SUS Lot-Genealogy Monitor — CPV Bridge for Single Use Systems

Pressure-Decay Integrity Test Capability — by SUS Assembly SKU

Histogram of pre-use post-sterilization integrity test results · Spec USL 5.0 mbar/30 min · Test method per documented porous-fleece-spacer methodology (Sartorius BioProcess International)

Sample Size (n) —

Mean (mbar) — σ = —

Cpk — target ≥ 1.33

Ppk — long-term capability

Supplier Scorecard — Q1 / Q2 2026 Snapshot

First-time-right rate · Lot acceptance · Mean days to disposition · Ranked composite score

Supplier signal — automatically detected Cytiva first-time-right rate dropped from 94% (Q2 2025) to 84% (Q1 2026). Disposition time doubled. Root-cause investigation tied to a film resin lot change introduced Sept 2025 — supplier change notification SCN-CYT-2025-0114.

Extractables & Leachables — Shewhart Control Chart, Sum of Extractables (ppm)

All qualified SUS lots, time-ordered by qualification date · CL ± 3σ derived from baseline 18-month window · USL 400 ppm per internal toxicological assessment · Per USP <665>

CUSUM — Cumulative Deviation from E&L Target

Tabular CUSUM, k=0.5σ, h=4σ · Detects sustained small shifts that Shewhart can miss

Hotelling T² — Multivariate Monitor of All SUS Quality Signals

Combines pressure-decay · E&L · gamma dose · days-to-disposition · supplier · UCL based on F-distribution α = 0.01

Lot Genealogy — Recent SUS Qualification Records

30 most-recent qualified or under-investigation SUS lots · Sortable, drillable to underlying integrity-test traces and E&L analytical reports

Lot ID	Supplier	SKU	Qual Date	Integrity (mbar)	E&L Sum (ppm)	γ Dose (kGy)	Days to Disp.	Status	Owner

Methodology · The SUS data programme an MSAT team should run from day one

The premise. Single Use Systems generate a continuous stream of lot-genealogy data — extractables & leachables, pre-use post-sterilization integrity tests, gamma-sterilization dose verification, supplier change notifications, days-to-disposition. Most sites only stitch this stream together after something fails. The SUS Lot-Genealogy Monitor stitches it from day one and surfaces drift before it becomes a deviation.

Data Pipeline

Bronze layer ingests SUPPLIER COA PDFs (via OCR for legacy, structured JSON for modern suppliers), INTEGRITY TEST traces from the test stations (HOLD pressure-decay logs), GAMMA SUBSTANTIATION reports per ISO 11137-2:2013, and QMS disposition records. Silver layer normalizes everything to one row per lot with the full lot-genealogy join (assembly SKU → bag SKU → film batch → resin batch). Gold layer is what this dashboard reads.

Pressure-Decay Capability

Cpk = min[(USL−μ)/3σ, (μ−LSL)/3σ] using within-subgroup σ. Ppk uses overall σ. For a one-sided spec (USL only), Cpk simplifies to (USL−μ)/3σ. Spec method: testing must use a porous fleece spacer against the bag, not a smooth plate — smooth-plate testing masks defects (Sartorius / BioProcess International). Detection limit: ~100 µm for a 200 L bag.

E&L Shewhart Control Chart

Individual values chart (n=1 per lot). CL = mean of baseline period. UCL/LCL = CL ± 3σ where σ is estimated from average moving range / d₂ (1.128 for n=2). Nelson rules 1-4 applied for special-cause detection. Spec USL set by toxicological assessment per USP <665> and the internal Permitted Daily Exposure (PDE) model.

CUSUM

Tabular CUSUM with reference value k = 0.5σ and decision interval h = 4σ. Detects sustained small shifts (≤ 1.5σ) that Shewhart misses. Reset to zero after each signal. The CUSUM signal in late Q4 2025 led the Shewhart signal by approximately two lots — typical Shewhart-vs-CUSUM relationship.

Hotelling T²

T² = (x − x̄)·S⁻¹·(x − x̄)' across p = 5 variables. UCL based on F-distribution at α = 0.01: UCL = [p(N+1)(N−1) / N(N−p)] × F_α,p,N−p. When a point exceeds UCL, the dashboard surfaces the contributing-variables decomposition so investigators see which dimension drove the multivariate signal. Paired with SPE (Q-statistic) for residual-space monitoring.

Supplier Scorecard

Composite rank score = 0.40 × FTR + 0.30 × lot acceptance + 0.20 × (1 − normalized disposition time) + 0.10 × E&L compliance. Weights configurable per site QMS policy. Rolling 6-month windows. Trend arrows indicate the direction relative to the prior period — drives the procurement & quality engagement cadence.

GAMP 5 Posture

Treated as Category 5 (custom configured application). URS specifies the CPV plan it serves. Risk assessment per ICH Q9(R1). DAX measures and Python pipeline versioned in Git, IQ on the runtime, OQ on the transformations, PQ on historical lots with known dispositions. Power BI and Databricks tenant audit logs preserve the change history. Source of truth is the gold layer — the dashboard never edits data.

Synthetic data disclosure

The 18 months of lot-genealogy data, supplier identifiers, lot IDs, integrity-test values, E&L results, gamma doses, and disposition records shown above are fabricated for portfolio demonstration purposes. The dataset is seeded with a deliberate Q4 2025 Cytiva film-resin shift so the demo produces a coherent, narratable signal across all four detection layers (Shewhart, CUSUM, T², supplier scorecard). No real GSK, Sartorius, Cytiva, or Thermo Fisher data is present.