---

DynNom Survival — User Guide

---

1. Page Structure

This app contains 6 tabs. Use the left panel for settings and the right panel for results:

• Data Preview — View the loaded data table (time, event, covariates)
• Nomogram — Display the Cox proportional hazards model nomogram
• Model Summary — Model summary with coefficients, HR, and P-values
• Model Performance — Time-dependent ROC and Calibration curves for model evaluation
• Prediction — Enter patient variable values to predict survival probabilities and Decision Curve Analysis
• Help — This page

2. Workflow

1. Select a data source (Demo Lung data or upload CSV/XLSX) on the left and click 'Load Data'
2. Select the time variable, event variable, and covariates (at least 2)
3. Set survival time evaluation points (default: event-time quantiles with min/max clamping)
4. Click 'Run Model' to build the Cox model
5. Switch to the 'Prediction' tab and enter patient variable values
6. Click 'Predict' to view survival probabilities and DCA at each time point with nomogram annotation

3. Prediction Details

In the Prediction tab:

• Enter values for each covariate (numeric or categorical)
• Supports custom evaluation time points, separated by commas or semicolons (e.g. 6,12,24,36)
• Results include: variable points, total points, linear predictor, survival probability AND event probability (1 − S(t)) at each time point
• The nomogram on the right uses red arrows to mark the current patient's position
• Decision Curve Analysis (DCA) shows the clinical utility of the model at each time point with adjustable threshold probability

4. Model Performance

In the Model Performance tab:

• ROC Curves: Time-dependent ROC (combined model linear predictor) at each evaluated time point, with AUC values
• Calibration Curves: Predicted vs. Observed (Kaplan-Meier) survival probabilities, grouped into 5 decile groups per time point

5. Data Format Requirements

• Time variable (time): Numeric, representing follow-up time (months)
• Event variable (event): 0/1 or TRUE/FALSE, 1 means event occurred (death)
• Covariates: Numeric or categorical, at least 2 required
• Missing values will be automatically removed

5. Model Principle

This app is based on the Cox proportional hazards model:

h(t|X) = h₀(t) × exp(β₁X₁ + β₂X₂ + ... + βₖXₖ)

Where h₀(t) is the baseline hazard and exp(βX) is the hazard ratio (HR).

The nomogram maps the Cox model's linear predictor to Points, then reads survival probabilities at each time point from the Total Points.

---

For questions, contact WeChat: icecoler