AEGIS
// LIFE SCIENCES
SIGNAL · 4.2σ · BATTERY · BATCH B-2026-04-AU
tomas.ribeiro@medtronic
STAGE 08 · POST-MARKET SURVEILLANCE
ORACLE detected a 4.2σ battery-fault cluster in Sydney. Sarah needs to call FSCA.
This stage runs for the commercial life of LINQ-NEXT — 10-year horizon, perpetual signal monitoring. ORACLE ingests 4 parallel data streams (PMCF outcomes · field complaints · vendor batch QA · post-market literature) and runs continuous statistical anomaly detection. Patient-level data tiers from hot (active monitoring) to warm (1-2 yr) to cold (3-10 yr) per Medtronic data retention policy. Stage instantiated on first sale (HK · 12 May 2026).
19Devices live
11Complaints 30d
1Active signal
0FSCA filed
!
SIGNAL DETECTED · 4.2σ · FIELD SAFETY CORRECTIVE ACTION DECISION
Sydney channel · battery-end-of-life arriving 28-34% earlier than spec on batch B-2026-04-AU.
PMS-SIG-2026-0008 · ORACLE flagged 17 May 22:14 SGT · TGA notification clock: 4d 8h
4.2σ
Anomaly strength
p < 0.0001 vs baseline
7 of 23
Affected units
batch B-2026-04-AU
28-34%
Early EOL signal
vs 4.5-yr spec
0
Patient harm
no AEs · no SAEs
Battery health telemetry · 28-day rolling Z-score · 19 active devices · batch overlay
+3σ threshold+4σ critical4.2σ · 13 May17 Feb19 Mar17 Apr13 May17 MayHK liveSG liveAU pre-load+4σ+2σμ−2σ
HK + SG baseline · within 1σ band
Sydney pre-launch channel-load (n=23) · diverging
⬢ ORACLE · signal analysisOn 17 May 22:14, the 28-day rolling Z-score on battery-end-of-life telemetry for batch B-2026-04-AU crossed +4σ (p < 0.0001 vs the locked baseline of 4.5-year EOL). 7 of 23 channel-loaded units in Sydney are tracking 28-34% early EOL signal at 60-day post-test-cycle telemetry. No patients are implanted with this batch yet — Australian launch is 28 May. HK and SG units (different batches: B-2026-03-HK, B-2026-03-SG) are within 1σ of baseline. Vendor traceability points to lithium cell vendor change in Q1 2026 (DEF-PMDA-2026-0120 in Stage 05 flagged this for biocompatibility). The new vendor's cells may have higher self-discharge under the 37°C body-temp simulation. Recommend QA hold on the batch and FSCA notification to TGA within the statutory 5-day window per POL-REPORTABLE-EVENT-V2.
Affected batch
B-2026-04-AU · 87 units manufactured · 23 channel-loaded Sydney pre-launch · 64 still in Memphis warehouse
Root cause hypothesis
Q1 2026 lithium cell vendor change · supplier: Tadiran TLI-1530A → Eagle Picher LTC-7P · self-discharge 1.8× spec at 37°C
Patient impact
Zero patient harm · no implants from this batch · no AEs · no SAEs · pre-launch QA caught before sale
Statutory clock
TGA notification within 5 business days of confirmed signal · clock started 17 May 22:14 · due 24 May 22:14
⬢ ORACLE · 3 paths surfaced for Sarah Chen (CRO) — recall authority
PATH A · QA HOLD · ORACLE RECOMMENDS
QA hold the batch · notify TGA voluntary FSCA
Pull 23 channel-loaded units from Sydney inventory + 64 from Memphis. File voluntary FSCA with TGA on day 5. Notify HSA & HK proactively (different batches but transparency). No patient communication needed (zero implants). Investigate vendor cell. Replace batch from B-2026-03-SG production. Australian launch slips ~2 weeks.
Launch delay~2 weeks
Revenue impact−$1.2M Q2
Brand trustPROTECTED
PATH B · MONITOR
Continue monitoring · no FSCA · QA hold pending more data
Hold the batch but do not file FSCA yet. Wait 14 days for additional telemetry to confirm pattern. Risk: TGA statutory clock runs whether or not we file. If TGA detects the signal independently (they receive identical telemetry under MDR 5.3.1), unreported event = compliance violation. If signal turns out spurious, we file nothing.
Launch delay~3 weeks if confirmed
Compliance riskHIGH
Brand trustEXPOSED
PATH C · FULL RECALL
Pre-emptive full recall of all B-2026-* batches
Recall all batches from Q1-Q2 vendor change (87 + 14 HK + 5 SG = 106 units). 19 patients with implanted devices (HK 14, SG 5) receive proactive replacement offers. Voluntary worldwide FSCA. Three-region launch pause. Strongest patient-safety posture but adds Class II recall to LINQ-NEXT's regulatory history (visible to PMDA, KFDA reviewers).
Launch delay~3 months
Recall cost−$3.8M direct
Reg historyCLASS II RECALL
POL-REPORTABLE-EVENT-V2 · policy://aegis.medtronic/pms/reportable-event/v2
"Any signal exceeding +3σ from locked baseline on a safety-critical parameter triggers mandatory CRO review within 24 hours and FSCA decision within 5 business days. Decision authority for FSCA filing and product recalls rests exclusively with the CRO. ORACLE may model, recommend, and pre-stage notifications but cannot file."
Sarah Chen (CRO) decision authority · TGA notification due in 4d 8h · Anika Reddy (VP Quality) acknowledgment required
4 parallel surveillance streams
always-on · statistical anomaly detection · 28-day rolling baselines
PMCF outcomes
NORMAL
19
In cohort
455
Target n
94.7%
AF sens · early
0
Signals
First 19 implants tracking AF detection sensitivity within 0.6 percentage points of locked CER v17 baseline (95.3%). 3-year follow-up window per PMCF v3 protocol. Cohort recruits as launches roll out across 7 markets.
Field complaints
NORMAL
11
30d total
2
Moderate
9
Minor
0
Serious
Complaints rate: 0.58 / device · month, vs Reveal LINQ legacy baseline 0.71. Within expected envelope. 2 moderate: 1 insertion-site irritation (resolved), 1 connectivity glitch (firmware patch shipped 14 May).
Vendor batch QA
SIGNAL 4.2σ
12
Batches 90d
1
Flagged
87
Units at risk
0
Implanted
PMS-SIG-2026-0008 · Batch B-2026-04-AU battery EOL 28-34% early. Channel-loaded Sydney units not yet implanted. Vendor change (Tadiran → Eagle Picher) Q1 2026 root cause hypothesis under investigation.
Literature monitoring
ELEVATED
38
Papers 90d
3
Critical
14
Cited LINQ-NEXT
2
Competitor
3 critical: Abbott Confirm Rx launched APAC 12 May (competitive), 2024 ESC consensus paper recommending broader ICM use in cryptogenic stroke (supportive), Korean Heart Journal published HSA-cohort data (visibility).
Field complaints · last 30 days · 11 total
2 moderate · 9 minor · 0 serious · 0 MDR-eligible filed
PMS-CMP-0011AUPre-launch QA — early battery EOL telemetry · 7 of 23 channel-loaded unitsB-2026-04-AUSIGNALPENDING17 MayT. Ribeiro
PMS-CMP-0010SGBluetooth pairing intermittent · patient phone iPhone 16 Pro · firmware patch v2.4B-2026-03-SGMODN/A14 MayT. Ribeiro
PMS-CMP-0009HKInsertion site mild erythema · resolved with topical · no infectionB-2026-03-HKMODN/A13 MayP. Nair
PMS-CMP-0008HKData sync delay 4-6 hr · uplink server load · resolved infra-sideB-2026-03-HKMINORN/A13 MayT. Ribeiro
PMS-CMP-0007HKPatient request — quieter operation expected · documentation expectation gapB-2026-03-HKMINORN/A12 MayT. Ribeiro
PMS-CMP-0006HKCompanion app login failure · OAuth refresh bug · patched v2.4.1B-2026-03-HKMINORN/A12 MayT. Ribeiro
PSUR · Periodic Safety Update Report
NEXT IN 184d
First PSUR17 Nov 2026 · 180-day
Annual after12 Nov each year through 2036
AuthoritiesPMDA · KFDA · TGA · HSA
Content lockPMCF + complaints + sales + literature
DraftingORACLE auto-drafts 14d before due
Sign-offA. Reddy · VP Quality
PMCF interim · 3-year follow-up cohort
YEAR 1 DUE 11 May 2027
Target cohort455 patients (WHO 2024 update)
Enrolled312 of 455 · 4.2%
Primary endpointAF sens at 36mo · real-world
Locked baseline95.3% · CER v17
Interim Y111 May 2027
Final Y311 May 2029
WORKFLOWORACLE's perpetual loop · 4 streams continuouslystream → baseline → anomaly check → escalate or log · 10-yr commercial horizon
Unlike sequential stages, PMS runs four parallel always-on detection loops for the commercial life of LINQ-NEXT. Each loop has its own statistical baseline locked from CER + Stage 02 reference data. Anomaly detection runs every 6 hours on hot data, daily on warm, weekly on cold.
STEP 1 · INGEST
4 data streams
PMCF telemetry · field complaints · vendor batch QA · post-market literature.
continuous
STEP 2 · BASELINE
Locked references
28-day rolling Z-score against CER v17 baselines + Reveal LINQ legacy norms.
~6 hr cycle
STEP 3 · DETECT
Statistical anomaly
+3σ → flag & review. +4σ → mandatory CRO escalation within 24h.
automatic
STEP 4 · DECIDE
CRO sign-off path
ORACLE pre-stages 3 paths. Sarah decides FSCA / monitor / recall.
5 biz days
STEP 5 · CLOSE
Action + receipt
FSCA filed · CAPA opened · baseline updated · receipt chain sealed.
never seals
DATA RETENTION TIERS
Hot (0-12 months) — full patient-level data, 6-hour anomaly cycles. Warm (1-2 years) — aggregated, daily cycles. Cold (3-10 years) — statistical summaries only, weekly cycles. All tiers retained for the full 10-year regulatory window per POL-PMS-RETENTION-V1.
Stage ThreadORACLE · SIGNAL
08 PMS · 68 events · 1 CRO escalation
ORACLE → ESCALATE17 May · 22:14
PMS-SIG-2026-0008 escalated to CRO. Sydney pre-launch channel-load batch B-2026-04-AU exceeded +4σ on battery-end-of-life telemetry. 7 of 23 channel-loaded units in Sydney showing 28-34% early EOL signal at 60-day post-test telemetry. No patients implanted. Vendor traceability indicates Q1 2026 lithium cell vendor change (Tadiran TLI-1530A → Eagle Picher LTC-7P) as root-cause hypothesis. New supplier cell self-discharge 1.8× spec at body-temp simulation. TGA statutory notification clock: 5 business days, due 24 May. Three paths surfaced: QA hold + voluntary FSCA (recommended), monitor only (risky), full recall (over-correcting). Sarah, per POL-REPORTABLE-EVENT-V2, this is non-delegable.
8e21...9c44 · CRO ESCALATE
ORACLE17 May · 22:08
Pre-staged TGA voluntary FSCA notification draft. Pre-staged HSA and HK courtesy notification drafts (different batches but transparency principle per Medtronic global PMS policy). Pre-staged CAPA opening template referencing Stage 05 DEF-PMDA-2026-0120 which flagged the same vendor change for biocompatibility — that was a different angle but same vendor change.
ORACLE13 May · 14:08
Batch B-2026-04-AU rolling 28-day Z-score on battery EOL telemetry crossed +3σ. 4 units flagged. Initiated continued monitoring per POL-REPORTABLE-EVENT-V2. Today (17 May), additional units brought the cluster to 7 of 23 channel-loaded units and Z crossed +4σ. Triggered escalation.
ORACLE12 May · 14:15
Stage 08 instantiated. HK first sale event received. PMCF cohort enrollment begins. 4 surveillance streams armed: PMCF outcomes, field complaints, vendor batch QA, post-market literature. Baselines locked from CER v17 + Reveal LINQ legacy norms. 10-year horizon initiated.