Introduction: We previously reported that severe periventricular white matter disease (WMD) was independently associated with poor functional outcome at 3 months after ischemic stroke when compared with WMD graded as none, mild, or moderate. In that report, using a cohort from 1999, WMD grading was based primarily on CT scans. We hypothesized that the grading scheme was crude enough to have missed a dose-response relationship between functional outcome and worsening WMD severity. In the present study, we examined data from a cohort of patients with ischemic strokes in 2005 having a higher proportion of MRIs to determine if grade of WMD is associated with functional outcome, or if the association is evident only when comparing severe WMD to other grades. This also allowed us to determine if our model was useful at predicting post-stroke outcomes in an independent data set.
Methods: In 2005, a cohort of 502 ischemic stroke patients was prospectively
identified in the same way as in 1999. Stroke team physicians reviewed the medical record abstract and all available imaging studies for each case. The degree of WMD was assessed using a four-level ordinal scale. The WMD grade from the first MRI (or CT, if no MRI was done) was used when multiple scans were available. The multivariable model to predict functional outcome (modified Rankin Scale, mRS) at 3 months previously derived from the 1999 cohort was applied to the 2005 cohort. We allowed the parameter estimates to vary but initially maintained the same variable structure. We then substituted ordinal WMD grade for the dichotomous variable and re-examined model performance.
Results: 402 patients were included in the analysis from 2005. Average age was 66. MRI was used for WMD grading in 78% of cases, and CT in 22% of cases. Distribution of WMD was 11% none, 49% mild, 22% moderate, and 18% severe. The model R2 for the 1999 cohort was 0.51, and 0.59 when the model was applied to the 2005 cohort. Including WMD as an ordinal variable did not significantly improve model performance. When considering WMD as a binary variable (severe vs. all other grades), there was a significant association with functional outcome (p=0.048). When considered as an ordinal variable, WMD grading was still statistically associated with functional outcome (p=0.035).
Conclusions: Severe WMD is associated with poorer functional outcomes 3 months after stroke when compared with patients who do not have it. Utilizing ordinal WMD information did not improve statistical model fit. It is possible that a “threshold” of severe WMD must be reached to be associated with poor outcome, although a dose-effect relationship may exist. Our results verify that WMD should be considered in studies of stroke outcome.