Does FDA require a specific apparatus for long-acting injectables?

FDA lists apparatus product by product for long-acting injectables and doesn't require them. Beyond the examples above, a 2026 review reproduces entries with reciprocating baskets for a dexamethasone intravitreal implant and a paddle for triptorelin and triamcinolone products.

How many replicates does a depot release study need?

Replicate counts for depot release studies have no published rule that we've found. FDA's 1997 guidance for oral extended-release products asks for profiles from at least 12 units per lot, with a coefficient of variation under 10% for the mean profile, when building an IVIVC. That text wasn't written for depots, so treat it as a reference point.

Is a Level A IVIVC the only useful correlation?

A Level A IVIVC is the most informative correlation, and authors on PLGA products differ on whether anything less helps. Wang and colleagues write that only Level A helps with regulatory compliance. Lusina Kregar and colleagues say lower-level correlations may still support regulatory decisions when they're justified, validated and fit for purpose.

Does a biorelevant medium make an IVIVC more likely?

A biorelevant medium doesn't guarantee an IVIVC, and Wang and colleagues write that an in vivo-mimicking medium isn't an absolute requirement for a sound one. They also call it highly improbable that any in vitro method will simulate in vivo conditions, which may vary over time.

In vitro release testing for depots

Start with the decision the test has to support, because no standardized compendial release method exists for long-acting injectables, a 2026 review says. Every choice is then yours to justify: method, medium, sampling, recovery. Then prove two things: the method separates a good batch from a bad one, and it runs long enough to describe the depot.

Why is there no standard release method for depots?

Compendial methods haven’t caught up with these products. FDA said so in its FY2016 regulatory science report, writing about biodegradable microspheres and in situ gelling implants: “Compendial in vitro release methods for these complex formulations are not well developed.” Ten years later, a 2026 review of PLGA products says there are “still no standardized compendial IVR methods” for them, and that no universal method is possible given how differently these products release.

USP does have a chapter. USP <1001> covers release test methods for parenteral preparations. Lusina Kregar and colleagues, who wrote that review, describe it as giving examples of what has worked, such as Apparatus 2, 4 and 7 and an incubation jar for in situ forming preparations and implants.

FDA’s Dissolution Methods Database shows how far apart the methods sit. Their review reproduces its entries:

Product	Setup	Medium and volume	Last sample
Risperidone extended-release suspension	Flow-through cell, closed loop	1,000 mL phosphate buffer, pH 7.4	264 hours
Goserelin implant, 10.8 mg	120 mL jar, swirled before sampling	50 mL PBS, pH 7.4, at 39 °C	84 days
Naltrexone injectable suspension	Sponsor develops a method on a flow-through cell, a paddle if applicable, or another appropriate method	PBS with 0.02% polysorbate 20 and 0.02% sodium azide, at 37 °C	None listed

Those entries are precedent. FDA’s disclaimer says the listed methods “are not required to be used by applicants”, and it’ll consider alternatives supported by appropriate data.

Which job is the release test doing?

A release test usually does one of three jobs, and a method built for one can fail another. Lusina Kregar and colleagues draw the first line: quality control methods stay simple and dependable, while development work often uses more complex biorelevant media.

Job	What the test needs	What it costs
Choosing between candidates	Real-time conditions held the same across candidates, long enough to see the whole curve	Weeks to months per run
Batch quality control	A simple, dependable method that discriminates, often accelerated	“It cannot be assumed” that an accelerated test catches every important failure, says the USP expert panel
Groundwork for a correlation	Formulations with different release rates, run under fixed conditions and over a period that matches the in vivo data	An in vivo arm, which the bench can’t supply

Write down which job you’re doing, and which parts of the test are meant to resemble the body. The panel recommends stating both.

Which release method should you start with?

Sample and separate is the most common starting point, the 2026 review says, and flow-through and dialysis are the usual alternatives. Our depot release methods comparison covers the choice in detail.

Method	How it runs	Where it struggles
Sample and separate	Depot in a vessel. Withdraw medium, filter or centrifuge, replace. Flexible on volume and suited to long runs	Microparticles float, aggregate and get pulled out with the sample
Flow-through cell (USP Apparatus 4)	Medium pumped through a cell, open or closed loop. Standardized hardware, easy sampling	Equipment cost, blocked filters, O-ring and filter failure in long runs
Dialysis	Depot in a membrane device with medium outside. Easy sampling	Poor mixing inside, sink lost inside the bag, slow equilibration, drug binding to the membrane

The choice can change the answer. Bao and colleagues tested medroxyprogesterone suspensions, which aren’t PLGA, modeled on a product given every 3 months. Release took about 1 week on a flow-through cell with semisolid adapters, about 2 weeks with enhancer cells, and over 100 days in dialysis sacs. The authors put their three retained methods at about 1 to 7 weeks, so the method with the least variability finished in a fraction of the product’s dosing interval. Wang and colleagues, reading 37 IVIVC studies on PLGA microparticles, say it’s hard to know in advance which method will correlate best.

Agitation belongs in the design too. Lusina Kregar and colleagues cite work in which vessel dimensions, agitation speed, beads and exchange volume each changed the profile of one formulation. A conference abstract on leuprolide microspheres reports higher release under continuous agitation than under once-a-week agitation. The USP panel recommends enough mixing to reach low-velocity regions, and cautions against assuming sink at very low flow rates.

An in situ forming depot adds a step. Lusina Kregar and colleagues call the way the implant forms “a primary determinant” of release performance, so fix how the depot forms and what shape it takes before you compare anything.

How do you choose the release medium?

Start from the drug’s solubility and stability, then add only what you can defend. Lusina Kregar and colleagues give the usual choices:

Buffer: pH 7.4 phosphate-buffered saline is the most common medium.
Surfactant: polysorbate at about 0.02% to 0.05% is frequent. It changes the medium the depot sits in, so check the effect on release.
Preservative: sodium azide at 0.01% to 0.15% limits microbial growth over long runs. Confirm it doesn’t interfere with your assay.
Drug stability: if the drug degrades in the medium, try another pH, an antioxidant, full medium replacement or an open-loop flow-through. Each one changes the test you’re running.
Temperature: 37 °C is usual for intramuscular and subcutaneous products. Rawat and colleagues, validating an accelerated flow-through method for commercial risperidone microspheres, needed tight control of temperature and pH, plus deaeration through the run.
Volume: set it against solubility and analytical sensitivity. Sink conditions has the calculation.

Don’t read a quality control medium as a model of the injection site. The USP expert panel notes that media in FDA’s database run from pH 3 to 7.4 at 25 to 47 °C for injectables and implants, designed for quality control “but not intended to be biorelevant”. It adds that PBS at pH 7.4 discriminates among formulations without being considered clinically relevant. As of 2022 it said no clinically relevant fluid existed for subcutaneous testing, and it called simulated subcutaneous interstitial fluid a first step toward a more biorelevant one.

How long should you run it, and how often should you sample?

Run long enough to see every release phase, and sample where the curve changes. The USP expert panel recommends a duration and frequency sufficient to characterize release kinetics and, where appropriate, degradation. It gives no schedule, and we haven’t found one that fits every depot.

PLGA products typically release in three phases, an initial burst, a slower middle and a later phase, Lusina Kregar and colleagues say, though some phases can be negligible. That points to dense sampling early and through the transitions, which is our reading rather than a published rule. Initial burst release covers the first hours.

Every sample takes something with it. Bao and colleagues counted drug lost to sampling and medium replacement in their cumulative release. Microspheres pulled out on filters or probes need to go back in, Lusina Kregar and colleagues add.

Accelerated runs buy time and give up certainty. Lusina Kregar and colleagues say acceleration should raise the rate without changing the mechanism, or at least keep the rank order between formulations. They note the general recommendation to stay below the glass transition temperature, since going above it may alter the mechanism, and they report that high temperature doesn’t suit every PLGA microparticle system. Run accelerated and real-time arms together until you’ve shown they agree.

How do you check recovery?

Account for all the drug: what’s in the medium, what’s left in the depot, and what you lost on the way. A plateau below 100% has several readings. The depot may still hold the drug, or the drug degraded, stuck to something or left with a sample.

Filter: determine how much drug it adsorbs. The USP panel lists this as a point to consider.
Membrane: load dissolved drug into the device and see how much comes through. Bao and colleagues reached almost 100% in 24 hours, which ruled out binding.
Stability: Lusina Kregar and colleagues suggest a nonspecific assay, or summing the peaks for drug and degradants, when the drug breaks down in the medium.
Residual drug: extract what’s left in the depot. It’s destructive and needs a lot of sample, so they say it isn’t generally preferred. The leuprolide abstract reported a satisfactory balance this way.
Evaporation: seal the vessel, or correct with an internal standard if you can’t. Then ask whether the smaller volume still holds sink.

How do you test discrimination between batches?

Make batches that differ on purpose and see whether the curves separate. Lusina Kregar and colleagues define discriminatory power as telling apart batches made with different critical process parameters or material attributes. Particle size is a common choice, says the USP panel.

Bao and colleagues show what that looks like. Release of a reference suspension and a test suspension overlapped in 100 kD dialysis sacs. Both separated with enhancer cells, and with a flow-through cell using semisolid adapters, which kept variability under 5% RSD against 15% for enhancer cells. Enhancer cells and the other apparatus 2 method gave the larger margin between formulations.

Two limits come with it:

Over-discrimination: methods that add agitation or flow sometimes exaggerate differences between formulations, the panel says. They also pick up slight differences more sensitively.
Relevance: “a discriminating method is not necessarily clinically relevant”, in the panel’s words.

For routine quality control, Lusina Kregar and colleagues say the method should be validated under ICH Q2(R2), including the analytical step.

How far can release data take you toward IVIVC?

Release data is one half of a correlation, and the other half is in vivo data. FDA’s 1997 guidance, written for oral extended-release forms, sets out the levels:

Level A: a point-to-point relationship between in vitro dissolution and in vivo input. FDA calls it the most informative and recommends it where possible.
Multiple Level C: relates pharmacokinetic parameters to the drug dissolved at several timepoints. It “can be as useful as” Level A, though FDA prefers Level A when multiple Level C is possible.
Level C: relates one dissolution figure to one pharmacokinetic parameter, which can help when pilot formulations are being selected.

The same guidance says human data should be supplied for regulatory consideration. It needs at least two formulations with different release rates and recommends three or more.

These products add their own problems. Wang and colleagues warn that FDA’s oral approaches may not extend readily to long-acting injectables without further scrutiny, and FDA’s FY2016 report describes Level A correlations for risperidone and naltrexone microspheres in a rabbit model. Wang and colleagues raise two more problems:

Reproducibility: no systematic studies have asked whether IVIVCs for these formulations are reproducible.
Duration: when in vitro and in vivo periods differ widely and neither nears completion, they call an IVIVC from those data “virtually impossible”.

Plan for a correlation before the first run: formulations with different release rates, conditions fixed once the method discriminates, and an in vitro period that matches the in vivo data.

What are the limits of a release test?

A release curve shows how much drug left the depot. It doesn’t show when the depot swelled, set or began to erode, and the release curve and the depot behind it explains what that gap costs. Polymer degradation testing covers the carrier’s side.

It can’t reproduce the injection site either. The USP panel describes a formulation meeting tissue and very little liquid, in an environment it calls difficult to simulate in vitro. If your question is about oral solids, or how the terms differ, dissolution testing is the page for it.

What a GelTech study measures on release testing

A GelTech study records depot mass and drug release on one timeline. Carousel runs six chambers per unit and weighs each sample to 1 mg at every timepoint, for weeks to months, from ambient to 60 °C in media from pH 1.5 to 12. Integrated instruments such as UV-Vis, HPLC and LC-MS read the drug in sampled medium.

Each 24 mL sample chamber sits inside a 50 mL incubation chamber and draws on its own bottle, in static, flow-through or circulate mode.

GelTech’s position is that weeks to months immersed at the temperature and pH you set sit closer to in vivo conditions than conventional bench methods, which gives IVIVC work a stronger basis. Carousel doesn’t replicate the human body, so an IVIVC still rests on the in vivo data it’s built against.

GelTech doesn’t run animal or clinical studies, or accredited biocompatibility or sterility testing. See drug delivery, how a pilot study works and the study brief checklist.

In vitro release testing for long-acting injectable depots