Hierarchical and Complex System Entropy Clustering Analysis Based Validation for Traditional Chinese Medicine Syndrome Patterns of Chronic Atrophic Gastritis

Abstract

Objectives:

Chronic atrophic gastritis (CAG) is the precancerous stage of gastric carcinoma. Traditional Chinese Medicine (TCM) has been widely used in treating CAG. This study aimed to reveal core pathogenesis of CAG by validating the TCM syndrome patterns and provide evidence for optimization of treatment strategies.

Design:

This is a cross-sectional study conducted in 4 hospitals in China. Hierarchical clustering analysis (HCA) and complex system entropy clustering analysis (CSECA) were performed, respectively, to achieve syndrome pattern validation.

Results:

Based on HCA, 15 common factors were assigned to 6 syndrome patterns: liver depression and spleen deficiency and blood stasis in the stomach collateral, internal harassment of phlegm-heat and blood stasis in the stomach collateral, phlegm-turbidity internal obstruction, spleen yang deficiency, internal harassment of phlegm-heat and spleen deficiency, and spleen qi deficiency. By CSECA, 22 common factors were assigned to 7 syndrome patterns: qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, and yin deficiency.

Conclusions:

Combination of qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, and yin deficiency may play a crucial role in CAG pathogenesis. In accord with this, treatment strategies by TCM herbal prescriptions should be targeted to regulating qi, activating blood, resolving turbidity, clearing heat, removing toxin, nourishing yin, and warming yang. Further explorations are needed to verify and expand the current conclusions.

Introduction

Chronic atrophic gastritis (CAG) is an inflammatory disease of gastric mucosa from various etiologies, which is defined as the precancerous stage of gastric carcinoma (GC).^1
–3 Global cancer statistics for 2012 estimated that there were nearly 1000,000 new GC cases reported worldwide annually, the incidence and mortality of which also have been increasing in China.⁴ The transition from CAG to GC is acknowledged as a typical disease model of uncontrolled inflammation leading to malignant transformation, the treatment of which is costly to individuals.^5,6

Active treatment of CAG arrests the transition from gastric mucosal atrophy to GC. Based on the theory of holism and time-honored principle of syndrome pattern differentiation, Traditional Chinese Medicine (TCM) has been widely applied in treating digestive disorders among Chinese people and has distinguished itself with satisfying cost and efficacy since ancient times. Modern-day TCM practitioners remain performing syndrome pattern differentiation to identify the pathogenesis in relation to CAG and thus treat the root.

The TCM concept of four diagnostic information subsumes manifold signs and symptoms. It is collected systematically in syndrome pattern differentiation process from the four diagnoses (symptoms, physical signs, tongue appearance, and pulse reading) and is regarded to have implications for disease pathogenesis. After assessing all the information collocated, the potential link between pathologic nature and disease in the current stage can be determined by TCM practitioners. Herbal prescriptions can then be formulated in accordance with conclusions drawn from the syndrome pattern differentiation process.

There has been growing literature demonstrating the notion that TCM herbal prescriptions can treat CAG effectively, including relieving symptoms, inhibiting mucosal inflammation, and reversing atrophy.^7

–13 However, no consensus exists across these studies on syndrome patterns and treatment strategies.^14
–16 The authors conducted the present study to ascertain core pathogenesis of CAG by validating the TCM syndrome patterns. It will provide evidence and potential translation for optimization of treatment strategies from the perspective of TCM.

Clustering analysis is one of the most important research fields of exploratory data mining. It has been used in many fields as a kind of unsupervised learning methods. In this study, syndrome pattern validation was achieved using exploratory factor analysis (EFA) based hierarchical clustering analysis (HCA) and association rule analysis (ARA) based complex system entropy clustering analysis (CSECA). Original TCM four diagnostic variables contain specific combinations and complex transactions of disease features, which have significantly higher dimensional characteristics.^17,18 The abovementioned unsupervised analytic methods can detect syndrome features by achieving multicollinearity elimination and dimensionality reduction.^19,20 Using EFA and ARA, large numbers of variables can be remodeled as combinations of linear and simpler common factors (variables with specific combination of features), which lay the foundation for further clustering analysis. HCA and CSECA can then be applied, respectively, based on the common factors revealed in previous steps to develop dimensional taxonomies and detect the relationships among variables.

Materials and Methods

Study population

This cross-sectional study was conducted in 4 medical centers in China, including Dongzhimen Hospital Affiliated to Beijing University of Chinese Medicine; Peking Union Medical College Hospital Affiliated to the Chinese Academy of Medical Sciences; Beijing Hospital Affiliated to the National Health and Family Planning Commission of China; and Xiamen Traditional Chinese Medicine Hospital. The study began in September 2010 based on participant survey and ended in October 2012. Patients 20–75 years old, who visited the aforementioned hospital for upper gastrointestinal endoscopy, met diagnostic criteria of CAG, and signed informed consent, were recruited.

Ascertainment of CAG

Diagnostic criteria of CAG were based on guidelines issued by Chinese Medical Association in 2012.²

Case report form

Case report form (CRF) was designed based on literature research, expert advice, and clinical guidelines.^21
–23 CRF content included general information (name, age, sex), disease feature and history (chief complaint, history of present illness, past medical history, family medical history), modern medicine diagnosis, TCM diagnosis, the TCM four diagnostic information, and copy of upper gastrointestinal endoscopy report. To reduce measurement bias, all the investigators who filled in the CRFs were trained in standard operating procedures (SOP). Each study participant was examined and followed up by at least two resident physicians or graduate students. At least two senior staff physicians supervised interview sessions to ensure consistency and authenticity of data collection.

Data analysis

HCA and CSECA were performed on data collected from the four diagnostic methods, based on EFA and ARA, respectively, to achieve syndrome pattern extraction (Fig. 1). All analyses in this study were processed by SPSS Statistics software (version 17.0; SPSS, Inc., Chicago, IL) and SPSS Clementine software (version 12.0; SPSS, Inc.).

FIG. 1.

Flow chart of clustering analysis based CAG syndrome pattern extraction. CAG, chronic atrophic gastritis.

Results

Characteristics of participants

One hundred thirty-five CRFs were distributed, and a total of 131 were gathered. One hundred twenty forms deemed eligible after eliminating 15 CRFs with incomplete information were answered. Of the 120 eligible participants, 58 were female and 62 were male, with an average age of 52.56 years. All the four diagnostic variables detected were preanalyzed, and 40 variables with more than 21% frequency of occurrence were chosen for subsequent analysis. The abovementioned variables were tabulated based on the distribution frequency (Table 1).

Table 1.

Four Diagnostic Frequency for Chronic Atrophic Gastritis (N = 120)

Sequence	Variable	Frequency (%)	Sequence	Variable	Frequency (%)
1	Wiry pulse	75.8	21	Constipation	32.5
2	Dark tongue	65	22	Enlarged tongue	32.5
3	Distention and fullness of the stomach	64.2	23	Sensation of coldness of the body and limbs	31.7
4	Thin tongue coating	61.7	24	Fetid mouth odor	30
5	White tongue coating	59.2	25	Choppy pulse	30
6	Greasy tongue coating	56.7	26	Stifling sensation in the chest	28.3
7	Belching	54.2	27	Poor appetite	28.3
8	Stomach aversion to cold	54.2	28	Night sweats	28.3
9	Acid regurgitation	50	29	Thick tongue coating	28.3
10	Lack of strength	50	30	Frequent sighing	25.8
11	Dry mouth	49.2	31	Abdominal distention	24.2
12	Heartburn	46.7	32	Palpitations	23.3
13	Irritability	45	33	Listlessness	23.3
14	Bitter taste in the mouth	45	34	Teeth-marked tongue	23.3
15	Yellow tongue coating	44.2	35	Fine pulse	23.3
16	Red tongue	39.2	36	Tinnitus	22.5
17	Insomnia	35.8	37	Sharp pain in stomach	21.8
18	Pale red tongue	34.2	38	Nausea	21.8
19	Loose stool	33.3	39	Tongue with petechiae	21.8
20	Profuse dreaming at night	33.3	40	Slippery pulse	21.8

EFA based HCA for syndrome pattern extraction

EFA based assessment of common factors

Before analysis, the Kaiser-Meyer-Olkin test and Bartlett's test of sphericity were used, respectively, to assess suitability of collocated diagnostic variables for EFA. Principal component analysis was then performed to extract common factors from the original variables. Through consultation with TCM experts, the diagnostic variables, nature of disease, and disease location obtained from the four diagnostic information were ultimately assigned to 15 common factors (Table 2).

Table 2.

Exploratory Factor Analysis Based Common Factor Extraction and Their Corresponding Four Diagnostic Variables, Disease Nature, and Disease Locations

Common factor	Four diagnostic variables	Syndrome pattern	Disease location
F1	Palpitations; frequent sighing; stomach aversion to cold; listlessness; sensation of coldness of the body and limbs; tinnitus	yang deficiency	Spleen
F1		yang deficiency	Stomach
F2	Teeth-marked tongue; thick tongue coating; greasy tongue coating	Phlegm turbidity	Spleen
F2		Phlegm turbidity	Stomach
F3	Severe stomach pain; red tongue; yellow tongue coating	Heat	Stomach
F3	Severe stomach pain; red tongue; yellow tongue coating	Blood stasis	Stomach
F4	Loose stool; constipation; listlessness; lack of strength; tinnitus; enlarged tongue	qi deficiency	Spleen
F5	Distention and fullness of the stomach; bitter taste in the mouth; belching; acid regurgitation; heartburn	qi stagnation	Liver
F5		qi stagnation	Stomach
F6	Irritability; bitter taste in the mouth; insomnia; profuse dreaming at night; dry mouth	Heat	Liver
F7	Sharp pain in stomach; tinnitus; dark tongue; tongue with petechiae; choppy pulse;	Blood stasis	Stomach
F8	Pale red tongue; enlarged tongue; teeth-marked tongue	qi deficiency	Spleen
F9	Loose stool; sallow complexion; insomnia; enlarged tongue; tongue with petechiae; teeth-marked tongue	qi deficiency	Spleen
F10	Distention and fullness in the stomach; constipation; poor appetite; dry mouth	qi stagnation	Spleen
F10		qi stagnation	Stomach
F11	Insomnia; slippery pulse	Phlegm turbidity	Spleen
F11	Insomnia; slippery pulse	Phlegm turbidity	Stomach
F12	Frequent sighing; abdominal distention; belching; nausea; fine pulse	qi stagnation	Liver
F12		qi stagnation	Stomach
F13	Constipation; abdominal distention; nausea; profuse dreaming at night	Phlegm turbidity	Spleen
F14	Palpitation; stifling sensation in the chest; nausea	Phlegm turbidity	Spleen
F14	Palpitation; stifling sensation in the chest; nausea	Phlegm turbidity	Stomach
F15	Abdominal distention; fetid mouth odor; enlarged tongue	Heat	Spleen
F15	Abdominal distention; fetid mouth odor; enlarged tongue	qi stagnation	Stomach

HCA based extraction of syndrome patterns

Based on HCA, 15 common factors were combined and 6 syndrome patterns were extracted. After consultation with experts, name of syndrome patterns, nature of disease, and disease locations were assigned (Fig. 2 and Table 3). The six syndrome patterns extracted by HCA were: (1) Liver depression and spleen deficiency and blood stasis in the stomach collateral; (2) Internal harassment of phlegm-heat and (blood) stasis in the stomach collateral; (3) Phlegm-turbidity internal obstruction; (4) Spleen yang deficiency; (5) Internal harassment of phlegm-heat and spleen deficiency; and (6) Spleen qi deficiency.

FIG. 2.

Tree diagram of HCA based syndrome pattern extraction. HCA, hierarchical clustering analysis.

Table 3.

Hierarchical Clustering Analysis Based Syndrome Pattern Extraction and Their Corresponding Four Diagnostic Variables, Disease Nature, and Disease Locations

Syndrome pattern
Disease nature	Disease location	Common factors	Four diagnostic variables
qi deficiency	Liver	F8, F12, F7	Pale red tongue, enlarged tongue, teeth-marked tongue, frequent sighing, abdominal distention, belching, nausea, fine pulse; stomach tingling, tinnitus, dark tongue, tongue with petechiae, choppy pulse
qi stagnation Blood stasis	Spleen
qi stagnation Blood stasis	Stomach
Phlegm	Spleen	F3, F11	Stomach tingling, red tongue, yellow tongue coating, insomnia, slippery pulse
Heat	Stomach
Blood stasis	Stomach
Phlegm turbidity	Spleen	F13	Constipation, abdominal distention, nausea, profuse dreaming at night
Phlegm turbidity	Stomach	F13
yang deficiency	Spleen	F1	Palpitations, frequent sighing, stomach aversion to cold, listlessness, sensation of coldness of the body and limbs, tinnitus, teeth-marked tongue
	Stomach
	Heart
Phlegm turbidity	Spleen	F2, F15, F5	Teeth-marked tongue, thick tongue coating, greasy tongue coating, distention and fullness of the stomach, bitter taste in the mouth, belching, acid regurgitation, heartburn, abdominal distention, fetid mouth odor, enlarged tongue
qi stagnation	Stomach
Heat	Liver
Phlegm turbidity	Spleen	F14, F6, F9	Irritability, bitter taste in the mouth, insomnia, profuse dreaming at night, dry mouth, loose stool, sallow complexion, insomnia, enlarged tongue, teeth-marked tongue, tongue with bruise, nausea, stifling sensation in the chest, palpitations
qi deficiency	Stomach
Heat	Liver
qi deficiency	Spleen	F4, F10	Loose stool, constipation, listlessness, lack of strength, tinnitus, enlarged tongue, Distention and fullness of the stomach, constipation, poor appetite, dry mouth
qi deficiency	Stomach	F4, F10

ARA based CSECA for syndrome pattern extraction

Based on ARA and CSECA, 22 common factors were combined, and 7 syndrome patterns were extracted. After consultation with experts, name of syndrome patterns, nature of disease, and disease locations were assigned (Tables 4, 5). Core combinations of four diagnostic variables were also revealed (Fig. 3). The 7 syndrome patterns extracted by ARA and CSECA were: (1) qi deficiency; (2) qi stagnation; (3) Blood stasis; (4) Phlegm turbidity; (5) Heat; (6) yang deficiency; and (7) yin deficiency.

FIG. 3.

Network diagram of core combinations of four diagnostic variables.

Table 4.

Association Rule Analysis and Complex System Entropy Clustering Analysis Based Common Factor Extraction and Their Corresponding Four Diagnostic Variables, Disease Nature, and Disease Locations

Common factor	Four diagnostic variables	Disease nature	Disease location(s)
F1	Insomnia, yellow urine, bland taste in the mouth, pale tongue	Heat	Stomach
F1		qi deficiency	Heart
F2	Wiry pulse, frequent sighing, night sweats, puffy swelling of the limbs	qi stagnation	Liver
		qi deficiency	Stomach
		yin deficiency	Spleen
F3	Wiry pulse, frequent sighing, bound pulse, thirst without desire to drink	qi stagnation	Liver
F3		qi stagnation	Stomach
F4	Bitter taste in the mouth, feeling of heaviness of the body, red tongue, enlarged tongue	Heat	Stomach
		Phlegm turbidity	Spleen
		Phlegm turbidity	Liver
F5	Greasy tongue coating, acid regurgitation, tidal fever, vomiting of phlegm-drool	Phlegm turbidity	Stomach
F5		Phlegm turbidity	Spleen
F6	Greasy tongue coating, thin tongue coating, yellow tongue fur, white tongue coating	Unable to judge	Unable to judge
F7	Acid regurgitation, feeling of heaviness of the body, tidal fever, sweet taste in the mouth	Phlegm turbidity	Stomach
F7		Phlegm turbidity	Spleen
F8	Palpitations, pale red tongue, gastric colic, vomiting of phlegm-drool	Phlegm turbidity	Stomach
F8		Phlegm turbidity	Heart
F9	Abdominal distention, frequent sighing, abdominal pain with desire of diarrhea, fine pulse	qi stagnation	Liver
F9		qi stagnation	Stomach
F10	Belching, feeling of heaviness of the body, stomach aversion to cold, shortage of qi, and laziness to speak	yang deficiency	Stomach
F10		yang deficiency	Spleen
F11	White tongue coating, dry mouth, stomach tingling, yellow tongue fur, thirst and desire to drink	Heat	Stomach
F11		Blood stasis	Stomach
F12	Stomach aversion to cold, pale red tongue, pale tongue, sensation of coldness of the body and limbs	yang deficiency	Stomach
F12		yang deficiency	Spleen
F13	Constipation, profuse dreaming at night, spontaneous sweating, borborygmus	Phlegm turbidity	Stomach
F13		Phlegm turbidity	Spleen
F14	Constipation, sticky mouth, spontaneous sweating, vomiting of phlegm-drool	Phlegm turbidity	Stomach
F14		Phlegm turbidity	Spleen
F15	Dark complexion, tinnitus, vomiting food, early satiety	Blood stasis	Stomach
F15	Dark complexion, tinnitus, vomiting food, early satiety	yin deficiency	Spleen
F16	Dry mouth, yellow tongue fur, red tongue, listlessness	Heat	Stomach
F16	Dry mouth, yellow tongue fur, red tongue, listlessness	Heat	Spleen
F17	Dry mouth, profuse dreaming at night, stomach tingling, thirst without desire to drink	Blood stasis	Stomach
F18	Dark tongue, sensation of coldness of the body and limbs, abdominal pain with desire of diarrhea, spontaneous sweating	yang deficiency	Stomach
F18		yang deficiency	Spleen
F19	Heartburn, profuse dreaming at night, stomach tingling, hiccups	Blood stasis	Stomach
F19		qi stagnation	Liver
F20	Choppy pulse, pale red tongue, pale tongue, dull ache in the stomach	Blood stasis	Stomach
F20		yin deficiency	Stomach
F21	Distending pain in the stomach, pale red tongue, pale tongue, dull ache in the stomach	qi stagnation	Stomach
F21		yin deficiency	Stomach
F22	Irritability, dry stool, loose stool, lack of strength	qi stagnation	Liver
F22	Irritability, dry stool, loose stool, lack of strength	qi deficiency	Spleen

Table 5.

Association Rule Analysis and Complex System Entropy Clustering Analysis Based Syndrome Pattern Extraction and Their Corresponding Disease Nature, Disease Locations, Common Factors, and Four Diagnostic Variables

Syndrome pattern
Disease nature	Disease locations	Common factors	Four diagnostic variables
qi deficiency	Stomach	F1, F2, F22	Lack of strength, bland taste in the mouth, frequent sighing, night sweats, puffy swelling of the limbs, insomnia, loose stool, dry stool, irritability, yellow urine, pale tongue, wiry pulse
	Liver
	Spleen
	Heart
qi stagnation	Stomach	F2, F3, F9, F19, F21, F22	Abdominal distention, hiccups, distending pain in stomach, dull ache in the stomach, stomach tingling, irritability, frequent sighing, abdominal pain with desire of diarrhea, heartburn, dry stool, loose stool, lack of strength, profuse dreaming at night, thirst without desire to drink, night sweats, puffy swelling of the limbs, wiry pulse, bound pulse, fine pulse, pale red tongue, pale tongue
	Liver
	Spleen
Blood stasis	Stomach	F11, F15, F17, F19, F20	Stomach tingling, dull ache in the stomach, dark complexion, thirst without desire to drink, dry mouth, thirst and desire to drink, hiccups, heartburn, tinnitus, vomiting food, early satiety, profuse dreaming at night, white tongue coating, yellow tongue fur, pale red tongue, pale tongue, choppy pulse
	Spleen
	Liver
Phlegm turbidity	Stomach	F4, F5, F7, F8, F13, F14	Vomiting of phlegm-drool, bitter taste in the mouth, feeling of heaviness of the body, sticky mouth, sweet taste in the mouth, palpitations, spontaneous sweating, constipation, borborygmus, acid regurgitation, gastric colic, profuse dreaming at night, tidal fever, red tongue, pale red tongue, greasy tongue coating, enlarged tongue
	Liver
	Spleen
	Heart
Heat	Stomach	F1, F4, F11, F16	Thirst and desire to drink, dry mouth, bitter taste in the mouth, stomach tingling, insomnia, yellow urine, bland taste in the mouth, feeling of heaviness of the body, listlessness, red tongue, pale tongue, enlarged tongue, yellow tongue fur, white tongue coating
	Liver
	Spleen
	Heart
yang deficiency	Spleen	F10, F12, F18	Stomach aversion to cold, belching, sensation of coldness of the body and limbs, shortage of qi and laziness to speak, feeling of heaviness of the body, abdominal pain with desire of diarrhea, spontaneous sweating, pale tongue, pale red tongue, dark tongue
	Stomach

yin deficiency	Stomach	F2, F15, F20, F21	Dull ache in the stomach, distending pain in the stomach, vomiting food, early satiety, frequent sighing, night sweats, dark complexion, puffy swelling of the limbs, tinnitus, pale red tongue, pale tongue, choppy pulse, wiry pulse
	Liver
	Spleen

Discussion

The holistic theory that TCM is based on determines the complex multivariate nonlinear relationship of the variables of the four diagnostic information. TCM syndrome can be regarded as a giant complex system with temporal dynamics and nonlinear high-dimensional interfaces. The fundamental concept of EFA based HCA is projecting high-dimensional TCM clinical information onto a lower plain. In the present study, HCA was first carried out based on EFA to measure the intimacy of variables (common factors) under a certain definition of “distance.”^24

–27 In this classical process, the number of clusters should be set artificially by data analyst and TCM practitioners. Ultimately, common factors with sufficient similarity were assigned in the same clusters through dimensionality reduction, by which syndrome patterns of CAG can be ascertained.

Complex systems represent new perspectives and solutions for investigating how relationships between variables contribute to the collective behaviors of a system and how the system interacts with each variable.^19,20 As one of the typical complex systems, the TCM concept syndrome is composed of various clinical information sets. Entropy can be applied in measuring the uncertainty and variety of information of a self-adjusted complex system in different dimensions and clusters precisely.^19,20 Because of the better adaptability for measuring the correlation between multidimensional and nonlinear variables compared with EFA based HCA, ARA based CSECA was also operated on data mining of TCM syndrome related variables in this study.^28,29

According to TCM theory, syndrome pattern is considered the most important unit for evaluating pathogenesis.^17,18 Results of emerging researches have not been consistent on CAG syndrome features, especially in disease nature, disease location, and differentiation of primary syndrome patterns, which indicate the complexities of pathogenesis.^19

–23 However, it may also be due to a lack of established reliability of observations and judgments in previous studies. It is critical to provide a much more detailed assessment by their present study to explored core syndrome patterns of CAG by applying HCA and CSECA. All information the common factors, syndrome patterns, disease location, and disease nature belonged to were appraised and determined based on professional knowledge.

HCA showed a total of 15 common factors composed of 6 syndrome patterns (liver depression and spleen deficiency and blood stasis in the stomach collateral, internal harassment of phlegm-heat and blood stasis in the stomach collateral, phlegm-turbidity internal obstruction, spleen yang deficiency, internal harassment of phlegm-heat and spleen deficiency, and spleen qi deficiency). CSECA showed a total of 22 common factors assigned to 7 syndrome patterns (qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, yin deficiency). Both HCA and CSECA results indicated that the TCM concept disease nature of CAG was a combination of qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, and yin deficiency, as well as disease location in the stomach, spleen, and liver.

Compared with HCA, CSECA results were more concordant with previous findings.^19

–23 CSECA suggests that combination of qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, and yin deficiency may form the core pathogenesis of CAG. Therefore, they should be taken into account during clinical treatment as the crucial diagnostic elements. The target organs of treatment should be the stomach, spleen, and liver. Treatment strategy by herbal prescriptions should hence be targeted to regulating qi, activating blood, resolving turbidity, clearing heat, removing toxin, nourishing yin, and warming yang. However, in terms of HCA results, yin deficiency was not included in the disease nature of CAG and, thus, may not be the most vital syndrome pattern in CAG patients. According to the seemingly contradictory results observed, their findings add support to the notion that herbs that nourish yin may receive less prominence or be prescribed in lesser amounts than herbs that focus on other abovementioned aspects when formulating a prescription, but it should not be totally underestimated when treating CAG.

The present study design has several strengths and limitations of note. Multidimensional and full-scale information was gathered in this research with strict quality control, for which the authors may adapt in multivariate data mining models. In addition, the whole study process is closely monitored and controlled with all the investigators trained in SOP and, therefore, has relatively high data quality. Several limitations are also noteworthy. Selection bias may exist, so that conclusions may be generalized only to CAG patients in TCM hospitals in Beijing and, therefore, are likely not representative of the disease features in patients in modern medicine hospitals and the rest of China. Moreover, their study is designed with a relatively small sample size. Given the limitations of their study, the current findings need to be repeated with a larger population and with guaranteed inter-rater reliability of all observations and judgments before the results can be accepted in clinical practice. Further explorations through multicentric and full-scale studies are needed to verify and expand the conclusions.

The study has demonstrated the suitability of using the new analytic methods as tools for development of TCM syndrome patterns and associated treatments of CAG. In terms of data mining methodology, there are also some potential advantages and limitations to be addressed. The results from both HCA and CSECA need to be validated and expanded with other supervised data mining methods. In HCA, each TCM variable has merely one chance to be assigned in a single common factor, thus resulting in an inability to describe multiple correlations between variables. Compared with conventional models, HCA can achieve rules more objectively and avoid the interference of noise variables to some extent. However, aforementioned flaws may still result in failure to describe the internal property and external relevance efficiently in this dimension-reducing process. In addition, a series of parameters must be set artificially in HCA, the analyst blinding can hardly be achieved.²⁶ In terms of CSECA, strengths include good algorithmic adaptation for describing relationship among multidimensional and nonlinear TCM variables. Flexible data presplitting in CSECA allows each variable a chance to be assigned in different common factors repeatedly, which is more in line with TCM theory.^30
–32 Application of additional methods is needed to address the broader problem of generalizability in future studies.

Conclusions

HCA and CSECA are important methodologies for clustering multidimensional and multilinear TCM syndrome-related variables, by which evidence for both pathogenesis and core treatment principles of CAG can be revealed. Their findings support that the TCM concept pathogenesis of CAG may be a combination of qi deficiency, qi stagnation, blood stasis, phlegm turbidity, heat, yang deficiency, and yin deficiency. Treatment of CAG by herbal prescriptions should therefore focus on regulating qi, activating blood, resolving turbidity, clearing heat, removing toxin, nourishing yin, and warming yang. Further explorations are needed to verify and expand the current conclusions.

Footnotes

Acknowledgments

Generous support for this study was provided by the National Natural Science Foundation of China (No. 81630080, 91129714, 81270466, 81173424, 81673793, and 81373796) and Specialized Research Fund for the Doctoral Program of Higher Education of China (No. 20120013110014). Yin Zhang was additionally supported by National Undergraduates Innovating Experimentation Project of the China Ministry of Education (No. 081002609) and Science Research Foundation of Beijing University of Chinese Medicine (No. 2014-JYBZZ-XS-134).

Xia Ding and Yin Zhang conceived the study conception and design. Yin Zhang, Yue Liu, Li Zhang, Shiyu Du, and Daming Liu trained investigators on SOP of this study. Yannan Li, Zeqi Su, and Cen Chen performed methodology development and data analysis. Yin Zhang, Yue Liu, Yannan Li, Xia Zhao, and Lin Zhuo performed writing and revision of the article with contributions from all other authors. All authors carried out the clinical investigation and approved the final article. Yin Zhang, Yue Liu, and Yannan Li contributed equally to this work. The authors thank all the participants for their contribution to the research.

Author Disclosure Statement

No competing financial interests exist.

References

Weck

, Brenner

. Prevalence of chronic atrophic gastritis in different parts of the world. Cancer Epidem Biomar, 2006; 15:1083–1094.

Fang

, Liu

, Li

, et al. Consensus on chronic gastritis in China. Chin J Gastroentero, 2013; 18:24–36. [Chinese]

Misdraji

, Lauwers

. Gastric epithelial dysplasia. Semin Diagno Pathol, 2002; 19:20–30.

Torre

, Bray

, Siegel

, et al. Global cancer statistics. CA-Cancer J Clin, 2015; 65:87–108.

Correa

, Piazuelo

. The gastric precancerous cascade. J Digest Dis, 2012; 13:2–9.

Chao

, Hellmich

. Gastrin, inflammation and carcinogenesis. Curr Opin Endocrinol, 2010; 17:33–39.

Zhao

, Meng

, Yang

. Effect of Xiaopi Granules on proliferation and apoptosis in the gastric mucosa of chronic atrophic gastritis rats with atypical hyperplasia. J Tradit Chin Med, 2012; 53:2118–2121. (Chinese)

, Li

, Liu

, et al. Improvement in gastric mucosal lesions in the rat model of chronic atrophic gastritis following administration of Shenqi Xiaopi Granules. Jiangsu J Tradit Chin Med, 2013; 45:69–71. (Chinese)

Gao

, Li

, Cui

. Efficacy of Huazhuo Jiedu Hewei Decoction in the treatment of chronic atrophic gastritis precancerous lesions. Chin J Gerontol, 2010; 30:460–462. (Chinese)

10.

Teng

. Yiqi Huoxue Decoction in the treatment of 40 cases of chronic atrophic gastritis. Yunnan J Tradit Chin Med, 2011; 32:56–57. (Chinese)

11.

Wei

, Yang

, et al. Effect of benefiting qi, transforming stasis, relieving toxicity strategy in the rat model of chronic atrophic gastritis with dysplasia. Chin J Gastroen Hepatol, 2011; 21:916–919. (Chinese)

12.

Liu

, Li

, Wang

, et al. Huazhuo Jiedu Formula for treatment of chronic atrophic gastritis with intestinal metaplasia. Hebei J Tradit Chin Med, 2011; 33:1139–1141. (Chinese)

13.

, Liu

, et al. Efficacy of Yiqi Huoxue Formula for treatment of chronic atrophic gastritis. Hebei J Tradit Chin Med, 2012; 34:14–16. (Chinese)

14.

, Ding

, Guo

, et al. Literature review of Chinese medicine syndromes and syndrome element distribution features of chronic gastritis. Liaoning J Tradit Chin Med, 2014; 41:2532–2534. (Chinese)

15.

Han

, Lin

, Gan

, et al. Literature review of Chinese medicine syndrome distribution of chronic gastritis. Chin J Tradit Chin Med Pharm, 2014; 29:900–903. (Chinese)

16.

Tang

, Li

, et al. Guideline of treatment and diagnosis of chronic gastritis. Chin J Mod Educ, 2011; 9:123–125. (Chinese)

17.

. Syndrome factors and formula factors. Chin J Tradit Chin Med Pharm, 2009; 24:117–121. (Chinese)

18.

Guo

, Wang

, Zhang

. Annotations on the concept of syndrome. J Beijing Univ Chin Med, 2003; 26, 2:5–8. (Chinese)

19.

Chen

. Complex system modeling of traditional Chinese medicine syndrome and its correlation between different diseases. Beijing: The Graduate School of the Chinese Academy of Sciences, 2008. (Chinese)

20.

Tang

, Chen

, Yang

, et al. The relationship between syndrome, entropy and formula - a new frontier of traditional Chinese medical studies. J Basic Tradit Chin Med, 2010; 16:267–270. (Chinese)

21.

State Bureau of Technical

Supervision

, State Administration of Traditional Chinese

Medicine

. Consensus statement on chronic gastritis. Beijing, China: China Medical Science Press, 1997. (Chinese)

22.

Zheng

. Clinical guideline of new drugs for traditional Chinese medicine (Trial). Beijing, China: China Medical Science Press, 1997. (Chinese)

23.

Spleen and Stomach Disease Association of China Association of Chinese Medicine, Consensus statement on traditional Chinese treatment of chronic atrophic gastritis. J Tradit Chin Med, 2010; 51:749–753. (Chinese)

24.

Reise

, Waller

, Comrey

. Factor analysis and scale revision. Psychol Asses, 2000; 12:287–297.

25.

Wirth

, Edwards

. Item factor analysis: Current approaches and future directions. Psychol Methods, 2007; 12:58–79.

26.

Kelton

, Leard Mann

, Smith

, et al. Exploratory factor analysis of self-reported symptoms in a large, population-based military cohort. BMC Med Res Methodol, 2010; 15:94.

27.

Eijk

, Rose

. Risky business: Factor analysis of survey data—assessing the probability of incorrect dimensionalisation. PLoS One, 2015; 10:e0118900.

28.

Pavón

, Viana

, Gómez

. Apriori: Speeding up the search for frequent patterns, Proceedings of 24th IASTED International Multi-Conference 2006; Innsbruck, Austria: 13–15.

29.

. Mining association rules on a huge number of items. In: Proceedings of the 8th Joint International Computer Conference, Oxford, England, 2002; 11:587–593.

30.

Yang

, Zhao

, Liu

. Application of entropy in traditional Chinese medicine syndrome studies. J Basic Tradit Chin Med, 2004; 10:16–19. (Chinese)

31.

Gao

, Wu

, Chen

, et al. Application of entropy based clustering in traditional Chinese medicine syndrome-related studies. J Syst Simulat, 2009; 21:1760–1768. (Chinese)

32.

Wang

, Wu

, Chen

, et al. Application of information entropy based association rules in extraction of traditional Chinese medicine syndrome patterns or syndrome factors. Tianjin J Tradit Chin Med, 2013; 26:274–276. (Chinese)