Introduction

Here is a scenario that ruins otherwise good datasets: a sample you expect to be highly positive reads lower than your mid-range standards. You repeat it. Same result. You blame the plate, the reader, the reagents — but the real culprit is the hook effect (also called the prozone effect or high-dose hook effect), a well-documented artifact of one-step sandwich immunoassays.

It is more common than most people think, especially in samples with a huge dynamic range of analyte — cytokines after strong stimulation, hormones, tumor markers, or concentrated bioprocess samples. This guide explains the mechanism, how to recognize it in your data, and the dilution strategy that fixes it.

What actually happens at the plate

A sandwich ELISA relies on two antibodies: a capture antibody immobilized on the well, and a labeled detection antibody. The signal is proportional to the number of capture–analyte–detection "sandwiches" formed.

In a one-step (simultaneous) assay, sample and detection antibody are incubated together. When the analyte is present at extreme excess:

The detection antibodies get saturated by free analyte in solution.
The capture antibodies get saturated by analyte molecules that never meet a detection antibody.
Very few complete sandwiches form — and the signal falls, even though the true concentration keeps rising.

The dose-response curve is no longer monotonic: it rises through the standard range, peaks, and then bends downward. On the descending arm, one signal value corresponds to two possible concentrations — one real, one falsely low. If your sample sits on the descending arm, your software happily interpolates the low answer.

Signal
  ^
  |            ____ peak
  |          _/    \___
  |        _/          \____   <- hook (prozone) region
  |      _/                 \____
  |   _/
  | _/
  +----------------------------> Concentration
     [ normal standard range ]

Who is at risk

One-step sandwich ELISAs — the classic case. Two-step assays (wash between sample and detection antibody) are far less susceptible.
Competitive ELISAs have their own inversion logic, but the prozone artifact is a sandwich-assay phenomenon.
Samples with extreme analyte levels: stimulated cell-culture supernatants, serum tumor markers, recombinant protein production, and any matrix where concentrations can exceed the highest standard by 10–1000×.

How to detect it in your data

The hook effect rarely announces itself. Look for these patterns:

A "hot" sample that dilutes up. You dilute a sample 1:10 and the back-calculated concentration (× dilution factor) comes out higher than the neat reading. That is the signature. In a healthy assay, dilution-corrected results agree within ~±20%.
Biologically implausible values. A strongly induced sample group reads lower than a mildly induced one.
Signal near or above the top standard. Anything at the upper asymptote of the 4PL curve is a candidate — you cannot distinguish "at the plateau" from "past the peak" with a single dilution.

The dilution linearity check

The definitive test is a dilution series of the suspect sample:

Dilution     Signal (OD)   Back-calculated conc. (corrected)
neat         1.95           850 ng/mL
1:4          2.40           2 600 ng/mL    <- rises: hook effect confirmed
1:16         2.10           10 400 ng/mL   <- still rising
1:64         0.85           18 000 ng/mL   <- plateaus: true value region

When dilution-corrected concentrations stop changing with further dilution, you have reached the linear region and can trust the number. If they keep climbing, keep diluting.

Why your curve fit won't save you

A 4PL or 5PL logistic model assumes a monotonic relationship — signal rises (or falls) with concentration, full stop. It has no way to represent the descending arm of a hooked response. So when a hooked sample's OD lands inside the standard range, the fit returns a confident, precisely wrong concentration, often with a narrow confidence interval. The math is fine; the biology of the assay lied to it.

This is also why fitting quality metrics (R², recovery of standards) tell you nothing about hook effects in unknowns — the standards themselves were never in the hooked range.

Prevention and good practice

Always run at least two dilutions for any sample expected to be concentrated (e.g. neat + 1:10). Agreement within ±20% means you are in the linear region.
Choose two-step assay formats when establishing a new method for high-abundance analytes.
Know your assay's hook threshold. Manufacturers of validated kits sometimes state it; if not, spike a sample well above ULOQ and check whether signal drops.
Report honestly. If a sample was re-assayed at higher dilution because of a suspected hook effect, say so in methods — reviewers increasingly ask.
Never report values above ULOQ from a single dilution. Flag them, dilute, re-run.

Speed it up

The Labreador ELISA module flags out-of-range and above-ULOQ samples automatically, tracks per-well dilution factors, and its dilution-linearity checks (together with spike-recovery analysis) make it straightforward to spot samples whose diluted and neat readings disagree — the earliest warning sign of a hook effect. Everything runs locally in your browser; no data is uploaded.

Related guides

If Labreador helps your work, please cite it: Helczman, M. (2026). Labreador - Bioassay Analysis Platform (Version 1.6.0) [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.21676846