Laboratory quality control and NABL

In short

Laboratory quality control is the set of statistical and procedural checks that prove a test result is close enough to the truth to be used for diagnosis. It has two arms: internal quality control (IQC), run daily with stable control material to detect error in your own run, and external quality assessment (EQA/PT), in which an outside agency sends blind samples and compares your answer with peers. Accreditation wraps both into a management system: in India, medical laboratories are accredited by NABL (a constituent board of the Quality Council of India) against ISO 15189:2022, applied through NABL's specific criteria document for medical laboratories. The topic carries heavy marks in B.Sc MLT clinical biochemistry and lab-management papers and is the single most asked area in technologist interviews at Dr Lal PathLabs, Metropolis, SRL, Thyrocare, Apollo, AIIMS and ICMR institutes, because NABL-accredited labs must show that every technologist can read a Levey-Jennings chart and act on a rule violation.

Explained

Core definitions

• Quality control (QC): operational techniques that monitor a process and detect error in real time. Quality assurance (QA): the whole planned system (people, SOPs, equipment, records) that gives confidence in results. Quality management system (QMS): the documented structure (quality manual → SOPs → formats → records) that runs QA.
• Accuracy / trueness: closeness of the measured value to the true (reference) value; expressed as bias. Precision: closeness of repeated measurements to each other; expressed as SD and CV. A result can be precise but wrong (constant bias) — the two are independent.
• Repeatability (within-run precision): same sample, same run, same operator. Intermediate precision (within-laboratory, between-day): across days, calibrations, operators — the realistic figure for QC planning. Reproducibility: between laboratories.
• Random error: unpredictable scatter — widens the SD (bubbles, pipetting variation, unstable power, clots). Systematic error (bias): a shift in one direction — wrong calibrator, wrong reagent lot, ageing lamp, dirty cuvette, wrong temperature.
• Analytical sensitivity: smallest amount detectable (LOD/LOQ). Analytical specificity: freedom from interference (haemolysis, icterus, lipaemia, drugs). Clinical sensitivity/specificity are different quantities and must not be confused with these.
• Traceability: an unbroken chain from the patient result through your calibrator to a higher-order reference material or method — IFCC reference methods, NIST SRMs, NGSP/IFCC standardisation for HbA1c, WHO international units for hormones.
• Three phases of error: pre-analytical (about two-thirds of all laboratory errors — wrong patient, wrong tube, wrong order of draw, wrong ratio, delay, haemolysis), analytical (what IQC monitors), post-analytical (transcription, units, critical value not telephoned, delayed report).

Internal quality control: material, frequency, charting

• Control material: a stable, matrix-matched pooled serum/blood/urine product, run as if it were a patient sample. Assayed controls come with a target range from the manufacturer; unassayed controls carry no values — the laboratory establishes its own. Third-party (independent) controls, not made by the analyser's manufacturer, are preferred because they are not biased towards one system's calibrators.
• Levels: at least two levels per analyte — one near the normal range, one in the medically important abnormal range (three for analytes with wide clinical span, e.g. glucose, TSH, digoxin).
• Establishing your own mean and SD: run the control for at least 20 different days (same lot, routine conditions), compute mean and SD, then draw limits at ±1, ±2 and ±3 SD. Do not use the insert range as your control limits — it is a consensus of many instruments and is too wide.
• Levey-Jennings (L-J) chart: days/run number on the x-axis, control value on the y-axis, horizontal lines at mean and ±1/2/3 SD. Points should be randomly scattered about the mean, roughly 68% within ±1 SD, 95% within ±2 SD, 99.7% within ±3 SD.
• Reading the chart: shift = sudden move of the mean to a new level, usually after calibration, a new reagent lot, a new lamp/electrode or maintenance. Trend = gradual drift over several days — reagent deterioration, calibrator evaporation, lamp ageing, slow electrode fouling, temperature creep. Increased scatter = random error.
• Frequency: at least once in each 24-hour period in which patient samples are tested, plus after calibration, after reagent lot change, after major maintenance, and whenever results look clinically improbable. Many labs run QC per shift or per batch for critical assays; the laboratory must define and justify the frequency in its SOP.
• Other IQC tools: delta check (compare with the same patient's previous result and flag an implausible change), average of normals / patient moving averages (Bull's algorithm for red cell indices MCV, MCH, MCHC in haematology), duplicate testing, reagent blank and standard absorbance monitoring, inter-instrument comparison when the same test runs on two analysers.

Westgard multirule QC and error interpretation

Each rule is written as A:L — A = number of control observations, L = the limit exceeded.

• Classic multirule combination: 1:3s / 2:2s / R:4s / 4:1s / 10x, with 1:2s as the warning gate — gives high error detection with low false rejection.
• Why not 1:2s as a rejection rule: with two controls, the chance of at least one point falling outside ±2 SD by pure chance is about 10% (1 − 0.95² = 9.75%), so a lab using 1:2s alone rejects good runs roughly one day in ten.
• Action on an out-of-control run: stop reporting → do not simply repeat the control and pass it → check the obvious (control expiry, reconstitution, storage, pipette, wrong level loaded) → identify random vs systematic from the rule broken → correct the cause (recalibrate, new reagent, maintenance) → re-run control → re-test and correct all patient samples run since the last acceptable QC → record the event, the root cause and the corrective action in the QC/non-conformity register. Repeating until the control "passes" is the classic malpractice and is an audit non-conformity.

External quality assessment, PT and inter-laboratory comparison

• EQA/PT: an external provider distributes identical blind samples to many laboratories; each returns its result; the provider reports your value against the peer group mean (same method/instrument) and the all-method mean, with an SDI (z-score) and often a % deviation.
• Interpretation: |SDI| ≤ 2 acceptable, 2–3 questionable, > 3 unacceptable. Repeated SDI of the same sign (e.g. always +1.5) is bias, even if "acceptable" — it needs calibration review. Large scatter of SDI across cycles is imprecision.
• Indian and international providers: CMC Vellore (clinical biochemistry, haematology, immunoassay), AIIMS New Delhi, NIB Noida (serology — HIV, HBV, HCV), ICMR schemes, Bio-Rad EQAS, Randox RIQAS, and NABL-recognised PT providers accredited to ISO/IEC 17043.
• NABL requirement: participation in EQA/PT for every accredited test where a scheme exists, for the required number of cycles per year, with documented root-cause analysis and CAPA for every unacceptable performance. Where no PT scheme exists, the lab must use an alternative assessment: inter-laboratory comparison with another accredited lab, split-sample testing, retesting of retained samples, or a certified reference material.
• Prohibited: discussing or sharing PT samples with another laboratory, referring the PT sample out, or treating it differently from a patient sample.

Method verification, calibration, uncertainty

• Verification (for a validated commercial method as used routinely): precision (within-run and between-day), accuracy/bias against a reference material or comparative method, analytical measurement range/linearity, LOD/LOQ, carry-over, interference, and verification of the reference interval for the local population.
• Validation (in-house, modified or non-standard methods) additionally covers specificity, robustness and, for qualitative tests, diagnostic sensitivity/specificity.
• Calibration: adjust the instrument against calibrators of stated traceability; calibration verification confirms the curve still holds. Calibrate after lot change, major service, or when QC shows a shift — not routinely "to make QC pass".
• Equipment control: balances, pipettes, centrifuges, thermometers, water baths, refrigerators and freezers must be calibrated by a NABL-accredited calibration laboratory at defined intervals, with temperature and humidity monitoring records; distilled water conductivity, biosafety cabinet airflow and autoclave validation are part of the same record set.
• Measurement uncertainty (MU): ISO 15189 requires labs to estimate MU for quantitative measurands and make it available on request. The practical top-down approach (ISO/TS 20914) combines long-term internal precision with the uncertainty of the calibrator/bias: u_c = √(u_Rw² + u_bias²), expanded as U = k·u_c with k = 2 for about 95% coverage.
• Allowable error goals: TEa from biological variation (Ricos/EFLM database), regulatory goals (CLIA; many CLIA acceptance limits were revised by the 2022 final rule and took effect on 1 January 2025, so quote the edition your syllabus uses), RCPA or clinician-agreed limits. State which source you are using.

NABL, ISO 15189 and the Indian regulatory frame

• NABL: National Accreditation Board for Testing and Calibration Laboratories, a constituent board of the Quality Council of India, signatory to the ILAC and APAC mutual recognition arrangements — so an NABL report is recognised internationally.
• Standard: ISO 15189:2022, Medical laboratories — Requirements for quality and competence (fourth edition; it is risk-based and now also covers point-of-care testing, whose separate standard ISO 22870 has been withdrawn). NABL applies it through its specific criteria document for medical laboratories (NABL 112) and operates a simpler scheme for small/basic composite medical laboratories; document numbers and editions are revised periodically, so check the current version on the NABL site.
• Two halves of ISO 15189: management system requirements (impartiality, confidentiality, organisation, personnel competence, documents and records control, nonconformity and CAPA, risk and improvement, internal audit, management review, complaints) and technical/process requirements (pre-examination, examination, post-examination, equipment, reagents, traceability, validation/verification, QC and EQA, reporting, release of results).
• Accreditation process, in order: gap analysis against the standard → build the QMS (quality manual, SOPs for every test, formats) → implement and generate at least a few months of records including IQC and EQA → train staff and record competence → internal audit → management review → apply to NABL with scope → document adequacy review → pre-assessment → final on-site assessment by a lead assessor plus technical assessors → close non-conformities (major/minor) with CAPA → accreditation granted, typically with surveillance assessments and renewal on a defined cycle (currently a two-year certificate validity with periodic surveillance — confirm the current cycle, as NABL revises it).
• Quality indicators (mandatory, monitored with targets): sample rejection rate, re-draw rate, haemolysed/clotted/insufficient sample rate, mislabelled sample rate, turnaround time compliance (routine and emergency, usually reported as % within target and the 90th/95th percentile), critical value notification rate and time, blood culture contamination rate, EQA performance, report amendment rate, customer complaints, uptime of analysers.
• Related Indian law and schemes: the Clinical Establishments (Registration and Regulation) Act, 2010 (in force in adopting States/UTs) requires registration and minimum standards; NABH accredits hospitals and blood banks; Drugs and Cosmetics rules and CDSCO govern diagnostic kits/reagents as medical devices (under the Medical Devices Rules, 2017); blood banks need a licence from the State/Central Drug authority; ICMR notifies testing protocols (as it did for COVID-19 RT-PCR labs). NABL accreditation is a condition for empanelment in CGHS and Ayushman Bharat PM-JAY laboratory panels and for most insurance and clinical-trial work.
• Discipline-specific QC: microbiology — ATCC reference strains for media performance and antibiotic disc potency, sterility and growth checks on each medium lot, AST QC per CLSI M100; haematology — commercial blood controls, Bull's moving average, slide review criteria; coagulation — normal and abnormal plasma controls, ISI/MNPT for INR; blood banking — QC of blood grouping antisera, anti-human globulin, blood component quality parameters; histopathology — positive and negative tissue controls for every IHC run, microtome and stain checks; molecular — positive control, negative control and no-template control in every run, plus internal amplification control to detect inhibition.

Key points

• IQC - own control material, detects error in today's run; EQA - external blind sample, detects bias against peers. Both are mandatory for accreditation.
• Precision = reproducibility (SD, CV); accuracy/trueness = closeness to true value (bias). Independent properties.
• SD and CV must be the laboratory's own, from at least 20 days of data, not the manufacturer's insert range.
• Levey-Jennings chart - mean with ±1/2/3 SD lines; 68/95/99.7% of points fall inside them when in control.
• Shift = abrupt level change (calibration, new lot); trend = gradual drift (reagent/lamp/electrode deterioration).
• 1:2s is a warning, never a stand-alone rejection rule; 1:3s and R:4s indicate random error, 2:2s, 4:1s and 10x systematic error.
• Out-of-control run - correct the cause, then re-test every patient sample since the last acceptable QC; record root cause and CAPA.
• SDI (z-score) in EQA: ≤2 acceptable, 2-3 questionable, >3 unacceptable; a persistent same-sign SDI means bias.
• Total error TE = |bias| + 1.65·SD must stay below TEa, the allowable total error for that analyte.
• Sigma metric = (TEa% − |bias%|)/CV%; ≥6 world class, 4-6 good, 3-4 marginal, <3 unacceptable and needs heavy multirule QC.
• Traceability - calibrators linked to IFCC/NIST/WHO higher-order materials; without it, accuracy cannot be claimed.
• Measurement uncertainty - u_c from precision and bias, reported as U = 2·u_c (about 95% coverage).
• ISO 15189:2022 is the standard for medical laboratories; NABL (under QCI) is the Indian accreditation body, ILAC-recognised; POCT is now inside ISO 15189 (ISO 22870 withdrawn).
• ISO/IEC 17025 is for testing/calibration labs, ISO/IEC 17043 for PT providers, NABH for hospitals - do not interchange them.
• Pre-analytical errors cause roughly two-thirds of laboratory errors; IQC cannot detect them, which is why quality indicators are tracked.
• Quality indicators - sample rejection rate, TAT compliance, critical value notification, blood culture contamination, EQA score.
• Clinical Establishments Act, 2010 governs registration in adopting States; NABL accreditation is needed for CGHS/PM-JAY empanelment.
• PT samples must be treated exactly like patient samples - no sharing, no referral, no repeat-until-right.

Formulas and examples

Example 1 — mean, SD, CV and control limits. Ten days of a glucose level-1 control (mg/dL), shortened for illustration - in practice at least 20 days are used to set limits: 96, 98, 102, 100, 104, 99, 101, 103, 97, 100 (Σ = 1,000).

• x̄ = 1,000/10 = 100 mg/dL.
• Deviations: −4, −2, +2, 0, +4, −1, +1, +3, −3, 0 → squares 16, 4, 4, 0, 16, 1, 1, 9, 9, 0 → Σ = 60.
• s = √(60/9) = √6.667 = 2.58 mg/dL; CV = (2.58/100) × 100 = 2.58%.
• Limits: ±1s = 97.4–102.6; ±2s = 94.8–105.2; ±3s = 92.3–107.7 mg/dL.

Example 2 — applying Westgard rules. Using the chart above: Day 21 control = 107 mg/dL → z = (107 − 100)/2.58 = +2.71 → outside ±2s but inside ±3s → 1:2s warning, run not rejected yet. Day 22 = 106 mg/dL → z = +2.33, same side → 2:2s violated → reject the run. Two consecutive points on the same side beyond 2s is systematic error; the probable causes are a new reagent lot or a shifted calibration, so recalibrate, re-run the control, and re-test the Day 21 and 22 patient samples.

Example 3 — total error and sigma metric (a marginal assay). Cholesterol: CV = 2.5%, bias = +3.0% from EQA, TEa = 9%.

• TE = 3.0 + (1.65 × 2.5) = 3.0 + 4.125 = 7.13% < 9% → the method is acceptable.
• σ = (9 − 3)/2.5 = 2.4 → below 3, i.e. poor performance despite passing the TE test, because almost all the error budget is eaten by bias.
• Action: find and remove the bias (recalibrate with a traceable calibrator) — if bias falls to 0.5%, σ = (9 − 0.5)/2.5 = 3.4. Until then, run a strict multirule (1:3s/2:2s/R:4s/4:1s) with three control measurements per run.

Example 4 — sigma-based QC design (a good assay). Glucose: TEa = 10%, bias = 1%, CV = 2% → σ = (10 − 1)/2 = 4.5. A sigma of 4–5 is adequately controlled by 1:3s with N = 2 (two controls per run), or 1:2.5s with N = 2; a σ ≥ 6 assay needs only 1:3s with N = 2 once a day, while σ < 3 demands N = 4 and multirules — or a change of method.

Example 5 — EQA interpretation and uncertainty. A lab reports creatinine 1.52 mg/dL on an EQA sample; the peer group mean is 1.40 mg/dL with SD 0.08.

• SDI = (1.52 − 1.40)/0.08 = +1.5 → acceptable (<2), but the previous three cycles were +1.4, +1.6 and +1.5 → a consistent positive bias of about +8.6% that must be investigated even though every single score "passed".
• Uncertainty for the same analyte: long-term within-lab CV (u_Rw) = 2.0%, uncertainty of calibrator/bias (u_bias) = 1.2% → u_c = √(2.0² + 1.2²) = √(4 + 1.44) = √5.44 = 2.33%; U = 2 × 2.33 = 4.7%. At a patient value of 1.00 mg/dL the result is 1.00 ± 0.05 mg/dL (k = 2).

Example 6 — quality indicators. In a month the lab received 2,000 samples and rejected 40 (haemolysed 22, clotted 10, insufficient 8) → rejection rate = (40/2,000) × 100 = 2.0% against a target of ≤1.5% → non-conformity, so CAPA = retrain two phlebotomy staff, replace a batch of short-fill EDTA tubes. Of 500 emergency samples, 460 were reported within 60 minutes → TAT compliance = (460/500) × 100 = 92% against a 90% target → met.

Where it is used

• Diagnostic chains and reference laboratories — Dr Lal PathLabs, Metropolis, SRL, Thyrocare, Agilus, Apollo Diagnostics: daily IQC, monthly EQAS submissions and NABL scope maintenance across hundreds of collection centres with one central reference lab.
• Hospital laboratories — AIIMS, PGIMER, CMC Vellore, Tata Memorial and NABH-accredited private hospitals: QC officer/quality manager roles, critical value protocols, TAT dashboards for ICU and casualty.
• Job roles — Lab Technologist/Technician (runs and documents QC), QC Executive / Quality Manager (L-J charts, EQAS, internal audits, NABL documentation), Internal Auditor and NABL technical assessor, Application/Service Specialist with IVD companies (Roche, Siemens Healthineers, Abbott, Beckman, Transasia, Mindray) who troubleshoot customer QC failures.
• IVD manufacturing and R&D — Transasia/Erba, Mylab, Molbio, Trivitron: lot release testing, control and calibrator value assignment, stability studies, CDSCO/ISO 13485 compliance.
• Public health and regulated testing — ICMR/NIB EQAS for HIV and viral hepatitis, NTEP (TB) culture and NAAT laboratory certification, Integrated Disease Surveillance Programme and State public health labs, food/water labs under ISO/IEC 17025.
• Clinical trials and insurance work — CRO central laboratories need NABL/CAP accreditation and documented MU and traceability before data are accepted; NABL status is also a condition for CGHS and PM-JAY panel empanelment.

Interview questions

• Q: What is the difference between accuracy and precision? A: Accuracy (trueness) is closeness to the true value, measured as bias; precision is closeness of repeated results to each other, measured as SD/CV. A method can be highly precise and still consistently wrong.
• Q: Why do we use CV instead of SD to compare two analytes? A: SD is in the units of the analyte and grows with concentration, so only the CV — SD as a percentage of the mean — allows comparison between analytes, levels or instruments.
• Q: How do you set your own control limits for a new control lot? A: Run the control under routine conditions for at least 20 separate days, calculate the mean and SD of those values, and draw the ±2 SD and ±3 SD limits from them; the insert range is used only as a sanity check.
• Q: One control value falls just outside 2 SD. Do you reject the run? A: No. 1:2s is only a warning; about 1 in 20 points falls outside 2 SD by chance. Apply the other rules (2:2s, R:4s, 4:1s, 10x) and inspect the chart before deciding.
• Q: Which rules point to random error and which to systematic error? A: 1:3s and R:4s indicate random error (increased imprecision); 2:2s, 4:1s, 10x and 7T indicate systematic error or drift.
• Q: Your chart shows ten points in a row just above the mean, all within 2 SD. Is that in control? A: No — that is a 10x violation, a small systematic shift. Statistically the points should scatter on both sides, so the method needs calibration review even though no point is outside limits.
• Q: What is the difference between a shift and a trend, and what causes each? A: A shift is an abrupt jump to a new level, typically after calibration, a reagent lot change or maintenance; a trend is a gradual drift, typically reagent or calibrator deterioration, an ageing lamp or a fouling electrode.
• Q: After an out-of-control run, what happens to the patient samples? A: They are held; after the cause is corrected and QC is acceptable, every patient sample run since the last acceptable QC is re-tested, and any already-released report is amended with documented notification.
• Q: What does an SDI of +2.8 in EQAS mean and what do you do? A: Your result is 2.8 peer SDs above the peer mean — questionable to unacceptable performance. Do root-cause analysis (calibration, reagent lot, sample handling, clerical error in submission), take corrective action, and verify in the next cycle.
• Q: Which standard are medical laboratories accredited to in India, and by whom? A: ISO 15189:2022, by NABL, a constituent board of the Quality Council of India, applied through NABL's specific criteria for medical laboratories; NABL is an ILAC MRA signatory.
• Q: Difference between ISO 15189 and ISO/IEC 17025? A: ISO 15189 is specific to medical laboratories and covers the full patient pathway including pre- and post-examination; ISO/IEC 17025 is the general standard for testing and calibration laboratories and has no patient-care requirements.
• Q: Give three quality indicators your laboratory must monitor. A: Sample rejection/haemolysis rate, turnaround time compliance for routine and emergency tests, and critical value notification rate and time — each with a defined target and CAPA when missed.
• Q: What is sigma metric and why is it useful? A: σ = (TEa% − |bias%|)/CV%; it converts method performance into a single number that tells you how many controls and which rules you need — high sigma means simple QC, low sigma means multirules or a better method.

Common mistakes

• Using the manufacturer's insert range as control limits, or re-calculating the SD after excluding "bad" days — both artificially widen or narrow the chart and hide real error.
• Repeating an out-of-range control until it passes and releasing the run, without root-cause analysis, CAPA record or re-testing the patient samples.
• Treating 1:2s as a rejection rule (causing constant false rejection) or ignoring in-limit patterns like 10x and 7T.
• Confusing analytical sensitivity/specificity (LOD, interference) with clinical sensitivity/specificity (TP/FN rates) in answers.
• Reporting SD without the mean, or comparing SDs at different concentrations instead of comparing CVs.
• Mixing up the standards: quoting ISO 9001 or ISO/IEC 17025 for a medical lab, or ISO 22870 for POCT (withdrawn; POCT is inside ISO 15189:2022), and calling NABL a government regulator rather than an accreditation body under QCI.
• Assuming good IQC proves a correct report — IQC cannot detect a mislabelled tube, a haemolysed sample or a transcription error, which is why pre- and post-analytical indicators exist.
• Forgetting n − 1 in the SD formula, or mixing units (mg/dL with mmol/L) in bias and total-error calculations.

Related topics

Specimen collection and phlebotomy · Laboratory instruments · Clinical biochemistry analytical methods · Haematology and coagulation testing · Clinical microbiology and AST · Laboratory management, documentation and audit