PROGNOSTIC AND PREDICTIVE FACTORS ON OVERALL SURVIVAL AND SURGICAL OUTCOMES IN PANCREATIC NEUROENDOCRINE NEOPLASMS: recent advances and controversies

Abstract

Introduction:

Recent advances in the diagnostic modalities and therapeutic agents have raised the importance of prognostic factors in predicting overall survival, as well as predictive factors for surgical outcomes, in tailoring therapeutic strategies of patients with pancreatic neuroendocrine neoplasms(panNENs).

Areas covered:

Numerous recent studies of panNEN patients report the prognostic values of a number of clinically-related factors(clinical, laboratory, imaging, treatment-related factors), pathological factors(histological, classification, grading) and molecular factors on long-term survival. In addition, increasing number of studies showed the usefulness of various factors, specifically biomarkers and molecular makers, in predicting recurrence and mortality related to surgical treatment. Recent findings(last 3 years) in each of these areas, as well as recent controversies, are reviewed.

Expert commentary:

The clinical importance of prognostic and predictive factors for panNENs is markedly increased for both overall outcome and post resection, as a result of recent advances in all aspects of the diagnosis, management and treatment of panNENs. Despite the proven prognostic utility of routinely-used tumor grading/classification and staging systems, further studies are required to establish these novel prognostic factors to support their routine clinical use.

1. Introduction

Pancreatic neuroendocrine neoplasms(panNENs) comprise a heterogeneous group of neoplasms that show wide spectrum of biologic behavior including some being associated with a functional syndrome, as well as wide variation in growth behavior[1–3]. PanNENs can vary widely in their disease stage, tumor differentiation, histological proliferative index, as well as various clinicopathological factors[4, 5]. Treatment strategy of individual patients is strongly affected by these factors, thus establishment of useful prognostic factors is becoming increasingly important in all aspects of their treatment/management [6, 7]. Histological grading systems and TNM staging classifications[2, 8, 9] as outlined by the World Health Organization(WHO), European Neuroendocrine Tumor Society(ENETS) and American Joint Committee on Cancer(AJCC), are now of established value in prediction of prognosis[2] and are now routinely used in the real clinical practice worldwide. In addition, a number of studies have investigated the diagnostic utility of various clinically-related factors and pathological factors, as well as biomarkers. Moreover, recent developments in diagnostic/imaging modalities, as well as therapeutic agents, have dramatically changed the situation surrounding patients with panNENs, by identifying numerous prognostic factors[6]. Recent advances in the diagnostic and imaging modalities have significantly increased the incidence and prevalence of patients with panNENs, especially for those in earlier/localized disease stage[4, 5]. Because such patients are usually treated with surgical resection, defining the high-risk group of disease recurrence is essential to decide the follow-up and management method(e.g. imaging modalities, intervals), which could also help tailor the use of medical resources[7]. To address these issues, a number of recent studies have investigated the possible prognostic and predictive values of numerous clinicopathological and genetic/molecular factors as well as biomarkers on surgical outcomes. However, to date, evidence that supports the clinical use of these prognostic/predictive factors in optimizing therapeutic and management/follow-up strategy for an individual patient with panNENs is still limited. In this paper, we review recent papers within the last 3–5 years, that have proposed prognostic factors for panNENs. We include both clinical, laboratory, pathological, histological and molecular factors reported to have prognostic value both for survival, as well as for surgical outcomes.

2. Methods

This review summarizes/discusses the recent insights into prognostic/predictive factors, by focusing on the studies that are published from the past 5 years, particularly concentrating on the last 3 years, which are available in the MEDLINE, abstracts, or Meeting proceedings. This review concentrates on the direct effect of various factors on overall prognosis and surgical outcomes, and thus we primarily dealt with the reports which provide exact prognostic information on progression-/disease-/recurrence-free survival(PFS/DFS/RFS), overall survival[OS] and/or disease-specific survival[DSS]. A number of studies reporting the predictive factors on other clinical variables or outcomes(i.e. histological grade, Ki-67 index, lymph nodes/distant metastases, etc) will not be discussed in detail in this review, although these can have uncertain/indirect predictive effects on overall prognosis, as well as post-surgical recurrence and mortality[10]. In most cases this review is limited to reports containing only patients with panNENs(pancreas origin) or reports that have a large proportion of panNENs patients; and that are reporting uni- and multi-variate analysis. In some sections of the review which includes large number of studies, we only dealt with the results of multi-variate analysis.

3. Factors associated with overall survival in panNENs(Tables 1–4)

Table 1.

Clinical factors and laboratory test/biomarkers associated with worse OS/DSS in panNENs

Factors	Metastatic/Advanced		All stage
	OS	DSS	OS	DSS
Clinical factor
Age, high	[5, 11–18]/[100, 101]	[11, 13, 16]/[100]	[19–32]/[38, 59, 94]	[26, 28, 33]/[34]
Gender, male	[17]/		[20, 21, 23–26, 28, 35] /[117]	[26, 28, 35] /[34]
Unmarried	[37]	[37]	[21, 23, 28, 35]/[26]	[26, 28, 35]
Non-function(vs function)	[19]		[38]/[22–24, 26, 28, 39]	[33]/[26, 28]
Race	[18]		[26]/[20, 28]	[26]
Performance status, poor			[39, 103, 119]	[98]
Year of diagnosis/treatment	[5, 15]	[16]	[22, 51]/[4]	[51]
Symptomatic			[118, 318, 319]
Comorbidity score, high	[101]
Poverty/low income			[32]/[26]	[26]
Hospital type, non-academic	[12, 101]		[32]
Laboratory test/Biomarker
CgA, high	[18, 54]		[39, 55]/[38, 56–59]
No CgA reduction			[59]
NSE, high	[14]		[55]
5HIAA, high				[62]
Serotonin secretion			[55]
AFP, CA19–9, high	[36]
CEA, high	[36]		[38]
NETest, high	[75](OS)[(PFS)
Number of VEGF SNP			[139]
Change in circulating tumor cells	[80]
Presence of circulating tumor cells	[81]
NLR, high			[118]
Albumin, low	[36]
Bilirubin, high	[36]
AST, high	[36]
ALP , high	[320](PFS)/[36]
LDH, high			[321]

Results of multi-variate analysis are shown in red and bold, while results of uni-variate analysis are shown in black.

AFP, alpha-fetoprotein; ALP, alkaline phosphatase; AST, aspartate aminotransferase; CA19–9, carbohydrate antigen 19–9; CEA, carcinoembryocnic antigen; CgA, chromogranin A; DSS, disease-specific survival; HIAA, hydroxyindoleacetic acid; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; NSE, neuron-specific enolase; OS, overall survival; PFS, progression-free survival; SNP, single nucleotide polymorphism; VEGF, vascular endothelial growth factor.

Table 4.

Molecular factors associated with worse OS/DSS in panNENs

Factors	Metastatic/Advanced		All stage
	OS	DSS	OS	DSS
Molecular factor
α-internexin, negative			[137]
ALT, negative	[102]
ALT, positive			[102, 150]/[151, 153](PFS)	[149, 152]/[150, 151]
BRAF, altered	[146]
CD117, positive			[123]
DAXX/ATRX , positive	[102, 128, 146, 147]		[147]
DAXX/ATRX, DAXX, negative,			[148](OS) [102, 149–151](PFS)	[149]/[150, 151]
DAXX/ATRX and MEN1, positive	[128]
MEN-1, positive	[146, 147]
MMRd, positive			[123]
PD-L1, positive	[332]
PR, loss			[136]
PTEN, low			[136]
SSTR2A, negative			[31, 120, 155]
SSTR5, negative			[155]
TP53, mutant			[154]

See Table 1. legend for explanation of data included.

ALT, alternative lengthening of telomeres; ATRX, alpha-thalassemia/mental retardation syndrome X-linked; BRAF, v-raf murine sarcoma viral oncogene homologue B1; DAXX, death domain-associated protein; DSS, disease-specific survival; MEN1, Multiple Endocrine Neoplasia-type 1; MMRd, mismatch repair deficiency; OS, overall survival; PD-L1, programmed death-ligand 1; PR, progesterone receptor; PTEN, phosphatase and tensin homolog deleted from chromosome 10; SSTR, somatostatin receptor.

A number of studies have reported the prognostic value of various clinically-related factors[clinical, laboratory(Table 1)], imaging and treatment-related factors(Table 2)] as well as pathologic factors[histological, classification, grading](Table 3) and molecular factors(Table 4)] affecting OS/DSS in series of patients with panNENs(i.e. from initial diagnosis to death). In general, factors were evaluated in panNENs patients with both advanced disease as well as all stages of disease, are reported to show similar significant prognostic power(Tables 1–4).

Table 2.

Imaging and treatment-related factors associated with worse OS/DSS in panNENs

Factors	Metastatic/Advanced		All stage
	OS	DSS	OS	DSS
Imaging factor
SUV/max, high(FDG-PET)	[91]		[322]
SUV/max, low(SSTR-PET)	[14]		[323]
Tumor volume, high(DOTATATE)				[62]
Enhancement/vascularity, low			[324]
MPD obstruction			[324]
Irregular margin			[324]
Treatment-related factor
Shorter PFS	[325]
No treatment	[115]
No Surgery	[18, 36, 93]		[20, 21, 23–26, 32, 41, 51, 94, 97]/[22, 147, 319, 326, 327]	[26, 33, 41, 51, 98]
No radical/aggressive surgery	[12] /[109]		[59]
No primary tumor resection	[11–13, 15–17, 37, 100–102]/[328–330]	[11, 13, 16, 37, 100]	[35, 103, 104]	[35]
No systemic therapy	[100, 111, 113]/[112]	[100]	[103]	[98]
No liver-directed therapy	[113]
No Radiotherapy			[20]

Results of multi-variate analysis are shown in red and bold, while results of uni-variate analysis are shown in black.

DSS, disease-specific survival; FDG, fluorodeoxyglucose ; MPD, main pancreatic duct; OS, overall survival; SSTR, somatostatin receptor; SUV, standardized uptake value.

Table 3.

Histological factors/classification/grading associated with worse OS/DSS in panNENs

Factors	Metastatic/Advanced		All stage
	OS	DSS	OS	DSS
Histological factor/Classification/Grading
Grade	[12, 14–16, 36, 80, 93, 101, 115]	[16]	[23–27, 30, 31, 39, 94, 104, 116–121]	[26]
Grade	[17, 113, 331]		[22, 29, 38, 56, 57, 147, 319]	[33, 62, 288]
Ki-67 index	[17, 54, 91, 102, 111]/[111]		[59, 103, 123, 137, 140]/[29, 38, 56, 57, 122, 321]
Higher Ki-67 in metastatic lesions	[113]
Mitosis			[38]/[29]
Differentiation, moderate	[5]		[28, 35, 51]/[20, 29]	[28, 51]
Differentiation, poorly	[5, 11, 13]/[124, 133]	[11, 13]	[20, 21, 28, 32, 35, 51, 122, 123]/[29, 129, 130, 132]	[28, 34, 35, 51, 98]
Stage(AJCC, ENETS)			[22–24, 26–28, 30, 116–119, 121–123, 136–138]	[26, 28, 33]
Stage(AJCC, ENETS)			[29, 38, 56, 59, 119, 120, 139, 319]
Size	[19]/[93]		[20, 26, 28, 139]/[29, 97, 120, 327]	[28]/[26]
T stage			[35]/[20, 94]	[34]
Adjacent invasion			[21]/[29, 38]
Lymphovascular invasion			[29]
Perineural invasion			[29]
Lymph nodes metastasis	[37]/[11, 15, 113]	[37]/[11]	[20, 21, 25, 97]/[94, 122, 136, 147, 319]	[34, 62]
Distant metastasis	[93, 111]		[19–21, 25, 26, 28, 35, 51, 104, 140, 141]/[29, 38, 57, 94, 120, 122, 147]	[28, 33–35, 51]/[26, 62, 288]
Hepatic tumor burden	[17, 54, 102, 113]		[140]
Number of metastatic sites			[39]/[104, 122]
Thymic tumor			[19]
Anatomical location, head/neck	[15, 100, 101]	[100]	[23, 27, 30, 32, 35, 51]/[20, 26, 28, 319]	[35, 51]/[26]

See Table 1. legend for explanation of data included.

AJCC, American Joint Committee on Cancer; DSS, disease-specific survival; ENETS, European Neuroendocrine Tumor Society; OS, overall survival.

3.1. Clinically-related prognostic factor; clinical, laboratory, imaging and treatment-related factor(Table 1, 2)

3.1.1. Clinical factors(Table 1)

Within the clinical prognostic factors, age, gender and marital status are the most frequent prognostic factors validated by multi-variate analysis in patients in series with either advanced disease or in all disease stages(Table 1). Higher age is associated with decreased OS[5, 11–32] as well as decreased DSS[11, 13, 16, 26, 28, 33]. The majority of those studies used the cut-off value of 60-years[5, 11, 13, 20, 23, 25, 28, 30], while others reported the usefulness of 40-[19, 24], 50-[33], 57-[34], 65-[27, 29, 32], and 75-years old[12].

Male gender[17, 20, 21, 23–26, 28, 35], except for one study[36], as well as unmarried status[21, 23, 26, 28, 35, 37] have worse prognosis from different studies. It is suggested that this may be partially due to a delayed diagnosis with advanced stage, as well as worse psychological mentation and socioeconomic status in unmarried patients[28].

The prognostic value of functional status of the panNENs, in which non-functioning tumors(NF-panNENs) generally are reported to have a worse prognosis, was found in several studies with multi-variate analyses[19, 33, 38], as well as with uni-variate analyses[22–24, 26, 28, 39]. This may be explained patients with hormone hypersecretory states, in the past, were generally diagnosed at an earlier disease stage, especially those with insulinoma, whereas patients with NF-panNENs were more likely to be diagnosed with larger tumors, as well as with advanced disease, mainly due to the absence of disease-specific symptom leading to earlier diagnosis[40]. However, recent changes in diagnostic modalities, as well as in screening methods, have resulted in an increased proportion of asymptomatic patients diagnosed with smaller tumors, at earlier disease stages[41], and thus the prognostic significance of hormone hypersecretion on survival in panNENs remains controversial and is not seen in some recent series[40, 42].

Not listed in Table 1, because of lack of uni-/multi-variate analysis, but reported in a number of series is the importance of whether the panNENs is sporadic or part of an inherited syndrome(i.e. Multiple Endocrine Neoplasia-type 1[MEN1], Von Hippel-Lindau[VHL]). In many series, panNENs in MEN1/VHL are generally reported to have a better prognosis and be less aggressive, however, particularly in the case of MEN1, this is not supported by all studies[43–50].

In addition, as a result of significant advances in the treatment/management modalities in advanced panNENs, recently treated/diagnosed patients are reported to have more prolonged survival, regardless of tumor stage[5, 15, 16, 22, 51].

3.1.2. Laboratory test/biomarkers(Table 1)

Although the diagnostic performance of several biomarkers, especially chromogranin A(CgA), has been extensively investigated, its prognostic utility still remains controversial, because its plasma levels are affected by a number of factors, particularly the use of proton pump inhibitors(PPIs)[52, 53]. Elevated baseline levels of serum CgA are reported to be an independent prognostic factor both in advanced stage[18, 54] and in all disease stages[39, 55], while other studies reported its limited role as a prognostic factor in panNENs[56, 57]. In addition, optimal cut-off values predicting prognosis vary between different studies, ranging from 1x to 10x the upper limit of normal(ULN)[18, 38, 39, 54–59]. Other studies also reported that early CgA response[59] as well as elevated neuron-specific enolase(NSE) levels[14, 55] were associated with decreased OS. Pancreastatin, another CgA degradation product, is reported to have greater specificity than CgA as a NEN marker and its levels not affected by PPIs[60, 61]. However, there is only limited studies on it at present in relatively small number of patients.

Urine 5-hydroxyindoleacetic acid(5-HIAA)[62] as well as serotonin secretion[55] is reported to have prognostic value, especially in patients with carcinoid syndrome, which will not be discussed in this review(refer[63] for detailed information). Older studies report the degree of elevation of a specific hormone level, such as gastrin in Zollinger-Ellison syndrome, also has predictive value and correlates with increasing disease extent[43, 64–68].

Recently, the diagnostic/prognostic utility of a blood-based multianalyte NET gene signature, the NETest, for panNENs, as well as for NENs from other tissues, is reported from several studies, demonstrating higher sensitivity and specificity than other existing diagnostic modalities, including CgA and imaging[69–73]. In a recent, prospective study[74] involving 140 NEN patients and 113 healthy volunteers, the NETest had a sensitivity of 93% and specificity of 56%, whereas CgA had a sensitivity and specificity for identifying patients with NENs of 56% and 83%, respectively. The CgA level correlated with tumor load(p<0.001), whereas the NETest did not, nor did it correlate with tumor location, grade or stage[74]. The authors[74] concluded that low specificity of the NETest precluded its use as a screening test for GEP-NENs. However, its greater sensitivity over CgA, irrespective of stage, enhanced its potential as a marker of disease progression in follow-up or as an indicator for residual disease after surgery. The proposal is consistent with other studies reporting the usefulness of the NETest in predicting disease progression and stabilization[71, 73–75]. The NETest has also been reported to be a sensitive method for detecting tumor progression/change with anti-tumor post-treatments(surgery, etc) as well as treatment response[73, 75–78], and some of these will be discussed in later sections regarding surgical treatment.

The prognostic utility of identifying circulating tumor cells(CTCs) was reported recently in a number of malignancies, including panNENs[79]. Although the proportion of panNEN patients in a recent study, of the value of CTCs in NENs, was low(31/138=22%)[80], the overall results showed the identification of CTCs change from the baseline after treatment was strongly associated with OS, but not tumor grade[80]. The results from a previous study with NEN patients, including a small proportion of panNENs(24%), showed CTC are an independent predictive factor for worse survival[81]. Significant correlation between the presence of CTC and bone metastases was reported from another study involving patients with panNENs(n=119) and midgut NENs(n=121)[82], while association with PFS/OS was not examined. Although these results suggest the possible role of these biomarkers in predicting prognosis, at present the clinical applicability is unclear and further large, prospective studies iare needed to support its routine clinical use.

3.1.3. Imaging factors(Table 2)

Recent advances in imaging/tumor localization modalities[83, 84] have increased its clinical importance beyond only assessing tumor location/extent, as was the case in the past with primarily cross-sectional imaging studies(ultrasound, CT scanning, MRI imaging), angiography or in rare cases with F-panNETs, tumor localization using assessment of hormonal gradients[83, 85–89].. Recently, for both cross-sectional imaging studies, as well as molecular imaging studies(¹⁸F-fluorodeoxyglucose positron emission tomography[¹⁸F-FDG PET], ⁶⁸Ga-DOTATOC PET/CT, ¹¹¹In-labeled somatostatin analogues[SSA] SPECT imaging, ⁶⁴Cu-DOTA-peptide imaging, ¹⁸F-DOPA PET/CT), an increasing number of studies are reporting strong correlations of imaging parameters with the tumor grading and show prognostic value[83, 84, 90–92]. For these different imaging modalities, important imaging prognostic factors have included texture analyses of the imaging results, various parameters from image tumor contrast patterns, computation of imaging modalities SUV/Max and other isotope parameters[83, 84]. Molecular imaging has become particularly important in this regard for its prognostic value and ability to change clinical management[84]. Particularly important is the finding that panNENs/NENs showing higher uptake on ¹⁸F-FDG PET correlate with a poorer OS and that this result correlates closely with advancing classification/grade[91]. These prognostic findings are not specifically covered here because they have been covered in recent reviews[83, 84, 90–92].

3.1.4. Treatment-related factors(Table 2)

Numerous studies report surgical resection significantly prolongs OS/DSS of patients with panNENs both in advanced stages[18, 36, 93] and in all stages[20, 21, 23–26, 32, 33, 41, 51, 94–99], regardless of functional status and tumor size[32], or tumor grade[98]. A number of studies report prolonged OS/DSS in patients with advanced disease who underwent primary tumor resection, even in advanced stages[11–13, 15–17, 35, 37, 100–104]. These results strongly support the conclusion that surgical resection should be initially considered if complete removal can be achieved. On the other hand, patients with small(<2 cm), asymptomatic, NF-panNENs are increasingly treated with a watch and wait approach[7], which will be discussed more detail in a later section. Several studies report increased survival benefit and prognostic value of aggressive/radical surgery(i.e. metastasectomy) in patients with advanced disease[12, 59, 105–108]. Although this approach is supported by a recent meta-analysis[109], there are no randomized, control studies and the approach remains controversial[106]. Further evidence is required to evaluate the true prognostic power of primary resection as well as radical/aggressive surgery, and thus, at present, its indication should be carefully considered for each patient. In patients with refractory F-panNENs syndromes with advanced disease, such an approach is reported to improve prognosis and help control the hormonal symptoms[109, 110].

As mentioned earlier, patients receiving various antitumor therapies including targeted therapy[111, 112], systemic therapy[113] and chemotherapy[98, 100, 103] live longer and have a better prognosis than those whom did not receive such treatment. This result suggests a significant contribution of the recent arrival of these novel therapeutic agents, in survival extension in advanced panNENs. In addition, the increasing importance of these systemic therapies in a multidisciplinary approach to advanced panNENs can decrease the importance of aggressive/palliative surgery among these population[6].

3.2. Pathological prognostic factor; histological factor, classification/grading, molecular factor(Tables 3, 4)

3.2.1. Histological/classification/grading factors(Table 3)

For various histological prognostic factors listed in Table 3, higher grade(WHO) and higher stage(AJCC, ENETS), in the majority of the studies, are the most prominent factors associated with worse OS/DSS. The 2010 WHO classification system categorizes gastroenteropancreatic NENs(GEP-NENs) into three groups based on the Ki-67 labeling index, mitotic counts and cellular differentiation including NET G1, NET G2 and G3/neuroendocrine carcinoma(NEC)[114]. The prognostic utility of this grading system was validated by numerous studies involving panNENs in advanced stages[12, 14–16, 36, 93, 101, 115] as well as series including patients in all stages[23–27, 30, 31, 39, 94, 104, 116–121]. Of these, numerous studies reported the poorer survival in patients with G3 tumors, showing an increased risk of mortality frequently ranging from 3.1 to 8.8-fold[12, 14–16, 23–27, 30, 31, 39, 80, 93, 94, 101, 104, 115–121]. These results strongly support the routine use of tumor grading in clinical practice in patients with panNENs, regardless of disease stage. Recent studies have emphasized the importance of heterogeneity in grade 3(G3) tumors, which is primarily being determined by differences in differentiation, which has a marked effect on prognosis, as well as treatment approaches[98, 122–124]. These studies show both well-differentiated and poorly-differentiated G3 panNENs(i.e. Ki-67 index >20% and/or mitotic index >20/10 HPF) occur and can not easily be differentiated by the Ki-67 values alone, and they vary markedly in prognosis[98, 122–124]. In addition, differences in the genetic profiles between poorly- and well-differentiated tumors are reported in a number of studies, with poorly-differentiated tumors characterized by higher frequencies of mutations of TP53, RB1, and KRAS than with well-differentiated tumors[124–131]. In contrast, DAXX, ATRX, MEN1 and PTEN are more frequently mutated in well-differentiated tumors[126–130]. Some of these mutations are reported to be associated with different clinical outcomes to chemotherapy[124], which will not be discussed in this review. The poorly-differentiated G3 tumors have a worse prognosis[98, 122–124, 129, 130, 132, 133], with HR for death ranging from 2.8 to 8.3-fold[98, 122, 123]. Accordingly, this issue has been reinforced in the latest 2017 WHO classification for panNENs, which divides the G3/NEC group into “G3NET”(well-differentiated) as a distinct subgroup from “G3NEC”(poorly-differentiated)[134, 135]. This distinction is clinically important not only because of the differing prognosis, but also because the treatment approaches differ and recent studies show they have a different pathogenesis[124, 128, 129].

Numerous studies report that patients with higher/advanced disease stage have a worse OS/DSS[22–24, 26–28, 30, 33, 116–119, 121–123, 136–138] with HR for mortality ranging from 1.6- to 17.2-fold. This is supported by other results that patients with increasing tumor size[19, 20, 26, 28, 139], lymph nodes metastases[20, 21, 25, 37, 97], and distant metastases[19–21, 25, 26, 28, 33–35, 35, 51, 93, 104, 111, 140, 141], show shorter OS/DSS. Among 16 studies reporting prognostic significance of these common factors(tumor grade/differentiation or disease stage) by multi-variate analysis, decreased OS was observed with advanced grade/ decreased differentiation in 7 studies[25, 26, 28, 35, 117, 120, 122, 123] and with advanced disease stage in 8 studies[23, 27, 30, 51, 116, 118, 119, 121], respectively, whereas another study showed no difference[35].

Another increasingly reported prognostic marker is the anatomical site of the primary tumor in the pancreas, with panNENs located in the pancreatic head/neck showing worse prognosis than those in the pancreatic body/tail[15, 23, 27, 30, 32, 35, 51, 100, 101]. The diversity of the malignant behavior between pancreatic head and body/tail has also been reported from the studies in pancreatic adenocarcinomas[142]. The difference in ontogeny between pancreas head and body/tail leads to significant differences in cell composition, blood supply, lymphatic and venous backflow, and innervations, which result in different chemo- and/or radio-resistance, invasiveness, as well as genetic profiles[142]. Although the effect of anatomical location on mortality risk in panNENs still needs to be validated, it may provide important information in treatment selection, particularly for patients undergoing surgical resection.

3.2.2. Molecular factors(Table 4)

Recently, in addition to the NETest evaluated from blood samples, numerous studies have investigated the prognostic utility of epigenetic profiles and tissue-based biomarkers, as well as the expression of various non-coding RNAs, including long non-coding RNAs(lcnRNAs) and microRNAs(miRNAs)[74, 143–145].

Initially, a whole-exome sequencing study of panNENs reported the high prevalent mutation of MEN1(44%) and DAXX(death-domain-associated protein)/ATRX(alpha thalassemia/mental retardation syndrome X-linked)(43%) in sporadic panNENs, and that their presence correlated with prolonged survival[128]. Recent studies report the loss/deletion of MEN-1[146, 147] was significantly associated with better OS, whereas loss/deletion of DAXX/ATRX is associated in most studies with a worse OS/DSS[148–151] and PFS[102, 149–151], although not all studies[147]. This DAXX/ATRX prognostic discrepancy could be explained by different tumor groups being involved in these studies with some studies containing primary panNENs showing ATRX/DAXX mutation/negativity is associated with aggressive behavior/reduced survival, whereas in patients with metastases, it is correlated with a better OS[102, 147]. Alternative lengthening of telomeres(ALT) is associated with mutations of DAXX/ATRX and is reported to be associated with shorter DSS[149–152], PFS[102, 150, 151, 153] and chromosomal instability[151], particularly when occurring in the primary tumor. This will be discussed more detail in the later section dealing with surgical resection. The prognostic significance of other major cancer-related gene mutations was also evaluated by uni-variate analysis, including BRAF(v-raf murine sarcoma viral oncogene homologue B1)[146] and TP53[154]. Positive expression of somatostatin receptor type 2A(SSTR2A)[31, 120, 155], as well as SSTR5[155], demonstrated significant association with better OS, suggesting its usefulness as prognostic marker, in addition to imaging/therapeutic target.

The prognostic utility of various non-coding RNAs, including long-noncoding RNAs(lncRNAs) and microRNAs[miRNAs], was investigated in various malignancies, including panNENs[74, 79, 145, 156, 157]. For lncRNAs, a recent study demonstrated high expression of either of two lncRNAs, MALAT1 or HOTAIR, was negatively associated with aggressiveness, including lower T stage and less frequent development of metastases, independent of histologic grade[158]. The results of a preclinical study in an insulinoma cell line reported the possible anti-tumor effect of lncRNA MEG3(maternally expressed gene) overexpression and subsequent downregulation of its target c-Met[159]. The prognostic value of circulating miRNAs has been recently reported for small intestine NENs[74, 160], however for panNENs, as well as for other NENs, almost all studies for the possible prognostic value of miRNAs are from histological assessments[74, 145, 161]. A number of studies have reported the potential prognostic value of different miRNAs found histologically, with correlations with expression of the different miRNAs with tumor aggressiveness and/or survival[74, 162, 163]. Some of these studies have been performed in regard to surgical outcome and will be discussed in a later section.

Although a number of other studies have suggested the potential of various molecular markers, as shown in Table 4, none of these are currently being used in the clinical practice, due to the lack of strong evidence to support their additional value to existing markers to justify their routine use.

4. Factors associated with DFS/OS after surgical resection in panNENs(Tables 5–7)

Table 5.

Clinically-related factors associated with worse DFS(recurrence) and OS/DSS after surgical resection in panNENs

Factors	DFS	OS	DSS
Clinical factor
Age	[102, 171–173]	[170, 174–187]	[181, 188]
Gender, male	[189–192]	[176–179, 181, 190]	[178, 181, 188]
Non-function(vs function)	[191, 193–195]
Presence of genetic syndrome	[196, 197]
Symptomatic	[277, 284]	[249]
Comorbidity score, high		[180]
Insurance type, government/medicare		[179, 180, 199]
Hospital type, non-academic		[180]
Laboratory test/Biological marker
CgA, high	[198]	[199]
NETest, high	[76]
Pancreastatin, high	[204]	[204]
Elastase-1, high	[205]
NLR, high	[191, 208–210]	[191, 209]
PLR, high	[211]
LMR, high	[212]
GLRI, high	[256]	[256]
Imaging factor
Enhancement/vascularity, low	[214–216]	[216]
MPD dilatation/involvement	[217, 218]
Apparent diffusion coefficient	[214]
Bile duct obstruction	[218]
Irregular tumor margin	[219, 220]	[220]
Treatment-related factor
Resection status, R1/R2	[42, 193, 225–230]	[179, 191, 225, 231–237]
Aggressive surgery, yes	[190, 196, 197, 210]	[170, 190]
Aggressive surgery, no		[234]
Recurrence		[141, 196, 245]
Adjuvant therapy	[175, 196, 246]
Re-admission		[232]
Intraoperative blood loss	[262]

There are a number of variables available from many studies reporting results of multi-variate analysis in addition to uni-variate analysis in this section, because of the limitation of space , only the results of multi-variant analysis are shown.

CgA, chromogranin A; DFS; disease-free survival; DSS, disease-specific survival; GLRI, gamma-glutamyltransferase-to-lymphocyte ratio index; LMR, lymphocyte-to-monocyte ratio; MPD, main pancreatic duct; NLR , neutrophil-to- lymphocyte ratio; OS, overall survival; PLR, platelet-to-lymphocyte ratio.

Table 7.

Pathological factors associated with worse DFS(recurrence) and OS/DSS after surgical resection in panNENs

Factors	DFS	OS	DSS
Molecular factor
ALT, positive	[102, 150]		[149]
CD133, positive	[253]
Cytokeratin19, high			[261]
CNPY2, positive	[286]
DAXX/ATRX , loss	[102, 149–151]	[148]	[149]
DAXX, loss + STC2, high	[299]
Endocan, high	[294]
Gastrin, positive		[184]
H3K36me3/ARID1A/CDKN2A, loss/deletion	[149]		[149]
HOPX , low	[336]
HOPX, hypermethylation	[336]
FoxP3+ tumor-infiltrating Lymphocyte, high		[187]
HDAC score, high		[290]
HuD(RNA binding protein), negative	[189]
IPM3, positive		[265]
MCM7 , high	[250]
miRNA-21, positive	[304]	[304]
miRNA-449a, positive(with/without HDAC4)	[293]	[293]
miRNA-196a , positive	[259]
PAX6, low	[337]
PHH3, positive	[230, 296]		[230, 296]
PR, loss and/or PTEN , low	[136]	[136]
SSTR2A, positive		[338]
STK33, positive	[192]
TPD52, low	[247]	[247]
Tumor associated macrophage, high	[251]	[251]
UCHL1/α-internexin, negative	[254]
16257 C/A SNP		[297]

See Table 3. legend for explanation of data included.

ALT , alternative lengthening of telomeres; ARID1A, AT-rich interactive domain 1A; ATRX, alpha-thalassemia/mental retardation syndrome X-linked; CDKN2A, cyclin-dependent kinase Inhibitor 2A; CNPY2 canopy FGF signaling regulator 2; DAXX, death domain-associated protein; DFS, disease-free survival; DSS, disease-specific survival; FoxP3, forkhead box P3; H3K36me3, histone H3 trimethylation at lysine 36; HDAC, histone deacetylase; HOPX, homeobox only protein homeobox; IPM3, insulin-like growth factor-II mRNA-binding protein 3; MCM7, minichromosome maintenance complex component 7; miRNA, microRNA; OS, overall survival; PAX6, paired box 6; PHH3, phosphohistone H3; PR, progesterone receptor; PTEN, phosphatase and tensin homolog deleted from chromosome 10; SNP, single nucleotide polymorphism; SSTR, somatostatin receptor; STC2, stanniocalcin-2; STK33, serine/threonine kinase 33; TPD52, tumor protein D52; UCHL1, ubiquitin carboxy-terminal hydrolase L1.

Despite increasingly sensitive preoperative imaging modalities[83, 84, 164–166], numerous intraoperative tumor localization methods such as intraoperative ultrasound[89, 164, 166–169] and more aggressive resections[107, 164, 166, 167], tumor recurrences, incomplete resections, and missed localized metastases with panNET resections are not uncommon[164, 166, 167]. Therefore, prognostic factors post-surgical treatment of panNENs for recurrence, prediction of aggressiveness, DFS and OS, are of increasing clinical importance. The availability of such prognostic factors allows tailoring follow-up times/procedures, the determination of when adjuvant/neoadjuvant treatments should be considered and what type of adjuvant/neoadjuvant treatments should be considered.

A number of studies have investigated the predictive value of various clinically-related factors[clinical, laboratory, imaging, treatment-related factors](Table 5) as well as pathologic factors[histological/classification/grading(Table 6), molecular factors(Table 7)] affecting recurrence and mortality in patients who underwent surgical resection of a panNEN. The majority of patients included in these studies are in early disease stage, which show more benign and indolent clinical courses than those in advanced stages. Thus, factors evaluated from these studies might be important not only for predicting recurrence and survival, but also in deciding when surgery is indicated, as in patients with small and asymptomatic panNENs who are frequently treated with careful surveillance, instead of surgical resection[7]. In general, the majority of factors which predict disease recurrence are also associated with the rate of mortality after surgical resection(Tables 5–7).

Table 6.

Pathological factors associated with worse DFS(recurrence) and OS/DSS after surgical resection in panNENs

Factors	DFS	OS	DSS
Histological factor/classification/grading
Grade	[42, 76, 102, 136, 151, 172, 174, 175, 190–192, 194, 210, 217, 218, 220, 225–227, 230, 231, 238, 241, 242]	[150, 151, 170, 173, 175, 177–179, 183, 184, 190, 191, 194, 198, 199, 225–227, 231–234]	[178, 181, 230, 271]
Grade	[245–262, 285]	[236, 249, 254, 256, 259, 261, 263–270]
Ki-67 index	[173, 182, 186, 195–197, 211, 226, 228, 231, 257, 260, 272–277]	[186, 226, 231, 235, 273, 278, 279]	[280]
Mitosis	[226]	[226]	[280]
Differentiation	[29, 185]	[176, 180, 185, 197, 202, 232, 237, 279]	[188]
Stage(AJCC/ENETS)	[136, 191, 192, 241, 250, 252–254, 256, 259, 261, 262, 281, 305]	[136, 170, 173, 174, 177, 180, 191, 227, 232, 234–237, 254, 259, 264, 268, 270, 279, 282]	[283]
Size	[76, 136, 149, 150, 173, 175, 189, 205, 208, 214–218, 227, 238, 248, 253, 272, 273, 276, 277, 284, 286, 287]	[175, 180, 235, 263, 265, 287]	[150, 271]
T stage	[151, 195, 251, 289]	[249, 290]
Adjacent invasion	[272]	[246, 282]
Lymphovascular invasion	[42, 172, 189, 194, 238, 251, 258, 259, 263, 295, 333]	[148, 174, 194, 267, 279]
Perineural invasion	[42, 174, 191, 196, 245, 259, 263, 299]	[148, 191, 197, 245]
Lymph nodes metastasis	[42, 76, 150, 172, 190, 193–195, 211, 231, 238, 245, 249, 251, 257, 258, 263, 272, 273, 276, 277]	[150, 176, 178–181, 190, 226, 231, 249, 256, 290, 293, 297]	[149, 178, 181, 230, 271, 280, 296]
Lymph nodes metastasis	[289–296]
Lymph nodes ratio	[171]	[177]
Distant metastasis	[173, 175, 216, 219, 225, 229, 230, 255, 257, 275, 284]	[170, 175, 176, 178, 179, 186, 194, 216, 227, 233, 256, 265, 278, 334]	[149, 178, 230]
Necrosis		[335]
Anatomical location, head/neck		[176–178, 180]

See Table 3. legend for explanation of data included.

AJCC, American Joint Committee on Cancer; DFS, disease-free survival; DSS, disease-specific survival; ENETS, European Neuroendocrine Tumor Society; OS, overall survival.

Although there are numerous studies reporting the results of both multi- and uni-variate analysis, because of the limited space, we primarily dealt with the results of multi-variate analysis in this section.

4.1. Clinically-related prognostic factors[clinical, laboratory, imaging and treatment-related factors](Table 5)

4.1.1. Post-surgical prognostic clinical factors(Table 5)

The most frequent clinical factors associated with recurrence and/or mortality after surgical resection in patients with panNENs were age[102, 170–181, 181–188], gender[43, 176–179, 181, 188–192] and functional status of the panNEN[191, 193–195], which are similar to that observed in the overall survival from diagnosis as described in Table 1. In terms of age, the most frequently used cut-off value was 60 years old[172, 175–179, 181, 182, 184, 185]. The majority of studies report male gender as a predictor for worse RFS[189, 191, 192] and OS/DSS[176–179, 181, 188]. In general, patients with a non-functioning tumor have greater risk of recurrence[191, 193–195]. This may be partially explained by an earlier disease stage at the time of diagnosis with patients with F-panNENs[40], the frequent inclusion of insulinomas in the various series which are more frequently non-malignant than other F-panNENs[40], as well as a significant proportion of small NF-panNENs treated without surgical resection[7]. It is reported that patients with panNENs associated with genetic syndromes, such as MEN1 or VHL, have significant higher risk of disease recurrence after curative resection[196, 197], while its prognostic value on survival still remains controversial(see section 3.1.1.).

4.1.2. Post-surgical prognostic laboratory test/biomarker factors(Table 5)

A few recent studies in panNENs reported the usefulness of CgA[198] in predicting DFS and OS[199] after surgical resection. Other groups reported comparative results evaluated from uni-variate analysis[200, 201], which were not confirmed by multi-variate analysis due to a small number of patients. In contrast, no significant association between basal CgA levels and DFS/OS was observed in other studies[202, 203]. This discrepancy underlies the controversy of its predictive power for DFS/OS in this setting. The efficacy of recently developed modality, the NETest, in predicting recurrence is reported in a study involving 35 patients with resectable panNENs[76]. Another study also showed superiority of the NETest in identifying recurrent/residual disease over that predicted from CgA assessment[77]. Other possible markers predicting surgical outcome of panNENs include serum pancreastatin[204] and elastase-1[205]. As mentioned earlier, some of these markers, particularly CgA and the NETest, may also be useful tools during the follow-up period in predicting residual disease as well as early detection of recurrence after surgical resection[71, 75, 206] and disease progression during surveillance Further detailed information of biomarkers, including findings from older studies, are reviewed in[79, 207].

Other potentially useful markers attracting increased attention in this area are the ratios of inflammatory markers in the blood, including the neutrophil-to-lymphocyte ratio(NLR)[191, 208–210], the platelet-to-lymphocyte ratio(PLR)[211] and the lymphocyte-to-monocyte ratio(LMR)[212]. Numerous studies in a wide range of neoplasms are reporting the prognostic significance of these ratios[213]. However, in panNENs, because of the limited data available at this time, the true prognostic significance of these markers, as well as optimal cut-off value to predict DFS and OS, still needs to be validated by future prospective studies. Nevertheless, the potential availability of these less-invasive and frequently-used prognostic markers might significantly contribute in the monitoring/managing patients undergoing surgical resections.

4.1.3. Post-surgical prognostic imaging factors(Table 5)

Common imaging factors associated with worse DFS/OS in regard to surgery in panNEN patients include tumoral hypo-enhancement/vascularity[214–216], the presence of main pancreatic duct involvement[217, 218], as well as the presence of irregular tumor margins[219, 220]. Several studies have compared disease recurrence in cystic panNENs versus non-cystic panNENs by uni-variate analysis, and showed longer DFS in cystic than in solid panNENs[221, 222], whereas others showed no difference in DFS/OS between these groups[223, 224]. Further information is covered by other recent reviews[83, 84, 90–92].

4.1.4. Post-surgical prognostic treatment-related factors(Table 5)

Resection status(R1 and/or R2) is the most frequent treatment-related factor reported associated with both DFS[42, 193, 225–230] and OS[179, 191, 225, 231–237], after surgical resection in patients with panNENs. Several studies report a decreased DFS in patients undergoing aggressive resection with advanced disease, including synchronous hepatic resection[210], multi-visceral resection[190], debulking-surgery[190], as well as major vascular resection[196, 197]. These are comparable to the results by uni-variate analyses from other studies[238–242]. In addition, several studies report that lymph nodes removal has limited survival benefit[199, 243]. At present, the basis for the decreased PFS in patients post these types of aggressive resections is unclear. In contrast, other studies report that complete/radical resection of distant metastases can significantly prolong the survival[190, 234, 235]. However, at present the prognostic value of aggressive surgery for metastatic disease still remains controversial. A recent meta-analysis showed the safety and feasibility of laparoscopic and/or robotic procedures, demonstrated significantly shorter length of hospital stay with equivalent rate of recurrence and post-operative complication compared to open surgery in panNEN patients[244]. Improved DFS in patients whom underwent laparoscopic/robotic resection were reported by uni-variate analysis from another group[197], suggesting the clinical utility of these minimally invasive surgeries in selected panNEN patients.

Disease recurrence is reported to be an independent prognostic factor for decreased survival after surgical resection[141, 196, 245]. Surprisingly, patients receiving adjuvant therapy are reported to have worse DFS[175, 196, , 246], mainly due to selection bias in these retrospective studies that patients with high risk of recurrence/mortality were more likely to receive adjuvant therapy, which likely underestimates its true value on disease recurrence. Currently, there is no evidence to support the efficacy of adjuvant therapy in panNENs, thus a high level of evidence, such as randomized control and prospective study, are warranted to address this issue.

4.2. Pathological prognostic factors[histological factors, classification/grading, molecular factors](Table 6, 7)

4.2.1. Post-surgical prognostic histological/classification/grading factors(Table 6)

A number of studies report the predictive value of various pathologic factors on DFS/OS/DSS in patients with panNENs(Table 6). The most important such factors by multi-variate analysis include including tumor grade, Ki-67 index, disease stage, tumor size, microscopic invasion, and the presence of lymph nodes/distant metastases(Table 6). The over-riding clinical importance of the recent panNENs classification/grading systems developed is attested to by the fact that of the most important pathological prognostic factors , tumor grade has now become the most frequent histologic factor validated by a large number of studies in predicting DFS[42, 76, 102, 136, 151, 172, 174, 175, 190–192, 194, 210, 217, 218, 220, 225–227, 230, 231, 238, 241, 242, 245–262], OS[150, 151, 170, 173, 175, 177–179, 183, 184, 190, 191, 194, 198, 199, 225–227, 231–234, 236, 249, 254, 256, 259, 261, 263–270] and DSS[178, 181, 230, 271]. Patients with G2 and G3 tumors have an increased risk of post-surgical recurrence, with HR ranging 1.4- to 7.3-fold[76, 136, 174, 175, 190, 194, 217, 218, 226, 241, 242, 245, 249, 251, 253] and 2.3- to 7.3-fold[42, 175, 190–192, 225–227, 238, 241, 246, 258], respectively. The risk of mortality is particularly increased in patients with G3 tumors(HR, 2.8–29.7)[ than with G2 tumors(HR, 1.2–7.9) [170, 179, 191, 194, 198, 209, 226, 230, 234, 249]. For studies reporting the predictive value of Ki-67 index on DFS[173, 182, 186, 195–197, 211, 226, 228, 231, 257, 260, 272–277] and OS/DSS[186, 226, 231, 235, 273, 278–280], the majority of those validate the cut-off value proposed by 2010 WHO classification to separate G1/G2[173, 182, 186, 195–197, 211, 226, 231, 257, 260, 275, 280] and G2/G3[195–197, 211, 226, 276, 280], while others showed the usefulness of Ki-67 threshold of 1% for G1 tumor[228], 5%[195, 260, 273] or 10%[274, 278] for G1/G2 tumor, as well as 55% for G3 tumor[235], respectively. Similar to the results described in the previous section 3.1.2., patients with poorly-differentiated tumor[29, 176, 180, 185, 185, 188, 197, 202, 232, 237, 279] and in advanced disease stages[136, 136, 174, 177, 180, 191, 191, 192, 227, 232, 234–237, 241, 250, 253, 256, 264, 268, 270, 279, 281–283] had poorer DFS and OS/DSS.

The predictive value of increased tumor size associated with decreased DFS[76, 136, 149, 150, 173, 175, 189, 205, 208, 214–218, 227, 238, 248, 253, 272, 273, 276, 277, 284–287], as well as poorer OS/DSS[150, 175, 180, 235, 263, 265, 271, 287] is reported by a number of studies. Many studies report the usefulness of a threshold of 2 cm[76, 149, 150, 150, 175, 180, 205, 208, 214, 238, 265, 271, 277]. This strongly supports the proposal that small(<2 cm), asymptomatic panNENs are good candidates for surveillance[32, 202, 218, 246, 288]. Moreover, an increased recurrence/mortality is observed with higher T-stage[151, 195, 249, 251, 289, 290], as well as with adjacent tumor invasion[246, 272, 282], raised the unanswered query whether locally advanced panNENs should initially be surgically resected or not[7].

Lymph node metastases are strongly correlated with increased recurrence[42, 76, 150, 172, 190, 193–195, 211, 231, 238, 245, 249, 251, 257, 258, 263, 272, 273, 276, 277, 289–296] as well as mortality[149, 150, 176, 178, 178–181, 181, 190, 226, 230, 231, 249, 256, 271, 280, 290, 293, 296–298] after surgical resection. Some studies also reported the predictive significance of the positive lymph nodes ratio or of the number of positive lymph nodes removed on DFS[171] and OS[177], raising the importance of lymph nodes sampling in predicting DFS/OS in panNEN patients. In addition, tumors located in pancreas head are reported to have worse OS[176–178, 180], which could suggest the diversity of malignant behavior related to anatomical location or possibly due to the need for more complex surgery in the pancreatic head area.

4.2.2. Post-surgical prognostic molecular factors(Table 7)

Recently, the predictive power of a number of tumor molecular factors on DFS/OS after surgical resection was reported(Table 7). The most prominent molecular factors include alterative lengthening of telomeres(ALT), as well as mutation/expression of ATRX and DAXX, which both are considered to be suppressors of ALT[102, 150]. As described in the previous section(3.1.2.), positivity of ALT expression histologically[102, 149, 150] and losses/deletions of ATRX/DAXX[102, 148–151, 299], are significantly associated with worse outcome after surgery. These results are consistent with the results of uni-variate analysis from some groups[299–301], but differ from other groups reporting the opposite results with ALT positivity[302] and DAXX/ATRX negativity[147] showing prolonged DFS. It is reported that ALT and DAXX/ATRX changes are correlated with higher disease stage and grade, and hence it has been proposed to be a late event in PanNET tumorigenesis[150]. The fact that the occurrence of these abnormalities appears to be linked with disease extent[102, 147] can result in conflicting prognostic power of these factors in different disease stages, as mentioned earlier. Besides this controversy, assessment of these molecular factors, in combination with WHO grading, can provide additional prognostic value in panNEN patients, specifically in patients with G3 tumors[129, 130].

Assessment of other reported prognostic molecular markers post-surgical resection include the assessment of PDX1(pancreatic and duodenal homeobox-1) and ARX(aristaless-related homeobox, X-linked), either of which are highly expressed in NF-panNENs(84–98%)[153, 303]. One study demonstrated that distant relapse after surgery almost entirely occurred in patients with PDX-negative/ARX-positive tumors[153]. In addition, the positivity of ALT was associated with increased recurrence, and its predictive value on DFS markedly increased when combined with ARX and PDX1 subtype, with patients with ARX-positive/ALT-positive tumors having disease recurrence[153]. The prognostic significance of PDX1 status was also examined in another study involving 46 panNEN patients(39 NET G1/2 NET, 6 NEC G3 and 1 MANEC). In addition to the negative correlation between PDX1 expression levels and tumor aggressiveness, markedly better DFS and OS in PDX1-positive patients was reported, with 5-yr DFS and OS rate of 95% and 100%, respectively[303]. Although the prognostic value of PDX1 status in a subset of NET(G1/2) still needs to be validated further in a prospective study including large number of patients[303], the above results suggest PDX1 assessment may have an important clinical role in predicting clinical/treatment outcomes.

Similar to other malignancies, predictive value of various non-coding RNAs, lnsRNAs and miRNAs, on surgical outcome have been recently studied in panNEN patients[152, 259, 293, 304]. In 39 patients with NF-panNENs, high expression of lncRNA H19 was associated with decreased DFS(p<0.001) and OS(p=0.059), although this was not validated by multi-variate analysis. In terms of miRNA, a recent in panNEN patients reported miRNA-21, miRNA-10a, and miRNA-106b were negatively associated with both PFS and OS, while miRNA-21 was the only independent prognostic factor confirmed by multi-variate analysis[304]. The expression of miRNA-449a was an independent predictor for poor OS from a study involving 57 patients with panNENs[293], whereas the combination of miRNA-449a and cytoplasmic positivity of histone deacetylase 4(HDAC4) is predictive of both decreased PFS and OS[293]. Another study investigating 37 panNEN patients reported that higher miRNA-196a expression was significantly correlated with poorer DFS(p<0.001) and OS(p=0.046)[259]. Although patients with panNENs with high miRNA-27b levels showed significantly shorter OS(p=0.016), its predictive value on DFS was not significant[259]. These results strongly support the potential value of the clinical utility of assessing non-coding RNAs in predicting surgical outcome. Additional information from older studies are well summarized in[79, 143].

Recently, the prognostic value of several immune-related markers is reported in panNET patients. The prognostic significance of the expression of forkhead box protein p3-positive regulatory-T cells was found a study including 101 panNEN patients[187]. A recent study involving 104 patients with G1/2 panNENs reported tumor-associated macrophage(TAM) infiltration was an independent predictor for decreased DFS(p=0.02) and DSS(p=0.02)[251]. The authors also report that high intra-tumoral CD8⁺ T cell infiltration correlated with prolonged DFS(p=0.05), whereas high peritumoral CD4⁺ cell and TAM infiltration were associated with a decreased DFS(p=0.02) and DSS(p=0.04)[251]. Another study reported that PD-1^high T cells(p=0.002) and PD-L1^high Type-II macrophages(p=0.004) were associated with decreased DFS, while PD-1^high T cells were predictive of decreased OS(p=0.013)[305]. The predictive value of the monocyte/macrophage marker, CD68, on disease recurrence was also reported from a study involving 97 NF-panNENs[262]. Although the predictive significance of most of these factors has not been confirmed by multi-variate analysis, these results suggest the potential of immune-related markers not only as a predictive factor for DFS/OS, but also as a novel therapeutic target in panNEN patients.

A number of studies in panNEN patients have reported the predictive value of various factors related to the PI3K(Phosphatidylinositol 3-kinase)/mTOR(mammalian target of rapamycin) signaling pathway on DFS/OS, including phospho-mTOR[281, 300], IGF-1R(insulin-like growth factor-1 receptor), PTEN(phosphatase and tensin homolog deleted from chromosome 10)[136], HIF(hypoxia inducible factor)-1α[292], TSC2(Tuberous Sclerosis Complex 2)[306], as well as CLEC3A(C-type lectin domain family 3 member A)[307]. Although the predictive value for surgical outcome of the majority of these factors has not been validated by multi-variate analyses, these may prove to be important in predicting responses to surgical and other treatments targeting such as everolimus[6]. This review will not explore this issue, due to the limited space. If interested, the reader is referred to [143, 144] for detailed information, especially for the findings from older studies, as well as results from uni-variate analysis.

5. Controversial and uncertain aspects of prognostic factors in panNENs

As reviewed in the previous paragraphs, there are a number of promising prognostic/predictive factors for survival and surgical outcomes reported for patients with panNENs. However, a number of areas of controversy and uncertainty remain.

One major such area which is not dealt with in this review is the identification of biomarkers predictive for diagnosis of NF-panNENs[52, 53, 69–73, 207]. While assessment of serum CgA has been widely used for this, numerous studies have shown not only does it lack sensitivity, but it also has major problem with specificity[52, 53, 207]. Not only is it elevated in numerous non-panNEN conditions, it is elevated in almost all patients taking PPIs[52, 53], which are of one of the most widely used drugs in the US and worldwide. Whether recent molecular tests, such as the NETest, will fill this need is unclear, with a recent prospective study[74] suggesting it lacks sufficient specificity to be useful as a diagnostic biomarker, and thus the use of imaging/histological studies remain the only established approaches at present.

In regard to the prognostic/predictive factors reported in the present study for overall survival and surgical outcomes for panNENs, a number of controversies/uncertainties also exist. For the various clinical features predictive of survival in studies, advanced age, unmarried status, male gender and the presence of a NF-panNEN over a F-panNEN are frequently reported. In regard to age, there is no agreement in the different studies on which age cut-off is the most appropriate to use. Furthermore, it remains unclear for both age, married status and gender, the determinants of their prognostic value. Higher age often correlates with poorer performance status, which affects efficacy and tolerability to each therapeutic option[35]. Male gender, as well as unmarried status, can be associated with worse psychological mentation and socioeconomic status[28]. However, the exact reasons for why these clinical factors are associated with shorter survival and worse surgical outcomes is still unclear. In the case of NF- vs F-panNENs having an effect on survival in numerous studies, this effect is generally attributed to the late diagnosis of NF-panNENs with patients only diagnosed late in their tumor course, with large primaries and advanced disease[40]. In recent studies with advances in diagnostic modalities and screening methods, an increasing proportion of patients with NF-panNENs are being detected early in their course[4, 5]. This will likely have impact on the prognostic value of functioning/symptomatic panNENs which will likely differ from results of older studies, and thus its true prognostic value needs to be re-evaluated in the future studies.

In regard to prognostic laboratory studies, the use of blood CgA or more recently, NETest levels, are the most frequently reported to be of value in predicting prognosis, as well as post-surgical recurrence and mortality(Table 1, 5). However, serum levels of CgA are frequently affected by non panNEN-related factors, as mentioned above. Therefore , many investigators do not routinely use serum CgA for prognostic purposes, primarily because of its low specificity, but also because of low sensitivity for detecting small tumors[52]. The NETest is receiving increasing attention as a useful biomarker in panNENs, as well as other NENs. Numerous studies report its superiority over other mono-analytes used and that it is not affected by the use of PPIs[69–73]. Although a number of studies report excellent sensitivity/specificity for the NETest, a recent prospective study[74] concluded its low specificity, which could affect its routine clinical use. At present it is not generally used routinely and its exact utility needs to be assessed by further prospective studies including not only different groups of panNENs, but also for prognostic value with different aspects of management. Other circulating biomarkers, including CTCs and non-coding RNAs(lncRNAs, miRNAs), also demonstrate promising results in numerous recent studies showing potential prognostic value in panNENs, although the true prognostic utility of these biomarkers needs to be prospectively validated to define their potential for routine clinical use.

For treatment-related factors with prognostic value, the most prominent factor in almost all studies is whether surgical resection was performed. However, the prognostic value of these surgical approaches in patients with advanced panNENs remains highly controversial[110]. This has occurred because of the lack of controlled, prospective studies in this area. Many current guidelines recommend cytoreductive surgery in patients with advanced panNENs if > 90% of the tumor can removed, however at present, its prognostic value in terms of survival remains in question[190, 196, 197, 210]. In the past this approach has been particularly important in patients with F-panNENs with hypersecretory states that could not be controlled medically, however, with the availability of new anti-secretory and anti-tumor treatments(everolimus, sunitinib, PRRT, SSA, liver-directed therapy, chemotherapy) this is becoming a decreasing indication[110].

Although the prognostic value of tumor grading/classification/staging systems is validated by numerous studies(Table 3, 6), there still remain several controversies in these systems, specifically related to heterogeneity in G3 tumors[98, 122–124] and for cut-off values for different grades/classes[111, 138]. A number of studies have proposed different cut-off values, particularly for the separation of grade G1 and G2, providing evidence that cut-off value of 5%[138, 195, 260, 273, 308] or 10%[111, 274, 278] for G1 from G2 provided significantly better separation of these groups. Furthermore, other studies propose dividing G3 tumors into two groups with the use of a Ki-67 < or ≥ 55% because of different overall survival of these two groups[103, 235, 309]. Numerous recent studies have demonstrated that G3 tumors identified by the original WHO classification(i.e. Ki-67 > 20%, mitotic rate > 20/10 HPF) are heterogeneous[98, 122–124]. This heterogeneity has been incorporated into the latest 2017 WHO classification for panNENs[134, 135]. The new G3 classification divides G3 tumors into a G3 NET grouping which are well-differentiated and G3NEC grouping which are poorly-differentiated with a small-cell and large-cell subtype[134, 135]. The importance of this re-classification is not only because the two G3 groups have a different prognosis and molecular pathogenesis, but it also distinguishes two groups that have been shown to respond better to different treatment regimens[124]. G3 NET and G3 NEC’s molecular pathogenesis differs with a higher rate of presence of ALT[149, 150] and mutations/losses in ATRX/DAXX[127, 129, 130] in NETG3, whereas mutations/losses of TP53, RB1, and KRAS are more frequent in NECG3[124–131]. The identification of these genomic factors is reported to have a prognostic effect among these patients, partially resulting from the different disease stages[102, 147]. Each of these proposed changes in the grading system has led to controversies/uncertainties and as demonstrated with the changes in the 2017 WHO classification, as more insights become available, there will be additional changes in the future. What will be particularly important is the prospective assessment of both the proposed changes in the classification system, as well as the molecular prognostic factors, in real clinical practice with the prospective validation done in a large cohort of panNEN patients.

In addition to the classification/grading systems, a number of other histological/pathological features, such as tumor size, presence of histological invasion, and extent of metastases(lymph nodes, liver, number of sites, location) have all been reported to have prognostic/predictive value(Table 3, 6). While these have been extensively used prior to the development of the classification/grading systems, particularly the presence and extent of liver metastases, they also are associated with controversies and uncertainties. Some, but not all, studies show lymph nodes involvement is predictive, but there is no agreement on what extent of involvement, number of positive lymph nodes or whether lymph nodes positive ratios are superior to number of lymph nodes involved[171, 177, 260]. These are particularly important because in relation to the degree of surgery proposed. The issue of tumor size as a prognostic determinant is particularly controversial. This is particularly illustrated by the controversy/uncertainty of the best treatment course in patients with small NF-panNENs or patients with inherited panNEN syndromes(MEN-1, VHL), where small NF-panNENs are increasingly treated with a watch and wait approach for tumors < 1.5–2cm in diameter rather than with surgery[7]. It is reported that only a small proportion of these tumors show aggressive growth[7], and at present there is no agreement on how these can be selected.

Targeting the immune system in patients with a number of common malignancies has become an increasing important novel therapeutic approach, therefore identifying useful molecular markers to predict prognosis and therapeutic response is of increasing clinical importance[213]. An increasing number of recent studies are reporting the prognostic values of blood-based ratios of inflammatory cells(Table 1, 5), as well as histologically-assessed immune-related markers(Table 4, 7), in patients with panNENs. However, the small cohort of patients included in these studies may overestimate the prognostic significant of these factors, and thus they need to be validated prospectively in future studies including large number of panNEN patients.

In terms of post-surgical prognostic factors predicting recurrence and mortality, numerous studies report the usefulness of tumor grade, staging, microscopic invasion, disease extent, as well as various molecular markers, however, a number of controversies and uncertainties still remain(Tables 5–7). Even though numerous studies established the predictive value of these parameters for recurrence and/or survival, at present there is no standardized approach in their use or agreement on how these should be used to affect management approaches. The latter would include their use in the timing of follow-up evaluations, helping to prioritize which investigations and imaging studies to perform on each visit relative to risks, or assisting in the selection of high-risk groups for neoadjuvant/adjuvant treatments. Furthermore, there is no generally accepted approach based on these prognostic factors which neoadjuvant/adjuvant therapies will be most appropriate for which patient. These issues each should be carefully evaluated by future systemic, prospective studies including a large number and all aspects of patients with panNENs.

In summary, although numerous studies reported the significant predictive values of various factors on overall survival and post-surgical outcomes, there remain a number of controversies and uncertainties of its true effect on prognosis/survival in panNEN patients which have affected their clinical routine use. All these controversies can only be resolved by future systemic, prospective, randomized control studies including a large cohort of panNEN patients.

6. Conclusions

The survival predictive value of the tumor grading , classification and staging systems on overall survival as well as on surgical outcomes has been extensively studied and well-validated by numerous studies, strongly supporting their prominent role as the most consistent independent prognostic factors for survival in the management/treatment in panNEN patients. An increasing number of studies provide data supporting the predictive utility of numerous clinically-related, pathological and molecular factors on prognosis/survival of panNEN patients. Particularly important clinically-related factors include higher age, male gender and unmarried status as predictive of poorer overall survival, as well as worse surgical outcomes. Prominent predictive biomarkers include CgA and the NETest, as well as a number of tumoral molecular factors, including the histological status of ALT, DAXX, ATRX, miRNAs, as well as immune-related markers, which all show promise in predicting long-term overall prognosis and surgical outcomes in panNEN patients. Furthermore, a number of recent studies suggest the usefulness of nomograms or biomathematical algorithms comprised of various parameters linked to prognosis/survival to calculate a prognostic score to predict long-term survival[11, 23–25, 38, 54], as well as post-surgical outcomes[113, 214, 245, 272, 310]. However, at present, except for the grading/staging systems, none of these prognostic factors is routinely used. Prospective studies are needed to support their routine use in real clinical practice. This review summarizes in detail the recent advances and remaining controversies/uncertainties in the establishment of effective prognostic/survival factors in the overall and post-operative management of panNEN patients.

7. Expert commentary

In the last few years, a number of factors have contributed to markedly changing the approaches to the diagnosis, management and treatment of panNENs , as well as NENs in other tissues, which have increased the need for reliable prognostic/predictive markers.

As a result of recent developments in diagnostic and imaging modalities, the incidence and prevalence of panNENs have markedly increased in the last few decades, an increasing fraction of all panNENs series is composed of patients with NF-panNENs, which are more likely to be diagnosed at an earlier disease stage, as well as with smaller tumors(< 2cm)[4, 5, 41]. These changes have led to controversies about the therapeutic and management strategies for these patients, particularly related to the surgical treatment[7]. In the past, all NF-panNEN patients with resectable tumors were surgically treated if complete resection could be achieved, particularly for patients with sporadic tumors[63, 110]. Because recent studies demonstrate that the most NF-panNENs with small tumors(< 2cm) show no growth or slow growth, these patients are now frequently followed without surgery[7]. This increasing incidence of patients not undergoing surgical resection emphasized the increasing importance of identifying prognostic factors in predicting which patients harbor potential aggressive tumors. The availability of sensitive predictive markers for tumor behavior could have significant impact on when and how-to follow-up these patients. Furthermore, the increasing tendency for a watch and wait approach in patients with small(< 1.5–2cm) NF-panNENs has led to controversy when and how-to follow-up these patients with the management tools currently available, as well as how to identify the high-risk group for tumor progression, and when the surgical resection should be indicated.

Other areas that have significantly changed in the last few years that are having a marked effect on the need, as well as potential availability of clinically useful predictive/prognostic factors, are the rapid advances in panNENs/NENs tumor imaging, the insights from molecular status of the pathogenesis of panNENs/NENs and determinants of their behavior, as well as the worldwide establishment of NEN centers/organizations which have provided diverse and large databases, as well as large, centralized patient groups for coordinated studies. In the recent past, all imaging was performed using cross-sectional studies(ultrasound, CT, MRI) which missed a large percentage of small lesions(< 1–2cm), as well as underestimated the true extent of the disease, affecting the reliability of classification/prognostic predicters[83]. With the availability of molecular imaging, primarily using ⁶⁸Ga-labeled SSAs and ¹⁸F-FDG PET imaging, the extent of the disease, changes with time and prognostic factors are now becoming better defined earlier in the disease course[311]. These increased abilities to assess disease extent/changes has increased the need for predicting disease course to tailor subsequent appropriate management[83]. The increased understanding of the molecular pathogenesis of panNENs/NENs is identifying molecular factors that are important in pathogenesis/tumor behavior which is providing a number of new factors which can be assessed in prognostic tests[143, 144]. Finally, the increasing development of NEN centers is providing increased information on the demographic and tumor behaviors in patients in different worldwide areas, which may result in differing prognostic factors.

Lastly, a major area of change contributing to need for sensitive prognostic factors is the proliferation of the number of treatment modalities that are now available for panNEN patients with advanced disease. This is represented by the recent approval of everolimus[6, 312], sunitinib[313], lanreotide[314], liver-directed therapies[315, 316] as well as PRRT[110, 317]. Each of these therapeutic agents demonstrate significant antitumor activity in panNEN patients with advanced disease, however, no study with a high level of evidence has directly and prospectively compared survival benefit of each of these agents. Prognostic markers to predict survival would have important clinical value for selection of the optimal treatment in a multidisciplinary approach to advanced panNENs, in addition to prognostic factors to predict which patient will benefit from each of the various therapeutic options. The latter area is not discussed in this review. In addition, recent developments in the medical treatment of advanced panNENs can have significant impact on the prognostic and predictive values of previously-reported prognostic factors in panNENs, as well as the role of traditional therapeutic approaches, including aggressive surgery for metastatic disease.

Each of the above areas are contributing to the current increased need to identify prognostic factors for all phases of the management of panNENs, as well as for their diagnosis. At present, no single prognostic factor fulfills all of these requirements. The different grading/staging systems are now well-established for their prognostic value, however, for diagnosis and various management decisions, none of the proposed clinical, laboratory, histological, pathological or molecular prognostic factors is widely used or has become well-established by prospective studies in different populations to result in their routine use. The recent development of algorithms involving combination of multiple prognostic variables may fulfill some of these needs, but at present none of these are well validated in prospective studies. Furthermore, the recent development of an increasing number of biomarkers/molecular prognostic factors including CTCs, the NETest, assessment of non-coding RNAs or gene mutations/alterations either alone or in combination may provide such a prognostic determinant.

The establishment of useful prognostic and predictive factors for panNENs will have significant clinical importance in the era of advanced diagnostic and imaging modalities which enable early tumor detection, as well as providing increasing therapeutic options with survival benefit in patients with advanced disease.

15. Key Issues.

• Numerous factors are leading to an increased need for prognostic/predictive factors in patients with panNENs

• These factors include: their increasing incidence, increasing incidental discovery of asymptomatic small tumors, the increased array of treatments for advanced disease, their variable courses, increasing insights into their pathogenesis and variable courses post-surgical treatments

• A number of factors have been identified that have prognostic/predictive for the overall survival or postsurgical disease course of patients with panNENs

• Prognostic factors identified for both survival and surgical outcomes include: clinically-related factors(clinical, laboratory, imaging, treatment-related factors), pathological factors(histological, classification, grading) and molecular factors prognostic predictive factors in panNETs

• Particularly important in all studies are the recent classification/grading systems, although there is still a need for additional prognostic factors

• Molecular insights are generating numerous predictive/prognostic factors, however generally, similar to the many of the other prognostic factors reviewed here, their evaluations are still limited, and large, prospective studies are needed to identify which factors should be routinely used

• These advances have generated a number of controversies and new unanswered questions.