World J Surg DOI 10.1007/s00268-015-3150-1
ORIGINAL SCIENTIFIC REPORT
Liver Resection for Advanced Intrahepatic Cholangiocarcinoma: A Cost-Utility Analysis Umberto Cillo1 • Gaya Spolverato2 • Alessandro Vitale1 • Aslam Ejaz2 Sara Lonardi3 • David Cosgrove4 • Timothy M. Pawlik2
•
Ó Socie´te´ Internationale de Chirurgie 2015
Abstract Background Data on cost-effectiveness and efficacy of hepatic resection (HR) for advanced intrahepatic cholangiocarcinoma (ICC) are lacking. We sought to estimate the cost-effectiveness of upfront HR resulting in an R1 resection (strategy A) relative to initial systemic chemotherapy (sCT) followed by possible curative HR (strategy B) for patients with advanced ICC. Methods A Markov model was developed using data from a systematic literature review. Three base cases were considered: (1) ICC [6 cm (2) ICC with vascular invasion (3) multi-focal ICC. A Monte Carlo simulation assessed outcomes including quality-adjusted life months (QALMs) and incremental cost-effectiveness ratio (ICER). Results The net health benefit (NHB) of strategy A versus strategy B was 1.4 QALMs for ICC [6 cm and 1.3 QALMs for ICC and vascular invasion; in contrast, there was a negative NHB for HR versus sCT for multi-focal ICC (-0.3 QALMs). In single nodule ICC [6 cm, the ICER of HR versus sCT was $22,482/quality-adjusted life years (QALY) and the ICER of HR versus sCT was $20,953/QALY for ICC with vascular invasion. In multi-focal ICC, the ICER of HR compared with sCT was $83,604/QALY. Patients with a higher American Society of Anesthesiologists score (coefficient 0.94), male sex (coefficient 0.43), low quality of life after sCT (coefficient -2.57) and T3 tumors (coefficient 0.53) had a better NHB for HR relative to sCT followed by potential surgery. Conclusions For patients with large ICC or ICC and vascular invasion, HR was more cost-effective than sCT. In contrast, HR was not associated with a positive NHB relative to sCT for patients with multi-focal ICC, and therefore these patients should be treated with sCT rather than HR.
Introduction
& Timothy M. Pawlik
[email protected] 1
Unita` di Chirurgia Epatobiliare e Trapianto Epatico, Azienda Ospedaliera-Universita` di Padova, Padua, Italy
2
Division of Surgical Oncology, Department of Surgery, The Johns Hopkins Hospital, Baltimore, MD, USA
3
Istituto Oncologico Veneto, Padua, Italy
4
Division of Medical Oncology, Department of Medicine, The Johns Hopkins Hospital, Baltimore, MD, USA
Intrahepatic cholangiocarcinoma (ICC) accounts for 20–25 % of all types of cholangiocarcinoma, which is the second most common primary liver malignancy after hepatocellular carcinoma [1–5]. The incidence of ICC increased from 3.2 per 1,000,000 in 1975 to 8.5 per 1,000,000 in 2000 and has continued to increase over the last decade [5–7]. Curative hepatic resection (HR) remains the most effective treatment for ICC with a 5-year survival ranging from 30 to 35 % [8]. Only about 20–40 % of patients with potentially operable disease are, however, offered surgical resection [9, 10]. Many patients with ICC
123
World J Surg
present with advanced disease and decisions about whether a patient is a candidate for surgery can therefore be challenging. In particular, some patients may present with a large ICC tumor that can be located in a difficult location requiring a technically complex operation with an associated increased risk of morbidity and an R1 margin [11]. Other patients with ICC will present with technically resectable disease, but have advanced tumor features such as multi-focal disease or vascular invasion, calling into question the oncological benefit of surgical resection [12]. When assessing whether HR is warranted in patients with advanced malignancies, the surgeon needs to consider both technical and oncological issues. For patients with large ICC, HR may result in an R1 resection in up to 24 % of patients [13]. In turn, R1 margin status is a strong independent predictor of worse long-term outcome. Specifically, Farges et al. reported that ICC patients who underwent an R1 resection or who had margins \5 mm had a worse survival compared with patients who had an R0 resection with wider margins [13]. Other studies have noted that patients with ICC characterized by multifocality or vascular invasion also do considerably worse than patients with earlier stage disease [14]. In the study by Nathan et al., the authors reported that patients with multi-focal disease or those patients with tumors characterized by vascular invasion had over a 1.4and 1.5-fold chance of death, respectively [12]. While the long-term survival of patients with advanced ICC who undergo HR, which may result in an R1 margin, is significantly worse than outcomes following surgery for early stage disease, alternative therapies for ICC are somewhat limited. For patients deemed to have inoperable ICC, systemic chemotherapy (sCT) is standard of care. Overall, there has been an improvement in the effectiveness of sCT from 5-fluorouracil (5FU) to gemcitabine-based regimens, with response rates increasing from 10–30 to 20–50 %, respectively [15–17]. More recently several studies have reported on the combination of gemcitabine with cisplatin, oxaliplatin, docetaxel, mitomycin C or 5-FU/leucovorin, with response rates ranging from 36 to 60 % [18–22]. Nonetheless the 5-year survival of patients with unresected ICC is only 5–10 %. As such, some surgeons advocate for resection of borderline, advanced ICC either with or without the use of neoadjuvant sCT [7]. For certain patients with large tumors at risk for an R1 resection, or those patients with multi-focal disease or vascular invasion, some surgeons proceed with ‘‘upfront’’ surgery, while other surgeons may choose to administer preoperative sCT in the hopes of facilitating a future operation. To the best of our knowledge, no previous study has evaluated these complex clinical scenarios involving patients with ICC. In particular, decision-making models that include co-factors such as anticipated quality of life (QoL), as well as long-term survival have not been undertaken for patients with advanced ICC. In addition to
123
assessing long-term survival outcomes, it is also important to relate the costs of care to the magnitude of the survival benefit. Whether the net health benefit (NHB) and costeffectiveness of offering HR to patients with advanced ICC is warranted remains poorly defined. As such, the objective of the current study was to estimate using Markov modeling the NHB and cost-effectiveness of upfront HR for patients with advanced ICC relative to initial sCT followed by possible subsequent curative hepatectomy.
Methods Definitions and endpoints The study focused on defining the NHB and cost-effectiveness of treating patients with advanced ICC with HR rather than sCT with possible subsequent liver resection (Fig. 1). We considered three base-case scenarios consisting of 61-years-old patients with (1) a large ([6 cm), solitary ICC with no vascular invasion, (2) uni-focal ICC with vascular invasion, and (3) multi-focal ICC with no vascular invasion. Since there are no conclusive cost-effectiveness studies measuring the efficacy of liver resection for advanced ICC (e.g., large tumor/R1 margin, presence of vascular invasion, multi-focal disease), the aim of this study was to compare two strategies: HR as primary treatment followed by adjuvant sCT (strategy A) and sCT with cisplatin (25 mg per square meter of body-surface area) and gemcitabine (1000 mg per square meter), each administered on days 1 and 8, every 3 weeks for four cycles followed by curative HR in those patients who respond to sCT (strategy B) [23]. In our model, the following endpoints were considered: 1.
2.
3. 4.
5.
Survival benefit. The survival of each strategy was measured in terms of life months (LMs). The survival benefit was defined as strategy A survival–strategy B survival. Incremental utility. The utility of each strategy was measured in terms of quality-adjusted survival and was expressed in quality-adjusted life months (QALMs). The incremental utility was defined as strategy A utility–strategy B utility. Incremental costs. Incremental costs were defined as strategy A cost—strategy B cost. Incremental cost-effectiveness ratio (ICER). ICER is the difference between the cost/utility of two competing strategies and was defined as strategy A cost/ utility—strategy B cost/utility. Net health benefit (NHB). The NHB of an alternative treatment was calculated using the formula employed by Stinnett et al. [24]: NHB = survival benefit—ICER/ WTP, where WTP is the willingness to pay. WTP is a
World J Surg Fig. 1 The event pathway: decision tree and states of health
fundamental cost-effectiveness endpoint representing the accepted limit to the additional cost per unit of effectiveness gained that a rational decision-maker will accept to allocate resources efficiently among competing priorities. In the present study, we adopted the most commonly cited cost-effectiveness threshold of $50,000/quality-adjusted life year [25]. Base-case clinical estimates We assumed that each patient in the strategy A group underwent R1 hepatic resection and that half of these patients received adjuvant sCT with cisplatin (25 mg per square meter of body-surface area) and gemcitabine (1000 mg per square meter) [26]. Five-year cumulative survival post-HR was derived from the literature (Table 1) [11, 27]. For strategy B, we assumed that each patient underwent preoperative sCT also with cisplatin and gemcitabine administered on days 1 and 8, every 3 weeks for four cycles [23]. Based on previously published data, we estimated that 26 % of patients achieved tumor down-sizing and were able to undergo subsequent curative HR [23]. Among the other patients who did not respond to sCT, the median time to progression was defined as: 8 months—(rate 9 8), where the rate was the proportion of patients with tumor down-sizing. For example, using the basecase value rate = 0.26, median time to progression becomes 8 - 0.26 9 8 = 6 months. Median survival of patients with tumor progression after sCT was considered 4 months [28]. Annual rates of survival and progression were transformed into monthly probabilities applying the declining
exponential approximation of life-expectancy (DEALE) approach: l = -1/t 9 ln(S). Transitional probabilities were varied within their relative 95 % confidence intervals assuming a triangular distribution. Utility and costs Due to the lack of data on QoL in ICC, we adopted the results of previous studies on colorectal cancer liver metastases (CLM), assuming a similar QoL after treatment for stage IV colorectal cancer with liver metastases and ICC [29]. The QoL after HR following complete recovery was considered 0.9. In contrast, the QoL after sCT in nonresponders was estimated to be similar to the QoL of patients with incurable CLM and scored 0.4 (Table 1). Base-case estimates for all utilities extracted from the literature are detailed in Table 1. Ranges were assumed to be within 20 % of the base-case values. Costs were assessed from the perspective of the health care providers. Rather than using estimated cost data such as those based on Medicare reimbursement data, we chose to use actual country-specific data. The costs and the utilities were discounted at annual rate of 3 % [30]. The costeffectiveness analysis was performed based on the EVEREST guidelines [31]. Sensitivity analyses and Monte Carlo simulation The correct calibration of the Markov model was checked for all patients, confirming that the predicted survivals of
123
World J Surg Table 1 Base-case value and sensitivity range extracted from literature for transition probabilities Variables
Base-case analysis
Range tested
Source
Background (all-cause) mortality
Age-specific
18–70
N/A
5-Year survival of single ICC without vascular invasion after curative resection
47 %
42–52 %
[11]
5-Year survival of ICC with vascular invasion after curative resection
25 %
20–30 %
[11]
5-Year survival of multi-focal ICC after curative resection
12 %
10–15 %
[11]
Down-staging rate after sCT
0.26
0.24–0.28
[23, 47]
Rate of patients undergoing adjuvant chemotherapy
0.45
0.3–0.6
[26]
Median survival in tumor progression after sCT (months)
4
3–5
[11]
Positive resection margin hazard ratio
2.2
1.5–3.2
[29]
Quality of life after surgery following complete recovery
1
0.9–1
[29]
Quality of life after non-curative resection
0.9
0.8–1
[29]
Quality of life of patients with incurable disease Hazard ratio of tumor size on survival
0.4 1.5
0–1 1.09–1.28
[34] [34]
Hazard ratio of male sex on survival
1.11
0.87–1.41
[34]
Hazard ratio of age [60 years on survival
1.31
1.10–1.56
[34]
1
6.2 %
1–10 %
2
1.6 %
1–10 %
3
6.9 %
1–10 %
4
27 %
20–40 %
Liver resection ($)
25,086
12,543–50,172
N/A
Chemotherapy ($)
13,120
10,496–15,744
N/A
Monthly follow up after chemotherapy ($)
1950
1560–2340
N/A
Monthly follow up after resection ($)
527
421–632
N/A
Life time
–
N/A
Postoperative mortality based on ASA score
[48]
COSTS
Time horizon (years)
ICC intrahepatic cholangiocarcinoma, sCT systemic chemotherapy, ASA American society of anesthesiologists
the model were within the 95 % confidence intervals of survivals derived from the literature. Such verification was repeated for each subgroup analysis showing a good calibration of the model for each tumor stage. One-way sensitivity analysis was performed for all transition probabilities, costs and utilities. A probabilistic sensitivity analysis, the Monte Carlo simulation, was performed. A total of 1000 ICC patients for each therapeutic strategy (A vs. B) were compared. Outcomes measured were LMs, QALMs, incremental costs, ICER, and NHB. Transitional probabilities were varied within their relative 95 % confidence intervals, while costs and utilities were varied within their plausible ranges, assuming a uniform distribution. The impact of variables on the NHB distribution of 1000 outcomes obtained from the Monte Carlo simulation was determined using the multivariate standard least square regression method. Statistical significance was set at p \ 0.05. A boosting forest tree method (partition modeling) was finally used to measure the contribution of each covariate to positive NHB [32]. Partition trees were constructed
123
using a training set (corresponding to 70 % of the whole cohort) and a validation set (corresponding to 30 % of the whole cohort) and the final model was that with the highest R square in both training and validation sets. The calculations were done with the JMP package version 9.0 (2010 SAS Institute Inc.) and TreeAge Pro version 2013 (1988–2013 TreeAge Software, Williamstown, MA).
Results Base-case analysis In the three possible base-case analyses, HR was cost-effective in base case 1 (large ICC with chance of R1 margin) and base case 2 (ICC with vascular invasion) but not base case 3 (multi-focal ICC). The NHB of strategy A was 1.4 QALMs compared with strategy B for patients with large ICC; similarly strategy A had a NHB of 1.3 QALMs compared with strategy B for patients with ICC and vascular invasion. In contrast, for base case 3, which involved
World J Surg
a patient with multi-focal ICC, HR was not cost-effective with a negative NHB (-0.3 QALMs) compared with strategy B. In single nodule ICC [6 cm, the ICER of HR compared with sCT was $22,482/QALY and the ICER of HR compared with sCT was $20,953/QALY for patients with ICC and vascular invasion. In multi-focal ICC, the ICER of HR compared with sCT was $83,604/QALY. Interestingly, while HR was therefore more cost-effective in large tumors ([6 cm) and in ICC with vascular invasion, sCT was more cost-effective in multi-focal ICC (Table 2). One-way sensitivity analysis In order to explore the impact of model covariates on costeffectiveness (ICER) of HR versus sCT, we performed a Tornado diagram (Fig. 2). On the univariate sensitivity Table 2 Incremental cost-effectiveness ratio and net health benefit comparing hepatic resection followed by systematic adjuvant chemotherapy (sCT) and sCT followed by liver resection at the base case Various costs
USA Strategy A
Strategy B
QALMs
19.7
17.1
Incremental QALMs gained Lifetime cost (US$)
– 41,532
2.6 36,586
Incremental cost (US$)
–
4946
ICER (US$/QALY)
–
22,482
ICC [6 cm
WTP (US$)
50,000
Net health benefit (QALMs)
1.4
Is HR cost-effective?
Yes
ICC with vascular invasion QALMs
16.5
14.2
Incremental QALMs gained
–
2.3
Lifetime cost (US$)
39,640
35,617
Incremental cost (US$)
–
4023
ICER (US$/QALY)
–
20,953
Net health benefit (QALMs)
1.3
Is HR cost-effective?
Yes
Multi-focal ICC QALMs Incremental QALMs gained
12.0 –
11.6 0.4
Lifetime cost (US$)
37,028
34,244
Incremental cost (US$)
–
2784
ICER (US$/QALY)
–
83,604
Net health benefit (QALMs)
-0.3
Is HR cost-effective?
No
ICC intrahepatic cholangiocarcinoma, QALMs quality-adjusted life months, QALY quality-adjusted life year, ICER incremental cost-effectiveness ratio, which is calculated as incremental cost per QALY gained, WTP willingness to pay, HR hepatic resection
analysis, the main factor influencing cost-effectiveness was margin hazard ratio, gender, and QoL after sCT. The impact of tumor-specific characteristics was apparently low. Monte Carlo simulation and analysis of the impact of main covariates on NHB When assessing the entire cohort, the acceptability curve (Fig. 3) demonstrated that strategy A generally had a higher probability to be more cost-effective than strategy B when a WTP above $24,000/QALY was accepted. To obtain a more complete assessment of the relative benefit of strategy A versus strategy B, a multivariable regression method was used to determine the impact of different model covariates on the NHB distribution of 1000 outcomes obtained from the Monte Carlo simulation. Specifically, the impact of a number of different factors on the NHB of patients with ICC undergoing resection was assessed. These factors included age, sex, American Society of Anesthesiologists (ASA) score, tumor size and number, presence of vascular invasion, hazard ratio of margin status, T stage, rate of patients who underwent sCT after R1 resection in strategy A, rate of response to sCT in strategy B, QoL of patients not responsive to sCT in strategy B. Of note, all the factors in the model had a significant impact on the NHB of HR (all p [ 0.05) (Table 3). In particular, patient-level factors such as age [60 (coefficient -1.12), low QoL after sCT (coefficient -2.57 if QoL after sCT \0.33 vs. 0.3–0.6 and -3.00 if QoL after sCT 0.3–0.6 vs. [0.6), and a low rate of patients who underwent sCT after R1 HR (coefficient -0.40) negatively impacted the NHB of HR. In addition, tumorand treatment-related variables including tumor size[6 cm (coefficient -1.67), presence of vascular invasion (coefficient -3.98), multi-focal ICC (coefficient -1.42), presence of periductal invasion (coefficient -0.48), and higher hazard ratio for an R1 margin (coefficient -4.23) all had a negative effect for the NHB of HR. Patients characterized by a higher ASA score (coefficient 0.94), male sex (coefficient 0.43), and T3 tumors (coefficient 0.53) were found to have a better NHB with HR (strategy A) relative to sCT followed by potential surgery (strategy B)(Table 3). Overall, the proportion of patients with advanced ICC who had a beneficial NHB associated with strategy A versus strategy B in the entire cohort of 1,000 patients in the Monte Carlo simulation was 524/1000 (52 %). A partition model was performed to identify which patient-, tumor-, and treatment-related variables had the highest contribution to the cost-effectiveness (positive NHB) of strategy A versus strategy B. The two covariates that were the most strongly associated with a positive NHB for HR relative to sCT were QoL utility of non-responders to sCT
123
World J Surg Fig. 2 One-way sensitivity analysis of factors associated with cost-effectiveness of strategy A versus strategy B (Tornado diagram)
Fig. 3 Cost-effectiveness acceptability curve of strategy A versus strategy B in patients with advanced ICC
and T stage (Fig. 4). These findings were different from the one-way sensitivity analysis in particular for tumor-specific characteristics, such as the T stage. Subsequent analyses were therefore focused on these two covariates. Strategy A was more cost-effective compared with strategy B for all
123
ICC T categories if QoL in non-responders to sCT was \0.3 (Fig. 5a). While strategy A was also cost-effective for T1 and T2a disease if QoL in non-responders to sCT was 0.3–0.6 (Fig. 5b), it was not for T2b or T3 ICC. In addition, strategy A was only cost-effective for patients
World J Surg Table 3 Multivariate analysis including variables with a significant impact on net health benefit of strategy A versus strategy B Variables
USA Coefficient
SE
p \0.0001
Age [60
-1.12
0.1124
Male gender
0.43
0.1123
0.0001
Tumor size [6 cm ASA score 3–4
-1.67 0.94
0.1119 0.1174
\0.0001 \0.0001
Positive marginal status hazard ratio
-4.23
0.1695
\0.0001
AJCC T2 vs. T1
-3.98
0.1732
\0.0001
AJCC T2b vs. T2a
-1.42
0.1790
\0.0001
AJCC T3 vs. T2b
0.53
0.1813
0.00036
AJCC T4 vs. T3
-0.48
0.1777
0.0075
QoL after sCT \0.33 vs. 0.3–0.6
-2.57
0.1426
\0.0001
QoL after sCT 0.3–0.6 vs. [0.6
-3.00
0.1304
\0.0001
sCT rate post R1 [45 vs. B45 %
-0.40
0.1120
0.0004
R0 rate after sCT in strategy B
-12.02
1.9559
\0.0001
SE standard error, ASA american society of anesthesiologists, QoL quality of life, sCT systemic chemotherapy
invasion) the cost-effectiveness of HR compared with sCT had a NHB and ICER of 12.5 QALMs and $4858/QALY, respectively.
Discussion
Fig. 4 Partition model showing the contribution of each covariate to positive net health benefit of strategy A versus strategy B
with early T1 disease when QoL in non-responders to sCT was [0.6 (Fig. 5c). Scenario analysis Given the overall modest NHB of HR, we sought to evaluate the best and worst possible scenario for HR versus sCT. To do this, we considered the covariates used in the multivariable analysis and in particular we defined a patient with a large[6 cm, multi-focal ICC with vascular invasion and an estimated QoL in non-responders to sCT[0.6 as the most pessimistic scenario for HR. The pessimistic scenario showed that HR was not cost-effective with a NHB of -8.5 QALMs and a negative ICER for strategy A versus strategy B. In contrast, when considering the most optimistic scenario (small, solitary ICC \6 cm with no vascular
Many patients with ICC may present with advanced disease, which can be characterized by large tumor size, vascular invasion, and multi-focal disease [8]. The management of patients with advanced ICC remains controversial. While some centers consider HR for patients with advanced ICC, other surgeons favor the use of sCT. Chemotherapy can be delivered either as definitive therapy or, in certain circumstances, with the hope to down-size the tumor to facilitate possible future surgery [33]. sCT— typically given as doublet therapy consisting of gemcitabine and cisplatin based on the Advance Biliary Cancer trial—is standard of care for patients when ICC is considered inoperable [23]. In contrast, surgery is clearly indicated for patients with small ICC in whom an R0 resection can be expected [13, 14, 34]. For those patients with advanced disease that is either technically (large tumor, risk of R1 margin) or oncologically (vascular invasion, multi-focal) borderline, optimal management is more controversial. Previous work from our group and others has noted that these factors are associated with a worse survival following HR [11, 14, 27]. While there have been several studies that have evaluated the relative risk and survival benefit of HR based on clinical and tumorspecific factors [34], no study has examined a broader set of outcomes that include NHB, QALY, and relative costs of the different therapies for advanced ICC. The current
123
World J Surg
Fig. 5 Net health benefit of strategy A over strategy B among different AJCC T stages. a Quality of life (QoL) after primary chemotherapy \0.3; b QoL after primary chemotherapy 0.3–0.6; c QoL after primary chemotherapy [0.6
study is unique and important in that we specifically examined the NHB of HR relative to sCT among patients with advanced ICC stratified based on different patient, tumor, and treatment characteristics (e.g., age, ASA score, large tumor [6 cm, vascular invasion, multi-focal disease, response to sCT). In addition, using Markov modeling in a United States geographical cost scenario, we were also able to evaluate the cost-effectiveness of HR versus sCT for patients with ICC. In other words, we simulated a
123
randomized clinical trial (RCT) that compares different treatments for advanced ICC. Half of patients were treated with HR and half with sCT. Survival, quality-adjusted survival, costs and cost-effectiveness of these two groups were compared. To better define which patients would have benefit the most of surgery versus sCT, we also performed scenario analysis, able to compare the two strategies in patients with different clinicopathologic characteristics (i.e., large tumors, lesions with vascular invasion, multifocal ICC). Overall, in examining the entire cohort of 1000 patients in the Monte Carlo simulation, we found that HR was generally cost-effective compared with sCT. HR was associated with a positive NHB for patients with large ICC and those patients with vascular invasion; in contrast, patients with multi-focal disease had a negative NHB with HR versus sCT. Interestingly, we did note that HR was particularly favorable compared with sCT in certain cases, such as younger patients who had solitary HCC without vascular invasion. The finding that HR was associated with a positive NHB compared with sCT for patients with large ICC is consistent with previous data suggesting that tumor size may not impact survival following resection [12]. Nathan et al. reported that tumor size was not associated with prognosis and the current 7th edition American Joint Committee on Cancer staging of ICC does not include tumor size as a prognostic factor [12, 35]. From a technical point of view, larger ICC lesions can, however, require complex liver resections thereby limiting the operability of these tumors. In turn, the risk of a close or microscopically positive (R1) margin can be higher for these large ICC lesions. In fact, several previous reports have noted the incidence of an R1 margin following resection of ICC to be as high as 15–20 % [13, 34]. While multiple studies have noted that R1 resection is associated with worse survival [8, 36–38], other investigators have reported no difference in the survival of patients who underwent an R0 versus R1 resection [39]. Whether patients with large ICC should undergo an attempt at resection with the risk of an R1 margin or be treated with preoperative sCT in the hopes of down-sizing the tumor remains unknown. For this reason, using Markov modeling, we compared the NHB of an R1 HR followed by adjuvant sCT versus preoperative sCT followed by an R0 HR in those patients who responded to sCT. Interestingly, we found that strategy A was associated with a positive NHB, as well as a lower cost relative to QALYs when compared with strategy B in patients with a large ICC. Specifically, the NHB for strategy A was 1.4 months better than strategy B, and the ICER of strategy A compared with strategy B was $22,482/QALY. The data therefore strongly suggest a NHB favoring HR for patients with large ICC even when there is a risk of an R1 margin.
World J Surg
While we also noted a positive NHB for patients with vascular invasion, the NHB favored sCT for patients with multi-focal disease. Specifically, the NHB of HR was 1.3 months for ICC with vascular invasion, but -0.3 months for ICC with multi-focal disease. The ICER of strategy A compared with strategy B was $20,953/ QALY and $83,604/QALY, respectively, for ICC with vascular invasion or multi-focal disease. While some previous data have suggested a 20–40 % increased risk of death for patients with ICC and vascular invasion, the role of HR in this clinical setting has not been completely defined [34, 40, 41]. Several authors have reported longterm survival among patients with ICC and vascular invasion following resection [40, 41]. Furthermore, resection of other primary liver tumors such as HCC in the setting of vascular invasion has been reported [42]. While associated with an overall worse long-term prognosis and a more modest NHB and ICER compared with patients with large tumors without vascular invasion, our data suggest that HR still may be preferable to sCT. Interestingly, we did note that HR was not cost-effective compared with sCT for multi-focal ICC. The prognostic importance of whether an ICC tumor is solitary or multiple has been recognized by its inclusion in the AJCC staging manual as one of the variables used to define the T category for ICC [14, 35]. Most studies, however, do not classify or define ‘‘multiple’’ lesions as satellitosis, ‘‘true’’ multi-focal disease, or intrahepatic metastasis [43]. As such, data on resection for multiple hepatic lesions likely represent a mix of these different patient groups, each of which may have a distinct prognosis [34]. Notwithstanding this fact, multi-focal disease seems to portend a particularly poor prognosis—reflecting either a field defect or intrahepatic spread of the disease. Similar to our findings for ICC, data on HR for multi-focal HCC have similarly demonstrated poor longterm results [44]. Taken together, data derived from our decision models would strongly suggest that sCT may be a better alternative to HR for patients with multi-focal ICC with regard to both NHB and QALYs. When considering therapeutic options for patients with advanced disease, the clinician needs to consider tumor characteristics such as lesion size and number, as well as vascular invasion, but also patient-specific factors. In the current study, multivariable analyses demonstrated that age and QoL were important patient-level factors when determining the NHB of the therapeutic strategy. In particular, strategy A was more cost-effective compared with strategy B for all ICC T categories if QoL in non-responders to sCT was \0.3. However, when the QoL in non-responders to sCT was better (e.g., C3), then HR had less NHB and was less cost-effective—especially among patients with more advanced disease (e.g., T2 or T3) (Fig. 4). QoL and performance status is recognized as a critical preoperative
factor in determining the relative anticipated benefit of almost any surgical intervention [45, 46]. These findings underline the importance of QoL in the already complex decision-making process of determining whether to offer HR for patients with advanced ICC. Several limitations should be considered when interpreting our data. As with all the decision-making Markov models, it was necessary to use data from the literature in making assumptions regarding the transition probabilities. To minimize potential bias we derived the probability estimates from a wide range of robust data sources including systematic reviews, meta-analyses and previous large multi-institutional studies [8, 11, 34]. Another limitation involved the estimation of utility for different health states, which were based on the available literature for treatment of colorectal cancer liver metastases. Unfortunately there are no data on QoL after liver resection for ICC or QoL of patients treated with sCT for ICC [29]. Moreover, it is important to note that mathematical models such as Markov decision-making and Monte Carlo simulation are not alternatives to prospective RCT, but are useful tools to guide clinical decision-making in the absence of such trials. Of course, in clinical practice, technical feasibility and clinical applicability should always guide treatment selection. In conclusion, in patients with large ([6 cm) ICC who may be at risk of R1 resection, and in patients with ICC and vascular invasion, HR appeared to be associated with a greater NHB and be more cost-effective than sCT. In contrast, in patients with multi-focal ICC HR had a negative NHB, and therefore they should be treated with sCT rather than HR. Compliance with Ethical Standards Conflict of interest
None.
References 1. Khan SA et al (2002) Guidelines for the diagnosis and treatment of cholangiocarcinoma: consensus document. Gut 51(Suppl 6):VI1–VI9 2. El-Serag HB et al (2009) Risk of hepatobiliary and pancreatic cancers after hepatitis C virus infection: a population-based study of U.S. veterans. Hepatology 49(1):116–123 3. Endo I et al (2008) Intrahepatic cholangiocarcinoma: rising frequency, improved survival, and determinants of outcome after resection. Ann Surg 248(1):84–96 4. Malhi H, Gores GJ (2006) Cholangiocarcinoma: modern advances in understanding a deadly old disease. J Hepatol 45(6): 856–867 5. Shaib YH et al (2004) Rising incidence of intrahepatic cholangiocarcinoma in the United States: a true increase? J Hepatol 40(3):472–477 6. Patel T (2001) Increasing incidence and mortality of primary intrahepatic cholangiocarcinoma in the United States. Hepatology 33(6):1353–1357
123
World J Surg 7. Shaib Y, El-Serag HB (2004) The epidemiology of cholangiocarcinoma. Semin Liver Dis 24(2):115–125 8. Mavros MN et al (2014) Treatment and prognosis for patients with intrahepatic cholangiocarcinoma: systematic review and meta-analysis. JAMA Surg 149(6):565–574 9. Weber SM et al (2001) Intrahepatic cholangiocarcinoma: resectability, recurrence pattern, and outcomes. J Am Coll Surg 193(4):384–391 10. Amini N et al (2014) Temporal trends in liver-directed therapy of patients with intrahepatic cholangiocarcinoma in the United States: a population-based analysis. J Surg Oncol 110(2): 163–170 11. de Jong MC et al (2011) Intrahepatic cholangiocarcinoma: an international multi-institutional analysis of prognostic factors and lymph node assessment. J Clin Oncol 29(23):3140–3145 12. Nathan H et al (2009) A proposed staging system for intrahepatic cholangiocarcinoma. Ann Surg Oncol 16(1):14–22 13. Farges O et al (2011) Influence of surgical margins on outcome in patients with intrahepatic cholangiocarcinoma: a multicenter study by the AFC-IHCC-2009 study group. Ann Surg 254(5): 824–829; discussion 830 14. Spolverato G et al (2014) Tumor size predicts vascular invasion and histologic grade among patients undergoing resection of intrahepatic cholangiocarcinoma. J Gastrointest Surg 18(7): 1284–1291 15. Raderer M et al (1999) Two consecutive phase II studies of 5-fluorouracil/leucovorin/mitomycin C and of gemcitabine in patients with advanced biliary cancer. Oncology 56(3):177–180 16. Kubicka S et al (2001) Phase II study of systemic gemcitabine chemotherapy for advanced unresectable hepatobiliary carcinomas. Hepatogastroenterology 48(39):783–789 17. Dodson RM et al (2013) Intrahepatic cholangiocarcinoma: management options and emerging therapies. J Am Coll Surg 217(4):736–750 e4 18. Kiba T et al (2006) Single-agent gemcitabine for biliary tract cancers. Study outcomes and systematic review of the literature. Oncology 70(5):358–365 19. Scheithauer W (2002) Review of gemcitabine in biliary tract carcinoma. Semin Oncol 29(6 Suppl 20):40–45 20. Lee GW et al (2006) Combination chemotherapy with gemcitabine and cisplatin as first-line treatment for immunohistochemically proven cholangiocarcinoma. Am J Clin Oncol 29(2):127–131 21. Knox JJ et al (2005) Combining gemcitabine and capecitabine in patients with advanced biliary cancer: a phase II trial. J Clin Oncol 23(10):2332–2338 22. Andre T et al (2004) Gemcitabine combined with oxaliplatin (GEMOX) in advanced biliary tract adenocarcinoma: a GERCOR study. Ann Oncol 15(9):1339–1343 23. Valle J et al (2010) Cisplatin plus gemcitabine versus gemcitabine for biliary tract cancer. N Engl J Med 362(14):1273–1281 24. Stinnett AA, Mullahy J (1998) Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis. Med Decis Making 18(2 Suppl):S68–S80 25. Grosse SD (2008) Assessing cost-effectiveness in healthcare: history of the $50,000 per QALY threshold. Expert Rev Pharmacoecon Outcomes Res 8(2):165–178 26. Zhang GW et al (2014) Identification of prognostic factors and the impact of palliative resection on survival of patients with stage IV hepatolithiasis-associated intrahepatic cholangiocarcinoma. J Surg Oncol 109(5):494–499 27. Farges O et al (2011) AJCC 7th edition of TNM staging accurately discriminates outcomes of patients with resectable intrahepatic cholangiocarcinoma: by the AFC-IHCC-2009 study group. Cancer 117(10):2170–2177
123
28. Park J et al (2009) Natural history and prognostic factors of advanced cholangiocarcinoma without surgery, chemotherapy, or radiotherapy: a large-scale observational study. Gut Liver 3(4):298–305 29. Khajanchee YS et al (2011) Hepatic resection vs minimally invasive radiofrequency ablation for the treatment of colorectal liver metastases: a Markov analysis. Arch Surg 146(12): 1416–1423 30. Lim KC et al (2014) Cost-effectiveness analysis of liver resection versus transplantation for early hepatocellular carcinoma within the Milan criteria. Hepatology 61(1):227–237 31. Siegel JE et al (1996) Recommendations for reporting cost-effectiveness analyses. Panel on cost-effectiveness in health and medicine. JAMA 276(16):1339–1341 32. Sexton J, Laake P (2007) Boosted regression trees with errors in variables. Biometrics 63(2):586–592 33. Valle JW et al (2009) Gemcitabine alone or in combination with cisplatin in patients with advanced or metastatic cholangiocarcinomas or other biliary tract tumours: a multicentre randomised phase II study—The UK ABC-01 Study. Br J Cancer 101(4): 621–627 34. Hyder O et al (2014) A nomogram to predict long-term survival after resection for intrahepatic cholangiocarcinoma: an eastern and western experience. JAMA Surg 149(5):432–438 35. Edge SBBD, Compton CC, Fritz AG, Greene FL, Trotti A (eds) (2010) AJCC cancer staging manual (7th edn). Springer: New York 36. Yamashita Y et al (2008) The impact of surgical treatment and poor prognostic factors for patients with intrahepatic cholangiocarcinoma: retrospective analysis of 60 patients. Anticancer Res 28(4C):2353–2359 37. Uenishi T et al (2008) Indications for surgical treatment of intrahepatic cholangiocarcinoma with lymph node metastases. J Hepatobiliary Pancreat Surg 15(4):417–422 38. Morimoto Y et al (2003) Long-term survival and prognostic factors in the surgical treatment for intrahepatic cholangiocarcinoma. J Hepatobiliary Pancreat Surg 10(6):432–440 39. Murakami S et al (2014) Factors affecting survival after resection of intrahepatic cholangiocarcinoma. Surg Today 44(10): 1847–1854 40. Akatsu T et al (2005) Long-term survival of intrahepatic cholangiocarcinoma with hilar lymph node metastasis and portal vein involvement. Hepatogastroenterology 52(62):603–605 41. Nakagohri T et al (2000) Extended right hepatic lobectomy with resection of inferior vena cava and portal vein for intrahepatic cholangiocarcinoma. J Hepatobiliary Pancreat Surg 7(6):599–602 42. Nathan H et al (2013) Surgical therapy for early hepatocellular carcinoma in the modern era: a 10-year SEER-medicare analysis. Ann Surg 258(6):1022–1027 43. Maithel SK et al (2013) Multidisciplinary approaches to intrahepatic cholangiocarcinoma. Cancer 119(22):3929–3942 44. Ng KK et al (2005) Is hepatic resection for large or multinodular hepatocellular carcinoma justified? Results from a multi-institutional database. Ann Surg Oncol 12(5):364–373 45. Martin RC et al (2007) Health-related quality of life: return to baseline after major and minor liver resection. Surgery 142(5): 676–684 46. Poon RT et al (2001) A prospective longitudinal study of quality of life after resection of hepatocellular carcinoma. Arch Surg 136(6):693–699 47. Sasaki T et al (2013) Current status of chemotherapy for the treatment of advanced biliary tract cancer. Korean J Intern Med 28(5):515–524 48. Reddy SK et al (2011) Major liver resection in elderly patients: a multi-institutional analysis. J Am Coll Surg 212(5):787–795
