Computed tomography perfusion as a predictor of delayed cerebral ischemia and functional outcome in spontaneous subarachnoid hemorrhage: A single center experience

Abstract

Background

Computed tomography (CT) perfusion has been studied as a tool to predict delayed cerebral ischemia (DCI) and clinical outcome in spontaneous subarachnoid hemorrhage (SAH). The purpose of the study was to determine whether quantitative CT perfusion performed within 72 hours after admission can predict the occurrence of DCI and clinical outcome as measured with a modified Rankin scale (mRS) at 3 months after ictus.

Methods

Cerebral perfusion was assessed in a prospective cohort of patients with acute SAH. CT perfusion parameters at <72 h post SAH were quantitatively measured in the main vascular territories and represented as whole-brain means. Spearman rank correlation coefficient and generalized additive regression models for binary outcome were used.

Results

A total of 66 patients underwent CT perfusion at <72 h. Poor clinical grade on admission was correlated with worse cerebral perfusion in all parameters. Multivariable analysis yielded an association of time to peak (TTP; odds ratio (OR) = 0.89; 95% confidence interval (CI): 0.77, 1.02; p = 0.083) with the occurrence of DCI. We also found an association of TTP values with poor outcome, with an 8% increase in the odds of mRS > 3 for each one second increase in TTP at admission (OR = 1.08; 95% CI: 1.00, 1.17; p = 0.061).

Conclusions

We identified an association of early TTP changes with DCI and poor clinical outcome. However, there were no associations with cerebral blood flow or mean transit time and DCI/clinical outcome. CT perfusion still remains to be validated as a tool in predicting outcome in SAH.

Introduction

Subarachnoid hemorrhage (SAH), usually caused by rupture of intracranial aneurysms, is a devastating clinical condition, with high mortality and morbidity, that reaches 50% and 20%, respectively.^1,2 Vasospasm and delayed cerebral ischemia (DCI) are the most frequent complications, negatively affecting clinical prognosis in 20–30% of patients^1,3–6 (Figure 1). In addition to these known complications, the occurrence of early brain injury at the time of hemorrhage is emerging as a cause of mortality,² and a contributor to DCI,⁷since it is now accepted that vasospasm alone cannot fully account for DCI development.⁸

Figure 1.

Imaging studies of a 38-year-old patient with acute SAH after rupture of a basilar tip aneurysm. (a) Admission CT showing diffuse SAH, Fisher grade 3. (b) Digital subtraction angiography showing a basilar tip saccular aneurysm. (c) CT Perfusion map of cerebral blood flow at admission showed no perfusion deficits. (d) Delayed ischemia at 9 days, affecting the posterior cerebral artery territory bilaterally.

CT: computed tomography; SAH: subarachnoid hemorrhage.

Cerebral perfusion is affected in patients with SAH, early post hemorrhage,^9–12 and also later, during the period of vasospasm.^12–14 However, it is known that vasospasm, as detected using measurements of blood flow velocity with transcranial Doppler ultrasonography (widely used as a bedside monitoring tool in patients with SAH) does not always correlate with cerebral blood flow reduction.¹⁵ Therefore, the interest in perfusion imaging studies has grown, to better identify patients with global and local cerebral blood flow deficits in patients with vasospasm and DCI.

There are several studies suggesting that early computed tomography (CT) perfusion might be useful also in predicting the occurrence of vasospasm and DCI, as soon as the first hours post ictus.^{13,14,16–21} Mean transit time (MTT), time to peak (TTP) and cerebral blood flow (CBF) were suggested as markers of vasospasm,²² DCI^10,22 and clinical outcome in SAH,²³ being the most commonly analyzed parameters. However, it remains to be proved whether early CT perfusion can be used as a prognostic tool when performed in the acute SAH phase. In a meta-analysis including 345 patients,¹⁴ a 23-fold increased probability of DCI was described in patients with CT perfusion changes measured either before or at the onset of symptoms of vasospasm. In another meta-analysis published on the same year¹⁶ including 11 studies and 570 patients early CT perfusion findings, before the time window for vasospasm, were not associated to the occurrence of DCI. In fact, perfusion changes were only consistently found at the time of neurological deterioration, in patients with DCI.

Association of CT perfusion in the first 10 days after SAH to long-term clinical prognosis has been studied by Mathys and colleagues,²³ who reported an MTT threshold above 4 seconds to predict an unfavorable outcome at 23 months. One other study reported an association of early elevated MTT together with a higher blood burden on admission CT with poor clinical outcome at 6 weeks.¹⁸

We decided to conduct a prospective study to investigate whether early (<72 h) quantitative CT perfusion could be useful in predicting DCI and clinical outcome at 3 months.

Material and methods

Sample

All patients with acute spontaneous SAH admitted at the Centro Hospitalar de Lisboa Central between May 2013 and November 2014 were enrolled in a prospective cohort study. Institutional review board approval was obtained. Inclusion criteria were as follows: (1) age > 18 years; (2) acute non-traumatic SAH diagnosed by CT and/or lumbar puncture (3); imaging studies performed within 72 hours after SAH; and (4) informed consent obtained from the patient or legal representative. Patients in a very poor clinical condition (Glasgow coma scale; GCS 3), pregnant women, patients with renal insufficiency, and patients whose time of onset of SAH was unknown were excluded.

Clinical and imaging data

For all included patients we recorded age, sex, clinical status at admission (according to the GCS, World Federation of Neurosurgeons scale (WFNS) and Hunt and Hess scale (HH)), the amount of blood on the admission brain CT scan (according to the modified Fisher scale²⁴ and the Hijdra scale²⁵), aneurysm location and treatment.

Occurrence of DCI and other relevant complications such as the presence of hydrocephalus, rebleed or vasospasm on transcranial Doppler ultrasonography were recorded.

Hydrocephalus was defined as a bicaudate index above the 95th percentile for age, occurring at any time between patient admission and discharge.²⁶ The presence of cerebral vasospasm was defined by a Lindegaard index²⁷ calculated as the ratio of the velocity measured at the middle cerebral artery and at the distal internal carotid artery above 3 (Lindegaard index 3–5: moderate vasospasm; Lindegaard index > 5: severe vasospasm) in transcranial Doppler ultrasonography studies after the 4th day after SAH.

Patients were classified as having DCI if: (1) presenting with a new focal neurological deficit/decrease in the level of consciousness non-attributable to other causes (for example, hydrocephalus, seizures, metabolic derangement, infection, or sedation); (2) presenting a new infarct on follow-up CT /magnetic resonance (MR) imaging; or (3) both 1 and 2, after 4 days after ictus.^28,29 New infarcts were either defined by areas with restricted diffusion on diffusion weighted imaging or defined hypodensities on CT, diagnosed by one staff neuroradiologist, and verified by another neuroradiologist (IF) blinded to clinical data.

CT perfusion protocol

CT perfusion studies were performed within 72 hours after SAH on the same 64-multidetector CT scanner (Lightspeed VCT; GE Healthcare, Milwaukee, WI, USA). For the CT perfusion scan, a bolus of 40 ml of nonionic contrast agent (Ultravist 300 mg iodine/ml, Schering, Berlin, Germany) was injected intravenously at a rate of 4 ml/s, followed by a 40 ml saline flush at the same rate, using a dual power injector. The following parameters were used: 80 kVp, 400 mAs, 64 × 0.625 collimation, 520 × 520 matrix, field of view (FOV) 250 mm, 272 images acquisition in 46 seconds.

Post processing

A CT perfusion post-processing quantitative evaluation was performed by a neuroradiologist (IF) with 10 years' experience, using OLEA software (La Ciotat, France). The arterial input function and the venous output function were manually defined at the A2 segment of the anterior cerebral artery, and at the superior sagittal sinus, respectively. Hemodynamic maps were calculated using the block circulant singular value decomposition deconvolution method.³⁰ Regions of interest (ROIs) were manually selected bilaterally and symmetrically covering arterial territories of the anterior, middle and posterior cerebral artery, and at the midpons and subcortical cerebellar hemispheres (Figure 2). For each ROI, mean quantitative values of CBF, cerebral blood volume (CBV), MTT, Tmax and TTP were obtained. Global cerebral perfusion was calculated using the mean of all ROIs for each parameter individually. Measurements avoided areas of parenchymal hematoma and ventricular drainage trajectory.

Figure 2.

Regions of interest drawn to include infratentorial vertebro-basilar territories at the pons and subcortical cerebellum (a) and cortical and deep territories of the anterior, middle and posterior cerebral arteries (b–d).

Outcome

We aimed to evaluate the association of CT perfusion parameters with: (1) occurrence of DCI and (2) functional prognosis at 3 months, assessed by the modified Rankin scale (mRS).³¹

The mRS was performed by two physicians with experience in stroke (APN, PF), blinded to clinical, CT perfusion and MR imaging data, by in-person or telephone interview.³² The outcome was dichotomized as good, if the patient remained independent 3 months after SAH (mRS ≤ 3), and poor, while remained dependent or died (mRS 3–6).

Statistical analysis

An exploratory analysis was carried out for all variables. Categorical data were presented as frequencies (percentages), and continuous variables as mean and standard deviation (SD) or median and range (minimum–maximum), as appropriate. To study the correlation between perfusion parameters at admission and severity of SAH (GCS, HH, WFNS and Fisher scales), Spearman rank correlation coefficient (r_S) was used.

To study the association between CT perfusion parameters and DCI, logistic regression models were used. To study the association between CT perfusion parameters and functional outcome, generalized additive regression models were applied due to the violation of the linearity assumption in logit regarding age (modeled with splines). The following variables were considered in univariable analyses: age, sex, clinical grade at admission (GCS grade, WFNS grade, HH grade), amount of blood on admission CT (modified Fisher scale score, cisternal Hijdra score, ventricular Hijdra score), SAH location (perimesencephalic versus non-perimesencephalic), presence of aneurysm (yes/no), hydrocephalus (yes/no), treatment (non-aneurysmal SAH, clipped aneurysm and coiled aneurysm), occurrence of vasospasm, and mean CT perfusion parameters; variables that attained a p value ≤ 0.25 were considered for multivariable analysis.

Discriminative ability of these models was assessed by the area under the receiver-operating characteristic curve (AUC), and refers to how well the model differentiates those with the outcome from those without the outcome. A level of significance α = 0.05 was considered. Data were analyzed using the Statistical Package for the Social Science for Windows, (SPSS, version 22.0, released 2013; IBM Corp., Armonk, NY, USA) and R software (a language and environment for statistical computing, R Core Team, R Foundation for Statistical Computing, Vienna, Austria, http://www.R-project.org.).

Results

Study sample

A total of 129 patients with spontaneous non-traumatic SAH were admitted at our tertiary center during the inclusion period. Overall, 80 patients fulfilled the inclusion criteria for the prospective cohort. For the current analysis, we selected 66 patients that performed CT perfusion within 72 h after SAH.

Patients characteristics

Demographic characteristics of patients are depicted in Table 1. A total of 15 patients (22.7%) developed DCI. At 3 months, 46 patients (69.7%) were independent (mRS < 3), and 5 patients (7.6%) died. No adverse events related to CT perfusion were reported.

Table 1.

Baseline characteristics of patients.

	Patients (n = 66)
Female, n (%)	38 (57.6)
Age (median, range)	57.5 (34–86)
WFNS score, n (%)
I	39 (59.1)
II	14 (21.2)
III	1 (1.5)
IV	9 (13.6)
V	3 (4.5)
Hunt and Hess grade, n (%)
1	25 (37.9)
2	17 (25.8)
3	14 (21.2)
4	6 (9.1)
5	4 (6.1)
GCS score (median, range)	15 (6–15)
Amount of blood: Modified Fisher Scale grade, n (%)
2	6 (9.1)
3	14 (21.2)
4	46 (69.7)
Amount of blood: Hijdra score (median, range)
Cisternal score	13.5 (0–28)
Ventricular score	2 (0–10)
Hydrocephalus, n (%)	8 (12.1)
Aneurysm location, n (%)
Anterior circulation	38 (57.6)
Posterior circulation	8 (12.1)
Perimesencephalic hemorrhage	16 (24.2)
Non-aneurysmal non-perimesencephalic	4 (6.1)
Aneurysm treatment, n (%)
Coiling	34 (26.1)
Clipping	14 (30.4)

GCS: Glasgow coma scale; WFNS: World Federation of Neurosurgeons scale.

Correlation of perfusion parameters to SAH severity

Some perfusion parameters were significantly correlated with clinical scores at admission: CBF was positively correlated with GCS (r_S = 0.342, p = 0.005), and negatively correlated with HH grade (r_S =− 0.354, p = 0.004) and WFNS score (r_S = −0.288, p = 0.019). CBV was positively correlated with GCS (r_S = 0.319, p = 0.009), and negatively correlated with HH grade (r_S = −0.275, p = 0.025) and WFNS score (r_S = −0.273, p = 0.026). MTT was negatively correlated with GCS (r_S = −0.350, p = 0.004), and positively correlated with HH grade (r_S = 0.379, p = 0.002) and WFNS score (r_S = 0.345, p = 0.005) scores. The remaining perfusion parameters were not correlated with SAH severity scores at admission. Moreover, no significant correlations were found between any of the perfusion parameters and the amount of blood on the CT scan (modified Fisher score; Table 2).

Table 2.

Correlation coefficient estimates between perfusion parameters and SAH severity assessed by clinical scales and amount of blood in brain CT at admission.

Perfusion parameter	GCS score		HH grade		WFNS score		Modified Fisher grade
	rS	p	rS	p	rS	p	rS	p
CBF	0.342	0.005	−0.354	0.004	−0.288	0.019	−0.071	0.572
CBV	0.319	0.009	−0.275	0.025	−0.273	0.026	−0.097	0.440
MTT	−0.350	0.004	0.379	0.002	0.345	0.005	0.081	0.519
TTP	0.059	0.639	0.048	0.700	0.029	0.820	−0.040	0.751
Tmax	0.084	0.503	−0.159	0.204	−0.053	0.674	0.147	0.237

CBF: cerebral blood flow; CBV: cerebral blood volume; GCS: Glasgow coma scale; HH: Hunt and Hess scale; MTT: mean transit time; r_SL Spearman's coefficient; SAH: subarachnoid hemorrhage; TTP: time to peak; Tmax: time to maximum of the tissue residue function; WFNS: World Federation of Neurosurgeons scale.

p values were obtained by Spearman rank correlation coefficient statistical significance test.

Early CT perfusion (CTP) perfusion parameters as indicators of DCI

The univariable analysis yielded the amount of cisternal blood on CT scan (cisternal Hijdra score), SAH location, endovascular treatment of the aneurysm, occurrence of vasospasm, and TTP as candidates for the multivariable model (Table 3).

Table 3.

Association of clinical variables and perfusion parameters with delayed cerebral ischemia (univariable regression analysis).

	No DCI (n = 51)	DCI (n = 15)	OR estimate (95% CI)	P value
Age (years)	58 (34–86)	56 (36–84)	0.95 (0.96, 1.03)	0.538
Male sex	21 (41.2)	7 (46.7)	1.25 (0.39, 3.98)	0.706
GCS	15 (6–15)	14 (8–15)	1.06 (0.82, 1.36)	0.657
WFNS score	1 (1–5)	2 (1–4)	1.03 (0.65, 1.63)	0.906
Hunt and Hess grade	2 (1–5)	2 (1–5)	1.06 (0.66, 1.70)	0.800
Modified Fisher scale grade	4 (2–4)	4 (3–4)	1.68 (0.61, 4.61)	0.313
Cisternal blood Hijdra score	12 (0–26)	17 (2–28)	1.12 (1.02, 1.23)	0.013
Ventricular blood Hijdra score	2 (0–10)	2 (0–7)	1.04 (0.83, 1.31)	0.742
SAH location Perimesencephalic	15 (31.9)	1 (6.7)	0.15 (0.02, 1.27)	0.082
Hydrocephalus*	8 (15.7)	0 (0)	–	–
Treatment
Clipped aneurysm	9 (17.6)	3 (23.1)	6.33 (0.58, 69.68)	0.131
Coiled aneurysm	23 (45.1)	9 (69.2)	7.44 (0.86, 64.05)	0.068
Vasospasm	11 (21.6)	7 (46.7)	3.18 (0.95, 10.72)	0.062
CBF (ml/100g/min)	24.0 (8.6)	26.5 (7.4)	1.04 (0.97, 1.11)	0.302
CBV (ml/100g)	2.9 (1.0)	3.1 (0.9)	1.29 (0.73, 2.28)	0.374
MTT (s)	6.9 (1.3)	6.7 (0.8)	0.90 (0.57, 1.42)	0.650
TTP (s)	24.1 (12.3)	19.5 (3.8)	0.90 (0.79, 1.02)	0.096
Tmax (s)	6.8 (4.9)	7.0 (2.6)	1.01 (0.89, 1.14)	0.915

*Computational problems emerging from a quasi-complete separation problem due to small sample size (DCI is perfectly determined by hydrocephalus); values are expressed as median and range (minimum–maximum), mean (SD) or n (%); p values were obtained by logistic regression models.

Reference categories: female sex, SAH location: non-perimesencephalic SAH, treatment: non-aneurysmal SAH/no treatment.

CBF: cerebral blood flow; CBV: cerebral blood volume; CI: confidence interval; DCI: delayed cerebral ischemia; GCS: Glasgow coma scale; MTT: mean transit time; OR: odds ratio; SAH: subarachnoid hemorrhage; TTP: time to peak; Tmax: time to maximum of the tissue residue function; WFNS: World Federation of Neurosurgeons scale.

We obtained a multivariable model that included TTP (OR = 0.89; 95% CI: 0.77, 1.02; p = 0.083) and cisternal Hijdra score (OR = 1.13; 95% CI: 1.03, 1.23; p = 0.012), showing an independent association of TTP values with the occurrence of DCI. Therefore, for each increase in 1 second in TTP, a decrease of 11% in the odds of DCI was found. Also, for each unit increase of cisternal Hijdra score there was a 13% increase in the odds of developing DCI. The ROC curve achieved an AUC = 0.76 (95% CI: 0.63–0.90; fair discriminative ability between patients with and without DCI).

Early CTP parameters as indicators of poor functional outcome

The univariable analysis showed a significant association of several clinical variables with mRS ≥ 3, of which the following were chosen as candidates in the multivariable model: GCS, WFNS score, HH grade, modified Fisher grade, cisternal Hijdra score, ventricular Hijdra score, SAH location, presence of hydrocephalus, and endovascular treatment of the aneurysm. Perfusion parameters selected in the univariable analysis included MTT and TTP (Table 4). Additionally, a non-linear association of age with poor outcome was identified and consequently modeled with splines through a generalized additive model (Figure 3).

Table 4.

Association of clinical variables and perfusion parameters with poor clinical outcome (mRS ≥ 3; univariable regression analysis).

	mRS < 3 (n = 46)	mRS ≥ 3 (n = 20)	OR estimate (95% CI)	p value
Age (years)	52 (35–84)	69 (34–86)	1.05 (1.01, 1.10)	0.012
Male sex	20 (43.5)	8 (40.0)	0.83 (0.29, 2.43)	0.739
GCS score	15 (6–15)	14 (6–15)	0.64 (0.49, 0.85)	0.002
WFNS score	1 (1–5)	3 (1–5)	2.70(1.59, 4.58)	<0.001
Hunt and Hess grade	2 (1–5)	3 (1–5)	2.45 (1.44, 4.15)	0.001
Modified Fisher scale grade	4 (2–4)	4 (3–4)	5.26 (1.26, 22.02)	0.023
Cisternal blood Hijdra scale	14 (2–26)	15.5 (0–28)	1.09 (1.01, 1.18)	0.032
Ventricular blood Hijdra scale	2 (0–6)	5 (0–10)	1.53 (1.17, 2.01)	0.002
SAH location Perimesencephalic	15 (32.6)	1 (5.0)	0.12 (0.02, 1.00)	0.050
Hydrocephalus	2 (4.3)	6 (30.0)	9.21 (1.67, 50.95)	0.011
Delayed cerebral ischemia	12 (26.7)	3 (15)	0.49 (0.12, 1.96)	0.309
Treatment
Clipped aneurysm	8 (18.2)	3 (15.8)	1.50 (0.27, 8.38)	0.644
Coiled aneurysm	20 (45.5)	12 (63.2)	2.40 (0.65, 8.88)	0.190
Vasospasm	13 (28.9)	4 (20.0)	0.62 (0.17, 2.19)	0.454
CBF (ml/100g/min)	24.6 (7.7)	24.4 (9.9)	1.00 (0.94, 1.07)	0.981
CBV (ml/100g)	2.9 (0.8)	3.0 (1.4)	1.17 (0.68, 2.00)	0.578
MTT (s)	6.6 (1.3)	7.4 (1.0)	2.09 (1.15, 3.78)	0.015
TTP (s)	21.5 (5.6)	26.6 (18.1)	1.05 (0.98, 1.11)	0.171
Tmax (s)	6.7 (2.5)	7.2 (7.4)	1.02 (0.92,1.15)	0.670

Values are expressed as median and range (minimum–maximum), mean (SD) or n (%); p values were obtained by logistic regression models.

Reference categories: female gender, SAH location: non-perimesencephalic SAH, Treatment: non-aneurysmal SAH/no treatment.

CBF: cerebral blood flow; CBV: cerebral blood volume; GCS: Glasgow coma scale; mRS: modified Rankin scale; MTT: mean transit time; SAH: subarachnoid hemorrhage; Tmax: time to maximum of the tissue residue function; TTP: time to peak; WFNS: World Federation of Neurosurgeons scale.

Figure 3.

Influence of age in mRS outcome at 3 months (black curve) and corresponding 95% confidence intervals (dashed lines), showing a non-linear association.

mRS: modified Rankin scale.

We obtained a multivariable model with TTP (OR = 1.08; 95% CI: 1.00, 1.17; p = 0.061), WFNS score (OR = 9.86; 95% CI: 2.04, 47.68; p = 0.004) and ventricular Hijdra score (OR = 3.57; 95% CI: 1.28, 9.98; p = 0.015). An association of TTP values with a poor outcome was found, with an 8% increase in the odds of having mRS ≥ 3 at 3 months for each 1 second increase in TTP at admission. Also, for each unit increase of ventricular Hijdra scale there was almost a four-fold increase in the odds of a poor outcome. A higher WFNS score was strongly associated with a poor outcome and for each increase of one unit in the WFNS, there was a ten-fold increase in the odds of a poor outcome. The ROC curve achieved an AUC = 0.97 (95% CI: 0.94–1.00) indicating an excellent discriminative ability between patients with and without a poor outcome.

Discussion

In this exploratory study, we assessed cerebral perfusion parameters in the first 72 h after SAH and their relation to DCI and clinical outcome at 3 months. We found an association between TTP values and the occurrence of DCI and a poor clinical outcome. We confirmed previous reports showing that worse cerebral perfusion in all parameters correlated with SAH severity at admission,^19,33 although we did not find any correlation between cerebral perfusion parameters and the amount of blood on CT, as other authors did.³⁴

Cerebral perfusion decreases acutely after SAH,²⁰ partly resulting from the elevated intracranial pressure, combined with impaired autoregulation and acute vasospasm.^35,36 From day 4 after SAH and for an approximate 2-week time window, delayed vasospasm occurs in up to 70% of patients,³⁷ leading to a persistent reduction in arterial diameter, and ultimately to a sustained reduction in cerebral perfusion.³⁸

Several studies carried out in the early phase of SAH have suggested that changes in cerebral perfusion can predict the clinical course and the outcome. However, many of those studies report heterogeneous findings, and many of them did not assess the functional outcome adjusting for variables known to influence the prognosis.

Most of the recent CT perfusion studies aimed at predicting DCI or vasospasm.

Some studies that used angiographic vasospasm as a measure of outcome found an association of reduced CBF and increased MTT with the occurrence of vasospasm.^13,39 In other studies however, only CBF and CBV were significantly lower in patients with vasospasm.⁴⁰ In our study, there were no significant differences in cerebral perfusion parameters between patients with and without vasospasm on Doppler (results not shown).

DCI has been preferred as an outcome measure in most studies on SAH, since it has been shown to have strong association to clinical outcomes.²⁸ Some authors have reported a reduction of CBF and increase in MTT, to be strongly predictive of DCI.^10,41–43 Lagares and colleagues³⁴ found that an increase in MTT at admission above 5.9 s increased the odds of a poor outcome by 20-fold, with a positive predictive value of 100% for DCI. However, in a meta-analysis on CT perfusion as a predictor of DCI, Cremers et al.¹⁶ concluded that none of the studies that quantified perfusion parameters at admission found significant differences in any parameter that could predict patients that later developed DCI. Only one study using semi-quantitative methods found a visual asymmetry in perfusion parameters to be predictive of DCI.⁴⁴ A recent study also failed to associate CT perfusion to DCI, but found an association of lower perfusion to lower neurological grade on admission.⁴⁵

Although we would expect to find higher TTP values in patients who ultimately develop DCI, we found a paradoxical decrease of TTP in those patients in the early phase of SAH, albeit not significant. One explanation could be a false positive finding, and a small number of patients with DCI in our sample; another explanation could be the presence of comorbidities, such as cardiac failure, that could affect the velocity of arrival of the contrast, although we have no data to support this. TTP is a simple to extract parameter, that represents the local arrival of the contrast bolus, and might be affected by hemodynamic conditions or even by the position of the venous catheter for the contrast injection, with high inter-individual and intra-individual variation.⁴⁶ Pham and colleagues⁴⁷ found TTP to be the only predictor of DCI, but other studies found no association of TTP values with DCI.³⁹ Pathophysiologic mechanisms in DCI are related to multiple other factors unrelated to blood flow and perfusion, and this could explain the lack of association between the remaining perfusion parameters and DCI in our sample.

Finally, regarding clinical outcome, although there are no available meta-analyses, previous studies indicate that a delay in tissue perfusion measured by elevation of MTT, in the first days after SAH, may signal patients with poor prognosis, but is probably not related to the occurrence of DCI.^18,23 Our results are in line with these other studies: we found that elevated TTP was an independent predictor of poor clinical outcome, and higher MTT values were also significantly associated with poor outcome, but only in the univariable analysis.

What stands out from all the studies and is highlighted in one meta-analysis¹⁶ is that CT perfusion is not reliable as a tool to predict DCI and outcome, in part because results are not always reproducible between different centers. Major limitations of these analyses are CT perfusion methodology and post-processing algorithms, which differ between centers, and the difficulty in finding validated thresholds for CT perfusion parameters, thus limiting global applicability of the technique. Our study further illustrates the limitations of generalizing CT perfusion as a tool for prognosis prediction: cutoff points of MTT = 5.9 s were reported to have high predictive value for DCI and poor outcome in one study, and MTT > 4.1 s was defined as a threshold for an unfavorable outcome in another study.²³ However, the lowest mean MTT in our sample was 6.7 s, which is significantly higher than the above thresholds, and still, not predictive of neither DCI nor poor outcome. Therefore, caution should to be taken when including CT perfusion results in clinical practice, especially in decision algorithms for vasospasm treatment.

Our study has limitations. Our sample, with only 15 patients developing DCI, is underpowered to evaluate a predictive effect of CT perfusion. However, this is an exploratory study, where we tried to apply CT perfusion as a predictor of outcome to our population. Some patients of the main cohort of 80 patients were excluded from this analysis because CT perfusion was not performed before 72 h, which may introduce a selection bias towards patients with better clinical status. Also, some low-grade patients were not stable enough to perform CT perfusion, contributing to this selection bias towards patients with better neurological status. As already discussed, reproducibility of CT perfusion quantification may vary according to post-processing software. In our study, all images were obtained in the same 64-slice scanner, and post processing was done using the same software, therefore intra-study reproducibility was not affected. We used global cerebral means of perfusion parameters, including all ROIs, which may have led to underestimation of perfusion deficits in specific regions. In addition, by including ROIs drawn in posterior fossa territories, where vasospasm and DCI are less often encountered, we may have underestimated hypoperfusion in patients who posteriorly developed DCI. Other factors that might influence CT perfusion parameters, such as body height and weight, pulse rate, atherosclerosis and cardiac function were not included in the analysis.

Conclusion

In this prospective study on CT perfusion on the first 72 h after SAH, we found an association between early cerebral hypoperfusion and poor clinical grade on admission. We were not able to reproduce previous findings of association of MTT and CBF to the occurrence of DCI. However, we found an increase in TTP in the first 72 h to be independently associated with poor clinical outcome at 3 months. The use of CT perfusion as a tool for predicting outcome in SAH still lacks validation, since results may not be reproducible in different settings.

Conflict of interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

Dr Fragata was supported by a grant of the Sociedade Portuguesa de AVC (SPAVC) sponsored by Tecnifar.