Antimicrobial susceptibility testing — MIC, breakpoints and data pipelines

Antimicrobial susceptibility testing (AST) is the laboratory result that directly drives whether a clinician prescribes ampicillin or a carbapenem. A MIC reported one dilution too high can mean an ineffective antibiotic for a septic patient; one dilution too low can mean undertreating a resistant organism. The margin for error is narrow and the data pipeline is rarely automated end-to-end. This piece covers the analytical, QC, and reporting discipline that keeps AST results defensible.

MIC determination — the core measurement

The minimum inhibitory concentration (MIC) is the lowest antibiotic concentration that visibly inhibits growth after 16–20 h incubation. Broth microdilution (BMD) per CLSI M07 / EUCAST E.Def 9.2 is the reference method; disk diffusion (Kirby-Bauer) and automated systems (VITEK 2, BD Phoenix, Sensititre) are calibrated against BMD in their breakpoint derivation.

BMD execution checklist

• Inoculum standardization: 0.5 McFarland turbidity (1–5 × 10⁸ CFU/mL), verified with a calibrated densitometer. Final well concentration 5 × 10⁵ CFU/mL ± 50%.
• Antibiotic dilution series: two-fold serial dilutions in cation-adjusted Mueller-Hinton Broth (CAMHB). For fastidious organisms (Streptococcus, Haemophilus) use the appropriate supplemented medium specified by EUCAST.
• Growth control: one well without antibiotic per plate to confirm growth.
• Sterility control: one well without inoculum to detect contamination.
• Reading: optical at OD600 or visual turbidity; record the exact dilution value, not categorical S/I/R.

QC reference strains — the non-negotiable

Every batch run must include a reference strain with CLSI/EUCAST-defined QC ranges per antibiotic–species combination. Common examples:

• Staphylococcus aureus ATCC 29213 — oxacillin, vancomycin, gentamicin control.
• Escherichia coli ATCC 25922 — ampicillin, ciprofloxacin, trimethoprim control.
• Pseudomonas aeruginosa ATCC 27853 — imipenem, meropenem, ceftazidime control.
• Enterococcus faecalis ATCC 29212 — vancomycin, linezolid, ampicillin control.

A QC MIC outside the published range invalidates the entire run — not just that organism. Trending QC MICs over time with a Levey-Jennings chart detects drift weeks before a failure invalidates a patient batch.

Breakpoint interpretation — EUCAST vs CLSI

Breakpoints are not fixed numbers — they change with EUCAST and CLSI annual updates. A result classified S (Susceptible) in 2024 may be reclassified SDD (Susceptible, dose-dependent) or I (Susceptible, Increased exposure) in 2025. The laboratory system must be tied to a specific version of the table, and version changes must trigger a documented impact assessment including re-analysis of historical results if the breakpoint changed for a clinically critical antibiotic.

EUCAST introduced the category "I = Susceptible, Increased Exposure" in 2019, replacing the old intermediate category. Many laboratory information systems still report the old three-category scheme. This is an active audit finding.

Automated reader systems — validating the instrument

VITEK 2, BD Phoenix and Sensititre read growth automatically, apply manufacturer-trained algorithms to call MIC, and apply breakpoints from their card/panel version. Validation requirements for these systems per ISO 15189:

• Essential agreement: ≥ 90% of MIC pairs within ± 1 dilution of BMD reference.
• Categorical agreement: ≥ 90% of S/I/R calls concordant with BMD at the relevant breakpoint version.
• Very major error rate: ≤ 3% (resistant organism classified susceptible — a patient safety failure).
• Major error rate: ≤ 3% (susceptible organism classified resistant).
• Minor error rate: ≤ 10% (one category discordant, excluding VME/ME).

Data pipeline considerations

AST generates large tables: dozens of antibiotics × hundreds of isolates per week, with QC data alongside. A reproducible data pipeline needs to handle:

1. MIC encoding: raw MIC values (e.g., 0.5, >128) must be stored as numerics, not strings — ">128" and "128" are different; choose a convention and enforce it at entry.
2. Breakpoint versioning: every result row must carry the EUCAST/CLSI table version used for interpretation at the time of reporting.
3. QC tracking: trend QC MICs per antibiotic-strain combination over time; flag and investigate any upward drift ≥ 1 dilution over 10 consecutive runs.
4. Resistance gene correlation: for labs running genotypic confirmation (PCR for mecA, blaKPC, NDM, OXA-48), the phenotypic MIC and the genotypic result must be co-reported and cross-checked.
5. Epidemiology outputs: antibiogram statistics (% susceptible per species-antibiotic pair) from the pipeline feed local empirical prescribing guidelines — these must exclude duplicates (one isolate per patient per episode).

Cumulative antibiogram reporting

CLSI M39 defines the methodology for cumulative antibiogram production. Key rules that data pipelines frequently violate:

• Include only the first isolate per species per patient per analysis period.
• Exclude screening isolates (surveillance swabs) from therapeutic isolate statistics.
• Report only if ≥ 30 isolates of a species-antibiotic pair — smaller denominators produce misleading percentages.
• Stratify by ward or infection type when sample size allows — ICU isolates behave very differently from outpatient isolates.

How AiLabrix fits

Drop the instrument export (VITEK LIS export, Sensititre SWIN, or a manual BMD spreadsheet). The pipeline imports MIC values, applies the specified breakpoint version (EUCAST or CLSI, by year), flags QC out-of-range, computes categorical agreement vs. reference, trends QC MICs on Levey-Jennings charts, and produces cumulative antibiograms per M39 rules with duplicate exclusion. The signed PDF covers every QC run, essential/categorical agreement tables, and the breakpoint version audit trail. [email protected] for a pilot.