Skip to content

Backup Comparison: 21 Feb 2026 Snapshot vs Current Production

Note: This is a temporary planning document that will be deleted when the investigation is complete.

Project: little DOMS (ac35058c-4767-4f9a-9ddc-5ba9d484c289) Date: 2026-03-07 Backup source: Atlas weekly snapshot 21 Feb 2026, restored to local container (mcp__mongodb-syrf-local) Production: Live syrftest database on prod Atlas cluster (mcp__mongodb-syrf-prod)


Method

Queried both databases for Althea's (CSUUID F5XWaUIex06j3KycXnpSUA==) annotations on the 6 affected studies, filtering to the 17 shared RoB question IDs. Compared answer values for each study × question combination.

The backup predates both re-submission events (25 Feb, 5 Mar 2026) and captures the state immediately after Althea completed Stage 2 (18–19 Feb).

Note on StageId: The StageId field on annotations records which stage the user was in when the annotation was created (provenance). Annotations do not belong to stages — sessions do.


High-Level Result

Metric Count
Stage 2 provenance annotations in backup 110
Stage 2 provenance annotations in current prod 0
RoB judgments compared ~78
RoB judgments changed 8
RoB annotations entirely lost (not re-created) 3
RoB judgments matching ~67

RoB Question ID Reference

Study-level items (one per study)

Short Name QuestionId (base64) RoB Item
Item 1 x0qAlFTE30SKuwcTf6VKbg== Selection bias: Sequence generation
Item 2 s3lHskqD0U+E6epC/TLzyQ== Selection bias: Baseline characteristics (shared variant)
Item 3 j0xK+34eDEuD9XcktXvDvQ== Selection bias: Allocation concealment
Item 4 iTiQBo9jrkSiRU2S/Zr5qg== Performance bias: Random housing
Item 5 J+Lhoi/piEW1fuCJDxGhgQ== Performance bias: Blinding
Item 9 T37in9k7mUW+qrBDBLyPug== Reporting bias: Selective outcome reporting
Item 10 xZm1vfctOkm7IrH+Clmn+w== Other sources of bias

Per-outcome items (one per outcome assessed)

Two sets of Items 6/⅞ exist — one "study-level" set and one "per-outcome" set with different QuestionIds:

Short Name QuestionId (base64) RoB Item
Item 6-SL 8KWtDbzlHEWjRiP/dYOxdA== Detection bias: Random outcome assessment (study-level)
Item 7-SL VsK19fbUIUOdMF2G0XXStg== Detection bias: Blinding (study-level)
Item 8-SL gYefGdoJIkGhL85yzOpEMw== Attrition bias: Incomplete outcome data (study-level)
Item 6-PO n3GUowpTSkmZVULHrc5/qw== Detection bias: Random outcome assessment (per-outcome)
Item 7-PO HADII2nna0CbQ1yyHpxJTA== Detection bias: Blinding (per-outcome)
Item 8-PO z+/34B1xEUyppzwZQbFQog== Attrition bias: Incomplete outcome data (per-outcome)

Other shared annotations

Short Name QuestionId (base64) Purpose
Outcome name 7+yzsWHUSUyCOcyt8fr55Q== "What is the name of the neurobehavioral outcome?"
Outcome label oIb48WLhUkq4N5PoK+74AQ== Outcome assessment procedure label
Outcome label 2 DHLi2wguR0+80D/krouMfw== Outcome label (full format)
Reconciliation LJJdVgjvsEuYFC6ooFHMNA== "Is this study reconciled" (BoolAnnotation, always false)

Per-Study Comparison

1. Contralesional angiotensin (20 Stage 2 annotations in backup)

Result: ALL MATCH

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 High High Yes
Item 10 Low Low Yes
Item 6-PO (Rotarod) Unclear Unclear Yes
Item 7-PO (Rotarod) Unclear Unclear Yes
Item 8-PO (Rotarod) Unclear Unclear Yes
Item 6-PO (Bederson) Unclear Unclear Yes
Item 7-PO (Bederson) Unclear Unclear Yes
Item 8-PO (Bederson) Unclear Unclear Yes

Notes: No SL variants (8KWt/VsK1/gYef) exist in either backup or prod for this study. Item 9 is notably "High" (with Notes referencing "data not shown" quote) — correctly preserved.


2. NLRP3 Inflammasome (13 Stage 2 annotations in backup)

Result: 1 CHANGED

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 Unclear Unclear Yes
Item 10 Low Low Yes
Item 6-SL Unclear Unclear Yes
Item 7-SL Unclear Unclear Yes
Item 8-SL Low High NO

Notes: Item 8-SL flipped from Low to High. Backup notes: "Was the number/percentage AND reason for dropout/mortality described PER GROUP? YES". Current notes likely say NO. Only 1 outcome (Infarct volume) so no PO items beyond SL.


3. Low-dose nifedipine (22 Stage 2 annotations in backup)

Result: ALL MATCH

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 Unclear Unclear Yes
Item 10 Low Low Yes
Item 6-SL Unclear Unclear Yes
Item 7-SL Unclear Unclear Yes
Item 8-SL Unclear Unclear Yes
Item 6-PO ×3 Unclear Unclear Yes
Item 7-PO ×3 Unclear Unclear Yes
Item 8-PO ×3 Unclear Unclear Yes

Notes: 3 outcomes (Infarct volume, Beam Test, Body Asymmetry Test). All Unclear across the board — fully consistent.


4. Huang-Lian-Jie-Du (17 Stage 2 annotations in backup)

Result: 1 CHANGED + 3 LOST

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 Unclear Unclear Yes
Item 10 Low Low Yes
Item 6-SL Unclear MISSING LOST
Item 7-SL Unclear MISSING LOST
Item 8-SL High MISSING LOST
Item 6-PO Unclear Unclear Yes
Item 7-PO Unclear Unclear Yes
Item 8-PO Low High NO

Notes: Item 8-PO flipped from Low→High. Backup notes: "Did the numbers of animals that started the experiment match the number of animals analyzed? YES". Current notes: same question but answer "NO". The SL variants (Items 6-SL, 7-SL, 8-SL) were not re-created during Stage 1 re-submission — entirely absent from current prod.


5. New alternative approaches (19 Stage 2 annotations in backup)

Result: 6 CHANGED (most affected study)

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 Unclear Unclear Yes
Item 10 Low Low Yes
Item 6-SL High Unclear NO
Item 7-SL Low Unclear NO
Item 8-SL Unclear Unclear Yes
Item 6-PO (Infarct) High Unclear NO
Item 7-PO (Infarct) Low Unclear NO
Item 8-PO (Infarct) Unclear Unclear Yes
Item 6-PO (Neuro) High Unclear NO
Item 7-PO (Neuro) Low Unclear NO
Item 8-PO (Neuro) Unclear Unclear Yes

Notes: This is the most significant study. Althea originally made specific judgments: - Item 6 (random outcome assessment): High — "Was randomization described in the paper? NO" - Item 7 (blinding detection): Low — "Was the described method of blinding appropriate? YES"

These were specific, evidenced assessments. In her re-entry, all were flattened to "Unclear", representing genuine loss of analytical precision.


6. DPP-4 Linagliptin (19 Stage 2 annotations in backup)

Result: ALL MATCH

RoB Item Backup Current Match?
Item 1 Unclear Unclear Yes
Item 2 Low Low Yes
Item 3 Unclear Unclear Yes
Item 4 Unclear Unclear Yes
Item 5 Unclear Unclear Yes
Item 9 Unclear Unclear Yes
Item 10 Low Low Yes
Item 6-SL Unclear Unclear Yes
Item 7-SL Unclear Unclear Yes
Item 8-SL Unclear Unclear Yes
Item 6-PO ×2 Unclear Unclear Yes
Item 7-PO ×2 Unclear Unclear Yes
Item 8-PO ×2 Unclear Unclear Yes

Notes: 2 outcomes (Infarct volume, Adhesive Removal Test). All values match.


Summary

What changed

# Study Items Changed Direction Significance
1 New alt. approaches Item 6-SL, 6-PO×2 High → Unclear Lost specific "no randomization" judgment
2 New alt. approaches Item 7-SL, 7-PO×2 Low → Unclear Lost specific "blinding appropriate" judgment
3 NLRP3 Item 8-SL Low → High Attrition rating flipped
4 Huang-Lian-Jie-Du Item 8-PO Low → High Attrition rating flipped, notes contradict

What was lost entirely

Study Items Original Values
Huang-Lian-Jie-Du Items 6-SL, 7-SL, 8-SL Unclear, Unclear, High

Pattern

  • Items 1–5, 9, 10 (study-level): 100% match across all 6 studies — Althea reliably recalled these
  • Items 6, 7, 8 (per-outcome/SL): where ALL changes and losses occurred — exactly the outcome-level items Althea reported as "disappeared" in issue #2337
  • Changes go in one direction: specific judgments (High/Low) → Unclear. This is consistent with a reviewer who can no longer see the original evidence and defaults to "Unclear"
  • Two attrition items flipped Low→High, with notes showing the underlying factual assessment reversed ("YES"→"NO")

Conclusion

The backup must be used for data recovery. Althea's re-entered values do not match her original Stage 2 work: - 8 RoB judgments have different answers - 3 annotations are entirely missing - The changes affect analytical precision — specific evidence-based judgments were replaced with "Unclear"

The 21 Feb Atlas backup contains the original Stage 2 annotations with correct provenance and Althea's original RoB assessments. Recovery procedure is documented in the main investigation doc (little-doms-investigation.md).