CT indicates computed tomography; SI-NETs, small intestinal neuroendocrine tumors.
Survival curves are shown for groups before and after propensity score matching. The locoregional (LRS) group underwent prophylactic up-front surgery combined with oncologic treatment, and the delayed LRS group underwent delayed surgery as needed combined with oncologic treatment. HR indicates hazard ratio.
eTable 1. Baseline Characteristics for the LRS and Delayed LRS Groups Before Propensity Score Matching
eTable 2. Baseline Characteristics for the LRS and Delayed LRS Groups After Propensity Score Matching
eTable 3. Liver-Targeted and Systemic Antitumoral Treatments in the LRS and Delayed LRS Groups After Propensity Score Matching
eFigure. Histograms of Propensity Score Before and After Matching
Customize your JAMA Network experience by selecting one or more topics from the list below.
Daskalakis K, Karakatsanis A, Hessman O, et al. Association of a Prophylactic Surgical Approach to Stage IV Small Intestinal Neuroendocrine Tumors With Survival. JAMA Oncol. 2018;4(2):183–189. doi:10.1001/jamaoncol.2017.3326
What is the association of prophylactic locoregional surgery at diagnosis with overall survival among asymptomatic patients with small intestinal neuroendocrine tumors (SI-NETs) and distant metastases?
In this cohort study of 363 asymptomatic patients with stage IV SI-NETs, no difference was found in overall survival after propensity score matching between patients who underwent prophylactic surgery within 6 months of the diagnosis and patients who were treated nonsurgically or with later surgery.
In asymptomatic patients with SI-NETs, the value of prophylactic, up-front locoregional surgery in the presence of distant metastases is challenged.
Primary tumor resection and mesenteric lymph node dissection in asymptomatic patients with stage IV small intestinal neuroendocrine tumors (SI-NETs) are controversial.
To determine the association of locoregional surgery (LRS) performed at diagnosis with outcomes in patients with asymptomatic SI-NETs and distant metastases.
Design, Setting, and Participants
This cohort study included asymptomatic patients with stage IV SI-NETs diagnosed from January 1, 1985, through December 31, 2015, and identified using the prospective database of SI-NETs from Uppsala University Hospital, Uppsala, Sweden. Patients included were treated at a tertiary referral center and followed up until May 31, 2016, with data from the Swedish National Patient Register. The 363 patients with stage IV SI-NETs without abdominal symptoms were divided between those who underwent prophylactic up-front surgery within 6 months from diagnosis combined with oncologic treatment (hereafter referred to as LRS group [n = 161]) and those who underwent nonsurgical treatment or delayed surgery as needed combined with oncologic treatment (hereafer referred to as delayed LRS group [n = 202]).
Prophylactic up-front surgery within 6 months from diagnosis combined with oncologic treatment vs nonsurgical treatment or delayed surgery as needed combined with oncologic treatment.
Main Outcomes and Measures
Overall survival (OS), length of hospital stay (LOS), postoperative morbidity and mortality, and reoperation rates measured from baseline. Propensity score matching was performed between the 2 groups.
The 363 patients included 173 women (47.7%) and 190 men (52.3%), with a mean (SD) age at diagnosis of 62.4 (11.1) years. Two isonumerical groups with 91 patients in each resulted after propensity score matching. The LRS and delayed LRS groups were comparable in median OS (7.9 years [range, 5.1-10.7 years] vs 7.6 years [range, 5.8-9.5 years]; hazard ratio [HR], 0.98; 95% CI, 0.70-1.37; log-rank P = .93) and cancer-specific survival (7.7 years [range, 4.5-10.8 years] vs 7.6 years [range, 5.6-9.7 years]; HR, 0.99; 95% CI, 0.71-1.40; log-rank P = .99). No difference was found in 30-day mortality (0 patients in both matched groups) or postoperative morbidity (2 [2.2%] vs 1 [1.1%]; P > .99), median LOS (73 days [range, 2-270 days] vs 76 days [range, 0-339 days]; P = .64) or LOS due to local tumor-related symptoms (7.0 days [range, 0-90 days] vs 11.5 days [range, 0-69 days]; P = .81), or incisional hernia repairs (4 patients [4.4%] in both groups; P > .99). Patients in the LRS group underwent more reoperative procedures (13 [14.3%]) compared with those in the delayed LRS group (3 [3.3%]) owing to intestinal obstruction (P < .001).
Conclusions and Relevance
Prophylactic up-front LRS conferred no survival advantage in asymptomatic patients with stage IV SI-NETs. Delayed surgery as needed was comparable in all examined outcomes and was associated with fewer reoperations for intestinal obstruction. The value of a priori LRS in the presence of distant metastases is challenged and needs to be elucidated in a randomized clinical study.
Small intestinal neuroendocrine tumors (SI-NETs) have an indolent clinical course and are often diagnosed at a late stage.1 Sixty percent of patients with SI-NETs present with distant metastases, most commonly in the liver; thus, a high proportion of patients are diagnosed with stage IV tumors.1,2
In patients with stages I to III disease, resection of the primary tumor and regional lymphadenectomy is indicated as a curative procedure. In stage IV disease, surgery is generally not considered to be curative, although some liver-directed therapies are performed with a curative intent after or before radical locoregional surgery (LRS).3,4 Even in the era of a broad panel of novel, targeted, and systemic therapies for SI-NET, recurrence after liver-directed therapy is still common, and neither resection nor ablative techniques have been unequivocally found to prolong survival.5 Nevertheless, stage IV SI-NETs have a relatively good prognosis, with a 5-year survival rate of 57%.1
Many patients with stage IV disease present with clinical signs of hormone excess, such as flushing and diarrhea, and/or local tumor-related symptoms owing to primary tumor, mesenteric lymph node metastases, and associated fibrosis causing abdominal pain, obstruction, and/or impaired blood supply to the intestine.6 Patients with local tumor-related symptoms generally undergo LRS at the time of diagnosis. However, some symptomatic patients have lymph node metastases located at the root of the mesentery, encasing the superior mesenteric vessels and rendering radical resection challenging.7,8 Palliative methods, such as stenting of the superior mesenteric vein, have shown some success for symptomatic cases when LRS is impossible or associated with unacceptable morbidity.9
In patients with stage IV disease without local tumor-related symptoms, prophylactic LRS has also been advocated to avoid future intestinal obstruction, ischemia, perforation, or bleeding.10,11 However, no data are available for these patients comparing prophylactic up-front LRS with delayed LRS as needed in regard to survival, morbidity, symptoms, and quality of life.
The objective of this study was to assess the outcome of LRS in asymptomatic patients with stage IV SI-NETs compared with delayed LRS as needed. The primary end point was overall survival (OS); secondary end points included 30-day postoperative mortality and morbidity, length of hospital stay (LOS), and rates of reoperation and incisional hernia repair.
Patients were identified from the SI-NET prospective database of Uppsala University Hospital, Uppsala, Sweden, who were admitted from January 1, 1985, through December 31, 2015. Only patients with a histopathologically confirmed diagnosis of SI-NET and radiologically confirmed distant metastases but no abdominal symptoms were included. Patients with neuroendocrine carcinomas (NET grade 3) at baseline were not included in this study. Exclusion criteria included nonretrievable follow-up data, contraindication to surgery owing to severe comorbidity or poor general condition, technically inoperable central mesenteric lymph node metastases (ie, encasing the superior mesenteric vessels at the root of the mesentery [lymph node stage 4]), and no evidence of locoregional disease on computed tomography scan.7 The study was conducted according to the 1975 Declaration of Helsinki and approved by the regional ethics review board in Uppsala, Sweden. Written informed consent was obtained from all the study participants.
Follow-up data were obtained from the National Patient Register, which covers approximately 99% of all health care in Sweden, and included LOS from baseline due to SI-NET disease and local tumor-related symptoms later in the disease course.12 Reoperation owing to bowel obstruction and incisional hernia repairs was also recorded. Patients were followed up until death or May 31, 2016.
Patients were selected for LRS or nonsurgical treatment after review by the multidisciplinary tumor board. The choice of treatment has to some extent been dictated by clinical expertise; therefore, despite similar clinical patient presentation, LRS had been offered to some patients, whereas nonsurgical treatment had been offered to others.
After data review, patients were divided between those who underwent prophylactic up-front surgery within 6 months of diagnosis combined with oncologic treatment (LRS group) and those who underwent no surgery or delayed surgery as needed combined with oncologic treatment (delayed LRS group). Patients in both groups received multimodality oncologic treatment, including somatostatin analogues (SSAs), interferon-alfa, liver-directed treatment, and/or peptide receptor radionuclide therapy with lutetium Lu 177–labeled DOTA-Tyr3-octreotate (PRRT) as indicated. To avoid immortal time bias, baseline for the LRS group was defined as the first date on which patients underwent LRS, whereas baseline for the delayed LRS group was the date of diagnosis.
The following variables were recorded at baseline: age, sex, calendar year at baseline, Charlson comorbidity index, carcinoid symptoms, carcinoid heart disease, liver tumor load, extrahepatic metastases, lymph node metastases on imaging, Ki-67 proliferation index, and urinary 5-hydroxyindoleacetic acid (5-HIAA) level. The Charlson comorbidity index score is a validated predictive factor for survival, ranging from 0 to 31 points (higher scores indicate more comorbidities), and was estimated for each patient in this study as described by Charlson et al.13
Computed tomography and/or abdominal ultrasonography images obtained at baseline were reviewed, and the highest liver tumor load was recorded. The following staging system was used to describe the stage of liver involvement: stage 1, fewer than 5 metastases confined in 1 lobe; stage 2, bilobar and/or 5 to 10 metastases; and stage 3, more than 10 metastases or diffuse metastatic disease. Octreoscan and gallium 68–positron emission tomography findings on liver tumor load were not used in this study because few patients had undergone these examinations at baseline.
Tumor grade was determined from primary and lymph node specimens and/or results of liver biopsies according to Ki67 proliferation index as described in European Neuroendocrine Tumor Society guidelines, which have remained unchanged during the past decade, with grade 1 indicating less than 3%; grade 2, 3% to 20%; and grade 3, higher than 20%.11,14,15 For urinary 5-HIAA levels, the reference value of 9.56 mg/24 hours (to convert to micromoles per day, multiply by 5.23) was used, and patients were arbitrarily divided into the following 5 groups for subsequent propensity score match: group 1, less than 9.56 mg/24 hours; group 2, 9.75 to 38.24 mg/24 hours; group 3, 38.43 to 114.72 mg/24 hours; group 4, greater than 114.72 mg/24 hours; and group 5, missing values. Liver surgery, radiofrequency ablation, hepatic artery embolization and radioembolization of liver metastases, and antitumoral treatment with PRRT, SSAs, and/or other antitumoral agents were noted at baseline and during follow-up. Morbidity was defined by Clavien-Dindo grade 3 and above.16 To ensure the quality of data reporting, the STROBE statement was followed.17
Variables are presented as medians with ranges or means with SDs as appropriate. Differences between groups were assessed using the Mann-Whitney test, χ2 test, and 2-sided t test for unmatched data or the Wilcoxon signed rank test and McNemar test for matched data as appropriate. Survival analyses were performed using the Kaplan-Meier method, and crude analysis of OS was computed using a log-rank or a matched log-rank test as appropriate. Hazard ratios (HRs) were calculated using a stratified Cox proportional hazards regression model. All tests were 2 sided unless stated otherwise. P < .05 was considered to be significant for all tests.
Power analysis was performed for the definition of sample size, and propensity score matching was performed to reduce bias between groups.18 The sample size calculation was based on the primary outcome (OS between matched groups), and a match ratio of 1:1 was used. The relative HR used in the sample size calculation between the matched groups was chosen to be greater than 2 (or <0.50). The probability of failure (death) in the cohort was projected to be 55% during follow-up based on previous data from this cohort.1 Therefore, calculations based on nonstratified Cox proportional hazards regression for 2 groups with equal sizes revealed that 120 patients were needed to achieve 65 events and thus retain 80% power with an α level of .05. Because stratified Cox proportional hazards regression was used, an additional 10% was projected in response to anticipated loss of power. Therefore, the required number of patients in the matched groups was increased to 132.
A 1:1 nearest-neighbor propensity score match with a caliper width of 0.1 was performed between the LRS and delayed LRS groups using the above-mentioned variables at baseline. Standardized differences were used to examine differences in baseline characteristics before and after matching, with a standardized difference of less than 10 considered as an adequate balance between groups.18
Of a total of 820 patients with SI-NETs, 363 filled the inclusion criteria (173 women [47.7%] and 190 men [52.3%]; mean [SD] age at diagnosis, 62.4 [11.1] years). Median overall follow-up was 5.4 years (range, 0-31 years). The study flow before propensity score matching is given in Figure 1.
Baseline variables of the unmatched LRS (n = 161) and delayed LRS (n = 202) groups are presented in eTable 1 in the Supplement. In the unmatched delayed LRS group, the patients were older with more advanced disease in terms of liver tumor load, extrahepatic metastases, and carcinoid heart disease and had higher 5-HIAA levels compared with the unmatched LRS group (eTable 1 in the Supplement).
In the unmatched delayed LRS group, a total of 89 patients underwent delayed LRS and 113 never underwent LRS. Thirty-two patients in the unmatched delayed LRS group underwent LRS for abdominal pain, obstruction, and/or acute ischemia, whereas 57 underwent delayed elective LRS with no obvious locoregional symptom or indication stated in the patient medical record.
For the 89 patients in the unmatched delayed LRS group who underwent surgery, 30-day postoperative morbidity was 1.0% (2 patients) and mortality was 0.5% (1 patient) compared with the unmatched LRS group, for whom 30-day morbidity was 4.3% (7 patients) with no mortality. The median OS in the unmatched LRS group (9.5 years; range, 7.5-11.6 years) was longer than that in the unmatched delayed LRS group (5.3 years; range, 4.1-6.6; HR, 1.38; 95% CI, 1.08-1.76; log-rank P = .01) (Figure 2).
Propensity score matching resulted in 2 isonumerical groups with similar baseline variables (Table and eTable 2 in the Supplement) and propensity score distributions (eFigure in the Supplement). In the matched delayed LRS group, 53 of 91 patients underwent LRS after more than 6 months, whereas 38 never underwent LRS. The median time until surgery for these 53 patients was 18 months (range, 6-168 months), and the indication for surgery was symptomatic disease in 20, including abdominal pain in 14, acute bowel obstruction in 5, or perforation in 1. In the remaining 33 patients undergoing delayed LRS, surgery was elective and locoregional symptoms were not stated in the patient medical records. The matched LRS and delayed LRS groups were comparable with regard to median OS (7.9 [range, 5.1-10.7 years] vs 7.6 years [range, 5.8-9.5 years]; HR, 0.98; 95% CI, 0.70-1.37; log-rank P = .93) (Figure 2) and cancer-specific survival (7.7 years [range, 4.5-10.8 years] vs 7.6 years [range, 5.6-9.7 years]; HR, 0.99; 95% CI, 0.71-1.40; log-rank P = .99). Similarly, we found no difference in postoperative morbidity in the matched groups at 30 days (LRS group, 2 [2.2%]; delayed LRS group, 1 [1.1%]; P > .99). No 30-day mortality was seen in either matched group.
No difference was found in median LOS owing to SI-NET disease between the matched LRS (73 days; range, 2-270 days) and the delayed LRS (76 days; range, 0-339 days; P = .64) groups. In addition, we found no difference between groups in median LOS specifically owing to local tumor-related symptoms later in the disease course (LRS group, 7.0 days [range, 0-90 days]; delayed LRS group, 11.5 days [range, 0-69 days]; P = .81). The number of reoperations owing to bowel obstruction (adhesiolysis, second resection, and/or intestinal bypass) was higher in the LRS group (13 [14.3%] vs 3 [3.3%]; P < .001), whereas the number of incisional hernia repairs was the same (4 [4.4%] in both matched groups; P > .99).
Liver surgery at baseline and during follow-up was more frequently applied in the matched LRS group compared with the matched delayed LRS group (22 [24.2%] vs 10 [11.0%]; P < .001) (eTable 3 in the Supplement). No difference was seen between the matched groups regarding the use of radiofrequency ablation of liver metastases, hepatic artery embolization, PRRT, SSA, interferon-alfa, or other antitumoral agents (eTable 3 in the Supplement).
This study assessed outcomes of asymptomatic patients with SI-NETs and distant metastases after prophylactic up-front LRS. No survival benefit was found after propensity score matching for these patients compared with matched control individuals with initial nonsurgical management. In addition, the 30-day postoperative morbidity and mortality in the LRS group was not different from that in the delayed LRS group. We found no difference in LOS due to SI-NET disease between the 2 matched groups even when locoregional symptoms were specifically addressed. A higher rate of reoperation owing to bowel obstruction was associated with prophylactic up-front LRS presumably because of the development of postoperative adhesions along with a lack of locoregional macroradicality and accompanying fibrotic manifestations in the mesentery of patients with stage IV disease.
Despite prophylactic LRS in asymptomatic patients with stage IV SI-NETs having been considered standard practice, it has not been evaluated in any randomized clinical trials, and the survival rates after LRS reported in retrospective cohort studies have likely been influenced by selection and immortal time bias.1,19-23 This possibility is supported by the present study, in which the unmatched groups varied in baseline variables (eTable 1 in the Supplement) and the use of propensity score matching eliminated these confounding factors. The results challenge current European Neuroendocrine Tumor Society guidelines,11,15 which suggest a possible benefit of prophylactic LRS in stage IV SI-NET; however, these guidelines are based on results from retrospective studies without attempts to correct for confounding factors. The benefit from resection of asymptomatic primary tumors in patients with advanced and nonresectable metastatic disease has been questioned in the literature24 and the National Comprehensive Cancer Network recommendations.25
Although surgery may alleviate carcinoid syndrome symptoms, flushing and diarrhea may also be efficiently controlled by antitumoral agents. Treatment with SSAs has been shown to control time to progression and stabilize the growth of metastases.26,27 These medical options provide an avenue to control symptoms while avoiding surgery and its associated risks. In addition, PRRT has been added to the therapeutic armamentarium for stage IV SI-NETs and may also alter the role of surgery.28
Limitations of the present study included the lack of mesenteric lymph node staging and assessment of macroradicality in the mesentery in patients who underwent LRS. However, most operations were performed at Uppsala University Hospital, where a standardized mesenteric dissection was undertaken as described by Ohrvall et al.7 Although the propensity score method may exacerbate imbalance in unmeasured covariates, the propensity score matching in this study controlled extensively for variables that were known or presumed confounders. However, critical variables may have been overlooked and thus instead increased hidden heterogeneity between groups. For example, pathologic T tumor categories, extent of mesenteric lymph node metastases, and peritoneal carcinomatosis were not included in the propensity score match because not all patients underwent surgery, and these variables could not be assessed in the same manner. Lymph node staging was impossible owing to lack of data, and peritoneal carcinomatosis was only assessable in the group undergoing LRS at baseline. Although antitumoral treatments were not controlled in the propensity score model, no differences were found in the use of SSAs, interferon-alfa, PRRT, or liver embolization between matched patients in our study. Only 3 matched patients received a mammalian target of rapamycin inhibitor, and only 1 matched patient was treated with a tyrosine kinase inhibitor; thus, it is implausible that these recently approved drugs influenced study outcome. The LRS group was more prone to undergo liver surgery, which may have influenced the results, although a previous propensity score–matched study5 showed that the presence or absence of liver surgery did not influence survival in locoregionally resected SI-NET.
Moreover, incomplete matching occurred in approximately 50% of the total cohort, which was accepted because including more patients in the matching would have led to a severe remaining imbalance in baseline covariates. Incomplete matching may limit the generalizability of the results; therefore, the results of this study may not apply to patients who were not matched owing to extreme propensity scores (eg, very young patients, mostly found in the LRS group, or patients with high comorbidity, who were overrepresented in the delayed LRS group). Furthermore, although our a priori calculated sample size was met, the study may have been underpowered to detect a smaller but still clinically significant difference in OS. However, the lack of difference in the survival estimates between the groups may indicate that the prophylactic surgical and delayed surgical as-needed approaches are similar with regard to OS.
In the delayed LRS group, 53 of the 91 patients eventually underwent surgery, and 20 of these patients would have benefited from prophylactic surgery because they developed symptoms over time. However, 33 patients may have had prophylactic surgery without symptoms. We can only speculate as to the reasons for this approach, including differences in clinical expert opinion, patient preference, and delay by physicians. Of importance, all patients included in this study received multimodality treatment, including SSAs, interferon-alfa, liver-directed treatment, and/or PRRT as indicated. Allowing for a 6-month interval between diagnosis and surgery may establish tumor biology and possible treatment effects before invasive therapy. For example, patients with growing metastases in the delayed LRS group may have been selected for surgery during follow-up. Finally, the patient population only included referrals to our tertiary center, which may include a referral bias.
Locoregional surgery retains its value in the treatment of patients with SI-NETs when radical resection is feasible or when symptomatic disease is present regardless of disease stage. However, present results challenge the traditional view for extensive LRS in patients with distant metastases in the absence of local tumor-related symptoms. No benefit in survival could be demonstrated, and patients had more reoperations for intestinal obstruction. On the contrary, a more conservative approach with delayed LRS as clinically indicated yielded comparable survival and did not confer risks in terms of increased morbidity and mortality or prolonged hospital stay.
Therefore, the approach of delayed LRS as needed seems reasonable for this subgroup of patients with SI-NETs and may complete the armamentarium of systemic and liver-directed treatments as indicated per patient. In the era of personalized treatment, maximalist surgery should be reconsidered and replaced by a comprehensive multidisciplinary approach for the optimal treatment of patients with SI-NETs. Finally, despite the sophisticated methods followed and the design of this study, a prospective randomized clinical trial is needed to further elucidate the value of prophylactic LRS and strengthen current recommendations. Therefore, a composite end point including reoperation owing to growing metastases or symptoms may be more feasible than OS as an end point in a future randomized clinical trial.
Corresponding Author: Kosmas Daskalakis, MD, Department of Surgical Sciences, Uppsala University, Akademiska sjukhuset Ingång 70 Trappa 1, 751 85 Uppsala, Sweden (firstname.lastname@example.org).
Accepted for Publication: August 8, 2017.
Published Online: October 19, 2017. doi:10.1001/jamaoncol.2017.3326
Author Contributions: Drs Norlén and Stålberg contributed equally to this study. Drs Daskalakis and Stålberg had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Daskalakis, Karakatsanis, Öberg, Hellman, Norlén, Stålberg.
Study concept and design: Daskalakis, Karakatsanis, Öberg, Hellman, Norlén, Stålberg.
Acquisition, analysis, or interpretation of data: Daskalakis, Karakatsanis, Hessman, Stuart, Welin, Tiensuu Janson, Hellman, Norlén, Stålberg.
Drafting of the manuscript: Daskalakis, Stuart, Hellman, Norlén, Stålberg.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Daskalakis, Hellman, Norlén.
Obtained funding: Daskalakis, Hellman, Norlén, Stålberg.
Administrative, technical, or material support: Karakatsanis, Hessman, Welin, Tiensuu Janson, Öberg, Hellman, Stålberg.
Study supervision: Norlén, Stålberg.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported by the Lions Cancer Foundation, the Göran Gustafsson Foundation for Research in Natural Sciences and Medicine, the Bengt Ihre Research Fellowship, Lennander and Selanders fund, and the Swedish Cancer Society.
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: