Dissolution Profile Fitting
On this page
Direct answer
Dissolution profile fitting turns percent-dissolved-time curves into mechanism and comparability judgements. Mechanistic models include zero-order (amount linear in time), first-order (log of undissolved fraction linear in time), Higuchi's square-root law Q = kH·t^0.5 for diffusion from an inert matrix, Korsmeyer-Peppas Mt/M∞ = k·tⁿ whose exponent classifies release (cylinders: n up to 0.45 Fickian, 0.45-0.89 anomalous transport mixing diffusion with polymer relaxation, 0.89 and above near zero-order case-II), Hixson-Crowell's cube-root law for dissolving particles, and the empirical Weibull with shape parameter beta describing exponential, sigmoid or parabolic curves. Comparability is judged model-independently by the similarity factor f2 = 50·log{[1 + (1/n)Σ(R − T)²]^−0.5 × 100}, where 50-100 means similar (difference factor f1 within 0-15); by convention f2 is unnecessary when both products dissolve 85 per cent or more within 15 minutes. All rests on sink conditions — medium volume at least three times saturation — and at least 12 units per profile.
What you must remember
- Model equations: zero-order Qt = Q0 + k0t; first-order log(Q∞ − Qt) versus t linear; Higuchi Q = kH√t; Korsmeyer-Peppas Mt/M∞ = k·tⁿ; Hixson-Crowell ∛W0 − ∛Wt = kt; Weibull F = 1 − exp[−(t/α)^β].
- Korsmeyer thresholds (cylinders): n ≤ 0.45 Fickian; 0.45-0.89 anomalous (diffusion plus chain relaxation); n ≥ 0.89 case-II, near zero-order — thresholds shift slightly for slabs and spheres.
- Fitting window: the Korsmeyer log-log plot uses only the early portion, conventionally Mt/M∞ below 0.6.
- f2 arithmetic: similarity factor 50-100 = similar; difference factor f1 between 0 and 15; a drop of a few points in f2 signals meaningful formulation change since f2 is logarithmic in squared differences.
- The 85-per-cent rule: when both products reach 85 per cent within 15 minutes, profiles are similar without f2 computation.
- Sink condition: medium dissolves at least three times the drug amount present (C well below Cs), otherwise dissolution is solubility-limited and meaningless as a formulation test.
- Weibull beta reading: β = 1 exponential, above 1 sigmoid with lag, below 1 initial-fast parabolic — descriptive power without mechanism.
- Application map: BCS biowaivers (Class I needs 85 per cent in 30 minutes), SUPAC-type post-approval change testing, and IVIVC levels A (point-to-point), B (statistical moments), C (single-point parameter).
Choosing a model for a modified-release matrix
A hydroxypropyl methylcellulose matrix tablet gives 30, 55, 72 and 84 per cent at 1, 4, 8 and 12 hours. Plot against the square root of time and the points fall on a line — Higuchi behaviour, the signature of a diffusion-controlled matrix where drug percolates through a gel layer that thickens as t^0.5. Refine with Korsmeyer-Peppas using only early points below 0.6 fraction released: the log-log slope comes out near 0.6, anomalous transport, meaning diffusion and polymer relaxation both carry release — telling the formulator that raising HPMC viscosity or grade will slow release through both channels. A Hixson-Crowell fit instead would mark erosion-led release.
Now the comparability question: a proposed site change shifts the early points, 22 versus 30 per cent at 1 hour, converging later. Compute f2 across matched time points: the squared differences sum modestly, f2 lands around 55 — inside 50-100, similar. But if the curves cross, one faster early and slower late, f2 can read deceptively; that is when the Weibull fit earns its keep, comparing α (scale, the time to 63.2 per cent) and β (shape) separately to expose the crossing. And the ultimate purpose surfaces in IVIVC: with a level A correlation, dissolution becomes a surrogate for bioavailability, and the laboratory beaker inherits the authority of a clinical study — the reason regulators scrutinise the methods, media and sink discipline behind every fitted curve.
Where students slip
The f1/f2 pair is remembered backwards: f2 is the similarity factor (higher is more similar, 50-100), f1 the difference factor (0-15) — writing "f2 below 15" is an instant marker error. Second, the Korsmeyer thresholds are quoted for cylinders but applied to any shape; slabs shift the Fickian boundary to 0.5 and spheres to 0.43, and stating the geometry alongside the number is what distinction answers do. Third, fitting the entire curve with Korsmeyer-Peppas: the model holds only below 0.6 Mt/M∞, and late points bend the log-log line meaninglessly. Fourth, model choice by best correlation coefficient alone — a Weibull will empirically fit almost anything, so pair statistical fit with mechanism (matrix diffusion, erosion, osmosis). Finally, forgetting sink conditions invalidates the entire discussion: a saturated medium measures solubility, not the dosage form.
Frequently asked questions
What does the Korsmeyer-Peppas exponent n indicate?
For cylindrical matrices, n ≤ 0.45 means Fickian diffusion, 0.45-0.89 anomalous transport combining diffusion with polymer relaxation, and n near or above 0.89 near zero-order case-II release.
How is the similarity factor f2 calculated and interpreted?
f2 = 50·log{[1 + (1/n)Σ(R − T)²]^−0.5 × 100}; values 50-100 declare similarity, computed over matched points from at least 12 units.
When is f2 calculation unnecessary?
When both formulations dissolve 85 per cent or more within 15 minutes, regulatory convention treats them as similar without modelling — very rapid dissolution makes comparison moot.
What is the Higuchi model and its assumption?
Q = kH·t^0.5 describes Fickian diffusion from an inert, porous matrix, assuming little dimensional change and constant diffusivity.
What does the Weibull shape parameter beta reveal?
β = 1 exponential, above 1 sigmoid with lag, below 1 fast-initial parabolic — descriptive flexibility where mechanism is unclear.