J Neurol DOI 10.1007/s00415-014-7284-0

ORIGINAL COMMUNICATION

Interpreting therapeutic effect in multiple sclerosis via MRI contrast enhancing lesions: now you see them, now you don’t Ilana R. Leppert • S. Narayanan • D. Arau´jo • P. S. Giacomini • Y. Lapierre • D. L. Arnold • G. B. Pike

Received: 11 November 2013 / Revised: 10 February 2014 / Accepted: 12 February 2014 Ó Springer-Verlag Berlin Heidelberg 2014

Abstract Gadolinium (Gd) enhancement of multiple sclerosis (MS) lesions on MRI scans is a commonly used outcome measure in therapeutic trials. However, enhancement depends on MRI acquisition parameters that might significantly alter detectability. We investigated how the difference in blood–brain barrier (BBB) permeability threshold between MRI protocols affects lesion detection and apparent enhancement time using dynamic-contrastenhanced (DCE) MRI. We examined fourty-four relapsingremitting MS patients with two MRI protocols: ‘standard sensitivity’ (SS) (1.5 T, single-dose Gd) and ‘high sensitivity’ (HS) (3 T, triple-dose Gd, delayed acquisition). Eleven patients had at least one enhancing lesion and completed the 1-month follow-up. We acquired DCE-MRI during the HS protocol and calculated BBB permeability. Sixty-five lesions were enhanced with the SS vs. 135 with the HS protocol. The detection threshold of the HS was significantly lower than that of the SS protocol (Ktrans = 2.64 vs. 4.00E-3 min-1, p \ 0.01). Most lesions (74 %) were in the recovery phase; none were in the onset phase and 26 % were at the peak of enhancement. The estimated duration of detectability with the HS protocol was significantly longer than for the SS protocol (6–12 weeks vs. 3 weeks). Our observations on the protocoldependent threshold for detection and time-course help explain discrepancies in the observed effects of antiinflammatory therapies on MS lesions.

I. R. Leppert (&)  S. Narayanan  D. Arau´jo  P. S. Giacomini  Y. Lapierre  D. L. Arnold  G. B. Pike Montreal Neurological Institute, 3801 University St, WB325, Montreal, QC H3A 2B4, Canada e-mail: [email protected]

Keywords Permeability  DCE-MRI  Lesion enhancement  Limit of detection  Therapy  Time-course

Introduction Multiple sclerosis (MS) is an inflammatory disease of the CNS characterized by demyelinating lesions. The acute phase of lesion formation involves local inflammation and an increase in the permeability of the blood–brain barrier (BBB) [12] that can be visualized with contrast-enhanced MRI. Specifically, gadolinium-based (Gd) contrast agents lead to a signal enhancement on T1-weighted MRI in areas of active inflammation and breakdown of the BBB. This technique is widely used and typically results in a classification of lesions as enhancing or non-enhancing. However, this conventional binary approach can lead to apparently conflicting results, particularly across differing MRI acquisition protocols. In fact, protocol variations, such as higher contrast doses, higher field strengths and delayed post-Gd acquisition [8, 9, 23–25, 28, 31], will increase the sensitivity to Gd-enhancement, leading to the classification of additional lesions that are not identified as ‘‘enhancing’’ using conventional protocols [8, 23, 30]. The use of contrast agents with different relaxivities, which directly affect the observed T1 relaxation time, might also affect lesion counts [22]. Understanding these protocolrelated differences is particularly important in the context of assessing therapeutic effects, since Gd-enhancing lesion frequency is commonly used as a marker of therapeutic efficacy [16, 21, 26, 27]. The effect of protocol sensitivity was exemplified in the BECOME study [2], where a 3 T triple-dose, delayed acquisition with background suppression showed little difference in the effect of two therapies that were expected to exhibit significantly different

123

J Neurol

Fig. 1 The effect on lesion count due to a a threshold difference between protocols and b the shape of the enhancement time course. a With the standard sensitivity (SS) protocol, both treatments (Rx1 and Rx2) affect lesion count, while only Rx2 has an effect on lesion

count with the high sensitivity (HS) protocol. b Most likely observation times (obs1 and obs2) depend on enhancement timecourse shape, and, thus, lead to different lesion classification (new, persistent or resolved) and count (see text for more explanations)

suppressions of BBB disruption [5, 33] and had shown different effects in multiple clinical trials [6, 11, 19, 20]. Moreover, this more sensitive protocol unexpectedly showed little effect of either therapy on Gd-lesion enhancement frequency. An illustration of how the difference in detection threshold of the protocols can affect lesion count is shown in Fig. 1a, where treatment 1 (Rx1) would show a therapeutic effect with a more conventional protocol (e.g., single dose 1.5 T) but not with a more sensitive protocol (e.g., triple-dose 3 T), whereas treatment 2 (Rx2) would show an effect with both. In addition, describing the shape of the lesion enhancement time-course would help determine whether the significant increase in the number of lesions observed with more sensitive protocols is explained by the presence of new lesions, or rather persistent lesions that have not yet resolved, but have a permeability below that detectable by conventional protocols. To illustrate this, Fig. 1b shows different possible enhancement time-course shapes. The most likely observation times (obs1 and obs2) correspond to the slowest part of each curve. If the enhancement time-course was characterized by an equally slow onset and recovery, a more sensitive protocol would enable the observation of lesions prior to and beyond their appearance on more conventional protocols (solid black line). Alternatively, if initial enhancement was rapid and recovery much slower, a protocol with a lower detection threshold would be more likely to detect lesions in the persistent recovery phase, which would appear to be resolved with a less sensitive protocol (black dashed line). Conversely, slow onset and quick recovery, would favor the detection of new lesions not yet visible with a conventional protocol (grey dashed line). Hence, the shape of the enhancement time-course would affect lesion count since it depends on lesion classification

(new, persistent or resolved) and could, thus, lead to different conclusions about therapeutic efficacy. The aim of this work was to use dynamic-contrastenhanced (DCE) MRI, which can be used to quantify the passing of contrast agent from blood plasma to lesion [14, 29], to compare the detection thresholds of a ‘standard sensitivity’ (SS) and a ‘high sensitivity’ (HS) protocol in order to estimate its impact on lesion detection and the apparent time-course of lesion enhancement. To this end, a HS protocol was defined [2, 18] that combines several measures previously shown to increase the sensitivity to Gd-enhancement, namely increased dose [9], added delay, and a magnetization transfer pulse for background suppression [9, 24, 25] in addition to increased field strength [7, 23, 30], as compared to more conventionally used protocols.

123

Materials and methods Subjects were examined using both SS and HS MRI protocols (described below) at baseline (m0) and at 1-month (m1) follow-up, resulting in a total of four exams per subject completing the study. Forty-four relapsing–remitting multiple sclerosis (RRMS) patients were recruited, who were expected to be stable in terms of disease-modifying therapy, meaning that initiation, cessation or change of therapy was not planned between the baseline and 1-month follow-up scans. In addition, MRI scanning was not performed within 60 days following steroid therapy. Only patients with at least one enhancing lesion visualized with the HS protocol at baseline were imaged again at 1-month follow-up. To increase the likelihood of detecting enhancing lesions, patients with a recent clinical relapse, a

J Neurol

short to moderate disease duration (mean 10 years) and relatively young age (mean 34 years) were recruited [1]. The study was approved by our Research Ethics Board and all subjects provided written informed consent. Standard sensitivity (SS) protocol While there are many different T1-weighted MRI protocols used for detection of Gd enhancing MS lesions, we chose a 3D GRE acquisition (1.5 T Siemens Sonata, TE/TR = 9/ 30 ms; FOV 192 9 256; 1 9 1 9 3 mm3, 60 slices, FA = 30°) and imaged prior to and 10 min after a singledose (0.1 mmol kg-1) injection of Gd-BT-DO3A (gadobutrol). This contrast agent is a macrocyclic chelate with high relaxivity and high kinetic stability, indicating a good safety profile. The overall protocol also included PD/T2 (TE1/TE2/ TR = 12/83/2,070 ms; 1 9 1 9 3 mm3) and FLAIR (TE/ TI/TR = 66/2,500/9,360 ms; 1 9 1 9 3 mm3) sequences, all acquired pre-contrast. A neuroradiologist manually outlined all enhancing lesions, defined to consist of at least three clustered voxels. The lesions were defined as being concurrently hyper-intense on T2, FLAIR and PD, hypo-intense on T1 pre-Gd, and having a well defined visual enhancement on T1 post-Gd. Enhancement and lesion definition criteria were the same for both acquisition protocols. High sensitivity (HS) protocol Using a 3 T TIM Trio Siemens scanner, a pre-injection 3D GRE (TE/TR = 6/30 ms, FOV 192 9 256, 1 9 1 9 3 mm3, 60 slices, FA = 27°) was acquired as well as a PD/ T2 pair (TE1/TE2/TR = 16/80/2,100 ms; 1 9 1 9 3 mm3). The 3D-FLASH volumes were acquired for the purpose of T1-mapping using the DESPOT1 technique (FA = 3°/15°, 192 9 256 9 40 matrix, 2 averages) [4]. Subsequently, DCE-MRI images were obtained prior to, during and after a single-dose injection of the contrast agent, which was flushed with a 30 ml bolus of saline. The dynamic images were acquired in the transverse plane using a dual-temporal and dual-spatial resolution (high temporal, low spatial resolution: TE/TR = 2.6/5.6 ms, 2 9 2 9 6 mm3, 96 9 128 matrix, 20 slices, FA = 20°, 5 s temporal resolution, 1:08 min; low temporal, high spatial resolution: 1 9 1 9 3 mm3, 192 9 256 matrix, 40 slices, FA = 20°, 32 s temporal resolution, 19:21 min) 3DFLASH acquisition (ending *20 min after injection) for a total of 85 dynamic images [15]. Immediately after the dynamic series, an additional double-dose was administered and a FLAIR (TE/TI/TR: 39.7/1,800/5,000 ms; 1 9 1 9 1 mm3) image was acquired. Finally, a postcontrast image, matched to the initial pre-injection 3D GRE, was acquired approximately 15 min after the final injection. Both the pre and post 3D GRE sequences

included a magnetization transfer pre-pulse for background suppression. Enhancing lesions were visually outlined by a neuroradiologist on the final triple-dose image, as well as on the final image of the dynamic series (note that the use of a double or triple-dose of this contrast agent is off-label in the USA). The SS and HS protocol scan sessions were acquired within a maximum interval of three days, in order to minimize temporal differences in lesion progression while allowing for contrast washout. The scanner sessions were randomized in order to limit any bias relative to the enhancement time-course of lesions, and the neuroradiologist was blinded to the nature of the acquisitions to be segmented. The T1-weighted time series was used together with the equilibrium magnetization map calculated with the DESPOT1 technique [4] to generate quantitative T1 4D volumes and were subsequently converted to concentration using the pre-contrast T1 maps and contrast agent relaxivity in the plasma (rGd-BT-DO3A = 5.0 ± 0.3 mM-1 s-1 [22]). Dynamic compartment modeling was performed using the Tofts–Kermode model [29]: Ct ðtÞ ¼ Ktrans

Zt

Cp ðuÞe

Ktrans ve ðtuÞ

du;

0

where Cp is the plasma concentration in a feeding artery, or arterial input function (AIF), ve is the extra-cellular, extravascular volume fraction, and the transfer constant Ktrans (proportional to BBB permeability) is in min-1. It was assumed that the vascular component is small and that the uptake is relatively slow, warranting the use of this simplified model. A manually segmented region of interest in the sagittal sinus was used for AIF estimation, with venous and arterial concentrations assumed equal at all times. Although the signal from the sagittal sinus is not arterial and, therefore, introduces deviations due to delay and dispersion, the large size of this vein makes it much less susceptible to partial volume effects, and the venous signal function has been shown to be both highly reproducible [17, 32] and a reasonable surrogate for the AIF in this context [10]. All fits with an R2 below 0.5 were rejected for further analysis. Prior to lesion segmentation, all images from both protocols were linearly registered to the final triple-dose 3 T image, such that the neuroradiologist carrying out the segmentation could visually verify correct alignment. To estimate the threshold for detection of each protocol, the mean Ktrans of voxels that were ‘‘barely detected’’ was used. These voxels were defined as having an intensity increase \30 % above the pre-injection intensity, within the segmented lesions. Although this limit is somewhat arbitrary, it was chosen to include all voxels that were ‘‘barely’’ visible by the rater, but high enough to overcome

123

J Neurol

Fig. 2 Lesion classification: case A (onset) enhances only with HS at m0, enhances with both protocols at m1, case B (peak) enhances at both time points, with both protocols, case C (recovery) enhances with both protocols at m0, but only with HS at m1

the differences in signal to noise (SNR) and contrast to noise (CNR) ratios of the protocols. In other words, an enhancement of 30 % is visible on both the HS and SS images, despite an approximate twofold increase in SNR in the HS (3 T) protocol. Despite the relatively low SNR of the last dynamic image of the DCE-MRI acquisition, it was considered as a pseudo ‘delayed 3 T single-dose’ acquisition, and the segmented lesions were, therefore, useful in differentiating the relative contributions of field strength and increased Gd dose to the increase in sensitivity of the HS protocol. Based on these thresholds and previous literature, we hypothesized a time-course of enhancement (Fig. 2). Lesions (and voxels) that enhance at least at one time-point on the SS protocol were classified into one of three epochs: onset (A), peak (B) and recovery (C). The onset state corresponds to lesions or voxels that only enhance with the HS protocol at m0 and enhance with both protocols at m1 (A in Fig. 2). Lesions (and voxels) were classified as being at the peak of enhancement, if they enhance at both time points, with both protocols (B in Fig. 2). Finally, the recovery phase corresponds to lesions (and voxels) that enhance with both protocols at m0 and are only visible with the HS protocol at m1 (C in Fig. 2). The lesions that are only visible with the increased sensitivity protocol at both time points can not be classified into any of these three states since it is impossible to determine on which side of the curve they reside.

Results Of the 44 patients that were recruited, 19 had at least one lesion that enhanced with the HS protocol at m0 (average age 34.2 years, range 18–49 years, 14 female, five male; average disease duration 5.5 years, range 0.25–21 years; mean time between last relapse and baseline scan 7.7 months, range 0–5.1 years) and 11 completed the baseline and follow-up scans. Figure 3 and Table 1 summarize the number of lesions and lesion voxels for subjects with both complete (m0 and m1) and incomplete (m0 only) datasets. All segmented lesions were concurrently hyperintense on T2, PD, FLAIR and T1 post-Gd images.

123

Incomplete datasets were only used for direct protocol comparisons while the complete datasets were also used for subsequent lesion time-course analysis. From Table 1, it is clear that the HS protocol provides a twofold to threefold increase in the number of detectable lesions and voxels. Figure 4 illustrates an example, where three lesions were visible on the post-Gd images with the HS protocol and only two were visible with the SS protocol. In total, 52 % of lesions detectable with the HS protocol had no detectable enhancement with the SS protocol. This corresponds to a 108 % increase in the number of detectable lesions with the HS protocol. Of the 44 patients, 43 % showed Gdenhancing lesion activity with the HS protocol whereas only 36 % showed activity with the SS protocol. In terms of scans, 31 % of the scan pairs that displayed an ‘‘active scan’’ (i.e., at least one enhancing lesion) with the HS protocol did not show Gd-enhancing activity on the corresponding SS protocol scan. There was never a lesion detected with the SS protocol that was not detected with the HS protocol, but for eight lesions, the volume segmented on the SS image was larger than on the HS image. However, in these cases, the SS protocol lesion volume was on average only 0.3 % larger, corresponding to relatively few voxels (96 in total). It was, thus, assumed that these voxels were most likely a result of slight misregistration or reslicing issues and were ignored for the remainder of the analysis. In order to compare the lesion counts with those outlined on the pseudo ‘delayed single-dose 3 T’ images, all masks were re-sampled to match the reduced coverage of this acquisition. On the pseudo ‘delayed single-dose 3 T’ images, 69 lesions were visible, which is only 7.8 % more than on the re-sampled SS images (64 lesions) and 49 % less than on the re-sampled HS images (134 lesions). Figure 5a illustrates the average Ktrans value associated with the percent enhancement between pre-Gad and postGad images for each protocol. From this figure, it is clear that the HS protocol provides greater signal enhancement and is, therefore, able to detect lesion voxels with lower permeability values. For example, a permeability of 0.005 min-1 leads to a 40 % signal increase with the SS protocol but nearly 150 % enhancement with the HS protocol, making it much more visible for segmentation. In Fig. 5b, which is a zoomed-in version of Fig. 5a, the

J Neurol Fig. 3 Lesion count for each subject time point and each protocol. Complete: number of lesions for patients with m0 and m1 data. With incomplete: number of lesions, including patients with baseline m0 only

Table 1 Number of enhancing lesions (voxels) for the SS and HS protocols SS

HS

Time-point m0

24 (1,845)

54 (3,563)

Time-point m1

15 (653)

38 (2,773)

Total (only complete: m0 and m1)

39 (2,498)

92 (6,336)

Total (including incomplete)

65 (4,028)

135 (9,829)

Data are presented for only complete [11 patients with baseline (m0) and follow-up (m1)] and including incomplete (eight patients with baseline only) datasets

horizontal bars denote the average Ktrans values of voxels that are barely detected (less than a 30 % difference between registered pre-Gad and post-Gad images) for each protocol. Based on this average, we determined the threshold to be 4.00E-3 min-1 (black horizontal bar) for the SS protocol, but observed a significantly lower (Mann– Whitney U test z = 7.0; p \ 0.01) detection threshold of 2.64E-3 min-1 for the HS protocol (grey horizontal bar). For the remainder of the analysis, only patients with complete datasets (four scans: both time points and both protocols) were included (11 patients; average age 35.1 years, range 18–49 years; eight female, three male). From Table 2, most lesions (74 %) were in the recovery phase (C) over the 1-month time interval studied, whereas none were captured in the acute onset phase (A) and only

26 % of lesions were at the peak (B) of enhancement. These values were computed as the average ratio on a persubject basis such that patients with more or larger lesions did not bias the result. Based on weekly measurements reported by Giovannoni et al. [13] and Cotton et al. [3], the duration of detectable enhancement was estimated to be approximately 3.07–3.5 weeks with the SS protocol. This would correspond to lesions labeled B in Fig. 2, with 26 % occurrence. Assuming that lesions are randomly sampled along their time-courses, the observed frequencies of occurrence should be proportional to the length of each epoch. In the limit that lesions are in the increasing phase, barely above the SS threshold (point 1 in Fig. 6), the duration of detectability with the HS protocol would be [3.07 9 (74/ 26) - 3.07] & 5.7 weeks. Alternatively, if the m0 timepoint was barely above the SS protocol threshold, but on the recovery side of the curve (point 2 in Fig. 6), the total duration would be [3.07 9 (74/26) ? 3.07] & 11.8 weeks. In other words, if 26 % corresponds to 3.07 weeks (peak), then 74 % corresponds to 8.74 weeks (recovery). Because the recovery overlaps to an unknown degree with the peak phase (B and C), the limiting cases are estimated by either subtracting (5.7 weeks) or adding (11.8 weeks) these values. Thus, the total estimated duration of detectability with the HS protocol ranges from 5.7 to 11.8 weeks.

123

J Neurol

Fig. 4 Example of the difference in lesion detection sensitivity a HS protocol: three lesions are visible, b SS protocol: only two lesions are visible

Discussion The direct impact of using a HS 3 T triple dose delayed [2] vs. a single dose, 1.5 T protocol is readily observable through the lesion count, which increased by approximately 108 %. The voxel count, and, thus, the enhancing lesion volume, increased by 144 %. The higher field strength alone led to only an approximate 7.8 % increase in the number of lesions detected. As such, it appears that the Gd dose is a much more important factor than field strength when attempting to compare or pool data across sites and studies. The values reported here are consistent with previously published results, where the effects of the field strength increasing from 1.5 to 3 T led to a 7.5–21 % increase in the number of contrast-enhanced lesions [23, 30] while an increase from single to triple dose with added delay (at 1.5 T) led to an approximate 126 % increase in the number of lesions detected [24, 25]. However, the

Fig. 5 Average Ktrans vs. percent enhancement a difference in enhancement between SS and HS protocols and b Ktrans threshold for detection with SS (4.00E-3 min-1) and with HS (2.64E-

123

results do not agree with a very recent study comparing similar protocols (i.e., standard 1.5 T vs. triple-dose delayed on 3 T), where an increase of only 38.2 % was reported (179.6 % for lesion volume) [18]. This discrepancy can partly be explained by the fact that the latter report only included patients having at least one Gdenhancing lesion on a 1.5 T screening scan. This criteria would have excluded at least 31 % of the scan pairs in our study that showed no disease activity with the SS protocol but did with the HS protocol. The DCE-MRI provides a quantitative tool for describing BBB impairment in MS lesions, and enabled us to estimate the BBB permeability increase required for lesions to be detected as Gd-enhancing with both protocols. This reinforces the notion that lesion enhancement is not a binary event, but rather a continuous process associated with varying degrees of BBB impairment. The finite sensitivity of the MRI protocol acts as a filter on the nature of

3 min-1) protocol, estimated as the average Ktrans across a signal enhancement of \30 % (b is zoomed-in version of a)

J Neurol Table 2 Relative occurrence of enhancement phase: onset (A), peak (B) and recovery (C) (see Fig. 2 for description)

A

B

C

% of total lesions (%)

0

26

74

% of total voxels (%)

1

11

88

therapies that differentially reduce BBB permeability could be shown to have different Gd-enhancing lesion counts using a SS protocol, yet exhibit the same number of Gdenhancing lesions using a protocol with higher sensitivity. Thus, a more sensitive protocol is not necessarily better for the task of measuring a treatment effect on Gd-enhancing lesion counts in a drug trial. More work is required to determine the optimal protocol sensitivity for this task or, alternatively, a better measure of suppression of inflammation is needed for use in therapeutic trials. Acknowledgments We would like to thank Rozalia Arnaoutelis for her invaluable help in patient recruitment and booking. This work was supported by the Multiple Sclerosis Society of Canada.

Fig. 6 Estimated time course of enhancement and relative occurrence of lesions in each phase. A onset, B peak, C recovery. Based on previous work, lesions last on average 3.07 weeks with the SS protocol, the HS protocol would enable detection from 5.7 to 11.8 weeks. The SS and HS protocol thresholds are based on the mean Ktrans of voxels at the detection level (see Fig. 5)

the lesions that are visualized, implying that some focal regions of BBB impairment are likely routinely missed on conventional imaging. The current work demonstrates the difference in the threshold for detection of enhancement; the HS protocol can detect nearly half the level of BBB impairment that would be required in order to be detected with the SS protocol, a finding that is critical for understanding the effects of drug therapy, where assessment based simply on (binary) enhancing lesion counts could be misleading across varying platforms and protocols (e.g., the BECOME [2] study described in the introduction). This paper also presented a probable time-course for enhancement of lesions based on the estimated thresholds, reinforcing the notion that lesion enhancement has a rapid onset and a much slower recovery phase. Based on these results, an average duration of enhancement for the HS protocol could be estimated to be two to three times that of the SS protocol (5.7–11.8 vs. 3.07–3.5 weeks). Since lesions are most often detected in the recovery phase, it would appear that lesions persist much longer when using a HS protocol. This would also imply that increasing the sensitivity of a protocol could lead to the detection of relatively more persistent lesions as opposed to new foci of inflammation. These findings have significant implications for interpreting studies investigating relative drug efficacies; immunomodulatory therapies are often assumed to eliminate the foci of white matter inflammation when Gdenhancing lesions are not detected after therapy, when, in fact, they could just be lowering the permeability below a protocol-dependent detectability threshold. Hence, two

Conflicts of interest Dr. Arnold has served on advisory boards, received speaker honoraria, served as a consultant or received research support from Bayer, Biogen Idec, Coronado Biosciences, Consortium of Multiple Sclerosis Centers, Eli Lilly, EMD Serono, Genentech, Genzyme, GlaxoSmithKline, MS Forum, NeuroRx Research, Novartis, Opexa Therapeutics, Roche, Merck Serono, S.A., Serono Symposia International Foundation, Teva, the Canadian Institutes of Health Research, and the Multiple Sclerosis Society of Canada; and holds stock in NeuroRx Research. Dr. Narayanan has received personal compensation from NeuroRx Research, Teva Neurosciences Canada and Biogen Idec Canada for consulting services. Dr. Giacomini has received personal compensation from NeuroRx Research, Allergan, Bayer, Biogen Idec, Genzyme, Novartis, EMD Serono and Teva Neuroscience for speaking, advisory board participation or consulting services.

References 1. Barkhof F, Held U, Simon JH, Daumer M, Fazekas F, Filippi M, Frank JA, Kappos L, Li D, Menzler S, Miller DH, Petkau J, Wolinsky J, for the Sylvia Lawry Centre for MSR (2005) Predicting gadolinium enhancement status in MS patients eligible for randomized clinical trials. Neurology 65:1447–1454 2. Cadavid D, Wolansky LJ, Skurnick J, Lincoln J, Cheriyan J, Szczepanowski K, Kamin SS, Pachner AR, Halper J, Cook SD (2009) Efficacy of treatment of MS with IFNb-1b or glatiramer acetate by monthly brain MRI in the BECOME study. Neurology 72:1976–1983 3. Cotton F, Weiner H, Jolesz F, Guttmann C (2003) MRI contrast uptake in new lesions in relapsing–remitting MS followed at weekly intervals. Neurology 60:640–646 4. Deoni SC, Rutt BK, Peters TM (2003) Rapid combined T1 and T2 mapping using gradient recalled acquisition in the steady state. Magn Reson Med 49:515–526 5. Dhib-Jalbut S (2002) Mechanisms of action of interferons and glatiramer acetate in multiple sclerosis. Neurology 58:S3–S9 6. Durelli L, Clerico M (2005) Comparison of immunomodulatory treatments for multiple sclerosis. Eur J Neurol Off J Eur Fed Neurol Soc 12:915 7. Erskine MK, Cook LL, Riddle KE, Mitchell JR, Karlik SJ (2005) Resolution-dependent estimates of multiple sclerosis lesion loads. Can J Neurol Sci 32:205–212 8. Filippi M, Mastronardo G, Bastianello S, Rocca MA, Rovaris M, Gasperini C, Pozzilli C, Comi G (1998) A longitudinal brain MRI study comparing the sensitivities of the conventional and a newer

123

J Neurol

9.

10.

11. 12.

13.

14.

15.

16.

17.

18.

19.

20.

approach for detecting active lesions in multiple sclerosis. J Neurol Sci 159:94–101 Filippi M, Rovaris M, Capra R, Gasperini C, Yousry TA, Sormani MP, Prandini F, Horsfield MA, Martinelli V, Bastianello S, Ku¨hne I, Pozzilli C, Comi G (1998) A multi-centre longitudinal study comparing the sensitivity of monthly MRI after standard and triple dose gadolinium-DTPA for monitoring disease activity in multiple sclerosis. Implications for phase II clinical trials. Brain 121:2011–2020 Footit C, Cron GO, Nguyen TV, Cameron I, Schweitzer ME, Sinclair J, Woulfe J, Hogan MJ, Nguyen TB (2011) Optimizing perfusion imaging of brain tumors: Validation of venous output function used as a surrogate AIF. In: 19th Proceedings of the international society for magnetic resonance in medicine. Abstract #2038 Francis G (2004) Importance of benefit-to-risk assessment for disease-modifying drugs used to treat MS. J Neurol 251(Suppl):5 Gay D, Esiri M (1991) Blood–brain barrier damage in acute multiple sclerosis plaques. An immunocytological study. Brain J Neurol 114(Pt 1B):557–572 Giovannoni G, Silver NC, Good CD, Miller DH, Thompson EJ (2000) Immunological time-course of gadolinium-enhancing MRI lesions in patients with multiple sclerosis. Eur Neurol 44:222–228 Ingrisch M, Sourbron S, Morhard D, Ertl-Wagner B, Ku¨mpfel T, Hohlfeld R, Reiser M, Glaser C (2012) Quantification of perfusion and permeability in multiple sclerosis: dynamic contrastenhanced MRI in 3D at 3 T. Invest Radiol 47:252–258 Jelescu IO, Leppert IR, Narayanan S, Arau´jo D, Arnold DL, Pike GB (2011) Dual-temporal resolution dynamic contrast-enhanced MRI protocol for blood–brain barrier permeability measurement in enhancing multiple sclerosis lesions. J Magn Reson Imaging 33:1291–1300 Kappos L, Moeri D, Radue EW, Schoetzau A, Schweikert K, Barkhof F, Miller D, Guttmann CR, Weiner HL, Gasperini C, Filippi M (1999) Predictive value of gadolinium-enhanced magnetic resonance imaging for relapse rate and changes in disability or impairment in multiple sclerosis: a meta-analysis. Gadolinium MRI meta-analysis group. Lancet 353:964–969 Lavini C, Verhoeff J (2010) Reproducibility of the gadolinium concentration measurements and of the fitting parameters of the vascular input function in the superior sagittal sinus in a patient population. Magn Reson Imaging Livshits I, Hussein S, Kennedy C, Weinstock-Guttman B, Hojnacki D, Zivadinov R (2012) Comparison of a 1.5 T standard vs. 3 T optimized protocols in multiple sclerosis patients. Minerva Med 103:97–102 Mikol D, Barkhof F, Chang P, Coyle P, Jeffery D, Schwid S, Stubinski B, Uitdehaag B (2008) Comparison of subcutaneous interferon beta-1a with glatiramer acetate in patients with relapsing multiple sclerosis (the REbif vs glatiramer acetate in relapsing MS disease [REGARD] study): a multicentre, randomised, parallel, open-label trial. Lancet Neurol 7:903–914 O’Connor P, Filippi M, Arnason B, Comi G, Cook S, Goodin D, Hartung H-P, Jeffery D, Kappos L, Boateng F, Filippov V, Groth M, Knappertz V, Kraus C, Sandbrink R, Pohl C, Bogumil T

123

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

(2009) 250 lg or 500 lg interferon beta-1b versus 20 mg glatiramer acetate in relapsing–remitting multiple sclerosis: a prospective, randomised, multicentre study. Lancet Neurol 8:889–897 Petkau J, Reingold SC, Held U, Cutter GR, Fleming TR, Hughes MD, Miller DH, McFarland HF, Wolinsky JS (2008) Magnetic resonance imaging as a surrogate outcome for multiple sclerosis relapses. Mult Scler 14:770–778 Rohrer M, Bauer H, Mintorovitch J, Requardt M, Weinmann H-J (2005) Comparison of magnetic properties of MRI contrast media solutions at different magnetic field strengths. Invest Radiol 40:715–724 Sicotte NL, Voskuhl RR, Bouvier S, Klutch R, Cohen MS, Mazziotta JC (2003) Comparison of multiple sclerosis lesions at 1.5 and 3.0 Tesla. Invest Radiol 38:423–427 Silver NC, Good CD, Barker GJ, MacManus DG, Thompson AJ, Moseley IF, McDonald WI, Miller DH (1997) Sensitivity of contrast enhanced MRI in multiple sclerosis. Effects of gadolinium dose, magnetization transfer contrast and delayed imaging. Brain 120:1149–1161 Silver NC, Good CD, Sormani MP, MacManus DG, Thompson AJ, Filippi M, Miller DH (2001) A modified protocol to improve the detection of enhancing brain and spinal cord lesions in multiple sclerosis. J Neurol 248:215–224 Sormani M, Bonzano L, Roccatagliata L, Cutter G, Mancardi G, Bruzzi P (2009) Magnetic resonance imaging as a potential surrogate for relapses in multiple sclerosis: a meta-analytic approach. Ann Neurol 65:268–275 Sormani MP, Li DK, Bruzzi P, Stubinski B, Cornelisse P, Rocak S, De Stefano N (2011) Combined MRI lesions and relapses as a surrogate for disability in multiple sclerosis. Neurology 77:1684–1690 Tofts PS (1996) Optimal detection of blood-brain barrier defects with Gd-DTPA MRI—the influences of delayed imaging and optimised repetition time. Magn Reson Imaging 14:373–380 Tofts PS, Kermode AG (1991) Measurement of the blood–brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 17:357–367 Wattjes M, Lutterbey G, Harzheim M, Gieseke J, Tra¨ber F, Klotz L, Klockgether T, Schild H (2006) Higher sensitivity in the detection of inflammatory brain lesions in patients with clinically isolated syndromes suggestive of multiple sclerosis using high field MRI: an intraindividual comparison of 1.5 T with 3.0 T. Eur Radiol 16:2067–2073 Wolansky LJ, Bardini JA, Cook SD, Zimmer AE, Sheffet A, Lee HJ (1994) Triple-dose versus single dose gadoteridol in multiple sclerosis patients. J Neuroimaging 4:141–145 Wu Y-Y, Wu C-H, Chiang C-M, Chen C-C, Chai J-W (2012) Consistency of permeability measurement using arterial input function and venous output function in DCE-MRI for metastatic brain tumors. In: 20th Proceedings of the international society for magnetic resonance in medicine. Abstract #3508 Yong W (2002) Differential mechanisms of action of interferonbeta and glatiramer acetate in MS. Neurology 59:802–808