Validation of new prognostic and predictive scores by sequential testing approach.

Area of Science:

• Radiation Oncology
• Clinical Prediction Tools
• Biostatistics

Background:

• Practitioners question the applicability of large-scale prediction scores to local patient populations.
• A sequential testing approach for in-clinic validation of new tools with limited patients was proposed.
• This study evaluates the sequential testing approach for radiation oncology scores.

Purpose of the Study:

• To evaluate the in-clinic validation of prediction scores in radiation oncology using a sequential testing approach.
• To determine if prediction tools are valid for local patient populations.
• To assess the feasibility of implementing new prediction tools at a local level.

Main Methods:

• Three scores predicting short overall survival after palliative radiotherapy were selected (bone metastases, brain metastases, spinal cord compression).
• A sequential testing approach was applied to consecutive patients for each score.
• Validation required a positive predictive value (PPV) exceeding 80%.

Main Results:

• The validity of two scores was confirmed locally with 13 and 17 patients, respectively.
• For the third score, no decision was reached even with 30 patients.
• Sequential testing allowed for timely validation in most cases.

Conclusions:

• In-clinic validation using sequential testing is preferable to uncritical adoption of prediction tools.
• This method can be implemented within reasonable time frames, even in small practices.
• Continuous validation occurs as data are collected, ensuring ongoing relevance.