis, and complex network analysis.

Results: We retrospectively analyzed 299 prescriptions of 166 patients with CKD receiving TCM treatment. The most frequently prescribed core herbs for CKD treatment were Rhizoma Dioscoreae (Shanyao), Spreading Hedyotis Herb (Baihuasheshecao), Root of Snow of June (Baimagu), Radix Astragali (Huangqi), Poria (Fulin), Rhizoma Atractylodis Macrocephalae (Baizhu), Radix Pseudostellariae (Taizishen), and Fructus Corni (Shanzhuyu). The TCM properties of the herbs were mainly being warm, mild, and cold. The tastes of the herbs were mainly sweet, followed by bitter. The main meridian tropisms were Spleen Meridian of Foot-Taiyin, Liver Meridian of Foot-Jueyi, Lung Meridian of Hand-Taiyin, Stomach Meridian of Foot-Yangming, and Kidney Meridian of Foot-Shaoyin. The top three categories were deficiency-tonifying, heat-clearing, and dampness-draining diuretic.

Conclusion: Using an integrated analysis method, we confirmed that the primary TCM pathogeneses of kidney disease were deficiency and dampness-heat. The primary treatment principles were tonifying deficiency and eliminating dampness-heat.

Figure 1

Data mining flowchart. The integrated data mining method included data processing, frequency statistics, association rules, cluster analysis, and complex network analysis.

Figure 2

Property, taste, and meridian tropism of herbs. (a) Herb properties. All herbs in each prescription were analyzed using a radar chart divided into five categories. (b) Herb taste. Tastes were divided into six categories using a radar chart. (c) Meridian tropism of herbs. We created a tree diagram of the meridian tropism of all herbs. Different meridian tropisms are indicated by different colors as shown at the bottom of the diagram. All images were analyzed using Microsoft Excel 2016.

Figure 3

Association rule combination matrix. The association rule combination matrix was analyzed using R-studio 3.5.3. Size indicatesconfidence (0.855–0.959) and color indicates lift (1.106–1.667).

Figure 4

Association rule diagram. Association rule learning was performed using R-studio 3.5.3. X-axis is the antecedent (or called lefthandside, LHS) and Y-axis is the consequent (or called right-hand side, RHS). Size indicates support and color indicates lift.

Figure 5

Cluster analysis tree diagram. The cluster analysis tree diagram was created using R-studio 3.5.3. The 32 most frequently prescribed herbs were analyzed. Each category is represented by a different color.

Figure 6

Core prescription network. The core prescription network was created using Liquorice. The weight represents the frequency with which two herbs appeared together.