By: Chad E Cook, PT, PhD, FAPTA
Background: In clinical research, especially in orthopedics, sports medicine, and rehabilitation, the minimally clinically important difference (MCID) has become a staple of outcome interpretation. By definition, an MCID is defined as the smallest change in an outcome that an individual patient perceives as beneficial. It’s patient-centered, intuitive, and easy to communicate to patients, clinicians, insurance providers, and policy-makers. When a patient improves by at least the MCID, we can say that the improvement was clinically meaningful to that individual. What’s not to like?
Well, there is a lot not to like, including the points that my colleagues [1,2] and I [3,4] have made since 2008. In summary, we found that the large number of methods for calculating MCIDs, as many as nine, produce widely varying thresholds. These thresholds are further influenced by differences in patient populations and baseline severity, which in turn lead to inconsistent predictive modeling and challenges in interpreting results. These issues have led to a litany of different MCID values and much consternation among clinicians who have to weed their way through this confusion.
More Problems: Recently, as I was assisting a colleague on the interpretation of data from their shoulder study, I encountered a statement in the discussion section of their paper: “Although the difference between groups was statistically significant, it did not exceed the MCID and therefore may not be clinically meaningful.” It sounded reasonable, it felt logical, but it was wrong. In fact, one of the most common errors in the literature is applying a within‑subject-derived MCID to judge the meaningfulness of between‑group differences. This blog unpacks why this practice is flawed, why it persists, and what researchers should be doing instead.
Why is This Wrong? MCID is typically derived using either anchor-based methods (e.g., global rating of change) or distribution-based methods (e.g., 0.5 SD, SEM). Sometimes they are calculated using a combination of both. Regardless of the method, the MCID is about change within a person, not differences between groups. Let’s look at an example involving individuals with shoulder pain as it will help frame this appropriately:
- Group A (exercise) improves by 15 points (on a patient-reported outcome).
- Group B (cognitive behavioral therapy) improves by 10 points.
- The published MCID for patients with shoulder pain who receive these interventions is 12 points (for the patient-reported outcome).
- The between-group difference is 5 points, and since 5 points is less than 12, the difference is not clinically meaningful.
This is a classic category error, as it applies a threshold designed for individual change to group mean differences. A group‑level difference of 5 points could still reflect a meaningful difference in the proportion of patients who achieve MCID. Conversely, a group‑level difference larger than the MCID does not guarantee that more patients in one group actually experienced meaningful improvement.
Misuse of MCID in comparative studies is extremely common. Numerous published examples show researchers applying a within‑subject MCID to judge between‑group differences, an approach that is conceptually incorrect [5]. In some cases, authors have concluded that a between‑group difference is not clinically meaningful simply because it does not exceed a within‑person MCID, even though MCID was never intended for group‑level interpretation [6].
The opposite error also appears frequently: studies claiming that a between‑group difference is clinically meaningful because one group exceeded a within‑group MCID threshold, despite the actual between‑group effect being trivial. This misapplication occurs because the MCID is not symmetric between individuals and groups. Using a within‑group MCID to judge between‑group effects can either inflate or diminish clinical claims, depending on the direction of the misuse [7]. This keeps being reported in the literature this way because clinicians understand MCID, reviewers like MCID, and it’s a simple way to say “this matters” or “this doesn’t.” It’s also wrong. It is so common, it makes me wonder if I’ve made this error in any of my previous papers!
What are the Correct Strategies to Show Meaningful True Between-Group Differences? The following are three correct strategies to reflect meaningful group differences
Example One: Report the between‑group effect size. Effect sizes, such as Cohen’s d, are specifically designed for comparing groups because they express the magnitude of the difference relative to the data’s variability. Depending on the effect‑size metric used, one can directly quantify how far apart two groups are. For example, a Cohen’s d of 0.3 indicates that approximately 62% of the comparison group scores below the average of the experimental group, whereas a Cohen’s d of 0.8 indicates that about 79% of the comparison group scores below the experimental‑group average [8].
Example Two: Report the proportion of patients who achieve the MCID. This is the appropriate way to use MCID in a comparative study because MCID is fundamentally a within‑person construct. By examining how many individuals in each group reach that threshold, you preserve the individual‑level meaning of MCID while still enabling a valid between‑group comparison. For example, you might find that 68% of patients in Group A achieve the MCID, compared with only 42% in Group B. This approach respects the within‑subject nature of MCID and provides a clear, interpretable basis for comparing treatment effectiveness across groups.
Example Three: Consider deriving a between‑group minimally important difference (MID). Some fields now report between‑group MIDs, which typically fall in the range of 0.3–0.5 SD of baseline scores. This provides a far more appropriate benchmark for interpreting group differences than a within‑subject MCID. Studies that derive between‑group MIDs (MID‑BG) consistently show that these thresholds are smaller than within‑person MCIDs—a pattern that makes intuitive sense. Detecting a meaningful difference between treatments generally requires less change than detecting a meaningful improvement within an individual patient.
Why This Matters: Misusing MCID for between‑group comparisons may seem harmless, but it can distort clinical interpretation. It can lead to:
- Dismissing meaningful differences between treatments since between-group differences are frequently smaller than MCID values
- Overstating the equivalence of interventions
- Muddying the evidence base
As the field moves toward more patient‑centered outcomes, it’s essential that we use our tools correctly. MCID is a powerful concept, when used correctly (which is almost never), as it helps clinicians understand whether patients truly improve, not just whether numbers change. But it was never meant to be a direct yardstick for comparing treatments.
The next time you read (or write) a study that says: “The between-group difference did not exceed the MCID…” Pause. Ask whether the authors are applying MCID in a way it was never intended to be used. And consider whether a more appropriate metric (e.g., effect size, comparative proportions who met the MCID, or between-group MID) would tell a clearer story. Because in the end, good research isn’t just about numbers. It’s about using the right numbers in the right way to answer the right questions
Key Points:
- MCID is not a between‑group metric. It reflects meaningful change within an individual, not differences between treatment groups.
- Its common misuse does not make it correct. The fact that many studies apply MCID to group comparisons simply reflects a widespread misunderstanding.
- Misapplying MCID blurs the distinction between statistical and clinical significance. It can lead researchers to overstate or understate the clinical relevance of a treatment effect.
- A statistically significant group difference may or may not be clinically meaningful—but MCID is not the appropriate tool for making that judgment at the group level.
References
- Cook CE. Clinimetrics Corner: The Minimal Clinically Important Change Score (MCID): A Necessary Pretense. J Man Manip Ther. 2008;16(4):E82-3.
- “Minimum Clinically Important Differences May Not Be So Meaningful…” (JOSPT Blog, 2022)
- Peacock JL, Lo J, Rees JR, Sauzet O. Minimal clinically important difference in means in vulnerable populations: challenges and solutions. BMJ Open. 2021 Nov 9;11(11):e052338.
- Copay AG, Subach BR, Glassman SD, Polly DW Jr, Schuler TC. Understanding the minimum clinically important difference: a review of concepts and methods. Spine J. 2007 Sep-Oct;7(5):541-6.
- Angst F, Aeschlimann A, Angst J. The minimal clinically important difference raised the significance of outcome effects above the statistical level, with methodological implications for future studies. J Clin Epidemiol. 2017 Feb;82:128-136.
- Cook C. Clinimetrics Corner: Use of Effect Sizes in Describing Data. Journal of Manual & Manipulative Therapy. 2008;16(3):54E-57E.
