Article Text
PDF
Abstract
Background The ICE trial demonstrated the efficacy of immune globulin intravenous (IGIV-C) over placebo in chronic inflammatory demyelinating polyradiculoneuropathy (CIDP). However, improving the interpretability of the results by analysing the minimum clinically important difference (MCID) had not been considered.
Objectives To identify MCID thresholds of various outcome measures using different methods and to test treatment differences (IGIV-C vs placebo) using these thresholds.
Methods One anchor-based (Short Form-36 question 2) and three distribution-based (½ SD, 1 SE of measurement, and effect size) techniques were employed to identify MCID cut-offs for various impairments (electromyographic parameters, Medical Research Council (MRC) sum score, grip strength, inflammatory neuropathy cause and treatment (INCAT) sensory sum score), disability (INCAT scale score, Rotterdam handicap scale (RHS) score) and quality of life (SF-36). IGIV-C or placebo was administered every 3 weeks for up to 24 weeks to 117 CIDP patients. Patients who did not improve by ≥1 point on the INCAT scale received alternate treatment. The proportion of patients with results exceeding identified MCID thresholds was compared.
Results MCID cut-offs for outcomes were determined using each method. For the INCAT disability scale (primary ICE-trial outcome), all MCID methods identified significantly more responders with IGIV-C than placebo. Significant differences favouring IGIV-C were also demonstrated for various nerve conduction parameters, MRC sum score, grip strength, RHS score and SF-36 physical component summary score.
Conclusion In addition to being statistically significant, all MCID analyses showed that CIDP improvements with IGIV-C are clinically meaningful. Consideration of MCID is recommended in future therapeutic trials.
Trial Registration Number NCT00220740 (http://ClinicalTrials.gov)
Statistics from Altmetric.com
Introduction
Chronic inflammatory demyelinating polyradiculoneuropathy (CIDP) is characterised by a clinical course that may show a continuous or stepwise progression or may have a more fluctuating pattern over months to years.1 2 Chronic inflammatory demyelinating polyradiculoneuropathy is characterised by a predominantly symmetrical distal weakness and sensory deficits that may lead to considerable long-term disability with decrement in health-related quality of life expectations.1–4 Recently, the immune globulin intravenous (IGIV-C) CIDP efficacy trial (ICE trial), the largest randomised clinical trial of any agent for the treatment of CIDP ever performed, demonstrated the efficacy and safety of immune globulin intravenous, 10% caprylate/chromatography purified (IGIV-C).4 Significant statistical differences in favour of IGIV-C were observed when comparing the results obtained with the inflammatory neuropathy cause and treatment (INCAT) disability scale as well as other outcome measures. However, despite these robust findings, no clinical trial in CIDP had ever examined whether the statistically significant results reflect clinically relevant changes for patients. Minimum clinically important difference (MCID) has been advanced as a concept that could enhance the interpretability of results of clinical studies and overcome the shortcomings of the ‘statistically significant difference.’5 6 Jaeschke and associates first defined MCID as being ‘the smallest difference in score in the domain of interest which 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.’7 Despite this seemingly clear definition, the MCID concept has been criticised because of difficulties with its calculation and because different methods produce different MCID values.8 A number of techniques have been used to address these issues.5 6 8
The current study systematically addresses, for the first time in any neurological condition, the concept of MCID in a wide variety of outcome measures that capture disease progression in CIDP. Data were analysed from the ICE trial data to discover whether MCID techniques capture the efficacy of IGIV-C compared with placebo during up to 24 weeks of therapy.
Methods
Patients and study design
Complete details of the patient population and the randomised, double-blind, placebo-controlled, response-conditional study design and methodology are described in the publication of the primary efficacy analysis.4 Please also see the online supplementary material for more information. Eligible patients developed CIDP, progressive or relapsing motor and sensory dysfunction resulting from neuropathy over at least 2 months before study entry, and clinical disability as defined by an overall INCAT disability score of 2–9.9 Eligible patients were randomised to receive either IGIV-C (Gamunex, Talecris Biotherapeutics, Inc, Research Triangle Park, North Carolina) or placebo (0.1% albumin). Patients received a baseline loading dose of 2 g/kg and then a maintenance infusion of 1 g/kg every 3 weeks for up to 24 weeks. The selected outcome measures applied in the ICE study were based on various consensus meetings addressing standardisation of outcome measures in immune-mediated neuropathies; the WHO’s international classification of functioning, disability and health; and the quality-of-life concept.10–13 All selected measures had demonstrated acceptable interpretability, good validity, reliability and responsive values in patients with CIDP, and scales used in the ICE study were applied in a standardised fashion.14 For the purpose of the current study, analyses were performed using the first-period data (up to 24 weeks) based on the intention-to-treat principle.
MCID determination
Because there is no global consensus on which technique is optimal for calculating MCID, this study employed four methods for quantitatively determining MCID, using either patient perception or mathematical modelling to calculate an MCID threshold value.6 8 One of the methods included in this study, an anchor-based method, compares changes in scale scores with those of another measurement or phenomenon that has known clinical relevance (eg, patient perception of a clinical change after therapy). Three other techniques for calculating MCID, distribution-based methods, are determined from the statistical characteristics of the scores obtained (eg, effect size).5 6 8 The various threshold MCID values were then applied to assess the clinical importance of changes in outcome data from the first period of the ICE study. A goal of the study was to determine whether the selected MCID techniques would consistently rank all the scales used in favour of the IGIV-C group compared with the placebo group.
Anchor-based method
The global change of health assessment used the answers to Short Form-36 (Medical Outcomes Trust, Boston, MA; SF-36) question number 2 in which patients were asked to compare their global health rating at entry and at the end of the first period (week 24 receiving IGIV-C or placebo in a blinded fashion) versus 1 year ago.15 This question is not used in scoring the domains and summary measure of the SF-36. This question asked patients to compare their current health with their health 1 year ago. Patient response options were ‘much better,’ ‘somewhat better,’ ‘about the same,’ ‘somewhat worse,’ or ‘much better’ now compared with 1 year ago. Various MCID techniques have been applied using the response options of the SF-36 question number 2.16 17 The MCID for the current study was based on a widely used approach and was determined as the mean score difference between the response options ‘somewhat better now than 1 year ago’ and ‘about the same as 1 year ago.’17 18 Separately for both the IGIV-C and placebo groups, the patients were classified into two subgroups depending on whether their change in health score from baseline visit to week 24 was 1 (corresponding to relevant improvement) or 0 (corresponding to no improvement).
Distribution-based methods
The one-half SD (½ SD) benchmark of an outcome measure has been proposed as the ‘unifying theory’ for assessing the MCID6 19 and has gained support in the literature.20 21 The SD for all outcome measures is based on their theoretical range. For example, the theoretical range for the grip strength, using the Vigorimeter (Martin, Tuttlingen, Germany), was based on available normative data.22
The SE of measurement (SEM) was used to evaluate the intraindividual changes in selected outcome measures in both patient groups.23 The SEM is equivalent to the SD of a scale multiplied by the square root of one minus its reliability coefficient (Cronbach α). A change of 1 SEM has empirically been found to correspond with the original MCID criterion of an outcome measure.23 24 For each scale, the Cronbach α was calculated using only the placebo group with the repeat measurements at the baseline and at week 24 or the last measurement of the first period.
The effect size is the mean change in scores divided by the SD of the baseline scores. In the current study for the first period, the equation was: effect size=(μ24−μ0)/SDμ0, where μ24=mean scale value at week 24 or the last visit of the first period, and μ0=mean scale value at baseline.25 The effect size was calculated for both the IGIV-C and placebo groups. An effect size of 0.5 was considered an appropriate definition of MCID and was used in the current study.6 26
Applying calculated MCID values
The MCID cut-off values were determined for each outcome measure applied in the ICE study based on all four MCID techniques. A patient was defined as a responder for a particular scale if the clinical condition of the patient improved enough to reach the MCID threshold for that scale. The percentage of patients in each treatment group was then calculated for those who reached the predefined MCID thresholds at the end of the first period, and the percentages were compared between the groups (Fisher exact two-tailed test). A p value <0.05 was considered statistically significant. SAS version 8.2 was used for all statistical analyses (SAS Institute, Cary, North Carolina).
Results
During the first period of the ICE study, 117 individuals were randomised; 59 patients received IGIV-C, and 58 patients received placebo.4 Baseline values for the impairment measures—electromyographic parameters, Medical Research Council (MRC) sum score, grip strength, INCAT sensory sum score, activity and participation scale scores (INCAT disability scale and Rotterdam handicap scale (RHS)), and quality-of-life component measures (physical component score (PCS) and mental component score (MCS))—were similar between the two treatment groups.3 4 27
MCID calculations
Determination of MCID was calculated using four methods, an anchor-based approach (SF-36, question 2) and three distribution-based approaches (½ SD, 1 SEM, and effect size). The calculated MCID cut-off scores of all measures for the anchor-based, ½ SD, and 1 SEM methods are shown in table 1. To strive for uniformity, the mean MCID for the grip strength of dominant and non-dominant hands was calculated and applied to both hands. As indicated in the methods, the fourth technique of identifying an effect size of 0.5 as a cut-off value for MCID was also applied.
Anchor-based*, ½ SD, and 1-SEM distribution-based minimum clinically important difference (MCID) techniques for the outcome measures applied in the immune globulin intravenous chronic inflammatory demyelinating polyradiculoneuropathy efficacy trial (ICE) study (first period of ICE study)
Applying MCID calculations to the ICE study
Anchor-based MCID method
All outcome measures except the MCS scores demonstrated higher percentages of improved values for the IGIV-C-treated patients compared with the placebo group when using the SF-36 question number 2 as the anchor (table 2). The INCAT disability score (the primary endpoint of the ICE study) showed a statistically significant difference in the percentage of patients reaching the MCID cut-off values in favour of the IGIV-C group. Significant differences in favour of IGIV-C were also observed for changes in average amplitude for all motor nerves, average conduction velocity for all motor nerves, percentage changes in conduction block, grip strength (non-dominant hand), MRC sum score and RHS score (table 2). A consistent trend in favour of IGIV-C was also observed for all outcome measures using the MCID cut-off values determined by ½ SD and 1 SEM methods (table 2).
Responder analyses defined by the minimum clinically important difference (MCID) cut-off values for anchor-based, ½ SD, and 1 SEM techniques for the immune globulin intravenous, 10% caprylate/chromatography purified (IGIV-C) and placebo groups of patients with chronic inflammatory demyelinating polyradiculoneuropathy (first period of the IGIV-C chronic inflammatory demyelinating polyradiculoneuropathy efficacy trial study)
Distribution-based MCID method
In the ½ SD method, significant differences were obtained in favour of the IGIV-C group for the nerve-conduction parameters, improvement in the average CMAP amplitude for all motor nerves, average conduction velocity for all motor nerves and reduction in percentage of conduction block (table 2). More patients in the IGIV-C group reached the MCID cut-off values for the MRC sum score and dominant hand-grip strength (p≤0.012). Applying the MCID cut-off value obtained for the INCAT disability scale using the ½ SD method demonstrated that a substantially higher percentage of patients in the IGIV-C group reached the MCID threshold (52.2% vs 22.4% in IGIV-C and placebo groups, respectively; p=0.001). A significant difference in favour of the IGIV-C group was also seen for the MCID findings for RHS score.
An almost similar pattern was seen using the 1 SEM method to determine the MCID thresholds for the applied outcome measures (table 2). A substantial significant difference in favour of the IGIV-C group was observed for improvement in average amplitude for all motor nerves, improvement in average conduction velocity for all motor nerves, MRC sum score, grip strength (dominant hand), INCAT disability score and RHS score. For the INCAT sensory score and the component quality-of-life measures, the differences were not statistically significant.
For the fourth technique, using effect size to determine MCID, the effect size values were consistently higher in the IGIV-C group than with the scores in the placebo group (table 3). None of the applied outcome measures in the placebo group reached the MCID cut-off value of 0.5. In the IGIV-C group, the MCID cut-off value was obtained for INCAT disability score, grip strength and PCS. The INCAT disability score had the highest effect size. The effect size for MRC sum score, RHS score and changes in amplitude for most severely affected motor nerve nearly reached the cut-off score of 0.5.
Responder analyses comparing effect size scores in the immune globulin intravenous, 10% caprylate/chromatography purified (IGIV-C) and placebo groups of patients with chronic inflammatory demyelinating polyradiculoneuropathy (first period of the IGIV-C chronic inflammatory demyelinating polyradiculoneuropathy efficacy trial study)
Discussion
The current study is the first in the field of peripheral nerve treatment trials to use MCID methods to evaluate treatment response. It considered measures of impairment, activity and participation, and quality of life, and used both anchor-based and distribution-based methods on the data of the ICE trial.4 The four methods evaluated resulted in varied MCID cut-off values, an observation that has been reported in other studies.5 6 8 28 However, in the current study, the threshold variation among the four MCID methods was smaller for the INCAT disability score and RHS score than for other outcome measures. Thus, for these scales, the MCID threshold is likely to be well defined and corroborated by all four methods used.
The INCAT disability score consistently demonstrated a higher percentage of responders reaching the calculated MCID threshold in the IGIV-C group compared with the placebo group. The differences for the INCAT disability score were always statistically significant regardless of the MCID method used, thereby neutralising any debate about the optimal method for the MCID calculation in this study. Using the ½ SD as the unified theory MCID technique, a total of 52.2% of the IGIV-C group reached the MCID threshold for the INCAT disability score compared with 22.4% in the placebo arm in the first period of treatment, which is almost identical to the calculation of the number improving by ≥1 point, the primary outcome in the trial.4 Similar percentage scores were found when applying the 1 SEM distribution and SF-36 question number 2 as methods for assessing MCID. Using the effect size method, the INCAT disability scale demonstrated the highest value in the IGIV-C group and can therefore be considered the most sensitive to capture responsiveness when compared with other outcome measures applied.
A systematic trend in favour of the IGIV-C group was also seen when evaluating the proportion of responders demonstrating changes greater than the MCID thresholds for the other outcome measures. These differences were robust in both the anchor-based and distribution-based techniques, with significant changes seen for various nerve-conduction study parameters, MRC sum score, grip strength, RHS score and effect size for the quality-of-life PCS.
The anchor-based method using the SF-36 question number 2 to identify changes in global health has been criticized because patient answers may be influenced by health issues unrelated to the illness examined. The retrospective rating using SF-36 question number 2 with an extended period of time of 1 year is probably susceptible to a number of known recall biases.29 In light of these biases, global ratings may not give an accurate picture of an individual's ‘true’ change. Also, global ratings may only explain parts of the health-related quality-of-life findings.18 However, despite these reported shortcomings, the results in the current study with the SF-36 question number 2 as the anchor were comparable with the results seen in the various distribution-based methods, despite the differences in their cut-off values for the various outcome measures. Similar comparable clinically meaningful changes were also reported in patients with lung cancer using anchor-based and distribution-based MCID methods.30
Another important issue is whether the baseline level of functioning, disability or health status affects MCID thresholds. Do more severely affected patients with CIDP require a greater change from baseline to be considered clinically meaningful than patients less severely affected?28 31 In a study of patients with low back pain, a greater MCID change was required in patients with initial higher pain intensity than those with less pain.32 According to modern clinimetric approaches, clinically meaningful changes may vary across the theoretical range of an outcome measure.33 Another important issue is whether to use a combination of an anchor-based and a distribution-based method to ascertain improvement in patients with CIDP.30 34 To compensate for the differences seen in cut-off values for some outcome measures in the current paper and in the light of lack of consensus on this topic, we propose for future trials that patients with CIDP should be considered improved only when they meet one anchor-based plus one distribution-based MCID technique for change (the so-called ‘combined MCID robustness approach’) as has been addressed by others.30 34 35 The combination of SF-36 question number 2 (anchor-based form) plus the unified distribution-based method of ½ SD is suggested in the absence of any current consensus.
There are some methodological limitations to the current study that should be addressed. One overall limitation of using MCID techniques is the applicability to nonlinear scales. Most of the outcome measures evaluated in the current study were nonlinear, except for grip strength and neurophysiological data. Therefore, the analyses in the current study should be interpreted with some caution because it is conceivable that a calculated MCID may vary over the range of a nonlinear scale. Also, the MCID approach in the current study should not be simply extrapolated to other illnesses. The current MCID results could, however, be useful in trials with a similar patient population. The results also serve as a first step and framework for defining appropriate MCIDs in other chronic neurological conditions. Importantly, the Peripheral Neuropathy Outcome Measures Standardization (PeriNomS) Study Group has developed a linearly weighted overall disability scale for immune-mediated peripheral neuropathies, aiming to bypass the above postulated shortcomings and to increase the future applicability of MCID techniques in these conditions.36
Rheumatologists have used consensus meetings to determine the optimal outcome measures and methods of analysis,31 37–39 and neurologists have begun to follow a similar approach.10 11 Additional consensus meetings should be organised to help researchers in the field to determine whether MCID should be based on the patients' or the clinicians' point of view, whether more emphasis should be put on objective measures rather than subjective measures, whether future studies should be based on a desirable amount of change rather than a minimal measurable change, and whether inferences about MCID are made with respect to individuals or groups of patients.5–7 18 28 31 37–39
In conclusion, the current paper shows how MCID techniques for a variety of outcome measures can be used to compare disease progression in a CIDP treatment trial. The MCID approach adds a dimension of clinical relevance to statistical significance and to the conclusions about the efficacy of IGIV-C. The concept of MCID is an important contribution to the interpretation of clinical trials and deserves further consideration in this and other fields of neurology.
Acknowledgments
We thank the ICE study steering committee and investigators for their participation in the trial. The ICE Study steering committee includes: V Bril, MC Dalakas, PD Donofrio, H-P Hartung, R Hughes, N Latov, I Merkies and PA van Doorn. ICE Study investigators include: B Adamova, S Apostolski, M Banach, F Barroso, H Bartosik-Psujek, I Basta, E Belniak, L Benedetti, A Buchman, J Caress, S Cho, U Chyrchel, G Comi, P Dacci, U del Carro, V Drory, A Dubrovsky, R Fazio, W Fryze, E Fulgenzi, S Gonzalez-Cornejo, JDJ Gonzalez-Jaime, M Grandis, J Haas, M Kaminski, H Kwiecinski, C Marchesoni, C Munch, S Nations, A Pardal, H Patwa, S Pavlovic, R Reisin, S Romero-Vargas, C Ruiz-Sandoval, JL Ruiz-Sandoval, Z Stelmasiak, A Szczudlik, FP Thomas, B Tsao, A Uncini, A Villa, S Vohanka, G Wolfe, O Zapletalova and M Zielinska. Editorial assistance was provided by MedThink Communications.
References
Supplementary materials
Web Only Data jnnp.2009.194324
Files in this Data Supplement:
Web Only Data
Files in this Data Supplement:
Footnotes
Funding Funding for the study was provided by Talecris Biotherapeutics, Center for Science and Education (Research Triangle Park, North Carolina, USA).
Competing interests RACH and ISJM received honoraria from Talecris Biotherapeutics. CD and KH are employees of Talecris Biotherapeutics.
Ethics approval Ethics approval was provided by all institutes involved as part of the ICE study.
Provenance and peer review Not commissioned; not externally peer reviewed.