1.3 Quality Control, Westgard Rules, Calibration, and Laboratory Mathematics

Quality Control, Westgard Rules, Calibration, and Laboratory Mathematics

Statistical Quality Control (QC) in clinical chemistry monitors analytical precision, detects method bias, and prevents erroneous patient data release. For the medical laboratory scientist, quality control is an active troubleshooting discipline rooted in metrology and mathematics.

---

1. Statistical Quality Control Parameters and Gaussian Distribution

Statistical QC assumes analytical measurements on stable control materials follow a Gaussian (normal) distribution—a symmetrical, bell-shaped probability density curve characterized by the mean ($\bar{x}$) and standard deviation ($s$).

                             Gaussian Distribution
                                      │
                                    ┌─┴─┐
                                  ┌─┘   └─┐
                                ┌─┘       └─┐
                              ┌─┘           └─┐
                            ┌─┘               └─┐
                         ───┴───────────────────┴───
                            -3s   -2s   -1s   x̄   +1s   +2s   +3s
                            ├────── 68.27% ─────┤
                            ├────────── 95.45% ─────────┤
                            ├────────────── 99.73% ─────────────┤

Statistical Parameters

• Mean ($\bar{x}$): The arithmetic average of all data points; represents analytical accuracy (trueness): $\bar{x} = \frac{\sum x_i}{n}$
• Standard Deviation ($s$ or SD): Quantifies data dispersion around the mean; represents analytical precision (or imprecision): $s = \sqrt{\frac{\sum (x_i - \bar{x})^2}{n - 1}}$
• Variance ($s^2$): The square of standard deviation, used in statistical comparisons (such as the $F$-test for method comparison).
• Coefficient of Variation ($CV\%$): The standard deviation expressed as a percentage of the mean: $CV\% = \left(\frac{s}{\bar{x}}\right) \times 100\%$ Because $CV\%$ is dimensionless, it enables direct comparison of analytical precision between different instruments, methodologies, or concentration levels (e.g., assessing if low-level troponin is as precise as high-level troponin).

Gaussian Confidence Intervals

• $\bar{x} \pm 1s$: Encloses $68.27\%$ of all observations.
• $\bar{x} \pm 2s$: Encloses $95.45\%$ of all observations. In single-rule $1_{2s}$ systems, approximately $4.55\%$ (1 in 22) of valid in-control runs fall outside $\pm 2s$ purely by random chance.
• $\bar{x} \pm 3s$: Encloses $99.73\%$ of all observations. Only $0.27\%$ (1 in 370 observations) will fall outside $\pm 3s$ when the method is stable. Points beyond $\pm 3s$ indicate true analytical error.

---

2. Levey-Jennings Control Charts and Error Classification

A Levey-Jennings (L-J) chart plots daily control measurements across time (x-axis) against established statistical limits (y-axis: $\bar{x}, \pm 1s, \pm 2s, \pm 3s$). Analytical errors manifest as characteristic patterns:

  +3s ──────────────────────────────────────────────────────── (Action Limit)
  +2s ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ (Warning Limit)
  +1s ························································
   x̄  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ (Target Mean)
  -1s ························································
  -2s ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ (Warning Limit)
  -3s ──────────────────────────────────────────────────────── (Action Limit)
       Day 1   Day 2   Day 3   Day 4   Day 5   Day 6   Day 7

Random Error (Imprecision)

• Nature: Unpredictable, non-directional deviations in either direction from the mean. Affects the reproducibility and spread ($s, CV\%$) of the assay.
• Underlying Causes: Air bubbles in reagent or sample lines, transient voltage spikes in electrical power lines, particulate contamination in cuvettes, pipetting tip leakage, or photometer lamp flicker.
• L-J Appearance: Increased scatter of individual points across both sides of the mean; violations of $1_{3s}$ or $R_{4s}$ rules.

Systematic Error (Inaccuracy / Bias)

• Nature: Predictable, unidirectional drift or step-shift in values away from the target mean. Shifts the observed mean while standard deviation may remain narrow.
• Patterns on Levey-Jennings:
  1. Trend: A gradual, continuous progression of control values in one direction away from the mean across $\ge 6$ consecutive runs. Caused by: slow deterioration of reconstituted reagents, aging of the photometer excitation lamp, gradual buildup of protein bio-film in flow cells, or slow evaporation of control material in open sample carousels.
  2. Shift: An abrupt, sustained change in control values that maintains a new mean above or below the target across $\ge 6$ consecutive runs. Caused by: introduction of a new reagent lot with an inaccurate calibrator assignment factor, sudden change in incubator chamber temperature, incorrect recalibration, or syringe pipettor slippage.

---

3. The Westgard Multirule Decision Algorithm

Formulated by Dr. James O. Westgard, the multirule algorithm combines statistical decision criteria to maximize the Probability of Error Detection ($P_{ed}$) while keeping the Probability of False Rejection ($P_{fr}$) below $1\%$.

                           ┌───────────────────────────────┐
                           │   Perform Daily Control Run   │
                           │    (e.g., Level 1, Level 2)   │
                           └───────────────┬───────────────┘
                                           │
                                           ▼
                           ┌───────────────────────────────┐
                           │  Does ANY control exceed 2s?  │
                           │           (Rule 1₂ₛ)          │
                           └───────┬───────────────┬───────┘
                                   │ NO            │ YES (Warning trigger)
                                   ▼               ▼
                          ┌────────────────┐  Inspect Multirules:
                          │   ACCEPT RUN   │  ├─ 1₃ₛ (Any value > 3s?) ──► REJECT (Random Error)
                          │ Report Results │  ├─ 2₂ₛ (2 in a row > 2s?) ──► REJECT (Systematic Error)
                          └────────────────┘  ├─ R₄ₛ (Span > 4s in run?) ─► REJECT (Random Error)
                                              ├─ 4₁ₛ (4 in a row > 1s?) ──► REJECT (Systematic Error)
                                              └─ 10ₓ (10 on same side?) ──► REJECT (Systematic Error)

• Application Scope: Rules such as $2_{2s}$, $4_{1s}$, and $10_x$ can be applied within-run across control levels (e.g., Level 1 at $+2.2s$ and Level 2 at $+2.4s$ violates $2_{2s}$) or across consecutive runs for a single control level.

---

4. Systematic Troubleshooting Protocol for QC Failures

When a Westgard rejection rule triggers, laboratory technologists must adhere to a structured corrective action cascade:

1. Do NOT blindly repeat the control: Re-running the identical vial hoping it falls within $\pm 2s$ creates false acceptance of an unstable analytical run. A single control repeated 20 times will statistically fall inside $2s$ purely by chance.
2. Do NOT average control results: Averaging disguises true random and systematic errors.
3. Step 1: Physical Inspection: Check reagent levels, fluidic tubing for air bubbles, sample probe alignment, wash block aspiration, and incubator bath temperature readouts.
4. Step 2: Re-run the Same Aliquot: Rules out an isolated physical artifact (e.g., tiny meniscus air bubble or temporary voltage dip).
5. Step 3: Test a Fresh Aliquot: Reconstitute or thaw a new, unopened vial of control material. If the fresh aliquot recovers on target, the problem was control degradation, evaporation, or vial contamination.
6. Step 4: Inspect Reagents: Check reagent open-vial expiration dates, lot changes, and physical appearance (turbidity, discoloration, precipitation). Load a fresh reagent cartridge.
7. Step 5: Perform Recalibration: Recalibrate the assay using fresh calibrator materials. Inspect the calibration curve parameters (slope, y-intercept, and $R^2$ fit).
8. Step 6: Hardware Maintenance: If fresh reagent and recalibration fail, replace photometer lamps, clean sample probe fluidics, replace peristaltic pump tubing, or contact manufacturer technical support.
9. Critical Patient Protection: After resolving the root cause and verifying QC acceptability, re-run all patient samples analyzed since the last documented acceptable QC run.

---

5. Calibration Hierarchy, Standards, and Metrological Traceability

• Calibrator versus Quality Control Material:
  • Calibrator: A substance with an established analyte concentration assigned by primary reference methods. It establishes the instrument response curve (signal-to-concentration mathematical relationship). Calibrators must never be used as daily QC controls, as doing so creates circular validation that masks calibration errors.
  • Control: A matrix-matched material (e.g., human serum base) with target means and standard deviations, treated identically to patient specimens to verify ongoing method stability.
• Traceability Hierarchy (ISO 17511): $\text{Primary Reference Method (e.g., ID-MS)} \longrightarrow \text{Primary Calibrator (Pure Substance)}$ $\downarrow$ $\text{Secondary Reference Measurement} \longrightarrow \text{Secondary Matrix-Matched Calibrator (WHO, NIST)}$ $\downarrow$ $\text{Manufacturer Working Calibrator} \longrightarrow \text{Manufacturer Commercial Product Calibrator}$ $\downarrow$ $\text{Routine End-User Clinical Laboratory Analyzer} \longrightarrow \text{Reportable Patient Result}$

---

6. Clinical Chemistry Mathematics and Diagnostic Formulas

Dilution Calculations

• The General Dilution Equation: $C_1 V_1 = C_2 V_2$ Where $C_1$ and $V_1$ are the initial concentration and volume, and $C_2$ and $V_2$ are the final concentration and volume.
• Dilution Factor (DF): The ratio of total final volume to original sample volume: $\text{DF} = \frac{V_{\text{final}}}{V_{\text{sample}}} = \frac{V_{\text{sample}} + V_{\text{diluent}}}{V_{\text{sample}}}$ To calculate patient concentration, multiply the instrument reading of the diluted sample by $\text{DF}$.
• Serial Dilutions: In a series of consecutive dilutions, the overall dilution factor is the mathematical product of the individual dilution steps: $\text{Total DF} = \text{DF}_1 \times \text{DF}_2 \times \text{DF}_3 \times \dots$

Molarity and Normality

• Molarity ($M$): Moles of solute per liter of total solution: $M = \frac{\text{moles of solute}}{\text{Volume in Liters}} = \frac{\text{mass (g)}}{\text{Molecular Weight (g/mol)} \times V \text{ (L)}}$
• Normality ($N$): Chemical equivalents of solute per liter of solution: $N = M \times \text{valence} = \frac{\text{mass (g)}}{\text{Equivalent Weight} \times V \text{ (L)}} \quad \text{where } \text{Equivalent Weight} = \frac{\text{MW}}{\text{valence}}$ For example: A $1.0\text{ M}$ solution of sulfuric acid ($\text{H}_2\text{SO}_4$, which has $2$ ionizable protons) has a normality of $1.0 \times 2 = 2.0\text{ N}$.

The Serum Anion Gap (AG)

The anion gap reflects the concentration of unmeasured anions (proteins, organic acids, phosphates, sulfates) relative to unmeasured cations ($Ca^{2+}, Mg^{2+}$, gamma-globulins) in serum.

• Formula 1 (Without Potassium—standard modern practice): $\text{AG} = [\text{Na}^+] - ([\text{Cl}^-] + [\text{HCO}_3^-]) \quad \text{Reference Interval: } 8\text{ to }16\text{ mmol/L}$
• Formula 2 (With Potassium): $\text{AG} = ([\text{Na}^+] + [\text{K}^+]) - ([\text{Cl}^-] + [\text{HCO}_3^-]) \quad \text{Reference Interval: } 10\text{ to }20\text{ mmol/L}$

Clinical Interpretations of the Anion Gap

• High Anion Gap Metabolic Acidosis (HAGMA): Occurs when unmeasured fixed organic acids accumulate, consuming bicarbonate buffer. Key etiologies are summarized by the classic mnemonics MUDPILES and GOLD MARK:
  • Methanol (metabolized to formic acid)
  • Uremia (renal failure causing phosphate/sulfate retention)
  • Diabetic Ketoacidosis (acetoacetic acid, $\beta$-hydroxybutyric acid)
  • Propylene glycol / Paraldehyde
  • Isoniazid / Iron intoxication
  • Lactic acidosis (hypoperfusion, tissue hypoxia, shock)
  • Ethylene glycol (metabolized to oxalic and glycolic acids)
  • Salicylate overdose (salicylic acid accumulation + uncoupled oxidative phosphorylation)
• Low Anion Gap ($<6\text{ mmol/L}$):
  • Severe Hypoalbuminemia: Albumin is the primary unmeasured anion in normal serum, accounting for $\approx 75\%$ of the normal anion gap. For every $1.0\text{ g/dL}$ drop in serum albumin below $4.0\text{ g/dL}$, the expected anion gap drops by approximately $2.5\text{ mmol/L}$: $\text{Corrected AG} = \text{Observed AG} + 2.5 \times (4.0 - [\text{Albumin in g/dL}])$
  • Multiple Myeloma: Accumulation of positively charged cationic paraproteins (IgG) increases unmeasured cations.
  • Bromide Intoxication: Bromide ions cross-react and bind avidly to chloride ISE membranes, producing spuriously elevated chloride readings that falsely collapse or yield negative anion gaps.