Do All ASPECT Score Regions have the Same Predictive Power for Functional Outcomes?

https://doi.org/10.1016/j.jstrokecerebrovasdis.2019.104516Get rights and content

Abstract

Background and Purpose

ASPECTS (Alberta Stroke Program Early CT Score) is a 10-point topographic CT scan score that has been shown to be a strong prognostic factor in acute ischemic stroke. We investigated whether all ASPECTS regions have the same prognostic value.

Methods

Clinical characteristics, ASPECTS, and 3-month modified Rankin Scale (mRS) data were retrospectively collected in 350 patients who were diagnosed with middle cerebral artery (MCA) territory stroke. To describe the 3-month mRS data, an ordered categorical approach was applied using a proportional odds model. Furthermore, external validation was performed using additional data from 30 patients.

Results

As expected, ASPECTS was an independently important predictor. However, when 10 regions were analyzed separately, the M1, M2, and M3 regions, related to MCA cortex, were not found to predict 3-month mRS scores in the final model. The odds ratios for ischemic change in other regions (except M1, M2, and M3) ranged from 2.6 to 3.8. Moreover, among clinical characteristics, only age was identified as a significant predictor. The sensitivity and specificity of the final model in the external validation were 91% and 88%, respectively.

Conclusions

All ASPECTS regions did not have the same predictive power for functional outcomes, defined as the 3-month mRS. The implementation of a proportional odds model allowed a proper description of the ordered categorical nature of the mRS and the identification of relevant predictors.

Introduction

ASPECTS (Alberta Stroke Program Early CT Score) is a 10-point topographic CT scan score designed to predict outcomes after intravenous thrombolysis in patients with middle cerebral artery (MCA) stroke.1,2 The MCA territory is assessed based on axial CT sections through the ganglionic and supraganglionic levels. Axial sections obtained at the level of the caudate head or below are classified as being in the ganglionic level, while those above are classified as the supraganglionic. The template used in this study consisted of 10 anatomically defined regions, including 4 in subcortical structures (the caudate, lentiform nucleus, internal capsule, and insular ribbon) and 6 in cortical structures in the MCA territory, labeled M1-M6. Early ischemic changes on CT were defined as intraparenchymal hypoattenuation and focal swelling. For each ASPECTS region that showed early ischemic changes, the overall score of 10 was reduced by 1. ASPECTS has been applied to various imaging modalities, such as diffusion-weighted MRI (DWI), cerebral blood volume, CT angiography source images and noncontrast CT, since it was introduced in 2000.3

ASPECTS is widely used in clinical trials and practice to assess the extent of early ischemic damage on brain imaging to facilitate acute stroke treatment. Barber et al. reported that ASPECTS can be a significant predictor of symptomatic intracranial hemorrhage after intravenous thrombolysis,2 and another early study suggested that a cutoff score of 7 or below was a predictor of functional independence in patients with intravenous thrombolysis within 3 hours of symptom onset.4 ASPECTS has been proposed as a fast and easy method to identify patients suitable for mechanical thrombectomy because ASPECTS does not need time consuming post-processing of CT or MRI data. Several studies have reported that patients with higher ASPECTS received a greater benefit from endovascular treatment, and ASPECTS has been used for patient selection in large stent-retriever thrombectomy trials, which have demonstrated that compared with standard intravenous thrombolysis alone, thrombectomy using stent-retriever devices exerted a strong positive mechanical effect on functional outcomes.5, 6, 7

One study reported that the correlation between ASPECTS and DWI lesion volume varied depended on lesion location.8 Because individual regions of ASPECTS cover different amounts of brain tissue and lesion volume is strongly correlated with functional outcomes,9,10 each region may have an unequal weighing for functional outcomes. When decisions are made regarding endovascular treatment based only on an ASPECTS threshold without considering the lesion location, the different prognostic values of the regions based on the template's unequal weighing can compromise clinical decisions and may lead to the unjustified exclusion of patients from clinical trials or treatment.

Many studies have used predictive models to explore functional outcomes defined by the modified Rankin Scale (mRS) in patients with ischemic stroke. However, with few exceptions, binary logistic regression models have been extensively used, and in these, the mRS is treated as a binary variable.11 Because the mRS is an ordinal and not a binary variable (range, 0 to 6), inappropriate model selection can lead to poor predictions and misleading results. Among the several regression models available for ordinal variables,12 a proportional odds model is commonly used.

Here, we developed a proportional odds model for functional outcomes to investigate whether all ASPECTS regions have the same prognostic value.

Section snippets

Patients

Consecutive patients over 5 years, who were diagnosed with MCA territory stroke and treated with intravenous thrombolysis and/or endovascular treatment, were retrospectively analyzed with regard to baseline demographic characteristics and functional outcomes. All patients underwent a baseline brain CT scan immediately when they reached the emergency department. Patients were excluded from the present study if they had transient ischemic stroke symptoms with negative imaging findings, had a

Results

The present study includes data from a model development cohort of 350 patients. The baseline characteristics of the model development and validation cohorts are presented in Table 1. The mean age of the patients was 63.8 years old, and 211 (60.3%) of patients were male. In all, 56.3% and 22.0% of the patients had a history of hypertension and diabetes mellitus, respectively. The prevalence of atrial fibrillation was 27.7%, and 13.1% of the patients had experienced a previous stroke. The median

Discussion and Summary

In the present study, we investigated whether the effect of ischemic changes on mRS scores at 3 months was equal across all ASPECTS regions in patients with ischemic stroke, and determined which predictors were associated with the mRS. Thus, a proportional odds model was developed, and the final model developed here provided an adequate fit for the data.

Our study showed that all ASPECTS regions did not have the same prognostic value for predicting mRS at 3 months and that M1, M2, and M3 were

Conflict of Interest

All authors declare that they have no conflict of interest.

References (30)

  • A.J. Yoo et al.

    Infarct volume is a pivotal biomarker after intra-arterial stroke therapy

    Stroke

    (2012)
  • S.F. Zaidi et al.

    Final infarct volume is a stronger predictor of outcome than recanalization in patients with proximal middle cerebral artery occlusion treated with endovascular therapy

    Stroke

    (2012)
  • H.S. Lim et al.

    Quantitative predictive models for the degree of disability after acute ischemic stroke

    J Clin Pharmacol

    (2018)
  • C.V. Ananth et al.

    Regression models for ordinal responses: a review of methods and applications

    Int J Epidemiol

    (1997)
  • R.H.B. Christensen

    Ordinal - regression models for ordinal data

    R Package Version

    (2019)
  • Cited by (9)

    • An automated ASPECTS method with atlas-based segmentation

      2021, Computer Methods and Programs in Biomedicine
      Citation Excerpt :

      However, the same score may be given by 2 raters attributing changes to different regions [14]. Different ASPECTS regions have different predictive power for prognosis [26]. The reproducing of ASPECT greatly limits its clinical value.

    View all citing articles on Scopus

    Funding: This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: HI17C1919).

    View full text