Lot-to-lot variations in a qualitative lateral-flow immunoassay for chronic pain drug monitoring.

Recently, Algeciras-Schimnich et al. highlighted the lot-to-lot variability of a quantitative immunoassay and the failure of frequently used protocols to detect these variations (1). We have discovered that lot-to-lot variations also present an important issue in qualitative immunoassays used for chronic pain monitoring. These variations impact the reliability of these assays in clinical practice. Over 7 months, we analyzed a total of 1681 urine samples across 5 lots of the same CLIA-waived, lateral-flow immunoassay (ABMC RapidTOX®), according to manufacturer instructions. We chose this device based on its low cost, its high negative predictive value, and that it includes an assay for the use of buprenorphine. We also analyzed these samples by liquid chromatographytandem mass spectrometry (LC-MS/ MS) (2). Each lot of immunoassay was used to analyze a similar number of different urine samples (mean across 5 lots 336, range 329–354) with a similar number of samples measured by each lot containing opiates (mean 150, range 140–156), oxycodone (mean 155, range 133–164), or methadone (mean 73, range 70 –78). As the comparative method, LC-MS/MS was deemed the gold standard and used to categorize samples that should test positive for opiates (morphine, hydromorphone, codeine, hydrocodone, 6-monoacetylmorphine, or glucuronide metabolites), oxycodone (oxycodone, oxymorphone, or oxymorphoneglucuronide), or methadone. The immunoassay was considered falselypositivewhentheimmunoassay was positive and there were no detectable relevant analytes by LC-MS/MS, and falsely negative when the immunoassay was negative and the LCMS/MS detected analytes present at concentrations above the cutoffs listed in the immunoassay package insert. Fig. 1 illustrates the falsepositive and false-negative rates for each of the 5 lots across 3 immunoassays on the lateral-flow device. For each immunoassay, there was statistically significant lot-to-lot variability in the false-positive and falsenegative rates, which were calculated as the number of false-positive results divided by the total number of positive results and the number of false-negative results divided by the total number of negative results, respectively. For example, the methadone false-positive rate jumped from a mean of 4% for the first 4 lots to 19% in lot 5 (P 0.001, G-test). As a result, the positive predictive value fell from 84% (mean across lots 1– 4) to 51% in lot 5. The prevalence of drug positivity would affect false-positive rates, false-negative rates, and the positive predictive value, but there were no between-lot differences in drug positivity rate for the analytes and samples analyzed in this study (P 0.57 for each immunoassay, G-test). Not all immunoassays on the device were affected similarly by lot-to-lot variations. For example, the oxycodone false-positive rate was 26% in lot 2 and 16%