Keywords

1 Hepatocellular Carcinoma (HCC)

Hepatocellular carcinoma (HCC) is the seventh most common cancer in the world and its incidence is increasing [1, 2]. The presentation of patients with HCC is broad, both in terms of tumor burden and extent of liver dysfunction. Tumor burden may range from small and solitary HCC, to multinodular and metastatic disease; the degree of liver dysfunction can also be extensive with cirrhosis, or be absent without any evidence of liver function compromise. The management of HCC must be directed to patients anywhere along this spectrum, and includes systemic chemotherapy, non-resection local therapies (NRLT), liverl resection (LR), and liver transplant (LT). Given the broad spectrum of presentation, combined with multiple emerging treatment modalities, expectedly, there are CER dilemmas.

1.1 Unresectable HCC: The Role of Local Therapies

Unfortunately, many patients with HCC present with disease that is unresectable secondary either to advanced stage, or with evidence of liver dysfunction that cannot tolerate resection. For those patients with advanced HCC that is unresectable, less invasive therapies like NRLT have been incorporated into the management of HCC. The common examples of NRLTs include percutaneous ethanol injection (PEI), radiofrequency ablation (RFA), transarterial chemoembolization (TACE), drug-eluting bead transarterial chemoembilzation (DEB-TACE), and transarterial radioembolization (TARE) with yttrium-90 (Y90). Over the past 20 years, each type of these therapies has been studied and applied to patients with varying levels of disease burden and liver dysfunction. Recently, the comparative data supporting the use of these less-invasive therapies in unresectable HCC has grown rapidly.

Prior to the 1990s, there was no evidence-based algorithm for applying local therapies in HCC and most guidelines were based on small retrospective reviews [3, 4]. This was especially true for unresectable patients with a low burden of disease, but who had evidence of significant cirrhosis. In 2003, Lencioni et al. [5] were one of the first groups to examine the role of NRLT in a prospective randomized study involving patients with HCC deemed to be not appropriate for LT or LR. In a sample of 102 patients, all patients with cirrhosis and single HCC <5 cm or three HCCs each <3 cm were randomized to PEI or RFA. At 2 years, the authors noted that patients treated with RFA had a trend toward improved overall survival and significantly better recurrence-free survival compared with patients treated with PEI (98 % vs. 88 %, p = 0.138, and 96 % vs. 62 %, RR = 0.17, p = 0.002, respectively). On multivariate analysis, RFA remained an independent prognostic factor associated with an improvement in local recurrence-free survival (RR 0.20, p = 0.015). Lin et al. [6] reported similar trends in a larger prospective trial with longer follow-up in which the authors noted improvements in both overall survival and recurrence-free survival with RFA. At 3 years, overall survival was 74 % versus 51 % (p = 0.31), with recurrence-free survival (43 % vs. 21 %; p = 0.038) also favoring the RFA versus PEI group. While overall survival was equivocal, the aggregate data seemed to support RFA over PEI for small HCC in terms of recurrence-free survival.

Some patients will present with HCC that has progressed to a more advanced stage where ablation cannot be utilized. In this scenario, the HCC typically has reached a larger size and transitioned to receive the majority of its blood flow from the hepatic artery [7]. With the HCC being larger in size, treatments like RFA and PEI do not have the same efficacy as when the lesion is smaller [8, 9]. As such, alternative NRLTs that utilize the vascular supply of the HCC to deliver therapy have been incorporated into the management of more advanced HCC. These intra-arterial therapies include bland embolization, embolization with chemotherapy (TACE), or embolization with drug-eluting beads with chemotherapy (DEB-TACE).

There are a number of CER issues relating to intra-arterial therapy of HCC, including but not limited to: (1) what type of chemotherapy (if any) should be given with TAE, (2) is there a role for DEB, (3) what agent should be used for embolization, (4) how many treatments or sessions should be offered, and (5) which patients will benefit from this type of therapy. Despite the many questions around the evolving treatment modalities for advanced HCC, some data do exist to guide our current understanding for the role of intra-arterial therapies for more advanced HCC. One early study, performed by the D’Etude [10] evaluated the effect of TACE on unresectable, larger HCC compared with conservative treatment/best supportive care. The authors noted that TACE reduced tumor growth (decreased >50, 16 % vs. 5 %, p = 0.001) and decreased serum AFP (decreased >50, 23 % vs. 8 %, p = 0.001). The effect of TACE on overall survival was not pronounced, with 4-year overall survival of 12 % versus 15 % (p = 0.13). Other studies have explored the role of TACE in unresectable HCC and similarly failed to demonstrate a dramatic improvement in survival with intra-arterial therapy [11–14]. In a separate study, however, Llovet et al. [15] demonstrated that TACE did indeed lead to a survival benefit. In this prospective study, patients with advanced HCC (i.e. lesions not amenable to resection or transplantation) were randomized to TACE, bland embolization, or conservative therapy/best supportive care. The investigators noted an overall survival benefit for TACE over bland embolization and conservative therapy (2-year overall survival: 63 % vs. 50 % vs. 27 %, respectively; p = 0.009). On multivariate analysis, TACE was the only variable independently associated with survival (OR 0.45, p = 0.02). The authors attributed this improvement to strict patient selection, gelfoam as their embolization agent, and doxorubicin as the chemotherapeutic agent. The trial was stopped early so that patients in this setting could receive TACE. Based on these data, TACE is now part of the standard therapeutic armamentarium for patients with advanced HCC [16–18].

1.2 Imaging and Tumor Response After Local Therapies

An area that has evolved dramatically both in terms of technological advancement and CER has been the adoption of standardized and objective radiological response criteria after NRLT. Radiologic response to local therapies is critical to the management of HCC as it may be a surrogate marker for survival [19]. The two earliest suggested recommendations for standardization of objective response to NRLT were the World Health Organization (WHO) criteria and the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines [20, 21]. Both of these guidelines were based on tumor response being correlated with changes in tumor size. However, treatment with these types of NRLT of HCC often results in change in tumor vascularity and viability, but not necessarily changes in tumor size. As such, the WHO and RECIST criteria have been criticized as being limited and unreliable. Subsequently, the European Association for the Study of the Liver (EASL) and the American Association for the Study of Liver Diseases (AASLD) proposed additional guidelines to assess tumor response following NRLT [22, 23]. These criteria specifically took into account tumor necrosis by examining the reduction in viable tumor area using contrast-enhanced radiologic imaging. Viable tumor was defined as the part of the tumor that took up contrast in the arterial phase, while the role of overall tumor size was made a secondary consideration in the assessment of the tumor response [23]. Another proposed set of criteria to assess response include the modified RECIST (mRECIST) criteria, which bases assessment of overall response on target lesions characteristics noted on contrast enhanced imaging, non-target lesions response, and the presence or absence of new lesions. The field of imaging assessment for tumor response after NRLT will continue to evolve as treatment and radiological modalities improve and will need to be a topic of future CER.

1.3 Resectable HCC: Non-resection Local Therapies Versus Resection

Patients with HCC may present with early stage disease/resectable lesion and optimal liver function with minimal to no comorbidities. For these patients, LR represents a potential therapeutic option. The long-term outcome with certain NRLT, such as ablation that can spare higher risk surgical patients from potential perioperative complications, is not well-defined and represents a subject of CER interest [24, 25]. Chen et al. [24] compared LR versus RFA for patients with a solitary small HCC. In this prospective series, 180 patients were randomized to RFA or LR. Inclusion criteria for the study were solitary HCC <5 cm in diameter, no vascular involvement, no evidence of liver dysfunction, and patients had to be suitable for either LR or RFA. At 4 years, overall survival was equivalent among patients undergoing RFA or LR at 65.9 % versus 64.0 %, respectively. In terms of disease-free survival, results were also similar with 4-year recurrence-free survival being 48.2 % versus 51.6 %, respectively. The authors concluded that RFA and LR were equally effective in the treatment of solitary and small HCC, with RFA being associated with decreased morbidity. Huang et al. [25] also prospectively examined the issue of LR versus RFA, but with expanded guidelines and came to different conclusions. In this study, patients with 3 lesions <3 cm or one lesion <5 cm were included. LR had improved survival and decreased recurrence over RFA. The 5-year overall survival was 54.8 % versus 75.7 % for RFA and LR, respectively (p = 0.001). In terms of disease-free survival, the same trend of better outcomes with LR over RFA was observed (5-year disease-free survival: 51.3 % vs. 28.7 %, respectively; p = 0.024). Direct comparisons of the different outcomes in the Chang and Huang studies is difficult as the studies varied in their inclusion criteria. Of note, in the Huang study, patients were not blinded to their treatment plan and 7 patients chose LR over RFA. In addition, the tumor size of the HCC was different between groups and the rate of loss of follow-up between groups was higher in the LR group (15.6 % vs. 6.1 %, p < 0.05). The comparison of these studies represents a key component to CER—understanding differences in study design, inclusion/exclusion criteria, as well as recognizing limitations in analysis. Furthermore, applying this CER perspective is necessary as more treatment modalities become available, e.g. the role of microwave versus RFA [26].

1.4 Resection Versus Transplant

Another source of treatment of CER interest is the debate over when to offer LR versus LT for early HCC. Theoretically, for most patients with HCC, LT represents the best treatment for survival because it removes both the tumor and underlying liver disease. There are, however, obstacles preventing LT from being offered to all patients with HCC including a limited availability of donor organs [27]. Historically in the 1980s and 1990s, the broad criteria utilized in organ allocation for patients with HCC led to a variety of outcomes for patients transplanted with HCC. In fact, the 90-day mortality, tumor recurrence, and long-term survival were not equivalent among all patients who were transplanted and some results were actually quite poor [28]. As such, attempts were made to identify the specific subset of patients who would benefit the most from LT [29]. In the seminal report by Mazzaferro et al. [30] the authors reported on a subset of patients with specific HCC characteristics who were proposed as a select patient population who would benefit from LT. This so-called “Milan criteria” included patients with one lesion smaller than 5 cm, up to 3 lesions smaller than 3 cm, no extra-hepatic manifestations, and no vascular invasion [30]. Since the adoption of the Milan criteria, additional guidelines have been suggested to broaden the inclusion of patients eligible for LT [31, 32]. These proposed guidelines include increasing the acceptable tumor size for transplantation. The ideal patient population to benefit from LT is still evolving and only with persistent CER re-assessment will refinements in the allocation system be possible.

Despite the identification of patient populations who benefit from LT, there are still limitations of donor organ availability, making application of LT to all patients with HCC not feasible. In turn, LR is a feasible alternative for some patients. Improvements in patient selection and perioperative management have made LR safe and relatively effective. While patients with end-stage liver disease and early stage tumors are most appropriate for LT, patients with compensated liver disease and early stage tumors can be appropriate for LT or LR. Choosing LT or LR for patients with compensated cirrhosis and early stage HCC remains challenging and often debated. In a retrospective review, Margarit et al. [33] examined the issue of when to offer LT versus LR for patients with early stage HCC and compensated cirrhosis. In this study, the authors reviewed patients with a single tumor <5 cm and Child’s class A liver disease and noted 10-year disease-free survival was worse after LR versus LT (18 % vs. 56 %, p = 0.001), with mean disease-free survival of 52 months versus 86 months, respectively (p = 0.04). Only 2.7 % in the LT group had local, hepatic recurrence versus 48.6 % in the LR cohort (p = 0.001). In terms of overall survival, there were no differences between the two groups (46 % vs. 36 %, LR vs. LT, p = 0.3) (Fig. 1a). Other studies have examined the same topic, with larger cohorts and intention-to-treat analyses [34–37] (Table 1). For example, Bellavance et al. [38] reported on 245 patients who underwent hepatic resection and 134 patients who underwent liver transplantation for early stage HCC. All patients had well-compensated cirrhosis. Compared with transplantation, patients undergoing resection had larger tumors and a higher incidence of microscopic vascular invasion. Transplantation was associated with better 5-year disease-free and overall survival compared with resection (Fig. 1b). Hepatitis status, presence of microscopic vascular invasion, and tumor size were predictors for recurrence, while the presence of microscopic vascular invasion and tumor size conferred an increased risk of death. The disease-free survival advantage with transplantation was more pronounced in hepatitis C patients compared with non-hepatitis and hepatitis B patients. The overall survival advantage with transplantation persisted in cases of solitary lesions ≤3 cm, but was attenuated in patients with a MELD score ≤8.

Fig. 1
figure 1

a, b Actuarial patient survival after liver resection and liver transplant. Used with permission [33, 38]

Full size image
Table 1 Summary of studies examining liver resection versus liver transplant for early hepatocellular cancer
Full size table

Final consensus on the comparative debate between LR and LT for early stage HCC with compensated liver disease remains lacking. While disease-free survival is clearly better among patients undergoing LT, the relative overall survival benefit of LT over LR remains ill defined. While LT has benefits over LR, it remains unclear whether patients who recur following LR can be salvaged with LT and experience the same long-term survival [39–41]. Prospective, randomized studies taking into account tumor size, multifocality, waitlist time and organ availability, and comorbidities, with appropriate long-term follow-up are needed to better address the LR versus LT debate.

2 Colorectal Liver Metastases

In the United States, colorectal metastases to the liver (CRLM) are  probably the most common secondary malignancy involving the liver [42]. Approximately 140,000 Americans are diagnosed with colon cancer annually, with more than half of these patients eventually developing metastases [43]. Most of these metastases are found in the liver, and the presentation may vary from disease that is isolated to the liver and resectable to disease with tumor burden that is extensive and unresectable [44]. Given the heterogeneity of this patient population with metastatic disease, combined with developments in NRLTs and a paucity of prospective data, numerous CER issues have arisen in the surgical oncology literature.

2.1 Role of Loco-regional Therapies in Patients with Unresectable Disease

Unfortunately, most patients with CRLM have unresectable disease [45]. There are numerous reasons why a patient may be unresectable and not be an appropriate candidate for LR, including multiple small tumors, vascular involvement of the tumor, a small future liver remnant (FLR), medical comorbidities, or extra-hepatic disease. Typically patients with unresectable disease are treated with systemic therapy, with an associated median survival of 2 years and 5-year survival around 10–15 % [46, 47]. For those patients with liver predominant or liver only disease that is unresectable, local therapies may have a possible therapeutic role over systemic therapy. These options include RFA, TARE with Yttrium-90 microspheres, TACE with irinotecan eluting beads (DEBIRI), and hepatic artery infusion (HAI) pumps [48, 49].

The role for RFA has been examined frequently as one of the more common alterative or adjunct therapeutic options for patients with unresectable advanced disease [50–55]. Siperstein et al. [56] examined a retrospective cohort of patients with unresectable CRLMs that were treated with RFA. A unique strength to this study was its extensive 10-year follow-up. In the cohort of 234 patients—all of whom were treated with RFA—the 5-year survival was 18.4 %, which the authors noted was better than the 5-year survival or 10 % for historical controls treated with systemic therapy alone [57]. In a separate study, as part of the Intergroup 40,004 trial contrast, Ruers et al. [58] reported their findings for patients with unresectable CRLM. In this prospective study, the investigators compared systemic FOLFOX-based chemotherapy combined with RFA versus systemic therapy alone for patients with advanced CRLM. The authors noted that 30-month overall survival was similar: 61.7 % in the combined arm versus 57.6 % in the systemic therapy alone arm (p = NS). Median overall survival was also similar: 40.5 months in the systemic arm versus 45.3 months in the combined treatment arm (p = NS). While overall survival was the same in both groups, 3-year progression free survival was worse in the systemic alone arm (10.6 % vs. 27.6 %, HR = 0.63, p = 0.025). Therefore, the authors concluded that overall survival for patients with unresectable disease treated with systemic therapy alone versus ablation plus systemic therapy was similar, while progression free survival was improved with the use of ablation.

Other local therapies have also shown promise, however the data supporting their use is not as robust and therefore is a focus on CER. One such therapy is the HAI pump. First proposed in 1984, the role for HAI has been controversial, and its efficacy has been compared to systemic therapies in multiple prospective studies [59–62]. Many of these studies, however, have been criticized due to low sample size, patient cross-over, and the single-center nature of the trials [63]. In an effort to address these issues, Kemeny et al. [63] prospectively compared HAI pump therapy with systemic therapy in a large multi-institutional trial that did not allow cross-over. The authors reported that overall survival was improved for patients who received HAI versus systemic chemotherapy (median, 24.4 months vs. 20 months, p = 0.0034). Additionally, response rates were higher (47 % vs. 24 %, p = 0.012) and time to hepatic progression was longer (9.8 months vs. 7.3 months, p = 0.034) with HAI therapy. While these data were promising, other studies have challenged the survival benefit of HAI. In a meta-analysis of 10 randomized controlled trials performed comparing HAI with systemic chemotherapy, Mocellin et al. recently suggested that there is no evidence supporting the use of HAI. In the pooled analysis, while tumor response rate was expectedly better in the HAI group (42.9 % vs. 18.4 %, RR 2.3, <0.001), overall survival was not different comparing HAI versus systemic therapy (15.9 months vs. 12.4 months, HR 0.9, p = 0.24, respectively). While HAI therapy may provide some benefit in the treatment of advanced colorectal liver metastasis, more CER is needed to determine the role for HAI.

Other newer local therapies such as TACE with irinotecan beads (DEBIRI) and TARE with Yttrium-90 (Y-90) have posed CER issues in the context of unresectable CRLM. There are emerging data for the use of transarterial DEBIRI in the treatment of unresectable liver metastasis [64–66]. Many studies, however, incorporate TACE or TARE only for patients who are refractory to systemic therapies. Martin et al. [66] reported 55 patients who had received prior systemic chemotherapy and who underwent DEBIRI treatment. In this series, response rates were 66 % at 6 months and 75 % at 12 months. TARE with Y-90 has also been investigated for patients refractory to chemotherapy [67, 68]. Cosimelli et al. [68] in a prospective multicenter phase II trial, evaluated the effect of TARE on patients who had failed previous oxaliplatin and ironotecan based chemotherapies. Based on RECIST criteria, 2 % had a complete response, 22 % a partial response, 24 % had stable disease, 44 % had progressive disease, and 8 % were non-evaluable. Because of these promising results, a phase III multicenter clinical trial, Efficacy Evaluation of TheraSphere following Failed First-Line Chemotherapy in Metastatic Colorectal Cancer (EPOCH) trial will soon open to elucidate the effect of TARE on not just response rates, but also overall survival. Final consensus on the optimal management of patients with unresectable CRLM is still developing, but progress is being made with emerging meta-analyses and prospective studies.

2.2 Ablation Versus Liver Resection

For the 10–25 % of patients with resectable CRLM, LR is the standard treatment approach with 5-year survival following surgery now approaching 60 % [42, 69, 70]. The role of ablation versus surgery among patients with CRLM potentially amenable to either therapy has been debated. The median overall survival associated with LR of CRLM reported in the literature ranges from 24 to 59 months whereas the data on survival following ablation are more limited [52, 71]. Several studies have sought to compare outcomes for patients who underwent ablation versus patients who underwent resection for CRLM [72–76]. Abdalla et al. [52] reported on 358 consecutive patients who underwent hepatic resection with or without RFA for CRLM. In this cohort, LR provided a significantly better overall 4-year survival over RFA alone, (65 % vs. 22 %, p < 0.001). Similarly, Hur et al. [77] noted a 5-year survival advantage for patients who underwent LR versus RFA (25.1 % for RFA vs. 50.0 % for LR). Based on the available data, it appears that patients managed with ablation have a worse outcome compared with patients who underwent hepatic resection (Table 2) [72, 74–76, 78].

Table 2 Summary of studies comparing RFA with resection for colorectal liver metastases
Full size table

The difference in the outcomes may, however, not be solely attributable to the type of therapy delivered (i.e. LR versus ablation), but also an issue of disparate underlying tumor biology among each patient population. Specifically, patients who undergo ablation as treatment for their CRLM often represent a distinct subgroup of patients with otherwise advanced disease who are not amenable to surgical extirpation [79]. In fact, many of the clinicopathologic features such as tumor size and number are often different in the group of patients receiving LR versus ablation. To achieve more comparable groups, subgroup analyses of patients undergoing either LR or ablation for CRLM have been performed, which have commonly been stratified by tumor number [78]. For example, Aloia et al. [78] examined a cohort of patients all of whom had only a solitary lesion. In this study, 150 patients treated with resection were compared with 30 patients treated with RFA. The authors reported that patients who underwent resection had a significantly better 5-year survival (resection: 71 % vs. RFA: 27 %; p < 0.001). However, patients managed with RFA likely had worse tumor biology as indicated by a higher proportion of patients with concomitant extrahepatic disease. In a separate study, Gleisner et al. [80] sought to examine how discordant clinicopathologic factors might play a crucial role in comparing patients who underwent resection versus ablation. Specifically, Gleisner et al. compared overall survival between patients who underwent resection with survival of patients who underwent RFA using three distinct statistical methods. The authors reported that patients managed with resection alone had an improved long-term overall survival compared with patients treated with resection plus ablation. The authors noted, however, that there were many differences in the clinicopathologic profile of each group. To examine the comparability of the baseline characteristics of the two treatment groups, Gleisner and colleagues utilized propensity score methodology. The authors noted that the aggregate distribution of the clinical and pathologic characteristics of patients undergoing resection alone versus RFA ± resection were markedly different and therefore direct comparisons of these groups may not be appropriate. The work of Gleisner and colleagues serves therefore to highlight the significant shortcomings of using retrospective data to compare outcomes following resection versus ablation in cohorts of patients who are very different and whose choice of treatment was undoubtedly based in part based of very different baseline characteristics.

In 2009, as part of an American Society of Clinical Oncology (ASCO) evidence-based review, Wong et al. [81] attempted to examine all data available at that time on ablation and LR for CRLM. The authors concluded that the available data were insufficient to form the basis of an evidence-based recommendation. Specifically, the authors noted that there was wide variability in 5-year survival (14–55 %) and local tumor recurrence (3.6–60 %) with ablation compared with LR. In turn, the investigators commented that the question of ablation versus LR could only be answered by a prospective, randomized trial. Such a trial, while ideal, would be challenging for a variety of reasons, most significantly, accrual would be required to be multi-institutional, and the ablation procedure itself would need to be standardized [45]. In an attempt to simulate such a trial, Khajanchee et al. [45] used a Markov and Monte Carlo analysis comparing RFA and LR. The authors reported that the model estimated 5-year survival among those patients who underwent LR over RFA alone to be 38.2 % versus 27.2 %, respectively. Five-year disease-free survival was also superior in the LR group (LR: 29.8 % vs. RFA: 15.5 %). While there are no prospective data comparing ablation with LR in the resectable population, from the limited data available, LR should remain the preferred approach with ablation being used as an adjunct second line therapy.

2.3 Systemic Therapy, Neoadjuvant Chemotherapy and the Disappearing Liver Metastasis

While local therapies such as ablation and resection are important treatment options for patients with CRLM, systemic chemotherapy plays a critical role in the multi-modal therapy of these patients. Systemic therapy has the potential of treating micrometastatic disease, evaluating tumor response, and downstaging unresectable tumors to resectability. The benefits of chemotherapy come with some possible consequences, including hepatotoxicity such as sinusoidal dilation, steatosis, or steatohepatitis [47, 57, 82, 83].

In an attempt identify which patients may benefit the most from systemic chemotherapy, several predictive models have been designed to identify patients at high risk for recurrence after hepatectomy. Fong et al. [42] were one of the first groups to propose a clinical risk score for predicting recurrence and survival. This study identified several prognostic factors for recurrence including: extra-hepatic disease, node-positive primary tumor, disease free interval from primary to metastases <12 months, CEA level >200, largest hepatic tumor >5 cm, and number of hepatic tumors >1. Similarly, Adam et al. [84] also created a prognostic model, examining initially unresectable CRLM that received chemotherapy and that were downstaged to resectability. In this study, the investigators identified a rectal primary, ≥3 CRLMs, maximum CRLM size of ≥10 cm, and CA19-9 >100 as independent factors of poor prognosis. Capussotti et al. [85] in 2007 suggested their own prognostic model. These authors identified patients with T4 primary colon cancers, metastases with infiltration of neighboring structures, and patients with more than three metastases as being potential indicators of poor prognosis and, in turn, may indicate a potential benefit from systemic chemotherapy. Despite these large retrospective series, there are still no specific consensus guidelines to indicate which patients with resectable CLRM should receive systemic chemotherapy either in the adjuvant or neoadjuvant setting [86].

Among the cohort of patients treated with neoadjuvant or preoperative systemic chemotherapy, there are several CER issues that remain debatable. Most data would suggest that those patients who have progressive disease on preoperative chemotherapy have a very poor prognosis [84]. Whether these patients should categorically be refused potential surgery even if the disease is technically still resectable remains controversial. Among patients who have a response to neoadjuvant/preoperative chemotherapy, surgery is typically performed with the goal of resecting all sites of disease. Up to 10–25 % of patients with CRLM who are treated with preoperative chemotherapy, however, will have a complete response with radiographic “disappearance” of some or all CRLM lesions in the liver [87, 88]. The so called “disappearing liver metastasis” (DLM) raises a number of CER issues.

From a radiologic perspective, there is no consensus regarding which imaging modality (CT, MRI, FDG-PET, or FDG-PET-CT) is most appropriate to determine whether the DLM is simply “missing” due to low sensitivity of the chosen imaging modality versus whether it has truly “disappeared.” Most medical oncologists and surgeons currently use CT in the treatment of patients with CRLM. The widespread use of dual phase helical CT is based on clinician familiarity and a high degree of reproducibility with excellent sensitivity and specificity up to 90 % when diagnosing CRLM [89, 90]. In the setting of a DLM, when the liver has been exposed to systemic chemotherapy—often many cycles—the background liver can appear darker with less contrast between the liver and any hypovascular metastases [91]. PET-CT has been considered as adjunct to CT alone, however PET-CT has limited sensitivity in its ability to detect lesions <1 cm and chemotherapy decreases hexokinase activity, thereby inhibiting glucose uptake for CRLM [92, 93]. Recently, there has been increasing data to suggest that MRI should be the imaging modality of choice in the setting of DLM. MRI, has increased sensitivity compared with CT, particularly in the setting of chemotherapy induced hepatic parenchymal changes (Fig. 2) [88]. In a recent meta-analysis, van Kessel et al. [94] compared various imaging modalities in the detection of CRLM after preoperative chemotherapy and found that the sensitivity of MRI was 85.7 % versus 69.9 % for CT, 54.5 % for PET, and 51.7 % for PET-CT. As such, MRI seems to be the imaging modality of choice for patients treated with preoperative chemotherapy—especially those with DLM. Future CER to understand better the role of different imaging modalities in treating patients with CRLM will be needed.

Fig. 2
figure 2

a Computed tomography (CT) image demonstrating colorectal liver metastases in segments II and VI (arrows) before systemic chemotherapy. b After 6 cycles of FOLFOX (folinic acid, 5-fluorouracil, oxaliplatin) therapy, CT showed that the lesion in segment II had ‘disappeared’, whereas the lesion in segment VI was significantly smaller and calcified. c Magnetic resonance imaging similarly identified the lesion in segment VI, but also demonstrated a residual 7-mm lesion in segment II. Used with permission [88]

Full size image

In addition to radiological issues in management of DLM, there is also is also a lack of consensus about the surgical management of DLM. As with other CER dilemmas, this primarily stems from the lack of reliable data. van Vledder et al. [87] attempted to examine the question of how to manage DLM using retrospective data; the authors noted that patients with untreated DLM had an increased local recurrence rate compared with patients who underwent LR of the DLM (p = 0.04). Despite these findings, the 1-, 3-, and 5-year overall survival was not different for patients undergoing LR versus those patients who had DLM left in situ (92.3 % vs. 93.8 %, 70.8 % vs. 63.5 %, 46.2 % vs. 63.5 % respectively) (Fig. 3a, b). The CER issues raised by the management of DLM were recently addressed by Bischof et al. [88]. In their review of DLM, the authors concluded that among patients who had a complete radiographic response, only 20–50 % had a durable long-term remission. In addition, among patients who had the DLM resected, residual tumor was present in 25–45 % of patients. Therefore, more CER is need to understand which patients need surgery for a DLM after receipt of preoperative chemotherapy.

Fig. 3
figure 3

a Kaplan–Meier curve of overall survival in patients with untreated DLM when compared to patients in whom all original disease sites were resected. b Kaplan–Meier curve showing overall survival in 99 patients with a complete or partial radiological response (RECIST) to preoperative chemotherapy stratified by the presence of untreated DLM. Used with permission [87]

Full size image

3 Neuroendocrine Liver Metastases

Neuroendocrine tumors (NETs) represent another important topic in the context of liver surgery and CER. NETs are of particular interest as these tumors are increasing in incidence and 40–95 % of cases are metastatic at diagnosis [95]. Treatment strategies for NETs once they have metastasized to the liver (NELM) are similar to those employed for the aforementioned liver tumors, and options include systemic chemotherapy, various NRLT, and LR. Also similar to other hepatic tumors, there is an absence of data from rigorous trials [96]. To further exacerbate the issue of reliable long-term data is the often indolent biologic behavior of these tumors compared to other liver tumors [97].

3.1 Cytoreductive Therapy—Liver Resection and Non-surgical Local Therapies

While the standard of care for NELM is LR, the data guiding this recommendation are surprisingly limited. In 2000, Chamberlain et al. [98] argued in presenting the results of their surgical series that LR improved survival. In this study, the authors demonstrated on multivariate analysis that LR prolonged 5-year survival versus NRLT (bland embolization) and best supportive care (76 % vs. 50 % vs. <25 %, respectively; p < 0.05). In a separate study, Sarmiento et al. [99] from the Mayo Clinic reported on an experience with 170 patients who underwent LR for NELM. The authors noted that overall 5- and 10-year survival were 61 and 35 %, respectively. Similarly, in a large multi-center study, Mayo et al. [100] reported excellent long-term results following surgery with 5- and 10-year survival of 74 and 51 %, respectively. As such, based on these data, surgical resection of NELM is widely utilized as it is believed to offer patients improved long-term survival. While most surgeons agree with an approach to resect patients with a low-disease burden, the role of surgical debulking of patients with larger disease-burdens is more controversial. For example, some groups have even suggested that patients with a high tumor burden may have a survival benefit after palliative debulking, as long as the majority (>75–80 %) of the liver disease can be removed [99, 101].

Despite the long-term survival associated with LR for NELM, recurrence following surgical management of NELM is almost universal. Specifically, in the largest retrospective review to date by Mayo et al. [100] the authors reported a 94 % recurrence at 5 years and 99 % recurrence rate at 10 years. Because of this remarkably high incidence of recurrence, there has been an increased interest in NRLT for NELM, including such intra-arterial therapies (IAT) as TACE and TARE. In a large, multi-institutional respective review, Mayo et al. [102] compared outcomes among patients with NELM based on treatment by LR versus IAT. Not surprisingly, the authors noted significant differences in the baseline characteristics of patients who underwent LR versus IAT, with the IAT group having more hormonally active tumors (48 % vs. 28 %, p < 0.001) and a larger hepatic tumor burden (>25 %: 76 % vs. 52 %, p < 0.001). The selection bias obviously calls into question any conclusions that can be drawn from retrospective comparisons of these two treatment modalities and highlights the CER challenges in answering this question. The authors did attempt to address the issue of selection bias by using propensity score matching. Propensity scoring provides a means to design and analyze a nonrandomized, retrospective dataset in an attempt to mimic some of the characteristics of a randomized controlled trial [103]. In the study, quintiles were created from their entire cohort with similar clinicopathologic characteristics and used in a matched analysis. With propensity matching, the authors noted that the analytic cohort comparing LR versus IAT groups now had much similar baseline characteristics. While LR was still associated with an improved survival over IAT, the difference was less pronounced (Fig. 4a, b). Furthermore, on stratified analyses, it was noted that symptomatic patients with a small burden of liver disease benefited the most from surgery. While symptomatic patients with a large burden of liver disease (>25 % hepatic tumor involvement) had improved median survival with LR over IAT (87 months vs. 51 months, p < 0.001, respectively), patients who were asymptomatic did not seemingly benefit from surgery resection (LR, 16.7 months vs. IAT, 18.5 months, p = 0.78). While propensity matching can assist with the comparison of groups with disparate baseline characteristics, more effective methodology and prospective trials will be necessary to answer better the CER question around which patients benefit from LR versus IAT.

Fig. 4
figure 4

a Histograms demonstrating the distribution of the propensity scores in the surgical and intra-arterial therapy (IAT) patient cohorts. The area of greatest overlap (quintile 3) corresponds to group of patients most likely have undergone either treatment based on baseline characteristics. b Overall survival of propensity-matched patients in quintile 3 stratified by receipt of surgery versus IAT. Used with permission [102]

Full size image

3.2 Role of Resecting the Small Bowel Primary in the Unresectable NELM Setting

An additional area of CER contention is the question whether to leave an asymptomatic small bowel (SB) primary NET in place in the setting of unresectable NELMs. For patients with SB-NET, this scenario is not uncommon, as 15–80 % of these primary tumors develop unresectable NELM [104]. The proposed goals of primary tumor resection are to provide relief from hormonal and local tumor-related symptoms (e.g., pain, perforation, bleeding, and obstruction), limit disease solely to the liver so that it may be treated with IAT, and potentially improve overall survival. Unfortunately, the data for the role of resection in this scenario is particularly sparse, and current management is based on personal experience and local practice patterns [95, 105–108]. Therefore, in an attempt to create evidence-based recommendation for this scenario, both meta-analyses and consensus panels have been used. Capurso et al. [109] recently performed a meta-analysis and examined the role of resecting SB-NET in the setting of unresectable NELM. The authors found that the only studies reported to date were solely retrospective and that the quality and type of data included in these small cohort studies did not meet the inclusion criteria for the meta-analysis (Table 3). In particular, Capuso et al. noted that some studies included patients with other primary tumor sites and the role of resection was also not appropriately analyzed in each study.

Table 3 Summary of studies examining resection or unresected primary midgut carcinoid tumors in patients with unresectable liver metastases
Full size table

Frilling et al. [96] also attempted to examine the role of primary tumor resection in a recent European-African Hepato-Pancreato-Biliary (E-AHPBA) consensus conference. Similar to the aforementioned meta-analysis, the authors concluded that many variables made a consensus statement impossible. The investigators noted that significant confounding factors included biases to operate on less advanced tumors, as well as a selection bias to operate on patients with better performance status—thereby making the actual benefit of the surgery itself impossible to discern. Because of the paucity of unbiased data, there are only weak evidence to recommend resection and more rigorous retrospective and prospective CER studies are needed.

4 Conclusion

Examining current treatment recommendations through the prism of CER is challenging and humbling, as it sheds light on areas that lack robust data and rigorous analysis. The management of liver malignancies is an ideal example of how CER has led to reliable treatments, but also where progress is urgently needed. When prospective, randomized controlled trials can be completed, these data will remain the gold standard for practice. However, in those scenarios where such trials are not feasible, clinicians must be cautious when adopting conclusions drawn from retrospective analyses.