MINIMAL CLINICALLY IMPORTANT DIFFERENCE IN MYASTHENIA GRAVIS: OUTCOMES FROM A RANDOMIZED TRIAL HANS D. KATZBERG, MD, MSc,1 CAROLINA BARNETT, MD,1 INGEMAR S.J. MERKIES, MD PhD,2 and VERA BRIL, MD1 1 University Health Network, Division of Neurology, Toronto General Hospital, 200 Elizabeth Street, 5ES-306, Toronto, Ontario M5G 2C4, Canada 2 Department of Neurology, Spaarne Hospital, Hoofddorp and Maastricht University Medical Centre, Maastricht, The Netherlands Accepted 1 August 2013 ABSTRACT: Introduction: The minimal clinically important difference (MCID) is the smallest outcome change that has clinical significance. Its use has not been established in the study of myasthenia gravis (MG). Methods: Patients from a published intravenous immunoglobulin (IVIg) vs. placebo study were studied. One anchor-based and 3 distribution-based techniques were used to identify quantitative myasthenia gravis score (QMGS), repetitive nerve stimulation (RNS), and single-fiber electromyography (SFEMG) MCID cut-offs. Patients with a change-score exceeding MCID cut-offs were compared. Results: MCID cut-offs were below a QMGS change of 3.0. Anchor-based and 1 3 SEM cut-offs showed 58.3% vs. 30.7% responders (P 5 0.017), 1=2 SD 54.2% vs. 19.2% responders (P 5 0.018), and effect size 0.519 vs. 0.164 (P 5 0.011) in IVIg vs. placebo. Anchor-based (P 5 0.73) and effect-size (P 5 0.41) MCID cut-offs did not show a difference between IVIg and placebo. MCID methods did not produce meaningful RNS cut-offs. Conclusions: QMGS MCID values provide clinically relevant information and are recommended in MG trials. MCID analysis shows that improvement in MG patients treated with IVIg reflects clinically meaningful changes. Muscle Nerve 49: 661–665, 2014
Clinical trials evaluating treatment effects in myasthenia gravis (MG) have used a variety of scales over the years, including the quantitative myasthenia gravis score (QMGS).1 Early studies of interrater reliability of the test determined that an improvement of >2.5 units is statistically significant, and trials have consistently used improvement of >3.5 units to differentiate between responders and non-responders.2 Despite the consensus, there has never been a comprehensive validation of these cut-offs, including intraobserver (test–retest) reliability studies, to determine the minimal detectable change and therefore whether the commonly used cutpoints are meaningful clinically. There have been several definitions of what should be considered as change when repeating a measurement. All measurement tools are subject Abbreviations: CIDP, chronic inflammatory demyelinating polyneuropathy; ICE, IVIg chronic inflammatory demyelinating polyneuropathy efficacy; IVIg, intravenous immunoglobulin; MCID, minimal clinical important differences; MG, myasthenia gravis; QMGS, quantitative myasthenia gravis score; RNS, repetitive nerve stimulation; SFEMG, single-fiber electromyography; VAS, visual analog scale Key words: MCID; myasthenia gravis; neuromuscular junction; outcomes; QMGS Correspondence to: H.D. Katzberg; e-mail: [email protected] C 2013 Wiley Periodicals, Inc. V
Published online 12 August 2013 in Wiley Online Library (wileyonlinelibrary. com). DOI 10.1002/mus.23988
MCID in MG
to measurement error, so for any change to be considered significant, it should be beyond that error. The minimal detectable change then is the smallest change that goes beyond the error with repeated measures.3 It is calculated by constructing a 90% or 95% confidence interval around the error of measurement with the following formula: 1.96 3 SD 3 冑2(1 2 r), where SD is the standard deviation of the scores and r is the intraclass coefficient (ICC) of reliability. This means that if a score change exceeds that range, we are 95% certain that it was not due to error. However, it is possible to have statistically significant change (beyond error) without it being relating to a clinical benefit. The shift from statistical to clinical significance has prompted the concept of the minimal clinically important difference (MCID), which is the smallest change score in a measure of interest that patients perceive as beneficial.4 This has been applied increasingly in neurological diseases, most recently in a large group of patients with chronic inflammatory demyelinating polyneuropathy (CIDP) who were treated with intravenous immunoglobulin (IVIg).5 There are several methods to calculate the MCID, such as anchor methods. These methods couple a clinically relevant definition of improvement, such as the patient’s perception of improvement, to the change in the measurement used. Distribution-based methods are based on statistical definitions of change and include using 1=2 SD or 1 3 SEM (standard error of measurement). The latter is mostly related to minimal detectable change; this is why multiple approaches are recommended, as the obtained MCID values may vary according to the methods used, but they also may be influenced by the baseline values, the intervention under study, and the time elapsed. The aim of this study was to identify the MCID for several outcome measures in MG by using data from a previously published clinical trial of the effects of IVIg vs. placebo.1 To determine whether results were clinically meaningful in addition to being statistically significant we then applied these cut-offs to the data with respect to: (a) the QMGS; (b) single-fiber electromyography (SFEMG); and (c) repetitive nerve stimulation (RNS). MUSCLE & NERVE
May 2014
661
METHODS
Patients with MG studied at our institution from 2004 to 2005, who participated in the IVIg vs. placebo study, were selected for inclusion in this study.1 The inclusion and exclusion criteria and methodology are described in the original article. In brief, patients 18 years of age with a diagnosis of MG and worsening weakness were enrolled in the study. Patients were excluded if they had respiratory distress requiring intensive care unit admission, a vital capacity 3.5 points (21.11 to 22.01). Using a cut-off of 3, all MCID methods for QMGS demonstrated significantly more responders in the IVIg group compared with placebo (Fig. 1a). The 1 3 SEM and anchor-based cut-offs both resulted in 58.3% vs. 30.7% responders (P 5 0.017) for IVIg vs. placebo. The 1=2 SD-based cut-offs resulted in 54.2% vs. 19.2% responders (P 5 0.018) (Fig. 1b). The effect-size method met MUSCLE & NERVE
May 2014
Table 2. Anchor- and statistical-based MCIDs in patients with MG treated with IVIg vs. placebo. Anchor-based approach
1/2 SD approach
Parameter
Unchanged*
Slightly better*
MCID cut-off
QMGS RNS baseline RNS postexercise RNS 1-min postexercise SFEMG jitter
20.41 6 1.9 28.2 6 15.4 25.1 6 10.0 210.1 6 16.1 5.3 6 25.2
22.08 6 3.5 0.45 6 9.8 0.0 6 6.3 20.1 6 9.7 28.9 6 72.9
21.67 12.70 5.10 10.00 214.20
SD 4.03 1.60 1.60 1.60 3.00
SEM
MCID cut-off
Cronbach alpha
MCID cut-off
22.01 0.80 0.80 0.80 1.50
0.954 0.752 0.720 0.819 0.735
–1.11 0.797 0.846 0.680 1.540
MCID, minimal clinical important differences; QMGS, quantitative myasthenia gravis score; RNS, repetitive nerve stimulation; SFEMG, single-fiber electromyography. *Data expressed as mean 6 standard deviation.
0.5 MCID in the IVIg group (0.519), but not in the placebo group (0.164) (P 5 0.011) (Fig. 1c). The ROC method showed that a cutpoint of 2 had a sensitivity of 60% and a specificity of 71%, and a cutpoint of 3 had a sensitivity of 56% and a specificity of 89% (area under the curve 5 0.73). To correct for baseline QMGS, patients were divided into 3 groups: QMGS 0–9 (n 5 19); QMGS 10–16 (n 5 19); and QMGS >16 (n 5 12). The QMGS MCID based on anchor methods for each of these tertiles was 21.6, 21.8, and 22.75, respectively. When the rate of response in the IVIg and placebo groups was reanalyzed using the cutpoints corrected for baseline values, the values were similar to the anchor and 1 3 SEM cutpoints: the IVIg group had 54% responders compared with 30% in the placebo group. However, this did not reach statistical significance (P 5 0.43), which may have been due to the small proportion of patients who had a baseline QMGS >16 (12 patients). Only anchor-based methods produced a meaningful SFEMG MCID cut-off (214.2 ls). Using this cut-off, a significant difference between IVIg and placebo was not identified (27% vs. 19% responders, P 5 0.73). A significant difference was also not found when the effect-size method was used (0.343 vs. 0.013, P 5 0.41). None of the MCID methods produced a meaningful cut-off for RNS outcomes at baseline and after exercise and thus were not applied. DISCUSSION
In this study we applied MCID cut-offs in patients with MG. A similar methodology has been applied to patients with CIDP using the data from the recent IVIg CIDP efficacy (ICE) study, which evaluated the effects of IVIg vs. placebo.5 Recent trials in patients with rheumatological and other chronic diseases used MCID and statistical methods in reporting outcomes.9–12 In our cohort, anchor- and distribution-based methods were consistent, distributed over a narrow range, and did not exceed the original2 or estabMCID in MG
lished cut-offs used in most clinical trials for QMGS. The consistency of the values obtained demonstrates the strength of the results and supports the recommendation that these values become incorporated into practice. The relevance of knowing the MCID for these outcome measures is that treatment efficacy can be assessed through changes that are relevant to the patients and not just statistically significant. The use of different methods to calculate the MCID is another strength of this study, as we looked not only at values relating to error of measurement but most importantly to patients’ perceptions. In a recent publication we showed that the only predictor of response to immunomodulation with IVIg or plasmapheresis in MG was baseline severity, but we used QMGS cut-offs of >3.5 to define responders.13 In the present study we found different MCID cutpoints depending on baseline severity: 2 for patients with QMGS between 0 and 16 and 3 for QMGS >16. If confirmed in a larger cohort, this correction should be taken into consideration in clinical trials when defining responders. This may also have implications in calculating sample size in MG clinical trials, as a different number of patients may be required to be able to detect clinically relevant treatment differences depending on disease severity at entry. We found an MCID jitter cut-off of 214.2 ls (16% change of the mean), which is slightly higher than previous approximations of clinically significant change in jitter of 10%14 and only slightly above a previously reported estimated jitter measurement error of 12.2%.15 Thus, jitter may be of questionable utility in clinical trials. This cut-off did not show significant differences between placebo and IVIg in our group of patients, which is not surprising given that the study was not powered to detect differences in electrophysiology. MCID methods did not yield clinically meaningful cut-offs for RNS methods and, as such, comparison of effectiveness in the IVIg vs. placebo group was not possible. Although additional MCID methods MUSCLE & NERVE
May 2014
663
FIGURE 1. (a) Change in QMGS in IVIg and placebo at 2 weeks compared with baseline. Each dot represents an individual patient. Dashes represent MCID cut-offs. (b) Percentage of responders to IVIg (squares) and placebo (triangles) using MCID cut-offs. (c) Effect size in patients receiving IVIg (squares) and placebo (triangles).
and larger data sets could be utilized in future studies to further evaluate RNS as a valid outcome measure, it is still not likely to yield useful results. This is also not surprising given that RNS is not a sensitive test in MG. There are a number of limitations to MCID analysis in our group of patients. MCID has been criticized due to a lack of agreement on which method is best to establish cut-offs, and there are no universally accepted methods for calculating sample size.16 Also, given the small number of patients in our study, it is difficult to confirm that the MCID cut-offs are reliable, particularly with regard to tertile analysis and the predictive ability of the cutpoints obtained by the ROC curve. These findings should be confirmed in a larger data set. A combination of anchors, including a quality-oflife scale should also be studied further. There were very few patients who reported worsening (n 5 8). Thus, the MCID for worsening could not be studied, as some studies showed that the significant values for worsening and improving tend to differ. Finally, we recognize that a limitation of the QMGS, as with any ordinal measure, is that the change scores in different patients are not necessarily comparable and that, in the future, transformation of the ordinal measures to interval ones may be studied to establish MCID cut-offs in a more appropriate way.17,18 In spite of these limitations, we recommend that the MCID be studied further for the QMGS in addition to other outcome measures in myasthenia, such as the MG composite,19 to have a better 664
MCID in MG
understanding of what constitutes clinically meaningful change. Although the cut-offs identified in this study will need to be confirmed in a larger data set, using a QMGS cut-off of 2 for patients with a baseline QMGS of 16 should be considered in future research studies. REFERENCES 1. Zinman L, Ng E, Bril V. IV immunoglobulin in patients with myasthenia gravis: a randomized controlled trial. Neurology 2007;68:837– 841. 2. Barohn RJ, McIntire D, Herbelin L, Wolfe GI, Nations S, Bryan WW. Reliability testing of the quantitative myasthenia gravis score. Ann NY Acad Sci 1998;841:769–772. 3. De Vet HCW, Terwee CB, Mokkink LB, Knol DL. Measurement in medicine. Cambridge, UK: Cambridge University Press; 2011. 4. Jaeschke R, Singer J, Guyatt GH. Measurement of health status: ascertaining the minimal clinically important difference. Control Clin Trials 1989;10:407–415. 5. Merkies IS, van Nes SI, Hanna K, Hughes RA, Deng C. Confirming the efficacy of intravenous immunoglobulin in CIDP through minimum clinically important differences: shifting from statistical significance to clinical relevance. J Neurol Neurosurg Psychiatry 2010;81: 1194–1199. 6. Juniper EF, Guyatt GH, Willan A, Griffith LE. Determining a minimal important change in a disease-specific quality of life questionnaire. J Clin Epidemiol 1994;47:81–87. 7. Copay AG, Glassman SD, Subach BR, Berven S, Schuler TC, Carreon LY. Minimum clinically important difference in lumbar spine surgery patients: a choice of methods using the Oswestry Disability Index, Medical Outcomes Study Questionnaire Short Form 36, and pain scales. Spine J 2008;8:968–974. 8. King MT. A point of minimal important difference (MID): a critique of terminology and methods. Expert Rev Pharmacoecon Outcomes Res 2011;11:171–184. 9. Wyrwich, KW, Tierney WM, Babu AN, Kroenke K, Wolinsky FD. A comparison of clinically important differences in health-related quality of life for patients with chronic lung disease, asthma, or heart disease. Health Serv Res 2005;40:577–591. 10. Stauffer ME, Taylor SD, Watson DJ, Peloso PM, Morrison A. Definition of nonresponse to analgesic treatment of arthritic pain: an analytical literature review of the smallest detectable difference, the minimal detectable change, and the minimal clinically important difference on the pain visual analog scale. Int J Inflam 2011;2011:231926.
MUSCLE & NERVE
May 2014
11. Tubach F, Ravaud P, Baron G, Falissard B, Logeart I, Bellamy N, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis 2005;64:29–33. 12. Rigby W, Ferraccioli G, Greenwald M, Zazueta-Montiel B, Fleischmann R, Wassenberg S, et al. Effect of rituximab on physical function and quality of life in patients with rheumatoid arthritis previously untreated with methotrexate. Arthritis Care Res 2011;63:711–720. 13. Katzberg HD, Barnett C, Bril V. Predictors of response to immunomodulation in patients with myasthenia gravis. Muscle Nerve 2012; 45:648–652. 14. Sanders DB, Stalberg EV. AANEM minimonograph #25: Single-fiber electromyography. Muscle Nerve 1996;19:1069–1083.
MCID in MG
15. Buchman AS, Garratt M. Determining neuromuscular jitter using a monopolar electrode. Muscle Nerve 1992;15:615–619. 16. Walters S. Comparison of the minimally important difference for two health state utility measures: EQ-5D and SF-6D. Qual Life Res 2005; 14:1523–1532. 17. Rasch G. Probabilistic models for some intelligence and attainment tests. Chicago: University of Chicago Press; 1980. 18. Vanhoutte EK, Faber CG, van Nes SI, Jacobs BC, van Doorn PA, van Koningsveld R, et al. Modifying the Medical Research Council grading system through Rasch analyses. Brain 2012;135:1639–1649. 19. Burns TM, Conaway MR, Sanders DB; MG Composite and MGQOL15 Study Group. The MG composite: a valid and reliable outcome measure for myasthenia gravis. Neurology 2010;74:1434–1440.
MUSCLE & NERVE
May 2014
665
Clinical trials evaluating treatment effects in myasthenia gravis (MG) have used a variety of scales over the years, including the quantitative myasthenia gravis score (QMGS).1 Early studies of interrater reliability of the test determined that an improvement of >2.5 units is statistically significant, and trials have consistently used improvement of >3.5 units to differentiate between responders and non-responders.2 Despite the consensus, there has never been a comprehensive validation of these cut-offs, including intraobserver (test–retest) reliability studies, to determine the minimal detectable change and therefore whether the commonly used cutpoints are meaningful clinically. There have been several definitions of what should be considered as change when repeating a measurement. All measurement tools are subject Abbreviations: CIDP, chronic inflammatory demyelinating polyneuropathy; ICE, IVIg chronic inflammatory demyelinating polyneuropathy efficacy; IVIg, intravenous immunoglobulin; MCID, minimal clinical important differences; MG, myasthenia gravis; QMGS, quantitative myasthenia gravis score; RNS, repetitive nerve stimulation; SFEMG, single-fiber electromyography; VAS, visual analog scale Key words: MCID; myasthenia gravis; neuromuscular junction; outcomes; QMGS Correspondence to: H.D. Katzberg; e-mail: [email protected] C 2013 Wiley Periodicals, Inc. V
Published online 12 August 2013 in Wiley Online Library (wileyonlinelibrary. com). DOI 10.1002/mus.23988
MCID in MG
to measurement error, so for any change to be considered significant, it should be beyond that error. The minimal detectable change then is the smallest change that goes beyond the error with repeated measures.3 It is calculated by constructing a 90% or 95% confidence interval around the error of measurement with the following formula: 1.96 3 SD 3 冑2(1 2 r), where SD is the standard deviation of the scores and r is the intraclass coefficient (ICC) of reliability. This means that if a score change exceeds that range, we are 95% certain that it was not due to error. However, it is possible to have statistically significant change (beyond error) without it being relating to a clinical benefit. The shift from statistical to clinical significance has prompted the concept of the minimal clinically important difference (MCID), which is the smallest change score in a measure of interest that patients perceive as beneficial.4 This has been applied increasingly in neurological diseases, most recently in a large group of patients with chronic inflammatory demyelinating polyneuropathy (CIDP) who were treated with intravenous immunoglobulin (IVIg).5 There are several methods to calculate the MCID, such as anchor methods. These methods couple a clinically relevant definition of improvement, such as the patient’s perception of improvement, to the change in the measurement used. Distribution-based methods are based on statistical definitions of change and include using 1=2 SD or 1 3 SEM (standard error of measurement). The latter is mostly related to minimal detectable change; this is why multiple approaches are recommended, as the obtained MCID values may vary according to the methods used, but they also may be influenced by the baseline values, the intervention under study, and the time elapsed. The aim of this study was to identify the MCID for several outcome measures in MG by using data from a previously published clinical trial of the effects of IVIg vs. placebo.1 To determine whether results were clinically meaningful in addition to being statistically significant we then applied these cut-offs to the data with respect to: (a) the QMGS; (b) single-fiber electromyography (SFEMG); and (c) repetitive nerve stimulation (RNS). MUSCLE & NERVE
May 2014
661
METHODS
Patients with MG studied at our institution from 2004 to 2005, who participated in the IVIg vs. placebo study, were selected for inclusion in this study.1 The inclusion and exclusion criteria and methodology are described in the original article. In brief, patients 18 years of age with a diagnosis of MG and worsening weakness were enrolled in the study. Patients were excluded if they had respiratory distress requiring intensive care unit admission, a vital capacity 3.5 points (21.11 to 22.01). Using a cut-off of 3, all MCID methods for QMGS demonstrated significantly more responders in the IVIg group compared with placebo (Fig. 1a). The 1 3 SEM and anchor-based cut-offs both resulted in 58.3% vs. 30.7% responders (P 5 0.017) for IVIg vs. placebo. The 1=2 SD-based cut-offs resulted in 54.2% vs. 19.2% responders (P 5 0.018) (Fig. 1b). The effect-size method met MUSCLE & NERVE
May 2014
Table 2. Anchor- and statistical-based MCIDs in patients with MG treated with IVIg vs. placebo. Anchor-based approach
1/2 SD approach
Parameter
Unchanged*
Slightly better*
MCID cut-off
QMGS RNS baseline RNS postexercise RNS 1-min postexercise SFEMG jitter
20.41 6 1.9 28.2 6 15.4 25.1 6 10.0 210.1 6 16.1 5.3 6 25.2
22.08 6 3.5 0.45 6 9.8 0.0 6 6.3 20.1 6 9.7 28.9 6 72.9
21.67 12.70 5.10 10.00 214.20
SD 4.03 1.60 1.60 1.60 3.00
SEM
MCID cut-off
Cronbach alpha
MCID cut-off
22.01 0.80 0.80 0.80 1.50
0.954 0.752 0.720 0.819 0.735
–1.11 0.797 0.846 0.680 1.540
MCID, minimal clinical important differences; QMGS, quantitative myasthenia gravis score; RNS, repetitive nerve stimulation; SFEMG, single-fiber electromyography. *Data expressed as mean 6 standard deviation.
0.5 MCID in the IVIg group (0.519), but not in the placebo group (0.164) (P 5 0.011) (Fig. 1c). The ROC method showed that a cutpoint of 2 had a sensitivity of 60% and a specificity of 71%, and a cutpoint of 3 had a sensitivity of 56% and a specificity of 89% (area under the curve 5 0.73). To correct for baseline QMGS, patients were divided into 3 groups: QMGS 0–9 (n 5 19); QMGS 10–16 (n 5 19); and QMGS >16 (n 5 12). The QMGS MCID based on anchor methods for each of these tertiles was 21.6, 21.8, and 22.75, respectively. When the rate of response in the IVIg and placebo groups was reanalyzed using the cutpoints corrected for baseline values, the values were similar to the anchor and 1 3 SEM cutpoints: the IVIg group had 54% responders compared with 30% in the placebo group. However, this did not reach statistical significance (P 5 0.43), which may have been due to the small proportion of patients who had a baseline QMGS >16 (12 patients). Only anchor-based methods produced a meaningful SFEMG MCID cut-off (214.2 ls). Using this cut-off, a significant difference between IVIg and placebo was not identified (27% vs. 19% responders, P 5 0.73). A significant difference was also not found when the effect-size method was used (0.343 vs. 0.013, P 5 0.41). None of the MCID methods produced a meaningful cut-off for RNS outcomes at baseline and after exercise and thus were not applied. DISCUSSION
In this study we applied MCID cut-offs in patients with MG. A similar methodology has been applied to patients with CIDP using the data from the recent IVIg CIDP efficacy (ICE) study, which evaluated the effects of IVIg vs. placebo.5 Recent trials in patients with rheumatological and other chronic diseases used MCID and statistical methods in reporting outcomes.9–12 In our cohort, anchor- and distribution-based methods were consistent, distributed over a narrow range, and did not exceed the original2 or estabMCID in MG
lished cut-offs used in most clinical trials for QMGS. The consistency of the values obtained demonstrates the strength of the results and supports the recommendation that these values become incorporated into practice. The relevance of knowing the MCID for these outcome measures is that treatment efficacy can be assessed through changes that are relevant to the patients and not just statistically significant. The use of different methods to calculate the MCID is another strength of this study, as we looked not only at values relating to error of measurement but most importantly to patients’ perceptions. In a recent publication we showed that the only predictor of response to immunomodulation with IVIg or plasmapheresis in MG was baseline severity, but we used QMGS cut-offs of >3.5 to define responders.13 In the present study we found different MCID cutpoints depending on baseline severity: 2 for patients with QMGS between 0 and 16 and 3 for QMGS >16. If confirmed in a larger cohort, this correction should be taken into consideration in clinical trials when defining responders. This may also have implications in calculating sample size in MG clinical trials, as a different number of patients may be required to be able to detect clinically relevant treatment differences depending on disease severity at entry. We found an MCID jitter cut-off of 214.2 ls (16% change of the mean), which is slightly higher than previous approximations of clinically significant change in jitter of 10%14 and only slightly above a previously reported estimated jitter measurement error of 12.2%.15 Thus, jitter may be of questionable utility in clinical trials. This cut-off did not show significant differences between placebo and IVIg in our group of patients, which is not surprising given that the study was not powered to detect differences in electrophysiology. MCID methods did not yield clinically meaningful cut-offs for RNS methods and, as such, comparison of effectiveness in the IVIg vs. placebo group was not possible. Although additional MCID methods MUSCLE & NERVE
May 2014
663
FIGURE 1. (a) Change in QMGS in IVIg and placebo at 2 weeks compared with baseline. Each dot represents an individual patient. Dashes represent MCID cut-offs. (b) Percentage of responders to IVIg (squares) and placebo (triangles) using MCID cut-offs. (c) Effect size in patients receiving IVIg (squares) and placebo (triangles).
and larger data sets could be utilized in future studies to further evaluate RNS as a valid outcome measure, it is still not likely to yield useful results. This is also not surprising given that RNS is not a sensitive test in MG. There are a number of limitations to MCID analysis in our group of patients. MCID has been criticized due to a lack of agreement on which method is best to establish cut-offs, and there are no universally accepted methods for calculating sample size.16 Also, given the small number of patients in our study, it is difficult to confirm that the MCID cut-offs are reliable, particularly with regard to tertile analysis and the predictive ability of the cutpoints obtained by the ROC curve. These findings should be confirmed in a larger data set. A combination of anchors, including a quality-oflife scale should also be studied further. There were very few patients who reported worsening (n 5 8). Thus, the MCID for worsening could not be studied, as some studies showed that the significant values for worsening and improving tend to differ. Finally, we recognize that a limitation of the QMGS, as with any ordinal measure, is that the change scores in different patients are not necessarily comparable and that, in the future, transformation of the ordinal measures to interval ones may be studied to establish MCID cut-offs in a more appropriate way.17,18 In spite of these limitations, we recommend that the MCID be studied further for the QMGS in addition to other outcome measures in myasthenia, such as the MG composite,19 to have a better 664
MCID in MG
understanding of what constitutes clinically meaningful change. Although the cut-offs identified in this study will need to be confirmed in a larger data set, using a QMGS cut-off of 2 for patients with a baseline QMGS of 16 should be considered in future research studies. REFERENCES 1. Zinman L, Ng E, Bril V. IV immunoglobulin in patients with myasthenia gravis: a randomized controlled trial. Neurology 2007;68:837– 841. 2. Barohn RJ, McIntire D, Herbelin L, Wolfe GI, Nations S, Bryan WW. Reliability testing of the quantitative myasthenia gravis score. Ann NY Acad Sci 1998;841:769–772. 3. De Vet HCW, Terwee CB, Mokkink LB, Knol DL. Measurement in medicine. Cambridge, UK: Cambridge University Press; 2011. 4. Jaeschke R, Singer J, Guyatt GH. Measurement of health status: ascertaining the minimal clinically important difference. Control Clin Trials 1989;10:407–415. 5. Merkies IS, van Nes SI, Hanna K, Hughes RA, Deng C. Confirming the efficacy of intravenous immunoglobulin in CIDP through minimum clinically important differences: shifting from statistical significance to clinical relevance. J Neurol Neurosurg Psychiatry 2010;81: 1194–1199. 6. Juniper EF, Guyatt GH, Willan A, Griffith LE. Determining a minimal important change in a disease-specific quality of life questionnaire. J Clin Epidemiol 1994;47:81–87. 7. Copay AG, Glassman SD, Subach BR, Berven S, Schuler TC, Carreon LY. Minimum clinically important difference in lumbar spine surgery patients: a choice of methods using the Oswestry Disability Index, Medical Outcomes Study Questionnaire Short Form 36, and pain scales. Spine J 2008;8:968–974. 8. King MT. A point of minimal important difference (MID): a critique of terminology and methods. Expert Rev Pharmacoecon Outcomes Res 2011;11:171–184. 9. Wyrwich, KW, Tierney WM, Babu AN, Kroenke K, Wolinsky FD. A comparison of clinically important differences in health-related quality of life for patients with chronic lung disease, asthma, or heart disease. Health Serv Res 2005;40:577–591. 10. Stauffer ME, Taylor SD, Watson DJ, Peloso PM, Morrison A. Definition of nonresponse to analgesic treatment of arthritic pain: an analytical literature review of the smallest detectable difference, the minimal detectable change, and the minimal clinically important difference on the pain visual analog scale. Int J Inflam 2011;2011:231926.
MUSCLE & NERVE
May 2014
11. Tubach F, Ravaud P, Baron G, Falissard B, Logeart I, Bellamy N, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis 2005;64:29–33. 12. Rigby W, Ferraccioli G, Greenwald M, Zazueta-Montiel B, Fleischmann R, Wassenberg S, et al. Effect of rituximab on physical function and quality of life in patients with rheumatoid arthritis previously untreated with methotrexate. Arthritis Care Res 2011;63:711–720. 13. Katzberg HD, Barnett C, Bril V. Predictors of response to immunomodulation in patients with myasthenia gravis. Muscle Nerve 2012; 45:648–652. 14. Sanders DB, Stalberg EV. AANEM minimonograph #25: Single-fiber electromyography. Muscle Nerve 1996;19:1069–1083.
MCID in MG
15. Buchman AS, Garratt M. Determining neuromuscular jitter using a monopolar electrode. Muscle Nerve 1992;15:615–619. 16. Walters S. Comparison of the minimally important difference for two health state utility measures: EQ-5D and SF-6D. Qual Life Res 2005; 14:1523–1532. 17. Rasch G. Probabilistic models for some intelligence and attainment tests. Chicago: University of Chicago Press; 1980. 18. Vanhoutte EK, Faber CG, van Nes SI, Jacobs BC, van Doorn PA, van Koningsveld R, et al. Modifying the Medical Research Council grading system through Rasch analyses. Brain 2012;135:1639–1649. 19. Burns TM, Conaway MR, Sanders DB; MG Composite and MGQOL15 Study Group. The MG composite: a valid and reliable outcome measure for myasthenia gravis. Neurology 2010;74:1434–1440.
MUSCLE & NERVE
May 2014
665
