multimodal treatment and neoadjuvant chemotherapy …
TRANSCRIPT
University of South Florida University of South Florida
Scholar Commons Scholar Commons
Graduate Theses and Dissertations Graduate School
November 2019
Multimodal Treatment and Neoadjuvant Chemotherapy Trends, Multimodal Treatment and Neoadjuvant Chemotherapy Trends,
Utilization and Survival Effects in Intrahepatic Utilization and Survival Effects in Intrahepatic
Cholangiocarcinoma – a Propensity Score Analysis Cholangiocarcinoma – a Propensity Score Analysis
Ovie Utuama University of South Florida
Follow this and additional works at: https://scholarcommons.usf.edu/etd
Part of the Oncology Commons
Scholar Commons Citation Scholar Commons Citation Utuama, Ovie, "Multimodal Treatment and Neoadjuvant Chemotherapy Trends, Utilization and Survival Effects in Intrahepatic Cholangiocarcinoma – a Propensity Score Analysis" (2019). Graduate Theses and Dissertations. https://scholarcommons.usf.edu/etd/8690
This Dissertation is brought to you for free and open access by the Graduate School at Scholar Commons. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Scholar Commons. For more information, please contact [email protected].
Multimodal Treatment and Neoadjuvant Chemotherapy Trends, Utilization and Survival
Effects in Intrahepatic Cholangiocarcinoma – a Propensity Score Analysis
by
Ovie Utuama
A dissertation submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy with a concentration in Epidemiology
College of Public Health
University of South Florida
Co-Major Professor: Aurora Sanchez-Anguiano, Ph.D Co-Major Professor: Jennifer Permuth, Ph.D
Getachew Dagne, Ph.D Amy Alman, Ph.D Daniel Anaya, MD
Date of Approval: November 1, 2019
Keywords: curative-intent surgery, pre-operative therapy, biliary cancer, elderly
Copyright © 2019, Ovie Utuama
DEDICATION
It takes a village to raise a child and as such this body of work is dedicated to the first and most
profound of my teachers -- my parents, Prof. Amos and Dr. Nelly Utuama -- who sowed the
seeds of curiosity and set me off on my life-long pursuit of knowledge. This endeavor would not
have been possible without your example and support.
To my precious wife, Ona, for embarking on this journey with me. Now that I’m all grown up, I
promise to be a responsible adult.
To Rovu, Rume and Ome, daddy loves you. May this work serve as a testament to striving to be
your best selves and leaving an indelible mark on the world, for the good of all humanity.
ACKNOWLEDGEMENTS
Though my instructors are many, all of whom important in shaping me, it is to my committee I
must now turn in gratitude. Dr. Sanchez-Anguiano, my longest collaborator, for life and
professional advice that helped center me during times of apparent stagnation. Dr. Permuth, for
demystifying the Ph.D., taking me by the hand, operationalizing and optimizing the process.
Your energy and can-do attitude are simply infectious! Drs. Dagne and Alman, for insight and
conversations that brought much clarity and encouragement. Dr. Anaya, for pushing the
boundaries of an idea, a question, a hypothesis. You have made me a better researcher.
I would be remiss to not mention Dozie and Bashir, who generously provided vital real estate in
Tampa, at a time when I couldn’t afford it, that enabled meaningful progress after many months
of vacillation. Fahad Mansuri, thank you for being the first guest at my dissertation defense. Dr.
Schwartz, you are a kindred spirit with whom I have enjoyed our many musings. Thank you for
always believing in me! To Jane, Donna, Chassity and Diana, for all the quiet but essential work
that you do that keep the cogs of the College and Gastrointestinal Oncology Department at
Moffitt turning – a warm and heartfelt thank you!
i
TABLE OF CONTENTS
LIST OF TABLES ......................................................................................................................... iii
LIST OF FIGURES....................................................................................................................... iv
ABSTRACT .................................................................................................................................. vi
CHAPTER 1: STATEMENT OF PROBLEM AND RATIONALE ..................................................... 1
CHAPTER 2: CURATIVE-INTENT SURGERY FOR ELDERLY PATIENTS WITH NON- METASTATIC INTRAHEPATIC CHOLANGIOCARCINOMA ................................................... 4
Abstract ........................................................................................................................... 4 Introduction ...................................................................................................................... 5 Methods .......................................................................................................................... 6
Study population and study design ....................................................................... 6 Inclusion/exclusion criteria ................................................................................... 6 Intervention/exposure........................................................................................... 7 Outcomes ............................................................................................................7 Covariates ............................................................................................................8 Propensity score development .............................................................................8 Statistical analysis ................................................................................................9
Results .......................................................................................................................... 10 Trends and patterns of treatment over time ........................................................ 10 Predictors of curative -- intent surgical treatment ................................................ 10 Survival analysis ................................................................................................ 11
Discussion ..................................................................................................................... 12 Supplemental Results .................................................................................................... 29
CHAPTER 3: NEOADJUVANT CHEMOTHERAPY FOR PATIENTS WITH INTRAHEPATIC CHOLANGIOCARCINOMA ........................................................................ 33
Abstract ......................................................................................................................... 33 Introduction .................................................................................................................... 34 Methods ........................................................................................................................ 35
Study design ...................................................................................................... 35 Inclusion/exclusion criteria ................................................................................. 35 Intervention – neoadjuvant chemotherapy .......................................................... 36 Outcomes .......................................................................................................... 36 Covariates .......................................................................................................... 36 Propensity score development ........................................................................... 37 Statistical analysis .............................................................................................. 38
ii
Results ..........................................................................................................................39 Neoadjuvant chemotherapy utilization – trends and predictors ...........................39 Survival analysis ................................................................................................40
Discussion .....................................................................................................................41
Supplemental Results ....................................................................................................58
CHAPTER 4: CONCLUSION AND RECOMMENDATIONS ......................................................62
REFERENCES .........................................................................................................................65
iii
LIST OF TABLES
TABLE 2.1 Sociodemographic, hospital, tumor and treatment characteristics of ICC
patients, n=3,653 ...............................................................................................17
TABLE 2.2 Predictors of surgical treatment ..........................................................................24
TABLE 2.3 Crude and propensity score (PS) stratified-adjusted Cox models for
overall survival- age and treatment effects examined .........................................28
TABLE S2.1 Standardized mean differences by propensity score strata and treatment
group..................................................................................................................30
TABLE 3.1 Baseline sociodemographic, hospital, clinical and tumor
characteristics of patients with non-metastatic intrahepatic
cholangiocarcinoma, treated with surgery, by treatment strategy
(n=881) ..............................................................................................................46
TABLE 3.2 Multivariable logistic regression analysis examining predictors of
neoadjuvant chemotherapy utilization (n=881) ...................................................52
TABLE 3.3 Results from Cox regression models examining the effect of neoadjuvant chemotherapy on survival ..............................................................57
TABLE S3.1 Baseline characteristic of patients in the neoadjuvant and no- neoadjuvant groups, from the propensity score matching – adequacy of matching expressed in p values and standardized differences .......................59
iv
LIST OF FIGURES
FIGURE 2.1 Treatment trend for stages 1-3 ICC patients, n=3,653 ........................................ 20
FIGURE 2.2A Treatment utilization for stages 1-3 ICC patients, n=3,653 .................................. 21
FIGURE 2.2B Treatment utilization for stages 1-3 ICC in patients < 70 years, n=2,337 ............ 22
FIGURE 2.2C Treatment utilization for stages 1-3 ICC in patients ≥ 70 years, n=1,316 ............ 23
FIGURE 2.3A KM plots for stages 1-3 ICC patients, n=3,651 ................................................... 25
FIGURE 2.3B KM plots for stages 1-3 ICC patients less than 70 years, n=2,337 ..................... 26
FIGURE 2.3C KM plots for stages 1-3 ICC patients at least 70 years, n=1,315 ........................ 27
FIGURE S2.1 Inclusion and exclusion criteria of study participants, NCDB (2004-2014) .......... 31
FIGURE S2.2 Treatment trend for stages 1-4 ICC patients, n=21,377 ..................................... 32
FIGURE 3.1A Annual proportion of patients with intrahepatic cholangiocarcinoma receiving neoadjuvant chemotherapy, in participating NCDB hospitals (n=881) ...............................................................................................50
FIGURE 3.1B Trend over time for neoadjuvant chemotherapy utilization among 26 hospitals with consistent patient reporting through study period (n=340) ............ 51
FIGURE 3.2A Unadjusted Kaplan-Meier curves depicting overall survival estimates for all patients with non-metastatic, intrahepatic cholangiocarcinoma by treatment strategy (n=881) ............................................................................54
FIGURE 3.2B Unadjusted Kaplan-Meier curves depicting overall survival estimates for patients with locally advanced stages (Stages II-III), intrahepatic cholangiocarcinoma - by treatment strategy (n=414) .......................................... 55
FIGURE 3.2C Unadjusted Kaplan-Meier curves depicting overall survival estimates for patients with early stage (Stages I), intrahepatic cholangiocarcinoma by treatment strategy (n=465) ............................................56
v
FIGURE S3.1 Study flow chart for selection criteria of patients with intrahepatic cholangiocarcinoma – NCDB (2006-2014) .........................................................58
FIGURE S3.2 Histogram illustrating the distribution of patients in each of the propensity score strata, by treatment group (n=881) .....................................................................60
FIGURE S3.3 Utilization of neoadjuvant chemotherapy among all ICC patients, n=21,018 ......61
vi
ABSTRACT
Intrahepatic Cholangiocarcinomas (ICC) are fatal malignancies common among the elderly.
Patients are diagnosed late and often relapse, even after curative-intent surgery (CIS). In this
context, additional systemic chemotherapy (multimodal treatment) is recommended for most
patients but reported survival benefits are minimal and are limited to small single institutional
studies. For this reason, using a national cancer registry, we sought to characterize multimodal
treatment trends and utilization in general and neoadjuvant chemotherapy, specifically, as well
as evaluate their survival effects among the larger population. We hypothesized that, 1). Elderly
ICC patients would have survival benefits equivalent to younger patients, 2). Neoadjuvant
chemotherapy (NC) provides improved survival over adjuvant chemotherapy or surgery alone.
Study participants were selected from the National Cancer Database (NCDB), a hospital-based
cancer registry which accounts for 70% of newly diagnosed cancer cases in the United States
annually. We examined trends from 2004 through 2014 using Cochran-Armitage trend tests,
identified independent predictors of multimodal treatment using logistic regression models,
evaluated survival using univariate KM plots with log-rank tests and adjusted for confounders
using propensity-score stratified and matched multivariable Cox regression models. Survival
benefit in elderly versus young patients was no different for CIS (HR 1.14 [0.92-1.41]) and for
CIS-multimodality treatment (1.35 [0.91-2.01]). NC utilization was associated with improved OS
(HR: 0.78 [95%CI 0.54-1.11]. Elderly patients were less likely to receive CIS than younger ones
but had an equivalent survival when treated. This study also demonstrated that patients with
more advanced disease may benefit from a multimodal approach using NC.
1
CHAPTER 1: STATEMENT OF PROBLEM AND RATIONALE
The prognosis of intrahepatic cholangiocarcinoma (ICC) has remained unchanged for more than
40 years. 1,2 Although a rare cancer with less than 10,000 cases annually in the United States, 3
together with hepatocellular carcinoma they represent the fourth largest cause of cancer
mortality with an ongoing and largely unexplained rise in incidence.4,5 Late disease presentation,
non-specific symptoms, non-specific disease markers and an absence of sensitive screening
tools have contributed to the fatality of the condition.6-8 Extensive surgery is the mainstay for
potential cure. 9,10
There is mounting evidence that surgery in combination with chemo- or radio-therapy, so-called
multimodal therapy, may be beneficial in late stage, non-metastatic ICC.11-13 However, until
recently only single patient case reviews and small observational studies from single institutions
have supported this benefit.14-16 Observational studies have been particularly subject to
selection biases in which early stage ICC is commonly treated with surgery alone while later
stage disease with multimodal therapy, making comparisons of the survival effects of
multimodal therapy and surgery alone confounded by disease stage. Additionally, the treatment
experience of elderly ICC patients, who make up more than half of all those diagnosed with the
disease, is unclear. 17-19 Against the backdrop of older cancer patients receiving substandard
care compared to younger ones, there is controversy about whether the former respond to
recommended treatment as well as young patients. Furthermore, the timing of the
administration of systemic chemotherapy in relation to surgery increasingly suggests that
neoadjuvant chemotherapy (chemotherapy before surgery) may be more beneficial than the
currently recommended adjuvant chemotherapy (chemotherapy after surgery).14,20
2
Therefore, the aims and objectives of the present dissertation are as follows.
Aim 1: To characterize the use and survival effects of multimodal therapy among elderly
ICC patients. We hypothesize that multimodal therapy has the same survival effect on older
patients as younger patients.
Aim 1 objectives were to, a) describe the utility and trends of multimodal therapy, b) elucidate
the predictors of multimodal therapy, and c) evaluate the survival effects of multimodal therapy
among the elderly in relation to younger patients.
Aim 2: To characterize the use and survival effects of neoadjuvant chemotherapy among
ICC patients. We hypothesize that the neoadjuvant chemotherapy approach has greater
survival effect than the adjuvant approach. Aim 2 objectives, therefore, were to, a) describe the
utility and trends of neoadjuvant chemotherapy, b) elucidate the predictors of neoadjuvant
chemotherapy, and c) evaluate the survival effects of neoadjuvant chemotherapy in relation to
non-neoadjuvant therapy.
To evaluate our hypotheses and objectives, we used the National Cancer Database (NCDB) as
a data source. The NCDB is a hospital-based cancer registry instituted in 1989 by the American
College of Surgeons and the American Cancer Society which tracks cancer patients, their
treatment and outcomes and now represents more than 70 percent of newly diagnosed cancer
cases annually, sourced from more than 1500 hospitals nationwide. 21 From 2004 through 2014,
the NCDB accumulated 23,273 cases of intrahepatic biliary malignancies which constitute the
source population for the retrospective cohort study design used throughout the dissertation.
This research is significant because it is the first to aggregate large numbers of incident ICC-
only patients with the aims of examining the association of multimodal therapy among the
elderly and neoadjuvant chemotherapy use and survival.
3
It is innovative because it has extensively employed causal inference techniques to adjust for
selection biases common in cancer treatment studies. This research hopes to present a
rigorous characterization of multimodal therapy (of which neoadjuvant chemotherapy is a
special case) and valid estimates of its survival effects that will advance our understanding of
best treatment practices for ICC, as an important step towards improving survival outcomes of
this lethal malignancy.
4
CHAPTER 2: CURATIVE-INTENT SURGERY FOR ELDERLY PATIENTS WITH NON-
METASTATIC INTRAHEPATIC CHOLANGIOCARCINOMA
Abstract [335 words]
Introduction: The incidence of intrahepatic cholangiocarcinoma (ICC) is increasing with most
patients (>50%) presenting over the age of 70 years. The use of curative-intent surgery
(hepatectomy - CIS) in elderly patients with ICC is currently unknown, and the survival benefit of
CIS for this population is unclear.
Methods: A retrospective cohort study of patients in the National Cancer Data Base with a
diagnosis of ICC was performed (2004-2014). A landmark approach was used to define the final
study cohort. Patients were categorized by age into young (<70) and elderly (≥70). Trends over
time for utilization of different treatment approaches were evaluated (CIS, CIS-multimodality,
other, no treatment). Using propensity score stratification and cox regression analysis, we
examined the association between treatment strategy and overall survival (OS), including
interaction for age and treatment type.
Results: Among the 3,653 patients in the study cohort, 1,316 (36%) were elderly. CIS was
performed in only 19% of patients, though its use increased over time for the whole population
(trend test, P=0.007). In the elderly group, there was a significant decrease over time in the “no-
treatment” group, primarily driven by an increase in the “other” treatment category (P<0.001).
On multivariable logistic regression, age≥70 was a predictor of not receiving CIS (OR 0.68
[0.56-0.81]; P<0.001). Median OS was significantly higher for those receiving CIS and CIS-
multimodality treatments as compared to those in the other and no-treatment groups (median
OS 46.5, 47.9, 25.3, and 16.3 months, respectively; logrank, P<0.001). Following propensity
5
score stratification, relative to no treatment, CIS and CIS-multimodality were each associated
with lower risk of death for the whole cohort (HR 0.28 [0.25-0.32] and 0.32 [0.27-0.38],
respectively), and for the elderly population (0.33 [0.27-0.40] and 0.42 [0.29-0.60], respectively).
Further, the survival benefit in elderly versus young patients was no different for CIS (HR 1.14
[0.92-1.41]) and for CIS-multimodality treatment (1.35 [0.91-2.01]).
Conclusion: Patients over the age of 70 years are less likely to receive surgical treatment for
ICC, despite having a significant survival benefit from curative-intent surgery (hepatectomy),
equivalent to that observed by younger patients.
Introduction
Intrahepatic cholangiocarcinomas (ICC) are typically adenocarcinomas arising from the biliary
ducts and ductules within the liver. ICC accounts for 10-15% biliary/hepatic malignancies and
represents the second commonest form of liver cancer. 5,22 In the United States, incidence is
approximately 1-2 per 100,000 persons, with a male preponderance of new cases. 22 About
5000-8000 Americans are diagnosed annually with a peak age between the fifth to seventh
decades of life. 5 Hispanic and Asian populations are reported to have the highest rate of
incidence (up to 3.3 cases per 100,000 persons) while non-Hispanic whites and African-
Americans have the lowest rates, both estimated at 2.1 cases per 100,000 persons. 23 Mortality
rates are highest among American Indian and Alaskan Natives (1.3-1.4 per 100,000) and lowest
among African Americans (0.7 per 100,000). 23,24 Despite its relative rarity, ICC remains of public
health significance because it is one of a few cancers with a five-year survival rate below 10%
and whose incidence is on the rise in many western countries, often without known risk
factors.25
Historically, there is abundant evidence that the elderly have not benefitted as much from
advances in cancer treatment and that they receive more substandard care than the young. 28,29
6
The reasons for this may be related to more comorbidities, reduced organ functional capacity,
perception of increased toxicities to systemic therapies among the elderly, and fewer treatment
guidelines for them. 30,31 More recently, several studies among patients with hepatobiliary
cancers have demonstrated that the elderly can achieve similar survival benefits as young
patients when stage-appropriate and patient-specific treatment is instituted. 32,33 There are only
a few studies investigating the treatment experience of the elderly patients with ICC. 17 We
therefore sought to characterize ICC treatment in the United States and investigate its survival
benefits among elderly patients.
Methods
Study population and study design
Using a retrospective cohort design, we abstracted data corresponding to ICC patients
diagnosed between January 1, 2004 through December 31, 2014 from the National Cancer
Database (NCDB). The NCDB is a hospital-based cancer registry that was created in 1989 and
maintained by the American College of Surgeons and the American Cancer Society. About
1500 Commission on Cancer (CoC) approved hospitals contribute to the registry annually,
representing more than 70% of all incident cancers reported in the United States. Case finding
and data collection items and procedures are detailed elsewhere. 34 The NCDB aggregates
anonymized patient information on tumor characteristics, disease stage and treatment with
additional masking of patient residential zip codes and hospital names. The current study did
not involve attempts to contact patients, identify them or link the database, or portions thereof,
with other data sources. This study was approved by the Moffitt Cancer Center and University of
South Florida IRB committee.
Inclusion/exclusion criteria
ICC cases were identified using the third edition of the WHO manual on the International
Classification of Diseases for Oncology, ICD-O-3, topographic (C22.1) and morphologic codes
7
(8140, 8160, 8255, 8260, 8453, 8480, 8481, 8503). Only patients who were diagnosed with
primary ICC, seen in one CoC hospital, and were worked up for treatment by the reporting
hospital were included in the analysis. We excluded stage 4 patients, patients with carcinoma-
in-situ and those without survival information, clinical stage or malignant tumor behavior.
Importantly, to limit the effects of survival bias among those who received treatment, we
additionally excluded all those who died within 90 days of diagnosis - landmark approach
(Supplemental Figure 2.1).
Intervention/exposure
Four intervention categories were defined among patients and served as the main exposures.
There were patients who received potential curative surgery alone; curative surgery with at least
chemotherapy or radiation of some form; other treatments, primarily chemotherapy with or
without external beam radiation, among others; and no treatment (no surgery, chemotherapy or
radiation therapy reported). Curative surgery was defined as liver resection with or without bile
duct excision in a patient who did not receive palliative care. No distinction was made among
patients who underwent surgery and received neoadjuvant or adjuvant chemotherapy.
Whenever the treatment groups required dichotomization, both surgical-based categories were
compared against those in whom patients did not receive any of the major interventions or
received other non-surgical treatment.
Outcomes
Overall survival was the primary outcome of interest and was measured in months from
diagnosis until death or date of last follow-up. Secondary outcomes included treatment
utilization using the categories listed above. Trends and patterns of treatment were
examined over time, and predictors of curative intent surgery were evaluated with logistic
regression.
8
Covariates
Age was an effect modifier and defined as a binary variable, less than 70 and equal to or
greater than 70 years of age. The year of diagnosis ranged from 2004 and 2014. Age was
dichotomized at the median age distribution of ICC for the ease of interpretation of results.
Patients were also categorized by race (white, black, others), comorbidity index (scores 0, 1, 2,
≥ 3) and primary type of insurance (no insurance, private, government). Other patient
characteristics, which were aggregated at the zip code level, included percent of residents with
a high school diploma (< 14%, 14-19.99%, 20-28.9%, ≥29%), median household income (<
$30,000, $30,000-34,999, $35,000-45,999, ≥$46,000 residency classification (urban, rural)
distance from residential zip code centroid to reporting hospital (miles). Hospitals were classified
by type (community cancer program, academic/research program, integrated network program)
and region (northeast, south, midwest, west) in which they were located.
Tumor characteristics were defined by grade and clinical stage. During the study period, the
sixth and seventh editions of the American Joint Committee on Cancer (AJCC) staging
conventions were used, the latter being adopted from 2010 onwards. To minimize bias from
differential staging, all staging information available was modified according to a system
proposed by Meng and colleagues. 35
Briefly, stages 1 and 2 were left unmodified; stages 3 and 3a of the sixth edition were recoded
as 3a while 3b remained unmodified; stages 4a of the seventh edition was recoded as 3b, with
4b remaining as stage 4; stage 4 of the sixth edition remained unmodified in so far as
metastasis was additionally documented.
Propensity score development
Propensity score stratification of all study participants was undertaken because there was a
need to preserve as many patients within the four treatment categories as possible. The choice
9
of baseline variables that were regressed on the odds of undergoing curative surgery were
guided in part by their level of prediction of surgical treatment and their identification in the
literature as confounders of the treatment-survival association, as well as being documented
predictors of survival in the absence of a relationship to treatment. We chose region, hospital
type, year of diagnosis, primary insurance, median household income, residency, Charlson
comorbidity index and clinical stage as baseline covariates for the generation of propensity
scores. The quality of the propensity score model was judged based on graphical demonstration
of the common support region between treatment groups, absolute standardized mean
differences of 0.25, and their associated p-values.
Statistical analysis
Sociodemographic, tumor and treatment characteristics were stratified by age and significance
testing with Chi-Square tests performed. Overall and age-group trend and treatment type
utilization over the study period were graphically displayed and tested for significance using the
Cochran-Armitage trend and Chi-Square tests, respectively. Logistic regressions models were
used to identify independent predictors of surgical treatment and to generate propensity scores
(probabilities) associated with receiving surgical treatment. Kaplan Meier survival plots stratified
by treatment categories were performed for the overall study sample, and for patients by age
group. Log rank tests were used to assess for differences in survival estimates across treatment
categories. A propensity score-stratified Cox model of treatment categories, age-groups and an
age-treatment interaction term was performed to evaluate hazard ratios (HRs) and estimate
95% confidence intervals (CIs) across levels of treatment and age. Only pooled HRs over the
five propensity-score strata were reported. All Cox models accounted for clustering of survival
among patients with the same facility by estimation of robust variance estimators; all analysis
was performed using SAS version 9.4 (SAS Institute, Cary, NC) and a type 1 error rate of
0.05% was specified a-priori.
10
Results
A total of 3,653 study participants were available for analysis, of which 1,316 (36%) were at
least 70 years of age (the elderly). The elderly were more likely than younger patients to seek
care in a community cancer program (33.9% elderly vs. 25.6% younger, p-value=<0.0001), to
present with earlier stage disease (56% elderly with stages 1 and 2 disease vs. 50% younger
with stages 1 and 2 disease, p-value=<0.0001), to not receive treatment (34.1% elderly vs
19.5% younger, p-value <0.0001) and to report some form of government insurance as the
primary source of payment (86.2% elderly vs. 37.0% younger). When elderly patients received
treatment, surgery-based multimodal treatment (with systemic chemo or radiation) was less
likely (3.9% elderly vs. 8.4% younger, p <0.0001) (Table 2.1).
Trends and patterns of treatment over time
Overall trends of surgery-based treatment demonstrated an increase from 2004 to 2010 (20% to
35% utilization), followed by a gradual decline (Figure 2.2a). When utilization was categorized
by the four treatment types and followed over time, other treatment use, such as chemotherapy
alone or in combination, increased from 42% in 2004 to 55% in 2014, while the level of patients
who did not receive treatment during the same period fell from 39% to 23%. The elderly
compared to younger patients experienced the largest drop in proportions of not treated from
the start of the study (56% old vs. 27.5% young) to study’s end (30% old vs. 17% young),
despite a third of them not receiving any of the major types of treatment. Conversely, over the
study duration, receipt of other treatment by the elderly went up from 30% to 45% while among
the younger patients from 49% to 59% (Figures 2.2b and 2.2c).
Predictors of curative-intent surgical treatment
After controlling for all other variables in a multivariable logistic regression, the elderly remained
less likely to receive any surgery-based form of treatment (OR=0.68, 95%CI: 0.56-0.81; p-value
11
<0.0001) as did patients diagnosed with later stages when compared to stage 1 disease
(OR=0.53, 0.43-0.64 stage 2; OR=0.27, 0.21-0.35 stage 3A; OR=0.17, 0.14-0.22 stage 3B; p-
value < 0.0001). Additionally, hospitals affiliated with academic/research and integrated cancer
programs were more likely to perform surgery-based treatment than community cancer
programs (OR=2.54, 2.09-3.08, OR=1.76, 1.30-2.40, respectively; p<0.0001) (Table 2.2).
Survival analysis
Overall median follow-up and survival of all study participants was 48.2 and 15.8 months,
respectively. There was a statistically significant difference in overall survival between treatment
groups (median OS: surgery alone 37.8 months, surgery/multimodality 33 months, other 14.4
months, and no treatment 9.2 months; p<0.01 – Figure 2.3a); this association persisted when
stratified by age group (Figures 2.3b-c).
When survival was examined by age-group using the propensity-score stratified model, the
hazard ratios among the younger treated age-group were similar among the corresponding
elderly treated when compared to their respective untreated peers (HR=0.36 [95%CI 0.31-0.43]
younger vs. HR= 0.33 [0.27-0.40] elderly for curative surgery; HR=0.39 [0.31-0.48] vs. HR=0.42
[0.29-0.60] for surgery-based multimodal therapy; HR=0.76 [0.68-0.87] vs. HR=0.75 [0.65-0.87]
for other treatment). When survival was further examined within treatment categories, no
significant survival difference was observed between the elderly and younger patients among
those who received surgery alone (HR=1.14 [0.92-1.41]) and among those who received
surgery-multimodality treatment (HR=1.35 [0.92-2.01]) (Table 2.3).
Propensity scores were generated for 3,431 (93.9%) patients, of which 941 patients underwent
surgery with or without additional chemotherapy or radiation (the treated, for propensity-score
generation purposes).
Among the treated, propensity scores were computed for 887 (94.2%) patients. Propensity
scores were grouped into non-overlapping quintiles, each of which contained approximately 177
12
treated patients and a variable number of control patients. Except for the hospital type variable
(absolute standardized mean difference=0.279, p-value=0.00) and stage 1 patients (p-
value=0.03) in stratum 1, treated and control groups were balanced within the standardized
mean difference and p-value boundaries for negligible differences (Supplemental Table 2.1).
Discussion
Using a large hospital-based cancer registry and employing a quasi-experimental design for the
purposes of causal inference, we examined treatment trends and evaluated average treatment
effects of various therapeutic modalities among ICC patients. During the study period, we
demonstrated a reduction in the proportion of patients who did not receive ICC-directed
treatment, especially among the elderly. In an age-stratified analysis, we observed that all
treatment types conferred significant survival benefits and there were no differences in overall
survival in a side by side comparison of elderly and younger patients, a result that underscores
the importance of instituting treatment in the elderly. Crucially, when we directly compared
younger patients to elderly ones stratified by treatment, we found neither clinically meaningful
nor statistically significant differences in overall survival among those who underwent curative
surgery alone. This finding suggests that the elderly with early-stage ICC disease should be
offered aggressive curative intention surgery based on already established criteria routinely
instituted in younger patients.
In assessing the role of multimodal treatment among the elderly, we observed a 35% increase in
the hazard rate among those who had received surgery in combination with another treatment
type, when compared to younger patients. This effect was larger than either that of elderly
patients receiving no treatment or receiving chemotherapy-based multimodal therapy, even as
the latter treatment is typically associated with later stage disease.
13
While the reason for this is not exactly known, it appeared to be related to the nature of the
disease treated by surgery-based multimodal therapy and not the treatment itself, as older and
younger ICC patients, respectively, experienced substantial benefit from it when compared to
their peers who did not receive any treatment. As potentially curative surgery is combined with
systemic chemotherapy in early stage disease with invasive features, this variant of ICC may
represent an unusually aggressive phenotype to which the elderly with limited functional organ-
system reserves are naturally more susceptible.
Unfortunately, the current results also corroborate many other studies that have reported
substandard care among elderly cancer patients. 32,33,36,37 We identified several indications of
this. First, despite an overall decline in the proportion of untreated elderly patients during the
study period, as many as a third remained untreated. Second, although elderly ICC patients
were diagnosed at an earlier stage than younger patients, there was no attendant increase in
the use of early stage surgery-based treatment. Instead, a large increase in the receipt of
chemotherapy-based treatment was found. Third, the older patients were more likely to seek
care in hospitals associated with community cancer programs, which we demonstrated were
less likely to offer surgical treatment. These findings were consistent with those of a large
population-based study among elderly patients with rectal cancer. 36 That study observed that
older rectal cancer patients were less likely to receive aggressive radical surgery and when
surgery was performed, it often led to local tumor excision without the recommended removal
of regional lymph nodes. In another large study of elderly patients with hepatocellular
carcinoma, a similar finding of reduced likelihood of surgical resection and increased receipt of
transarterial chemoembolization for unresectable cancer was reported, despite the elderly
having smaller and fewer multiple tumors. 38 Lastly, in one of the few large ICC studies, older
patients who had undergone surgery were less likely to receive adjuvant chemotherapy or
radiotherapy. 17
14
The decision to treat the elderly cancer patient is complicated. Several provider and patient
factors play an important role in the decision-making process. Limited treatment guidelines for
the elderly exist due to underrepresentation in clinical trials, the presence of comorbidities, and
pharmacist and physician perception of increased treatment toxicity among the elderly. Other
factors that may impede optimal care are poor life expectancy, time constraints, the focus of
multidisciplinary teams on cancer pathology only, the lack of physician experience with geriatric
care and/or lack of capacity for referral to relevant specialists. 30,39,40 Specifically, physician
expertise and clinicians in academic institutions have been identified as more likely to offer
personalized, multimodal treatment to cancer patients, observations that may explain, in part,
lower receipt of surgery-based treatment in hospitals affiliated with community cancer programs.
41 Much less studied patient-level factors that may also affect treatment decisions include trust in
physicians, the level and quality of communication with physicians, presence of cognitive
impairment, health literacy and numeracy. However, we were unable to explore reasons for
non-treatment in our study beyond the explanation that a specific form of therapy was not part of
the first-line of treatment.
The NCDB has several limitations. First, it reports only all-cause deaths and we were therefore
unable to demonstrate ICC-specific survival. The elderly are more likely to die than the young at
any given time in the larger population, and among our study participants, we estimated that the
elderly had an instantaneous background death rate 20-25% times faster than the young at any
given time point, irrespective of treatment. The implications of this are that our treatment-
stratified survival estimates would be biased in favor of younger patients; the reported HRs may
therefore represent an upper limit of disease-specific mortality rate ratios. We attempted to limit
this bias to that which would occur in a healthy aging population by accounting for the severity of
comorbid conditions among study participants by including the Charlson Comorbidity Index in
15
the propensity score model. Second, the NCDB may not be representative of patients who
reside in rural areas and are diagnosed with early stage disease in physician offices, as CoC
hospitals tend to be clustered in metro and urban areas.
Additionally, propensity score techniques have their own limitations. Most notably, they are
unable to adjust for unmeasured variables and our results are therefore subject to bias from
potential confounders not included in the propensity score model. Furthermore, we were unable
to include 6% of those who received surgical treatment in our results as propensity scores were
not generated for them. While the effect of this on the validity of our results is unknown, we
believe it is negligible.
On the other hand, the present study leverages the strengths of the NCDB in a way that few
others examining cholangiocarcinoma (CC) have. We have aggregated one of the largest
patient cohorts of ICC and excluded extrahepatic cholangiocarcinomas (ECC). Other studies
have investigated CC as a single entity, therefore obscuring prognostic insights into the
heterogenous group of diseases. 42 The well-established observation of the rising incidence of
ICC in developed countries and the unassociated decline of ECC incidence suggest, at a
minimum, both groups of CC have different etiologic drivers. 43 We have been careful to
minimize several potential biases common to retrospective cohort studies assessing cancer
treatment. First, we deduplicated ICC cases by restricting analysis to patients who sought care
in one CoC hospital. Second, we excluded patients with multiple primary tumors preventing
confounding by use of different treatment regimen. Third, by excluding patients who died within
90 days of diagnosis we minimized survival bias. Fourth, by adopting a standard staging
definition across the entire study duration, we mitigated the differential exclusion of stage 3
patients with lymph node invasion that would have been reclassified as stage 4 in the seventh
AJCC edition from 2010 onwards. Had the seventh edition AJCC staging remained unmodified,
this would have artificially biased reported HRs towards the null. Fifth, by including stage in
16
generating our propensity scores, we mitigated the effects of confounding by indication as
stage is both a proxy for treatment and survival outcomes.
In conclusion, we demonstrated that a substantial proportion of elderly ICC patients either
received no treatment or were undertreated. When they did receive treatment, after adjusting for
stage, year of diagnosis, comorbidities, and other potential confounders using propensity score
stratification for the purpose of causal inference, they experienced treatment benefits equivalent
to younger patients for all treatment types. Surgery-based treatment, alone or in combination,
offer the best survival benefit for the elderly with non-metastatic disease. However, there remain
institutional barriers to optimal treatment of elderly ICC patients that need to be overcome.
17
Table 2.1. Sociodemographic, hospital, tumor and treatment characteristics of ICC
patients, n=3,653.
CHARACTERISTICS All
N=3,653 < 70 years
n=2,337 ≥ 70 years
n=1,316
*P
n (%)
Region 0.0001
Northeast 914 (25) 531 (22.7) 383 (29.1)
South 1271 (34.8) 820 (35.1) 451 (34.3)
Midwest 900 (24.6) 606 (25.9) 294 (22.3)
West 568 (15.5) 380 (16.3) 188 (14.3) Hospital Type <.0001
Community cancer program
1044 (28.6)
598 (25.6)
446 (33.9)
Academic/research program 2269 (62.1) 1531 (65.5) 738 (56.1)
Integrated network program 340 (9.3) 208 (8.9) 132 (10.0) Year of Diagnosis 0.6493
2004
119 (3.3)
73 (3.1)
46 (3.5)
2005 128 (3.5) 84 (3.6) 44 (3.3)
2006 141 (3.9) 103 (4.4) 38 (2.9)
2007 212 (5.8) 131 (5.6) 81 (6.2)
2008 279 (7.6) 178 (7.6) 101 (7.7)
2009 327 (9.0) 207 (8.9) 120 (9.1)
2010 436 (11.9) 275 (11.8) 161 (12.2)
2011 430 (11.8) 275 (11.8) 155 (11.8)
2012 469 (12.8) 289 (12.4) 180 (13.7)
2013 502 (13.7) 327 (14.0) 175 (13.3)
2014 610 (16.7) 395 (16.9) 215 (16.3) Sex 0.0025 Male 1754
(48.0) 1166
(49.9) 588
(44.7)
Female 1899 (52.0)
1171 (50.1)
728 (55.3)
Race <.0001
White 3045 (83.4)
1899 (81.3)
1146 (87.1)
Black 293 (8.0)
226 (9.7)
67 (5.1)
Others 315
(8.6) 212
(9.1) 103
(7.8)
18
Table 2.1. (Continued).
Primary Insurance <.0001
No Insurance 247 (6.8)
205 (8.8)
42 (3.2)
Private 1408 (38.5)
1268 (54.3)
140 (10.6)
Government Insurance 1998
(54.7) 864
(37.0) 1134
(86.2) % without HSD 0.1104
Missing 144 (3.9)
101 (4.3)
43 (3.3)
>=29% 622
(17.0) 411
(17.6) 211
(16.0) 20-28.9% 778
(21.3) 511
(21.9) 267
(20.3) 14-19.9% 787
(21.5) 503
(21.5) 284
(21.6) < 14% 1322
(36.2) 811
(34.7) 511
(38.8) Median Household Income 0.0863
Missing 143 (3.9)
100 (4.3)
43 (3.3)
< $30,000 484 (13.2)
320 (13.7)
164 (12.5)
$30,000 - $34,999 607
(16.6) 407
(17.4) 200
(15.2) $35,000 - $45,999 919
(25.2) 564
(24.1) 355
(27.0) $46,000 + 1500
(41.1) 946
(40.5) 554
(42.1) Residency Type 0.1052
Missing 144 (3.9)
91 (3.9)
53 (4.0)
Metro 2928
(80.2) 1857
(79.5) 1071
(81.4) Urban/Rural 581
(15.9) 389
(16.6) 192
(14.6) Charlson Comorbidity Index 0.0002
0 2557 (70.0)
1683 (72.0)
874 (66.4)
1 755 (20.7)
441 (18.9)
314 (23.9)
2 205 (5.6)
118 (5.0)
87 (6.6)
≥3
136 (3.7)
95 (4.1)
41 (3.1)
Tumor Grade 0.0009
Missing 1892 (51.8)
1162 (49.7)
730 (55.5)
Well Differentiated, Differentiated, NOS 204 (5.6)
125 (5.3)
79 (6.0)
Moderately Differentiated, Moderately Well Differentiated, Intermediate Differentiation
874 (23.9)
604 (25.8)
270 (20.5)
Poor/Differentiated 683 (18.7)
446 (19.1)
237 (18.0)
19
Table 2.1. (Continued).
Clinical Stage <.0001
1 1049 (28.7)
607
(26.0)
442 (33.6)
2 856 (23.4)
560 (24.0)
296 (22.5)
3A 515 (14.1)
324 (13.9)
191 (14.5)
3B 1233 (33.8)
846 (36.2)
387 (29.4)
Treatment Type <.0001
No Surgery, No Chemo, No Radiation 905
(24.8) 456
(19.5) 449
(34.1) Curative Surgery Alone 694
(19.0) 445
(19.0) 249
(18.9) Curative Surgery +/- Chemo 247
(6.8) 196
(8.4) 51
(3.9) Other (Chemo Alone, External Beam Radiotherapy +/- Chemo, Etc)
1807 (49.5)
1240 (53.1)
567 (43.1)
*Chi-Square test p-value
20
Figure 2.1. Treatment trend for stages 1-3 ICC patients, n=3,653. Cochran-Armitage: p
= 0.0072
21
Figure 2.2a. Treatment utilization for stages 1-3 ICC patients, n=3,653. Chi-square: p<0.001
22
Figure 2.2b Treatment utilization for stages 1-3 ICC in patients < 70 years, n=2,337.
Chi- square: p<0.001
23
Figure 2.2c. Treatment utilization for stages 1-3 ICC in patients ≥ 70 years, n=1,316.
Chi- square: p=0.0001
24
Table 2.2. Predictors of surgical treatment.
Unadjusted Adjusted*
Odds Ratio
95% CI
Odds Ratio
95%CI
**P
Age ≥ 70 years
0.67
0.58
0.77
0.68
0.56
0.81
<.0001
Region: midwest vs west 1.69 1.32 2.16 1.51 1.14 1.98 0.00
northeast vs west 1.65 1.29 2.12 1.69 1.28 2.22
south vs west 1.56 1.23 1.97 1.53 1.18 2.00
Hospital Type: academic/research program vs community cancer program
2.63 2.20 3.14 2.54 2.09 3.08 <.0001
integrated network program vs community cancer program
1.89 1.43 2.49 1.76 1.30 2.40
Year of Diagnosis: 2005 vs 2004 1.06 0.58 1.91 0.99 0.52 1.90 <.0001
2006 vs 2004 0.73 0.39 1.36 0.68 0.34 1.34
2007 vs 2004 1.20 0.70 2.04 1.15 0.64 2.07
2008 vs 2004 1.81 1.10 2.98 1.81 1.05 3.13
2009 vs 2004 1.41 0.86 2.31 1.21 0.70 2.09
2010 vs 2004 2.60 1.63 4.15 2.31 1.38 3.88
2011 vs 2004 2.04 1.27 3.27 1.76 1.05 2.97
2012 vs 2004 1.85 1.16 2.97 1.81 1.07 3.04
2013 vs 2004 1.71 1.07 2.74 1.43 0.85 2.41
2014 vs 2004 1.58 0.99 2.52 1.39 0.83 2.32
Sex: Female vs Male 1.13 0.99 1.30 1.19 1.01 1.39 0.03
Race: black vs white 0.79 0.60 1.03 0.73 0.54 0.98 0.08
others vs white 0.90 0.69 1.16 0.86 0.64 1.15
Insurance Status: government insurance vs no insurance
1.06 0.78 1.44 1.27 0.90 1.80 0.00
private vs no insurance 1.68 1.23 2.29 1.76 1.25 2.48
% without HSD 1.07 1.00 1.14 0.97 0.88 1.07 0.49
Median Household Income 1.11 1.04 1.19 1.17 1.05 1.30 0.00
Residency Type 1.03 0.99 1.07 1.06 1.01 1.11 0.01
Charlson Comorbidity Index 0.97 0.89 1.06 0.95 0.86 1.04 0.24
Clinical Stage: 2 vs 1 0.56 0.47 0.67 0.53 0.43 0.64 <.0001
3A vs 1 0.28 0.22 0.36 0.27 0.21 0.35
3B vs 1 0.19 0.16 0.24 0.17 0.14 0.22
*adjusted for all other variables **Wald Chi-Square test p-value
25
Figure 2.3a. KM plots for stages 1-3 ICC patients, n=3,651.
26
Figure 2.3b. KM plots for stages 1-3 ICC patients less than 70 years, n=2,337.
27
Figure 2.3c. KM plots for stages 1-3 ICC patients at least 70 years, n=1,315.
28
Table 2.3. Crude and propensity score (PS) stratified-adjusted Cox models for overall survival- age and treatment effects examined.
HR LCL UCL HR LCL UCL
Unadjusted model
Curative surgery vs no treatment 0.28 0.25 0.32 Surgery +/- chemo or RT vs no
treatment
0.32
0.27
0.38
Other treatment vs no treatment 0.70 0.64 0.76
PS stratified model – by age group PS stratified model – by treatment
< 70 years no treatment
Curative surgery vs no treatment 0.36 0.31 0.43 ≥ 70 vs < 70 years 1.26 1.09 1.46 Surgery +/- chemo or RT vs no
treatment
0.39
0.31
0.48
Other treatment vs no treatment 0.76 0.68 0.87 curative surgery
≥ 70 vs < 70 years 1.14 0.92 1.41
≥ 70 years
Curative surgery vs no treatment 0.33 0.27 0.40 surgery +/- chemo or RT Surgery +/- chemo or RT vs no
treatment
0.42
0.29
0.60
≥ 70 vs < 70 years
1.35
0.91
2.01
Other treatment vs no treatment 0.75 0.65 0.86
other treatment
≥ 70 vs < 70 years 1.23 1.10 1.38
Abbreviations: HR – hazard ratio, LCL –95% lower confidence limit, UCL – 95% upper confidence limit
29
Supplemental results
Table S2.1. Standardized mean differences by propensity score strata and treatment group.
Treated (curative surgery)
Control (no curative surgery)
Treated - Control
Stratum Index
N Mean N Mean Standardized Mean
**p- value
Difference
Propensity score 1 177 0.14 1358 0.12 0.133 <0.01
2 178 0.25 483 0.24 0.021 0.21
3 177 0.35 335 0.35 0.011 0.51
4 178 0.46 257 0.46 0.008 0.64
5 177 0.58 111 0.57 0.098 0.02
Region 1 177 4.82 1358 4.84 -0.008 0.93
2 178 4.52 483 4.38 0.058 0.41
3 177 4.52 335 4.32 0.083 0.39
4 178 4.05 257 4.20 -0.063 0.48
5 177 3.85 111 4.01 -0.067 0.57
Hospital type 1 177 2.77 1358 2.60 0.279 0.00
2 178 2.82 483 2.77 0.089 0.62
3 177 2.90 335 2.88 0.039 0.89
4 178 2.99 257 3.05 -0.102 0.11
5 177 3.04 111 3.01 0.050 0.36
Year of diagnosis 1 177 2010.63 1358 2010.20 0.157 0.10
2 178 2010.53 483 2010.64 -0.038 0.64
3 177 2010.61 335 2010.90 -0.108 0.12
4 178 2010.99 257 2011.03 -0.014 0.90
5 177 2011.02 111 2011.07 -0.018 0.97
Primary insurance 1 177 2.41 1358 2.41 0.001 0.41
2 178 2.39 483 2.12 0.168 0.32
3 177 2.19 335 2.34 -0.099 0.43
4 178 2.14 257 2.21 -0.042 0.95
5 177 1.73 111 1.79 -0.037 0.34
Median Household Income
1 177 3.02 1358 2.88 0.134 0.10
2 178 3.12 483 3.04 0.083 0.28
3 177
2.81 335 2.93 -0.110 0.31
30
Table S2.1. (Continued).
* mean represents proportion of patients not of a given stage within each stratum
**Z-score test p-value
4 178 2.97 257 3.09 -0.112 0.26
5 177 3.26 111 3.30 -0.035 0.63
Residency type 1 177 1.97 1358 2.02 -0.031 0.31
2 178 1.90 483 2.19 -0.150 0.18
3 177 2.38 335 2.17 0.108 0.45
4 178 2.49 257 2.31 0.094 0.95
5 177 2.80 111 2.61 0.100 0.55
Charlson Comorbidity Index
1 177 0.46 1358 0.39 0.096 0.22
2 178 0.44 483 0.42 0.029 0.60
3 177 0.46 335 0.46 0.005 0.75
4 178 0.47 257 0.52 -0.066 0.51
5 177 0.49 111 0.59 -0.122 0.52
Stage 1* 1 177 0.95 1358 0.98 -0.066 0.03
2 178 0.78 483 0.83 -0.111 0.14
3 177 0.56 335 0.53 0.054 0.59
4 178 0.26 257 0.27 -0.001 0.99
5 177 0.09 111 0.07 0.028 0.70
Stage 2* 1 177 0.86 1358 0.89 -0.060 0.31
2 178 0.62 483 0.59 0.055 0.58
3 177 0.51 335 0.55 -0.081 0.45
4 178 0.74 257 0.74 0.001 0.99
5 177 0.92 111 0.93 -0.029 0.70
Stage 3A* 1 177 0.80 1358 0.79 0.013 0.90
2 178 0.81 483 0.80 0.030 0.78
3 177 0.94 335 0.94 0.002 0.98
4 178 1.00 257 1.00 0.000
5 177 1.00 111 1.00 0.000
Stage 3B* 1 177 0.39 1358 0.34 0.119 0.17
2 178 0.80 483 0.78 0.040 0.63
3 177 0.99 335 0.98 0.022 0.43
4 178 1.00 257 1.00 0.000
5 177 1.00 111 1.00 0.000
31
[Type a quote from the
document or the summary of an interesting
point. You can position the text box
Intrahepatic biliary malignancies in NCDB
n=23,273
ICD-O-3 morphologies not defined as intrahepatic
cholangiocarcinoma
n=1,896
Intrahepatic cholangiocarcinoma (ICC) n=21,377
Other patient exclusion criteria: ▪ Patients seen for
diagnostic purposes only: 2,675
▪ multiple primary site tumors: n=3,544
▪ single primary tumors seen in more than 1 CoC facility: n=2,153
▪ borderline malignant behavior: n=11
▪ carcinoma-in-situ: n=8 ▪ unknown vital status or
follow-up time: 1,906 ▪ missing or unknown
clinical stage: 3,124 ▪ patient deaths within 90
days of diagnosis: n=1,110
ICC patients seen in 1 CoC facility for treatment and included in analysis
n=3,653
Patients with stage IV disease
excluded: n=3,194
Figure S2.1. Inclusion and exclusion criteria of study participants, NCDB (2004-2014).
32
Figure S2.2. Treatment trend for stages 1-4 ICC patients, n=21,377. Cochran-Armitage: p
=0.702
33
CHAPTER 3: NEOADJUVANT CHEMOTHERAPY FOR PATIENTS WITH INTRAHEPATIC
CHOLANGIOCARCINOMA
Abstract [276 words]
Introduction: Although liver resection provides the only potentially-curative approach for
intrahepatic cholangiocarcinoma (ICC), locoregional and systemic recurrence remain common.
Despite recent studies supporting the use of multimodality therapy for resectable ICC, the
survival benefit in the adjuvant setting is small. Neoadjuvant chemotherapy (NC) is a potential
alternative, though its role for resectable disease is controversial and has not been well
characterized.
Methods: We performed a retrospective cohort study of ICC patients in the National Cancer
Data Base who were treated with curative-intent surgery (2006-2014). NC utilization over time
was evaluated across participating hospitals, and predictors of NC use were identified using
multivariable logistic regression. The effect of NC on overall survival (OS) was examined using
propensity-score matched Cox regression models.
Results: A total of 881 patients met inclusion criteria for the study cohort. NC was used in 8.3 %
of the population. On multivariable analysis, increasing stage (p<0.001) and year of diagnosis
(p=0.03) were independent predictors of NC utilization. In the unadjusted model, there was no
difference in OS between the NC versus non-NC groups (median OS of 51.8 months versus
35.6 months, respectively; p=0.51), however there was a trend towards improved survival in the
NC group for the high-risk population (stages 2-3) (median OS 35.7 versus 26.4 months,
respectively; p=0.1). After adjusting using the propensity-score 1:4 matched cohort, NC
utilization was associated with a trend towards improved OS (HR 0.78 [95%CI 0.54-1.11];
p=0.16).
34
Conclusion: Overall NC utilization for resectable ICC is low, although its use has increased over
time and for those with more advanced disease. Despite no clear survival benefit, this study
shows that patients with more advanced disease may benefit from a multimodal approach using
preoperative chemotherapy.
Introduction The prognosis of intrahepatic cholangiocarcinoma (ICC) is dismal, with median survival of less
than three years and five-year survival of less than 10%. 44,45 Complete resection of small single
tumors without evidence of lymph node, neural, or blood vessel involvement remains the best
option for cure among the 15% of patients eligible for surgery.46-48 Additionally, the National
Comprehensive Cancer Network (NCCN) suggests the use of adjuvant chemotherapy among
patients with early stage ICC in whom lymph node metastasis are present or negative resection
margins cannot be achieved. 49,50 Although neoadjuvant chemotherapy is the lesser examined
of the systemic chemotherapy approaches, there is recent evidence that it may offer better
survival than the current standard of care among patients with locally invasive disease. 51,52
Theoretically, its advantages over adjuvant chemotherapy include better compliance, improved
delivery of drugs to a better vascularized presurgical tumor bed and sterilization of micro-
metastasis, potential tumor downstaging and preselection of patients who may benefit from
surgery and further chemotherapy. 51
Only a few studies have quantified the effect of neoadjuvant chemotherapy in ICC patients and
these have been based on small samples from single institutions. 53-55 Only now are some
randomized clinical trials such as ACTICCA, PRODIGE-12 and BILCAP suggesting survival
benefits of adjuvant chemotherapy over post-operative observation. 56,57 There are no published
neoadjuvant chemotherapy clinical trials that we are aware of. Furthermore, there is an inherent
selection bias of patients in observational studies as those who receive neoadjuvant
chemotherapy tend to be younger albeit diagnosed with later stage disease. 58 As a result,
35
causal inferences of the neoadjuvant chemotherapy effect on ICC survival cannot often be
made.
Based on this context, the aims of the current study are to characterize neoadjuvant
chemotherapy use across a hospital-based cancer registry and evaluate its treatment effect on
survival among patients with ICC having surgical treatment.
Methods Study Design We carried out a retrospective cohort design among ICC patients diagnosed between January
1, 2006 through December 31, 2014 in the National Cancer Database (NCDB). The NCDB is a
hospital-based cancer registry jointly established in 1989 and maintained by the Commission on
Cancer (CoC) of the American College of Surgeons and the American Cancer Society.
Annually, more than 1500 CoC hospitals contribute anonymized patient records to the
database, which currently represents more than 70% of newly diagnosed cancers,
accumulating 34 million cancer cases across the United States. 21The NCDB collects data on
patient characteristics, staging information, tumor histology, first-line treatment and outcomes
from participating hospitals using nationally standardized data item and coding definitions
specified in CoC oncology registry data standards. 59,60 The current study was approved by the
Moffitt Cancer Center and University of South Florida IRB committee.
Inclusion/exclusion criteria
We extracted data on patients who met our criteria for ICC diagnosis during the study period.
Using the third edition of the International Classification of Diseases for Oncology (ICD-O-3), we
defined ICC as topographic code C22.1 and morphological codes 8160, 8140, 8255, 8260,
8453, 8480, 8481 and 8503. We excluded patients with multiple primary tumors, non-malignant
tumors, stage IV disease and for which clinical staging information was missing, unknown or
non-applicable. To minimize duplication of records, we also excluded patients with single
36
primary ICC tumors seen in more than one CoC hospital. Our analysis was restricted to patients
who had potentially curative surgical procedures in a CoC hospital. Curative surgery was
defined as a biliary tract-specific procedure involving a hepatectomy and/or biliary tract
resection in a patient who did not undergo palliative treatment, a tumor destructive procedure or
liver transplant (Supplemental Figure 3.1).
Intervention - neoadjuvant chemotherapy Neoadjuvant chemotherapy was our main exposure and was defined as any receipt of systemic
chemotherapy preceding a curative surgical procedure, irrespective of the presence of a record
of chemotherapy after surgery. Two mutually exclusive non-neoadjuvant chemotherapy
interventions, or unexposed groups, were also defined: adjuvant chemotherapy and surgery
alone. Adjuvant chemotherapy was defined as a receipt date of systemic chemotherapy
following that of curative surgery, while surgery alone was the absence of chemotherapy as part
of the surgical treatment. For analytic purposes, patient treatment in which chemotherapy was
administered between two surgical procedures were neither defined nor included in the present
study. Except for the descriptive analysis of the study sample, adjuvant chemotherapy and
surgery alone interventions were considered under a single referent non-neoadjuvant category.
Outcomes
Overall survival was our primary outcome of interest and was measured from time of diagnosis
until death or date of last follow-up. The secondary outcome was neoadjuvant chemotherapy
utilization rates at the hospital level. A trend analysis was performed to examine changes over
time within the NCDB registry and within consistent participating hospitals during the study
period, and multivariable logistic regression was performed to identify predictors of neoadjuvant
chemotherapy utilization.
Covariates Covariates for assessing independent predictors of neoadjuvant chemotherapy use were
broadly grouped into hospital-, patient- and tumor/treatment-based. The hospital-based
37
covariates included hospital location by region (Northeast, South, Midwest, West) and cancer
program type of hospital (community, academic/research, integrated network). Patient-based
covariates included year of cancer diagnosis, age (measured as both continuous and binary:
< 70 & ≥ 70 years), sex, race (white, black, other), primary insurance type (uninsured, private
insurance, government insurance), Charlson comorbid severity index (0, 1, 2, ≥ 3 point-
score) and distance of residence from the hospital (miles). To enhance the anonymity of
small easily identifiable patient subgroups, the following socioeconomic variables were
collected at the zip code level, which was itself masked: residence type (metro, urban/rural);
percent of residents without high school diploma (<14%, 14-19.9%, 20-28.9%, ≥ 29%);
median household income (<$30,000, ≥$30,000-34,999, >$34,999-45,999, >$45,999).
Finally, tumor grade (well differentiated, moderately differentiated, undifferentiated, poorly
differentiated) and stage (I, II, IIIA, IIIB) comprised the tumor-based covariates. Because of the
small numbers of poorly differentiated ICC, this category was collapsed into that of the
undifferentiated during analysis. Crucially, two editions of the American Joint Committee on
Cancer (AJCC) staging were used across the study time span: the sixth edition from 2006 to
2009 and the seventh from 2010 onwards; to minimize misclassification bias from differential
diagnostic criteria, we recoded all clinical staging information to reflect the current eighth edition
AJCC using a framework proposed by Meng and colleagues.35 In summary, stages 1 and 2
were left unmodified across the study period; stages 3 and 3A of the sixth AJCC edition were
recoded as stage 3A, while stages 3B and 3C were recoded as 3B; stage 4A of the seventh
edition was recoded as 3B while stage 4 remained unmodified.
Propensity score development
To mitigate the effects of selection biases inherent in treatment decisions, we used logistic
regression models to calculate probabilities/propensity scores (PS) of receiving neoadjuvant
chemotherapy among all study participants under the assumption that grouping similarly-scored
38
patients with different treatment would provide a counterfactual for the estimation of an average
treatment effect among treated ICC patients. Baseline variable selection for PS generation was
guided by significant predictors of neoadjuvant chemotherapy use in our study that were also
well-established determinants of survival. However, because of a small event size and the
further loss of neoadjuvant chemotherapy patients when we attempted to include all covariates
in our logistic models, we selected only the following variables for PS generation: hospital type,
patient age, comorbidity index, clinical stage of disease and year of diagnosis. Several PS
techniques were subsequently applied to quantify neoadjuvant chemotherapy effects on
survival, including matching and stratification. We additionally ensured that all patients treated
with neoadjuvant chemotherapy were used in the PS techniques by identifying and linking
uniquely anonymized hospital IDs associated with missing observations of hospital type to non-
missing ones.
Statistical analysis Baseline hospital, patient socioeconomic and tumor characteristics were stratified by treatment
categories, with differences in the distribution of proportions and medians of these
characteristics assessed by chi-square and Kruskal-Wallis tests, respectively. Trends of
neoadjuvant chemotherapy utilization overtime was examined at the hospital-level using the
Cochran-Armitage trend test for assessment of statistical significance, while logistic regression
with odds ratios (ORs) and 95% confidence intervals (CIs) were constructed to identify
independent predictors of neoadjuvant chemotherapy use. Kaplan-Meier (KM) plots were used
to describe survival probabilities among the entire sample, and subset populations stratified by
stage. Several Cox models were used to estimate hazard ratios (HRs) and 95% CIs associated
with neoadjuvant chemotherapy use on survival. The first of the Cox models was an unadjusted
estimate of the neoadjuvant chemotherapy effect while the others incorporated propensity score
techniques. The second Cox model PS-matched neoadjuvant to non-neoadjuvant recipients in a
fixed 1:1 ratio within a 0.25 caliper and specifying the shortest Malahanobis distance of the age
39
and year of diagnosis variables between exposed and unexposed groups. The third Cox model
PS-matched recipients in a variable 1:4 ratio while specifying a matching region defined by all
neoadjuvant chemotherapy patients using an optimal algorithm. The fourth Cox model
estimated a pooled neoadjuvant chemotherapy effect across five propensity score strata, into
which study participants were grouped. All Cox models additionally accounted for potential
clustering of survival effects within unique hospitals by specifying robust estimates of the
potential correlation of such effects within hospitals. Other than attempts to identify missing
hospital types, the results presented here were analyzed by complete case analysis. A 0.05
level of significance was used and SAS version 9.4 (Cary, North Carolina) was used for all
analyses.
Results A total of 881 ICC patients with known clinical staging information who had undergone
potentially curative surgery in a CoC-accredited hospital met inclusion criteria and represented
our study sample (2006 – 2014). Most of these patients presented with stage I disease (52.8%),
were white (85.0%), female (55.1%), less than 70 years of age (65.7%) and were treated in an
academic/research setting (66.6%) with surgery alone (66.1%). Overall, only 8.3% of patients
received neoadjuvant chemotherapy (NC). Patients who received NC tended to be younger
(79.7% less than 70 years vs 65.7% in overall sample, p < 0.01), sought treatment farther away
from residence (median 38 miles vs 25 in overall sample, p < 0.01) and presented with more
stage III disease (26.7% vs 15.1% in overall sample, p < 0.01) (Table 3.1).
Neoadjuvant chemotherapy utilization – trends and predictors
NC use was reported by a total of 266 unique hospitals in every year except 2007, but rarely
accounted for more than 10% of the treatment strategy utilized annually. The overall pattern of
use over time was not statistically significant. The annual number of reporting hospitals ranged
from 25 in 2006 to 101 in 2014. When we restricted the trend analysis to hospitals that
consistently reported patients throughout the study period, NC use approached 20% of the
40
annual treatment mix and the trend moved towards statistical significance without reaching it.
Of 27 such hospitals, 26 (96.2%) were affiliated with academic/research programs (Figures
3.1a & 3.1b). On multivariable logistic regression, region, year of diagnosis and stage
independently predicted NC use (Table 3.2). The mid-western region was least likely than the
others to use NC (OR=0.31, 95% CI=0.09-1.01; p=0.04). NC use increased with advancing
year of diagnosis (OR=1.20, 1.01-1.43; p=0.03) and patients with stage 3A disease were more
likely to receive NC (OR=4.33, 1.44-13.04; p < 0.001).
Survival analysis The median follow-up time and overall survival for the whole population was 50.9 months and
36.4 months, respectively. On unadjusted analysis, although there was no statistically significant
difference between the NC and non-NC groups, there was a trend towards improved survival for
the intervention group (median OS 50.8 vs. 35.6 months, respectively; p=0.5). When stratifying
by stage of disease, the trend favoring neoadjuvant chemotherapy utilization was stronger for
those with advanced disease (stages 2 and 3) with median survival of 47.6 vs. 25.9 months
(p=0.1), while it disappeared for those with early stage presentation (stage 1) (Figures 3.2a-c).
Table 3.2 lists the results of the Cox regression models. On univariate analysis, neoadjuvant
chemotherapy was associated with a non-significant trend of improved survival (HR 0.92 [95%CI
0.64-1.31]; p=0.66). Each of the propensity score models included all 74 patients who received
NC. For both the matched (1:1 and 1:4) and the propensity stratified models, the trend persisted,
though it was not statistically significant. Notably, the performance of the matched cohorts was
excellent; using the 1:1 PS matched cohort, all standardized mean differences between NC and
non-NC patients associated with propensity-adjusted variables were below 0.1, except for the
academic category of the hospital type variable (Supplemental Table 3.1). Similarly, for the
stratified model, there were approximately 15 NC patients in each of the five strata. The
frequency distribution of NC and non-NC patients in each of the five strata for the PS stratified
model is show in Supplemental Figure 3.2.
41
Discussion
In this large retrospective study of ICC patients that received potentially curative surgery, we
observed that there was a small but increasing trend towards NC use and, on average, its use
appeared to have a small association-trend to improved survival over the receipt of surgery
alone or the use of adjuvant chemotherapy. NC use was primarily driven by hospitals
associated with academic/research cancer programs, a finding that reflects the complexity of
the current recommendation that NC be considered in initially unresectable disease. 50 Our
robust findings are compatible with the notion that some patients, who seek care in non-
academic/research settings, may benefit from NC use and the improved outcomes observed in
some hospitals may be related, in part, to it. Furthermore, we observed what appears to be a
delayed survival benefit for NC patients who survived at least two years post-diagnosis, the
veracity and the exact nature of which are unknown and require further investigation.
Until recently, large study evidence of the survival benefits of NC has come from its use in other
gastrointestinal malignancies. In a 2016 study of 12,857 pancreatic cancer patients, 58 of which
12% of those who were eligible for surgical resection used NC, a stage-dependent advantage of
NC over adjuvant chemotherapy (AC) use was observed. Stage III patients had a median
overall survival (OS) of 22.9 vs 17.3 months for the AC group, while stages II and I patients
receiving NC experienced longer median OS (26.2 vs. 25.7 months AC use and 23.3 vs. 23.0
months AC use, respectively). Similarly, in evaluating two approaches of NC use among 10,086
patients with esophageal cancer, when compared to that to which watchful waiting was
employed post-operatively, a 2017 study demonstrated the group that continued chemotherapy
after surgery had a 21% lower eight-year risk of death. 61 Smaller, older studies involving
neoadjuvant chemoradiation in extrahepatic cholangiocarcinoma (ECC) have also reported
advantages. 62,63 In one of such studies, 12 (26%) of 45 patients were treated with neoadjuvant
therapy with fluoropyrimidine-based chemotherapy and radiotherapy. The reported five-year
42
survival rates were 53% and 23% for the NC and AC patients, respectively, despite the former
initially having unresectable disease. 63
Our main finding cannot be directly compared to prior studies. To our knowledge, the present
study is the first to aggregate only ICC patients and attempt to evaluate chemotherapeutic
treatment effects using causal inference methods. A study involving all types of
cholangiocarcinoma was recently undertaken in which NC use was compared directly to AC
use, while applying a one-to-many propensity score matching technique. 42 Despite the inclusion
of a non-baseline radiotherapy use variable in generating propensity scores and not
distinguishing between ICC and other types of cholangiocarcinoma, they estimated an average
NC treatment hazard ratio of 0.78 (95% CI=0.64-0.94), a finding compatible with those of the
present study. Specifically, our one-to-many matching estimate is nearly identical to their result,
an observation that suggests NC may have the same effect on different types of CC, albeit we
argue this estimate may be biased. 64 Nonetheless, that study was able to achieve better
precision in its confidence limits due to its larger sample size. The Malahanobis distance-
adjusted one-to-one propensity score match, however, may have offered better validity in
sacrifice of precision by ensuring instead only the best matches for the NC group were used.
When compared to ICC-only studies from western centers, our other findings appear to fall
outside the range of reported results. In a small French study of 74 surgical ICC patients with
locally advanced disease, of which 39 (53%) were treated with NC, it was found that the NC
group had a median survival of 24.1 vs. 25.7 months among those who received surgery
alone.58 Among 45 ICC patients who underwent surgical resection in a single German
institution, 5 (11.1%) were given NC while 23 received AC. Although survival was not stratified
by chemotherapy sequence in relation to surgery, among those who had a resection with
negative tumor margins, chemotherapy use was associated with a median survival of only 15
43
months. 65 It is pertinent to report the recent and first-ever findings of phase III clinical trials of
AC use among patients with advanced biliary cancers (ABC trials). 66 A total of 109 ICC
patients from the United Kingdom ABC-01, -02 and -03 trials who received first line treatment
with Cisplatin and Gemcitabine, of whom 52 had liver-only disease, were retrospectively
reviewed and observed to have a median OS of 15.4 months after a median follow-up of 12.2
months from a larger cohort of biliary cancer patients. Importantly, of the 52 patients, 44% had
non- metastatic disease while 26.9% received prior treatment including surgery. We, however,
demonstrated higher median OS with NC use in general and among patients with stage 3
ICC, which would correspond to locally invasive non-distant metastatic disease. This
departure from other western centers is likely a result of the use of NC in earlier stages of
disease and more aggressive treatment with surgery among all patients with later stage
disease in our study.
Additionally, we cannot rule out the survival effects of hospital-related structural and process of
care factors, such as the presence of multidisciplinary specialists and monitoring of operating
times and blood loss, which have been shown to vary widely in the treatment of pancreatic
cancer and may further explain systematic patient survival variation in closely related ICC. 67
The current findings should be interpreted with caution. It is worth noting that the NCDB is a
hospital-based cancer registry. As such, it is likely to aggregate more severe ICC cases and
possibly underestimate its survival in the larger population. Also, CoC hospitals which contribute
cancer cases to NCDB have been identified as being more urban, larger and providing more
cancer services than non-CoC hospitals.68 This suggests that patients treated in non-CoC
hospitals may have poorer outcomes than those treated in CoC hospitals, possibly resulting in
less of a survival underestimation by NCDB. Not all patients who used NC are captured in the
present study, namely, those who may have received NC prior to liver transplantation. These
patients have limited functional liver reserves to begin with, the survival effect of NC in this
44
setting has not been comprehensively evaluated and as such our findings cannot be
generalized to them. Furthermore, the results presented here are subject to bias from
confounders not included in the propensity-score generating model.
However, this study has several key advantages. NCDB accounts for some 70% of incident
cancer cases in the United States, making it one of the most representative cancer registries
available. We have been careful to account for one of the most important biases in ICC
treatment studies, i.e. treatment selection of patients by disease stage, which itself is a predictor
of survival. By applying propensity score techniques, we were able to compare patients
diagnosed with the same clinical stage but receiving different surgery-based therapies.
Crucially, we used all patients who received NC in the setting of potentially curative surgery in
our propensity score matching, and therefore minimized bias associated with partial matching.
In conclusion, in this propensity score-adjusted study using NCDB registry data, we
demonstrated a trend towards a survival advantage of NC use among ICC patients over the
more prevalent practices of surgery alone and in combination with AC. The use of systemic
chemotherapy in the treatment of ICC, irrespective of its sequence in relation to surgery,
presupposes disease with poor prognostic features such as invasion of nearby major blood or
lymphatic vessels. For many patients with advanced non-metastatic disease, NC use may offer
distinct advantages over AC by improving drug compliance and delivery which are often
hampered by post-operative complications in the AC setting, and by providing an opportunity for
pre-operative tumor downstaging and subsequent R0 resections among patients with initially
unresectable disease. Despite the use of large data, we were unable to aggregate a large
sample of patients who received NC. As the same drugs with the same safety profile are used in
the NC and AC approaches, it seems reasonable to offer more patients NC in situations where
contraindications are not present, especially in a randomized clinical trial setting. It is likely that
NC affects prognosis by working in concert with institutional best practice and patient factors.
45
There is a need to identify these factors and quantify their survival effects. In particular, the
tumor microenvironment and its many cellular and molecular interactions remain mostly
mysterious.
46 9
Table 3.1. Baseline sociodemographic, hospital, clinical and tumor characteristics of patients with non-metastatic intrahepatic
cholangiocarcinoma, treated with surgery, by treatment strategy (n=881).
Treatment Strategy
All
(n=881)
Surgery
alone
(n=583)
Surgery +
adjuvant
therapy
(n=224)
Neoadjuvan
t therapy +
Surgery
(n=74)
N (%) N (%) N (%) N (%) *p-
value
Region 0.01
Missing 28 (3.2) 10 (1.7) 12 (5.4) 6 (8.1)
Northeast 219 (24.9) 126 (21.6) 71 (31.7) 22 (29.7)
South 312 (35.4) 226 (38.8) 68 (30.4) 18 (24.3)
Midwest 219 (24.9) 155 (26.6) 47 (21.0) 17 (23.0)
West 103 (11.7) 66 (11.3) 26 (11.6) 11 (14.9)
Hospital type 0.25
Missing 28 (3.2) 10 (1.7) 12 (5.4) 6 (8.1)
Community cancer program 181 (20.5) 119 (20.4) 51 (22.8) 11 (14.9)
Academic/research program 587 (66.6) 400 (68.6) 135 (60.3) 52 (70.3)
Integrated network program 85 (9.6) 54 (9.3) 26 (11.6) 5 (6.8)
Year of diagnosis 0.08
2006 28 (3.2) 13 (2.2) 12 (5.4) 3 (4.1)
2007 46 (5.2) 32 (5.5) 14 (6.3) 0 0.0
2008 83 (9.4) 55 (9.4) 24 (10.7) 4 (5.4)
2009 80 (9.1) 48 (8.2) 23 (10.3) 9 (12.2)
47 9
Table 3.1. (Continued).
2010 144 (16.3) 94 (16.1) 41 (18.3) 9 (12.2)
2011 119 (13.5) 77 (13.2) 32 (14.3) 10 (13.5)
2012 121 (13.7) 81 (13.9) 23 (10.3) 17 (23.0)
2013 114 (12.9) 82 (14.1) 21 (9.4) 11 (14.9)
2014 146 (16.6) 101 (17.3) 34 (15.2) 11 (14.9)
Median age at diagnosis (25%, 75%) 64(57,72)
66(59, 73) 62(54, 69) 62 (53, 67) <0.01
Median age at diagnosis - groups <0.01
< 70 years 579 (65.7) 347 (59.5) 173 (77.2) 59 (79.7)
70 years + 302 (34.3) 236 (40.5) 51 (22.8) 15 (20.3)
Gender 0.02
Male 396 (44.9) 281 (48.2) 88 (39.3) 27 (36.5)
Female 485 (55.1) 302 (51.8) 136 (60.7) 47 (63.5)
Race 0.20
White 749 (85.0) 495 (84.9) 187 (83.5) 67 (90.5)
Black 60 (6.8) 44 (7.5) 12 (5.4) 4 (5.4)
other 72 (8.2) 44 (7.5) 25 (11.2) 3 (4.1)
Primary insurance <0.01
Uninsured 54 (6.1) 44 (7.5) 6 (2.7) 4 (5.4)
Private insurance 366 (41.5) 202 (34.6) 127 (56.7) 37 (50.0)
Government insurance 461 (52.3) 337 (57.8) 91 (40.6) 33 (44.6)
% without high-school education 0.13
Missing 33 (3.7) 20 (3.4) 12 (5.4) 1 (1.4)
>=29% 147 (16.7) 112 (19.2) 27 (12.1) 8 (10.8)
48
Table 3.1. (Continued).
20-28.9% 178 (20.2) 120 (20.6) 40 (17.9) 18 (24.3)
14-19.9% 198 (22.5) 131 (22.5) 51 (22.8) 16 (21.6)
< 14% 325 (36.9) 200 (34.3) 94 (42.0) 31 (41.9)
Median household income 0.30 Missing 33 (3.7) 20 (3.4) 12 (5.4) 1 (1.4)
< $30,000 111 (12.6) 81 (13.9) 22 (9.8) 8 (10.8)
$30,000 - $34,999 138 (15.7) 94 (16.1) 32 (14.3) 12 (16.2)
$35,000 - $45,999 219 (24.9) 151 (25.9) 46 (20.5) 22 (29.7)
$46,000 + 380 (43.1) 237 (40.7) 112 (50.0) 31 (41.9)
Residency location 0.36
Missing 31 (3.5) 21 (3.6) 8 (3.6) 2 (2.7)
Metro 678 (77.0) 446 (76.5) 178 (79.5) 54 (73.0)
Urban/rural 172 (19.5) 116 (19.9) 38 (17.0) 18 (24.3)
Median distance to treating hospital,
in miles (25%, 75%) <0.01
25(9,73) 26(9,82) 18(8,48) 38(18,85)
Charlson comorbidity index 0.15
0 599 (68.0) 379 (65.0) 168 (75.0) 52 (70.3)
1 183 (20.8) 129 (22.1) 37 (16.5) 17 (23.0)
2 65 (7.4) 49 (8.4) 13 (5.8) 3 (4.1)
≥3 34 (3.9) 26 (4.5) 6 (2.7) 2 (2.7)
Tumor Grade <0.01
Missing 139 (15.8) 97 (16.6) 22 (9.8) 20 (27.0)
Well differentiated, differentiated, NOS 92 (10.4) 67 (11.5) 19 (8.5) 6 (8.1)
49
Table 3.1. (Continued).
Moderately differentiated, moderately
well differentiated, intermediate
differentiation
425 (48.2) 287 (49.2) 108 (48.2) 30 (40.5)
Poorly differentiated 213 (24.2) 123 (21.1) 73 (32.6) 17 (23.0)
Undifferentiated, anaplastic 12 (1.4) 9 (1.5) 2 (0.9) 1 (1.4)
Clinical stage <0.01
0 2 (0.2) 1 (0.2) 1 (0.4) 0 0.0
1 465 (52.8) 346 (59.3) 94 (42.0) 25 (33.8)
2 281 (31.9) 161 (27.6) 90 (40.2) 30 (40.5)
3A 93 (10.6) 50 (8.6) 28 (12.5) 15 (21.3)
3B 40 (4.5) 25 (4.3) 11 (4.9) 4 (5.4)
Lymph node resection 0.01
No 416 (47.2) 295 (50.6) 88 (39.2) 33 (44.5)
Yes 465 (52.8) 288 (49.4) 136 (60.7) 41 (55.5)
Surgical margins
R0 655 (74.3) 468 (80.3) 139 (62.1) 48 (64.9)
R1/R2 176 (20.0) 86 (14.8) 72 (32.1) 18 (24.3)
margin status unknown 50 (5.7) 29 (5.0) 13 (5.8) 8 (10.8)
Radiation treatment <0.01
None 751 (85.2) 570 (97.8) 125 (55.8) 56 (75.7)
External beam 114 (12.9) 7 (1.2) 93 (41.5) 14 (18.9)
Other 16 (1.8) 6 (1.0) 6 (2.7) 4 (5.4)
*Chi-Square test p-values for categorical variables and Kruskal-Wallis test p-values for continuous variables
50
Figure 3.1a. Annual proportion of patients with intrahepatic cholangiocarcinoma
receiving neoadjuvant chemotherapy, in participating NCDB hospitals (n=881). Cochran-
Armitage p- value=0.17
51
Figure 3.1b. Trend over time for neoadjuvant chemotherapy utilization among 26 hospitals with
consistent patient reporting through study period (n=340). Cochran-Armitage p-value= 0.1
52
Table 3.2. Multivariable logistic regression analysis examining predictors of neoadjuvant chemotherapy utilization (n=881).
Crude Odds Ratio
95% Confidence
Interval
*Adjusted Odds Ratio
95% Confidence Interval
**p-value adjusted
Odds Ratio Region 0.04
Midwest vs west 0.46 0.16, 1.32 0.31 0.09, 1.01
Northeast vs west 0.93 0.36, 2.37 0.82 0.28, 2.38 South vs west 0.55 0.21, 1.44 0.51 0.17, 1.53
Hospital type 0.12 Academic vs community
1.62 0.70, 3.73 1.95 0.76, 4.97
Integrated vs community
-- --, -- -- --, --
Year of diagnosis 1.10 0.95, 1.27 1.20 1.01, 1.43 0.03
Age 70+ vs <70
0.59 0.29, 1.19 0.53 0.24, 1.19 0.07
Sex
Female vs male 1.34 0.72, 2.52 1.61 0.79, 3.28 0.14
Race 0.24 B vs. W 0.88 0.26, 2.97 0.87 0.22, 3.34
Other vs. W 0.23 0.03, 1.73 0.14 0.01, 1.17 Primary Insurance
0.36
Government vs no insurance
1.57 0.20, 12.18 1.41 0.16, 12.13
Private vs no insurance
1.88 0.24, 14.69 1.29 0.15, 11.02
Charlson Comorbidity Index
0.64
0 vs 3 1.82 0.23, 13.92 1.22 0.14, 10.29
1 vs 3 2.67 0.33, 21.35 2.11 0.24, 18.45 2 vs 3 1.80 0.17, 18.18 1.22 0.10, 13.08
Residency Metro vs.
urban/rural
0.95 0.44, 2.04 1.06 0.41, 2.71 0.70
Distance to hospital
1.00 0.99, 1.00 1.00 0.99, 1.00 0.63
Grade Moderate vs well
differentiated 1.16 0.43, 3.11 0.84 0.28, 2.51
Poor vs well differentiated
1.17 0.40, 3.41 0.94 0.29, 3.05
Undifferentiated vs well differentiated
-- -- , -- -- --, --
Stage 0.00 2 vs 1 2.34 1.18, 4.63 2.72 1.27, 5.82
3A vs 1 3.00 1.17, 7.65 4.33 1.44, 13.04
53
Table 3.2. (Continued).
3B vs 1 0.85 0.10, 6.73 2.70 0.26, 27.44
No high school diploma
0.59
<14% vs >=29 1.45 0.51, 4.12 1.40 0.35, 5.50 14%-19.9 vs >=29 1.99 0.68, 5.81 1.80 0.49, 6.51 20%-28.9 vs >=29 2.55 0.88, 7.39 2.17 0.63, 7.40 Median household income
0.56
$30,000-34,9999 vs <30,000
1.48 0.43, 5.09 1.13 0.28, 4.59
$35,000-45,999 vs <30,000
2.02 0.65, 6.31 2.20 0.53, 9.04
$46,000 + vs <30,000
1.27 0.41, 3.89 1.08 0.23, 4.98
*Adjusted for all other variables **Wald Chi-Square test p-value
54
Figure 3.2a. Unadjusted Kaplan-Meier curves depicting overall survival estimates for all patients with non-metastatic, intrahepatic cholangiocarcinoma - by treatment strategy (n=881).
55
Figure 3.2b. Unadjusted Kaplan-Meier curves depicting overall survival estimates for patients with locally advanced stages (Stages II-III), intrahepatic cholangiocarcinoma - by treatment strategy (n=414).
56
Figure 3.2c. Unadjusted Kaplan-Meier curves depicting overall survival estimates for patients with early stage (Stages I), intrahepatic cholangiocarcinoma - by treatment strategy (n=465).
57
Table 3.3. Results from Cox regression models examining the effect of neoadjuvant chemotherapy on survival.
Model Description
HR
95% CI
**p-value
Unadjusted with facility as cluster variable
0.92
0.64, 1.31
0.66
Adjusted using 1:1 propensity score matching*
0.85
0.58, 1.25
0.42
Adjusted using 1:4 propensity score matching†
0.78
0.54, 1.11
0.16
Adjusted using propensity score stratification‡
0.82
0.60, 1.13
0.24
Abbreviations: CI - confidence interval, HR - hazard ratio **Wald Chi-Square test p-value *1:1 matching using Mahalanobis distance matching of age, year of diagnosis with facility as cluster variable (n=128)
†1:4 optimal variable matching with facility as cluster variable (n=259) ‡Pooled estimate of stratified matched propensity scores (strata=5) with facility as cluster variable (n=878)
58
Supplemental results
Figure S3.1. Study flow chart for selection criteria of patients with intrahepatic cholangiocarcinoma – NCDB (2006-2014).
ICD-O-3 morphologies not defined as intrahepatic
cholangiocarcinoma n=1,896
▪ no surgical
procedure:
n=2,519
▪ destructive tumor/non-
specific surgical
procedures: n=130
▪ transplant procedures:
n=48
Other patient exclusion criteria: ▪ multiple primary site
tumors: n=2,825 ▪ single primary tumors
seen in more than 1 CoC facility: n=1,520
▪ borderline malignant behavior: n=12
▪ missing a clinical stage: 4,272
▪ absent or unknown surgery-systemic treatment information: n=415
▪ intraoperative systemic therapy n=5
▪ unknown vital status or
follow-up time: 600
Intrahepatic biliary malignancies in NCDB
n=23,273
Intrahepatic cholangiocarcinoma (ICC)
n=21,377
Exclusion of
8,150 stage
IV patients
ICC patients seen in 1 CoC facility for diagnosis and/or treatment and
included in analysis n=3,578
ICC patients seen in 1 CoC facility for surgical treatment and included for propensity score matching analysis
n=1,059
ICC patients seen in 1 CoC facility
who underwent curative surgery:
n=881
59
Table S3.1. Baseline characteristic of patients in the neoadjuvant and no-neoadjuvant groups, from the propensity score matching – adequacy of matching expressed in p values and standardized differences.
No neoadjuvant chemotherapy
(n= 74)
Neoadjuvant
chemotherapy (n=74)
*p-value
Standardized difference
Age
1.0
Age: <70 yrs 59(79.7) 59(79.7) 0.00000
Age: ≥70 yrs 15(20.3) 15(20.3) 0.00000
Hospital type
0.95
Community program 10 (13.5) 12(16.2) -0.06924 Academic program 61(82.4) 57(77.0) 0.12193
Integrated network program 3(4.0) 5(6.7) -0.09753
Stage
0.93
Stage 0/I 25(33.7) 25(33.7) 0.00000 Stage II 30(40.5) 30(40.5) 0.00000 Stage III 19(25.6) 19(25.6) 0.00000
Charlson comorbidity index
0.88
0 56(75.6) 52(70.2) -0.03507 1 14(18.9) 17(22.9) 0.03274 2 4(5.4) 3(4.0) -0.01152
≥3 0(0.0) 2(2.7) 0.03010
Year of diagnosis
1.0
2006 2(2.7) 3(4.0) -0.07336 2007 1(1.3) 0(0.0) -- 2008 4(5.4) 4(5.4) 0.00000 2009 9(12.1) 9(12.1) 0.00000 2010 10(13.5) 9(12.1) 0.03849 2011 9(12.1) 10(13.5) -0.03950 2012 17(22.9) 17(22.9) 0.00000 2013 10(13.5) 11(14.8) -0.03915 2014 12(16.2) 11(14.8) 0.03709
*Z-score test p-value
60
Figure S3.2. Histogram illustrating the distribution of patients in each of the propensity score strata, by treatment group (n=881).
61
Figure S3.3 Utilization of neoadjuvant chemotherapy among all ICC patients, n=21,018
62
CHAPTER 4: CONCLUSION AND RECOMMENDATIONS
In summary, most patients will require multimodal treatment for non-metastatic ICC because of
the late presentation of disease. In this scenario, we have demonstrated that potentially curative
surgery should be an integral part of treatment for best survival outcomes. Specifically,
neoadjuvant chemotherapy offers several advantages over the more prevalent adjuvant
chemotherapy, namely, it promotes better drug compliance and pre-selects patients who have
chemo-responsive tumors. Elderly patients experience a treatment response equivalent to
younger ones. However, against the backdrop of a naturally reduced life expectancy and
perception of increased toxicity to chemotherapy, when compared to younger ones, older
patients are less likely to receive stage-appropriate treatment. The receipt of substandard care
among the elderly is an important and often overlooked driver of poor ICC outcomes.
It is important to note that the present dissertation does not measure or account for quality of life
post treatment, as no surrogates were available in NCDB at the time of analysis. Survival does
not necessarily equate to quality of life. Quality of life is a subjective and multidimensional
concept that is difficult to quantify. Although as much as 5-7% of patients in small studies have
experienced a post-surgical complication which may contribute to substantial disability and
reduced quality of life, there appears to be the perception of improved overall health despite
reported worsening of digestive and liver-related symptoms beyond three months after stage-
appropriate treatment.69,70 There is a need for large prospective study evaluation of quality of life
among ICC patients who have undergone treatment.
In the United States, it is estimated that the 2000 CENSUS population of persons 65 years and
older would double by 2050. 71 As ICC risk increases with age, it is therefore expected that ICC
63
incidence and its prevalence will likely increase in the future. There is a need to address several
current shortcomings in order to meet this challenge.
Participation in clinical trials: Low enrollment of older clinical trial participants is one of the most
consistently cited reasons for substandard care in practice among this population. As a result,
there are few treatment guidelines for the elderly among whom comorbidity and polypharmacy
further complicate chemotherapy use. It has been suggested that incentivizing the relaxation of
stringent trial inclusion criteria may be helpful. 31,72
Chemotherapy and pharmaco- dynamic and kinetic studies: Healthy aging is associated with a
host of system-wide changes including reduced liver volume and renal function as measured
by glomerular filtration rate, bone marrow function and general muscle loss. 73 These changes
potentially affect the dose, dosing interval and choice of chemotherapeutic agents or
supportive care that elder ICC patients may require. 74 There is a continuing need to update the
effects of new chemotherapeutic agents on the aging body as well as the older body’s ability to
absorb, distribute and eliminate these agents. More importantly, these physiologic changes
occur at different rates for different individuals and older age should not necessarily imply
substantial organ impairment. 75
Standardized geriatric assessment: With few guidelines for elderly patients, there is a need to
standardize and validate ICC-specific geriatric assessments that comprehensively assess risk of
adverse effects of therapeutic chemotherapy and complications of surgery, especially in settings
where a multidisciplinary team is not available. While a few geriatric assessments for cancer
patients exist, they are yet to be routinely used as part of patient work-up. 30
64
Integration and coordination of care: Not only will more of a multidisciplinary approach to
managing older cancer patients be required, but primary care providers will also need to
effectively compile and transmit complex health information to and from specialists and patients
alike, in a manner that encourages active participation and decision making. 76
65
REFERENCES
1. Shaib Y, El-Serag HB. The epidemiology of cholangiocarcinoma. Paper presented
at: Seminars in liver disease 2004.
2. Khan SA, Thomas HC, Davidson BR, Taylor-Robinson SD. Cholangiocarcinoma.
The Lancet. 2005;366(9493):1303-1314.
3. Maithel SK, Gamblin TC, Kamel I, Corona‐Villalobos CP, Thomas M, Pawlik
TM. Multidisciplinary approaches to intrahepatic cholangiocarcinoma. Cancer.
2013;119(22):3929-3942.
4. NCI. National Cancer Institute/Surveillance, Epidemiology, End Results: Cancer Stat
Facts: Liver and Intrahepatic Bile Duct Cancer:
https://seer.cancer.gov/statfacts/html/livibd.html , accessed from the Internet on
October 22, 2019
5. Gupta A, Dixon E. Epidemiology and risk factors: intrahepatic cholangiocarcinoma.
Hepatobiliary Surg Nutr. 2017;6(2):101-104. 6. Razumilava N, Gores GJ. Cholangiocarcinoma. The Lancet. 2014;383(9935):2168-2179.
7. Goodman ZD. Neoplasms of the liver. Modern pathology: an official journal of the
United States and Canadian Academy of Pathology, Inc. 2007;20 Suppl 1:S49-
60.
8. Patel AH, Harnois DM, Klee GG, LaRusso NF, Gores GJ. The utility of CA 19-9 in the
diagnoses of cholangiocarcinoma in patients without primary sclerosing cholangitis.
The American journal of gastroenterology. 2000;95(1):204-207.
9. Bridgewater J, Galle PR, Khan SA, et al. Guidelines for the diagnosis and management
66
of intrahepatic cholangiocarcinoma. Journal of hepatology. 2014;60(6):1268-1289.
10. Mavros MN, Economopoulos KP, Alexiou VG, Pawlik TM. Treatment and prognosis for
patients with intrahepatic cholangiocarcinoma: systematic review and meta-analysis.
JAMA surgery. 2014;149(6):565-574.
11. Ghidini M, Tomasello G, Botticelli A, et al. Adjuvant chemotherapy for resected biliary
tract cancers: a systematic review and meta-analysis. HPB. 2017;19(9):741-748.
12. Simo KA, Halpin LE, McBrier NM, et al. Multimodality treatment of intrahepatic
cholangiocarcinoma: a review. Journal of surgical oncology. 2016;113(1):62-83.
13. Morise Z, Sugioka A, Tokoro T, et al. Surgery and chemotherapy for intrahepatic
cholangiocarcinoma. World J Hepatol. 2010;2(2):58-64.
14. Tran TB, Bal CK, Schaberg K, Longacre TA, Chatrath BS, Poultsides GA. Locally
Advanced Intrahepatic Cholangiocarcinoma: Complete Pathologic Response to
Neoadjuvant Chemotherapy Followed by Left Hepatic Trisectionectomy and Caudate
Lobectomy. Digestive Diseases and Sciences. 2015;60(11):3226-3229.
15. Morise Z, Sugioka A, Hoshimoto S, et al. Patient with advanced intrahepatic
cholangiocarcinoma with long‐term survival successfully treated with a combination of
surgery and chemotherapy. Journal of Hepato‐Biliary‐Pancreatic Surgery.
2008;15(5):545-548.
16. Howell M, Valle JW. The role of adjuvant chemotherapy and radiotherapy for
cholangiocarcinoma. Best Practice & Research Clinical Gastroenterology.
2015;29(2):333-343.
17. Vitale A, Spolverato G, Bagante F, et al. A multi‐institutional analysis of elderly patients
undergoing a liver resection for intrahepatic cholangiocarcinoma. Journal of surgical
oncology. 2016;113(4):420-426.
18. Yeh C-N, Jan Y-Y, Chen M-F. Hepatectomy for Peripheral Cholangiocarcinoma in
67
Elderly Patients. Annals of Surgical Oncology. 2006;13(12):1553-1559.
19. De La Fuente SG, Bennett KM, Scarborough JE. Functional status determines
postoperative outcomes in elderly patients undergoing hepatic resections. Journal
of surgical oncology. 2013;107(8):865-870.
20. Grendar J, Grendarova P, Sinha R, Dixon E. Neoadjuvant therapy for downstaging of
locally advanced hilar cholangiocarcinoma: a systematic review. HPB.
2014;16(4):297- 303.
21. NCDB. About the National Cancer Database: https://www.facs.org/quality-
programs/cancer/ncdb/about, accessed from the Internet on July 30, 2019.
22. Massarweh NN, El-Serag HB. Epidemiology of hepatocellular carcinoma and
intrahepatic cholangiocarcinoma. Cancer Control,
2017;24(3):1073274817729245.
23. Everhart JE, Ruhl CE. Burden of digestive diseases in the United States Part III:
Liver, biliary tract, and pancreas. Gastroenterology 2009;136(4):1134-1144.
24. McLean L, Patel T. Racial and ethnic variations in the epidemiology of intrahepatic
cholangiocarcinoma in the United States. Liver International, 2006;26(9):1047-
1053.
25. Khan SA, Thomas HC, Davidson BR, Taylor-Robinson SD. Cholangiocarcinoma.
The Lancet 2005;366(9493):1303-1314.
26. Tyson GL, El‐Serag HB. Risk factors for cholangiocarcinoma.
Hepatology, 2011;54(1):173-184.
27. Simo KA, Halpin LE, McBrier NM, et al. Multimodality treatment of
intrahepatic cholangiocarcinoma: a review. Journal of surgical oncology
2016;113(1):62-83.
28. Bellury LM, Ellington L, Beck SL, Stein K, Pett M, Clark J. Elderly cancer survivorship:
69
an integrative review and conceptual framework. European Journal of Oncology
Nursing. 2011;15(3):233-242.
29. Berger NA, Savvides P, Koroukian SM, et al. Cancer in the elderly. Trans Am Clin
Climatol Assoc. 2006;117:147-156.
30. Mohile SG, Dale W, Somerfield MR, et al. Practical assessment and management of
vulnerabilities in older patients receiving chemotherapy: American Society of Clinical
Oncology Guideline for Geriatric Oncology. Journal of clinical oncology: official journal
of the American Society of Clinical Oncology. 2018;36(22):2326.
31. Sarfati D, Koczwara B, Jackson C. The impact of comorbidity on cancer and
its treatment. CA: a cancer journal for clinicians. 2016;66(4):337-350.
32. DePeralta DK, Frakes J, Mahipal A, et al. Multidisciplinary Management of Liver,
Pancreatic, and Gastric Malignancies in Older Adults. Geriatric Oncology 2019:1-
28.
33. Hung AK, Guy J. Hepatocellular carcinoma in the elderly: Meta-analysis and
systematic literature review. World J Gastroenterol. 2015;21(42):12197-12210.
34. NCDB. Facility Oncology Registry Data Standards: https://www.facs.org/-
/media/files/quality-programs/cancer/ncdb/fords-2016.ashx?la=en, accessed from the
internet on: July 30th, 2019.
35. Meng Z-W, Pan W, Hong H-J, Chen J-Z, Chen Y-L. Modified staging classification
for intrahepatic cholangiocarcinoma based on the sixth and seventh editions of the
AJCC/UICC TNM staging systems. Medicine (Baltimore). 2017;96(34):e7891-e7891.
36. Chang GJ, Skibber JM, Feig BW, Rodriguez-Bigas M. Are we undertreating
rectal cancer in the elderly? An epidemiologic study. Ann Surg. 2007;246(2):215-
221.
70
37. Tranvåg EJ, Norheim OF, Ottersen TJBc. Clinical decision making in cancer care:
a review of current and future roles of patient age. 2018;18(1):546.
38. Liu P-H, Hsu C-Y, Lee Y-H, et al. Uncompromised treatment efficacy in elderly
patients with hepatocellular carcinoma: a propensity score analysis. Medicine
(Baltimore). 2014;93(28):e264-e264.
39. Sarfati D, Gurney J, Stanley J, Koea J. A retrospective cohort study of patients with
stomach and liver cancers: the impact of comorbidity and ethnicity on cancer care
and outcomes. J BMC cancer. 2014;14(1):821.
40. Bridges J, Hughes J, Farrington N, Richardson A. Cancer treatment decision-
making processes for older patients with complex needs: a qualitative study. BMJ
open. 2015;5(12):e009674.
41. Tariman JD, Berry DL, Cochrane B, Doorenbos A, Schepp KG. Physician, patient, and
contextual factors affecting treatment decisions in older adults with cancer and models
of decision making: a literature review. Oncol Nurs Forum. 2012;39(1):E70-E83.
42. Yadav S, Xie H, Bin-Riaz I, et al. Neoadjuvant vs. adjuvant chemotherapy for
cholangiocarcinoma: A propensity score matched analysis. European Journal of
Surgical Oncology. 2019.
43. Cardinale V, Semeraro R, Torrice A, et al. Intra-hepatic and extra-hepatic
cholangiocarcinoma: New insight into epidemiology and risk factors. World
J Gastrointest Oncol. 2010;2(11):407-416.
44. Mavros MN, Economopoulos KP, Alexiou VG, Pawlik TM. Treatment and Prognosis
for Patients With Intrahepatic Cholangiocarcinoma: Systematic Review and Meta-
analysisTreatment of Intrahepatic CholangiocarcinomaTreatment of Intrahepatic
Cholangiocarcinoma. JAMA Surgery. 2014;149(6):565-574
71
45. Buettner S, van Vugt JLA, Ijzermans JN, Groot Koerkamp B. Intrahepatic
cholangiocarcinoma: current perspectives. Onco Targets Ther. 2017;10:1131-
1142.
46. Maithel SK, Gamblin TC, Kamel I, Corona-Villalobos CP, Thomas M, Pawlik
TM. Multidisciplinary approaches to intrahepatic cholangiocarcinoma. Cancer.
2013;119(22):3929-3942.
47. Wang K, Zhang H, Xia Y, Liu J, Shen F. Surgical options for
intrahepatic cholangiocarcinoma. Hepatobiliary Surg Nutr.
2017;6(2):79-90.
48. Bagante F, Spolverato G, Weiss M, et al. Defining Long-Term Survivors Following
Resection of Intrahepatic Cholangiocarcinoma. Journal of Gastrointestinal Surgery.
2017;21(11):1888-1897.
49. Ramírez-Merino N, Aix SP, Cortés-Funes H. Chemotherapy for cholangiocarcinoma:
An update. World J Gastrointest Oncol. 2013;5(7):171-176.
50. Bagante F, Gani F, Beal EW, et al. Prognosis and adherence with the National
Comprehensive Cancer Network Guidelines of patients with biliary tract cancers:
an analysis of the National Cancer Database. Journal of Gastrointestinal Surgery.
2019;23(3):518-528.
51. Cantrell CK, White J. Successful Management of an "Unresectable" Intrahepatic
Cholangiocarcinoma with Neoadjuvant Systemic Therapy, Chemoembolization, and
Extended Hepatectomy with Portal Vein Reconstruction. Cureus.
2018;10(5):e2696- e2696.
72
52. Kato A, Shimizu H, Ohtsuka M, et al. Downsizing Chemotherapy for Initially
Unresectable Locally Advanced Biliary Tract Cancer Patients Treated with
Gemcitabine Plus Cisplatin Combination Therapy Followed by Radical Surgery. Annals
of Surgical Oncology. 2015;22(3):1093-1099.
53. Le Roy B, Gelli M, Pittau G, et al. Neoadjuvant chemotherapy for initially
unresectable intrahepatic cholangiocarcinoma. BJS. 2018;105(7):839-847.
54. Kato A, Shimizu H, Ohtsuka M, et al. Surgical resection after downsizing
chemotherapy for initially unresectable locally advanced biliary tract cancer: a
retrospective single- center study. Annals of surgical oncology. 2013;20(1):318-324.
55. Lunsford KE, Javle M, Heyne K, et al. Liver transplantation for locally advanced
intrahepatic cholangiocarcinoma treated with neoadjuvant therapy: a prospective
case- series. The lancet Gastroenterology & hepatology. 2018;3(5):337-348.
56. Cai Y, Cheng N, Ye H, Li F, Song P, Tang W. The current management of
cholangiocarcinoma: A comparison of current guidelines. Bioscience trends. 2016.
57. Shroff RT, Kennedy EB, Bachini M, et al. Adjuvant therapy for resected biliary tract
cancer: ASCO clinical practice guideline. Journal of Clinical Oncology. 2019:
JCO1802178.
58. Le Roy B, Gelli M, Pittau G, et al. Neoadjuvant chemotherapy for initially unresectable
intrahepatic cholangiocarcinoma. British Journal of Surgery. 2018;105(7):839-847.
59. NCDB. Facility Oncology Registry Data Standards: https://www.facs.org/-
/media/files/quality-programs/cancer/ncdb/fords-2016.ashx?la=en, accessed from the
internet on July 30, 2019.
73
60. Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Data Base: a
powerful initiative to improve cancer care in the United States. Ann Surg Oncol.
2008;15(3):683-690.
61. Mokdad AA, Yopp AC, Polanco PM, et al. Adjuvant Chemotherapy vs Postoperative
Observation Following Preoperative Chemoradiotherapy and Resection in
Gastroesophageal Cancer: A Propensity Score–Matched Analysis. JAMA oncology.
2018;4(1):31-38.
62. McMasters KM, Tuttle TM, Leach SD, et al. Neoadjuvant chemoradiation for
extrahepatic cholangiocarcinoma. The American journal of surgery. 1997;174(6):605-
609.
63. Nelson JW, Ghafoori AP, Willett CG, et al. Concurrent chemoradiotherapy in resected
extrahepatic cholangiocarcinoma. International Journal of Radiation Oncology* Biology*
Physics. 2009;73(1):148-153.
64. Rassen JA, Shelat AA, Myers J, Glynn RJ, Rothman KJ, Schneeweiss S. One‐to‐many
propensity score matching in cohort studies. Pharmacoepidemiology and drug safety.
2012;21:69-80.
65. Yedibela S, Demir R, Zhang W, Meyer T, Hohenberger W, Schönleben F. Surgical
treatment of mass-forming intrahepatic cholangiocarcinoma: an 11-year Western
single- center experience in 107 patients. Annals of Surgical Oncology.
2009;16(2):404.
66. Lamarca A, Ross P, Wasan HS, et al. Advanced intrahepatic cholangiocarcinoma:
post- hoc analysis of the ABC-01,-02 and-03 clinical trials. JNCI: Journal of the
National Cancer Institute. 2019.
74
67. Bilimoria KY, Bentrem DJ, Lillemoe KD, Talamonti MS, Ko CY, on behalf of the
American College of Surgeons' Pancreatic Cancer Quality Indicator Development
Expert
P. Assessment of Pancreatic Cancer Care in the United States Based on
Formally Developed Quality Indicators. JNCI: Journal of the National Cancer
Institute. 2009;101(12):848-859.
68. Bilimoria KY, Bentrem DJ, Stewart AK, Winchester DP, Ko CYJJoCO. Comparison
of Commission on Cancer–approved and–nonapproved hospitals in the United
States: implications for studies that use the National Cancer Data Base.
2009;27(25):4177- 4181.
69. Ma KW, Cheung TT, She WH, et al. Major postoperative complications
compromise oncological outcomes of patients with intrahepatic
cholangiocarcinoma after curative resection – A 13-year cohort in a tertiary center.
Asian Journal of Surgery. 2019;42(1):164-171.
70. Raoof M, Lewis A, Goldstein L, et al. Timing and severity of post-discharge
morbidity after hepatectomy. HPB. 2017;19(4):371-377.
71. Cinar D, Tas D. Cancer in the elderly. North Clin Istanb. 2015;2(1):73-80. 72. Lewis JH, Kilgore ML, Goldman DP, et al. Participation of patients 65 years of age
or older in cancer clinical trials. Journal of clinical oncology. 2003;21(7):1383-1389.
73. Mangoni AA, Jackson SHD. Age-related changes in pharmacokinetics and
pharmacodynamics: basic principles and practical applications. British Journal of
Clinical Pharmacology. 2004;57(1):6-14.
75
74. Lichtman SM, Wildiers H, Launay-Vacher V, Steer C, Chatelut E, Aapro M.
International Society of Geriatric Oncology (SIOG) recommendations for the
adjustment of dosing in elderly cancer patients with renal insufficiency. European
Journal of Cancer. 2007;43(1):14-34.
75. Crombag BSM-R, Joerger M, Thürlimann B, Schellens HMJ, Beijnen HJ, Huitema
DRA. Pharmacokinetics of Selected Anticancer Drugs in Elderly Cancer Patients:
Focus on Breast Cancer. Cancers. 2016;8(1).
76. Walsh J, Harrison JD, Young JM, Butow PN, Solomon MJ, Masya L. What are the
current barriers to effective cancer care coordination? A qualitative study. BMC
Health Services Research. 2010;10(1):132.