Failed Duplicate Pair on a 5-Day Charcoal Test
You are conducting a 5-day charcoal canister test in a 2,400-square-foot ranch with a finished basement. Outdoor temperature is 12°F and the furnace runs constantly. You deploy two side-by-side canisters in the basement family room, 6 inches apart and 20 inches above the floor, plus one field blank that stays sealed until analysis. Both canisters are opened within one minute of each other; both sit within 3 feet of a cold-air return. The lab reports 4.8 pCi/L for canister A, 6.4 pCi/L for canister B, and 0.3 pCi/L for the blank. Your QA plan requires 10% duplicates, 5% blanks, and 3% spikes. This is the second duplicate pair this month that has failed to agree, and the client needs to close in 10 days. The lab's lower limit of detection for charcoal is 0.6 pCi/L.
1. Compute the relative percent error (RPE) for the duplicate pair and decide whether the pair passes your QA criteria. Show the calculation.
Read the worked answer
RPE = |A − B| ÷ [(A + B) ÷ 2] × 100 = |4.8 − 6.4| ÷ [(4.8 + 6.4) ÷ 2] × 100 = 1.6 ÷ 5.6 × 100 = 28.6%. The 25% duplicate criterion is exceeded, so the pair fails. The blank at 0.3 pCi/L is below the lab's 0.6 pCi/L lower limit of detection, so it does not indicate contamination, but it should still be recorded. Before deciding the cause, check the obvious: were the canisters actually side-by-side at the same height, were both opened and closed within the allowed window, was one closer to the cold-air return, and did the lab report any holding-time or transit issue? Six inches apart is acceptable spacing, but proximity to a return can create local variability. Because this is the second failing pair this month, also look for a batch, shipping, or lab pattern rather than assuming the house is the problem.
2. What do you do with the result - report 4.8, report 6.4, average to 5.6, or invalidate? Justify your choice.
Read the worked answer
Do not simply average 4.8 and 6.4 and report 5.6 pCi/L as if it were a valid result - averaging masks the disagreement and makes the uncertainty disappear. The defensible options are: (1) treat the duplicate pair as a failed QA check and repeat the test in the same location with a new pair of devices or a CRM; (2) if a result must be reported while the retest is pending, report both values and state that the duplicate agreement criterion was not met; (3) if your written protocol allows a conservative interpretation, use the higher value for decision-making, but say so explicitly. With a 10-day closing, a new 48-hour CRM test is feasible and is the cleanest path. Note the QA failure in the report and keep the lab sheets.
3. Your company is planning a 40-home project with one device per home. How many duplicate, blank, and spike devices should the QA plan include, and what else should you audit given the repeat failure?
Read the worked answer
For 40 homes with one device each, the plan needs at least 10% duplicates = 4 duplicate devices, 5% blanks = 2 blanks, and about 3% spikes = 1 or 2 spikes, so budget roughly 47 devices, not 40. Duplicates go side-by-side in the same location for the same exposure period, blanks stay sealed and travel with the shipment, and spikes are prepared by the lab. Also verify the lab's LLD, canister shelf life, chain-of-custody, and holding times. Because two duplicate pairs failed in one month, audit the batch: the same lab lot, the same courier, or the same field technician could be contributing. Review whether the second failure also involved a cold-air return or a low concentration, where precision is inherently harder.