Table 2.
Predictors of parenchymal hematoma: multivariate analysis.
| PH at follow-up MRI | OR (95% CI); p | OR (95% CI); p |
|---|---|---|
| Model A | Model B | |
| VLCBV regions | 2.816 (1.146–6.918), p = 0.024 | 1.592 (1.133–2.237), p = 0.007 |
| Pretreatment glucose (per IQR) | 1.049 (0.755–1.456), p = 0.777 | 1.072 (0.771–1.491), p = 0.678 |
| Baseline NIHSS (per IQR) | 1.337 (0.948–1.887), p = 0.098 | 1.260 (0.885–1.795), p = 0.200 |
| Rescue MT (vs primary) | 1.395 (0.668–2.916), p = 0.376 | 1.415 (0.674–2.971), p = 0.359 |
| Recanalization status | ||
| > 4′5 h from stroke onset (vs < 4′5 h) | 6.656 (2.444–18.130), p < 0.001 | 7.552 (2.745–20.779), p < 0.001 |
| No recanalization (vs < 4′5 h) | 2.848 (0.867–9.351), p = 0.084 | 2.810 (0.865–9.132), p = 0.086 |
| Cardioembolic etiology (vs no) | 2.584 (1.197–5.577), p = 0.016 | 2.812 (1.285–6.153), p = 0.010 |
| Sex (females vs males) | 0.500 (0.228–1.097), p = 0.084 | 0.522 (0.237–1.150), p = 0.107 |
| Good collaterals (vs poor) | 1.256 (0.550–2.867), p = 0.588 | 1.369 (0.599–3.129), p = 0.456 |
Very low cerebral blood volume (VLCBV) was included as a dichotomic variable (yes vs no) in model A, and as a continuous quantitative variable (estimations per IQR of VLCBV increase) in model B. MT: Mechanical thrombectomy; NIHSS: National Institutes of Health Stroke Scale; PH: parenchymal hematoma. The Hosmer–Lemeshow test showed an adequate goodness-of-fit of the final models (Model A: X2 = 4.028, p = 0.855; Model B: X2 = 9.329, p = 0.315), and the models classified correctly a total of 84% (Model A) and 85% (Model B) of cases.