Surgery Versus Surveillance for Well‐Differentiated, Nonfunctional Pancreatic Neuroendocrine Tumors: An 11‐Year Analysis of the National Cancer Database

Abstract

Background

Pancreatic neuroendocrine tumors (panNETs) are a rare group of tumors that make up 2%–3% of pancreatic tumors. Recommended treatment for panNETs generally consists of resection for symptomatic or large asymptomatic tumors; however, optimal management for localized disease is still controversial, with conflicting recommendations in established guidelines. Our study aim is to compare surgical intervention versus active surveillance in nonmetastatic panNETs by size of primary tumor.

Materials and Methods

Using the National Cancer Database, we identified 2,004 patients diagnosed with localized well‐differentiated, nonfunctional panNETs (NF‐panNETs) between 2004 and 2015. Patients’ clinicopathologic characteristics, treatment modalities, and overall survival (OS) were analyzed using frequency statistics, chi‐square, and Kaplan‐Meier curves. The objective of the study is to assess the outcome of surgical resection versus nonoperative management in patients with panNETs with different tumor sizes.

Results

Tumor sizes were divided into three categories: <1 cm, 1–2 cm, and >2 cm. The number of patients with tumor size <1 cm, 1–2 cm, and >2 cm was 220 (11%), 794 (39.6%), and 990 (49.4%), respectively. Overall, 1,781 underwent surgical resection, whereas 223 patients did not. Median follow‐up was 25.9 months. After adjusting for covariates, surgical resection was associated with improved OS in patients with tumor size 1–2 cm (hazard ratio [HR] = 0.37) and >2c m (HR = 0.30) but not <1 cm (HR = 2.81). Independent prognostic factors were age at diagnosis, Charlson‐Deyo comorbidity score, stage, tumor location, and surgical resection. Higher tumor grade was not associated with worse OS.

Conclusion

Our findings suggest that active surveillance is potentially a safe approach for NF‐panNETs <1 cm. Larger tumors likely need active intervention. Intermediate‐grade tumors did not result in worse survival outcome compared with low‐grade tumors. Future studies might consider prospective randomized clinical trials to validate our findings.

Implications for Practice

The present study seeks to address the discrepancy in treatment recommendations in the management of nonfunctional pancreatic neuroendocrine tumors (NF‐panNETs) by evaluating whether surgical resection is associated with improved overall survival in different tumor size groups as well as elucidating independent prognostic factors in patients with NF‐panNETs. Data from the National Cancer Database were reviewed. This study's findings suggest that active surveillance is potentially a safe approach for NF‐panNETs <1 cm. Larger tumors likely need active intervention. Independent prognostic factors include age at diagnosis, Charlson‐Deyo comorbidity score, stage, tumor location, and surgical resection. These findings will help guide medical and surgical oncologists when formulating treatment plans for patients with small NF‐panNETs.

Short abstract

Optimal management for pancreatic neuroendocrine tumors is controversial. This article evaluates whether surgical resection is associated with improved overall survival in different tumor size groups.

Introduction

Pancreatic neuroendocrine tumors (panNETs) are rare neoplasms of endocrine origin that make up about 7% of all neuroendocrine tumors, and only 1%–2% of all pancreatic lesions 1, 2, 3. They are typically classified as functional (10%–50%) and nonfunctional tumors (50%–90%), based on their ability to secrete hormones 3, 4, 5. Despite affecting approximately 1 in 100,000 individuals in the general U.S. population, postmortem studies suggest that rates could be as high as 1%–10% of the total population 3, 6. This discrepancy, coupled with an increase in quantity and quality of diagnostic imaging, raises the concern for overdiagnosis and, subsequently, overuse of surgery for these tumors 6, 7, 8. Indeed, there has been a sevenfold rise in panNET incidence in the U.S. over the past 2 decades 6, 9. Another complicating factor is the fact that panNETs encompass a heterogeneous spectrum of tumors with varying natural history, clinical symptoms, and prognosis 10. The 5‐year survival ranges from 25% up to 100% in some studies 10. This heterogeneity makes it challenging to standardize the surgical management of localized disease. Therefore, many studies have attempted to elucidate prognostic factors of tumor growth and metastasis that would guide treatment recommendations, with inconsistent results 4, 8, 11, 12. These studies suffered from being mainly single‐centered, having small numbers of patients, and having substantial heterogeneity in inclusion criteria with different tumor size cutoffs.

Because of these conflicting findings, treatment by surgical resection versus active surveillance of small nonfunctional panNETs (NF‐panNETs) remains a matter of debate, with different guidelines suggesting varying approaches. The European Neuroendocrine Tumor Society (ENETS) and the North American Neuroendocrine Tumor Society suggest consideration of active surveillance in small NF‐panNETs <2 cm in size 2, 13, 14. On the other hand, the National Comprehensive Cancer Network (NCCN) states that observation can be considered for low‐grade, incidentally discovered NF‐panNETs <1 cm in size 15.

The present study seeks to address the discrepancy in treatment recommendations in the management of NF‐panNETs by evaluating whether surgical resection is associated with improved overall survival (OS) in different tumor size groups as well as elucidating independent prognostic factors in patients with NF‐panNETs.

Materials and Methods

Data Source

The National Cancer Database (NCDB) is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. The NCDB captures information from approximately 1,500 CoC‐accredited hospitals and more than 70% of all newly diagnosed malignancies in the U.S. All data within the NCDB are deidentified of specific patient and hospital factors and are thus in compliance with the Health Insurance Portability and Accountability Act 16.

Study Design

An NCDB Participant Use Data File containing data on patients with pancreatic cancer was obtained. The study protocol was approved by the local institutional review board at the University of Oklahoma. Patients diagnosed with panNETs were identified using the diagnostic histology code associated with the neuroendocrine tumors as per the International Classification of Diseases for Oncology 3rd edition (ICD‐O‐3 code) 17. Nonfunctional tumors were represented by codes 8150 (pancreatic endocrine tumor) and 8246 (neuroendocrine carcinoma) Patients with histology code 8246 were excluded from this study as they denote neuroendocrine carcinomas. Information regarding tumor grade was obtained through the individual institutions reporting to the NCDB program and was not standardized among these participant sites. Using the NCDB data, we identified 5,057 patients diagnosed with biopsy‐proven NF‐panNETs between 2004 and 2015. After excluding patients with grade 3 tumors, metastatic disease, and missing survival data, we analyzed 2,004 patients with nonmetastatic, well‐differentiated NF‐panNETs. Covariates included age, sex, race, Charlson‐Deyo (CD) score, primary tumor site, tumor size, stage, grade, regional lymph node (LN) surgery, chemotherapy, and radiation therapy. The CD score is a weighted score derived from the sum of the scores for each of the comorbid conditions listed in the Charlson Comorbidity Score Mapping Table (source: http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?conceptID=109). Tumor sizes were divided into three categories: <1 cm, 1–2 cm, and >2 cm. OS was defined as the number of months between date of diagnosis and date of death (if known) or last follow‐up (last known alive date). Follow‐up time was calculated from the date of diagnosis to the last date of contact. This study was determined by the institutional review board to be exempt as a result of the deidentified nature of the data. Our primary objective is to compare OS between patients with NF‐panNETs who were treated with primary surgical resection versus nonoperative management, stratified by tumor size. The secondary objective is to evaluate independent prognostic factors in patients with NF‐panNETs.

Statistical Analysis

Patient's clinicopathologic categorical characteristics were reported using frequency statistics (n and %) and compared across treatment modalities (surgery vs. no surgery) using chi‐square analysis. Because age was continuous, median and interquartile range were reported with Wilcoxon rank sum test performed to compare medians. Kaplan‐Meier survival analysis and plots were created for OS by surgical group, stratified by tumor size group. Kaplan‐Meier survival analysis was also done to explore how each covariate was associated with OS. Variables that showed an association with OS on a univariate basis or were potential confounders were included in the Cox proportional hazards model for multivariable survival analysis. A variable was considered a potential confounder if in the descriptive analysis it was not equally represented across surgery versus nonsurgery group. Stratified analysis was also done by tumor size group for both unadjusted and adjusted Cox survival analysis and is also reported. Hazard ratios with p values <.05 were considered to be significant. SAS 9.4 (SAS Institute, Cary, NC) was used to perform all analysis.

Results

Patient and Tumor Characteristics

We identified a total of 2,004 patients who met the inclusion criteria of age 18 or older with nonmetastatic, well‐differentiated, nonfunctional panNETs (Table 1). By the World Health Organization (WHO) neuroendocrine tumor grading system, all patients had grade 1 or 2 tumors. Among those patients, 89.1% had grade 1 tumors. Patients were categorized into a surgical and nonsurgical group. Overall, 1,781 underwent surgical resection of the primary tumor, whereas 223 patients received no surgery. Fifty‐one percent of patients were male. The nonsurgical group was significantly older (median 68 vs. 60 years of age, p < .0001) and had a shorter median time of follow‐up (25.1 vs. 27.8 months, p < .0001). Sex, race, and CD score were not significantly different between the two groups. The vast majority of patients were white (82.1%), with the remaining patients being black (13.4%) and other races (4.5%).

Table 1.

Descriptive statistics for patients with nonfunctional pancreatic neuroendocrine tumors in the National Cancer Database registry between 2004 and 2015

Variable	Overall	No surgery (n = 223)	Surgery (n = 1781)	p value
Age, years				<.0001
Median	61.0	68	60
Interquartile range	51–69	57–76	50–68
Sex				.9179
Male	1,022 (51.0)	113 (50.7)	909 (51.0)
Female	982 (49.0)	110 (49.3)	872 (49.0)
Race				.0907
White	1,627 (82.1)	171 (77.4)	1,456 (82.7)
Black	265 (13.4)	39 (17.6)	226 (12.8)
Other	90 (4.5)	11 (5.0)	79 (4.5)
Charlson‐Deyo score				.1379
0	1428 (71.3)	153 (68.6)	1275 (71.6)
1	434 (21.7)	46 (20.6)	388 (21.8)
2	113 (5.6)	20 (9.0)	93 (5.2)
3	29 (1.5)	4 (1.8)	25 (1.4)
Primary site				<.0001
Head of pancreas	518 (25.9)	83 (37.2)	435 (24.4)
Body of pancreas	377 (18.8)	52 (23.3)	325 (18.2)
Tail of pancreas	778 (38.8)	52 (23.3)	726 (40.8)
Other	331 (16.5)	36 (16.1)	295 (16.6)
Tumor size, cm				.1701
<1	220 (11.0)	31 (13.9)	189 (10.6)
1–2	794 (39.6)	93 (41.7)	701 (39.4)
>2	990 (49.4)	99 (44.4)	891 (50.0)
Stage				<.0001
1	1,494 (74.6)	177 (79.4)	1,317 (73.9)
2	487 (24.3)	30 (13.5)	457 (25.7)
3	23 (1.2)	16 (7.2)	7 (0.4)
Grade				.3933
1	1,786 (89.1)	195 (87.4)	1,591 (89.3)
2	218 (10.9)	28 (12.6)	190 (10.7)
Regional nodes positivity				<.0001
No positive LNs	1,200 (60.0)	5 (2.3)	1,195 (67.2)
Positive LNs	273 (13.7)	8 (3.6)	265 (14.9)
LNs not examined	526 (26.3)	208 (94.1)	318 (17.9)
Chemotherapy				<.0001
None administered	1,887 (94.2)	196 (87.9)	1,691 (95.0)
Administered	51 (2.5)	22 (9.9)	29 (1.6)
Unknown	66 (3.3)	5 (2.2)	61 (3.4)
Radiation therapy				.0018
None administered	1,972 (98.4)	214 (96.0)	1,758 (98.7)
Administered	27 (1.4)	9 (4.0)	18 (1.0)
Unknown	5 (0.3)	0 (0.0)	5 (0.3)

Data are presented as n (%).

Abbreviation: LN, lymph node.

With respect to tumor location, tail of pancreas tumors constituted a higher proportion in the surgical group compared with the nonsurgical group (40.8% vs. 23.3%, p < .0001). The reverse was true for head of pancreas lesions (24.4% vs. 37.2%, p < .0001). Tumor size was equally distributed between the two groups. A higher percentage of stage III tumors did not undergo surgery as opposed to nonoperative management (7.2% vs. 0.4%, p < .0001). Use of chemotherapy and radiation therapy was significantly higher in the surgical group compared with the nonsurgical group (9.9% vs. 1.6% and 4% vs. 1%, respectively).

Overall Survival by Tumor Size Group

Using Kaplan‐Meier survival analysis, the 5‐year unadjusted overall survival (OS) was calculated for surgical and nonsurgical patients, stratified by tumor size group, <1 cm (Fig. 1), 1–2 cm (Fig. 2), and >2 cm (Fig. 3). Among patients with tumor size <1 cm, the 5‐year OS was not significantly different between surgical and nonsurgical group (100% vs. 85.0%, p = .1411). For patients undergoing surgical management, however, a higher 5‐year OS was observed when compared with nonoperative management in patients with tumors >2 cm (89.3% vs. 73.1%, p < .0001) as well as those with tumor size 1–2 cm (91.9% vs. 70.0%, p = .0003; Table 2).

Figure 1.

Kaplan‐Meier curves of overall survival by surgery—tumor <1 cm.

Figure 2.

Kaplan‐Meier curves of overall survival by surgery—tumor 1–2 cm.

Figure 3.

Kaplan‐Meier curves of overall survival by surgery—tumor >2 cm.

Table 2.

Estimated unadjusted 5‐year overall survival percentage (95% confidence interval) by tumor size group

Tumor size, cm	No surgery (n = 223)	Surgery (n = 1,781)	p value
<1	100.0	85.0 (73.3–96.7)	.1411
1–2	70.0 (52.6–87.5)	91.9 (88.0–95.9)	.0003
>2	73.1 (61.9–84.3)	89.3 (85.4–93.2)	<.0001

After adjusting for age, CD comorbidity score, the primary site of tumor within the pancreas, and stage, patients who underwent surgery with tumor sizes 1–2 cm had a 63% lower hazard of death relative to those who did not undergo surgery (p = .0077). Similarly, those who underwent surgery with tumors sizes >2 cm had a 70% lower hazard of death (p < .0001). In contrast, patients with tumor size <1 cm did not have a significant difference in the hazard of death whether they underwent surgical resection or not, after adjustment (p = .4234; Table 3).

Table 3.

Unadjusted and adjusted HRs of death for surgery by tumor size—HR (95% confidence interval) for surgery versus nonsurgery

Tumor size	<1 cm	p value	1–2 cm	p value	>2 cm	p value
Unadjusted HR	2.79 (0.35–22.20)	.3320	0.32 (0.16–0.62)	.0007	0.29 (0.17–0.48)	<.0001
Adjusted HRa	2.81 (0.22–35.13)	.4234	0.37 (0.18–0.77)	.0077	0.30 (0.17–0.53)	<.0001

^aAdjusted for age, Charlson‐Deyo score, tumor location, stage, grade, chemotherapy, and radiation.

Abbreviation: HR, hazard ratio.

Prognostic Factors

In order to assess what variables might be individually associated with OS, we carried out a multivariate Cox proportional hazards analysis (Table 4). Variables that were independently associated with OS were age at diagnosis, CD comorbidity score, tumor location, stage, and surgical resection. Patients who underwent surgical resection had a better prognosis than patients who did not after adjusting for other variables, with those receiving surgery having a 56% reduction in the hazard of death relative to those who did not (p = .0054). Higher tumor grade (grade 2 vs. 1) was not an independent prognostic factor (p = .4362). Race, tumor size, chemotherapy, radiation, and regional LN positivity were also not associated with OS.

Table 4.

Cox proportional hazards model—multivariable analysis

Characteristic	Hazard ratio	95% CI	p value
Age, years
<60	Ref	Ref
≥60	3.71	2.41–5.72	<.0001
Sex
Male	Ref	Ref
Female	0.76	0.54–1.06	.108
Race
White	Ref	Ref
Black	0.93	0.55–1.57	.7726
Other	1.00	0.40–2.46	.9930
Charlson‐Deyo score
0	Ref	Ref
1	1.59	1.07–2.36	.0218
2	1.91	1.07–3.42	.0280
3	4.16	1.77–9.78	.0011
Primary site
Head of pancreas	Ref	Ref
Body of pancreas	0.47	0.27–0.82	.0080
Tail of pancreas	0.65	0.43–1.00	.0496
Other	0.85	0.53–1.37	.5054
Tumor size, cm
<1	Ref	Ref
1–2	0.64	0.36–1.15	.1347
>2	0.78	0.44–1.40	.4038
Stage
1	Ref	Ref
2	1.67	1.01–2.79	.0478
3	0.69	0.19–2.50	.5742
Grade
1	Ref	Ref
2	1.22	0.74–2.00	.4362
Regional nodes status
Negative	Ref	Ref
Positive	1.02	0.56–1.84	.9541
Not examined	1.17	0.71–1.95	.5400
Surgical resection
Yes	Ref
No	0.44	0.25–0.79	.0054
Chemotherapy
No	Ref
Yes	1.49	0.65–3.45	.3455
Unknown	0.98	0.54–1.79	.9802
Radiation therapy
No	Ref
Yes	0.60	0.14–2.63	.5014

Abbreviations: CI, confidence interval; Ref, reference.

Discussion

Optimal management of small NF‐panNETs is unknown. Several factors explain the lack of consensus of European‐ and U.S.‐based treatment guidelines. First, panNETs are rare neoplasms, making large prospective randomized trials very difficult to conduct. Moreover, there is heterogeneity in the clinical behavior and aggressiveness of panNETs, irrespective of size; whereas most small panNETs confer an indolent course, some studies have shown that some cases can show aggressive behavior 7, 8. A study by Haynes et al. showed that even small NF‐panNETs <2 cm in size that underwent surgical resection had a 7.7% rate of recurrence 7. Differences in WHO grade, Ki‐67, mitotic index, degree of differentiation, and molecular profile may all contribute to the heterogeneity. For example, in vitro panNET cell lines were found to harbor a wide array of different mutations, including MEN1, ATRX, DAXX, and TP53 18. Complicating the issue further is the fact that technological advancements in diagnostic imaging have led to overdiagnosis of small, indolent panNETs. In this study, we present what we believe is the largest cancer registry–based study reporting on outcomes of pancreatic neuroendocrine tumors.

Surgical resection of panNETs has historically been the mainstay of treatment of localized disease, in an era in which most of these tumors were functional and/or symptomatic 19. However, pancreatic surgery is associated with significant postoperative morbidity and non‐negligible mortality. In one series, 50% of the patients had complications including postsurgical infection, pancreatic fistula, endocrine and exocrine insufficiency, and severe hemorrhage 4, 20. The location of the tumor within the pancreas and the surgical approach are also factors to be taken into consideration. A systematic review showed an overall pooled in‐hospital mortality rate of 3% after enucleation, 4% after distal pancreatectomy, and up to 6% after pancreatoduodenectomy 21. A meta‐analysis comparing laparoscopic versus open pancreas resection showed a lower complication rate, intraoperative blood loss, and hospital length of stay in the former group, with no difference in mortality or pancreatic fistula rates 22. Consequently, active surveillance or observation have been proposed as reasonable management strategies 4, 10. Although surgical resection is universally recommended for functional or large panNETs in most current consensus guidelines, the management of small NF‐panNETs remains controversial 23. Some studies demonstrated improved outcomes with surgical resection in all panNETs 7, 8, whereas other studies advocate for observation in select small NF‐panNETs as a safe management approach 6, 24, 25. Thus, many single‐center retrospective studies have attempted to better understand the natural disease course of NF‐panNETs by evaluating tumor growth rate and rate of nodal/distant metastasis as surrogates for survival outcomes.

A review of the literature on predictive and prognostic factors that may guide treatment guidelines in NF‐panNETs has shown conflicting findings. One study showed that tumors >4 cm had a worse overall survival than those 2–4 cm and <4 cm 26. However, this was a univariate analysis, and the study population included patients with neuroendocrine carcinoma. A more recent study of 128 patients with NF‐panNETs in which a multivariate analysis was done failed to validate tumor size as a prognostic factor 8. However, despite being a multi‐institutional study, only patients undergoing pancreatectomy were included, therefore introducing selection bias against patients treated conservatively. Yet another French multicenter study suggests using a size cutoff of 1.7 cm based on a retrospective analysis of 80 patients with NF‐panNETs 27. With respect to nodal metastasis, many studies have shown that LN metastasis was associated with decreased disease‐free survival 12. However, when considering small panNETs <2 cm, an analysis of the Surveillance, Epidemiology, and End Results data from 1988 to 2009 did not predict a higher disease‐specific mortality with lymph node positivity 28. Tumor grade is another potential prognostic factor with uncertain significance. Although most studies concur that grade 3 tumors have worse outcomes, some studies even suggest that grade 2 tumors have significantly worse survival than grade 1 tumors 29. In fact, the NCCN guidelines recommend considering observation specifically for low‐grade, incidentally discovered panNETs <1 cm in size 15. Thus, the management of intermediate‐grade small tumors remains controversial.

The only prognostic factor that has been validated in most studies is older age 3, 8. In our study, age, CD comorbidity score, location of the tumor within the pancreas, stage, and surgery of primary site were found to be independent prognostic factors in NF‐panNETs. Interestingly, intermediate‐grade tumors did not result in worse survival outcome compared with low‐grade tumors. A potential explanation for this lack in survival difference could be attributed to insufficient events as well as a relatively small sample of patients with grade 2 tumors.

One of the very few prospective studies to specifically address small NF‐panNETs <2 cm was done by Ganjoux et al. 24. All 46 patients were followed with serial imaging, with the option of undergoing surgical resection if symptomatic or tumor growth >20% or >2 cm. After a median follow‐up of 34 months, 17% eventually underwent surgical resection. Overall median tumor growth was 0.12 mm per year. During the study, no patients developed nodal or distant metastasis. The authors concluded that there was no survival outcome difference between surgically treated patients and those undergoing nonoperative management. In our study, we found that patients with NF‐panNETs <1 cm were the only group in which surgical resection was not associated with an improved overall survival. Interestingly, patients with tumor size 1–2 cm who underwent surgical resection had a higher OS, after adjusting for other variables. This finding is noteworthy given the fact that ENETS and NCCN guidelines recommend different management for this group of patients 15, 30.

Another factor that has been evaluated is the incidental discovery of these tumors. In a case series, up to 82.2% of all resected NF‐panNETs were incidentally discovered 7. Another study provides a lower estimate of 45% 19. A retrospective study of patients with incidental NF‐panNETs showed a 20.1% lymph node positivity rate, and 10.8% of cases had distant metastasis at the time of surgery 7. Nevertheless, this study did not stratify patients by tumor size, as 72% had a tumor size >2 cm. After adjusting for multiple covariates including tumor stage, grade, and location, a study reported a 4.67‐fold increase in the hazard of death in symptomatic patients compared with those diagnosed incidentally 19. A caveat to consider is that this study included functional tumors, which precludes any conclusion about NF‐panNETs specifically. Moreover, a multivariate analysis of predictors of progression‐free survival in 124 patients with incidental panNETs did not elucidate any significant factor, including surgical resection 29.

Our study has several limitations. Most importantly, the retrospective nature of the data precludes the ability to make definite recommendations on the management of NF‐panNETs based on tumor size. Patients following nonoperative management may have done so as a result of comorbidities, which would likely negatively affect survival rates in this group. Patients in the surgical group must have survived from the time of diagnosis to the time of surgery, thus introducing immortal time bias. The effect of this bias is likely negligible given the fact that the median time from diagnosis to surgery was only 26 days. In addition, the National Cancer Database does not provide data on symptomatology; hence, extrapolation from the ICD‐O‐3 code had to be made. Whether these tumors were incidentally discovered is also unknown. Moreover, the NCDB only records overall survival; progression‐free survival and disease‐specific mortality would have been interesting outcomes to analyze. Another limitation is the grading of the tumors, which lacked a centralized pathology review. Additionally, the use of octreotide is not captured by the database.

Conclusion

To our knowledge, this retrospective analysis is one of the largest on well‐differentiated NF‐panNETs. Active surveillance is potentially a safe approach for panNETs <1 cm. Larger tumors likely need surgical intervention. Our study showed that age at diagnosis, CD comorbidity score, location of the tumor within the pancreas, and disease stage were independent prognostic factors whereas tumor grade (grade 1 vs. 2) was not in our cohort. Prospective randomized clinical trials are needed in order to conquer the current challenges of patients with panNETs. A potentially promising study would be a multicenter clinical trial randomizing patients with small panNETs to surgical intervention versus active surveillance, evaluating disease‐free survival as a primary outcome and using prognostic factors identified in our study as covariates. Although grade was not a prognostic factor in our analysis, we strongly recommend including it in future studies.

Author Contributions

Conception/design: Hussein A. Assi, Sarbajit Mukherjee, Vipul Pareek, Hassan Hatoum

Provision of study material or patients: Hussein A. Assi, Vipul Pareek, Hassan Hatoum

Collection and/or assembly of data: Hussein A. Assi, Sarbajit Mukherjee, Vipul Pareek, Hassan Hatoum

Data analysis and interpretation: Hussein A. Assi, Sarbajit Mukherjee, Michael Machiorlatti, Sara Vesely, Hassan Hatoum

Manuscript writing: Hussein A. Assi, Sarbajit Mukherjee, Pamela Kunz, Michael Machiorlatti, Sara Vesely, Vipul Pareek, Hassan Hatoum

Final approval of manuscript: Hussein A. Assi, Sarbajit Mukherjee, Pamela Kunz, Michael Machiorlatti, Sara Vesely, Vipul Pareek, Hassan Hatoum

Disclosures

Pamela Kunz: Advanced Accelerator Applications, Ipsen, Lexicon, Novartis (C/A), Advanced Accelerator Applications, Ipsen, Lexicon, Xencor, Brahams (RF). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board