A new prognosis factor analysis based on nonhomogeneous Markov description.

Area of Science:

• Oncology
• Biostatistics
• Medical Informatics

Background:

• Traditional Cox proportional hazard models show limitations in analyzing breast cancer prognosis factors.
• Accurate identification of prognosis factors is crucial for effective treatment strategies in Stage II breast cancer.

Purpose of the Study:

• To develop and validate a new evaluation method for estimating prognosis factors in Stage II breast cancer.
• To combine Markov chain models and multiple logistic regression analysis for enhanced analytical ability.

Main Methods:

• A Markov chain model was constructed to illustrate breast cancer state transitions.
• Nonparametric tests identified patients with poor prognosis based on early recurrence (within 2.5 years).
• Multiple logistic regression analysis identified key prognosis factors: pathological diagnosis (n classification), ductal spread, and estrogen receptor status.

Main Results:

• The combined Markov chain and logistic regression model demonstrated effectiveness in identifying prognosis factors.
• The model's findings aligned with clinical experience, validating its practical utility.
• Significant prognosis factors for early death in Stage II breast cancer were identified.

Conclusions:

• The proposed method offers superior analytical capability compared to traditional models for breast cancer prognosis.
• This approach provides scientific evidence supporting clinical observations regarding prognosis factors.
• The identified factors (pathological diagnosis, ductal spread, estrogen receptor) are critical for predicting outcomes in Stage II breast cancer.