Analytical Phase

The analytical phase covers nucleic acid extraction, the assay itself, and the per-run quality control that determines whether a run is valid. The overarching clinical question is whether the assay's analytical performance (accuracy, precision, sensitivity, specificity, limit of detection) is fit for the clinical question, and whether the controls demonstrate that the run actually achieved that performance on this particular day on this particular instrument.

• Instrument calibration and verification: instruments are calibrated per manufacturer schedule and verified after maintenance, lamp/laser changes, or reagent lot changes. Calibration verification confirms that reportable range is preserved.
• Controls run with each batch:
  • Positive control: a sample with the expected analyte/variant; confirms the assay can detect the target.
  • Negative control: a known-negative sample; confirms the assay does not call false positives.
  • No-template (blank/NTC) control: water or buffer in place of DNA; detects reagent or environmental contamination, especially critical in PCR.
  • Internal control: an endogenous or spiked-in target run within the patient sample to confirm extraction and amplification worked.
• Westgard rules (originally for clinical chemistry, broadly applied):
  • 1-2s: one control beyond +/-2 SD - warning, not rejection.
  • 1-3s: one control beyond +/-3 SD - reject the run (random error).
  • R-4s: range between two controls in the same run exceeds 4 SD - reject (random error).
  • 2-2s: two consecutive controls beyond the same +/-2 SD limit - reject (systematic error).
  • 4-1s: four consecutive controls on the same side beyond 1 SD - reject (systematic shift).
  • 10x: ten consecutive controls on the same side of the mean - reject (systematic shift, even if within 1 SD).
• Analytical sensitivity vs analytical specificity: sensitivity is the ability to detect the analyte when present (true positive rate at the assay level); specificity is the ability to call negative when the analyte is absent (true negative rate). Distinct from clinical sensitivity/specificity, which incorporate disease prevalence and phenotype.
• Limit of detection (LoD): the lowest analyte level reliably distinguished from background. Critical for somatic variant calling, mosaicism, NIPS fetal fraction, and minimal residual disease, where the variant of interest may be present at 1% allele fraction or less.
• Interferences: heterophile antibodies (false positives in immunoassays), high-dose biotin (interferes with streptavidin-biotin immunoassays, classically falsely lowering TSH and falsely elevating troponin), hemoglobin and lipemia (spectrophotometric interference), and heparin carryover (PCR inhibition).
• Batch vs random-access: batch testing groups samples into a single run for cost efficiency (typical for NGS, MLPA, microarray); random-access analyzers run samples on demand (typical for clinical chemistry). Batch workflows have longer turnaround but tighter control of run-to-run variability.
• Reflex testing rules: predefined algorithms by which an initial result triggers an additional confirmatory or follow-on test (positive NIPS reflex to diagnostic karyotype/CMA; abnormal newborn screen reflex to confirmatory biochemical or molecular test; positive HPV reflex to genotyping). Reflex rules must be defined in the SOP and signed off by the director.
• Run validity criteria: each SOP defines what must be true for a run to be reported - controls within range, internal control amplified, no-template clean, instrument QC passed, coverage thresholds met (e.g., 20x at >=98% of target bases for germline NGS). If criteria fail, the run is rejected and samples are repeated.

• A patient on high-dose biotin supplements (often for hair/skin) can produce spuriously normal or abnormal immunoassay results; labs increasingly add biotin warnings to requisitions.
• A 1-3s control failure on a single run mandates repeat of all patient samples in that run, not selective repeat.
• For somatic NGS, an LoD of 5% variant allele frequency is inadequate for detecting low-level resistance mutations or measurable residual disease; deeper coverage and lower LoD assays are required.
• An NTC that amplifies in a PCR run indicates contamination and invalidates the entire run, even if patient samples look "clean."