When a formulation changes — a new excipient supplier, a different granulation route, a scale-up from pilot to production — you have to show the new product releases drug like the old one. Comparing the two curves by eye is not acceptable to a regulator. The standard tools are the difference factor f1 and the similarity factor f2, and they appear in FDA and EMA guidance on dissolution testing and biowaivers.
The formulas. With n time points, R for reference and T for test:
- f1 = [ Σ|Rt − Tt| ÷ ΣRt ] × 100 — a percentage difference, so smaller is better, and the limit is f1 ≤ 15
- f2 = 50 × log₁₀ { 100 ÷ √[ 1 + (1/n)Σ(Rt − Tt)² ] } — a similarity score out of 100, so bigger is better, and the limit is f2 ≥ 50
f2 rests on the mean squared difference, which is the design choice that matters: one badly deviating time point hurts f2 far more than several small ones, because squaring punishes outliers.
Choose the time points before you calculate. The reference profile is:
| Time (min) | 15 | 30 | 45 | 60 | 75 | 90 |
|---|
| Reference % dissolved | 35 | 58 | 74 | 85 | 92 | 96 |
Guidance allows only one measurement after 85% dissolution to be included. Here both 92% (75 min) and 96% (90 min) are above 85%, so the 90-minute point is dropped and five points remain. This is not a technicality to skip — beyond 85% both products are nearly fully dissolved, the differences shrink towards zero, and including those points inflates f2 artificially. A student who uses all six points gets a higher, friendlier number and loses the mark.
Test A — a minor process change.
| Time (min) | 15 | 30 | 45 | 60 | 75 |
|---|
| Reference | 35 | 58 | 74 | 85 | 92 |
| Test A | 30 | 52 | 70 | 83 | 91 |
| Difference | 5 | 6 | 4 | 2 | 1 |
| Difference² | 25 | 36 | 16 | 4 | 1 |
Σ(R−T)² = 82, so (1/5)(82) = 16.40, and √(1 + 16.40) = 4.1713.
f2 = 50 × log₁₀(100 ÷ 4.1713) = 50 × log₁₀(23.973) = 50 × 1.37979 = 69.0
f1 = (5 + 6 + 4 + 2 + 1) ÷ 344 × 100 = 18 ÷ 344 × 100 = 5.23%
Both pass comfortably — f2 = 69 against a floor of 50, f1 = 5.2% against a ceiling of 15. The profiles are similar and the change is supportable.
Test B — a different granulation route.
| Time (min) | 15 | 30 | 45 | 60 | 75 |
|---|
| Reference | 35 | 58 | 74 | 85 | 92 |
| Test B | 22 | 40 | 58 | 72 | 83 |
| Difference | 13 | 18 | 16 | 13 | 9 |
| Difference² | 169 | 324 | 256 | 169 | 81 |
Σ(R−T)² = 999, so (1/5)(999) = 199.80, and √200.80 = 14.1704.
f2 = 50 × log₁₀(100 ÷ 14.1704) = 50 × log₁₀(7.0569) = 50 × 0.84858 = 42.4
f1 = 69 ÷ 344 × 100 = 20.06%
Both fail — f2 is below 50 and f1 is above 15. The two tests agree, which they usually do; when they disagree, f2 is the one regulators quote.
Read the shape, not just the verdict. Test B is slower at every single point and the gap is widest in the middle of the curve, peaking at 18% at 30 minutes. That pattern — slow early and mid-release, converging late — points at the granules rather than the drug: harder granules, a higher compression force, or a more hydrophobic binder distribution slowing wetting and disintegration. A profile that instead matched early and diverged late would point somewhere else entirely, usually at a solubility or sink-condition problem in the vessel.
Three conditions that must hold before either number means anything.
- Same conditions for both profiles — same apparatus, speed, medium, volume and temperature. Comparing a test run in pH 6.8 phosphate buffer with a reference run in 0.1 M HCl produces an f2 value that is arithmetically correct and scientifically worthless.
- At least 12 units per profile, with a coefficient of variation no more than 20% at the early time points and 10% later. f2 is calculated on mean values, so it hides variability completely; a test batch where individual tablets range from 40% to 95% at 45 minutes can still average into a passing f2.
- At least three time points, excluding zero, and the same points for both profiles. Two points cannot describe a curve, and f2 on two points is not a defensible submission.
Finally, note what f2 does not do. It compares release in a beaker. It is evidence for a biowaiver only when the drug is highly soluble and the dosage form is immediate-release — a BCS Class I or III argument. For a poorly soluble Class II drug, similar dissolution in one medium does not establish bioequivalence, and the examiner's expected answer is that in vivo data may still be required.