QA Review: Duplicate, Blank, and Spike Failures in One Charcoal Batch
At annual QA review for a small measurement firm, you examine 120 charcoal short-term tests, 12 duplicates, 6 field blanks, and 3 spikes. One duplicate pair from a basement reads 4.6 and 7.8 pCi/L. A field blank reads 0.8 pCi/L. One spike returns 78% recovery; the other two are 96% and 102%. Duplicate relative percent differences (RPDs) for the other 11 pairs are 3–8%. The lab says the charcoal batch is new and the CRM used for the spike has not been cross-checked since last year. You have 14 client results from the same batch pending release and one real-estate closing in 24 h. The firm's SOP says do not release results until QA review is complete. The owner wants to release the 14 results because only one duplicate failed.
1. Calculate the duplicate RPD and evaluate the QC data.
Read the worked answer
RPD = |4.6 − 7.8| ÷ ((4.6 + 7.8)/2) × 100 = 3.2 ÷ 6.2 × 100 = 51.6%. That is far above the typical 10% acceptance criterion for results at or above 4 pCi/L, so the duplicate fails. The field blank of 0.8 pCi/L also fails the usual 0.5 pCi/L blank criterion, suggesting contamination or background in the batch, shipping, or analysis. The 78% spike recovery is outside a typical 90–110% recovery window, while the 96% and 102% spikes pass. Because the other 11 duplicates are good, the problem may not be universal, but a failed blank and a failed spike in the same batch indicate a batch-level or system-level issue rather than a single random outlier.
2. Should the 14 pending results be released? Justify your decision.
Read the worked answer
No. The SOP says hold results until QA review is complete, and the QA data do not support release. A single duplicate failure might be investigated while other results are released, but here the blank and spike also fail. Those failures affect the batch's accuracy and background, not just one location's precision. Releasing the 14 results could give clients invalid numbers and could force an unnecessary or inadequate real-estate decision. Hold the results, notify the affected clients and the closing agent that release is delayed for QA, and give a realistic timeline. The closing deadline is a business pressure, not a technical reason to release unverified data.
3. What corrective actions should be prioritized, and how should affected clients be handled?
Read the worked answer
Prioritize the blank first because it can bias every result in the batch: inspect storage, shipping, and analysis background; run new blanks. Next address the spike failure by cross-checking or recalibrating the CRM and reviewing spike preparation; a 78% recovery suggests a low bias. Then investigate the duplicate by checking placement differences, device handling, and lab precision. Suspend release of the batch, re-analyze or re-test affected locations as appropriate, and document the corrective action. For clients, explain that QA failed and that their results will be reissued only after correction. Do not simply average the duplicate pair or subtract the blank without a validated correction.