Allograft liver biopsy is the gold standard in assessing transplant recipients for graft dysfunction. The impact of biopsy sample size on the diagnosis of acute cellular rejection (ACR) has not been studied.
To assess the relationship of biopsy sample length with the diagnosis and determine optimal biopsy sample size in the transplant setting.
We retrospectively reviewed 68 core biopsies from patients with a history of liver transplant. Each biopsy sample was read, on 5 different occasions with differing lengths, to assess for ACR per Banff criteria. Categorical agreement was calculated from rejection severity.
The length of biopsy sample strongly correlated with the number of portal tracts. ACR rates increased from 73.5% to 79.4% with increase in length from 1 cm to 2 cm, and moderate rejection increased from 27.9% to 33.82%. At 1.0 and 1.5 cm, no cases of severe rejection were detected; at 2.0 cm, 1 case was detected; and at 3.0 cm, 2 cases were detected. The major error rate was reduced to less than 10% with a length of 2.0 cm, at which length the average number of complete and partial portal triads was 10 and 13, respectively.
The likelihood of diagnosing ACR and rejection grade increased substantially with increase in biopsy sample length. This study suggests that a minimum length of 2 cm, 10 complete portal triads, or 13 partial/complete portal triads should be obtained for confident exclusion and grading of ACR.
Liver biopsy is the gold standard for assessment of liver rejection, and it is often undertaken in liver transplant recipients when graft dysfunction is suspected. With advancements in techniques, the risks associated with the procedure have been reduced, making it a relatively safe technique to assess rejection.1,2 Furthermore, biopsy findings are associated with an alteration in treatment in up to 41% of patients.3
Acute cellular rejection (ACR) is the most common diagnosis made in allograft biopsies, affecting up to 40% of transplant recipients.4–7 Chronic rejection occurs in up to 17% of transplant recipients.8 Appropriate treatment influences graft loss. ACR has a successful response rate of more than 80%, and chronic rejection can be reversible, especially if diagnosed early.4,9 This study was performed to assess the relationship of biopsy sample length with the diagnosis of ACR and to determine optimal biopsy sample size in the transplant setting.
This is a retrospective study in which a search of the anatomic pathology information system was conducted to generate a list of consecutive allograft liver biopsies during a 3-year period. Institutional review board approval was obtained. Biopsies from patients younger than 18 years were excluded. Biopsy samples less than 3 cm in aggregate length were excluded. At our institution, percutaneous liver biopsies are performed with a 16-gauge needle. Slides are routinely stained with hematoxylin-eosin, trichrome, reticulin, periodic acid–Schiff with diastase, and Prussian blue. Portal tracts with all 3 elements, namely, hepatic artery, portal vein, and bile duct, were identified as complete portal tracts. Lack of one of the elements in the portal tract was identified as partial portal tract.
Slides were retrieved and total length of biopsy samples was measured. One investigator blinded the slides and bracketed segments of biopsy sample at intervals of 1, 1.5, 2, and 3 cm. The slides were evaluated by a second investigator. Each slide was read 5 times at 1-week intervals. Slides were evaluated for enumeration of portal tracts (complete and partial), central veins, bile ducts, senescent bile ducts, and arterioles. In addition, an assessment was made, according to Banff criteria for ACR, of portal inflammation, ductitis, venulitis, total rejection activity index (RAI) score, and fibrosis (0–9 Banff suggested staging system). The percentage of senescent bile ducts, percentage of portal tracts with bile ducts, and percentage of tracts with arterioles were calculated. To assess severity of ACR, we used categorical assessments recommended by the International Liver Transplantation Society; specifically, RAI of 1 to 3 is categorized as mild, RAI of 4 to 6 as moderate, and RAI of 7 to 9 as severe rejection.10
For comparison, the reference diagnosis was taken as that which would be rendered on the fullness of available tissue. We then evaluated the effect of number of portal tracts and biopsy sample length on the diagnosis of ACR, and on grading of the severity of rejection. We also evaluated the correlation between biopsy sample length and number of portal tracts.
An assessment was made to look for disagreement between different categories of ACR (category error). For this, we designated 3 major categories of diagnosis, each associated with specific treatment plans: negative/indeterminate for ACR, mild/moderate ACR, and severe ACR. In practice, these categories may be more fluid, owing to other clinicopathologic considerations. That is, in some instances a patient with a diagnosis categorized as indefinite may receive treatment for mild ACR, and some patients with a diagnosis categorized as moderate may be treated aggressively, as though they had severe ACR. Nevertheless, for the purposes of this analysis, and based on prevailing treatment principles, we chose the 3 categories listed above.11
For each case, the diagnosis at the final sample length was used as the gold standard to look for agreement between the diagnosis at different lengths. For example, a case diagnosed as mild ACR at both 1 cm and the final length was considered in agreement at the length of 1 cm. However, if the case was diagnosed as mild ACR at 1 cm but moderate ACR at the final length it was considered an error.
In regard to category error, we chose to adopt a description of minor errors as those resulting in no significant alteration in therapy and major errors as those in which therapy would have been significantly impacted.12 Assignment of errors was defined as in Table 1. By this definition, there are no minor errors among the 3 categories described, as all would result in significant treatment implications. Therefore, to further elucidate the magnitude of errors, we have adopted terminology borrowed from quality management in clinical pathology, in which errors may be classified as major errors, which denotes a 1-category difference, and very major errors, which denotes 2 or more.13,14 For example, if a case was diagnosed as mild at 1 cm but moderate ACR at final length it was considered a major error. A case diagnosed as mild ACR at 1 cm but severe ACR at final length is a very major error.
A total of 68 patients were included in the study. The final diagnosis, based upon examination of all available tissue, was negative or indeterminate for ACR in 19.1% of patients (n = 13) and positive for ACR in 80.9% of patients (n = 55). Mild ACR was diagnosed in 32.4% of patients (n = 22), moderate ACR in 30.9% (n = 21), and severe ACR in 17.6% (n = 12).
Correlation of Length With Number of Portal Tracts
There was an increase in mean number of complete portal tracts from 4.9 at 1 cm to 15.4 at 3 cm. An increase was also seen in mean of total (complete plus incomplete) portal tracts from 6.3 at 1 cm to 22.1 at 3 cm. The length of the biopsy sample positively and strongly correlated with the number of portal tracts (Pearson correlation efficient: 0.741–0.768, P < .001) (Table 2).
Acute Cellular Rejection
Effect of Number of Portal Tracts on Diagnosis and Grading of ACR
We evaluated the percentage of cases in which the diagnosis of ACR was missed, or the severity was underestimated, as compared to the number of portal tracts (Figure 1). No cases of mild ACR were missed when the total (complete and incomplete) portal tracts numbered 11. However, 23.81% (5 of 21) of moderate ACR cases and 91.67% (11 of 12) of severe ACR cases were undercalled at that number. When the total number of portal tracts was 20, all mild ACR cases were identified, with 4.76% (1 of 21) of moderate ACR and 16.67% (2 of 12 cases) of severe ACR cases misclassified. A total portal tract count of 28 correctly identified the ACR in all the categories, as seen in Figure 1.
Effect of Length on Diagnosis and Grading of ACR
Biopsy samples of 1 cm in length identified ACR in 73.5% of cases, as compared to 80.9% in biopsy samples measuring more than 3 cm; that is, ACR was correctly identified in 90.8% of cases with only 1 cm of biopsy sample length (Table 3). However, the assessment of ACR severity increased progressively with length. As seen in Table 3, detection of moderate-severe ACR cases increased by 20.6% with an increase in length from 1 to 3 cm or more.
Some smaller biopsy samples would have resulted in underestimation of ACR severity (Figure 2). At 1 cm, 66.67% (14 of 21) of moderate ACR cases and 91.67% (11 of 12) of severe ACR cases would have been underestimated. At 2 cm, 23.8% (5 of 21) of moderate ACR and 41.67% (5 of 12) of severe ACR cases would have been underestimated.
Categorical Agreement in ACR
Categorical agreement improved from 57.4% at 1 cm to 85.4% at 2.0 cm and 95.6% at 3 cm (Table 4). There were no very major errors identified. Major error rates reduced from 22% at 1 cm to less than 10% at 2 cm (8.8%) and 3 cm (4.4%). To look for an adequate biopsy sample length, major errors was considered as the most important category, as minor errors did not result in change in patient treatment.
A scatterplot of the relationship between all bile ducts to portal triad ratio and biopsy sample length showed tight clustering of cases at 3.0 cm (Figure 3). Medians of RAI, duct to triad ratio, and fibrosis scores differed across size groups, as determined by Kruskal-Wallis 1-way analysis of variance test (P < .05). Similar findings were noted when comparing the number of portal tracts with these variables (P < .001).
Limitations of the Study
The liver biopsies included in the study were performed by percutaneous methods only. The study includes a total of 68 cases that were assessed 5 times each. There were no cases identified of plasma cell–rich T-cell–mediated rejection.
One important question in the liver allograft setting is the impact, if any, of biopsy sample size on the accuracy of assessment. This question has been studied in other settings, such as chronic viral hepatitis.15,16 In these studies, significant changes in staging and grading were observed with incremental increases in length of the biopsy sample as well as with a corresponding increase in the number of portal triads. While no definitive threshold was identified, a variety of sources appear to coalesce around recommending a threshold of at least 1 to 1.5 cm and a minimum of 6 to 11 portal tracts in a biopsy sample at least 1 to 1.4 mm in width.16–20 The Banff guideline recommended that a biopsy sample should contain at least 5 portal tracts, while a more recent update concerning antibody-mediated rejection suggests 2 passes with a 16-gauge needle.21,22
However, there are no studies to support these recommendations, and, to our knowledge, none that have been designed specifically to address this question, namely, what is the minimal amount of tissue that allows confident exclusion of or proper grading of rejection in liver allograft biopsies? We drew from studies conducted in other settings regarding liver biopsy adequacy in our design, assessing a variety of histologic parameters in specimens of varying size.
This study demonstrates a significant relationship between biopsy sample size and accuracy in the diagnosis of ACR in liver allograft biopsies. Our data suggest that smaller biopsy samples may result in the underdiagnosis of ACR or in underestimation of its severity. This effect was profound in biopsy samples measuring less than 1 cm but was relatively small in samples measuring at least 2 cm, as compared to samples measuring more than 3 cm in length. This effect persisted when categorical agreement was considered and would thereby be expected to have clinical significance.
Our design was similar to that adopted previously in the setting of chronic hepatitis.15,16,19,23 Unlike in those studies, the effect of biopsy sample diameter was not considered, since our study was conducted in a setting wherein 16-guage needles are routinely used. This is commonplace and conforms to international recommendations.24–26
In addition to biopsy sample length, we examined the number of portal triads as a way of providing context when examining biopsies obtained by other methods. It is unclear whether the effect of biopsy sample size is related more strongly to sample length or to number of portal tracts. Of note, we considered the number of complete portal tracts in addition to the number of compete and partial portal tracts in order to enhance the usefulness of the results to local practices.
The results show a marked tendency toward underdiagnosis of ACR in biopsy samples measuring only 1 cm. In samples measuring 2 cm, very few cases of ACR were missed entirely; however, at this length there was a marked tendency to underestimate severity.
There was an increase in mean number of complete portal tracts from 4.9 at 1 cm to 15.4 at 3 cm. An increase was also seen in the mean of total (partial and complete) portal tracts from 6.3 at 1 cm to 22.1 at 3 cm. No cases of ACR were missed when the total (complete and incomplete) portal tracts numbered 11. However, severity tended to be underestimated at that number. Even when the total number of portal tracts was 20, 16.67% of severe ACR cases were misclassified as mild or moderate. A total portal tract count of 28 correctly identified the ACR cases in all the respective categories.
Diagnostic accuracy is strongly influenced by biopsy sample size. In this study, we found that 1-cm biopsy samples provided high sensitivity for the presence of ACR. The mean number of complete portal triads at this length was 4.9, and the mean number of total (complete and partial) portal triads was 6.3. However, biopsy samples of this length tended to markedly underestimate the degree of ACR. Assessment of severity was much more accurate at a length of 2 cm, with a mean of 10.2 complete portal triads and 13.4 total (complete and partial) triads. Nevertheless, additional clinically relevant information continued to be gained in biopsy samples greater than 2 cm, and these data suggest that a 3-cm sample may be needed for complete confidence in the degree of severity assessed. At 3 cm, our biopsy samples had a mean of 15.4 complete portal triads and 20.2 total (complete and partial) triads.
Presented in part at the College of American Pathologists annual meeting; September 22, 2019; Orlando, Florida.
The authors have no relevant financial interest in the products or companies described in this article.