Baseline NIH Stroke Scale score strongly predicts outcome after stroke

Methods.

In a trial testing the usefulness of the low-molecular weight heparinoid, danaparoid (Org 10172), TOAST enrolled 1,281 patients within 24 hours of onset of acute ischemic stroke at 37 centers in the United States.13 Data collected prospectively included the patients’ age, sex, ethnicity, history of previous stroke, baseline neurologic impairments as rated on the NIHSS, TOAST subtypes of ischemic stroke, and outcomes at 7 days and 3 months after stroke. The NIHSS scores were quantified by local investigators who had been certified in using the scale.14 Because physicians often disagree about the diagnosis of subtype of ischemic stroke, all TOAST subtype diagnoses were made by one physician at the University of Iowa who had the results of ancillary diagnostic tests available and who used a previously described algorithm.13,15 Outcomes at 7 days and 3 months were determined using the Glasgow Outcome Scale (GOS) and the modified Barthel Index (BI). For the purposes of this analysis, patients who had a GOS score of 1 and a BI score of 19 or 20 (on a scale of 0 to 20) were rated as having excellent outcomes. Patients who had a GOS score of 1 or 2 and a BI score of 12 to 20 but who did not meet the criteria for an excellent outcome were considered to have reached a good outcome. Survivors who did not have both a GOS score of 1 or 2 and a BI score of 12 to 20, and patients who died, were categorized as having poor outcomes.

Of 1,275 patients included in the trial’s intention-to-treat population, 7 did not have a stroke subtype diagnosis made at the University of Iowa. Therefore, 1,268 patients (637 treated with danaparoid and 631 given placebo) were included in this analysis. For each outcome measure, a logistic regression model was fit with excellent or good outcome as the dependent variable, baseline NIHSS score (linear, quadratic, cubic, and quartic terms) and age as continuous predictor variables, and stroke subtype, history of previous stroke, sex, and race (white versus nonwhite) as categoric predictor variables. For the purposes of the analysis of the TOAST subtypes, the patients included in the categories of stroke of other or undetermined etiology were used as the reference group against which patients with strokes of other causes were compared. Treatment groups were combined for these analyses because the overall results of the trial did not show significant differences in these two groups. The estimated probabilities were calculated only for scores that occurred in the population; no extrapolation to other scores was performed. Estimated probabilities were plotted for any stroke subtype that was a significant predictor of an excellent outcome.

Results.

Marked differences in the baseline NIHSS scores among patients with strokes attributed to small-artery occlusion (lacunes) were present in comparison with patients with strokes secondary to other causes (table 1). Although approximately two thirds of patients with stroke secondary to large-artery atherosclerosis or cardioembolism had an NIHSS score of 7 or greater, two thirds of patients with small-artery occlusion had scores of 6 or less. The baseline NIHSS score strongly predicted outcomes at 7 days and 3 months (figures 1 and 2⇓). By 7 days, almost two thirds of patients who had a score of 3 or less had already achieved an excellent outcome, but by 3 months very few patients with a baseline NIHSS score of >15 had achieved excellent results.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Effect of baseline NIH Stroke Scale score on outcome at 7 days. Patients’ outcomes are rated as excellent, good, poor, or dead. The number of points that define each group and the number of patients included in the group are listed at the bottom of each column. The accumulated percentages are listed on the left side of the figure.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Effect of baseline NIH Stroke Scale score on outcome at 3 months. Patients’ outcomes are rated as excellent, good, poor, or dead. The number of points that define each group and the number of patients included in the group are listed at the bottom of each column. The accumulated percentages are listed on the left side of the figure.

The odds ratios and CIs from the logistic regression analysis of excellent or good outcomes are presented in tables 2 and 3⇓. The higher order terms for NIHSS were not significant and were dropped from the model. No significant differences were noted between men and women at 7 days or 3 months. No significant differences between nonwhite patients and white patients was found at 7 days. However, at 3 months, nonwhite patients had approximately 34% less likelihood of having an excellent outcome—a result that is significantly different. Patients with a history of a previous stroke had approximately 30% less likelihood of an excellent outcome at 3 months. For each increase of 1 year in age, the odds for an excellent outcome at 7 days declined approximately 4% and at 3 months declined approximately 2%. For each additional point on the NIHSS, the likelihood of an excellent outcome at 7 days dropped by approximately 24% and at 3 months dropped by approximately 17%.

When adjusted for the baseline NIHSS score, patients with small-artery occlusion (lacunar infarction) had a significantly higher likelihood of an excellent outcome at 7 days and 3 months than did patients with nonlacunar strokes (see table 2). When adjusted for the baseline score on the NIHSS, the odds of an excellent outcome were not significantly different among patients with stroke of other or undetermined cause versus those with large-artery atherosclerosis or cardioembolism. The slopes of the decline in the likelihood of an excellent outcome were steep. with each increment of the NIHSS score in the range of scores of 5 to 15 (see table 2 and figure 3). The slope was more precipitous among patients with lacunar stroke than among those with nonlacunar strokes (see figure 3). At lower scores, patients with lacunes had a higher likelihood of an excellent outcome but their odds were poorer when the scores were 10 or greater. Similar results and slopes were noted for the probability to achieve either a good or excellent outcome at 7 days or 3 months (see table 3 and figure 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Probability of an excellent outcome at 7 days and 3 months as influenced by the baseline score on the NIH Stroke Scale. Results among patients with lacunar strokes are noted with dotted lines and the outcomes among patients with nonlacunar strokes are shown with non-dotted lines.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Probability of an excellent or good outcome at 7 days and 3 months as influenced by the baseline score on the NIH Stroke Scale. Results among patients with lacunar strokes are noted with dotted lines and the outcomes among patients with nonlacunar strokes are shown with non-dotted lines.

Discussion.

This project was performed after the final results of TOAST were known. Thus, these data should be considered exploratory. The data are from a clinical trial that had specific inclusion criteria and may not be representative of all patients with ischemic stroke. Still, the data collected prospectively from a large clinical trial of patients enrolled across the United States provide clinically important details that can guide clinicians. In particular, the measures to ensure the quality of the data in regard to the NIHSS scores and the TOAST subtype diagnoses mean that results of this project can provide information about these two rating instruments.

Our data show that the NIHSS score is a robust predictor of outcomes after stroke even after adjusting for other variables, including the patient’s age, race, gender, and history of a previous stroke. Although most clinicians are not surprised that a rating instrument based on the severity of the patient’s neurologic deficits correlates with the odds of favorable recovery, the steep decline in the likelihood of an excellent outcome with each increase in the NIHSS score is striking. The changes in the scores in the NIHSS were more powerful than any other factor, including stroke subtype, in predicting an excellent outcome after stroke. The NIHSS score provides prognostic information that could be useful to physicians, patients, and families. The total score can influence decisions about emergent management. The NIHSS score can also be used as an exclusion or inclusion criterion for enrollment of patients testing new treatments for stroke.

The data show that patients with severe impairments (NIHSS score > 15) have less than a 20% chance of achieving an excellent outcome. A baseline NIHSS score of approximately 15 probably corresponds to the grouping of total anterior cerebral infarction included in the classification of Bamford et al.16 These investigators reported that only 4% of patients with total anterior cerebral infarction were independent at 6 months after stroke. Our results also are similar to those of Muir et al.,4 who noted that an NIHSS score of 13 was the break point for predicting outcomes after stroke. The trial of rt-PA also found that the likelihood of a nearly complete recovery at 3 months was influenced by the NIHSS score. Regardless of age or treatment, most patients with a score of 15 or greater did not recover. The probability for either a good or excellent outcome is also affected by the NIHSS score. Approximately 90% of patients with a score of 4 to 6 have a good or excellent outcome at 3 months whereas approximately 40% of patients with a score of 16 to 22 have a similar rate of recovery. At present, most patients with severe stroke (NIHSS score > 15) do not recover, and therapies that can improve the likelihood of recovery are needed to treat these critically ill patients.

Conversely, our data show that most patients with a low NIHSS score have a favorable prognosis. Thus, recruitment of patients with low NIHSS scores into a trial testing a promising intervention likely will dilute any treatment effect. Enrollment of mildly affected patients could increase the likelihood of rejecting a potentially beneficial therapy. The negative results of a recent European trial of rt-PA were ascribed partially to the low NIHSS scores among patients enrolled in the study.17 Patients with a low NIHSS score generally do well regardless of treatment; consequently, they might not need to be treated with agents that have the potential for serious complications.

The data show that the baseline NIHSS score correlates strongly with the TOAST subtype diagnosis. In general, patients with lacunar strokes had a lower NIHSS score than persons with strokes of other causes. Still, even after adjustment for the NIHSS score, patients with lacunar stroke have a threefold higher likelihood of an excellent outcome at 3 months than persons with strokes of other causes. Emergent diagnosis of subtype of ischemic stroke can be difficult. Subtype diagnoses change following stroke as the results of ancillary diagnostic tests become available.18 In the study of Madden et al.,18 approximately one-third of patients diagnosed initially as lacunar infarction had other causes of stroke. In addition, a sizable proportion of patients whose strokes were attributed to other causes were later diagnosed as having lacunar strokes. Thus, because the baseline stroke subtype diagnoses are unreliable, we did not use these diagnoses for this project. Rather, this project used the diagnosis after the results of ancillary tests were acquired. Even when presented with the same data, physicians often disagree.15 To compensate for the potential for interobserver disagreements, the subtype diagnoses used in this analysis were made by one physician who used a standard algorithm.12 Thus, the subtype diagnoses used in this study do not replicate those that are made in an emergency situation. At present, the usefulness of the emergent diagnosis of subtype of ischemic stroke in swaying decisions in early treatment is unknown. The trial of rt-PA did not demonstrate a difference in response to treatment between persons with different subtypes of stroke.19 Until definitive data show the usefulness of subtype diagnoses in emergent management, the rapid diagnosis of subtype should not be mandated. On the other hand, the subtype diagnosis probably is critical in determining long-term therapies to prevent recurrent stroke. In this situation, the TOAST classification might be useful.

Clinicians and researchers recognize the shortcomings of the neurologic examination and related scales in forecasting prognosis. Other ways are needed to improve the ability to predict outcomes and to monitor responses to treatment. Tests such as diffusion and perfusion MRI could be effective adjunctive measures.20-23 However, to justify the widespread use of these expensive technologies, their usefulness in predicting both prognosis and responses to any intervention must be compared with clinical rating instruments, such as the NIHSS.24 In particular, brain imaging tests may provide additional information to the clinical assessments in those patients who have NIHSS scores at the extremes of the scale (<5 or >15). At present, the NIHSS score remains the most powerful predictor of outcome after ischemic stroke.