Correlations Between Traditional Chinese Medicine Syndromes and Clinical Indexes in Health Check-Up Population

Routine health check-up is associated with improved lifespan and reduced medical cost. Traditional Chinese medicine (TCM) serves as a cost-effective modality in healthcare system. We examined the correlations of TCM syndromes with modern medical indicators in health check-up population. Methods We studied 5231 subjects undergoing health check-up between January 1st 2008 and December 31, 2016. Physical indexes such as body weight and blood pressure and biomedical indicators like live function and tumour markers were measured. All subjects underwent colonoscopy. All subjects were classied and differentiated into ve different TCM syndromes. An articial neural network (ANN) was employed to evaluate the predictive value of TCM syndrome differentiation. good predictive of A recent study has reported that the concentration of adiponectin was correlated with TCM deciency factors in dyslipidaemia patients. 13 The correlations of TCM syndrome pattern with clinical indexes including renal function, immunologic function, and blood coagulation in patients with Henoch-Schönlein purpura nephritis have been reported. 8 Blood stasis syndrome has positive correlation with triglyceride and cholesterol in patients with primary glomerular disease. 14 In our study, we found that syndrome of accumulated dampness-heat (SADH) is correlated with level of cholesterol in the subjects undergoing health check-up. The nding is consistent with the result of Chen’s study showing accumulation of dampness in dyslipidaemia patients. 13 Our data highlight that yin deciency of liver and kidney (YiDLK) is associated with abnormality of liver function indexes. This is in an agreement with recent study in which common TCM syndromes in primary liver cancer are dampness, blood-stasis, heat, Yang deciency of spleen and kidney, and Yin deciency of liver and kidney. 15 Our ndings suggest the usefulness of TCM syndrome differentiation for non-invasive screening at routine health examinations.

TCM fundamentally includes herbal remedies, acupuncture and massages, which is prescribed based on its own diagnosis.
Diagnostic approaches employed in TCM are non-invasive including inspection, auscultation, inquiry and palpation. Through the TCM diagnoses, the conditions of individuals are categorized into different constitution types according to perceived features such as disharmony and functional disturbance. In TCM, syndromes are differentiated according to the eight principles that are yin, yang, exterior, interior, cold, heat, de ciency (xu) and excess (shi). Recent studies have shown that TCM syndrome differentiation is correlated with western medical laboratory indicators in several disorders. [1][2][3][4] Taken well acceptance and feasibility of TCM in practice, TCM syndrome differentiation holds great promise when it is implemented to routine health examination at a screening phase.
In this study, we investigated the association of TCM syndrome differentiation with medical indicators in a health check-up setting. The TCM syndromes used in this study were identi ed with focus on lower gastrointestinal malignancies. We applied TCM syndrome differentiation to identify physiological indexes that act as predictors for colorectal cancer.

Data source
We analysed the data collected from individuals who received self-paid health examinations including undergone colonoscopy at Taiwan Adventist Hospital from January 1st 2008 to December 31, 2016. The study was reviewed and approved by Institutional review board of Taiwan Adventist Hospital. Informed written consent was waived due to a retrospective study design.

Differentiation Of Tcm Syndrome
TCM syndrome differentiations of all cases were performed by TCM doctors according to the Standard of TCM Diagnosis and Curative Effect of Disease-Syndrome. TCM syndromes were identi ed as follows: (1) syndrome of accumulated dampness-heat (SADH), (2) intermittent dysentery with syndrome of internal blockade of static blood (IDSIBSB), (3) yang de ciency of spleen and kidney (YaDSK), (4) yin de ciency of liver and kidney (YiDLK) and (5) de ciency of both QI and blood (DQB). TCM syndromes were con rmed for each subject in the condition of two main symptoms or one main symptom with two secondary symptoms. If two or more than two kinds of syndromes occur in one subject, the main syndromes were judged by subsequent discussion.

Arti cial Neural Networks Model
Back-propagation neural (BPN) network used in this study was developed per previous study. 5 The BPN had three layers of neurons, which were input, hidden and output layer. Hidden layer consisted of 10 neurons. The nal network output layer consisted of only one neuron for each TCM syndrome. The learning rate was 0.2. The training termination condition was when the root mean square error (RMSE) was lower or equal to 0.0001 or at most 1000 times training repetitions.

Statistical analysis
All statistical analyses were performed using SPSS 17.0 software. The Pearson χ2 test was used to determine the correlation between TCM syndromes and clinicopathologic indexes, and the data were presented as the frequency and percentage. p < 0.05 was considered statistical signi cant.

Results
A total of 5231 subjects undergoing health check-up were included and assessed in this study. 55.9% (2922) of enrolled subjects were male and the mean age was 47.9 ± 10.7 years. Of the enrolled subjects, 56.7% had BMI values classi ed as overweight or obesity and 3108 subjects (59.4%) were detected HBsAb positive. Among subjects undergoing colonoscopy examination, haemorrhoid was found in 3000 subjects (57.4%) and 29 subjects (0.6%) were found with cancers. Regarding TCM syndrome differentiation, subjects diagnosed with SADH, IDSIBSB, YaDSK and YiDLK were 4490 (85.8%), 4576 (87.5%), 5207 (99.5%) and 4232 (80.9%) respectively. There were only 30 (0.6%) diagnosed with DQB. The clinical characteristics of the enrolled subjects are summarized in Table 1. We next determined the correlation of 5 TCM syndromes with clinical indexes. The results showed that SADH had signi cant correlations with BMI, the symptoms in following systems including respiratory system, cardiovascular system, abdomen system, neurologic system, skeletomuscular system and colonoscopy (Table 2). We found that IDSIBSB was signi cantly correlated with tumour markers including CEA, CA199 and AFP. In addition, statistical correlations were found between IDSIBSB and symptoms in many systems (Table 3). YaDSK was found only to have signi cant correlations with colonoscopy and symptoms in cardiovascular system, abdomen system and skin (Table 4). YiDLK was correlated with the levels of diagnostic markers that were associated with liver including AST, ALT, rGT, AFP, HBsAb and HBsAg. It also had high correlations with the symptoms in following systems including head and neck/lymphatic, respiratory, cardiovascular, abdomen, neurologic, skin and skeletomuscular systems (Table 5). Statistical correlations of DQB with symptoms in several systems was observed, including head and neck/lymphatic, respiratory, cardiovascular, abdomen, neurologic, skin and skeletomuscular systems (Table 6).     We also evaluated the predictive potential of clinical indexes for 5 TCM syndromes with focus on colorectal malignancies using arti cial neural networks approach. The strength of correlation of syndrome and variables was shown as the width of lines. The results showed that the predictive ANN model showed a good tting with an accuracy of 100%. The three most predictive factors for SADH were colonoscopy, skin and neural system with importance of 0.24, 0.09 and 0.08 respectively. The interrelationships between predictor factor (input layer), hidden factors (hidden layer), and SADH are shown in Fig. 1. For IDSIBSB, the predictive model exhibited an accuracy of 99.7% and the most predictive factors was colonoscopy with importance of 0.24 (Fig. 2). We found that the accuracy of predictive model for YaDSK was 100% and three factors with high importance were colonoscopy (0.24), digestive system (0.13) and CEA (0.05) (Fig. 3). For YiDLK, the prediction accuracy of ANN model was 100%. The factor with highest importance was cardiovascular system (0.16) followed by digestive system (0.08) and age (0.07) (Fig. 4). The data showed that 3 factors with highest importance in the ANN predictive model for DQB were cardiovascular system (0.26), digestive system (0.2) and skeletomuscular system (0.07).

Discussions
In the present study, we investigated the correlation of TCM syndrome differentiations with clinical indicators in a setting of health screening examination. The results indicated that 5 TCM syndromes had different degrees of correlations with certain clinical indexes and markers. We found that SADH, IDSIBSB, YaDSK and YiDLK had signi cant correlation with colonoscopy which is considered as gold standard for colorectal cancer screening.
TCM is practiced theoretically based on syndrome differentiation of which information is obtained through a series of diagnostic procedures. TCM has received tremendous attention as complementary approach in modern medicine practice. The integration of TCM syndrome differentiation and modern clinical diagnosis has been extensively studied. TCM syndromes have been translationally applied and veri ed in many diseases including irritable bowel syndrome 6 , coronary artery disease, hepatitis B cirrhosis 7 , purpura nephritis 8 , colorectal carcinoma 3 and post-stroke depression 9 . Recently, correlations between TCM syndromes and disorders have been explored and validated. Chen et al reported the association of kidney-vacuity syndrome with osteoporosis in postmenopausal women. 10 Qi de ciency and blood stasis syndrome has been reported to have a correlation with the bleeding after receiving percutaneous coronary intervention. 11 A study has shown that Qi-de ciency is associated with cancer related fatigue. 12 In addition to the association with disorders, TCM syndrome differentiation has been demonstrated to correlate with many measureable clinical indexes in modern medicine in disease setting.
A recent study has reported that the concentration of adiponectin was correlated with TCM de ciency factors in dyslipidaemia patients. 13 The correlations of TCM syndrome pattern with clinical indexes including renal function, immunologic function, and blood coagulation in patients with Henoch-Schönlein purpura nephritis have been reported. 8 Blood stasis syndrome has positive correlation with triglyceride and cholesterol in patients with primary glomerular disease. 14 In our study, we found that syndrome of accumulated dampness-heat (SADH) is correlated with level of cholesterol in the subjects undergoing health check-up. The nding is consistent with the result of Chen's study showing accumulation of dampness in dyslipidaemia patients. 13 Our data highlight that yin de ciency of liver and kidney (YiDLK) is associated with abnormality of liver function indexes. This is in an agreement with recent study in which common TCM syndromes in primary liver cancer are dampness, blood-stasis, heat, Yang de ciency of spleen and kidney, and Yin de ciency of liver and kidney. 15 Our ndings suggest the usefulness of TCM syndrome differentiation for non-invasive screening at routine health examinations.
Colonoscopy is the gold standard of colorectal cancer screening, which allows physician to examine the anatomic structure of colon and rectum. This procedure is invasive and performed commonly in routine health check-up. We evaluated the correlation between TCM syndromes and colonoscopy ndings. We found that 4 TCM syndrome had high correlation with colonoscopy in individual having health check-up, which were SADH, IDSIBSB, YaDSK and YiDLK. Our nding is similar to the results of Yang et al study showing that liver and kidney yin de ciency syndrome, spleen de ciency syndrome, spleen de cient qi stagnation syndrome, and damp heat syndrome were main syndromes in colorectal cancer patients. 16 Li and his colleagues reported that the TCM patterns of patients with colorectal cancer were blood de ciency, Qi de ciency, and Yin de ciency. 3 In our study, yang de ciency of spleen and kidney and yin de ciency of liver and kidney were correlated with colonoscopy, nevertheless, de ciency of both QI and blood had no correlation with colonoscopy. A possible explanation for this different observation is the physical status of enrolled individuals in each study. Further studies are necessarily required to explore the TCM syndrome distribution in different health status and to justify the difference among studies.

Conclusions
In conclusion, the results demonstrated that TCM syndromes were closely correlated with clinical laboratory indexes regardless of health status. TCM syndrome differentiation is suggested to contribute to routine health examination as screening measure with its non-invasive nature. TCM syndrome represents a promising diagnostic means for screening in routine health examination process.