Changing Control Lots: Reset, Continue or Merge? What Happens to Your Levey-Jennings Charts

Category: Basic QC Concepts | Estimated reading time: 11 minutes

Questions That Always Arise When Changing Lots

Open the QC interpretation notes in many laboratories, and "new lot" is almost always on the list of causes. Changing lots does often shift the chart. But there are questions that are rarely fully answered:

• Should the control lines (TV and SD) be reset?
• Should the history of the old lot's chart be continued?
• Or should the data from both lots be combined?

The Preliminary Test article explains how to establish TV and SD from scratch, and the Target Value & SD article explains the consequences of incorrect numbers. This article discusses one specific situation: the control material has already been in use, and then the lot changes.

Short Answer

• TV: reset. The new lot mean is established from new lot data.
• CV: continued from the old lot. No need to re-estimate from scratch, provided the requirements are met (discussed below).
• SD: derived from old CV × new mean. The SD number indeed changes because the mean changes, but what is inherited is the CV. The old SD is not simply copied.
• Chart history: continued as an archive, with a lot-change marker. Data from two lots are not mixed in a single mean calculation.
• Combine: only for estimating long-term CV or SD, for example, pooled SD, not for determining daily control lines.

The reasoning behind each point is in the following sections.

Why a New Lot Shifts the Chart

A new lot of control material is a different production run. The target values and manufacturer ranges are almost always different from the previous lot, even if the material and manufacturer are the same. This means the old lot mean is not the new lot mean.

If the old lot's lines are still used for the new lot, the chart will reflect that mean difference, not the instrument's performance:

• Suppose the old lot has a mean of 120 with an SD of 3.6, and the new lot actually has a mean of 124. A difference of 4 units is equal to about 1.1 SD. All new lot points will sit above the old mean, so the 4₁s and 10x rules will trigger even if the instrument is performing well.
• If the difference is larger, for example, the new lot has a mean of 142, the points will immediately fall far outside ±3SD, and the 1₃s rule will trigger from the first day.

From the simulation in the Target Value & SD article, a reference mean that is off by only 0.5 SD increases the probability of rejection in 20 runs from about 13% to 33% (1₃s+2₂s rule), and from about 29% to 62% (multirule). A mean that is off by 1 SD makes rejection almost certain. One source we reference cites using old lot ranges for a new lot without verification as one of the most common errors in QC management.

Step-by-Step Procedure

1. Prepare the new lot before the old lot runs out. Verification of the new lot should ideally be done before the expiration date of the old lot so that there is an overlap period.
2. Run in parallel. Measure the new lot alongside the lot currently in use. The old lot continues to be used for daily decisions. New lot data are collected separately (example laboratory SOPs suggest separate QC files) so as not to influence decisions.
3. Calculate the new mean. The CLSI C24 guidelines updated in 2016 allow for 10 points for the new lot mean, for example, one measurement per day for 10 days. Older guidelines mention 20 points over 2–4 weeks, so check your laboratory's SOP. For control material that is brand new and has no history, stick to 20 points in 20 days for mean and SD.
4. Derive the SD from the old lot CV.

New SD = New Mean × Old Lot CV

1. Perform a reasonableness check (discussed in a separate section below).
2. Use new control limits. Start using the new mean and SD for QC decisions.
3. Update as data increases. The mean from 10 points is still an estimate. Update it cumulatively as new lot data continue to be collected, and compare the resulting CV with the old lot CV.

Calculation Example

The old lot CV (calculated from stable data) is 3.0%. The old lot mean is 120, so the old SD = 120 × 3.0% = 3.6. From 10 parallel data points, the new lot mean = 142.

• New SD = 142 × 3.0% = 4.26
• ±2SD limits: 133.5 to 150.5
• ±3SD limits: 129.2 to 154.8

If the old SD (3.6) were simply copied to the new lot, that SD would be 15% smaller than it should be, resulting in control limits that are too narrow and more frequent false alarms. That is why Bio-Rad suggests copying the CV, not the SD: SD depends on concentration.

Requirements for Using an Old Lot CV

Inheriting a CV may only be done if:

1. The CV history comes from a stable period. Periods with unresolved issues should not be used as a basis.
2. The CV does not exceed the standard CV (target imprecision or the CV allowed by the laboratory, e.g., derived from TEa or biological variation; see the Preliminary Test and Accuracy, Precision & Monthly Evaluation articles). If the historical CV is already above the standard, that is a precision issue that must be solved first, not inherited.
3. The new lot formulation is similar to the old lot. Bio-Rad guidelines limit the use of old CVs to lots with similar formulations.
4. The CV is derived from calculated QC data, not from an SD that was manually altered in the chart settings. If the old SD was ever widened to stop flags (see the "chasing data" section in the Target Value & SD article), the defect will also be inherited by the new lot.
5. (Author's addition, not a source quote) The new lot mean is not vastly different in concentration level from the old lot. CV is not always the same at all concentrations, so if the new lot is at a very different level, the assumption that the CV is the same needs to be checked.

If the requirements are not met, or the lot is brand new with no history, revert to self-estimation: 20 points in 20 days for mean and SD.

Reasonableness Check Before Using New Limits

• Laboratory mean vs. manufacturer mean. The example laboratory SOP we refer to sets the target laboratory mean to be within 1 SD of the manufacturer's mean. Differences in manufacturer ranges between lots are not a problem as long as they are used for comparison, not as a target (see the Target Value & SD article).
• New CV vs. old CV. Once there is enough new lot data, compare the resulting CV with the old lot CV. If it is significantly larger, find the cause first; do not immediately accept or widen the SD.
• CV vs. standard CV. Ensure it remains below the target imprecision.
• Control material function. This stage also ensures the new lot is working like control material: not degraded, not showing unreasonable scatter.

How Certain Is the Mean from 10 Points?

A mean from 10 points is still an estimate. Assuming ideal conditions (normal and independent data), the standard error of the mean from 10 points is about 0.32 SD. The probability of an estimate being off by more than 0.5 SD is roughly 11%, whereas with 20 points it is about 2.5%. This is the author's calculation, not a figure from a publication.

Implication: in the early weeks of a new lot, 10x or 4₁s rules that appear could be from a temporary mean that is off, rather than from the process itself. Therefore, the mean needs to be updated cumulatively as data increases, and patterns that appear early need to be read with this awareness (see the Levey-Jennings Chart Patterns article).

If There Is No Time for Overlap

Sometimes the old lot has run out or there is no overlap period. Some options discussed in the sources:

• Old Westgard QC articles suggest estimating the new lot mean from a small number of measurements (about 9 to 16), using the SD from the previous lot for temporary limits, then updating the mean as measurements increase.
• For control materials with short shelf lives (difficult to collect long-term data), Bio-Rad mentions pooled SD to combine data from several lots.
• A molecular laboratory study (2025) mentions that historical mean, CV, and SD may be temporarily used for a new lot assuming the lots are comparable. This article leans more toward a more cautious approach: do not use the old lot mean for the new lot, even temporarily, because a mean that is off immediately triggers 10x. What may be temporarily inherited is the CV (for SD), not the mean.

Bio-Rad also notes that the manufacturer's CV can be used initially if there is no time to establish one's own CV, but it is always better to establish your own SD or CV.

Reset, Continue, Combine: What Happens to the Chart

• Reset control lines. Every lot has a new TV and a new SD (result of CV conversion). Old lot lines are not used for the new lot.
• Continue history as an archive. Display a continuous chart with a "lot change" marker (vertical line). This helps during investigations: did the shift coincide with the lot change?
• Do not mix data from two lots for one mean. A combined mean does not represent either lot, and the SD will bloat because it includes the differences between lots. (Author's explanation based on statistical logic.)
• Across-run rules are evaluated within one lot. Rules like 2₂s across runs, 4₁s, and 10x should not calculate points from both the old lot and the new lot in a single sequence, because the reference lines are different. (Author's advice; check the behavior of your QC software.)
• Combine only for long-term estimation. Combining data from several lots (pooled SD) is useful for assessing long-term CV, not for daily control lines.

What About Changing Reagent or Calibrator Lots?

A different situation: the control material is the same, but the analytical system has changed lots. Bio-Rad QCNet reminds us that the mean and SD might need to be re-established when changing reagent or calibrator lots.

Be careful here, as re-establishing the TV can hide real problems:

• If control values shift after changing reagent lots, verify with patient samples (see CLSI EP26 in the New Control Material Arrives article).
• If patient samples do not shift but the control does, the difference is likely characteristic of the control material (matrix effect), and the TV may be updated.
• If patient samples also shift, that is a real bias in patient results. Resolve it through calibration or lot acceptance, not by resetting the TV (the principle is the same as "chasing data" in the Target Value & SD article).

(This flow is the author's synthesis of QCNet statements and EP26 logic, not a standard procedure from a single source.)

Common Errors

• Using old lot control limits for a new lot.
• Using the manufacturer's mean as the TV without your own data.
• Copying the old SD (rather than calculating from CV × new mean).
• Treating the mean from 10 points as a final figure forever.
• Mixing data from two lots to calculate one mean.
• Widening the SD because the new lot triggers flags.
• Not recording the date of the lot change, making shifts difficult to trace.

Summary

• TV is reset from new lot data, SD is calculated from old CV × new mean.
• CV may only be inherited if stable, does not exceed the standard CV, formulation is similar, and it is calculated from real data.
• Run the new lot in parallel with the current lot before use.
• The mean from 10 points is still an estimate; update as data increases.
• History is continued as an archive with a lot-change marker, not mixed.
• Do not use the old lot mean for the new lot, and do not widen the SD to make flags stop.
• Changing reagent or calibrator lots requires verification with patient samples before the TV is updated.

Conclusion

A lot change is the moment when an LJ chart is most easily misleading, in two directions: too many alarms because old lines are used for a new lot, or too few because the SD is widened to dampen them. A simple procedure—new TV from new data and SD from a proven stable CV—keeps detection capabilities consistent from lot to lot and ensures that changes in the chart truly reflect changes in the process.