In vitro cytotoxicity testing fails peer review more often for design reasons than for technique. The wrong assay answered the question, or the concentration range could never have supported the IC50 that was reported.
This guide covers what each assay physically measures, how to pick a cell line and seeding density, how to build a concentration series that produces a fittable curve, where interference comes from, and what ISO 10993-5 requires when the work is regulatory rather than academic.
Written from how we run in vitro cytotoxicity testing at ACME Research Solutions, across 300+ assays for research groups in six countries.
What in vitro cytotoxicity testing actually measures
The term covers at least four different measurements, and journals increasingly ask authors to be specific about which one they made.
Metabolic activity. MTT, MTS, WST-8 and resazurin all report the reductive capacity of the cell population. Cells that are alive but metabolically suppressed read as dead. Cells that are stressed and hypermetabolic can read as more viable than untreated controls, which is where those confusing viability points above 100% at low concentrations come from.
Membrane integrity. Trypan blue, LDH release and propidium iodide report whether the plasma membrane has been breached. This is a late event. A compound that triggers apoptosis at 24 h may show almost no LDH release until secondary necrosis sets in.
Total biomass. Sulforhodamine B and crystal violet stain protein or adherent cell mass. No metabolic component at all, and the endpoint is stable for weeks once fixed. The NCI-60 screen uses SRB for this reason.
Growth inhibition. This needs a time-zero reference plate. Without one you cannot distinguish a compound that killed half the cells from a compound that simply stopped them dividing.
A compound can suppress MTT signal by 60% while LDH release stays at baseline. Both results are correct. They are answering different questions, and reporting only one of them is the most common reason reviewers ask for additional experiments.
Choosing the assay for in vitro cytotoxicity testing
Assay choice is the first real decision in any in vitro cytotoxicity testing workflow, and it is the one that quietly determines what the rest of the study can claim.
| Assay | Endpoint | Read | Destructive | Notes |
|---|---|---|---|---|
| MTT | NAD(P)H-dependent reduction | 570 / 630 nm | Yes | Insoluble formazan needs a solubilisation step |
| MTS / XTT | Same, soluble product | 490 nm | No | Needs PMS electron coupling; higher reagent cost |
| WST-8 (CCK-8) | Extracellular reduction | 450 nm | No | Add-and-read, low reagent toxicity, good for kinetics |
| Resazurin | Reduction to resorufin | Ex 560 / Em 590 | No | Same wells readable at 24, 48 and 72 h |
| LDH release | Membrane rupture | 490 nm | Supernatant | Needs a maximum-release lysis control |
| SRB | Total cellular protein | 510–565 nm | Yes | Fixed plates stable for weeks; adherent cells only |
| Neutral red | Lysosomal accumulation | 540 nm | Yes | Basis of OECD TG 432 phototoxicity |
| ATP luminescence | Intracellular ATP | Luminescence | Yes | Most sensitive; detects very small cell numbers |
MTT remains the default in academic pharmacology because it is inexpensive and every reviewer recognises it. It also has the narrowest interference tolerance of anything on this list, which matters if you are testing plant extracts or nanoparticles.
For suspension cells, skip MTT. The aspiration step before solubilisation loses cells, and the loss is not uniform across a plate. Use WST-8 or resazurin instead.
For a 72 h exposure where you also want intermediate timepoints, resazurin lets you read the same wells repeatedly. That single change removes plate-to-plate variability from your time course.
We use resazurin for time-course work, SRB for growth inhibition studies against a T0 plate, and MTT where the study needs to sit alongside published comparators. Details of how we run each are on our in vitro research services page.
Send us the compound class and your research question. We will tell you which endpoint holds up in review, and which one will waste your budget.
Ask a questionCell line and passage
Two questions decide the cell line for in vitro cytotoxicity testing: what tissue does your compound target, and what will a reviewer accept as a comparator.
For anticancer screening, most groups run a paired design: one or two cancer lines plus one non-cancerous line to calculate a selectivity index. MCF-7 against MCF-10A, or HepG2 against a normal hepatocyte line, tells a reader far more than an IC50 against a single cancer line ever will.
For skin, cosmetic and topical formulation work, HaCaT keratinocytes and 3T3 fibroblasts are the conventional pairing. For device extracts, ISO 10993-5 points to L929.
Approximate doubling times, which you will need for the seeding calculation:
| Cell line | Doubling time | Cell line | Doubling time |
|---|---|---|---|
| L929 | 15–20 h | Vero | ~24 h |
| 3T3 | ~20 h | HaCaT | 24–28 h |
| A549 | ~22 h | MCF-7 | 29–38 h |
| HeLa | 20–24 h | HepG2 | 48–60 h |
These are literature ranges. Your own line under your own serum lot will differ, sometimes by a third. Measure it once and use your own number. Our cell doubling time calculator does the fit from a simple cell-count series.
On passage number: record it, report it, and keep the working range narrow. HepG2 and MCF-7 both drift in drug sensitivity across extended passage. A study that runs from passage 12 to passage 45 has an uncontrolled variable in it, and if your replicates disagree, this is the first place to look.
Seeding density decides your in vitro cytotoxicity testing result
More in vitro cytotoxicity testing is ruined by seeding density than by any other single parameter, and it is almost never reported in enough detail to reproduce.
Overseed and the control saturates. Contact inhibition kicks in, metabolic activity per cell falls, and your treated wells look less different from control than they are. The dose-response curve flattens and the fitted IC50 shifts upward. Underseed and the signal-to-background ratio collapses, well-to-well variation climbs, and the bottom plateau of your curve becomes noise.
Starting points for a 96-well plate, 100 µL per well:
| Exposure | Fast-dividing lines | Slow-dividing lines |
|---|---|---|
| 24 h | 1.0 × 10⁴ cells/well | 1.5 × 10⁴ cells/well |
| 48 h | 5 × 10³ – 7.5 × 10³ | 1.0 × 10⁴ |
| 72 h | 2.5 × 10³ – 5 × 10³ | 7.5 × 10³ |
Run a density optimisation plate once per line, per serum lot, before any in vitro cytotoxicity testing that matters. Seed a range, incubate for your intended exposure duration plus the 24 h attachment period, and read. Pick the density that lands in the linear part of the signal-versus-density plot with an optical density around 0.8 to 1.2 for MTT. That plate takes an afternoon — and it is the difference between a clean sigmoid and three months of replicates that will not agree.
Two practical points. Allow 24 h for attachment before dosing, so the cells are in log phase when the compound arrives. And do not use the perimeter wells. Evaporation across a 72 h incubation produces a concentration gradient in row A, row H and columns 1 and 12 that has nothing to do with your compound. Fill them with sterile PBS and treat the inner 60 wells as your plate.
Seeding density is the first thing we check. We run optimisation plates as a standalone service before the main study, so the real experiment only runs once.
Tell us about your cell lineExposure duration
Set it from the doubling time, not from convention.
A 24 h exposure on HepG2 covers less than half a division cycle. Any antiproliferative compound will look weak, because the mechanism has not had time to express itself. On L929, 24 h covers more than one full cycle. The same compound at the same concentration will produce different apparent potency in the two systems, and neither number is wrong.
Standard practice for in vitro cytotoxicity testing is 24, 48 and 72 h. If you can only justify one, pick the duration that covers two doubling times for your line, and say in the methods why you chose it.
Solvent, vehicle and the controls that matter
DMSO above 0.5% is cytotoxic to most adherent lines. Above 1% it is cytotoxic to all of them. Keep the final concentration at or below 0.1% where the compound’s solubility allows, and cap it at 0.5%.
The vehicle control must contain the DMSO concentration present in your highest dose well, not an average. If your top concentration carries 0.5% DMSO and your vehicle control carries 0.1%, part of the effect you are measuring is solvent.
Every in vitro cytotoxicity testing plate needs the same minimum control set:
- Blank wells. Medium and reagent, no cells. Subtract this from everything.
- Untreated control. Cells, medium, no compound, no solvent.
- Vehicle control. Cells plus solvent at the highest working concentration.
- Positive control. A compound with a known IC50 in your line. Doxorubicin, cisplatin or 5-fluorouracil for cancer lines; sodium dodecyl sulfate or hydrogen peroxide for general cytotoxicity.
- Compound interference control. Compound plus reagent, no cells, at every concentration tested.
That last one is the control most often skipped, and the one that most often explains an impossible result.
Normalise to the vehicle control, not the untreated control. Report both.
Interference: where impossible results come from
Tetrazolium reduction is a redox reaction. Anything in the well capable of reducing MTT will produce formazan whether or not a cell is involved. This is the failure mode that most often invalidates in vitro cytotoxicity testing on natural products.
Reducing compounds. Ascorbic acid, glutathione, NADH, reducing sugars and most polyphenols reduce tetrazolium salts directly. Test a polyphenol-rich plant extract by MTT without the cell-free control and you can get viability rising above 150% of control at high concentrations. The extract has not stimulated proliferation. It has reduced your dye.
This is a routine problem for anyone working with herbal extracts, and the reason we pair MTT with an orthogonal endpoint on extract work. Quantifying antioxidant capacity separately, by DPPH or a cell-based antioxidant assay, tells you how much interference to expect before you commit to a design.
Optical interference. Curcumin, anthocyanin-rich extracts and any coloured or turbid sample absorb in the 500 to 600 nm window. Phenol red in the medium absorbs near 560 nm. Use phenol red-free medium for the assay step, and always take a reference reading at 630 to 690 nm.
Nanoparticles. Silver, gold, iron oxide and carbon nanomaterials interfere by more than one route. They adsorb formazan onto their surface, they catalyse tetrazolium reduction directly, and they scatter light at the read wavelength. For nanoparticle work, run the cell-free control at every concentration, and confirm with a non-tetrazolium endpoint such as LDH or ATP. If you are characterising nanoemulsions or SEDDS, the same caution applies to the surfactant system, which can be cytotoxic independently of your active. Our nanoemulsion HLB calculator and Box-Behnken design tool are built for that optimisation step.
Serum binding. Highly lipophilic compounds bind albumin in FBS. Free concentration at the cell surface is lower than nominal concentration in the well, sometimes by an order of magnitude. Two labs using 5% and 10% FBS will report different IC50 values for the same molecule. Report your serum percentage.
Precipitation. A compound that comes out of solution above a certain concentration produces a curve that flattens at the top for physical rather than biological reasons. Check the wells under the microscope before you read the plate.
These are the three categories where MTT alone will mislead you. We design the orthogonal endpoint and the interference controls into the study from the start.
Discuss your sampleIC50, GI50, LC50
In vitro cytotoxicity testing produces all three, and they are not synonyms. Reviewers in oncology and toxicology will catch the substitution.
IC50 is the concentration producing 50% of the maximum inhibitory response for whatever you measured. It is defined against your own control, so it is only interpretable if you state the assay, the exposure duration and the cell line.
GI50 is the concentration causing 50% reduction in net cell growth. It requires a time-zero plate, read at the moment of dosing, and it separates growth inhibition from cell killing.
LC50 is the concentration producing 50% net cell loss relative to time zero. Below the TGI value, the population has shrunk.
If you did not run a T0 plate, you measured IC50. Report it as IC50.
Building a curve that can be fitted
A four-parameter logistic model needs data to constrain four parameters. That imposes real requirements on the concentration series:
- Eight concentrations minimum, spaced logarithmically or half-logarithmically. Linear spacing wastes points where nothing happens and leaves the transition region undersampled.
- At least two concentrations on each plateau. Without an upper plateau the model cannot estimate Top, and the fitted IC50 becomes an extrapolation. This is the single most common defect in submitted in vitro cytotoxicity testing data.
- A range that brackets the transition. If your lowest concentration already gives 40% inhibition, your IC50 sits below your tested range. Run a pilot at wide log spacing first, then a narrow series around the transition.
- Three independent biological replicates, run on separate days from separate passages. Triplicate wells within one plate are technical replicates and cannot substitute.
The four-parameter logistic:
Fit on log-transformed concentration. Constrain Bottom to 0 and Top to 100 only when the data show genuine plateaus; forcing constraints onto data that does not plateau produces a confident-looking IC50 that means nothing.
Report the IC50 with its 95% confidence interval, the R2, the Hill slope and n. A Hill slope far from 1, particularly above 2 or below 0.5, is worth a sentence in the discussion. It usually indicates cooperative binding, a mixed mechanism, or a solubility artefact.
Assay quality
Two numbers tell you whether a plate is usable before you look at the biology.
Coefficient of variation across your control wells should sit below 10 to 15%. Higher than that and pipetting, seeding or edge effects are contributing more variance than your compound.
The Z′-factor compares signal separation against combined variability:
Above 0.5 the assay is reliable. Between 0 and 0.5 is workable for a small focused study but marginal for screening. Below 0 means the positive and negative populations overlap and the plate should be repeated.
Comparing IC50 values
Two IC50 values are rarely compared correctly. Authors report 12.4 µM against 18.7 µM, call the first more potent, and move on.
An IC50 is a fitted parameter with an uncertainty attached, and the comparison has to account for it. Work on the log scale: logIC50 is approximately normally distributed, while IC50 itself is not, so a t-test on raw IC50 values from three replicates is the wrong test. Compare logIC50 values with their standard errors instead.
For two curves from the same experiment, the extra sum-of-squares F test asks directly whether one shared IC50 fits the combined data as well as two separate ones. That is the question you actually want answered, and it is more informative than checking whether confidence intervals overlap. Non-overlapping intervals do imply a difference, but overlapping intervals do not imply the absence of one.
Selectivity index has the same problem. An SI of 3.2 built from two IC50 values each carrying a wide confidence interval is not a meaningful number. Propagate the uncertainty, or report the two values and let the reader judge.
Send the plate readings. We will fit the 4PL, return IC50 with confidence intervals and Hill slope, and tell you whether your concentration range actually supports the number.
Send your dataISO 10993-5 in vitro cytotoxicity testing for devices and cosmetics
Academic and regulatory in vitro cytotoxicity testing are different exercises with different acceptance criteria.
ISO 10993-5:2009 is the third edition of the standard, and it remains current: ISO reviewed and confirmed it in 2022. The European version picked up amendment A11 in 2025, which revised the European foreword and Annex ZA to map the standard against EU MDR 2017/745. That amendment changed the regulatory framing, not the test methods.
The standard defines three categories of test:
Extract test. The device or material is extracted into culture medium, and cells are exposed to the extract. Extraction conditions come from ISO 10993-12: 37 °C for 24 h is the common condition, with 37 °C/72 h, 50 °C/72 h, 70 °C/24 h and 121 °C/1 h available depending on the material and the exaggeration required. Surface-area-to-volume ratios are specified: 6 cm²/mL for material up to 0.5 mm thick, 3 cm²/mL above 0.5 mm, and 0.2 g/mL for irregular shapes.
Direct contact test. The material is placed on a subconfluent monolayer and the zone beneath and around it is assessed.
Indirect contact test. Agar diffusion or filter diffusion, where the material is separated from the cells by a permeable layer.
Reference materials. ISO 10993-12 defines RM-A (ZDEC polyurethane film) as a positive control and RM-C (high-density polyethylene) as a negative control. Running both is what makes the result defensible.
For qualitative grading, the standard uses a 0 to 4 scale, from no reactivity to severe reactivity. Grades above 2 are not acceptable.
How much latitude the standard leaves
This is not a theoretical concern. In an interlaboratory comparison, 52 laboratories ran the elution test on the same two materials under the same extraction specifications, choosing the remaining parameters themselves within what the standard permits. Only 58% correctly identified the cytotoxic potential of both samples. Two variables drove most of the difference: adding 10% serum to the extraction medium, and extending the exposure period, both raised sensitivity considerably.
If an extract passes marginally in your hands, those are the first two parameters to interrogate. It is also why we hold both fixed and documented across every regulatory study we run, rather than treating them as lab preference.
Cosmetics. EU Regulation 1223/2009 prohibits animal testing for cosmetic ingredients and finished products, which makes in vitro cytotoxicity testing the base layer of the entire safety file. The validated OECD methods that sit above it — TG 439 for reconstructed human epidermis skin irritation, TG 431 for corrosion, TG 492 for eye irritation and TG 432 for 3T3 NRU phototoxicity — all use a cytotoxicity endpoint as the readout. TG 439 uses MTT directly, with viability below 50% defining an irritant. So an error in seeding density or interference control propagates upward into a regulatory classification.
L929, RM-A and RM-C reference controls, extraction per ISO 10993-12, and a documented report against the 70% viability criterion.
Request a study design and quoteWhat to report from in vitro cytotoxicity testing
Journals in pharmacology and toxicology are rejecting cytotoxicity data for incomplete methods more often than they were five years ago. Everything below belongs in the methods section of any in vitro cytotoxicity testing you publish:
- Cell line, source, and passage range used
- Seeding density per well and plate format
- Attachment period before dosing
- Exposure duration, and the reason for it
- Complete medium composition including serum percentage and phenol red status
- Vehicle, final solvent concentration, and how the vehicle control was matched
- Assay reagent, working concentration, incubation time, solubilisation method
- Read wavelength and reference wavelength
- Full concentration series with spacing
- Technical replicates per plate and independent biological replicates
- Positive control and its observed IC50 in your hands
- Cell-free interference control and its result
- Curve-fitting model, constraints applied, software
- IC50 with 95% confidence interval, R2, Hill slope
The positive control value is the item most often omitted and the one that does the most work. A reviewer who sees doxorubicin at an IC50 consistent with published values in your line will accept the rest of your dataset with far less friction.
In vitro cytotoxicity testing: questions researchers ask us
What is in vitro cytotoxicity testing?
In vitro cytotoxicity testing measures how a compound, extract or material affects cultured cells, using metabolic activity, membrane integrity, total biomass or growth rate as the readout. It is the entry point for anticancer screening, cosmetic and topical safety assessment, and medical device biocompatibility under ISO 10993-5.
Why is my viability above 100% at low concentrations?
Three usual causes. The compound is reducing the tetrazolium reagent directly, which your cell-free control will confirm. The compound is causing a hormetic metabolic increase without increasing cell number, which SRB or crystal violet will distinguish. Or the control wells are overconfluent and metabolically suppressed, which means the seeding density needs work.
How many concentrations do I need for a reliable IC50?
Eight, logarithmically spaced, with at least two sitting on each plateau. Six can work if the transition is steep and well sampled, but the confidence interval widens noticeably.
Can I use MTT for suspension cells?
It is possible with a centrifugation step, but cell loss during aspiration is uneven across the plate. WST-8 or resazurin avoid the problem because the formazan product stays in solution.
What is the difference between IC50 and GI50?
GI50 is measured against a time-zero plate and reflects net growth inhibition. IC50 is measured against a concurrent control and reflects reduction in the assay signal. If you did not run a T0 plate, you have an IC50.
Does ISO 10993-5 require L929?
L929 is recommended, not mandated. Other established cell lines are permitted with justification, though most notified bodies expect L929 unless there is a documented reason.
How long can I keep MTT stock?
Prepare at 5 mg/mL in PBS, filter through 0.22 µm, and store in aliquots at −20 °C protected from light. Discard any stock that has gone from yellow toward green or purple; it has already begun reducing.
More on in vitro cytotoxicity testing
Start a project
Tell us the problem. We will send back a study design.
Not a sales call. Describe the compound, the sample type and what your in vitro cytotoxicity testing has to prove, and you get a written design with the assay, controls, replicates and timeline, plus a quote.
- 300+ in vitro cytotoxicity testing and related assays completed
- 45+ partner institutions across 6 countries
- ISO GMP and ISO 9001:2015 certified
- ISEF Qualified Scientist at Regeneron ISEF 2026, the world’s largest pre-college science fair
Put this in your message
- Sample typePure compound, plant extract, nanoformulation, cosmetic, device material, or raw data you already have.
- Cell line, if you have one in mindIf you do not, say what tissue or indication you are working on and we will propose one.
- What you need to proveAn IC50 for a manuscript, a selectivity index, or a pass against the ISO 10993-5 70% criterion are different studies.
- Deadline and destinationThesis submission, journal revision, or a regulatory file. This changes the controls we build in.
Typical reply within one working day. Study designs are quoted before any work begins.
