# Antidepressant maintenance × complete relapse prevention — evidence sheet distinguishing remaining 52-week events

Verdict 2937 · Evidence Check

Stored-material review date: Not recorded  
Original evidence cut-off: 2026-08-26  
Review-criteria version: 1.0  
Public page/material preparation date: 2026-10-11

Clinical papers were not read directly; existing review summaries were checked. No new clinical-literature search was performed. Unverified does not mean that research is absent or that there is no effect.

## Assessment scope

- Population: Adults on long-term treatment who felt well enough to consider discontinuation; at least two previous episodes or at least two years of treatment; UK primary care
- Ingredient, formulation and route: Historical study exposure to long-term maintenance of four specific antidepressants; not a recommendation to maintain treatment
- Dose: Original included medications and doses are preserved in the complete summary; no individual medication, tapering, discontinuation or restarting instructions are created
- Comparator: Tapering to matching placebo in the original trial; usual concomitant treatment and detailed actual exposure remain unread
- Measured outcome: Selected complete-relapse-prevention claim; distinguish average reduction from the direct original finding of 92/238 (39%) maintenance participants relapsing by 52 weeks
- Duration: Original duration 52 weeks; the original 1~2-month taper and 12-week withdrawal measurement are historical study conditions and measurement times, not instructions
- Questions outside this scope: Repurposing average reduction as complete prevention for every person or duration; Mixing the 468-participant time-to-event denominator with the 478-participant absolute-event denominator; Merging withdrawal, relapse, restarting and SAE as the same events; Certifying all medications, populations, lifelong prevention or individual prediction; Recommendations to maintain, discontinue, taper or restart treatment

## Checklist items

### N01 · Have the complete-prevention claim, subclaim grade D and parent relapse-reduction grade B been kept separate?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The selected existing claim is ‘Maintenance completely prevents relapse’, with original grade D and a note that 39% relapsed despite maintenance. The parent concerns average relapse reduction, B with 70 points and Caution.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Absent subclaim score, status and axes are not filled with parent values. This is not a new grade assessment.
- Reason for missing information: Absent subclaim score, status and axes are not filled with parent values. This is not a new grade assessment.
- Required materials and conditions: Separate preservation of the subclaim wording, grade and note and the parent verdict
- Materials to obtain first: Absent subclaim score, status and axes are not filled with parent values. This is not a new grade assessment.
- Further research: Proposal: Link additional materials while distinguishing complete prevention from average relapse reduction.
- Conditions for verification: Separate preservation of the subclaim wording, grade and note and the parent verdict
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N02 · What were the original population's remission, long-term use and consideration of discontinuation conditions?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Participants were adults who felt well enough to consider discontinuation, had at least two previous depressive episodes or at least two years of treatment, and were long-term users. The trial included 150 UK general practices.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. The findings are not automatically applied to other diseases, treatment durations, remission states or populations. Detailed baseline and eligibility tables remain unread.
- Reason for missing information: The findings are not automatically applied to other diseases, treatment durations, remission states or populations. Detailed baseline and eligibility tables remain unread.
- Required materials and conditions: Age, remission, previous episodes, treatment history, eligibility and denominators
- Materials to obtain first: The findings are not automatically applied to other diseases, treatment durations, remission states or populations. Detailed baseline and eligibility tables remain unread.
- Further research: Proposal: Obtain original eligibility, baseline characteristics, remission assessment and relapse history.
- Conditions for verification: Age, remission, previous episodes, treatment history, eligibility and denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N03 · Have the four antidepressants' specific study conditions been extended to all medications?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: ANTLER included users of specified doses of citalopram, fluoxetine, sertraline and mirtazapine. Original doses are preserved in the complete existing review summary.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not extend the finding to complete prevention with other medications, doses or patient groups or create instructions for taking, maintaining or discontinuing medication. Original medication-specific denominators and effects remain unread.
- Reason for missing information: Do not extend the finding to complete prevention with other medications, doses or patient groups or create instructions for taking, maintaining or discontinuing medication. Original medication-specific denominators and effects remain unread.
- Required materials and conditions: Recording included medications, original study doses, exposure and group composition
- Materials to obtain first: Do not extend the finding to complete prevention with other medications, doses or patient groups or create instructions for taking, maintaining or discontinuing medication. Original medication-specific denominators and effects remain unread.
- Further research: Proposal: Obtain medication-specific inclusion, baseline exposure, denominators and prespecified subgroup materials.
- Conditions for verification: Recording included medications, original study doses, exposure and group composition
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N04 · What was the actual comparison between maintenance and tapering to matching placebo?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The randomized comparison was long-term maintenance versus tapering and switching to matching placebo. The original study's 1~2-month taper is preserved only as a historical condition.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. This is not equated with abrupt discontinuation or every discontinuation method. No new taper schedule, medication instructions or treatment substitution is created.
- Reason for missing information: This is not equated with abrupt discontinuation or every discontinuation method. No new taper schedule, medication instructions or treatment substitution is created.
- Required materials and conditions: Historical interventions, placebo switching, common treatments and actual exposure
- Materials to obtain first: This is not equated with abrupt discontinuation or every discontinuation method. No new taper schedule, medication instructions or treatment substitution is created.
- Further research: Proposal: Obtain original intervention, placebo, actual exposure, common-treatment and change materials.
- Conditions for verification: Historical interventions, placebo switching, common treatments and actual exposure
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N05 · Are the 478 randomized participants and group-specific absolute relapse denominators preserved?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The 478 participants were allocated to maintenance, 238, or tapering to placebo, 240. Absolute relapse counts were 92/238 versus 135/240.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Two reports of the same trial, multiple medications and other subclaims are not added as new participants.
- Reason for missing information: Two reports of the same trial, multiple medications and other subclaims are not added as new participants.
- Required materials and conditions: Randomization, analysis, group-specific absolute events and follow-up denominators
- Materials to obtain first: Two reports of the same trial, multiple medications and other subclaims are not added as new participants.
- Further research: Proposal: Obtain original participant flow, allocation, absolute event tables and missing-data mapping by group.
- Conditions for verification: Randomization, analysis, group-specific absolute events and follow-up denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N06 · Does the complete-prevention claim agree with the actual maintenance-group events?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The existing review summary and note record relapse within 52 weeks in 92/238 (39%) of the maintenance group. The preserved summary explicitly states that maintenance did not eliminate relapse.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Average reduction is not converted into complete prevention for every person or duration. No new individual prediction or guarantee is made.
- Reason for missing information: Average reduction is not converted into complete prevention for every person or duration. No new individual prediction or guarantee is made.
- Required materials and conditions: Direct evidence for the selected duration, group, relapse definition and event counts
- Materials to obtain first: Average reduction is not converted into complete prevention for every person or duration. No new individual prediction or guarantee is made.
- Further research: Proposal: Link original relapse definitions, ascertainment, events by duration and scope of individual applicability.
- Conditions for verification: Direct evidence for the selected duration, group, relapse definition and event counts
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N07 · What were the relapse symptom instrument, assessment and ascertainment times?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The preserved summary records relapse assessed with a depression symptom instrument as a patient-important treatment target and analyzes time to first relapse over 52 weeks.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not invent an instrument name, exact diagnostic criteria, measurement schedule or assessor details absent from the original material.
- Reason for missing information: Do not invent an instrument name, exact diagnostic criteria, measurement schedule or assessor details absent from the original material.
- Required materials and conditions: Instrument, episode definition, ascertainment, measurement times and group denominators
- Materials to obtain first: Do not invent an instrument name, exact diagnostic criteria, measurement schedule or assessor details absent from the original material.
- Further research: Proposal: Obtain original relapse definitions, instrument, assessments, ascertainment times and tables.
- Conditions for verification: Instrument, episode definition, ascertainment, measurement times and group denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N08 · Are 39% versus 56% and the absolute 17%p difference preserved as originally recorded?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Preserve the original wording: maintenance 92/238 (39%) versus discontinuation 135/240 (56%), an absolute difference of 17%p.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. No new proportions, risk differences, RR, NNT, absolute-difference CI or event examples are calculated. Average reduction is not complete prevention.
- Reason for missing information: No new proportions, risk differences, RR, NNT, absolute-difference CI or event examples are calculated. Average reduction is not complete prevention.
- Required materials and conditions: Relapsing participants, denominators, original proportions, absolute difference and uncertainty by group
- Materials to obtain first: No new proportions, risk differences, RR, NNT, absolute-difference CI or event examples are calculated. Average reduction is not complete prevention.
- Further research: Proposal: Obtain original absolute-difference, CI, denominator and test materials.
- Conditions for verification: Relapsing participants, denominators, original proportions, absolute difference and uncertainty by group
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N09 · Are the discontinuation-versus-maintenance HR direction, CI and P preserved?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The discontinuation-versus-maintenance HR is 2.06 (95% CI 1.56~2.70), P<0.001.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not invert this into a maintenance-versus-discontinuation HR or convert it into a relapse risk ratio or individual complete-protection rate. Detailed model and adjustment variables remain unread.
- Reason for missing information: Do not invert this into a maintenance-versus-discontinuation HR or convert it into a relapse risk ratio or individual complete-protection rate. Detailed model and adjustment variables remain unread.
- Required materials and conditions: Time-to-event effect, comparison direction, CI, test and model
- Materials to obtain first: Do not invert this into a maintenance-versus-discontinuation HR or convert it into a relapse risk ratio or individual complete-protection rate. Detailed model and adjustment variables remain unread.
- Further research: Proposal: Obtain the original effect of interest, HR direction, model, tests and denominators.
- Conditions for verification: Time-to-event effect, comparison direction, CI, test and model
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N10 · Are the 468 participants in the primary time-to-event analysis the same denominator as the 478 for absolute relapse counts?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The 10 participants without relapse timing, 6 maintenance and 4 discontinuation, were excluded from the primary time-to-event analysis but included in absolute relapse counts. The evidence list records 468 participants in the primary analysis.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not change the absolute-event denominator to 468 or newly certify the primary analysis as an all-participant ITT analysis. No new group-specific complete-case counts are calculated.
- Reason for missing information: Do not change the absolute-event denominator to 468 or newly certify the primary analysis as an all-participant ITT analysis. No new group-specific complete-case counts are calculated.
- Required materials and conditions: Mapping relapse timing, event occurrence, complete-case exclusions and denominators
- Materials to obtain first: Do not change the absolute-event denominator to 468 or newly certify the primary analysis as an all-participant ITT analysis. No new group-specific complete-case counts are calculated.
- Further research: Proposal: Obtain original analysis populations, missing timing, group denominators and event mapping.
- Conditions for verification: Mapping relapse timing, event occurrence, complete-case exclusions and denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N11 · Are the best- and worst-case sensitivity analyses and gaps in effect magnitude retained?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The preserved summary records that sensitivity analyses using best- and worst-case assumptions remained in the same direction.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Consistent direction is not used to certify resolution of all missing-data bias, an exact effect or complete follow-up. Detailed numbers remain unread.
- Reason for missing information: Consistent direction is not used to certify resolution of all missing-data bias, an exact effect or complete follow-up. Detailed numbers remain unread.
- Required materials and conditions: Missing-data assumptions, omissions by group, sensitivity models, effects and intervals
- Materials to obtain first: Consistent direction is not used to certify resolution of all missing-data bias, an exact effect or complete follow-up. Detailed numbers remain unread.
- Further research: Proposal: Obtain original sensitivity-analysis plans, assumptions, effects, intervals and denominators.
- Conditions for verification: Missing-data assumptions, omissions by group, sensitivity models, effects and intervals
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N12 · Are adherence of 70% and 52% and restarting of 39% distinguished from relapse proportions?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Adherence was 70% in maintenance and 52% in discontinuation; 39% of the discontinuation group restarted a clinician-prescribed antidepressant before trial end.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not merge 39% restarting in discontinuation with 39% relapse in maintenance as the same event. No new adherer effects, reasons for restarting or denominators are calculated.
- Reason for missing information: Do not merge 39% restarting in discontinuation with 39% relapse in maintenance as the same event. No new adherer effects, reasons for restarting or denominators are calculated.
- Required materials and conditions: Actual exposure, adherence definitions, restarting, duration and event relationships
- Materials to obtain first: Do not merge 39% restarting in discontinuation with 39% relapse in maintenance as the same event. No new adherer effects, reasons for restarting or denominators are calculated.
- Further research: Proposal: Obtain original exposure, adherence, restarting, timing, event mapping and effects of interest.
- Conditions for verification: Actual exposure, adherence definitions, restarting, duration and event relationships
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N13 · Has the magnitude of possible withdrawal-versus-relapse misclassification been established?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The preserved summary states that withdrawal symptoms were greater at 12 weeks in discontinuation and possible relapse misclassification could explain some findings, but its magnitude was not established.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not assert that withdrawal explains the entire difference or newly certify that there was no misclassification.
- Reason for missing information: Do not assert that withdrawal explains the entire difference or newly certify that there was no misclassification.
- Required materials and conditions: Distinguishing withdrawal and relapse, measurement, original models and sensitivity analyses
- Materials to obtain first: Do not assert that withdrawal explains the entire difference or newly certify that there was no misclassification.
- Further research: Proposal: Obtain original withdrawal and relapse definitions, timing, assessment, sensitivity and distinction analyses.
- Conditions for verification: Distinguishing withdrawal and relapse, measurement, original models and sensitivity analyses
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N14 · Have the original modified DESS means, difference, CI and direction been kept separate?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: At 12 weeks, modified DESS was 3.1 in discontinuation versus 1.3 in maintenance, with the original between-group difference 1.9 (95% CI 1.5~2.3).
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not subtract the means again to revise the recorded difference or convert it into a new P, CI, denominator or relapse count. Adjustment and instrument definitions remain unread.
- Reason for missing information: Do not subtract the means again to revise the recorded difference or convert it into a new P, CI, denominator or relapse count. Adjustment and instrument definitions remain unread.
- Required materials and conditions: Withdrawal instrument, time point, group order, original difference, interval and denominators
- Materials to obtain first: Do not subtract the means again to revise the recorded difference or convert it into a new P, CI, denominator or relapse count. Adjustment and instrument definitions remain unread.
- Further research: Proposal: Obtain original DESS measurements, model, recorded difference, group denominators and repeated time points.
- Conditions for verification: Withdrawal instrument, time point, group order, original difference, interval and denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N15 · Have matching-tablet blinding and limitations of actual allocation guesses been distinguished?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The existing review summary records a double-blind trial using matching tablets, but discontinuation participants guessed allocation better.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Nominal blinding alone does not certify complete actual blinding or resolution of all bias. Guessing proportions and allocation-concealment details remain unread.
- Reason for missing information: Nominal blinding alone does not certify complete actual blinding or resolution of all bias. Guessing proportions and allocation-concealment details remain unread.
- Required materials and conditions: Allocation, concealment, blinding, allocation guesses and analyst procedures
- Materials to obtain first: Nominal blinding alone does not certify complete actual blinding or resolution of all bias. Guessing proportions and allocation-concealment details remain unread.
- Further research: Proposal: Obtain original blinding, guessing, allocation, measurement and analysis procedures.
- Conditions for verification: Allocation, concealment, blinding, allocation guesses and analyst procedures
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N16 · Are the prespecified first-relapse endpoint, registration and analysis hierarchy actually linked?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The evidence-list endpoint is time to first relapse over 52 weeks, and the registration identifiers are ISRCTN15969819 · EudraCT2015-004210-26.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. No new registration or SAP reading or registration URL construction is performed. Complete prevention is not created as a separate prespecified endpoint of the original trial.
- Reason for missing information: No new registration or SAP reading or registration URL construction is performed. Complete prevention is not created as a separate prespecified endpoint of the original trial.
- Required materials and conditions: Registration, protocol, SAP, original primary endpoint, changes and multiplicity plans
- Materials to obtain first: No new registration or SAP reading or registration URL construction is performed. Complete prevention is not created as a separate prespecified endpoint of the original trial.
- Further research: Proposal: Obtain original registration, plans, analysis hierarchy and change history.
- Conditions for verification: Registration, protocol, SAP, original primary endpoint, changes and multiplicity plans
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N17 · Do gaps in secondary outcomes, medication or baseline-history subgroups and individual prediction remain?

Verification status: Unverified  
Reading scope: Summary read; external materials not read

- Current findings: The preserved summary states that the trial did not identify who could discontinue safely.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not create universal protection with maintenance, individual safe discontinuation, medication rankings or baseline-history effects.
- Reason for missing information: Do not create universal protection with maintenance, individual safe discontinuation, medication rankings or baseline-history effects.
- Required materials and conditions: Prespecified subgroups, interactions, denominators, effects and prediction materials
- Materials to obtain first: Do not create universal protection with maintenance, individual safe discontinuation, medication rankings or baseline-history effects.
- Further research: Proposal: Obtain original subgroup, interaction, individual-prediction and secondary-outcome plans and denominators.
- Conditions for verification: Prespecified subgroups, interactions, denominators, effects and prediction materials
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N18 · Is the relationship to the other claim about 44% without relapse in discontinuation retained?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The existing review summary and sibling subclaim preserve the wording that 44% of the discontinuation group did not relapse within 52 weeks.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. This value is not newly calculated or converted into advice that every person can or cannot discontinue safely.
- Reason for missing information: This value is not newly calculated or converted into advice that every person can or cannot discontinue safely.
- Required materials and conditions: Relationships among shared participants, relapse or non-relapse and other subclaims
- Materials to obtain first: This value is not newly calculated or converted into advice that every person can or cannot discontinue safely.
- Further research: Proposal: Link the two subclaims while retaining their relationship to the same trial, duration and denominators.
- Conditions for verification: Relationships among shared participants, relapse or non-relapse and other subclaims
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N19 · Has the 52-week effect been kept from extending to permanent prevention or other clinical outcomes?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The selected original materials' central duration is 52 weeks, and maintenance participants also relapsed.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. The 52-week reduction is not extended to lifelong absence of relapse, function, quality of life, survival or treatment benefit for every disease.
- Reason for missing information: The 52-week reduction is not extended to lifelong absence of relapse, function, quality of life, survival or treatment benefit for every disease.
- Required materials and conditions: Direct outcomes including long-term follow-up, episodes, function and deaths
- Materials to obtain first: The 52-week reduction is not extended to lifelong absence of relapse, function, quality of life, survival or treatment benefit for every disease.
- Further research: Proposal: Obtain direct relapse, clinical-function and long-term outcome materials separately by duration.
- Conditions for verification: Direct outcomes including long-term follow-up, episodes, function and deaths
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N20 · Do Kato's 40 studies, 8,890 participants and original responses concern the same population as ANTLER?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Kato pooled 40 studies and 8,890 participants from varied post-acute-treatment maintenance-versus-placebo-switch trials.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. This is not treated as the same direct comparison as the real discontinuation decision among long-term primary-care users. Populations, durations and overlap by included trial remain unread.
- Reason for missing information: This is not treated as the same direct comparison as the real discontinuation decision among long-term primary-care users. Populations, durations and overlap by included trial remain unread.
- Required materials and conditions: Review populations, post-acute maintenance, durations, denominators and trial mapping
- Materials to obtain first: This is not treated as the same direct comparison as the real discontinuation decision among long-term primary-care users. Populations, durations and overlap by included trial remain unread.
- Further research: Proposal: Obtain the review's original eligibility, included trials, durations, denominators and relationship to ANTLER.
- Conditions for verification: Review populations, post-acute maintenance, durations, denominators and trial mapping
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R03

### N21 · Are review results of 20.9% maintenance versus 39.7% placebo and the OR, CI and P preserved within their original scope?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The review reports maintenance 20.9% versus placebo switch 39.7%, OR 0.38 (95% CI 0.33~0.43), P<0.00001.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Maintenance 20.9% also does not mean complete prevention. The original OR is not converted into ANTLER's HR or an individual protection rate. No new proportions or differences are calculated.
- Reason for missing information: Maintenance 20.9% also does not mean complete prevention. The original OR is not converted into ANTLER's HR or an individual protection rate. No new proportions or differences are calculated.
- Required materials and conditions: Pooled relapse effects, group proportions, OR, intervals, tests and denominators
- Materials to obtain first: Maintenance 20.9% also does not mean complete prevention. The original OR is not converted into ANTLER's HR or an individual protection rate. No new proportions or differences are calculated.
- Further research: Proposal: Obtain original pooled models, trial-specific effects, denominators, definitions and heterogeneity.
- Conditions for verification: Pooled relapse effects, group proportions, OR, intervals, tests and denominators
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R03

### N22 · Have NEJM, the detailed NIHR report and the review been kept from being added as new independent trials?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: NEJM and NIHR HTA are two reports of the same ANTLER trial. The preserved summary states that the review supports the same direction but was not counted as an independent repeat of the ANTLER question.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Reports, medications, subclaims and the review's 40 studies are not added as new independent trials or participants. Final independent counts and total participant sums remain unrecorded.
- Reason for missing information: Reports, medications, subclaims and the review's 40 studies are not added as new independent trials or participants. Final independent counts and total participant sums remain unrecorded.
- Required materials and conditions: Relationships among original trials, additional reports, pooled materials, shared participants and sibling questions
- Materials to obtain first: Reports, medications, subclaims and the review's 40 studies are not added as new independent trials or participants. Final independent counts and total participant sums remain unrecorded.
- Further research: Proposal: Link additional reports through original identifiers and preserve relationships among shared populations and included trials.
- Conditions for verification: Relationships among original trials, additional reports, pooled materials, shared participants and sibling questions
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02, R03

### N23 · Have gaps in public funding, sponsorship, product supply, affiliations and external analysis been distinguished?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Preserve NIHR HTA 13/115/48, PRIMENT NIHR core support and UCL sponsorship. The preserved summary records that donated trial medication or placebo from a pharmaceutical company was not described in the NEJM or NIHR reports.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Original non-reporting is not converted into certification of no donated supply. Detailed company affiliations and external-analysis roles remain unread.
- Reason for missing information: Original non-reporting is not converted into certification of no donated supply. Detailed company affiliations and external-analysis roles remain unread.
- Required materials and conditions: Cash support, medication or placebo supply, affiliations, external analysis and design or writing roles
- Materials to obtain first: Original non-reporting is not converted into certification of no donated supply. Detailed company affiliations and external-analysis roles remain unread.
- Further research: Proposal: Obtain original support, supply, affiliations, research roles and analysis disclosures separately.
- Conditions for verification: Cash support, medication or placebo supply, affiliations, external analysis and design or writing roles
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N24 · Is the original information level for review funding and involvement in included trials retained?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The existing review summary records Japanese Ministry of Health, Labour and Welfare grants H29-SEISHIN-ippan-001 and 19GC1012 for the review.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Review support is not extended to certification that all 40 included studies had public funding, independence or no product supply.
- Reason for missing information: Review support is not extended to certification that all 40 included studies had public funding, independence or no product supply.
- Required materials and conditions: Review support, funding by original trial, authors and product or analysis roles
- Materials to obtain first: Review support is not extended to certification that all 40 included studies had public funding, independence or no product supply.
- Further research: Proposal: Obtain original review and trial-specific support, affiliations, products and analysis relationships.
- Conditions for verification: Review support, funding by original trial, authors and product or analysis roles
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R03

### N25 · Are serious-adverse-event participant counts, proportions and denominators distinguished from relapse?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Original safety information records serious adverse events in 9 maintenance participants (4%) versus 8 discontinuation participants (3%).
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not convert these into event counts, medication-related harms or statistically significant safety differences. No new differences, P, CI or proportions are calculated. Denominator, collection and relatedness details remain unread.
- Reason for missing information: Do not convert these into event counts, medication-related harms or statistically significant safety differences. No new differences, P, CI or proportions are calculated. Denominator, collection and relatedness details remain unread.
- Required materials and conditions: SAE definitions, participants versus events, group denominators, relatedness and exposure tables
- Materials to obtain first: Do not convert these into event counts, medication-related harms or statistically significant safety differences. No new differences, P, CI or proportions are calculated. Denominator, collection and relatedness details remain unread.
- Further research: Proposal: Obtain original SAE definitions, group-specific events and participants, denominators, relatedness, exposure and discontinuation.
- Conditions for verification: SAE definitions, participants versus events, group denominators, relatedness and exposure tables
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N26 · Have the original duration and scope of 0 deaths or suicide attempts been distinguished from long-term absence of risk?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Original safety information records 0 deaths and suicide attempts in both groups.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Zero events are not extended to absence of every rare harm, long-term safety or longevity benefit. No new risk CI or numerical examples are created.
- Reason for missing information: Zero events are not extended to absence of every rare harm, long-term safety or longevity benefit. No new risk CI or numerical examples are created.
- Required materials and conditions: Event definitions, collection, duration, exposure, denominators and rare-harm materials
- Materials to obtain first: Zero events are not extended to absence of every rare harm, long-term safety or longevity benefit. No new risk CI or numerical examples are created.
- Further research: Proposal: Obtain original event collection, exposure, denominators, duration and long-term rare-harm materials.
- Conditions for verification: Event definitions, collection, duration, exposure, denominators and rare-harm materials
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N27 · Do gaps remain in withdrawal symptoms, other maintenance-treatment harms and discontinuation applicability?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The preserved summary records increased withdrawal symptoms 12 weeks after discontinuation and possible misclassification of some relapse events. The parent safety status is Caution.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Withdrawal is not wholly equated with relapse or maintenance-medication harm. No long-term harm rates or maintenance or discontinuation instructions outside the original materials are created.
- Reason for missing information: Withdrawal is not wholly equated with relapse or maintenance-medication harm. No long-term harm rates or maintenance or discontinuation instructions outside the original materials are created.
- Required materials and conditions: Separate records of withdrawal, medication harms, discontinuation, exposure and long-term follow-up
- Materials to obtain first: Withdrawal is not wholly equated with relapse or maintenance-medication harm. No long-term harm rates or maintenance or discontinuation instructions outside the original materials are created.
- Further research: Proposal: Obtain original withdrawal, maintenance-medication harm, exposure, discontinuation and long-term materials separately.
- Conditions for verification: Separate records of withdrawal, medication harms, discontinuation, exposure and long-term follow-up
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02

### N28 · Have repeated harms and applicability to other medications, populations and exposures been kept from new certification?

Verification status: Unverified  
Reading scope: Summary read; external materials not read

- Current findings: Current materials compare maintenance with tapering to placebo among particular long-term users in remission; the review concerns varied maintenance trials.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Do not advise an individual to maintain, discontinue, taper or restart medication or certify complete prevention or safety for all medications and populations.
- Reason for missing information: Do not advise an individual to maintain, discontinue, taper or restart medication or certify complete prevention or safety for all medications and populations.
- Required materials and conditions: Direct effects or harms and report relationships by population, medication and duration
- Materials to obtain first: Do not advise an individual to maintain, discontinue, taper or restart medication or certify complete prevention or safety for all medications and populations.
- Further research: Proposal: Link direct comparisons and harm materials for new populations, medications and exposures separately.
- Conditions for verification: Direct effects or harms and report relationships by population, medication and duration
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02, R03

### N29 · Are the summary, three article URLs, registration records and original-material access distinguished?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: The complete parent verdict summary was read for production, and three PubMed URLs, original DOIs and registration strings were linked.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. No new URLs are constructed from identifiers. External abstracts, full texts, registrations, protocols and SAP remain unread, with no separate link test.
- Reason for missing information: No new URLs are constructed from identifiers. External abstracts, full texts, registrations, protocols and SAP remain unread, with no separate link test.
- Required materials and conditions: Individual bibliography, DOI, summary, external abstracts or full texts and registration or plan records
- Materials to obtain first: No new URLs are constructed from identifiers. External abstracts, full texts, registrations, protocols and SAP remain unread, with no separate link test.
- Further research: Proposal: Link original materials supplied by the user with individual types and actual access status.
- Conditions for verification: Individual bibliography, DOI, summary, external abstracts or full texts and registration or plan records
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02, R03

### N30 · Are the cutoff date, languages, empty values, original verdict, candidate history and production status preserved?

Verification status: Partly verified  
Reading scope: Summary read; external materials not read

- Current findings: Preserve 2026-08-26, empty correction and deliberation values, Korean and English drafts, parent B70, Caution and axes, and subclaim grade D. Original review-pending and production-eligibility markers are historical records.
- Why unverified: Only relevant preserved-summary information has been recorded; detailed original materials remain unread and existing content review is incomplete. Absent subclaim scores are not filled. Existing content review remains incomplete; review-pending activation and completion markers are false, new units are 0 and the final independent total is unrecorded.
- Reason for missing information: Absent subclaim scores are not filled. Existing content review remains incomplete; review-pending activation and completion markers are false, new units are 0 and the final independent total is unrecorded.
- Required materials and conditions: Distinguishing the complete preserved summary, subclaim record, parent values, corrections and current review
- Materials to obtain first: Absent subclaim scores are not filled. Existing content review remains incomplete; review-pending activation and completion markers are false, new units are 0 and the final independent total is unrecorded.
- Further research: Proposal: Reflect additional materials and decisions through preservation of original wording and separate events.
- Conditions for verification: Distinguishing the complete preserved summary, subclaim record, parent values, corrections and current review
- Why this condition is needed: Individual production condition recording the selected claim, original events, denominators, duration and interpretive boundaries
- Application exceptions: Unread external material, other questions or out-of-scope outcomes do not substitute for satisfied conditions, zero effect or safety.
- Remaining uncertainty: Unread/unverified does not mean zero effect, zero events, safety or new clinical certification.
- Referenced methods: Maintenance completely prevents relapse
- Related sources: R01, R02, R03

## Evidence and sources

### R01

Lewis G, Marston L, Duffy L, et al. Maintenance or Discontinuation of Antidepressants in Primary Care. N Engl J Med. 2021;385:1257-1267. PMID: 34587384. DOI: 10.1056/NEJMoa2106356.

[Publisher](https://doi.org/10.1056/NEJMoa2106356) · [PubMed](https://pubmed.ncbi.nlm.nih.gov/34587384/)

Materials actually read: Summary read; external materials not read  
Findings recorded in the summary: Relapse 92/238 (39%) versus 135/240 (56%); discontinuation-versus-maintenance HR 2.06 (95% CI 1.56~2.70), P<0.001; absolute difference 17%p  
Record level: Preserved summary; not the result of individually comparing the original texts of the two ANTLER reports  
Independent replication: NEJM and NIHR HTA reports of the same ANTLER trial share the same participants across three subclaims. The 40-study meta-analysis was not originally counted as an independent repeat of the ANTLER question; overlap among included trials and participants remains unread. Study-family identifiers link relationships and do not certify unique trial counts. Additional reports, medications, sibling questions, reviews and shared participants are not newly added as independent evidence.

| Material | Public availability | Access status | Actually read | Notes |
|---|---|---|---|---|
| local_summary | Existing link preserved | Summary read | Summary read | Stored parent-verdict summary |
| abstract | Existing link preserved | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| fulltext | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| registration | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| protocol_sap | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |

### R02

Duffy L, Clarke CS, Lewis G, et al. Antidepressant medication to prevent depression relapse in primary care: the ANTLER RCT. Health Technol Assess. 2021;25(69):1-62. PMID: 34842135. DOI: 10.3310/hta25690.

[Publisher](https://doi.org/10.3310/hta25690) · [PubMed](https://pubmed.ncbi.nlm.nih.gov/34842135/)

Materials actually read: Summary read; external materials not read  
Findings recorded in the summary: Relapse 92/238 (39%) versus 135/240 (56%); discontinuation-versus-maintenance HR 2.06 (95% CI 1.56~2.70), P<0.001; absolute difference 17%p  
Record level: Preserved summary; not the result of individually comparing the original texts of the two ANTLER reports  
Independent replication: NEJM and NIHR HTA reports of the same ANTLER trial share the same participants across three subclaims. The 40-study meta-analysis was not originally counted as an independent repeat of the ANTLER question; overlap among included trials and participants remains unread. Study-family identifiers link relationships and do not certify unique trial counts. Additional reports, medications, sibling questions, reviews and shared participants are not newly added as independent evidence.

| Material | Public availability | Access status | Actually read | Notes |
|---|---|---|---|---|
| local_summary | Existing link preserved | Summary read | Summary read | Stored parent-verdict summary |
| abstract | Existing link preserved | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| fulltext | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| registration | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| protocol_sap | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |

### R03

Kato M, Hori H, Inoue T, et al. Discontinuation of antidepressants after remission with antidepressant medication in major depressive disorder: a systematic review and meta-analysis. Mol Psychiatry. 2021;26:118-133. PMID: 32704061. DOI: 10.1038/s41380-020-0843-0.

[Publisher](https://doi.org/10.1038/s41380-020-0843-0) · [PubMed](https://pubmed.ncbi.nlm.nih.gov/32704061/)

Materials actually read: Summary read; external materials not read  
Findings recorded in the summary: Maintenance 20.9% versus placebo switch 39.7%, OR 0.38 (95% CI 0.33~0.43), P<0.00001  
Record level: Preserved summary; not the result of individually comparing the original texts of the two ANTLER reports  
Independent replication: NEJM and NIHR HTA reports of the same ANTLER trial share the same participants across three subclaims. The 40-study meta-analysis was not originally counted as an independent repeat of the ANTLER question; overlap among included trials and participants remains unread. Study-family identifiers link relationships and do not certify unique trial counts. Additional reports, medications, sibling questions, reviews and shared participants are not newly added as independent evidence.

| Material | Public availability | Access status | Actually read | Notes |
|---|---|---|---|---|
| local_summary | Existing link preserved | Summary read | Summary read | Stored parent-verdict summary |
| abstract | Existing link preserved | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| fulltext | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| registration | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |
| protocol_sap | No separate URL in the original material | Not read | Not read | External material unread; a URL does not mean acquisition, reading or completed review |

## Review criteria and methods

These are Chamgap's own application conditions, informed by the relevant sections of official methodological guidance. They do not constitute official assessment-tool certification or a new efficacy score.

- M03 [Cochrane Handbook 6.5 (2024), chapter updated 2023-08](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-03) · §3.1, §3.2.1–3.2.3, MECIR C5/C7/C13 · Clearly record the question, eligibility conditions and reasons for changes, and distinguish variations in interventions. · Preserved methodology connections were reused for preparation. No new review of methodological content, reassessment or external-link checks were performed.
- M04 [Cochrane Handbook 6.5.1 (2025), chapter updated 2025-03](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-04) · §4.5, §4.6.1–4.6.3, MECIR C42 · Distinguish studies from reports and link them; do not omit a study solely because results are unreported. · Preserved methodology connections were reused for preparation. No new review of methodological content, reassessment or external-link checks were performed.
- M08 [Cochrane Handbook 6.5 (2024), chapter updated 2019-10](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08) · §8.1–8.2.3, Table 8.2.a · Define the outcome and effect of interest, then examine randomization, deviations, missing data, measurement and selective reporting. · Preserved methodology connections were reused for preparation. No new review of methodological content, reassessment or external-link checks were performed.
- M15 [Cochrane Handbook 6.5 (2024), chapter updated 2023-08](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-15) · §15.2.4, §15.3.1–15.3.3 · Distinguish effect estimates, intervals, importance and applicability. Sample size alone does not determine precision. · Preserved methodology connections were reused for preparation. No new review of methodological content, reassessment or external-link checks were performed.
- M19 [Cochrane Handbook 6.5 (2024), chapter updated 2019-10](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-19) · §19.1.2, §19.2.1–19.2.4 · Harm questions may require different designs, populations and follow-up from efficacy questions. Case reports do not establish incidence rates. · Preserved methodology connections were reused for preparation. No new review of methodological content, reassessment or external-link checks were performed.

## What the statuses mean

- Partly verified: Some information exists in the preserved review summary; detailed original materials remain unverified. This does not certify a passed review or clinical efficacy.
- Unverified: Detailed materials for the selected condition could not be established within the stored scope; this does not mean zero effect, zero studies or safety
- Review incomplete: Existing content review pending
- Met: Use only within the relevant conditions after existing content review; no automatic designation in this preparation
- Not reported: Distinguish only within materials actually read; do not turn unread external materials into unreported findings
- Inaccessible: Not applied during this preparation because actual access was not tested
- Not applicable: Use with a reason for non-applicability based on an explicit question boundary
- Conflicting data: Judge after actual material comparison; review was not performed during this preparation
- Insufficient evidence for the conditions: Do not use as a clinical grade encompassing currently unread materials

## Change history

2026-10-11 · Public page/materials 1.0. The public page and data were prepared from the existing 30 items. No new clinical search, verification-status change or scoring was performed.