Estimating LD50 and LC50 is a common requirement in toxicology, environmental science, pharmacology, pesticide research, aquatic toxicity studies, entomology, and other experimental research involving quantal mortality responses.

When mortality changes across increasing doses or concentrations, researchers need a statistical method that can convert those grouped mortality observations into an estimated dose or concentration associated with 50% mortality.

That is where probit analysis is commonly used.

The Research Utility LD50 / LC50 Probit Analyzer provides a browser-based way to analyze grouped mortality data using binomial maximum-likelihood probit analysis, estimate the LD50/LC50, calculate a 95% confidence interval, display the fitted response curve, and download the results for research documentation.

What Is LD50?

LD50 means median lethal dose.

It is the dose of a substance expected to produce death in approximately 50% of a defined test population under specified experimental conditions.

The dose is normally expressed in units appropriate to the experiment, such as:

• mg/kg body weight
• µg/animal
• mg/organism
• another specified dose unit

LD50 is therefore not simply the dose at which one particular subject dies. It is a statistically estimated population-level endpoint.

The U.S. EPA describes LD50/LC50 as statistically or graphically estimated dose or concentration expected to be lethal to 50% of a group under specified conditions.

What Is LC50?

LC50 means median lethal concentration.

It is conceptually similar to LD50, except that the exposure is expressed as a concentration rather than an administered dose.

LC50 is frequently encountered in:

• aquatic toxicology
• environmental toxicology
• chemical toxicity studies
• pesticide research
• wastewater/effluent toxicity testing
• ecotoxicology

For example, an aquatic toxicity experiment might expose organisms to several concentrations of a chemical and record mortality at each concentration.

The resulting analysis may report:

LC50 = 8.4 mg/L

meaning that the estimated concentration associated with 50% mortality under the specified experimental conditions is 8.4 mg/L.

OECD guidance for acute fish toxicity describes estimation of LC50 and notes that probit or logit maximum-likelihood models can be used when the data contain sufficient partial mortalities.

LD50 vs LC50

The underlying statistical approach can be similar; the major distinction is how exposure is quantified.

What Is Probit Analysis?

Probit analysis is a statistical method used for analyzing quantal responses, such as:

• dead/alive
• affected/not affected
• responder/non-responder
• mortality/survival

Instead of treating mortality percentage as an ordinary continuous measurement, probit analysis models the probability of response as a function of exposure.

A typical mortality experiment might look like this:

The objective is not merely to identify the tested dose having exactly 50% mortality.

Instead, the statistical model uses all the available dose groups to estimate the underlying dose–mortality relationship and determine where the fitted response reaches 50%.

EPA documentation describes the classical probit approach as transforming observed mortality proportions using a probit transformation and transforming concentrations to log10 scale, with the resulting relationship estimated iteratively.

Why Not Simply Pick the Dose With 50% Mortality?

This is an important distinction.

Suppose your experiment produces:

• 4 mg → 20% mortality
• 8 mg → 50% mortality
• 16 mg → 85% mortality

You could report the observed 50% mortality dose as 8 mg.

But that ignores the information contained in the other dose groups.

Probit analysis instead estimates a continuous dose–response curve.

The resulting model might estimate:

LD50 = 7.39 mg

rather than simply reporting 8 mg.

The fitted estimate can therefore fall between tested dose levels.

That is one of the major advantages of model-based dose–response analysis.

What Does the Research Utility LD50 / LC50 Probit Analyzer Do?

The Research Utility tool is designed for grouped mortality data.

It allows researchers to:

1. Upload CSV data

You can upload your mortality dataset as a CSV file.

2. Select the analysis variables

The tool uses the relevant exposure and mortality information from your dataset.

3. Calculate mortality proportions

The analyzer works with the number exposed and number dead at each dose/concentration.

4. Fit a probit model

The tool uses a binomial maximum-likelihood probit model.

5. Estimate LD50 or LC50

The model produces the estimated 50% lethal endpoint.

6. Calculate a 95% confidence interval

The tool provides an uncertainty interval around the estimated endpoint when the model permits estimation.

7. Display the fitted curve

The graphical output allows you to visually inspect the observed mortality pattern and fitted dose–response relationship.

8. Report model parameters

The output includes relevant model statistics such as the slope and intercept.

9. Assess goodness of fit

The analyzer reports a Pearson chi-square goodness-of-fit statistic, degrees of freedom, and p-value.

10. Download results

The results and graph can be saved for further analysis, reporting, or documentation.

How to Prepare Your CSV Dataset

The quality of the analysis depends heavily on the quality of the experimental data.

A typical grouped mortality dataset should contain information such as:

The exact column names depend on how your Research Utility tool is configured, but conceptually you need:

Exposure level + number exposed + number responding/dead

What Does “Exposed” Mean?

The Exposed column represents the number of experimental subjects exposed to a particular dose or concentration.

For example:

Exposed = 20

means that 20 organisms/subjects were included in that treatment group.

What Does “Dead” Mean?

The Dead column represents the number of subjects that died under the specified experimental conditions.

For example:

Exposed = 20
Dead = 10

corresponds to:

10 / 20 = 50% mortality

Why Replication Matters

Suppose two dose groups both show 50% mortality.

Group A

10 exposed, 5 dead

Group B

100 exposed, 50 dead

Both have 50% observed mortality, but they do not contain the same amount of information.

The second group has considerably more observations.

This is one reason why a binomial model is appropriate for grouped mortality data: it accounts for the number of subjects exposed at each dose rather than treating every percentage as equally informative.

How to Use the LD50 / LC50 Probit Analyzer

Step 1 — Prepare your data

Create a CSV file containing your dose/concentration and mortality information.

For example:

Dose,Exposed,Dead
1,20,0
2,20,1
4,20,4
8,20,10
16,20,17
32,20,20

Step 2 — Open the tool

Go to the Research Utility Data Analysis Tools page and open:

LD50 / LC50 Probit Analyzer

Step 3 — Upload your CSV

Select your CSV file.

The tool will process the dataset in the browser.

Step 4 — Select the appropriate variables

Make sure the correct variables correspond to:

• dose/concentration
• number exposed
• number dead/mortality

Always verify that the selected columns represent the intended experimental variables.

Step 5 — Run the analysis

Click:

Run Probit Analysis

The analyzer will calculate the fitted model.

Understanding the Results

The tool provides several important outputs.

1. LD50 / LC50

This is the primary result.

For example:

LD50 = 7.3925

This means the fitted model estimates that the dose corresponding to a 50% lethal response is approximately 7.39 dose units, under the conditions represented by the data.

The number must always be reported together with its units and experimental context.

2. 95% Confidence Interval

For example:

95% CI: 5.8777–9.2978

The confidence interval communicates uncertainty surrounding the estimated LD50/LC50.

It is much more informative to report:

LD50 = 7.39 mg/kg (95% CI: 5.88–9.30 mg/kg)

than simply:

LD50 = 7.39 mg/kg

OECD guidance specifically discusses confidence limits around LD50 estimates and notes maximum-likelihood and profile-likelihood approaches for obtaining confidence limits under appropriate conditions.

3. Probit Slope

The probit slope describes how rapidly the probability of mortality changes with increasing exposure on the model scale.

A steeper slope indicates a more rapid transition from low to high predicted mortality over the exposure range.

For example:

Probit slope = 3.2538

The slope should be interpreted alongside the fitted curve and experimental design rather than in isolation.

4. Intercept

The intercept is another parameter of the fitted probit relationship.

For example:

Intercept = −2.8269

It is a model parameter used in determining the fitted dose–response relationship.

Researchers generally should not interpret the intercept as a standalone biological endpoint.

5. Pearson χ² Goodness-of-Fit

The tool also provides a Pearson goodness-of-fit statistic.

For example:

χ² = 0.8296
df = 4
p = 0.9344

The p-value is used to assess whether there is statistically detectable evidence of lack of fit under the specified goodness-of-fit procedure.

A large p-value does not prove that the model is true or that the model is perfect.

A better interpretation is:

There is no statistically significant evidence of lack of fit at the 0.05 significance level.

This distinction is important when reporting results in a thesis, dissertation, manuscript, or scientific report.

Understanding the Probit Dose–Response Curve

The graphical output is one of the most useful parts of the analyzer.

The graph typically contains:

• observed mortality points
• fitted probit curve
• LD50/LC50 marker
• 50% mortality reference
• exposure axis
• mortality/response axis

The graph allows you to determine whether the fitted curve is broadly consistent with the observed data.

Why the Exposure Axis Is Logarithmic

Dose–response experiments frequently span several orders of magnitude.

For example:

1 → 2 → 4 → 8 → 16 → 32

A logarithmic exposure axis makes such data easier to visualize and is consistent with the conventional representation used in probit dose–response analysis.

EPA documentation describes the classical probit procedure using log10-transformed concentration together with a probit transformation of observed mortality.

What Does an LD50 Inside the Tested Range Mean?

Suppose your tested doses are:

1, 2, 4, 8, 16, 32

and the analyzer estimates:

LD50 = 7.39

The estimate lies within the tested range.

This is generally preferable to a result that requires substantial extrapolation.

What If the LD50 Is Outside the Tested Range?

Suppose the highest tested dose is 10 mg/kg but the model estimates:

LD50 = 45 mg/kg

That estimate is outside the observed range.

This means the model is extrapolating beyond the experimental data.

It should therefore be interpreted cautiously.

Similarly, if mortality never reaches approximately 50% across the tested range, a definitive LD50/LC50 estimate may not be supported by the experiment.

EPA guidance notes that when 50% mortality is not reached at the highest concentration, the endpoint may instead be reported as a non-definitive value such as LC50 > highest concentration tested, depending on the study and applicable method.

Why Intermediate Mortality Levels Are Important

This is one of the most important considerations in probit analysis.

If every dose produces:

0% mortality

or:

100% mortality

there is very little information about where the 50% response occurs.

You ideally want exposure levels producing a range of mortality responses.

For example:

0%, 5%, 20%, 50%, 85%, 100%

contains substantially more information about the transition than:

0%, 0%, 0%, 100%, 100%, 100%.

OECD guidance notes that classical maximum-likelihood approaches require sufficient partial mortalities; its fish toxicity guidance specifically notes the importance of having at least two concentrations with mortality between 0% and 100% for classical probit/logit estimation of LC50 and its confidence limits.

What Are Partial Mortalities?

A partial mortality is a dose group where:

0% < mortality < 100%

Examples:

• 5%
• 20%
• 50%
• 75%
• 90%

These groups are particularly informative for estimating the shape and midpoint of a dose–response curve.

What If I Have 0% and 100% Mortality?

0% and 100% mortality groups are not automatically useless.

They provide information about the lower and upper portions of the response relationship.

However, having intermediate mortality groups is important for estimating the transition around 50%.

A dataset consisting almost entirely of 0% and 100% mortality can make the fitted endpoint and its uncertainty less stable.

What If There Is Control Mortality?

Control mortality requires careful consideration.

For many toxicity experiments, a control group is included to establish background mortality.

The appropriate treatment of control mortality depends on the experimental protocol and analytical method.

Do not automatically enter control mortality as an ordinary treated dose group.

Depending on the study design, control mortality may need to be adjusted or handled according to the relevant guideline or statistical procedure.

For environmental toxicity testing, EPA methods specifically discuss adjustment of observed mortality for control mortality in relevant procedures.

When Should You Use LD50?

LD50 analysis may be relevant when your experimental endpoint is mortality following an administered dose.

Examples can include research involving:

• toxicology
• pesticide toxicity
• pharmacological toxicology
• chemical safety studies
• entomological bioassays
• experimental toxicology

However, the appropriate endpoint and experimental design depend on the research question and applicable regulatory or institutional guidelines.

When Should You Use LC50?

LC50 is more appropriate when the exposure is represented as a concentration.

Common applications include:

• aquatic toxicity
• environmental toxicology
• wastewater toxicity
• chemical ecotoxicology
• fish toxicity
• invertebrate toxicity
• algal or other organism exposure studies where concentration is the exposure metric

LD50 Is Not the Same as EC50

This is an important distinction.

LD50

Dose producing 50% lethality.

LC50

Concentration producing 50% lethality.

EC50

Concentration or dose producing 50% of a specified effect, which does not necessarily mean death.

For example, EC50 could refer to:

• 50% growth inhibition
• 50% reduction in activity
• 50% maximal pharmacological response

Therefore, you should never automatically label every 50% response as LD50.

LD50 Is Not the Same as NOAEL

These endpoints answer different questions.

LD50 concerns a lethal response in 50% of the modeled population.

NOAEL refers to the highest tested exposure level at which no statistically significant adverse effect is observed under the specified study conditions.

They should not be substituted for one another.

How to Report LD50 Results in a Research Paper

A concise reporting format could be:

Probit analysis using a binomial maximum-likelihood model estimated an LD50 of 7.39 mg/kg (95% CI: 5.88–9.30 mg/kg). The fitted model had a probit slope of 3.25. Pearson’s goodness-of-fit test indicated no statistically significant evidence of lack of fit (χ² = 0.83, df = 4, p = 0.934).

Notice that the units must be included in the final scientific report.

The software cannot determine the scientific unit for you unless it is explicitly supplied as part of your research context.

Example Interpretation

Suppose your analysis produces:

• LD50 = 7.3925
• 95% CI = 5.8777–9.2978
• slope = 3.2538
• χ² = 0.8296
• df = 4
• p = 0.9344

A suitable interpretation could be:

The binomial maximum-likelihood probit model estimated an LD50 of 7.39 dose units, with a 95% confidence interval of 5.88–9.30 dose units. The estimated probit slope was 3.25, indicating the steepness of the modeled mortality transition. Pearson’s goodness-of-fit test produced χ² = 0.83 with 4 degrees of freedom (p = 0.934), indicating no statistically significant evidence of lack of fit at the 0.05 level.

This should still be adapted to your actual species, substance, route of exposure, duration, units, and experimental design.

How to Use the Results in a Thesis

For a dissertation or PhD thesis, I recommend reporting:

Methods

Describe:

• experimental organism
• test substance
• dose/concentration units
• number of subjects per group
• exposure duration
• mortality endpoint
• statistical model
• software/tool used
• confidence level
• goodness-of-fit procedure

Results

Report:

• estimated LD50/LC50
• confidence interval
• slope
• goodness-of-fit statistics
• relevant mortality observations
• fitted curve

Discussion

Discuss:

• biological interpretation
• uncertainty
• dose spacing
• mortality pattern
• model adequacy
• comparison with previous studies
• experimental limitations

Common Mistakes to Avoid

Mistake 1: Treating LD50 as an observed dose

LD50 is generally a model-derived endpoint, not necessarily one of the tested doses.

Mistake 2: Ignoring confidence intervals

An LD50 without uncertainty provides incomplete information.

Always report the confidence interval when it is estimable and appropriate.

Mistake 3: Calling LD50 “50% lethal dose” without context

LD50 is conditional on the experimental system.

Species, strain, sex, exposure route, exposure duration, formulation, environmental conditions, and other factors can affect the result.

Mistake 4: Assuming p > 0.05 proves the model

A non-significant goodness-of-fit result does not prove that the model is correct.

It means the test did not detect statistically significant evidence of lack of fit under the specified test.

Mistake 5: Ignoring dose units

An LD50 of:

7.39

is incomplete.

You need something like:

7.39 mg/kg body weight

if that is the actual unit used in the experiment.

Mistake 6: Using LD50 when the endpoint is not mortality

If your response is inhibition, growth, activity, enzyme response, or another non-lethal endpoint, an EC50 or another appropriate endpoint may be more suitable.

Mistake 7: Ignoring experimental design

Statistical software cannot compensate for poorly designed experiments.

The quality of the estimated endpoint depends on the underlying experimental data.

Important Limitations of Probit Analysis

Probit analysis is powerful, but it is not universally appropriate.

You should carefully evaluate:

• the response definition
• dose/concentration measurement
• sample size
• number of exposure levels
• dose spacing
• partial mortality levels
• control mortality
• independence of observations
• biological plausibility
• model assumptions

Different dose–response models may be appropriate for different datasets.

For example, logit models can also be used for quantal dose–response data. OECD guidance explicitly recognizes both probit and logit approaches in relevant acute toxicity contexts.

Probit vs Logit

Both models can be used for binary/quantal response data.

The choice should be justified by the study methodology rather than made simply because one model produces a more favorable result.

Why Your Experimental Dose Range Matters

A good dose range should provide meaningful information about the transition from low mortality to high mortality.

For example:

Poorly informative design:

0%, 0%, 0%, 100%, 100%

More informative design:

0%, 5%, 20%, 50%, 85%, 100%

The second pattern provides much better information about the shape of the response around the 50% region.

EPA and OECD guidance both emphasize the importance of appropriate dose/concentration ranges and intermediate response information for estimating lethal endpoints.

Can This Tool Replace Statistical Software?

For many exploratory and routine research calculations, a browser-based analyzer can be extremely convenient.

However, for a publication, regulatory submission, or high-stakes toxicological assessment, you should independently verify the analysis using an appropriate statistical package or validated method when required by your study protocol or guideline.

The Research Utility tool should be viewed as a research analysis aid, not a substitute for methodological judgment.

Why Local Browser Processing Is Useful

The Research Utility analyzer is designed so that the uploaded CSV is processed locally in the browser.

That provides a useful workflow for researchers working with:

• experimental datasets
• laboratory measurements
• unpublished research
• thesis datasets
• preliminary analyses

You should nevertheless follow your institution’s policies for handling confidential or sensitive research data.

Frequently Asked Questions

What is LD50?

LD50 is the estimated dose associated with 50% lethality in a defined population under specified conditions.

What is LC50?

LC50 is the estimated concentration associated with 50% lethality under specified conditions.

What is probit analysis?

Probit analysis is a statistical method for modeling quantal responses such as mortality as a function of dose or concentration.

Can I calculate LD50 from a CSV file?

Yes, provided your CSV contains appropriately structured grouped mortality data supported by the analyzer.

Does LD50 have to equal one of my tested doses?

No. A model-based LD50 can fall between tested dose levels.

Why do I need several dose levels?

Multiple dose levels help characterize the dose–mortality relationship and improve estimation of the 50% endpoint.

Why are partial mortalities important?

Groups with mortality between 0% and 100% provide important information about the transition around the LD50/LC50.

What does the 95% CI mean?

It provides an interval describing uncertainty around the estimated lethal endpoint under the statistical method used.

What does a high goodness-of-fit p-value mean?

It means the goodness-of-fit test did not detect statistically significant evidence of lack of fit at the chosen significance level. It does not prove that the model is correct.

Can I use the analyzer for aquatic toxicity?

Yes, LC50 analysis is commonly used in aquatic toxicity research, provided the data and study design are appropriate.

Can I use it for pesticide research?

It can be useful for appropriate mortality bioassay data, but you should follow the applicable regulatory or methodological guidance for your specific study.

Can I use it for EC50?

Not automatically. EC50 concerns a 50% effect and is not synonymous with a lethal endpoint. Use an analysis specifically appropriate to the response being studied.

Best Practices for Researchers

Before accepting an LD50/LC50 result, ask:

Data quality

• Are the doses/concentrations correct?
• Are the mortality counts correct?
• Are exposed and dead counts internally consistent?
• Are there transcription errors?

Experimental design

• Is the dose range appropriate?
• Are there enough dose levels?
• Are there adequate partial mortality groups?
• Was the exposure duration consistent?

Statistical analysis

• Is probit analysis appropriate?
• Is the fitted curve biologically plausible?
• Is the confidence interval reasonably precise?
• Is there evidence of lack of fit?

Reporting

• Did you report units?
• Did you report the 95% CI?
• Did you describe the experimental population?
• Did you specify the model?
• Did you avoid interpreting association/model fit as causation?

A Practical Workflow for PhD Researchers

If you are using the analyzer for thesis or publication work, a good workflow is:

Raw experimental data

↓

Data cleaning

↓

Check dose/concentration values

↓

Check exposed/dead counts

↓

Calculate mortality proportions

↓

Inspect dose–mortality pattern

↓

Run probit analysis

↓

Review LD50/LC50

↓

Review 95% CI

↓

Review fitted curve

↓

Review goodness of fit

↓

Validate results independently when required

↓

Report statistical method and results

This is much safer than simply copying the LD50 number into a thesis table.

Research Utility LD50 / LC50 Probit Analyzer

The Research Utility LD50 / LC50 Probit Analyzer brings this workflow into a simple browser-based research environment.

It is particularly useful when you need to quickly:

• analyze grouped mortality data
• estimate LD50
• estimate LC50
• visualize mortality dose–response relationships
• obtain confidence intervals
• inspect model fit
• prepare results for research reporting
• download analysis outputs

Try the LD50 / LC50 Probit Analyzer

For researchers working with mortality bioassay data, the tool provides a practical starting point for probit-based LD50/LC50 estimation without requiring manual calculation of the underlying model.

Use your experimental data, verify the selected variables and units, examine the fitted curve and confidence interval, and interpret the result in the context of your experimental design.

Remember: the numerical result is only one part of a scientific analysis. The experimental design, biological context, model assumptions, uncertainty, and applicable research guidelines should all be considered before drawing conclusions.

References and Further Reading

• OECD Guidance Document on Acute Oral Toxicity Testing — discusses LD50 estimation, confidence limits, and dose-effect curves.
• OECD Test No. 203: Fish, Acute Toxicity Test — discusses LC50 estimation and probit/logit methods.
• U.S. EPA Methods for Measuring Acute Toxicity — provides detailed discussion of the probit method and its requirements.
• U.S. EPA Dose-Response Assessment — provides background on dose-response and concentration-response relationships.