Compare two dissolution profiles the way a regulator does. The engine computes the f2 similarity and f1 difference factors, enforces the 85% rule and RSD eligibility checks, and — when variability is too high for a direct f2 — runs a bootstrap 90% confidence interval of f2 (5,000 resamples). Supplementary Weibull / first-order / Higuchi / Korsmeyer-Peppas fits characterise the release kinetics.
Two dissolution profiles are compared with the model-independent similarity factor f2 and difference factor f1 on the mean profiles. An f2 of 50 corresponds to an average difference of about 10% across the timepoints, so f2 ≥ 50 is the accepted threshold for "similar" under FDA SUPAC and EMA guidance.
f2 = 50 · log₁₀{ [1 + (1/n)·Σ(Rₜ − Tₜ)²]−0.5 · 100 } | f1 = 100 · Σ|Rₜ − Tₜ| / ΣRₜRₜ and Tₜ are the mean percent dissolved of the reference and test at each timepoint. Similar profiles give f2 ≥ 50 and f1 ≤ 15.
Before f2 is trusted, the data must pass strict rules:
| Rule | Requirement |
|---|---|
| 85% rule | Only one timepoint after either profile reaches 85% may enter the calculation. |
| RSD at first point | ≤ 20% between vessels. |
| RSD at later points | ≤ 10% between vessels. |
| Timepoints | ≥ 3 comparison points (excluding time zero); n = 12 vessels per product. |
Complex generics and modified-release products often fail the RSD limits. In that case a single f2 value is unreliable, and FDA/EMA expect a multivariate confidence interval. The engine resamples the 12 vessels 5,000 times and reports the lower bound of the 90% confidence interval of f2 — profiles are declared similar only if that lower bound is ≥ 50.
Decision (high variability): lower 90% CI of bootstrap f2 ≥ 50 → equivalentTwelve reference and twelve test vessels are measured at 10, 20, 30, 45 and 60 min. Both reach 85% by 45 min, so 60 min is dropped by the 85% rule:
Reference: FDA SUPAC-IR / Guidance on dissolution testing; EMA guideline on the investigation of bioequivalence (f2). Supportive of, not a replacement for, a full biowaiver assessment.
Alongside f2, the mean profiles are fitted to standard release models — first-order (immediate release), Weibull (the industry standard for complex curves), Higuchi (matrix systems) and Korsmeyer-Peppas (polymeric systems) — each reported with its R² and mean dissolution time, to characterise the mechanism.
An f2 of at least 50 (with f1 no greater than 15) indicates the two profiles differ by about 10% or less on average, which FDA and EMA accept as similar. Below 50 the profiles are considered different.
Once a product is nearly fully dissolved, extra timepoints add little information but heavily inflate f2. The 85% rule keeps at most one timepoint after either profile passes 85%, so the comparison reflects the informative early-to-mid region.
When between-vessel variability exceeds the RSD limits (over 20% at the first point or 10% later). A single f2 is then unreliable, so the FDA/EMA-recommended approach is the lower bound of the 90% confidence interval of f2 from 5,000 vessel resamples.
FDA and EMA expect 12 units per product for a compliant f2 comparison. The tool will still compute f2 with fewer, but it flags a regulatory warning and cannot run a reliable bootstrap below two vessels per profile.