Skip to main content
. 2021 Dec 2;33:102904. doi: 10.1016/j.nicl.2021.102904

Table 4.

Prediction of future disability using SPACE-MS metrics.

Clinical variable at follow-up# (dependent variable)
EDSS score Inverse of TWT Inverse of 9HPT PASAT score SDMT score
Best a priori models
Model R2 0.6205 0.5607 0.7528 0.6522 0.8063
Main predictors (below)
Lesion load at baseline in mLRC (95%CI), p-value 3.22•10-06(-1.91•10-06 to 8.35•10-06), p = 0.219 −3.66•10-07(-6.50•10-07 to−8.20•10-08), p = 0.012 −3.56•10-08(-8.31•10-08 to 1.18•10-08), p = 0.140 −0.00025(−0.00033 to−0.00017), p < 0.001 −0.00019(−0.00026 to−0.00013), p < 0.001
WM volume at baseline in mLRC (95%CI), p-value 2.65•10-08(-2.21•10-06 to 2.27•10-06), p = 0.981 −4.83•10-08(-1.70•10-07 to 7.37•10-08), p = 0.437 −3.96•10-09(-2.44•10-08 to 1.64•10-08), p = 0.703 0.00004(2.28•10-06 to 0.00007), p = 0.036 5.14•10-06(−0.00002 to 0.00003), p = 0.686
GM volume at baseline in mL
RC (95%CI), p-value
−4.49•10-07(−2.42•10-06 to 1.52•10-06), p = 0.654 5.07•10-08(−5.87•10-08 to 1.60•10-07), p = 0.363 1.28•10-08(−5.13•10-09 to 3.07•10-08), p = 0.161 4.95•10-06(−0.00002 to 0.00003), p = 0.743 9.73•10-06(−0.00002 to 0.00003), p = 0.445
Age at baseline in years
RC (95%CI), p-value
−0.008(−0.017 to 0.002), p = 0.116 0.0004(−0.0001 to 0.001), p = 0.144 0.00006(−0.00003 to 0.00015), p = 0.177 0.168 (0.022 to 0.314), p = 0.024 0.090 (−0.022 to 0.202), p = 0.113
Male sex
RC (95%CI), p-value
0.014 (−0.119 to 0.146), p = 0.840 0.001 (−0.007 to 0.008), p = 0.883 −0.0002 (−0.001 to 0.001), p = 0.734 2.177 (0.179 to 4.175), p = 0.033 −0.541 (-2.096 to 1.013), p = 0.494
Disease duration at baseline in years
RC (95%CI), p-value
−0.007 (−0.015 to −0.0003), p = 0.042 0.00049 (0.00009 to 0.00089), p = 0.017 0.00003(−0.00004 to 0.00009), p = 0.403 −0.039 (−0.145 to 0.068), p = 0.474 −0.042 (−0.120 to 0.036), p = 0.287
Clinical variable at baseline
RC (95%CI), p-value
1.027 (0.950 to 1.105), p < 0.001 0.860 (0.777 to 0.943), p < 0.001 0.961 (0.899 to 1.023), p < 0.001 0.720 (0.651 to 0.789), p < 0.001 0.843 (0.776 to 0.911), p < 0.000



SPACE-MS models of future disability&
Baseline SPACE-MS metrics (below)$
NCIRC (95%CI), p-value
Model R2 (%I)
1.114 (0.217 to 2.012), p = 0.015
0.6253 (0.8%)
−0.044 (−0.094 to 0.005), p = 0.079
0.5641 (0.6%)
0.001 (−0.008 to 0.009), p = 0.902
0.7528 (0%)
4.084 (-9.477 to 17.644), p = 0.554
0.6525 (0.1%)
−5.118 (-14.567 to 4.331), p = 0.287
0.8070 (0.1%)
Maximum lesion NCIRC (95%CI), p-value
Model R2 (%I)
0.453 (−0.033 to 0.939), p = 0.068
0.6232 (0.4%)
−0.023 (−0.051 to 0.005), p = 0.109
0.5635 (0.5%)
−0.004 (−0.008 to 0.001), p = 0.118
0.7542 (0.2%)
1.659 (-5.825 to 9.143), p = 0.6603
0.6523 (0.02%)
−8.669 (-14.040 to −3.299), p = 0.002
0.8122 (0.7%)
MCIRC (95%CI), p-value
Model R2 (%I)
−0.0002(−0.0008 to 0.0004), p = 0.503
0.6209 (0.1%)
0.00001(−0.00002 to 0.00004), p = 0.523
0.5611 (0.1%)
−5.39•10−07(-6.19•10−06 to 5.11•10−06), p = 0.851
0.7528 (0%)
0.004 (−0.006 to 0.013), p = 0.439
0.6527 (0.1%)
−0.001 (−0.008 to 0.005), p = 0.664
0.8064 (0.01%)
CAIRC (95%CI), p-value
Model R2 (%I)
−0.523 (-1.149 to 0.103), p = 0.101
0.6227 (0.4%)
0.017 (−0.019 to 0.052), p = 0.352
0.5616 (0.2%)
0.007 (0.001 to 0.013), p = 0.018
0.7560 (0.4%)
−0.188 (-9.802 to 9.427), p = 0.969
0.6522 (0%)
7.970 (1.237 to 14.703), p = 0.020
0.8095 (0.4%)
CPIRC (95%CI), p-value
Model R2 (%I)
−0.172 (−0.686 to 0.343), p = 0.513
0.6208 (0.1%)
0.023 (−0.006 to 0.052), p = 0.125
0.5633 (0.5%)
0.004 (−0.001 to 0.008), p = 0.123
0.7542 (0.2%)
−4.428 (-12.093 to 3.237), p = 0.257
0.6532 (0.2%)
3.198 (-2.489 to 8.885), p = 0.269
0.8071 (0.1%)
CSIRC (95%CI), p-value
Model R2 (%I)
0.498 (−0.057 to 1.054), p = 0.079
0.6230 (0.4%)
−0.028 (−0.060 to 0.004), p = 0.082
0.5640 (0.6%)
−0.008 (−0.013 to −0.003), p = 0.002
0.7585 (0.8%)
3.640 (-4.759 to 12.039), p = 0.395
0.6528 (0.1%)
−8.317 (-14.378 to −2.257), p = 0.007
0.8106 (0.5%)



Best SPACE-MS model of future disability
Best SPACE-MS metric(s) at baselineRC (95%CI), p-value
Model R2 (%I)
NCI: 1.114 (0.217 to 2.012), p = 0.015
0.6253 (0.8%)
NCI: −0.044(−0.094 to 0.005), p = 0.079
0.5641 (0.6%)
CSI: −0.008(−0.013 to−0.003), p = 0.002
0.7585 (0.8%)
CPI: −4.428(-12.093 to 3.237), p = 0.257
0.6532 (0.2%)
Maximum lesion NCI: −8.669 (-14.040 to −3.299), p = 0.002
0.8122 (0.7%)

The ‘best a priori model’ included, as covariates, baseline lesion volume, baseline WM volume, baseline GM volume, age in years at baseline, male sex, disease duration at baseline, centre, and clinical variable at baseline. Of note, the regression coefficients for ‘study centre’ are not shown (for simplicity). The ‘SPACE-MS models’ included all the variables of the ‘best a priori model’ plus one SPACE-MS metric at baseline at a time. See methods for full details. #: the EDSS score is measured in EDSS score units; the inverse of TWT and the inverse of 9HPT, in 1/s; and the PASAT and SDMT scores, in number of correct answers; &: all ‘SPACE-MS models’ included, as covariates, the variables included in the ‘best a priori models’. However, for simplicity, the RC (95% CI) of all these extra variables are not shown; $: all spatial distribution metrics are measured in dimensionless units except for MCI, which is measured in mm2. Significant results are highlighted. Abbreviations (in alphabetical order): %I: percentage of model improvement based on R2; 9HPT: nine-hole peg test; CAI: covariance anisotropy index; CI: Confidence Interval; CPI: covariance planarity index; CSI: covariance sphericity index; EDSS: expanded disability status scale; GM: grey matter; Max: maximum; MCI: mean covariance index; NAWM: normal-appearing white matter; NCI: neuraxis caudality index; PASAT: paced auditory serial addition test (measured in number of correct answers); RC: regression coefficient; SD: standard deviation; SDMT: symbol digit modalities test (measured in number of correct answers); SPACE-MS: spatial patterns (of MS lesions) assessed through covariance estimations of lesional voxels; TWT: 25-foot timed walk test.