Lower NIH Stroke Scale Scores Are Required to Accurately Predict a Good Prognosis in Posterior Circulation Stroke

Abstract

Background

The NIH stroke scale (NIHSS) is an indispensable tool that aids in the determination of acute stroke prognosis and decision making. Patients with posterior circulation (PC) strokes often present with lower NIHSS scores, which may result in the withholding of thrombolytic treatment from these patients. However, whether these lower initial NIHSS scores predict better long-term prognoses is uncertain. We aimed to assess the utility of the NIHSS at presentation for predicting the functional outcome at 3 months in anterior circulation (AC) versus PC strokes.

Methods

This was a retrospective analysis of a large prospectively collected database of adults with acute ischemic stroke. Univariate and multivariate analyses were conducted to identify factors associated with outcome. Additional analyses were performed to determine the receiver operating characteristic (ROC) curves for NIHSS scores and outcomes in AC and PC infarctions. Both the optimal cutoffs for maximal diagnostic accuracy and the cutoffs to obtain >80% sensitivity for poor outcomes were determined in AC and PC strokes.

Results

The analysis included 1,197 patients with AC stroke and 372 with PC stroke. The median initial NIHSS score for patients with AC strokes was 7 and for PC strokes it was 2. The majority (71%) of PC stroke patients had baseline NIHSS scores ≤ 4, and 15% of these ‘minor’ stroke patients had a poor outcome at 3 months. ROC analysis identified that the optimal NIHSS cutoff for outcome prediction after infarction in the AC was 8 and for infarction in the PC it was 4. To achieve >80% sensitivity for detecting patients with a subsequent poor outcome, the NIHSS cutoff for infarctions in the AC was 4 and for infarctions in the PC it was 2.

Conclusion

The NIHSS cutoff that most accurately predicts outcomes is 4 points higher in AC compared to PC infarctions. There is potential for poor outcomes in patients with PC strokes and low NIHSS scores, suggesting that thrombolytic treatment should not be withheld from these patients based solely on the NIHSS.

Introduction

The NIH stroke scale (NIHSS) is the most commonly used scoring system for quantification of neurologic deficits after acute stroke [1, 2]. It is an excellent predictor of long-term outcomes after stroke [3, 4]. Of the elements present on the current scale, motor function and cortical signs receive the highest punctuation. Symptoms such as limb ataxia and cranial nerve palsies receive fewer points and other posterior circulation (PC) findings such as truncal ataxia and nystagmus are not measured. The initial NIHSS score is often lower in patients with PC strokes since the scale is strongly weighted towards deficits caused by anterior circulation (AC) lesions [3, 5]. In a recent cohort study, 76% of patients with infarctions in the PC presented with baseline NIHSS scores ≤ 5 [6].

Administration of i.v. tissue plasminogen activator (tPA) remains the standard of care for patients with acute ischemic stroke [7]. An NIHSS score of 4–5 or less is frequently used as a relative exclusion criterion for treatment based on the original NINDS tPA trial exclusion of minor, nondisabling symptoms [7, 8]. Based on the tendency of patients with PC strokes to present with lower NIHSS scores, many patients with PC strokes may not receive acute thrombolysis. This would be appropriate if the NIHSS cutoffs that optimally predict outcomes were the same for AC and PC strokes, but this is unlikely based on the limited existing data [9]. The objective of this study was to assess the predictive value of the NIHSS in AC versus PC strokes on the functional outcome at 3 months.

Methods

This was a retrospective analysis of a prospectively collected database from a Joint Commission-certified comprehensive stroke center.

Patient Selection

Adult patients admitted to the stroke service with acute arterial infarction from January 2005 through December 2011 were included. Patients were excluded if they had a preexisting baseline disability [defined as a modified Barthel index (mBI) <15], missing clinical data, or infarctions involving both the AC and the and PC, or if the stroke was a complication of surgery or a major medical illness that had been the primary cause for admission.

Data Collection

A full-time stroke data abstracter prospectively collected demographics, initial NIHSS scores, prestroke functional status, medical comorbidities, vascular distribution of the infarct, treatment with i.v. tPA or endovascular procedures, complications, and outcome assessments at discharge and at 3 months. Stroke distribution was diagnosed by head CT, by diffusion-weighted MRI, or clinically (in a few cases). Outcomes were assessed at 3 months ± 2 weeks using the mBI administered either during a clinic visit or via a structured telephone interview. Since the study focused on minor strokes, a poor outcome was defined as death or a score ≤ 18 on the mBI scale, indicating the patient had at least some difficulty with activities of daily living.

Statistics

A χ² test, a t test, and Wilcoxon’s rank-sum test were used for univariate analyses based on the distribution of the data. Receiver operating characteristic (ROC) curves were constructed for admission NIHSS scores and outcome, with optimal cutoffs determined at Youden’s point. Logistic regression was employed for multivariate analyses using backwards stepwise selection with the threshold for elimination of a variable set at p > 0.10. All potential confounding variables identified in the univariate analyses (p < 0.10) as well as prestroke mBI and history of diabetes were included in the initial model. Analyses were performed using Stata v11.2.

Results

There were 1,569 acute arterial ischemic stroke patients who met the inclusion criteria. This included 1,197 patients with AC strokes and 372 patients with PC strokes. The characteristics of the patient cohort are shown in table 1. Patients with PC stroke were more likely to be younger and male and less likely to have atrial fibrillation or congestive heart failure. The median NIHSS score on admission was 5 points lower in PC strokes. Almost three quarters of PC stroke patients presented with NIHSS scores ≤4 compared to less than half of AC stroke patients (p < 0.001).

Table 1.

Patient characteristics by location of infarction

	AC (n = 1,197)	PC (n = 372)	p
Age	73 (60–81)	68 (57–79)	<0.001
Male sex	49	57	0.005
Ethnicity
White	84	82	n.s.
African-American	7	6	n.s.
Hispanic	6	9	n.s.
Other	3	4	n.s.
Past medical history
Atrial fibrillation	30	19	<0.001
Congestive heart failure	12	7	0.008
Hypertension	74	76	n.s.
Coronary artery disease	35	31	n.s.
Diabetes	24	28	n.s.
Hyperlipidemia	48	52	n.s.
Prestroke mBI	20 (19–20)	20 (19–20)	n.s.
Initial NIHSS score	7 (2–16)	2 (1–5)	<0.001
NIHSS score ≤4	41	71	<0.001
Acute treatments
Any thrombolysis	34	14	<0.001
i.v. tPA	28	11	<0.001
Endovascular treatment	10	3	<0.001
Inpatient mortality	21	6	<0.001
Poor outcome at 3 months	48	26	<0.001

Data are presented as percents or medians (IQR). n.s. = Not significant.

Consistent with their higher NIHSS scores, patients with AC strokes were more likely to be treated with i.v. and/or endovascular thrombolysis (table 1). Overall, 80% of the patients with NIHSS scores ≤ 4 achieved an excellent outcome. However, 15% of patients with PC stroke and an admission NIHSS score ≤ 4 were dead or disabled at 3 months. Five patients (2%) with PC stroke and an initial NIHSS score ≤ 4 died or were transitioned to hospice care as a direct consequence of the stroke. Only one patient (0.2%) with AC stroke and an NIHSS score ≤ 4 died, and the death was not due to the stroke. No patient in this category was transitioned to hospice care.

In multivariate analyses, the initial NIHSS score was an independent predictor of excellent outcomes at 3 months in both AC and PC stroke (table 2). In order to determine whether the accuracy of the NIHSS for predicting outcomes differed depending on the location of the infarction, ROC curves for the two stroke distributions were created and the area under the curve (AUC) was calculated (fig. 1). Youden’s point for maximal diagnostic accuracy was determined. In the AC, NIHSS scores ≥ 8 correctly classified 72% of patient outcomes (95% CI 0.77–0.82, sensitivity 71.5%, specificity 73.8%). In the PC, NIHSS scores ≥ 4 correctly classified 71% of patient outcomes (95% CI 0.67–0.79, sensitivity 65.6%, specificity 72.4%). The cutoff to achieve >80% sensitivity for poor outcomes was ≥ 4 in the AC and ≥ 2 in the PC.

Table 2.

Multivariate analyses of factors associated with excellent outcomes at 90 days by location of infarction

Variable	OR	95% CI	p
AC
NIHSS score (per point)	0.86	0.84–0.89	<0.001
Prestroke Barthel index (per point)	1.63	1.28–2.09	<0.001
Age (per year)	0.97	0.96–0.98	<0.001
History of diabetes	0.36	0.25–0.52	<0.001
Any thrombolysis	1.46	0.98–2.19	0.065
PC
NIHSS score (per point)	0.88	0.83–0.94	<0.001
Prestroke Barthel index (per point)	1.47	1.02–2.11	0.04
Age (per year)	0.96	0.94–0.99	0.004
Male sex	2.04	1.10–3.78	0.024
History of diabetes	0.35	0.19–0.65	0.001

Fig. 1.

ROC analysis of admission NIHSS scores and excellent outcomes in both AC and PC strokes. Youden’s point (sensitivity = specificity) is shown by the arrows. The sensitivity, specificity, and likelihood ratios (LR) for each NIHSS cutoff are shown below the ROC curves.

Discussion

This study investigated the NIHSS score at presentation as a predictor of the prognosis in AC versus PC stroke. The finding of lower NIHSS scores in PC strokes is consistent with the analysis of the TOAST trial [5] and several cohort studies [10–12], in which the mean NIHSS scores in PC strokes were 3–5 points lower than those in AC strokes. For analyses of outcomes, a definition of excellent outcome equivalent to that in the NINDS tPA trial [7] was utilized. We focused on lower NIHSS scores since those scores are often interpreted as assurance that the patient will achieve a good outcome and therefore does not require acute thrombolytic treatment. While this perspective has been challenged by recent work showing surprisingly high rates of poor outcomes in ‘mild’ strokes [13–16], limited work has focused on the specific challenge of interpreting low NIHSS scores in PC stroke patients. One of the most relevant findings of our study is that 15% of patients with PC stroke with NIHSS scores ≤ 4 had death or disability at 3 months as a direct consequence of their stroke. This suggests that an NIHSS score ≤ 4 does not reliably predict an excellent outcome in the PC.

The NIHSS score was independently associated with outcome in both AC and PC strokes in the multivariate analyses, with similar OR for excellent outcome. We also found that the NIHSS performs moderately well in outcome prediction for both AC and PC strokes based on the AUC analysis of the ROC curves. However, both the AUC and the cutoff for greatest diagnostic accuracy were lower in PC strokes. These results concur with a previous cohort study that also determined that the optimal cutoff NIHSS scores to predict a favorable outcome differed in patients with PC and AC strokes [9]. Using mRS 0–2 as a good outcome, in PC stroke the optimal cutoff of the NIHSS score was ≤ 5 (sensitivity 84%, specificity 81%) and for AC stroke patients the cutoff was ≤ 8 (sensitivity 80%, specificity 82%). These results are very similar to ours despite different definitions of good outcome, our inclusion of patients treated with thrombolysis, and the larger sample size in our cohort.

Identification of Youden’s point is helpful for determining the diagnostic accuracy of a test, but one could argue that the NIHSS score cutoff for the selection of patients who should be treated with tPA should be sufficiently low to maximize the sensitivity of the test and reduce the possibility of withholding treatment from a patient with chances of a poor outcome. From this perspective, one would accept the risk of treating ‘false positives’, a view that is supported by the relatively low rate of hemorrhagic transformation in small [17] and PC [12] strokes. In our cohort, NIHSS scores ≥ 4 in the AC yielded a sensitivity of 84% for poor outcomes, while in the PC the cutoff needed to be ≥ 2 to achieve the same sensitivity. This data suggest that if an NIHSS cutoff is used to guide tPA treatment decisions, the criteria should likely be lower in suspected strokes in the PC.

The data were derived from a large, prospectively collected database at a high-volume center, yielding over 300 PC strokes to analyze. However, as this was a single-center study, it is possible that local biases in treatment decisions impacted the results. In addition, despite the high treatment rates in the overall cohort, only 52 patients with PC strokes were treated with thrombolysis, limiting our power to detect a treatment effect in this subgroup.

Conclusions

The NIHSS score, which identifies patients at risk for poor outcomes, is lower in patients with PC strokes. The commonly used NIHSS cutoffs to select patients for tPA treatment should be viewed with caution in suspected PC strokes.