A sandwich ELISA giving discordant results on high-analyte and interfering sera
A sandwich ELISA for a serum biomarker uses a capture antibody, a biotinylated detection antibody and streptavidin-HRP, calibrated from 0 to 500 ng/mL with a four-parameter logistic fit; the lower limit of quantification is 15 ng/mL and repeatability at 100 ng/mL is 8% CV. During validation, three serum samples behave oddly. Sample A gives 38 ng/mL neat, 1,180 ng/mL when corrected from a 1:10 dilution and 1,100 ng/mL when corrected from a 1:100 dilution. Sample B gives 90, 95 and 88 ng/mL at the same three dilutions. Sample C gives 460 ng/mL neat and 470 ng/mL corrected from 1:10, but only 44 ng/mL after pre-treatment with a heterophile/rheumatoid factor blocking reagent. The clinical cut-off is 150 ng/mL. The analyst proposes reporting a single 1:10 dilution for every sample to standardise the workflow.
1. Explain the mechanism behind sample A's pattern, name the phenomenon, and state the true concentration and what must be done to report it.
Read the worked answer
Pattern. In a sandwich assay, signal requires antigen to bind capture and detection antibodies simultaneously. At very high antigen concentration, capture sites and detection antibodies are each saturated by separate antigen molecules, so few sandwiches form and the signal falls. The neat sample therefore reads falsely low (38 ng/mL), while dilution brings the antigen into the range where sandwich formation is efficient, giving 1,180 and 1,100 ng/mL at 1:10 and 1:100. The agreement between the two dilutions is the key evidence: the analyte is genuinely present, and the neat result was suppressed.
Name. This is the hook effect (prozone phenomenon), specific to two-site immunometric assays. It is not present in competitive formats.
True concentration. Approximately 1,100–1,200 ng/mL, far above the 500 ng/mL upper calibration limit, so neither the 1:10 nor the 1:100 result is inside the calibrated range. To report a valid number, the sample must be diluted further (for example 1:50 and 1:100) until two dilutions agree within about 20% and both fall on the curve. The concentration must then be reported with the dilution factor and the assay's upper reporting limit noted, or as greater than the highest validated value if dilution linearity has not been demonstrated for this analyte.
2. Distinguish sample A from sample C mechanistically, and state the additional test that confirms each diagnosis.
Read the worked answer
Sample A. Dilution produces proportionality: correcting for dilution gives values that agree (1,180 versus 1,100). This is concentration-dependent suppression, which is what hook kinetics predict. Confirmation is dilution linearity plus spike recovery: adding a known amount of analyte to the diluted sample should recover close to 100%, and running the sample on a different antibody pair should give the same high value. A further confirmation is to show that signal rises then falls as a serially diluted high sample is run neat and at increasing dilutions.
Sample C. Dilution is also linear (460 versus 470 corrected), so this is not a hook. The decisive observation is that a heterophile/rheumatoid factor blocking reagent drops the signal from 460 to 44 ng/mL. Rheumatoid factor and heterophile antibodies bridge the capture and detection antibodies directly, creating signal without antigen; the blocker absorbs these antibodies and abolishes the bridge. Confirmation is to re-run the sample in a different assay format or with a non-competing antibody pair, and to show that the blocker does not change results for a panel of true positive and true negative samples.
Practical distinction. Hook is an analyte-excess artefact of the format and is corrected by dilution; interference is a sample-matrix artefact and is corrected by blocker, re-assay in another format, or rejecting the sample. Dilution alone would not have revealed sample C.
3. Critique the proposal to report a single 1:10 dilution for all samples and give a reporting rule that would prevent both errors.
Read the worked answer
Why the proposal fails. A fixed 1:10 dilution pushes low-concentration samples below the lower limit of quantification: a true 20 ng/mL sample becomes 2 ng/mL, well under the 15 ng/mL LLoQ, so it cannot be quantified and will be reported as either undetectable or, worse, extrapolated. It also risks introducing dilution error into samples that never needed it, and it does not fix sample A: at 1:10 the sample still reads about 118 ng/mL, below the true 1,100 ng/mL and below the 150 ng/mL cut-off, so the clinical classification flips. A universal dilution therefore creates false negatives at both ends of the range.
Reporting rule. Screen every sample neat first. If the neat result lies comfortably within the calibration range and dilutes linearly in a confirmatory dilution, report the neat value. If the neat result exceeds the upper calibration limit, or lies near it, dilute stepwise until two dilutions agree within 20% and both lie on the curve, then report the corrected value with the dilution factor. If the neat and diluted results are discordant, do not report a number until hook is excluded by dilution linearity and interference is excluded by blocker treatment or re-assay in a second format.
Validation support. Establish the hook threshold experimentally by spiking analyte to 10,000 ng/mL and identifying the concentration above which signal falls, and include that threshold in the SOP as the trigger for reflex dilution.