bs_bs_banner
C 2017 International Center for Artificial Organs and Transplantation and Wiley Periodicals, Inc. Copyright V
Effects of Orthostatism and Hemodialysis on Mean Heart Period and Fractal Heart Rate Properties of Chronic Renal Failure Patients *Juan C. Echeverrıa, †Oscar Infante, ‡H ector P erez-Grovas, §Hortensia Gonz alez, ¶Marco V. Jos e, and †Claudia Lerma *Departamento de Ingenierıa Electrica, Universidad Aut onoma Metropolitana Unidad Iztapalapa, Iztapalapa; †Departamento de Instrumentaci on Electromecanica, Instituto Nacional de Cardiologıa Ignacio Chavez, Tlalpan; ‡Departamento de Nefrologıa, Instituto Nacional de Cardiologıa Ignacio Chavez, Tlalpan; §Laboratorio de Biofısica de Sistemas Excitables, Facultad de Ciencias, Universidad Nacional Aut onoma de Mexico, Coyoacan; and ¶Theoretical Biology Group, Instituto de Investigaciones Biomedicas, Universidad Nacional Aut onoma de Mexico, Coyoacan, Ciudad de Mexico, Mexico
Abstract: The aim of this work was to evaluate the shortterm fractal index (a1) of heart rate variability (HRV) in chronic renal failure (CRF) patients by identifying the effects of orthostatism and hemodialysis (HD), and by evaluating the correlation between a1 and the mean RR interval from sinus beats (meanNN). HRV time series were derived from ECG data of 19 CRF patients and 20 age-matched healthy subjects obtained at supine and orthostatic positions (lasting 5 min each). Data from CRF patients were collected before and after HD. a1 was calculated from each time series and compared by analysis of variance. Pearson’s correlations between meanNN and a1 were calculated using the data from both positions by considering three groups: healthy subjects, CRF before HD and CRF after HD. At supine position, a1 of CRF patients after HD (1.17 6 0.30) was larger (P < 0.05) than in healthy subjects (0.89 6 0.28) but not before HD (1.10 6 0.34). a1 increased (P < 0.05) in response to orthostatism in healthy subjects (1.29 6 0.26) and CRF patients after HD
(1.34 6 0.31), but not before HD (1.25 6 0.37). Whereas a1 was correlated (P < 0.05) with the meanNN of healthy subjects (r 5 20.562) and CRF patients after HD (r 5 20.388), no significance in CRF patients before HD was identified (r 5 0.003). Multiple regression analysis confirmed that a1 was mainly predicted by the orthostatic position (in all groups) and meanNN (healthy subjects and patients after HD), showing no association with the renal disease condition in itself. In conclusion, as in healthy subjects, a1 of CRF patients correlates with meanNN after HD (indicating a more irregular-like HRV behavior at slower heart rates). This suggests that CRF patients with stable blood pressure preserve a regulatory adaptability despite a shifted setting point of the heart period (i.e., higher heart rate) in comparison with healthy subjects. Key Words: Autonomic regulation—Chronic renal failure—Hemodialysis—Mean heart rate—Orthostatism—Scaling behavior—Sympathetic predominance.
Cardiovascular mortality shows high prevalence among patients with chronic renal failure (CRF) receiving hemodialysis (HD) (1). Persistent sympathetic overactivity, as evidenced by the analysis of
heart rate variability (HRV), is common for CRF patients (2,3). In many patients, the linear HRV indexes indicate accelerated heart rate (i.e., shorter mean RR intervals or meanNN), reduced parasympathetic influence towards the heart (e.g., reduced root-mean-squared of successive differences [RMSSD] or reduced spectral power within the high frequency band [HF]), and increased preponderance of sympathetic activity (e.g., increased low frequency band [LF] and LF/HF indexes) (4,5). Although sympathetic overactivity increases the
doi: 10.1111/aor.12887 Received June 2016; revised August 2016; accepted October 2016. Address correspondence and reprint requests to Claudia Lerma, Ph.D., Departamento de Instrumentaci on Electromec anica, Instituto Nacional de Cardiologıa Ignacio Ch avez, Juan Badiano 1, Secci on 16, Tlalpan D.F. 14080, M exico. E-mail: dr.
[email protected] Artificial Organs 2017, 41(11):1026–1034
EFFECTS OF ORTHOSTATISM AND HEMODIALYSIS long-term risk of mortality, sustained sympathetic predominance becomes important to prevent hypotension during HD (6) and it is also linked to major cardiovascular adaptations associated to blood pressure stability in CRF (i.e., accelerated heart rate, increased peripheral resistance, increased baroreflex delay and reduced baroreflex sensivity) (7). All these adaptations have been considered as evidence that the cardiovascular mechanisms, such as the baroreflex control, are operating by a regulatory set point shifted towards a state of “saturation” (7). However, this adjustment does not imply the loss of a regulatory capacity for some CRF patients because they still exhibit a clear response to external stimulus such as orthostatism and hemodialysis (4), and because following a renal transplant several cardiovascular abnormalities may be reversed (8,9). A strong and consistent correlation has been described between most linear HRV indexes and the mean heart rate (10). The relationship has been observed in humans (healthy or having chronic pathologies such as essential hypertension), living animals, isolated hearts and even in a single sinoatrial nodal cell (11). This covariance has important implications regarding the interpretation of the linear HRV indexes. The relationship has been explained as the result of an intrinsic biophysical property, restricting the use of HRV indexes to perform a simple assessment of the autonomic modulation of heart rate (11). Thus, the influence of mean heart rate upon the diagnostic and prognostic value of linear HRV indexes should be taken into consideration (10,11). As a consequence, some changes in the linear HRV indexes of CRF patients (either due to renal disease or as a response to stimulus such as orthostatism and HD) (4) could be then basically explained by variations in the mean heart rate associated to a shifted operating setting points of cardiovascular control mechanisms. Other HRV indexes have been proposed to estimate the dynamical behavior of RR intervals (12). The fractal-like scaling exponent, a1, quantifies the HRV irregularity along different time scales within a short term (between 4 and 11 heartbeats) (12), and it is considered by some researches as clinically useful (13). For CRF patients, the scaling exponent a1 evaluated from 24-h recordings is an independent predictor of mortality (14,15). For short-term recordings (e.g., 5–15 min), a1 shows changes in the HRV behavior in response to different physiological stimuli (16,17). A correlation between a1 and the mean heart rate has been observed in several physiological and experimental conditions as well
1027
(17–19). Yet in CRF patients, a1 has not been evaluated in short-term recordings (i.e., 5 min), and it is unknown if a1 is correlated with changes in the mean heart rate. The aim of this work was to evaluate a1 of CRF patients from 5-min recordings to identify the effect of orthostatism and HD, and to evaluate the correlation between a1 and the mean RR intervals from sinus beats (meanNN) for these patients. Following a previous study (4), we hypothesized that changes of the scaling exponent driven by an orthostatic challenge in CRF patients with stable blood pressure, either before or after HD, are similar to those shown by healthy subjects once considering the actual differences in meanNN between groups. PATIENTS AND METHODS Participants The study included 19 CRF patients treated with HD three times per week with stable hemodynamic response (i.e., no more than three events of intradialysis hypotension within the last month). Intradialysis hypotension was defined as a decreased systolic blood pressure by 20 mm Hg or a decrement of mean blood pressure by 10 mm Hg associated with symptoms such as nausea, vomiting, muscle cramps, and dizziness or fainting. The patients’ time per HD session was 3.6 6 0.5 h with intradialytic weight gain of 3.1 6 1.1 L. HD vintage was 12.5 6 10.2 months and a residual renal function of 0.9 6 1.5 mL/min. From the clinical record the most recent echocardiograms (within 6 months prior to study) showed left ventricular ejection fractions of 65 6 8% and laboratory results (with samples taken on a day where hemodialysis was not performed, within 1 month prior to study) showed creatinine 5 8.7 6 2.5 mg/dL, potassium 5 4.9 6 0.7 mEq/L, phosphorous 5 5.1 6 1.5 mEq/dL, calcium 5 8.9 6 1.1 mg/dL, hemoglobin 5 8.3 6 2.7 g/dL, albumin 5 3.9 6 0.5 g/dL, cholesterol 5 165 6 41 mg/dL, and triglycerides 5 145 6 86 mg/ dL. The mean age of patients was 33 6 10 years old, who had body mass indices at the time of enrollment of 22.5 6 2.7 kg/m2 (12 patients were female). The CRF etiology was systemic lupus erythematosus (n 5 1), focal segmental glomerulosclerosis (n 5 1), or unknown (n 5 17). None of the patients had clinical manifestations of amyloidosis. Patients had no collagen disease or other comorbidities that could have affected the autonomic nervous system (such as diabetes mellitus), neither supraventricular arrhythmias nor electrical conduction disorders. Artif Organs, Vol. 41, No. 11, 2017
1028
J.C. ECHEVERRIA ET AL.
A group of 20 healthy subjects were recruited with similar mean age (27 6 8 years old, P 5 0.07), sex (11 were female, P 5 0.35) and body mass index (24.3 6 3.8 kg/m2, P 5 0.11) in comparison with patients. Their condition as healthy participants was confirmed with a routine clinical exam including an electrocardiogram (ECG). The study protocol complied with the principles outlined in the Declaration of Helsinki and was approved by the Research and Ethics Committee of the National Institute of Cardiology Ignacio Chavez (protocol number 12-673). Hemodialysis prescription HD sessions were delivered with volumetric dialysis machines (4008 H, Fresenius Medical Care, Bad Homburg, Germany), using ultrapure dialysate (HCO3 5 35 mmol/L, Na1 5 138 mmol/L, K1 5 2 mmol/L, Ca21 5 3.5 mEq/L, Mg21 5 1.0 mEq/L) and polysulfone membranes (F-60 y F-80, Fresenius Medical Care, Walnut Creek, CA, USA). Hypertension was controlled by strict prescription of dry body weight without using anti-hypertensive drugs, following an approach of extracellular volume control by convection (20). Patients were on a non-restrictive diet, did not use erythropoietin, and were in a program of aerobic exercise during all HD sessions (cycling in recumbent position with modified bicycles). Study protocol ECG recordings were obtained with a protocol described previously (4,21). Briefly, a continuous ECG recording (lead II) was obtained during 16 min in supine position, and a subsequent recording was also collected during 16 min of active orthostatism. Patients maintained spontaneous breathing during all protocols and blood pressure was measured using a sphygmomanometer at the end of each recording. In CRF patients the recordings were obtained both before HD and after HD. In healthy subjects, orthostatism caused significant increments (P < 0.05) of systolic blood pressure (107 6 10 vs. 108 6 10 mm Hg) and diastolic blood pressure (73 6 9 vs. 79 6 10 mm Hg), while CRF patients showed higher systolic blood pressure than healthy subjects during supine position (before HD: 141 6 14 mm Hg, after HD: 137 6 24 mm Hg) and diastolic blood pressure was similar to healthy subjects (before HD: 77 6 10 mm Hg, after HD: 73 6 7 mm Hg). Compared to supine position, CRF patients during orthostatism had similar systolic blood pressure (before HD: 143 6 121 mm Hg, after HD: 137 6 24 mm Hg) and similar diastolic Artif Organs, Vol. 41, No. 11, 2017
blood pressure (before HD: 86 6 13 mm Hg, after HD: 80 6 12 mm Hg). Systolic blood pressure of CRF patients during orthostatism was higher than in healthy subjects (both before HD and after HD), while diastolic blood pressure was not different. Comparing before HD and after HD, there were no significant changes in both systolic blood pressure and diastolic blood pressure. ECG signal processing The ECG recordings were digitized at 250 samples per second with a 12-bits resolution using and equipment previously validated (22). The last 5 min of the recording at each position were selected to obtain the HRV or RR interval time series. The QRS complex of each heart period was identified by a second derivative algorithm (23), followed by manual inspection to delete artifacts and replace ectopic beats with interpolated intervals (24). Linear HRV indexes estimation The following HRV indexes were calculated with the computer program Kubios HRV version 2.0 (25): meanNN, SDNN, RMSSD, LF (mean power in the low frequency band from 0.04 to 0.15 Hz), HF (mean power in the high frequency band from 0.15 to 0.4 Hz) and the LF/HF ratio. LF and HF indexes were estimated in absolute units (ms2) and normalized units (n.u.) (24). Power spectral indexes were obtained with the Fourier transform method after eliminating linear trend, resampling at 3 Hz and applying a nonoverlapped Hamming window of 100 data points. Scaling exponent (a1) as provided by detrended fluctuation analysis To quantify self-affinity in the NN time series of healthy subjects and CRF patients we used detrended fluctuation analysis (DFA) (12). The suitability of this analysis for obtaining scaling exponents from short sets of HRV data, in particular those involving only 300 samples, has been assessed in detail (26). When applying DFA, the raw NN time series is initially integrated by: Y ðkÞ5
k X
½NN ðiÞ2meanNN
(1)
i51
where Y(k) is the k-th sample of the resulting integration (k5 1,2, . . .., L), NN(i) is the i-th sample of the raw time series of length L, and meanNN is the raw NN time series mean value. Y(k) is then divided into boxes of equal number n of NN intervals. A local trend (Yn) is obtained for each box by a leastsquared-linear fit, which is subtracted from Y(k) to
EFFECTS OF ORTHOSTATISM AND HEMODIALYSIS
1029
FIG. 1. Typical heart rate variability (HRV) time series obtained in supine position (clinostatism) and during active standing (orthostatism) from a healthy participant, and chronic renal failure (CRF) patient before hemodialysis (HD) and after HD. HR 5 heart rate shown in beats per minute (bpm). [Color figure can be viewed at wileyonlinelibrary.com]
reduce non-stationary deviations within boxes. The average root-mean-square fluctuation, F(n), is subsequently calculated as: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u L u1 X ½Y ðkÞ2Yn ðkÞ2 F ðnÞ5t L k51
(2)
This calculation is repeated for different size boxes or time scales n. Finally, a double-log relationship between F(n) and n is approximately evaluated by a linear model F(n) na that provides the scaling exponent a. Values > 0.5 of this exponent indicate a time series with persistent or even fractal-like correlations; resulting values < 0.5 show an anticorrelated behavior and a 5 0.5 indicates a time series with uncorrelated fluctuations. In agreement with previous studies of CRF patients we selected the range 4 to 11 NN intervals to estimate the short-range scaling exponent a1 (14,15), which appears as a reliable fitting range
(26) as it covers time scales within the scope of the autonomic response (12,13). Importantly, to avoid introducing variations in the estimation of a1 owing to the different number of NN intervals on each time series (26), we restricted and used the same number of samples from all series before applying DFA (i.e., L 5 300). Statistical analysis Categorical variables are reported as absolute values or percentages and were compared among groups by chi-squared tests or exact Fisher’s tests. Continuous variables are reported as mean 6 standard deviation. These variables were transformed to a normal distribution by a natural logarithm in case of not showing a normal distribution (Kolmogorov-Smirnov test). Mean values were compared by t tests (two groups) or analysis of variance (ANOVA) for repeated measures (three or more groups) followed by post-hoc tests adjusted by the Bonferroni method. Pearson’s correlation Artif Organs, Vol. 41, No. 11, 2017
J.C. ECHEVERRIA ET AL.
1030
coefficients were calculated between pairs of HRV indexes using all data from both positions in three groups: healthy subjects, CRF before HD, and CRF after HD. Given the use of repeated observation per subject in the calculation of correlation coefficients, we also applied a multiple regression considering the short-term scaling exponent a1 as the outcome variable as well as the meanNN index, the orthostatic position, and the CRF group membership as the predictor variables. The orthostatic position and the fact of belonging to the CRF group were treated as categorical factors using dichotomous dummy variables of a stepwise multiple regression. The statistical analysis was performed with the Statistical Package for the Social Sciences (SPSS) version 15.0 (SPSS Inc., Chicago, IL, USA), and P values < 0.05 were considered significant. RESULTS Typical NN time series from a healthy subject and a CRF patient are illustrated in Fig. 1. In comparison, the healthy subject during clinostatism shows relatively larger meanNN and SDNN, and reduced LF/HF. In response to orthostatism, decrements in the meanNN and SDNN, and an increment in LF/HF are appreciated. In contrast, the CRF patient during clinostatism (both before and after HD) exhibits smaller meanNN and SDNN, and larger LF/HF than the healthy subject. In response to orthostatism, the CRF patient shows a clear reduction of the meanNN and an increased LF/HF, both before and after HD. Before HD the SDNN augments in response to orthostatism, whereas after HD the SDNN decreases in response to orthostatism. These examples are consistent with the response to orthostatism of the linear HRV indexes in healthy subjects and CRF patients, and with the response to HD reported before in these patients having stable hemodynamic variables during HD (4,21). Table 1 shows results of meanNN, RMSSD, and a1 evaluated from all participants in this study using 300 data points from each RR time series. During clinostatism, CRF patients present a shorter meanNN (i.e., higher mean heart rate) and lower variability (smaller RMSSD) than the healthy group. These differences were observed both before and after HD. The mean value of a1 during clinostatism in healthy subjects is smaller than 1, while for CRF patients approximates 1 before HD and becomes larger than 1 after HD. CRF patients after HD show larger a1 than healthy subjects, but Artif Organs, Vol. 41, No. 11, 2017
TABLE 1. Heart rate variability indexes evaluated in 20 healthy individuals and 19 chronic renal failure (CRF) patients under hemodialysis (HD) treatment. Healthy group CRF before HD CRF after HD Clinostatism meanNN (s)
0.909 6 0.126* 0.732 6 0.110*,† (67 6 10 bpm) (84 6 12 bpm) RMSSD (ms) 54.3 6 22.8* 16.0 6 14.2† a1 0.886 6 0.283* 1.095 6 0.342 Orthostatism meanNN (s) 0.728 6 0.117 0.709 6 0.118 (85 6 14 bpm) (87 6 13 bpm) RMSSD (ms) 23.6 6 9.5 17.1 6 15.6 a1 1.290 6 0.261 1.248 6 0.368
0.688 6 0.113*,† (89 6 13 bpm) 14.0 6 12.2† 1.166 6 0.302*,† 0.599 6 0.101†,‡ (103 6 15 bpm) 12.0 6 14.8† 1.336 6 0.309
*P < 0.05 clinostatism versus orthostatism (same group). †P < 0.05 CRF (either before HD or after HD) versus healthy group (same position). ‡P < 0.05 before HD versus after HD (same position). Data are shown as mean 6 standard deviation. bpm, beats per minute.
no significant differences were observed before HD. In response to orthostatism, the healthy subjects manifest significant reductions in meanNN and RMSSD accompanied by an increment in a1, while CRF patients exhibit also a reduction in the meanNN (with no change in RMSSD) both before and after HD. The a1 value increases in response to orthostatism only after HD. At orthostatism, CRF patients after HD shows smaller meanNN and RMSSD than healthy subjects, but no significant differences were observed between the healthy subjects and CRF patients before HD. The mean value of a1 during orthostatism becomes larger than 1.2 with no significant differences among groups. The dispersion plots of Fig. 2 depict a significant correlation between a1 and meanNN in both the healthy group and the CRF group after HD, while no significant correlation was observed in the CRF group before HD (Fig. 2). In contrast, the RMSSD evaluated from the same time series correlated with meanNN in the three groups (Table 2). In fact, all linear HRV indexes (evaluated from 5-min length time series) also involve a significant correlation with the meanNN in all groups, except for the LF (n.u.) of the healthy group (Table 3). The multiple regression analysis results summarized in Table 4 show that the changes of a1 in CRF patients before HD are explained by the orthostatic challenge showing no correlation with meanNN (Model 1), while after HD the changes of a1 are explained by both the orthostatic challenge and changes in the meanNN (Model 2). This analysis also reveals that the condition of renal disease in itself has no association with the differences of
EFFECTS OF ORTHOSTATISM AND HEMODIALYSIS
1031
TABLE 2. Pearson’s correlation analysis between meanNN, RMSSD and the short-term scaling exponent (a1) estimated from time series of 300 data points. The estimation of each correlation coefficient included data during both clinostatism and orthostatism Healthy group r
P-value
CRF before HD r
P-value
CRF after HD r
P-value
RMSSD (ms) 0.844 < 0.001 0.609 < 0.001 0.619 < 0.001 20.562 < 0.001 0.003 0.984 20.388 0.016 a1
FIG. 2. Dispersion plots between meanNN and short-term scaling index (a1) for 20 healthy controls and 19 CRF patients. The estimation of each correlation coefficient (r) includes data during both clinostatism and orthostatism. [Color figure can be viewed at wileyonlinelibrary.com]
a1 between the healthy group and the CRF patients once considering the differences in the meanNN. DISCUSSION CRF patients with overall reduced HRV (e.g., decreased SDNN) (27) or altered behavior in HRV dynamics (e.g., decreased a1) have a higher mortality risk (14). Moreover, a large reduction of the spectral power within LF after HD also predicts both overall and cardiovascular mortality (28). One of the main findings of this study is that CRF patients with stable blood pressure response during
HD show dynamical changes introduced by an orthostatic challenge (active standing) of the shortterm scaling exponent (a1) that are similar (i.e., in the same direction) to those exhibited by healthy age-matched subjects. In both groups, such challenge provoked significant increments of a1 in conjunction with a reduction of the meanNN interval (meanNN) and the RMSSD parameter (Table 1). In comparison with clinostatism, the increased values of a1 (i.e., a1 1.3 for both groups) reveal smoother and less irregular dynamics of the NN interval series during orthostatism, which are in accordance with the findings of Tulppo et al. (2001) who reported an increment in a1 towards correlated values during a passive tilt-test for healthy subjects (17). Given that an orthostatic challenge redistributes blood from central- to lower-body vascular beds and unloads cardiopulmonary receptors (29), it is generally considered that this central circulatory hypovolemia prompts the sympathetic activation and the so-called vagal withdrawal (30). The smoother and less irregular fluctuations of the NN intervals during orthostatism, as suggested by the increased scaling exponent (a1 > 1), become coincident with such presumed autonomic response
TABLE 3. Pearson’s correlation analysis between meanNN and linear HRV indexes estimated from time series of 300 s length. The estimation of each correlation coefficient included data during both clinostatism and orthostatism Healthy group r
P
CRF before HD r
SDNN (ms) 0.756 < 0.001 0.663 0.545 < 0.001 0.683 LF (ms2) 2 Ln (HF [ms ]) 0.799 < 0.001 0.699 LF (n.u.) 20.219 0.174 20.415 HF (n.u.) 0.625 < 0.001 0.360 Ln (LF/HF) 20.544 < 0.001 20.388
P < < <
