According to cadaver studies of gallbladder lymphatics, the lymphatic drainage of the hepatic pedicle can be divided into three pathways: (1) the cholecysto-retropancreatic pathway (right descending pathway), (2) the cholecysto-celiac pathway (left oblique pathway), and (3) The cholecysto-mesenteric pathway (mesenteric pathway) [8, 9]. These three pathways converge with the para-aortic lymph nodes (PALN) near the left renal vein. The lymphatic vessels of the right descending pathway (to the right of the hepatoduodenal ligament) first drain into the cystic node and through several lymphatic vessels eventually into the epiploic foramen LN (foramen of Winslow), located to the right of the common hepatic duct (Fig. 22.2a). The epiploic foramen LN drains into the superior retro-pancreaticoduodenal (Rouvière) LN and from there either directly or via posterior pancreaticoduodenal LNs into the PALN at the level of the left renal vein (Fig. 22.2a). The lymphatic vessels of the left oblique pathway run along the cystic artery and hepatic artery to reach the nodes around the celiac trunk via the common hepatic artery (CHA) LN (Fig. 22.2b). The mesenteric pathway is composed of many thin lymph vessels originating from the porta hepatis and the gallbladder neck. These lymph vessels drain into the “principal portal node”, located in front of the portal vein and at the confluence between portal vein and splenic vein. From this node, lymphatic vessels connect into LN surrounding the superior mesenteric artery (Fig. 22.2c). Several reports have argued that the right descending pathway is the dominant lymphatic pathway [10, 11].

In summary, the vast majority of hepatic lymphatic drainage is directed towards the HPLN.

The frequency of involvement of distant LN such as para-aortic, supradiaphragmatic and mediastitinal LN in patients with CLM is not well known. This is not surprising as lymph node dissection is rarely indicated and remains technically difficult. Interestingly the frequency of mediastinal LN involvement in colorectal cancer patients with lung metastasis was reported to be 12–33% [12, 13].

The analysis of HPLN involvement is further complicated by the fact that LN involvement can be either microsocpic or macroscopic. Studies looking only at macroscopic LN involvement, mostly by using the technique of “cherry picking”, do not take into account the number of microscopically involved LNs and therefore may grossly underestimate the frequency of HPLN involvement. A few studies on the frequency of HPLN involvement are reliable, as they use systematic lymph node dissection in consecutive patients with microscopic analysis [14, 15]. The study from Beaujon Hospital [15] analyzed 76 patients who underwent systematic HPLN dissection simultaneously with CLM resection. Macroscopic palpable LN involvement (1 cm in a diameter and/or firm on palpation) was suspected in 23 patients during surgery. Of these, only ten patients had microscopically proven metastatic disease in the LN, and five patients who had microscopically node-positive disease were misdiagnosed as node-negative disease during surgery. In a series of 114 patients undergoing CLM resection reported by Elias et al. [14], 22 patients had microscopic HPLN involvement on final pathology, but only eight patients (40%) were suspected to have positive nodal disease intraoperatively. Therefore, the sensitivity and positive predictive value (PPV) of intraoperative diagnosis was quite low (67% and 43% respectively) [15]. In this context, frozen section analysis of macroscopically suspected LN would have less value. Current evidence suggests that accurate diagnosis of HPLN and other perihepatic LN involvement in patients with CLM during surgery is very difficult to obtain.

Studies using postoperative results of LN involvement after systematic HPLN dissection in patients with CLM report a frequency of 8.9–20% (Table 22.1). Elias et al. [14] studied the pattern of LN involvement in patients with CLM according to the anatomic location of respective LN basins. They divided LN specimens into six groups: (1) antero-superior LN in the hepatic pedicle, (2) antero-inferior LN in the hepatic pedicle, (3) postero-superior LN in the hepatic pedicle, (4) postero-inferior LN in the hepatic pedicle, (5) common hepatic artery LN, and (6) celiac LN. They showed that the most frequently invaded group of LNs was the common hepatic artery LN basin, and that the positive LN location was highly variable even in patients with a single CLM. This result suggests the possibility of “skip metastasis” and a random pattern of LN involvement in CLM patients.

The ability to predict the presence of HPLN involvement in patients with CLM was evaluated by Jaeck et al. [16]. In this study the clinicopathological variables associated with HPLN involvement in 160 patients undergoing systematic HPLN dissection during liver resection for CLM were reviewed. Five relevant clinico-pathological factors were extracted by univariate analysis; (1) the presence of more than three CLM, (2) CLM located in segment 4 and / or segment 5, (3) synchronous CLM, (4) the presence of a resectable peritoneal deposit, and (5) poorly differentiated histology of CLM. In a similar study Elias et al. [17] evaluated 100 patients who underwent systematic HPLN dissection concomitantly with liver resection for CLM. They found that HPLN involvement was significantly correlated with (1) more than three metastases, (2) a greater than 15% tumor burden relative to total liver volume, and (3) a CEA level > 118 ng/L. In contrast, a study from the group at Beaujon Hospital with 76 patients who underwent systematic HPLN dissection simultaneously with liver resection for CLM (multiple CLM, 74%; bilobar CLM, 49%) did not find any correlation between HPLN involvement and any clinicopathological variables [15]. Based on these studies, it is reasonable to expect a higher chance of HPLN involvement in patients with higher tumor burden, multiple CLM, poor histology, and presence of extrahepatic disease.

Although very desirable, preoperative prediction of HPLN or other perihepatic LN involvement in patients with CLM remains very difficult. Current state-of-the-art imaging modalities such as 64-slice multi-detector computed tomography (MDCT), MRI, and PET scans lack sufficient diagnostic accuracy to be relied upon in clinical practice. The report from Beaujon Hospital team evaluated the ability of preoperative CT imaging to detect HPLN involvement compared to intraoperative assessment [15]. They defined preoperative LN involvement as greater than 1 cm in the short axis diameter, round-shaped, irregular contoured and/or heterogeneous LN in CT appearance. Of 15 patients with pathological proven LN involvement, only five patients fulfilled the preoperative imaging criteria for positive LN disease, resulting in a low sensitivity and positive predictive value (PPV) of 33% and 56% respectively. Positron emission tomography (PET) scan appears to be more accurate than CT for detection of HPLN involvement. A study from Memorial Sloan Kettering Cancer Center (MSKCC) evaluated 100 patients with metastatic hepatic malignancies who underwent liver resection and HPLN sampling [18]. In their study, CT scan had a high negative predictive value (NPV) of 95% and a low PPV of 39%, compared to PET scan with a NPV of 88% and a PPV of 100%. In patients demonstrating both a negative CT and PET scan, HPLN involvement was very unlikely (only one patient; 2.1%). However, systemic HPLN dissection was not performed, which limits the ability to assess the true denominator for occult metastases in this study and therefore prevents a true meaningful comparison of preoperative imaging and postoperative pathologic results. Although PET has a reasonable specificity for the presence of colorectal cancer and CLM (87%), it was found to be unreliable to detect LN metastases with a small tumor burden, leading to an overall low sensitivity of 37% for primary LN staging [19]. The authors experienced a CLM patient with HPLN involvement with a small tumor burden in whom both CT and PET scan were negative (Fig. 22.3a–f).

In conclusion, current imaging modalities cannot predict the presence of perihepatic LN involvement in patients with CLM with sufficient accuracy to be clinically valuable.

There have been few reports about the prognosis in patients with CLM and distant LN involvement. Two large studies analyzed the prognosis of CLM patients with resectable extrahepatic disease (EHD). An international multi-institutional database evaluated 1,629 patients with CLM who underwent resection for CLM [20]. Of these 1,629 patients, 171 (10.4%) had resectable EHD, and all malignant foci including CLM were removed. The common EHD sites were the lung (n = 62), HPLN (n = 41) and peritoneum (n = 25). PALN involvement was observed in 14 patients. The median survival and 5-year actual overall survival in patients without EHD who underwent CLM resection (n = 1,458) were 77 months and 57%, respectively, while the median overall survival was 39 months and 5-year overall survival rate was 26% in patients with successful resection of EHD. The prognosis of patients with HPLN involvement (median, 29 months; 5-year, 27%) was comparable to that of patients with lung metastasis (median, 46 months; 5-year, 33%) and peritoneal carcinomatosis (median, 32 months; 5-year, 26%). Patients with PALN involvement had a particularly poor survival (median, 13 months; 5-year, 7%). A second single institutional study evaluated 1,369 patients with CLM who underwent hepatic resection [21]. Of these, 127 (9%) underwent concomitant resection of EHD. The most common EHD site was the lung (n = 34) followed by HPLN (n = 27). Nine patients had ovarian metastasis. PALN and mediastinal LN involvement were observed in five patients and one patient respectively. The median and 5-year survival rate in 1,242 patients without EHD were 55 months and 49% respectively, compared to 36 months median survival and 26% 5-year survival in patients with EHD. Patients with HPLN involvement had poor survival (median, 26 months; 5-year, 12%), compared to 45 months median survival and 28% 5-year survival for patients with lung metastasis and 82 months median survival and 51% 5-year survival for patients with ovarian metastasis. The worst survival was seen in the five patients with PALN (median, 16 months). All five patients relapsed, and three of five patients died within 16 months of surgery.