# A comparative analysis of TOAST and ASCOD criteria in etiologic subtyping of acute ischemic stroke at a tertiary hospital in Tanzania

## TL;DR

This study compares two methods for classifying the causes of ischemic stroke in Tanzania and finds they agree well, except in some cases.

## Contribution

The study evaluates the agreement between TOAST and ASCOD criteria for stroke subtyping in an East African setting.

## Key findings

- TOAST and ASCOD1 showed good to very good agreement in most stroke subtypes.
- TOAST and ASCOD1 failed to determine a definitive etiology in 34.6% and 48.5% of strokes, respectively.
- Discordance was observed in the cardioembolic subtype when comparing TOAST and ASCOD1,2.

## Abstract

Etiologic subtyping of ischemic stroke is crucial for determining its treatment, prognosis, and prevention. However, data on the widely utilized TOAST and ASCOD criteria remain scarce in the East African region.

The study aimed to compare the performance of the TOAST and ASCOD systems in subtyping ischemic stroke among stroke patients at a tertiary hospital in Tanzania.

This was an institutional cross-sectional study. All adults (≥18 years) admitted with a diagnosis of stroke over a six-year period were selected from the registry, and their clinical details reviewed retrospectively. One hundred and thirty (130) patients with first- or second-time acute stroke (as defined by the World Health Organization) were included. Acute stroke was confirmed as ischemic by magnetic resonance imaging. For each index stroke, TOAST and ASCOD criteria were applied. The discordance and level of agreement between the approaches were assessed using McNemar's test χ2 (P-value) and Cohen's kappa coefficient (κ), respectively. The value of κ was interpreted as moderate (0.41–0.6), good (0.61–0.8), very good (0.81–0.9), or excellent (0.91–1.0). Statistical significance was set at P < 0.05.

There was no significant discordance between TOAST and the grade 1 level of evidence of ASCOD (ASCOD1) in assigning stroke to all subtypes, except for undetermined etiology χ2 (P = 0.023). Agreement between these systems was good to very good (κ = 0.601 to 0.843, P < 0.01) across the subtypes. TOAST and ASCOD1 failed to determine a definitive etiology in 34.6% and 48.5% of strokes, respectively. On comparing TOAST vs. combined grade of evidence 1 and 2 of ASCOD (ASCOD1,2), there was a discordance in allocation of strokes to the cardioembolic subtype χ2 (P < 0.001), and agreement was moderate (κ = 0.471, P = 0.001). However, the agreement across other identified subtypes was good to very good (κ = 0.601 to 0.875, P ≤ 0.001).

There was a good to very good agreement between TOAST and ASCOD1 in etiologic subtyping of ischemic stroke. Further research is warranted to evaluate their consistency across diverse local settings and to explore factors influencing their performance.