Validation of the peak bilirubin criterion for outcome after partial hepatectomy

Abstract

Introduction

Postoperative liver failure (PLF) is a dreaded complication after partial hepatectomy. The peak bilirubin criterion (>7.0 mg/dL or ≥120 μmol/L) is used to define PLF. This study aimed to validate the peak bilirubin criterion as postoperative risk indicator for 90-day liver-related mortality.

Methods

Characteristics of 956 consecutive patients who underwent partial hepatectomy at the Maastricht University Medical Centre or Royal Free London between 2005 and 2012 were analyzed by uni- and multivariable analyses with odds ratios (OR) and 95% confidence intervals (95%CI).

Results

Thirty-five patients (3.7%) met the postoperative peak bilirubin criterion at median day 19 with a median bilirubin level of 183 [121–588] μmol/L. Sensitivity and specificity for liver-related mortality after major hepatectomy were 41.2% and 94.6%, respectively. The positive predictive value was 22.6%. Predictors of liver-related mortality were the peak bilirubin criterion (p < 0.001, OR = 15.9 [95%CI 5.2–48.7]), moderate-severe steatosis and fibrosis (p = 0.013, OR = 8.5 [95%CI 1.6–46.6]), ASA 3–4 (p = 0.047, OR = 3.0 [95%CI 1.0–8.8]) and age (p = 0.044, OR = 1.1 [95%CI 1.0–1.1]).

Conclusion

The peak bilirubin criterion has a low sensitivity and positive predictive value for 90-day liver-related mortality after major hepatectomy.

Introduction

Postoperative liver failure (PLF) is a serious complication after partial hepatectomy for benign or malignant liver disease.¹ PLF occurs in approximately 5% of all patients undergoing partial liver resection.^{2, 3} A liver remnant volume of 25% is regarded as the minimum to prevent PLF.² In patients with compromised liver function, up to 40% needs to be preserved.¹

PLF is characterized by an impaired synthetic, secretory, and detoxifying function of the liver, and accounts for the majority of mortality after extensive partial hepatectomy.¹ Strategies to avoid an insufficient remnant liver include staged resection (two-stage hepatectomy,⁴ associating liver partition and portal vein ligation for staged hepatectomy (ALPPS)),⁵ and portal vein embolization (PVE).⁶ Postoperative treatments lack and only intensive support can be provided when PLF occurs. It is therefore essential to have an accurate postoperative clinical risk indicator that can predict PLF to provide early optimal support.

Three postoperative clinical risk indices that are currently used are the 50–50 criteria (prothrombin time <50% and serum bilirubin >50 μmol/L on postoperative day 5),^{7, 8} the definition of PLF by the International Study Group of Liver Surgery,⁹ and the Model for End-stage Liver Disease.^{10, 11} However, these parameters have revealed to be suboptimal to detect patients with developing PLF based on their definitions as shown by recent validation studies.^{12, 13, 14}

In 2007, Mullen et al. proposed a definition for PLF based on analysis of 1059 patients without cirrhosis who underwent major hepatectomy between 1995 and 2005 at three hepatobiliary centres in the United States and Italy.¹⁵ The authors stated that the occurrence of a systemic bilirubin level of >7.0 mg/dL (≥120 μmol/L, ‘peak bilirubin criterion’) within 90 days after major hepatectomy provides a sensitivity of 93.3% for liver-related death and an odds ratio (OR) of 250 (95% confidence interval, 25.0 to >1000) for 90-day liver-related mortality.

Recently, this risk indicator has been retrospectively validated in a prospectively constructed single centre European database.¹⁴ Analysis of 680 hepatectomies in patients without cirrhosis resulted in a positive predictive value of 61.4% for major morbidity and 40.5% for overall mortality when the peak bilirubin criterion was applied within 10 days after hepatectomy.¹⁴ However, this is a rather short period of time since the majority of complications occur within 90 days after liver resection.¹⁶

The peak bilirubin criterion is one of the main predictors used in daily practice,¹⁷ but modern practice has changed in the last decade. With the increasing incidence of non-alcoholic fatty liver disease, extending indications for resection,¹⁸ and complex vascular procedures,¹⁹ postoperative morbidity and mortality are negatively affected.^{20, 21} In addition, chemotherapy-associated liver injury (e.g. sinusoidal obstruction syndrome, steatohepatitis) seems to increase liver-related morbidity and 90-day mortality after liver resection.^{21, 22}

We hypothesized that following current practice, the criterion would be met more often and mortality rates would be higher. Therefore this study aimed to validate the peak bilirubin criterion as a postoperative clinical risk indicator regarding major morbidity and liver-related death within 90 days after partial hepatectomy in two European tertiary hepatobiliary referral centres.

Methods

Patients

Prospectively collected data from all consecutive patients who had undergone minor or major hepatic resection between January 2005 and December 2012 at the Maastricht University Medical Centre (the Netherlands) or the Royal Free Hospital London (United Kingdom) were reviewed. All patients with age >18 years who underwent liver resection were included, independent of ethnicity, quality of the background liver and preoperative systemic levels of liver-related parameters (e.g. bilirubin, alanine aminotransferase). Before surgery, a complete medical history was taken and patients underwent physical examination and liver-related blood tests. All patients were discussed at a multidisciplinary meeting to determine optimal medical and surgical treatment. Patients who underwent extended hepatectomy received computed tomography-volumetry prior to surgery. Preoperative PVE was performed in patients with a limited predicted functional liver remnant.

Surgical techniques and perioperative management

All liver resections were classified in accordance with the International Hepato-Pancreato-Biliary Association Brisbane nomenclature.²³ Liver resection was performed as described before.²⁴ To prevent excessive blood loss, central venous pressure was maintained below 5 cm H₂O during transection. Intermittent Pringle maneuver was performed in case of increased bleeding risk. Hemostasis was achieved using bipolar coagulation and argon beam coagulation, sutures and clips. Intraoperative transfusion of packed red blood cells (RBCs) and fresh frozen plasma (FFP) was performed according to hospital protocols. Postoperatively, patients were admitted to the overnight recovery room or intensive care unit and transferred to the surgical ward the next day if clinically stable.

Definitions

Liver resections were divided into minor (<3 Couinaud segments) and major (≥3 Couinaud segments) resections.²⁵ Background steatosis was morphologically quantified by histopathologists using a four-graded scale (0–3) as defined by Kleiner et al.²⁶ Scores 1–3 were considered to correspond with fat deposition in 5–33%, 33–66% and >66% of hepatocytes. Preoperative chemotherapy consisted of adjuvant chemotherapy for the primary tumor, or neo-adjuvant chemotherapy to downsize liver metastases within six months before liver surgery. The number of units of transfusion of RBCs or FFP perioperatively and within 24 h after surgery was documented. Liver failure was defined according to the peak bilirubin criterion of Mullen et al.¹⁵ Postoperative 90-day morbidity was defined in accordance with the surgical Dindo–Clavien classification,²⁷ and clinically relevant (major) morbidity was defined as a Dindo–Clavien score of 3 or more. Overall mortality was defined as all cause death within 90 days after liver resection. Ninety-day morbidity and mortality were included irrespective of whether this occurred during first admission, after discharge or during readmission. Mortality was subdivided into overall and liver-related (i.e. liver failure, multi-organ failure including liver failure) mortality for validation analyses. Moreover, the liver surgery-specific composite endpoint (LSSCEP) composed of ascites, postoperative liver failure, bile leakage, intra-abdominal hemorrhage, intra-abdominal abscess, and mortality, was used to assess liver surgery-specific complications.²⁸ Due to implementation of enhanced recovery after surgery (ERAS) programs in liver surgery,²⁹ bilirubin levels were determined on postoperative day 1, 3, and if deemed necessary, day 5 or beyond. Medical records of all patients were checked up to 90 days after partial hepatectomy.

Statistical analyses

Statistical analyses were performed using SPSS^® version 20 (IBM, Armonk, New York, USA). Data are expressed as median [range] and percentages. Sensitivity, specificity and predictive values for major postoperative morbidity and 90-day liver-related mortality were computed. Clinicopathologic variables associated with morbidity and mortality were examined using univariable and, where applicable, multivariable analyses using logistic regression. For multivariable analysis, a univariable inclusion criterion of p ≤ 0.15 was used. Statistical significance was considered at a p < 0.05. Results were depicted in p-values with odds ratio (OR) and 95% confidence interval (95%CI).

Results

Patient characteristics

A total of 956 patients were included in this study (Table 1). Four hundred forty-two patients (45.0%) underwent liver resection at the Maastricht University Medical Centre, whereas 514 patients (55.0%) underwent liver surgery at the Royal Free Hospital. Median age of the patients was 64 [20–88] years. Cardiovascular co-morbidity (e.g. hypertension, previous myocardial infarction) was present in 442 patients (46.2%) and diabetes mellitus was present in 118 patients (12.3%). ‘Other co-morbidity’ (228 patients, 23.8%) comprised conditions such as hypothyroidism, a history of deep venous thrombosis and/or pulmonary embolism.

Table 1.

Demographic and clinical characteristics of the patient cohort. Values in parentheses are percentages, values in square brackets depict range

Characteristic	n = 956
Age (years)	64 [20–88]
Female	409 (42.8)
Co-morbidity
Cardiovascular	442 (46.2)
Pulmonary	107 (11.2)
Diabetes mellitus	118 (12.3)
Renal	16 (1.7)
Other	228 (23.8)
ASA 3 or 4	148 (15.5)
Indication
Primary hepatic tumor	123 (12.9)
Hepatocellular carcinoma	61 (6.4)
Cholangiocarcinoma	38 (4.0)
Gallbladder carcinoma	21 (2.2)
Other	3 (0.3)
Metastasis	718 (75.1)
Primary colorectal carcinoma	657 (68.7)
Other primary tumor	61 (6.4)
Benign	115 (12.0)
Histology
Moderate/severe steatosis	219 (22.9)
Moderate/severe fibrosis	77 (8.1)
Cirrhosis	24 (2.5)

ASA, American Society of Anesthesiologists

Malignant liver disease was the indication for resection in 841 patients (88.0%), with hepatocellular carcinoma as the most common indication in primary liver disease (61 patients, 6.4%), and colorectal liver metastases in metastatic disease (657 patients, 68.7%). Hepatocellular adenoma and cavernous hemangioma (both 19 patients, 2.0%) were the most common benign indications for hepatic resection. Impaired background liver quality was present in 320 patients (219 patients with moderate to severe steatosis, 77 patients with moderate to severe fibrosis and 24 patients with cirrhosis).

Operative details

A total of 490 patients (51.3%) underwent limited resection and 466 patients (48.7%) underwent resection of 3 or more segments (Table 2). Multi-segmentectomies (303 patients, 31.7%) and wedge resections (288 patients, 30.1%) were the most commonly performed surgical procedures followed by a right hepatectomy in 196 patients (20.5%). A total of 426 patients (44.9%) received neo-adjuvant chemotherapy. Median surgery time was 267 [45–1200] minutes and median perioperative blood loss was 600 [20–11,600] mL. Perioperative transfusion of RBCs and FFP was carried out in 235 (27.5%) and 90 (10.6%) patients, respectively.

Table 2.

Peri -and postoperative details of the cohort. Values in parentheses are percentages, values in square brackets depict range

Characteristic	n = 956
Procedure
Right hepatectomy	196 (20.5)
Extended right hepatectomy	62 (6.5)
Left hepatectomy	56 (5.9)
Extended left hepatectomy	22 (2.3)
Central resection	27 (2.8)
Wedge/segmentectomy	288 (30.1)
Multisegmentectomy	303 (31.7)
Surgery type
Minor (<3 Couinaud segments)	490 (51.3)
Major (≥3 Couinaud segments)	466 (48.7)
Preoperative procedures
Portal vein embolization	59 (6.2)
Chemotherapy	426 (44.9)
Previous liver resection	106 (11.1)
Intraoperative details
Laparoscopy	95 (9.9)
Operation time (minutes)	267 [45–1200]
Intraoperative blood loss (mL)	600 [20–11600]
Vascular clamping
Patients (number)	214 (25.2)
Clamping time (minutes)	30 [8–75]
Transfusion RBCs (units)
Patients (number)	235 (27.5)
Units (number)	0 [0–25]
Transfusion FFP (units)
Patients (number)	90 (10.6)
Units (number)	0 [0–13]
Length of stay
Initial length of stay in hospital	9 [2–167]
Complication grade
Complications present	453 (47.4)
Dindo–Clavien grade 1	67 (7.0)
Dindo–Clavien grade 2	142 (14.9)
Dindo–Clavien grade 3	134 (14.0)
Dindo–Clavien grade 4	73 (7.6)
Dindo–Clavien grade 5	37 (3.9)
Liver surgery-specific composite endpoint
Liver surgery-specific composite endpoint present	194 (20.3)
Ascites	19 (2.0)
Liver failure	32 (3.3)
Bile leak	70 (7.3)
Intra-abdominal hemorrhage	25 (2.6)
Intra-abdominal abscess	77 (8.1)
90-day mortality	37 (3.9)
Liver-related mortality	23 (2.4)
Re-admission
30-day re-admission	97 (10.1)

RBCs, packed red blood cells; FFP, fresh frozen plasma

Postoperative details

The median length of hospital stay following partial hepatectomy was 9 [2–167] days (Table 2). Overall complications were present in 453 patients (47.4%) and death occurred in 37 patients (3.9%) within 90 days after surgery. Ninety-day major morbidity was present in 244 patients (25.5%) and liver-related mortality in 23 patients (2.4%). Non liver-related causes of death were of pulmonary (respiratory insufficiency, pneumonia, pulmonary embolus), renal (renal failure), cardiac (myocardial infarction) and gastroenterological (hemorrhage or ischemia) origin. A total of 194 patients (20.3%) met at least one of the criteria of the liver-surgery specific composite endpoint,²⁸ with intra-abdominal abscess (77 patients, 8.1%) and bile leakage (70 patients, 7.3%) as most frequent complications.

Validation of the peak bilirubin criterion

During the postoperative course, the median of the highest bilirubin level measured in all patients was 25 [0–588] μmol/L and occurred on postoperative day 1 [0–53]. The peak bilirubin criterion was met in 35 patients (3.7%), with a median bilirubin level of 183 [121–588] μmol/L on postoperative day 19 [5–53]. The positive predictive value of the peak bilirubin criterion for liver-related death within 90 days after major liver resection (n = 458) was 22.6% (Table 3). Specificity for liver-related death within 90 days was 94.6% after major liver surgery, but sensitivity was only 41.2%. In addition, of all patients with severely elevated bilirubin levels (for the present study defined as >250 μmol/L), 10 patients (66.7%) out of 15 survived and bilirubin levels normalized (Fig. 1).

Table 3.

Sensitivity, specificity and predictive values of the peak bilirubin criterion for postoperative overall and liver-related 90-day mortality. Values are depicted in percentages

	Overall 90-day mortality		Liver-related 90-day mortality
	All resections (n = 936)	Major liver resections (n = 459)	All resections (n = 934)	Major liver resections (n = 458)
Sensitivity	28.6	34.8	36.4	41.2
Specificity	97.2	94.7	97.0	94.6
PPV	28.6	25.8	22.9	22.6
NPV	97.2	96.5	98.4	97.7

PPV, positive predictive value; NPV, negative predictive value

Figure 1.

Highest postoperative serum bilirubin levels in μmol/L from all patients who met the peak bilirubin criterion

Predictors for liver-related 90-day mortality

In univariable analysis (Table 4), five factors (age, major liver resection, duration of surgery, Pringle maneuver, and peak bilirubin criterion) were identified as significant prognostic factors affecting liver-related 90-day mortality. Two additional variables (American Society of Anesthesiologists (ASA) 3–4, and the co-occurrence of moderate-severe steatosis and moderate-severe fibrosis/cirrhosis, p < 0.15) were included in the multivariable analysis. Four independent predictors for liver-related 90-day mortality were identified, with the peak bilirubin criterion being the strongest predictor (p < 0.001, OR = 15.9 [95%CI 5.2–48.7]). The other significant predictors for 90-day liver-related mortality were the co-occurrence of moderate-severe steatosis and moderate-severe fibrosis/cirrhosis (p = 0.013, OR = 8.5 [95%CI 1.6–46.6]), ASA 3–4 (p = 0.047, OR = 3.0 [95%CI 1.0–8.8]) and age (p = 0.044, OR = 1.1 [95%CI 1.0–1.1]).

Table 4.

Uni- and multivariable analyses on postoperative liver-related 90-day mortality in the patient cohort

Prognostic factor	Univariable		Multivariable
	p-Value	OR [95%CI]	p-Value	OR [95%CI]
Agea	0.002	1.1 [1.0–1.1]	0.044	1.1 [1.0–1.1]
Female sex	0.232	0.6 [0.2–1.4]
ASA 3–4	0.052	2.5 [1.0–6.1]	0.047	3.0 [1.0–8.8]
Presence of co-morbidities	0.783	1.1 [0.5–2.7]
Malignant indication (vs benign)	0.275	3.1 [0.4–23.0]
Preoperative chemotherapy	0.566	0.8 [0.3–1.8]
Background liver histology
Normal	Reference		Reference
Moderate-severe steatosis	0.499	0.7 [0.2–2.3]	0.754	0.8 [0.2–3.7]
Moderate-severe fibrosis or cirrhosis	0.504	1.7 [0.4–7.5]	0.446	1.9 [0.4–9.8]
Moderate-severe steatosis and fibrosis or cirrhosis	0.116	3.4 [0.7–15.6]	0.013	8.5 [1.6–46.6]
Open liver resection (vs. laparoscopic)	0.385	0.4 [0.1–3.1]
Major liver resection (vs. minor)	0.008	3.9 [1.4–10.6]	0.093	2.9 [0.8–9.7]
Duration of surgery (hours)b	0.026	1.2 [1.0–1.3]
Blood transfusionb	0.064	2.2 [1.0–5.3]
Pringle maneuver	0.032	2.5 [1.1–6.0]	0.097	2.3 [0.9–5.9]
Peak bilirubin ≥120 μmol/L	<0.001	18.7 [7.3–48.4]	<0.001	15.9 [5.2–48.7]

P-values depicted in bold are significant in multivariable analysis (p < 0.05).

OR, odds ratio; CI, confidence interval; ASA, American Society of Anesthesiologists.

^aAge as nominal variable (per year).

^bLeft out due to potential collinearity.

Discussion

In the present patient cohort, sensitivity and specificity of the peak bilirubin criterion for 90-day liver-related death after major liver resection were 41.2% and 94.6%, respectively, whereas the positive predictive value only reached 22.6%. In multivariable analysis, the peak bilirubin criterion (p < 0.001, OR = 15.9 [95%CI 5.2–48.7]), co-existing moderate-severe steatosis and moderate-severe fibrosis/cirrhosis (p = 0.013), ASA 3–4 (p = 0.047), and age (p = 0.044) were independent predictors of liver-related death.

The study of Mullen et al. suggests that patients who meet the criterion of the peak bilirubin have a more than 30% chance of dying from liver failure after major hepatectomy, in addition to a sensitivity of 93.3% and specificity of 94.3% for liver failure-related mortality. The recent validation study of Skrzypczyk et al. showed a positive predictive value of 40.5% for overall mortality when the peak bilirubin criterion was met within 10 days after partial hepatectomy, with a sensitivity and specificity for overall mortality of 56.7% and 96.1%, respectively.

We hypothesized that in the current era of extensive surgeries in patients with (hepatic) co-morbidity, more patients would meet the peak bilirubin criterion postoperatively and 90-day mortality could be increased due to a lack of treatment. In addition, more patients were expected to meet the criterion when the observation period was 90 days instead of 10 days as in Skrzypczyk et al. Whereas we indeed saw that combined moderate to severe steatosis and severe fibrosis was a significant risk factor for 90-day liver-related mortality after partial hepatectomy, we observed a much lower positive predictive value of the peak bilirubin criterion. Moreover, of the 15 patients with a severely elevated bilirubin level, 10 patients survived and bilirubin levels normalized. This confirms the statement of Mullen et al. that the peak bilirubin criterion is a turning point instead of a point of no return.¹⁵ In addition, improvements in surgical techniques, preoperative PVE and perioperative care might have led to better support and timely transfer to the intensive care ward.

Despite the fact that the present prognostic values were lower than reported by others, the peak bilirubin was still identified as conferring the highest risk for 90-day postoperative liver-related mortality in multivariable analysis. Whereas other risk indicators such as the 50–50 criteria or the validation study of Skrzypczyk et al. focus on detection of PLF in the early postoperative course, the present analysis showed that bilirubin levels peaked on median postoperative day 19 indicating that 10 days is too short to detect the majority of patients with this complication.

Although its clinical use is widespread, serum bilirubin reflecting secretory liver function might not be the optimal indicator for poor outcome after partial hepatectomy as shown in a systematic literature review.³⁰ In the past decades, human and animal studies have shown that plasma and hepatic bile salts are direct indicators of secretory liver function.³¹ The relative hepatic overload of bile salts after liver resection³² in combination with toxicity of excess hydrophobic bile salts and ensuing impairment of secretory function,³³ can cause a vicious cycle of hepatotoxicity in the remnant liver. Moreover, a delicate intrahepatic balance of bile salts is needed for proper liver regeneration as shown in animal studies.³⁴ It may be worthwhile to explore the prognostic potential of bile salts in PLF.

Although the clinical characteristics of the present study resemble those of the cohort of Mullen et al., some dissimilarities were present. Whereas having a cirrhotic liver was an exclusion criterion in the latter study, we included patients with cirrhosis (n = 24). Furthermore, we included patients with a preoperative serum bilirubin level >2.0 mg/dL (or >34 μmol, n = 25), whereas these patients were excluded in Mullen's cohort. Statistical testing, however, did not reveal differences in sensitivity, specificity and predictive values for the peak bilirubin criterion when aforementioned patient groups were excluded (data not shown).

While prospectively collected, data of this study were retrospectively analyzed. Due to implementation of enhanced recovery after surgery (ERAS) programs in liver surgery, several systemic parameters such as the international normalized ratio and C-reactive protein were not measured daily.^{29, 35} Validation of for instance the 50–50 criteria was thus not possible in this cohort.

In conclusion, the present study found a rather low positive predictive value and sensitivity of the peak bilirubin criterion for liver-related mortality within 90 days after major liver resection. Nevertheless, it was still identified as the most risk-bearing factor for postoperative liver-related mortality within 90 days after partial liver resection in multivariable analysis. Prospective studies should focus on novel liver function-related parameters such as bile salts, and/or combined functional and volumetric criteria.³⁶

Funding

This study was established without the use of any funding.

Conflict of interest

None declared.