Calibration Curve Failure in a Bioanalytical Run: Impact on Sample Reporting
A bioanalytical run for drug Z in human plasma includes a calibration curve (8 non-zero standards, 1–500 ng/mL), blank and zero samples, and 40 study samples. The run also includes QCs at 5, 50, and 400 ng/mL (n=6 each). After analysis, the calibration curve back-calculated concentrations show that the 1 ng/mL standard is −25% (outside ±20% at LLOQ), and the 5 ng/mL standard is −18% (within ±15%? Actually ±15% for non-LLOQ). The 500 ng/mL standard is +16% (outside ±15%). The regression line using all standards has r=0.995, but the residuals show a pattern. The QC results: low QC (5 ng/mL) mean accuracy −12%, precision 8%; mid QC (50 ng/mL) mean accuracy +5%, precision 6%; high QC (400 ng/mL) mean accuracy +14%, precision 7%. The run was repeated with a new calibration curve and new QCs, but the same study samples were re-injected. The repeat run had all standards within ±15% (LLOQ within ±20%), and QCs within ±10%. The lab wants to report the repeat run results. The original run had 3 samples above the upper limit of quantification (ULOQ) that were diluted 1:10 and re-assayed; those diluted samples were within the calibration range in the repeat run.
1. Evaluate the acceptability of the original run. What are the issues with the calibration curve and QCs? Should the original run be rejected?
Read the worked answer
The original run has multiple acceptance criteria failures. For the calibration curve: the 1 ng/mL standard is −25%, exceeding the ±20% allowance at LLOQ; the 500 ng/mL standard is +16%, exceeding ±15% for non-LLOQ standards. Typically, at least 75% of standards and a minimum of 6 non-zero standards must be within ±15% (±20% at LLOQ). Here, 6 out of 8 standards are within criteria if we exclude the two failures, which meets the 75% (6/8=75%) and minimum 6. However, the LLOQ standard failure is critical because it defines the lower limit; without an acceptable LLOQ standard, the LLOQ is not validated in that run. Additionally, the pattern of residuals suggests a non-linear response or weighing issue. The QCs: high QC mean accuracy +14% is within ±15%, but individual QC values may exceed ±15%; the low QC −12% is within ±15%. However, the high QC is near the limit. More importantly, the calibration curve failures at the extremes affect the accuracy of samples at low and high concentrations. The run should be rejected because the LLOQ standard failed and the high standard failed, compromising the quantitative range. Regulatory guidance requires that calibration standards meet criteria; failing standards cannot be used to anchor the curve. Although 75% of standards passed, the LLOQ failure means the curve cannot be used to report concentrations near the LLOQ. Therefore, the original run is not acceptable for reporting sample concentrations.
2. The lab proposes to report the repeat run results. Discuss the validity of using re-injected samples from the original run. What additional considerations are needed for the diluted samples?
Read the worked answer
Re-injecting the same sample extracts in a repeat run is acceptable only if the samples were stored under conditions that ensure stability and if the re-injection is part of an investigation. However, the repeat run used the same extracts, which may have undergone degradation or evaporation during storage. The lab must demonstrate that the extracts were stable during the storage period between runs (e.g., in the autosampler). If stability of extracts is not established, the repeat results may not be valid. Additionally, the repeat run's calibration curve and QCs were acceptable, which supports the analytical system's performance at the time of re-injection. But the sample integrity is the key concern. For the diluted samples: the original run had 3 samples above ULOQ that were diluted 1:10 and re-assayed. In the repeat run, these diluted samples were within the calibration range. The dilution integrity must be validated: the dilution QC (e.g., a sample spiked at high concentration and diluted 1:10) should be within ±15% of nominal. If dilution integrity was established during validation, and the repeat run included dilution QCs that passed, then the diluted results can be reported. However, if the repeat run did not include dilution QCs, the dilution step is not verified in that run. Also, the original dilution results cannot be used if the original run is rejected. The lab should ensure that the repeat run includes appropriate dilution QCs and that the extracts were stable. If these conditions are met, the repeat run results can be reported, but the investigation should be documented, and the original run should be invalidated with a clear rationale.