Objective: 

To investigate the classification and modeling of traditional Chinese medicine (TCM) syndromes in posthepatitic cirrhosis by partial least-squares (PLS) method, and to study the clinical application of PLS method in TCM research.

Methods: 

Inclusion criteria, exclusion criteria and criteria for harvesting and inputting of the biological information such as symptoms, signs, tongue, pulse and biological parameters were established. Stepwise regression was used to analyze the clinical information, including clinical symptoms and biological parameters obtained from 278 patients with posthepatitic cirrhosis by clinical epidemiological approach. TCM syndrome types were set up as dependent variable matrix, and the related clinical information was screened by stepwise regression as independent variable matrix. With the column maximization, all the variables were transformed into the PLS program. Mathematical models were obtained by Leave-One-Out in PLS program.

Results: 

The determination coefficients R ² of 5 regression equations were above 0.7. The practical accuracy rates of the models in TCM patterns including liver stagnation and spleen asthenia, liver-kidney yin deficiency, spleen-kidney yang deficiency, retention of dampness-heat, and blood stasis-heat accumulation were 74.02%, 68.71%, 92.45%, 84.17% and 70.50% by PLS, respectively.

Conclusion: 

The utilization of clinical information can be improved by PLS, which is definitely useful in the classification and modeling of TCM complicated syndromes.

Key words: hepatitis, liver cirrhosis, symptom complex, mathematics, models, partial least-squares