Quality control in the lab

A number from a lab is only useful if you can trust it, and quality control is how labs show that. It runs alongside every batch of samples, not just once.

A method blank goes through every step with no sample in it, to catch contamination from reagents or glassware. A check standard of known concentration, run every so often, shows the instrument hasn't drifted. A spiked sample has a known amount added; recovering close to 100% of the spike shows the sample's other contents (its matrix) aren't throwing the result off. Duplicates show precision.

Every method has a limit of detection, LOD, the smallest amount it can tell apart from zero. A common estimate is three times the standard deviation of blank readings, divided by the calibration slope. If blanks scatter with s = 0.002 absorbance units and the slope is 0.10 per mg/L, LOD = 3 × 0.002 ÷ 0.10 = 0.06 mg/L. The limit of quantitation, LOQ, where you can give a reliable number, is usually about ten times s divided by the slope. Below the LOD you report "not detected" (or "less than" the LOD), not zero.

A control chart plots the check standard's result batch after batch, with a centre line at its mean and limits at ±2s (warning) and ±3s (action). A point outside the action limits, or a run of points creeping one way, means stop and find the cause before reporting anything.

Records make it traceable: which instrument, which standards and their batch numbers, who ran it and when, and the raw data. Labs work to written procedures and are audited against standards such as ISO/IEC 17025 or Good Laboratory Practice, so another analyst could repeat the work and get the same answer.