Articles

Surgical versus non-surgical management for patients with

malignant bowel obstruction (S1316): a pragmatic
comparative effectiveness trial
Robert S Krouse, Garnet L Anderson, Kathryn B Arnold, Cynthia A Thomson, Valentine N Nfonsam, Mazin F Al-Kasspooles, Joan L Walker, Virginia Sun,
Angeles Alvarez Secord, Ernest S Han, Alberto M Leon-Takahashi, David Isla-Ortiz, Phillip Rodgers, Samantha Hendren, Marco Sanchez Salcedo,
Jonathan A Laryea, Whitney S Graybill, Devin C Flaherty, Harveshp Mogal, Thomas J Miner, Jose M Pimiento, Mio Kitano, Brian Badgwell, Giles Whalen,
Jeffrey P Lamont, Oscar A Guevara, Maheswari S Senthil, Summer B Dewdney, Eric Silberfein, Jason D Wright, Bret Friday, Bridget Fahy,
Sandeep Anantha Sathyanarayana, Mark O’Rourke, Marie Bakitas, Jeff Sloan, Marcia Grant, Gary B Deutsch, Jeremiah L Deneve

Summary
Lancet Gastroenterol Hepatol Background Malignant small bowel obstruction has a poor prognosis and is associated with multiple related
2023; 8: 908–18 symptoms. The optimal treatment approach is often unclear. We aimed to compare surgical versus non-surgical
Published Online management with the aim to determine the optimal approach for managing malignant bowel obstruction.
August 1, 2023
https://doi.org/10.1016/
S2468-1253(23)00191-7 Methods S1316 was a pragmatic comparative effectiveness trial done within the National Cancer Trials Network at
See Comment page 863
30 hospital and cancer research centres in the USA, Mexico, Peru, and Colombia. Participants had an intra-abdominal
For the Spanish translation of
or retroperitoneal primary cancer confirmed via pathological report and malignant bowel disease; were aged 18 years
the abstract see Online for or older with a Zubrod performance status 0–2 within 1 week before admission; had a surgical indication; and
appendix 1 treatment equipoise. Participants were randomly assigned (1:1) to surgical or non-surgical treatment using a dynamic
Department of Surgery, balancing algorithm, balancing on primary tumour type. Patients who declined consent for random assignment were
University of Pennsylvania, offered a prospective observational patient choice pathway. The primary outcome was the number of days alive and
Philadelphia, PA, USA
(Prof R S Krouse MD); Corporal
out of the hospital (good days) at 91 days. Analyses were based on intention-to-treat linear, logistic, and Cox regression
Michael J Crescenz Veterans models combining data from both pathways and adjusting for potential confounders. Treatment complications were
Affairs Medical Center, assessed in all analysed patients in the study. This completed study is registered with ClinicalTrials.gov, NCT02270450.
Philadelphia, PA, USA
(Prof R S Krouse); SWOG
Statistics and Data
Findings From May 11, 2015, to April 27, 2020, 221 patients were enrolled (143 [65%] were female and 78 [35%] were
Management Center, male). There were 199 evaluable participants: 49 in the randomised pathway (24 surgery and 25 non-surgery) and
Fred Hutchinson Cancer Center, 150 in the patient choice pathway (58 surgery and 92 non-surgery). No difference was seen between surgery and non-
Seattle, WA, USA surgery for the primary outcome of good days: mean 42·6 days (SD 32·2) in the randomised surgery group, 43·9 days
(Prof G L Anderson PhD,
K B Arnold MS); Department of
(29·5) in the randomised non-surgery group, 54·8 days (27·0) in the patient choice surgery group, and 52·7 days
Health Promotion Sciences, (30·7) in the patient choice non-surgery group (adjusted mean difference 2·9 additional good days in surgical versus
Mel and Enid Zuckerman non-surgical treatment [95% CI –5·5 to 11·3]; p=0·50). During their initial hospital stay, six participants died, five due
College of Public Health, to cancer progression (four patients from the randomised pathway, two in each treatment group, and one from the
University of Arizona,
Tucson, AZ, USA
patient choice pathway, in the surgery group) and one due to malignant bowel obstruction treatment complications
(Prof C A Thomson PhD); (patient choice pathway, non-surgery). The most common grade 3–4 malignant bowel obstruction treatment
Department of Surgery, complication was anaemia (three [6%] patients in the randomised pathway, all in the surgical group, and five [3%]
University of Arizona, Tucson,
patients in the patient choice pathway, four in the surgical group and one in the non-surgical group).
AZ, USA (Prof V N Nfonsam MD);
Department of Surgery,
Louisiana State University, Interpretation In our study, whether patients received a surgical or non-surgical treatment approach did not influence
New Orleans, LA, USA good days during the first 91 days after registration. These findings should inform treatment decisions for patients
(Prof V N Nfonsam);
hospitalised with malignant bowel obstruction.
Department of Surgery,
University of Kansas Cancer
Center, Kansas City, KS, USA Funding Agency for Healthcare Research and Quality and the National Cancer Institute.
(M F Al-Kasspooles MD);
Stephenson Cancer Center,
Copyright © 2023 Elsevier Ltd. All rights reserved.
University of Oklahoma,
Oklahoma City, OK, USA
(Prof J L Walker MD); Division of Introduction treatment-related complications due to radiotherapy or
Nursing Research and Palliation in the setting of advanced cancer is an essential surgery. Patients typically present with symptoms
Education, City of Hope
component of care. One common clinical diagnosis including nausea, vomiting, bloating, and pain. Survival
National Medical Center,
Duarte, CA, USA influencing health-related quality of life (HRQOL) and for patients presenting with a malignant bowel
(Prof V Sun PhD, survival in patients with advanced cancer is malignant obstruction is often poor. A Cochrane review has shown
Prof M Grant PhD); Department bowel obstruction.1 This is most commonly a small that those undergoing surgery have a median survival
of Obstetrics and Gynecology,
Division of Gynecologic
bowel obstruction,2 and might be directly due to cancer, range of 2 to 8·4 months, and those treated non-surgically
adhesions in a patient with advanced cancer, or other of 28 to 69 days.3

908 www.thelancet.com/gastrohep Vol 8 October 2023

 Articles

Oncology, Duke Cancer

Research in context Institute, Durham, NC, USA
(A Alvarez Secord MD); Division
Evidence before this study in patients with malignant bowel obstruction. In addition, there of Gynecologic Oncology,
We did an extensive PubMed literature search from Jan 1, 1983, are no other prospectively collected quality of life, diet, and Department of Surgery, City
to July 14, 2023, using terms such as “malignant small bowl clinical outcome data in this population of cancer survivors. of Hope National Medical
Center, Duarte, CA, USA
obstruction”, “palliative care”, and “cancer survivors”. There Our study showed no difference in surgery versus non-surgery (E S Han MD); Department of
were no language restrictions. To our knowledge, there has for days in hospital and alive or for overall survival, although it Surgical Oncology, National
never been a prospective trial comparing surgery versus non- suggested that there is a benefit in quality of life outcomes for Cancer Institute, Tlalpan,
surgery in the setting of malignant bowel obstruction with the those who receive surgery. Mexico City, Mexico
(A M Leon-Takahashi MD,
same rigorous inclusion criteria and a randomised component.
Implications of all the available evidence D Isla-Ortiz MD); Department
There have been many retrospective reports attempting to of Family Medicine, University
Evidence from this study could help in making difficult decisions
examine surgical or non-surgical approaches, and a few of Michigan, Ann Arbor, MI,
regarding treatment for this population of patients. We believe USA (Prof P Rodgers MD,
attempts to retrospectively compare surgical versus
that surgically eligible patients, often deemed non-operative, S Hendren MD); Department
non-surgical approaches. These studies suffer from bias
should be offered an operation earlier in their hospital stay to of Surgery, Instituto Nacional
selection and patient populations that differ between surgical de Enfermedades Neoplásicas,
improve gastrointestinal symptoms, although not to improve
and non-surgical groups (eg, patients in non-surgery groups Surquillo, Peru
survival or to increase the number of days alive and out of the (M Sanchez Salcedo MD);
were deemed non-operative).
hospital in the first few months after admittance. This could Department of Surgery,
Added value of this study be practice-changing for many surgeons. In addition, the University of Arkansas for
Medical Sciences, Little Rock,
To our knowledge, this is the first prospective trial using a completion of this trial should facilitate future study in patients
AR, USA (J A Laryea MD);
randomised component to compare surgery versus non-surgery with complex advanced cancer. Department of Gynecologic
Oncology, Medical University
of South Carolina, Charleston,
SC, USA (W S Graybill MD);
The main treatment decision for clinicians and patients approaches. A difference in either direction was possible Department of Surgical
is operative or non-operative care. Even when a patient is and, if large enough, would be of considerable value to Oncology, Valley Health,
Winchester, VA, USA
surgically eligible, there is often uncertainty as to the patients and providers.
(D C Flaherty DO); Department
benefit of surgery. The inherent risk–benefit ratio of an of Surgery, Medical College of
operation must be considered, but this is more difficult to Methods Wisconsin, Milwaukee, WI, USA
assess when the benefits are unclear. Complications can Study design and participants (H Mogal MD); University of
Washington, Seattle, WA, USA
occur even with the most fastidious operation, especially S1316 was a pragmatic comparative effectiveness trial
(H Mogal); Department of
if the patient is debilitated secondary to cancer, previous done within the National Cancer Trials Network (NCTN) Surgery, Rhode Island Hospital,
cancer treatments, malnutrition, or other medical with SWOG Cancer Research Network (hereafter referred Providence, RI, USA
conditions. Surgical morbidity might lead to longer to as SWOG) as the lead group. Recognising that accrual (T J Miner MD); Department of
Gastrointestinal Oncology,
hospital stays, further reducing quality of life. Importantly, would be difficult and multiple sites would be needed, the H Lee Moffitt Cancer Center,
an operation might hasten death, which is the ultimate study includes 27 hospitals and cancer research centres in Tampa, FL, USA
poor outcome. Most surgical consultations for malignant the USA, and one centre each in Mexico, Peru, and (J M Pimiento MD); Mays Cancer
bowel obstructions result in non-surgical treatments.2,4 Colombia. To compare outcomes of surgical versus non- Center, University of Texas
Health San Antonio,
However, when evaluating the optimal treatment approach surgical management of malignant bowel obstruction, we San Antonio, TX, USA
for malignant bowel obstruction in the literature, it is used a hybrid design involving a randomised pathway and (M Kitano MD); Department of
often unclear if patients deemed not eligible for surgery a parallel patient choice pathway comparing the same two Surgical Oncology, MD
were ever evaluated by a surgical team. treatment groups (surgery or non-surgery). Eligibility Anderson Cancer Center,
Houston, TX, USA
Although multiple outcomes are of great importance to and data collection were uniform across these pathways, (B Badgwell MD); Department
those facing the end of life, being out of the acute care except for consent to randomisation. This study was under of Surgical Oncology, Umass
(hospital) setting is among the most important.5,6 For the oversight of a data and safety committee (SWOG data Memorial Medical Center,
patients with malignant bowel obstruction, other goals of and safety monitoring committee). The protocol was Worcester, MA, USA
(G Whalen MD); Department of
care include the ability to eat and relief of constitutional approved by institutional review boards at each site, and Surgery, Baylor University
symptoms such as nausea, vomiting, bloating, and pain. all partici­pants gave written informed consent. Consent Medical Center, Dallas, TX, USA
Outcomes for patients with malignant bowel obstruction forms, questionnaires, and interviews were made available (J P Lamont MD); Division of
are variable,3 and those related to symptom burden are in Spanish to assist in recruitment of and data collection Gastrointestinal Surgery,
Instituto Nacional de
often not clearly defined and documented. Patient reported from Spanish-only speakers. The full trial protocol can be Cancerologia, Bogota,
outcomes can identify severe symptoms that impact global accessed on the Cancer Trials Support Unit website. Colombia (O A Guevara MD);
HRQOL issues in the setting of palliative surgery.7 Eligible participants had an intra-abdominal or Department of Surgery,
Loma Linda University Health,
This pragmatic study aimed to compare two standard retroperitoneal primary cancer confirmed via pathological
Loma Linda, CA, USA
approaches—surgical and non-surgical treatment—to report and malignant bowel disease defined as adapted (M S Senthil MD); University of
determine the optimal approach for managing malignant from a previous report8 by clinical and radiological California-Irvine, Orange, CA,
bowel obstruction. We launched this as a truly two-sided criteria as a small bowel obstruction below the ligament USA (M S Senthil); Department
of Obstetrics and Gynecology
hypothesis test comparing two standard practice of Treitz, and an intra-abdominal, retroperitoneal, or

www.thelancet.com/gastrohep Vol 8 October 2023 909

 Articles
Oncology, Rush University
Medical Center, Chicago, IL,
USA (S B Dewdney MD);
Department of Surgery, Baylor
College of Medicine, Houston,
TX, USA (Prof E Silberfein MD);
Division of Gynecologic
Oncology, Columbia University
Medical Center, New York, NY,
USA (J D Wright MD);
Department of Hematology/
Oncology Essentia Health
Cancer Center, Duluth, MN,
USA (B Friday MD); Department
of Surgery, University of
New Mexico, Albuquerque,
NM, USA (Prof B Fahy MD);
Northwell Health
Cancer Institute,
Lake Success, NY, USA
(S Anantha Sathyanarayana MD,
G B Deutsch MD); Center for
Integrative Oncology and
Survivorship, Greenville Health
System, Clemson, SC, USA
(M O’Rourke MD); School of
Nursing, University of Alabama
at Birmingham, Birmingham,
AL, USA (M Bakitas DNSc);
Mayo Clinic Rochester,
Rochester, MN, USA
(J Sloan PhD*); Department of
Surgery, University of
Tennessee Health Science
Center, Memphis, TN, USA
(J L Deneve DO)
*Dr Sloan died in May, 2022
Correspondence to:
Prof Robert S Krouse,
Department of Surgery,
University of Pennsylvania,
PA 19104, USA
robert.krouse@pennmedicine.
upenn.edu
For the Cancer Trials Support
Unit see https://ctsu.org
910
gastro-intestinal tumour with incurable disease. All
radiological reports were reviewed to document
obstruction; disease status was recorded when possible.
Eligible participants were aged 18 years or older with a
Zubrod performance status 0–2 within 1 week before
admission. No specified previous treatments were an
accrual criterion. Before registration, the surgical team
confirmed that there was a surgical indication, the
patient could tolerate an operation, and there was
equipoise (no evidence to prefer one treatment approach
over the other). Patients with signs of potentially needing
an emergency procedure with signs of perforation or
peritonitis were excluded. A radiological complete
obstruction was not an exclusion criterion because
physical examination took precedence. Patients either
agreed to be randomly assigned or, after declining
random assignment, were offered registration in the
patient choice pathway.
Randomisation and masking
Participants were registered by site staff within 3 days of
eligibility for the trial. For randomly assigned patients,
the unmasked treatment assignment (surgery or non-
surgery) was computer-generated at registration in the
SWOG database using a dynamic balancing algorithm9 to
randomly assign patients (1:1) to either surgery or non-
surgery, balancing on primary tumour type (colorectal,
ovarian, or other). The randomisation sequence was
concealed until assignment. In the patient choice
pathway, the treatment group selected by the patient and
their physician at the time of consent was considered the
assigned treatment.
Procedures
Treatment was initiated within 48 h of registration. A
somatostatin analogue was recommended for the non-
surgical groups. Somatostatin analogue use was initially a
mandatory treatment in the non-surgical groups, but
supply issues meant it was not possible to continue the
study with this treatment requirement. Because this was a
pragmatic trial, crossover was allowed as circumstances
changed but the initial study group assignment or choice
was retained regardless of actual treatment received for
analytical purposes. The operation performed in the
surgical groups was at the discretion of the surgical team
based on findings at the time of surgery.
Once discharged, hospital admission status was
assessed weekly via telephone up to 91 days, and then
monthly for up to 1 year. Participants were removed from
the study when they completed the prespecified 53 weeks
of follow-up, developed a medical condition that precluded
them from study participation, or they refused follow-up.
All demographic information, including patient sex,
were collected by chart review at each accrual site.
Patient-reported outcomes and dietary recalls were
reported by the patients; all other data, including sex, were
reported by the site.
The trial continued until completion of accrual targets
and conclusion of the planned primary outcome
completion (follow-up day 91) and up to 1 year to assess
longer term effects.
Outcomes
The primary outcome, selected on the basis of an expert
panel recommendation, was the number of good days,
defined as the number of days alive and out of the hospital
from registration (day 0) through to day 91. Additional
prespecified outcomes measured during the initial
hospital stay included days with a nasogastric tube,
medications taken including somatostatin analogue and
steroid use, complications based on the National Cancer
Institute Common Terminology Criteria for Adverse
Events criteria, and length of hospital stay. Other pre­
specified measures collected at baseline in the hospital
and at weeks 5, 9, and 13, included the ability to consume
food (based on self-reported or caregiver-reported 24-h
dietary recalls), inpatient treatment-related complications
(recorded at the end of every hospital admission), and
overall survival. Dietary recalls were collected by trained
assessors at the University of Arizona Cancer Center
Behavioral Measurement and Interventions Shared
Resource via telephone using standardised protocols and
a nutrient database for research.10 Nausea, vomiting, pain,
bloating, and constipation were assessed using the MD
Anderson Symptom Inventory for Gastrointestinal Cancer
(MDASI-GI),11,12 administered at baseline and during
follow-up calls (weekly through to week 13, then every
4 weeks through to week 53). The MDASI-GI measures
symptom severity on an 11-point scale, from 0 (not
present) to 10 (as bad as you can imagine). A mean
difference of at least 1·2 points was considered a clinically
meaningful difference between groups for MDASI-GI
outcomes.13,14 Overall survival was documented using
medical records reports. Prespecified exploratory out­
comes were ascites volume, albumin levels, and presence
of carcinomatosis. Future prespecified exploratory analyses
will examine diet, intravenous hydration, 1-year survival,
and quality of life (EQ-5D).
Statistical analysis
This was a truly two-sided hypothesis test comparing
two standard practice approaches (surgery vs non-
surgical treatments) for malignant bowel obstruction. A
sample size of 180, including 50 randomly assigned
participants and 130 enrolled in the patient choice
pathway, balanced by treatment selection was needed
to provide 90% power to detect a mean difference of
14 good days (assuming a standard deviation of 29 days)
between surgical and non-surgical treatment modalities.
The analysis plan pre­ specified that data from both
pathways (randomisation and patient choice) would be
combined in a regression-based analysis. To address
anticipated loss of statistical power from probable
imbalance in treatment selection in the patient choice
www.thelancet.com/gastrohep Vol 8 October 2023
 Articles

pathway, missing data, and ineligible patients, the final
target sample size was set at 220.
Patients were not assessable if they were deemed
ineligible or withdrew consent for follow-up before the
end of their initial hospital stay on study. Only assessable
patients were included in the analyses.
Baseline patient characteristics were compared
between the randomised and patient choice pathway and
between the treatment groups of the patient choice
pathway using t-tests for continuous variables and χ²
tests for categorical data.
The primary outcome was analysed using a linear
regression model applied to the combined data from the
randomised and patient choice pathways, using the
assigned treatment indicator variable (surgical vs non-
surgical management) regardless of treatment received,
consistent with the intention-to-treat principle. The
statistical analysis plan indicated that transformations of
the outcome variable could be made to improve adherence
to normality assumptions.
To establish potential confounders of treatment selection
in the patient choice pathway, two supportive analyses
were conducted. First a stepwise logistic regression model
was used to assess which baseline factors were associated
with selected treatment in the patient choice pathway. A
separate stepwise model evaluated which of these same
baseline factors were associated with good days in the
patient choice pathway, irrespective of treatment selection.
Variables found to be statistically significant (p<0·10) in
either of these models were considered as potential
confounders in the primary analysis.
In addition to potential confounders, the model
included indicator variables for treatment group
(1=surgical; 0=non-surgical) and study pathway
(1=patient choice; 0=randomised) and an interaction
term between these two variables, the latter providing
an estimate of the residual bias in the patient choice
treatment effect estimate. Potential confounders, as
identified in the stepwise models, were retained in the
final model if their inclusion helped to explain the
difference between pathways as assessed by any
noteworthy reduction in the estimated bias (interaction
term). The analysis plan indicated that inference
regarding the treatment effect would be based on the

221 patients enrolled

56 randomly assigned 165 patient choice

29 allocated to surgery 27 allocated to non-surgical management 65 chose surgery 100 chose non-surgical management
25 received surgery 20 received non-surgical management 63 received surgery 88 received non-surgical management
4 did not receive surgery 7 did not receive non-surgical 2 did not receive surgery 12 did not receive non-surgical
2 refused management 1 refused management
2 symptoms improved before 6 symptoms failed to resolve with 1 symptoms improved before 12 symptoms failed to resolve with
scheduled surgery non-surgical management so scheduled surgery non-surgical management so
received surgery received surgery
1 site error*

29 followed up at 91 days 27 followed up at 91 days 65 followed up at 91 days 100 followed up at 91 days

11 alive 8 alive 39 alive 46 alive
2 refusal 1 refusal 2 refusal 9 refusal
16 deceased 18 deceased 24 deceased 45 deceased

5 ineligible 2 ineligible 7 ineligible 8 ineligible

4 no small bowel 2 no small bowel 3 no small bowel 3 no small bowel
obstruction obstruction obstruction obstruction
1 no confirmed bowel 1 no confirmed bowel 1 no confirmed bowel
obstruction obstruction obstruction
1 treated before 2 treated before
registration registration
1 no intra-abdominal 2 withdrew consent
cancer before end of first
1 withdrew consent hospital stay
before end of first
hospital stay

24 included in analysis 25 included in analysis 58 included in analysis 92 included in analysis

Figure 1: Trial profile

*This patient consented to the surgery group in the patient choice pathway and was registered to the randomised pathway in error.

www.thelancet.com/gastrohep Vol 8 October 2023 911

 Articles

regression model without the interaction term if there

was no evidence of effect modification by study pathway.
Randomised pathway Patient choice pathway
If the interaction term was significant, the main effect
Surgery Non-surgery Surgery Non-surgery
of treatment was reported from the full regression
(n=24) (n=25) (n=58) (n=92)
model (including the interaction term), reflecting the
Demographics
treatment effect in the randomised component within
Age, years this model.
Median 63·4 61·2 61·1 59·3 To examine the sensitivity of these findings to actual
(54·6–71·1) (56·5–69·7) (51·8–69·2) (50·5–68·1)
treatment received, two sensitivity analyses were
<65 13 (54%) 17 (68%) 36 (62%) 65 (71%)
conducted, one using actual treatment received in place
Sex
of treatment assignment or selection, and one including
Female 16 (67%) 15 (60%) 36 (62%) 62 (67%)
only patients who received their assigned or selected
Male 8 (33%) 10 (40%) 23 (38%) 30 (33%)
treatment.
NCI race category
Kaplan-Meier curves were used to show overall survival
Black or African American 6 (25%) 7 (28%) 13 (22%) 16 (17%)
by treatment group and study pathway. The treatment
White 9 (38%) 12 (48%) 39 (67%) 69 (75%)
effect on overall survival was estimated using a Cox
Other* 1 (4%) 1 (4%) 1 (2%) 3 (3%) proportional hazards regression model stratified on
Unknown 0 0 5 (9%) 4 (4%) pathway and controlling for the same variables as in the
Not applicable† 8 (33%) 5 (20%) 0 0 model for good days. Time zero was defined as date of
Ethnicity random assignment or, in the patient choice pathway,
Hispanic 9 (38%) 5 (20%) 6 (10%) 13 (14%) date of registration. Follow-up was censored at last
Non-Hispanic 14 (58%) 18 (72%) 51 (88%) 78 (85%) contact date or day 91, whichever was earlier. This same
Unknown 1 (4%) 2 (8%) 1 (2%) 1 (1%) regression framework was used with linear models to
Clinical characteristics compare treatment groups with respect to the length of
Primary cancer site initial hospital stay, the days of nasogastric tube use and
Colorectal 6 (25%) 7 (28%) 19 (33%) 25 (27%) severity of nausea, vomiting, bloating, constipation, and
Gynaecological 10 (42%) 12 (48%) 14 (24%) 38 (41%) pain at week 4. A parallel logistic regression model was
Other 8 (33%) 6 (24%) 25 (43%) 29 (32%) used to assess nasogastric tube use and ability to eat at
BMI, kg/m² week 5. Treatment-related complications were evaluated
Median 23·8 25·7 24·4 24·7 on the basis of treatment received and are reported
(20·8–28·5) (20·5–29·4) (21·5–27·8) (21·3–28·6) descriptively. We conduced a prespecified exploratory
<18·5 3 (13%) 2 (8%) 7 (12%) 6 (7%) analysis of the effect of baseline ascites volume, albumin
18·5 to <25 10 (42%) 10 (40%) 26 (45%) 45 (49%) levels, and presence of carcinomatosis on good days and
25 to <30 6 (25%) 7 (28%) 12 (21%) 20 (22%) overall survival in all participants. Due to the change in
≥30 5 (21%) 6 (24%) 13 (22%) 21 (23%) protocol-specified use of somatostatin analogues during
Albumin, g/dL the study, we conducted a post-hoc analysis of the effect
Mean 3·1 (0·9) 3·1 (0·7) 3·2 (0·7) 3·2 (0·7) of somatostatin use on good days in the non-surgery
<2·8 7 (29%) 7 (28%) 18 (31%) 25 (27%) groups. We conducted post-hoc quantile analyses to
2·8–3·5 9 (38%) 11 (44%) 20 (34%) 35 (38%) address a lack of normality for the primary endpoint. We
>3·5 8 (33%) 7 (28%) 20 (34%) 31 (34%) performed quantile regressions across a wide range of
Missing 0 0 0 1 (1%) quantiles to determine whether the conclusions drawn
Ascites from the prespecified linear regression analyses would
No evidence 8 (33%) 12 (48%) 26 (45%) 43 (47%) persist in an analysis better attuned to the data. The two
Evidence of ascites: trace or small amount 11 (46%) 7 (28%) 24 (41%) 29 (32%) analytic approaches provided the same overall conclusion
Evidence of ascites: large or massive 5 (21%) 6 (24%) 8 (14%) 20 (22%) (no substantial difference between surgical and non-
amount surgical groups) and so the originally planned analyses
Carcinomatosis are reported.
No evidence 6 (25%) 5 (20%) 25 (43%) 31 (34%) Missing data for the secondary HRQOL outcomes
Suspicion of carcinomatosis 2 (8%) 2 (8%) 4 (7%) 8 (9%) were relatively common, expected, and unavoidable in
Clear evidence: trace or small amount 2 (8%) 4 (16%) 3 (5%) 0 the setting of advanced cancer. The analyses presented
Clear evidence: large or massive amount 14 (58%) 14 (56%) 26 (45%) 53 (58%) here use all available data without reweighting and must
Zubrod performance status therefore be viewed as conditional on survival and ability
0 8 (33%) 5 (20%) 9 (16%) 15 (16%) to respond. A p value less than 0·05 was considered
1 6 (25%) 9 (36%) 29 (50%) 50 (54%) statistically significant for the primary outcome. For
2 10 (42%) 11 (44%) 20 (34%) 27 (29%) other analyses, p values should be interpreted cautiously
(Table 1 continues on next page) as no adjustments for multiple comparisons were made.
With nine secondary outcomes, fewer than one of these

912 www.thelancet.com/gastrohep Vol 8 October 2023

 Articles

outcomes would be expected to be significant at the

Randomised pathway Patient choice pathway
0·05 level by chance alone.
Data analyses were generated using SAS/STAT software, Surgery Non-surgery Surgery Non-surgery
(n=24) (n=25) (n=58) (n=92)
version 9.4 of the SAS system.
This trial is registered with ClinicalTrials.gov, (Continued from previous page)

NCT02270450. Number of previous malignant bowel obstructions

None 17 (71%) 16 (64%) 32 (55%) 51 (55%)
Role of the funding source At least one 5 (21%) 4 (16%) 21 (36%) 31 (34%)
The funders of the study had no role in study design, data Unknown 2 (8%) 4 (16%) 4 (7%) 9 (10%)
collection, data analysis, data interpretation, or writing of Missing 0 1 (4%) 1 (2%) 1 (1%)
the report. Care characteristics
Care providers (might have more than one)
Results Family member 22 (92%) 21 (84%) 51 (88%) 77 (84%)
From May 11, 2015, to April 27, 2020, 221 patients were Hospice or skilled nursing 1 (4%) 1 (4%) 0 6 (7%)
enrolled (56 in the randomisation pathway [29 surgery Other 0 2 (8%) 4 (7%) 3 (3%)
and 27 non-surgery] and 165 in the patient choice pathway No-one 1 (4%) 2 (8%) 5 (9%) 11 (12%)
[65 surgery and 100 non-surgery]; figure 1). 143 (65%) Admitting attending physician specialty
participants were female and 78 (35%) were male. Medical specialty‡ 0 2 (8%) 4 (7%) 6 (7%)
22 patients were subsequently found to be ineligible or General surgery 0 1 (4%) 10 (17%) 7 (8%)
not analysable and were excluded from analyses. Thus Gynaecological oncology 3 (13%) 4 (16%) 5 (9%) 30 (33%)
199 patients were evaluable, including 49 in the Surgical oncology 20 (83%) 16 (64%) 39 (67%) 49 (53%)
randomised pathway (24 surgery and 25 non-surgery) and Other 1 (4%) 2 (8%) 0 0
150 in the patient choice pathway (58 surgery and 92 non-
surgery). The median age was 60 years (IQR 52–69). Data are median (IQR), n (%), or mean (SD). NCI=National Cancer Institute. *Other race catergories were American Indian
or Alaskan Native, Asian, and Native Hawaiian or other Pacific Islander. †NCI race categories collected for this study apply
Baseline patient characteristics are shown in table 1. only to participants in the USA, as mandated by the NCI; Mexico and Peru have their own racial designations, which were
Baseline characteristics were balanced between the not collected as part of this study. ‡Includes family practice, internal medicine, and medical oncology.
randomly assigned groups, but modest differences were
Table 1: Baseline patient characteristics by study group
seen between groups in the patient choice pathway. Larger
differences were seen between the two study pathways on
several factors, including race, ethnicity, and admitting Randomised pathway Patient choice pathway
attending physician specialty (table 1 and appendix 2 p 1).
Surgery (n=24) Non-surgery (n=25) Surgery (n=58) Non-surgery (n=92)
175 (88%) of patients adhered to the designated treat­
ment group (76 [93%] of 82 in the surgery groups and Surgical treatment
99 [85%] of 117 in the non-surgery groups). Only six (7%) Yes 20 (83%) 7 (28%) 56 (97%) 11 (12%)
of the 82 patients in the surgical groups did not have an No 4 (17%) 18 (72%) 2 (3%) 81 (88%)
operation (table 2). Similarly, 18 (15%) of 117 patients in Somatostatin analogue use*
the non-surgery groups had surgery during their initial Yes 3 (13%) 10 (40%) 5 (9%) 39 (42%)
hospital stay for malignant bowel obstruction (table 2). No 21 (88%) 15 (60%) 51 (88%) 53 (58%)
49 (42%) of 117 patients in the non-surgery groups used a Missing 0 0 2 (3%) 0
somatostatin analogue compared with eight (10%) of 82 Steroid use
in the surgical groups. Opioid use was not common in Yes 2 (8%) 5 (20%) 12 (21%) 28 (30%)
Latin America, but was frequently used by US patients No 22 (92%) 20 (80%) 44 (76%) 64 (70%)
(>80% for all groups; data not shown). Missing 0 0 2 (3%) 0
In unadjusted analyses, the mean number of good days Percutaneous endoscopic gastrostomy tube use
during the first 13 weeks after enrolment was 51·0 Yes 2 (8%) 4 (16%) 6 (10%) 22 (24%)
(SD 29·8), and was lower in the randomised pathway No 22 (92%) 21 (84%) 50 (86%) 70 (76%)
(43·3 days [30·6]) than in the patient choice pathway Missing 0 0 2 (3%) 0
(53·5 days [29·2]). The number of good days in the
*Somatostatin analogues were required for non-surgical management until revision 6 of the protocol on Sept 1, 2016.
two treatment groups in the randomised pathway were
similar, as were the number of good days in the Table 2: Malignant bowel obstruction-related treatment during the first hospital stay by study group
two treatment groups in the patient choice pathway
(table 3).
In analyses to identify potential confounders, the only with good days. The full model for good days controlled See Online for appendix 2
factor associated with treatment selection in the patient for these potential confounders as well as pathway
choice pathway was admitting attending physician (randomised vs patient care) to examine both the effect of
specialty (table 1; appendix 2 pp 3–18). Performance treatment and the potential for treatment effect to vary by
status, albumin levels, and type of cancer were associated pathway. In this model, there was no evidence for effect

www.thelancet.com/gastrohep Vol 8 October 2023 913

 Articles

Randomised pathway Patient choice pathway Adjusted combined

comparison*
Surgery (n=24) Non-surgery Unadjusted comparison Surgery (n=58) Non-surgery Unadjusted comparison Regression p value
(n=25) (n=92) estimate (95% CI)
Difference (95% CI) p value Difference (95% CI) p value
Primary outcome
Good days† 42·6 (32·2) 43·9 (29·5) –1·3 (–19·0 to 16·5) 0·88 54·8 (27·0) 52·7 (30·7) 2·1 (–7·6 to 11·9) 0·66 2·9 (–5·5 to 11·3) 0·50
Secondary outcomes: initial hospital stay
Time in hospital from registration, days
Mean 12·3 (9·3) 12·8 (12·2) –0·5 (–6·8 to 5·8) 0·87 11·6 (7·8) 6·2 (6·6) 5·4 (3·0 to 7·7) <0·0001 –1·1‡ (–5·7 to 3·5) 0·64
0–6 7 (29%) 12 (48%) ·· ·· 14 (24%) 65 (71%) ·· ·· ··
7–3 9 (38%) 5 (20%) ·· ·· 27 (47%) 11 (12%) ·· ·· ·· ··
≥14 8 (33%) 8 (32%) ·· ·· 16 (28%) 14 (15%) ·· ·· ·· ··
Missing 0 0 ·· ·· 1 (2%) 2 (2%) ·· ·· ·· ··
Nasogastric tube use
Yes 19 (79%) 20 (80%) ·· 0·94 42 (72%) 69 (75%) ·· 0·97 0·8§ (0·4 to 1·7) 0·60
No 5 (21%) 5 (20%) ·· ·· 13 (22%) 21 (23%) ·· ·· ·· ··
Missing 0 0 ·· ·· 3 (5%) 2 (2%) ·· ·· ·· ··
Time using nasogastric tube, days¶
Mean 5·8 (4·1) 7·7 (8·4) –1·9 (–5·7 to 1·9) 0·32 6·5 (5·6) 5·1 (5·3) 1·4 (–0·5 to 3·2) 0·14 0·6 (–1·1 to 2·4) 0·47
0 5 (21%) 5 (20%) ·· ·· 13 (22%) 21 (23%) ·· ·· ·· ··
1–6 8 (33%) 9 (36%) ·· ·· 15 (26%) 45 (49%) ·· ·· ·· ··
7–13 11 (46%) 7 (28%) ·· ·· 20 (34%) 18 (20%) ·· ·· ·· ··
≥14 0 4 (16%) ·· ·· 7 (12%) 6 (7%) ·· ·· ·· ··
Missing 0 0 ·· ·· 3 (5%) 2 (2%) ·· ·· ·· ··
MDASI-GI symptom assessment||
Dead 4 (17%) 5 (20%) ·· ·· 6 (10%) 14 (15%) ·· ·· ·· ··
Alive with data 11 (46%) 11 (44%) ·· ·· 31 (53%) 41 (45%) ·· ·· ·· ··
Nausea severity 0·64 (1·3) 4·6 (3·8) –4·0 (–6·6 to –1·4) 0·0061 1·9 (3·1) 2·3 (3·0) –0·4 (–1·8 to 1·1) 0·61 –4·1‡ (–6·5 to –1·7) 0·0012
Vomiting severity 0·64 (2·1) 2·9 (3·5) –2·3 (–4·8 to 0·3) 0·079 0·58 (1·5) 1·9 (3·0) –1·3 (–2·4 to –0·2) 0·024 –1·6 (–2·8 to –0·5) 0·0075
Bloating severity 1·5 (2·3) 3·9 (3·2) –2·5 (–4·9 to 0·0) 0·054 1·6 (2·4) 2·3 (2·9) –0·7 (–2·0 to 0·6) 0·27 –1·3 (–2·6 to –0·05) 0·042
Pain severity 2·9 (3·1) 5·7 (3·6) –2·8 (–5·8 to 0·2) 0·062 4·1 (3·3) 3·8 (3·6) 0·3 (–1·3 to 2·0) 0·70 –3·6‡ (–6·4 to –0·8) 0·012
Constipation severity 0·55 (1·3) 2·7 (3·9) –2·2 (–4·9 to 0·5) 0·10 0·77 (2·2) 2·2 (3·4) –1·4 (–2·8 to –0·1) 0·033 –1·9 (–3·3 to –0·6) 0·0066
Ability to eat**
Dead 4 (17%) 5 (20%) ·· ·· 6 (10%) 14 (15%) ·· ·· ·· ··
Alive with data 12 (50%) 12 (60%) ·· ·· 35 (60%) 47 (51%) ·· ·· ·· ··
Eating 12 (100%) 12 (100%) ·· 1·00 29 (83%) 40 (85%) ·· 0·78 0·7†† (0·2 to 2·5) 0·55
Not eating 0 0 ·· ·· 6 (17%) 7 (15%) ·· ·· ·· ··

Data are mean (SD) or n (%). MDASI-GI=MD Anderson Symptom Inventory for Gastrointestinal Cancer. *Comparison between surgical and non-surgical management. All models were based on the intention-to-
treat principle and were adjusted for pathway, attending physician specialty, Zubrod performance status, baseline albumin, and primary cancer site. †Defined as number of days alive and out of the hospital
during the first 91 days after registration. ‡Main effect of treatment from the full regression model (including the interaction term). §Odds ratio for treatment effect for comparing nasogastric tube use vs no
nasogastric tube use. ¶During entire hospital stay (admission to discharge). ||Assessed at week 4; Higher scores indicate greater symptom severity. **Assessed at week 5. ††Odds ratio for treatment effect for
comparing eating vs not eating—model does not include pathway.

Table 3: Study outcomes and comparisons by treatment group within pathway and by treatment group overall

modification by pathway (interaction p value=0·5), so the
interaction term was excluded from the final model.
In the final adjusted model, there was no significant
difference in the number of good days between the
surgical and non-surgical groups (table 3). However, the
distribution of the primary endpoint was bimodal rather
than normal, raising concerns about the robustness of
these results. Transformations prespecified in the
analysis plan were not effective in improving the
distribution.
Sensitivity analyses that used the same framework as
the adjusted analysis but were restricted to patients who
were treated only according to the initially assigned
treatment, showed no significant difference in good days
between surgically and non-surgically treated patients
(adjusted mean difference 4·1 [95% CI –4·8 to 13·1];
p=0·37). Using an as-treated approach, the adjusted
mean difference was 3·3 (95% CI –5·0 to 11·6; p=0·43).
Reported in-patient complications were more
common among surgically treated patients than among

914 www.thelancet.com/gastrohep Vol 8 October 2023

 Articles

non-surgically treated patients (appendix 2 p 19). During 100 Non-surgical management

their initial hospital stay, six participants died, five due to Surgery
cancer progression (four patients from the randomised
pathway, two in each treatment group, and one patient 75
from the patient choice pathway, in the surgery group)

Overall survival (%)

and one due to malignant bowel obstruction treatment
complications (patient choice pathway, non-surgery). The 50
most common grade 3–4 treatment-related complication
was anaemia (three [6%] patients in the randomised Events at 3 months 3-month survival estimate
pathway, all in the surgical group, and five [3%] patients in 25 Non-surgical management (n=117) 92 48% (95% CI 38–56)
the patient choice pathway, four in the surgical group and Surgery (n=82) 57 55% (95% CI 43–65)

one in the non-surgical group). All other complications
were rare or lower grade (appendix 2 p 19).
As expected, overall survival was poor during the
primary follow-up period with only 94 (47%) of 199 patients
surviving 91 days (median overall survival 86 days [IQR
41–91]; figure 2). Survival was higher in the patient choice
pathway than the randomised pathway (data not shown),
but there was no interaction between pathway and
assigned treatment in the Cox proportional hazards
regression (p value for interaction=0·63; appendix 2 p 4).
In the final model without the interaction term, there was
no difference in overall survival between the patients in the
surgical and non-surgical treatment groups (figure 2).
Length of the initial hospital stay was similar in the
surgical and non-surgical groups in the randomised
pathway (table 3). The unadjusted mean length of stay
for patients in the surgical group in the patient choice
pathway was similar to patients in the randomised
pathway. However, mean length of stay was significantly
shorter for patients in the non-surgical group than in the
surgical group in the patient choice pathway (table 3). In
the full regression model, the difference in length of stay
between the randomised and patient choice pathway
was significant, even after controlling for potential
confounders (adjusted mean difference –6·8 days
[95% CI –10·5 to –3·0] in the patient choice pathway vs
in the randomised pathway) and the interaction between
treatment and pathway was also significant (p=0·026).
The adjusted mean difference between treatments from
this model was not significant (–1·1 days [95% CI
–5·7 to 3·5] for surgery vs for non-surgery).
There were no significant interactions between
treatment and pathway in models of nasogastric tube
use. There was also no significant difference between
surgery and non-surgery for either use of nasogastric
tube or days of nasogastric tube use (table 3).
At week 4, significant improvement was observed with
surgery compared with non-surgery in severity scores
for vomiting and constipation (table 3). Surgery was
associated with improved nausea scores compared with
non-surgery in the full model, in which a significant
interaction between treatment and pathway reflected a
larger difference in the randomised pathway than the
patient choice pathway. Severity scores for bloating were
lower in the surgical than the non-surgical groups
(table 3). Surgery was associated with reduced pain
HR 0·70 (95% CI 0·45–1·09; p=0·12)
0
0 15 30 45 60 75 90
Time since registration (days)
Number at risk
(number censored)
Non-surgical 117 (0) 109 (2) 96 (2) 84 (3) 69 (6) 62 (7) 52 (7)
management
Surgery 82 (0) 76 (1) 70 (1) 64 (1) 54 (3) 49 (3) 43 (3)
Figure 2: Overall survival
severity compared with non-surgery in the full model
including the nominally significant treatment–pathway
interaction.
Of the 170 patients alive at week 5, 106 responded to
dietary recall data collection, 93 (88%) of whom reported
oral intake. All patients in the randomised pathway
reported being able to eat. In the patient choice pathway,
29 (83%) of 35 in the surgical group and 40 (85%) of
47 in the non-surgical group reported being able to eat.
In a logistic regression model that did not include the
pathway indicator, ability to eat did not vary between
surgical and non-surgical treatment groups (table 3).
Prespecified exploratory analyses explored whether
baseline albumin levels, large quantities of ascites, and
presence of carcinomatosis were predictors. No evidence
of effect modification was seen with any of these factors
for either good days or overall survival (data not shown).
In a post-hoc regression analysis of somatostatin
analogue use in the pooled non-surgical group, there was
no effect of somatostatin analogue use on good days
(data not shown).
Discussion
S1316 was designed to prospectively evaluate which
initial treatment was superior, surgical or non-surgical
management, for surgically eligible patients with
malignant bowel obstruction. It was a pragmatic trial in
that once the pathway of care was determined, treatments
were documented but not mandated. The main outcome
measure, good days, defined as days out of the hospital
and alive at 91 days, was not significantly different between
surgical and non-surgical treatment groups. Furthermore,
the analysis did not show a difference between treatments
in ability to consume food orally or in overall survival.
There was a significant advantage for the surgical groups
for symptoms related to malignant bowel obstruction. This

www.thelancet.com/gastrohep Vol 8 October 2023 915

 Articles
indicates that an initial surgical approach might provide
some meaningful HRQOL benefit among surgically
eligible patients with malignant bowel obstruction.
Surgical complications might be quite high for patients
with malignant bowel obstruction, although wide ranges
are reported (5–87%).3 During their first hospital
admission during the current study, 15% of surgical
patients reported a treatment-related complication versus
4% of the non-surgery patients, but length of initial
hospital stay remained similar across treatment groups
in the randomised pathway. Although surgical comp­
lications might be a surrogate for worse HRQOL and
longer hospital stays, this was not observed in the study.
Therefore, although surgical complications must be
carefully considered, they should not be the most
important factor in treatment decision making.
Non-surgical approaches to treat malignant bowel
obstruction have typically been studied in patients
described as being non-operative. There is an inherent
bias for many of these patients, as some might be
surgically eligible but are never evaluated by a surgical
team. In addition, there might be surgical bias owing to
the type of surgeon, institution, or training. Non-operative
approaches typically include nasogastric de­compression,
intravenous hydration, medications to treat symptoms
such as nausea and pain, and antisecretory agents,
most notably somatostatin analogue and steroids.15 A
combination or multidrug therapy of medications is likely
to confer a benefit when compared with individual
medications for malignant bowel obstruction.16 In
practice, there are multiple non-surgical strategies
available for patients with malignant bowel obstruction.
Treatment should be individualised and palliative care
specialists should be sought.15 Although somatostatin
analogues are a mainstay in the treatment of malignant
bowel obstruction, their availability or clinician familiarity
is not universal. Some providers do not use somatostatin
analogues due to concerns regarding side-effects, drug
interactions, and cost. In a randomised study, somatostatin
analogues did not show benefit in preventing emesis
within 72 h after administration.17 In a review of all studies
evaluating the efficacy of somatostatin analogues in the
setting of malignant bowel obstruction, there was no high-
level evidence of benefit, although further study might
be warranted.18 Somatostatin analogues were initially a
mandatory treatment in the non-surgical groups for this
trial, but it was not possible to complete accrual in this
study with this as a criterion. As expected, more patients
in the non-surgical groups were given somatostatin
analogues and steroids. In a post-hoc regression analysis
of somatostatin analogue use in the pooled non-surgical
group, there was no effect of somatostatin analogue use
on good days, suggesting that the non-standardised use of
somatostatin analogues probably did not play a role in our
results. Venting percutaneous endoscopic gastronomy
tubes are a widely accepted, well tolerated non-operative
treatment approach for patients with malignant bowel
916
obstruction.19–21 A venting percutaneous endoscopic
gastronomy tube was used more frequently in the non-
surgery groups than the surgery groups.
The ability to consume food orally is of considerable
importance for patients with a malignant bowel
obstruction and should not be overlooked. All the
patients in the randomised pathway and more than 80%
of those in the patient choice pathway who were still
contributing data at week 5 were able to eat, and no
difference between treatment approaches was observed.
The large amount of missing data, which might reflect
bias in response as those in poorer health were less likely
to complete recalls and more likely to have difficulty
consuming food orally, precludes a definitive assessment
based on these analyses. Potential differences related to
eating will be examined in future reports, including the
relationship to good days and HRQOL outcomes.
There are limited reports of prospective HRQOL-related
symptom measurements for patients treated for malignant
bowel obstruction, and none in the setting of a comparative
trial of surgical versus non-surgical approaches. The
patient experience, which includes HRQOL measures, is
under-reported and should be included in studies of
malignant bowel obstruction.22 These symptoms might be
of primary importance to patients with a terminal
diagnosis who are making end-of-life decisions when
determining treatment choices. Our data support multiple
benefits of an initial surgery approach for malignant bowel
obstruction, including improvements for nausea and
vomiting, constipation, and pain. Bloating was improved,
although the evidence was less compelling. In the setting
of an advanced cancer diagnosis, patients who are unable
to have a bowel movement often equate that symptom to
constipation, when it might in fact be due to a bowel
obstruction. The discomfort of the symptoms of malignant
bowel obstruction is often substantial and a source of
suffering for many patients. Therefore, the findings of
HRQOL symptom benefits in this study should be a major
factor in decision making for patients considering surgery
or a non-surgical treatment approach for malignant bowel
obstruction. Although symptoms alone do not capture
important associations of individual HRQOL items
with outcome domains such as emotional or social
functioning, which are crucial to defining the patient
experience with malignant bowel obstruction,23 they are
clearly important to patients and their families.
Although survival is not the primary goal for patients
admitted to the hospital with a malignant bowel
obstruction, it is always an important concern for patients
and families. An expectation of patients’ long-term
outcome allows for greater understanding of their
disease process and might play a role when considering
such divergent treatment options. The findings of this
trial did not show significant survival differences by the
91-day timepoint, although the hazard ratio was
consistent with lower survival in the non-surgical group
compared with the surgical group. Longer-term survival
www.thelancet.com/gastrohep Vol 8 October 2023

data might provide clarification that would be helpful in
counselling patients and families regarding goals of
treatment, palliation, and end-of-life care.
Limitations to this study include potential residual
confounding, departure from original design assumptions
regarding recruitment patterns requiring additional
enrolment centres, no standardisation of somatostatin
analogue use, treatment crossover, and missing data for
secondary outcomes not missing at random, which
introduces potential survivor bias. Statistical approaches
to account for these were used wherever possible but
these might not fully address the issues. Although there
was a greater proportion of patients with admitting
attending gynaecological oncologists in the patient choice
pathway, the analysis did not show evidence of noteworthy
confounding of treatment effect. Patients with previous
malignant bowel obstruction were more likely to select
the patient choice pathway, but there was no difference
in their choice between surgical and non-surgical
management. Latin American sites entered the trial later
when our focus was on the randomised pathway and the
modest sample size precluded us from examining
variation in treatment or pathway effects by site. Another
potential limitation was that we did not collect data on
palliative care team involvement—we focused on the
primary team during the hospitalisation, which could be
a factor in non-surgical care. Finally, further examining
the effect of HRQOL issues on patient priorities and
preferences is imperative.
There are multiple important achievements associated
with this trial, including a large proportion of Black
and Hispanic patients accrued from wide geographic
locations, and completion of accrual in an understudied
population of patients treated for advanced cancer.
Inclusion of Latin American sites enhanced accrual to
the randomised pathway and increased the general­
isability of results. Although there have been many
reports related to malignant bowel obstruction and
treatment, there are very few prospective studies, and
none as large as S1316, that include a comparison of
surgical and non-surgical approaches with the rigorous
inclusion criteria of our study and a randomised
component. Participants in both pathways were similar
in clinical characteristics, and therefore equally ill.
Finally, as accruing to a trial with patients admitted to
the hospital with end-stage disease is difficult, the ability
to complete accrual is a notable accomplishment; what
was learned in the process will be helpful in attempts to
initiate trials for patients with cancer with difficult
palliative care issues.24
Although there was no difference between treatment
groups in the primary outcome of good days (days out of
the hospital and alive), our findings suggest that patients
who have initial surgery have improved malignant
bowel obstruction-related symptoms. The study did not
show a significant effect on survival by day 91, but the
hazard ratio was consistent with a possible benefit for
www.thelancet.com/gastrohep Vol 8 October 2023
Articles
surgery when compared with non-operative approaches;
longer-term follow-up might clarify this association.
Although management of patients with malignant
bowel obstruction will remain challenging for patients
and clinicians, these findings are of great importance to
patients and their families. The results of S1316 could
allow for more informed conversations between
clinicians and patients and their families.
Contributors
RSK, JS, and MG designed the study. RSK worked on the
conceptualisation, investigation, and funding acquisition. CAT, KBA,
GLA, and RSK provided the methodology. RSK, VNN, MFA-K, VS, AAS,
ESH, AML-T, DI-O, PR, SH, MSS, JAL, WSG, DCF, HM, TJM, JPM,
MK, BB, GW, JPL, OAG, MSS, SBD, ES, JDW, BFr, BFa, SAS, GBD, and
JLD administered the project. RSK wrote the original draft and provided
resources. Formal data analysis was performed by GLA, KBA, and JS.
GLA and KBA provided data validation and accessed and verified the
data. Not all the authors had access to all the underlying data because
SWOG policy is that the statistics and data management centre and the
statistic centres are the only ones with full access to raw data. Data
interpretation was performed by RSK, GLA, KBA, VS, AAS, JS, MG,
GBD, and JLD. Data was collected by VNN, MFA-K, AAS, ESH, AML-T,
DI-O, PR, SH, MSS, JAL, WSG, DCF, HM, TJM, JMP, MK, BB, GW,
JPL, OAG, MSS, SBD, ES, JDW, BFr, BFa, SAS, GBD, and JLD. Data
was curated by KBA and CAT. The study tables and figures were created
by KBA. AAS performed the literature search and reviewing. Patient
accrual was performed by VNN, MFA-K, JLW, AAS, ESH, AML-T, DI-O,
PR, SH, MSS, JAL, WSG, DCF, HM, TJM, JMP, MK, BB, GW, JPL,
OAG, MSS, SBD, ES, JDW, BFr, BFa, SAS, GBD, and JLD. All the
authors reviewed and edited the manuscript except JS (passed away).
The manuscript and the study’s data were approved in their final
version for submission for publication by all authors, except JS.
Declaration of interests
GLA received payments to Fred Hutch for support for the present
manuscript. AAS has grants or contracts from AbbVie, Aravive,
AstraZeneca, Boehringer, Ingelheim, Clovis, Eisai, Ellipses,
Immunogen, Merck, Oncoquest, Roche/Genentech, SeagenInc,
TapImmune, Tesaro/GSK, VBL Therapeutics, and National Cancer
Trial Network; has received honoraria for educational presentations or
lectures from @Point of Care, Clinical Care Optios, Curio Science,
Peerview, Bio ASCEND, RTP, and GOG Foundation; has received
honorarium from Myriad; and has received support for attending
meetings or travel from GOG, Society of Gynecologic Oncology, and
NRG. DCF received support for the present manuscript in the form of
payments through S1316 to Valley Health for patient enrolment;
received payment or honoraria for lectures from Intuitive for the
resident lecture programme; and has stock options in Intuitive.
JDW has grants or contracts from Merck; receives royalties or licenses
from UptoDate; receives payments for expert testimony from
Medicolegal for consulting on gynaecological cancer; is a journal
editor for the American College of Obstetricians and Gynecologists.
BFa received visiting professor honoraria in April, 2022, at the Ohio
State University Department of Surgery, October, 2021, at the
University of Nebraska Department of Surgery, and December, 2020, at
the University of Tennessee Department of Surgery; and has stock or
stock options in Align Tech, Biogen, Bristol Myers Squibb, DexCom,
Editas Medicine, Fulgent Genetics, GoodRx Holdings, Guardant
Health, Globus Medical, Healthequity, HCA Healthcare, IDEXX
Laboratories, Illumina, Intuitive Surgical, Invitae, Masimo, Moderna,
Neurocrine Bioscience, Novocure, Quidel, Repligen, Seagen,
Shockwave Medical, STAAR Surgical, UnitedHealth Group, Veeva
Systems, Teladoc Health, and ResMed. All other authors declare no
competing interests.
Data sharing For SWOG’s policy
SWOG makes research data available to investigators and memorandum number 43 see
pharmaceutical companies, as required by the policies of the National https://www.swog.org/sites/
Institutes of Health (NIH). Any SWOG research data may be default/files/docs/2019-12/
requested, but requests for data including study endpoints will only be Policy43_0.pdf
917
 Articles
considered once the primary study analyses are published. In rare
cases, study endpoint data may be provided before publication with
approval of the group chair’s office. Procedures and details for data
sharing can be found in SWOG’s policy memorandum number 43.
Acknowledgments
This original work was supported under grant awards by the Agency
for Healthcare Research and Quality (grant number HS021491), the
National Cancer Institute (NCI), Division of Cancer Prevention,
National Cancer Institute Community Oncology Research Program
Research Base (grant number UG1CA189974), and Behavioral
Measurement and Interventions Shared Resource (grant number
P30CA023074) NCI/NIH. The content is solely the responsibility of the
authors and does not necessarily represent the official views of the
NIH.
References
1 Badgwell BD, Smith K, Liu P, Bruera E, Curley SA, Cormier JN.
Indicators of surgery and survival in oncology inpatients requiring
surgical evaluation for palliation. Support Care Cancer 2009;
17: 727–34.
2 Badgwell BD, Contreras C, Askew R, Krouse R, Feig B,
Cormier JN. Radiographic and clinical factors associated with
improved outcomes in advanced cancer patients with bowel
obstruction. J Palliat Med 2011; 14: 990–96.
3 Cousins SE, Tempest E, Feuer DJ: Surgery for the resolution of
symptoms in malignant bowel obstruction in advanced
gynaecological and gastrointestinal cancer.
Cochrane Database Syst Rev 2016; 2016: CD002764.
4 Pujara D, Chiang YJ, Cormier JN, Bruera E, Badgwell B.
Selective Approach for Patients with Advanced Malignancy and
Gastrointestinal Obstruction. J Am Coll Surg 2017; 225: 53–59.
5 Higginson IJ, Sen-Gupta GJA. Place of care in advanced cancer:
a qualitative systematic literature review of patient preferences.
J Palliat Med 2000; 3: 287–300.
6 Last Acts. Means to a better end: a report on dying in America
today. https://folio.iupui.edu/handle/10244/469 (accessed
March 25, 2013).
7 Badgwell B, Bruera E, Klimberg SV. Can patient reported
outcomes help identify the optimal outcome in palliative surgery?
J Surg Oncol 2014; 109: 145–50.
8 Anthony T, Baron T, Mercadante S, et al. Report of the clinical
protocol committee: development of randomized trials for
malignant bowel obstruction. J Pain Symptom Manage 2007;
34 (suppl): S49–59.
9 Pocock SJ, Simon R. Sequential treatment assignment with
balancing for prognostic factors in the controlled clinical trial.
Biometrics 1975; 31: 103–15.
10 Schakel SF. Maintaining a nutrient database in a changing
marketplace: keeping pace with changing food products—
a research perspective. J Food Compost Anal 2001; 14: 315–22.
11 Wang XS, Williams LA, Eng C, et al. Validation and application of
a module of the M. D. Anderson Symptom Inventory for
measuring multiple symptoms in patients with gastrointestinal
cancer (the MDASI-GI). Cancer 2010; 116: 2053–63.
918
12 Cleeland CS, Mendoza TR, Wang XS, et al. Assessing symptom
distress in cancer patients: the M.D. Anderson Symptom
Inventory. Cancer 2000; 89: 1634–46.
13 Cleeland CS. MDASI user guide, version 1. Nov 19, 2009.
https://www.mdanderson.org/content/dam/mdanderson/
documents/Departments-and-Divisions/Symptom-Research/
MDASI_userguide.pdf (accessed March 28, 2022).
14 Sloan JA, Dueck A. Issues for statisticians in conducting analyses
and translating results for quality of life end points in clinical
trials. J Biopharm Stat 2004; 14: 73–96.
15 Boland JW, Boland EG. Malignant bowel obstruction.
In: Cherny NI, Fallon MT, Kaasa S, et al, eds. Oxford textbook of
palliative medicine, 6th edn. New York, NY: Oxford University
Press, 2021: 904–17.
16 Wey W, Mian M, Calabrese R, et al. Palliative medical management
of inoperable malignant bowel obstruction with “triple therapy”:
dexamethasone, octreotide, and metoclopramide.
Am J Hosp Palliat Care 2021; 38: 340–45.
17 Currow DC, Quinn S, Agar M, et al. Double-blind, placebo-
controlled, randomized trial of octreotide in malignant bowel
obstruction. J Pain Symptom Manage 2015; 49: 814–21.
18 Obita GP, Boland EG, Currow DC, Johnson MJ, Boland JW.
Somatostatin analogues compared with placebo and other
pharmacologic agents in the management of symptoms of
inoperable malignant bowel obstruction: a systematic review.
J Pain Symptom Manage 2016; 52: 901–19.
19 Thampy S, Najran P, Mullan D, Laasch H-U. Safety and efficacy of
venting gastrostomy in malignant bowel obstruction: a systematic
review. J Palliat Care 2020; 35: 93–102.
20 Zucchi E, Fornasarig M, Martella L, et al. Decompressive
percutaneous endoscopic gastrostomy in advanced cancer patients
with small-bowel obstruction is feasible and effective: a large
prospective study. Support Care Cancer 2016; 24: 2877–82.
21 Dittrich A, Schubert B, Kramer M, et al. Benefits and risks of a
percutaneous endoscopic gastrostomy (PEG) for decompression in
patients with malignant gastrointestinal obstruction.
Support Care Cancer 2017; 25: 2849–56.
22 Baddeley E, Mann M, Bravington A, et al. Symptom burden and
lived experiences of patients, caregivers and healthcare
professionals on the management of malignant bowel obstruction:
a qualitative systematic review. Palliat Med 2022; 36: 895–911.
23 Bravington A, Obita G, Baddeley E, et al. The range and suitability
of outcome measures used in the assessment of palliative
treatment for inoperable malignant bowel obstruction:
a systematic review. Palliat Med 2022; 36: 1336–50.
24 Deutsch GB, Deneve JL, Al-Kasspooles MF, et al. Intellectual
equipoise and challenges: accruing patients with advanced cancer
to a trial randomizing to surgical or nonsurgical management
(SWOG S1316). Am J Hosp Palliat Care 2020; 37: 12–18.
www.thelancet.com/gastrohep Vol 8 October 2023