Introduction
The earliest Chinese medical record was written during the Shang Dynasty (ca. 11th century BC). From then on, the fundamental theory-system and future development-model of traditional Chinese medicine (TCM) were established gradually. The progress of the establishment was noted down in Huangdi’s Internal Classic and Classic of Difficult Issues, the current existing earliest works on basic TCM theory, and meridian and collateral theory (both were published around 475–221 BC), to Shennong’s Classic of Materia Medica, the earliest material medica in China (published during the West Han Dynasty, 5 BC), and Treatise on Cold Damage and Miscellaneous Diseases at the end of the East Han Dynasty (published around 200 AD). Clinical sub-disciplines, such as pediatrics and gynecology, appeared during the East Han Dynasty, indicating that the clinical diagnosis-treatment system of TCM has developed into a mature stage.
TCM is a distinguishing theory-system formed by constant summarizing, inducing, and refining of the experiences accumulated in preventing and treating diseases from daily life, production, and medical practice. It is one of the rarely existing ancient traditional medicines that hold systematic theories as well as preventative and therapeutic methods for diseases in practice. Due to its unique theory and vast applications in medical practice, TCM has gradually become the hot topic of application and research around the world [
1]. From the 1950s, such research methods as mathematics, statistics, and data-mining (DM) have been gradually introduced to the studies of TCM. Due to the separation of disciplines, the interaction between TCM and other subjects such as statistics is not sufficient. Nevertheless, statistics has been applied widely in the field of TCM [
2]. Meanwhile, the distinguishing features of TCM theories and related diagnostic and treatment models have constantly challenged the methodology of statistics. Some basic scientific problems remain to be solved. Hereon, this current work introduces the scientific features of traditional Chinese therapeutics, and discusses the major problems that statistics may face in TCM study.
Traditional Chinese therapeutics is based on regulating the function and condition of the human body. The process of describing, measuring, and evaluating the functions and conditions of the human body comprises the most urgent issue that must be resolved by statistics as a field of study
Due to the limitation of technology, TCM doctors in ancient times understood the condition of human body only from the external manifestations of an individual. TCM pays more attention to the external physiologic and pathological appearances of human beings. Its theoretical system is based on observations made regarding the functions and conditions of the human body. For example, the formation of the theory of visceral manifestation, although based on some direct observations of Chinese ancient anatomy practice, mostly came from the observations on different kinds of signs reflecting functional activities of the viscera. In addition, when treating diseases, TCM also focuses on adjusting the functional status of the body and aims at regulating its function and condition by sticking to the principle of restoring balance in the body (e.g. treating cold syndromes with heat therapy or treating heat syndromes with cold therapy). How to describe, measure, and evaluate the function and condition of human body is the most urgent issue that must be addressed by statistics as a field of study.
A comprehensive measurement and evaluation system, which is multi-dimensional and multi-index, is needed for describing and measuring the functions and status of the human body. Often, what are being evaluated are latent variables. This is because the process of describing and estimating is based mostly on clinical symptoms and signs that are usually subjective descriptions and not objective measurements. Thus, when making the statistical plan, there is a need to focus on the selection, quantization, and measurement of subjective and latent variables. In addition, when making an analysis, a statistical method that can analyze subjective and latent variables is needed. For subjective indicators, some of the conventional statistical methods, such as t-test, variance analysis and general linear regression, are not suitable because of the limitation of their application conditions. For latent variables, some methods that are commonly used, such as discriminant analysis, cluster analysis and multiple linear regression, are only suitable for linear correlation which is simple, single-dimensional and independent, but not suitable for multi-dimensional and complicated data. Therefore, appropriate methods are needed to solve the problems traditional Chinese therapeutics is facing. At the same time, these methods must be used properly.
Traditional Chinese therapeutics emphasizes on holism and individualization. Thus, the parameters that must be analyzed are very complicated and are nonlinear, multi-dimensional, multi-phase, polymorphic, and multi-granular
The conception of holism in TCM is emphasized from the aspects of integrity of the body and the unity of the human and his natural environment as well as that between the human and the social environment. In TCM, the human body is considered as an organic whole or a physiologic and pathological unity. It takes the five zang-viscera as its core and links the six Fu-viscera, orifices, and limbs as a complete, organic whole. The system not only explains that all constituent parts of the human body are structurally inseparable, it also shows that they are functionally coordinative and interactive. Second, according to TCM theory, human body is closely related with natural and social environments. Correspondences between “man and nature” and “nature and man in one” explain that humans and nature have a connection. The body maintains the unity of internal and external environment in the process of adapting to and transforming the nature environment in order to perform normal life activities. Accordingly, the influences of pathological changes, the application of the four diagnostic methods, and the establishment of syndrome differentiation and therapeutic measures should be considered upon the integrity of human body, the unity of human and nature, and human and social environment.
The patients treated by TCM are not only considered as biologic beings, but are also considered people within a social and natural environment. Thus, the evaluation indicators involved in TCM studies include not only biologic indicators that can be easily quantified, but also indicators of psychological function, social and environmental attributes, and so on. In TCM studies, complicated parameters need to be analyzed, and these parameters have the features of nonlinearity, multi-dimension, multi-phase, polymorphism, and multi-granularity [
3]. Consequently, when analyzing the relation between these parameters, we need the complex ideas as well as appropriate data analysis methods for this complex system. Currently, conventional statistical analysis methods, such as
t-test, one-way ANOVA, rank test, chi-square test, and so on, are used separately in TCM studies. These methods have their own application requirements. Misuse of such methods without considering the characteristics of data in TCM studies may lead to wrong conclusions. Some studies used statistical methods that are more complicated, such as multiple regression (including multiple linear regression, logistic regression, Cox proportional hazards model), cluster analysis, discriminant analysis, principal component analysis, and factor analysis. These methods also have their own data requirements. They may require the data to be independent, linear, or compliant with the distribution of some particular function. Some complex issues can be simplified by reducing the dimension and the mean of linear simplification using these methods. However, the indicators in TCM studies usually have multi-collinearity and coordination relationship that may cause the invalidation of regression models and lead to unreliable results. Therefore, these methods are still not sufficient in terms of meeting the need for the complex data analysis in TCM studies.
During the process of clinical treatment, TCM stresses that diagnosis and treatment should be based on the person, time, and place. In doing so, the practitioner can identify an individualized tactic, choose prescriptions, correct abnormality, and adjust unbalance to harmonize the function of the body. The same patient often has different syndromes in different stages of disease, so the prescription should be changed accordingly. Even if the syndrome differentiations are the same, the medication can be different because of different seasons and body constitutions. The individualized diagnosis and treatment mode of TCM raises requirements in complex analysis methods for unbalanced design.1
Individualized therapy has two connotations. First, the evidences for clinical diagnosis and treatment in TCM typically include many case studies. Finding the regularity of diagnosis and treatment from the numerous case studies is a great challenge for statistics. This is because it is usually difficult for people to evaluate the reliability of the conclusion due to the lack of external parallel control and possible selection bias in existing case studies. Second, the adjustments of intervention also bring problems to data analysis in the process of individualized therapy. Common issues in clinical practice include: parallel control studies cannot be conducted and interventions must be adjusted according to the individual differences. Multivariate statistical methods (cluster analysis and regression analysis) and data mining methods (neural network and Bayesian network) are all possible methods for obtaining scientific data and finding regularities from those individualized information [
4]. Several study design methods, such as pragmatic randomized controlled trials, might be more suitable for individualized studies than the classic methods [
5]. However, the application of these methods is still in the exploratory stages. The process of using them to solve problems existing in TCM clinical works calls for further studies.
The evaluation mode of traditional Chinese therapeutics is a circle of syndrome diagnosis, treatment, and evaluation. Thus, methods that can analyze longitudinal and repeating data are required
In TCM, treatment according to syndrome differentiation follows the instruction of holism. This requires examining a patient carefully with four diagnostic methods of TCM, grasping the characteristics of symptoms and signs caused by the disease, deciding on the cause of the disease according to the rule of “finding the cause with syndrome differentiation,” and conducting an analysis considering the environment, time, climate, and constitution, gender, and occupation of the patient, so as to identify the essence of the disease and determine the treatment. The condition of the patient and TCM syndrome may change in time; thus, doctors must regulate the therapeutic plan according to the changed syndrome. Therefore, the process of treatment according to syndrome differentiation is a circle of syndrome diagnosis, treatment, and evaluation.
In view of the dynamic feature of the TCM syndrome, the observation of corresponding indexes should not be static. One measure result may be too unilateral; hence, we need to observe the indexes repeatedly in different times. Otherwise, the changes in the diseases cannot be reflected accurately. For this, we need to set appropriate observational points and decide how many times we should make observations according to the specific features of the indexes. Unfortunately, arriving at a decision is still a problem waiting to be solved. At the same time, when choosing a statistical analysis method, the feature of TCM data (longitudinal, dynamic, and have correlations between different times of observation) should be considered. At present, there are a few statistical methods that are generally used when dealing with this kind of data, including repeated measure ANOVA, linear mixed effect model, generalized linear mixed effect model, generalized estimating equations, and so on. Further studies need to be done to provide more references for the use of these methods.
The compound, natural, and abundant TCM therapy brings multiple-target therapeutic effects. Statistical methods by which to evaluate these comprehensively are required
TCM treatments include drug therapy, such as herbal medicinal, as well as abundant non-drug therapies, such as acupuncture and moxibustion, massage, etc. Most Chinese medicines are compound formulas emphasizing the combination of “sovereign, minister, assistant, and courier.” When selecting different acupoints with different therapeutic effects in the treatment, compatibility is also considered in acupuncture treatment. Even if a single medicine or acupoint is used for treatment, it is the composite component that is expected to take effect. At the same time, the TCM interventions are mostly taken from nature, instead of modern artificial synthetic chemical substances. This is probably one of the reasons why treatments in TCM have fewer side-effects.
Almost all the TCM interventions are compound ones, and their effects often reflect the character of multiple targets. This is different from chemical drugs that work on a distinct and single target. Therefore, we have to make an effective comparison between the comprehensive multiple-target effect evaluation and single-target effect evaluation [
6]. We often have to explain: Is the combination reasonable in the compound intervention? How can the integrated multiple-target effect be revealed by simple and specific indexes? How can we compare different comprehensive effects? Is the weak combined effect more beneficial to patients than a significant single-target effect?
In conducting statistical analysis, humanistic thought in TCM requires more consideration on ecological features of data
Any science is a part of human culture, which may influence the choice of disease prevention and treatment methods. In TCM, humanity and science are even more evidently combined. In fact, there have always been cultural problems in medicine. In recent years, there has been an increasing awareness of considering humanity factors in scientific studies. This not only requires us to be alert to their influence when we analyze the relationship between the data, it also directly affects our analysis results. This is exemplified in the cases of the behaviors of the doctor and patient during the analysis of the placebo effect, influences of the researchers on the analysis of evaluation measurement, and so on. Thus far, there have been studies confirming the influence humanity factors have on interventional effect. For example, in the analysis of central effects in a diabetes clinical trial with syndrome of dual deficiency of
qi and
yin, the researchers found that even in the same clinical trial, the central effects of the objective testing indexes and subjective measuring indexes may not be the same, suggesting that there might be other effects caused by researchers [
7,
8]. Further studies must be done to measure the influence caused by the researchers and identify ways to avoid or reduce this influence. In addition, pragmatic randomized controlled trials [
5] and Zelen’s design [
9], etc. have been accepted by an increasing number of researchers while considering ethic reasons. The current statistical measures are often inadequate when dealing with specific issues in sociology. These issues are often divided or dealt with through reductionism. However, this is inappropriate. In recent years, statistics has been criticized as ecological fallacy for its neglect and inefficiency in dealing with data from complex social environment.
As a complex procedure of cognition and intervention, several rules of TCM diagnostics and treatment must be handled by newer statistical methods, so as to analyze and summarize data to provide scientific bases for the establishment of evidence-based diagnostics and treatment programs and clinical pathway. In fact, as modern medicine changes from the biomedical model to a biopsychosocial medical model, the main diseases that threaten the health of human beings have also changed from infectious diseases to chronic and non-infectious diseases. The integrity of human social function, the concept of holism, compound and individualized intervention, and humanity factors have all gradually become common foci of modern medicine and TCM. The problems discussed above are what both TCM and western medicine must face as of the moment. However, for TCM, these problems are more obvious. To all the statisticians, carrying out application studies on methodology or discussing new methods to solve scientific issues of TCM has become the common expectations of both TCM and modern medicine.
Higher Education Press and Springer-Verlag Berlin Heidelberg
PDF
(91KB)
