What a PAMPA membrane really measures for supersaturated drug solutions: water layers, receiver sinks and drug-rich nanodroplets

Background and motivation

Many enabling formulations, amorphous solid dispersions (ASDs) among them, dissolve to concentrations far above the crystalline solubility of the drug. Such supersaturated solutions are not simple. The highest concentration of molecularly dissolved drug that can be reached is the amorphous solubility; when a formulation delivers more drug than that, the excess separates by liquid–liquid phase separation (LLPS) into a drug-rich phase dispersed as nanodroplets, in equilibrium with a water-rich phase that stays saturated at the amorphous solubility. Free drug therefore stops rising at the amorphous solubility, and the nanodroplets can act as a reservoir (Ilevbare and Taylor 2013; Mosquera-Giraldo and Taylor 2015). C1

Because only free drug crosses a membrane by passive diffusion, membrane transport, measured as mass flow from a donor into a receiver, has become a standard way to read the thermodynamics of these solutions. In side-by-side diffusion cells, mass flow rises with donor concentration up to the amorphous solubility and then levels off (Raina et al. 2014; Indulkar et al. 2016). Mass flow measurements have shown that surfactants lower the free drug available for transport (Raina et al. 2015), have separated ASD formulations by polymer and drug loading, and have shown that an absorptive compartment changes dissolution and crystallization compared with simple dissolution (Hate et al. 2018). C2·C3·C4

Most of this work used regenerated cellulose membranes, whose biorelevance is debated. The parallel artificial membrane permeability assay (PAMPA) is meant to be closer to a cell membrane: a porous synthetic filter is impregnated with phospholipid in an organic solvent. But PAMPA comes in many variants. The filter support is usually hydrophobic, although hydrophilic supports were introduced to speed up transport and to let hydrophilic or ionized compounds through; the lipid may be 2–20% phosphatidylcholine in n-dodecane or other mixtures; and in the widely used double-sink PAMPA (DS-PAMPA), surfactant in the receiver creates a "high sink" that removes back diffusion (Avdeef et al., US patent). C5·C6·C7·C8 Stewart et al. (2017) had argued that micelles and other colloids can raise mass flow by diffusing into the unstirred water layer next to the membrane. C9

The authors' point is that PAMPA is increasingly used to judge supersaturating formulations without a thorough analysis of how its set-up affects the result, particularly when drug-rich nanodroplets are present. The paper therefore sets out to determine how solute type and concentration, membrane material, lipid content and receiver composition change the mass flow of highly supersaturated solutions with and without nanodroplets, and to explain why. Atazanavir (ATZ) is the main drug because its drug-rich phase had been characterized extensively. C16

Background concepts

Amorphous solubility and drug-rich nanodroplets. Above the amorphous solubility, adding more drug does not raise the free drug concentration; it adds drug-rich nanodroplets. For ATZ, earlier work showed that more drug gives more nanodroplets rather than larger ones, and drug-rich aggregates in such solutions are typically 200–300 nm in size. C1·R2·R4 The amorphous solubilities the paper relies on are cited, not measured here: 75 µg/mL for ATZ, 90 µg/mL for telaprevir, 72 µg/mL for nifedipine (NIF), 6 µg/mL for felodipine and 28 µg/mL for ritonavir (Figure 2 caption). M1·M2·M3·M4·M5

Mass flow rate. In this paper mass flow is the amount of drug arriving in the receiver per minute (µg/min), taken from the slope of the receiver concentration over 90 minutes and the receiver volume. It is not normalized by membrane area, so values compare only within this set-up. E2·P1

Three barriers in series. A solute crossing a lipid membrane meets three resistances: the unstirred water layer (UWL) on the donor side, the membrane itself, and the UWL on the receiver side. The paper writes the overall (effective) permeability as

In plain terms: resistances add, so the overall permeability $P_{\mathrm{e}\mathrm{f}\mathrm{f}}$ is controlled by the smallest of the membrane permeability $P_m$ and the two UWL permeabilities $P_{\mathrm{U}\mathrm{W}\mathrm{L}}$. A UWL is the thin film of still water next to the membrane that drug can cross only by diffusion. Filter hydrophobicity and lipid content act on $P_m$; the media in the two compartments act on all three terms (Equation 1). E1·C10

Sink. A receiver is a "sink" when it keeps the free drug concentration there close to zero. Here this is done either with 2% SDS, whose micelles take up the arriving drug, or with pH 3 buffer, at which the weak base ATZ (pKa 4.48) ionizes and is far more soluble than at pH 6.8. P9·M1·C52

Membrane asymmetry. With the same buffer on both sides, the rate constants for drug entering the membrane from solution ($k_1$) and leaving it ($k_2$) are the same on the donor (d) and receiver (r) sides. A solubilizer on one side only makes them differ, so the membrane is "asymmetric". The membrane partition coefficient is the ratio

In plain terms: $K$ measures how strongly the drug prefers the membrane over the surrounding water. For an asymmetric membrane the paper, citing Kocherginsky and Zhang (2003), states that the overall partition coefficient becomes the logarithmic mean of the donor-side value $K_d$ and the receiver-side value $K_r$ rather than a single $K$. E4·E5·C50

Two ways in which colloids could raise mass flow. The first is the "shuttling" or particle-drifting effect: nanoparticles or micelles diffuse into the UWL and carry drug across it, raising the effective diffusivity there. Since the diffusion coefficient of a particle falls as its size rises (the Stokes–Einstein equation), this effect should weaken for larger particles. C37·C55·E3 The second, which the authors adopt, is a reservoir effect at the membrane surface: colloids that dissolve fast enough replace drug as it is absorbed, so the concentration gradient across the UWL shrinks. Westergaard and Dietschy (1976) proposed this for bile acid micelles feeding fatty acids and cholesterol to the gut wall. C57·C58

Methods in detail

The study changes one element of the PAMPA set-up at a time and reads the effect on mass flow. Its contributions are: a concentration series for five lipophilic drugs across a lipid membrane; a test of filter type, pore size and lipid for ATZ; two independent membrane-integrity tests with and without a surfactant sink; a comparison of a micellar sink with an ionization sink and of a lipophilic with a hydrophilic solute; a test of nanodroplet size and surface charge; and an estimate of the drug concentration at the membrane surface from the mass flow data (Figure 1 summarizes the experimental space). A1·A3·A4·A5·A6·A7·A8

The diffusion set-up and its variants

All mass flow measurements used a side-by-side diffusion cell (PermeGear) with 34 mL compartments and a 30 mm orifice, each side filled with 32 mL of pH 6.8 phosphate buffer (50 or 100 mM) at 25 or 37 °C. A PVDF filter separated the two compartments, either bare or impregnated with 100–150 µL of lipid solution. Drug stock solution was added to the donor, and a UV–visible dip probe followed the receiver concentration for 90 minutes. For ATZ, 1 mg/mL HPMCAS was dissolved in the donor to maintain supersaturation unless stated otherwise. Which buffer strength, temperature and stock concentration belong to which experiment is listed in SI Table S1, which was not available. P1·X10

Three PVDF filters were used: hydrophilic with 0.45 µm pores (HPI0.45), hydrophilic with 0.22 µm pores (HPI0.22) and hydrophobic with 0.45 µm pores (HPO0.45). Two lipid solutions were used: 2% w/v lecithin in n-dodecane, prepared in house, for the first concentration series (Figure 2), and the commercial Pion GIT-0 solution with 20% w/v phosphatidylcholine for all later work, chosen to remove variability from in-house preparation. Regenerated cellulose membranes served as the classic comparator. M18·M19·P2

Testing membrane integrity: Lucifer yellow and D2O

Surfactants can dissolve or disorder lipid bilayers, so the authors had to show that SDS in the receiver did not simply damage the membrane. The usual probe is Lucifer yellow (LY), a hydrophilic dye that should not cross an intact lipid layer; its transport was followed for 16 hours across hydrophilic and hydrophobic PVDF, bare or with 20% lipid, with and without SDS in the receiver. P3 Because LY may be rejected by a hydrophobic filter even without lipid, they added a second probe: water itself. In a 1 mL side-by-side cell the donor held 100% D2O and the receiver H2O, and after 60 minutes the D2O band near 2480 cm−1 (area 2400–2600 cm−1) in the receiver was measured by FTIR-ATR; for the hydrophobic lipid membrane the test was extended to 16 hours. P4·P5

Probing the sink and the nanodroplets

The critical micelle concentrations (CMCs) of SDS and CTAB were measured with pyrene fluorescence (ratio of the 373 and 383 nm peaks). SDS was then placed in the receiver below, at and above its CMC, up to 2% w/v, with 150 µg/mL ATZ in the donor. To separate the effect of micelles from that of solubilizing capacity in general, a pH 3 buffer receiver was compared with the SDS receiver. P7·P9 To compare a membrane-limited with a UWL-limited solute, nifedipine at 75 µg/mL and acetaminophen (APAP) at 500 µg/mL were measured across bare, lipid-impregnated and lipid-impregnated hydrophobic PVDF with an SDS receiver. A6·R16

To change nanodroplet size and surface charge, ATZ at 150 µg/mL was prepared with HPMCAS, HPMC, PVPVA or a Eudragit grade at 1 mg/mL in the donor, or with SDS or CTAB at their CMC (the figure labels show that these two were combined with HPMCAS), with a 2% SDS receiver. Droplet size (Z-average, by dynamic light scattering) and zeta potential were measured at 37 °C on donor samples. The text calls the Eudragit grade "Eudragit L100", while the materials list and the figure name Eudragit E PO; the text describes the additive as anionic, yet the zeta potential plotted for it is positive, and the paper does not resolve the conflict. P10·P6·X1·X17

From mass flow to the concentration at the membrane surface

To test their reservoir idea, the authors turned mass flow into an estimate of the drug concentration in the water at the donor side of the membrane. At steady state, the mass flow across the donor-side UWL equals that across the membrane (Equation 2):

In plain terms: the concentration drop from the bulk donor $C_b$ to the membrane surface $C_{m1}$ grows with the mass flow $J$ and with the UWL thickness $d_{\mathrm{U}\mathrm{W}\mathrm{L}}$, and falls with the solute diffusion coefficient $D$. Across the membrane itself (Equation 3):

In plain terms: the concentration drop across the membrane, from $C_{m1}$ to the receiver-side surface concentration $C_{m2}$, grows with mass flow, partition coefficient $K$ and membrane thickness $d_m$. With a receiver sink, $C_{m2}$ is far smaller than $C_{m1}$, so with constant membrane thickness, diffusivity and area, mass flow is directly proportional to $C_{m1}$. E6·E7·C60 The authors then set $C_{m1}$ equal to the ATZ amorphous solubility, 75 µg/mL, at the maximum mass flow ("~8 µg/min") and scaled the mass flow at each lower donor concentration to a $C_{m1}$ value (Figure 10b). E8·P12

Note by the AI writer (not discussed in the paper): two details of this step matter for reading Figure 10b. First, the equations define $J$ as a mass flow rate, while dimensionally they need a flux per unit area; since the area is constant, the proportionality is unaffected. X18 Second, the plotted $C_{m1}$ values are reproduced when the anchor is the mass flow at 1000 µg/mL (about 8.8 µg/min) rather than the "~8 µg/min" in the text: at 80 µg/mL, 75 × 5.0/8.8 gives about 42.6 µg/mL against about 42 plotted, whereas 8 µg/min would give about 47. The 1000 µg/mL point therefore equals 75 µg/mL by construction. X7·E8·R19

Experimental analysis

Mass flow below and above the amorphous solubility

Across hydrophilic PVDF with 2% lipid and a buffer receiver, the mass flow of all five drugs first rose with donor concentration and then stopped rising (Figure 2). The authors describe a linear increase up to the amorphous solubility, after which "the mass flow rate seemed to approach a constant value", and the break "coincides" with the amorphous solubility, where solute thermodynamic activity becomes constant after LLPS. R1·C11

Table A. Mass flow across hydrophilic PVDF with 2% lipid, buffer receiver; read from Figure 2 (approximate). Amorphous solubilities from the Figure 2 caption (cited). R1·M1·M2·M3·M4·M5

Because the mass flow did not keep rising with added drug, the authors conclude that drug-rich aggregates are "unlikely to permeate" the 0.45 µm filter and that only free drug crosses; 0.22 and 0.45 µm filters gave similar mass flow, which they take as further support, "presumably" because the pores are filled with lipid. C13·C14·C15·R3 They note that absolute mass flow differs between cellulose and PVDF membranes, while the break falls at about the same concentration. C11·C12

Note by the AI writer (not discussed in the paper): the break can be placed only between tested concentrations. Felodipine is already at its plateau at about 3 µg/mL, below its cited 6 µg/mL, and nifedipine has no point between 70 and 90 µg/mL, so its break could lie anywhere in that interval. X13

Filter type, pore size and lipid hardly matter for ATZ

With a buffer receiver, ATZ mass flow was measured on bare and lipid-impregnated hydrophilic and hydrophobic filters from 80 to 1000 µg/mL, well above the amorphous solubility (Figure 3). At each concentration the four values lay within about a factor of 1.7, and very high donor concentrations gave higher mass flow than 80 µg/mL on both supports. R5·C17

Table B. ATZ mass flow (µg/min) with a buffer receiver; read from Figure 3 (approximate). * marks the no-lipid/lipid pairs flagged as significant (p < 0.05) in the figure. R5

The authors write that no significant difference appeared between hydrophilic and hydrophobic filters and that mass flow was similar with or without lipid, and they read this as showing that "membrane permeability is not the rate-limiting step". C17·C18 Note by the AI writer: Figure 3b itself marks two lipid/no-lipid pairs on the hydrophobic support as significant, and no test across filter types is shown, although at 80 µg/mL the lipid-impregnated filters differ by about 1.7-fold. The conclusion "little effect" holds; "no difference" is stronger than the figure. X2

The hydrophilic comparator makes the point clearer (Figure 7). On hydrophobic PVDF, NIF at 75 µg/mL gave about 2.77 µg/min without lipid and 3.50 with lipid, whereas APAP at 500 µg/mL gave about 0.19 without and 0.50 with lipid. R16 The authors conclude that ATZ and NIF transport is "most likely UWL-limited", "a typical rate-limiting transport barrier for lipophilic compounds", and that the APAP results are "strongly suggesting" a membrane-limited solute. They attribute the low bare-filter APAP value to the hydrophobic filter rejecting hydrophilic molecules, as it did LY, and the much higher ATZ and NIF mass flows to lipophilicity. C32·C33·C34·C36 By Equation 1, ATZ (log P ~4.5) should have a high membrane permeability but a low UWL permeability. C44·M1 The supporting information reports calculated $P_m$ and $P_{\mathrm{U}\mathrm{W}\mathrm{L}}$ values with and without the SDS sink (Tables S2, S3), which could not be checked here. X8

What SDS does to the membrane

LY transport over 16 hours (Table 1) separates the two filter types sharply. R8

Table C. Lucifer yellow mass flow across 0.45 µm PVDF with 20% lipid, over 16 h (from Table 1). R8

On the hydrophilic support, LY crossed the lipid membrane again once SDS was in the receiver, at 0.82 µg/min against 1.01 for the bare filter, and SDS was found in the donor. The authors read this as SDS leaching the lipid out of the pores, "suggesting that SDS might have caused the lipids to leach from the filter support", which "could be because of weaker interactions between lipids and the hydrophilic membrane". R8·R9·C20·C21 On the hydrophobic support LY never crossed, but since it did not pass even the bare hydrophobic filter, the authors call it "questionable" as an integrity probe there. C22·C23

The D2O test resolves this (Figure 5). After 60 minutes, D2O had crossed both bare filters (D2O band peak about 0.20 for hydrophobic and 0.185 for hydrophilic, against about 0.34 for pure D2O), the lipid-impregnated hydrophilic membrane slightly (about 0.086, which "could possibly be due to incomplete pore filling"), and the same membrane with SDS in the receiver almost as much as the bare filter (about 0.17). No D2O band appeared for the lipid-impregnated hydrophobic membrane with or without SDS, and after 16 hours its receiver held 3.0% D2O without and 2.2% with SDS. R10·R11·C24 The authors conclude that SDS impaired only the hydrophilic lipid membrane, so the larger mass flow seen with SDS on the hydrophobic support "cannot be explained by surfactant-induced changes in membrane properties", and that monitoring water transport "may be an improved approach" over LY. C25·C26·C27

The receiver sink

With the membrane shown to be intact, the hydrophobic support was used to compare receivers (Figure 6). R15

Table D. ATZ mass flow (µg/min) across hydrophobic PVDF with 20% lipid; read from Figure 6 (approximate; the pH 3 value at 150 µg/mL has a large error bar). R15

Three observations follow. A sink raised mass flow both without nanodroplets (20 µg/mL) and with them (80 µg/mL and above); micellar and pH 3 sinks gave similar gains; and under a sink, mass flow kept rising as donor ATZ increased above the amorphous solubility. The authors conclude that the gain comes from "the solubilizing capacity of the receiver media", "not due to the presence of SDS micelles per se". C30 SDS raised mass flow only marginally below its CMC (0.5 mg/mL, 0.05% w/v), significantly at the CMC, and not further at 2% w/v. R12·R13·C29 Note by the AI writer: the text then places the full sink effect "at 0.15% SDS", three times the measured CMC, and the SDS levels tested are only in Figure S6; whether 0.15% is a tested level or a misprint cannot be settled from the main text. X6

The same sink raised NIF mass flow from about 3.50 to 9.44 µg/min but APAP only from about 0.50 to 0.55 µg/min (Figure 7). R16·C31·C35 On the hydrophilic support, where SDS leaches the lipid, the picture is mixed: mass flow with SDS was lower at 80 and 150 µg/mL (about 0.88 against 1.31, and 1.15 against 1.66 µg/min) and higher at 500 and 1000 µg/mL (about 3.75 against 2.44, and 4.5 against 3.68 µg/min). The text speaks of "a small increase". R7·C19·X3

Nanodroplets under a sink

To follow the nanodroplet effect further, ATZ mass flow was measured from 80 to 1000 µg/mL with HPMCAS in the donor and 2% SDS plus 1 mg/mL HPMCAS in the receiver (Figure 10). P11·R19

Table E. ATZ mass flow under an SDS sink and the donor-side interface concentration derived from it; read from Figure 10 (approximate). R19·E8

Mass flow rose from about 5.0 to 8.8 µg/min, by about three quarters, between 80 and 1000 µg/mL. With a buffer receiver it rose much less, from about 2.2 to 3.2–3.4 µg/min over the same range (Figure 4b), and across cellulose it stayed essentially constant (Figure S2). R19·R7·R6 The authors describe the derived $C_{m1}$ as rising "asymptotically" to "a near-plateau at donor concentration of 400 µg/mL", "quantitatively agreeing" with nanodroplets replenishing drug in the UWL. C61 Note by the AI writer: $C_{m1}$ is the mass flow rescaled so that the 1000 µg/mL point equals 75 µg/mL, so it carries no information beyond Figure 10a; and the values still rise from about 64 at 400 to 75 at 1000 µg/mL. X7·X8 The text also states that ATZ mass flow with each additive "seemed to eventually plateau at ~ 8 µg/min" (Figure S7). R18·C59

Nanodroplet size and surface charge

At 150 µg/mL ATZ with an SDS receiver, the donor additive changed nanodroplet size, charge and mass flow (Figure 8). R17

Table F. Effect of donor additives on ATZ nanodroplets and mass flow; read from Figure 8 (approximate). R17·X1·X15

Neutral HPMC and PVPVA gave the highest mass flow, and the four charged systems gave lower values. Size did not order mass flow: CTAB formed large agglomerates yet matched HPMCAS, and HPMC gave larger droplets than HPMCAS yet higher mass flow. C38·C39 The authors conclude that nanodroplet surface properties (charge and polymer type) "may impact the mass flow", "although additional experiments are warranted". C40 At constant ATZ (150 µg/mL), more HPMCAS in the donor gave lower mass flow (Figure S9). R21·C65

Discussion

The authors' mechanism

The authors build one picture from these results (Figures 9 and 11). For lipophilic drugs the UWL is the smallest permeability in Equation 1, so changing the filter or the lipid changes little, while water (D2O) and APAP, which are membrane-limited, respond to the membrane. C42·C45 A solubilizing receiver makes the membrane asymmetric: drug entering the receiver is taken up by micelles or ionized "immediately" at pH 3, receiver free drug stays near zero, the exit rate constant on the receiver side rises, the membrane concentration falls and the donor concentration declines faster. "As a result of this asymmetry, both the membrane and UWL permeabilities increase". C50·C51·C52 The effect is "expected to be significant" when the membrane partition coefficient is large, which fits the gain for ATZ and NIF and its absence for APAP (log P given as 0.46 and as 0.51 in two places). C53·X5

For the gain above the amorphous solubility, where donor free drug is constant, the authors rule out membrane damage and transport of nanodroplets through the pores, and conclude there are two contributions: the sink with its asymmetry, and the nanodroplets. C54 Because size did not order the gain, they set aside shuttling and turn to the reservoir idea: nanodroplets "can potentially dissolve virtually instantaneously", keeping the water at the membrane surface near the amorphous solubility and removing the UWL gradient, so that the system becomes membrane-limited. C56·C58

The authors add that the effect "would be more pronounced for larger molecules", whose UWL permeability is lower, and that additives which adsorb onto or partition into nanodroplets may change their dissolution rate or even their amorphous solubility, which "can potentially" explain why HPMC and PVPVA reached the maximum mass flow at a lower ATZ excess than HPMCAS and Eudragit. C70·C66·R18 In the conclusions this becomes a firmer statement: drug-rich nanodroplets "were observed to serve as drug reservoirs", reducing the UWL concentration gradient and increasing the driving force across the membrane. C76

What it means for evaluating formulations

The authors state that PAMPA mass flow of highly supersaturated solutions of lipophilic drugs is "highly dependent on the assay conditions, most notably the composition of the receiver fluid", that understanding these differences is needed to choose the best system for evaluating supersaturating formulations, and that the effect of polymer or surfactant type on nanodroplet properties "certainly warrants careful consideration". C73·C46·C68 The nanodroplet effect "may have a significance in in vivo settings", although the mucus layer "may play an important role" against nanoparticle transport to the cell membrane. C72

Note by the AI writer (not discussed in the paper): for a formulation scientist the practical reading is that the receiver defines the question. A buffer receiver reports mainly the free-drug ceiling set by the amorphous solubility; a strong sink also rewards formulations that generate many drug-rich nanodroplets. Two ASDs with the same free drug can therefore rank alike in one set-up and differently in the other. Within the sink set-up, the donor polymer alone shifted ATZ mass flow by up to about 1.5-fold at the same drug concentration (HPMC about 7.7 against Eudragit about 5.05 µg/min; against HPMCAS, about 5.7, the factor is about 1.35), so flux rankings of ASD polymers may partly reflect nanodroplet surface chemistry. R17·R15 Finally, the hydrophilic PVDF support lost its lipid with an SDS receiver, so the hydrophobic support is the safer choice for DS-PAMPA of such formulations under these conditions. R8·R10

Limitations

Stated by the authors. They call the role of nanodroplet surface properties a matter where "additional experiments are warranted", describe the effect of polymer or surfactant type on nanodroplets as "still an active area of research", and caution that in vivo the mucus layer may hinder nanoparticle transport. C40·C68·C72 They also flag Lucifer yellow as a questionable integrity probe on hydrophobic supports. C23

The AI writer's own judgement.

• The reservoir mechanism is an interpretation, not an observation: $C_{m1}$ was derived by rescaling mass flow to 75 µg/mL at the highest point, and no interface concentration, UWL thickness or nanodroplet dissolution rate was measured. The membrane asymmetry itself was not measured in the main text either. X7·X8
• Several text statements are stronger than the figures: "no significant difference" between supports and lipid states despite two significant pairs (Figure 3b); "a small increase" with SDS on the hydrophilic support despite lower values at 80 and 150 µg/mL (Figure 4a); and a low-sink "plateau" in the Figure 1 summary although Figure 3 shows mass flow rising up to 1000 µg/mL. X2·X3·X12
• Identical bars recur in Figures 3, 4 and 6, so these figures do not add independent replicates; replicate number and error-bar type are not stated in the main text. X9·X10
• "Lipid composition" is said to have little effect, but the 2% lecithin and 20% Pion lipid solutions were never compared side by side in the main text; membrane fluidity is said to be unchanged, but only bulk D2O leakage was measured. X11·X16
• Small inconsistencies: NIF at 75 µg/mL is called "below its amorphous solubility" although the cited value is 72 µg/mL; APAP log P is 0.46 in one place and 0.51 in another; "Fig. 9a" in the discussion should read Figure 8a; the CTAB droplet size sits at a round 1000 nm without error bar. X4·X5·X14·X15
• The nanodroplet and sink work rests on one drug (ATZ), one default polymer (HPMCAS) and one hydrophobic support; NIF and APAP were tested at one concentration each, and the supporting information with per-experiment conditions and the $P_m$ and $P_{\mathrm{U}\mathrm{W}\mathrm{L}}$ calculations was not available. P10·P11·R16·X10

Conclusion

This paper asks what a PAMPA mass flow measurement actually reports when the donor is a supersaturated solution that may contain drug-rich nanodroplets. Its answer, for lipophilic drugs, is that the unstirred water layers set the pace: filter type, pore size and lipid change mass flow little, while a hydrophilic solute such as acetaminophen depends on the lipid. R5·R16·C32·C33 SDS in the receiver strips lipid from a hydrophilic PVDF support but not from a hydrophobic one, and water transport is a more informative integrity test than Lucifer yellow. R8·R10·C25·C27 A solubilizing receiver, micellar or pH 3, raises mass flow two- to threefold below and above the amorphous solubility, which the authors attribute to membrane asymmetry. R15·C30·C75 Only under such a sink does mass flow keep rising with more nanodroplets, from about 5.0 to 8.8 µg/min for ATZ between 80 and 1000 µg/mL, which the authors attribute to nanodroplets refilling the water layer at the membrane surface; that last link rests on a calculated, not a measured, interface concentration. R19·C76·X8