Abstract
Objective: To determine whether the MS Functional Composite (MSFC) can predict future disease progression in patients with relapsing remitting MS (RR-MS).
Background: The MSFC was recommended by the Clinical Outcomes Assessment Task Force of the National MS Society as a new clinical outcome measure for clinical trials. The MSFC, which contains a test of walking speed, arm dexterity, and cognitive function, is expressed as a single score on a continuous scale. It was thought to offer improved reliability and responsiveness compared with traditional clinical MS outcome measures. The predictive value of MSFC scores in RR-MS has not been determined.
Methods: The authors conducted a follow-up study of patients with RR-MS who participated in a phase III study of interferon β-1a (AVONEX) to determine the predictive value of MSFC scores. MSFC scores were constructed from data obtained during the phase III trial. Patients were evaluated by neurologic and MRI examinations after an average interval of 8.1 years from the start of the clinical trial. The relationships between MSFC scores during the clinical trial and follow-up status were determined.
Results: MSFC scores from the phase III clinical trial strongly predicted clinical and MRI status at the follow-up visit. Baseline MSFC scores, and change in MSFC score over 2 years correlated with both disability status and the severity of whole brain atrophy at follow-up. There were also significant correlations between MSFC scores during the clinical trial and patient-reported quality of life at follow-up. The correlation with whole brain atrophy at follow-up was stronger for baseline MSFC than for baseline EDSS.
Conclusion: MSFC scores in patients with RR-MS predict the level of disability and extent of brain atrophy 6 to 8 years later. MSFC scores may prove useful to assign prognosis, monitor patients during early stages of MS, and to assess treatment effects.
In 1994, the National MS Society (NMSS) convened an international workshop to review the status of outcome measures for MS clinical trials. Proceedings from the meeting indicated that available clinical outcome scales were not satisfactory.1 As a result, the NMSS appointed a task force to recommend improved clinical outcome measures. The Clinical Outcomes Assessment Task Force consisted of neurologists, neuropsychologists, biostatisticians, and industry representatives from five countries. The task force defined the attributes of an idealized clinical outcome measure2 and conducted post hoc data analysis by using placebo arms from completed MS clinical trials, and natural history data sets.3 Based on the desired attributes and results from the data analysis, the Task Force recommended an approach to clinical outcomes assessment as well as a specific clinical outcome measure, the MS Functional Composite (MSFC).4
The MSFC consists of three timed tests of neurologic function, which are combined into a single score that is expressed along a continuous scale. The tests include a timed 25-foot walk, the nine-hole peg test,5 and the 3-second version of the Paced Serial Addition Test (PASAT).6 Each of the test results is standardized by using a reference population, and resulting scores are averaged to provide a single score that indicates performance relative to the reference population. A score of +1 indicates that on average a given subject scored 1 SD better than the reference population, whereas a score of −1 means on average a subject scored 1 SD worse than the reference population. A performance and scoring manual was published to promote consistent use of the MSFC.7 The MSFC is administered by a trained technician and requires approximately 15 minutes of testing time.
MSFC scores were found to correlate with EDSS,3,8⇓ MRI lesion load,9 and self-reported quality of life.10 Furthermore, MSFC change correlated with change on the Expanded Disability Status Scale (EDSS).3 The Clinical Outcomes Task Force reported that MSFC change preceded EDSS change during a 2-year follow-up period,3 and recommended that the predictive value for long-term disability be determined in prospective studies.4 We report the predictive value of MSFC scores derived from a phase III clinical trial of interferon β-1a (AVONEX, Biogen, Inc., Cambridge, MA).
Methods.
Study objective.
The primary objective was to test the hypothesis that MSFC scores, obtained during a phase III clinical trial of IFNβ-1a initiated in 1990, were correlated with clinically relevant disability, whole brain atrophy,11 and self-reported quality of life determined at follow-up in 1999. An exploratory study objective was to investigate use of contrast sensitivity12 as a visual measure for possible inclusion in the MSFC. Because treatment was controlled only for the initial 2 years after entering the phase III clinical trial, and patients in both treatment arms were untreated for a period while awaiting the outcome analyses, long-term effects of treatment with IFNβ-1a could not be determined from this follow-up study. Therefore, long-term treatment effects were not a predetermined objective of the study, and the results reported herein represent the combined cohort followed without reference to duration of therapy. This report deals with the predictive value of the MSFC. Subsequent reports will address the use of contrast sensitivity, and potential methods for assessing treatment effects during open-label uncontrolled studies, such as this follow-up study.
Patients.
To be eligible for the study, patients with MS had to be enrolled in the original IFNβ-1a phase III study, and based on their randomization date have follow-up data for at least 2 years during the phase III trial. The latter requirement was used so that each patient had baseline, year 1, and year 2 study visits. As planned in advance, follow-up time in the original phase III study was variable.13 Although 301 patients were enrolled in the original study, 172 were enrolled early enough to have follow-up for at least 2 years. These 172 patients at four clinical sites were eligible for this follow-up study.
Study protocol.
The study was reviewed and approved by institutional review boards at each site, and study participants signed informed consent documents. The study consisted of a single patient evaluation done during the summer of 1999, comprising a clinic visit, a self-report for patients unable or unwilling to return to clinic for an evaluation, or an interview with families of those patients who were eligible for the study but had died. At the clinic visit, an examining neurologist classified the patient as having relapsing-remitting MS (RR-MS) or secondary progressive MS (SP-MS), recorded the type and dates of disease-modifying drug therapy since the end of the phase III clinical trial, and recorded EDSS and functional system scores at the time of this follow-up. A trained technician performed the MSFC and administered the contrast sensitivity test.14 The patient completed the Sickness Impact Profile (SIP),15,16⇓ a validated measure of disease-related quality of life. A cranial MRI scan was done, using acquisition parameters as similar as possible to the original clinical trial.17 Brain parenchymal fraction (BPF) was calculated as described.11,12⇓ The components to compute the MSFC during the phase III clinical trial were available for comparison. For patients unable to return to the clinic, disease-modifying drug therapy was recorded, the site neurologist assigned disease category, and the patient completed a self-report EDSS18 and the SIP. For patients who died, the examining physician interviewed relatives and assigned an EDSS score based on the patient’s condition before the terminal event.
Analysis approach.
The predictive value was determined for MSFC scores at entry into the phase III study, and for change in MSFC scores between baseline and week 104. Poor outcome at the follow-up was defined in advance in four ways: 1) severe disability based on EDSS score, defined as EDSS ≥ 6.0; 2) patients categorized as SP-MS; 3) the presence of severe brain atrophy, defined as BPF < 0.80 (this definition was selected because BPF of 0.80 corresponded to approximately the lowest 5% of the baseline scores from the phase III trial); and 4) patient self-report above the median physical SIP score at follow-up. All available data from all patients were used for the analyses.
Statistical methods.
Spearman rank correlation coefficients were computed to assess associations between the outcome measures. Multivariate logistic regression models were used to determine the predictive value of MSFC or MSFC component tests for the predefined outcomes. In the models, MSFC scores (or individual MSFC component scores) at baseline and MSFC change between baseline and 2 years were independent (predictor) variables, and EDSS category, BPF category, disease category, and physical SIP score category at the follow-up visit were used as the dependent variables. Odds ratios (OR) with their respective 95% confidence limits were calculated to determine the predictive value of the MSFC. All statistical analyses were performed by using SAS 6.12 (SAS Institute, Inc., Cary, NC) for PC.
Results.
Case ascertainment.
Information was collected on 160 (93%) of the 172 eligible patients. One hundred thirty-seven patients were examined at the clinics, 16 patients provided self-reports, and families of the seven patients who had died were interviewed. Twelve patients could not be located or refused to participate. Of 137 patients who were reexamined at a clinic visit, 134 had follow-up MRI scans. At the time of follow-up, the average time from randomization into the phase III study was 8.1 years, and the average disease duration was 14.3 years.
Characteristics of study participants.
Table 1 shows baseline characteristics of the 160 patients who participated in this study. Patients in this study were very similar to the 301 patients enrolled in the original phase III study.19 Age (mean ± 1 SD) was 36.1 ± 6.8 years; disease duration, 6.3 ± 5.5 years; EDSS, 2.3 ± 0.8; and 77% were women. Mean BPF was 0.83 (± 0.16), and mean MSFC was 0.00 ± 0.72. At the follow-up visit, mean EDSS was 4.4 ± 2.3, BPF was 0.80 ± 0.028, and MSFC was −1.1 ± 2.2. At entry into the phase III study, EDSS was 1.0 to 3.5. At the follow-up examination, 35% of patients were EDSS ≥ 6. At study entry, BPF scores were normally distributed with a mean of 0.83; fewer than 5% of patients had BPF below 0.80. At the follow-up evaluation, 42% of patients had BPF < 0.80. At entry into the study, all of the patients were classified as RR-MS20; at the follow-up examination, 35% of the patients were classified as SP-MS.
Characteristics of patients at baseline and follow-up
Relationships between study variables.
Table 2 shows correlation coefficients testing the relationship between EDSS, MSFC, and BPF at baseline, year 2, and at the follow-up visit. EDSS at baseline was moderately correlated with EDSS at year 2 and at follow-up. MSFC at baseline was somewhat more strongly correlated with MSFC at year 2 and at follow-up than EDSS. The strongest correlations were observed between baseline BPF and year 2 BPF (r = 0.84) and BPF at follow-up (r = 0.71). Significant cross-sectional correlations occurred between EDSS, MSFC, and BPF. For example, cross-sectional rank correlations between MSFC and EDSS ranged from −0.42 and −0.85 over time; between MSFC and BPF from 0.42 and 0.50; and between EDSS and BPF from −0.29 to −0.42. MSFC at study baseline was significantly correlated with EDSS at year 2 and at follow-up, and with BPF at year 2 and at follow-up. Also shown in table 2 are correlation coefficients for concurrent change in EDSS, MSFC, and BPF during the 2-year clinical trial. There was a moderate correlation between EDSS change and MSFC change, and a weaker but still significant correlation between EDSS change or MSFC change and BPF change.
Correlations between study variables
Predictive validity of MSFC for later EDSS status.
Figure 1A shows the relationship between MSFC and the percentage of patients with EDSS ≥ 6.0 at follow-up. Overall, 35% of the patients were found to be at EDSS ≥ 6.0 at follow-up. The figure shows that patients in the most favorable quartile of MSFC at baseline, and the most favorable category of MSFC change between baseline and 2 years had a considerably lower likelihood of being in the EDSS ≥ 6.0 category at follow-up. In contrast, patients in the worst quartile of MSFC at baseline, and MSFC change between baseline and 2 years had a much higher likelihood of the unfavorable EDSS category at follow-up.
Figure 1. (A) Percentage of patients at Expanded Disability Status Scale score ≥6.0 at the follow-up examination, according to the baseline MS Functional Composite (MSFC) score from the phase III clinical trial, and the amount of MSFC change during the 2 years of the clinical trial. (B) Percentage of patients classified as severe brain atrophy, defined as brain parenchymal fraction < 0.80 at the follow-up examination, according to the baseline MSFC score from the phase III clinical trial, and the amount of MSFC change during the 2 years of the clinical trial. For each panel, the most favorable quartile is shown to the left, and the least favorable to the right. Quartile 1: MSFC baseline ≥ 0.51, MSFC change ≥ 0.35. Quartile 2: MSFC baseline ≥ 0.06 and < 0.51; MSFC change ≥ 0.08 and < 0.35; Quartile 3: MSFC baseline ≥ −0.32 and < 0.06, MSFC change ≥ −0.34 and < 0.08; Quartile 4: MSFC baseline < −0.31; MSFC change < −0.34. White bars = MSFC year 0; hatched bars = MSFC change, year 0–2.
Logistic regression analyses showed that the relationships between baseline MSFC and MSFC change between baseline and year 2 and EDSS status were significant when both were included in the same model. The OR (95% CI) for baseline MSFC predicting outcome EDSS status was 2.72 (1.42 to 5.21), p = 0.002, after adjusting for the effect of MSFC change, and the OR for MSFC change predicting outcome EDSS status after adjusting for the effect of baseline MSFC was 3.05 (1.61 to 5.78), p < 0.001.
Predictive validity of MSFC for later brain atrophy.
Figure 1B shows the relationship between MSFC and the proportion of patients with severe brain atrophy at follow-up. Patients in the most favorable quartile of baseline MSFC scores and MSFC change had the lowest likelihood of having BPF < 0.80 at follow-up, whereas patients with the least favorable quartile of baseline MSFC and MSFC change had the highest likelihood.
Logistic regression analyses showed that the relationships between baseline MSFC and MSFC change between baseline and year 2 and severe brain atrophy were significant when both were included in the same model. The OR (95% CI) for baseline MSFC predicting severe brain atrophy at the follow-up was 4.37 (1.96 to 9.71), p < 0.001, after adjusting for MSFC change. The OR of MSFC change predicting severe brain atrophy was 3.10 (1.45 to 6.62), p = 0.004, after adjusting for baseline MSFC.
Predictive validity of MSFC for later SP-MS disease category.
Figure 2A shows the relationship between MSFC and the proportion of patients classified as SP-MS at follow-up. Patients in the most favorable quartile of baseline MSFC scores and MSFC change had the lowest likelihood of being classified as SP-MS at the follow-up visit, whereas patients with the least favorable quartile of baseline MSFC and MSFC change had the highest likelihood.
Figure 2. (A) Percentage of patients classified as secondary progressive MS at the follow-up examination, according to the baseline MS Functional Composite (MSFC) score from the phase III clinical trial, and the amount of MSFC change during the 2 years of the clinical trial. (B) Percentage of patients above the median of the physical Sickness Impact Profile score at the follow-up examination, according to the baseline MSFC score from the phase III clinical trial, and the amount of MSFC change during the 2 years of the clinical trial. For each panel, the most favorable quartile is shown to the left, and the least favorable to the right. Quartile 1: MSFC baseline ≥ 0.51, MSFC change ≥ 0.35. Quartile 2: MSFC baseline ≥ 0.06 and < 0.51; MSFC change ≥ 0.08 and < 0.35; Quartile 3: MSFC baseline ≥ −0.32 and < 0.06, MSFC change ≥ −0.34 and < 0.08; Quartile 4: MSFC baseline < −0.31; MSFC change < −0.34. White bars = MSFC year 0; hatched bars = MSFC change, year 0 to 2.
Logistic regression analyses showed that the relationships between baseline MSFC and MSFC change between baseline and year 2 and disease category were significant when both were included in the same model. The OR (95% CI) for baseline MSFC predicting secondary progressive MS status at the follow-up was 2.20 (1.13 to 4.27), p = 0.02, after adjusting for the effect of MSFC change. The OR for MSFC change predicting secondary progressive MS was 3.86 (1.89 to 7.94), p < 0.001, after adjusting for the effect of baseline MSFC.
Predictive validity of MSFC for self-reported physical functioning.
Figure 2B shows the predictive relationship between MSFC and the self-reported quality of life score based on the physical subscore of the SIP at the follow-up examination. Outcome was defined as being above (bad outcome) or below (good outcome) the median physical SIP score at follow-up. Patients in the most favorable quartile of baseline MSFC scores and MSFC change had the lowest likelihood of being above the median physical SIP score, whereas patients with the least favorable quartile of baseline MSFC and MSFC change had the highest likelihood. The OR (95% CI) for baseline MSFC predicting the unfavorable physical SIP status at the follow-up was 1.64 (0.94 to 2.84), p = 0.08, after adjusting for MSFC change. The OR for MSFC change predicting the unfavorable physical SIP status at follow-up was 1.64 (0.95 to 2.84), p = 0.05, after adjusting for baseline MSFC.
Relationship between baseline MSFC or EDSS and brain atrophy at follow-up.
Figure 3A shows the relationship between baseline EDSS and the BPF at follow-up, whereas figure 3B shows baseline MSFC as it relates to follow-up BPF. Baseline MSFC accounted for 31% of the variance in subsequent brain atrophy, whereas baseline EDSS accounted for only 6% of the variance in brain atrophy.
Figure 3. (A) Relationship between baseline Expanded Disability Status Scale score and BPF at the follow-up (FU) examination. (B) Relationship between baseline MS Functional Composite (MSFC) and BPF at the follow-up examination in the same patients.
Predictive validity of MSFC component measures for disability status and severe brain atrophy.
Logistic regression models were constructed to determine the predictive value of the MSFC component measures and status at follow-up. The baseline scores for each MSFC component measure, as well as change on each component score from baseline to year 2, were tested as independent variables, and EDSS or BPF status at follow-up were used as dependent variables. Table 3 shows the model χ2 values. The overall model χ2 shows that the baseline and 2-year change in each of the individual MSFC component measures were significantly correlated with outcome status at the follow-up visit. For EDSS outcome, timed walk was most strongly correlated, whereas the predictive relationship was somewhat weaker, though still statistically significant, with the nine-hole peg test and PASAT. For the brain atrophy outcome, the nine-hole peg test and PASAT were most strongly correlated with brain atrophy outcome, whereas the timed walk was a slightly weaker predictor. Importantly, for both the EDSS and brain atrophy outcomes, performance during the phase III study on each component of the MSFC correlated significantly with the outcome at follow-up.
Relationship between MSFC component measures and outcome
Discussion.
In this study, MSFC scores from patients with RR-MS participating in a clinical trial of IFNβ-1a were shown to be strongly predictive of clinical and MRI status in the same patients an average of 8.1 years later. Both baseline MSFC scores and the change in MSFC scores during 2 years were predictors of clinical and MRI status at the follow-up examination. Less favorable performance on the single MSFC score done at study entry and more deterioration in MSFC scores during the original phase III trial were found to be significant risk factors for severe physical disability, SP-MS status, severe brain atrophy, and lower self-reported quality of life 6 to 8 years later. These findings suggest that the MSFC can be used during the RR-MS stage of disease to identify patients at high risk for progressive MS. In addition to the composite measure, individual component measures contained within the MSFC—walking, arm function, and cognitive function—also correlated with both EDSS and MRI outcome, suggesting that each component of the MSFC had predictive value and contributed to the value of the composite measure.
The MSFC was proposed as a new outcome measure for MS clinical trials because it was thought to have advantages over existing neurologic rating scales traditionally used as clinical outcome measures for MS trials. Existing neurologic rating scales are relatively imprecise, insensitive, and weighted toward walking, to the exclusion of other disease manifestations. Because the MSFC consists of simple quantitative tests of neurologic function, this measure was predicted to be highly reproducible. This has been shown to be correct in a prospective study.21 Improved measurement error would be expected to improve the power of a clinical trial to detect biological changes at any sample size. It was hoped that the MSFC would be more sensitive than existing clinical measures, not only because of reduced measurement error, but also because the MSFC includes other dimensions of MS and is expressed along a continuous scale. The NMSS Task Force obtained preliminary evidence for this. In an analysis of data contained within existing clinical trials data sets, MSFC change during a 1-year period was observed in a proportion of patients who did not change on EDSS during the same period. Patients who showed MSFC worsening were found to be at significantly higher risk of EDSS worsening in the next year of observation.3 This suggested that MSFC change occurred earlier than EDSS change and supported the Task Force recommendation to use MSFC in future MS clinical trials.4 Finally, a significant advantage of the MSFC was thought to be low cost. A trained technician can perform the test with an individual patient in less than 15 minutes, and necessary equipment is minimal.
The current study provides important new information on the predictive validity of MSFC in patients with RR-MS. The patients in this study had an average disease duration of 6.1 years and average EDSS of 2.3 at entry into the clinical trial. MSFC scores in these patients were predictive of outcome, defined in a variety of ways, an average of 8.1 years later, when 35% of the patients were EDSS ≥ 6.0, 35% were classified as SP-MS, and 42% had BPF < 0.80. This suggests that MSFC will be useful in assigning prognoses to individual patients with RR-MS early in the disease course. This is important because current clinical prognostic markers in the RR-MS disease stage are weak, and MRI prognostic markers are largely undeveloped.
Second, MSFC may prove to be a meaningful way to monitor individual patients with RR-MS. MSFC change was both a significant predictor of clinical outcome and also provided independent information beyond the baseline score. Most important, the study results suggest that favorable effects of treatment on MSFC may translate into clinically meaningful therapeutic effects, increasing the evidence in support of the MSFC as an endpoint in clinical trials.
A number of important questions remain for future studies. At the time the IFNβ-1a study was conducted, the MSFC had not been proposed, and there were no procedural measures taken to optimize the use of the individual MSFC components. In particular, significant practice effects have been documented for the MSFC, particularly the PASAT.21 Patients in the IFNβ-1a trial were simply given instructions, and their baseline MSFC score was recorded. The highly significant correlations observed in this study were found despite this limitation. With more consistent use of MSFC, and particularly with the use of procedures taken to attenuate the learning effect, correlations might be stronger than were observed in this study and the estimated changes that occur over time more accurate. Second, it was pointed out by the NMSS Task Force that a visual measure should be developed for consideration of inclusion in the MSFC, because some patients experience progressive visual impairment during the course of MS. Visual acuity was assessed by the Task Force. However, there were no other candidate visual measures included in the original IFNβ-1a clinical trial. Third, we have not defined a “clinically significant” amount of MSFC change through this study. The results show highly significant relative risk of poor outcomes based on poor performance on MSFC, but extrapolating this study result to individual patients is considered premature.
Finally, because data in this report consisted of both treatment arms combined, the relationships between MSFC and outcome were evident despite variable treatment with disease-modifying drugs. Predictive validity of baseline MSFC, obtained before disease-modifying drug therapy was started, indicates that MSFC predictive value is independent of treatment. This is an important characteristic of a surrogate outcome, which should show treatment effects but not be a surrogate for treatment itself. Analysis of long-term effects of disease-modifying drugs is an important, but complex undertaking, because untreated concurrent control groups are not feasible. Comparison of long-term outcomes in cohorts such as this with untreated historical control groups might provide insight into the long-term effects of treatment with disease-modifying drugs, but with the limitations inherent in historical controls.
Acknowledgments
The original clinical trial for this study was supported by NINDS RO1 26321, NMSS RG3099, and Biogen, Inc., Cambridge, MA. The long-term follow up study was supported by Biogen, Inc., and by PO1 NS38667.
Acknowledgment
The MSFC was developed by the Clinical Outcomes Assessment Task Force of the National Multiple Sclerosis Society. A manual describing the MSFC testing and scoring procedures is available through the NMSS.
Footnotes
-
The results were presented at the annual meeting of the American Academy of Neurology; San Diego, CA; May 3, 2000.
-
N.A.S. is an employee of Biogen, Inc. None of the other authors has a personal financial investment, ownership, equity, or other financial holdings with Biogen, Inc.
- Received September 12, 2000.
- Accepted January 30, 2001.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Hastening the Diagnosis of Amyotrophic Lateral Sclerosis
Dr. Brian Callaghan and Dr. Kellen Quigg