High NIHSS Values Predict Impairment of Cardiovascular Autonomic Control

Autonomic cardiovascular dysfunction is common after stroke.^1–8 Sympathetic hyperactivity and parasympathetic dysfunction⁹ may cause tachy- or bradyarrhythmias,^6,7 troponin T increase,¹⁰ myocardial infarction, or sudden death^11,12 depending on brain area affected by the stroke.^8,12,13

Altered or reduced heart rate variability during acute stroke may be prognostically unfavorable.^9,14,15 Sykora et al¹⁶ showed reduced baroreflex sensitivity (BRS), ie, compromised autonomic adjustment of heart rate and vascular tone to sudden blood pressure (BP) changes, in acute and subacute stroke patients.^5,17 They concluded that sympathetic overactivity and blunted BRS predict poor prognosis after stroke.^5,15,18 Thus, early diagnosis of autonomic dysregulation has prognostic and therapeutic relevance in acute stroke.^5,18

However, diagnosis of impaired autonomic BP and heart rate modulation requires specific techniques and expertise that is not widely available. Therefore, easily determined clinical surrogate markers of autonomic failure are desirable.

Based on previously reported correlations between autonomic impairment and clinical deficits,^4,19 we hypothesize that readily available clinical stroke scale scores may serve as a surrogate measure of increased autonomic risk in acute stroke.

To determine whether acute clinical deficits reflect risk of autonomic cardiovascular dysregulation, we studied correlations between parameters of autonomic modulation and the National Institutes of Health Stroke Scale (NIHSS) scores²⁰ in acute stroke patients.

Patients and Methods

In 50 patients (25 women, 25 men; age 48–84 years; mean age, 66±13 years) with acute, first-ever ischemic stroke in the middle cerebral artery territory (28 left-hemispheric and 22 right-hemispheric strokes), we assessed clinical stroke severity by means of NIHSS (range, 0–42 points)²⁰; we also monitored cardiovascular autonomic modulation within 50 minutes to 23 hours (mean, 589±444 minutes) after stroke onset. Patients with other diseases and medication that affect the autonomic nervous system were excluded from the study. Patient data were compared with those of 32 age-matched healthy controls (20 women, 12 men; mean age, 61±8 years). We recruited healthy volunteers among unaffected relatives and friends of patients and among members of our research team. The study was approved by the ethics committee of the University of Erlangen-Nuremberg.

To derive parameters of cardiovascular autonomic modulation, we recorded 5-minute time-series of R-R-interval (RRI), BP, respiratory frequency, and transcutaneous oxygen saturation (SatO₂). RRIs were recorded by conventional 3-lead electrocardiography. Beat-to-beat systolic and diastolic blood pressures (BP_sys, BP_dia) were measured noninvasively at the index or middle finger of the nonparetic hand, using the vascular unloading technique²¹ (CNAP^TM, Dräger Medical), then were calibrated against ipsilateral brachial artery BP.²¹ Respiratory frequency was recorded by chest impedance measurements. SatO₂ was measured by pulse-oximetry (Dräger Medical).

All signals were sampled at 200 Hz, digitized, and stored for analysis on a custom-designed data acquisition and analysis system (SUEmpathy^TM, SUESS Medizin-Technik).²²

From 5-minute recordings without artifacts, we extracted the most stationary 90-second epochs, then calculated mean values and SD of all signals. To avoid a bias regarding the signal epoch selected for data analysis, we extracted the most stationary 90-second period from the 5-minute recordings while blinded to the participant's status (eg, sex, age, healthy control or patient, NIHSS score).

As autonomic parameters, we determined the coefficient of variation of RRIs (RRI-CV). RRI-CV and RRI-SD reflect sympathetic and parasympathetic cardiac modulation.^23,24 We calculated square root of the mean squared differences of successive RRIs (RMSSD), reflecting parasympathetic cardiac modulation.^23,24

We performed trigonometric regressive spectral analyses²⁵ of slow, underlying RRI and BP oscillations in frequency ranges reflecting sympathetic and parasympathetic influences on RRI and BP.²⁵

We identified peaks of oscillations in the so-called low-frequency (LF; 0.04–0.14 Hz) and high-frequency (HF; 0.15–0.50 Hz) ranges of RRI and BP modulation.^23,24

LF oscillations of RRI at rest are considered to be mediated by sympathetic outflow and, to an undetermined degree, also by parasympathetic activity; meanwhile, LF oscillations of BP are related to sympathetic outflow only.^23,24 HF oscillations in RRI reflect parasympathetic activity,^23,24 whereas BP fluctuations in the HF range are primarily a mechanical consequence of respiration-induced fluctuations in venous return and cardiac output.^23,24

The magnitude of LF and HF oscillations was determined as the integral under the power spectral density curves of RRI (ms²/Hz) and BP (mm Hg²/Hz) for the 2 frequency bands, and was expressed as LF- and HF-powers of RRI (ms²) and BP (mm Hg²).^23,24

In addition, we calculated RRI-LF/HF-ratios as an index of sympatho-vagal balance, and the sum of LF- and HF-powers as an approximation of the total power of RRI oscillations and index of overall autonomic cardiac modulation.^23,24 We normalized RRI-LF- and RRI-HF- powers to reduce effects of interindividual differences in total powers on absolute RRI-LF- and RRI-HF-powers,²⁶ where RRI-LFnu=(RRI-LF/[RRI-LF+RRI−HF])×100%, and RRI-HFnu=(RRI-HF/[RRI-LF+RRI−HF])×100%.^23,24 To determine BRS, the trigonometric regressive spectral software selected pairs of LF and HF oscillations of BP_sys and RRI with high coherence (>0.7).²⁷ With high coherence, the sensitivity of the baroreflex loop (ms×mm Hg⁻¹) can be derived as gain values from changes in RRIs (ms) in relation to changes in BP_sys (mm Hg).²⁷

Statistics

For data analysis, we used a commercially available statistical program (SPSS 18.0, SPSS Inc.). We tested data for normal distribution by the Shapiro-Wilk test.

Normally distributed patient and control data were compared using the t test for unpaired samples. Non-normally distributed data were compared using the Mann-Whitney U test.

Correlations between NIHSS scores and bio-signals as well as autonomic parameters and BRS were assessed with the Spearman rank correlation test.

Using the Spearman rank correlation test, we also calculated correlations between the interval from stroke onset to autonomic testing and the NIHSS scores, and we correlated the interval with values of the recorded bio-signals and with parameters of autonomic modulation. Significance was assumed for P<0.05.

Results

In 50 stroke patients, NIHSS scores ranged from 1 to 21 (median, 5; lower quartile, 3; upper quartile, 11). Table 1 summarizes data of patients and controls.

In patients, BP_sys and respiratory frequency were significantly higher than they were in controls, whereas RRIs, RRI-SD, RRI-CV, and RMSSD were lower in patients than they were in controls (Table 1).

Similarly, patients had lower RRI-LF-powers, RRI-HF-powers, RRI-total powers, and BRS than did controls (Table 1).

BP_dia values were not quite significantly higher in patients than in controls (P=0.07), while BP_sys-LF-powers, BP_sys-HF-powers, normalized RRI-LF-powers, normalized RRI-HF-powers, and RRI-LF/HF-ratios did not differ between patients and controls (P>0.05).

NIHSS scores correlated significantly with normalized RRI-LF-powers and RRI-LF/HF-ratios, while there were an inverse correlations between NIHSS scores and RRIs, RRI-SDs, RMSSD, RRI-HF-powers, normalized RRI-HF-powers, RRI-total powers, and BRS (for Spearman-Rho-values, see Table 2).

There were no significant correlations between NIHSS scores and BP_sys, BP_dia, SatO₂, respiratory frequency, RRI-CV, absolute RRI-LF-powers, and BP_sys-LF-powers.

There was no significant correlation between NIHSS scores and the interval from stroke onset to autonomic testing (Spearman Rho, 0.218; P=0.128).

Moreover, there were no significant correlations between this interval and RRIs, BP_sys and BP_dia, respiratory frequency, SatO₂, RRI-SDs, RRI-CVs, RRI-LF-powers, RRI-LF/HF-ratios, normalized RRI-LF- and RRI-HF-powers, BP_sys-LF-powers, BP_sys-HF-powers, and BRS. In contrast, the interval correlated with the parasympathetic indices RMSSD of RRIs (Rho=0.308; P=0.029), RRI-HF-powers (Rho=0.415; P=0.003), and with the index of overall autonomic cardiac modulation, the sum of RRI-LF-powers and RRI-HF-powers (RRI-total powers; Rho=0.284; P=0.046).

Discussion

Our stroke patients had higher BP, heart rate, and respiratory frequency than did controls, indicating increased sympathetic cardiovascular modulation.^{6,7,9,28–30} However, the lower RRI-LF-powers, lower sympathetically and parasympathetically mediated RRI-SDs, RRI-CVs, and RRI-total powers^23,24 in patients than in controls show a general loss of autonomic cardiac modulation; this has been reported in previous stroke studies.^6,7,28 In contrast to the increase in BP, heart rate, and respiratory frequency of our patients, similar RRI-LF/HF-ratios between patients and controls seem to suggest that there is no major change in sympatho-vagal balance after stroke. Yet, increasing RRI-LF/HF-ratios in patients with higher NIHSS scores, as well as the lower RMSSDs and RRI-HF-powers in patients than in controls, confirm a loss in parasympathetic modulation after stroke, and predominant sympathetic tone with increasing stroke severity.

Previous studies support the conclusion that autonomic imbalance depends on stroke severity. Korpelainen et al⁴ report no increase in RRI-LF/HF-ratios in patients with stroke severity similar to that of our patients.^4,31 In contrast, Tokgözoglu et al found increased RRI-LF/HF-ratios in patients who had higher NIHSS scores than did our patients.⁶

The correlations seen in our patients between NIHSS scores and parameters of autonomic modulation (Figure) indicate a higher risk of autonomic complications in patients with more severe strokes. Tokgözoglu et al observed an association between sudden death and reduced parasympathetic, but increased sympathetic activity in their 62 stroke patients.⁶ The 7 patients who died unexpectedly during hospitalization had higher RRI-LF/HF-ratios than did surviving patients.⁶ Among 44 stroke patients, Orlandi et al found increased RRI-LF/HF-ratios in the 31 patients with arrhythmias.⁷

• Download figure
• Download PowerPoint

Sympathetic predominance increases the risk of poststroke tachyarrhythmias,^6,7 myocardial infarctions,^10,12 myofibrillary necrosis, perivascular and interstitial fibrosis, and myocyte vacuolization^10,32; it additionally increases the risk of secondary brain injury and edema caused by sympathetically driven inflammation with fever, hyperglycemia, polycythemia, and increased blood-brain barrier permeability.^18,33 Consequently, increased sympathetic outflow compromises stroke outcome.²⁹

The progressive decline in parasympathetic activity in our patients with more severe strokes adds to the risk of cardiovascular and cerebral complications.³⁴ Parasympathetic deficiency promotes malignant tachyarrhythmias^8,13,34,35 and mortality,³⁶ reduces cerebral vasodilatation in animal stroke studies, and subsequently furthers cerebral vasoconstriction³⁷ and secondary brain damage.³⁸

The overall loss in autonomic modulation, ie, the decreasing RRI-SDs and RRI-total powers in patients with higher NIHSS scores, is associated with a growing risk of cardiac complications and sudden death.^14,39

Declining autonomic modulation predicts poor outcome, as shown in patients with myocardial infarction,⁴⁰ chronic heart failure,⁴¹ multiple organ dysfunction syndrome,⁴² and in ischemic stroke.^4,14

Progressive loss in autonomic modulation in patients with more severe stroke also causes deteriorating heart rate and BP adjustment to instantaneous changes of either parameter because of declining BRS.¹⁷ Sykora et al showed that BRS impairment depends on the volume of the stroke and involvement of the insula¹⁶; they confirm the conclusions of Robinson et al that BRS deterioration after stroke reflects central autonomic dysfunction.¹⁵ Similar to our results, Sykora et al found correlations between decreasing BRS and increasing NIHSS scores.¹⁹

Reduced BRS is associated with poor outcome in cardiac, renal, or metabolic diseases,^40,43 and in stroke.^5,15 According to Robinson et al, BRS impairment during acute stroke is associated with a 4.5-fold increase in mortality rates.¹⁵ Baroreflex failure results in increased BP fluctuations⁴⁴ that may exceed cerebral autoregulation buffering capacity⁴⁵; this causes secondary cerebral lesions,⁴⁶ particularly in patients with more severe stroke and more deficient BRS. BP fluctuations worsen stroke outcome, as they cause more severe end-organ damage than does nonfluctuating arterial hypertension.⁴⁷

In our patients, coefficients of correlation between increasing NIHSS scores and deteriorating autonomic parameters range from Spearman Rho values of 0.29 to 0.47. Still, the high consistency of correlations between stroke severity and all measures of autonomic dysregulation confirms that more severe stroke is associated with more pronounced autonomic failure and subsequent risk of secondary cardiovascular^{6,7,10,12,18,29,30} or cerebral complications.^18,33,37,38

Study Limitations

The rather wide interval between stroke onset and autonomic testing, from 50 minutes to 23 hours, might bias our results. However, NIHSS scores were not dependent on the interval between stroke onset and autonomic evaluation. In contrast, there seems to be inconsistent correlations between this interval and some of the autonomic parameters. Particularly, the positive correlation of the interval between stroke onset and autonomic testing with the parasympathetic parameters RMSSD and RRI-HF-powers suggests that parasympathetic modulation recovers with increasing time since stroke onset. Moreover, the correlation of the interval with overall autonomic modulation points toward the potential for regaining autonomic control over time. The findings encourage follow-up assessments of autonomic control to determine the duration and time course of autonomic dysfunction.

Although we found significant correlations between NIHSS scores and parameters of cardiovascular autonomic dysfunction, there is substantial variability within these correlations. We assume that this variability is because of the effects of age and sex on autonomic parameters, and because of the difference between discontinuous NIHSS scoring and continuous values of autonomic function.

In contrast to continuous values of autonomic parameters, the NIHSS is designed as a straightforward scoring system that assigns noncontinuous scores to the major clinical deficiencies without reflecting the entire scope of deficits in an individual stroke patient (eg, apraxias and neurocognitive deficits).^20,48 Consequently, stroke severity may be categorized by the same NIHSS score in patients with a different extent or location of the neurological lesion. In contrast, involvement of different parts of the central autonomic network most likely accounts for differences in autonomic dysfunction and thus different values of parameters reflecting dysautonomia.⁴⁹

Moreover, most autonomic parameters vary with differences in age and sex,^50,51 while NIHSS scores are independent of the patient's sex or age. The age range of our patients was rather wide (48 to 84 years) and very likely contributed to the variation in autonomic parameters, regardless of NIHSS scores. Similarly, differences in sex, with 25 male and 25 female stroke patients, contribute to the variation in autonomic parameters, again regardless of the NIHSS score.^50,51

The variability of autonomic parameters demonstrates the need for refined autonomic testing in stroke patients. Yet, the methodology is not readily available. Despite the rather wide variability of autonomic parameters for a given NIHSS score, the consistency of correlations between the autonomic parameters and NIHSS scores still supports the conclusion that NIHSS scoring may serve as a coarse substitute for sophisticated autonomic assessment.

In summary and in conformity with previous studies,^{4–7,14,15,28} our results demonstrate the need for autonomic monitoring of stroke patients to prevent complications caused by autonomic failure.

However, autonomic monitoring is not widely available; and yet, NIHSS scores are easily taken. From the correlations seen in our patients, we suggest that NIHSS scores may serve as surrogate markers of progressive autonomic failure. Deteriorating NIHSS scores require close observation of heart rate, BP, and the variabilities of those measures. Loss of heart rate variability and increasing BP variability indicate growing autonomic risk and predict the need for interventions to stabilize the cardiovascular system.

Perspective

There are many reports about differences in autonomic dysfunction after left- and right-sided stroke.^{2,3,6,8,29,52} Although many studies found a shift toward more prominent sympathetic modulation after right-hemispheric stroke,^{8,11,29,52,53} there are also reports that only found a decrease in total autonomic modulation⁴ or a decrease in parasympathetic outflow after right-hemispheric stroke.^2,54 Moreover, NIHSS scores are higher with left-sided than with right-sided stroke.^48,55–59

Therefore, we assume that the correlations seen between NIHSS scores and autonomic parameters might be hemisphere-dependent. Hemispheric predominance of autonomic modulation³⁵ might account for discrepancies of autonomic dysfunction and of its correlation with NIHSS scores between patients with right and left middle cerebral artery stroke. A preliminary analysis of our hemisphere-specific data suggests there are quite complex and intricate interactions between the side of the lesion and the dysautonomia. Yet, it is beyond the scope of this article to present and discuss the hemisphere-specific data. We, however, intend to provide a separate detailed analysis of hemispheric correlations and differences.

Sources of Funding

The study was supported by Sanofi-Aventis , GmbH , Germany, the Rolf und Hubertine Schiffbauer-Stiftung , Hof, Germany, and the International Brain Research Foundation , Edison, NJ.

Disclosures

None.

Footnotes