Correlation Is Not Prediction: Reassessing Predictive MRI Evidence in Guidelines for Persons With Relapsing-Remitting Multiple Sclerosis | JoVE Visualize

Area of Science:

Neurology
Medical Imaging
Biostatistics

Background:

Accurate prediction of disease course in relapsing-remitting multiple sclerosis (RRMS) is crucial for effective treatment monitoring and decision-making.
Magnetic Resonance Imaging (MRI) outcomes are frequently cited in major MS guidelines (MAGNIMS, CMSWG) as predictive, yet the methodological rigor of this evidence is questionable.

Purpose of the Study:

To critically evaluate the methodological standards of predictive claims regarding MRI outcomes within four key multiple sclerosis (MS) guidelines.
To assess the quality of evidence supporting MRI-based predictions in RRMS treatment guidelines.

Main Methods:

A content review of citations within the MAGNIMS (2015, 2021) and CMSWG (2013, 2020) guideline publications was performed.
Evaluated sources for quantitative predictive evidence, including predictive values with confidence intervals, Kaplan-Meier estimates, or validated prediction models (assessing calibration and discrimination).
Examined the use of statistical measures such as correlations, odds ratios, hazard ratios, Prentice criteria, and likelihood ratio tests.

Main Results:

Most predictive statements in the guidelines relied on secondary citations and association-based measures (e.g., odds ratios, hazard ratios, correlations).
While some studies reported predictive values, confidence intervals were often missing. Validated prediction models were rarely cited, with only one undergoing full external validation.
Advanced statistical methods for prediction accuracy were largely absent in the cited evidence.

Conclusions:

Guideline statements on MRI prediction in RRMS currently emphasize associations over validated, individualized risk predictions.
The evidence lacks quantification of individual risk and robust assessment of accuracy, calibration, discrimination, or robustness.
Future guidelines should mandate prospective risk estimates with confidence intervals, externally validated models, and evaluation of clinical utility for trustworthy evidence.