Home

Current Issue

Archives

Guidelines for Authors

Classified Ads

Links

Search PubMed

Subscriptions

Subscriber Registration

Guidelines for Website Users

JRheum Update Service

Contact Info

Definition of Clinical Differences

To the Editor:

The article by Wolfe, et al¹ compares the mean scores on the Health Assessment Questionnaire (HAQ) and the Physical Component Scores (PCS) of the Short Form-36 (SF-36) in different patient groups, using 6 different anchors to define these groups. The mean difference found between these groups is called Really Important Differences (RID). They chose as anchors for determining RID several clinical and societal outcomes, such as independence in participation activities versus not, or total joint replacement versus no total joint replacement. In this way they attach meaningful anchors to scores on the HAQ and the PCS of the SF-36. This is a very useful contribution to the interpretation of scores on the HAQ-DI and PCS by clinicians and patients. We would like to comment on 2 points.

Extrapolation of group differences to individual changes. We criticize the extrapolation of their cross-sectional findings about differences between groups to improvements in individual patients in randomized controlled trials (RCT) or clinical practice. First, the authors make no distinction between really important difference (RID) and really important change (RIC). They use these concepts interchangeably, which leads to confusion. An illustrative example appears in their Table 2: patients without a total joint replacement have more favorable scores on the HAQ and PCS than patients with a total joint replacement. This would suggest that surgery to replace a joint leads to deterioration of health. This shows that "really important change" cannot be derived directly from "really important difference." RID considers the differences between (groups of) individuals, while important improvements refer to changes within (groups of) individuals. To assess RID, different groups of patients are compared cross-sectionally, while RIC concerns intra-individual changes, which are assessed longitudinally by following the same individuals over time.

Second, Wolfe, et al present the mean scores and standard deviations (SD) on the HAQ and PCS of the SF-36 for different groups of patients. All SD are rather large, which means that there are substantial differences between patients. Consider for instance a patient scoring 0.80 on the HAQ, exactly the mean score of patients who are not work-disabled (Table 2 of Wolfe, et al). Is it safe to assume that this patient is really able to work? Not at all, as the 95% confidence interval of the work-disabled patients runs from 0.32 to 2.78 (i.e., 1.55 ± 1.96*0.63), provided the scores are normally distributed. It is obvious that a given work-disabled patient does not necessarily have to experience a "really important" improvement of 0.74 points (100% of RID) on the HAQ in order to regain the ability to work. In fact, in individual cases the necessary improvement, in terms of HAQ score change, may be smaller or much larger. Therefore, one should be cautious in extrapolating these group findings to individual patients.

Objections against the minimal clinically important difference (MCID). The authors point out a number of objections to MCID, which are in their opinion not encountered by defining RID. As we do not agree with any of these, we will react to them one by one.

Wolfe, et al state that "MCID represents a minimally clinically important (or detectable) change, which may be neither clinically meaningful nor useful." This objection has to do with the fact that they consider the MCID to be synonymous with the minimally detectable difference. They state that "A widely adopted approach for defining meaningful change is to identify the minimally clinically important difference (MCID), or the minimal detectable improvement perceptible to patients." However, the widely adopted definition of Jaeschke, et al² for MCID being "the smallest difference in score in the domain of interest which the patients perceive as beneficial and which would mandate, in the absence of troublesome side-effects and excessive cost, a change in the patient's management," clearly points to the importance of the change for patients and/or importance for clinical management. Note that in this definition MCID concerns a clinically meaningful change, and it does not contain any reference to what is minimally detectable².

Furthermore, Wolfe, et al see the dependence of MCID on baseline values as a disadvantage. However, this dependence is real and functional. For instance, if "independence in participation" is considered to be an important goal, patients who are close to this value need a smaller change than those patients whose baseline values are far from that intended score. This also holds for really important changes or differences.

As second objection, Wolfe, et al mention that "as MCID identifies a minimal detectable improvement rather than deterioration, it is not possible to interpret the magnitude of change patients perceive to be important." MCID might focus on improvements and deteriorations separately, and various studies have shown that these do not need to be the same³. Typically, Wolfe, et al determine "real important differences" instead of "real important changes" as they perform a cross-sectional analysis. Intuitively, one would say that the magnitude of "important differences" will be the same as magnitudes of "important changes." The example of total joint replacement shows that this is not true. In addition, the fact that minimally clinically important change (MCIC) for improvement and MCIC for deterioration sometimes differ, clearly implies that "important differences" are not the same as "important changes." Note that this holds for both MCID and RID.

Wolfe, et al state that "When applied to RCT, it is not always clear whether MCID should refer to absolute change from baseline or if one should subtract the result of placebo or comparator treatment." Often in RCT differences in change in different arms are determined. Using MCIC, one might report the percentage of patients showing a change larger than the MCIC value in each treatment arm (success rate), and analyze whether these percentages differ significantly between the arms. For RID, Wolfe, et al propose to express the results in percentage of RID achieved in each arm. Also in this case one should take the decision to refer to absolute change from baseline or subtract the result of placebo or comparator treatment.

Wolfe, et al state that "MCID does not offer clinicians an appropriate goal for improvement, based on patient's perceptions of realistic and desirable HAQ-DI or SF-36 PCS scores." We do not agree with this statement. In our opinion, MCID and MCIC values are far more realistic than RID values. One important point is that the large differences reported in this cross-sectional study may not be attainable (any more) for most patients thinking about improvement. This is illustrated by the fact that Wolfe, et al suggest expression of the change as a percentage of RID in RCT and clinical practice.

As a last objection, Wolfe, et al state that it is not clear how definitions of MCID should be used to interpret results from RCT or applied to clinical practice. In our opinion the MCID or MCIC defines the boundary between success and failure. The analysis of RCT as described above, comparing the percentages of patients that reach the MCIC or MCID in both treatment arms, is a very realistic approach. The ways in which MCID and RID are applied are the same in this respect.

In summary, in our opinion the RIC is a subclass of MCIC concerning the extreme situation that patients perceive only large changes as minimally important. The OMERACT group⁴ has presented a cube with 3 dimensions that are all at issue in this report: a dimension indicating change or difference, a dimension indicating group or individual level, and a dimension indicating the type of difference/change being assessed. In the latter they distinguish minimum potentially detectable, minimum actually detectable beyond error, observed in a population, observed in those estimated to differ/to have changed, and observed in those estimated to have an important difference/change. Information about all these anchors on the scale of a measurement helps to interpret the (change in) scores.

We would plead for rehabilitation of the term MCIC (not MCID) as a change that patients would consider important to reach in their situation, dependent on baseline values or severity of disease, on the type of intervention, and on the duration of the followup period. This MCIC value would be an important parameter to be used in power calculations in the design of clinical studies.

HENRICA C.W. de VET; HELEEN BECKERMAN; CAROLINE B. TERWEE; BEREND TERLUIN, MD, and LEX M. BOUTER, for the Clinimetrics Working Group, VU University Medical Center, Amsterdam, The Netherlands. E-mail: hcw.devet@vumc.nl

REFERENCES

Search PubMed for: