Postdoctoral PDDM921 ⭐ 3 Credits Semester 3

Process Engineering, Scale-Up, Hold Times, Production Operations, and Process Validation

This course takes an arbitrary, already-classified dermatologist formula and turns it into a controlled, scalable, validated semisolid manufacturing process. The formula comes from PDDM901 (classification), PDDM902 (materials) and PDDM903 (operating envelope). The course is API-agnostic and built around five process families: anhydrous hydrocarbon ointment, water-soluble (PEG) ointment, O/W cream, lotion, and aqueous or hydroalcoholic gel. The process flow starts at 21 CFR 211.101 weighing and dispensing and runs through: - phase preparation; - emulsification order and rate; - rotor-stator homogenization, in the vessel or in an external in-line loop under vacuum; - scraped-wall controlled cooling; - staged addition of the API and heat-sensitive materials; - vacuum deaeration; - bulk hold and transfer to the filler. CPPs are taken from the FDA-authored reviews by Chang et al.: phase-mixing temperature, mixing speed, rotor-stator gap, rotor speed, powder addition rate, homogenization time, and cooling rate. Equipment expectations come from the FDA Guide to Inspections of Topical Drug Products (7/94): - walls scraped by design, not by hand spatula; - dead spots either recirculated or the material from them not used; - recording thermometers where temperature is critical; - continuous mixing of suspensions during filling, with a plan for the tank heel. Scale-up uses SUPAC-SS (May 1997) only as a scientific change language. It is written for NDA/ANDA products, and its change levels are tied to equipment design and operating principle. Engineering rules such as tip speed, energy per volume, shear rate and heat-transfer scaling are standard practice but are not FDA-specified, so they are taught as (verify) hypotheses that must be confirmed at each scale. Operations content covers: - bulk hold and transfer studies with chemical, physical and microbiological arms; - equipment occupancy and scheduling, including campaign versus dedicated lines (for example, an oxidizer line for benzoyl peroxide); - transfer losses, hold-up, yields and reconciliation; - maintenance downtime. Process validation is taught as a justified design, not a fixed batch count. Students set the number and scale of batches with a risk rationale, define worst-case conditions, write bulk and filling-run sampling plans, set acceptance criteria tied to specifications and IPCs, and plan continued monitoring. This responds directly to the §211.100(a) process-validation citation in the 2025 Columbia Cosmetics warning letter. Peptide Partners owns no semisolid manufacturing equipment. Every equipment step is therefore framed as an evidence-gated purchase (bench kit → CMO pilot as a bridge → pilot or commercial vacuum emulsifier), backed by user requirement specifications (URSs). If the bench kit has been bought, placebo training batches are made and characterized on the existing HPLC-DAD and through contract tests, then handed to PDDM924 for contract stability.

📅 Studied by Skipper on 2026-10-16

📖 Core Concepts

Claude Opus 5.5

Core Concepts: Process Engineering, Scale-Up, Hold Times, Production Operations, and Process Validation

PDDM921 · Skipper's personal textbook · Program 2 (OTC dermatological semisolids)

0. What this subject is really about

In the PDQC program you learned to measure a peptide that already existed. A lyophilized peptide's quality is mostly fixed at the molecular level: sequence, stereochemistry, counter-ion, water, and related substances. The process that made it matters, but you could take the cake and characterize it almost completely with HPLC, MS, KF, and a few physical tests.

A cream behaves differently, and that difference is the core of this course. A semisolid is a structure, not a solution. An O/W cream is oil droplets, sized by the shear history. It also contains a crystalline gel network of fatty alcohol and surfactant, whose crystal form and lamellar spacing were set by how fast it cooled. It carries entrapped air if deaeration was poor, and dissolved or suspended API whose physical state depends on when the API was added and at what temperature. Two batches with the same quantitative formula can differ in viscosity, release rate, stability, and fill weight. They can even differ in whether they pass USP <3> uniformity. The only difference needed is that one was cooled in 8 minutes and the other in 40.

For that reason FDA's generic topical reviewers write that a topical product's performance depends on both "Q1/Q2" (qualitative and quantitative sameness of the formula) and "Q3", meaning microstructure and physicochemical arrangement. They explicitly name process parameters as determinants of Q3 [Chang et al., 2013a][Chang et al., 2013b]. The biologics mantra "the process is the product" applies almost literally to semisolids, even with a small-molecule API.

Analogy: think of a semisolid like steel rather than like table salt. Salt is salt however you crystallize it, at least to a first approximation. Steel with one composition can be soft, hard, or brittle depending on its quench and temper history. A cream also has a metallurgy: its properties are path-dependent. This course teaches you to:

  • specify the path;
  • control it;
  • prove it is reproduced at scale and across time;
  • write it down so a regulator can verify it.

The regulatory spine is short. Learn to hear each clause as an engineering requirement:

"There shall be written procedures for production and process control designed to assure that the drug products have the identity, strength, quality, and purity they purport or are represented to possess."
— 21 CFR 211.100(a) (wording to be verified against current eCFR)

The key word is designed. A procedure cannot assure quality unless its design is based on knowledge of which steps matter. The 2025 warning letters studied in Week 1 (e.g., Columbia Cosmetics, cited under §211.100(a) for lacking process validation) show what happens when a firm writes procedures that describe what operators do but were never shown to produce consistent product. Several other sections convert into design obligations:

  • §211.110 (in-process controls): its listed examples include adequacy of mixing to assure uniformity and homogeneity (verify paragraph numbering).
  • §211.111 (time limitations on production): requires time limits for each phase of production "when appropriate". This is the regulatory root of hold-time studies (verify wording).
  • §211.103: calculation of yield.
  • §211.188: batch records.
  • §211.192: investigation of discrepancies.

Read together, these sections say: know your process, put limits on it, record that you stayed inside them, and investigate when you didn't.

1. The process flow map: CQAs, CPPs, IPCs, and hold points as a single graph

Every semisolid process in this course reduces to one generic chain:

dispense → phase prep (aqueous | oil) → combine/emulsify → homogenize → cool
        → late additions (API, fragrance, heat-labile preservative) → deaerate
        → hold / IPC → transfer → fill

The families (hydrocarbon ointment, PEG ointment, O/W cream, lotion, gel) are deletions or substitutions on this chain:

  • A hydrocarbon ointment has no aqueous phase.
  • A carbomer gel replaces "emulsify" with "disperse → hydrate/swell → neutralize."

Each node gets four tags:

  1. CQAs affected: the product properties that must be within limits.
  2. CPPs: the process parameters at that node whose variation moves a CQA.
  3. IPCs: the measurements taken at or after the node.
  4. Hold point: whether material may wait here, for how long, and under what conditions.

Use the ICH definition of a CPP precisely, because people misuse the term constantly:

A critical process parameter is "a process parameter whose variability has an impact on a critical quality attribute and therefore should be monitored or controlled to ensure the process produces the desired quality." — ICH Q8(R2) glossary (verify exact wording)

Criticality is therefore an empirical claim about sensitivity, not a statement about importance. Homogenizer speed is a CPP only if your DoE shows that globule size or viscosity moves when speed moves within plausible operating variation. If the response is flat across the range you will actually run, speed is a key parameter that you control, but it may not be critical. The Chang et al. QbD paper lists candidate CPPs for creams:

  • phase-mixing temperature;
  • mixing speed;
  • rotor-stator gap and rotor speed;
  • powder addition rate;
  • homogenization time.

They are named as candidates to be studied, not as automatic CPPs [Chang et al., 2013b].

The flow map is the backbone of three deliverables: the MBR template (each node becomes a section with CPP fields), the risk assessment (each node-CQA pair is scored), and the PV sampling plan (each node with a hold point becomes a sampling opportunity). Build it once and carefully.

Why the drug standard governs the whole line. If a cosmetic-only product regulated under MoCRA shares a vessel with an OTC drug, that vessel's cleaning, maintenance, and qualification must meet §211.67 and §211.42 in any case. You cannot run a "cosmetic-grade" cleaning on Tuesday and a drug-grade cleaning on Wednesday on the same equipment. The drug's carryover acceptance criteria apply to whatever ran before it. The stricter regime absorbs the shared asset.

2. Emulsification physics: why droplets are the size they are

Homogenization is a contest between disruptive stress, which pulls a droplet apart, and interfacial (Laplace) stress, which holds it together. The Laplace pressure inside a droplet of radius r is:

ΔP_Laplace = 2σ / r          σ = interfacial tension (N/m)

Smaller droplets resist more strongly, so each halving of size requires more than double the stress. The physics depends on the flow regime.

Turbulent-inertial regime (lotions and low-viscosity premixes in high-shear rotor-stators). Breakup is governed by a Weber number. Hinze's classic result gives the largest stable droplet in terms of energy dissipation rate ε (W/kg):

We = ρ_c · u² · d / σ

d_max = C · (σ / ρ_c)^(3/5) · ε^(−2/5)          (Hinze–Kolmogorov; valid when d_max ≫ Kolmogorov scale)

The exponent −2/5 is a lesson in diminishing returns. Doubling the local power dissipation shrinks d_max by only about 24% (2−0.4 ≈ 0.76). The other 76% of the droplet size remains, and the extra energy becomes heat [Paul, Atiemo-Obeng & Kresta, 2004].

Viscous (laminar) regime (most creams once they thicken). Breakup is governed by a capillary number and the viscosity ratio p = μ_d/μ_c:

Ca = μ_c · γ̇ · r / σ           breakup when Ca > Ca_crit(p, flow type)

One result here is counterintuitive and has practical consequences. In simple shear, Ca_crit rises steeply as p increases. Above roughly p ≈ 3–4, a droplet cannot be broken by simple shear at all, because it rotates and stretches without rupturing. Elongational flow has no such ceiling. This fact, from the Grace curve and summarized in the mixing handbook, explains three practical observations:

  • Rotor-stator heads (strong elongational and turbulent flow in the slots) outperform an anchor for droplet size.
  • Emulsifying a high-viscosity wax phase into a thin aqueous phase is harder than the reverse.
  • Emulsification temperature matters partly through the viscosity ratio. Molten waxes at 75 °C are much less viscous than at 60 °C, so a cooler "matched" phase temperature can quietly push the system across the breakup ceiling.

Order of addition and phase inversion. Adding the discontinuous phase into the continuous phase (oil into water for O/W) keeps the intended continuous phase continuous throughout. The reverse, water into oil for an O/W product, passes through a W/O state and then inverts. Inversion can be:

  • catastrophic, driven by the dispersed-phase volume fraction exceeding a critical value; or
  • transitional, driven by a change in surfactant affinity, e.g., temperature crossing the phase-inversion temperature for ethoxylated nonionics.

Inversion processes can give very fine droplets when well controlled, which is why some formulators use them deliberately. The droplet size, however, becomes strongly sensitive to addition rate and temperature at the inversion point. That sensitivity is why Chang et al. treat addition rate as a controlled parameter rather than "add slowly" [Chang et al., 2013a]. The mechanism details (catastrophic vs. transitional, PIT applicability) are marked "verify" in the syllabus. Confirm against the Disperse Systems volumes [Lieberman, Rieger & Banker, eds., 2nd ed.] before writing them into an MBR rationale.

Over-processing is real. More homogenization time is not monotonically better. Excess shear:

  • heats the batch, raising local temperature at the stator, which matters for heat-labile preservatives;
  • entrains air;
  • can break the crystalline gel network as it forms during cooling;
  • in gels, can mechanically degrade polymer and permanently lower viscosity.

The DoE in Week 3 must therefore span enough range to find the upper edge of the operating envelope as well as the lower one.

The IPC that sees this is microscopy: globule size distribution and absence of crystals above a product-specific size. The lab has no microscope (PDDM912 gap). Until one is purchased, the image-based IPC is either contracted out or performed by the CMO during pilot work. In either case the acceptance criterion and the image record format belong in the MBR now.

3. Mixing scale-up rules: there is no single invariant

When moving from a 5 L bench vessel to a 500 L commercial vessel, you must choose what to hold constant. The candidate invariants are below [Paul, Atiemo-Obeng & Kresta, 2004][Levin, ed., 3rd ed.]; the syllabus correctly flags all of these as engineering rules, not FDA-specified rules.

Tip speed:          v_tip = π · D · N                      (m/s; D = rotor diameter, N = rev/s)
Reynolds number:    Re = ρ · N · D² / μ                    (impeller Re; μ → μ_apparent for non-Newtonian)
Power (turbulent):  P = N_p · ρ · N³ · D⁵                  (N_p ≈ constant for Re ≳ 10⁴)
Power (laminar):    P = K_p · μ · N² · D³                  (Re ≲ 10)
Power per volume (turbulent, geometric similarity, V ∝ D³):
                    P/V ∝ N³ · D²
Metzner–Otto average shear rate in stirred vessel:
                    γ̇_avg ≈ k_s · N                        (k_s ≈ 10–13 for many impellers; verify for your geometry)

Under geometric similarity, the invariants conflict:

  • Constant tip speed (N ∝ 1/D): P/V ∝ N³D² ∝ D−1, so power per volume falls with scale, and bulk turnover slows.
  • Constant P/V (N ∝ D−2/3): tip speed ∝ ND ∝ D1/3, so peak shear at the blade rises with scale.
  • Constant Metzner–Otto shear rate (N constant): tip speed and P/V both rise sharply, which is usually impractical at commercial scale.

The right choice depends on what controls your CQA:

  • Droplet breakup in the rotor-stator gap responds to the maximum local stress, approximated by tip speed and gap. For an emulsion whose globule size is set at the homogenizer head, tip speed (with gap width held constant) is the usual first choice.
  • Bulk blending and heat removal during cooling respond to anchor turnover and wall scraping.
  • Polymer hydration in a gel responds to dispersion quality at the moment of powder wetting (powder addition rate, induction under vacuum) much more than to steady-state shear.

Most real vessels separate these functions: a high-shear head for droplets and a slow anchor for bulk motion and heat transfer. Scaling each function by its own invariant is legitimate.

Worked calculation: matching tip speed bench → pilot.

Bench probe rotor: D = 25 mm, N = 10,000 rpm = 166.7 rev/s
  v_tip = π × 0.025 m × 166.7 s⁻¹ = 13.1 m/s

Pilot in-line rotor-stator: D = 150 mm
  N required = v_tip / (π D) = 13.1 / (π × 0.150) = 27.8 rev/s ≈ 1,670 rpm

Check: does the pilot unit's VFD range include 1,670 rpm?  → URS line item.

Tip speed alone is not enough for an in-line rotor-stator in a recirculation loop. In that configuration, each fluid element sees the high-shear zone only while passing through it. The relevant dose is the number of passes (tank turnovers):

turnovers = Q · t / V          Q = loop flow rate, t = homogenization time, V = batch volume

A bench probe immersed in a beaker treats nearly the whole volume continuously. A 500 L tank with a 50 L/min loop turns over only once every 10 minutes. "Homogenize 10 minutes" therefore does not transfer unchanged. The transferable statement is "homogenize at v_tip = 13 m/s, gap 0.5 mm, for ≥ N turnovers." This is a typical example of a parameter that is scale-dependent versus one that is scale-independent, and the tech-transfer package must state CPPs in scale-independent form wherever possible.

Evidence that this matters. Experimental work on emulsification parameters (Calvo et al., 2022, Chem Eng Res Des 186:407–415, content to be verified) reports that homogenizer speed, time, and temperature measurably change rheology, texture, and stability. The practical consequence is non-negotiable: comparative CQA testing at each scale is mandatory. Scale-up rules generate hypotheses, and comparative batches test them.

4. Thermal history: the slowest variable, and the one scale-up breaks

Heating the phases to about 70–75 °C (product-specific; verify per formula) melts waxes and fatty alcohols, lowers the viscosity ratio for emulsification, and keeps phases matched so that wax does not flash-crystallize at the interface. The subtler variable is cooling. As an O/W cream with cetyl/stearyl alcohol cools, the excess fatty alcohol and surfactant form a lamellar gel (α-crystalline) network in the continuous phase. That network is most of the cream's viscosity and yield stress. Its structure depends on cooling rate and on the shear applied while it forms. A dissolved API near saturation at 75 °C can also become supersaturated on cooling and crystallize, creating particles that change uniformity, appearance, and release.

The 7/94 FDA Guide to Inspections of Topical Drug Products expects jacket heat and high-energy-mixer heat to be controlled. Its reasons are that excess heat can degrade API, vehicle, or preservatives and shift particle size or crystal form. It also prefers recording thermometers over hand-held readings where temperature is critical. Here is why scale makes this hard.

Lumped heat-transfer model. Treat a well-mixed batch cooled by a jacket at constant temperature T_j:

m · c_p · dT/dt = − U · A · (T − T_j)

⇒  (T − T_j) / (T₀ − T_j) = exp(−t / τ),      τ = m·c_p / (U·A) = ρ·c_p / [U·(A/V)]

For a cylinder with H = D and a jacketed side plus bottom, A ≈ 1.25πD² and V = πD³/4, so A/V ≈ 5/D. The surface-to-volume ratio falls linearly with vessel diameter. At geometric similarity, a 100× increase in volume is a 1001/3 ≈ 4.64× increase in diameter.

Worked calculation: cooling time, bench vs. commercial.

Assume ρ = 1000 kg/m³, c_p = 3500 J/(kg·K), U = 300 W/(m²·K) (same U — optimistic at scale)
Cool from T₀ = 75 °C to T = 35 °C with jacket T_j = 15 °C
  ln[(T₀−T_j)/(T−T_j)] = ln(60/20) = ln 3 = 1.099

Bench (5 L):    D ≈ 0.185 m → A/V ≈ 27 m⁻¹
  τ = 3.5×10⁶ / (300 × 27) ≈ 432 s;   t_cool ≈ 1.099 × 432 ≈ 475 s ≈ 8 min

Commercial (500 L):  D ≈ 0.86 m → A/V ≈ 5.8 m⁻¹
  τ = 3.5×10⁶ / (300 × 5.8) ≈ 2010 s;  t_cool ≈ 1.099 × 2010 ≈ 2210 s ≈ 37 min

The same recipe cools almost five times more slowly at scale, and this is the optimistic case. In reality, U falls as the cream thickens, because the viscous boundary layer at the wall grows. That is why wall scrapers matter: they renew the boundary layer. The tempting fix is to drop the jacket temperature lower, but that creates a large wall-to-bulk gradient. Material at the wall chills and crystallizes into a skin while the core is still hot. If it is not scraped off continuously, you get lumps and non-uniformity. The 7/94 guide's observation that hand-spatula scraping causes non-uniformity is the regulatory record of this failure.

The design response has four parts:

  1. Define the cooling profile (temperature vs. time, including any isothermal hold during network formation) as the thing to reproduce, not the jacket setpoint.
  2. Use jacket setpoint ramps and cooling time at scale to approximate that profile.
  3. Record it on a chart attached to the batch record.
  4. Verify the product by comparative CQAs (viscosity, microscopy for crystals, specific gravity).

Vessel temperature mapping with multiple calibrated probes (Week 5 OQ/PQ) is needed because one probe near the jacket does not represent the batch.

Late additions. The API, fragrance, and heat-labile preservatives are added below a validated temperature. "Validated" means you have data showing that degradation above that temperature is unacceptable and that dispersion below it is complete. The API-addition stage may be a CPP, with an IPC confirming the addition was complete and dispersed. Special-handling flags from PDDM902 attach here, and they stay API-agnostic:

  • photolabile actives need amber or actinic lighting;
  • oxidizable actives need inert-gas blanketing and chelators;
  • oxidizer actives such as benzoyl peroxide need a dedicated line and benzene control.

5. Rheology of structured systems: viscosity is a history, not a number

Semisolids are non-Newtonian. A useful constitutive model is Herschel–Bulkley:

τ = τ₀ + K · γ̇ⁿ          τ₀ = yield stress, K = consistency, n < 1 for shear-thinning
apparent viscosity  η_app = τ / γ̇ = τ₀/γ̇ + K · γ̇^(n−1)

Three consequences follow for process engineering.

(a) An apparent viscosity is meaningless without its shear rate. The single-point "viscosity" IPC under USP <912> is defined by spindle, speed, temperature, and time-at-speed. Change any of these and you have a different number. The method, not the instrument, carries the meaning (the PDQC802 lifecycle logic again).

(b) Structure rebuilds after the process stops, and sometimes doesn't. Many creams are thixotropic and also continue to build structure for 24–48 h as the gel network matures. Viscosity measured at the IPC hold point is therefore not the release value, and the two need separate limits. Some structure damage is irreversible, such as polymer chain scission in gels or crystalline network fragments that never re-knit. Pumps and in-line shear during transfer can permanently lower viscosity. That is why the transfer study (Week 10) measures viscosity and specific gravity before and after transfer, and why pump type matters. Positive-displacement low-shear pumps (lobe, progressive cavity) are typical for structured bulks; centrifugal pumps generally are not.

(c) Carbomer gels are a special, forgiving case. Neutralized carbomer forms swollen, crosslinked microgels that pack into a yield-stress, pseudoplastic network. Islam et al. found that the rheology of topical carbomer gels changes relatively little between about pH 5.0 and 8.0 [Islam et al., 2004]. The robust plateau means neutralization endpoint pH can be a CPP with a reasonably wide range inside it. Off the plateau, and especially at low pH where neutralization is incomplete, viscosity is steeply sensitive. The CPPs that are not forgiving are dispersion and swelling. Polymer added too fast forms "fish-eyes": lumps whose wetted, gelled outer skin blocks water from the dry core. Once formed, they are very hard to remove without high shear, which may damage the polymer. Powder addition rate, induction method, and swelling time before neutralization are the real CPPs. The Dec 2023 FDA carbomer guidance (benzene-free grades, ≤2 ppm benzene per course materials; verify) reaches back to the dispensary and supplier qualification.

(d) Classification depends on rheology. Buhse et al. proposed a topical classification decision tree that uses composition (water and volatiles content, hydrocarbon/polyol content) and flow behavior to distinguish ointments, creams, gels, and lotions [Buhse et al., 2005]. Verify the specific numeric cut-offs against the paper before quoting them. This links back to PDDM901/903: the process family is assigned from classification, and classification is partly a rheological fact.

6. Air, density, and fill weight: a lesson in what "assay" means

Entrapped air lowers the observed density of the bulk:

ρ_obs = ρ_true · (1 − φ)          φ = volume fraction of air

Many fillers dose by volume (piston fillers). If the filler is set from the theoretical density but the bulk carries air, each container receives less mass.

Worked calculation.

ρ_true = 0.980 g/mL; target fill 30.0 g → piston set to 30.0/0.980 = 30.61 mL
Bulk has 4% air (φ = 0.04): ρ_obs = 0.980 × 0.96 = 0.941 g/mL
Delivered mass = 30.61 × 0.941 = 28.8 g   → 4% underfill (1.2 g per tube)

Assay reported as % w/w on a weighed sample is not changed by air, because air has negligible mass. 4% air by volume does not dilute the API per gram. The course statement that air "can shift assay and fill weight" is correct, but the mechanism must be understood precisely. Air shifts these results in three ways:

  • Amount per container: fill weight falls, so label claim per tube falls.
  • Volumetric sampling: any sample taken by volume (e.g., a syringe draw for an in-process aliquot) is short on mass.
  • Chemistry: dispersed air is a dissolved-oxygen reservoir distributed throughout the matrix, which can accelerate oxidative degradation and show up as real assay loss over stability.

Air is also often non-uniform: it migrates and coalesces during hold, so top-of-tank and bottom-of-tank material differ.

That is why specific gravity is an IPC and why the March 2026 FDA Q3 characterization guidance (final; course reference to item 9, verify numbering) lists specific gravity among characterization attributes. Measurement uses a pycnometer or weight-per-gallon cup, which the lab does not own; this is a tracked gap.

Vacuum deaeration with low-speed mixing removes air. Vacuum level and time are parameters, with failure modes in both directions:

  • Too little vacuum or time: residual air.
  • Too much: foaming (bubbles expand faster than they escape through a yield-stress matrix) and stripping of volatiles. Alcohol in a hydroalcoholic gel, fragrance components, and water itself at elevated temperature follow vapor-pressure physics. Deep vacuum on a warm batch can boil the aqueous phase.

The deaeration window therefore depends on temperature as well as vacuum.

7. Suspensions, creaming, and uniformity: where gravity meets the yield stress

For a dispersed droplet or particle in a Newtonian continuous phase, Stokes' law gives the terminal velocity:

v = 2 · r² · Δρ · g / (9 · μ)

Worked calculation: lotion creaming during hold.

Oil droplet d = 5 µm (r = 2.5×10⁻⁶ m), Δρ = 100 kg/m³ (oil lighter), μ = 0.10 Pa·s
v = 2 × (2.5×10⁻⁶)² × 100 × 9.81 / (9 × 0.10) ≈ 1.4×10⁻⁸ m/s ≈ 1.2 mm/day
Over a 7-day hold: ≈ 8 mm of creaming.  If droplets flocculate to effective r = 5 µm: ×4 → ≈ 3 cm.

The r² dependence is the important lesson. Flocculation, Ostwald ripening, and coalescence during hold all raise r, so creaming accelerates over time. A lotion that looks fine at T₀ can stratify by day 5. This is why lotion hold studies need top/middle/bottom sampling and why lotions often need mixing during hold and transfer.

Creams and gels with a real yield stress behave differently. A particle remains suspended if the stress it exerts on the surrounding matrix is below τ₀. A commonly used dimensionless criterion is:

Y = τ₀ / (Δρ · g · d)          particles held if Y exceeds a critical value (literature ~0.05–0.1; verify for your system)

A well-structured cream can hold a suspended API indefinitely at rest, which is why creams usually have fewer settling problems than lotions. The catch is that pumping and filling shear the matrix. While it is sheared, τ₀ effectively drops and settling can occur in hoppers, hose low points, and filler nozzles. The 7/94 guide's suspension expectations follow directly: continuous mixing during filling, a tank-heel plan, and demonstrated uniformity across the fill run. The required sampling has two dimensions:

  • Beginning/middle/end of the fill run: drift across time.
  • Top/middle/bottom within a container: per USP <3> uniformity in containers (verify current chapter text).

The tank heel needs explicit treatment because the last 2–5% of a vessel often behaves unlike the rest. Settled particles concentrate there, air-rich foam may be on top, the scrapers can no longer reach, and the filler may draw differently. The MBR states whether the heel is filled, discarded, or tested, and the yield calculation accounts for it.

8. Hold time as a validated state, not an idle period

Hold time is a process step whose parameter is time, and §211.111 makes time limits a cGMP expectation where quality could be affected (verify wording). A bulk cream waiting 72 h in a vessel before filling is still being processed, by chemistry, physics, and microbiology. A hold-time study answers: for how long, in what container, at what temperature, with what headspace and closure, does the bulk remain equivalent to T₀?

Worst-case design. Test the longest hold you intend to claim, at the warmest credible temperature. In Sarasota that means checking whether the hold area is truly controlled. Use the container with the worst surface-to-volume and headspace ratio, typically a partially emptied tote with more air contact. Sample at T₀ and planned intervals up to the maximum. You may claim only what you have tested.

Three arms:

  1. Chemical: assay, related substances, preservative and antioxidant content by HPLC-DAD. These can run on the Agilent 1260s, reusing PDDM912 sample prep (solvent dispersion, heat, centrifugation). Antioxidant depletion is often the earliest signal, before the API moves.
  2. Physical: appearance, phase separation, viscosity, pH (<791> at 25 ± 2 °C), specific gravity, and microscopy. Most of these are contracted until the lab gaps close.
  3. Microbiological (aqueous bulks): TAMC/TYMC per <61> against the cutaneous-use criteria cited in <1111> (TAMC 10² and TYMC 10¹ cfu/g per course materials; verify), specified organisms per <62>, and Burkholderia cepacia complex per <60>. Use a contract lab until the PDDM913 own-vs-outsource decision. BacT/ALERT cannot do this, because it reports presence or absence and cannot count.

For anhydrous bulks (hydrocarbon or PEG ointments), a water-activity rationale under <1112> can justify a reduced micro arm, because organisms cannot proliferate at sufficiently low aw. The rationale requires a measured or well-supported aw value, which means an instrument or contract measurement. Note one subtlety for PEG ointments: PEGs are hygroscopic, so aw can drift upward during an open or humid hold. That is a Sarasota-specific risk to test, not assume away.

Why "it was heated to 75 °C" is not a micro argument. Heating a phase reduces some vegetative organisms but does not sterilize. Spore-formers survive, and contamination introduced after cooling (late additions, sampling, transfer, Purified Water of marginal quality) is untouched by the earlier heat. The Strukmyer 2025 warning letter linked water-system failures to Bcc. In aqueous bulks, the hold period is when an initial low-level contamination has the time to grow.

9. Change language: SUPAC-SS as a borrowed grammar

SUPAC-SS (FDA, May 1997) was written for NDA and ANDA products. For an OTC monograph product it has no legal force, but it provides a carefully reasoned vocabulary for describing how big a change is and what evidence would show the product was unaffected. You adopt it as a scientific reference in your own change-control system. Its key distinctions:

  • Equipment of the same design and operating principle: scale-up within the guidance's limit (up to 10× the biobatch per course materials; verify) is treated as Level 1, with minimal comparative data.
  • Equipment of a different design or operating principle: e.g., high-shear to low-shear mixer, or a different homogenizer type. This is a Level 2 change and requires comparative in-vitro release testing (IVRT).

The IVRT comparison logic (developed fully in PDDM924) works as follows:

Stage 1: 6 cells test + 6 cells reference → release-rate slopes (amount vs √t)
         compute 90% CI of the ratio T/R (nonparametric, per SUPAC-SS procedure)
Pass if 90% CI ⊂ [75%, 133.33%]
Stage 2 (if stage 1 fails): 12 more cells each (18 total per product); recompute

The interval is asymmetric, 75–133.33% rather than 75–125%, because it is symmetric on the log scale (0.75 = 1/1.3333). The biobatch (≥100 kg or 10% of production, whichever larger; verify) is an NDA construct tied to a clinical or bioequivalence batch. A monograph product has none, so you must define a reference batch: a well-characterized batch, made at known scale with recorded CPPs, against which future changes are compared. Without a defined reference, "comparative" data have nothing to compare to. The lab has no Franz cells, so IVRT is contract work. Voluntary adoption for monograph products is justified when a change plausibly alters microstructure, such as a change of mixer principle or a new emulsifier grade.

The change-control template is linear: proposed change → SUPAC-style level → required comparative data (CQAs, stability commitment, IVRT if warranted) → QU approval. The scale-up rules from §3 and the cooling-profile logic from §4 supply the scientific rationale section.

10. Process validation as a lifecycle and a statistical argument

FDA's 2011 guidance (Process Validation: General Principles and Practices; verify current status) defines validation as:

"the collection and evaluation of data, from the process design stage through commercial production, which establishes scientific evidence that a process is capable of consistently delivering quality product." — FDA, 2011 (verify wording)

It has three stages:

  • Stage 1, Process Design: bench DoE, CPP identification, scale-down models, pilot data.
  • Stage 2, Process Qualification: facility/equipment qualification plus PPQ batches under a protocol.
  • Stage 3, Continued Process Verification: ongoing trending of CPPs and CQAs, with triggers.

Validation, then, is an evidentiary argument that runs for the life of the product, not an event [Agalloco & Carleton, eds., 3rd ed.].

There is no universal batch count. The guidance does not mandate three batches. The number and scale of PPQ batches must be justified from risk, Stage 1 knowledge, and family similarity. A well-characterized O/W cream platform with a broad DoE and pilot history may justify fewer than an unfamiliar suspension whose uniformity has never been tested at scale. Bracketing across family members (e.g., strengths or fills that share a bulk) needs explicit justification; verify the approach against current guidance and your inspectorate's expectations.

Worst case is part of the design. PPQ should include, or be bracketed by, the following:

  • maximum batch size, which gives the slowest cooling;
  • maximum hold and transfer times;
  • edge-of-envelope temperatures and cooling rates;
  • maximum filling-run duration, which allows the most settling or drift.

Running PPQ at nominal conditions only demonstrates that the center works, which you knew from the DoE.

Sampling is spatial and temporal. Bulk: top/middle/bottom and near-wall versus center, since the wall is where cooling and scraping effects concentrate. Fill run: beginning/middle/end. Within containers: per <3>. Aqueous families: micro samples.

Acceptance criteria are statistical. Meeting the specification on every sample tested is necessary but not sufficient. You want confidence that the untested units also conform. Two tools from PDQC802/831:

Two-sided tolerance interval:   x̄ ± k · s
   k from tables for chosen confidence (e.g., 95%) and coverage (e.g., 99%) and n
Process capability:             Cpk = min(USL − μ, μ − LSL) / (3σ)

Worked example. Thirty bulk and fill assay results give mean 100.2% LC and s = 0.9%. For n = 30 at 95% confidence and 99% coverage, k ≈ 3.35 (verify against tables). The interval is 100.2 ± 3.0, or 97.2–103.2%. Against a hypothetical product-specific specification of 90.0–110.0%, the tolerance interval sits comfortably inside, and Cpk = (100.2 − 90.0)/(3 × 0.9) ≈ 3.8.

Watch what the statistic is computed over. If the 30 results come from three batches with distinct batch means, pooling them underestimates between-batch variance. A variance-components view (within-batch vs. between-batch) is more honest. This reuses ANOVA thinking from PDQC802 intermediate precision.

Failures during PPQ are investigated under §211.192. A PPQ batch is not quietly discarded and replaced with "batch 4." The investigation outcome decides whether the protocol continues, is amended, or returns to Stage 1.

CPV trends CPPs (cooling time, homogenizer turnovers, deaeration vacuum) as well as CQAs. Parameters drift before products fail. Review frequency and triggers for change control or revalidation are written into the CPV plan, drawing on ICH Q10 concepts (verify specific language).

11. Batch records, yield, and reconciliation: the accounting identity

§211.188 requires the batch record to reproduce the master record and document:

  • each significant step, with dates, equipment IDs, and component lot numbers;
  • weights and measures;
  • in-process results;
  • packaging and labeling inspection;
  • yields;
  • signatures of the person performing and the person checking or supervising each step.

§211.101 adds dispensing control: second-person verification or a validated automated system, and only components released under §211.84(d), meaning an identity test on every lot. For high-risk components (glycerin, propylene glycol, sorbitol), this includes DEG/EG testing to NMT 0.10% per the FDA high-risk-components guidance (May 2023 Rev. 1; verify title).

Yield is a mass balance, the same conceptual move as peptide-content accounting in PDQC812:

Theoretical yield = Σ dispensed component masses (adjusted for any validated water-loss q.s.)
Expected actual   = Theoretical − (vessel hold-up + loop/hose hold-up + heel + sampling + known transfer losses)
% yield           = Actual / Theoretical × 100

Worked calculation.

Theoretical 200.0 kg
Known losses: vessel walls/scrapers 3.2 kg, loop + hoses 1.1 kg, IPC/PV samples 0.4 kg  → 4.7 kg
Expected actual ≈ 195.3 kg (97.65%)
Observed actual 194.8 kg → 97.4%  → within a historically justified limit (e.g., 96.0–99.0%)

A yield above the expected value is a stronger red flag than one slightly below. Mass cannot appear from nowhere, so a high yield suggests a dispensing error or uncontrolled water addition. Hot O/W phases lose water by evaporation, and many processes add "water to compensate for evaporation, q.s." That compensation is acceptable only if the MBR defines it, with a method (weigh the batch, add to target mass) and a limit. Otherwise it is uncontrolled reprocessing. The same principle governs IPC failures. Pre-approved adjust-and-retest rules in the MBR (e.g., a pH adjustment within a defined amount of a defined neutralizer) are part of the validated process. Anything outside them is a deviation under §211.192, and reprocessing requires its own approved procedure (§211.115; verify).

Recording-chart and EBR data (cooling curves, homogenizer speeds) are electronic records. Their audit trails, exports, and review follow PDQC814/801 data-integrity controls. A cooling chart attached as a printout without review of the underlying record does not count as reviewed data.

12. Production operations: capacity is set by the slowest resource, including cleaning

An equipment-occupancy model treats each batch as a sequence of steps that occupy resources. A plausible illustrative cycle for a 500 L O/W cream vessel:

charge + heat 1.5 h | emulsify + homogenize 0.75 h | cool 2.5 h (scale-dependent, §4)
| deaerate 0.5 h | IPC hold 1.0 h | transfer 0.75 h | clean + changeover 3.0 h
= 10.0 h vessel occupancy per batch

Several points are easy to miss:

  • Cleaning is often 25–35% of occupancy, with its duration supplied by PDDM922 cleaning validation. A shorter validated cleaning raises capacity as much as a faster process.
  • The bottleneck can move. At 1.2 kg/min, a filler needs about 14 h for a 1000 kg batch. If the vessel cycle is 10 h, the filler becomes the bottleneck. A qualified hold vessel then decouples vessel and filler, which is only possible with a hold-time study behind it.
  • Dedicated vs. campaign: a BPO line is dedicated by chemistry, since an oxidizer is incompatible with many residues and benzene control is needed. Dedication removes changeover but idles the asset between demands. Campaigning products sharing a line reduces cleaning frequency but lengthens holds and cycle time.
  • Downtime: planned maintenance, calibration, and requalification windows are scheduled losses. Unplanned downtime is a probabilistic loss and should be a model input with sensitivity, not a single number.
  • Humid-climate coupling (Sarasota): open hot kettles release water vapor that adds latent load to HVAC already stressed by ambient humidity (WHO TRS 1010 Annex 8). Scheduling several hot steps in the same room at once can push room RH out of specification. HVAC capacity is therefore a scheduling constraint.

The spreadsheet model feeds the own-vs-CMO sensitivity analysis for the PDDM925 gate. The model's assumptions (cooling time, cleaning duration, downtime rate) are exactly the parameters the rest of this course measures.

13. Scale-down models and the CMO bridge

The lab owns no semisolid manufacturing equipment, so the real development path is bench kit (if purchased) → CMO pilot → commercial. The bench kit consists of:

  • a jacketed beaker or small vacuum vessel;
  • an overhead stirrer with anchor;
  • a rotor-stator probe;
  • a circulating bath.

It is useful only to the extent it is a qualified scale-down model. To qualify it:

  1. Declare matching criteria in advance: tip speed and gap, turnovers or equivalent shear dose, the cooling profile rather than jacket setpoint, deaeration vacuum and time.
  2. Make comparative batches at bench and pilot.
  3. Show that the CQAs (viscosity at defined shear, globule size distribution, specific gravity, pH, assay and preservative content) agree within pre-set equivalence margins.

Where they disagree, as with cooling (the bench cools too fast), you deliberately slow the bench to mimic scale. This is a counterintuitive point: a good scale-down model sometimes reproduces the large vessel's limitations on purpose.

The CMO pilot is the bridge. The technical-transfer package states CPPs in scale-independent form, specifies IPCs and acceptance, and defines hold points. A person-in-plant plan puts your scientist at the CMO during critical steps. The quality agreement allocates MBR authorship, IPC execution, and deviation handling. The facility's QU cannot delegate its cGMP responsibility (§211.22; PDDM914 cross-reference). You may outsource the work, but you remain responsible for it.

14. Connections to Peptide Chemistry

This course is not about peptides, yet nearly every idea maps onto something you already understand from the BCHM and PDQC programs.

  • "The process is the product" came from biologics. In BCHM616 and PDQC815 you saw that aggregation state, glycoform distribution, and higher-order structure depend on process history. Q3 microstructure in a cream is the semisolid version of higher-order structure: an attribute the formula alone does not determine, measured by orthogonal physical methods.
  • Resin swelling ↔ polymer hydration. In SPPS (BCHM602), resin swelling in DMF/NMP sets reagent access and coupling kinetics, and poorly swollen resin gives deletion sequences. Carbomer dispersion is the same physics: an under-swollen or fish-eyed polymer gives lumps and viscosity deficits. In both cases solvation is a rate-limiting process step, not a formula attribute.
  • Exotherms and heat transfer at scale. Scaling an SPPS activation or a large cleavage cocktail runs into the same A/V ∝ 1/D problem, with heat removal failing at scale. The cooling-time calculation in §4 is the same lumped model you would use for an exothermic coupling at kilogram scale.
  • Shear and interfaces. PDQC815 showed that peptides and proteins aggregate at air-water interfaces and under shear. If a future formula contains a peptide active (topical cosmetic and OTC peptides exist), the homogenizer, the deaeration step, and the transfer pump become chemical-stability hazards as well as physical ones. Air entrainment would raise a peptide-specific aggregation risk on top of the oxidation risk in §6. Methionine oxidation, Asn deamidation, and Asp isomerization are temperature-, pH- and oxygen-dependent. All of them would make the late-addition temperature, pH, and blanketing CPPs in a principled way. (This stays API-agnostic: these are the hooks to activate if classification brings a peptide into scope.)
  • Partitioning. Paraben partitioning into oil and micelles (Week 2) is the same equilibrium thinking as peptide retention on reversed-phase media or peptide binding to surfactant micelles. Preservative free aqueous concentration, not total, drives antimicrobial efficacy. A peptide's free fraction similarly drives what you measure and what is active.
  • Mass balance. Peptide content accounting in PDQC812 (peptide + water + counter-ion + impurities ≈ 100%) and batch yield reconciliation are the same accounting identity. Unexplained excess is as suspicious as unexplained loss.
  • Hold times. In peptide manufacture you held cleaved crude, purified pools, and pre-lyophilization solutions, each with a validated hold. The three-arm hold study (chemical/physical/micro) generalizes what you did for peptide intermediates, with the physical arm grown larger because the product is a structure.
  • Analytics carry over directly. The three Agilent 1260 HPLC-DADs, the stability-indicating method logic of PDQC803, and the validation statistics of PDQC802 cover preservative and antioxidant assay, related substances, and in-process uniformity. The new work is sample prep from a semisolid matrix (PDDM912) and the physical IPCs the lab does not yet own.

15. Common misconceptions to watch out for

  1. "Process validation means three batches." No regulation or the 2011 guidance requires three. The batch count is a justified, risk-based conclusion from Stage 1 knowledge.
  2. "Keep tip speed constant and the scale-up is done." Tip speed may govern droplet breakup at the head. It does not govern bulk turnover, heat transfer, or shear dose in an in-line loop. Each function has its own invariant, and comparative CQAs at each scale are still required.
  3. "More homogenization is always better." The Hinze exponent gives diminishing returns, and over-shearing heats the batch, entrains air, breaks gel networks, and degrades polymers. The envelope has an upper boundary.
  4. "Same recipe and same setpoints gives the same product at scale." Cooling time grows roughly in proportion to vessel diameter. Recipes must reproduce the temperature–time profile and its effects, not the jacket setpoint.
  5. "Viscosity at the IPC equals release viscosity." Structure keeps building, often for 24–48 h, and shear damage from transfer may be irreversible. IPC and release limits are distinct and both need data.
  6. "Air doesn't matter because assay is w/w." Air does not dilute a weighed w/w assay. It does reduce fill weight with volumetric fillers, bias volumetric samples, stratify, and supply oxygen for degradation. Specific gravity is the IPC that catches it.
  7. "The batch was heated to 75 °C, so hold-time micro is unnecessary." Heat does not sterilize, and post-cooling additions, sampling, transfer, and water quality all introduce organisms that the hold period lets grow. Aqueous bulks need a micro arm, by plate counts at a contract lab, since BacT/ALERT cannot enumerate.
  8. "Anhydrous means no micro risk, permanently." The low-aw rationale (<1112>) holds only while the bulk stays anhydrous. Hygroscopic PEG bases in a humid climate can pick up water during holds.
  9. "SUPAC-SS applies to our monograph products." Legally it does not; it is an NDA/ANDA guidance. We adopt its change levels and IVRT logic voluntarily as a scientific reference, and we define our own reference batch because there is no biobatch.
  10. "Cosmetic products on the same line follow cosmetic rules." On shared equipment, drug cGMP governs cleaning, maintenance, and qualification of the asset.
  11. "One temperature probe tells you the batch temperature." Viscous batches carry wall-to-core gradients. Temperature mapping in OQ/PQ establishes whether a single control probe is representative and where it must sit.
  12. "Adding water to hit target weight, or tweaking pH until it passes, is routine practice." Only adjustments pre-approved and bounded in the MBR are part of the validated process. Anything else is a deviation (§211.192) and possibly reprocessing.
  13. "A yield above 100% is good news." It indicates a dispensing error, unaccounted water, or a recording error, and must be investigated.
  14. "The bench kit is representative because it's the same formula." A scale-down model is representative only after comparative batches against declared matching criteria have shown it.
  15. "Validation freezes the process." Stage 3 CPV is designed to detect drift and drive change control. A validated process is one under continuing evidence.

The idea to keep from this chapter: for semisolids, the formula specifies the ingredients and the process determines what the product becomes. Every deliverable in this course (MBRs, URSs, DoE, hold studies, PV protocols, capacity models) is a way of writing that process down precisely enough that it happens the same way every time, at every scale, and of proving that it did.

⚙️ Mechanisms & Applications

GPT-6 Astra

Mechanisms & Applications — Process Engineering, Scale-Up, Hold Times, Production Operations, and Process Validation

My central insight: manufacturing a semisolid is not simply mixing ingredients until the assay is correct. It is constructing a physical architecture—and then preserving that architecture through cooling, storage, pumping, filling, and shelf life. Two creams can contain the same ingredients at the same concentrations yet differ in droplet size, crystal population, rheology, preservative availability, and drug release. The process is part of the product.

This extends what I learned in protein science. A protein’s amino-acid composition does not establish its folded state; a cream’s composition does not establish its microstructure. The analogy has limits: most emulsions are kinetically stabilized, nonequilibrium systems rather than structures approaching a unique equilibrium state. Nevertheless, both require control of history, not merely composition. [1–5]

Scope of these notes: numerical examples below are engineering illustrations, not approved manufacturing instructions or universal specifications. Actual operating ranges, sampling plans, and acceptance criteria must be established for the classified formulation, equipment, analytical methods, and intended packaging. Peptide Partners currently lacks semisolid manufacturing equipment and several essential physical-testing capabilities; bench execution is conditional on acquisition and qualification, otherwise it belongs with a qualified CMO.

1. Process Architecture: Turning a Formula into a Sequence of Controlled State Changes

A useful process map describes what changes physically at each step, why that change matters, and what evidence permits the next step. “Mix for 20 minutes” is an instruction. “Disperse the polymer until wetting and hydration are adequate, without persistent agglomerates” describes the engineering purpose. A defensible master batch record needs both.

Released components + qualified utilities
    ↓ Dispense and independently verify
Correct identities, quantities, and traceability
    ↓ Prepare phases or disperse solids
Ingredients dissolved, melted, wetted, or hydrated as intended
    ↓ Combine phases / homogenize
Droplet and particle populations established
    ↓ Controlled cooling / neutralization / late additions
Final continuous-phase structure develops
    ↓ Deaerate
Entrained gas reduced without unacceptable volatile loss
    ↓ Bulk hold and in-process testing
Chemical, physical, and microbiological condition remains acceptable
    ↓ Transfer and fill
Structure and composition preserved across the filling run
    ↓ Quality-unit review
Release decision based on the complete manufacturing evidence

The family assignment changes the mechanism. An O/W cream requires formation and stabilization of oil droplets in an aqueous continuous phase. A carbomer gel requires wetting, hydration, and controlled neutralization. A hydrocarbon ointment may require melt blending followed by controlled crystallization of its structuring components. Forcing all three into one generic “heat–homogenize–cool” recipe hides the most important risks.

  • O/W cream: phase preparation → controlled combination → droplet-size development → cooling-dependent network formation.
  • Hydrocarbon ointment: fusion or mechanical blending → incorporation of dissolved or suspended materials → controlled congealing. A suitable levigating liquid must be compatible with the intended phase; glycerin is not a universal choice for a hydrocarbon base.
  • PEG ointment: melt blending or appropriate dissolution → cooling and structure formation. Hygroscopicity and water content can materially affect properties.
  • Gel: powder wetting → hydration → neutralization or other gelation mechanism → deaeration.
  • Lotion or suspension: dispersion formation → protection against creaming or sedimentation during holding and filling.

Regulatory connection: 21 CFR 211.100(a) requires written production and process-control procedures designed to assure the drug’s identity, strength, quality, and purity, with appropriate review and quality-unit approval. Section 211.100(b) addresses following those procedures, documenting execution, and recording and justifying deviations. Section 211.110 requires appropriate in-process controls, including monitoring processes that may cause variability. Section 211.188 requires the executed batch record. These are related obligations, not interchangeable citations. [6]

A second important distinction is between a candidate CPP and an established CPP. A parameter becomes critical because its variability can affect a critical quality attribute—not because the equipment has a display for it. Conversely, an unmeasured cooling rate can be critical even when the batch record records only the final temperature.

2. Phase Preparation and Emulsification: Creating Interface Faster Than It Disappears

2.1 Why phase preparation comes before high shear

  1. Establish the intended location of each ingredient. Solubility, ionization, thermal stability, and partitioning determine whether an ingredient enters the aqueous phase, oil phase, or a later addition.
  2. Complete melting, dissolution, or wetting. Undissolved preservative, unmelted wax, and poorly wetted polymer are different defects and need different checks.
  3. Bring phases to justified combining conditions. Some wax-containing systems use temperatures around 70–75 °C, but this is a formulation-dependent example—not a default. The temperature must accommodate melting requirements without unacceptable degradation or volatile loss.
  4. Combine at a controlled rate and location. Addition must not locally overwhelm the mixing capacity or create regions with unintended phase ratios.
  5. Apply sufficient dispersive stress. Droplets must deform and break while emulsifier adsorption limits recoalescence.

“Add the dispersed phase to the continuous phase” is often useful, but not a universal law. Some processes deliberately pass through an inversion route. Accidental inversion can occur when phase volume, temperature-dependent surfactant behavior, or mixing conditions change which phase is continuous. The resulting product may look superficially acceptable while possessing a fundamentally different structure. [1–3,5]

2.2 Droplet breakup is a competition between stress and interfacial restoration

Interfacial tension resists creation of additional interface. Flow supplies deforming stress. For approximately spherical droplets, reducing diameter increases interfacial area enormously:

A ≈ 6φV / d

A = total droplet interfacial area
φ = dispersed-phase volume fraction
V = total emulsion volume
d = representative droplet diameter

Example:
V = 1 L = 0.001 m³
φ = 0.20

At d = 10 μm: A ≈ 120 m²
At d =  2 μm: A ≈ 600 m²

The surprising consequence: making droplets five times smaller creates approximately five times more interface. If emulsifier adsorption is too slow or coverage becomes inadequate, additional homogenization can promote recoalescence rather than produce a permanently finer emulsion.

Two dimensionless groups help organize the physics:

Capillary number:
Ca = μc γ̇ R / σ

Weber number:
We = ρc u² d / σ

μc = continuous-phase viscosity
γ̇ = characteristic shear rate
R = droplet radius
σ = interfacial tension
ρc = continuous-phase density
u = characteristic relative or fluctuating velocity

The capillary number compares viscous deformation with interfacial restoration; the Weber number compares inertial stress with interfacial restoration. Breakup thresholds depend on viscosity ratio, flow type, surfactant behavior, and exposure time. Neither equation supplies a universal manufacturing setpoint. [2]

Experimental application: use microscopy or a qualified particle-size method to measure the result of homogenization, rather than treating motor speed as proof of success. Microscopy methods need defined sample preparation, calibration, image-selection rules, and size metrics. Dilution, compression under a coverslip, or warming during preparation can alter the structure being measured.

At Peptide Partners, the microscope is an acquisition or outsourcing gap. An HPLC chromatogram cannot reveal a coarse-droplet tail, undispersed crystals, or incomplete polymer wetting.

2.3 Preservatives make phase behavior a microbiological issue

As in membrane partitioning from biochemistry, total concentration and concentration in the relevant environment are not equivalent. A preservative may partition into oil droplets, associate with surfactant assemblies, bind to formulation components, or precipitate. Therefore, an acceptable total preservative assay does not establish adequate antimicrobial preservation.

Control implication: the addition stage, solvent or phase, mixing conditions, and cooling history must be documented. Preservation effectiveness must be established using the appropriate microbiological study; HPLC supports that assessment but cannot replace it. [3,5]

3. Cooling and Gelation: The Product Continues to Form After Homogenization

3.1 Cooling determines where molecules and crystals end up

  1. At elevated temperature, an ingredient may be dissolved or a structurant may be fully molten.
  2. Cooling changes solubility and phase equilibria.
  3. Supersaturation can develop, followed by nucleation and crystal growth.
  4. Crystals and lamellar structures may form a load-bearing network.
  5. The resulting structure determines yield behavior, apparent viscosity, spreadability, and sometimes release performance.

Two batches with the same final temperature can therefore be different products if their cooling histories differ. Fast cooling can favor more numerous nuclei; slow cooling can favor greater crystal growth, although the actual outcome depends on nucleation kinetics, polymorphism, composition, and shear. This is a close conceptual relative of controlling crystallization in protein purification or small-molecule API manufacture. [1,3]

Simplified stirred-vessel cooling model:

m cp dT/dt = −UA(T − Tj) + Pmix

m = batch mass
cp = specific heat capacity
U = overall heat-transfer coefficient
A = effective heat-transfer area
Tj = jacket temperature
Pmix = mechanical power dissipated as heat

If Tj is constant, Pmix is negligible, and the batch is well mixed:

T − Tj = (T0 − Tj) exp(−t/τ)

τ = m cp / UA

This model excludes crystallization heat, variable viscosity, changing heat-transfer coefficients, and internal temperature gradients. Those exclusions explain why a larger kettle cannot be qualified by checking only whether it eventually reaches the target temperature.

Experimental technique: temperature mapping uses multiple calibrated probes to detect warm cores, cold walls, stagnant regions, and lag during cooling. Qualification should address meaningful operating conditions; a water-only test does not establish mixing and heat-transfer performance for a viscous cream. The FDA’s 1994 topical-products inspection guide emphasizes temperature control, mixing, and the consequences of inadequate equipment design. [7]

3.2 Carbomer gels: neutralization is a structural operation

Carbomers are crosslinked poly(acrylic acid) materials. Neutralization increases carboxylate ionization, changing electrostatic interactions and osmotic swelling. The swollen microgel particles interact to create macroscopic yield behavior and shear thinning.

Poorly controlled powder addition
    → rapidly hydrated outer layer around a dry core
    → persistent agglomerates ("fish-eyes")
    → nonuniform hydration and misleading viscosity

Controlled dispersion and hydration
    → appropriately timed, distributed neutralizer addition
    → swelling and network development
    → defined equilibration
    → pH and rheology measured under standardized conditions

More neutralizer is not a general remedy for low viscosity. The underlying cause could be insufficient hydration, electrolyte screening, polymer damage, an incompatible ingredient, or an incorrect measurement history. Islam and colleagues demonstrated the importance of pH-dependent characterization in particular carbomer systems; their reported behavior should not be extrapolated into a universal pH 5–8 invariance rule for all grades and formulations. [8]

3.3 Rheology measures structure through its response to force

Herschel–Bulkley model:

τ = τy + Kγ̇ⁿ

τ  = shear stress
τy = fitted yield stress
K  = consistency index
n  = flow index

For shear-thinning behavior: n < 1

Apparent viscosity:
ηapp = τ / γ̇

A statement such as “viscosity = 30,000 mPa·s” is incomplete without geometry or spindle, speed or shear rate, temperature, preparation, rest time, and reading time. Thixotropic materials also depend on prior shear exposure. Wall slip can make a structured sample appear easier to flow than it really is.

Application: a rotation-based QC method can monitor consistency if it is appropriately qualified. A rheometer provides more mechanistic information: flow curves, oscillatory storage and loss moduli, and controlled recovery after shear. These tools help predict whether a product can be pumped, whether it recovers after filling, and whether suspended material remains supported. USP <912> provides the relevant compendial framework for rotational viscosity measurements; current official procedures must be checked when developing the method.

4. Scale-Up: Why Matching RPM Is Usually the Wrong Question

The engineering question is not “What speed did the bench mixer use?” It is “Which physical mechanisms controlled product formation, and how will their relative rates change at larger scale?”

Impeller or rotor tip speed:
utip = πDN

Mixing Reynolds number, for a Newtonian fluid:
Re = ρND² / μ

Power number:
Np = P / (ρN³D⁵)

Specific mechanical energy:
E/V = ∫P dt / V

D = characteristic mixer diameter
N = rotational frequency in s⁻¹, not rpm
P = power delivered to the fluid

For non-Newtonian fluids, an appropriate effective viscosity and mixer-specific characterization are needed. Motor nameplate power is not the same as power delivered to the batch, and average energy per volume does not describe local rotor–stator stress or the distribution of exposures. [1,2]

4.1 Worked example: constant tip speed does not preserve power per volume

Bench rotor:
D1 = 0.10 m
N1 = 3,000 rpm = 50 s⁻¹

utip,1 = π × 0.10 × 50 = 15.7 m/s

Pilot rotor:
D2 = 0.30 m

To preserve tip speed:
N2 = 15.7 / (π × 0.30)
   = 16.7 s⁻¹
   ≈ 1,000 rpm

For geometrically similar turbulent mixing systems with approximately constant power number:

P/V ∝ N³D²

At constant tip speed:
N ∝ D⁻¹

Therefore:
P/V ∝ D⁻¹

Tripling D gives approximately one-third the power per volume.

This is a useful warning, not an exact prediction for every rotor–stator installation. Rotor–stator gap, stator geometry, pumping capacity, viscosity, and recirculation arrangement can break the simple similarity assumptions.

The practical lesson: constant tip speed, constant power per volume, constant energy per volume, and constant processing time generally cannot all be maintained simultaneously. Choose a scale-up hypothesis based on the mechanism, then test the resulting CQAs.

4.2 Heat transfer becomes relatively weaker as equipment grows

For geometrically similar vessels with characteristic size L:

Volume ∝ L³
Heat-transfer area ∝ L²
Area/volume ∝ 1/L

If U and fluid properties remain comparable:
Cooling time constant τ ∝ L

A threefold increase in vessel dimensions can therefore imply roughly threefold longer cooling time under these simplified assumptions. Longer time in a crystallization window may alter particle size or vehicle structure even when the emulsification stage scales successfully.

Scale-up dossier: document the selected similarity criteria, criteria knowingly not preserved, expected risks, comparative analytical results, and conditions requiring investigation. SUPAC-SS provides useful change-control language, but its postapproval framework was written for applicable NDA/ANDA products—not as a direct filing framework for OTC monograph products. Its equipment categories, batch-size provisions, and IVRT comparison rules must be consulted in the original guidance before formal use. [9]

5. Deaeration, Transfer, and Filling: Preserving What the Kettle Created

5.1 Air is a hidden dispersed phase

Entrained gas changes bulk density, volumetric filling behavior, pump performance, appearance, and potentially oxidative exposure. Vacuum expands bubbles and promotes their removal, while gentle mixing renews the exposed surface. Excessive vacuum or agitation can instead cause foaming or loss of volatile ingredients.

Approximate density relation, neglecting gas mass:

ρbulk ≈ (1 − ε)ρair-free

ε = entrained-gas volume fraction

If ρair-free = 1.00 g/mL and ε = 0.04:
ρbulk ≈ 0.96 g/mL

A nominal 50 mL volumetric fill then contains:
m ≈ 48 g, rather than 50 g.

Important correction: air alone does not change the true API mass fraction in a homogeneous formulation. It can change mass delivered by a volumetric filler and bias volumetric sampling or sample preparation.

A density cup or pycnometer method needs controlled temperature, defined filling and leveling, and a procedure that does not unpredictably remove bubbles. Specific gravity is a dimensionless density ratio at stated temperatures; it is not simply another name for density.

5.2 Pumping is an additional processing step

  • A pump can disrupt a weak gel network or alter droplet and particle populations.
  • A restrictive valve can create greater local stress than the nominal pump setting suggests.
  • Recirculation can repeatedly process some material while other regions remain poorly exchanged.
  • Hoses and tank heels can retain a compositionally different fraction.
  • Uncontrolled hopper heating can continue crystal growth or accelerate degradation.

Transfer-study mechanism: compare matched samples before transfer, after transfer, and across the fill run, controlling sample age and temperature. Include a fixed recovery period for rheology. A lower immediate viscosity followed by reproducible recovery is different from a persistent structural change.

For dilute Newtonian suspensions, Stokes’ law illustrates why particle size matters:

Settling velocity:
v = Δρ g d² / (18μ)

Doubling particle diameter increases predicted settling velocity fourfold.

Concentrated, flocculated, and yield-stress systems may violate the assumptions, but the equation explains why a small coarse-particle fraction can dominate segregation. Mixing during filling must be sufficient to preserve uniformity without causing air entrainment or structural damage.

6. Hold-Time Studies: Stability Under Manufacturing Conditions

A bulk hold is a period during which the product continues to evolve. It differs from packaged stability because the vessel, headspace, surface-to-volume ratio, closure, agitation, and exposure history differ. A passing preservative-effectiveness study does not justify an unlimited manufacturing hold.

6.1 Build the study around the actual manufacturing clock

  1. Define time zero. For example, completion of final compounding. If an earlier aqueous phase is held, it needs its own justified limit.
  2. Specify the held state. Vessel or tote, fill level, headspace, closure, temperature, light exposure, and agitation state.
  3. Identify credible worst cases. These may include maximum time, a warm condition, minimal agitation, a low fill level, or an interrupted transfer.
  4. Use planned time points. An illustrative 24-hour hold study might sample at 0, 8, 16, and 24 hours, with a justified later point to characterize margin.
  5. Include spatial and filling-run sampling. Top/middle/bottom bulk locations and beginning/middle/end fill samples answer different questions.
  6. Evaluate chemical, physical, and microbiological results together. Passing assay does not rescue failed physical uniformity or unacceptable microbiology.
Chemical arm:
Assay + relevant degradants + preservative/antioxidant content

Physical arm:
Appearance + pH + controlled rheology + density/specific gravity
+ droplet/particle observations

Microbiological arm:
Suitable enumeration + specified/objectionable organism assessment
+ risk-based additional testing

USP <61>, <62>, <60>, <1111>, and <1112> provide relevant frameworks. Common cutaneous-product criteria include TAMC 10² and TYMC 10¹ CFU/g or mL, subject to the applicable product requirements and USP interpretation of these numerical limits. Bcc testing is especially relevant to susceptible aqueous products but must be established through product and process risk assessment rather than assumed to be an identical requirement for every aqueous bulk.

Local laboratory consequence: BacT/ALERT provides presence/absence information in its qualified application. It does not supply compendial enumeration or preservative-challenge log reductions. Appropriate contract microbiology remains necessary.

6.2 Water loss can mimic a potency increase

Initial bulk:
500 kg containing 5 kg API
API concentration = 1.000% w/w

After loss of 10 kg water:
Remaining mass = 490 kg
API concentration = 5/490 × 100 = 1.0204% w/w

Relative concentration increase ≈ 2.04%

The API has not increased; the denominator has decreased. This connects directly to the reporting-basis discipline learned in peptide QC. Assay trends, density, vessel mass, closure integrity, and water-loss evidence should be interpreted together.

Hold limits must cover the sequence actually used: bulk storage, transfer interruptions, hopper residence, and the last unit filled. Moving material to another container does not reset its age.

7. Experimental Design: Separating Causation from Convenient Stories

7.1 A bench DoE should test the mechanism

For a cream, candidate factors might be homogenizer speed, homogenization time, phase-combination temperature, and phase-addition rate. Responses should include droplet-size distribution, controlled rheology, air content or density, uniformity, appearance, and appropriate short-term physical stability observations—not viscosity alone.

Illustrative coded-factor model:

y = β0 + β1x1 + β2x2 + β3x3
    + β12x1x2 + β13x1x3 + β23x2x3
    + ε

An interaction means:
The effect of one factor depends on the level of another.

For example, longer homogenization may reduce droplet size at one temperature but produce little benefit at another because viscosity, interfacial behavior, or recoalescence kinetics differ. One-factor-at-a-time trials can miss that relationship.

  • Randomize when feasible to prevent time trends from masquerading as factor effects.
  • Block by meaningful sources such as day or raw-material lot.
  • Replicate independent batches. Repeated instrument readings are not independent process replicates.
  • Include center points when appropriate to detect curvature.
  • Confirm predictions with independent runs, including challenging conditions.

A bench-supported operating range is not automatically an ICH Q8(R2) design space. Its relevance must survive equipment, scale, raw-material, and measurement variation. ICH Q9(R1) supplies the risk-management framework; ICH Q10 places that knowledge within the pharmaceutical quality system.

7.2 Measurement methods must not erase the defect

HPLC sample preparation often intentionally dissolves the formulation. That is appropriate for total assay, but it destroys evidence about the original API location, particle size, and phase distribution. Similarly, warming a sample to make it easier to handle may remove a crystal population or alter rheology.

The existing Agilent 1260 HPLC-DAD instruments can support suitable assay and impurity procedures, but only after formulation-specific extraction recovery, specificity, and sample stability are established. HPLC-DAD is not universally suitable for every unknown active or degradant. Physical characterization therefore complements chemical analysis; it is not redundant testing.

8. Process Validation: Establishing Control Rather Than Celebrating Three Passing Batches

FDA’s Process Validation: General Principles and Practices describes a lifecycle: process design, process qualification, and continued process verification. It does not impose a universal three-batch rule. The number of PPQ batches and sampling intensity must be scientifically justified. [10]

  1. Process design: identify the important mechanisms, material attributes, parameter ranges, and measurement limitations.
  2. Facility, utility, and equipment qualification: establish that systems perform their intended functions. Equipment qualification alone does not validate the product process.
  3. Process performance qualification: demonstrate reproducible performance using the commercial process, trained personnel, approved procedures, and a justified sampling plan.
  4. Continued process verification: monitor whether the process remains in control as materials, equipment condition, operators, and production patterns change.

Worst-case selection must be mechanistic. Maximum batch size may be worst for cooling; minimum fill may be worst for mixer immersion. Maximum viscosity may challenge transfer, while minimum viscosity may worsen sedimentation. A single “largest batch” is not automatically the worst case for everything.

Placebo work is useful for training, basic equipment characterization, and vehicle-process development. It does not establish API distribution, degradation control, active-containing microbiological behavior, or validation of the final drug-product process.

8.1 Sampling should reveal variability rather than average it away

Conceptual nested model:

yijk = μ + Bi + Lj(i) + εijk

Bi    = between-batch contribution
Lj(i) = location or filling-position contribution within batch
εijk  = analytical and residual contribution

A composite sample can conceal a high top concentration and a low bottom concentration. A batch mean can conceal the last part of a filling run drifting out of control. Preserve individual results where uniformity is the question.

A confidence interval for the mean describes uncertainty about the mean. A tolerance interval addresses a specified fraction of the population with specified confidence. They answer different validation questions, and their use requires defensible distributional and independence assumptions.

Preapprove acceptance criteria and investigation rules. Do not widen criteria after observing results or repeatedly remix and retest until the batch passes. Sections 211.110 and 211.192 connect in-process control, batch review, and investigation of discrepancies. [6]

8.2 Continued verification should detect drift before specification failure

  • Trend cooling time, phase-addition rate, homogenization exposure, vacuum behavior, and transfer duration.
  • Trend pH, density, rheology, particle/droplet metrics, assay uniformity, microbiological findings, and yield.
  • Investigate changes in variability as well as shifts in the average.
  • Define triggers for maintenance, revised sampling, change control, and additional qualification.

A control limit describes observed process behavior; a specification describes an acceptance requirement. A stable process can be incapable, and an unstable process can temporarily produce passing batches. This is why continued verification cannot be reduced to counting out-of-specification results.

9. Production Operations: Capacity Is Governed by Occupancy and Queues

A kettle is unavailable while it contains acceptable bulk awaiting testing, even when its agitator is off. Outsourced testing can therefore determine production capacity as strongly as mixer size.

Equipment occupancy per batch =
Charging + heating + combining + homogenizing + cooling
+ deaeration + IPC/hold + transfer + cleaning + changeover

Illustrative kettle occupancy:
1.0 + 1.5 + 0.5 + 0.5 + 2.5 + 0.5 + 2.0 + 1.0 + 2.0
= 11.5 h/batch

Scheduled time = 80 h/week
Illustrative usable-time factor = 0.85

Usable time = 68 h/week

Simple steady-state estimate:
68 / 11.5 ≈ 5.9 batch-equivalents/week

This is not a finished schedule. Staffing, batch boundaries, cleaning windows, utility limits, maintenance, and filler availability may reduce completed batches. Do not apply a downtime factor that counts cleaning losses already included in occupancy.

Little’s law links average work in process, throughput, and flow time:

WIP = throughput × average flow time

If testing delays increase flow time, the same throughput requires more material tied up in vessels or totes. Additional hold vessels may relieve a kettle bottleneck—but introduce transfer steps, cleaning requirements, hold-time obligations, and contamination opportunities.

Sarasota application: heat and moisture released during processing must be evaluated against HVAC capacity. Campaigning similar products may reduce cleaning and changeover, but segregation and equipment dedication must follow incompatibility and cross-contamination assessments. A dedicated oxidizer line can be a prudent design control; it should not be described as a universal statutory requirement for every oxidizer product.

10. Equipment Qualification and the CMO Bridge

A useful user requirement is testable. “Good mixing” is not. “Demonstrate adequate mixing at the justified minimum and maximum working volumes, including defined near-wall and outlet regions” can become an OQ or product-performance test.

  • Scrapers and agitators: verify installation, clearance or contact as designed, speed accuracy, and performance under representative loading.
  • Temperature control: verify sensor calibration, control response, mapping, recording, and alarms.
  • Vacuum: verify pressure measurement, leakage performance, control, and behavior with representative foaming conditions.
  • Transfer systems: characterize flow range, pressure, retained volume, air ingress, and product effects.
  • Cleanability: assess seals, dead legs, valves, outlets, and access; carry the findings into cleaning validation.

The 1994 FDA inspection guide provides important inspection-oriented expectations, but it is not itself a regulation or a prescription for one universal emulsifier design. Each URS should trace the risk to a design feature and then to objective evidence. [7]

For Larissa and Peptide Partners: the immediate deliverable is a credible technical-transfer package, not a claim of in-house manufacturing readiness. It should contain the process-family rationale, draft master record, material requirements, proposed CPPs, sampling and IPC methods, hold strategy, transfer risks, deviation rules, and comparability plan. The quality agreement must allocate responsibilities without suggesting that either party can contract away its applicable cGMP obligations. [11]

11. Historical Context: How These Ideas Became Engineering

  • Osborne Reynolds, 1883: experiments on flow transition established the importance of a dimensionless comparison of inertial and viscous effects. The lasting contribution to scale-up is the idea that physical similarity depends on governing ratios—not on matching one dimensional setting such as rpm.
  • G. I. Taylor, early 1930s: theoretical and experimental studies connected the deformation of liquid drops with surrounding flow, viscosity ratio, and interfacial tension. These ideas explain why changing the continuous-phase viscosity changes emulsification behavior even at the same mixer speed.
  • A. N. Kolmogorov, 1941, and J. O. Hinze, 1955: turbulent-energy concepts and droplet-breakup analysis connected local fluid stresses to dispersion size. Hinze’s work is foundational to understanding why average mixing energy alone does not fully specify an emulsion process. [12]
  • Walter A. Shewhart, 1920s–1930s: statistical control charts distinguished routine process variation from signals of change. Modern continued process verification retains this central insight: understand the process generating the data, not merely whether individual results pass.
  • George E. P. Box and K. B. Wilson, 1951: response-surface methodology gave industrial researchers a systematic way to learn about interacting factors and curvature through planned experiments. This is the intellectual foundation of the bench DoE rather than trial-and-error optimization. [13]
  • FDA, 1997 and 2011: SUPAC-SS formalized an important framework for evaluating postapproval semisolid changes; the 2011 process-validation guidance articulated the modern lifecycle approach. Together they emphasize accumulated evidence rather than recipe repetition. [9,10]

What I should be able to defend: for an unfamiliar, properly classified formula, I should explain what structure must be created, which competing rates govern its formation, why scale changes those rates, how the structure will be measured, and what evidence establishes that it survives the complete manufacturing timeline. If my explanation stops at “the assay passed,” I have not yet understood the process.

Sources and Official-Text Checks

  1. Levin, M., ed. Pharmaceutical Process Scale-Up, 3rd ed. Engineering scale-up, process development, and transfer principles.
  2. Paul, E. L.; Atiemo-Obeng, V. A.; Kresta, S. M., eds. Handbook of Industrial Mixing: Science and Practice, 1st ed. Mixing similarity, power input, dispersion, and non-Newtonian processing.
  3. Lieberman, H. A.; Rieger, M. M.; Banker, G. S., eds. Pharmaceutical Dosage Forms: Disperse Systems, 2nd ed. Emulsions, suspensions, semisolid structure, and formulation–process interactions.
  4. Agalloco, J.; Carleton, F. J., eds. Validation of Pharmaceutical Processes, 3rd ed. Qualification, validation strategy, and supporting evidence.
  5. Chang, R. K.; Raw, A.; Lionberger, R.; Yu, L. “Generic Development of Topical Dermatologic Products: Formulation Development, Process Development, and Testing of Topical Dermatologic Products.” AAPS Journal (2013). doi:10.1208/s12248-012-9411-0. Also: “Generic Development of Topical Dermatologic Products, Part II: Quality by Design for Topical Semisolid Products.” doi:10.1208/s12248-013-9472-8.
  6. 21 CFR Part 211, particularly §§211.22, 211.67, 211.84, 211.100, 211.101, 211.110, 211.113(a), 211.186, 211.188, and 211.192. Consult current eCFR text when drafting controlled procedures; these notes paraphrase rather than quote the regulations.
  7. FDA. Guide to Inspections of Topical Drug Products (July 1994). Historical inspection guidance; distinguish its recommendations from binding regulatory text.
  8. Islam, M. T.; Rodríguez-Hornedo, N.; Ciotti, S.; Ackermann, C. “Rheological characterization of topical carbomer gels neutralized to different pH.” Pharmaceutical Research (2004). doi:10.1023/B:PHAM.0000033006.11619.07.
  9. FDA. SUPAC-SS: Nonsterile Semisolid Dosage Forms—Scale-Up and Postapproval Changes: Chemistry, Manufacturing, and Controls; In Vitro Release Testing and In Vivo Bioequivalence Documentation (May 1997).
  10. FDA. Process Validation: General Principles and Practices (January 2011). Lifecycle framework used here; confirm current status when approving a validation program.
  11. FDA. Contract Manufacturing Arrangements for Drugs: Quality Agreements (November 2016).
  12. Hinze, J. O. “Fundamentals of the Hydrodynamic Mechanism of Splitting in Dispersion Processes.” AIChE Journal 1 (1955): 289–295.
  13. Box, G. E. P.; Wilson, K. B. “On the Experimental Attainment of Optimum Conditions.” Journal of the Royal Statistical Society, Series B 13 (1951): 1–45.

Checks before operational use: consult current USP–NF text for <3>, <60>, <61>, <62>, <791>, <912>, <1111>, and <1112>, including applicability, methods, and numerical-limit interpretation. The course identifies a March 2026 final FDA Q3 characterization guidance; verify the official version and its exact recommendations before citing specific item numbers. ANDA-focused characterization guidance can inform the science for monograph products without automatically imposing its submission framework on them.

🪬 Skipper's Reflection

Gemini 3.1 Pro

Skipper’s Reflection: Process Engineering, Scale-Up, Hold Times, Production Operations, and Process Validation

Coming from the precise, sterile, and largely liquid-phase world of peptide quality control, my transition into semisolid process engineering has been a shock to the system. In analytical chemistry, a solution is a solution. But in the world of creams, ointments, and gels, the physical process is the product. The exact same ingredients, mixed at different speeds or cooled at different rates, yield entirely different dosage forms. PDDM921 has completely rewired how I view manufacturing.

1. What Surprised Me Most

I was most surprised by the tyranny of geometry and thermodynamics during scale-up. On the benchtop, if I want to cool a 1-liter beaker of a newly emulsified O/W cream, it takes minutes. At pilot scale (100 L) or commercial scale (1,000 L), the volume (mass to be cooled) scales cubically, while the surface area of the jacket (cooling area) scales only quadratically. This fundamental math means cooling a commercial batch can take hours.

Why does this matter? Because while the batch sits at 60 °C, the waxes are crystallizing differently, viscosity is building at a different rate, and heat-labile preservatives might be degrading. Reading the FDA’s 7/94 Guide to Inspections of Topical Drug Products made me realize that a tank's physical anatomy—scraped-surface agitators, dead spots, tank heels, bottom valves—is just as heavily regulated under 21 CFR 211.100(a) as our chemical test methods. The fact that a hand-spatula scraping a kettle wall instead of a designed scraper blade is a compliance failure (causing non-uniformity) blew my mind. Form follows compliance.

2. Conceptual Breakthroughs

The biggest "aha" moment was understanding the SUPAC-SS (Scale-Up and Postapproval Changes for Nonsterile Semisolid Dosage Forms) guidelines and the concept of engineering invariants. I used to think scaling up just meant building a bigger motor. But you cannot keep all parameters constant when moving from the bench to a commercial vacuum emulsifier.

If you want to maintain the same shear rate, your mixing time has to change. If you keep the rotor tip speed constant ($\pi DN$), the energy dissipation per unit volume drops. The breakthrough was realizing that we don't seek a mathematically perfect scale-up; we seek an Operating Envelope. Through Design of Experiments (DoE), we purposely push the rotor speed, phase-addition rate, and temperature to their boundaries on the bench to see where the emulsion fails (e.g., phase inversion or globule sizes exceeding our microscopic IPC limits). We then map those boundaries to the commercial vessel. It makes process validation (FDA 2011 lifecycle guidance) feel much less like a guessing game and more like a verified mathematical model.

A secondary breakthrough was the concept of theoretical vs. actual yield in semisolids (§211.192). In our LC-MS/MS peptide work, a 2-microliter injection loss is nothing. In a semisolid manufacturing plant, tons of viscous product gets stuck in the transfer hoses, homogenizer loops, and tank heel. Managing these physical losses without triggering a deviation investigation requires meticulous mass-balance engineering.

3. Connections to Peptides (Retatrutide, BPC-157, etc.)

Even though this course strictly focuses on API-agnostic, OTC dermatological topicals, my mind kept drifting to the peptides we analyze at Peptide Partners. If Larissa and I ever had to scale a topical hydrogel of BPC-157 or TB-500, the engineering rules I learned this week would be our saving grace—or our downfall.

Peptides are highly susceptible to mechanical and thermal degradation. If we added a peptide to the aqueous phase before a high-shear rotor-stator homogenization step at 70 °C, the shear forces and heat would likely shred or aggregate the peptide before the emulsion even formed. This course taught me the strategy of "late additions" (API-addition stage as a CPP). For a peptide topical, we would process the base vehicle, cool it to a validated temperature (say, 30 °C), and only then fold in the API under low-shear sweep agitation. We would also need to carefully evaluate vacuum deaeration; entrapped air can oxidize sensitive molecules, but pulling too hard of a vacuum could pull off volatile antimicrobial preservatives required for multidose gels.

4. Questions for Further Investigation

First, how exactly are we going to construct a representative "bench kit" scale-down model? If commercial cooling is slow, can we program the circulating water bath on our jacketed bench vessel to artificially retard the cooling rate, mimicking the commercial tank’s thermal profile?

Second, I am highly curious about the upcoming finalization of the FDA Q3 Characterization guidance (March 2026). It elevates specific gravity and rheology (<912>) to critical characterization attributes. Since our lab currently lacks a pycnometer, a rotational viscometer, and a microscope for globule-size IPCs, how quickly can we justify acquiring these? Contract labs can handle bulk hold-time studies, but we cannot rely on CMOs for real-time, in-process control (IPC) adjustments during phase prep.

Third, for SUPAC Level 2 equipment changes (like changing from an in-line to an in-tank homogenizer), In Vitro Release Testing (IVRT) is required. I need to dive deeper into Franz diffusion cells. How do we statistically prove that the release rate of our API hasn't changed just because we used a different mixer?

5. Confidence Rating

Most Confident: I feel extremely solid on the compliance frameworks, hold-time protocol design, and chemical IPCs. Taking the sampling logic from PDQC802 and applying it to multi-location bulk sampling (top/middle/bottom) or filling-run uniformity (<3>) makes perfect sense. I also feel great about utilizing our Agilent 1260 HPLCs for assay and preservative partitioning checks, and writing the 21 CFR 211.188 Master Batch Records (MBRs).

Least Confident: The pure mechanical engineering and rheological physics. Fluid dynamics surrounding homogenizer rotor-stator gaps, predicting exact shear rates, and the nuances of pseudoplastic yield-stress behaviors in carbomer gels (like the Islam 2004 paper) are abstract to me right now. Without having a rheometer on our bench in Sarasota to actually feel the difference between a gel neutralized to pH 5.0 versus 8.0, the physics of semisolid flow remains my biggest blind spot.

🗺️ Visual Aids & Key Takeaways

Claude Opus 5.5

Diagram 1: Generic Semisolid Process Flow — Nodes, CPPs, IPCs, and Hold Points

╔══════════════════════════════════════════════════════════════════════════════════════════════════╗
║          SEMISOLID MANUFACTURING PROCESS FLOW — CPPs, IPCs, AND HOLD POINTS                     ║
╚══════════════════════════════════════════════════════════════════════════════════════════════════╝

  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  DISPENSE / WEIGH                                                                           │
  │  • §211.101: second-person verify or validated automated system                             │
  │  • §211.84(d): identity test on every component lot before release                          │
  │  • DEG/EG limit test on glycerin, PG, sorbitol (NMT 0.10% each shipment)                   │
  │  • Benzene-free carbomer grade confirmed (≤ 2 ppm benzene)                                  │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       │
               ┌───────────────────────┴───────────────────────┐
               ↓                                               ↓
  ┌────────────────────────┐                     ┌─────────────────────────────┐
  │  AQUEOUS PHASE PREP    │                     │   OIL PHASE PREP            │
  │  • Purified Water from │                     │   • Waxes, fatty alcohols,  │
  │    qualified POU       │                     │     emulsifiers melted       │
  │  • Humectants,         │                     │   • Target ~70–75 °C        │
  │    water-soluble       │                     │     (product-specific)      │
  │    preservatives,      │                     │   • Visual check: full melt │
  │    thickeners added    │                     │     confirmed before combine │
  │  • Heat ~70–75 °C      │                     │   CPP: temperature, time    │
  │  CPP: temp, time       │                     └────────────┬────────────────┘
  └────────────┬───────────┘                                  │
               └───────────────────────┬───────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  EMULSIFY                                                                                   │
  │  • Phase-addition order (discontinuous into continuous) and rate are CPPs                   │
  │  • CPPs: mixing temperature, mixing speed, addition rate, rotor-stator gap                  │
  │  • Phase-inversion risk: monitor continuously; rate deviations logged                       │
  │  IPC: visual appearance, temperature match confirmed                                        │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  HOMOGENIZE                                                                                 │
  │  • Rotor-stator in-tank or in-line loop under vacuum                                        │
  │  CPPs: rotor speed, gap, homogenization time, vacuum level                                  │
  │  IPC: microscopy — no globules or crystals above product-specific size limit                │
  │        (image-based IPC; acceptance criterion in MBR)                                       │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  COOL                                                                                       │
  │  • Jacket setpoint, low-shear sweep during cooling                                          │
  │  CPPs: cooling rate profile, endpoint temperature                                           │
  │  Risk: API recrystallization if dissolved hot; crystal form shift                           │
  │  IPC: temperature recorded continuously (recording thermometer preferred over hand-held)    │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  LATE ADDITIONS  (below validated temperature — confirmed by IPC before proceeding)         │
  │  • API added here if heat-labile                                                            │
  │  • Fragrance, heat-labile preservatives                                                     │
  │  CPPs: addition temperature limit, mixing speed, addition rate                              │
  │  IPC: temperature at time of addition recorded; visual check dispersion complete            │
  │  Special flags: photolabile → actinic lighting; oxidizable → inert-gas blanket;            │
  │                 BPO → dedicated line, benzene control                                       │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  DEAERATE                                                                                   │
  │  • Vacuum with low-speed mixing                                                             │
  │  CPPs: vacuum level, deaeration time                                                        │
  │  Risk: foaming; loss of volatiles (alcohol, fragrance)                                      │
  │  IPC: specific gravity (pycnometer/weight-per-gallon cup); target range in MBR             │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  FINISHING IPC PANEL  ◄─── HOLD POINT — batch cannot proceed until panel passes            │
  │  • Appearance          (visual, defined standard)                                           │
  │  • pH                  USP <791>, 25 ± 2 °C                                                │
  │  • Apparent viscosity  USP <912>                                                            │
  │  • Specific gravity    pycnometer or weight-per-gallon cup                                  │
  │  • Microscopy          globule/crystal size, contract until scope purchased                 │
  │  • In-process assay    HPLC-DAD on existing Agilent 1260                                   │
  │  IPC FAILURE: pre-approved adjust-and-retest rules in MBR only;                            │
  │               uncontrolled reprocessing prohibited; deviation opened if outside rules       │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  BULK HOLD                                                                                  │
  │  • Documented hold time, temperature, container type, headspace/closure                    │
  │  • Hold-time study arms: chemical (assay, RS, preservative), physical (appearance,          │
  │    viscosity, phase sep., pH, SG, microscopy), microbiological (TAMC/TYMC, specified       │
  │    organisms, Bcc for aqueous)                                                              │
  │  • TAMC ≤ 10² cfu/g; TYMC ≤ 10¹ cfu/g (USP <1111> cutaneous criteria)                   │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  TRANSFER                                                                                   │
  │  • Pump type and shear effects on structured systems documented                             │
  │  • Hose/line dedication; temperature during transfer; air entrainment risk                  │
  │  IPC: viscosity and specific gravity before and after transfer                              │
  └────────────────────────────────────┬────────────────────────────────────────────────────────┘
                                       ↓
  ┌─────────────────────────────────────────────────────────────────────────────────────────────┐
  │  FILL                                                                                       │
  │  • Suspensions: continuous mixing during fill; tank-heel plan documented                   │
  │  • Heated or stirred hopper for viscous systems                                             │
  │  IPC: fill weight beginning / middle / end of run                                           │
  │       uniformity within container: top / middle / bottom per USP <3>                        │
  └─────────────────────────────────────────────────────────────────────────────────────────────┘

End-to-end semisolid process map showing every major node with its critical process parameters, in-process controls, hold points, and regulatory obligations, from dispensing through filling.

---

Diagram 2: Scale-Up Decision Tree — Bench → Pilot → Commercial

╔══════════════════════════════════════════════════════════════════════════════════════════════╗
║        SCALE-UP DECISION TREE: BENCH → PILOT (CMO) → COMMERCIAL                            ║
╚══════════════════════════════════════════════════════════════════════════════════════════════╝

  ┌────────────────────────────────────────────────────────────────────────────┐
  │  BENCH SCALE  (lab kit: jacketed beaker, rotor-stator probe, circulating  │
  │               water bath, overhead stirrer)                                │
  │                                                                            │
  │  Activities:                                                               │
  │  • DoE: factors = rotor speed, time, temperature, addition rate           │
  │  • Responses = viscosity, globule size, SG, appearance, uniformity        │
  │  • Establish design space → operating envelope → draft CPPs               │
  │  • Qualify bench kit as scale-down model (matching criteria, comparative  │
  │    batches vs. pilot; geometric similarity documented)                     │
  └──────────────────────────────────┬─────────────────────────────────────────┘
                                     │
                    Is mixer type SAME design and principle
                    as the next scale vessel?
                                     │
                    ┌────────────────┴────────────────┐
                    ↓ YES                             ↓ NO
        ┌───────────────────────┐         ┌────────────────────────────────────┐
        │  LEVEL 1 CHANGE       │         │  LEVEL 2 EQUIPMENT CHANGE          │
        │  (SUPAC-SS reference) │         │  (SUPAC-SS reference)              │
        │  Same design/principle│         │  Mixer type changed (e.g.,         │
        │  up to 10× biobatch  │         │  high-shear ↔ low-shear)          │
        │  No IVRT required by  │         │  IVRT comparison required:         │
        │  SUPAC-SS per se      │         │  90% CI of slope ratio            │
        └───────────┬───────────┘         │  within 75–133.33%;               │
                    │                     │  Stage 1 = 6 cells;                │
                    │                     │  Stage 2 adds 12 (18 total)        │
                    │                     └────────────────┬───────────────────┘
                    └─────────────────────┬───────────────-┘
                                          ↓
  ┌───────────────────────────────────────────────────────────────────────────────┐
  │  PILOT SCALE  (CMO vessel — vacuum emulsifier, matched design principle)     │
  │                                                                               │
  │  Engineering rules applied (verify — not FDA-specified):                     │
  │  ┌──────────────────────────────────────────────────────────────────────┐    │
  │  │  Criterion             │ Formula               │ Trade-off            │    │
  │  │────────────────────────│───────────────────────│──────────────────── │    │
  │  │  Constant tip speed    │ π × D × N = constant  │ Good for emulsions  │    │
  │  │  Constant E/V          │ P/V = constant        │ Good for dispersions │    │
  │  │  Constant shear rate   │ N = constant          │ Rarely achievable   │    │
  │  └──────────────────────────────────────────────────────────────────────┘    │
  │                                                                               │
  │  Heat-transfer warning:                                                       │
  │    Jacket area / volume FALLS with scale → cooling SLOWS                     │
  │    Adjust jacket setpoints or cooling time to reproduce bench cooling profile │
  │    Failure to adjust → recrystallization, viscosity shift                    │
  │                                                                               │
  │  Tech-transfer package to CMO includes:                                      │
  │    MBR draft, CPP ranges, IPC methods, person-in-plant plan,                │
  │    quality agreement allocating MBR and IPC responsibilities                 │
  │    (facility QU cannot delegate cGMP responsibility)                         │
  └──────────────────────────────────┬────────────────────────────────────────────┘
                                     │
               Comparative CQA testing at pilot scale:
               viscosity, globule size, assay, pH, SG, stability
               Must meet acceptance criteria before proceeding
                                     │
                    ┌────────────────┴──────────────────┐
                    ↓ DATA ACCEPTABLE                  ↓ DATA FAILS
        ┌───────────────────────────┐     ┌────────────────────────────────────┐
        │  COMMERCIAL SCALE         │     │  ITERATE: adjust CPPs, re-run      │
        │  Process Validation (PV)  │     │  pilot; update design space;       │
        │  Justified batch number   │     │  document all changes in dossier   │
        │  Worst-case matrix        │     └────────────────────────────────────┘
        │  Bulk + fill sampling     │
        │  Continued process        │
        │  verification (CPV)       │
        └───────────────────────────┘

Decision tree governing the bench-to-commercial scale-up path, integrating SUPAC-SS change-level logic, engineering scale-up criteria, heat-transfer warnings, and the CMO technical-transfer requirements.

---

Diagram 3: Process Validation Lifecycle and Sampling Architecture

╔═══════════════════════════════════════════════════════════════════════════════════════════════════╗
║      PROCESS VALIDATION LIFECYCLE  (FDA 2011 Three-Stage Framework — verify current status)      ║
╚═══════════════════════════════════════════════════════════════════════════════════════════════════╝

  STAGE 1 — PROCESS DESIGN
  ┌──────────────────────────────────────────────────────────────────────────────────────────────┐
  │  Sources of knowledge:                                                                        │
  │    Bench DoE  →  design space  →  CPP ranges  →  operating envelope (PDDM903)               │
  │    PDDM902 materials survey  →  special-handling flags (BPO, photolabile, oxidizable)        │
  │    7/94 guide expectations  →  equipment URS line items                                      │
  │    Scale-up engineering rules (tip speed, E/V, shear rate) with documented rationale         │
  │  Output: draft MBR, CPP/CQA risk assessment, scale-up dossier                               │
  └──────────────────────────────────────────────────────────────────────────────────────────────┘
                                          ↓
  STAGE 2 — PROCESS QUALIFICATION
  ┌──────────────────────────────────────────────────────────────────────────────────────────────┐
  │  Batch number: JUSTIFIED — no universal count; risk + process knowledge + family bracketing  │
  │  (FDA 2011 PV guidance: verify current language)                                             │
  │                                                                                              │
  │  WORST-CASE MATRIX:                                                                          │
  │    ┌─────────────────────────────────────────────────────────────────────────┐               │
  │    │  Variable              │ Worst-case condition                            │               │
  │    │────────────────────────│─────────────────────────────────────────────── │               │
  │    │  Batch size            │ Maximum (largest vessel, full charge)           │               │
  │    │  Hold time             │ Maximum proposed duration, worst-case container │               │
  │    │  Transfer time         │ Maximum; slowest pump speed                     │               │
  │    │  Cooling rate          │ Slowest (largest vessel, lowest jacket ΔT)     │               │
  │    │  Filling run duration  │ Longest run; most starts and stops             │               │
  │    │  Temperature           │ Edge of validated envelope                      │               │
  │    └─────────────────────────────────────────────────────────────────────────┘               │
  │                                                                                              │
  │  BULK SAMPLING PLAN:                                                                         │
  │                                                                                              │
  │         ┌──────────────────────────────────────┐                                             │
  │         │           BULK VESSEL                │                                             │
  │         │                                      │                                             │
  │         │   ● TOP sample (near surface)        │  ← T, M, B at minimum                     │
  │         │         │                            │                                             │
  │         │   ● MID sample (center depth)        │  ← near-wall vs. center                   │
  │         │         │                            │  ← if vessel large enough                  │
  │         │   ● BOTTOM sample (near drain)       │                                             │
  │         └──────────────────────────────────────┘                                             │
  │                                                                                              │
  │  FILLING-RUN SAMPLING PLAN:                                                                  │
  │                                                                                              │
  │    ┌──────────┐      ┌──────────┐      ┌──────────┐                                         │
  │    │ BEGIN    │      │ MIDDLE   │      │  END     │   ← at least 3 time points              │
  │    │ of run   │      │ of run   │      │  of run  │                                         │
  │    └────┬─────┘      └────┬─────┘      └────┬─────┘                                         │
  │         │                │                  │                                                │
  │         ↓                ↓                  ↓                                                │
  │    Within each container: TOP / MIDDLE / BOTTOM per USP <3> Uniformity in Containers        │
  │                                                                                              │
  │  ACCEPTANCE CRITERIA:                                                                        │
  │    Specifications + tighter IPC and uniformity criteria                                      │
  │    Set using statistical tools: tolerance intervals (PDQC802, PDQC831 — verify)             │
  │    Deviation / failure during PV handled per §211.192                                        │
  └──────────────────────────────────────────────────────────────────────────────────────────────┘
                                          ↓
  STAGE 3 — CONTINUED PROCESS VERIFICATION (CPV)
  ┌──────────────────────────────────────────────────────────────────────────────────────────────┐
  │  CPP and CQA trending  →  defined review frequency  →  control charts                       │
  │  Triggers for revalidation: CPP drift, OOS trend, change control, complaint signal           │
  │  ICH Q10 concepts: product lifecycle management (verify current ICH Q10 text)                │
  │  Links back to: PDQC831 (stability spec lifecycle), PDDM914 (QS deltas)                     │
  └──────────────────────────────────────────────────────────────────────────────────────────────┘

The FDA 2011 three-stage process validation lifecycle mapped to semisolid manufacturing, with the worst-case matrix, bulk and filling-run sampling architecture, and continued process verification triggers.

---

Diagram 4: Hold-Time Study — Three-Arm Protocol and Decision Logic

╔══════════════════════════════════════════════════════════════════════════════════════════════╗
║         BULK HOLD-TIME STUDY: THREE-ARM PROTOCOL AND DECISION LOGIC                         ║
╚══════════════════════════════════════════════════════════════════════════════════════════════╝

  STUDY DESIGN INPUTS:
  ┌──────────────────────────────────────────────────────────────────────────────────────────┐
  │  Worst-case hold conditions:                                                              │
  │    Duration  → maximum proposed hold time                                                │
  │    Temp      → highest ambient or controlled temperature in hold area                   │
  │    Container → vessel vs. totes (document both if used)                                 │
  │    Headspace → air vs. nitrogen blanket (oxidizable actives)                            │
  │    Closure   → sealed vs. loosely covered                                               │
  │  Sampling timepoints: T0, then planned intervals → Tmax                                 │
  └──────────────────────────────────────────────────────────────────────────────────────────┘
                                         │
              ┌──────────────────────────┼──────────────────────────┐
              ↓                          ↓                          ↓
  ┌───────────────────────┐  ┌───────────────────────┐  ┌──────────────────────────────────┐
  │   CHEMICAL ARM        │  │   PHYSICAL ARM        │  │   MICROBIOLOGICAL ARM            │
  │                       │  │                       │  │   (AQUEOUS FAMILIES ONLY)        │
  │  • Assay (HPLC-DAD)   │  │  • Appearance         │  │                                  │
  │  • Related substances │  │  • Phase separation   │  │  TAMC ≤ 10² cfu/g               │
  │  • Preservative       │  │  • Viscosity          │  │  TYMC ≤ 10¹ cfu/g               │
  │    content (HPLC-DAD) │  │  • pH (<791>)         │  │  USP <1111> cutaneous criteria  │
  │  • Antioxidant        │  │  • Specific gravity   │  │                                  │
  │    content (if used)  │  │  • Microscopy         │  │  Specified organisms: USP <62>   │
  │                       │  │    (globule size)     │  │  Bcc: USP <60>                  │
  │  In-house: Agilent    │  │                       │  │                                  │
  │  1260 HPLC-DAD        │  │  In-house gaps:       │  │  IN-HOUSE GAP:                  │
  │                       │  │  viscometer, SG cup,  │  │  BacT/ALERT = presence/absence  │
  │                       │  │  microscope → contract│  │  only; CANNOT do USP <61>       │
  │                       │  │                       │  │  counts → CONTRACT LAB           │
  └──────────┬────────────┘  └──────────┬────────────┘  └──────────────┬───────────────────┘
             │                          │                               │
             └──────────────────────────┼───────────────────────────────┘
                                        ↓
  ┌──────────────────────────────────────────────────────────────────────────────────────────┐
  │  ANHYDROUS FAMILIES (ointments, PEG ointments):                                          │
  │    Water-activity rationale: low aw supports omission of micro arm                       │
  │    Document per USP <1112>                                                               │
  └──────────────────────────────────────────────────────────────────────────────────────────┘
                                        ↓
  ┌──────────────────────────────────────────────────────────────────────────────────────────┐
  │  DECISION GATE AT EACH TIMEPOINT:                                                        │
  │                                                                                          │
  │    ALL ARMS PASS at Tmax?                                                               │
  │         │                                                                                │
  │    ┌────┴────────────────────────────────────────┐                                       │
  │    ↓ YES                                        ↓ NO                                    │
  │  Tmax becomes approved hold time          Investigate failing arm:                       │
  │  Enter in MBR as a hold-time limit        Shorten hold limit to last                    │
  │                                           passing timepoint, OR                         │
  │                                           identify root cause and                       │
  │                                           modify container/conditions                   │
  └──────────────────────────────────────────────────────────────────────────────────────────┘
                                        ↓
  ┌──────────────────────────────────────────────────────────────────────────────────────────┐
  │  TRANSFER ARM (separate but linked study):                                               │
  │    Pre-transfer: viscosity, SG, appearance, pH, assay                                   │
  │    Transfer: pump type (peristaltic vs. centrifugal — shear effects), temp, air          │
  │    Post-transfer: same panel → compare to pre-transfer values                            │
  │    Filling hopper: uniformity at beginning / middle / end of fill run                   │
  └──────────────────────────────────────────────────────────────────────────────────────────┘

The three-arm hold-time study protocol for semisolid bulk, showing the chemical, physical, and microbiological testing requirements, in-house versus contract-lab gaps, the anhydrous water-activity rationale, and the decision logic for setting maximum hold times.

---

Key Takeaways

  1. The seven named CPPs in the FDA QbD review for semisolids are: phase-mixing temperature, mixing speed, rotor-stator gap, rotor speed, powder addition rate, and homogenization time. These are not generic engineering intuition — they were specifically identified in the FDA review of generic topical development (Chang et al., AAPS Journal 2013). Any bench DoE must treat these as primary factors before exploring secondary ones.
  2. Phase-addition order and rate are controlled parameters, not just procedural notes. Adding the discontinuous phase into the continuous phase is the standard approach; reversing it or adding too rapidly risks phase inversion, producing a water-in-oil emulsion when an oil-in-water product is intended. Because inversion is not always visible immediately and can produce a product that passes a gross appearance check while failing microscopically, addition rate must be a documented, monitored CPP with an alarm limit — not merely a verbal instruction.
  3. Jacket area per unit volume decreases as vessel scale increases, so cooling becomes progressively slower at larger scale. This is a geometric reality: surface area scales as the square of a linear dimension, volume as the cube. The consequence for semisolids is that an API dissolved at elevated temperature may have much more time to recrystallize during scale-up if the cooling profile is not re-optimized by adjusting jacket setpoints or cooling time. This is perhaps the single most common and underappreciated source of scale-up failure in semisolid manufacturing.
  4. Constant rotor tip speed (π × D × N = constant) is the most commonly applied criterion for scaling rotor-stator homogenizers, but it is not FDA-specified and involves trade-offs. Maintaining constant tip speed as diameter D increases requires reducing rotational speed N. This preserves the velocity of the rotor surface against the fluid but does not conserve energy per unit volume or shear rate. For emulsions, tip speed is usually the best single criterion; for gels under different rheology, energy per volume (P/V = constant) may be more appropriate. The choice must be documented and justified in the scale-up dossier with supporting comparative CQA data.
  5. SUPAC-SS (May 1997) is an NDA/ANDA document and is used only as a scientific reference for OTC monograph products — it does not create automatic regulatory pathways for monograph changes. Its change-level logic is nevertheless the most systematic framework available: scaling up to 10× the biobatch on equipment of the same design and operating principle is Level 1; changing mixer type (e.g., high-shear to low-shear) is Level 2 and triggers an IVRT comparison requirement. The IVRT acceptance criterion is that the 90% confidence interval of the ratio of release-rate slopes falls within 75–133.33%, using 6 cells in Stage 1 and 18 total in Stage 2.
  6. The FDA 2011 Process Validation guidance (verify current title and status) does not specify a universal batch number for process qualification. It requires a scientifically justified number based on risk, process knowledge accumulated from the bench DoE and pilot data, and the use of family similarity and bracketing logic. This is a profound departure from the historical "three-batch" rule, which had no regulatory or statistical basis. The practical implication is that a well-characterized process with extensive bench and pilot data may need fewer PQ batches, while a novel or high-risk process may need more.
  7. Entrapped air reduces specific gravity and can shift assay and fill weight, making specific gravity a characterization attribute under FDA Q3 guidance (final, March 2026). The mechanism is straightforward: a product that should weigh, for example, 1.02 g/mL will weigh less per unit volume if it contains air bubbles, so a fill target set by weight will deliver less active ingredient per container. Deaeration with low-speed mixing under vacuum is the control, and specific gravity measured by pycnometer or weight-per-gallon cup is the IPC. The lab does not currently own either instrument — this is a documented gap requiring contract or purchase resolution.
  8. For aqueous semisolid bulks, the USP <1111> cutaneous limits are TAMC ≤ 10² cfu/g and TYMC ≤ 10¹ cfu/g, and Bcc testing per USP <60> is required. Bcc (Burkholderia cepacia complex) is singled out because it is an intrinsically multi-drug-resistant opportunistic pathogen capable of surviving in preserved aqueous environments that would kill less hardy organisms. The existing BacT/ALERT instrument cannot perform USP <61> enumeration or <51> log-reduction measurements — it does presence/absence only. All microbiological hold-time arms for aqueous families must go to a qualified contract laboratory.
  9. For anhydrous families (hydrocarbon ointments, PEG ointments), the low water activity rationale documented per USP <1112> supports omission of the microbiological arm of hold-time and validation studies. This is not a waiver granted by a regulator; it is a scientific argument the manufacturer must construct and document: the water activity of the bulk is measured, shown to be below the threshold that supports microbial growth, and the reasoning is recorded in the validation protocol. If the water activity is not measured or documented, the argument cannot be made.
  10. The 7/94 FDA Inspection Guide requires that vessel walls be scraped by mechanical design, not by a hand spatula. This is an underappreciated design requirement: manual spatula scraping introduces non-uniformity because it is inconsistent, it may not be documented in the batch record with sufficient detail, and it represents a process step that cannot be validated in the conventional sense. The engineering solution is a counter-rotating anchor with wall scrapers that contact the vessel surface by design, verified during OQ by a scraper-contact test. Any dead spots must be recirculated or excluded from the process volume.
  11. IPC failure management requires pre-approved adjust-and-retest rules written into the MBR; uncontrolled reprocessing — adjusting a batch without a pre-approved protocol — is a cGMP violation. The distinction is critical: if a batch falls outside a pH IPC limit and the MBR says "if pH is X.X–Y.Y, add Z g of neutralizing agent and retest once," that is controlled reprocessing. If a supervisor decides on the floor to add more base without that pre-approved language, that is an unauthorized deviation regardless of outcome. All adjust-and-retest rules must be reviewed and approved by the QU before the MBR is approved for use.
  12. Carbomer gels show a rheological plateau between approximately pH 5.0 and 8.0, where viscosity changes relatively little with pH, but the neutralization step itself (adding base to swell the polymer) is a CPP because the rate and completeness of swelling determine the final gel structure. Islam et al. (2004) demonstrated this plateau behavior experimentally. The practical consequence is that gel CPPs are dispersion pH, polymer swelling time, and neutralization completeness — not the final pH value across the plateau range. An undergelled batch (incomplete swelling) will pass a pH check but fail viscosity and uniformity.
  13. Suspensions require continuous mixing during filling and a documented tank-heel plan because sedimentation or creaming during filling produces a non-uniform fill across the run. The beginning of a filling run may be enriched in the dispersed phase if it has settled, and the end depleted, or the reverse. The control strategy is mechanical: continuous low-shear mixing in the hopper, a defined minimum stir speed, and beginning/middle/end sampling of fill weight and assay. The tank-heel plan defines the minimum residual volume that must be discarded or reprocessed, because the last material in a settling suspension is compositionally different from the bulk.
  14. The theoretical versus actual yield calculation must account for hold-up in vessels, loops, hoses, the tank heel, transfer losses, and sampling withdrawals — and yield limits that trigger investigation are required by §211.192. A batch that loses 3% to hose hold-up, 1% to sampling, and 2% to the tank heel is at 94% yield before a single container is filled; if the approved MBR states that yields below 92% require a deviation, that batch is compliant. If yields fall below the investigation threshold and no investigation is opened, that is a §211.192 finding. Yield limits must be set on real manufacturing data, not theoretical calculations alone.
  15. A vacuum emulsifier equipped with a bottom-mounted homogenizer, counter-rotating anchor with wall scrapers, recirculation loop, and powder/liquid induction ports represents the engineering minimum for a compliant semisolid manufacturing vessel, per the 7/94 guide and cGMP expectations. Each feature resolves a specific failure mode: the anchor with scrapers eliminates dead zones at the wall (where unmixed material accumulates and later introduces non-uniformity into the batch), the recirculation loop addresses dead spots at the bottom valve, the homogenizer provides controlled globule size reduction, and the induction port allows powder addition without opening the vessel to atmosphere (preserving deaeration and preventing contamination). Every one of these must trace to a URS line item verified during IQ/OQ.
---

References

  1. Levin, M. (ed.), Pharmaceutical Process Scale-Up, 3rd ed. Informa Healthcare, New York, 2011.
  2. Paul, E.L.; Atiemo-Obeng, V.A.; Kresta, S.M. (eds.), Handbook of Industrial Mixing: Science and Practice, 1st ed. Wiley-Interscience, Hoboken, NJ, 2004.
  3. Lieberman, H.A.; Rieger, M.M.; Banker, G.S. (eds.), Pharmaceutical Dosage Forms: Disperse Systems, 2nd ed. Marcel Dekker, New York, 1996.
  4. Agalloco, J.; Carleton, F.J. (eds.), Validation of Pharmaceutical Processes, 3rd ed. Informa Healthcare, New York, 2008.
  5. Chang, R.K.; Raw, A.; Lionberger, R.; Yu, L., 'Generic Development of Topical Dermatologic Products: Formulation Development, Process Development, and Testing of Topical Dermatologic Products,' AAPS Journal, 2013, 15(1), 41–52. DOI: 10.1208/s12248-012-9411-0
  6. Chang, R.K.; Raw, A.; Lionberger, R.; Yu, L., 'Generic Development of Topical Dermatologic Products, Part II: Quality by Design for Topical Semisolid Products,' AAPS Journal, 2013, 15(3), 674–683. DOI: 10.1208/s12248-013-9472-8
  7. Buhse, L.; Kolinski, R.; Westenberger, B.; et al., 'Topical drug classification,' International Journal of Pharmaceutics, 2005, 295(1–2), 101–112. DOI: 10.1016/j.ijpharm.2005.01.032
  8. Islam, M.T.; Rodríguez-Hornedo, N.; Ciotti, S.; Ackermann, C., 'Rheological characterization of topical carbomer gels neutralized to different pH,' Pharmaceutical Research, 2004, 21(7), 1192–1199. DOI: 10.1023/b:pham.0000033006.11619.07
  9. U.S. Food and Drug Administration, Guide to Inspections of Topical Drug Products, July 1994. Available at: https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/inspection-guides/topical-drug-products-794
  10. U.S. Food and Drug Administration, SUPAC-SS: Nonsterile Semisolid Dosage Forms; Scale-Up and Postapproval Changes: Chemistry, Manufacturing, and Controls; In Vitro Release Testing and In Vivo Bioequivalence Documentation, May 1997. Docket 97D-0225.
  11. U.S. Food and Drug Administration, Physicochemical and Structural (Q3) Characterization of Topical Drug Products Submitted in ANDAs, final guidance, March 2026. [Verify current availability on FDA.gov at time of use.]
  12. U.S. Food and Drug Administration, Process Validation: General Principles and Practices, January 2011. [Verify current title and status at FDA.gov.]
  13. U.S. Food and Drug Administration, Microbiological Quality Considerations in Non-Sterile Drug Manufacturing, draft guidance, September 2021.
  14. U.S. Food and Drug Administration, Reformulating Drug Products That Contain Carbomers Manufactured With Benzene, December 2023.
  15. U.S. Food and Drug Administration, Testing of Glycerin, Propylene Glycol, Maltitol Solution, Hydrogenated Starch Hydrolysate, Sorbitol Solution, and Other High-Risk Drug Components for Diethylene Glycol and Ethylene Glycol, guidance for industry, Rev. 1, May 2023. [Verify exact current title at FDA.gov.]
  16. U.S. Pharmacopeial Convention, USP General Chapter <3> Topical and Transdermal Drug Products — Product Quality Tests. In: United States Pharmacopeia. Rockville, MD: USP. [Verify current edition.]
  17. U.S. Pharmacopeial Convention, USP General Chapters <61> Microbiological Examination of Nonsterile Products: Microbial Enumeration Tests; <62> Microbiological Examination of Nonsterile Products: Tests for Specified Microorganisms; <60> Microbiological Examination of Nonsterile Products: Acceptance Criteria for Pharmaceutical Preparations and Substances for Pharmaceutical Use; <1111> Microbiological Examination of Nonsterile Products: Acceptance Criteria for Pharmaceutical Preparations and Substances for Pharmaceutical Use; <1112> Application of Water Activity Determination to Nonsterile Pharmaceutical Products. In: United States Pharmacopeia. Rockville, MD: USP. [Verify current edition.]
  18. U.S. Pharmacopeial Convention, USP General Chapter <791> pH; <912> Viscosity — Rotational Methods. In: United States Pharmacopeia. Rockville, MD: USP. [Verify current edition.]
  19. World Health Organization, WHO Technical Report Series No. 1010, Annex 8: Guidelines on Heating, Ventilation and Air-Conditioning Systems for Non-Sterile Pharmaceutical Products, 2018.
  20. ICH, Q8(R2): Pharmaceutical Development; Q9(R1): Quality Risk Management; Q10: Pharmaceutical Quality System. International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use. [Verify current versions at ich.org.]
  21. Code of Federal Regulations, Title 21, Parts 211.22, 211.42, 211.67, 211.84, 211.100, 211.101, 211.110, 211.113, 211.188, 211.192. U.S. Government Publishing Office. [Verify current wording at ecfr.gov.]
  22. Calvo, N.L.; Asociación Española de Químicos y Coloristas Textiles; et al. [Note to Skipper: The Calvo et al. 2022 reference cited in Week 9 course materials as "Chem Eng Res Des 186:407–415" regarding emulsification parameters should be independently verified before citing — confirm authors, full title, journal volume, and page numbers in the current literature before including in any formal document.]