Bootstrap f2 software built for ICH M13B
StatMod computes f2 and bootstrap f2 for dissolution profile comparison and writes a report an assessor can repeat. The ICH M13B release, scheduled before the rules apply in the EU, adds the new time point and acceptance rules. Audit-trailed, and in production with a leading global generics manufacturer since 2023.
A product of 
The f2 formula is one line. Assessors also check everything around it.
The f2 similarity factor fits on a sticky note. A biowaiver dossier also needs everything around that number: which time points were used and why, how many bootstrap sets were drawn, the seed, how the confidence interval was built, and proof that the software computing it was validated.
Most teams still run bootstrap f2 in an R script or a spreadsheet add-in kept by one colleague. That works until a deficiency letter asks for the algorithm, two tools disagree on the same data set, or the script has to change for a new guideline and be validated again.
“Evidence that the statistical software has been validated should be provided as well.”EMA assessor request, quoted at BioBridges 2019
Reproducibility
Runs without a fixed, reported seed cannot be repeated. The ICH M13B Q&A asks for a report detailed enough to repeat the analysis.
Tool disagreement
Published comparisons show free bootstrap tools reaching different similarity conclusions on the same profiles.
Validation cost
Under EU GMP Annex 11 and 21 CFR Part 11 an in-house script is a computerised system you validate, version and defend.
Key-person risk
When the one person who understood the script leaves, the method leaves with them.
What ICH M13B changes in dissolution similarity
ICH M13B harmonises the additional-strength biowaiver across ICH regions. It reached Step 4 on 4 August 2026. The EMA adopted it on 17 September 2026, effective 17 March 2027. For anyone who compares dissolution profiles, these are the rules that move.
| Topic | EMA guideline (2010) and Q&A practice | ICH M13B |
|---|---|---|
| Variability trigger | Coefficient of variation: up to 20% at early points, up to 10% at the others | Standard deviation: SD ≤ 8% at all time points for both strengths allows plain f2. SD > 8% at any point is high variability. |
| Last time point | Not more than one mean value above 85% for either product | Final point when either strength reaches 85%, or just after both reach a plateau below 85% |
| Plateau | Not defined | Three successive points within 5% across the three. Normally only the first plateau point is included. Slow dissolvers: check for artefact plateaus up to two hours. |
| Time points | At least three, zero excluded | At least three, zero excluded, and in principle not more than six, identical for both strengths |
| High-variability method | Bootstrap of the expected f2, lower bound of the 90% CI ≥ 50 | Bootstrap f2: median f2 ≥ 50 and lower bound of the two-sided 90% CI ≥ 46 |
| Bootstrap details | At least 5,000 samples, software validation evidence on request | At least 5,000 data sets, percentile interval, number of sets and seed reported, truncation algorithm described for each bootstrap data set (M13B Q&A) |
| No test needed | Both ≥ 85% within 15 minutes | Both ≥ 85% within 15 minutes, or both plateau below 10% |
Sources: ICH M13B guideline, EMA/CHMP/ICH/85092/2025, Step 5, sections 2.3.1 and 2.3.2; ICH M13B Q&A on bootstrap methodology, on the ICH multidisciplinary guidelines page; EMA Guideline on the investigation of bioequivalence (CPMP/EWP/QWP/1401/98 Rev. 1). Read the guideline itself before relying on any summary, this one included.
From the dissolution sheet to a documented result
Upload the Excel template
Unit-level dissolution data for the reference and the test profile, 6 to 24 units and up to 20 time points. StatMod checks that the template is complete and numeric and stores the data with the run.
Run the bootstrap
The bootf2BCA engine draws the bootstrap data sets with a fixed seed and applies the 85% truncation. The calculation runs in the background, and the result stays in the history with its inputs and parameters. The ICH M13B release will add the plateau rule, the 5,000-set minimum and the new acceptance gate.
Download the report
A result page, a PDF report and an Excel export with every bootstrapped f2 value. Observed f2, bootstrap mean and median, 5th and 95th percentiles, percentile, basic, normal and BCa intervals, the seed and the verdict. The ICH M13B release will add a methods section with the settings the Q&A asks for.
Coming in the ICH M13B release
- Acceptance: median bootstrapped f2 ≥ 50 and lower bound of the 90% percentile CI ≥ 46, shown on the result page, PDF and Excel export
- Truncation always applied: cut at ≥ 85% instead of > 85%, one rule instead of selectable options
- Plateau rule: three successive points spanning 5% or less, applied automatically, also inside each bootstrap replicate
- Check for false plateaus in slow-dissolving products, flagged in the report
- Plain f2 on each bootstrap data set instead of the expected f2
- Whole tablet profiles as the bootstrap unit, zero time point excluded
- At least 5,000 bootstrap data sets; replicates left with fewer than three time points are dropped and counted
- The Figure 1 cases: no calculation needed for very rapid dissolution or plateaus below 10%, and a note when SD ≤ 8% at every time point
- A report section listing the settings the Q&A asks for: bootstrap method, unit, zero point, truncation algorithm, count, seed
The ICH M13B release is in development, scheduled before 17 March 2027, the date the guideline applies in the EU. Its test set covers every path of the guideline's Figure 1 decision tree with reference data sets, each checked against an independent calculation. On request, the pre-M13B method stays available on a separate instance for filings to agencies that have not yet implemented the guideline.
Audit-trailed, documented for validation, hosted on Azure
Validation documentation
Functional requirement specification, design specification and test plan, with requirement IDs traced through design and tests, written to support your computerised system validation. The ICH M13B release will come with reference data sets for every path of the Figure 1 decision tree.
Audit trail and roles
Calculations, data entry, uploads, result views and every PDF and Excel download are written to the audit trail, which the application offers no way to edit or delete. Sign-ins come from Microsoft Entra ID. A separate Auditor role has read access to the audit screens. Each calculation keeps its inputs, parameters and outputs together.
Hosting and access
Managed service on Microsoft Azure in the West Europe region, single sign-on through Microsoft Entra ID. A dedicated instance in your own Azure tenant is possible as a project. Annual licence independent of user count, with L2 and L3 support under an SLA.
A method you can cite
The bootstrap engine is bootf2BCA, the open method from Aleksander Mendyk's group at Jagiellonian University and one of the platforms compared in the peer-reviewed literature on bootstrap f2 confidence intervals. StatMod adds data checks, reporting, access control and the audit trail around it; the ICH M13B release adds the M13B time point and decision logic. Two years of regulatory scientists' feedback shaped the labels, the report and the verdict wording.
StatMod compared with scripts, spreadsheets and general statistics suites
| Capability | R script or open-source package | Spreadsheet add-in | General statistics suite | StatMod |
|---|---|---|---|---|
| Bootstrap f2 with percentile 90% CI | Yes | Some | Yes | Yes |
| ICH M13B plateau rule and median gate | Not found as of Sept 2026 | Not found as of Sept 2026 | No guideline verdict | In the M13B release |
| Seed, sets retained and methods in the report | If you write it | Varies | Partly | Seed and count today, methods section in the M13B release |
| Validation documents | Yours to write | Varies by vendor | Platform level | Requirements, design spec, test plan supplied |
| Audit trail, roles, history | If you build it | Rarely | Varies | Yes |
| Support under SLA | Community | Varies by vendor | Vendor | L2 and L3 |
Typical for each category, not a verdict on any single product: individual tools vary. Based on public documentation of the free and commercial tools we reviewed in September 2026. Tell us if something has changed and we will correct the table.
For the people who sign off dissolution similarity
Generic manufacturers
Biopharmaceutics, clinical pharmacology and regulatory CMC teams filing additional-strength biowaivers in the EU, US, Canada and Japan.
BA/BE CROs
Biostatistics teams who run similarity evaluations for many sponsors and need one documented method, one report format and a clean audit trail per client.
Regulatory consultancies
Advisers who prepare Module 3 and Module 5 dossiers and answer deficiency letters on dissolution comparisons.
ICH M13B and bootstrap f2: common questions
What is ICH M13B?
ICH M13B is the ICH guideline on bioequivalence for immediate-release solid oral dosage forms that covers the additional-strength biowaiver: when a strength can be approved without its own bioequivalence study. Comparative dissolution and the f2 similarity factor are central to it. The ICH Assembly endorsed it at Step 4 on 4 August 2026.
When does ICH M13B apply in the EU?
The EMA adopted ICH M13B on 17 September 2026 with a date for coming into effect of 17 March 2027. From then it supersedes the additional-strength biowaiver provisions of the EMA bioequivalence guideline. Other ICH regulators set their own implementation dates.
What is the bootstrap f2 acceptance criterion in ICH M13B?
When either strength shows high variability (an SD above 8% at any time point), the guideline recommends a bootstrapped 90% confidence interval for f2. Similarity is shown when the lower bound of the two-sided 90% bootstrapped CI is at least 46 and the median f2 of the bootstrapped samples is at least 50.
How does ICH M13B define a dissolution plateau?
A plateau is three successive time points whose mean dissolution values differ by 5% or less across the three points. Normally only the first time point of a plateau is included in the similarity calculation. For slow dissolution the guideline asks you to check whether an apparent plateau is an artefact, with profiles collected up to two hours.
Which time points go into the f2 calculation under ICH M13B?
Identical time points for both strengths, zero excluded, at least three and in principle not more than six. The final time point is where either strength reaches 85% dissolved, or just after both have reached a plateau below 85%.
When is no f2 calculation needed?
When both strengths dissolve 85% or more within 15 minutes (very rapid dissolution), similarity can be concluded without mathematical evaluation. When both plateau below 10% mean dissolution, similarity can be assumed.
Is StatMod validated?
StatMod is in production for regulatory work with a leading global generics manufacturer since 2023. It ships with the documentation your computerised system validation needs: functional requirement specification, design specification and test plan, with requirement IDs traced through design and tests. Calculations, data entry, uploads and downloads are written to an audit trail with no edit or delete function in the application, and a separate Auditor role has read access to it. Validation of the system for your use stays with your quality unit; we support it.
Can StatMod run in our own environment?
StatMod runs as a managed service on Microsoft Azure in the West Europe region, with single sign-on through Microsoft Entra ID. A dedicated instance in your own Azure tenant is possible as a project. Ask us about your requirements.
See StatMod and what ICH M13B changes, in 45 minutes.
We run StatMod on example data, show the report and the audit trail, and walk through which time points and acceptance rules change under ICH M13B. No data needed. If you like, send two anonymised data sets you have already evaluated and we run them in StatMod as well. Bring your questions; we skip the slides.
- A full run: upload, result page, PDF report, audit trail
- Which time points and acceptance rules change under ICH M13B, and why
- What validation and onboarding look like for your site
Or write to sales@wecan.technology.
Built and supported by