SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES

An accurate clinical assessment of hepatic steatosis before transplantation is critical for successful outcomes after liver transplantation, especially if a pathologist is not available at the time of procurement. This prospective study investigated the surgeon's accuracy in predicting hepatic steatosis and organ quality in 201 adult donor livers. A steatosis assessment by a blinded expert pathologist served as the reference gold standard. The surgeon's steatosis estimate correlated more strongly with large-droplet macrovesicular steatosis [ld-MaS; nonparametric Spearman correlation coefficient (rS) = 0.504] versus small-droplet macrovesicular steatosis (sd-MaS; rS = 0.398). True microvesicular steatosis was present in only 2 donors (1%). Liver texture criteria (yellowness, absence of scratch marks, and round edges) were mainly associated with ld-MaS (variance = 0.619) and were less associated with sd-MaS (variance = 0.264). The prediction of ≥30% ld-MaS versus <30% ld-MaS was excellent when liver texture criteria were used (accuracy = 86.2%), but it was less accurate when the surgeon's direct estimation of the steatosis percentage was used (accuracy = 75.5%). The surgeon's quality grading correlated with the degree of ld-MaS and the surgeon's steatosis estimate as well as the incidence of poor initial function and primary nonfunction. In conclusion, the precise estimation of steatosis remains challenging even in experienced hands. Liver texture characteristics are more helpful in identifying macrosteatotic organs than the surgeon's actual perception of steatosis. These findings are especially important when histological assessment is not available at the donor's hospital. Liver Transpl 19:437–449, 2013. © 2013 AASLD.

Abbreviations
ALT

alanine aminotransferase

ANOVA

analysis of variance

AST

aspartate aminotransferase

CART

classification and regression tree

H&E

hematoxylin and eosin

ICC

interclass correlation coefficient

ld-MaS

large-droplet macrovesicular steatosis

MaS

macrovesicular steatosis

MELD

Model for End-Stage Liver Disease

MiS

microvesicular steatosis

OLT

orthotopic liver transplantation

rS

nonparametric Spearman correlation coefficient

sd-MaS

small-droplet macrovesicular steatosis

UCLA

University of California Los Angeles

UW

University of Wisconsin.

The growing discrepancy between the demand for donor organs and their availability presents one of the greatest hurdles to liver transplantation today.[1] Therefore, marginal grafts or extended criteria donor organs have to be considered to overcome this shortcoming and save the lives of patients on the waiting list. In addition to donor age and prolonged donor hospital stays, hepatic steatosis—especially macrovesicular steatosis (MaS)—is one of the most important criteria defining extended criteria donor organs. Mild MaS (<30%) is considered a transplantable condition, whereas the utilization of donor livers with moderate MaS (30%-60%) remains challenging.[2] A recently published national analysis of the Scientific Registry of Transplant Recipients demonstrated that MaS > 30% was an independent predictor of reduced 1-year graft survival.[3]

Although steatosis is a risk factor for reduced outcomes after orthotopic liver transplantation (OLT),[3-5] the utilization of steatotic donor livers significantly contributes to the expansion of the donor pool. Furthermore, the epidemic of diabetes and obesity in the United States during the past 2 decades has resulted in an increased number of donors with steatosis.[6, 7] Thus, the accurate assessment of hepatic steatosis is critical for the selection of these high-risk organs. The microscopic evaluation of a liver biopsy sample by an experienced pathologist is still the gold standard for steatosis assessment. Because a pathologist is not always available at the time of procurement, many centers, including ours, rely mainly on the assessment of steatosis and organ quality by the procurement surgeon.

Therefore, the objective of this study was to investigate how accurately a surgeon could predict hepatic steatosis and organ quality. The qualitative and quantitative assessment of steatosis by a pathologist served as the reference gold standard. We further investigated whether the surgeon's gross steatosis estimation correlated with different types of steatosis seen on biopsy. In addition, we studied liver texture criteria as potential predictors of pathological steatosis.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES

Study Design

From November 2009 to March 2011, all donation after brain death donors from whom a liver was procured by the University of California Los Angeles (UCLA) liver transplant team were assessed for study eligibility. Eligible donors were prospectively enrolled, and donor variables were prospectively recorded from the donor summary reports. Each enrolled donor liver was assessed by the senior UCLA procurement surgeon at the donor hospital according to standard quality criteria (described later). After a macroscopic assessment, 2 liver core needle biopsy samples were taken in a standardized manner (Tru-Cut, 14G × 15 cm, Cardinal Health) from the left lobe (segment 3) and the right lobe (segment 5). Each biopsy was separately fixed in 10% formalin until processing for permanent sections and staining. The study had a double-blind design: the assessing procurement surgeon had no histological information, and the pathologist was unaware of any clinical donor data. The study was approved by the UCLA institutional review board (UCLA-IRB-07-11-096) and by the institutional review board of OneLegacy, the local organ procurement organization (RS-10-08-1).

Inclusion and Exclusion Criteria

Only whole donation after brain death organs that were procured by the UCLA team were included in the study. Further inclusion criteria were a donor age ≥ 16 years and a recipient age ≥ 18 years. Signed informed consent from the recipient and, when a liver was declined for transplantation, signed research consent from the donor were mandatory for enrollment in the study. Any knowledge of the histology of preprocurement biopsy samples was an exclusion criterion. Living related, multivisceral, pediatric, split, and donation after cardiac death livers were not included.

Liver Procurement

All liver procurement procedures were performed with a standardized technique.[8] We exclusively used University of Wisconsin (UW) solution for procurement. Immediately after the cross-clamping and transection of the suprahepatic vena cava, dual perfusion with ice-cold UW solution was started via the abdominal aorta and inferior mesenteric vein. Slush ice was placed around the liver, including the subdiaphragmatic space and hilum, and anterior and posterior to the liver. Approximately 2000 to 3000 mL of UW solution was infused via the aortic cannula, and 1000 mL was infused via the inferior mesenteric vein.

Macroscopic Assessment of Steatosis and Donor Organ Quality

If both the donor and the recipient were eligible for the study, the senior surgeon of the procurement team assessed the liver graft quality after organ flushing at the donor hospital. If intraoperative frozen-section biopsy samples were taken for determining acceptance or rejection during procurement, the macroscopic evaluation was always performed before the announcement of the biopsy results in order to exclude any influence of histological donor organ information on the judgment of the procurement surgeon. Each organ was evaluated in a standardized fashion according to the following parenchymal texture criteria: degree of yellowness (normal, mild, moderate, or severe), degree of firmness (too soft, normal, mild, moderate, or severe), round liver edges (present or absent), and scratch marks (present or absent). We included the criterion scratch marks in the assessment on the basis of the longstanding observation by our senior procurement surgeon (H.Y.) that livers displaying faint scratch lines on the liver capsule are frequently associated with favorable outcomes. Furthermore, the degree of steatosis as a percentage was estimated. Finally, the overall liver quality was ranked as excellent, good, average, poor, or unacceptable. Livers that were ranked as unacceptable were always declined for OLT.

Microscopic Assessment of Steatosis

Hepatic steatosis was assessed with prospectively collected biopsy samples, which were taken from segments 3 and 5 after organ flushing. All biopsy samples were fixed in 10% formalin and processed for permanent hematoxylin and eosin (H&E)–stained sections. The evaluation was performed by a board-certified expert pathologist (C.R.L.) with extensive experience in hepatobiliary and liver transplant pathology. Each biopsy sample was linked to a United Network for Organ Sharing number, which ensured the blinding of the pathologist. Frozen-section biopsy samples, which were used for decision making at the time of procurement, and postreperfusion biopsy samples were not used for this study, and the pathologist was unaware of those findings. MaS was separately assessed for small-droplet macrovesicular steatosis (sd-MaS) and large-droplet macrovesicular steatosis (ld-MaS).[9, 10] We used the term ld-MaS when a single fat vacuole was larger than the half of the cell displacing the nucleus to the periphery of the cell (Fig. 1B). In contrast, fat vacuoles that were smaller than half of the cell and did not displace the nucleolus were considered sd-MaS (Fig. 1C). Microvesicular steatosis (MiS) was defined as innumerable tiny lipid vesicles that were diffusely distributed and caused a foamy appearance of the cytoplasm. The extent of steatosis was quantified by an estimation of the percentage of liver parenchyma occupied by small- and large-droplet steatotic vacuoles. The assessment was performed for left lobe (segment 3) and right lobe (segment 5) biopsy samples, and the total percentages of sd-MaS and ld-MaS were calculated as averages of both biopsy samples. Furthermore, we determined the interobserver variability for sd-MaS and ld-MaS between the reference pathologist (C.R.L.) and another expert hepatobiliary pathologist (B.V.N.) in a randomly selected study sample of 51 cases.

image

Figure 1. Different types of steatosis in donor livers: (A) no steatosis, (B) predominantly ld-MaS (70%), (C) predominantly sd-MaS (50%), and (D) mixed sd-MaS (40%) and ld-MaS (30%; H&E stain, ×400).

Download figure to PowerPoint

Histological Characterization of Scratch Marks

In order to characterize microscopic scratch marks, cube-shaped wedge biopsy samples (approximately 1.5 cm × 1.5 cm) were taken from areas with and without surface scratch marks of livers that were declined by all centers for transplantation (Fig. 2A). For this purpose, we used livers with good or acceptable parenchymal quality that were declined for nonparenchymal reasons. Tissue blocks were fixed and processed according to the methods described previously for biopsy samples. On paraffin-embedded blocks of tissue with scratch marks, serial sections were performed in the axis perpendicular to the grossly identified scratch marks. H&E-stained sections of tissue with and without scratch marks were microscopically compared.

image

Figure 2. Characterization of capsular scratch marks. (A) The visual appearance of faint surface scratch marks is illustrated. Liver wedge biopsy samples with and without surface scratch marks were processed for H&E histology. (B,C) A characteristic example of the histological appearance of scratch marks (arrow) is presented. (C) The scratch marks were caused by focal ruptures of Glisson's capsule. (D) Areas without scratch marks displayed an intact Glisson's capsule. The original magnifications were (A) ×8, (B) ×100, (C) ×400, and (D) ×100.

Download figure to PowerPoint

Initial Allograft Function

Initial allograft function was classified as normal function, poor function, or nonfunction. Poor initial function was defined as an aspartate aminotransferase (AST) or alanine aminotransferase (ALT) level > 2500 U/L during the first 3 days and a total bilirubin level > 10 mg/dL on day 7 after OLT. Nonfunction was defined as primary and delayed nonfunction resulting in retransplantation or death from graft dysfunction within 30 days of OLT. Normal initial function was defined as no poor function or nonfunction.

Statistical Methods

The nonparametric Spearman correlation coefficient (rS) was computed to assess associations between the different measures of the steatosis percentage, other continuous variables, and ordinal measures. The κ statistic was computed for assessing beyond-chance agreement between categorical variables. Robust multiway analysis of variance (ANOVA) methods were used to compare median steatosis percentage scores by combinations of yellowness, firmness, scratch signs and/or round edges, and firmness. Nonparametric and robust methods were used for steatosis percentage scores because these scores did not follow a normal distribution. The R2 statistic was computed for the ANOVA model as a measure of the variation for which the model accounted. The interobserver agreement between 2 pathologists was assessed with the interclass correlation coefficient (ICC). ICC values range from 0 to 1, and an ICC ≥ 0.7 reflects acceptable agreement. A multivariate classification tree [classification and regression tree (CART)] model was used to predict the ld-MaS category (normal, 0%-10%; mild, 11%-29%; moderate, 30%-60%; and severe, >60%) from yellowness, firmness, scratch-sign, and round-edge candidates.[11] Another classification tree model was used to predict a binary ld-MaS outcome (<30% versus ≥30%). The average accuracy was computed as a measure of the tree model's predictive performance and was defined as the average of the percentages correctly classified in each category.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES

Study Population

There were 315 procured donor livers, but 67 donors were excluded from evaluation because they did not meet the inclusion criteria (n = 61), recipient consent was refused (n = 1), or there was knowledge of the donor's liver histology before the evaluation (n = 5). One hundred seventy-one of the 247 evaluated donors were accepted for transplantation, and 76 were declined because of inappropriate quality. Thirty of the 76 declined livers were included in the study, whereas the other 46 donor livers were excluded because no donor research consent was given and/or the backup center accepted the donor organ. Finally, the donor study population was composed of 171 accepted donor livers and 30 declined donor livers. The donor and recipient characteristics are shown in Table 1. The most common reason for declining a liver was poor parenchymal quality (70%) due to cirrhosis (n = 5) or severe steatosis (n = 16). Only 7 of the 23 donor livers with moderate or severe ld-MaS were transplanted. All cirrhotic livers were excluded from the analyses.

Table 1. Study Population Characteristics
CharacteristicDonors (n = 201)
  1. NOTE: Poor initial function was defined as an AST or ALT level > 2500 U/L within 3 days and a total bilirubin level < 10 mg/dL on day 7 after OLT. Primary nonfunction was defined as retransplantation or death within 30 days of OLT. Normal initial function was defined as no poor initial function or primary nonfunction.

  2. a

    The data are presented as medians and ranges.

  3. b

    The data are presented as means and standard deviations.

  4. c

    Laboratory MELD score without exception points.

Age (years)a44 (17-79)
Female sex [n (%)]82 (41)
Body mass index (kg/m2)b27.7 ± 6.1
Hospital stay (days)a4.0 (0-32)
Accepted for OLT (n)171
Declined for OLT (n)30
Cirrhosis5
Steatosis16
Cold ischemia time1
Tumor2
Recipient too sick2
OLT aborted3
CharacteristicRecipients (n = 171)
Age (years)a58 (16-78)
Female sex [n (%)]57 (33)
MELD scoreac37 (6-48)
Peak AST (U/L)a761 (108-10,940)
Primary OLT/re-OLT (n/n)156/15
Normal initial function [n (%)]124 (72.5)
Poor initial function [n (%)]36 (21.1)
Primary nonfunction [n (%)]8 (4.7)

Assessment of Steatosis by the Reference Expert Pathologist

The majority of the donor liver biopsy samples (65%) were assessed with minimal sd-MaS and ld-MaS (0%-10%), and 24% of the cases had sd-MaS and/or ld-MaS ≥ 30%. Figure 3A illustrates the heterogeneous composition of sd-MaS and ld-MaS in donor liver biopsy samples, which ranged from mixed steatosis (Fig. 1D) to predominantly sd-MaS (Fig. 1C) or ld-MaS (Fig. 1B). There was only a moderate correlation (rS = 0.537, P < 0.001) between sd-MaS and ld-MaS. True MiS was found in only 2 donor livers (1%).

image

Figure 3. Nonparametric Spearman correlations for hepatic steatosis: (A) the correlation of sd-MaS with ld-MaS as assessed by an expert pathologist and correlations of (B) sd-MaS, (C) ld-MaS, and (D) total MaS as assessed by an expert pathologist with a surgeon's steatosis estimates. All correlations were statistically significant at P < 0.001.

Download figure to PowerPoint

Interobserver Variability of Steatosis Between Expert Pathologists

An assessment of the interobserver variability for sd-MaS and ld-MaS between 2 expert pathologists was performed with a representative study sample of 51 cases. The sample population (n = 51) was comparable to the entire study population (n = 196), and this was reflected by similar medians and ranges for sd-MaS [10% (0%-80%) versus 4.0% (0%-80%)] and ld-MaS [3.8% (0%-70%) versus 4.0% (0%-90%)]. The median ld-MaS values were 4% and 5% according to pathologists 1 and 2, respectively, and the corresponding median sd-MaS values were 10% and 10%. The 5th to 95th percentiles of the differences in steatosis percent units between the 2 pathologists were −8.5 to 5 for ld-MaS and −25.5 to 10 for sd-MaS. The median difference was zero for both ld-MaS and sd-MaS. There was excellent agreement for ld-MaS (ICC = 0.962) but less agreement for sd-MaS (ICC = 0.775; Fig. 4).

image

Figure 4. Interobserver variability of steatosis between 2 expert pathologists: an assessment of (A) sd-MaS and (B) ld-MaS by 2 expert pathologists in a representative study sample of 51 cases. The ICC values showed excellent agreement for ld-MaS and lesser but acceptable agreement for sd-MaS.

Download figure to PowerPoint

Steatosis Estimation by Surgeons at 2 Different Experience Levels

The liver assessment was performed in 83 cases by the senior faculty surgeon and in 113 cases by a second-year fellow who had experience with at least 80 liver procurements. The Pearson and Spearman correlations did not show any significant differences between the 2 levels of experience in evaluating sd-MaS and ld-MaS (Table 2).

Table 2. Steatosis Correlations Between an Expert Pathologist and Surgeons at 2 Different Experience Levels
Expert PathologistSteatosis According to SurgeonsP Valuea
Level I (n = 113)Level II (n = 83)
  1. NOTE: All correlations were statistically significant at P < 0.001. For experience level I, evaluations were performed by second-year fellows who had experience with at least 80 procurement procedures. For experience level II, evaluations were performed by a senior faculty surgeon.

  2. a

    Group I versus group II (Z test based on a Fisher transformation for independent correlations).

Pearson correlation   
sd-MaS0.3450.3880.74
ld-MaS0.6890.7870.14
Spearman correlation   
sd-MaS0.4030.3620.74
ld-MaS0.5130.4700.70

Histological Characterization of Scratch Marks

A histological assessment of wedge specimens with (Fig. 2A) and without surface scratch marks showed that scratched tissue had focal disruption of the capsular connective tissue, which was characterized by the separation of collagen fibrils (Fig. 2B,C). Tissue areas without scratch marks revealed an intact capsule that adhered to the underlying hepatocyte cell layer (Fig. 2D). Tissue with scratch marks had no associated acute or chronic inflammation, and the underlying parenchyma was identical to the surrounding parenchyma and the parenchyma in the tissue sample from areas without scratch marks.

Relationship Between Liver Texture and Steatosis According to an Expert Pathologist

Scratch marks and sharp edges were observed in 68% and 73% of the donor livers, respectively. In terms of yellowness, the majority were classified as normal (n = 128), and the rest were classified as mild (n = 45), moderate (n = 15), or severe (n = 8). Donor organs without scratch marks or round edges had significantly higher mean percentages of ld-MaS versus donor organs with scratch marks (17.3% versus 5.1%, P < 0.001) or sharp edges (18.1% versus 5.6%, P < 0.001). Although the expression of scratch marks decreased, round edges were more frequently present with an increasing degree of ld-MaS (Fig. 5A). Scratch marks were absent in all livers with severe ld-MaS. This constellation was not observed for sd-MaS (Fig. 5B). The mean percentage of ld-MaS increased as yellowness increased from normal to severe (Fig. 5D). This relationship was not observed for sd-MaS, for which the average percentages were similar for mild, moderate, and severe yellowness. When we controlled for yellowness, ld-MaS was lowest for those with scratch marks and no round edges and was highest for those with no scratch marks or round edges (2.8% versus 23.1%, P < 0.001). Although the orders of sd-MaS and ld-MaS were the same for yellowness, the order of sd-MaS was not the same as the order of ld-MaS for the category of scratch marks and round edges (Table 3). Moreover, the variance for which these factors accounted was only 0.264 for sd-MaS and 0.619 for ld-MaS. These data show that sd-MaS was less influenced by yellowness, scratch marks, and round edges than ld-MaS. Although firmness was also considered, we did not retain firmness in these models because 180 of the 196 livers had a normal firmness, and firmness was not statistically significant (P = 0.10 for ld-MaS and P = 0.09 for sd-MaS).

Table 3. Relationships of Characteristic Predictors of ld-MaS and sd-MaS
Degree of YellownessLd-MaS (%)
Total*No Scratches or Round EdgesNo Scratches and Round EdgesScratches and Round EdgesScratches and No Round Edges
Normal2.12.5 (0-25)3.5 (0-4)2.5 (0-25)0.5 (0-30)
Mild6.87.5 (0-37.5)6 (0-30)7.5 (0-30)6 (0-40)
Moderate20.520a30 (0-70)30 (9-40)2 (0-4)
Severe62.562.5 (35-90)50 (7.5-80)No observationNo observation
Total*16.423.122.413.32.8
Degree of YellownessSd-MaS (%)
Total*No Scratches or Round EdgesNo Scratches and Round EdgesScratches and Round EdgesScratches and No Round Edges
  1. NOTE: The data are presented as medians and ranges or as means of medians (*). The Spearman correlations are 0.514 for ld-MaS and 0.421 for sd-MaS, the Pearson correlations are 0.787 for ld-MaS and 0.514 for sd-MaS, and the variance R2 values are 0.619 for ld-MaS and 0.264 for sd-MaS. The ANOVA table (robust) for ld-MaS is as follows: χ2 for yellowness = 969 (P < 0.001), χ2 for scratches/round edges = 0.042 (P = 0.94), and χ2 for yellowness × scratches/round edges = 206 (P < 0.001). The ANOVA table (robust) for sd-MaS is as follows: χ2 for yellowness = 2.90 (P = 0.41), χ2 for scratches/round edges = 19.0 (P = 0.003), and χ2 for yellowness × scratches/round edges = 45.2 (P < 0.001).

  2. a

    There was 1 observation.

Normal3.83 (0-4)4 (0-50)4 (0-35)4 (0-20)
Mild20.615 (0-45)22.5 (0-75)25 (0-60)20 (0-80)
Moderate19.415 (0-80)17.5 (10-25)25 (9-50)20a
Severe27.315 (10-50)No observationNo observation39.5 (4-75)
Total*16.412.014.718.020.9
image

Figure 5. Liver texture versus different types of steatosis. Expression of round edges and scratch marks according to different degrees of (A) ld-MaS and (B) sd-MaS. The degree of sd-MaS and ld-MaS was graded as normal (0%-10%), mild (11%-29%), moderate (30%-60%), or severe (>60%). (C) Mean percentages of steatosis for the 4 different combinations of round edges (R indicates round edges; S indicates scratch marks; and + and − indicate presence and absence, respectively). (D) Mean steatosis percentages for the different degrees of yellowness.

Download figure to PowerPoint

Prediction Models of Steatosis

The surgeon's steatosis estimate correlated moderately with ld-MaS (rS = 0.504) and total MaS (rS = 0.527) and was weak for sd-MaS (rS = 0.398; Fig. 3). We used the multivariate CART model to predict the ld-MaS category with yellowness, scratch marks, and round edges as potential simultaneous predictors. The tree for binary ld-MaS (<30% versus ≥30%) using yellowness as the most important predictor had an accuracy of 86.2% (Table 4 and Fig. 6). The more complex CART analysis (not shown) using 4 categories of ld-MaS (normal, mild, moderate, and severe) had correct prediction rates of 68.5% for normal ld-MaS, 57.1% for mild ld-MaS, 33.3% for moderate ld-MaS, and 100% for severe ld-MaS (Table 5). The low average accuracy of 64.7% is somewhat misleading because this model misclassified 8 of the true moderate livers as mild and classified some of the normal livers as mild and some of the mild livers as normal.

Table 4. Steatosis Predictions for Surgeon's Steatosis Estimates and CART Analysis with MaS Cutoffs of <30% and ≥30%
Steatosis PredictionActual ld-MaS (Expert Pathologist)Total
<30%≥30%
  1. NOTE: For the CART analysis, see Fig. 6.

  2. a

    rS = 0.637 ± 0.089 (P < 0.001) and κ = 0.614 ± 0.097 (P < 0.001).

  3. b

    Average accuracy.

  4. c

    rS = 0.555 ± 0.070 (P < 0.001) and κ = 0.513 ± 0.077 (P < 0.001).

Surgeon's estimatea   
<30% (n)17111182
≥30% (n)21214
Total (n)17323196
Correct prediction (%)98.852.275.5b
CART analysisc   
<30% ld-MaS (n)1483151
≥30% ld-MaS (n)252045
Total (n)17323196
Correct prediction (%)85.587.086.2b
Table 5. Steatosis Predictions for Surgeon's Steatosis Estimates and CART Analysis for 4 Different Degrees of MaS
Steatosis PredictionActual ld-MaS (Expert Pathologist)Total
NormalMildModerateSevere
  1. NOTE: The degree of ld-MaS was defined as normal (0%-10%), mild (11%-29%), moderate (30%-60%), or severe (>60%).

  2. a

    rS = 0.609 ± 0.063 (P < 0.001) and κ = 0.669 ± 0.056 (P < 0.01).

  3. b

    Average accuracy.

  4. c

    rS = 0.557 ± 0.059 (P < 0.01) and κ = 0.609 ± 0.066 (P < 0.01).

Surgeon's estimatea     
Normal (n)131330137
Mild (n)26118045
Moderate (n)204511
Severe (n)00303
Total (n)15914185196
Correct prediction (%)82.478.622.20.045.8b
CART analysisc     
Normal (n)109310113
Mild (n)4084052
Moderate (n)736016
Severe (n)307515
Total (n)15914185196
Correct prediction (%)68.657.133.3100.064.7b
image

Figure 6. Binary CART model. Yellowness, round edges, and scratch marks were used as simultaneous predictors of ld-MaS (<30% versus ≥30%). This model creates 6 predictive groups (terminal nodes) and has an accuracy of 86.2%.

Download figure to PowerPoint

The same categories of ld-MaS were used to assess the accuracy of the surgeon's steatosis estimate. The prediction of <30% ld-MaS by the surgeon was correct 98.8% of the time, but the rate was significantly lower for predicting ld-MaS ≥ 30% (52.2%; Table 4). Predictions of ld-MaS for the 4 categories (normal, mild, moderate, and severe) by the surgeon had an average accuracy of 45.8% (Table 5). Although the predictions of no steatosis and mild steatosis were good (82.4% and 78.6%), the surgeon's classification of moderate and severe MaS was poor (22.2% and 0.0%). All 5 livers with severe ld-MaS were classified as having moderate steatosis. Livers with moderate ld-MaS were 61% underestimated and 17% overestimated by the surgeon. Overall, the CART analysis had better prediction accuracy than the surgeon's steatosis estimate (Tables 4 and 5).

Associations With the Surgeon's Liver Quality Assessment

The surgeon's steatosis estimate had the strongest correlation with liver quality (Table 6). The surgeon's quality assessment had a higher correlation with ld-MaS versus total steatosis or sd-MaS. Although many other factors, including recipient variables, had a significant impact on initial allograft function, there was a statistically significant association between the surgeon's quality grading and initial allograft function (Table 6). Eighty percent of the 67 livers classified as excellent by the surgeon had normal initial function, whereas 74% of those livers classified as good and only 53% of those livers classified as average had normal function; an inverse relationship was observed for the poor initial function and primary nonfunction categories (Kruskal-Wallis P = 0.035). The correlation of quality with the peak serum AST level was statistically significant but weak (rS = 0.192, P = 0.012).

Table 6. Organ Quality Grading by the Surgeon With Respect to Steatosis, Graft Injury, and Initial Allograft Function
Organ Quality GradingOLT/No OLT (n/n)Median MELD Scoreasd-MaS (%)bld-MaS (%)bsd-MaS + ld-MaS (%)bSurgeon's Steatosis Estimate (%)bPeak AST (U/L)cNormal Initial Function [n (%)]Poor Initial Function [n (%)]Primary Nonfunction [n (%)]
  1. a

    Median laboratory MELD score without exception points at the time of OLT.

  2. b

    The data are presented as medians and ranges.

  3. c

    The data are presented as means and standard errors of the mean.

  4. d

    There was 1 observation.

  5. e

    All livers that were ranked as unacceptable were declined for OLT for various reasons, including poor quality, recipient too sick for OLT, and vascular anatomy.

Excellent67/1384 (0-50)0 (0-30)5 (0-70)8.3 (0-15)1089 ± 18252 (78)12 (18)1 (1)
Good84/4374 (0-80)4 (0-40)8 (0-100)10 (4-30)1354 ± 18562 (74)17 (20)4 (5)
Average19/43220 (0-60)5 (0-60)25 (0-70)20 (5-40)1857 ± 52710 (53)6 (32)3 (16)
Poor1/428d9 (0-75)35 (9-90)70 (18-100)30 (20-50)6222d0 (0)1 (100)0 (0)
Unacceptablee0/1220 (0-80)25 (0-80)68 (0-90)45 (15-90)
Spearman correlation−0.206 (P = 0.004)−0.436 (P < 0.001)−0.368 (P < 0.001)−0.607 (P < 0.001)0.196 (P = 0.012)

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES

This prospective study compared systematic assessments of steatosis in donor livers by the procuring transplant surgeon and an expert pathologist. The evaluation of steatosis at the time of procurement was mainly influenced by the degree of ld-MaS rather than the degree of sd-MaS. Liver texture criteria (yellowness, round edges, and an absence of capsular scratch marks) were associated with underlying ld-MaS and were more helpful in identifying macrosteatotic organs than the actual steatosis estimation by the surgeon.

The nature of a prospective and double-blind study design provides the highest level of evidence for the objective of the study. We always took 2 biopsy samples from 2 separate but same anatomic sites in each donor liver because it has been shown that biopsy samples from 2 different sites are sufficient for evaluating donor livers.[12] Furthermore, the liver procurement technique, the surgical donor organ evaluation, the biopsy processing, and the staining techniques were highly standardized. On the other hand, the present study is associated with a few shortcomings. Forty-six of the 76 donor livers that were declined at the donor hospital were not included in the study because no donor research consent was given and/or the backup center accepted the organ (this precluded biopsy). Another limitation is related to the nature of a single-center study, the results of which may reflect the practice of a single institution. Furthermore, we did not use frozen sections or specific fat staining because these methods require immediate freezing of liver biopsy samples, which is logistically difficult at donor hospitals. Although the use of frozen sections is the method of choice for gaining histological information at the time of procurement, this technique is associated with some shortcomings such as Venetian blind–like artifacts and freezing artifacts, which appear as holes. We chose H&E staining on permanent sections because this technique is one of the methods most commonly used in clinical practice for assessing steatosis.[13-16]

Steatosis is classified as MiS or MaS. MaS can be divided into sd-MaS and ld-MaS.[9, 10] The definition of ld-MaS in the present study is concordant with the terms MaS and macrosteatosis reported in the surgical literature,[5, 13, 17-21] which are associated with inferior outcomes.[3-5] All of these terms have been used interchangeably and define the condition of displacement of the cell nucleolus by single large fat vacuoles. However, the term microsteatosis or MiS is often incorrectly used and confused with sd-MaS.[5, 10, 18] True MiS is a very rare condition characterized by innumerable tiny lipid vesicles that are diffusely distributed, whereas the fat vacuoles in sd-MaS are larger and do not show a diffuse distribution pattern. In the present study, only 2 of the 201 donor livers had true MiS. Although cells with MiS and sd-MaS have a centrally located nucleolus, sd-MaS is a form of MaS and has to be differentiated from true MiS. In the surgical transplant literature, sd-MaS is often incorrectly labeled as MiS.[5, 18]

In the present study, the qualitative and quantitative assessment of steatosis by a blinded pathologist served as the reference for the surgeon's evaluation. Although the microscopic assessment of steatosis has been recently challenged by noninvasive techniques such as magnetic resonance imaging,[22, 23] computed tomography,[22-24] ultrasonic elastography,[25, 26] and bioelectrical impedance measurement,[27] liver biopsy remains the gold standard for the evaluation of steatosis in donor livers because it allows differentiation between different types of steatosis and provides additional important information on inflammation, necrosis, and fibrosis. On the other hand, conflicting data have been published on the variability of steatosis assessment among pathologists.[13, 17, 28-30] In contrast to others,[13] we found excellent agreement for ld-MaS and lesser but still acceptable agreement for sd-MaS between 2 expert pathologists.

Steatosis (especially ld-MaS) has been proposed as an independent risk factor for reduced graft survival[3-5] and early hepatitis C virus recurrence in hepatitis C virus–positive recipients.[31, 32] Although steatotic donor organs can be used for liver transplantation[33, 34] and steatosis is reversible after OLT,[18, 34, 35] an appropriate donor-recipient match is of paramount importance for these high-risk organs.[33, 34] The practice of many transplant centers is to allocate steatotic organs to patients with lower Model for End-Stage Liver Disease (MELD) scores or hepatocellular carcinoma. The idea behind this policy is that critically ill recipients with high MELD scores may not tolerate initial graft dysfunction and postoperative complications. This policy is also mirrored in the present study, in which transplanted livers that were classified as excellent or good had higher recipient MELD scores, lower degrees of steatosis, and better initial graft function than those ranked as average or poor (Table 6).

One of the central findings of this study is the prediction of steatosis by the liver texture criteria of yellowness, round edges, and an absence of scratch marks. Yellowness and round edges are widely used criteria for steatosis,[29, 36-38] but surprisingly, published data on the association of these criteria with the type and quantity of steatosis are sparse. Two studies have reported conflicting data on the correlation of liver color assessments with the degree of steatosis in donor livers.[29, 36] One of the important observations of our study was that both criteria were predictors for ld-MaS, but they had no predictive value for sd-MaS (Fig. 5). We also included scratch marks in the systematic assessment of steatosis. Scratch marks are faint scratch lines on the liver capsule that microscopically appear as focal disruptions of the capsular connective tissue (Fig. 2). These marks can be observed on both the convexity and concavity of the liver surface. Our hypothesis is that scratch marks are caused by crystals from slush ice placed around the liver at the time of organ flushing during procurement. Similarly to yellowness and round edges, the absence of scratch marks was associated with ld-MaS and was less influenced by sd-MaS. Why macrosteatotic livers are less associated with scratch marks remains unknown. The combination of scratch marks and sharp edges identified livers without ld-MaS, regardless of the degree of yellowness. On the basis of these observations, we propose the incorporation of scratch marks into the surgical evaluation of steatosis.

Another central goal of this study was the prediction of steatosis based on the surgical assessment. This was approached in 2 ways: first by the surgeon's steatosis estimate and second by the assessment of the liver texture. Because sd-MaS and ld-MaS are microscopically defined conditions, the surgeon's estimation of the steatosis percentage can reflect only an estimate of fatty infiltration. The finding that the surgeon's steatosis estimate correlated moderately with ld-MaS and was weak for sd-MaS showed that the surgeon's estimation was mainly influenced by ld-MaS. This observation and the fact that graft failure is mainly associated with ld-MaS rather than sd-MaS[3-5] led us to restrict further prediction analysis to ld-MaS. With the clinically important cutoff value of 30% for ld-MaS,[2, 3] the surgeon was excellent in predicting ld-MaS below the cutoff, whereas the prediction of ld-MaS ≥ 30% was correct only for the half of the cases with a significant underestimation of ld-MaS. The argument that differences in the clinical assessment of steatosis are related to the different experience levels of evaluating surgeon cannot be supported by the present study, in which the surgical assessment of sd-MaS and ld-MaS was challenging even in experienced hands.

In a second approach, we used CART analysis to predict ld-MaS through the use of liver texture characteristics (yellowness, round edges, and scratch marks).[11] The CART analysis was better in predicting moderate and severe ld-MaS but less precise in predicting no or mild ld-MaS in comparison with the surgeon's steatosis estimate (Tables 4 and 5). This implies that liver texture characteristics are more important in identifying moderately and severely macrosteatotic organs than the actual steatosis estimation by the surgeon. In contrast to other studies that have relied mainly on yellowness,[29, 36] it is important to take all 3 liver texture criteria into account for steatosis assessment. The constellation of moderate to severe yellowness, round edges, and an absence of scratch marks should alert the surgeon to potential underlying high-degree ld-MaS, whereas normal/mild yellowness or the combination of sharp edges and scratch marks is an excellent predictor for no or minimal ld-MaS.

The findings of the present study bring up the question of which donor livers need to be biopsied on the basis of visualization. Currently, biopsies are recorded for only 23% of transplanted donor livers in the United States.[3] A transatlantic survey has revealed that 47% and 38% of liver transplant surgeons in the United States and United Kingdom, respectively, perform biopsy if there is serious concern on visualization.[14] Because of the surgical underestimation of ld-MaS, we strongly recommend biopsy if there is visual suspicion of steatosis. Organs with scratch marks and sharp edges are rarely associated with steatosis and do not require biopsy for steatosis evaluation. In the scenario in which a pathologist is logistically not available at the time of procurement, the surgeon's assessment of steatosis should primarily rely on the appearance of the 3 proposed liver texture criteria.

In conclusion, the present study shows that the precise estimation of steatosis in steatotic organs remains challenging even in experienced hands. The liver texture and the estimation of steatosis reflect mainly the degree of ld-MaS rather than the degree of sd-MaS. This study has demonstrated for the first time that the absence of scratch marks is associated with ld-MaS. The findings of this study provide practical guidelines for the clinical evaluation of the degree of steatosis in donor livers before transplantation, especially when histological assessment at the donor hospital is not available.

ACKNOWLEDGMENT

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES

The authors thank Christopher Jones, M.D., Kiran Dhanireddy, M.D., and Ali Cheaito, M.D., for their help in conducting the study. They also thank Stephanie Okimoto for her excellent assistance with data management and OneLegacy for its great cooperation.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. ACKNOWLEDGMENT
  7. REFERENCES