Author + information
- Received July 25, 2016
- Revision received January 17, 2017
- Accepted January 18, 2017
- Published online December 4, 2017.
- Ethan J. Anderson, PhDa,b,
- Jimmy T. Efird, PhD, MScc,d,e,f,∗ (, )
- Andy C. Kiser, MDd,
- Patricia B. Crane, PhD, RNe,
- Wesley T. O’Neal, MD, MPHg,
- T. Bruce Ferguson, MDd,
- Hazaim Alwair, MDd,
- Kendal Carterc,
- J. Mark Williams, MDd,
- Anil K. Gehi, MDh and
- Alan P. Kypson, MDi
- aDepartment of Pharmaceutical Sciences and Experimental Therapeutics, College of Pharmacy, Fraternal Order of Eagles Diabetes Research Center, University of Iowa, Iowa City, Iowa
- bDepartment of Pharmacology and Toxicology, Brody School of Medicine, East Carolina University, Greenville, North Carolina
- cCenter for Epidemiology and Outcomes Research, East Carolina Heart Institute, Greenville, North Carolina
- dDepartment of Cardiovascular Sciences, Brody School of Medicine, East Carolina University, Greenville, North Carolina
- eOffice of the Dean of Research, College of Nursing, East Carolina University, Greenville, North Carolina
- fCenter for Clinical Epidemiology and Biostatistics, School of Medicine and Public Health, University of Newcastle, Newcastle, New South Wales, Australia
- gDepartment of Medicine, Division of Cardiology, Emory University, Atlanta, Georgia
- hDepartment of Medicine, Division of Cardiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
- iREX Cardiac Surgical Specialists, University of North Carolina Health Care, Raleigh, North Carolina
- ↵∗Address for correspondence:
Dr. Jimmy T. Efird, Center for Epidemiology and Outcomes Research, East Carolina Heart Institute, 115 Heart Boulevard, Greenville, North Carolina 27834.
Objectives This study sought to determine whether plasma catecholamines and monoamine oxidase-B (MOA-B) are associated with post-operative atrial fibrillation (POAF) in patients undergoing elective cardiac surgery.
Background Although intra- and post-operative adrenergic tone has been demonstrated to be an causative factor for POAF, the role and association of pre-operative plasma catecholamines remains unclear.
Methods Prior to administration of anesthesia on the morning of surgery, blood samples were obtained from 324 patients undergoing nonemergent coronary artery bypass graft and/or aortic valve surgery with cardiopulmonary bypass at East Carolina Heart Institute. The concentrations of norepinephrine (NE), dopamine (DA), epinephrine (EPI), and enzyme MAO-B were assessed in platelet-rich plasma. A log-binomial regression model was used to determine the association between quartiles of these variables and POAF.
Results Levels of NE (p = 0.0006) and EPI (p = 0.047) in the 4th quartile () were positively associated with POAF, whereas DA (p = 0.0034) levels in the 4th quartile ( were inversely associated with POAF. Adjusting for age, heart failure (HF), and history of atrial fibrillation, the composite pre-operative (adrenergic) plasma marker () was associated with a 4-fold increased occurrence of POAF (adjusted p = 0.0001). No association between plasma MAO-B and POAF was observed.
Conclusions Our results suggest that pre-operative adrenergic tone is an important factor underlying POAF. This information provides evidence that assessment of plasma catecholamines may be a low-cost method that is easy to implement for predicting which patients are likely to develop POAF. More investigation in a multicentric setting is needed to validate our results.
Atrial fibrillation (AF) is a prevalent and costly complication of open-heart surgery (1). The benefits of patients remaining in sinus rhythm include lower hospital readmission rates, reduced length of stay, and fewer post-operative complications (e.g., stroke, myocardial infarction, heart failure, ventricular arrhythmias, and death) (2–6). Thus, there is a need for a convenient, noninvasive, and cost-effective means for predicting which patients may develop post-operative atrial fibrillation (POAF). This is especially important during the pre-operative window when maximum benefit of pharmacologic therapy management of POAF can still be achieved.
Catecholamines are vital to cardiac electromechanical function owing to regulation of intracellular Ca2+ through β-adrenoceptor activation. Increased catecholamines are arrhythmogenic and have been reported to be key factors underlying POAF (2,7,8). Although anesthesia dampens the effect of catecholamines on the heart during surgery, a rebound effect is thought to occur following cardiopulmonary bypass, leading to a sympathetic surge that can propel the heart into AF (7,9,10). For this reason, contemporary post-operative standard of care includes use of β-blockers following cardiac surgery.
In the intra- and post-operative states, catecholamine overload, inflammatory cytokines, metabolic imbalance, and impaired cardiomyocyte energetics have been implicated as causal factors in POAF (8,11–13). Independent of their inotropic effects, catecholamines also are capable of generating substantial oxidative stress through their metabolism by monoamine oxidase (MAO) (14,15). In a previous study, we established that MAO activity in atrial tissue is associated with POAF in patients undergoing cardiac surgery (16). Because MAO is the principal enzyme responsible for catecholamine metabolism in the myocardium, we hypothesized that pre-operative plasma MAO isoform B (MAO-B), expressed in platelets and catecholamine levels (dopamine, epinephrine, norepinephrine), also would be associated with POAF.
The study was approved by the Institutional Review Board of Brody School of Medicine at East Carolina University (UMCIRB09-0669). Study design followed STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines for observational cohort studies (17).
A total of 324 consecutive patients undergoing on-pump elective (nonemergent, nonurgent) coronary artery bypass graft or aortic valve replacement surgeries between August 2014 and January 2016 at the East Carolina Heart Institute were included in the analysis. This cohort included patients who received routine concurrent procedures (such as coronary endarterectomy, internal cardiac defibrillation, femoral-femoral cardiopulmonary bypass, pacemaker insertion, pericardiectomy, repair or restoration of the heart or pericardium, or transmyocardial laser revascularization). Patients undergoing surgical or catheter maze procedures and those with pre-operative shock were excluded from the study. Participants gave written consent in person to 2 members of the research team experienced in hospital patient care. Basic demographics were recorded for patients who did not give consent.
Blood sample collection
Pre-operative blood samples were obtained prior to anesthesia through central intravenous lines placed upon arrival to the perianesthesia care unit (PACU) on the morning of surgery (to account for potential diurnal and nutritional variations in plasma catecholamine levels). Vacutainer tubes containing sodium citrate were used to collect the blood samples. The tubes were placed in a bag of ice and transported to the laboratory within 30 min. The samples were centrifuged for 10 min at 4,000 rpm. The plasma and buffy coat were transferred to another tube and centrifuged at 8,000 rpm for an additional 10 min. Two-thirds of the plasma volume was divided into 2-ml cryotube aliquots for storage. The other one-third of the plasma volume was mixed with the protein-rich pellet and transferred to a 2-ml cryotube. To prevent degradation, ethylenediaminetetraacetic acid was added to the protein-rich plasma. Samples were stored at −80°C.
Diagnostic criteria for AF, POAF, HF, and renal failure
Atrial fibrillation was defined in accordance with American Heart Association/American College of Cardiology (AHA/ACC) 2014 AF Guidelines as any chaotic or irregular supraventricular rhythm that included the absence of a P wave before the QRS complex, a variable rate and rhythm in the atria, and may include an irregular ventricular rhythm (18). Atrial fibrillation was confirmed by 12-lead electrocardiography (ECG). Post-operatively, cardiac rhythm and rate were continuously monitored by ECG in the cardiac intensive care or stepdown unit until hospital discharge. POAF was defined as any episode of AF confirmed by ECG lasting >5 min. Cases of POAF occurring after hospital discharge were not recorded in our database. Because some patients with a history of AF do not develop POAF, history of AF was considered an important prognostic variable for POAF in our analysis. Heart failure (HF) was similarly defined in accordance with AHA/ACC-2013 HF Guidelines as impaired left ventricular function (19). This includes patients with HF with reduced ejection fraction ≤40 and also HF with preserved ejection fraction of >40. Patients with a serum creatinine concentration of ≥4.0 mg/dl, a glomerular filtration rate of <15 ml per minute, and/or who are receiving dialysis therapy were considered to be in renal failure.
Catecholamine measurements in platelet-rich plasma
Pre-operative plasma concentrations of epinephrine (EPI), dopamine (DA), and norepinephrine (NE) were measured using the competitive enzyme-linked immunosorbent assay (ELISA) method with a commercially available assay (Labor Diagnostika Nord, Nordhorn, Germany) according to the manufacturer’s instructions. Briefly, catecholamines were extracted from platelet-rich plasma using a cis-diol-specific boronate affinity gel-coated plate. Following extraction, the catecholamine standards and plasma samples were acylated and enzymatically derivatized. Samples were then transferred to ELISA plates with the appropriate primary antiserum and incubated overnight at 4°C. The assay was completed with a series of wash steps, followed by incubation with a supplied colorimetric substrate. The reaction was stopped with sulfuric acid (0.25 M) prior to reading absorbance at 405 nm. Final concentrations of catecholamine in the plasma samples were determined by fitting values to a standard curve obtained with known concentrations of EPI, DA, and NE, simultaneously determined on each plate and normalized to plasma volume.
Quantitative analysis of MAO-B in platelet-rich plasma
Absolute quantity of MAO-B in platelet-rich plasma was determined using a sensitive ELISA developed in our laboratory. Levels of MAO isoform A (MAO-A) in platelets and blood are undetectable with standard ELISA, so this isoform was not measured in this study. A standard curve of known concentrations of MAO-B was generated with recombinant MAO-B (Sigma-Aldrich, St. Louis, Missouri) and incubated on Immulon-coated 96-well assay plates (Thermo Fisher Scientific; Waltham, Massachusetts) along with diluted plasma samples. Samples were incubated overnight at 4°C and subsequently washed with a solution of phosphate-buffered saline (PBS) plus 0.05% Tween-20 and blocked for 2 hours with 5% bovine serum albumin at 37°C. Samples were then incubated with primary antibody for MAO-B (Abcam, Cambridge, England) for 2 h at 37°C (or overnight at 4°C). Following another wash with PBS plus 0.05% Tween-20, samples were incubated with horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology, Inc.; Dallas, Texas) for 2 h at 37°C. Following an additional wash, samples were incubated with Amplex Red (10 μM; Thermo Fisher). Concentrations of MAO-B enzyme within each sample were determined by fitting to the standard curve of recombinant MAO-B within each plate and normalized to plasma volume.
Categorical variables are frequency and percentages, whereas continuous variables are medians and interquartile ranges (IQR). Statistical significance for categorical variables was computed using the Fisher exact test and the Deuchler-Wilcoxon procedure for continuous variables.
Log-binomial regression was used to directly estimate relative risk (RR) for POAF (20). p Values were computed using a likelihood ratio test and were reported in place of Wald-based confidence intervals (which are known to have poor reliability for small samples). Goodness of fit was assessed by examining Akaike's information criteria and leverage or case-wise diagnostic statistics, generalized to log-binomial regression (21). Stable convergence was achieved for maximal likelihood RR estimates and p values (22). All models satisfied admissibility criteria (i.e., linear predictor constrained to be negative) (20).
The union of plasma DA and NE was identified as the most parsimonious 2-factor composite plasma marker of POAF. Multivariate models were adjusted for age, HF, and history of AF based on their statistical significance in Table 1 (p < 0.001). Other demographic variables and comorbidities were entered into the model in a post hoc fashion. Medications and cardiac function characteristics were not included in our multivariate models because of their potential to be an intermediate variable on a causal path from exposure to outcome (or a descending proxy for an intermediate variable) (23). An iterative expectation-maximization procedure was used to account for missing values, assuming missing-at-random sampling (imputation efficiency of >97%) (24). Conducting a power analysis following study completion was considered inconsistent with standard statistical practice (25). Rounding was performed using the method of Holly and Whittemore (26). SAS version 9.4 software (SAS, Cary, North Carolina) was used for all analyses.
A total of 74 cases of POAF were identified (Table 1). The median age for participants was 64 years of age (IQR = 14), with the majority being male (72%) and white (77%). On average, participants tended to be overweight (median body mass index 29 kg/m2; IQR = 7.6 [not shown in tables]). These characteristics did not substantively differ from those in patients who declined to participate in this study. Age >65 years, history of AF, and HF were positively associated with POAF (p < 0.001). The median pre-operative values for NE and MAO-B significantly differed between patients with and those without diabetes mellitus (Table 2). Patients remaining in sinus rhythm were hospitalized a median of 3 days less than those who developed POAF (p < 0.0001) (Table 3). Use of intra- and post-operative inotrope was more prevalent in the POAF (35%) group than the sinus rhythm (19%) group.
No association between plasma MAO-B content and POAF was observed (unadjusted RR: 0.85; p = 0.35). Plasma NE (unadjusted RR: 2.0; p = 0.0006) and EPI (unadjusted RR: 1.5; p = 0.047) concentrations in the 4th quartile () were positively associated with POAF, whereas DA (unadjusted RR: 2.2; p = 0.0034) in the 4th quartile ( was inversely associated with POAF (not shown in tables). Controlling for age, HF, and history of AF, patients in the 4th quartile for plasma NE () were most likely to develop POAF, whereas those in the 4th quartile for plasma DA () were least likely (Figures 1A and 1B, Table 4). The composite pre-operative marker () was associated with a 4-fold increased incidence of POAF (adjusted p = 0.0001). Patients who gave consent did not significantly differ from those who did not with respect to key demographic characteristics. The post hoc pairwise inclusion of other demographic variables and comorbidities (listed in Table 1) into our multivariate log-binomial models did not substantively change our results.
The current study represents the first to examine the association between pre-operative plasma catecholamines and POAF before the administration of anesthesia. Although several reports have documented an association between sympathetic nervous system activity and POAF, those studies have focused predominantly on the intra- and post-operative period (8,11–13). This is a limitation in the context of POAF prevention, because the time for intervention has already passed.
High levels of catecholamines can be arrhythmogenic, particularly in patients >65 years of age and those with underlying comorbidities such as HF and history of AF, when electrophysiological and structural remodeling (e.g., changes in expression of ion channels, fibrosis) have already occurred in the myocardium (27). Previous studies have shown that these factors are associated with increased circulating NE levels (28–30). When adjusting for these variables in our multivariate analysis, NE remained an independent predictor of POAF. This suggests that, in the pre-operative patient, increased NE levels may reflect an underlying biological predisposition that leaves the patient vulnerable to arrhythmogenic triggers in the post-operative state. Differences in these biological variables, in theory, may explain why certain patients develop POAF when exposed to the stress of cardiac surgery, whereas others remain in sinus rhythm.
Considering the role of catecholamines in the heart, it is necessary to interpret our findings in the context of the cause for POAF, as variations in autonomic tone are known to precede the onset of AF after open-heart surgery (10,31). Catecholamines bind to and activate cardiomyocyte sarcolemmal adrenoceptors, initiating an intracellular signaling cascade that ultimately leads to increases in intracellular Ca+2 and contractility (i.e., inotropic effect). Consequences of this inotropic effect include increased myocardial ATP demand and O2 consumption, increased automaticity and heart rate, and oxidative stress within the cardiomyocytes (Figure 2). However, the net effect of this process depends on the catecholamine type and the density and sensitivity of the adrenoceptors they activate, with sometimes opposing effects on electromechanical function (as was observed with the inverse relationship between NE and DA with POAF in our study) (7).
Both NE and DA share a pathway of synthesis and metabolism. Because DA is the direct precursor to NE, the enzyme dopamine β-hydroxylase is responsible for the final rate-limiting step of NE synthesis (29). Possibly, this conversion is occurring more rapidly in patients exhibiting high NE concentrations, thereby resulting in correspondingly lower DA levels (Figure 1C). Alternatively, there may be different rates of catecholamine re-uptake among patients, which underlie the divergence in their association with POAF. For example, patients with HF have decreased rates and efficiency of catecholamine re-uptake, in addition to the increased sympathetic discharge, that contribute to the high circulating catecholamine levels in these patients (32). In any case, the extent to which catecholamine synthesis, conversion and re-uptake mechanisms contribute to the differences observed in this study is unknown and worthy of future investigation.
MAO is the primary enzyme responsible for metabolism of both of the catecholamines. In our analysis, plasma MAO-B concentration was not found to be associated with POAF. Although activity was not measured, it is conceivable that DA is preferentially being metabolized in systemic organs to greater extent than NE, leading to the inverse association of NE and DA levels with POAF.
We deemed that measuring plasma catecholamines in patients presenting to the PACU would best reflect their adrenergic tone immediately prior to surgery. In this study, pre-operative catecholamines were measured directly from plasma samples. Intuitively, catecholamines at first glance may not appear to be reliable predictive biomarkers of POAF given the many influences determining their levels and activity over time (e.g., pre-operative medications, anxious state of the patient, diurnal and nutritional status, rapid turnover, and other influences). Although these factors do indeed influence catecholamine levels, our objective was to determine whether the actual catecholamine and/or MAO-B level at a single time point closest to surgery (i.e., when patients are in PACU) was associated with POAF, regardless of contributing factors. This would allow for risk stratification and potential use of prophylactic therapy for POAF.
Previous POAF models have been inconclusive or yielded conflicting results (33–35). For example, some were insufficiently powered, whereas others were limited by the exclusion of patients with renal failure and/or HF and other complex conditions (e.g., aortic valve surgery, uncontrolled diabetes) (34). The association between POAF and intra- and post-operative factors such as use of inotrope and cardiopulmonary bypass time is well known, similar to our results (36–38). However, the current study further illustrates the utility of plasma catecholamines in the pre-operative setting.
We recently reported that MAO activity in atrial tissue was associated with POAF (16). In the current study, no association between plasma MAO-B concentrations and POAF was observed. Several factors may explain this result. Only MAO-B concentration was measured in the present analysis, not activity. Conceivably, activity of the enzyme may be independent of its content. Additionally, in our earlier report, we measured MAO activity directly in the atrial tissue, which may indicate a divergence in control of MAO expression and activity in cardiac tissue compared with circulation. Because tyramine was the MAO substrate used in that analysis, the kinetics of MAO activity with physiological monoamine substrates such as NE and DA may be different.
Many patients are anxious prior to open-heart surgery (39). Heightened state of anxiety can lead to increased production and release of stress hormones (e.g., cortisol) and subsequent catecholamine release (9). Although serum cortisol was not directly measured at the time of surgery, we did systematically collect information on catecholamine levels. Nevertheless, because anxiety may, in part, be driving the high pre-operative catecholamine levels in patients who develop POAF, this information would be useful in the context of implementing relaxation strategies among suitable patients. Further studies are needed to evaluate this possibility.
Although our multivariate models were adjusted for history of AF, information was not available for antiarrhythmic drug use in our analysis dataset, as well as data pertaining to duration or type of AF (i.e., atrial flutter, paroxysmal versus persistent AF). The delineation between compensated and decompensated HF also was not taken into account. Although this might have biased our results toward the null, a history of both AF and HF were significant predictive variables in our model. In future studies, it may be informative to delineate the differential effects of various types of atrial arrhythmia and HF on catecholamine levels and POAF. The combined effect of inflammatory cytokines and/or cortisol with catecholamines also would be of potential interest. Similarly, estimated glomerular filtration rate has been associated with increased inflammation and POAF following open-heart surgery (1). Because few patients had renal failure, this variable lacked statistical power to detect a statistically significant association with POAF.
In our analysis, we did not compute sensitivity and specificity because the unbiased estimation of these predictive classification measures would have required that outcome-related covariates (known and unknown) be equally distributed between the POAF and sinus rhythm groups. To our knowledge, there is no robust analytic method free of collapsibility bias to account for such imbalances when computing these classification estimates. Additionally, prediction probability estimates were not provided in our manuscript. We used a binomial family generalized linear model with a log link to allow for direct estimation of RR. This method is generally preferred to logistic regression for estimating RR because of the above-mentioned noncollapsibility issues associated with adjusted odds ratio estimates. Although log-binomial estimates have a meaningful causative interpretation, this method can on occasion yield prediction probabilities outside the allowable domain space.
Clinical implications and future directions
Our results have important scientific and clinical implications. Foremost, they are further evidence of a connection between sympathetic nervous system activity and arrhythmogenesis whereby pre-operative adrenergic tone (plasma NE and/or DA levels) predisposes certain patients to develop POAF. Subsequent research efforts may focus on assessing the utility of strategies such as: 1) selectively administering prophylactic antiarrhythmic medication, particularly β-blockers; 2) identifying patients who would benefit from pharmacological management and stabilization prior to surgery; and 3) prompting closer monitoring of patients during and after surgery. Additionally, this study has potential significance for reducing the number of complications associated with POAF (e.g., stroke, myocardial infarction, HF, other thromboembolic events, and ventricular arrhythmias), decreasing hospital length of stay and lowering healthcare costs.
In our study, a composite pre-operative catecholamine marker measured from blood samples on the morning of surgery was observed to significantly predict the occurrence of POAF. Further investigation is needed to independently validate our results.
COMPETENCY IN MEDICAL KNOWLEDGE: The need to develop convenient, quickly measured, reproducible biomarkers of POAF risk in patients prior to cardio-thoracic surgery remains a challenge. Pre-operative plasma NE and DA levels are plausible candidates that satisfy these criteria and may be useful in developing POAF risk stratification models. Implementation of individualized perioperative drug therapy (e.g., β-blockers, amiodarone) based on plasma NE and DA levels may be beneficial.
TRANSLATIONAL OUTLOOK: Owing in part to their inotropic effects, catecholamines are factors in the post-operative period that contribute to development of POAF. Pharmacologic interventions that specifically target adrenoceptor and catecholamine synthesis/metabolism would elucidate the precise pathways and mechanisms by which catecholamines contribute to POAF. Follow-on studies in multiple institutions are necessary to validate the present study.
The authors thank Cherese Beatty, Preeti Gudimella, and Kathleen Thayne for assistance with consenting patients, tissue and blood collection, data collection, and manuscript editing. The authors also thank all the nurses and staff of the East Carolina Heart Institute, particularly the post-anesthesia care unit.
This study was supported by U.S. National Institutes of Health grant R01HL122863 to Drs. Anderson, Efird, and Kypson. Dr. Gehi is a speaker for St. Jude Medical and Biotronik. All the other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- angiotensin converting enzyme inhibitor
- atrial fibrillation
- angiotensin receptor blocker
- calcium channel blocker
- chronic obstructive pulmonary disease
- enzyme-linked immunosorbent assay
- heart failure
- intensive care unit
- length of stay
- monoamine oxidase
- myocardial infarction
- MV E/A
- ratio of early to late ventricular filling velocities across the mitral valve
- peri-anesthesia care unit
- post-operative atrial fibrillation
- fourth quartile
- not fourth quartile
- Received July 25, 2016.
- Revision received January 17, 2017.
- Accepted January 18, 2017.
- 2017 American College of Cardiology Foundation
- Echahidi N.,
- Pibarot P.,
- O'Hara G.,
- Mathieu P.
- Haider A.H.,
- Dankwa-Mullan I.,
- Maragh-Bass A.C.,
- et al.
- Wu J.H.,
- Marchioli R.,
- Silletta M.G.,
- et al.
- Zakkar M.,
- Ascione R.,
- James A.F.,
- Angelini G.D.,
- Suleiman M.S.
- Sturza A.,
- Duicu O.M.,
- Vaduva A.,
- et al.
- Anderson E.J.,
- Efird J.T.,
- Davies S.W.,
- et al.
- January C.T.,
- Wann L.S.,
- Alpert J.S.,
- et al.
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- Blizzard L.,
- Hosmer D.W.
- Harel O.,
- Stratton J.,
- Aseltine R.
- Goldstein D.S.,
- Eisenhofer G.,
- Kopin I.J.
- Hasking G.J.,
- Esler M.D.,
- Jennings G.L.,
- Burton D.,
- Johns J.A.,
- Korner P.I.
- Eisenhofer G.,
- Friberg P.,
- Rundqvist B.,
- et al.
- Mariscalco G.,
- Biancari F.,
- Zanobini M.,
- et al.
- Zaman A.G.,
- Archbold R.A.,
- Helft G.,
- Paul E.A.,
- Curzen N.P.,
- Mills P.G.
- Fleming G.A.,
- Murray K.T.,
- Yu C.,
- et al.
- O'Neal W.T.,
- Qureshi W.,
- Judd S.E.,
- et al.