Patients with left ventricular assist devices presenting for noncardiac surgery are increasingly commonplace; however, little is known about their outcomes. Accordingly, the authors sought to determine the frequency of complications, risk factors, and staffing patterns.
The authors performed a retrospective study at their academic tertiary care center, investigating all adult left ventricular assist device patients undergoing noncardiac surgery from 2006 to 2015. The authors described perioperative profiles of noncardiac surgery cases, including patient, left ventricular assist device, surgical case, and anesthetic characteristics, as well as staffing by cardiac/noncardiac anesthesiologists. Through univariate and multivariable analyses, the authors studied acute kidney injury as a primary outcome; secondary outcomes included elevated serum lactate dehydrogenase suggestive of left ventricular assist device thrombosis, intraoperative bleeding complication, and intraoperative hypotension. The authors additionally studied major perioperative complications and mortality.
Two hundred and forty-six patients underwent 702 procedures. Of 607 index cases, 110 (18%) experienced postoperative acute kidney injury, and 16 (2.6%) had elevated lactate dehydrogenase. Of cases with complete blood pressure data, 176 (27%) experienced intraoperative hypotension. Bleeding complications occurred in 45 cases (6.4%). Thirteen (5.3%) patients died within 30 days of surgery. Independent risk factors associated with acute kidney injury included major surgical procedures (adjusted odds ratio, 4.4; 95% CI, 1.1 to 17.3; P = 0.03) and cases prompting invasive arterial line monitoring (adjusted odds ratio, 3.6; 95% CI, 1.3 to 10.3; P = 0.02) or preoperative fresh frozen plasma transfusion (adjusted odds ratio, 1.7; 95% CI, 1.1 to 2.8; P = 0.02).
Intraoperative hypotension and acute kidney injury were the most common complications in left ventricular assist device patients presenting for noncardiac surgery; perioperative management remains a challenge.
Patients with left ventricular assist devices presenting for noncardiac surgery are increasingly common and are increasingly managed by noncardiac anesthesiologists. However, little is known about perioperative outcomes.
This study characterized complications, risk factors, and staffing patterns for 246 patients with left ventricular assist devices undergoing 702 noncardiac surgical procedures.
Intraoperative hypotension and acute kidney injury were the most common complications.
THE left ventricular assist device (LVAD) was first successfully implemented in 1988. Over the ensuing three decades, numerous modifications and improvements have been made, allowing LVAD therapy to become increasingly commonplace.1 More recent durable LVAD technologies have utilized smaller devices with continuous flow using either centrifugal or axial flow designs, leading to improved survival.2 Used as either a bridge to heart transplant or a destination therapy (i.e., permanent pump placement), patients with LVADs have made successful recoveries and are managed as outpatients. However, owing to illness and complications, related or unrelated to LVAD therapy, patients have inevitably presented for a wide array of noncardiac surgeries.3–14
Within the patient population presenting for noncardiac surgery (NCS), patients with LVADs present unique perioperative challenges to the anesthesiologist. Common anesthetic concerns include management of anticoagulation,15,16 patient monitoring limitations inherent to minimally pulsatile blood flow,10,17,18 modified surgical access to the operative site,19–23 concern for intraoperative cardiovascular and cerebrovascular events,24–26 management of intraoperative device malfunction,6 and postoperative complications including bleeding15,26–28 and infection.29–34
Currently, numerous studies have characterized the LVAD patient population presenting for NCS and have described management within the preoperative, intraoperative, and postoperative periods.3–5,7–14,17,18,22,35–37 Profiles of this patient population have been described in terms of surgical procedure type, admission status, elective versus urgent/emergent status, as well as the frequency of complications. Although complications have been described, studies have been limited in sample size, with the largest study to date comprising 271 adult cases,13 and have been greatly underpowered when investigating adverse event frequency. Due to small study sizes, the range of surgical procedures described within the LVAD population has also been limited. Additionally, no study to date has currently addressed risk factors for perioperative complications within this patient population. Calls for such a study have been made, as identifying perioperative risk factors would provide a valuable evidence basis for clinical decision-making in this uniquely high-risk population.38
To characterize the LVAD patient population presenting for NCS, we performed this retrospective observational study at our tertiary care facility. We hypothesized that patients with LVADs undergo a steadily increasing frequency of NCS procedures with anesthesia care increasingly from noncardiac anesthesiologists and that variations in perioperative management exist. Additionally, we hypothesized that specific perioperative characteristics exist that place this patient population at increased risk of complications.
Materials and Methods
The institutional review board approved this retrospective study and waived informed consent (HUM00092710; Ann Arbor, Michigan). The primary outcome, data collection process, and statistical analysis plan were prospectively established, presented at a departmental peer-review forum, and registered on August 27, 2014. The Strengthening the Reporting of Observational studies in Epidemiology Statement checklist for reporting observational studies was followed. Data, including demographics, laboratory values, discharge diagnoses, anesthesia history, and physical and intraoperative record, were extracted from our enterprise and departmental electronic health record systems (EPIC© EPIC Systems, USA and Centricity© General Electric, USA). Methods used for data validation, recording, storage, and extraction from a local, single-center database within the Multicenter Perioperative Outcomes Group enterprise have been described elsewhere and utilized in previous studies.39–42
Using the local Society for Thoracic Surgeons LVAD registry (which tracks all LVAD patients regardless of primary implantation hospital), we identified all adult (greater than or equal to 18 yr) patients with durable LVADs of any type receiving care in our health system. We integrated this patient list with our local Multicenter Perioperative Outcomes Group database to identify LVAD patients undergoing NCS procedures requiring an anesthesiologist between January 1, 2006 and August 13, 2015. NCS was determined on the basis of Anesthesiology Current Procedural Terminology (CPT) code; additionally, CPT code data were used to define NCS as minor (2011 American Society of Anesthesiologists [ASA] base unit values less than or equal to 5) or major (base unit values more than 5) procedures. All data were extracted from the electronic health record. Data quality assurance was maintained by a hand review of all cases included in the study, including all perioperative complications, intraoperative vital signs, and preoperative patient anesthesia history and physical variables (table 1; complete details are provided in Supplemental Digital Content 1, http://links.lww.com/ALN/B356).
For descriptive purposes, variables studied included an array of surgical, anesthetic, and patient characteristics (table 1). As has been investigated in previous similar studies, the involvement of a cardiac anesthesia provider, defined by faculty appointment and practice pattern within the cardiac anesthesia division of our department, was included as a study variable.13,38 To account for evolving trends in LVAD management over the study period, the year of surgery, type of LVAD, and duration since LVAD placement were also included as study variables. Perioperative complications studied included those known to the ventricular assist device population (table 2).6,9–13
Based upon previous studies assessing perioperative LVAD complications, expected complication frequencies, and clinical relevance, we chose acute kidney injury (AKI) as the primary outcome for this study. AKI was defined by established clinical practice guidelines as a postoperative increase in serum creatinine greater than or equal to 1.5 times preoperative baseline within 7 days as the primary outcome.6,9–13,43 Secondary outcomes included elevated serum lactate dehydrogenase (LDH) concerning for possible LVAD-associated hemolysis (defined by the current literature as a value greater than 600 IU/l, i.e., 2.5 times the upper limit of laboratory normal),44–46 intraoperative bleeding complication (defined as an estimated blood loss greater than 500 ml or blood product transfusion intraoperatively), and clinically significant intraoperative hypotension. Given the 7-day postoperative measurement window defining AKI and elevated LDH, patients incurring multiple NCS procedures were restricted to one (index) case evaluated per 7-day period to prevent the duplication of a single outcome. To analyze clinically significant intraoperative hypotension as a secondary outcome, we selected mean arterial pressure (MAP) thresholds and durations established in the current LVAD,47,48 NCS,49 and AKI50 literature. For purposes of multivariable analyses, we planned a priori to select the MAP threshold and duration most associated with our primary outcome, AKI. This selected hypotension definition was studied as a covariate for primary and secondary outcomes, as well as an outcome itself.
Basic descriptive statistics, including percentages, means, SDs, medians, and interquartile ranges (IQRs), were calculated for all factors listed in table 1; complete details are available in Supplemental Digital Content 1 (http://links.lww.com/ALN/B356). Factors were univariately compared using Fisher exact or Pearson chi-square tests for categorical variables and Student’s t tests for continuous variables. To calculate adjusted odds ratios (aORs) via multivariable analysis, conditional binary logistic regressions were performed with patient as the strata and adjusting for factors with univariate P < 0.20, followed by stepwise variable selection by Akaike information criterion. A 95% CI excluding 1 or P < 0.05 was deemed statistically significant. Model discrimination was determined by Harrell c-statistic. Internal model validation was evaluated using 1,000 bootstrapped resamples. The resamples were chosen by patients with replacement. For each resample, a full model was fit, and then, a reduced model was selected by stepwise variable selection by Akaike information criterion. All analyses were done using R, version 3.2.2 (R Foundation for Statistical Computing, Austria).
An a priori power analysis was performed for the study to determine the adequacy of sample size. Based upon an expected AKI incidence of 10%, a multivariable analysis including up to 10 variables would require approximately 100 primary outcomes to be observed, and thus, 1,000 cases for analysis.
We identified 246 patients with LVADs presenting for 702 unique procedures. As stratified by anesthesia CPT code, 270 (38%) were major surgical procedures and 432 (62%) were minor (complete details are provided in Supplemental Digital Content 2, http://links.lww.com/ALN/B357). Upper or lower gastrointestinal endoscopies (n = 263; 37%) and implantable cardioverter-defibrillator procedures (n = 248, 35%) were the most common procedures.
Most patients were males (n = 197; 80%) and had continuous flow devices (n = 227; 92%) including the HeartMate II (n = 170; 69%; St. Jude, USA) and HVAD® (n = 50; 20%; HeartWare, USA). Five hundred and thirty-nine of the NCS (77%) were performed on an inpatient basis; 47 were performed emergently (ASA E status). The median duration from LVAD placement to NCS date was 370 days (IQR, 87–827) and preoperative MAP averaged 82 mmHg (SD, ± 12 mmHg; table 1).
Given a transition to HeartMate II and HVAD® devices as routinely implanted LVADs after 2010 at our institution, we describe preoperative management before and after this transition (early and late study periods). Between the early and late study periods, cardiac anesthesiologist involvement decreased from 57 to 33% (P < 0.001). The use of general anesthesia similarly decreased from 31 to 23% (P = 0.03). Across study cases, 316 (45%) received preoperative hospital-administered anticoagulation of any type (unfractionated heparin , warfarin , enoxaparin , argatroban , clopidogrel , prasugrel , fondaparinux , eptifibatide , or bivalirudin ). Anticoagulation was rarely reversed preoperatively: fresh frozen plasma (FFP) was given in only 26 (3.7%) cases, vitamin K in 15 (2.1%) cases, platelets in 3 (0.4%) cases, and prothrombin complex concentrate in 2 (0.3%) cases. Cryoprecipitate, recombinant coagulation factor VIIa, and recombinant fibrinogen were not administered preoperatively for any case. Arterial line invasive blood pressure monitoring was used in 141 (20%) cases. Arterial line monitoring was less commonly used for minor procedures (n = 50; 12% of minor procedures; P < 0.001) and cardiology procedures (n = 39; 14% of cardiology procedures; P < 0.001) compared to noncardiology major procedures (n = 55; 72% of noncardiology major procedures; P < 0.001). Arterial line monitoring was also more common for general anesthetics (n = 114; 64%) compared to cases without general anesthesia (n = 27; 5.1%; P < 0.001). Between the early and late study periods, arterial line blood pressure monitoring decreased from an overall rate of 31 to 16% (P < 0.001; table 3); among minor procedures, arterial line monitoring decreased from an overall rate of 30 to 7.6% (P < 0.001) from early to late study periods.
In more than half of NCS cases (n = 386; 55%), there was a more than 20-min intraoperative gap without recorded blood pressures. Furthermore, among the 141 cases with arterial line monitoring, 45 (32%) demonstrated a monitoring gap of more than 20 min, most commonly after anesthetic induction before placement of the arterial line. Among 561 cases without arterial line monitoring, 31 (5.5%) were devoid of any recorded intraoperative blood pressure (table 4). In such cases, measures approximating vital organ perfusion were documented, including patient responsiveness (e.g., patient following commands, patient alert, etc.) in 11 cases and/or serial documentation of stable LVAD parameters (i.e., flow, power, and pulsatility index) in 29 cases. In one remaining case, documentation-describing attempts at blood pressure monitoring were made; in all cases lacking blood pressure recording, anesthesiologists subsequently documented a postoperative anesthesia evaluation including an assessment of mental function before recovery unit discharge.
The use of vasoactive medications beyond phenylephrine was rare: only four cases received epinephrine more than 10 μg and 15 cases received vasopressin more than 1 unit. In three cases, 150 mg amiodarone was administered: one administration occurred in the setting of ventricular tachycardia, another in the setting of supraventricular tachycardia, and a third prophylactically. Malignant dysrhythmias were rare: ventricular tachycardia occurred in seven cases; ventricular fibrillation and asystole once each. All were successfully managed pharmacologically or with electrical defibrillation; chest compressions were used in one patient.
Among primary and secondary outcomes studied, AKI and intraoperative hypotension regularly occurred; elevated LDH and bleeding complications were less common (tables 1, 2; Supplemental Digital Content 1, http://links.lww.com/ALN/B356). Among the 607 index cases, 110 (18%) developed AKI. Patients who developed AKI were more likely to have preoperative organ dysfunction, comorbidities, and lower blood pressures (table 1). After adjusting for other factors, three processes of care demonstrated statistically significant independent associations with AKI: major procedures (aOR, 4.4; 95% CI, 1.1 to 17.3; P = 0.03), invasive arterial blood pressure monitoring (aOR, 3.6; 95% CI, 1.3 to 10.3; P = 0.02), and FFP transfusion preoperatively (aOR, 1.7; 95% CI, 1.1 to 2.8; P = 0.02; fig. 1). The model had fair predictive discrimination (Harrell c-statistic = 0.645). Of the 1,000 bootstrapped models, 875 of the full and reduced models converged. All of the variables in the reduced model remained in greater than 45% of the bootstrapped reduced models, with major procedures remaining in 66%, invasive arterial blood pressure monitoring remaining in 70%, and FFP transfusion preoperatively remaining in 59% of the models. All of the CIs for the original model estimates overlapped with the 95% percentile bootstrapped CIs.
Elevated LDH occurred in 16 of 607 index cases (2.6%) and bleeding complications in 45 of 702 cases (6.4%). Intraoperative hypotension defined by an MAP less than 70 mmHg for more than 20 min demonstrated the strongest univariate association with AKI (P = 0.02); using this definition, clinically significant intraoperative hypotension occurred in 176 of 664 cases (27%). While there were several risk factors associated with elevated LDH, bleeding complication, or intraoperative hypotension (including surgery characteristics, emergency cases, longer cases, cases with general anesthesia, and cases with preoperative elevated LDH [Supplemental Digital Content 1, http://links.lww.com/ALN/B356]), multivariable analyses failed to demonstrate any significant independent associations. Among cases with preoperative and postoperative LDH values available, absolute and relative changes in LDH levels were varied (median, −12 IU/l; IQR, −49 to 20 IU/l and median, −9%; IQR, −26 to 15%, for absolute and relative changes, respectively). Furthermore, 13 of 16 cases with postoperative elevated LDH had LDH elevated (more than 600 IU/l) preoperatively. A post hoc sensitivity analysis defining elevated LDH as more than 480 IU/l demonstrated similar results through univariate and multivariate analyses (available as Supplemental Digital Content 3, http://links.lww.com/ALN/B358).
We compared cases for which the majority of care was performed by a cardiac versus a noncardiac anesthesiologist (described in Supplemental Digital Content 4, http://links.lww.com/ALN/B359). Through univariate analyses performed, cardiac anesthesiologists were associated with a population of patients more likely to have HeartMate II devices, more frequently undergoing cardiology procedures, and less frequently having had a recent surgery within 30 days. Cardiac anesthesiologists were more frequently associated with elevated troponin levels but less frequently associated with bleeding complications. Of note, cardiology procedures were primarily managed by cardiac anesthesiologists (n = 212; 75% of cardiology procedures (284); P < 0.001), whereas gastroenterology procedures were overwhelmingly managed by noncardiac anesthesiologists (n = 255; 97% of gastroenterology procedures ; P < 0.001); among gastroenterology procedures, seven patients received intraoperative packed erythrocyte transfusions. Among major surgeries, unadjusted 30-day mortality rates were 7.1% among noncardiac anesthesiologists versus 2.2% among cardiac anesthesiologists; however, this did not meet statistical significance (P = 0.07). For all study outcomes including AKI, elevated LDH, bleeding complication, intraoperative hypotension, and mortality, no independent associations between cardiac anesthesiologist versus noncardiac anesthesiologist were demonstrated via multivariable analysis.
Intraoperative bleeding complications occurred in 45 cases (6.4%). Among these cases, 30 (4.2%) required packed erythrocyte transfusion, 21 (3.0%) FFP transfusion, 14 (2.0%) platelet transfusion, 2 (0.3%) desmopressin administration, and 1 (0.1%) recombinant coagulation factor VIIa administration. In no cases were cryoprecipitate, prothrombin complex concentrate, or recombinant fibrinogen administered. Fifteen cases were minor NCS, seven of which were gastrointestinal endoscopies for bleeding. The 30 major NCS cases with bleeding complications were most commonly general surgical procedures (n = 8), neurosurgical procedures (n = 8), and vascular procedures (n = 4). In the postoperative period, blood product administration occurred in 92 cases; 56 were gastrointestinal endoscopies. Among 29 major NCS receiving postoperative transfusions, general surgical (n = 11) and cardiology (n = 7) procedures were the most common.
We also found several other less frequent adverse events (table 2). LVAD thrombosis, defined by administration of hemolytics and/or requiring LVAD exchange before hospital discharge, was observed in four cases (0.6%). All patients received tissue plasminogen activator, and none required LVAD exchange. LVAD device malfunction occurred in one case, in which an LVAD device controller engineering defect required controller replacement. One patient also required device exchange for major driveline infection not resolved with medical therapy. While there were no intraoperative deaths, 13 patients (5.3%) died within 30 days postoperatively. Among these patients, most (n = 10) had been operated on as emergency salvage procedures.
Enabled by improvements in device technology and LVAD-specific care infrastructure, the LVAD patient presenting for NCS is increasingly common. Descriptive analyses of LVAD patients safely managed for NCS are well documented. Our study builds upon the previous literature by providing greater perioperative detail and performing risk factor analyses for which previous studies have been underpowered. We report 246 LVAD patients presenting for 702 NCS procedures, with complications including perioperative mortality (5.3%), AKI (18%), elevated LDH (2.6%), intraoperative bleeding (6.4%), and hypotension (MAP less than 70 mmHg more than 20 min; 27%).
We found that perioperative characteristics independently associated with AKI included major procedures, invasive arterial blood pressure monitoring, and preoperative FFP transfusion; no clinically significant independent associations with secondary outcomes were identified. The independent association between arterial line monitoring and AKI underscores arterial line use as a marker of case complexity not measured by other covariates and reinforces the importance of anesthesiologist vigilance for hemodynamic instability during cases in which arterial line monitoring is deemed necessary. FFP transfusion, also independently associated with AKI, may serve as a similar marker, reflecting cases requiring rapid anticoagulation reversal or patients with ongoing hemorrhage and decreased renal perfusion. The acuity of such cases may explain increased AKI risk; however, it is also possible that a causal relationship between FFP transfusion and AKI may exist. This association has been demonstrated in patients undergoing cardiothoracic surgery.51,52
Among hypotension measures studied, an MAP less than 70 mmHg more than 20 min demonstrated the strongest univariate association with AKI (P < 0.001). This finding is in contrast to previous studies of the general surgical population involving primarily non-LVAD patients, in which MAP values less than 55 to 60 mmHg have been established as thresholds for postoperative complications.49,50,53 Given altered hemodynamics imparted by an LVAD as evidenced by changes in organ perfusion after implantation,54,55 it is reasonable that intraoperative hemodynamic goals, including target MAP range, may differ from the general surgical population as well. Furthermore, the etiology of hypotension in LVAD patients commonly differs from the general surgical population (e.g., thrombus, suction events, right ventricular failure, and limited compensatory response to vasodilation), and thus, treatment may differ as well. Our study represents an initial step toward defining the impact of intraoperative hypotension on outcomes in these patients; further studies are needed to investigate hypotension treatment in the LVAD population. Our study results support the current recommended minimum MAP target of 70 mmHg; current recommended maximum MAP values range from 80 to 90 mmHg.47,48
In agreement with Stone et al.,13 we found that noncardiac anesthesiologists provided an increased proportion of care over time. This can be attributed to a developing familiarity with LVADs, as well as an increasing LVAD patient caseload, with a shift toward minor procedures, requiring a wider range of anesthesiology staff resources. Although limited by provider-specific case clustering, our study offers evidence justifying this trend, demonstrating no independent association with perioperative complications among noncardiac anesthesiologists when adjusted by the perioperative characteristics in Supplemental Digital Content 4 (http://links.lww.com/ALN/B359). Among major surgeries, it is possible that our study failed to demonstrate a univariate association between cardiac anesthesiologists and decreased perioperative mortality (P = 0.07) due to limited sample size. After adjustment for case complexity via multivariable analysis, however, we continued to find no independent association. Our provider-specific findings should be interpreted with caution, as the extent of cardiac anesthesiologist involvement continues to vary widely among institutions, due to a lack of consensus on LVAD patient management, as well as intra- and interinstitutional variation in anesthesia provider skills and experience.13,14,37
Similar to the previous literature, we found a decrease in invasive arterial blood pressure monitoring over the study period.13 Much of this has been attributed to the increase in minor NCS procedures, most notably those in the endoscopy suite.37,38,56 However, our study offers caution in this regard. As we observed intraoperative blood pressure monitoring gaps more than 20 min in a majority of cases despite anesthesiologist attempts to obtain blood pressure measurements or document vital organ perfusion, opportunities for unrecognized hypotension were common. Our study suggests a need for improved blood pressure monitoring for detection and treatment of hypotension in LVAD patients and sheds light on the previous conflicting literature either supporting9,37 or refuting14,56 this concept.
Our results differ from studies noting no difficulty with noninvasive blood pressure monitoring.14,56 This may relate to differences in monitor type, patient and surgical characteristics, and anesthetic management. In particular, the use of a Doppler probe with a manual sphygomomanometer, as well as a slow-deflation cuff (e.g., Elemano, Terumo Medical Corp, USA), has been shown to increase success rates.57–59 Within the continuous-flow LVAD population, noninvasive monitoring success rates improved from 50 to 63% for automated measurements to 91 to 100% when utilizing such methods.57–59 Although shortcomings have been described, including limitations in accuracy,57 as well as lack of widespread availability, these alternative noninvasive monitoring devices allow for a blood pressure to be measured in patients for which automated noninvasive monitoring would otherwise be unsuccessful. Of note, our institution in 2012 implemented the elective use of a Doppler probe with a sphygmomanometer among patients with continuous-flow LVADs; however, documentation of this method was frequently not present. In our study, cases lacking a recorded blood pressure documented other measures approximating organ perfusion, including patient responsiveness and LVAD flow parameters. In light of these findings, our study illustrates how the LVAD patient can create difficulties in observing ASA basic monitoring standards, including systemic blood pressure monitoring every 5 min, and suggests a need for more widespread use of alternative blood pressure monitoring techniques in order to meet these standards.60
Anticoagulation management during the perioperative period remains controversial for LVAD patients. Although manufacturer guidelines and institutional standards exist for target international normalized ratio (INR) levels postoperatively, no clear guidelines exist for the preoperative and intraoperative periods. Postoperatively, HeartMate II manufacturer guidelines recommend an INR range of 2.0 to 3.0.61 Among studies investigating patients transitioned to vitamin K antagonists postoperatively, goal INR values vary from 1.5 to 3.5.62 Whereas previously a lower INR range (1.5 to 2.5) was recommended in LVAD patients,63 recent reports of hemolysis and thrombotic complications64 have prompted some centers to adopt higher ranges (2.0 to 3.0).44,62,65 In the preoperative context among patients in our study, we observed a mean INR of 1.7. In some cases, INR values were below target ranges and unaccompanied by hospital-administered perioperative bridging anticoagulation within 24 h of surgery. These findings shed light on the need for consensus guidelines for preoperative anticoagulation management in the LVAD population.66,67 Of additional consideration is the emerging role of point-of-care viscoelastic coagulation testing. Although studies within the LVAD population are limited, such testing has shown associations with thromboembolic complications after NCS in the non-LVAD population.68 In the absence of prospective studies, anticoagulation management for LVAD patients remains challenging for the perioperative clinician; current management most likely should be determined on a case-by-case basis.
Management of LVAD-induced hemolysis also remains challenging for the perioperative clinician; we found postoperative LDH levels more than 600 IU/l in 2.6% of cases. Previous studies have cited LDH thresholds ranging from 250 to 1,000 IU/l to define LVAD-associated hemolysis, among other accompanying markers, including elevated plasma-free hemoglobin and bilirubin, decreased haptoglobin, anemia, hemoglobinuria, and altered pump parameters.65,69,70 We chose an LDH more than 600 IU/l based upon the recent literature and expert guidelines.44,46 In addition to increased risk of device thrombosis, elevated LDH is also a marker of increased cell-free heme, which has been causally implicated in AKI.71,72 While we found a univariate association among elevated LDH and AKI, this did not persist after multivariable adjustment. Although in only four instances did patients receive postoperative thrombolytics, our study is limited by the short-term follow-up. Some patients with suspected thrombosis may not have been treated due to increased risk of hemorrhage.
Our study has several limitations. As characteristic to retrospective reviews, no specific interventions were mandated although standard clinical practices were employed. Additionally, data were limited to that which was recorded for clinical care purposes. Our study was performed at a single tertiary care center, and thus, variation related to our practice patterns and patient population may limit generalizability. In our characterization of cases performed primarily by noncardiac anesthesiologists, it is possible that a cardiac anesthesiologist participated in patient care, either for a minority of the case duration or through informal case discussion. As the 702 cases studied were comprised of 246 patients, patients undergoing multiple procedures may have skewed our results although our AKI and elevated LDH outcomes were restricted to index cases to minimize this impact. Finally, as most complications were rare, it is possible that multivariable analyses were underpowered to detect risk factors with small effect size.
Despite limitations, our review of LVAD patients presenting for NCS represents the largest, most in-depth study to date. The increased prevalence of such patients demands a knowledgebase and familiarity with LVADs among cardiac and noncardiac anesthesiologists alike. Through reporting trends in clinical decision-making, complications, and risk factors, our study offers unparalleled insight into perioperative management of LVAD patients. In conclusion, we report AKI, hypotension, elevated LDH, and bleeding to comprise the leading major complications within our LVAD patient population and a 30-day postoperative mortality rate of 5.3%.
The authors acknowledge Anik Sinha, M.S. (Department of Anesthesiology, University of Michigan Health System, Ann Arbor, Michigan) for his contributions in data acquisition and electronic search query programming for this project.
All work and partial funding attributed to the Department of Anesthesiology, University of Michigan Medical School (Ann Arbor, Michigan). The University of Michigan performs contract research with St. Jude (Pleasanton, California) and HeartWare (Framingham, Massachusetts). The project described was supported in part by the National Center for Research Resources, grant no. UL1RR024986, which is now at the National Center for Advancing Translational Sciences (Bethesda, Maryland), grant no. 2UL1TR000433. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The authors declare no competing interests.