What is calculated?

Two dissolution profiles are compared point by point. With R the mean percentage dissolved of the reference and T that of the test product at a given time:

• f1 (difference factor) = 100 · Σ|R−T| / ΣR — the mean relative difference between the profiles. The closer to zero, the more the profiles overlap.
• f2 (similarity factor) = 50 · log₁₀(100 / √(1 + Σ(R−T)²/n)) — equals 100 for identical profiles.

The commonly accepted criteria are f2 ≥ 50 and f1 ≤ 15. f2 = 50 corresponds to an average difference of 10 % between the profiles.

Where is it used in pharmaceutical technology?

• Comparison of a generic product's dissolution profile with that of the reference product
• After a change in formulation or process (components, manufacturing site, equipment, scale-up), comparison with the pre-change profile
• Demonstration of profile similarity in different dissolution media (e.g. pH 1.2 / 4.5 / 6.8)
• Comparison of the end-of-shelf-life profile with the initial profile
• Profile similarity between different dosage strengths of the same formulation

Prerequisites

• At least 12 units per profile.
• The same time points and the same conditions (medium, apparatus, agitation speed, temperature) for both products.
• Only a single point should be taken after 85 % has been exceeded. The 85 % cut-off rule option on the page applies this: it stops at the first time point at which either the reference or the test product reaches 85 % and includes that point in the calculation.
• At least 3 time points should remain after the cut-off; an f2 calculated with fewer is not meaningful. The application warns when fewer points are used.
• Variability limits: RSD ≤ 20 % at the first time point, ≤ 10 % at subsequent points. The RSD column in the grid shows the value; if a limit is exceeded, the report produced after the calculation states this together with the time point.
• If both products exceed 85 % within 15 minutes, no profile comparison is needed; the profiles are considered similar.

Limitations

• f2 is a point estimate without a confidence interval. It is a regulatory acceptance criterion, not a test of statistical significance. When the variability limits cannot be met, the guidelines recommend bootstrap f2; this application does not provide it.
• It is sensitive to the choice of time points. f2 changes with the number and spacing of the points. The points should therefore be defined before the study, not chosen with the result in view.
• It measures the point-by-point difference, not the shape of the profile. Two curves of different shape may come out similar as long as their differences stay small.
• Only mean profiles are compared; between-unit variability does not enter f2. This is why the RSD prerequisite matters.
• With the "12 replicates required" option enabled, time points that do not have exactly 12 replicates are excluded from the analysis. The excluded points are listed in the report; it is still worth checking the reported n.
• Time points whose mean reaches 100 % are also excluded. In a profile that reaches complete dissolution the plateau points drop out for this reason; the report states this too.