When undissolved drug looks dissolved: particle absorption in in-situ UV/Vis concentration monitoring

Background and motivation

Dissolution testing needs the concentration of dissolved drug as a function of time. The classical route is to withdraw a sample, filter it and analyse the filtrate by UV spectroscopy or HPLC-UV. An increasingly common alternative is to leave a fiber-optic UV/Vis probe in the vessel and record spectra in situ (Introduction). The probes remove sampling consumables and offline analysis and give frequent, real-time data; they are used for miniaturised intrinsic dissolution rates, for particle size estimates from dissolution data and for precipitation studies in artificial stomach–duodenum models (refs 1–5). The technique was first explored in the late 1980s, with felodipine tablets as the model system (Josefson 1988, ref 6), and has since gained diode arrays, multiple probes, CCD detectors, robotics, several probe designs and validation schemes (refs 7–26). C1

Without filtration, anything in the light path contributes to the spectrum. Dissolved excipients can interfere, and chemometrics or three-wavelength methods have been used to handle that (refs 27–29). For undissolved particles, whether excipient or drug, the literature the authors review (refs 1–29) treats scattering as the only interference. Scattering raises the apparent absorbance, either evenly across the spectrum, which a baseline offset at a compensation wavelength removes, or increasingly toward short wavelengths (Tyndall scattering), which a second derivative of the spectrum removes. Such corrections have produced dissolution curves that agree with conventional sampling. C2

The authors point out that this rests on an assumption particle optics does not support in general. Light removed from a beam by a particle (its extinction) is the sum of scattered and absorbed light, and a molecule that absorbs light does not need to be dissolved to do so (Van de Hulst 1981, ref 30; Bohren and Huffman 1983, ref 31). Absorption by particles depends on size and, as a general rule, dominates the extinction of sufficiently small particles, while scattering takes over as size grows. C3 The scattering-only assumption is therefore "not necessarily valid", and scattering corrections "may not be adequate" when small particles are present. C4 The optical absorption of metal nanoparticles has been described in detail (refs 32–34), but that of organic nanoparticles very little (Horn 2001, ref 35), and, "to the best of our knowledge", it had never been connected to fiber-optic concentration measurement. C5

Small drug particles are not exotic in dissolution media. They are present when a drug is formulated as a nanosuspension (refs 36–38), they appear as larger particles dissolve, and they form when drug precipitates from a supersaturating delivery system (Brouwers 2009, ref 39) or after a pH-induced solubility change; poorly soluble candidates (Lipinski 2002, ref 40) increasingly need exactly such formulations. C6 The paper therefore sets out to characterise how particle size and concentration affect absorption and scattering of UV/Vis light, using felodipine suspensions with particle sizes from about 300 nm to 400 µm read with fiber-optic probes, to interpret the results with Mie theory, and to discuss what they mean for in-situ dissolution monitoring (end of the Introduction). A1·A2·A5

Background concepts

Absorbance and the Beer–Lambert form. A spectrometer reports absorbance, defined from the incident intensity $I_0$ and the transmitted intensity $I$ (Eq. 3):

In plain terms: absorbance is the decadic logarithm of how much the light is weakened at wavelength $\lambda$ over a path of thickness $h$. E3 For a dissolved drug, absorbance is proportional to concentration, which is why a calibration with standard solutions turns a spectrum into a concentration. The method assumes that every bit of absorbance in the drug's band comes from dissolved molecules. P9

Extinction, absorption and scattering. When the light path contains particles, the light lost is described by an extinction cross-section $C_{\mathrm{e}\mathrm{x}\mathrm{t}}$, the effective area a particle removes from the beam. It has two parts (Eq. 8):

In plain terms: a particle removes light either by absorbing it ($C_{\mathrm{a}\mathrm{b}\mathrm{s}}$, absorption cross-section) or by sending it in another direction ($C_{\mathrm{s}\mathrm{c}\mathrm{a}}$, scattering cross-section). E8 If light scattered once does not get scattered again (negligible multiple scattering), a beam crossing a distance $h$ of a suspension with $N$ particles per unit volume is attenuated exponentially (Eqs. 4 and 5, quoted by the authors from Bohren and Huffman):

In plain terms: the attenuation coefficient ${\alpha }_{\mathrm{e}\mathrm{x}\mathrm{t}}$ is the number of particles per volume times the light each one removes, and the transmitted light falls off exponentially with path length. The authors stress that ${\alpha }_{\mathrm{e}\mathrm{x}\mathrm{t}}$ is not the same as the absorption coefficient $\alpha$ of a homogeneous material defined below. E4·E5·C8

The complex refractive index. How strongly a material absorbs is captured by the imaginary part $k$ of its refractive index; the real part $n$ governs refraction and, together with size, scattering. For a homogeneous absorbing film, $k$ and the absorption coefficient $\alpha$ are linked by (Eq. 1):

In plain terms: a material with $k=0$ absorbs nothing; the larger $k$, the more light is absorbed per unit path length, and the effect is stronger at shorter wavelengths. E1 A particle with $k=0$ can still remove light, but only by scattering.

Mie theory. Mie theory is the exact solution for the absorption and scattering of light by a homogeneous sphere of any size. Its inputs are the real refractive indices of particle and medium, the wavelength, the particle diameter and the particle's imaginary refractive index. It returns $C_{\mathrm{a}\mathrm{b}\mathrm{s}}$ and $C_{\mathrm{s}\mathrm{c}\mathrm{a}}$ separately, which is exactly what is needed to ask whether particle absorption matters. Laser diffraction instruments also use Mie theory to turn diffraction patterns into size distributions, as this paper does for its suspensions. C3·E13·P7

Scattering corrections. Two corrections are standard in fiber-optic dissolution work. A constant offset is removed by subtracting the absorbance at a compensation wavelength where only scattering contributes. A baseline that rises smoothly toward short wavelengths (Tyndall scattering) is suppressed by taking the second derivative of the spectrum, which removes constant and linearly sloping baselines but keeps narrower features such as an absorption band. C2 The authors used the second derivative between 270 and 440 nm for both calibration and sample spectra, after Savitzky–Golay smoothing (second-order polynomial, 40-point window), which they report did not alter peaks or intensities. P9·R11 The point of the paper is that a band produced by particle absorption sits at the same wavelengths as the dissolved-drug band, so a second derivative cannot tell them apart.

Size distributions. Laser diffraction reports volume-weighted percentiles: $d(v,0.1)$, $d(v,0.5)$ and $d(v,0.9)$ are the diameters below which 10, 50 and 90% of the particle volume lies. The width of the distribution is summarised by the span (Section on laser diffraction):

In plain terms: a span near 1 is a narrow distribution; a span of 4 or more means the distribution stretches far beyond its median. E12 Dynamic light scattering, used for the nanosuspension, reports instead an intensity-weighted mean, the Z-average, and a polydispersity index. P6·R1

Methods in detail

The study has three parts: probe spectra of felodipine suspensions that differ in particle size and concentration; a set of checks that the extra signal of the nanosuspension is not dissolved drug; and Mie calculations, with felodipine's optical constants measured on amorphous films, that test whether particle absorption explains the spectra. The flow below summarises them.

Contribution 1: probe spectra across three orders of magnitude in particle size

Problem. If particles disturbed probe readings only by scattering, the corrected reading of any felodipine suspension should equal the solubility, whatever the particle size and however much solid is present. The test therefore needs one drug, a wide range of particle sizes, and solid concentrations well above the solubility. Felodipine was a good choice: its aqueous solubility is only about 0.81 µg/ml at room temperature (Lindfors 2006, ref 41), so 10–50 µg/ml suspensions are 12 to 60 times above it. M1·A2

How the suspensions were made. For the nanosuspension, 0.5 g felodipine was milled in 5 ml water containing 0.5 g PVP (Kollidon 30, chosen from a stabilizer screen that is not shown) with 5 g of 0.5 mm zirconia beads and a magnetic stirring bar at 100 rpm, in eight batches, until DLS showed a mean size around 300 nm; the pooled batches were left overnight in a cylinder so that large particles and agglomerates settled, and the supernatant was kept. Its total felodipine content was measured by HPLC, and dilutions in water held PVP at 1 mg/ml. P1·M2·M6 Jet-milled felodipine was sieved through 53 µm, and hand-ground felodipine was sieved into three fractions. These powders were dispersed in aqueous PVP (equal weight to the drug) by 10 min sonication, left at least 60 min and diluted to the same 1 mg/ml PVP. P2·P3·P4 DSC, XRPD, Raman and mid-IR showed crystalline felodipine and no amorphous form in all powders and in the freeze-dried nanosuspension (data not shown). R10

Sizes. DLS gave the nanosuspension a Z-average diameter of 300 ± 3 nm with a polydispersity index of 0.141 ± 0.006. R1 Laser diffraction (Mastersizer 2000, Mie analysis, triplicate) gave median sizes of 0.38, 10.19, 38.39, 105.03 and 380.21 µm for the nanosuspension, the jetmilled powder and sieved fractions 1, 2 and 3 (Table I; Table A below). R2·P7 The text lists the sieve cuts as 300–500, 106–250 and 53–106 µm "Sieved Fraction 1, 2 and 3", the reverse of the order Table I implies; the fractions are best identified by their Table I medians. P3·X1

Spectra and quantification. A six-channel Ocean Optics system with 1 cm path-length transflectance dip probes recorded 200–700 nm at 25 °C. P9 Calibration used 0.5–20 µg/ml felodipine in methanol, and the apparent dissolved concentration of each sample came from the second derivative of its spectrum between 270 and 440 nm, compared with the second-derivative calibration. P9 The paper does not say which derivative feature or wavelength was used for the comparison, and the calibration was made in methanol while the samples were aqueous. X8

Contribution 2: ruling out a real solubility increase

Problem. A nanosuspension that reads high might genuinely hold more dissolved drug: small crystals are more soluble than large ones, an amorphous fraction would be more soluble still, and PVP could solubilise felodipine. Each of these had to be excluded before attributing the signal to the particles. C10

What they did. For small crystals, the authors cite the Ostwald–Freundlich equation and an earlier experimental study, which put the solubility gain of crystalline nanoparticles of this size at 100–115% of the equilibrium value (refs 46, 47). R16·C11·E14 They then measured it: excess nanosuspension or coarse felodipine was stirred for 24 h in 25/75 ethanol/water (chosen to raise solubility and signal-to-noise), and the dissolved felodipine was quantified by 1H NMR at 25 °C from its aromatic peaks, with the water and ethanol signals suppressed, in triplicate. P8 For the solid form, they relied on DSC, XRPD, Raman and mid-IR (no amorphous form found) and on earlier work showing that amorphous felodipine recrystallises in water (refs 48, 49). R10·C13 For PVP, they measured felodipine solubility in PVP solutions (data not shown). R9

Contribution 3: optical constants from films and Mie extinction spectra

Problem. Mie theory needs the imaginary refractive index of felodipine across the UV band. Crystalline powders scatter too strongly to measure it, so the authors used amorphous films, for which they found scattering negligible. R12

Films. Four films with average thicknesses from 0.52 to 1.58 µm were spin-coated from dichloromethane onto quartz cover slips, with thickness from the weight gain (method of ref 43), and read at 361 nm, the band maximum, three times each. P10 Assuming no reflection at the film surfaces, the absorbance of a film of thickness $h$ gives the absorption coefficient (Eq. 2): C7

In plain terms: for a homogeneous absorbing film, the natural logarithm of the attenuation grows linearly with thickness, and the slope is $\alpha$. E2 This gave $\alpha$ = 4.0 ± 0.4 µm⁻¹ at 361 nm and, through Eq. 1, $k$ = 0.116 ± 0.013. R12 Note by the AI writer: Eq. 1 with these inputs gives $k$ = 4.0 × 0.361/(4π) ≈ 0.115, which matches the printed 0.116 within the rounding of $\alpha$. E1 Values of $k$ from 275 to 500 nm, every 5 nm, were then taken from the shape of the film spectrum; they are not printed. R12

From cross-sections to a probe reading. MiePlot computed extinction cross-sections for spheres with the felodipine $k$ spectrum and a real refractive index of 1.57 (ref 44) in water (1.33), and for nonabsorbing spheres with $k=0$. P11·M1·M7 To compare these with measured spectra, the authors rewrite Eqs. 4 and 5 in decadic units. Dividing the cross-section by the particle volume $V_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}}$ gives a cross-section per unit volume, $C_{\mathrm{e}\mathrm{x}\mathrm{t},V}=C_{\mathrm{e}\mathrm{x}\mathrm{t}}/V_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}}$, so that (Eqs. 9–11):

In plain terms: $N{V}_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}}$ is simply the particle volume fraction ${\varphi }_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}}$, so the absorbance a probe would read equals the volume fraction times the path length times the extinction per unit particle volume, divided by ln 10. E9·E10·E11 With a 1 cm path, 50 µg/ml of particles and the density of crystalline felodipine, 1.451 (from the crystal structure DONTIJ, ref 45), every calculated spectrum refers to the same mass of drug as the measured 50 µg/ml nanosuspension. P11 Note by the AI writer: the paper prints no value of $K$; with these inputs, reading the density as g/cm³, ${\varphi }_{\mathrm{p}\mathrm{a}\mathrm{r}\mathrm{t}}$ ≈ 3.45 × 10⁻⁵ and $K$ ≈ 1.50 × 10⁻⁵ cm (0.150 µm). Equation 7 is printed with a Latin "a" where the other equations have ${\alpha }_{\mathrm{e}\mathrm{x}\mathrm{t}}$, a typographical slip. E11·X10

Why it works, and what it assumes. The calculation isolates particle absorption: the difference between an absorbing and a nonabsorbing sphere of the same size is what felodipine's absorption adds. It assumes homogeneous monodisperse spheres, $k$ from amorphous rather than crystalline felodipine, a real refractive index that is estimated and taken as wavelength-independent, and no Kramers–Kronig coupling between real and imaginary parts; multiple scattering is neglected. C39·C8 The sizes covered were 30, 60, 100, 200 and 300 nm (Figure 4A), 400–1000 nm (Figure 4B) and 1200–4000 nm (Figure 4C), all at 50 µg/ml. P11

Experimental analysis

The five suspensions. Table A reproduces the laser-diffraction data. The medians span three orders of magnitude, as the authors note. R2·C19 Two details matter later. The nanosuspension is by far the broadest distribution (span 4.61): its median is 0.38 µm but 10% of its volume lies above 1.90 µm. R2 And no suspension has a median between 0.38 and 10.19 µm, so the size range in which, according to the Mie calculations, the spectra are most complex was not covered experimentally. R2·X5

Table A. Particle size by laser diffraction, mean ± SD, n = 3 (from Table I). R2

Main result 1: the nanosuspension reads far above the solubility. Figure 1 shows the probe spectra of the nanosuspension at 10, 25, 40 and 50 µg/ml total felodipine. Every spectrum has a baseline that rises toward short wavelengths, the Tyndall shape of scattering, and a felodipine band whose height grows with the solids: at the band maximum near 375 nm the absorbance rises from about 0.20 to 1.04, and at 276 nm from about 0.24 to 1.28. R3·C9 The methanolic solution at 0.81 µg/ml, which shows what a saturated solution should give, peaks at only about 0.015. R3 The authors note that, read conventionally, this would mean the dissolved concentration rises with the amount of suspended drug, and that one might be tempted to explain it by the higher solubility of highly curved nanoparticles. C10 After the second-derivative correction the apparent dissolved concentration was about 3.4, 5.75, 7.85 and 8.95 µg/ml at 10, 25, 40 and 50 µg/ml total (Figure 3; Table B). R6 Linear regression of these four points gives a slope of 0.14 with R² = 0.995: each additional 10 µg/ml of particles adds 1.4 µg/ml of apparent dissolved drug. R7 The authors call the impact "dramatic" and conclude that, with or without the usual corrections, such measurements give much higher apparent concentrations than are actually present, an interference that "clearly cannot be ignored". C23·C25·C27

Note by the AI writer (not discussed in the paper): the band maximum of the aqueous nanosuspension lies near 375 nm, with a shoulder near 390 nm, whereas the methanolic reference and the jetmilled suspensions peak near 355–365 nm. R3·R4 The quantification was calibrated with methanolic standards, so the second-derivative comparison involves bands at slightly different positions. X8 A straight line through the four nanosuspension points read from Figure 3 reproduces the printed slope (about 0.140) but has an intercept of about 2.1 µg/ml; even at 10 µg/ml total the reading is about four times the solubility, more than the slope alone accounts for. X9

Table B. Apparent dissolved felodipine (µg/ml) from second-derivative spectra, read from Figure 3 (approximate). Cited aqueous solubility: about 0.81 µg/ml. In Figure 3 no sieved fraction 1 marker is visible at 50 µg/ml. R6·M1·X11

Main result 2: micron-sized particles add a nearly flat offset. For the jetmilled suspension (Figure 2, on a y-axis about 17 times smaller than Figure 1) the absorbance stays below about 0.075. R4 Scattering raises the whole spectrum fairly evenly, from about 0.011 to 0.046 at 425 nm between 10 and 50 µg/ml, rather than rising steeply toward short wavelengths, and the felodipine band above that baseline "appears to be relatively constant and similar to" the methanolic reference, whatever the total concentration. R4·C16·C17 The authors report similar spectra for the other particle sizes but do not show them. C18 In Figure 3 the four coarser suspensions read about 0.9–2.1 µg/ml, "relatively flat" with a "slight trend" upward (Table B). R6·C20 Note by the AI writer: reading the band height above the 425 nm baseline off Figure 2 gives about 0.018 at 10 µg/ml and 0.027 at 50 µg/ml, against about 0.016 for the reference, which is the same slight upward drift. X13

These coarse-suspension values should equal the solubility, but reach up to 2.10 µg/ml, about 2.6 times the cited 0.81 µg/ml. R6·M1·C21 The authors think this "most likely" reflects the small absolute absorbance of a saturated solution, quoting a band maximum of "only 0.034 a.u.", so that small absolute errors in the quantification become large relative errors; they leave the cause unexplored because the deviations are small next to those of the nanosuspension. C22·R5 Note by the AI writer: the fiber-optic spectrum of the 0.81 µg/ml methanolic reference plotted in Figures 1 and 2 peaks at about 0.016, not 0.034; the paper does not say which measurement the 0.034 a.u. refers to. X2

Main result 3: the extra signal is not dissolved drug. NMR in 25/75 ethanol/water gave a nanosuspension solubility of 89 ± 10% of that of coarse felodipine, consistent with the cited expectation of at most 100–115% for crystalline nanoparticles of this size; the authors say this "independently confirms" that solubility increases are not responsible. R8·R16·C11·C15 DSC, XRPD, Raman and mid-IR found no amorphous felodipine, which they take as "eliminating" a high-energy form as the cause; even a small amorphous fraction "would be expected" to recrystallise. R10·C12·C13 PVP had no significant effect on felodipine solubility below 5 mg/ml, five times the 1 mg/ml used (data not shown). R9·C14 With these alternatives excluded, the authors write that the observation "strongly suggests" that submicron felodipine particles absorb light much as dissolved felodipine molecules do; in the abstract the same point is worded more cautiously, as nanoparticles that "might absorb UV light", and the Mie section speaks of what the experiments "demonstrate". C24

Note by the AI writer (not discussed in the paper): the solubility comparison was made in ethanol/water, not in the 1 mg/ml PVP aqueous medium that the probe read, and no particle-free aqueous reference (for example a spectrum after ultracentrifugation) was recorded. The particle-absorption explanation therefore rests on excluding alternatives and on the Mie calculation, not on a direct separation of particle and dissolved-drug signals. X4

Main result 4: Mie theory reproduces the size of the effect. Figure 4A compares calculated spectra of 30–300 nm spheres at 50 µg/ml with the measured 50 µg/ml nanosuspension. For nonabsorbing spheres the extinction is scattering only; it rises steeply with size (at 276 nm, about 0.01 at 30 nm, 0.22 at 100 nm and 0.93 at 300 nm) and toward short wavelengths, the Tyndall pattern. R13·C28 Spheres with felodipine's absorption add a band where felodipine absorbs. In absolute and relative terms this addition shrinks as the spheres grow: at 30 nm absorption is "the dominant form of extinction", as for dissolved molecules, while at 300 nm scattering dominates but the absorption contribution is "certainly still significant" (Table C). R13·C29·C26 The calculated 300 nm spectrum and the measured nanosuspension agree in magnitude, which the authors call "good agreement": band maxima of about 0.94 and about 1.04. R13·R3·C30 Their shapes differ: the calculated band peaks near 350–355 nm, the measured one near 375 nm. The authors think differences in the local environment of felodipine molecules in crystalline nanoparticles, compared with the amorphous films used for $k$, "most likely underlie" this, and note that the calculated band shape resembles the film spectrum (shown scaled to 0.5 in Figure 4A). C31·R12 They conclude that the simulations "clearly illustrate" the cause of the high absorbance: absorption by the particles cannot be neglected. C32

Note by the AI writer: the comparison uses one monodisperse size for a suspension whose volume distribution runs from 0.15 to 1.90 µm (10th to 90th percentile), and no polydisperse calculation was made, so the agreement is best read as one of order of magnitude. X3

Table C. Calculated extinction −log(I/I₀) of 50 µg/ml felodipine spheres, 1 cm path, near 360 nm: band maxima (350–360 nm) for absorbing 30–300 nm spheres, values at 360 nm otherwise; read from Figures 4A–C (approximate). The difference column is the AI writer's arithmetic, not a value given in the paper. R13·R14·R15

Main result 5: larger particles give complex spectra, then a flat baseline. For 400 and 500 nm spheres the absorption contribution falls further (Figure 4B). R14·C33 At 600, 800 and 1000 nm the patterns become complex: nonabsorbing spheres lose the Tyndall shape, with extinction toward short wavelengths that rises only moderately (600 nm), levels off (800 nm) or even falls (1000 nm), and absorbing spheres now show lower extinction than nonabsorbing ones in the felodipine band (Table C). R14·C33 The authors warn that this could make the usual quantification routines problematic even for nonabsorbing excipient particles of these sizes ("may even"), and that for drug particles it is a potential source of error that "may not be easily identified". C34·C35 From 1200 to 4000 nm (Figure 4C, same y-scale) extinction falls with size, from about 0.65 to about 0.13 at 500 nm, its dependence on wavelength weakens, and absorbing and nonabsorbing spheres converge; above 2000 nm the spectrum "appears" to be the horizontal baseline that the literature assumes. R15·C36·C37 Hence, "upwards of a certain particle size, offsetting baselines or taking a second derivative might be appropriate". C38 These regimes are calculated only; no suspension with a median between 0.38 and 10 µm was measured, and the calculations were made at one concentration. X5

Discussion

What the paper establishes. Taken together, the measurements and calculations support one picture. Undissolved particles remove light by scattering and by absorbing it. For particles of a few micrometres and more, extinction is mostly a flat scattering baseline, which the usual corrections handle, and the measured 10–380 µm suspensions indeed read close to the solubility. C37·C38·R6 For submicron particles, absorption is a large part of the extinction and sits in the drug's own band, so it looks like dissolved drug and survives a second derivative; the 300 nm nanosuspension read up to about 9 µg/ml against 0.81 µg/ml. C24·C29·R6·M1 In the authors' summary, the spectra of submicron felodipine suspensions combine particle absorption with dissolved-drug absorption and particle scattering. C46 Because these contributions overlap and change shape with size, the authors judge the spectral influence of small particles "much more complex" than that of dissolved excipients or large particles, which makes separating dissolved-drug and particle contributions "extremely challenging". C41

What sets the size of the error. The authors expect the relative importance of particle absorption to grow with (i) lower solubility of the drug in the medium, because the true dissolved signal is then small, (ii) higher solids concentration and (iii) smaller particle size ("can be expected"). C44 Solids and size were varied here: the error grew steeply with the solids for the nanosuspension and was small for the coarse suspensions. R6·R7 Solubility was not varied; felodipine in water is one very poorly soluble case. X6

Where it applies. The authors' recommendation is broad: caution with in-situ UV/Vis probes whenever small (submicron) absorbing particles are present, whether formulated as nanosuspensions or formed during dissolution, for example when larger particles dissolve or when drug precipitates from a supersaturating delivery system ("caution should be exerted"). C43 As a practical tool they propose Mie calculations, which in their words "proved successful in modeling the effect of particle size and suspension concentration", as a first evaluation of whether particle absorption matters for a given system. C45·C32 They also flag that nonabsorbing excipient particles of 600–1000 nm may upset scattering corrections. C34

Implications for formulation work (note by the AI writer, not discussed in the paper). The paper predates much of the later work on amorphous solid dispersions, but its message carries over. Supersaturating formulations often produce drug-rich nanodroplets or nanoparticle precipitates during dissolution, and in-situ UV is a common way to follow them. A probe reading taken while such particles are present may overstate the dissolved drug, and a second-derivative correction is not a safeguard. A separation-based measurement on the same samples (filtration or centrifugation followed by HPLC), or an independent method that sees only dissolved drug, shows whether the effect matters. A reading that keeps rising with the amount of solid added, well above the known solubility, is the signature to look for. The slope of 0.14 found here belongs to this felodipine nanosuspension, probe and calibration and should not be used to correct other systems. R7·X9

Limitations

Limitations stated by the authors. The authors list the simplifications of their Mie calculations: homogeneous monodisperse spheres; an imaginary refractive index estimated on amorphous films rather than crystals; a real refractive index that is an estimate and is taken as wavelength-independent; and no Kramers–Kronig relation between the real and imaginary parts. C39 They attribute the band-shape difference between calculation and measurement to the crystalline versus amorphous environment of the molecules ("most likely"). C31 They also leave open why the coarse suspensions read above the solubility, suspecting quantification error on a weak band. C22

The AI writer's own judgement (1): how direct is the evidence? The existence of the artifact is well supported: a large, linear excess over four concentrations for the nanosuspension, absent for four coarser suspensions measured the same way. R6·R7 Its attribution to particle absorption is indirect. The solubility check was made in 25/75 ethanol/water, not in the aqueous medium, and no particle-free aqueous reading was taken, so particle absorption was never separated experimentally from dissolved-drug absorption. X4 The Mie support comes from one monodisperse size compared with a broad distribution, at one concentration, with band maxima about 20–25 nm apart. X3·X5 The conclusion that the Mie approach "proved successful" for the effect of suspension concentration goes beyond what was calculated, since only 50 µg/ml was modelled. C32·X5

The AI writer's own judgement (2): reporting details. Several details need care when reusing the numbers. The sieve ranges in the text are listed in the reverse order to the fractions in Table I. X1 The 0.034 a.u. given for a saturated solution does not match the plotted reference spectrum, which peaks at about 0.016. X2 How the second-derivative spectra were turned into concentrations (which feature or wavelength) is not stated, and the calibration was in methanol while the samples were aqueous. X8 The laser-diffraction sizes of the micron powders were measured in 100 mg/ml PVP after 15 min of sonication, not under the 1 mg/ml PVP conditions of the UV samples. X12 The recommendation extends to precipitated and partly dissolved particles, but only milled or ground crystalline felodipine was measured. X7 None of these changes the main observation, but they limit how far the numbers can be transferred.

Conclusion

The paper shows, for one poorly soluble drug, that an in-situ fiber-optic UV probe can badly overstate dissolved concentration when submicron drug particles are present: a felodipine nanosuspension read about 3.4–9.0 µg/ml at 10–50 µg/ml total against a solubility of 0.81 µg/ml, after the standard second-derivative correction, while 10–380 µm suspensions read about 0.9–2.1 µg/ml. R6·R7·M1 With a real solubility increase, an amorphous form and PVP solubilisation excluded, and Mie theory reproducing the magnitude, the authors conclude that the particles themselves absorb UV light in the drug's band ("strongly suggests"), that this absorption fades with size until a flat scattering baseline remains above about 2 µm, and that in-situ probes should be used with caution whenever small absorbing particles are present or formed. C24·C29·C36·C43