Fax: (608) 265-5963;
Histologic grade is correlated with outcome after resection of hepatic neuroendocrine neoplasms
Article first published online: 5 MAY 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 1, pages 126–134, 1 July 2008
How to Cite
Cho, C. S., Labow, D. M., Tang, L., Klimstra, D. S., Loeffler, A. G., Leverson, G. E., Fong, Y., Jarnagin, W. R., D'Angelica, M. I., Weber, S. M., Blumgart, L. H. and DeMatteo, R. P. (2008), Histologic grade is correlated with outcome after resection of hepatic neuroendocrine neoplasms. Cancer, 113: 126–134. doi: 10.1002/cncr.23523
- Issue published online: 20 JUN 2008
- Article first published online: 5 MAY 2008
- Manuscript Accepted: 4 FEB 2008
- Manuscript Revised: 13 DEC 2007
- Manuscript Received: 2 AUG 2007
The behavior of neuroendocrine neoplasms is poorly defined, and predictors of outcome after surgical resection have yet to be identified. Consequently, guidelines for treatment remain unclear. Current pathologic classification systems do not permit meaningful discrimination of hepatic neuroendocrine neoplasms.
The authors reviewed prospectively maintained databases from 2 institutions of patients who underwent hepatic resection for neuroendocrine neoplasms between 1990 and 2006. Patient, tumor, and operative characteristics were analyzed to identify factors associated with overall survival, progression-free survival, and symptom control. Hepatic neoplasms were stratified by using a 3-tier pathologic classification system based on the number of mitotic figures and the presence of tumor necrosis that was recently validated for pancreatic neuroendocrine neoplasms.
Seventy patients were identified from the databases. Low-grade, intermediate-grade, and high-grade neoplasms were identified in 53%, 37%, and 10% of patients, respectively. After a median follow-up of 51 months, the median overall survival for all patients was 91 months, and it was 108 months when 7 patients with high-grade neuroendocrine carcinomas were excluded. Progressive disease was eventually observed in 81% of patients, and the median progression-free survival was 17 months. The median time to the onset of symptoms was 39 months for patients who presented with hormonal symptoms and 80 months for all patients. Histologic grade was associated with poorer overall and progression-free survival.
When performed in a context of aggressive multimodality therapy, long-term outcomes after partial hepatectomy for hepatic neuroendocrine neoplasms were favorable; however, disease progression was eventually observed in the majority of patients. Several oncologic variables were associated with significant differences in survival after resection. A novel pathologic classification system appears to enhance prognostic stratification of patients with hepatic neuroendocrine neoplasms. Cancer 2008. © 2008 American Cancer Society.
Neuroendocrine neoplasms can present as well‒differentiated tumors or poorly differentiated carcinomas arising from various organs. They commonly metastasize to the liver, and it is not uncommon for the liver to be the only site of metastatic disease. The relative rarity and clinical indolence of hepatic neuroendocrine neoplasms, as well their multimodal treatment options, obfuscate their rational clinical management.1 Although their natural history is greatly variable, the clinical course of the majority of patients with hepatic metastases of well‒differentiated neuroendocrine neoplasms is typically prolonged but occasionally marked by major symptoms referable to hormone overproduction or tumor bulk. Significant improvements in surgical technique and perioperative management have vastly extended the application of surgical resection for these patients. However, it remains unclear when and how surgical intervention ought to be used—particularly in a setting in which multiple nonoperative modalities of therapy, such as octreotide hormone suppression, hepatic arterial embolization, and percutaneous ablation, are also available. The published experience to date suggests that surgical resection of hepatic neuroendocrine neoplasms can be associated with favorable outcomes, but clinical and oncologic variables that can distinguish patient cohorts that would benefit most from such aggressive therapy have not been identified.2-7 In addition, contemporary staging systems do not permit meaningful categorization of neuroendocrine neoplasms that have metastasized to the liver. Current pathologic criteria stratify neuroendocrine neoplasms as well‒differentiated neuroendocrine tumors (which, by definition, have not demonstrated clinical evidence of metastasis), well‒differentiated neuroendocrine carcinomas (which, by definition, have demonstrated local invasion or distant metastasis), and poorly differentiated neuroendocrine carcinomas (defined by the presence of high mitotic activity and characterized by markedly more aggressive tumor biology).8 The objective of the current analysis was to examine our experience with partial hepatectomy for patients with hepatic neuroendocrine neoplasms to identify the patient, tumor, and surgical characteristics associated with improved survival, disease control, and symptom relief. In part, this is a follow-up of some of the data presented in a previous report from Memorial Sloan-Kettering Cancer Center that had evaluated patients who had hepatic neuroendocrine metastases managed both surgically and nonsurgically.2 Furthermore, we wanted to determine whether a novel pathologic classification system, which was recently validated for use with pancreatic neuroendocrine neoplasms,9 could provide useful prognostic discrimination for neuroendocrine neoplasms of the liver.
MATERIALS AND METHODS
Patients who underwent surgical resection for hepatic neuroendocrine neoplasms by the Hepatobiliary Service at Memorial Sloan-Kettering Cancer Center between 1992 and 2004 and by the Section of Surgical Oncology of the University of Wisconsin School of Medicine and Public Health between 1990 and 2006 were identified from prospectively collected databases with institutional approvals. For preoperative imaging, standard cross-sectional modalities such as computed tomography or magnetic resonance imaging with or without nuclear medicine octreotide scans were used. Wherever technically feasible, attempts were made to resect all macroscopic disease. Intraoperative liver ultrasonography was used to exclude the presence of previously undiagnosed lesions, and surgical resection was performed using low central venous pressure anesthetic management and inflow/outflow control with parenchymal transection during intermittent Pringle maneuvers.10 Reoperative resections and adjunctive treatment modalities, such as octreotide injections, bland hepatic transarterial embolization, and radiofrequency tumor ablation, were used at the discretion of the attending surgeons after repeat multidisciplinary review; repeat hepatic resections were not analyzed in the current study.
Patient demographic characteristics and tumor and operative characteristics (of both hepatic resection and, if applicable, primary tumor resection) were recorded, and all pathologic specimens were reanalyzed by a pathologist (L.T., or D.S.K., or A.G.L.) to verify the histologic diagnosis of neuroendocrine neoplasm. Neoplasms were categorized as low grade (<2 mitotic figures per 50 high-power fields and no necrosis), intermediate grade (2–50 mitotic figures per 50 high-power fields and/or focal necrosis), or high grade (>50 mitotic figures per 50 high-power fields and/or extensive necrosis) using the classification system recently described for use with pancreatic neuroendocrine neoplasms. In this scheme, low-grade and intermediate-grade neoplasms are considered well differentiated, and high-grade neoplasms are considered poorly differentiated (Table 1).9 Primary tumors outside the liver that were resected completely at the time of partial hepatectomy were not considered extrahepatic foci of disease for this analysis. Patients with negative margins and no evidence of residual disease after resection were considered to have undergone complete resection (R0); resections that resulted in microscopic or macroscopic residual disease were categorized as R1 or R2 resections, respectively. Patients who underwent complete resection of all macroscopic disease (R0 or R1 resections) were considered to have undergone complete macroscopic resections. Medical records were reviewed to obtain follow-up information regarding survival, disease status, and symptom status. Perioperative complications were graded on a 5-point system previously defined as follows: grade 1, complication requiring oral medical therapy or bedside care; grade 2, complication requiring intravenous medical therapy; grade 3, complication requiring surgical or radiologic intervention; grade 4, complication resulting in unintended organ resection or chronic disability; and grade 5, complication resulting in death.11 The date of disease progression was defined as the date of earliest radiographic evidence of new tumor recurrence after complete resection or of tumor increase or new tumor growth after incomplete resection. The date of symptomatic recurrence was defined as the earliest date of documented postoperative onset of symptoms because of hormone overproduction or tumor bulk; postoperative hormone- or tumor-related symptoms that differed qualitatively from symptoms that were present preoperatively, were considered evidence of symptomatic recurrence. Overall survival, progression-free survival, and symptom-free survival were calculated by using the Kaplan-Meier method. Univariate analysis was performed with the log-rank test for categorical variables and with Cox regression analysis for continuous variables; variable associations with P < .05 on univariate analysis were analyzed by multivariate analysis. Statistical significance was defined as P < .05.
Patient Demographics and Primary Tumor Characteristics
Seventy patients underwent partial hepatectomy for pathologically confirmed hepatic neuroendocrine neoplasms between 1990 and 2006 (56 patients at Memorial Sloan-Kettering Cancer Center, 14 patients at the University of Wisconsin). Seventy-five percent of the patients in this population were women, and the mean age was 53 years (range, 25–81 years). Twenty-two patients presented with hormone-related symptoms referable to their neuroendocrine neoplasms. The most common sites of primary tumor were in the alimentary tract (43%) and the pancreas (33%). Other sites included the lung (4%), ovary (2%), and gallbladder (2%); the primary site was not known in 10 patients. The median primary tumor size was 3 cm (range, 0.5–15 cm). Of the 60 patients with known primary site, the primary tumor was resected in 50 patients. Sixteen of those patients underwent primary tumor resection before liver resection; the remaining 34 patients underwent primary tumor resection concurrently at the time of liver resection.
The median interval from initial diagnosis to partial hepatectomy was 4 months (range, 0–121 months). The median percentage of hepatic parenchymal involvement by tumor, as estimated by a review of cross-sectional imaging studies, was 40% (range, 10%–75%); the median number of hepatic segments involved with tumor foci was 4 (range, 1–8 segments). Sixty percent of patients underwent at least a hemihepatectomy; overall, a median of 4 segments was resected (range, 0–6 segments). Complete resection of all macroscopic disease (R0 or R1 resection) was performed in 69% of patients; the remaining 31% of patients had macroscopic residual disease after resection (R2 resection). In 15 patients (21%), extrahepatic foci of disease were identified. Eight of those 15 patients underwent complete macroscopic resection of intrahepatic and extrahepatic tumors. The median operative time was 230 minutes (range, 115–448 minutes), and the median estimated operative blood loss was 800 mL (range, 100–5500 mL). Thirty patients (43%) experienced at least 1 postoperative complication; 5 patients (17%) had complications that were classified as grade 1, 6 patients (20%) patients had complications that were classified as grade 2, 15 patients (50%) had complications that were classified as grade 3, and 2 patients died in the postoperative period (resulting in a 30-day operative mortality rate of 3%). The 60-day operative mortality rate was 4% (3 patients).
Pathologic analysis of resected partial hepatectomy specimens identified well‒differentiated neuroendocrine neoplasms in 63 of 70 patients (90%), including 37 neoplasms that were classified as low grade and 26 neoplasms that were classified as intermediate grade. The remaining 7 neoplasms were poorly differentiated, high-grade neuroendocrine carcinomas. The median number of tumors resected was 2 (range, 1–15 tumors), and the median size of the largest hepatic tumor resected was 6.7 cm (range, 0.7–28 cm). Of the 63 partial hepatectomy specimens for which margin status was documented, 40 specimens (63%) had microscopically negative margins (R0 resection), and the remaining 23 specimens harbored microscopically positive margins (R1 resection). Vascular invasion was observed in 12 specimens (17%), and lymph node metastases in the hepatic pedicle were encountered in 10 of 31 specimens that had evaluable lymph nodes. A comparison of clinicopathologic variables demonstrated that patients with poorly differentiated (high grade) hepatic neuroendocrine neoplasms more often were men and white and were more likely to harbor bilateral hepatic lesions than patients with well‒differentiated (low or intermediate grade) neoplasms.
Fifty-one patients (73%) received some additional form of oncologic therapy after recovery from their partial hepatectomy. Thirty patients underwent at least 1 round of bland hepatic arterial embolization, with a median of 2 embolizations per patient (range, 1–6 embolizations). Additional forms of postoperative treatment included octreotide in 22 patients, cytotoxic chemotherapy in 16 patients, reoperation in 12 patients, radiotherapy in 2 patients, and percutaneous radiofrequency ablation in 1 patient. After a median follow-up of 51 months, the 5-year actuarial overall survival rate was 61%, and the median overall survival was 91 months. When the 7 patients with high-grade neuroendocrine carcinomas were excluded from the analysis, the 5-year actuarial overall survival rate for patients with well‒differentiated hepatic neuroendocrine neoplasms was 65%, and the median overall survival was 108 months (Fig. 1A). Fifty-seven patients (81%) developed some form of disease progression during postoperative follow-up. Of these, de novo disease recurrence was the initial event in 44 patients, and interval growth of residual disease was the initial event in 13 patients. Of the 23 patients who underwent R2 resection, 20 patients developed disease recurrence (as the initial event in 6 patients) or growth (as the initial event in 13 patients); in 3 patients, no disease progression was observed after R2 resection after a mean follow-up of 15 months. The liver was the most common site of initial progression (80%); subsequent areas of disease progression were distributed more evenly between liver, lung, lymph nodes, bone, and brain. Six patients (10%) demonstrated initial progression in multiple organ sites simultaneously. The median time to disease progression was 17 months (range, 1–113 months for patients who progressed). When the 7 patients who had high-grade neuroendocrine carcinomas were excluded, the median progression-free survival was 19 months (Fig. 1B). Of the 22 patients who presented with symptoms referable to hormone overproduction, 100% experienced symptom relief after partial hepatectomy. However, 16 of those patients eventually developed recurrent symptoms during follow-up; the median time to first symptom recurrence was 39 months (range, 1–121 months for patients with recurrent symptoms). For the entire cohort of patients, including those who presented with and without symptoms, the median time to onset of any symptoms (endocrine and nonendocrine in nature) was 80 months for all patients and 80 months when the 7 patients with high-grade neuroendocrine carcinomas were excluded (Fig. 1C).
When the entire cohort of 70 patients was analyzed, the variables associated with poorer overall survival were histologic grade (P < .0001), maximal hepatic tumor size ≥6.5 cm (P = .001), primary tumor size ≥3 cm (P = .015), extrahepatic disease (P = .028), and the inability to undergo complete macroscopic resection (P = .036). On multivariate analysis, only histologic grade (median, 6 months for high grade, 63 months for intermediate grade, not reached for low grade; P = .004), maximal hepatic tumor size ≥6.5 cm (median, 49 months vs 137 months; P = .006), and primary tumor size ≥3 cm (median, 59 months vs 137 months; P = .015) retained an independent association with worse overall survival. Variables that were associated with poorer progression-free survival were histologic grade (P < .0001), inability to undergo complete macroscopic resection (P = .00018), maximal hepatic tumor size ≥6.5 cm (P = .005), and pancreatic primary (P = .047); histologic grade (median, 5 months for high grade, 14 months for intermediate grade, 26 months for low grade; P < .0001), inability to undergo complete resection (median, 17 months vs 26 months; P = .00,023), and maximal hepatic tumor size ≥6.5 cm (median, 12 months vs 30 months; P = .006) remained significant factors on multivariate analysis. Overall and progression-free survival estimates stratified by histologic grade are shown in Figure 2A,B, respectively. No significant associations with symptom-free survival were identified.
These analyses were repeated for the cohort of 63 patients with well‒differentiated (low and intermediate grade) neoplasms, excluding the 7 patients who had high-grade neuroendocrine carcinomas. Primary tumor size ≥3 cm (P = .012), pancreatic primary (P = .035), and maximal hepatic tumor size ≥6.5 cm (P = .0014) were associated with poorer overall survival; the association between tumor grade (low vs intermediate) and overall survival neared statistical significance (P = .054), as did the association between ability to undergo complete macroscopic resection and overall survival (P = .061). None of these variables remained significant on multivariate analysis (Table 2). Extent of resection (R0, R1, or R2; P = .0002), maximal tumor size ≥6.5 cm (P = .0039), tumor grade (low vs intermediate) (P = .0072), inability to undergo complete macroscopic resection (P = .036), and pancreatic primary (P = .012) were associated with poorer progression-free survival; the inability to undergo complete resection (P = .0003), pancreatic primary (P = .003), and tumor grade (P = .048) retained an association with poorer progression-free survival for patients with well‒differentiated neuroendocrine neoplasms on multivariate analysis (Table 3). Tumor grade was the only variable associated with symptom-free survival (measured as the time from surgery to the onset of any hormone- or tumor-related symptoms) for all patients with well‒differentiated neoplasms (P = .037) (data not shown). The prognostic variables that were associated by multivariate analysis with overall and progression-free survival for all patients and for patients with well‒differentiated neoplasms are summarized in Table 4.
|Variable||No. of patients||Median OS, mo||P|
|Age at hepatectomy||.43|
|Primary tumor size, cm|
|Primary tumor previously resected|
|Interval from diagnosis to hepatectomy||.18|
|Size of largest hepatic tumor, cm|
|No. of hepatic tumors||.93|
|Percentage of liver involvement by tumor||.27|
|No. of liver segments involved||.83|
|No. of liver segments resected||.30|
|Complete macroscopic resection|
|Extent of resection|
|Operative estimated blood loss||.10|
|Variable||No.||Median PFS, mo||P|
|Age at hepatectomy||.70|
|Primary tumor size, cm|
|Primary tumor resected|
|Interval from diagnosis to hepatectomy||.07|
|Size of largest hepatic tumor, cm|
|No. of hepatic tumors||.19|
|Percentage of liver involvement by tumor||.06|
|No. of liver segments involved||.29|
|No. of liver segments resected||.21|
|Complete macroscopic resection|
|Extent of resection|
|Operative estimated blood loss||.39|
|All neoplasms (n=70)||Well-differentiated neoplasms (n=63)|
|Overall survival||Overall survival|
|Maximum hepatic tumor size ≥6.5 cm||.006||None|
|Primary tumor size ≥3 cm||.015|
|Progression-free survival||Progression-free survival|
|Histologic grade||<.0001||Histologic grade||<.0001|
|Ability to perform complete macroscopic resection||.00023||Ability to perform complete macroscopic resection||.00023|
|Maximum hepatic tumor size||.006||Pancreatic primary||.003|
Understanding the role of partial hepatectomy in the management of hepatic neuroendocrine neoplasms is confounded by the rarity and occasionally indolent nature of this disease. Indeed, evidence that surgical resection may prolong survival for patients with neuroendocrine metastases can be deduced only indirectly from biased, retrospective comparisons with nonoperatively managed historic controls. Patients with unresected or unresectable metastases from well‒differentiated neuroendocrine neoplasms often can expect to live for years, with 5-year survival rates ranging from 30% to 40% and with a median survival between 20 months and 35 months.1, 3, 4, 12–16 In contrast, most contemporary series describe 5-year survival rates of up to 76% for patients who present with disease biology and medical comorbidities that permit surgical resection.2–7 We observed a 5-year overall survival likelihood of 61% and a median survival of 91 months among patients who underwent resection in conjunction with a multidisciplinary approach to adjuvant care. When we excluded the small but prognostically distinct cohort of 7 patients who had unexpected high-grade neuroendocrine carcinoma, the 5-year overall survival likelihood was 67%, and the median survival was 108 months. The median time to disease progression (defined either by the onset of a new recurrence in patients who underwent complete macroscopic resection or by the development of tumor growth in patients who underwent incomplete resection) was only 17 months. Although estimating disease progression rates can be subject to variations in the intensity of follow-up, this finding is in agreement with previous observations.5–7 Clearly, most patients who undergo resection of hepatic neuroendocrine metastases can expect a reasonable likelihood of durable survival, but it is unlikely that hepatic metastasectomy for this disease will be curative. In this study, we sought to more rigorously identify the variables associated with improved outcomes after partial hepatectomy.
Many surgeons have adopted the strategy of surgical cytoreduction, or “debulking,” in their management of hepatic neuroendocrine metastases.5–7, 16 Although patients who were able to undergo complete macroscopic resection demonstrated prolonged overall survival (median, 108 months) compared with patients who could not undergo complete macroscopic resection (median, 47 months), the association between completeness of resection and survival did not remain significant on multivariate analysis. Indeed, when the patients with high-grade neoplasms were excluded, we were unable to identify a strict statistical correlation between completeness of resection and overall survival (P = .061). An association may have been missed given the limited sample size of this study. It is worth noting that differences in the extent of disease burden within the liver (quantified by number of hepatic tumors, estimated percentage of hepatic parenchymal radiographic involvement, number of hepatic segments harboring tumor, or extent of hepatic resection) did not appear to be correlated with overall survival. Thus, our observations do not provide sufficient evidence to explicitly refute the rationale for cytoreduction for this disease. Among the 48 patients whose disease burden was amenable to complete macroscopic resection, histologic grade remained highly associated with overall survival (P < .00001) and progression-free survival (P = .001) (data not shown).
Current pathologic criteria segregate primary neuroendocrine neoplasms into “well‒differentiated neuroendocrine tumors” (defined by small tumor size and lack of local invasion and categorized, based on angioinvasion or mitotic activity, as either benign or of uncertain malignant potential), “well‒differentiated neuroendocrine carcinomas” (defined by demonstrable local invasion or distant metastasis), and “poorly differentiated neuroendocrine carcinomas” (defined by high mitotic activity).8 Although it may be useful for primary neuroendocrine neoplasms, this classification scheme does not permit the stratification of hepatic neuroendocrine metastases (which, by definition, are all well‒differentiated carcinomas or, rarely, poorly differentiated neuroendocrine carcinomas). Recent work with pancreatic neuroendocrine neoplasms has suggested that stratification into low, intermediate, and high grades based on mitotic activity and histologic necrosis can greatly enhance prognostic discrimination.9 Classification of a separate intermediate grade category already is well accepted for atypical carcinoid tumors of the lung.17 Our findings strongly suggest that pathologic classification of hepatic neuroendocrine neoplasms using the criteria of mitotic activity and histologic necrosis may provide a meaningful and accurate means of prognostication for patients with neuroendocrine neoplasms of the liver. Seven of 70 patients (10%) who underwent partial hepatectomy ultimately were diagnosed with high-grade neuroendocrine neoplasms, and their median overall survival was only 6 months; indeed, this can be considered an entirely separate disease entity.9, 17, 18 None of those 7 patients had known high-grade neoplasms preoperatively, and none presented with hormone-related symptoms. Although the patients with high-grade neoplasms were more likely to be men, white, and to harbor bilateral hepatic lesions, no surrogate variables could reliably predict the presence of high-grade histology without direct pathologic analysis. Theoretically, preoperative determination of tumor grade by core needle biopsy may enhance our ability to select patients properly for surgical intervention by excluding the small subset of patients with high-grade carcinomas, because they do not appear to benefit from surgical therapy. When this cohort was excluded from our analysis, none of the clinicopathologic variables examined were associated independently with differences in overall survival, but the difference between patients with low-grade and intermediate-grade neoplasms approached statistical significance (P = .054). However, tumor grade (intermediate vs low) retained a strong association with differences in progression-free survival for patients with well‒differentiated neoplasms.
All patients who presented with hormone-related symptoms experienced complete symptom resolution postoperatively, concordant with previous observations.2, 4, 6, 7 Symptom control was durable; the median time to symptomatic recurrence (of either the original symptoms or different symptoms) was 39 months. The median time from operation to the onset of any tumor-related symptoms (hormonal, or pain or obstructive symptoms related to tumor bulk) for the entire cohort of patients (regardless of whether or not they presented initially with hormonal symptoms) was 80 months. Limitations in sample size precluded any meaningful analysis of the variables associated with symptom-free survival for the 22 patients who presented with hormone symptoms, but tumor grade was the only variable associated with differences in symptom-free survival for the entire cohort of patients.
The conduct of the current study did not permit precise retrospective standardization or definition of the indications for and timing of the nonoperative therapies used. Nevertheless, these nonoperative treatments may have contributed to the survival outcomes of our patients. For example, the impact of incomplete resection on overall survival theoretically may have been more pronounced in the absence of postoperative hepatic arterial embolization. Given the array of contemporary treatment modalities in wide use today,2, 19–21 it would be difficult to measure the efficacy of surgical therapy alone in isolation from these multimodality therapies.
In conclusion, we have identified several variables that appear to have prognostic relevance for patients who undergo partial hepatectomy of hepatic neuroendocrine neoplasms. Whereas the median survival for patients who have favorable oncologic characteristics (eg, smaller primary or hepatic tumor size) exceeds 9 years, the median survival is typically less than half that for those with unfavorable features. Further investigation will be necessary to determine whether combined operative and nonoperative therapies for patients with the adverse oncologic variables identified in this study provide greater benefit than might be achieved by nonoperative therapies alone.