Hello, folks! 
Today I’d like to discuss two things:
- Briefly introduce MCID (Minimal Clinically Important Difference) for those who don’t routinely use it in their research.
- Hear from those who do use it: What have you learned from incorporating MCID into your analyses?
MCID emerged as a way to assess the clinical relevance of changes in outcomes. Its central premise is that statistical significance does not necessarily imply clinical significance.
This distinction becomes especially important in studies with very large sample sizes, where even small differences may achieve statistical significance. For example, imagine a longitudinal study comparing Therapeutic X versus Therapeutic Y that finds a 1-point difference in the MDS-UPDRS total score with a p-value < 0.001. The result is statistically significant—but does a 1-point improvement actually translate into a meaningful benefit in a patient’s daily life?
MCID was developed to help answer that question. Several studies have proposed thresholds for commonly used Parkinson’s disease outcome measures. For example, Horváth et al. (2015) proposed an MCID of −3.25 points for minimal clinically important improvement and +4.63 points for minimal clinically important worsening in MDS-UPDRS Part III. Later, Horváth et al. (2017) proposed thresholds for the PDQ-39 Summary Index (−4.72/+4.22), PDQ-8 Summary Index (−5.94/+4.91), and then again in 2017 for MDS-UPDRS Part I (−2.64/+2.45), Part II (−3.05/+2.51), and the combined Parts I+II score (−5.73/+4.70), representing the smallest changes considered clinically meaningful for improvement and worsening, respectively.
Beyond Parkinson’s disease, MCID has also been described for other movement disorders, including Huntington’s disease and dystonia, highlighting its broader applicability across neurological research.
Another interesting paper, Can Testing Clinical Significance Reduce False Positive Rates in Randomized Controlled Trials? A Snap Review, reviewed 50 randomized controlled trials and found that only about 20% of trials with statistically significant findings also met the predefined threshold for clinical significance based on the minimum clinically important effect used in their sample size calculations. The authors also noted that none of the reviewed studies explicitly incorporated MCID into the interpretation of their findings, suggesting there may be opportunities to better integrate clinical significance into how we communicate and interpret research results.
That said, there are notable examples where clinically meaningful change has been incorporated into the interpretation of trial results. The EARLYSTIM study, for instance, complemented its primary analyses by evaluating the Minimally Important Change (MIC) in health-related quality of life (PDQ-39 Summary Index) using both anchor-based methods (Patient Global Impression of Change) and distribution-based approaches. Beyond demonstrating statistically significant improvements with deep brain stimulation, the study showed that the observed changes exceeded the estimated threshold for clinically meaningful improvement and that a substantially greater proportion of patients experienced benefits that were meaningful from the patient’s perspective. I think this is a great example of how incorporating clinical significance can provide a richer and more patient-centered interpretation of trial outcomes.
Of course, MCID is just one approach. Other methods can also help us evaluate whether a finding truly matters to patients. For example:
- Clinical Global Impression (CGI) scales, which capture clinicians’ overall assessment of improvement or worsening.
- Patient Global Impression of Change (PGIC), which directly asks patients whether they perceive a meaningful change.
- Goal Attainment Scaling (GAS), which evaluates success based on individualized patient goals.
- Measures of quality of life, activities of daily living, and patient-reported outcomes (PROMs).
- Functional outcomes such as independence, caregiver burden, healthcare utilization, or time to clinically meaningful milestones.
I’d love to hear your thoughts and experiences:
- Have you used MCID in your own research? If so, what insights has it provided that conventional statistical analyses might have missed?
- Are there other metrics—or even different conceptual approaches—that you use to determine whether a statistically significant finding is also clinically meaningful?
And finally, I’d like to end with a broader question.
As research continues to move toward increasingly large datasets, artificial intelligence, and the pressure to produce more and more results, how do you keep patients at the center of your work?
Not just through a metric, but through the way you formulate your research questions, interpret your findings, and define what success actually means (and whom should it benefit the most).
I’d be very interested to hear how others navigate this balance in this rapidly advancing research world! 
