Hematology analyzer validation — CBC, differential and coagulation under ISO 15189

CBC is ordered more often than any other lab test. The irony is that hematology validation is also one of the most abbreviated — "we ran 20 samples, results matched the old analyzer, signed off." An ISO 15189 audit does not accept that any more. This piece covers a validation package that does.

The CBC parameter set (and where precision matters)

• WBC, RBC, Hb, Hct, MCV, MCH, MCHC, RDW, Plt, MPV — quantitative, all need precision + trueness + reference intervals.
• 5-part differential — Neutrophils, Lymphocytes, Monocytes, Eosinophils, Basophils (absolute + percentage). Challenging at low counts — validate per decade of concentration.
• Nucleated RBC, reticulocytes, IPF — optional per panel; separately validated.

Validation package

1. Precision (EP05)

Two or three control levels, 20 days, 2 runs, 2 reps. Acceptance per EFLM biological variation — WBC SL ≤ 5.6%, Hb ≤ 1.8%, platelets ≤ 5.3% at normal levels. Thresholds loosen at clinical decision points (platelet count at 50 is different from at 250).

2. Method comparison (EP09)

≥ 40 samples spanning the AMR, with at least 30% outside the reference interval. Passing-Bablok and Bland-Altman per parameter. Acceptance from ICSH recommendations, not the manufacturer's marketing brief.

3. Linearity (EP06)

Commercial linearity material across the claimed range. Weighted regression, residuals check, highest and lowest points specifically validated (this is where analyzers fall out of spec).

4. Carry-over

High → low → low sequence, three repeats. Carry-over for WBC, RBC, Hb, Plt should be < 1% per ICSH. Do not skip this — it is the single most common hematology pre-analytical failure mode.

5. Reference intervals (EP28)

Age + sex partitions are mandatory for CBC. Pediatric reference intervals (CALIPER, KiGGS) change steeply in the first year — a single "pediatric" interval is not enough. Pregnancy-specific intervals for Hb, platelets, WBC.

6. Flagging logic

Hematology analyzers produce ~30 morphology flags (blast, immature granulocyte, left shift, atypical lymph, NRBC, RBC fragments). Validate flag sensitivity + specificity against microscopic review on ≥ 200 samples enriched for abnormal. A blast flag with 60% sensitivity is not safe for an outpatient population.

7. Sample stability

Document stability at room temperature and refrigerated for CBC + diff. WBC diff shifts within 4 hours at room temperature — reporting a differential on an overnight sample is a common silent error.

Coagulation — the separate discipline

PT/INR/APTT on the same platform need their own validation loop:

• ISI verification per reagent lot — direct ISI calibration with WHO-certified plasmas is preferred over accepting the manufacturer's ISI.
• INR reference plane — anticoagulated and normal patient pools to confirm the ISI and geometric mean of the normal range.
• APTT reagent response — heparin responsiveness curve, lupus-anticoagulant sensitivity claim.
• Clot detection — optical vs mechanical impacts lipemic samples differently.

QC program

• Two control levels per shift, Levey-Jennings, Westgard multi-rule.
• Retained patient means (Bull algorithm) as a secondary drift sentinel — it catches systematic shifts that controls miss.
• Peer-group comparison (manufacturer QC program or external) monthly.
• EQA participation (NEQAS, CAP, RIQAS) with documented response to any outlier.

Where AiLabrix fits

Drop the analyzer CSV plus the microscopic review log. The pipeline runs EP05 precision, EP09 method comparison per parameter, linearity, carry-over, flag sensitivity/specificity from the morphology review, and age/sex-partitioned reference interval verification. Output: signed PDF with all plots and the Westgard + Bull monitoring configured for day-one operations. [email protected].