Intensive Care Med (2013) 39:13591367 DOI 10.

1007/s00134-013-2937-5

ORIGINAL

Guillermo Gutierrez Aparna Das Guillermo Ballarino Arshan Beyzaei-Arani lya Tu rkan Hu Marian Wulf-Gutierrez Katherine Rider Hatice Kaya Richard Amdur

Decreased respiratory rate variability during mechanical ventilation is associated with increased mortality

Received: 22 February 2013 Accepted: 14 April 2013 Published online: 7 June 2013 Springer-Verlag Berlin Heidelberg and ESICM 2013 Electronic supplementary material The online version of this article (doi:10.1007/s00134-013-2937-5) contains supplementary material, which is available to authorized users.

R. Amdur Department of Surgery, The George Washington University MFA, Washington DC, USA R. Amdur VA Medical Center, Washington DC, USA

Abstract Objective: Patients on ventilatory support often experience signicant changes in respiratory rate. G. Gutierrez ()) A. Das Our aim was to determine the possiG. Ballarino A. Beyzaei-Arani ble association between respiratory K. Rider H. Kaya rate variability (RRV) and outcomes Pulmonary, Critical Care and Sleep Medicine Division, The George Washington in these patients. Design: A longitudinal, prospective, observational University MFA, 2150 Pennsylvania Ave, study of patients mechanically ventiNW, Washington DC 20037, USA e-mail: ggutierrez@mfa.gwu.edu lated for at least 12 h performed in a Fax: ?1-202-7412238 medical-surgical intensive care unit. Patients were enrolled within 24 h of rkan H. Tu Department of Anesthesiology, TUBITAK the initiation of ventilatory support. We measured airway signals contin lhane Military Research Scholar, Gu uously for the duration of ventilatory Medical Faculty, Ankara, Turkey support and calculated expiratory M. Wulf-Gutierrez ow frequency spectra at 2.5-min Department of Obstetrics and Gynecology, intervals. We assessed RRV using the Georgetown University, Washington DC, amplitude ratio of the ow spectrums Keywords Continuous monitoring USA rst harmonic to the zero frequency Patient-ventilator asynchrony component. Measures of the ampliH. Kaya Sedation Neuromuscular blockers tude ratio were averaged over the lhane Military Pulmonary Division, Gu total monitored time. Patients with Medical Faculty, Ankara, Turkey

time-averaged amplitude ratios \40 % were classied as high RRV and those C40 % as low RRV. Allcause mortality rates were assessed at 28 and 180 days from enrollment with a Cox proportional hazards model adjusted for disease acuity by the simplied acute physiology score II. Results: We enrolled 178 patients, of whom 47 had high RRV and 131 low RRV. Both groups had similar disease acuity upon enrollment. The 28- and 180-day mortality rates were greater for low RRV patients with hazard ratios of 4.81 (95 % CI 1.8512.65, p = 0.001) and 2.26 (95 % CI 1.214.20, p = 0.01), respectively. Independent predictors of 28-day mortality were low RRV, i.v. vasopressin, and SAPS II. Conclusions: Decreased RRV during ventilatory support is associated with increased mortality. The mechanisms responsible for this nding remain to be determined.

Introduction

the patients intrinsic breathing rhythm is an accepted clinical goal. This ventilatory strategy, however, is During mechanical ventilation, achieving a harmonious, in marked contrast with the considerable respiratory regular breathing pattern by entraining machine cycling to rate variability (RRV) characteristic of spontaneously

1360

breathing humans [1], and usually requires the administration of analgesics, sedatives, and occasionally of neuromuscular blockers [2]. The impact of RRV on clinical outcomes, in particular short- and long-term mortality, is unknown. A major challenge in trying to assess the relationship of RRV to outcomes is its chaotic nature, since alterations in respiratory patterns occur haphazardly with unpredictable duration. To effectively classify patients as having either high or low RRV requires knowledge of respiratory ow patterns for the duration of ventilator support, information only available through continuous monitoring of airway signals. An accurate method to determine alterations in periodicity of recurrent signals, such as those generated during breathing, is time series spectral analysis [3]. We previously described a method to monitor RRV continuously and non-invasively [4] by applying spectral analysis to the expiratory portion of the ow signal to generate sequential frequency spectra. Figure 1 shows changes in expiratory ow spectra according to RRV. During quiet, regular breathing, the frequency spectrum is characterized by an ensemble of sharply dened peaks centered at frequency multiples of the mean respiratory rate (Fig. 1a). Increases in RRV dissipate signal power, resulting in wider and lower peaks (Fig. 1b). With further increases in RRV the spectrum loses all major identifying characteristics (Fig. 1c). Previously, we proposed using the amplitude ratio of the spectrums rst harmonic (H1) to its zero frequency or DC component (H1/DC) as a parameter to characterize the degree of RRV [4]. The aim of the present study was to determine the relationship of RRV, dened by alterations in expiratory ow spectra, to clinical outcomes in mechanically ventilated patients. We monitored RRV by computing H1/DC at 2.5-min intervals during most of their time on ventilatory support. All-cause mortality rates were determined at 28 and 180 days from the time of enrollment [5].

mechanical ventilation, death or transfer to a ventilator long-term care facility (VLTCF). Patients were ventilated with Maquet Servoi or Servos ventilators (Maquet Critical Care, Solna, Sweden). Physicians not involved in the study determined all treatment modalities, including ventilatory parameters. We excluded from analysis patients who either died or who were extubated \12 h from enrollment. Patient demographics, ICU admission diagnoses, simplied acute physiology score II (SAPS II) [6], sequential organ failure assessment score (SOFA) [7], Glasgow coma scale (GCS) score modied for intubated patients (verbal score as one) [8], hemodynamic data, laboratory values, and ventilator settings were determined at the time of enrollment. All data were de-identied at the conclusion of the study. Data acquisition and frequency spectral analysis were performed in real time with software written specically for this purpose. Respiratory ow and pressure were sampled digitally at 30 Hz using the servo ventilator computer interface emulator [9]. The signals were not ltered. Consecutive respiratory ow frequency spectra were generated every 2.5 min using the CooleyTukey fast Fourier transform [10]. For each spectrum, we calculated H1/DC with a peak detection algorithm. The mode of ventilation, inspired O2 fraction (FIO2), tidal volume, minute ventilation, positive end-expiratory pressure (PEEP), and mean and peak airway pressures were also recorded at 2.5 min intervals. Hourly hemodynamic measurements, i.v. medications, and laboratory data were transcribed from the ICU chart. Cardiac output was measured in a subset of patients with an arterial pulse-wave contour device (FloTrac; Edwards Lifesciences, Irvine, CA, USA). All hemodynamic and ventilatory data, including H1/DC, were averaged for the duration of monitoring, resulting in a time-averaged value for each recorded variable. Mean values for blood gases were calculated for patients having two or more arterial blood gases drawn while monitored. Patient classication We classied patients prospectively as having high RRV (RRVHigh) when their time-averaged H1/DC was \40 % and with low RRV (RRVLow) for time-averaged H1/DC C40 %. This H1/DC value was previously noted to have 90 % specicity in identifying asynchronous ventilation dened as asynchrony index C10 % [4]. Outcomes Primary outcomes were all-cause mortality at 28 and 180 days from the day of enrollment. Survival status was determined from hospital from clinic records, and from

Methods
This was a longitudinal, prospective, observational study of adult mechanically ventilated patients conducted from March 2011 to January 2012 at The George Washington University Hospital ICU. The study was approved by The George Washington University Institutional Review Board (IRB No. 110910). Given the observational nature of the project, the IRB determined that informed consent was implied, provided that patients or their surrogates could opt for removal from the study. We enrolled patients consecutively within 24 h of initiation of mechanical ventilation, or in the case of postsurgical patients, within 24 h of ICU admission. Patients were monitored from enrollment until weaning from

1361

Fig. 1 Respiratory ow and corresponding frequency spectra obtained at different times from a mechanically ventilated patient with pancreatitis and ARDS. a The patient breathes with monotonic regularity at a mean respiratory rate (RR) of 18 bpm. The frequency spectrum shows a series of sharply dened peaks, with the rst harmonic (H1) having the greatest amplitude and centered at the mean RR (H1/DC = 64 %). b One hour later, there is some loss of breathing regularity resulting in broader spectral peaks of

lower amplitude. H1 has shifted leftward and is now centered at the new RR of 14 bpm (H1/DC = 40 %). c Fifteen minutes later, there is a noticeable increase in RRV (mean RR = 17 bpm), accompanied by changes in the expiratory portion of the signal suggestive of ineffective triggering (downward pointing arrows). The spectrum has lost much of its characteristic peaked pattern as H1/DC declines to 27 %

telephone interviews. Secondary outcomes were ICU and hospital lengths of stay, measured from the time of enrollment, and time on mechanical ventilation, measured from the time of initiation of mechanical ventilation to

weaning or VLTCF transfer. We also assessed the incidence of ventilator-associated pneumonia (pneumonia that arises more than 4872 h after endotracheal intubation) [11], barotrauma (newly developed pneumothorax,

1362

pneumomediastinum or subcutaneous emphysema), sep- Results sis or septic shock [12], tracheostomy, unplanned extubation, failure to wean (reintubation within 48 h of We enrolled 215 patients and excluded from analysis 37 weaning), and transfer to a VLTCF. patients who died or were extubated within 12 h of the initiation of monitoring (Online Supplement, eFig. 1). These patients (n = 178) were monitored for 29,343 h, generating 602,784 frequency spectra. Individual patients Statistics were monitored for 76 (30164) h; minimum 15 h and maximum 1,100 h. Time from intubation to enrollment The MannWhitney test was used to determine signicant was 13.1 (818) h, with total time on mechanical ventidifferences between independent samples. A worst-case lation of 101 (48201) h. Overall monitoring efciency sensitivity analysis of primary outcome variables was was 82 (6790) % of mechanical ventilation time. performed to account for missing data [13]. Serious There were 47 RRVHigh and 131 RRVLow patients, for a adverse events and medications were compared in the two prevalence of RRV High of 26.4 %. RRVHigh patients spent 66 2 groups with Pearsons v test [14]. The Benjamini (5681) % of the monitored time with H /DC \40, com1 Hochberg procedure [15] with a false discovery rate of pared to 14 (551) % for the RRV Low group (p \ 0.001). 0.20 was used to control for multiple comparisons. Other than a lower body mass index for the RRV High group, A logistic regression analysis was performed to predict there were no differences in demographics (Table 1). Both mortality at 28 days with initial predictors that included groups had comparable disease acuity at enrollment in the variables with univariate p \ 0.30. A manual stepwise study, as evidenced by their similar SAPS II, SOFA scores, procedure was used to eliminate variables with p [ 0.40 at modied GCS scores and arterial PO /F O . 2 I 2 each step. We determined the effect of RRV on primary There were no differences between the groups in terms outcomes with a Cox multivariate proportional hazards of medical diagnoses or surgical procedures, except for a model, with enrollment SAPS II as a covariate to account greater percentage of cardio-thoracic surgical procedures for disease acuity. Crude mortality curves at 28, 60, 90, performed in the RRV High group (Online Supplement, and 180 days were compared using GehanBreslow sur- eTable 1). The ethnic composition of the patient popuvival analysis. A sample size of 164 patients was projected lation, as well as medical and lifestyle risk factors were to detect a 20 % difference in mortality rates between the similar (Online Supplement, eTable 2). There were no groups with 90 % power and a two-sided p value of 0.05. differences in laboratory values at the time of enrollment Unless otherwise specied, data are shown as median and between the groups (Online Supplement, eTable 3). interquartile range. All reported p values are two-sided The duration of mechanical ventilation and monitored with p \ 0.05 considered signicant. time were similar for both groups (Table 2). Most patients
Table 1 Demographics and clinical description of patients at enrollment RRVHigh (n = 47) Demographics Age (years) BMI Female gender (%) Disease acuity SAPS II score SOFA score GCS score Enrollment PaO2/FIO2 Admission type (%)a Medical Scheduled surgical Unscheduled surgical Reason for ventilation (%)a Respiratory failure Airway protection Post-op complication 64 (5175) 24.9 (21.628.4) 29.8 44 (3752) 6.0 (3.5 8.0) 8 (710) 327 (234459) 83.0 8.5 8.5 63.8 27.7 8.5 RRVLow (n = 131) 57 (4572) 27.3 (23.031.0) 44.3 41 (3352) 6.0 (4.0-8.5) 8 (610) 303 (163439) 75.6 7.6 16.8 66.4 26.7 6.9 p value 0.13 0.032* 0.08 0.36 0.41 0.66 0.27 0.30 1.00 0.17 0.80 0.75 0.88

modied to include only best eye and best motor response, with Unless stated, gures are median (interquartile range) RRVLow patients with low respiratory rate variability, RRVHigh best verbal response = 1 patients with high respiratory rate variability, BMI body-mass * p \ 0.05 index, SAPS II simplied acute physiology score, SOFA the sepsis- a Percent of patients in the group related organ failure assessment, GCS Glasgow coma scale

1363

were ventilated with two or more ventilatory modes. Pressure-regulated assist volume control (PRVC) was the most frequently used mode. Airway pressure release ventilation (APRV) was used exclusively in 16 % of RRVLow patients, accounting for 6.8 % of the monitored time (p = 0.003). There were no signicant differences between the groups in the proportion of patients receiving other ventilatory modes. The RRVLow group had lower respiratory rates and greater airway peak pressures. There were no differences in the other monitored ventilatory or hemodynamic variables or in arterial blood gases while on ventilatory support (Online Only Supplement, eTable 4). Outcomes Two patients were lost to follow-up prior to 28 days and another four patients prior to 180 days. All missing patients belonged to the RRVLow group. Missing patients were excluded in the calculation of mortality rates. A KaplanMeier survival plot comparing both groups is shown in Fig. 2. The survival curves separate early, with crude all-cause mortality being greater for the RRVLow group at 28 days (28.7 vs. 10.6 %, p = 0.013) by GehanBreslow survival analysis. A worst-case sensitivity analysis did not alter the statistical inference. Assuming that all missing persons were either alive or
Table 2 Mechanical ventilation data

dead at 28 days resulted in p values of 0.015 and 0.009, respectively. There were no differences between the groups in any of the secondary outcome measures (Online Supplement, eTable 5). A Cox multivariate proportional hazards model with SAPS II as a covariate yielded hazard ratios for RRVLow patients of 4.81 (95 % CI 1.8512.65, p = 0.001) for death at 28 days and 2.26 (95 % CI 1.214.20, p = 0.010) for death at 180 days. Predictors independently associated with greater 28-day mortality were RRVLow with OR 6.08 (1.6722.20, p = 0.006), i.v. vasopressin OR 4.59 (1.7312.15, p = 0.002), and SAPS II OR = 1.08 (1.031.13, p = 0.001). Predictors not independently associated with greater 28-day mortality were age, BMI, mean and peak airway pressures, respiratory rate, PEEP and the use of i.v. cisatracurium (Online Supplement, eTable 6). Patients with the lowest RRV also had the greatest 28-day mortality rates. The upper panel of Fig. 3 shows the percentage of patients with known 28-day outcome (n = 176) distributed according to their H1/DC values. The histogram does not deviate signicantly from normality (Skewness = 0.7, Kurtosis = 0.7, KolmogorovSmirnov p [ 0.15) and is centered on H1/DC4049 %. Conversely, 28-day mortality rates increase exponentially with increases in H1/DC (Mortality = e0:06H1 =DC , R2 = 0.80), being greatest in patients with H1/DC C60 % (Fig. 3, bottom panel).

RRVHigh (n = 47) Duration of mechanical ventilation (h) Ventilatory support time Monitored time Range of monitored time Mode of ventilationa PRVC AVC APC PS APRV NAVA Average ventilatory variablesb FIO2 PEEP (cm-H2O) Tidal volume (mL kg-1 ideal body weight) Respiratory rate (bpm) Mean airway pressure (cm-H2O) Peak airway pressure (cm-H2O) Minute ventilation (L min-1) 129.1 (48.5225.9) 77.0 (29.5197.5) 16759 85.1 (58.6) 31.9 (20.3) 46.8 (14.0) 76.6 (7.1) 0.0 (0.0) 4.3 (0.0) 41.2 (39.947.4) 4.9 (4.75.2) 6.6 (6.08.0) 17.5 (15.519.6) 10.1 (9.112.1) 23.2 (20.826.7) 9.4 (7.610.4)

RRVLow (n = 131) 96.2 (48.0200.9) 75.0 (29.0167.5) 191100 72.5 (60.0) 38.2 (23.9) 39.7 (7.7) 67.9 (1.1) 16.0 (6.8) 1.5 (0.0) 41.4 (40.049.6) 5.0 (4.95.3) 6.5 (5.57.8) 16.4 (14.619.0) 10.6 (9.512.9) 26.4 (23.129.4) 8.6 (7.410.1)

p value 0.64 0.64 0.08 0.45 0.40 0.27 0.003** 0.28 0.26 0.06 0.22 0.021* 0.09 0.001** 0.10

a Unless stated, gures are median (interquartile range) Percentage of patients in which a specic mode of ventilation was RRVLow low respiratory rate variability, RRVHigh high respiratory used; in parentheses are the percentages of monitored time that the rate variability, FIO2 inspired O2 fraction, PEEP positive end- mode of ventilation was used. Statistics refer to Pearsons v2 test expiratory pressure, PRVC pressure-regulated volume control, AVC comparing percent of patients assist volume control, APC assist pressure control, PS pressure b Mean values computed during the time patients were monitored support, APRV airway pressure release ventilation, NAVA neurally adjusted ventilatory assist * p \ 0.05, ** p \ 0.01

1364

Fig. 2 KaplanMeier survival plot for RRVLow (dashed line, n = 126) and RRVHigh patients (solid line, n = 47). A Cox multivariate proportional hazards model with SAPS II as a covariate yielded hazard ratios for RRVLow patients of 4.81 (95 % CI 1.8512.65, p = 0.001) for death at 28 days and 2.26 (95 % CI 1.214.20, p = 0.010) for death at 180 days from the day of enrollment. The survival curves separate early, suggesting the effect of an acute therapeutic intervention

There were no signicant differences in i.v. medications administered during the time patients were monitored, except for greater use of the neuromuscular blocker cisatracurium in RRVLow patients (Online Supplement, eTable 7). Similar proportions of patients in both groups were treated with vasoactive agents (49.6 vs. 46.8 %, p = 0.74) and with i.v. sedatives (96.2 vs. 97.9 %, p = 0.58). We performed a survival analysis using multivariate proportional hazards model with SAPS II as a covariate in a subset of patients not treated with cisatracurium (n = 155). Once more, we found signicant hazard ratios in RRVLow patients for death at 28 days, 4.16 (95 % CI 1.5411.27, p = 0.005) and death at 180 days, 2.11 (95 % CI 1.094.07, p = 0.027). There were no differences in enrollment disease acuity, the use of vasopressor agents, nor i.v. sedatives between the low and high RRV groups in this patient subset.

Fig. 3 Distribution of patients and their associated 28-day mortality rate according to H1/DC value (upper panel). The percentage of patients with known 28-day outcome (n = 176) distributed according to their H1/DC values. The shape of the histogram approximates a Gaussian distribution centered on H1/DC4049 % (lower panel) 28-day mortality rates according to H1/DC interval. Mortality rates increase nearly exponentially as a function of H1/ DC, being greatest in patients displaying a highly regular breathing pattern (H1/DC [60 %)

Discussion
Ours is the rst study relating alterations in RRV to mortality in mechanically ventilated patients. We noted that patients with low RRV during most of their time on ventilatory support were at greater risk of dying than

those with high RRV. This observation runs counter to clinical practice promoting the maintenance of regular, periodic breathing during mechanical ventilation [16, 17]. We characterized alterations in RRV from successive respiratory ow frequency spectra calculated at 2.5-min intervals for[80 % of the time patients were on ventilatory support. By time-averaging H1/DC values, we could classify patients as primarily high or primarily low RRV with a high degree of certainty. Since high RRV was present in all patients at one time or another during mechanical ventilation, this method avoided misclassifying them on the basis of data acquired during short time-windows. This approach, however, does not preclude the possibility of future studies

1365

using shorter time-windows to explore the relationship of RRV to specic outcome measures. We found that 26.4 % of our patients could be classied as RRVHigh patients, a prevalence similar to that reported for asynchronous ventilation by others [1820]. There were no signicant differences between the groups in secondary outcomes, but we noted a trend towards longer duration of mechanical ventilation and hospital stay in the RRVHigh group, a nding consonant with the results of Thille et al. [18] and de Wit et al. [19] in asynchronous patients. We chose H1/DC = 40 % prospectively as the boundary separating high and low RRV patients. Given the paucity of clinical data on the use of spectral analysis in mechanically ventilated patients, we based this choice on our previous work comparing H1/DC to the presence of asynchronous ventilation [4]. The validity of H1/ DC = 40 % as the boundary limit was conrmed post hoc by a detailed sensitivity analysis based on the mortality data from the present study (see Online Supplement). According to that sensitivity analysis, any boundary limit for H1/DC in the range of 34.942.3 % would have yielded a statistically signicant difference (p \ 0.05) in crude 28-day mortality rate between low and high RRV groups, with H1/DC = 41. 6 % being the optimal boundary limit (p = 0.003). Given the observational nature of the study, causality for the greater mortality rates noted in RRVLow patients cannot be established from our results. Both groups were alike in most respects, including demographics, admission type, and enrollment disease acuity. They were ventilated with similar modes for more than 90 % of monitored time. Although RRVLow patients had signicantly lower respiratory rates and greater peak airway pressures, these and other measured hemodynamic or ventilatory variables were not independent predictors of mortality. There were no differences in disease acuity at enrollment between the RRVLow and RRVHigh groups. A glaring difference was the greater use of i.v. cisatracurium besylate in RRVLow patients. It is possible that patients treated with cisatracurium may have comprised a sicker population, one not differentiated at enrollment by SAPS II, SOFA, GCS scores, or PaO2/FIO2 ratio. To guard against this possibility, we performed a secondary analysis excluding patients treated with cisatracurium at any point during their time on ventilatory support, and found hazard ratios to be mostly unaltered. RRVLow patients never treated with neuromuscular blockers also experienced a signicantly greater risk of death at 28 and 180 days. This nding, as well as that of cisatracurium administration not being independently associated with greater mortality, suggests that a decrease in RRV per se, rather than the use of neuromuscular blockers, may have been primarily responsible for the greater mortality rates noted in the RRVHigh group. Subsequent randomized studies are required to test this hypothesis.

Patients may breathe irregularly for a variety of reasons. In fact, variability is a common feature of physiological processes [21]. Normal heart function is characterized by irregularities in heart rate, with decreased heart rate variability being associated with severe coronary artery disease and poorer outcome following a myocardial infarction [22]. In utero, the absence of heart rate variability can be associated with profound acidosis [23]. Vagus nerve stimulation not only promotes heart rate variability [24], it also appears to attenuate the systemic inammatory response as macrophage activation is inhibited through the parasympathetic outow tract [25, 26]. Spontaneously breathing patients with greater RRV show improved weaning success [27]. Perhaps patients with increased RRV benet from greater parasympathetic tone, which in turn could lead to the upregulation of cholinergic anti-inammatory pathways [28]. Endotoxin administration to healthy volunteers is known to decrease RRV, a response possibly mediated through the cyclooxygenase pathway [29]. Another possibility to consider is that irregular breathing may promote capillary recruitment and redistribution of perfusion towards better aerated lung tissue [30], thus inhibiting the release of proinammatory cytokines as regional oxygenation improves [31]. Perhaps a high RRV represents a favorable response to critical illness, and blunting this response with neuromuscular blockers or heavy i.v. sedation may be counterproductive. Further insight into these mechanisms must await future studies. Major strengths of the study are its prospective, longitudinal nature and the large number of patients enrolled. We recorded airway signals and modes of ventilation continuously for over 80 % of the time patients were on ventilatory support. During that time, we also documented changes in hemodynamic data, medications, and laboratory results on an hourly basis. Measuring respiratory ow using the data acquisition system of the mechanical ventilator proved to be a robust, practical technique that lends clinical validity to our data. Among the weaknesses of the study are its observational nature and being a single-center study. Care must, therefore, be exercised in generalizing these ndings until corroborated by a larger multicenter trial. It is possible that our ndings may not apply to patients treated with neuromuscular blockers other than cisatracurium besylate, since the latter was the only paralytic agent used by the treating physicians in the study. Moreover, our method is restricted to measuring RRV from changes in expiratory airway ow. Perhaps other descriptors of breathing pattern, such as tidal volume, could also have similar utility. In conclusion, we found that decreased RRV in mechanically ventilated patients is associated with greater mortality rates. The degree and duration of RRV that may be safely achieved with i.v. sedatives in order to promote patient comfort remains to be determined.

1366

lya Turkan, MD, Beyzaei-Arani, MD: No conicts of interest. Hu PhD: No conicts of interest. Marian Wulf-Gutierrez, MD: No conicts of interest. Katherine Rider, MD: No conicts of interest. Hatice Kaya, MD: No conicts of interest. Richard Amdur, PhD: No conicts of interest. Data Access and Responsibility: Guillermo Gutierrez, MD, PhD and Richard Amdur, PhD had full access to all the data in the study and take responsibility for the integrity of the Conicts of Interest Guillermo Gutierrez, MD, PhD: Has applied data and the accuracy of the data analysis. for a US patent regarding the method used to determine RRV. No other conicts of interest. Aparna Das, MD: No conicts of interest. Guillermo Ballarino, MD: No conicts of interest. Arshan Acknowledgments The authors wish to recognize C.S.H., a former patient and friend, whose remarkably keen observations while undergoing mechanical ventilation provided much of the inspiration that led to this research. Portions of this work were presented in abstract form at the American Thoracic Society International Conference, May 2012, San Francisco.

References
1. Tobin MJ, Mador J, Guenther SM, Lodato RF, Sackner MA (1988) Variability of resting respiratory drive and timing in healthy subjects. J Appl Physiol 65:309317 2. Hall JB, Schweickert W, Kress JP (2009) Role of analgesics, sedatives, neuromuscular blockers, and delirium. Crit Care Med 37:S416S421 3. Priestley MB (1983) Spectral analysis and time series. volumes I and II (11th printing). Academic: San Diego, pp 39 4. Gutierrez G, Ballarino GJ, Turkan H, Abril J, De La Cruz L, Edsall C, George B, Gutierrez S, Jha V, Ahari J (2011) Automatic detection of patientventilator asynchrony by spectral analysis of respiratory ow. Crit Care 15:R167 5. Gutierrez G, Ballarino G, Das A, Turkan H (2012) Decreased mortality in asynchronous mechanically ventilated patients. Am J Respir Crit Care Med 185:A1704 6. Le Gall JR, Lemeshow S, Saulnier F (1993) A new simplied acute physiology score (SAPS II) based on a European/North American multicenter study. JAMA 270:29572963 7. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonc a A, Bruining H, Reinhart CK, Suter PM, Thijs LG (1996) The SOFA (Sepsis-related organ failure assessment) score to describe organ dysfunction/failure. On behalf of the working group on sepsis-related problems of the European society of intensive care medicine. Intensive Care Med 22:707710 8. Wu YK, Tsai YH, Lan CC, Huang CY, Lee CH, Kao KC, Fu JY (2010) Prolonged mechanical ventilation in a respiratory care setting: a comparison of outcome between tracheotomized and translaryngeal intubated patients. Crit Care 14:R26 9. http://www.maquet.com/content/ Documents/Connectivity/SERVOi_ SERVREFMANComputerInterface Emulator002ReferenceManual_EN_ All.pdf. Last Accessed March 11, 2013 10. Duhamel P, Vetterli M (1990) Fast Fourier transforms: a tutorial review and a state of the art. Signal Process 19:259299 11. American Thoracic Society; Infectious Diseases Society of America (2005) Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcareassociated pneumonia. Am J Respir Crit Care Med 171:388416 12. American College of Chest Physicians/ Society of Critical Care Medicine (1992) Consensus conference: denitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 20:864874 13. Unnebrink K, Windeler J (1999) Sensitivity analysis by worst and best case assessment: is it really sensitive? Drug Inf J 33:835839 14. Zar J (2009) Biostatistical analysis, 5th edn. Prentice Hall, Upper Saddle River, pp 452459 15. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289300 16. Tobin MJ (2001) Advances in mechanical ventilation. N Engl J Med 344:19861996 17. MacIntyre NR (2011) Patient-ventilator interactions: optimizing conventional ventilation modes. Respir Care 56:7384 18. Thille AW, Rodriguez P, Cabello B, Lellouche F, Brochard L (2006) Patient-ventilator asynchrony during assisted mechanical ventilation. Intensive Care Med 32:15151522 19. de Wit M, Pedram S, Best AM, Epstein SK (2009) Observational study of patient-ventilator asynchrony and relationship to sedation level. J Crit Care 24:7480 20. Georgopoulos D, Prinianakis G, Kondili E (2006) Bedside waveforms interpretation as a tool to identify patient-ventilator asynchronies. Intensive Care Med 32:3447 21. Goldberger AL (2001) Heartbeats, hormones, and health: is variability the spice of life? Am J Respir Crit Care Med 163:12891290 22. Kleiger RE, Miller JP, Bigger JT Jr, Moss AJ, The Multicenter PostInfarction Research Group (1987) Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. Am J Cardiol 59:256262 23. Macones GA, Hankins GD, Spong CY, Hauth J, Moore T (2008) The 2008 National institute of child health and human development workshop report on electronic fetal monitoring: update on denitions, interpretation, and research guidelines. Obstet Gynecol 112:661666 24. Sloan RP, McCreath H, Tracey K, Sidney S, Liu K, Seeman T (2007) RR interval variability is inversely related to inammatory markers: the CARDIA study. Mol Med 13:178184 25. Haensel A, Mills PJ, Nelesen RA, Ziegler MG, Dimsdale JE (2008) The relationship between heart rate variability and inammatory markers in cardiovascular diseases. Psychoneuroendocrinology 33:13051312 26. Borovikova LV, Ivanova S, Zhang M, Yang H, Botchkina GI, Watkins LR, Wang H, Abumrad N, Eaton JW, Tracey KJ (2000) Vagus nerve stimulation attenuates the systemic inammatory response to endotoxin. Nature 405:458462 27. Wysocki M, Cracco C, Teixeira A, Mercat A, Diehl JL, Lefort Y, Derenne JP, Similowski T (2006) Reduced respiratory rate variability as a predictor of unsuccessful patient separation from mechanical ventilation. Crit Care Med 34:20762083 28. Suki B, Alencar AM, Sujeer MK, Lutchen KR, Collins JJ, Andrade JS Jr, Ingenito EP, Zapperi S, Stanley HE (1998) Life-support system benets from noise. Nature 393:127128

1367

29. Preas HL 2nd, Jubran A, Vandivier RW, Reda D, Godin PJ, Banks SM, Tobin MJ, Suffredini AF (2001) Effect of endotoxin on ventilation and breath variability: role of cyclooxygenase pathway. Am J Respir Crit Care Med 164:620626

30. Gama de Abreu M, Spieth PM, Pelosi P, 31. Mutch WA, Eschun GM, Kowalski SE, Graham MR, Girling LG, Lefevre GR Carvalho AR, Walter C, Schreiber bler (2000) Biologically variable or Ferstl A, Aikele P, Neykova B, Hu naturally noisy mechanical ventilation M, Koch T (2008) Noisy pressure recruits atelectatic lung. Am J Respir support ventilation: a pilot study on a Crit Care Med 162:319323 new assisted ventilation mode in experimental lung injury. Crit Care Med 36:818827