Future remnant liver function estimated by combining liver volumetry on magnetic resonance imaging with total liver function on ^99mTc-mebrofenin hepatobiliary scintigraphy: can this tool predict post-hepatectomy liver failure?

Abstract

Introduction

Posthepatectomy liver failure (PHLF) is a major complication after hepatectomy with a high mortality rate and is likely to happen in insufficient liver remnant. We hypothesize that assessment of the estimated future liver remnant function (eFLRF), combining future remnant liver volume (FLRV) with total liver function (TLF), is an accurate formula for prediction of PHLF.

Methods

88 patients undergoing hepatectomy were included. The ratio of the future liver remnant volume (FLRV%) was measured on MRI. TLF was estimated by liver clearance of ^99mTechnetium (Tc)-mebrofenin on hepatobiliary scintigraphy (HBS). eFLRF was calculated by multiplying FLRV% by TLF. Cut-off values of FLRV% and eFLRF predicting PHLF, were defined by receiver-operating-characteristic (ROC) analysis.

Results

PHLF occurred in 12 patients (13%). Perioperative mortality was 5/12 (41%). Multivariate analysis showed that FLRV% cut off at 40% was not an independent predictive factor. eFLRF cut off at 2.3%/min/m² was the only independent predictive factor for PHLF. For FLRV% vs. eFLRF, positive predictive value was 41% vs. 92% and Odds Ratio 26 vs. 836.

Conclusion

FRLF measured by combining FLRV% and TLF is a more valuable tool to predict PHLF than FLRV% alone. The cutoff of eFLRF can be used in clinical decision making.

Introduction

Liver resection remains a cornerstone in the treatment of liver tumors. Due to advances in surgical techniques, in preoperative management and in understanding liver regeneration, liver resection became safer over years.^{1, 2} Despite this, morbidity and mortality due to posthepatectomy liver failure (PHLF) are still a concern for hepatobiliary surgeons, as PHLF is associated with a mortality rate between 1.2% and 32%.³ Due to enhanced safety of liver resection, more extended hepatic resections with a consequently smaller remnant liver volume are currently performed.

Moreover, liver resections are commonly undertaken in patients with underlying liver parenchymal disease, such as steatosis, steatohepatitis, chemotherapy associated liver injury (CALI), cirrhosis and in obstructive jaundice.^{4, 5} Below a critical liver volume, the remnant liver cannot sustain metabolic, synthetic, and detoxifying functions,⁶ resulting in PHLF. Whereas liver function correlates well with liver volume in uncompromised livers, this correlation in patients with parenchymal disease is less straightforward. Also in extended liver resections, function rather than volume alone, should dictate the minimal future liver remnant.⁷ Therefore, future remnant liver volume (FLRV) is not equivalent to the future remnant liver function (eFLRF).^{8, 9} Determining the maximum extent of resection compatible with a safe postoperative outcome remains unsure with current techniques.^{10, 11}

The ideal tool to measure eFLRF should combine information on remnant liver volume and on actual liver function. It should be applicable to all kind of resections regardless of the type or severity of liver parenchymal disease. It should be easy to perform in daily practice and it should be harmless to the patient. Finally, this tool should have a cut-off value, above which PHLF is not likely to occur. Actually, there is no good and all-inclusive tool to predict PHLF.

In this study, a novel method is presented using ^99mTc-mebrofenin hepatobiliary scintigraphy (HBS) for estimation of the function of the total liver and using MRI for gradual measuring of the total and future remnant liver volumes. Function and volume are combined in one formula. The predictive value of this formula for the occurrence of eFLRF is examined. Impact of eFLRF on PHLF and its mortality in a cohort of patients with normal and underlying liver parenchymal disease was evaluated. Predicting PHLF by measurement of eFLRF could help the hepatobiliary surgeon to decide whether the hepatectomy can be performed safely or whether additional procedures should be undertaken prior to the planned hepatectomy.

Objective of this study

This study aimed at developing an accurate tool to predict PHLF, taking into account both the volume ratio of the future liver remnant liver (FLRV%) and the actual functional capacity of the total liver or total liver function (TLF).

The FLRV% is calculated as the ratio between the future liver remnant volume (FLRV, in ml) and the total functional liver volume (TLV, in ml). Both volumes are measured on MRI.

The TLF was measured by ^99mTc-mebrofenin HBS and corrected to the Body Surface Area (BSA) of the patient. In the formula, the FLRV% will be multiplied by the TLF.

Our hypothesis is that this formula combining future liver remnant volume and total liver function gives a more precise and more individual tool to predict PHLF, compared to future liver remnant volume alone.

Methods

Inclusion criteria

Patients undergoing hepatectomy between October 2008 and December 2010 have been consecutively selected. Indications for liver resection were: Benign Liver Tumor (BLT), Colorectal Liver Metastasis (CRLM), non-Colorectal Liver Metastasis (non-CRLM), intrahepatic cholangiocarcinoma (ICCC), perihilar cholangiocarcinoma (PHCCC) and hepatocellular carcinoma (HCC), all confirmed by pathology of the resection specimen. Patients with cirrhosis and Child Pugh B or C were considered an absolute contraindication for liver resection. Other abnormal liver parenchyma (cirrhosis Child Pugh A, suspected CALI or steatosis/steatohepatitis) was never considered as a contraindication for resection.

According to the policy of the department, only patients with a FLRV% >25% were resected and included in the study, regardless of the type of hepatectomy. Exclusion criteria were: patient under the age of 18 year and pregnancy. The study was approved by the Medical Ethics Committee of the University Hospital of Antwerp. Written informed consent was obtained from each patient before participating in this study.

Preoperative evaluation

In all patients, age and Body Mass Index (BMI, kg/m²) were recorded. BSA was calculated according the Mosteller formula (BSA (m²) = ([Height(cm) × Weight(kg)]/3600)^½).¹² The physical status was estimated by using the American Society of Anesthesiologists score (ASA), and graded 1 to 4: 1 = normal, healthy; 2 = mild systemic disease. 3 = severe systemic disease; 4 = severe systemic disease that is a constant threat to life.

Blood analysis (with normal values at our laboratory between brackets) was performed the day before hepatectomy as potential predictive factor of PHLF: International Normalized Ratio (INR, 1–1.5), serum bilirubin (0.3–1.2 mg/dl), serum creatinine (0.62–1.1 mg/dl), serum glutamic oxaloacetic transaminase (SGOT, <35 U/l), serum glutamic pyruvic transaminase (SGPT, <45 U/l), lactate dehydrogenase (LDH, 84–246 U/l) and gamma glutamyltransferase (GGT, 15–55 U/l).

Underlying liver parenchyma was clinically estimated preoperatively and reported as follows: normal liver parenchyma, cirrhosis and risk for CALI, based on administration of preoperative chemotherapy.

Magnetic resonance imaging (MRI) of the liver was used as an oncologic-diagnostic standard a few weeks before surgery. On the same MRI examination and, an expert radiologist performed liver volumetry on contingent slices. The volume to be resected was defined in close collaboration with the operating hepatobiliary surgeon. It was calculated as follows: the surface area on each slice is multiplied by slice thickness (area × thickness = volume). Afterwards, all the separate slice volumes are added and expressed in ml. TLV, tumor volume (TV) and the volume to be resected (RV) were measured. In case of multiple tumors and/or multiple resections, TV and RV were calculated by respectively adding the different tumor volumes and/or volumes to be resected.

^99mTc-mebrofenin HBS was performed as measure for the total liver function (TLF) and expressed in %/min, according to the protocol of Ekman et al.^{13, 14}

After intravenous injection of 7 mCi (245MBq) of ^99mTc-mebrofenin (Bridatec, GE Healthcare), dynamic and static images are obtained with a large Field of View gamma camera. Dynamic acquisition starts immediately after IV bolus administration of the tracer and lasts for 30 min at 180 frames (10 s per frame, matrix 128 × 128) with the liver and the heart in the field of view. Static acquisition of 1000 kCTs starts 30 min post injection with anterior and posterior view centered on the liver (256 × 256 matrix). Regions of interest (ROI) were drawn manually over the heart and over the liver on the anterior images. From those ROI's two different clearance rates were calculated (Ekman et al.). First, the liver clearance measures the rate of uptake of ^99mTc-mebrofenin by the liver. Secondly, the total blood clearance represents the rate of elimination of ^99mTc-mebrofenin from the blood by the liver, the kidneys and by dilution in the extracellular compartment. To compensate for variations in individual metabolic needs, the clearance was divided by the Body Surface Area (BSA), expressed in m², leading to TLF^Mebro/BSA, expressed in %/min/m². TLF^Mebro/BSA will be mentioned as TLF.

Preoperative calculations

The total liver volume (=the total functional, nontumoral liver) was calculated by subtracting the TV from the TLV^MRI (TLV^MRI − TV), and will be mentioned as TLV.

The total liver volume was also calculated (so called “standardized” TLV) by the formula of Urata based on the BSA (TLV^BSA = −794.41 + 1267.28 × BSA).¹⁵

The volume of the future liver remnant (FLRV) was calculated by subtracting the RV from the TLV (FLRV = TLV − RV).

The ratio of the future liver remnant volume (FLRV%) was calculated by dividing the FLRV by the TLV (FLRV/TLV) and expressed as %.

The estimated future remnant liver function (eFLRF) was calculated by multiplying the future liver remnant volume ratio by the corrected total liver function (eFLRF = FLRV% × TLF).

Intraoperative measurements

The hepatectomy was performed by one of the 3 senior hepatobiliary surgeons, all with large expertise in liver resections. Intraoperative blood loss was measured and reported in ml. The amount (n segments) and the numbering of the resected liver segments, according to Couinaud, were reported.

At the end of the surgical procedure, the volume of the resected liver specimen was measured by means of water displacement method in a water bath, according to the principle of Archimedes (${\text{FLRV}}^{{\text{H}}_2\text{O}}$) and expressed in ml.

Postoperative evaluation

PHLF was diagnosed according to the ISGLS criteria (International Study Group of Liver Surgery),^{3, 16} characterized by an increased INR and hyperbilirubinemia (according to the normal cut-off defined by the local laboratory) on or after postoperative day 5. In patients with preoperatively increased INR or increased serum bilirubin concentration, PHLF is defined by an increasing serum bilirubin concentration and increasing INR on or after postoperative day 5 compared with the values of the previous day. Furthermore, the need for clotting factors such as fresh frozen plasma to maintain normal INR on or after postoperative day 5 in combination with hyperbilirubinemia is considered PHLF. Cutoffs at our laboratory were set: INR >1.5 and serum bilirubin >1.2 mg/dl. PT and serum bilirubin values beyond these limits at or after the 5th postoperative day were considered as an impaired ability of the liver to maintain its synthetic, excretory and detoxifying functions and subsequently diagnosed as PHLF.

The severity of PHLF is graded by its impact on clinical management. Grade A PHLF requires no change of the patient's clinical management and was not considered clinical PHLF in this study In grade B PHLF, clinical management deviates from the regular course but does not require invasive therapy. The need for invasive treatment defines grade C PHLF. In this study, clinical significant PHLF was recorded as PHLF grade B or C.

Perioperative mortality was defined as mortality within 3 months after hepatectomy. The cause of death was recorded as related to PHLF or not.

Statistical analysis

Correlation between TLV and TLV^BSA and between FLRV^MRI and ${\text{FLRV}}^{{\text{H}}_2\text{O}}$ (the final resected volume) was analyzed with Pearson's regression coefficient.

Student t test was used after confirmation of normal distribution of data with Kolmogorov–Smirnov method.

Continuous variables were expressed as mean with standard deviation.

In all statistical analysis, results were considered significant at a P-value <0.05. The predictive value for PHLF of FLRV% and eFLRF was analyzed using univariate logistic regression model.

A cut-off value predictive for PHLF was defined for FLRV% and FRLF by means of ROC analysis. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratios (LR+, LR−) and odds ratio (OR) for this cut-off value of eFLRF was calculated.

Results

A good correlation was shown in Figure 2. Between FLRV^MRI (volume to be resected predicted on MRI) and ${\text{FLRV}}^{{\text{H}}_2\text{O}}$ (final resected volume measured in water bath), with a Pearson's regression coefficient of 0.950. This confirms that the volume of liver intended to be resected (FLRV^MRI) was very similar to the final resected volume (${\text{FLRV}}^{{\text{H}}_2\text{O}}$).

Figure 2.

Correlation between FLRV^MRI and ${\text{FLRV}}^{{\text{H}}_2\text{O}}$, Pearson's regression coefficient = 0.9504

At the contrary, a worse correlation with a Pearson's correlation coefficient of 0.714 was seen between TLV (on MRI) and TLV^BSA, suggesting overestimation of the total liver volume by the BSA based formula. For this reason, TLV^BSA was not retained for further calculations.

Baseline characteristics of the patient cohort are shown in Table 1.

Table 1.

Baseline characteristics of the total patient cohort according to the absence or occurrence of PHLF grade B or C

		Total	No PHLF	PHLF (B/C)	P value
Patients	n	88	76 (86.4%)	12 (13.6%)
Age	years		62 ± 11	62 ± 11	0.798
BMI	kg/m2		26.1 ± 42.5	25.5 ± 4.1	0.701
ASA, Grade
	1	10	8	1
	2	54	48	8
	3	22	18	3
	4	2	2	0
Preoperative blood analysis
INR	1–1.5		1.1 (±0.4)	1.1 (±0.1)	0.804
Serum bilirubin	0.3–1.2 mg/dl		0.5 (±0.85)	0.9 (±0.65)	0.015
Serum creatinine	0.62–1.1 mg/dl		1.1 ± 0.473	0.9 ± 0.301	0.098
SGOT	<35 U/l		61.8 ± 118.8	39.7 ± 27.1	0.755
SGPT	<45 U/l		58.3 ± 94.3	45.7 ± 20.7	0.907
LDH	84–246 U/l		479 ± 357	472 ± 242	0.151
GGT	15–55 U/l		82 ± 62	95 ± 50	0.151
Intraoperative blood loss ml			923 ± 1250	955 ± 1047	0.497
Indication for hepatectomy
BLT	n	10	9	1 (10%)
CRLM	n	48	46	2 (4.2%)
non-CRLM	n	8	7	1 (12.5%)
ICCC	n	5	3	2 (40%)
PHCCC	n	5	2	3 (60%)
HCC	n	12	9	3 (25%)
Perioperative mortality (<3 months)
Total	n	5	0	5 (5.7%)
BLT	n	0	0	0
CRLM	n	0	0	0
non-CRLM	n	0	0	0
ICCC	n	1	0	1 (20%)
PHCCC	n	2	0	2 (40%)
HCC	n	2	0	2 (16.7%)
Liver parenchyma, clinical estimation
Normal	n	33	27	6
Cirrhosis	n	9	6	3
CALI	n	46	43	3

Abbreviations: see text.

Eighty-eight patients were consecutively included. The mean age was 62 years and the gender ratio was 52/36. Preoperative factors such as age, BMI or ASA grade had no impact on the occurrence of PHLF. Of all preoperative blood analysis only preoperative serum bilirubin was significantly different between patients with or without PHLF. Five of the 88 patients (5.7%) died postoperatively and all were in the PHLF group. There were no other reasons for preoperative mortality.

In the 48 patients resected for CRLM, only 2/48 (4.2%) developed PHLF, although 45/48 received preoperative chemotherapy and were potentially at risk for having CALI. Eight resections were performed for non-CRLM (primary neuro-endocrine, breast, renal and laryngeal) and of the 12 resections for HCC, nine were cirrhosis Child A, of whom three developed PHLF. The details of indications for resection and outcome are detailed in Table 1.

The characteristics of the 12 patients with PHLF grade B or C are described in detail in Supplementary Table 2.

Grade B PHLF occurred in six patients without mortality: 3 resections for liver metastasis (2 CRLM and 1 non-CRLM). Grade C PHLF occurred in 6/12, all for primary liver tumors: 1/6 in ICCC, 2/6 in PHCCC, and 3/6 in HCC; all but one died. 5 died of PHLF grade C.

Preoperative obstructive jaundice was diagnosed in 3 patients with PHCCC: all had biliary drainage, resulting in a normal serum bilirubin at the time of the resection in two of these. One patient resected for PHCCC had a residual elevation of preoperative serum bilirubin because of preoperative external drainage of only the left liver. On final pathology, all PHCCC had cholestatic hepatitis with no to mild fibrosis (F0–F2). Parenchyma after obstructive jaundice was initially estimated as normal and these patients had a much higher incidence of PHLF and related mortality.

In hindsight, the negative impact of the cholestasis on postoperative liver function could have been predicted. In Table 3, volumetric, functional and combined volume-functional means with standard deviation for the total cohort are described according to the indications for hepatectomy. Liver parenchyma in BLT was considered as normal with a normal TLF, so BLT was used as control group. TLF in PHCCC was significantly worse compared to TLF in BLT. TLF in PHCCC was similar to TLF in HCC patients, suggesting that liver function at the time of resection was as bad as in cirrhosis. There was a tendency to resect more liver segments in PHCCC than in HCC, resulting in lower but not significant future liver remnant (FLRV%) in PHCCC than in HCC. The combination of a worse TLF with a lower FLRV%, resulted in a significantly lower eFLRF in PHCCC compared to eFLRF in BLT and compared to eFLRF in HCC. This may explain the higher incidence of PHLF and related mortality in PHCCC.

Table 3.

Volumetric, functional and combined volume-functional data according to the indication for hepatectomy.2

The three clinical suspected cirrhotic livers were all confirmed by final pathology (Supplementary Table 2).

Three of the deaths occurred in patient undergoing right or extended right hepatectomies. Two PHLF related deaths occurred in more limited resections in cirrhosis.

Seven of 12 patients with FLRV% >40% developed PHLF and three of these died. PHLF occurred in five of 12 patients. With a FLRV% >50%, and two of these died. Almost all patients (11/12) who developed PHLF grade B or C (5/6 grade B and 6/6) grade C had a eFLRF of less than 2.3%/min/m².

Potential predictive factors for PHLF from univariate and multivariate analysis are detailed in Table 4. In univariate analysis, only serum bilirubin was not predictive for PHLF, but in multivariate analysis, eFLRF was the only independent predictive factor for PHLF.

Table 4.

Preoperative measurements predictive for PHLF: univariate and multivariate analysis

	No PHLF	PHLF (B or C)	P value	Univariate P=	Multivariate P=
Serum Bilirubin (mg/dl)	0.5 (±0.85)	0.9 (±0.65)	0.015	0.992	/
Liver segments resected (n)	2.6 ± 1.4	3.9 ± 1.5	0.013	0.04	0.503
FLRV% (%)	76 (±17.4)	49 (±21)	<0.001	<0.001	0.653
TLF (%/min/m2)	6.2 (±1.6)	5 (±1.8)	0.020	0.027	0.25
eFLRF (%/min/m2)	4.7 (±1.6)	2.2 (±0.7)	<0.001	<0.001	0.001

Abbreviations: PHLF, posthepatectomy liver failure; INR, International Normalized Ratio; FLRV%, ratio of the future liver remnant volume; TLF, total liver function; eFLRF, estimated future liver remnant function.

Receptor-operating-characteristics (ROC) are shown in Figure 1. The cut-off values, representing the optimal combination of sensitivity and specificity, were calculated and set at 40% for FLRV% and at 2.3%/min/m² for eFLRF.

Figure 1.

ROC curve defining cut-off values for FLRV% and eFLRF and their predictive indices for PHLF

Sensitivity, specificity, positive predictive value, likelihood ratio and odds ratio of the eFLRF cut-off value of 2.3%/min/m² are better than those of the FLRV% cut-off value of 40%.

If a FLRV% cut-off value of 40% had been used preoperatively, five of the 12 patients with PHLF and two of the five deceased patients would have been avoided. With the eFLRF cut-off value of 2.3%/min/m², five of six patients with grade B PHLF and all with grade C PHLF including all peroperative mortality could have been prevented.

Discussion

In 88 patients undergoing hepatectomy, the predictability of PHLF by a formula was analyzed. In this formula, the ratio of the future liver remnant volume (in ml) to the total functional liver volume (in ml) was multiplied by the total liver function measured by ^99mTc-mebrofenin HBS, corrected to the patient's BSA. This formula was developed to predict PHLF in patients with various tumor types, with different types of liver parenchymal disorder and with a whole range of liver resections, from limited to extended hepatectomy.

Liver volumetry

In many centers, FLRV% cut off values of 20%, 25% or 30% are used as preoperative selective tool before hepatectomy. In this study, FLRV% >25% was selected as inclusion criterion for liver resection for several reasons. When calculating FLRV% (FLRV% = FLRV/TLV × 100), total liver volume can be estimated by two means: first, the “standardized” TLV can be calculated by a formula based on the Body Surface Area (BSA) (TLV^BSA = −794.41 + 1267.28 × BSA). Secondly, TLV can be estimated on CT or MRI by cross-sectional measuring of liver surface areas on each slice multiplied by slice thickness and by adding all separate slice volumes. Tumor volume has to be calculated separately and subtracted from the TLV measured on CT or MRI. Both TLV^BSA and TLV on CT or MRI are validated procedures and are supposed to be predictive for each other, although overestimation of TLV^BSA was been described¹⁵ and was confirmed in this study. Overestimation of TLV^BSA, leads to underestimation of FLRV% calculated with TLV^BSA in the proposed formula. FLRV% >20% as under limit is applicable when the standardized TLV^BSA is used¹⁷ but when using TLV on CT or MRI, lower limit of 25% seems reasonable. Most centers use currently a FLRV% lower limit of 30%.¹⁷ Using a FLRV% cutoff at 25% is somewhat intermediary between the theoretical lower limit of 20% and the safe “current practice” in many centers of 30%. Moreover, using the same measurement methodology for both FLRV and TLV on the same MRI images and by the same radiologist, similar measurement biases for both FLRV and TLV could be annihilated.

Liver volumetry can be performed either by manual or automated delineation on computed tomography (CT) or MRI of liver volumes (gradual method), or by counting segments to be resected (segmental/non-gradual method). In this study, liver volumetry was performed on MRI instead of on CT, as MRI is the standard oncologic liver imaging and is performed prior to every liver resection at our institution. As far as same dedication to the technical precision in performing volumetry is guaranteed, it is likely that similar results could be obtained with volumetry on CT. A good correlation between volumetry on CT and on MRI has been demonstrated.¹⁸

In this study, the volumes have been estimated precisely by manual delineation of cross sectional areas on MRI and summation of all areas to be resected, resulting in gradual measurement of liver volumes expressed in ml. This gradual delineation method is time-consuming, although faster work out in our recent patients was experienced compared to the initial experience. The method is operator-dependent and close interaction between the radiologist and the hepatobiliary surgeon is mandatory.

Segmental/non-gradual estimation of liver volumes by simply counting the amount of segments to be resected is less time consuming but also less accurate than the gradual delineation method for several reasons. First, there is a significant variation in volume among different liver segments in one individual as well as variation among individuals. Mise et al demonstrated that on average segment 8 accounts for 26% of the TLV, whereas segments 2 and 3 together hardly reach 10% of TLV.¹⁹ Second, there is a trend towards multiple, non-anatomical resections, especially in liver surgery for metastases. Those multiple resections can cumulate in a serious amount of resected liver parenchyma, with a consequent risk for PHLF, but these multiple, non-anatomical resections cannot be described as an amount of resected segments. Third, most algorithms combining liver function with segmental/non-gradual estimation of liver volumes to be resected, were published for HCC in cirrhotic livers. Makuuchi et al. proposed a segmental/non-gradual algorithm to liver resection for HCC: major hepatectomy should be safe with an Indocyanine Green Retention Test at 15 min (ICGR₁₅) <10%; when ICGR₁₅ <20%, only 2 segments could be resected safely, when ICGR₁₅ >30% only one segment could be resected and enucleation is still possible when ICGR₁₅ is >40%.²⁰ Similarly, Poon proposed a segmental/non-gradual approach for resections in Child A cirrhosis, combining ICGR with the number of segments that can be resected safely.²¹ In an expert consensus report concerning liver resection for HCC, Vauthey proposed also a segmental/non-gradual approach: the future liver remnant volume should be at least 20% in normal liver, 30% in an injured liver and 40% in a well-compensated liver fibrosis or cirrhosis.²² Although these segmental/non-gradual approaches have proven their value in avoiding PHLF in cirrhosis, they have not been used in liver resection in non-cirrhotic liver parenchyma disorder and cannot be extrapolated to other causes of impaired liver function. For all those reasons, the gradual delineation method of liver volumes seems to be more adapted to the current type of hepatic resections than the segmental/non-gradual approach.

In this study, the predicted liver volume on MRI (FLRV^MRI) correlates well with the final resected volume measured according to the water displacement principle in a water bad (${\text{FLRV}}^{{\text{H}}_2\text{O}}$). Several biases could occur: first, the surgeon could choose or be forced to resect more or less non tumoral liver parenchyma than predicted on preoperative imaging. Unexpected findings during the surgery, such as more or bigger nodules or more difficult localization close to vascular liver structures or even vanished lesions can alter the intraoperative strategy from the preoperative assessment. Second, volumetry on MRI is performed on a perfused liver, while the volumetry in the water bath is performed on an exsanguinated resection specimen, resulting in a somewhat lower water bath volumetry. Despite these possible biases, the correlation between the FLRV^MRI and ${\text{FLRV}}^{{\text{H}}_2\text{O}}$ was excellent. It can therefore be concluded that gradual volumetry on MRI provides an adequate estimation of the real resected liver volumes. This accuracy presumably contributes to the accuracy of our predictive formula.

Liver function evaluation

Estimation of the preoperative liver function is important as a majority of resections are performed in livers with reduced function. Parenchymal disorders can reduce liver function and regeneration capacity and are a particular risk for developing PHLF. Major partial hepatectomy in cirrhosis or obstructive jaundice correlates well with increased morbidity and mortality rates.^{4, 23, 24, 25, 26} In cirrhosis, PHLF after liver resection for HCC has a mortality rate of 25%–100%.^{27, 28, 29, 30, 31, 32, 33}

In this study, PHLF was seen more often after resection for HCC, ICCC and PHCCC. This can be expected after resection for HCC in cirrhosis which was confirmed on final pathology. The higher incidence of PHLF after resection for PHCCC can be explained by ongoing parenchymal disorder³⁴ even after preoperative biliary drainage and normalization of serum bilirubin. In severe steatosis, a higher risk of PHLF has been described.^{4, 35, 36, 37, 38} In liver resection for CRLM, 50% of postoperative mortality can be attributed to PHLF.^{39, 40, 41} Most chemotherapeutic agents can cause CALI⁴²: 5-Fluorouracil can cause steatosis.^{23, 43, 44, 45} Irinotecan is associated with the development of steatohepatitis^{46, 47} whereas Oxaliplatin causes sinusoidal obstruction.⁴⁸

Estimation of liver function by means of preoperative blood analysis^{7, 49, 50, 51} or mainly on blood analysis based clinical scoring systems⁵² are not considered adequate predictive tools for PHLF. In this study, INR neither serum bilirubin were predictive for PHLF.

In many centers, ICGR₁₅ is used as estimation tool for preoperative liver function. Because a good correlation between ICGR₁₅ and ^99mTc-mebrofenin HBS has been demonstrated, ^99mTc-mebrofenin HBS has been validated as a tool to measure total liver function and functional remnant liver after liver resection.^{14 99m}Tc-mebrofenin HBS can be used in normal livers, as well as in livers with parenchymal disorder, regardless the type and the grade of parenchymal liver disorder.^{53 99m}Tc-mebrofenin HBS was used for the estimation of eFLRF by De Graaf et al. by delineating areas to be resected as “regions of interest” on the images of a ^99mTc-mebrofenin HBS. They calculated a cutoff of 2.69%/min/m² for eFLRF by ROC analysis. A higher risk for PHLF was shown below this cutoff. A better sensitivity and specificity was demonstrated for eFLRF compared to volumetry of the future remnant.⁹ However, on ^99mTc-mebrofenin HBS, only rudimental delineation of liver volumes can be performed. Delineation is not possible in more complex or less anatomic resections, like in atypical resections and multiple segmentectomies, limiting this technique to the use in planned hemihepatectomy. Recently, the same group showed that ^99mTc-mebrofenin Single Photon Emission Computed Tomography (SPECT) could provide valuable visual and quantitative information on total and segmental liver function. However, volumetric measurement on SPECT has similar limitations regarding delineation in more complex resections.⁵⁴

Besides these arguments, ^99mTc-mebrofenin HBS was chosen as evaluation method for total liver function in this study because of high availability and the low costs of scintigraphy at our institution.

Evaluation by combining liver volumetry and liver function

As ^99mTc-mebrofenin HBS measures the total liver function regardless the type or grade of underlying parenchymal disorder as far as no biliary obstruction is still present, other evaluations of parenchymal disorder seems useless when estimating the future liver function. However no precise volumetry can be performed on ^99mTc-mebrofenin HBS. As FLRV% does not provide information about the future liver function and as, combination of both volume and function in one formula could give a stronger prediction of the future liver function after resection (eFLRF). eFLRF is a better tool in avoiding PHLF.^{7, 20, 21, 22}

In this study, although both FLRV% and eFLRF were lower in the PHLF group, only eFLRF seems to predict accurately PHLF in multivariate analysis. By using the cut off value of 40% for FLRV%, only few PHLF would have been avoided. At the contrary, using the cut off value of 2.3%/min/m² for eFLRF would have prevented all mortalities related to PHLF and almost all PHLF incidences. The superiority of eFLRF over FLRV% as cut off to prevent PHLF is supported by a higher positive predictive value, positive likelihood ratio and Odds ratio.

Conclusion

By using a formula combining liver remnant volumetry on MRI and liver function measurement on ^99mTc-mebrofenin HBS, a more accurate tool to predict PHLF than the current available methods, has been developed. Preoperative evaluation of eFLRF should become the standard for the hepatobiliary surgeon, especially when planning resections in diseased liver parenchyma.

Conflicts of interest

None declared.

Appendix A. Supplementary material

The following is the supplementary material related to this article: