During recent years, the intensity of preoperative systemic chemotherapy for patients with colorectal liver metastases (CLM) has increased significantly. Patients with unresectable metastatic disease frequently receive prolonged chemotherapy treatment in an attempt to convert them to resectability. With this approach, long-term survival can be achieved when liver resection becomes feasible after tumoral downsizing.1 In addition, neoadjuvant chemotherapy is applied for resectable liver metastases to facilitate margin-free resections, and this approach has shown recently to improve progression-free survival after hepatectomy.2,3

Our group and others have reported a relationship between the use of preoperative chemotherapy and histopathologic changes of the nontumoral liver with consequently an increased risk of perioperative morbidity.3–11 This mainly concerns the prolonged use of oxaliplatin and associated vascular lesions. However, close evaluation of direct relations between specific vascular lesions and postoperative outcome remains limited (Table 1).3,6–15 Only three studies have correlated specific chemotherapy-related vascular changes in the nontumoral liver with an increased intraoperative transfusion rate or longer hospital stay.8–10

Table 1 Review of publications evaluating the effect of preoperative chemotherapy and hepatic chemotoxicity on short-term outcome after resection of colorectal liver metastases
Full size table

Regenerative nodular hyperplasia (RNH) is considered the end-stage of vascular lesions induced by chemotherapy, but its effect on the outcome of hepatic resection for colorectal metastases remains unclear. However, with the increasing indications of preoperative chemotherapy, especially oxaliplatin, RNH is observed more frequently, necessitating an evaluation of its consequences. Furthermore, with the high incidence of recurrences observed in patients resected of CLM, repeat hepatectomies are increasingly performed.16–18 Knowledge concerning the consequences of RNH, as well as its potential to regress, is crucial in evaluating the risks of repeat surgery with the continuing administration of chemotherapy.

In this study, we evaluated the incidence of RNH and its impact on postoperative outcome in patients resected of CLM. In addition, we assessed the evolution of RNH by analyzing the pathological specimens of patients submitted to repeat hepatectomy.

Patients and Methods

Patients

From January 1990 to November 2006, 856 consecutive patients underwent partial hepatectomy for colorectal metastases at our institute; 771 (90%) of these patients were treated by preoperative chemotherapy, whereas 85 patients (10%) underwent hepatic resection without preoperative chemotherapy treatment. Of all 771 patients treated by preoperative chemotherapy, this study focused only on patients who received six or more cycles of first-line therapy. In addition, patients treated with preoperative intra-arterial chemotherapy were excluded.

Preoperative Chemotherapy

Chemotherapy was most often administered before surgery for patients with initially unresectable metastases. Technical unresectability was defined as the inability to completely resect all metastases while leaving at least 30% of normal liver parenchyma, resulting from a multinodular tumor distribution, large tumor size, or a close relationship with major vascular or biliary structures. The presence of extrahepatic metastases determined oncological unresectability. The rationale to administer preoperative chemotherapy to patients with upfront resectable metastases was to assess tumor chemoresponsiveness and to facilitate margin-negative resections.

Response to chemotherapy was evaluated in a multidisciplinary meeting with surgeons, oncologists, and radiologists, and surgery was only performed when the overall strategy could result in complete intra- and extrahepatic tumor clearance.

Liver Resection

The goal of liver surgery was to resect completely all detectable lesions. Detailed inspection, palpation, and intraoperative ultrasound of the liver were routinely performed in each patient. Local ablation, portal vein embolization, and two-stage hepatectomy were used as described before to increase the possibility of radical tumor resection.19–21 General and local hepatic complications occurring within 2 months after surgery were recorded and classified.22,23

Histopathologic Examination

Detailed histopathologic assessment of the nontumoral liver was performed by a single hepatobiliary pathologist, blinded for the information regarding preoperative chemotherapy and perioperative outcome. Liver tissue was analyzed according to a standard format previously described.8 Briefly, vascular lesions were categorized as sinusoidal alterations (vasodilatation and congestion), peliosis, hemorrhagic centrilobular necrosis (HCN), RNH, and veno-occlusive disease. The presence of macrovacuolar steatosis was graded as mild (<30% of hepatocytes), moderate (30–60%), or severe (>60%). Steatohepatitis included steatosis with signs of local inflammation and apoptotic hepatocytes. Fibrosis was divided into portal fibrosis, porto-portal fibrosis, septal fibrosis, and cirrhosis. Surgical necrosis also was noted.

Repeat Surgery

The development of recurrences was assessed by physical examination, serum CEA and CA 19.9 levels, and abdominal ultrasound at 4-month intervals after hepatectomy. CT imaging of the chest, abdomen, and pelvis was performed every 8 months. Repeat resection of intra- and/or extrahepatic recurrences was only considered if it could be macroscopically complete.17 For patients who underwent repeat liver surgery, histopathologic examination of the nontumoral liver was performed in a similar way as described above to evaluate the evolution of initial lesions.

Statistical Analysis

All statistical analyses were performed using SPSS® software version 13.0 (SPSS Inc., Chicago, IL). Categorical data were reported as the number of patients with percentages and compared by the χ2 test. For continuous data, reported as means ± standard deviation, the independent-samples t test was used to compare groups. Logistic regression was done to define independent predictive factors of hepatic morbidity as well as preoperative predictive factors of RNH. Factors with P ≤ 0.10 at univariate analysis were included. P values ≤0.05 were considered significant.

Results

Of all 771 consecutively resected patients treated by preoperative chemotherapy, 155 received six or more cycles of first-line therapy, delivered by intravenous route. Due to an insufficient amount of nontumoral liver parenchyma available for histopathological analysis, 9 patients were excluded, resulting in a cohort of 146 patients (Fig. 1).

Fig. 1
figure 1

Flowchart of patient selection

Full size image

Patient and Tumor Characteristics

Included patients had a median age of 61 (range, 34–79) years and 76% presented with synchronous liver metastases (Table 2). Most patients (54%) had >3 metastases at diagnosis with a median diameter of 40 (range, 6–160) mm. Metastases were located in both liver lobes in 70% of patients and 20 patients (14%) had concomitant extrahepatic disease.

Table 2 Characteristics of 146 included patients
Full size table

Preoperative Chemotherapy

Chemotherapy was indicated for initially unresectable metastases in the majority of patients (72%). Unresectability was related to multinodular disease (59%), large tumor size (29%), close vascular relationship (10%), and extrahepatic disease (3%). The remaining 28% of patients received preoperative chemotherapy for resectable disease. The median number of administered cycles for the total group was 8 (range, 6–21) and chemotherapy delivery was chronomodulated in 41% of patients.24 Twenty-four patients (16%) received 5-fluorouracil (5-FU) and leucovorin (LV) alone (9.0 ± 2.0 cycles), 92 patients (63%) had 5-FU/LV and oxaliplatin (8.6 ± 2.8 cycles), 18 patients (12%) had 5-FU/LV and irinotecan (8.9 ± 3.0 cycles), and 12 patients (8%) were treated by 5-FU/LV, oxaliplatin, and irinotecan (9.6 ± 3.6 cycles). The number of chemotherapy cycles did not differ between different regimens (P = 0.70).

Hepatectomy Characteristics

Major hepatectomies (≥3 segments) were performed in 50% of patients (Table 2). Red blood cell transfusions were required in 41% of patients, of whom 94% needed more than 1 unit of blood. Postoperative morbidity occurred in 43% of patients and one patient (1%) died within 60 days after surgery. Hepatic complications were classified as grade III or IV complications in 34% of patients. Median duration of hospital stay was 11 (range, 6–42) days.

Nontumoral Liver Parenchyma

Vascular liver lesions constituted the most frequent type of histopathological lesion and were present in 82 patients (56%; Table 2). Peliosis was most often observed (31%). RNH occurred in 22 of 146 patients (15%) and was more frequent than sinusoidal alterations (11%; Fig. 2). Of note, steatohepatitis occurred in only one patient (1%).

Fig. 2
figure 2

Example of regenerative nodular hyperplasia. Nodules of hyperplastic hepatocytes replace the normal liver parenchyma and are surrounded by atrophic plates without evidence of fibrosis (note the hemorrhagic changes close to atrophic plates). a Gordon and Sweet stain (×20); b Hematoxylin-eosin stain (×10); c Picrosirius stain (×20); d Picrosirius stain (×10)

Full size image

RNH Versus Non-RNH Patients

Patients with RNH more often presented with >3 metastases at diagnosis compared with patients without RNH (78 vs. 50%; P = 0.03; Table 3). Twenty RNH patients (91%) preoperatively received 5-FU/LV and oxaliplatin (9.2 ± 2.6 cycles). The two remaining patients were treated by 5-FU/LV and irinotecan (12 cycles; N = 1) and 5-FU/LV, oxaliplatin, and irinotecan (6 cycles; N = 1). Chemotherapy was chronomodulated in six patients (27%; all oxaliplatin). RNH occurred in 22% of patients treated by oxaliplatin compared with 4% of oxaliplatin-naïve patients (P = 0.003). The number of chemotherapy cycles was not increased in RNH patients compared with the control group (9.1 ± 2.7 vs. 8.8 ± 2.8; P = 0.55).

Table 3 Characteristics of patients with and without RNH
Full size table

RNH patients had lower platelet counts at hospital admission (≤150 × 103/μL: 48 vs. 17%; P = 0.002). Mean alkaline phosphatase, gamma-glutamyltransferase (GGT), and total bilirubin levels before surgery were higher in RNH patients (Table 3).

Major hepatectomies were performed in a similar percentage of patients with and without RNH (55 vs. 49%, respectively; P = 0.64; Table 3). None of the RNH patients died within 60 days postoperatively. However, hepatic complications occurred in 50% of RNH patients compared with 29% of patients without RNH (P = 0.05). This difference was mainly caused by an increased incidence of biliary leaks (27 vs. 0%).

Uni- and Multivariate Analysis of Hepatic Morbidity

Seven factors, including RNH, were associated with hepatic morbidity at univariate analysis (Table 4). However, only four factors were independent predictors at multivariate analysis: a preoperative platelet count of <150 × 103/μL, major hepatectomy, two-stage hepatectomy, and intraoperative red blood cell transfusion.

Table 4 Univariate and multivariate analysis of hepatic morbidity
Full size table

Predictive Factors of RNH

Multivariate logistic regression analysis identified elevated preoperative GGT (>80 U/L; >1N) and total bilirubin levels (>15 μmol/L; >1N) as independent factors predictive for the presence of RNH. Risk ratios were 6.6 (95% confidence interval (CI), 2–21.4) for GGT (P = 0.002) and 3.3 (95% CI, 1.1–10.0) for total bilirubin (P = 0.04).

Evolution of RNH Within Time

Fifteen of 82 patients (18%) with vascular changes of the nontumoral liver at first hepatectomy underwent repeat liver surgery. This included 2 of 22 patients (9%) with RNH at first hepatectomy.

RNH was replaced by HCN at second hepatectomy in both patients following interruption of oxaliplatin and subsequent treatment with irinotecan. These patients received 11 and 12 cycles of irinotecan-based chemotherapy between both hepatectomies, respectively. No new cases of RNH were found at repeat hepatectomy in the remaining cases.

Discussion

Although previous reports have correlated preoperative chemotherapy for CLM with increased postoperative complications, evidence for a direct relation between specific nontumoral liver lesions and postoperative morbidity remains preliminary.7–10,25,26 With the increasing use of preoperative chemotherapy, especially oxaliplatin, it is nevertheless important to know the incidence and impact of different vascular lesions on postoperative outcome and to know how these lesions can be predicted to adjust patient monitoring and to identify patients at risk of increased morbidity.

Our present study shows that RNH may occur in 15% of patients treated with preoperative chemotherapy. RNH is associated with increased hepatic morbidity and occurs most frequently in patients receiving oxaliplatin. Interestingly, its presence can be predicted preoperatively by elevated levels of GGT and total bilirubin.

The fact that RNH was related with increased postoperative hepatic morbidity was an important finding of our study. However, only a preoperative platelet count of <150 × 103/μL, major hepatectomy, two-stage hepatectomy, and intraoperative red blood cell transfusion were independent predictors of hepatic morbidity at multivariate analysis in the total study population. Major hepatectomy, two-stage hepatectomy, and intraoperative red blood cell transfusions were equally distributed between RNH and non-RNH patients. However, RNH patients had relatively low platelet counts compared with non-RNH patients. We may assume that a low platelet count was related to splenomegaly due to portal hypertension caused by RNH, with subsequent platelet trapping. These results all strengthen the association of RNH with increased hepatic morbidity observed in our study.

In a recent study, sinusoidal liver injury was related with increased morbidity after major hepatectomy for CLM after preoperative chemotherapy.10 Our inclusion of both minor and major hepatectomies confirms the importance of recognizing RNH in all patients scheduled for hepatectomy after preoperative chemotherapy treatment. Furthermore, our result was independent of the number of chemotherapy cycles.

Interestingly, we identified preoperative elevated levels of GGT and total bilirubin as predictive factors of RNH. A recent study also found that high levels of GGT predicted the presence of sinusoidal lesions.27 Surprisingly, mean ICG-R15 values, known to be more sensitive and reliable for hepatic injury, were not altered in our patients with RNH. For patients at risk for RNH, efforts should be made to reduce the risks of liver surgery. Techniques, such as portal vein embolization and two-stage hepatectomy, may be helpful to spare the highest amount of liver parenchyma as possible, thereby maximizing the chances of an uneventful postoperative course.

In relation with the increased risk of hepatic morbidity and the enlarging number of patients who undergo repeat hepatectomy with perioperative chemotherapy, it is important to consider the evolution of RNH within time. RNH may have deleterious long-term consequences related to the development of portal hypertension. One case study reported the development of RNH and portal hypertension in three patients treated with oxaliplatin that finally contraindicated curative liver surgery.28 Recently, the development of portal hypertension in patients with RNH with deleterious postoperative complications and even death was reported by another group.29 Other reports on RNH as a result of preoperative chemotherapy are rare.30 When we evaluated the evolution of vascular lesions in patients that underwent repeat hepatectomy, previously diagnosed RNH was replaced by HCN in two patients. Because RNH is distributed throughout the liver in a regular pattern, sample variation is unlikely to cause the absence of RNH at subsequent hepatectomies.31 Furthermore, all nontumoral liver specimens were evaluated by the same hepatobiliary pathologist. The natural history of RNH remains largely unknown.28 However, because it is a noncirrhotic liver disease without fibrosis, RNH can theoretically regress, as was demonstrated in our study.27

Interestingly, in both patients in whom RNH disappeared, oxaliplatin was stopped and irinotecan was administered before the second hepatectomy. This may suggest that irinotecan may be a good alternative of oxaliplatin to treat these patients. Previously, RNH had already been associated with the use of oxaliplatin.4 Recently, different authors have suggested a protective effect of bevacizumab on the development of vascular toxicity.32–34 Therefore, its addition to conventional chemotherapy may reduce the risk of RNH and associated morbidity. However, this issue lies beyond the scope of the present study and needs further evaluation. Conclusions on the evolution of vascular lesions other than RNH into less or more severe types at repeat hepatectomy are difficult because of their irregular distribution throughout the liver with the subsequent risk of sample variation.

Our study represents a selected patient group that received only one line of chemotherapy. By this way, we were able to correlate RNH with different chemotherapy regimens most accurately. However, with the large amount of patients receiving multiple chemotherapy regimens before surgery, RNH may be even more frequent in daily practice. The potential negative effect of portal hypertension related to RNH on patient outcome should not be underestimated.

A final interesting remark of our study is that we observed only one patient with steatohepatitis, who received oxaliplatin before hepatectomy. Previous large series have associated steatohepatitis mainly with irinotecan, one of whom even found that steatohepatitis was related with an increased 90-day mortality rate.7,14 The low incidence of obese patients and patients with diabetes probably is one of the reasons for the low frequency of steatohepatitis in our current study. The precise causes and consequences of this entity should nevertheless be investigated more extensively.

Conclusions

An increasing number of patients with CLM currently receive oxaliplatin-based chemotherapy, including adjuvant treatment after stage III colon cancer, induction therapy to convert extensive metastases to resectability, or perioperative treatment in patients with resectable metastases.1,3,35 RNH may occur in one of five patients, with an increased risk of postoperative morbidity after hepatectomy. Elevated serum GGT and bilirubin are useful markers to detect RNH that does not contraindicate hepatic resection. Clinical recommendations regarding preoperative chemotherapy treatment based on these results should be evaluated further, taking into account the availability and consequences of new biological agents.