- Email: [email protected]
How does brain MRI lesion volume change on serial scans in patients with multiple sclerosis?
Magnetic Resonance Imaging, Vol. 16, No. 10, pp. 1181–1183, 1998 © 1998 Elsevier Science Inc. All rights reserved. Printed in the USA. 0730-725X/98 $19.00 1 .00
PII S0730-725X(98)00083-6
● Original Contribution
HOW DOES BRAIN MRI LESION VOLUME CHANGE ON SERIAL SCANS IN PATIENTS WITH MULTIPLE SCLEROSIS? M. FILIPPI,* M.P. SORMANI,*† M. ROVARIS,*
AND
G. COMI*
*MS Biosignal Analysis Center, Department of Neurology, Scientific Institute Ospedale San Raffaele, University of Milan, Milan; and †Unit of Clinical Epidemiology and Trials, National Institute for Cancer Research, Genoa, Italy Although lesion load changes on conventional T2-weighted brain magnetic resonance imaging (MRI) scans from patients with multiple sclerosis (MS) are used to monitor the effect of treatment, there is no clear definition of how lesion load changes over years according to the lesion load present at a baseline evaluation. In the present study, we evaluated the relationship between lesion load changes over time and lesion load at a baseline evaluation in a group of untreated patients with MS. We scanned nineteen patients on two separate occasions with a mean interval 16.4 months between the two examinations. In each scanning session, a scan with forty contiguous 3-mm-thick axial slices was acquired. We assessed MRI lesion loads using a semi-automated local thresholding technique. Both a linear (p < 0.0001) and a quadratic component (p 5 0.0008) of the baseline volume were significant in describing the follow-up volume. The equation to model this finding was as follows: Vf 5 b0 Vb 1 b1 (Vb)2, where Vf is the lesion volume at follow-up, Vb is the lesion volume at baseline, b0 5 0.834 (SE 5 0.098), and b1 5 0.014 (SE 5 0.003) (mL)21. Our data indicate that lesion volume changes detectable on serial brain MRI studies from patients with MS are dependent on the extent of lesion burden present on the baseline MRI scans. This finding has to be considered when planning phase III trials. © 1998 Elsevier Science Inc. Keywords: Multiple sclerosis; Brain; Lesion volume measurements; Clinical trials.
INTRODUCTION
tween lesion load changes over time and lesion load at a baseline evaluation in a group of untreated patients with MS.
Changes of lesion loads on conventional T2-weighted magnetic resonance imaging (MRI) scans of the brain from patients with multiple sclerosis (MS) are used in virtually all long-term natural history studies and treatment trials as a measure of disease evolution.1 From the placebo arms of previous clinical trials, it has been estimated that the expected change of lesion load is around 5–10% per annum.2 At present, however, no attempt has been made to provide a model to predict the changes of lesion load over time on the basis of the lesion load detected at a baseline evaluation. This is clearly important when planning clinical trials, because it is conceivable that the rate of lesion load accumulation might be different according to the amount of lesions present on the entry scan. In the present study, we assessed which model best explains the relationship be-
MATERIALS AND METHODS Patients Nineteen consecutive patients (thirteen women, five men) with clinically definite MS3 entered the study. Thirteen had a relapsing-remitting and five a secondaryprogressive MS disease course.4 Their mean age was 33.4 years (SD 5 7.9 years), median duration of the disease was 6.0 years (range: 2–15 years), and median expanded disability status scale (EDSS) was 2.0 (range: 1.0 –7.5). To be included, patients must have had neither clinical relapses nor steroid treatment during the 3 months preceding the study. In addition, they were not treated with immunosuppressive or immunomodulating drugs for a period starting 6 months prior to inclusion in
RECEIVED 11/1/97; ACCEPTED 2/2/98. Address correspondence to Dr. Massimo Filippi, MS Biosignal Analysis Center, Department of Neurology, Scientific
Institute Ospedale San Raffaele, via Olgettina 60 –20132, Milan, Italy. E-mail address: [email protected] 1181
Magnetic Resonance Imaging ● Volume 16, Number 10, 1998
1182
the study and lasting until the end of the study. In case of relapses during the follow-up period, only steroid treatment was allowed (usually methylprednisolone i.v. 1 g/day for 5 days). The exit MRI scans were always obtained at least 3 months after the last relapse or steroid treatment. Written informed consent was obtained from all patients before inclusion in the study. MRI Brain MRI scans were performed using a 1.5 Tesla machine (Siemens Magnetom SP63). The pulse sequence used (SE 2000/50) provided a moderate T2 weighting and gave good definition of the MS lesions, with some suppression of the cerebrospinal fluid signal. Brain MRI scans consisted of forty contiguous interleaved 3-mmthick axial slices, with a field of view of 22 cm and a 256 3 256 matrix size. For all patients, MRI scans were performed both at entry to the study and follow-up, with a mean interval of 16.4 months (range: 13–20 months) between the two scans. For both the entry and the follow-up scans, the MRI slices were carefully positioned according to published guidelines.5 MRI Quantification A single observer (MR), unaware of the scan timepoint, identified and marked the lesions on hard copies. Then lesion volumes were measured on two different occasions (separated by one month), using the hard copies as a reference, by three independent technicians using a semi-automated local thresholding technique for lesion segmentation.1,6 Statistical Analysis The mean lesion volumes obtained for the baseline and the follow-up scans from each patient by the three technicians on the two occasions entered the analysis. The comparison between lesion load at baseline and follow-up was performed using the Wilcoxon signed rank test. To evaluate the relationship between the baseline lesion volume and the changes over time, data were fitted using a multiple linear regression model (backward selection): the follow-up volume (Vf) was treated as the dependent variable; the covariates entering the model were a linear (Vb) and a quadratic (Vb2) component of the baseline volume; and a time variable (Dt) correcting for the deviance of each subject from the mean follow-up time (Dt 5 t 2 16 months), according to the following formula: V f 5 b0 z Vb 1 b1zVb 1 b2 z Dt. 2
(1)
RESULTS The mean (6SD) lesion volumes obtained from the two measurements performed by the three technicians
Fig. 1. Correlation between baseline and follow-up volumes from nineteen patients with MS.
was 16.8 (611.1) mL for entry and 19.4 (615.4) mL for exit scans (p 5 0.036). The mean volume ratio (i.e., lesion volume at follow-up divided by lesion volume at baseline) was equal to 1.05 (SD 5 0.22). The percentage increase in lesion volume measured at follow-up was not constant as the baseline volume increased (Fig. 1). When data were fitted by a multiple regression model, the time-correcting factor was not significant and quit the model. Both a linear (p , 0.0001) and a quadratic component (p 5 0.0008) of the baseline volume were significant in describing the follow-up volume. The final equation to model this finding was as follows (Fig. 2): V f 5 b0 Vb 1 b1 (Vb)2,
(2)
where Vf is the lesion volume at follow-up and Vb is the lesion volume at baseline; the two parameters were estimated as b0 5 0.834 (SE 5 0.098) and b1 5 0.014 (SE 5 0.003) (mL)21. This finding indicates that a linear relation describes the dependence of the ratio of lesion volume on lesion volume at baseline: V r 5 0.834 1 0.014 Vb,
(3)
where Vr is the ratio of lesion volumes. DISCUSSION The main result of the present study is the demonstration that lesion volume changes detectable on serialbrain MRI studies from patients with MS are dependent on the extent of lesion burden present on the entry MRI scans. The final model which fits with our data indicates
Brain MRI lesion volume changes in MS ● M. FILIPPI
Fig. 2. Scatterplot of the volume ratios versus lesion volumes at baseline from nineteen patients with MS. The dotted line indicates no changes in lesion volumes between baseline and follow-up.
that if the baseline lesion volume is around 10 mL, no significant changes of the lesion volume are to be expected in the next 1–2 years, whereas if the baseline lesion load is around 30 – 40 mL, an average increase of about 20 –35% has to be expected in the same period of time. The first consequence of this observation is that lesion load accumulation is different in different phases of disease evolution. Our data suggest that in patients with low lesion load there might be a balance between mechanisms leading to lesion load increase (i.e., new lesion formation and enlargement of pre-existing lesions) and mechanisms leading to repair. This balance seems to be in patients with higher lesion loads. Whether the geometric increase of lesion load is due to a failure of the reparative mechanisms or to an increase of the pathological damage needs to be elucidated. Our sample size was constituted only by patients with low to medium lesion loads. Possibly in patients with higher lesion loads this tendency to a continuous progressive increase in lesion load would be lost. Once there is diffuse damage of the white matter, it is difficult to measure any other pathological change using conventional T2-weighted scans. The presence of continuous activity and recurrent damage within the diseased white matter is perhaps better monitored by the use of enhanced MRI or newer non-conventional MR techniques, such as magnetization transfer imaging7 or magnetic resonance spectroscopy.8 Because the amount of lesion load changes which can be expected to occur in a period of time similar to the
ET AL.
1183
follow-up periods of clinical trials is dependent on the lesion volumes present on the baseline scans, our results have some implications also for the use of MRI lesion load in phase III trials. First, patients should be randomized to placebo or treatment according to their lesion load at entry, because it is conceivable that patients with low lesion loads will tend to have a stable lesion load regardless of being actively treated or not. Second, the duration of follow-up when patients with low lesion loads are studied should be longer than the average period of 2–3 years generally used in trials if any significant change of their lesion load has to be detected. In conclusion, our data indicate that the total lesion load seen on a baseline scan of patients with MS might be a good predictor of lesion load changes over the next few years. Further studies are now needed to confirm these data and to evaluate the evolution of lesion load in patients with higher lesion loads on a baseline scan. Acknowledgements—Dr. Maria Pia Sormani is supported by a grant from TEVA Italy.
REFERENCES 1. Filippi, M.; Horsfield, M.A.; Tofts, P.S.; Barkhof, F.; Thompson, A.J.; Miller, D.H. Quantitative assessment of MRI lesion load in monitoring the evolution of multiple sclerosis. Brain 118:1601–1612; 1995. 2. Paty, D.W.; Li, D.K.B.; Oger, J.J.F.; et al. Magnetic resonance imaging in the evaluation of clinical trials in multiple sclerosis. Ann. Neurol. 36:S95–S96; 1994. 3. Kurtzke, J.F. Rating neurological impairment in multiple sclerosis: An expanded disability status scale (EDSS). Neurology 33:1444 –1452; 1983. 4. Lublin, F.D.; Reingold, S.C. The National Multiple Sclerosis Society (USA) Advisory Board Committee on clinical trials of new agents in multiple sclerosis. Defining the clinical course of multiple sclerosis: Results of an international survey. Neurology 46:907–911; 1996. 5. Miller, D.H.; Barkhof, F.; Berry, I.; Kappos, L.; Scotti, G.; Thompson, A.J. Magnetic resonance imaging in monitoring the treatment of multiple sclerosis: Concerted action guidelines. J. Neurol. Neurosurg. Psychiatry 54:683– 688; 1991. 6. Grimaud, J.; Lai, M.; Thorpe, J.W.; et al. Quantification of MRI lesion load in multiple sclerosis: A comparison of three computer-assisted techniques. Magn. Reson. Imaging 14:495–505; 1996. 7. Grossman, R.I. Magnetization transfer in multiple sclerosis. Ann. Neurol. 36:S97–S99; 1994. 8. Grossman, R.I.; Lenkinski, E.; Ramer, N.K.; GonzalesScarano, F.; Cohen, J.A. MR proton spectroscopy in multiple sclerosis. AJNR 13:1535–1543; 1992.
PII S0730-725X(98)00083-6
● Original Contribution
HOW DOES BRAIN MRI LESION VOLUME CHANGE ON SERIAL SCANS IN PATIENTS WITH MULTIPLE SCLEROSIS? M. FILIPPI,* M.P. SORMANI,*† M. ROVARIS,*
AND
G. COMI*
*MS Biosignal Analysis Center, Department of Neurology, Scientific Institute Ospedale San Raffaele, University of Milan, Milan; and †Unit of Clinical Epidemiology and Trials, National Institute for Cancer Research, Genoa, Italy Although lesion load changes on conventional T2-weighted brain magnetic resonance imaging (MRI) scans from patients with multiple sclerosis (MS) are used to monitor the effect of treatment, there is no clear definition of how lesion load changes over years according to the lesion load present at a baseline evaluation. In the present study, we evaluated the relationship between lesion load changes over time and lesion load at a baseline evaluation in a group of untreated patients with MS. We scanned nineteen patients on two separate occasions with a mean interval 16.4 months between the two examinations. In each scanning session, a scan with forty contiguous 3-mm-thick axial slices was acquired. We assessed MRI lesion loads using a semi-automated local thresholding technique. Both a linear (p < 0.0001) and a quadratic component (p 5 0.0008) of the baseline volume were significant in describing the follow-up volume. The equation to model this finding was as follows: Vf 5 b0 Vb 1 b1 (Vb)2, where Vf is the lesion volume at follow-up, Vb is the lesion volume at baseline, b0 5 0.834 (SE 5 0.098), and b1 5 0.014 (SE 5 0.003) (mL)21. Our data indicate that lesion volume changes detectable on serial brain MRI studies from patients with MS are dependent on the extent of lesion burden present on the baseline MRI scans. This finding has to be considered when planning phase III trials. © 1998 Elsevier Science Inc. Keywords: Multiple sclerosis; Brain; Lesion volume measurements; Clinical trials.
INTRODUCTION
tween lesion load changes over time and lesion load at a baseline evaluation in a group of untreated patients with MS.
Changes of lesion loads on conventional T2-weighted magnetic resonance imaging (MRI) scans of the brain from patients with multiple sclerosis (MS) are used in virtually all long-term natural history studies and treatment trials as a measure of disease evolution.1 From the placebo arms of previous clinical trials, it has been estimated that the expected change of lesion load is around 5–10% per annum.2 At present, however, no attempt has been made to provide a model to predict the changes of lesion load over time on the basis of the lesion load detected at a baseline evaluation. This is clearly important when planning clinical trials, because it is conceivable that the rate of lesion load accumulation might be different according to the amount of lesions present on the entry scan. In the present study, we assessed which model best explains the relationship be-
MATERIALS AND METHODS Patients Nineteen consecutive patients (thirteen women, five men) with clinically definite MS3 entered the study. Thirteen had a relapsing-remitting and five a secondaryprogressive MS disease course.4 Their mean age was 33.4 years (SD 5 7.9 years), median duration of the disease was 6.0 years (range: 2–15 years), and median expanded disability status scale (EDSS) was 2.0 (range: 1.0 –7.5). To be included, patients must have had neither clinical relapses nor steroid treatment during the 3 months preceding the study. In addition, they were not treated with immunosuppressive or immunomodulating drugs for a period starting 6 months prior to inclusion in
RECEIVED 11/1/97; ACCEPTED 2/2/98. Address correspondence to Dr. Massimo Filippi, MS Biosignal Analysis Center, Department of Neurology, Scientific
Institute Ospedale San Raffaele, via Olgettina 60 –20132, Milan, Italy. E-mail address: [email protected] 1181
Magnetic Resonance Imaging ● Volume 16, Number 10, 1998
1182
the study and lasting until the end of the study. In case of relapses during the follow-up period, only steroid treatment was allowed (usually methylprednisolone i.v. 1 g/day for 5 days). The exit MRI scans were always obtained at least 3 months after the last relapse or steroid treatment. Written informed consent was obtained from all patients before inclusion in the study. MRI Brain MRI scans were performed using a 1.5 Tesla machine (Siemens Magnetom SP63). The pulse sequence used (SE 2000/50) provided a moderate T2 weighting and gave good definition of the MS lesions, with some suppression of the cerebrospinal fluid signal. Brain MRI scans consisted of forty contiguous interleaved 3-mmthick axial slices, with a field of view of 22 cm and a 256 3 256 matrix size. For all patients, MRI scans were performed both at entry to the study and follow-up, with a mean interval of 16.4 months (range: 13–20 months) between the two scans. For both the entry and the follow-up scans, the MRI slices were carefully positioned according to published guidelines.5 MRI Quantification A single observer (MR), unaware of the scan timepoint, identified and marked the lesions on hard copies. Then lesion volumes were measured on two different occasions (separated by one month), using the hard copies as a reference, by three independent technicians using a semi-automated local thresholding technique for lesion segmentation.1,6 Statistical Analysis The mean lesion volumes obtained for the baseline and the follow-up scans from each patient by the three technicians on the two occasions entered the analysis. The comparison between lesion load at baseline and follow-up was performed using the Wilcoxon signed rank test. To evaluate the relationship between the baseline lesion volume and the changes over time, data were fitted using a multiple linear regression model (backward selection): the follow-up volume (Vf) was treated as the dependent variable; the covariates entering the model were a linear (Vb) and a quadratic (Vb2) component of the baseline volume; and a time variable (Dt) correcting for the deviance of each subject from the mean follow-up time (Dt 5 t 2 16 months), according to the following formula: V f 5 b0 z Vb 1 b1zVb 1 b2 z Dt. 2
(1)
RESULTS The mean (6SD) lesion volumes obtained from the two measurements performed by the three technicians
Fig. 1. Correlation between baseline and follow-up volumes from nineteen patients with MS.
was 16.8 (611.1) mL for entry and 19.4 (615.4) mL for exit scans (p 5 0.036). The mean volume ratio (i.e., lesion volume at follow-up divided by lesion volume at baseline) was equal to 1.05 (SD 5 0.22). The percentage increase in lesion volume measured at follow-up was not constant as the baseline volume increased (Fig. 1). When data were fitted by a multiple regression model, the time-correcting factor was not significant and quit the model. Both a linear (p , 0.0001) and a quadratic component (p 5 0.0008) of the baseline volume were significant in describing the follow-up volume. The final equation to model this finding was as follows (Fig. 2): V f 5 b0 Vb 1 b1 (Vb)2,
(2)
where Vf is the lesion volume at follow-up and Vb is the lesion volume at baseline; the two parameters were estimated as b0 5 0.834 (SE 5 0.098) and b1 5 0.014 (SE 5 0.003) (mL)21. This finding indicates that a linear relation describes the dependence of the ratio of lesion volume on lesion volume at baseline: V r 5 0.834 1 0.014 Vb,
(3)
where Vr is the ratio of lesion volumes. DISCUSSION The main result of the present study is the demonstration that lesion volume changes detectable on serialbrain MRI studies from patients with MS are dependent on the extent of lesion burden present on the entry MRI scans. The final model which fits with our data indicates
Brain MRI lesion volume changes in MS ● M. FILIPPI
Fig. 2. Scatterplot of the volume ratios versus lesion volumes at baseline from nineteen patients with MS. The dotted line indicates no changes in lesion volumes between baseline and follow-up.
that if the baseline lesion volume is around 10 mL, no significant changes of the lesion volume are to be expected in the next 1–2 years, whereas if the baseline lesion load is around 30 – 40 mL, an average increase of about 20 –35% has to be expected in the same period of time. The first consequence of this observation is that lesion load accumulation is different in different phases of disease evolution. Our data suggest that in patients with low lesion load there might be a balance between mechanisms leading to lesion load increase (i.e., new lesion formation and enlargement of pre-existing lesions) and mechanisms leading to repair. This balance seems to be in patients with higher lesion loads. Whether the geometric increase of lesion load is due to a failure of the reparative mechanisms or to an increase of the pathological damage needs to be elucidated. Our sample size was constituted only by patients with low to medium lesion loads. Possibly in patients with higher lesion loads this tendency to a continuous progressive increase in lesion load would be lost. Once there is diffuse damage of the white matter, it is difficult to measure any other pathological change using conventional T2-weighted scans. The presence of continuous activity and recurrent damage within the diseased white matter is perhaps better monitored by the use of enhanced MRI or newer non-conventional MR techniques, such as magnetization transfer imaging7 or magnetic resonance spectroscopy.8 Because the amount of lesion load changes which can be expected to occur in a period of time similar to the
ET AL.
1183
follow-up periods of clinical trials is dependent on the lesion volumes present on the baseline scans, our results have some implications also for the use of MRI lesion load in phase III trials. First, patients should be randomized to placebo or treatment according to their lesion load at entry, because it is conceivable that patients with low lesion loads will tend to have a stable lesion load regardless of being actively treated or not. Second, the duration of follow-up when patients with low lesion loads are studied should be longer than the average period of 2–3 years generally used in trials if any significant change of their lesion load has to be detected. In conclusion, our data indicate that the total lesion load seen on a baseline scan of patients with MS might be a good predictor of lesion load changes over the next few years. Further studies are now needed to confirm these data and to evaluate the evolution of lesion load in patients with higher lesion loads on a baseline scan. Acknowledgements—Dr. Maria Pia Sormani is supported by a grant from TEVA Italy.
REFERENCES 1. Filippi, M.; Horsfield, M.A.; Tofts, P.S.; Barkhof, F.; Thompson, A.J.; Miller, D.H. Quantitative assessment of MRI lesion load in monitoring the evolution of multiple sclerosis. Brain 118:1601–1612; 1995. 2. Paty, D.W.; Li, D.K.B.; Oger, J.J.F.; et al. Magnetic resonance imaging in the evaluation of clinical trials in multiple sclerosis. Ann. Neurol. 36:S95–S96; 1994. 3. Kurtzke, J.F. Rating neurological impairment in multiple sclerosis: An expanded disability status scale (EDSS). Neurology 33:1444 –1452; 1983. 4. Lublin, F.D.; Reingold, S.C. The National Multiple Sclerosis Society (USA) Advisory Board Committee on clinical trials of new agents in multiple sclerosis. Defining the clinical course of multiple sclerosis: Results of an international survey. Neurology 46:907–911; 1996. 5. Miller, D.H.; Barkhof, F.; Berry, I.; Kappos, L.; Scotti, G.; Thompson, A.J. Magnetic resonance imaging in monitoring the treatment of multiple sclerosis: Concerted action guidelines. J. Neurol. Neurosurg. Psychiatry 54:683– 688; 1991. 6. Grimaud, J.; Lai, M.; Thorpe, J.W.; et al. Quantification of MRI lesion load in multiple sclerosis: A comparison of three computer-assisted techniques. Magn. Reson. Imaging 14:495–505; 1996. 7. Grossman, R.I. Magnetization transfer in multiple sclerosis. Ann. Neurol. 36:S97–S99; 1994. 8. Grossman, R.I.; Lenkinski, E.; Ramer, N.K.; GonzalesScarano, F.; Cohen, J.A. MR proton spectroscopy in multiple sclerosis. AJNR 13:1535–1543; 1992.