Predicting Language Improvement in Acute Stroke Patients Presenting with Aphasia: A Multivariate Logistic Model Using Location-Weighted Atlas-Based Analysis of Admission CT Perfusion Scans

S. Payabvash; S. Kamalian; S. Fung; Y. Wang; J. Passanese; S. Kamalian; L.C.S. Souza; A. Kemmling; G.J. Harris; E.F. Halpern; R.G. González; K.L. Furie; M.H. Lev

doi:10.3174/ajnr.A2125

Article Figures & Data

Figures

Tables

Download figure
Open in new tab
Download powerpoint
Fig 1.
Of 119 patients with acute ischemic stroke who underwent CTP at our hospital between December 2006 and April 2008, 58 were included in this study.
Download figure
Open in new tab
Download powerpoint
Fig 2.
Upper row: orthogonal sections of the lower third of the sublobar insular ribbon. Middle row: orthogonal sections passing through the angular gyrus GM of BA 39. The mean relative rCBF values of these 2 regions were the only CTP parameters that were independent predictors of early aphasia improvement. Lower row: a 3D standard MNI-152 brain space reflecting the main cortical BAs directly involved in language function (labeled by corresponding number), as well as the lower third subinsular ribbon and angular gyrus GM, as noted above.
Download figure
Open in new tab
Download powerpoint
Fig 3.
A, The ROC curve for the forced-entry multiple logistic regression model derived from the admission CTP/CTA imaging and the clinically independent predictors of language improvement (Table 3). B, Boxplot graph shows how this model can distinguish between those patients with improvement of language function and those without; the y-axis represents the values of the regression equation (Table 3). Patients with aphasic with positive regression scores (>0) have a >50% probability of language improvement by discharge.

Tables

Figures

View popup

Table 1:

Comparison of clinical/CTA characteristics between patients with and without improvement of aphasia (univariate analysis)

	Aphasia Improvement (n = 21)	No Improvement (n = 37)	P Value
Age	69.5 ± 3.8	76.8 ± 1.9	.10
Female	11 (52%)	21 (57%)	.78
Admission NIHSS score	8 (2–23)	19 (5–28)	.001^a
Admission aphasia NIHSS score	2 (1–3)	3 (1–3)	.002^a
Major proximal arterial occlusion	7 (33%)	30 (81%)	.001^a
Thrombolytic therapy (IV)	5 (24%)	11 (29%)	.76
IA thrombectomy^b	1 (5%)	3 (8%)	1.00
Ictus to IV treatment interval (hour)	2.11 ± 0.25	1.93 ± 0.25	.67
IV treatment to CTP interval (hour)^c	1.13 ± 0.49	1.51 ± 0.52	.68
Ictus to CTP interval (hour)	4.46 ± 0.57	4.78 ± 0.48	.68

a Significant.
b Of the 4 patients who received thrombectomy, only 1, who had complete recanalization, subsequently showed language improvement. All patients who received thrombectomy had first received IV thrombolysis.
c Only 1 patient with and 4 without language improvement received IV thrombolysis following CTP; negative treatment-to-CTP-time values were calculated for these patients.

View popup

Table 2:

Details of the forced-entry multiple logistic regression model derived only from the admission CTP/CTA independent predictors of language improvement

	B	P Value	EXP(B)	95% CI for EXP(B)
rCBF of sublobar insular ribbon (lower third)	2.89	.01	16.76	1.87-149.80
rCBF of angular gyrus GM BA 39	1.76	.27	5.82	0.24-139.82
Presence of proximal cerebral artery occlusion	−1.46	.08	.23	0.04-1.21
Constant	−3.87	.05	0.21

View popup

Table 3:

Details of the forced-entry multiple logistic regression model derived from both the admission CTP/CTA imaging and the clinically independent predictors of language improvement

	B	P Value	EXP(B)	95% CI for EXP(B)
Admission aphasia score (NIHSS, section 9)	−1.21	.03	0.29	0.09-0.90
rCBF of sublobar insular ribbon (lower third)	2.23	.04	9.34	1.09-80.02
Presence of proximal cerebral artery occlusion	−1.94	.04	0.14	0.02-0.93
rCBF of angular gyrus GM BA 39	1.64	.37	5.16	0.14-191.01
Constant	−0.47	.84	0.62

View popup

Table 4:

Calculation of the 8-point aphasia improvement score^a

Variables		Points
Aphasia score on admission NIHSS examination		1–3
Proximal cerebral artery occlusion	Absent	0
on admission CTA	Present	2
rCBF of the sublobar insular ribbon	>1.5	−2
(lower third)	0.66–1.5	0
	0.34–0.66	1
	<0.34	2
rCBF of angular gyrus GM (BA 39)	>0.66	0
	≤0.66	1

a The aphasia outcome score is the sum of all points based on the admission language component of the NIHSS score, CTP (2 regions), and the presence or absence of CTA primal intracranial occlusion. Note that an rCBF > 1.5 corresponds to a region of marked hyperemia compared with the baseline normal cerebral blood flow in the rCBF of that region.

View popup

Table 5:

Predictive value of the aphasia improvement score

Degree of Language Improvement	Aphasia Improvement Score^a	No. of Patients	Probability of Improvement^b	No. of Patients with This Score Showing Improvement
Excellent	1 or 2	13	65%–99% (1 FP)	12 (93%)
Fair	3 or 4	20	18%–96% (2 FP, 2 FN)	9 (45%)
Poor	5 or 6	12	3%–32% (1 FN)	1 (12%)
Dismal	7 or 8	13	1%–6%	0 (0%)

a The aphasia improvement score is calculated as described in Table 4.
b The probability of improvement is estimated on the basis of the logistic regression equation of Table 3; FP and FN cases are defined on the basis of the results of a multivariate model.