Paperonium

← Back to the plain-language case

Registered 6-week re-intervention count reported instead as a 6-month total procedure count, without disclosure

Gut (2018) Β· doi:10.1136/gutjnl-2017-315335 Β· PMC6582745 Β· NCT02685865

Summary

This paper is the primary report of a single-centre randomised trial comparing lumen-apposing metal stents (LAMS) with plastic stents for EUS-guided drainage of walled-off necrosis (WON). In every ClinicalTrials.gov version (v0, 2016-02-18, to v7, 2019-02-08) the primary outcome has a 6-week time frame, and its description defines the counted events as re-interventions: repeat EUS-guided drainage, direct endoscopic necrosectomy (DEN) or a percutaneous catheter "due to inadequate treatment response". v0 to v3 title it "The number of re-interventions performed between metal and plastic stent groups to achieve treatment success". v4 (2017-09-13) changed "re-interventions" to "interventions" in the measure title and in the first sentence of the description, but left the definition, the hypothesis and the time frame unchanged; v4 to v7, including the current record, carry this wording. The paper's primary outcome is instead the "total number of procedures needed to achieve treatment success at 6-month follow-up". This count also includes the index procedure, ERCP, enteral feeding tubes and procedures for adverse events, so it differs from the registered outcome in both time point and construct. The paper does not mention the difference, and it reports no per-patient result or between-group test for the registered outcome.

What was registered

The record has eight versions, v0 to v7 (indices from the ClinicalTrials.gov history API). The record was first submitted on 2015-11-24. The earliest version, v0, is dated 2016-02-18 (first posted 2016-02-19), six days after the actual study start of 2016-02-12. v0 gave the start only as "2016-02", and from v3 (2017-04-21) the start is "2016-02-12" (actual). No version exists on or before the study start.

v0, 2016-02-18 (status Not yet recruiting; estimated enrolment 62). One primary outcome:

Measure: "The number of re-interventions performed between metal and plastic stent groups to achieve treatment success" (time frame: "6 weeks")

Description: "The primary objective of this study is to compare the number of re-interventions performed between metal and plastic stent groups to achieve treatment success at 6 week follow-up. Re-intervention is defined as the need for repeat EUS-guided drainage of WON using the same stent type and/or DEN and/or insertion of percutaneous catheter due to inadequate treatment response following initial WON drainage. As FCSEMS have larger diameter than plastic stents, the investigators hypothesize that the no. of re-interventions performed will be lower in the FCSEMS group compared to the plastic stent group."

The registered secondary outcome for treatment success was:

"Treatment success" (time frame: "6 week"): "Decrease in the size of WON to <33% of the original size and resolution of symptoms at 6-week outpatient follow-up (from day of hospital discharge) WITHOUT the need for cross-over to minimally invasive surgical debridement"

The other registered secondaries were technical success (day 0), treatment failure (6 months), recurrence (6 months), procedure duration (day 0), adverse events (6 months), duration of hospitalisation (6 months) and cost analysis (6 months).

v1 (2016-04-12), v2 (2017-02-13), v3 (2017-04-21). No change to the outcome measures. v1 changed the Study Identification module (brief title from "Trial Comparing FCSEMS and Plastic Stents" to "Comparison of FCSEMS and Plastic Stents") and the status, description, conditions and contacts modules; v2 and v3 changed the status module (status and dates) only.

v4, 2017-09-13 (status Active, not recruiting; actual enrolment 60; primary completion "2017-07-10", actual). The module labels include "Outcome Measures". The primary measure title became:

"The number of interventions performed between metal and plastic stent groups to achieve treatment success" (time frame: "6 weeks")

In the first sentence of the description, "re-interventions" became "interventions" ("...to compare the number of interventions performed between metal and plastic stent groups to achieve treatment success at 6 week follow-up."). The rest of the description was not changed: the definition still begins "Re-intervention is defined as the need for repeat EUS-guided drainage of WON using the same stent type and/or DEN and/or insertion of percutaneous catheter due to inadequate treatment response following initial WON drainage", and the hypothesis still refers to "the no. of re-interventions". The secondary treatment-success description became "Resolution of WON on CT scan and resolution of symptoms at 6-week outpatient follow-up (from day of hospital discharge) WITHOUT the need for cross-over to minimally invasive surgical debridement", still at "6 week". v4 also made wording edits to two other secondary-outcome descriptions, with no change to their time frames: treatment failure gained the words "or underlying diseaseor underlying disease." (as written), and procedure duration now starts when the "endoscope [is] inserted into the gastrointestinal lumen to perform transmural drainage" instead of "inserted into the oropharynx". This edit was made after the actual primary completion date then registered (2017-07-10; v6 later changed it to 2018-07-10) and nine days before the manuscript was received (2017-09-22).

v5, 2018-05-30. This is the last version before online publication (2018-06-01). The outcomes are the same as in v4.

v6, 2018-07-31, and v7, 2019-02-08. No outcome change. v6 changed the actual primary completion date from 2017-07-10 to 2018-07-10. v7 set the status to Completed. The current record (fetched from /api/v2/studies/NCT02685865 on 2026-10-23; last update posted 2019-02-11) matches v7, and it still gives the primary time frame as "6 weeks".

No version gives a 6-month time frame for the primary outcome. None defines the primary outcome as a count that includes the index procedure, ERCP, enteral feeding tubes or procedures for adverse events.

What the paper reports

The primary outcome, Abstract (Design):

"Primary outcome was comparing total number of procedures to achieve treatment success defined as symptom relief in conjunction with WON resolution on CT at 6 months."

Introduction, final paragraph:

"The primary objective of this randomised trial was to evaluate the effectiveness of LAMS by comparing the total number of procedures required to achieve treatment success between patients undergoing LAMS and plastic stent placement at 6-month follow-up."

Methods > Outcome measures:

"The main outcome measure was to compare the total number of procedures needed to achieve treatment success at 6-month follow-up."

Methods > Definitions:

"The total number of procedures required to achieve treatment success included the index endoscopic intervention, endoscopic reinterventions, percutaneous drain placement, ERCP, enteral feeding tube placement and any procedures performed for the management of adverse events or disease-specific events in each cohort during the 6-month follow-up period."

"Treatment success was defined as resolution of WON on CT scan in association with clinical resolution of symptoms at 6-month follow-up."

Methods > Sample size calculation:

"A two-sided sample size calculation was performed to detect a difference of one in the total number of procedures needed to achieve treatment success (assuming one for LAMS and two for plastic stents)"

Methods > Patient follow-up. This is the only role the 6-week visit has in the paper:

"On initial outpatient follow-up at 6 weeks postintervention, all patients underwent CT to assess treatment response prior to stent removal and an ERCP to assess for pancreatic duct integrity."

The primary result, Abstract (Results) and Table 3 ("Total no. of procedures for treatment success": mean 2.8 (SD 1.2) vs 3.2 (SD 1.5); median 2 vs 3; range 2–7 vs 2–7; p=0.192):

"There was no significant difference in total number of procedures performed (median 2 (range 2–7) LAMS vs 3 (range 2–7) plastic, p=0.192)"

The paper's statements about the registration and the protocol:

The online supplement (gutjnl-2017-315335supp001.pdf, retrieved from the Europe PMC supplementary-files archive) has four tables: costs; "Procedures performed to achieve treatment success"; Poisson regression; and linear regression. Supplementary Table 2 lists procedure totals by type (LAMS vs plastic): "Repeat EUS-guided drainage 2 / 5", "Endoscopic necrosectomy 4 / 6", "Percutaneous drain placement 2 / 5", "Enteral feeding tube placement 7 / 13", "ERCP for pancreatic duct integrity 31 / 29" and "Management of adverse events 14 / 3". The supplement says nothing about the registration or any outcome change.

The discrepancy

Outcome Registered (version, date) In the paper Status
Number of re-interventions (repeat EUS drainage with the same stent type, DEN or percutaneous catheter, for inadequate response) to treatment success Primary in v0–v3 (2016-02-18 to 2017-04-21). From v4 (2017-09-13) the title says "interventions", but the description keeps the re-intervention definition (through v7, 2019-02-08) Not an outcome. The Results text and Supp. Table 2 give only procedure totals by type (LAMS 2+4+2 = 8; plastic 5+6+5 = 16), with no per-patient summary and no test not reported
Time point of the primary outcome "6 weeks" in every version, v0–v7; description "at 6 week follow-up" "at 6-month follow-up" (Abstract, Introduction, Methods) time point changed
Total number of procedures to treatment success (index procedure, endoscopic reinterventions, percutaneous drains, ERCP, enteral feeding tubes, procedures for adverse or disease-specific events) In no version Primary outcome and basis of the sample size calculation; median 2 vs 3, p=0.192 added (as primary)
Treatment success Secondary, "6 week"; v0: "<33% of the original size" plus symptoms; v4 (2017-09-13): "Resolution of WON on CT scan" plus symptoms Secondary, "at 6-month follow-up"; 29 (93.5%) vs 28 (96.6%), p=0.999; also the end point of the primary count time point changed
Technical success, procedure duration, adverse events, length of stay, costs, recurrence, treatment failure Secondary, v0–v7 (day 0 or 6 months) Secondary outcomes or definitions in Methods; reported in Table 3 and the text matched

Figures

Timeline of registration, registry versions, enrolment and publication for NCT02685865

Figure 1. Timeline of NCT02685865 and the paper. Blue marks are registry events (above the axis). Filled circles are the first submission (2015-11-24), v0 (2016-02-18, six days after the actual study start of 2016-02-12) and v4. The diamond is the registered actual primary completion (2017-07-10). Open squares are registry versions that did not change the outcomes (v1–v3, v5, v6, v7); v5 (2018-05-30) sits beside the publication marker. The ringed marker is v4 (2017-09-13), the only version that edited the outcome measures. It falls after the actual primary completion date then registered (2017-07-10; v6 later changed it to 2018-07-10) and nine days before the manuscript was received. Orange marks are the paper (below the axis): the randomisation period it reports (February 2016 to March 2017), receipt (2017-09-22; its axis marker sits under the v4 marker nine days before it, and its stem leads down to the label), online publication (2018-06-01) and print (2019-07). The paper's 6-month primary outcome appears at no point in the registry. v6 (2018-07-31), after publication, moved the actual primary completion date to 2018-07-10. Source: make_figures.py.txt.

Registered versus reported primary outcomes

Figure 2. On the left, in blue, are the registered outcomes. On the right, in orange, are what the paper reports. Each connector carries a label for its status. Top row: the registered primary (re-interventions to treatment success at 6 weeks) was replaced by a 6-month total procedure count. That count adds the index procedure, ERCP, feeding tubes and procedures for adverse or disease-specific events. Its result was median 2 vs 3, p=0.192, with a minimum of 2 in each arm. The v4 title edit is noted in grey. Middle row (dashed outline: an outcome with no reported result): for the registered primary as defined, the paper gives only procedure totals for its three components (8 vs 16), with no per-patient summary, no test and no 6-week split. Bottom row: the registered secondary, treatment success at 6 weeks, is reported at 6 months. The footer lists the other procedure types in the paper's count (Supplementary Table 2) and the paper's only registration statement and only protocol amendment.

Why it matters

The paper reports no result for the registered primary outcome. The registered measure is a per-patient count of re-interventions done for inadequate treatment response, reaching treatment success at 6 weeks, and the paper gives no between-group comparison of it at any time point. The closest data are procedure totals for the three registered components: 8 in the LAMS arm (2 extra transmural tracts, 4 necrosectomies, 2 percutaneous drains) against 16 in the plastic arm (5, 6 and 5). These come from Results > LAMS cohort and Plastic stent cohort ("Additional procedures undertaken due to suboptimal treatment response at 72 hours (for ongoing SIRS) included creation of additional transmural tracts (n=2), direct endoscopic necrosectomy (n=4) and percutaneous drainage catheter placement (n=2)" and "...(n=5), direct endoscopic necrosectomy (n=6) and percutaneous drainage catheter placement (n=5)") and Supplementary Table 2. These are event totals, not patient-level data. They are not split at 6 weeks and are not tested, so the registered comparison cannot be reconstructed from the paper.

The substituted outcome is built differently in ways that bear on its result:

So the paper's null primary result (p=0.192) and its "non-superiority" headline concern a different end point from the one registered. The totals for the registered components are lower in the LAMS arm (8 vs 16), but no test is given and these are not per-patient data. Whether the registered comparison would have been null or not cannot be judged from what is published. The registered hypothesis itself ("the no. of re-interventions performed will be lower in the FCSEMS group") is never tested as stated.

Two further points make the published primary harder to interpret:

This is not a case where a null registered result became a positive headline: the published primary is itself null, and both outcomes are procedure counts. But the paper presents the 6-month total count as the trial's primary outcome. Readers, reviewers and meta-analysts who take it at face value would not know that the registered primary, measured at 6 weeks and restricted to re-interventions, is not reported.

Explanations considered

Status

As of 2026-10-23:

Suggested correction

The authors and the journal could add a statement to Methods > Outcome measures, along these lines:

"The primary outcome registered at ClinicalTrials.gov (NCT02685865; first submitted 24 November 2015) was the number of re-interventions (repeat EUS-guided drainage using the same stent type, direct endoscopic necrosectomy, or percutaneous catheter insertion for inadequate treatment response) needed to achieve treatment success at 6-week follow-up. The primary outcome reported here, the total number of procedures (including the index procedure, ERCP, enteral feeding tubes and procedures for adverse events) needed to achieve treatment success at 6-month follow-up, differs from the registered outcome in its definition and time point. [Date of and reason for the change.]"

The correction could also report the registered primary outcome as prespecified: the per-patient number of re-interventions to treatment success at 6 weeks, by arm, with the planned between-group test, or say why it cannot be computed. Treatment success at 6 weeks, the registered secondary, could be reported too. If a protocol amendment or statistical analysis plan set up the 6-month total count, making that document and its date available would let readers check when the end point was fixed relative to unblinded data. The registry record could be updated to give the outcome that was analysed, with a note on when it changed.