|Home | About | Journals | Submit | Contact Us | Français|
Biliary tract cancers (BTC), which encompass intra- and extrahepatic cholangiocarcinomas and gallbladder carcinomas, are a genetically diverse collection of cancers. Evidence suggests distinct models of molecular and pathologic progression, and a growing body of genetics data points to a heterogeneous collection of underlying mutations in key oncogenes and tumor suppressor genes. Although tumor genetics have been used to tailor individual treatment regimens and guide clinical decision making in other cancers, these principles have not been applied in BTC. Recent clinical trials with targeted therapies seem promising, although the relationships between subsets of patients with positive responses to therapy and tumor genetics remain unexplored. Here, we summarize the molecular pathogenesis and genetics of BTCs and animal modeling and relate these to recent and ongoing clinical trials with targeted agents.
Biliary tract cancers (BTCs) include a spectrum of invasive adenocarcinomas encompassing both cholangiocarcinoma (CC), which refers to cancers arising in the intrahepatic, perihilar, or distal biliary tree, and gallbladder carcinoma (GBC). BTCs are characterized by early lymph node and distant metastases. Only 10% of patients present with early-stage disease and are considered candidates for surgical resection, which offers the only chance for cure. The prognosis is poor for the majority of patients with locally advanced or metastatic BTC, with a median survival of less than 1 year. BTC afflicts approximately 12,000 people in the United States annually, with a recently appreciated increasing incidence of intrahepatic CC (IHCC).1 The few clear risk factors share a common theme of chronic tissue damage and inflammation such as gallstones, chronic hepatitis, primary sclerosing cholangitis, and liver fluke infection.2,3 Although this collective of diseases (IHCC, extrahepatic CC [EHCC], and GBCs) shares a similar anatomic origin in the biliary system, there are important differences in disease behavior, molecular profiles, and sensitivity to therapy. In general, GBC carries a worse prognosis than CC; however, GBC tends to exhibit a greater response rate (RR) to chemotherapy.4 After surgery, GBC is more likely recur distantly than EHCC, for which local recurrence is an appreciable risk.5 These differences in clinical behavior are echoed in the molecular features of the disease, which, although overlapping, are clearly distinct. As the landscape of solid tumor genetics, animal modeling, and clinical research evolves, we see great opportunity to make headway in our basic understanding of BTC biology and in treatment paradigms. Here, we review key molecular features of BTC. We then focus on emerging therapies and discuss the opportunities and challenges presented by the growing ranks of targeted therapies. Other reviews have recently summarized clinical aspects of BTCs including epidemiology, staging, surgical management, and chemotherapeutic treatment of advanced disease to which we refer the reader.1,6–12
Models of stepwise transformation from a normal epithelial compartment into carcinoma with corresponding accumulation of molecular changes is delineated in colon, brain, breast, and pancreatic cancers.13,14 Similarly, such models are in development in BTC, where precancerous changes in the gallbladder and biliary tract are beginning to be linked with molecular changes in established oncogenes and tumor suppressor genes. There are important differences in the pathogenesis of GBC and CC; thus, each entity is discussed separately.
There are two appreciated precursors to CC—biliary intraepithelial carcinoma neoplasia (BilIN) and intraductal papillary neoplasm of the bile duct; the incidence of the former is significantly greater than the latter (Figs 1A to to1C).1C). BilIN, which only recently acquired an internationally agreed on classification system, is graded based on the extent of cellular atypia ranging from grades 1 to 3.15 This mirrors the spectrum of precursor lesions of pancreatic carcinoma, with increasing cellular atypia and architectural aberrancy as the stage advances. Biliary intraductal papillary neoplasm is a macroscopic lesion similar to its pancreatic counterpart intraductal papillary-mucinous neoplasm (Fig 1C).16–19 Molecularly, these precursors of carcinoma remain poorly characterized, although immunohistochemical analysis suggests that these harbor mutations in p53 and loss of SMAD4 (SMAD family member 4).20,21 The striking similarity of the precursor lesions between the pancreatic ductal system and the biliary tract raises questions about a shared molecular pathogenesis. For instance, are KRAS mutations, which are detected in advanced cancers (discussed later) and are also an early genetic event in the formation of GBC and pancreatic cancer, detectable in the early stages of BilIN? A detailed molecular profile of is not yet fully established.
Gallbladder adenomas and dysplasia are two distinct pathologic lesions of the epithelial mucosa (Figs 1D and and1E).1E). Both lesions are found in association with carcinomas, although the frequency of dysplastic changes in association with carcinoma seems to be far greater. In pathologic studies most invasive carcinomas are associated with dysplasia and carcinoma in situ, whereas adenomas are rarely found in the context of such dysplastic changes.22–25 Thus, based on present evidence, a metaplasia-dysplasia → carcinoma in situ → carcinoma sequence seems to be the more prevalent route over an adenoma → carcinoma sequence.26 Genetic evidence, although limited, also suggests that adenomas, which tend to harbor mutations in CTNNB1 (catenin β1; more commonly known as β-catenin), and epithelial dysplasias, which have been found to harbor KRAS mutations, may represent two distinct processes (Table 1). KRAS mutations have been identified in dysplastic and hyperplastic lesions associated with anomalous junction of the pancreaticobiliary duct (AJPBD). This anatomic variant of the biliary tract is strongly associated with the risk for the development of GBC, and these tumors consistently demonstrate KRAS mutations.27,34 Conversely, CTNNB1 mutations are found in a high number of adenomas but are rare in GBCs and never found in dysplastic lesions.28–30 The reciprocal relationship between CTNNB1 and KRAS mutations among adenomas and dysplasia/carcinomas, the frequent association of dysplasia with carcinoma, and the less common finding of adenoma surrounding cancer all suggest that gallbladder adenomas may represent a distinct biologic process. Furthermore, clinically, it is observed that incidentally identified adenomas rarely harbor cancerous changes and seem to represent an entity with negligible clinical risk. This might suggest a paradigm where CTNNB1-mutated adenomas represent a relatively low malignant potential, whereas KRAS-mutated dysplastic lesions associated with inflammation (such as in cases of AJPBD) may have a greater malignant potential.
BTCs have a spectrum of mutations in established oncogenes and tumor suppressor genes, although the actual frequency of these events has been hard to pinpoint. This is because of heterogeneous sample sets, differing methodologies used to detect mutations, and the high degree of stromal cells, or desmoplasia, which can dilute out cancer-specific nucleic acids and limit analysis. Despite these confounding factors, established mutations and amplifications of known oncogenes have been clearly demonstrated (Table 2). Beyond those genes and associated pathways that harbor mutations in BTC, a number of other molecules that have demonstrated importance in the disease and represent potential therapeutic targets are also discussed.
KRAS is a member of the RAS/RAC family that functions to propagate growth signals via downstream effectors such as RAF and PI3K.46 Activation of KRAS is found in subsets of both GBC and CCs with reported rates that range from 3% to 100% among patients with GBC in the setting of AJPBD.27,35,36,38,47 The frequency of KRAS mutation is consistently between 40% and 50% among IHCC (Table 2). BRAF (v-raf murine sarcoma viral oncogene homolog B1), which sits at the upstream of the MEK/MAPK signaling pathways and is a key effecter of the oncogenic activity of KRAS, was found to be mutated in approximately 20% of patients in two European BTC collections including both GBCs and IHCCs.35,40 Importantly, these mutations were mutually exclusive of KRAS mutations, which occurred in 25% and 45% of GBCs and IHCCs, respectively, suggesting that activation of the RAF/MAPK signaling pathway is a key event within the majority of cancers.35 GBCs and CCs from North America and Chile demonstrated differing results. Despite microdissection of specimens and the use of three methods to detect mutations, no BRAF mutations were found.41 Whether these discrepant results represent a true regional difference in the genetics of BTC in Europe versus the Americas or sampling bias is not clear. Mutational analysis of BRAF among additional larger cohorts or in a prospective manner with newer, more sensitive technologies could resolve this question.
Both the EGFR (epidermal growth factor receptor) and HER2/NEU (v-erb-b2 erythroblastic leukemia viral oncogene homolog 2, neuro/glioblastoma-derived oncogene homolog [avian]) receptor tyrosine kinases have been implicated in BTC pathogenesis. EGFR mutations are found in a subset (13.6% to 15%) of BTCs; however, the results of the mutation analyses from the two studies were quite different.42,43 The first study identified previously described mutations in the tyrosine kinase domains that have been observed to confer sensitivity to small-molecule inhibitors of EGFR in lung cancer.43 The subsequent report identifies mutations not previously described; thus presently, we do not know the biologic significance of these mutations or to what extent they may confer sensitivity to currently available small-molecule inhibitors.42 EGFR amplifications are also seen in approximately 6% of all BTCs, further prompting consideration of inhibitors of this pathway.44 ERBB2/HER2/NEU overexpression and gene amplification were also evaluated in this study,44 and excellent concordance (79%) between expression measured by immunohistochemistry and gene amplification was observed. GBC had the highest incidence of overexpression at approximately 15% compared with EHCC at approximately 5% and IHCC at 0%. The relevance of this gene in GBC pathogenesis is further demonstrated by transgenic mouse models (described in detail later) created through expression of Erbb2 in the gallbladder epithelium.48
PIK3CA (phosphoinositide 3-kinase, catalytic, α-polypeptide) hotspot mutations, which are commonly found in colon, breast, gastric, and brain cancers, are rarely found in BTC.45 Despite a low prevalence of activating mutations, immunohistochemical evaluation of downstream PIK3CA targets eIF4-E and 4E-BP1, looking for overexpression and evidence of phosphorylation, respectively, suggests that additional mechanisms may be at play in positively regulating this pathway. Mutations in PTEN (phosphatase and tensin homolog) have not been reported. Expression profiling of BTC compared with normal biliary epithelium has identified AKT/mTOR signaling components as being upregulated, including the potential drug target insulin-like growth factor 1 receptor (IGF1-R).49 Expression of IGF1-R and its ligands is seen in the majority of GBCs and metastases.50,51 Treatment of BTC cell lines with a small-molecule inhibitor of the IGF1-R suggested the efficacy of targeting this pathway.52
The spectrum of tumor suppressor gene mutations seen in BTC largely mirrors that of other GI malignancies (Table 3) and includes loss of function of CDKN2A (cyclin-dependent kinase inhibitor 2A; commonly known as P16INK4A),58 TP53 (tumor protein p53),59 and SMAD4.60 CDKN2A, which is located at the 9p21 locus, encodes two proteins through splicing of an alternative first exon and is frequently silenced in BTC through hypermethylation, homozygous deletion, and inactivating point mutation.38,53,61 Evidence of TP53 inactivation through either overexpression or mutation is identified in a similarly high frequency of BTCs, irrespective of site of origin.36,37,54 SMAD4 mutations, which function to abrogate the tumor suppressor activities of transforming growth factor β (TGF-β), have been identified in a proportion of BTCs, including both GBC and CC.55,56 Concurrent mutations of KRAS and SMAD4 were identified in an adenocarcinoma arising from a choledochal cyst, suggesting cooperation of two events in cancer progression,34 and engineered mutants of both p53 and Smad4 have been used in concert to model IHCC in mice.62,63 These studies are discussed in detail later in this article. Additional tumor suppressor genes have been implicated in BTC pathogenesis, including STK11 (serine/threonine kinase 11; also known as LKB1), in which case one homozygous deletion was identified among 16 tumors of the common bile duct.57 STK11, a key regulator of metabolism and cellular polarity, is frequently mutated in non–small-cell lung cancer (NSCLC) and less frequently mutated in pancreatic cancers.64,65
Beyond mutations in established oncogenes and tumor suppressor genes, genomic studies using array comparative genomic hybridization demonstrate widespread genomic instability of BTC and many additional alterations. These include recurrent loss of the 9p21 locus encoding the P16INK4A/P19ARF tumor suppressor gene and a number of recurrent amplifications including 5p, 7p, and 19q–, all loci amplified in subsets of other cancers66–68; 7p and 19q harbor the oncogenes EGFR and AKT, respectively. MYCN (v-myc myelocytomatosis viral related oncogene) is rarely amplified.69 Recently, mutations in KEAP1 (kelch-like ECH-associated protein 1) have been identified. KEAP1 suppresses NFE2L2 (nuclear factor [erythroid-derived 2]-like 2; also known as NRF2) activity, which transcriptionally controls expression of many genes involved in attenuating oxidative damage.70 The expression of additional proteins implicated in the pathogenesis of cancer is similarly characterized in BTC and is reviewed elsewhere.9,71
Beyond the vasculature supporting cancer growth, the microenvironment, which includes activated fibroblasts, macrophages, and T regulatory cells, has been shown to both influence tumor growth and impact the delivery of therapeutics. Hepatic stellate cells (resident myofibroblasts within the hepatic parenchyma) are associated with a poor prognosis in IHCC and enhance growth of BTC cell lines in vitro.72,73 T regulatory cells and mast cells, which modulate immune response to cancer growth and may be of prognostic significance in colon cancer, are found in IHCC.74,75 The TGF-β pathway, which may regulate both immune response to cancer and fibrosis, has been shown to be upregulated in IHCC. This pathway may serve to both drive stromal proliferation and dampen immune response.76 TGF-β may also modulate angiogenesis through the transcriptional activation of vascular endothelial growth factor (VEGF).77 The role of tumor stroma and the additional factors involved are reviewed in further depth elsewhere.78 Several studies have examined the expression of VEGFs and other angiogenic molecules and their correlation with clinical outcomes. VEGF is expressed by most tumors,79,80 and high VEGF expression seems to correlate with hematogenous metastasis in IHCC,79 nodal metastases, and poor survival.81,82
Animal models of cancer have proved invaluable in furthering our understanding of cancer genetics and biologic behavior and are increasingly being used in drug development. There are a number of spontaneous models of BTC in mammalian systems, including mice, rats, and guinea pigs (reviewed in Kiguchi et al48), that rely on chemical mutagens. Among these, a furan-induced rat CC model has yielded insight into the potential roles of c-met and ErbB2 oncogenes.83–87 Immortalized rat cholangiocyte lines have proved to be valuable to understanding ErbB2 signaling in CC. Furthermore, these cells may be transplanted into syngeneic animals, offering a genetically malleable in vivo model of advanced disease.88,89
Engineered mouse models have proved valuable in the study of cancer genetics, pathogenesis, and biology. Recently, such models have become fundamental to drug development efforts.90–92 Three models of BTC—one of GBC and two of CC—are based on key genetic events observed in human tumors. As discussed earlier, amplification of ERBB2 is seen in approximately 15% of GBCs. A transgenic mutant with constitutive expression of ErbB2 in the gallbladder and biliary tract epithelium develops GBC and CC with a 100% penetrance.48 Tumors from this model exhibit elevated MAPK signaling and COX2 expression and have been used to preclinically evaluate the mTOR inhibitor rapamycin.93 Mutations in key tumor suppressor genes have also enabled modeling of BTC. Liver fibrosis and inflammation have been associated with IHCC in humans,94,95 and carbon tetrachloride is used as a hepatotoxin in several animal models of both cirrhosis and hepatocellular carcinoma.96 With regular carbon tetrachloride exposure causing liver injury and fibrosis, p53-mutant mice develop IHCC.63 Mirroring human tumors, expression of c-met, cox2, and Erbb2 and loss of E-cadherin are elevated. Because chronic liver injury is significant in the pathogenesis of these tumors, this model may offer some advantage over approaches relying solely on targeted mutations. Homozygous somatic mutation of both Smad4 and Pten in the liver leads to IHCC.62 Although deletion of Smad4 alone does not lead to any hepatic phenotype, mutation of Pten does cause biliary hyperplasia and IHCC, albeit with a significantly longer latency. Although Pten itself is not mutated in human BTC, deregulation of the pathway may occur through alternative mechanisms such as AKT amplification, PI3K mutations, and so on, as discussed earlier.
These models provide a critical foundation for further efforts focused on understanding other relevant genetic events involved in BTC pathogenesis. The major oncogenic mutation observed in humans, KRAS, has not yet been integrated into modeling efforts, nor have other less common mutations, such as PIK3CA, BRAF, or EGFR. The availability of a spectrum of models mirroring the genetics of the human disease would facilitate both a basic understanding of disease biology and the clinical development of drugs with specific activity given the genetics of each tumor type.
Chemotherapy in locally advanced and metastatic BTC has historically been loosely defined based on reports of activity of single agents and combinations of agents such as gemcitabine, cisplatin, oxaliplatin, and fluorouracil in phase II trials.12 Only recently has an optimal strategy for the management of advanced disease been defined; a randomized phase III trial demonstrated improved overall survival (OS) with gemcitabine and cisplatin compared with gemcitabine alone.97 The trial, which included both CC and GBC, demonstrated that the addition of cisplatin to gemcitabine, compared with gemcitabine alone, afforded a significant progression-free survival (PFS; median, 8.4 v 6.5 months, respectively) and OS benefit (median, 11.7 v 8.3 months, respectively). This study has set the new benchmark against which future studies may be compared.
A number of novel inhibitors of key oncogenic pathways are in development and have been evaluated in BTCs as single agents, as combined targeted agents, and in combination with classic chemotherapy (Tables 4 and and5).5). All of these early-stage trials include patients with locally advanced or metastatic disease, both GBC and CC. Amplifications and rare activating mutations in EGFR are observed in BTC and seem to predict responsiveness to drugs targeting EGFR in other cancers.105,106 Both small-molecule inhibitors of the kinase domain and blocking antibodies have been used in BTC.44,107,108 A randomized phase II study evaluating the impact of cetuximab when added to gemcitabine and oxaliplatin (GEMOX) among patients who did not receive prior chemotherapy suggests a benefit in terms of PFS at 4 months.98 One hundred one randomly assigned patients were stratified according to tumor stage, location, and prior treatments. Thirty-six patients were available for an interim analysis. The 4-month PFS rates were 44% (95% CI, 20% to 70%) and 61% (95% CI, 36% to 83%) among patients receiving GEMOX and GEMOX-cetuximab, respectively. Toxicity was balanced between arms with the exception of rash, which was more common among patients on cetuximab. The difference in PFS rate at 4 months seems promising. The study did not report the status of KRAS mutation for patients receiving cetuximab-based treatment. In a trial of 42 patients with BTC treated with single-agent erlotinib, three patients achieved responses to therapy, indicating efficacy of the drug in at least some patients.99 Many of these patients had already received first-line therapy. Because EGFR mutation was not tested in these patients, we do not know whether responses correlated with EGFR mutation status. Lapatinib, a dual EGFR1 and ERBB2 (HER2) inhibitor, has been tested in 17 patients with BTCs.100 No responses were observed, although five patients had stable disease.
AZD6244, a potent inhibitor of MEK1/2, downstream of RAS and RAF, has been tested in 29 patients including 22 who were evaluable for response at the time of reporting.101 Thirty-nine percent of patients had received prior therapy. Three patients achieved responses (one complete response and two partial responses) for an RR of 14%. Median PFS was 5.4 months, and median OS was 8.2 months. Correlation between clinical responses and p-ERK staining was observed. Genetic analysis of BRAF and KRAS mutation status is ongoing. Among single agents tested in a mixed population of pretreated and treatment-naive patients, this relatively high RR and PFS are encouraging. Sorafenib, which targets BRAF and VEGF receptors, has been evaluated in two different phase II studies. A single-agent trial of sorafenib in 46 patients included patients with an Eastern Cooperative Oncology Group performance status of 2, of whom 35% needed to discontinue treatment as a result of toxicity. One patient had a partial response (2%), and the median PFS was 2.3 months.102 Another phase II study of sorafenib as a single agent in 36 treatment-naive patients reported an RR of 6%, median PFS of 2 months, and median OS of 6 months.103
In a single-arm phase II study, the efficacy and safety of bevacizumab, a humanized monoclonal antibody against VEGF, in combination with GEMOX in patients with advanced BTC was established.104 Of the 35 patients enrolled, 14 (40%) had a partial response, and an additional 10 (29%) had stable disease. The median OS was 12.7 months, and the median PFS was 7.0 months. Treatment was generally well tolerated, and grade 3 and 4 toxicities included neutropenia, elevation of transaminases, peripheral neuropathy, hypertension, anorexia, and thrombocytopenia. Despite the encouraging results, because of the single-arm study design, patient selection bias, and well-known efficacy of GEMOX alone, the relative contribution of bevacizumab remains to be defined. Combining targeted agents that inhibit different pathways critical to cancer growth and survival represents an attractive strategy. An interim report of a phase II study of bevacizumab in combination with erlotinib in 34 patients with CC and GBC observed a promising RR of 20% and a time to progression of more than 7 months.109
The data on human BTC provide a picture of a collection of distinct diseases that could be classified both by the anatomic location of the tumor's origin (ie, gallbladder, extrahepatic or intrahepatic ducts) and underlying genetics. This genetic heterogeneity is mirrored by differing clinical behavior and responses to therapy.4 KRAS mutation seems to have the highest incidence, at approximately 40% to 50% in IHCC and lower in GBC and EHCC (Table 2). ERBB2/HER2 amplifications are found in subsets of GBC. Given the evidence of efficacy of trastuzumab in gastric cancer with such amplifications, the relevance of testing ERBB2/HER2 should be similarly evaluated in BTC patients. Regarding BRAF, the real mutation rate seems impossible to speculate on. However, common activation of the downstream MEK/MAPK pathway and promising single-agent activity of a potent MEK inhibitor should help to prioritize further testing and relate activity to underlying tumor genetics. Rare EGFR mutations raise questions concerning responses to small-molecule inhibitors of EGFR, as are observed among NSCLC patients with these mutations.105 Additional mutations in the PI3K pathway point to that pathway's importance.
Improved animal models not only may help to define the biology and behavior of cancers with specific genetics, but may also provide insights into the optimum strategy for using targeted agents and thus streamline clinical efforts. As discussed earlier, engineered mouse models of lung and pancreatic cancers, among others, have proved useful in testing emerging drugs. Creating models of BTC harboring the relevant genetic mutations (eg, KRAS, PIK3CA) would be a first step. Therefore, the key therapeutic questions (eg, how to best select patients based on tumor genetics and combine new targeted therapies) are similar to those that exist in many other diseases, such as lung, colon, and breast cancer. As these questions are answered in these more common malignancies, we should see progress regarding these issues in BTC.
A newly defined standard of care of gemcitabine in combination with cisplatin in BTC provides a foundation on which to add targeted agents.97 Despite the lack of randomized controlled trials comparing gemcitabine plus cisplatin versus GEMOX versus gemcitabine plus capecitabine, GEMOX and gemcitabine plus capecitabine may represent reasonable alternatives based on phase II studies.110–116 One immediate question prompted by early promising data, as well as success in colon cancer, is to what extent EGFR-targeting antibodies will offer benefit. The early report of an improved PFS with GEMOX-cetuximab compared with GEMOX alone gives this question further weight.98 It will be crucial to retrospectively evaluate KRAS status among patients enrolled onto this randomized effort to see whether any benefit is restricted to a KRAS wild-type subgroup of patients. Additionally, ongoing studies in genetically selected patients will offer insight into any added benefit of EGFR-targeted agents in KRAS wild-type populations. Because methodologies for screening such patients are presently clinically adapted and the numbers of patients with BTC with wild-type KRAS is approximately 50%, a prospective trial evaluating cetuximab or panitumumab in screened patients is certainly feasible. Even if EGFR-targeting agents prove beneficial in KRAS wild-type patients, we will need to have trials available for the many patients with KRAS and/or BRAF mutations. The potential benefits of bevacizumab or other antiangiogenic agents when added to chemotherapy should be further examined in randomized studies. Will MEK inhibitors have a role in KRAS-mutant BTC, possibly in combination with PI3K inhibitors, as was shown to be beneficial in a mouse model of NSCLC?91 Proving the value of enriching for specific populations of BTC for targeted therapy will prompt us to expedite the routine use of genetic screening in these patients. This is already occurring in selected trials (Table 5). Given the smaller number of patients with BTC compared with other common solid tumors, coordination of trials among institutions and cooperative groups, both nationally and internationally, will be the key to progress.
We thank Jeff Clark and Alok Khorana for critical reading of the manuscript.
Supported by a Howard Hughes Medical Institute Early Career Development Award (A.F.H.) and NIHK08 Career Development Award (A.F.H.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Employment or Leadership Position: None Consultant or Advisory Role: Aram F. Hezel, Bayer (C), Andrew X. Zhu, Genentech (C), Bayer (C), sanofi-aventis (C) Stock Ownership: None Honoraria: None Research Funding: None Expert Testimony: None Other Remuneration: None
Conception and design: Aram F. Hezel, Vikram Deshpande, Andrew X. Zhu
Collection and assembly of data: Aram F. Hezel, Vikram Deshpande, Andrew X. Zhu
Data analysis and interpretation: Aram F. Hezel, Vikram Deshpande, Andrew X. Zhu
Manuscript writing: Aram F. Hezel, Vikram Deshpande, Andrew X. Zhu
Final approval of manuscript: Aram F. Hezel, Vikram Deshpande, Andrew X. Zhu