How to Measure Cytokines: Key Methods and Best Practices

Cytokine measurement spans a wide toolkit, from classic single-target immunoassays to platforms that profile hundreds of proteins at once, and the “best” method depends on what you need: sensitivity, throughput, cost, or the ability to look at many cytokines simultaneously. The most established approach remains the enzyme-linked immunosorbent assay, but newer technologies like digital immunoassays and proximity extension assays are reshaping what is possible. Choosing the right method is only half the challenge, though, because how you collect, store, and handle your samples can distort results just as much as picking the wrong platform.

The Classic Workhorse: ELISA

ELISA has been the go-to method for decades and remains the most widely used and best-validated approach for cytokine quantification.¹ The sandwich ELISA format works by capturing a target cytokine between two antibodies: one fixed to a plate surface and one carrying a detectable label. When the label generates a signal proportional to the amount of captured protein, you get a concentration readout. This method delivers strong signal amplification with high sensitivity and specificity in complex biological fluids like serum, plasma, and cell culture supernatant.²

The main limitation is that a standard ELISA measures only one cytokine per well.¹ If you need to profile a dozen cytokines, that means a dozen separate assays, each consuming sample volume and bench time. For studies with limited sample or tight budgets, this quickly becomes impractical. ELISA also typically detects cytokines down to about the low picogram-per-milliliter range, which is sufficient for many inflamed or stimulated conditions but often falls short for healthy baseline concentrations, where many cytokines circulate well below the assay’s detection threshold.

Multiplex Bead-Based Assays

Multiplex bead array assays solve ELISA’s one-at-a-time problem by coating different populations of color-coded microspheres with different capture antibodies, then reading them on a flow-based analyzer. This lets you quantify large numbers of analytes from a single small sample in an automated 96-well plate format.³ Luminex-based platforms, the most common example, routinely handle panels of 20 to 50 or more targets at once.

The trade-off is that multiplexing introduces new variables. The specific antibody clones used for capture and detection matter enormously; head-to-head studies show that getting results comparable to ELISA depends on carefully matching those clones.³ A multisite comparison of four high-sensitivity multiplex kits found that no single panel detected all cytokines and that there were highly significant differences between laboratories and between manufacturing lots.⁴ That study cautioned that current multiplex assays may not be reproducible enough for repeated measurements across long-term studies or multiple labs, but they work well for longitudinal studies in which relative changes matter more than absolute concentrations.

Recovery rates also differ by platform. A comparison of multiplex kits in a pediatric cohort found that among ten shared analytes, one high-sensitivity kit had high recovery for all of them, while two competitors achieved high recovery for only about half or fewer.⁵ So platform choice at this level is not just about branding; it directly affects data quality on an analyte-by-analyte basis.

Ultra-Sensitive Digital Immunoassays

Some cytokines circulate at concentrations far below what conventional ELISA or multiplex beads can detect, particularly in healthy individuals or in cerebrospinal fluid. Digital ELISA technologies were developed to break through that floor. The Single Molecule Array platform, known as Simoa, isolates individual immunocomplexes in femtoliter-sized reaction chambers and counts them digitally rather than measuring a bulk signal. This approach achieves detection sensitivity in the attomolar range.⁶ Optimized Simoa assays for ten common cytokines have achieved limits of detection as low as 90 attomolar, enabling reliable quantification in healthy serum at levels far below what conventional ELISA can see.⁷

Researchers have pushed sensitivity even further with droplet digital ELISA, which compartmentalizes reactions into picoliter-sized droplets using microfluidics. This method detects proteins in the low attomolar range and is up to 25 times more sensitive than Simoa, which itself is roughly a thousand times more sensitive than conventional ELISA.⁸ These platforms remain specialized and relatively expensive, so they are not the default choice for every study. But for biomarker discovery, low-abundance analytes, and clinical contexts where baseline concentrations matter, they fill a gap that nothing else can.

Proximity Extension Assay and High-Throughput Proteomics

The Proximity Extension Assay, commercialized as the Olink platform, takes a fundamentally different approach. Instead of relying on a sandwich of two antibodies to generate a signal, it attaches short DNA tags to paired antibodies. When both antibodies bind the same target protein, the DNA tags come close enough to hybridize and extend, producing a unique DNA sequence that gets amplified and read by sequencing. One report demonstrated parallel measurement of nearly 1,500 proteins across 96 samples in a single run, generating close to 150,000 data points.⁹

This scale moves cytokine measurement into proteome-wide territory. Rather than selecting a fixed panel of targets in advance, you can screen broadly and then focus on whatever signals emerge. Comparative analyses have recognized Olink’s combination of sensitivity and multiplexing capacity, particularly when many cytokines need quantitation simultaneously.¹⁰ The trade-off is that standard Olink readouts report relative protein expression values rather than absolute concentrations. Newer quantitative versions are narrowing that gap, but if your study requires exact picogram-per-milliliter values for regulatory submission or clinical decision-making, you need to verify that the specific assay format provides them.

Single-Cell and Functional Assays

All of the methods above measure cytokines in bulk: you analyze a tube of serum, plasma, or culture supernatant and get a total concentration. That tells you how much cytokine is present but not which cells produced it. For questions about immune function, particularly in vaccine research, transplant monitoring, and T-cell therapy, you need cell-level resolution.

The two main options are ELISpot and intracellular cytokine staining by flow cytometry. ELISpot captures secreted cytokines on an antibody-coated membrane, producing visible spots that each represent a single responding cell. Intracellular staining blocks secretion with a transport inhibitor, then labels cytokines inside the cell and reads them on a flow cytometer, which also captures surface markers and can identify the specific cell subtype responsible. Both methods are highly sensitive for detecting antigen-specific T-cell responses.¹¹

The two approaches do not always give equivalent results, however. Head-to-head comparisons for viral-specific T cells, including SARS-CoV-2-specific responses, have shown that the methods sometimes diverge, likely because ELISpot reflects cumulative secretion over hours while intracellular staining captures a snapshot at the time of fixation.¹¹ A separate study of cytomegalovirus-specific T cells found strong correlation between ELISpot spot counts and intracellular staining of CD3+ cells, suggesting that for some antigens the two track well together.¹² In practice, the choice often comes down to whether you need phenotypic data on the producing cell (flow cytometry wins) or simply a sensitive count of responding cells (ELISpot is simpler and cheaper).

Mass Spectrometry as an Antibody-Free Alternative

Every method discussed so far relies on antibodies to recognize cytokines, which means you are limited by antibody availability, cross-reactivity, and the accuracy of your standards. Targeted mass spectrometry, specifically multiple reaction monitoring, offers an antibody-free route. Instead of recognizing a protein by its binding to an antibody, it identifies proteins by the mass of their component peptides after enzymatic digestion.

One group developed an MRM assay targeting 79 peptides representing 23 human cytokines and was able to identify all 23 in spiked cell culture media.¹³ Using nano-flow chromatography coupled to a triple-quadrupole instrument, another team achieved calibration curves spanning four orders of magnitude with detection limits in the low attomol-per-microliter range and a median technical coefficient of variation of just 2.2%.¹⁴

The appeal of mass spectrometry is its independence from antibodies: you do not need matched pairs of monoclonals, you can confirm protein identity by sequence, and multiplexing is limited mainly by chromatographic separation rather than by bead colors or plate wells. The downside is instrumentation cost, throughput, and the fact that sensitivity, while excellent for concentrated culture supernatants, still falls short for serum cytokines at their low resting concentrations in healthy individuals.¹⁵ For now, mass spectrometry occupies a niche as a confirmatory or discovery tool rather than a routine clinical assay.

Point-of-Care Biosensors

Most cytokine assays require a laboratory, trained staff, and hours or days of turnaround. For acute clinical settings like sepsis management or cytokine release syndrome in cell therapy, that timeline can be too slow. Point-of-care electrochemical biosensors aim to bring cytokine measurement to the bedside. These devices integrate microfluidic sample handling with miniaturized sensors, enabling low sample consumption and multiplexed detection in a compact format.¹⁶

One prototype designed for sepsis endotyping simultaneously measures five cytokine biomarkers from a single drop of undiluted plasma, with a detection limit of about 1 picogram per milliliter and results in under five minutes.¹⁷ These devices are not yet in widespread clinical use, but they represent where the field is heading: fast, small-volume, multiplexed, and deployable outside the central lab.

Sample Handling Can Make or Break Your Data

No matter how sophisticated the measurement platform, pre-analytical variables routinely introduce more error than the assay itself. The type of collection tube, time to centrifugation, storage temperature, and number of freeze-thaw cycles all affect measured cytokine levels.

Serum and plasma are not interchangeable. Some cytokines released by platelets during clotting are much higher in serum than in plasma, and the choice of anticoagulant also matters. A study comparing serum with three types of plasma found that platelet-derived mediators like CCL5 and CD40L were markedly elevated in serum, while for many other cytokines the two matrices showed strong correlations in relative levels even when absolute concentrations differed.¹⁸ Unsupervised clustering in that study showed that the effect of sample type was strongest for serum and heparin plasma samples, meaning these two tend to drift the furthest from each other.

Freeze-thaw cycles are another well-known hazard. One study found that certain cytokines, including IL-1β, IL-4, and IL-10, increased significantly after a single freeze-thaw cycle and then stayed at those elevated levels through nine more cycles, while other cytokines were less affected.¹⁹ A larger analysis of circulating inflammation markers reported that the first additional freeze-thaw cycle introduced only small, generally non-significant changes, but by the third cycle the absolute concentrations shifted enough to be statistically detectable, though the rank ordering across subjects was usually preserved and the median change was around 10%.²⁰ The practical takeaway: aliquot samples before freezing so that each vial is thawed only once before measurement, and separate serum or plasma from the cellular fraction as soon as possible after collection.¹⁹

Time of Day Matters More Than You Might Expect

Cytokines are not static. Their circulating levels follow circadian patterns driven largely by the daily cortisol cycle. Cortisol peaks in the early morning and suppresses production of pro-inflammatory cytokines, so those cytokines tend to be at their lowest in morning blood draws and climb later in the day as cortisol falls.²¹ Stimulated whole-blood production of IFN-γ, TNF-α, IL-1, and IL-12 has been shown to peak in the early morning hours and inversely track cortisol.²²

For IL-6 in particular, a meta-analysis found that the most consistent diurnal feature was a morning trough, in contrast to earlier reports that had pointed to an evening or nighttime peak.²³ If you are comparing cytokine levels between groups or across time points, inconsistent sampling times can easily confound the results. Standardizing blood draws to the same clock time, or at minimum documenting and adjusting for it, is a simple step that prevents a surprising amount of noise.

Why Different Kits Give Different Numbers

One of the most frustrating aspects of cytokine measurement is that two labs using different commercial kits will often get different absolute concentrations from the same sample. A study that standardized three commercially available multiplex kits against WHO international reference standards found striking differences in reported values between manufacturers.²⁴ The differences traced in part to the calibration standards shipped with each kit: recombinant proteins produced in different expression systems carry different glycosylation patterns, and those modifications can change how the protein interacts with antibodies, altering the dose-response curve. The broader implication is that quantitative statements about cytokine levels need to be interpreted in the context of the specific assay platform used.

Fitting the standard curve matters too. Most cytokine immunoassays generate sigmoidal dose-response curves that are fitted with four-parameter or five-parameter logistic models. The five-parameter model adds an asymmetry factor. When the true curve is symmetric, using the five-parameter version costs essentially nothing in accuracy, but when the curve is genuinely asymmetric, the four-parameter model can sometimes perform better because of how bias and variance trade off at the extremes.²⁵ Software defaults vary, so it is worth checking which model your analysis pipeline uses and whether it fits your specific data well, particularly for samples near the lower or upper limits of quantification.

Interference You May Not See Coming

Even a perfectly run assay can give misleading results if the sample itself contains interfering substances. Endogenous anti-cytokine antibodies and soluble cytokine receptors can bind the target protein before the assay antibodies do, lowering the apparent concentration. One study found evidence of significant negative interference from either endogenous antibodies or soluble receptors in about 10% of samples for IL-6 and about 26% for TNF-α.²⁶ This kind of interference is invisible unless you run spike-recovery experiments, where you add a known amount of cytokine to the sample and check whether you get it back. Routine spike-recovery checks are not standard in many research labs, which means a fraction of published cytokine values may be artificially low without anyone realizing it.

Therapeutic antibodies introduce another layer of interference. Patients receiving IL-6 pathway blockers like tocilizumab or siltuximab for cytokine release syndrome often have their IL-6 levels monitored, but the drugs themselves can interfere with the detection antibodies in the assay, giving falsely high or falsely low readings depending on which epitopes overlap.²⁷ Clinicians using cytokine panels to guide treatment decisions need to know whether the assay has been validated in the presence of the specific drug their patient is receiving.

Choosing the Right Platform for Your Question

With so many options, the decision usually comes down to a few practical questions: how many cytokines do you need, how sensitive does the assay need to be, how many samples are you running, and what is your budget? For high-throughput drug screening, cytometric bead arrays have been highlighted as the most cost-effective option, particularly with reduced-volume protocols, owing to their homogeneous format and strong robustness metrics.²⁸ For ultra-sensitive clinical work, Simoa or electrochemiluminescence-based S-plex assays are recommended as priority platforms.¹⁰ For broad proteome-wide screening, Olink’s proximity extension assay offers unmatched multiplexing. And for a quick, reliable answer on one or two well-characterized cytokines, classic ELISA still delivers.

Whatever platform you choose, fit-for-purpose validation on a per-analyte basis is essential. The same platform that excels for IL-6 may underperform for IL-4, and the same kit lot that worked last month may drift the next. Running your own spike-recovery experiments, including quality-control samples on every plate, and documenting the time of sample collection will protect your data far more than choosing the most expensive technology on the market.

Cytokine Monitoring in Cell Therapy

One area where accurate, rapid cytokine measurement has direct clinical stakes is chimeric antigen receptor T-cell therapy. CAR-T therapy can trigger cytokine release syndrome, a potentially life-threatening inflammatory cascade. A study of patients treated for acute lymphoblastic leukemia found that peak levels of 24 cytokines in the first month after infusion were strongly associated with severe CRS, and a regression model built on just three cytokines could accurately predict which patients would develop the severe form. Conventional biochemical markers like C-reactive protein and ferritin tracked with CRS but could not predict its severity.²⁹ This kind of application is driving demand for faster turnaround and bedside-capable assays, since waiting overnight for a multiplex result is not practical when a patient’s condition can deteriorate within hours.

Measuring Cytokines in Animal Models and Tissues

Much of what we know about cytokines comes from animal research, but measuring them outside of human serum brings additional complications. Antibody reagents developed for human cytokines do not automatically work in other species. An investigation of 11 monoclonal antibodies raised against ovine, bovine, and human cytokines found that only four cross-reacted broadly across species including sheep, cattle, goats, swine, horses, dogs, and rabbits. Others cross-reacted weakly or not at all.³⁰ Researchers working with non-standard species often need to validate each antibody pair independently, which adds considerable front-end work.

Tissue-level cytokine measurement adds a physical extraction step before the assay even begins. For mouse brain tissue, for example, published protocols call for mechanical homogenization with zirconia beads in lysis buffer, followed by high-speed centrifugation to remove debris, before the clarified extract goes onto a multiplex bead array.³¹ The lysis conditions, protein content of the homogenate, and normalization strategy all introduce variability on top of whatever the assay contributes. Tissue measurements are therefore harder to compare across studies than serum values, and reporting raw concentrations without normalizing to total protein or tissue weight makes the numbers nearly uninterpretable.

The Standard Curve Problem Nobody Talks About Enough

Every immunoassay relies on a standard curve built from recombinant protein diluted into buffer. The assumption is that the recombinant protein behaves the same way as the native cytokine in your biological sample. It often does not. Native human cytokines carry sugar modifications whose patterns differ from those on recombinant proteins produced in bacterial or insect cells. These glycosylation differences can alter how the protein folds, how antibodies recognize it, and even how it behaves during storage. When the standard and the sample contain structurally different versions of the same protein, the reported concentration becomes an approximation rather than a true measurement. This is one reason why results from different kit manufacturers diverge even when measuring the same sample, and why absolute cytokine values should be treated with more caution than relative comparisons within the same assay run.