Font Size




Menu Style

Analysis of IC50 data

Statistical analysis

The PowerPoint presentations provided below are annotated with notes.

The first provides guidance on interpretation of raw data for NA activity and IC50 assays. It covers the following topics: determining standard virus dose; validation and interpretation of NA activity and IC50 analysis; how to calculate IC50 values, and examples of fluorescence and chemiluminescence data

Interpretation of raw data powerpoint icon2

The second addresses a number of the issues encountered when undertaking statistical analysis of IC50 data. It covers the impact of using different methods for determining IC50 values; how to identify outlying results (above the seasonal norm), and how to monitor trends in IC50 data:

Statistical analysis of IC50s powerpoint icon2

Examples of both correct and incorrect curve fitting and point-to-point graphs for both sensitive and resistant isolates are provided below. The files contain graphical examples and accompanying explanatory notes to aid interpretation.

Troubleshooting & Analysis of Curve Fitting powerpoint icon2

Troubleshooting & Analysis of Point-to-Point IC50 determination powerpoint icon2


Frequently asked questions on IC50 data analysis

Q. What IC50 values can I expect to get?

A. Susceptibility to NI drugs is not absolute.  There are typical ranges of IC50 values that differ between influenza subtypes and between oseltamivir and zanamivir for each subtype. Therefore this subtype and drug specific data should not be compared.

Influenza B viruses tend to have IC50 values 10-100 fold higher than influenza A viruses. This is normal for influenza B and this lower susceptibility does not appear to have a significant clinical impact.  IC50 values for influenza B viruses tend to be higher for oseltamivir than zanamivir.  IC50 values for H1N1 viruses tend to be higher for zanamivir than oseltamivir and IC50 values for H3N2 viruses tend to be higher for oseltamivir than zanamivir.

Also, values generated by fluorescence and chemiluminescence methods should not be compared.  Typically, values generated by chemiluminescence are lower than those for the same virus and drug in the fluorescence test.  IC50 values generated from the different assay methods should not be directly compared.

Q. How do I determine IC50?

A. Raw data (relative fluorescence or luminescence units) are plotted against the drug concentration. IC50 values can be determined in two ways; using curve fitting software, or by a point to point analysis.

Curve Fitting: Several curve fitting software are commercially available:  Graphpad Prism, Grafit

CDC have developed a curve fitting software which is available from them directly

Point to Point: This method uses in house excel templates. Raw data (relative fluorescence or luminescence units) are plotted against the drug concentration.  Guidance on how to undertake point-to-point analysis is provided above in the IC50 analysis section. 

Q:  A 2-3-fold difference is seen when the same virus is re-tested.  Is this normal?

A.  A 2-3-fold difference in IC50 values is common, in all assay methodologies.
Q:  What IC50 value obtained in a neuraminidase inhibition assay can be used as a cut-off value for clinically relevant resistance?

A. There is no firm definition of a resistant IC50.  Commonly used criteria to identify isolates that outside the normal range are either:

1. A value greater than 3SD from the mean (or median) value for the given subtype and drug.

 2. A value 10-fold or greater than the mean (or median) value for the given subtype and drug.

True resistant isolates with one of the currently known and characterised mutations tend to have IC50 values 100-10,000 times higher than the normal range for that subtype and therefore are instantly recognisable.  Confirmation of a resistant phenotype should be carried out by sequencing of the NA gene where possible.

Q. What should I do with intermediate IC50 results?

A. Intermediate results are those which fall over the minor cut off (whether using box and whisker or SMAD to monitor IC50s) and under the major outlier cut off and therefore are not strictly termed as resistant.

These viruses could potentially be mixtures, containing quasi-species of sensitive and resistant virus or may have reduced susceptibility to drug.  The IC50 test for such isolates should be repeated to confirm the result, and further characterised where possible, for example, sequencing of the NA gene.

Q: Could drug susceptibility assessment be based solely on the result of the neuraminidase inhibition assay provided that the IC50 value is very high or rather low?
A. No, such a practice is not recommended.  Genotypic analysis such as Sanger sequencing, or pyrosequencing may be needed, including on matching clinical specimens.
Q: Some ‘drug-resistant’ reference influenza A viruses (e.g., E119V in H3N2) have IC50 values that are not very different from the IC50 values of some ‘drug-sensitive’ influenza B viruses. Why is this?
A. A comparison needs to be done for each specific subtype and type.  The chemiluminescent assay is less discriminative compared to the fluorescent assay.  In addition, influenza B viruses have been shown to be less susceptible to oseltamivir in a clinical trial.

Q:  Some influenza viruses show an elevated IC50 value but they do not have any known or unique changes in the NA.  What is the reason for the elevated IC50 value?
A. There could be several reasons for this:  1. Repeat the test.  2. Virus preparation may contain a mix of influenza A and B viruses and influenza B viruses typically have higher IC50 values.  It is prudent to perform real time RT-PCR testing to subtype the virus isolate.  3. There could be a mix of virus variants and cloning may help to separate a resistant variant from a sensitive one. 

Q. How do I analyse my IC50 values over time?

A. Once you have tested a number of isolates of a given subtype against a drug, you can determine the ‘normal range’.  This can be done in two ways: Box Plot Analysis and robust statistics using the standard deviation of the median absolute deviation of the median (SMAD). 

Further information on using robust statistics is available from AMC Technical Briefs.  An MS-Excel add-in for calculating the SMAD is available.  Usually, the minor outlier cut-off is chosen as the median + 1.65 SMAD, and the major outlier cut-off is chosen as the median +3 SMAD.    It is important to note that all calculations for box plot and robust statistics are performed on log data.  Only after all calculations have been completed are the data back-transformed by taking the antilog.

Q. How should I monitor trends?

A. Trends from one influenza season to the next can be monitored using the box and whisker or SMAD method to determine whether the mean/median values and the minor and major cut off criteria are changing or remaining approximately constant.

top of page