Why Sampling Design Matters
A dissolution profile is observed only at the times when samples are collected. A schedule that begins too late can miss rapid early release; one with widely spaced observations can conceal a steep transition; and one that stops too soon may not establish whether release has approached completion. Meanwhile, each withdrawn sample can alter the vessel’s volume and drug content. Sampling design is therefore part of the measurement method, not a clerical detail.
The design should follow the purpose of the test, the expected release behavior, the apparatus and method conditions, and the precision needed for the decision. The earlier material on sink conditions and hydrodynamics and on selecting a dissolution apparatus establishes conditions that must be held consistent. This tutorial focuses on when to sample and how many independent units and assay replicates to use; it does not prescribe a universal schedule or a universal sample size.
Build the Schedule Around the Release Behavior
Begin with the question the dissolution test must answer and a reasonable expectation of the profile. A pilot run, prior formulation knowledge, or a development target can indicate the likely onset, rapid-release interval, slower transition, and approach to completion. These expectations guide a candidate schedule; they do not justify treating an unobserved feature as known.
A useful schedule usually gives observations in three regions. Early samples establish whether release begins promptly and whether there is a lag. Several samples across the interval of expected rapid change help describe the profile’s shape rather than merely its beginning and end. Later samples show whether release continues and whether a near-complete level is reached within the test duration. The intervals need not be equally spaced: short intervals may be needed during rapid change, while longer intervals may suffice when the profile changes slowly.
Time points should be operationally feasible as well as scientifically useful. If withdrawal, filtration, and sample handling take appreciable time, the recorded sampling time should represent the actual collection time, using a consistent convention. The protocol should specify withdrawal volume, replacement medium and volume, temperature handling, filtration, and assay procedures. Sampling and replacement should not disrupt agitation or create inconsistent delays among vessels.
Worked Example: Placing Times Around a Rapid Release Interval
Suppose a hypothetical pilot for an immediate-release product gives approximate mean release values of 8%, 27%, 54%, 73%, 89%, 96%, and 98% at 5, 10, 15, 20, 30, 45, and 60 minutes, respectively. The development objective is to characterize early release, the steep middle portion, and the approach to completion.
The changes are 19 percentage points from 5 to 10 minutes, 27 points from 10 to 15, and 19 points from 15 to 20. By comparison, the change from 45 to 60 minutes is only 2 points. This pattern supports relatively close sampling early, where the profile changes rapidly, and less dense sampling late, where the observed change is small.
A candidate schedule is 5, 10, 15, 20, 30, 45, and 60 minutes. It includes multiple early observations, several observations across the rapid transition, and later observations to check for a near-complete level. The schedule is a design candidate, not proof that every time is necessary: the intended decision, method requirements, operational burden, and variability should be considered before finalizing it.
A schedule containing only 5, 20, and 60 minutes would be easier to execute but would leave the rapid interval sparsely observed. It could not show whether the increase occurred steadily or was concentrated in a short interval. Conversely, adding many closely spaced late samples after the profile has nearly stopped changing may add workload without useful information.
Correct for Drug Removed in Earlier Samples
When an aliquot is withdrawn and replaced with drug-free medium, some dissolved drug leaves the vessel. If the calculation uses only the concentration measured at the current time multiplied by the vessel volume, it omits drug removed in earlier aliquots. Under a simple, explicitly stated sampling model, cumulative amount can be corrected by mass balance.
Let \(C_i\) be the measured concentration at sampling time \(t_i\). Immediately before the \(n\)th sample, the vessel contains \(VC_n\) of dissolved drug. Each earlier sample removed \(vC_i\) of drug, which must be added back to estimate the cumulative amount released by time \(t_n\). Thus:
Here \(Q_n\) and the terms on the right are amounts of drug; when \(C_i\) is in mg/mL and volumes are in mL, \(Q_n\) is in mg. If the label amount is \(D\), the corresponding percentage released is \(100Q_n/D\). The correction is cumulative: it includes concentrations from all earlier withdrawals, not the concentration at the current time a second time.
Worked Example: Correcting for Replaced Aliquots
In a hypothetical test, the vessel volume is \(V=900\) mL and a \(v=5\) mL aliquot is withdrawn and replaced at each time. Measured concentrations at 5, 10, and 15 minutes are 0.006, 0.012, and 0.018 mg/mL. Calculate cumulative amount at 15 minutes and percentage released from a 100 mg dose.
At 15 minutes, \(C_3=0.018\) mg/mL. The current vessel amount is \(900(0.018)=16.2\) mg. The first two samples removed \(5(0.006)=0.030\) mg and \(5(0.012)=0.060\) mg.
The estimated percentage released is \(100(16.29/100)=16.29\%\). Using only \(VC_3\) would give 16.2 mg, underestimating the cumulative amount by 0.090 mg, the amount carried away in the two earlier aliquots. This calculation depends on the replacement and mixing assumptions stated above.
Choose Replicates for the Objective
Independent dosage units provide information about unit-to-unit variation under the specified test conditions. Repeated assay determinations from one aliquot can help quantify analytical repeatability, but they do not create additional independent dosage units. Likewise, repeated time points from one vessel are measurements on the same unit, not independent replicates of that unit.
The number of units should be justified by the purpose of the study, expected variability, desired precision, applicable method or regulatory requirements, and practical constraints. A compendial or approved method may specify a minimum number of units or a staged procedure; those requirements take precedence. In development work, a pilot estimate of between-unit standard deviation can support a precision-based plan. A calculation at one time point is a planning aid, not a guarantee of precision over the entire profile.
For a simple planning calculation, if the outcome at a prespecified time is approximately independent across units with standard deviation \(s\), the estimated mean has standard error \(s/\sqrt{n}\). An approximate two-sided confidence-interval half-width is \(t_{1-\alpha/2,df}s/\sqrt{n}\), where \(df=n-1\). This calculation assumes independent units and a suitable t-based interval for the mean; a small pilot can give an uncertain estimate of \(s\), so the resulting \(n\) should not be treated as exact.
Worked Example: Planning Units for a Precision Target
A hypothetical pilot suggests a between-unit standard deviation of 6 percentage points at a prespecified sampling time. The goal is an approximate 95% confidence-interval half-width no greater than 4 percentage points for the mean at that time. For planning, compare candidate values of \(n\) using \(t_{0.975,n-1}s/\sqrt{n}\).
For \(n=10\), \(df=9\) and \(t_{0.975,9}\) is approximately 2.262. The estimated half-width is \(2.262(6/\sqrt{10})=4.29\) percentage points, which exceeds the target.
For \(n=12\), \(df=11\) and \(t_{0.975,11}\) is approximately 2.201. The estimated half-width is \(2.201(6/\sqrt{12})=3.81\) percentage points, which is below the target. Under these planning assumptions, 12 independent units per formulation would meet the stated precision target at this time point.
This does not establish that 12 is required or sufficient for every purpose. The target concerns the mean at one time, the pilot standard deviation may be imprecise, and variability may differ at other times. The design must also comply with the applicable method and account for the study’s broader objective.
Check the Design and Its Failure Modes
Before a run, verify that every scheduled time can be collected consistently and that the chosen number of units is feasible. During and after a run, examine whether actual sampling times, withdrawal volumes, replacement volumes, and sample handling followed the protocol. Unplanned timing differences matter most when release is changing rapidly. Large or inconsistent withdrawals, incomplete replacement, poor mixing, or inconsistent filtration can make the measured profile reflect procedure as well as formulation.
The basic mass-balance correction is not a remedy for all sampling effects. Replacing an aliquot restores liquid volume but does not restore the drug removed, which is why the concentration correction is needed. Nor does the equation undo changes to hydrodynamics, temperature, or sink conditions caused by a poorly designed procedure. If sample volume is a material fraction of vessel volume or repeated withdrawals could influence conditions, consider reducing aliquot volume, using a suitable sampling arrangement, or otherwise validating the procedure. There is no single universal percentage threshold that makes sampling effects negligible.
A fixed schedule chosen before testing supports consistent comparison and reduces the temptation to select times after seeing favorable results. An adaptive schedule can be useful in exploratory work, where early observations guide later collection, but it requires a prespecified decision rule and careful recording; it is not interchangeable with a fixed schedule for every intended comparison. Similarly, adding technical assay replicates can improve precision of concentration measurement, but if the principal uncertainty is between dosage units, increasing independent units is the more relevant response.
Sampling also has limits as a description of release. A finite set of observations cannot reveal every change between collection times, and apparent completion depends on the assay, method conditions, and observation window. The design should therefore support the intended interpretation without implying that unmeasured portions of the profile are known exactly.
Practical Design Sequence
Specify whether the test must resolve early release, a transition, late release, or another prespecified feature.
Use expected behavior or pilot observations to place closer times where change is rapid and later observations where completion must be assessed.
Set aliquot volume, replacement procedure, timing convention, handling, and the applicable cumulative-amount calculation.
Use the study objective, pilot variability, precision needs, and applicable method requirements; distinguish these units from assay repeats.
Confirm reproducible collection and assess whether sampling or handling could alter conditions or obscure the intended profile.
Check Your Understanding
Answer each question, then compare with the solutions below.
- Why might a schedule use shorter intervals during an expected rapid-release period than near a late plateau?
- A vessel has \(V=500\) mL. Samples of \(v=4\) mL are replaced. Concentrations at the first two times are 0.010 and 0.020 mg/mL. What is the corrected amount at the second time?
- Why do duplicate assays on one aliquot not count as two independent dosage-unit replicates?
- Name two assumptions required for the stated cumulative-amount correction.
Solutions. (1) Closer intervals provide more observations where release changes quickly, while dense sampling of a nearly unchanged late profile may add little information. (2) \(Q_2=500(0.020)+4(0.010)=10.04\) mg: 10.0 mg remains in the vessel, and 0.040 mg was removed in the earlier sample. (3) Both determinations measure the same collected material from one dosage unit, so they inform assay repeatability rather than between-unit variation. (4) For example, the vessel is well mixed, the equal-volume aliquot is replaced with drug-free medium, and concentration is representative; the correction also assumes no relevant unmeasured drug loss.