Figure 1. Postoperative fluctuations in serum hemoglobin level (A) and serum hematocrit level (B).
Figure 2. Postoperative fluctuations in serum hemoglobin level (A) and serum hematocrit level (B) in patients with and without cirrhosis. Differences between groups were not significant.
Torzilli G, Gambetti A, Del Fabbro D, Leoni P, Olivari N, Donadon M, Montorsi M, Makuuchi M. Techniques for Hepatectomies Without Blood Transfusion, Focusing on Interpretation of Postoperative Anemia. Arch Surg. 2004;139(10):1061–1065. doi:10.1001/archsurg.139.10.1061
Transient postoperative anemia is partially a physiologic phenomenon, and variations in blood transfusion rates after liver resection in different series in part are due to different interpretations of postoperative anemia. Based on the hypothesis that transient postoperative anemia is partially a physiologic phenomenon, we analyzed serum hemoglobin and hematocrit values in patients who underwent liver resection without blood transfusion to check fluctuations.
Prospective cohort study.
Forty-six consecutive patients with primary and metastatic liver tumors.
Surgical treatment consisting of dissection technique performed under intermittent warm ischemia, using intraoperative ultrasonography, and without blood transfusion.
Main Outcome Measures
Hematocrit and hemoglobin concentrations in serum sampled preoperatively and on the first, third, fifth, and seventh postoperative days.
No postoperative mortality and major morbidity were observed. No patient received a blood transfusion. The hematocrit and hemoglobin concentrations in serum were significantly lower on the third postoperative day than on the first, fifth, and seventh postoperative days; differences among the first, fifth, and seventh postoperative days were not significant.
The fluctuations of hemoglobin and hematocrit levels after liver resection showed a steady and significant decrease until the third postoperative day and then an increase. Therefore, a decrease in the hemoglobin and hematocrit levels between first and fifth postoperative days without evidence of active bleeding from drain discharge or any other possible source of bleeding does not justify blood administration.
Liver resection has historically been considered the treatment of choice for liver tumors, and recent studies have shown that, by following well-established policies, it is possible to perform hepatectomy without mortality or major morbidity.1,2 Several techniques have been advocated to yield these favorable results, and there is almost total agreement on issues such as the need for bloodless dissection and operation without blood transfusion; massive blood loss and blood transfusions impair short- and long-term patient outcome.3- 7
Nonetheless, several authors have recently reported relatively high rates of patients receiving blood transfusions,7- 11 although there are also large series in which fewer than 10% of patients received transfusions.1,12 Possible reasons for these discrepancies fall into 3 categories: surgical strategy, surgical technique, and administration of blood in relation to the occurrence of postoperative anemia.
To verify the natural course of postoperative anemia, we performed a prospective cohort analysis of the fluctuations of serum hemoglobin and hematocrit values in patients who underwent liver resection.
Between May 1, 2001, and September 30, 2003, 46 consecutive patients with liver tumors were enrolled. The patients' characteristics are given in Table 1.
Cirrhotic patients were selected for surgery according to a previously published flowchart1 on the basis of the presence of ascites, serum total bilirubin level, and indocyanine green retention rate at 15 minutes. In all of the other patients, indication for surgery was established on the basis of tumor stage, technical feasibility, and volume of the remnant liver.13 Portal vein embolization was never needed.
The terminology for liver anatomy and resections is based on the recommendations of the International Hepato-Pancreato-Biliary Association.14 A J-shaped laparotomy was performed in 36 patients, and a J-shaped thoracophrenolaparotomy was performed in the remaining 10, who had tumors involving segments 1, 7, and 8. Major resection (bisegmentectomy, sectionectomy, and larger resections) was performed in 9 patients (20%); segmentectomy was performed in 11 (24%); and limited resection, including strictly nonanatomic resections, which means incomplete exposure of the hepatic veins on the liver cut surface even though a segment or a section was almost completely removed, was performed in 26 (56%) (Table 2). Limited resections were also associated with segmentectomies in 4 cases. Eight of the 11 segmentectomies were performed in cirrhotic patients, which accounted for 38% of liver resections in this group; the remaining 13 patients underwent major resection (n = 3 [14%]) or limited resection (n = 10 [48%]). In 24 patients, tumors were located in segments 1, 4b, 7, and 8, accounting for 52% of the entire series: 11 patients had colorectal liver metastases on fatty liver; 8 patients had hepatocellular carcinoma on cirrhotic liver; 3 had mass-forming cholangiocellular carcinoma, gastrointestinal stromal tumor metastases, and multiple giant hemangiomas on normal liver; and 2 had hepatocellular carcinoma on chronic hepatitis. Vascular invasion was present in 10 patients (22%) who underwent resection for colorectal cancer liver metastases (n = 6), multiple hepatocellular carcinoma (n = 3), or mass-forming cholangiocellular carcinoma (n = 1). Among them, the inferior vena cava was reconstructed in one and the right hepatic vein in another; in both patients, reconstruction was accomplished by direct suture. In 2 patients hepatectomies were associated with other organs: the distal part of the stomach in an 87-year-old patient with a 28-cm hepatocellular carcinoma in liver cirrhosis that infiltrated the stomach, and the descending colon in a patient with synchronous colorectal liver metastases. In one patient, who was a carrier of colorectal cancer liver metastases located in segments 6 and 7, the diaphragm was resected because of tumor infiltration.
Liver dissection was performed under intermittent clamping by the Pringle maneuver,15 using Pean forceps and a bipolar electrocautery for vessel coagulation; each vessel thicker than 2 mm was ligated with thin (2/3-0) sutures. Preconditioning and total vascular exclusion were not performed. The cut surface of the liver was secured by electrocautery, argon beam, fibrillar-oxidated regenerated cellulose, and fibrin glue. Patients underwent careful examination to rule out bile leakage, and cholangiography was not performed for this purpose.16
In cirrhotic patients, blood loss and ascites production during the operation were balanced by infusing 10% to 20% more fresh frozen plasma than the volume of blood lost. Intraoperative blood transfusions would have been given only if the hematocrit had been less than 30%. A total volume of 4 to 5 mL of fluid per kilogram per hour was infused. Equal volumes of crystalloid and 5% to 10% glucose solution were administered.
The level of anesthesia was maintained by general and epidural anesthetic, thereby reducing the quantity of inhalation agents and intravenous drugs. Central venous pressure was maintained between 0 and 4 cm H2O to reduce the back-bleeding from the hepatic veins and/or their tributaries. Sodium hemisuccinate hydrocortisone (100 mg) was injected intravenously before vascular occlusion was started to protect the liver during warm ischemia.
The patients received a total of about 40 to 45 mL of fluid per kilogram per day by infusion for the first 2 postoperative days. Glucose solution was given for calories, hydrocortisone was also given intravenously for 3 days postoperatively, and short-term antibiotic prophylaxis, omeprazole, and low-molecular-weight heparin sodium were administered. Appropriate oral intake was restored from the third postoperative day, and fluid administration was stopped at that time.
If the postoperative course was uneventful, the chest drain (if placed) was removed as soon as possible (during the first 2-4 postoperative days). Closed-suction drains were always left in the peritoneal cavity, with the tip facing the cut surface of the liver. In accordance with a previous report,1 the drain was carefully checked for blood discharge; an emergency laparotomy would have been performed if such discharge exceeded 100 mL/h. In patients with stable hemodynamic measures, blood transfusions would have been administered only to patients with hematocrit values below 20%; in this case, patients would have been dropped from the study. If no bile leakage was detected and the bilirubin concentration was less than that of the control on the fifth postoperative day, the drains were removed on the seventh postoperative day; conversely, the tube was maintained without closed suction until the bilirubin concentration was less than that in the previous control. After drain removal, ultrasonography was carried out in each patient.
Two-tailed t test for unpaired homoscedastic data was used to compare the mean serum hemoglobin and hematocrit levels sampled the day before operation and on the first, third, fifth, and seventh postoperative days. Similarly, the mean serum hemoglobin and hematocrit levels on the first, third, fifth, and seventh postoperative days were compared between cirrhotic and noncirrhotic patients. P value for significance was set at .05.
No 30-day mortality was seen, and all of the patients returned to their normal daily life. The overall morbidity rate was 17.4%. No major complication occurred, and no reoperation was required. Morbidity consisted of transient ascites in 5 cases, superficial wound dehiscence in 2 cases, and transient fever in 1 case. All cases of transient postoperative ascites occurred in patients with cirrhosis.
Mean clamping time was 43 minutes (median, 35 minutes; range, 15-90 minutes). Overall mean blood loss was 296 mL (median, 200 mL; range, 50-800 mL). No blood transfusions were administered, and no patients were excluded because of the aforementioned criteria for blood administration either intraoperatively (hematocrit <30%) or postoperatively (hematocrit <20%).
The overall mean operation time was 326 minutes (median, 330 minutes; range, 150-600 minutes). Closed-suction drains were placed in all patients, with an average of 1.7 drains per patient (median, 2; range, 1-3); they were removed on the seventh postoperative day in all but 3 cases because of the lack of decrement in the bilirubin concentration in the drain discharge. In these last patients, drains were removed on the ninth postoperative day for 2 patients and the 11th postoperative day for the third patient when the bilirubin level had fallen.
Overall mean fluid balances on the day of the operation and on the first and second postoperative days were, respectively, +1350 mL (median, +1100 mL; range, +200 to +2200 mL), −793.6 mL (median, −800 mL; range, +900 to −1800 mL), and −829.5 mL (median, −850 mL; range, +600 to −1700 mL). Fluid administration was stopped on the third postoperative day in all patients.
Serum hemoglobin levels in the overall patient groups as well as those with and without cirrhosis are given in Table 3. Overall serum hemoglobin levels are shown in Figure 1A. The serum hemoglobin levels in samples taken postoperatively were all significantly lower than that on the day before surgery (P<.001). The serum hemoglobin level in samples on the first postoperative day was significantly higher than that on the third postoperative day (P = .03); this latter value was significantly lower than that on the fifth (P = .01) and seventh (P = .03) postoperative days. The other comparisons were not significant: first vs fifth postoperative day (P = .70), first vs seventh postoperative day (P = .74), and fifth vs seventh postoperative day (P = .5). All differences between mean serum hemoglobin levels on the same postoperative day between cirrhotic and noncirrhotic patients were not significant (Figure 2A).
The serum hematocrit levels (Table 3) (Figure 1B) in samples obtained postoperatively were all significantly lower than those on the day before surgery (P<.001). The serum hematocrit level in samples on the first postoperative day was significantly higher than that on the third postoperative day (P = .02); the latter value was significantly lower than that on the fifth and seventh postoperative days (P = .03 for both). The other comparisons were not significant: first vs fifth postoperative day (P = .80), first vs seventh postoperative day (P = .10), and fifth vs seventh postoperative day (P = .60). All differences between mean serum hematocrit levels on the same postoperative day between cirrhotic and noncirrhotic patients were not significant (Figure 2B).
The mean hospital stay was 10.3 days (median, 10 days; range, 8-12 days).
Limitation of intraoperative blood loss and the consequent lack of necessity for blood transfusion is a widely accepted goal,3,4 because these factors impair the patients' outcome5 and prognosis.6 One possible reason for these findings comes from experimental studies in which impairment in liver regeneration by transfused blood has been observed17; conversely the same authors reported that administration of fresh frozen plasma seems to reduce this negative effect. However, in several centers, perioperative blood transfusions are still administered in a substantial proportion of patients (31%-91%).7- 11
In this study, adopting the aforementioned policy, no patients underwent blood transfusion and none were excluded; this result is in line with those of recent reports based on Eastern series1,12 and is well below the maximum transfusion rate of 30% that is generally considered acceptable.18 Our indications for blood transfusion are strict, both intraoperatively and postoperatively. However, the lack of major morbidity and mortality is indirect proof that this policy does not increase patients' perioperative risk.
A possible explanation for this low rate of transfusion is the limited number of major hepatectomies in our series and in previous series that adopted similar policies,1,12 despite the fact that in all of these series, major resections included resection of just one section. This is mainly related to the extensive use of the ultrasound guidance to accomplish liver resection, which is probably enhanced by the habit to face with a high rate of cirrhotic patients; in fact, in these patients, sparing of functioning parenchyma is mandatory.19 The mean tumor size, mean number of lesions, percentage of patients with cirrhosis, percentage of patients with tumors located in segments 1, 4b, 7, and 8, and percentage of patients with invasion of major vessels were consistent with those in major series (Table 1). Furthermore, the rate of patients who underwent surgery among all cases referred to our outpatient clinic for primary or metastatic liver tumors was relatively high (73%).20 These considerations should help to argue against a major selection bias.
Some authors have attributed the different amount of blood loss and the consequent need for blood transfusion to the dissection technique, and to limit blood loss, they proposed a new method for liver dissection.21 However, Takayama et al22 showed in a controlled randomized trial that dissection technique is not a key factor in determining the blood loss and, consequently, the need for blood transfusion. Furthermore, the present series and previous ones have demonstrated that with the classic and inexpensive Pean-clasia crush clamping of the hepatic parenchyma, it is possible to carry out bloodless liver resection, with less than 10% of patients receiving blood transfusion1,12; to our knowledge, this rate remains the lowest ever published. Therefore, irrespective of surgical technique and vascular clamping, the reason for such widely different rates of blood transfusion among different series may be explained not only by differences in surgical strategy but also by different interpretations of postoperative anemia.
However, the natural postoperative course of serum hematocrit and hemoglobin levels is still not well known. Furthermore, among the factors that contribute to postoperative anemia, the amount of intraoperative fluid administration should not be neglected. In fact, liver mobilization and traction compress the inferior vena cava, reducing the cardiac inflow and outflow, which, if not recognized, may simulate a sudden hypovolemia. This misinterpreted phenomenon may induce the anesthesiologist to administer massive amount of fluids, with consequent hemodilution mimicking postoperative anemia and thus justifying blood transfusion. In the present series, we have demonstrated that in uncomplicated cases, with a fixed amount of intraoperative and postoperative fluid infusion per kilogram per hour, in patients with or without cirrhosis, serum hemoglobin and hematocrit levels fluctuate physiologically between the first and fifth postoperative days (Figure 1) (Table 3). The lowest levels occur on the third day, and after this we observed a constant and progressive increment until patient discharge in patients with and without cirrhosis (Figure 2). This happens in the absence of blood loss confirmed by drain discharges and postoperative ultrasound. Furthermore, serum hemoglobin and hematocrit values would be expected to be lower on the first postoperative day rather than on the third based on hemodilution; in fact, the fluid balance was positive on the day of the operation and negative on the first 2 postoperative days.
The evidence of these fluctuations should induce surgeons to consider that postoperative blood transfusion solely because of anemia is avoidable in most cases, excluding those with otherwise untreatable circulatory instability. Furthermore, it has been well stated that transient postoperative anemia is of some benefit for cirrhotic patients by reducing blood viscosity5; in the present series, it occurred in cirrhotic and noncirrhotic patients, with no significant differences (Figure 2; Table 3).
In conclusion, our results are consistent with previous reports concerning the safety of liver surgery. To our knowledge, this is the first report that focuses on the fluctuations of hemoglobin and hematocrit levels after liver resection; Suc et al23 described the biochemical changes after hepatectomy, limiting their observations to a retrospectively selected group of patients who did not have underlying liver disease. Our series consisted of a consecutively enrolled cohort of patients, and among them only 13% had normal livers. Based on our results, a decrement in the hemoglobin and hematocrit levels between the first and fifth postoperative days without evidence of active bleeding from drain discharge or collection on ultrasound examination or any other possible source of bleeding does not justify blood administration. With this perspective, blood transfusion can be drastically limited and reserved for only patients with objective signs of massive bleeding or with circulatory instability otherwise not treatable.
Correspondence: Guido Torzilli, MD, PhD, Hepatobiliary Surgery Unit, Reparto di Chirurgia Generale 1, Azienda Ospedaliera della Provincia di Lodi, Largo Donatori del Sangue, 2, I-26900 Lodi, Italy (firstname.lastname@example.org or email@example.com).
Accepted for publication February 4, 2004.