Control Charts in Laboratory Quality Control: Levey-Jennings Plots and Westgard Rules

Control charts are graphical tools used in laboratory quality control (QC) to monitor analytical performance over time, enabling detection of systematic drift, bias, random error, and rule violations in routine testing. The Levey-Jennings plot—a specialized form of the Shewhart control chart—displays sequential QC measurements against established mean and standard deviation limits, while Westgard rules provide a structured decision framework for interpreting these plots. This method is most useful when laboratories need to distinguish between acceptable analytical variation and performance degradation that could compromise result accuracy, particularly in clinical chemistry, immunoassay, and molecular diagnostics where consistent output is critical for patient care or research reproducibility.

At a Glance

Scientific Principle of Control Charts

Control charts operate on the fundamental statistical principle that all analytical measurements contain inherent random error. When a process is "in control," this error follows a predictable distribution—typically Gaussian—around a target value. The Levey-Jennings plot visualizes this distribution by plotting individual QC results chronologically, with horizontal lines representing the mean and multiples of the standard deviation (commonly ±1 SD, ±2 SD, and ±3 SD) [1].

The underlying assumption is that approximately 95% of measurements should fall within ±2 SD of the mean, and 99.7% within ±3 SD, when only random error is present. Systematic errors—such as reagent degradation, calibration drift, or instrument malfunction—cause QC values to shift persistently above or below the mean, producing patterns that deviate from this expected distribution. The Westgard rules operationalize this statistical reasoning by defining specific patterns that indicate the process is "out of control" and requires investigation before patient or experimental results are reported [2].

The mathematical foundation relies on calculating the mean (x̄) and standard deviation (s) from an initial set of QC measurements. For a new assay, laboratories typically collect at least 20 data points across multiple runs under stable conditions. The mean serves as the target value, while the SD quantifies the expected dispersion. The coefficient of variation (CV = s/x̄ × 100%) provides a normalized measure of imprecision that facilitates comparison across analytes with different concentration ranges [2].

Materials and Instrumentation Choices

QC Materials

The choice of quality control material significantly influences chart reliability. Commercial lyophilized or liquid controls offer known target ranges and matrix compatibility but can be expensive and may not perfectly match patient sample matrices. Pooled patient sera or residual clinical specimens provide matrix-relevant alternatives at lower cost, though they require careful preparation and stability validation [3].

For molecular assays, synthetic nucleic acid controls or plasmid-based standards are common. These must be verified for homogeneity, stability during storage, and absence of interfering substances. The Triple-Point Pooled Sera (TriPPS) approach demonstrates that in-house pooled materials can achieve stability for at least 60 consecutive analytical days when properly prepared and stored [3].

Instrumentation

Any analytical instrument capable of producing quantitative results can generate data for Levey-Jennings charts. The key requirement is consistent measurement conditions across runs. Automated clinical chemistry analyzers (e.g., Roche Cobas series) typically include built-in QC software, while ELISA readers, PCR platforms, and spectrophotometers may require external data processing [1][2].

For laboratories without commercial QC software, open-source solutions exist. A Python program developed for ELISA QC can batch-process standard curve data, generate Levey-Jennings plots, and flag values exceeding two standard deviations [1]. This approach democratizes QC implementation for academic settings with limited budgets.

Data Management Systems

Modern laboratories increasingly use middleware or laboratory information systems (LIS) that automatically capture QC results and apply Westgard rules. Patient-based real-time quality control (PBRTQC) platforms, such as those using exponentially weighted moving average (EWMA) algorithms, can complement traditional control charts by detecting shifts earlier than conventional methods [4]. However, these advanced systems require careful validation and may not replace Levey-Jennings plots for regulatory compliance.

Controls and Standards

Internal Quality Control (IQC)

Internal quality control involves analyzing materials of known composition alongside patient or experimental samples. Two concentration levels—typically normal and abnormal (or low and high)—are recommended to detect errors across the analytical measurement range. Each level should be run at least once per analytical batch or every 24 hours, depending on laboratory policy and regulatory requirements [2].

The IQC material must be treated identically to patient samples throughout the analytical process. This includes storage conditions, thawing procedures, and pipetting volumes. Any deviation in QC material handling can introduce pre-analytical variation that masquerades as analytical error.

Establishing Initial Control Limits

Before routine monitoring begins, laboratories must establish baseline statistics. The process involves:

1. Running the QC material across at least 20 separate analytical runs under optimal conditions
2. Calculating the mean and standard deviation for each level
3. Verifying that no outliers or trends exist in the initial dataset
4. Setting control limits at ±1 SD (warning), ±2 SD (alert), and ±3 SD (action)

If the initial dataset contains outliers, they should be investigated and removed before calculating limits. Some laboratories use robust statistics (e.g., median and median absolute deviation) to minimize the influence of extreme values during the establishment phase.

External Quality Assessment (EQA)

While not directly part of Levey-Jennings charts, participation in external quality assessment schemes (proficiency testing) provides independent verification of accuracy. Bias estimates from EQA programs can be incorporated into sigma metric calculations, which in turn inform the selection of appropriate Westgard rules [2].

Conceptual Workflow

Step 1: Data Collection

Record each QC measurement with its date, time, instrument identifier, reagent lot number, and operator. This metadata is essential for troubleshooting when rule violations occur. For each analytical run, analyze at least two levels of QC material.

Step 2: Plot Construction

Create a Levey-Jennings chart with the x-axis representing chronological order (run number or date) and the y-axis representing the measured QC value. Draw horizontal lines at the mean, mean ± 1 SD, mean ± 2 SD, and mean ± 3 SD. Commercial software or spreadsheet templates can automate this process.

Step 3: Visual Inspection

Before applying formal rules, visually inspect the chart for:

• Trends: Six or more consecutive points moving in one direction (upward or downward)
• Shifts: A sudden change where multiple consecutive points fall on one side of the mean
• Random scatter: Excessive variability or clustering near control limits

Step 4: Apply Westgard Rules

Apply the following rules in sequence, starting with those most sensitive to random error:

The 1₂s rule serves as a warning flag that triggers inspection of other rules. If only the 1₂s rule is violated without other rule violations, the run may be accepted after verifying no other problems exist. Violation of 1₃s, 2₂s, R₄s, 4₁s, or 10x typically requires rejection of the run and investigation [2].

Step 5: Decision and Documentation

When a rule violation occurs:

1. Do not report patient results from the affected run
2. Investigate potential causes (reagent, calibrator, instrument, operator)
3. Document the violation, investigation, and corrective action
4. Repeat QC after corrective action before resuming patient testing

Quality Checks and Validation

Initial Validation of Control Limits

After establishing initial limits, verify their appropriateness by running an additional 20 QC measurements. If more than 5% of values fall outside ±2 SD, or if any values fall outside ±3 SD, the limits may need recalculation or the assay may require optimization before routine use.

Ongoing Monitoring

Control limits should be recalculated periodically (e.g., every 6 months or after reagent lot changes) using the most recent 20-30 data points. However, limits should not be recalculated so frequently that they mask gradual performance degradation. Some laboratories use cumulative statistics that incorporate all historical data while weighting recent observations more heavily.

Sigma Metric Integration

The sigma metric (σ = [TEa - bias] / CV) quantifies assay performance relative to allowable error. Higher sigma values indicate more robust assays that can tolerate less stringent QC rules. For assays with σ ≥ 6, simple 1₃s or 2₂s rules may suffice. For assays with σ < 3, more complex multi-rule procedures or increased QC frequency may be necessary [2].

Result Interpretation

In-Control vs. Out-of-Control

An assay is considered "in control" when QC values fall within ±2 SD and no Westgard rule violations occur. Random scatter around the mean with approximately equal numbers of points above and below indicates stable performance.

An "out-of-control" condition exists when any Westgard rule is violated. The specific rule violated provides clues about the error type:

• 1₃s violation: Suggests a gross error (e.g., wrong QC material, massive pipetting error, instrument malfunction)
• 2₂s violation: Indicates systematic bias (e.g., calibration drift, reagent deterioration)
• R₄s violation: Points to increased random error (e.g., pipetting inconsistency, temperature fluctuation)
• 10x violation: Reflects gradual shift (e.g., reagent evaporation, aging calibrator)

Trend and Shift Detection

A trend (six or more consecutive points moving in one direction) often precedes a formal rule violation. Early detection allows proactive intervention before results become unacceptable. Common causes include reagent evaporation, gradual calibration drift, or accumulating sample residue in the instrument.

A shift (abrupt change in mean level) typically indicates a discrete event such as reagent lot change, recalibration, or instrument maintenance. When a shift occurs, the laboratory must decide whether to accept the new mean or investigate further.

Troubleshooting

Limitations

Retrospective Detection

Traditional Levey-Jennings charts with Westgard rules detect errors after they have occurred. By the time a rule violation is identified, patient results from the affected run may already be reported. This limitation is particularly concerning for assays where small shifts can have clinical significance before triggering rule violations [3].

Sensitivity to Small Shifts

Conventional rule-based systems exhibit limited sensitivity to small but clinically meaningful shifts. For example, a 1% daily bias may accumulate over many days before violating the 10x rule. Machine learning approaches, such as Gaussian Process Regression for systematic bias modeling, can detect such subtle changes earlier, but these methods require computational infrastructure and validation [3].

QC Material Limitations

Commercial QC materials may not perfectly mimic patient sample behavior, particularly for complex matrices or analytes with nonlinear dilution characteristics. Matrix effects can cause QC results to appear stable while patient results drift. Patient-based QC methods, such as moving averages of patient results, can complement traditional approaches by monitoring the actual analytical process [4].

Statistical Assumptions

Control charts assume normally distributed data with constant variance. Some assays exhibit heteroscedasticity (increasing variance with concentration) or non-Gaussian distributions. In such cases, transformation (e.g., logarithmic) or nonparametric control limits may be more appropriate.

Documentation Requirements

Essential Records

Each QC event should generate:

• Date and time of analysis
• QC material lot number and expiration date
• Measured value for each level
• Operator identification
• Instrument and reagent lot information
• Any rule violations observed
• Corrective actions taken

Long-Term Records

Maintain cumulative Levey-Jennings charts for each analyte and QC level. These charts provide historical context for evaluating long-term trends and supporting proficiency testing investigations. Regulatory agencies typically require retention of QC records for at least two years, though institutional policies may specify longer periods.

Change Control

Document any changes that could affect QC performance, including:

• Reagent lot changes
• Calibrator lot changes
• Instrument maintenance or repair
• Software updates
• Operator training events

When a reagent lot changes, run parallel QC with old and new lots for at least 5-10 days to establish new baseline statistics before discontinuing the old lot.

Biosafety Considerations

While control charts themselves are purely statistical tools, the QC materials used in their construction may pose biological hazards. Pooled patient sera, residual clinical specimens, or QC materials derived from human sources should be handled according to institutional biosafety guidelines and the principles outlined in the Biosafety in Microbiological and Biomedical Laboratories (BMBL) manual [5].

For molecular biology laboratories working with recombinant or synthetic nucleic acid molecules, the NIH Guidelines for Research Involving Recombinant or Synthetic Nucleic Acid Molecules provide the regulatory framework for containment and risk assessment [6]. QC materials containing recombinant DNA or RNA must be handled at the appropriate biosafety level determined by institutional biosafety committee review.

General precautions include:

• Treat all QC materials of human origin as potentially infectious
• Use appropriate personal protective equipment (gloves, lab coat, eye protection)
• Perform QC material preparation and aliquoting in a biosafety cabinet when handling infectious materials
• Decontaminate work surfaces after each QC run
• Dispose of QC materials according to institutional biohazard waste protocols

For teaching laboratories operating at BSL-1, QC materials should be limited to non-infectious, commercially prepared controls or synthetic standards that present no biological hazard. The BMBL guidelines emphasize that risk assessment should be performed for each material and procedure, with containment measures proportional to the identified risk [5].

Frequently Asked Questions

How many QC measurements are needed to establish initial control limits?

A minimum of 20 measurements collected over separate analytical runs under stable conditions is standard. Fewer than 20 data points produce unreliable estimates of the mean and standard deviation, potentially leading to control limits that are too narrow (causing excessive false rejections) or too wide (missing true errors). If 20 runs cannot be completed quickly, some laboratories use 10-15 initial points with provisional limits, then recalculate after accumulating 20 points.

Can Levey-Jennings charts be used for qualitative or semi-quantitative assays?

Yes, but with modifications. For semi-quantitative assays like ELISA, the optical density values or calculated concentrations can be plotted on Levey-Jennings charts. The key is to use a consistent QC material with a known target range. The Python program developed for ELISA QC demonstrates that this approach works effectively for semi-quantitative methods, identifying outliers consistent with ANOVA results [1]. For purely qualitative assays (positive/negative), control charts are less applicable, though frequency of positive controls can be monitored using p-charts or u-charts.

What should I do when a Westgard rule violation occurs but I cannot identify the cause?

First, repeat the QC measurement using a fresh aliquot of QC material. If the repeat value is acceptable, the original violation may have been due to random error or a pre-analytical issue with that specific aliquot. If the repeat also violates rules, check instrument function (calibration, maintenance status), reagent integrity (expiration, storage conditions), and operator technique. If no cause is found after systematic investigation, consider contacting the instrument manufacturer or reagent supplier for technical support. Document all steps taken, as regulatory inspectors may review these records.

How do I choose which Westgard rules to apply for my assay?

Rule selection depends on assay performance, measured by the sigma metric. For high-sigma assays (σ ≥ 6), simple rules like 1₃s or 2₂s provide adequate error detection with low false rejection rates. For moderate-sigma assays (σ = 3-5), multi-rule procedures (e.g., 1₂s/2₂s/R₄s/4₁s/10x) offer better sensitivity. For low-sigma assays (σ < 3), the assay itself may need improvement before routine use, or more frequent QC with additional rules may be necessary [2]. The normalized operational specification chart can help visualize the optimal rule set for a given sigma level.

References and Further Reading

1. Wetzel HN, Cohen C, Norman AB, Webster RP. A novel Python program for implementation of quality control in the ELISA. 2017. https://pubmed.ncbi.nlm.nih.gov/28579365/
2. Ahmad MI, Ahmed FS, Rao D, Begum M, Joan S. A comparative study of sigma metrics and statistical quality control rules in clinical biochemistry laboratories using a Cobas pure analyzer. 2026. https://pubmed.ncbi.nlm.nih.gov/42023312/
3. Dash P, Rout S, Koppisetty BK, Panda CR, Priyadarshini D, Roy T, Nayak S. Triple Point Pooled Sera (TriPPS) QC for Laboratory Analyte Error Detection: A Machine Learning based Quality Control in Laboratory. 2026. https://pubmed.ncbi.nlm.nih.gov/42006498/
4. Yang X, Chen Q, Pan Z, Cheng J, Zheng W, Liang Y, Chen H, Chen G, Wang W. Application of Patient-Based Real-Time Quality Control Based on Artificial Intelligence Monitoring Platform in Continuously Quality Risk Monitoring of Down Syndrome Serum Screening. 2024. https://pubmed.ncbi.nlm.nih.gov/38468408/
5. CDC and NIH. Biosafety in Microbiological and Biomedical Laboratories (BMBL), 6th Edition. U.S. Department of Health and Human Services, 2020. https://www.cdc.gov/labs/bmbl/index.html
6. National Institutes of Health. NIH Guidelines for Research Involving Recombinant or Synthetic Nucleic Acid Molecules. https://osp.od.nih.gov/policies/biosafety-and-biosecurity-policy/nih-guidelines-for-research-involving-recombinant-or-synthetic-nucleic-acid-molecules/
7. National Center for Biotechnology Information. NCBI Bookshelf: Molecular Biology and Laboratory Methods. https://www.ncbi.nlm.nih.gov/books/

Related Articles

• Quality Control Analysis: Methods for Monitoring Lab Performance
• Understanding Laboratory Balance Calibration: Types, Procedures, and Quality Checks
• Laboratory Terminology for Beginners: Aliquot, Blank, Control, Replicate, and Standard
• What Is a Laboratory Protocol? Required Sections, Version Control, and Deviations
• Quality Laboratory: Implementing a Quality Management System in Academic Research Labs
• Procedure for Quality Control: Step-by-Step Implementation in a Molecular Lab