The hazards of severe hyperkalemia in patients undergoing orthotopic liver transplantation (OLT) are well recognized and warrant vigilant monitoring and acute management (1–3). A number of interventions can be used to decrease serum potassium (K+) (4–8). However, discovering a method to identify patients who are at high risk of developing intraoperative hyperkalemia would be appealing. Such knowledge would allow early planning and effective use of the interventions on targeted patients, rather than universal, costly, and potentially harmful interventions.
Previous studies of intraoperative hyperkalemia during OLT included only a small number of patients (9,10), and some included both pediatric and adult patients who may have different patterns of intraoperative K+ (8,11). A few studies used multivariate analysis, which has distinctive advantages over univariate analysis (10,12). In addition, previous studies focused on hyperkalemia immediately after reperfusion and left other periods largely unexamined (12). As physiological and hemodynamic changes vary significantly in different periods of the procedure, identification of risk factors for hyperkalemia at one period may not apply to others. Although hyperkalemic episodes immediately after postreperfusion are most frequent and significant, hyperkalemia in other periods during OLT can be equally serious (2,13).
We undertook this retrospective observational study to identify predictors associated with hyperkalemia in the prereperfusion, early postreperfusion, and late postreperfusion periods in adult patients undergoing OLT. We hypothesized that predictors of hyperkalemia may vary among the different periods, and that some donor characteristics may be associated with hyperkalemia in the postreperfusion periods during adult OLT.
The IRB of the University of California, Los Angeles (UCLA) approved the study. The study population consisted of all adult (age, ≥18 yr) patients who underwent OLT at the UCLA Medical Center between January 1, 1998, and December 31, 2004. Data were collected from the UCLA transplant database and electronic medical records.
Anesthesia was typically induced with IV fentanyl, propofol or etomidate, and succinylcholine, followed by endotracheal intubation. If patients were intubated before OLT, combined inhaled and IV anesthetics were used for induction. Anesthesia was maintained with isoflurane, fentanyl, and neuromuscular blocking drugs. Venovenous bypass was used selectively during OLT. Selection criteria for venovenous bypass included history of cardiac disease, poorly controlled systemic hypertension, hypotension, pretransplant vasopressor requirement, persistent hypotension after a trial of portal vein or vena cava cross-clamping or the opinion of anesthesiologist or surgeon that the patient would benefit from venovenous bypass. Packed red blood cells (RBCs) were administered to maintain the hematocrit at 30%. RBCs were not routinely washed or treated to reduce K+ levels unless requested either by surgeon or anesthesiologist. Cell salvage was not used. The graft liver was routinely flushed with cold lactated Ringer's solution with albumin. Blood (300–500 mL) was selectively vented via the portal vein before graft reperfusion. Serum K+ concentrations were monitored at approximately hourly intervals except immediately before and after reperfusion, when more frequent assessments were made. The anesthesiologist selected the interventions to prevent or treat intraoperative hyperkalemia.
Three periods (prereperfusion, early postreperfusion, and late postreperfusion) were studied. The prereperfusion period was defined as the 2 h immediately before reperfusion, including the preanhepatic and anhepatic stages. The early postreperfusion period was the first 15 min after reperfusion, and the late postreperfusion period started 1 h after reperfusion until the end of the case. The 45-min period after reperfusion was not examined because it is a period of instability, when reperfusion metabolic changes are clearing. Hyperkalemia was defined as serum K+ concentration ≥5.5 mmol/L as used in previous studies (12). Patients were divided into two groups, hyperkalemic and nonhyperkalemic, according to their K+ levels.
Recipient variables selected for analysis included age, gender, weight, Model for End-Stage Liver Disease (MELD) score, OLT year (pre- vesus post-MELD era), United Network for Organ Sharing status, etiology of liver diseases, prior upper abdominal surgery, prior nonshunt surgery, prior shunt surgery, presence of ascites at surgery, recent prior OLT (within 6 mo), remote prior OLT (>6 mo), and combined organ transplant. Donor variables included donor age, gender, peak serum sodium, number of vasopressor(s) used, donor hospital stay, donation after cardiac death (DCD), living-related donor, split graft, and graft cold and warm ischemia times. Laboratory variables were baseline serum concentration of K+, creatinine, hematocrit, platelets, fibrinogen, international normalized ratio of prothrombin time, base excess, and prereperfusion K+ concentrations. Baseline laboratory values were defined as the first intraoperative laboratory values or immediately before surgery if the first intraoperative value was during the prereperfusion period. Prereperfusion K+ was defined as the highest K+ in the prereperfusion period. Intraoperative variables included the use of venovenous bypass, the piggyback technique, units of transfused RBCs and fresh frozen plasma, intraoperative urine output, and surgical time.
Data were expressed as mean ± sd and median values for continuous variables, or as proportions for nonparametric variables. Before bivariate screening, each continuous variable was dichotomized at its median, at a meaningful value indicated by a gap in its histogram or was grouped by quartile. First, we did univariate χ2 or Fisher's exact tests to identify variables that were significantly different between patients with and without hyperkalemia in each period. Then, we included variables that showed a potential significance (P < 0.10) during univariate analyses in a multivariate forward and backward-stepwise logistic regression modeling process. The final logistic regression model was given for prereperfusion, early postreperfusion, and late postreperfusion hyperkalemia individually. Selected continuous variables in the final logistic regression models were grouped by quantile to assess a trend of increasing odds with increasing values of variables. The odds ratio and 95% confidence interval with associated P values for each variable were reported in the logistic models. P < 0.05 was considered statistically significant in the multivariate models. To provide a less biased estimate of the discrimination accuracy (receiver operating characteristic, ROC), the dataset was randomly divided into five equal subsets, and ROC was re-estimated after five-fold validation, using four-fifths of the data as the training set and the remaining one-fifth as the validation set. This was repeated five times, so that all subsets were used as a validation set once and the average of ROC values of validation sets is reported. Statistical analyses were performed using the Statistical Package for the Social Sciences 14.0 for Windows (SPSS, Inc., Chicago, IL) and SAS 9.0 (SAS Inc., Cary, NC).
Data from 1124 adult patients who underwent OLT from 1998 to 2004 were analyzed. Table 1 shows the distribution of selected recipient, donor, laboratory, and intraoperative variables. The majority of recipients (62.7%) and donors (63.4%) were men. The mean age was 52 ± 10 (mean ± sd) years for recipients and 41 ± 17.0 for donors. Chronic cirrhosis caused by hepatitis B or C was the predominant indication for OLT.
Hyperkalemia occurred in 10.2% of patients. The highest recorded K+ in the prereperfusion period was 7.6 mmol/L. As expected, the largest percentage of hyperkalemic patients (19.1%) and the highest K+ (8.9 mmol/L) was found in the early postreperfusion period. Although less frequent compared with those in other periods, a significant number of patients (8.3%) were hyperkalemic in the late postreperfusion period, and the highest K+ was 7.6 mmol/L in this period.
In univariate analysis, we found that higher baseline K+ was consistently associated with hyperkalemia in all three periods. Other significant variables were found in the various periods (Table 2). Specifically, in the prereperfusion period, patients who had higher baseline serum K+, higher number of transfused RBCs, higher baseline serum creatinine, lower baseline base excess, remote prior OLT, combined organ transplant, post-MELD era OLT, and no prior shunt surgery had a significantly higher incidence of hyperkalemia. In the early postreperfusion period, a significantly positive association was noted between hyperkalemia and higher baseline (or prereperfusion) K+, nonliving-related donor, post-MELD era OLT, and DCD donor. In the late reperfusion period, we also found that higher baseline (or prereperfusion) K+, higher baseline creatinine, use of venovenous bypass, longer warm ischemia time, older donor age, living-related donor, lower intraoperative urine output, and longer donor hospital stay were associated with significantly increased risk of hyperkalemia.
Results of forward and backward stepwise multivariate logistic regression analysis for prereperfusion hyperkalemia are presented in Table 3. Baseline K+ and RBC transfusion were the two significant predictors of hyperkalemia in the prereperfusion period. As baseline K+ concentrations and the volume of transfused RBCs increased, the odds of prereperfusion hyperkalemia were greater. Compared with patients with baseline K+ <3.5 mmol/L, patients with higher baseline K+ had 3–36 fold increased odds of developing hyperkalemia. Similarly, compared with patients who received ≤5 U of RBCs, the odds of having hyperkalemia in patients with more RBC transfusions were 5–12-fold higher.
In multivariate logistic regression analysis, higher baseline (or prereperfusion) K+ and DCD donor were the only two predictors (Table 4, baseline or prereperfusion K+ was used in Model 1 or 2, respectively) of early postreperfusion hyperkalemia. The higher baseline (or prereperfusion) K+ resulted in higher odds of developing early postreperfusion hyperkalemia. Patients who received DCD had three-fold greater odds of developing early postreperfusion hyperkalemia, compared with recipients of brain-dead donors. Both models using either baseline or prereperfusion K+ concentrations had similar accuracy, indicated by the areas under ROC (0.675 vs 0.713 for baseline versus prereperfusion K+).
In multivariate logistic regression analysis, independent predictors of late reperfusion hyperkalemia included baseline K+, two intraoperative variables (intraoperative urine output and the use of venovenous bypass), and two donor variables (warm ischemia time and donor hospital stay, Table 5). Similar to its effects in other periods, higher baseline K+ resulted in higher odds of late postreperfusion hyperkalemia. Among donor variables, longer warm ischemia time and donor hospital stay posed significant risks of hyperkalemia in the late postreperfusion period. A model including baseline K+, 2 intraoperative predictors, and 2 donor predictors yielded the area under ROC of 0.820, indicating a highly predictable event. Replacing baseline K+ with prereperfusion K+ did not increase the accuracy of prediction (ROC from 0.820 to 0.794).
In this study, we identified several predictors of hyperkalemia in each period of OLT. Higher baseline K+ was the only consistent and most significant predictor of hyperkalemia in all three periods of OLT. Although many factors, including blood pH, osmolarity, insulin, and catecholamine, affect serum K+ concentrations (14–16), baseline K+ provides vital information for hyperkalemia prediction during OLT. These findings suggest that measurement of baseline K+ concentrations at the beginning of surgery is useful for predicting hyperkalemia during OLT. Although not surprising, this conclusion has not been highlighted in other studies (5).
Hyperkalemia is more common in the early postreperfusion period than at other times. The nature of rapidly increasing and decreasing serum K+ and many contributing factors associated with reperfusion of the liver graft make this event unpredictable. In agreement with other studies (12,17), we confirmed that a higher prereperfusion K+ level was significantly associated with hyperkalemia in the early postreperfusion period. We also demonstrated that by using baseline K+ a similar prediction in a logistic regression model could be achieved. The obvious advantages of using baseline over prereperfusion K+ is the time afforded to treat anticipated hyperkalemia.
In addition to baseline serum K+, other predictors of hyperkalemia vary in different periods. Three intraoperative variables, one for prereperfusion and two for postreperfusion hyperkalemia, were identified. RBC transfusion is a well-known risk factor for intraoperative hyperkalemia, and such an association has been demonstrated in a variety of clinical settings (1,18). The finding that RBC transfusion was associated with hyperkalemia only in the prereperfusion period may suggest that the grafted liver plays a significant role in serum K+ concentrations after reperfusion. Patients with intraoperative urine output <500 mL had a three-fold increased risk of hyperkalemia in the late postreperfusion period compared with patients with higher intraoperative urine output. It is reasonable to speculate that decreased urine output causes decreasing urinary K+ excretion, which then leads to serum K+ retention and serum K+ increase. The association between the use of venovenous bypass and hyperkalemia in the late postreperfusion period was unexpected. Several factors might be involved. These include hemolysis caused by the venovenous bypass machine, which can be demonstrated in similar machines used in cardiopulmonary bypass (19,20), and intracellular to extracellular K+ shift during the venovenous bypass warming process (21). Finally, venovenous bypass was used in selected, often sicker, patients, raising the possibility of a selection bias.
A persistent organ shortage has led to novel strategies to expand the donor pool of livers for transplantation (22). Although some characteristics of extended criteria donors have been correlated with inferior short- and long-term outcomes, and animal studies have shown that poorly preserved liver grafts are associated with increased concentration of serum K+ after reperfusion (23), no human studies have concluded that hyperkalemia is increased in recipients of extended criteria donors (12).
The findings of the association between intraoperative hyperkalemia and donor characteristics have significant clinical implications that are of particular interest to anesthesiologists. With increasing use of DCD donors and other extended criteria donors, more intraoperative complications, including hyperkalemia, may be anticipated. Furthermore, intraoperative management may need to be modified to meet challenges presented by extended criteria donors in the MELD era (24,25). As a result of this study, we changed our practice in two areas: 1) Since higher baseline K+ and RBC transfusion are two predictors of prereperfusion hyperkalemia, insulin is administered IV as soon as RBC transfusion begins (1–2 international units of regular insulin for each unit of RBC) in patients with baseline K+ ≥4.0 mmol/L; 2) If patients receive extended criteria donations, especially DCD or donations with multiple extended criteria, we will initiate early and aggressive preemptive therapies, including insulin, alkalizing drugs, hyperventilation, furosemide, and calcium chloride, to decrease the risk of postreperfusion hyperkalemia.
We acknowledge several limitations of this study. First, this is a retrospective study, and diagnosis and management of hyperkalemia were not standardized. Second, the data are derived from only one center, and practices at other centers may result in different patterns of hyperkalemia during OLT. Third, although the association between hyperkalemia and its predictors can be demonstrated in multivariate analyses, a cause-and-effect relationship cannot be assumed.
In summary, our study identified several predictors of hyperkalemia in different periods of adult OLT. These findings should allow anesthesiologists to implement preemptive therapies on selected patients at risk of developing hyperkalemia during OLT.
We thank Dr. Jeff Gornbein, the UCLA Department of Biomathematics, for statistical support and helpful discussion.
1. Merritt WT. Metabolism and liver transplantation: review of perioperative issues. Liver Transpl 2000;6:S76–84
2. Huang YR, Gu MN, Deng RJ, Xiao JF, Liang SW, Xu JS, Chen YM. [Management of hyperkalemia-induced cardiac arrest in anhepatic stage of orthotopic liver transplantation without venovenous bypass: report of 3 cases.] Di Yi Jun Yi Da Xue Xue Bao 2004;24:1210–14
3. Shi XY, Xu ZD, Xu HT, Jiang JJ, Liu G. Cardiac arrest after graft reperfusion during liver transplantation. Hepatobiliary Pancreat Dis Int 2006;5:185–9
4. Acosta F, Diaz J, Sansano T, Bento M, Cura S, Contreas RF, Belmonte JG, Parrilla P, Bueno FS, Robles R. Prophylactic treatment of metabolic alterations during revascularization in liver transplantation. Transplant Proc 1994;26:3667–8
5. De Wolf A, Frenette L, Kang Y, Tang C. Insulin decreases the serum potassium concentration during the anhepatic stage of liver transplantation. Anesthesiology 1993;78:677–82
6. Ellis R, Beeston JT, Witherington SS, Witheington SS, Allen EL, Keenan RL. Liver transplantation: effect of washing bank blood on intraoperative control of hyperkalemia. Transplant Proc 1987;19:73–4
7. Ziemann-Gimmel P, Pygon B, Hurley F, Albrechet RF, Schwartz DE. Treatment of life-threatening hyperkalemia using hemoconcentration in parallel to venovenous bypass during orthotopic liver transplantation. Anesth Analg 2003;96:680–2
8. Belani KG, Reardon RF, Payne WD, Williams GW, Palahniuk RJ. Acute hypokalemia is well tolerated in infants during liver transplantation. Transplant Proc 1994;26:147–50
9. Carmichael FJ, Lindop MJ, Farman JV. Anesthesia for hepatic transplantation: cardiovascular and metabolic alterations and their management. Anesth Analg 1985;64:108–16
10. Chapin JW, Wood RP, Hurlbert BJ, Shaw BW, Kennedy EM, Cuke DJ, Markin RS, Peter KR, Newland MC. Sources of increased serum potassium following reperfusion of liver allografts. Transplant Proc 1987;19:51–3
11. Xia VW, Du B, Tran A, Liu L, Hu KQ, Hiatt JR, Busuttil RW, Steadman RH. Intraoperative hypokalemia in pediatric liver transplantation: incidence and risk factors. Anesth Analg 2006;103:587–93
12. Nakasuji M, Bookallil MJ. Pathophysiological mechanisms of postrevascularization hyperkalemia in orthotopic liver transplantation. Anesth Analg 2000;91:1351–5
13. Kaku R, Matsumi M, Fujii H, Ohashi I, Mizobuchi S, Katayama H, Morita K, Hirakawa M. [A case of severe acute hyperkalemia during pre-anhepatic stage in living-related liver transplantation]. Masui 2002;51:1003–6
14. Halperin ML, Kamel KS. Potassium. Lancet 1998;352:135–40
15. Evans KJ, Greenberg A. Hyperkalemia: a review. J Intensive Care Med 2005;20:272–90
16. Mahoney BA, Smith WA, Lo DS, Tsoi K, Tonelli M, Clase CM. Emergency interventions for hyperkalaemia. Cochrane Database Syst Rev 2005;CD003235
17. Acosta F, Sansano T, Contreras RF, Reche M, Beltran R, Roques V, Rodriguez MA, Robles R, Bueno FG, Ramirez P, Parrilla P. Changes in serum potassium during reperfusion in liver transplantation. Transplant Proc 1999;31:2382–3
18. Brown KA, Bissonnette B, MacDonald M, Poon AO. Hyperkalaemia during massive blood transfusion in paediatric craniofacial surgery. Can J Anaesth 1990;37:401–8
19. Paparella D, Galeone A, Venneri MT, Coviello M, Visicchio G, Cappabianca G, Maselli G, Marraudino N, Wuaranta M, Luca Tupputi SL. Blood damage related to cardiopulmonary bypass: in vivo and in vitro comparison of two different centrifugal pumps. ASAIO J 2004;50:473–8
20. Martin R, McKenty S, Thisdale Y, Lavallee P, Teijeira J, Bonneau D, Tetrault JP. Hemolysis during cardiopulmonary bypass. J Cardiothorac Anesth 1989;3:737–40
21. Sprung J, Gamulin S, Bosnjak ZJ, Kampine JP. Potassium correction of hypothermic hypokalemia induces hyperkalemia after rewarming. Can J Anaesth 1990;37:S69
22. Cameron AM, Ghobrial RM, Yersiz H, Farmer DG, Lipshutz GS, Gordon SA, Zimmerman M, Hong J, Collins TE, Gornbein J, Amersi F, Weaver M, Cao C, Chen T, Hiatt JR, Busuttil RW. Optimal utilization of donor grafts with extended criteria: a single-center experience in over 1000 liver transplants. Ann Surg 2006;243:748–53
23. Abouna GM, Aldrete JA, Starzl TE. Changes in serum potassium and pH during clinical and experimental liver transplantation. Surgery 1971;69:419–26
24. Xia VW, Du B, Braunfeld M, Neelakanta G, Hu KQ, Nourmand H, Levin P, Enriquez R, Hiatt JR, Ghobrial RM, Farmer DG, Busuttil RW, Steadman RH. Preoperative characteristics and intraoperative transfusion and vasopressor requirements in patients with low vs. high MELD scores. Liver Transpl 2006;12:614–20
25. Cameron A, Busuttil RW. AASLD/ILTS transplant course: is there an extended donor suitable for everyone? Liver Transpl 2005;11:S2–5