Hepatic epithelioid haemangioendothelioma: is transplantation the only treatment option?

Abstract

Background

Hepatic epithelioid haemangioendothelioma (HEH) is a rare vascular neoplasm with unpredictable clinical behaviour.

Aim

To compare overall survival (OS) and disease-free survival (DFS) between liver resection (LR) and orthotopic liver transplantation (OLT) for the treatment of HEH.

Methods

Retrospective review of 30 patients with HEH treated at Mayo Clinic during 1984 and 2007.

Results

Median age was 46 years with a female predominance of 2 : 1. Treatment included LR (n = 11), OLT (n = 11), chemotherapy (n = 5) and no treatment (n = 3). LR was associated with a 1-, 3- and 5-year OS of 100%, 86% and 86% and a DFS of 78%, 62% and 62%, respectively. OLT was associated with a 1-, 3- and 5-year OS of 91%, 73% and 73% and a DFS 64%, 46% and 46%, respectively. Metastases were present in 37% of patients but did not significantly affect OS. Important predictors of a favourable OS and DFS were largest tumour ≤ 10 cm and multifocal disease with ≤10 nodules.

Conclusion

LR and OLT achieve comparable results in the treatment of HEH. LR is appropriate for patients with resectable disease and favourable prognostic factors. OLT is appropriate for patients with unresectable disease and possibly those with unfavourable prognostic factors. Metastases may not be a contraindication to surgical treatment.

Introduction

Hepatic epithelioid haemangioendothelioma (HEH) is a rare vascular neoplasm accounting for less than 1% of all hepatic malignancies.¹ Since initially described in 1982 by Weiss and Enzinger,² less than 500 cases have been reported.³ HEH primarily occurs in women in the mid fifth decade of life without underlying chronic liver disease. Aetiologic risk factors are unknown, although HEH has occurred in association with oral contraceptive pills (OCPs), exposure to vinyl chloride or major hepatic trauma.^3–7 Clinical presentation is variable and non-specific and ranges from an indolent course^6,7 to a rapidly progressive disease.^8–12 In nearly 25% of patients, HEH is diagnosed incidentally. The diagnosis of HEH is confirmed by immunostaining for factor VIII-related antigens and other endothelial markers on biopsy.² No imaging features of HEH are pathognomonic.

The optimal treatment for HEH is disputed. In fact, all current treatment recommendations are based on evidence from limited retrospective clinical series. Moreover, treatment is influenced by the clinical behaviour of the HEH encountered, and the practice expertise of the managing surgeons or physicians. Although both liver resection (LR) and orthoptic liver transplantation (OLT) have been used as treatment options, neither has proven superior. Initial reports of LR were disappointing with a 5-year survival of only 28%.⁶ Outcome after OLT is reportedly better than LR with a 5-year survival of 64%;¹³ however, whether this clinical difference is attributable to OLT as a more optimal treatment, patient selection bias or biological behaviour of the HEH is unknown. Moreover, the lack of underlying liver disease and the extent of intrahepatic disease also impact the recommendation of OLT for HEH.

We reviewed our experience to further document outcomes of both LR and OLT in the treatment of HEH. We sought to provide additional data on outcomes for both surgical options, define selection factors for surgical treatment and determine whether any clinicopathological features of HEH have prognostic value.

Methods

Between 1984 and 2007, 33 consecutive adult patients with HEH were evaluated and treated at the Mayo Clinic. Three patients were lost to follow-up and were excluded from the analysis. Of the remaining 30 patients, four patients were treated at Mayo Clinic Arizona; five patients were treated at Mayo Clinic Florida; and the remaining 21 patients were treated at Mayo Clinic in Rochester, Minnesota. We retrospectively reviewed their medical records, radiological studies, laboratory values, operative reports and pathological specimens.

The diagnosis of HEH in all patients was based on positive staining for factor VIII-related antigen and other endothelial cell markers as described by Ishak et al.⁶ When patients were referred from other institutions, the liver biopsy was reviewed for pathological confirmation. Ultrasonography (US), computed tomography (CT) and magnetic resonance imaging (MRI) were reviewed to determine site, number and size of HEH. Chest X-ray (CXR), positron emission tomography (PET) and CT were used to detect distant metastases.

Hepatic epithelioid hemangioendothelioma burden was divided into groups based on number and size of nodules: ≤10 nodules and >10 nodules, largest HEH size <5 cm, 5–10 cm and >10 cm. The intrahepatic morphology of HEH was classified as nodular if masses were discreet and well circumscribed or diffuse if masses were widespread and coalescing. The type of LR and tumour location was classified according to the International Hepato-Pancreato-Biliary Association (IHPBA) classification.⁷ Post-operative mortality and morbidity were defined at 30 days or during the hospitalization of the incident operation. The severity of post-operative complications was categorized according to the Clavien et al. classification.¹⁴

A follow-up survey was mailed to 23 patients, who did not have follow-up within the past 6 months, and 12 (52%) responded. The follow-up survey addressed current symptoms, disease status, frequency of follow-up and additional treatments rendered outside of Mayo Clinic. Patients who chose to follow-up at the Mayo Clinic were assessed by clinical and radiological examination.

Frequency distribution among groups was determined using the χ²-test. Survival was estimated by Kaplan–Maier. Cox proportional hazards regression model was used to identify risk. A P-value of <0.05 was considered statistically significant.

Results

Demographics

There were 30 patients with a 2:1 female-to-male ratio. The median age was 46 years (range 21–79) (Table 1). Of the three treatment groups, 37% (n = 11) of patients underwent LR, 37% (n = 11) underwent OLT and the remaining 27% (n = 8) were treated non-operatively: chemotherapy in five and no treatment in three. The most common symptom at time of diagnosis was right upper quadrant pain (67%). Two patients presented with liver failure, both were secondary to Budd–Chiari syndrome. Twenty-seven per cent of patients were completely asymptomatic and the diagnosis was made incidentally during imaging for other indications. None of our patients had underlying chronic liver disease, cirrhosis, hepatic trauma or known exposure to vinyl chloride or contraceptive hormones.

Table 1.

Demographics and presentation

	Resection (n = 11)	Transplant (n = 11)	Palliative treatment (n = 8)	Total (n = 30)	P-value
Gender					0.6955a
Female	6 (54%)	8 (77%)	6 (75%)	20 (67%)
Male	5 (45%)	3 (27%)	2 (25%)	10 (33%)
Age at diagnosis					0.0782b
Mean	40.1	46.9	52.2	45.8
Size of largest nodule at initial surgery					0.0065ac
0–10 cm	11 (100%)	8 (73%)	3 (37%)	22 (73%)
>10 cm	0 (0%)	3 (27%)	5 (62%)	8 (27%)
Number of nodules at initial surgery					0.0039acd
≤10	10 (91%)	3 (27%)	2 (25%)	15 (50%)
>10	1 (9%)	8 (73%)	6 (75%)	15 (50%)
Number of segments involved					0.0096acd
≤4	8 (73%)	2 (18%)	1 (12%)	11 (37%)
>4	3 (27%)	9 (82%)	7 (87%)	19 (63%)
Extrahepatic involvement at initial surgery					0.0031ace
N	9 (82%)	9 (82%)	1 (12%)	19 (63%)
Y	2 (18%)	2 (18%)	7 (87%)	11 (37%)

^aFisher's exact.

^bKruskal–Wallis.

^cSignificant difference observed in a pair-wise comparison between resection and palliative treatment.

^dSignificant difference observed in a pair-wise comparison between transplant and resection.

^eSignificant difference observed in a pair-wise comparison between transplant and palliative treatment.

A pre-operative biopsy was obtained in 28 patients (93%); two (7%) patients were diagnosed at time of resection. Consequently, interval from diagnosis to treatment was quite disparate with a median of 368 days (range 0–7300 days). One patient initially had no treatment after diagnosis until disease progression prompted treatment 20 years later. Patients with available tests for tumour markers including alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), cancer antigen 19-9 (CA 19-9), revealed that all were within normal limits. Alkaline phosphatase was the most frequent abnormal pre-operative laboratory value and was consistently increased in 63% (n = 19) of patients.

Pre-operative imaging included CT in all patients, US in 50%, MRI in 40% and PET in 20%. Disease burden was based on imaging and was extremely variable. Patients were divided equally between groups with regard to number of HEH nodules, 50% had ≤10 nodules and 50% had >10 nodules. The largest HEH was <5 cm in 43% of patients, between 5 and 10 cm in 30% and >10 cm in 27%. Extrahepatic metastatic disease was present at diagnosis in 37% of patients; however, this did not have a significant impact on overall survival (OS) (P = 0.5). The sites of metastases included: lung (n = 8), peritoneum (n = 2), bone (n = 2), brain (n = 1) and skin (n = 1). Multiple sites of metastases were present in one-third of patients who had metastatic disease.

Post-operative complications

Complications after LR occurred in 36% of patients. One patient had a grade IV complication (perihepatic abscess requiring operative drainage). Median post-operative hospitalization was 6 days (range: 2–15 days) after LR. Complications after OLT occurred in 54% of patients. There were three grade IV complications which included a portal vein thrombosis, hepatic artery thrombosis requiring thrombectomy and anastomotic revision and post-operative bleeding requiring operative exploration and evacuation of a haematoma. The last patient subsequently developed a small bowel obstruction from an internal hernia that necessitated re-exploration and repair of a mesenteric defect. One patient died post-operatively (grade V complication) in a hospice on post-operative day 16 after OLT was aborted as a result of diffuse intraperitoneal metastases. Median post-operative hospitalization was 19 days (range: 4–120 days) after OLT but not significantly longer than LR (P = 0.095).

Overall survival

Overall median follow-up from time of diagnosis was 41.6 months (range 0–243 months). Overall survival for all patients at 5 years was 64% (95% CI: 47.6-86.3%). Overall, 12 out of the 30 (40%) patients died during follow-up: one patient died within 2 years of LR, six patients died from 1 month to 11 years after OLT and five of the eight patients who received no treatment died within 4 years of diagnosis. Interval from treatment to death was 0.5 to 11 years after OLT. Death in the no treatment group occurred within 1 month to 4 years after diagnosis. The early death in the OLT treatment group was as a result of brain metastases that were undiagnosed at the time of transplantation. Regardless of treatment, almost all the patients who are deceased had more than 10 tumour nodules and a diffuse disease pattern.

Of the surviving patients, the disease status at last follow-up was no evidence of disease (NED) in 68%, stable disease in 26% and progressive disease in 5%. There was no significant difference in OS (P = 0.128) after LR and OLT (Fig. 1). In the LR group, overall survival at 1-, 3- and 5-years was 100%, 86% and 86%, respectively. Among patients treated with OLT, overall survival at 1-, 3- and 5-years was 91%, 73% and 73%, respectively. In contrast, OS for patients who had medical or no treatment at 1-, 3- and 5-years was 57%, 43% and 29%, respectively. The long-term survival (10 years) for LR and OLT was 85.7% and 42.4%, respectively.

Figure 1.

Kaplan–Meier curve demonstrating overall survival among treatment groups. LR, liver resection; OLT, orthotopic liver transplantation

Disease-free survival

There was no significant difference in disease-free survival (DFS) after LR and OLT (P = 0.405). Disease-free survival at 1-, 3- and 5-years was 78%, 62% and 62% after LR and 64%, 46% and 46% after OLT (Fig. 2). Forty per cent of patients underwent a subsequent intervention for recurrence after initial LR and OLT. Interventions included chemotherapy, radiofrequency ablation (RFA), repeat LR and extrahepatic metastasectomy. There was no difference in number of subsequent interventions between LR (n = 4) and OLT (n = 5).

Figure 2.

Kaplan–Meier curve demonstrating disease-free survival among treatment groups. LR, liver resection; OLT, orthotopic liver transplantation

Clinicopathological factors as prognosticators

Several pathological and morphological features of HEH correlated with survival (Table 2, Fig. 3a–d). Largest HEH size ≤ 10 cm (P = 0.003) and nodular disease pattern (P = 0.01) were associated with prolonged DFS. Number of tumour nodules ≤10 approached significance (P = 0.052) for prolonged DFS. Clinicopathological factors associated with prolonged OS were tumour size ≤ 10 cm, ≤ 10 nodules and nodular disease ≤ 4 hepatic segments (P < 0.02). Using a univariate Cox Proportional Hazard model, the hazard ratio (HR) for mortality in patients for >10 nodules was 5.83 (1.27–26.80, P = 0.023) and the HR for HEH >10 cm was 10.976 (2.76–43.68, P = 0.0007). The HR for diffuse disease pattern was 8.14 (1.75–37.91, P = 0.0076) and that for >4 liver segments involvement was 4.92 (1.06–22.77, P = 0.041) (Table 1).

Table 2.

Correlation of clinicopathological factors to overall survival

Variable	Survival – years (%)
	1	3	5	P value
Treatment option				0.1
Transplant	91	73	73
Resection	100	86	86
Palliative	57	43	29
Metastasis at diagnosis				0.5
Yes	70	50	50
No	94	80	73
Number of nodules				0.01
≤10	100	91	91
>10	73	51	44
Size of nodules				0.0001
<5	100	89	89
5–10	100	87	87
>10	50	25	12.5
Disease type				0.0015
Nodular	100	92	92
Diffuse	67	42	33
Hepatic segments involved				0.02
≤4	100	90	90
>4	75	55	48

Figure 3.

Kaplan–Meier curve demonstrating overall survival among patients with (a) ≤10 and >10 tumour nodules, (b) varying tumour sizes, (c) varying disease pattern and (d) number of liver segments involved (≤4 vs. >4)

The clinicopathological factors were unevenly distributed across the three treatment groups. Patients who underwent LR had significantly fewer tumour nodules and liver segments involved with HEH compared with those that underwent OLT (P = 0.004). Patients who had extrahepatic disease had more tumour nodules, large tumour nodules and more liver segments involved and were more likely to receive no treatment at all (P < 0.009).

Discussion

In the present study, we sought to determine the appropriate roles for LR and OLT in the treatment of patients with HEH. In addition, we aimed to define clinicopathological characteristics that lead to favourable outcomes. We identified size ≤ 10 cm, number ≤10 and extent of hepatic involvement ≤ 4 segments as associated with prolonged DFS. Moreover, our study suggests that patients with these findings are candidates for LR rather than OLT; thus, further clarifying which operative approach should be utilized in patients with HEH. Also our study reaffirmed the wide spectrum of biological behaviour of HEH which can affect reported outcomes and emphasized the need to further identify better markers of its diverse tumour biology.

We identified clinicopathological factors that can stratify patients prognostically. These factors, if confirmed, potentially can direct therapeutic approaches and provide clinicians with prognostic information to consult individual patients regarding management of HEH which has variable tumour biology. The gross pathological features of HEH which correlated with clinical behaviour (or manifestations) included: tumour size, number of nodules, disease type and number of hepatic segments. Survival by largest tumour size for patients with HEH < 5 cm and 5–10 cm was similar with 1-, 3- and 5-year OS rates of 100%, 89% and 89% and 100%, 88% and 88%, respectively. Survival for patients with HEH > 10 cm had significantly less OS with 1-, 3- and 5-year survival of 5%, 25% and 13%. Therefore, patients were dichotomized by largest tumour size of ≤10 cm and >10 cm. We also showed that >10 HEH nodules, diffuse intrahepatic morphology and >4 segments involvement was associated with a significantly poorer DFS and OS. Interestingly, extrahepatic metastasis at the time of diagnosis was not associated with a worse overall survival and in fact two of the three patients with extrahepatic metastasis are alive at 3 and 10 years after operative resection.

A recent review of the literature has suggested that LR can play an important role in the treatment of patients with localized unilobar HEH.³ Liver resection was not widely recommended for patients with bilobar or infiltrating HEH because such disease was viewed as biologically aggressive based on the frequency of recurrence.¹⁵ Consequently, OLT was recommended to address bilobar HEH despite the absence of underlying cirrhosis, a limited donor pool and the potential sequelae of long-term immunosuppression.^3,15–17 Indeed, the recommendation of OLT for HEH is somewhat confounding given the findings that patients undergoing OLT for HEH have more advanced intrahepatic disease and occasionally even extrahepatic HEH which is distinctly disparate from the recommendations for OLT for less advanced hepatocellular carcinoma (HCC) as evident by the Milan (or even extended Milan) criteria used for HCC. The fact that OLT can provide prolonged survival for advanced intrahepatic HEH likely implies that intrahepatic morphology or extent of HEH does not fully reflect aggressive systemic tumour biology.

Indeed, our findings and others^6,7 of prolonged survival of patients with HEH regardless of treatment show that the natural history of HEH is variable and, in part, has biased reported outcomes. Regardless of unknown factors affecting natural history, our findings support LR for HEH despite the presence of bilobar disease. We found that the overall 5-year survival after LR and OLT was similar (86% and 64%, respectively). Furthermore, we found similar 1- and 5-year DFS after LR (78%, and 62%) and OLT (64% and 46%). Thus, LR is a surgical option for many patients with HEH regardless of bilobar distribution provided the hepatic disease can be resected. Clearly, LR is not applicable for all patients with HEH, and OLT will remain an important treatment alternative. Our patients who underwent OLT had significantly more nodules of HEH and hepatic segments involved (Table 2). Both number of HEH nodules and involved hepatic segments were important predictors of resectability of our patients. We believe that OLT is a more appropriate treatment for patients with >10 nodules or >4 involved hepatic segments. Thus, careful assessment of patients pre-operatively for these findings may provide more reliable criteria for selecting surgical approach and achieving similar oncological outcomes.

Moreover, the two patient populations treated with LR and OLT are not similar, as our patients who underwent OLT had significantly more nodules of HEH and hepatic segments involved (Table 2). All of the patients treated with OLT who are deceased had diffuse disease and either > 10 nodules, largest nodule > 10 cm or both. Both OLT patients with favourable prognostic factors (≤10 nodules, ≤10 cm, and nodular disease) are alive with NED at a minimum of 5 years. The more advanced disease in the OLT patients may explain the slightly worse, although not statistically significant, OS and DFS compared with LR.

Extrahepatic HEH at presentation occurred in 37% of our patients and is consistent with the literature.³ However, our findings and that of several others^3,15,18,19 suggest that metastatic disease is not associated with a decreased OS. Although a recent study did find that extrahepatic disease decreased median survival from 202 months to 59 months, none of these patients reportedly underwent systemic chemotherapy or metastasectomy.²⁰ All of our patients with extrahepatic disease had some systemic chemotherapy and/or metastasectomy, although chemotherapy was not associated with improved survival (P = 0.42). One (9% of all OLT patients) patient with stable biopsy proven bilateral lung metastases and favourable tumour factors (size, number of nodules, and liver segments) was offered transplantation and is currently alive with stable pulmonary metastasis 10 years later. Two (18% of all LR patients) other patients, one with metastasis to the skin and the other with metastases to the skin and lung were offered resection. The former did well and is alive without evidence of disease 3 years after resection of primary and metastasis, and the latter died just over a year after surgery. It is unclear why metastatic HEH is not the harbinger of imminent demise that is so often observed in other hepatic and biliary malignancies. The biology of metastatic HEH more closely resembles neuroendocrine tumours where patients have stable metastatic disease for years. Given the heterogeneity of the disease future studies are needed to determine the underlying biology that is the cause of this clinical observation. Given the current data we believe that limited and stable extrahepatic disease should not be a contraindication to LR or OLT, in very selected patients, particularly when extrahepatic disease is resectable. Most patients in our study with extrahepatic disease underwent adjunctive chemotherapy. Chemotherapeutic agents included: doxorubicin, ifosfamide, interferon-alpha, bevacizumab, gemcitabine, carboplatin, docetaxel and paclitaxel. A complete or partial response was not observed in any patients receiving chemotherapy. As a result of the variability of agents and regimens of chemotherapy used, we cannot currently recommend chemotherapy. Collaborative multi-institutional studies would be required.

Despite numerous case series of HEH, no prognostic clinical or histopathological features have been identified to reliably stage or predict tumour aggressiveness. Traditional indicators of biological activity, such as mitotic rate, nuclear atypia and capsular penetration are unrelated to outcome.⁷ Even distant metastases have not consistently correlated with survival.³ This discordance among biology and clinical outcome has made it difficult to construct treatment algorithms and assess treatment efficacy. In addition, the natural history of HEH is quite heterogeneous. Although the natural history of patients with untreated HEH is typically poor with only a 5%, 5-year survival,³ long-term survival without treatment has been reported^6,7,21 as evident by one of our patients who survived over 20 years with stable disease before progression prompted treatment. Finally, no staging system for HEH exists, thus, comparisons of outcomes between series with different treatment approaches are difficult. We have proposed that indicators of tumour burden could be used to stage HEH: largest tumour size, number of tumours, number of hepatic segments involved and pattern of disease. These factors can predict prognosis as well as direct surgical treatment. Selection of surgical approaches based on the above may allow development of a staged-based treatment approach in the future.

In conclusion, HEH remains an uncommon hepatic malignancy with a varied clinical presentation and natural history.^6,7,21 Consequently, a single therapeutic approach for all patients with HEH is not appropriate. Although usually distributed multicentrically in the liver, partial hepatic resection clearly can result in long-term survival. We believe that both partial hepatectomy and liver transplantation are selectively warranted by both our results and others^3,7,22,23Fig. 4. HEH size, number and extent in the liver can be used for selection of treatment for patients. Conversely, the presence of metastasis should not be used to exclude patients from surgical management, both LR and OLT. Finally, further studies, ideally using a large prospective multicenter database, to corroborate our findings and to more clearly define tumour stage, predictors of survival, outcomes after resection of disease in patients with extrahepatic metastasis and natural history are urgently needed.

Figure 4.

Suggested therapeutic strategy in patients with hepatic epithelioid haemangioendothelioma

Conflicts of interest

None declared.