Background: Needle biopsy of a suspicious liver lesion could guide management in the setting of equivocal imaging and serology, although it is not recommended generally because there is the possibility of tumour dissemination outside the liver. The incidence of needle track seeding following biopsy of a suspicious liver lesion is ill-defined, however.
Methods: A systematic review and meta-analysis of observational studies published before March 2007 was performed. Studies that reported on needle tract seeding following biopsy of suspicious liver lesions were identified. Lesions suspected of being hepatocelleular cancer (HCC) were considered. Data on the type of needle biopsy, diagnosis, incidence of needle track seeding duration to seeding, follow-up and impact on outcome were tabulated.
Results: Eight studies identified by systematic review on biopsy of HCC were included in a meta-analysis. The pooled estimate of a patient with seeding per 100 patients with HCC was 0.027 (95% confidence interval (CI) 0.018 to 0.040). There was no difference whether a fixed or random effects model was used. Q was 4.802 with 7 degrees of freedom, p = 0.684; thus the observed heterogeneity was compatible with variation by chance alone. The pooled estimate of a patient with seeding per 100 patients per year was 0.009 (95% CI 0.006 to 0.013), p = 0.686.
Conclusions: In this systematic review we have shown that the incidence of needle tract tumour seeding following biopsy of a HCC is 2.7% overall, or 0.9% per year.
Statistics from Altmetric.com
A worldwide shortage of liver donors and long waiting time for a liver transplant have resulted in patients on waiting lists with chronic liver disease being screened frequently to detect premalignant or early stage malignancies.1 2 The detection of high-grade dysplastic nodules (DNs) or small hepatocellular carcinomas (HCCs) usually results in therapeutic options including tumour ablation or transplantation.1 It is well recognised that early detection of HCC with meticulous pretransplant staging based on the Milan3 or University of California, San Francisco (UCSF) criteria results in excellent post-transplant survival.4 With the adaptation of the model for end stage liver disease (MELD) and patients with HCCs satisfying accepted criteria for early disease, allocation of additional or MELD exception scores has resulted in an increased likelihood of these patients undergoing a transplant.5–7
A high serum α-fetoprotein (AFP) level (>400 ng/ml) in the presence of a liver lesion is highly suggestive of HCC. Unfortunately, AFP has a low specificity, with elevated levels found in pregnancy, germ cell tumour, and acute or chronic hepatitis. Two-thirds of patients with cirrhosis may have indeterminate or significantly elevated AFP levels. AFP alone is therefore an inadequate method to screen for HCC in at-risk groups.8 Pretransplant diagnosis of HCC in patients with cirrhosis is therefore often based solely on imaging. According to the United Network for Organ Sharing (UNOS) criteria “pre listing biopsy is not mandatory”. Seven per cent of organs allocated for transplantation in the USA, however, occur in patients misdiagnosed as having HCC. This accounts for approximately 300 livers per year being transplanted in patients with false positive results in the USA.9
Advances in cross-sectional imaging have led to a drastic limitation in the requirement for liver nodule biopsy.10 11 Good-quality, contrast-enhanced, computed tomography (CT) and magnetic resonance imaging (MRI) is generally highly accurate for the diagnosis of malignant liver tumours.12–14 It is believed that MRI is more accurate and sensitive than CT in the detection and characterisation of focal liver lesions.1 The accuracy of cross-sectional imaging in diagnosing small malignant liver tumours (<2 cm), especially in the cirrhotic liver, however, remains problematic with sensitivities for MRI detection of such lesions being 13–67% and approaching 100% for lesions >3 cm.12–14 Therefore the European Association for the Study of the Liver (EASL) permits needle biopsy of lesions ranging from 10 to 20 mm in diameter in patients with cirrhosis. Smaller nodules require closer follow-up with imaging coupled with serology for tumour makers.15
A tissue diagnosis of a suspicious liver lesion would therefore be an ideal guide in the setting of equivocal imaging and serology to the best line of treatment. This could be achieved by image-guided percutaneous, transjugular or endoscopic needle aspirations or core biopsies.16 Liver surgeons, however, recommend strongly against such biopsies due to the fear of tumour dissemination outside the liver along the needle track. The incidence of such tumour seeding is yet to be established with certainty, however. While modest rates of tumour dissemination of 0.003–0.009% have been described,17–20 reviews on the subject have described incidences of 5%.21 There also are reports of incidences of 12.5% following needle biopsy followed by radio-frequency ablation.22 This incidence will also vary with the diameter of the needle used, the number of passes, and the amount of normal parenchyma traversed by the needle.16
The sensitivity of fine needle aspiration biopsy (FNA) of liver lesions reported range from 67 to 100%, and the specificity 80 to 100%.23–30 In addition to needle tract seeding of tumour, distinctions between hepatocellular nodular lesions and reactive hepatocytes, between well-differentiated HCC and benign hepatocellular nodular lesions, between poorly differentiated HCC from cholangiocarcinoma and metastatic carcinomas are other issues that need consideration.31 32 There have been no robust controlled clinical trials comparing FNA and core biopsy, although it is believed that both methods are complementary.33–35
In an attempt to accurately ascertain the incidence of tumour seeding following needle biopsy of a primary liver malignancy we systematically reviewed the evidence. This study was therefore designed as a systematic review and meta-analysis of observational studies based on the guidelines described by the MOOSE group,36 since no randomised controlled trials on the subject exist. The hypothesis explored in this study is that needle biopsy of a primary liver malignancy results in tumour seeding that would affect treatment options and impact patient survival. This study focused solely on those patients who have had needle biopsies of liver lesions for the diagnosis of HCC. The outcomes were the occurrence of tumour seeding and patient survival. Studies using existing databases retrospectively, cross-sectional studies, case series and case–control studies with or without historical controls were included. Prior to the outset of the study we confirmed that there were no existing systematic reviews on this subject.
What is the incidence of needle track seeding after biopsy of a liver lesion suspected of being hepatocellular cancer?
Three authors (MS, BH, CH) independently performed a systematic literature search of the following databases using the methods of the Cochrane collaboration; National Centre for Biotechnology Information (NCBI; part of the United States National Library of Medicine (NLM), a branch of the National Institutes of Health), National Library of Medicine, PubMed (January 1966 to March 2007) and EMBASE. Key words used were: needle biopsy, needle aspiration, core biopsy, liver lesions, track, seeding, implantation, spread, extension, proliferation, hepatocellular cancer. The bibliographic database search was supplemented with general Internet searches, examination of personal paper collections and checking the reference lists of included studies. There was no language restriction. Unpublished material was not specifically looked for.
Cross-sectional case studies where needle biopsy was used in the cytopathological diagnosis of liver lesions were included. This also included historical case controlled trials. It was a requirement that malignancy needed to be proven by formal histology. Details of type of needle used, and level of follow-up of the patient was also a requirement. Case reports, seeding following percutaneous ethanol injection and radio-frequency ablation were excluded. We also excluded reports on intrahepatic seeding, since it was difficult to identify if true seeding had occurred or a new lesion had occurred as a result of field cancerisation.
Data were abstracted by one reviewer and checked by a second. Information was recorded on the nature of the population, the level of follow-up, the type of needle used, number of passes and method of biopsy (eg, free hand or guided). In addition, where seeding occurred, details were extracted on how this was diagnosed (imaging or histology), time to seeding, tumour biology and seeding site.
Results for the studies included were first calculated both as patients with seeding per 100 patients and patients with seeding per 100 patients per year of follow-up. These values were then tabulated. Meta-analysis was also undertaken using the “meta_rate” command in Stata. Two different approaches were used to deal with zero events (exclusion of study from the meta-analysis or replacing 0 with 0.5) but this did not make a difference to the overall estimates. The sensitivity of the result to a different statistical approach was also explored using Poisson regression. Again the results were very similar and so for simplicity only the results of the “meta_rate” analysis are presented. Heterogeneity was examined using the Q statistic with a p value <0.05 being taken as evidence of heterogeneity unlikely to be accounted for by chance alone.
Number of included studies
Seven hundred and fifty-nine potentially relevant abstracts were identified. Five hundred and ninety-eight abstracts were excluded as clearly irrelevant on the basis of the title and abstract. One hundred and sixty-one articles were retrieved for detailed scrutiny. One hundred and fifty-three articles were excluded for failure to meet inclusion criteria. Finally, eight articles were selected for systematic review (table 1).
The first three authors independently reviewed the eight articles included in the meta-analysis in full, confirming that the papers described studies in which needle biopsy was used in the cytopathological diagnosis of liver lesions, with malignancy being diagnosed by histology. The quality of each paper was assessed in terms of study design (retrospective, cross-sectional, historical case–control), outcome measures, potential biases and whether it identified for case mix differences (differences in patient demographics, stage of disease between groups that may have biased results). Outcome measures were categorised into rates of seeding.
Details of the studies included in the survey
Huang et al37 performed 455 ultrasound guided cutting needle biopsies on 420 patients with HCC. In this series ultrasound-guided biopsy seems to have been done on all patients who were suspected to have HCC, irrespective of size of the lesion (<1 cm to >5 cm and for diffuse lesions). Those patients with multiple lesions had all lesions biopsied resulting in 455 biopsy sessions. Multiple passes were performed until tumour and non-tumour tissue was obtained as assessed grossly.37 Of the 455 biopsies, 391 (86%) confirmed HCC. All 420 had a final histological diagnosis of HCC.37 There were nine cases of spread of HCC to the chest wall along needle track in this series 37 (table 1). Among the 455 biopsy sessions, 126 (28%) had Vim Silverman needles used of which four had needle track seeding (3.2%). Five of 329 (1.5%) biopsied with disposable Tru-cut needles had implantations. The incidence of implantation using different needles was not significant (χ2 test p>0.01). Needle tract seeding occurred in 2%. Interestingly, eight of the nine tumours which resulted in seeding following biopsy were small lesions (⩽2 cm). Needle track seeding in these patients, however, did not impact survival.37
In the series described by Takamori et al,38 51 of 91 patients with HCC underwent image-guided biopsy of lesions suspected of being HCC with three seedings reported. It is not clear, however, in this paper as to what the indications were for needle biopsy since one of the lesions biopsied was associated with grossly elevated AFP levels in a lesion >10 cm.38 Another lesion which resulted in seeding in this series also went on to have alcohol injection into a 3 cm lesion, which could be a confounding factor.38
Using a coaxial cutting needle technique with a 17G introducer and an 18G biopsy needle Maturen et al39 biopsied 1012 liver lesions of which 128 were later confirmed to be HCC. In the technique described, the introducer stayed in place while multiple passes were made with the needle (three to four). They report no incidence of needle tract seeding in their series of patients. Citing previous reviews these authors argue that seeding occurs as early as 21 days post-procedure.39–41 They therefore used a cut-off of 30 days follow-up as adequate in their cohort of patients. In the 101 patients which constituted the final study population, the mean follow-up was 410 days (33–1989), 27 being excluded since the follow-up was less than 30 days.39
In another series, the diagnosis of HCC was established in 122 patients following needle biopsy.40 Fifteen patients had negative biopsies of which 13 were found to have HCC following surgery.40 This series reports two needle tract seeding in 135 patients with HCC.40 Tumour seeding commonly is reported along the needle tract in the subcutaneous tissue, in the intercostal muscles, and in the peritoneal cavity.40 42–44 These occur as discrete nodules or contiguous elongated lesions.42
Caturelli et al45 have published on the detection of HCC in patients with chronic liver disease, and again on ultrasound-guided FNA in the diagnosis of HCC in patients with cirrhosis.46 There is overlap in the durations both studies were conducted and therefore invariably some patients would be common to both papers. We therefore excluded the first paper by these authors45 and included the 2004 paper.46 This was a prospective multicentre study conducted over 9 years (January 1992 to December 2001), and included four large healthcare centres in Italy. It involved ultrasound surveillance of 4375 patients with cirrhosis. In the absence of raised AFP ultrasound-guided FNA was performed on 294 newly identified nodules with diameters of 2 cm or less. Of this number, 245 (89%) went on to receive correct pathological diagnoses following needle biopsy of which 210 were confirmed HCC, four others were found to be lymphomas and 31 nodules were benign. In the remaining 49 (17%) a diagnosis was not established.46 Of this 49, 29 (10%) later proved to be HCC, where FNA failed to yield a correct diagnosis. In these 29, eight were not adequate for making a diagnosis, 11 were incorrectly diagnosed regenerative nodules and 10 were dysplastic.46 These authors go on to state that confident exclusion of HCC is admittedly more difficult based on biopsy alone, and that the “gold standard” for diagnosis of regenerative and dysplastic liver nodules is pathological examination of a whole resected section.46 This multicentre study did not report a single needle tract seeding.46
Tumour seeding following FNA biopsy of HCC in the series included in this systematic review were treated with either local excision with clear margins and/or local radiotherapy.38 43 44 47 HCC seeding along needle tracts in patients who later went on to have definitive surgery in the form of liver resection also did not impact patient survival.37 40 42–44 47 There are no reports in the eight papers included in this study nor in published case reports and series excluded from this analysis of widespread intraperitoneal dissemination of HCC following FNA.
In the literature search undertaken we did not encounter any patients who developed needle tract seeding following biopsy, subsequently undergoing liver transplantation. In the eight studies included in the meta-analysis, three studies report 44 patients with FNA-diagnosed HCC went on to have successful liver transplants.37 39 40 47 There is no mention of any of these patients developing tumour seeding either before or after transplantation.37 39 40 47
The pooled estimate of patient with seeding per 100 patients was 0.027 (95% confidence interval (CI) 0.018 to 0.040) (see fig 1). There was no difference whether a fixed or random effects model was used. Q was 4.802 with 7 degrees of freedom, p = 0.684, thus the observed heterogeneity was compatible with variation by chance alone.
The pooled estimate of patient with seeding per 100 patients per year was 0.009 (95% CI 0.006 to 0.013) (see fig 2). There was no difference whether a fixed or random effects model was used. Q was 4.790 with 7 degrees of freedom, p = 0.686, thus again observed heterogeneity was compatible with variation by chance alone.
It has been shown previously that fine needle aspiration of a tumour could induce implantation of 1000 to 100 000 cells along the needle tract.48 The possibility therefore of needle tract seeding of tumour is a justifiable concern. This is especially so if such a seeding could impact patient survival and also management options. In this systematic review we have shown that the incidence of needle tract tumour seeding following biopsy of a HCC is 2.7% overall, or 0.9% per year. The 95% CI indicates that there is uncertainty, but that the highest rates with which the results of the included studies are compatible are 4.0% or 1.3% per year. In none of the reported cases of seeding did the event impact patient survival with the lesion being treated successfully by resection and local ablation.37 40 42–44 47 Comparative long-term survival data between those with and without seeding if it were available would have certainly have benefitted this meta-analysis. Another limitation of this review is that techniques used for needle biopsy varied between the studies included as would be the completeness in detecting all seedings following needle biopsy. All diagnosed seedings included in this review, however, were confirmed histologically (table 1). It is of interest to note that there are no reports of patients who develop needle tract seeding following biopsy going on to have a liver transplant or being taken off the waiting list.
Many transplant centres currently report 1 year survival rates in excess of 90% following liver transplantation for elective cases. The overall 1 year survival rate following liver transplantation for all indication which includes transplantation for end stage liver disease in non-fulminant, fulminant and those patients with HCC approaches 80%.49 50 Therefore there is a 10–20% mortality rate at 1 year following a liver transplant. The cause for this mortality rate is multifactorial with sepsis, multiorgan failure and disease recurrence being important. It could be weighed against a 2.7% risk of needle tract seeding following a biopsy of a HCC. This would be done in the setting of detection of a new small liver nodule (<2 cm) in a patient with cirrhosis with inconclusive AFP levels. While percutaneous image-guided biopsies of liver lesions should not be performed indiscriminately, at the risk of transplanting for a false positive, there exists a place for such a guided biopsy. However, it is our view that this should be undertaken only in specialised liver transplant centres on a case-by-case basis following multi-disciplinary review.
The reported time interval between FNA and the detection of an implanted tumour has a wide range and is from 3 months to 4 years.42 Many factors, however, must contribute towards this time interval: the initial number of spilled cells, the doubling time, and the microenvironment, which would include the local blood supply, are but a few. The median time to seeding in this study was 17 months.5–48 This therefore would invariably present in most cases once the definitive treatment procedure of either resection or transplant has taken place. It can also be argued that patients with small or well-differentiated HCCs are expected to have a longer life span than those with large or poorly differentiated HCCs.43 Therefore, the chances of detection of needle tract seeding is higher in the former.37 43
In the prospectively designed study reported by Caturelli et al, ultrasound-guided biopsy was done on 294 newly diagnosed nodules with diameters of 2 mm or less in the absence of a raised AFP. Two hundred and forty-five (89%) received correct pathological diagnoses. This gives an indication of the diagnostic role needle biopsy could have in the early diagnosis of HCC. The remaining seven papers focus mainly on needle biopsy and the incidence of seeding. In some of the studies it would seem that needle biopsy was carried out despite the diagnosis being obvious on the basis of imaging and serum AFP levels.37 39 The diagnostic role needle biopsy had over imaging in conjunction with AFP cannot therefore be clearly ascertained from these retrospective studies.
It has been suggested by Maturen et al39 that the use of a coaxial cutting needle; with a biopsy needle within an introducer is a method that could prevent needle tract seeding. This was following biopsy of 1012 liver lesions of which 128 were confirmed to be HCC. In the technique described, the introducer stayed in place while multiple passes were made with the biopsy needle.39 These authors report no incidence of seeding. Although this seems plausible, we feel that this could only be proven by a well-designed, prospective, randomised, controlled trial.
In the series reviewed false negative results due to inadequate sampling, mistaken HCC for regeneration or dysplastic nodules occurred. It is clear therefore that FNA is highly operator dependent, and that there is a requirement for close clinicopathological correlation.32 It is also important to combine cytology with micro histology especially in the diagnosis of well-differentiated HCC and the use of an ever-increasing array of discriminant panels of immuno stains may help in making a diagnosis.32
Following a negative or benign diagnosis of a biopsied liver lesion Caturelli et al suggest a second imaging study to follow. If this too is negative for radiological features of HCC an increased frequency of follow-up in the form of an ultrasound scan, AFP and clinical evaluation every 3–4 months is suggested for 12 months. If, during this period, no malignant change is detected, the non-malignant diagnosis is considered definitive, and HCC surveillance is resumed at the original frequency.46
We thank A Fry-Smith for assistance with searches.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.