Why the polymer decides how much ritonavir survives hot melt extrusion

Background and motivation

Hot melt extrusion (HME) makes an amorphous solid dispersion (ASD) by melting and mixing a drug with a polymer in a heated barrel. The drug and the polymer are exposed to high temperature and shear, and both can degrade; how much depends on the cumulative exposure, that is, on time and temperature together (Section 1, refs 1–4). For crystalline drugs the usual advice is to keep the process temperature as low as possible, because many drugs degrade rapidly once they melt. But the melt must also be processable, which typically needs 15–60 °C above the polymer's glass transition temperature ($T_g$). For polymers such as PVP/VA (poly(vinylpyrrolidone) vinyl acetate copolymer) and HPMCAS (hydroxypropyl methylcellulose acetate succinate) this places the practical processing range at around 130 °C and above.

Ritonavir (RTV) shows the dilemma clearly. It melts at 126 °C, it is manufactured commercially as an HME dispersion, and it is known to degrade under typical extrusion conditions (refs 9–11). Choosing a polymer that allows extrusion while limiting degradation is therefore, in the authors' words, "inherently challenging".

Polymer selection for ASDs is normally driven by crystallization inhibition during storage or dissolution, with melt viscosity, thermoplasticity and thermal stability considered for processing. The effect of the polymer on the chemical stability of the drug during processing is often overlooked. Earlier work shows that it can go either way. In the group's own TGA study (Moseson et al., ref 3), polymers lowered the degradation onset of posaconazole, with PVP/VA possibly more protective than HPMCAS, while PVP/VA delayed the degradation onset of indomethacin. Vyazovkin and colleagues found that poly(vinylpyrrolidone) (PVP) could protect or destabilize indomethacin depending on intermolecular interactions and drug–polymer ratio (refs 14, 15), and Ferreira et al. found that cellulose acetate protected ibuprofen and naproxen, which they suggested was through hydrogen bonding (ref 16). The authors conclude that the stabilizing mechanism of a polymer "is complex and warrants further investigation".

The paper therefore asks how the polymer changes the thermal degradation of RTV, and why. RTV was dispersed at 20% drug loading (DL), close to the commercial Norvir formulation, in four polymer families: PVP, PVP/VA, HPMCAS and hydroxypropyl methylcellulose (HPMC). The dispersions were heated isothermally in a thermogravimetric analyzer (TGA) and in an oven at HME-like temperatures, and compared with dispersions made by HME. The degradation products were identified by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Three possible explanations were weighed: molecular mobility (represented by $T_g$), intermolecular interactions, and drug–polymer compatibility.

Background concepts

Assay versus mass loss. A TGA records the mass of a sample as it is heated. Mass loss is a common quick measure of thermal degradation, but it only sees products that evaporate. The more direct measure is the assay: the sample is dissolved after heating and the remaining RTV is quantified by HPLC against an unheated reference sample. In this paper the assay is the main read-out, and the comparison between the two is itself one of the results.

Degradation onset, nonisothermal and isothermal tests. In a nonisothermal TGA scan the temperature rises at a constant rate and the degradation onset temperature, $T_{onset}$, is read where the mass starts to fall. The value depends on the heating rate and on how it is read: here at 0.1, 0.5 or 1.0% weight loss after normalizing the curve to 100% at 120 °C, or at the intersection of tangents to the baseline and the falling curve. Because degradation depends on both time and temperature, a scan at one rate cannot tell how much drug will survive a fixed residence time in an extruder. Isothermal tests, in which the sample is held at one temperature for a fixed time, are closer to the process.

Glass transition temperature and molecular mobility. An amorphous solid is a glass below its $T_g$ and a viscous liquid above it, and mobility rises steeply above $T_g$. Mobility can matter for degradation because reacting groups have to move. The authors use the ratio of the hold temperature to the dispersion's $T_g$, both in kelvin, as a mobility scale:

In plain terms: the higher the ratio, the further above its glass transition the dispersion is held. For RTV-HPMC (ASD $T_g$ 54.3 °C, 327.45 K) held at 180 °C (453.15 K) the ratio is 1.38; for RTV-PVP K-30 ($T_g$ 126.5 °C) at the same temperature it is about 1.13. The authors themselves note that $T_g$ is "only a crude indicator of molecular mobility".

Protic and aprotic polymers. The paper groups the polymers by their substituents. PVP and PVP/VA carry pyrrolidone rings (and, for PVP/VA, vinyl acetate units) and are called aprotic. HPMC carries methoxy and hydroxypropoxy groups, and HPMCAS adds acetyl and succinoyl esters, the succinoyl group ending in a carboxylic acid; both are called protic (Figure 1). HPMCAS LF has more succinoyl (14–18%) and less acetyl (5–9%) than HPMCAS MF (10–14% and 7–11%); HPMC has 28–30% methoxy and 7–12% hydroxypropoxy substitution (Section 4.1).

Dehydration and hydrolysis. RTV contains a carbamate, a urea and amide groups (Figure 1). A dehydration reaction removes a water molecule from the drug; a hydrolysis reaction uses water to cleave a bond. Both can be catalyzed by hydrogen or hydroxyl ions. The paper's central mechanistic idea links the two: a dehydration product forms first and releases water, which then hydrolyses more RTV.

First-order kinetics. If the degradation rate is proportional to the amount of drug left, the logarithm of the assay falls linearly with time:

In plain terms: $A$ is the RTV assay (%), $t$ the time in seconds, $k$ the first-order rate constant and $b$ the intercept; a straight line in a plot of $\ln A$ against $t$ is taken as first-order behavior, and a steeper line means faster degradation (SI Figure S13).

Exact mass and ppm error. High-resolution mass spectrometry measures the mass of the protonated molecule, $[M+H{]}^{+}$, to four decimal places. Comparing it with the mass calculated for a candidate molecular formula gives an error in parts per million (ppm); errors below about 5 ppm support the assigned formula. Product ions formed when the molecule is fragmented (MS/MS) then show which parts of the structure were kept or lost.

Methods in detail

The study has four parts: rank the polymers by how much RTV survives a short isothermal hold; test whether mobility explains the ranking by varying $T_g$ within a polymer family; identify the degradation products; and check how well the TGA predicts degradation in a real extruder. The workflow is shown below.

Contribution 1: An isothermal assay screen that ranks the polymers

The problem. Nonisothermal TGA gives a degradation onset that moves with heating rate, and mass loss misses degradation products that do not evaporate. Neither tells how much RTV survives a few minutes in an extruder.

What they did. ASDs were prepared by dissolving RTV and polymer in methanol and removing the solvent in a rotary evaporator at 45 °C, drying overnight under vacuum and cryomilling for 30 s under liquid nitrogen. All ten polymers were used at 20% DL: PVP K-12, K-17, K-30 and K-90; PVP/VA 37, 64 and 73; HPMCAS LF and MF; and HPMC. PVP K-30 and PVP/VA 64 were also made at 10, 30, 40 and 50% DL. Solvent-evaporated dispersions were chosen because they had no previous thermal history but already contained the drug–polymer interactions that "may prevent or prompt" chemical instability (Section 3.2.2). Each dispersion and neat RTV was held for 5 min at 120, 130, 140, 150, 160, 170 or 180 °C in a TGA under a nitrogen purge, then dissolved in acetonitrile with sonication and assayed by HPLC (C18 column, 0.1% formic acid:acetonitrile 35:65, 1.0 mL/min, 240 nm). The 5 min hold was chosen to be short but controlled, to limit the effect of temperature equilibration, and to match the HME residence time. The HPLC was calibrated from 5 to 350 µg/mL (SI Figure S2):

In plain terms: the peak area (mAU·s) rises linearly with the RTV concentration in µg/mL, so the remaining RTV can be read from the peak area. The limits of detection and quantification were set from the standard deviation of ten blanks (std) and the slope $S$:

In plain terms: the smallest detectable and quantifiable concentrations are three and ten times the blank noise expressed in concentration units; they were 7.2 × 10⁻³ and 2.4 × 10⁻² µg/mL.

In separate runs, RTV, the neat polymers and the dispersions were held for 60 min at 120–180 °C (heating at 50 °C/min, duplicate), and the mass curves were normalized to remove early losses attributed to moisture or solvent. Neat RTV was also scanned nonisothermally at 1, 2, 5 and 10 °C/min.

Why it works and how it differs. The assay counts all RTV that has reacted, whether or not the products evaporate, at a fixed time and temperature, so the polymers can be compared under one exposure. It extends the group's earlier TGA work (ref 3), which relied on onsets and mass loss.

Contribution 2: Separating mobility from chemistry

The problem. Polymers that differ in chemistry also differ in $T_g$, so a ranking by polymer type could reflect mobility rather than chemistry.

What they did. Two kinds of series change $T_g$ while keeping the chemistry: polymer grades of different molecular weight (PVP K-12 to K-90) or different VP:VA content (PVP/VA 37, 64, 73), and drug loading (10–50%) in PVP K-30 and PVP/VA 64. $T_g$ was measured by modulated DSC (0–200 °C at 5 °C/min ± 0.796 °C every 60 s, second heating, inflection point, n = 3). The drop in $T_g$ on adding drug was reported as

In plain terms: how many degrees the dispersion's $T_g$ lies below that of the neat polymer (Table 2). The assays of each series were compared by one-way ANOVA at the 95% level, and all assays were plotted against $T/T_g$ (Figure 6). Interactions were probed by transmission FT-IR of spin-coated films on ZnSe (4500–600 cm⁻¹, 4 cm⁻¹, 64 scans, dry-air purge).

Why it works. If mobility controlled degradation, the grade and loading series, which span wide $T_g$ ranges with the same chemistry, should show clear differences, and all systems should fall on one curve against $T/T_g$.

Contribution 3: Identifying the degradation products

What they did. The RTV-HPMCAS LF dispersion held 5 min at 170 °C was chosen because its chromatogram showed the largest number of degradation products while RTV was still present. It was analysed on an Agilent Q-ToF 6550 with a ZORBAX Extend-C18 column, positive electrospray ionization and a scan range of m/z 100–1700; exact masses gave molecular formulas, and product-ion spectra gave fragments (Table 3, SI Figures S15–S17). Structures were proposed by comparing the fragmentation of each product with that of RTV, on the stated assumption that "any changes in the structural content of the parent compound would certainly produce characteristic ions". The temperatures used are above those of typical forced degradation studies, which is why new products could appear.

Contribution 4: Testing the TGA against a convection oven and real extrusion

What they did. Physical mixtures of RTV (20% DL) with PVP/VA 64, HPMCAS MF or HPMC were extruded in a 2 mL Xplore Pharma Melt Extruder with co-rotating conveying screws and a recirculation channel (SI Figure S1), at 50 rpm, with a 5 min residence time counted from the beginning of feeding to the opening of the recirculation valve, and at product melt temperatures of 130, 140, 150 and 160 °C measured by an in-line thermocouple (triplicate preparations). PVP K-30 was not extruded because its high $T_g$ and viscosity would need much higher processing temperatures, and HPMCAS LF was dropped because it was worse than MF. In parallel, about 7 mg of neat RTV, RTV-PVP/VA 64 and RTV-HPMC were held at 160 °C for 2–30 min in a mechanical convection oven (atmosphere not stated) and in the TGA (nitrogen purge), in triplicate, and $\ln A$ was fitted against time.

Why it matters. The TGA uses milligrams of material, so it is attractive for screening. The comparison shows whether its numbers, or only its rank order, can be carried over to the process.

Experimental analysis

All dispersions were single-phase and amorphous

Neat polymers had $T_g$ values from 64.3 °C (PVP/VA 37) to 177.9 °C (PVP K-90), and RTV 50.0 °C (Table 1). Each 20% DL dispersion showed a single $T_g$, between 54.3 °C (RTV-HPMC) and 135.4 °C (RTV-PVP K-90), which the authors read as a single-phase amorphous system; XRPD showed only halos for every solvent-evaporated dispersion and for every extrudate, while neat RTV gave sharp crystalline peaks (SI Figures S3–S5). Adding drug lowered $T_g$ steadily: from 146.9 to 80.4 °C for RTV-PVP K-30 and from 97.1 to 53.2 °C for RTV-PVP/VA 64 between 10 and 50% DL, a $\Delta T_g$ of up to 89.9 and 57.3 °C (Table 2). The grades and loadings therefore cover a wide $T_g$ range, which is what the mobility test needs.

Card check: twice the text gives $T_g$ ranges for the PVP-grade dispersions that do not match Table 1: "64 to 177°C" in Section 3.2.4 and "80 to 170 °C" in Section 4.1, whereas Table 1 gives 86.1–135.4 °C for the 20% DL PVP-grade dispersions; 64–178 °C is the range of the neat polymers (note X1).

Neat ritonavir degrades within minutes once it is above about 150 °C

In nonisothermal TGA, neat RTV started to lose mass at temperatures that depended strongly on the heating rate and the criterion: the 0.1% weight-loss onset stayed at about 134.5–137 °C at every rate, while the 1.0% onset moved from about 165 °C at 1 °C/min to about 195 °C at 10 °C/min, and the tangent onset from about 183 to 214 °C (SI Figure S6; the text gives the overall range as 135–214 °C). In line with their earlier multi-drug study, the authors conclude that judgements about RTV's suitability for HME based on nonisothermal heating alone "may be misleading", and that degradation at high temperature depends on cumulative exposure. In 5 min isothermal holds (Figure 3A), neat RTV kept about 100% up to 140 °C, about 95.5% at 150 °C and then fell to about 82, 63 and 38% at 160, 170 and 180 °C.

Mass loss does not measure degradation

After 60 min at 180 °C, neat RTV had lost 36.4% of its mass and RTV-HPMC 9.1%, with the other dispersions losing less; after 5 min the two had lost 8.0 and 6.4%, and all other dispersions no more than 1.5% (Figure 2B). Yet after 5 min at 180 °C the HPMCAS dispersions had almost no RTV left by assay (below). Over 60 min at 120–140 °C RTV-HPMC lost as much mass as neat RTV or more, but above 150 °C much less (about 91 against 67.5% remaining at 170 °C; SI Figure S7). The neat polymers kept at least about 96.4% of their mass (SI Figure S9). The authors conclude that mass loss "is not a reliable estimate" of RTV degradation, in neat or formulated samples, and also note that the small mass losses do not significantly affect the assay in most systems, which justified the 5 min hold.

The polymer ranking

Table A gives the assay after 5 min at the four highest temperatures, where the systems separate.

Table A. RTV assay (%) after 5 min isothermal TGA under nitrogen (read from Figure 3A, approximate); $T_g$ from Table 1.

The order is PVP K-30 > PVP/VA 64 > neat RTV > HPMCAS MF > HPMCAS LF > HPMC at every temperature from 160 °C. The cellulose polymers are worse than neat drug from the lowest temperature: at 120 °C RTV-HPMC kept only about 88% and RTV-HPMCAS LF about 93%. The authors state that the PVP and PVP/VA dispersions showed "distinctly less degradation than neat amorphous RTV over the entire temperature range", while the HPMCAS and HPMC dispersions were less stable than neat RTV and reached complete or near-complete degradation at 180 °C. The chromatograms (SI Figure S10) show the same: degradant peaks grow with temperature; in the PVP K-30 and PVP/VA 64 sets they stay small even at 180 °C, while for HPMCAS and HPMC the RTV peak has become small or vanished by 180 °C.

Card check: up to 150 °C neat RTV, RTV-PVP K-30 and RTV-PVP/VA 64 all keep about 95.5–102%, so the protection over neat drug is visible from 160 °C, not over the entire range (note X13). The paper calls the heated reference "neat amorphous RTV", but the RTV was bought crystalline and its form during heating is not stated, which matters at 120 °C, below its melting point (note X11).

Drug loading, grade and $T_g$ do not explain the ranking

Across 10–50% DL the assays of RTV-PVP K-30 and RTV-PVP/VA 64 did not differ significantly (one-way ANOVA: F 0.445 and 0.041 against an F critical of 2.690; P 0.775 and 0.997; SI Table S1), and the authors conclude that the protection "was not found to be dependent on DL". The four PVP grades kept about 78–83% at 180 °C (Figure 5A) with no significant difference (P 0.935, SI Table S2). Because the grades differ in molecular weight and so in viscosity, "we infer that polymer global mobility differences are not a major factor". Among the PVP/VA grades, PVP/VA 64 and 73 did not differ (P 0.852), but PVP/VA 37, the grade with the most vinyl acetate, fell furthest: about 50.5% at 180 °C against about 59 and 63% (Figure 5B). The authors read this as "indicating that the PVP portion of the polymer may improve chemical stability, suggesting favorable intermolecular interactions".

Table B. Grade series: ASD $T_g$ from Table 1; $T/T_g$ and assay read from Figures 5 and 6 (approximate).

Against $T/T_g$ (Figure 6) the systems do not fall on one curve. Across polymer types, stability "generally increases as $T_g$ increases", but HPMCAS MF ($T_g$ 72.6 °C) was more stable than HPMCAS LF (75.0 °C), and within the grade and loading series a wide range of $T_g$ gave similar degradation. The authors conclude that there is "no clear correlation" between $T_g$ and degradation, add that $T_g$ is a crude indicator of mobility which specific interactions and high API fraction can distort, and say the data "hint that stability differences are more likely related to polymer chemistry".

Card check: each ANOVA pools all seven temperatures in each group (35, 28 and 14 observations), so the within-group variance includes the temperature effect and differences at one temperature are not tested; at 180 °C the RTV-PVP/VA 64 assay falls from about 58–59.5% at 10–20% DL to about 48% at 50% DL (Figure 4B) (note X7). PVP/VA 37 was not included in any test (note X6). The same RTV-PVP K-30 sample is plotted at about 101.5% and 76% (140 and 180 °C) in Figures 3A and 4A, but about 97.5% and 78% in Figures 5A and 6 (note X4).

FT-IR shows only small changes

Amorphous RTV shows bands at 3294 cm⁻¹ (N–H stretching), 1716 cm⁻¹ (carbamate C=O), 1633 cm⁻¹ (amide C=O) and 1525 cm⁻¹ (aromatic C=C); PVP K-30 has its C=O band at 1681 cm⁻¹, and PVP/VA 64 bands at 1735 cm⁻¹ (vinyl acetate C=O) and 1683 cm⁻¹ (vinyl pyrrole C=O). In the dispersions it was "not possible to detect notable changes". The HPMCAS dispersions showed no shifts or new bands. RTV-HPMC showed RTV's two carbonyl bands at 1639 and 1722 cm⁻¹, which "might indicate" drug–polymer interactions that disrupt RTV's own interactions. RTV-PVP K-30 showed a shoulder at 1720 cm⁻¹ that might reflect changed hydrogen bonding or "a simple overlap" of RTV and PVP bands, and RTV-PVP/VA 64 a slightly shifted and broadened 1681 cm⁻¹ band that "may suggest" interaction at the vinyl pyrrole C=O (Figure 9, SI Figure S14).

Seven degradation products, led by ritonavir minus water

The chromatogram of RTV-HPMCAS LF after 5 min at 170 °C (Figure 3B) shows RTV (about 19.5 mAU) and seven degradation products, the tallest being P4 (about 47.5 mAU) and P2 (about 39 mAU). P4 is the largest degradant peak in every degraded set (SI Figure S10). Table C summarizes the mass spectrometry.

Table C. Exact masses and formulas from Table 3; product ions and routes from Section 3.4 and Figure 10.

All products except P4 are assigned to hydrolysis; P4, with the main fragment ions of RTV and a mass 18 units lower, is assigned to loss of water. P1, P5, P6 and P7 had been reported in forced degradation studies; P2–P4 are reported as new.

Card check: the printed theoretical masses of P5 (551.2315) and P7 (606.3103) differ from those of their printed formulas (551.2323 and 606.3109, recalculated), and the printed P1 error (1.17 ppm) differs from the 0.58 ppm given by its printed masses; all errors stay below 5 ppm (note X3). The P2 retention time is 1.76 min in the text but 1.925 min in the header of SI Figure S15C (note X16).

Extrusion and the oven: same order, far less degradation

All three systems extruded at 130–160 °C and stayed amorphous (SI Figure S5). The PVP/VA 64 extrudates were clear and colourless, while those with HPMCAS MF and HPMC were yellow and darkened slightly with temperature, which "may suggest" drug or polymer degradation (SI Figure S11). Assays stayed between about 95.7 and 100.5%, and the summed degradant areas rose with temperature to about 1.1% (PVP/VA 64), 2.3% (HPMCAS MF) and 6.3% (HPMC) at 160 °C (Figure 7); only PVP/VA 64 extrudates made at 130 and 140 °C stayed under 0.5% total degradants. P4, P5 and P6 were found in the extrudates. The authors rank stability PVP/VA 64 > HPMCAS MF > HPMC "at all temperatures", the same order as the TGA.

Table D. RTV assay (%) after 5 min exposure at 160 °C in three heating modes, as stated in Section 3.2.6.

Over 30 min at 160 °C (Figure 8), the oven left about 86% of RTV in RTV-PVP/VA 64, 77.5% in neat RTV and 18.5% in RTV-HPMC; the TGA left about 76, 23 and 0%. Plots of $\ln A$ against time were straight for all samples (R² 0.9824–0.9989), and the fitted rate constants show the TGA effect: for neat RTV $k$ was 8.15 × 10⁻⁴ s⁻¹ in the TGA against 1.52 × 10⁻⁴ s⁻¹ in the oven, the "up to 5 times faster" rate the text reports; for RTV-HPMC 4.54 × 10⁻³ against 1.01 × 10⁻³ s⁻¹, and for RTV-PVP/VA 64 1.43 × 10⁻⁴ against 9.24 × 10⁻⁵ s⁻¹ (SI Figure S13). The authors find the oven results "similar" to those of HME.

Card check: Figure 8 has no 5 min point; the oven and TGA values in Table D reproduce the fitted lines of SI Figure S13 (for example $e^{4.72-1.01\times {10}^{-3}\times 300}$ = 82.8% for the oven RTV-HPMC). They also differ from the separate 5 min TGA assays at 160 °C in Figure 3A (about 91% for RTV-PVP/VA 64 and 22% for RTV-HPMC) (note X12). At 130 °C the HME assays of the three systems (about 99.9, 98.9 and 99.7%) overlap within their error bars; the rank order holds clearly in the degradant areas (note X15).

Discussion

The authors' mechanism: dehydration first, then hydrolysis

The authors weigh their three hypotheses in Section 4.1. Molecular mobility "was not found to have a direct relationship to degradation extent". Intermolecular interactions "may offer some support" for the protection by PVP and PVP/VA, but the spectral evidence is "somewhat tenuous". Drug–polymer compatibility "was found to have the most consequential influence". Their chain of reasoning is as follows.

Two pieces of evidence support the first two steps: P4 is the dominant degradant in every degraded sample, and the other products have formulas and fragments consistent with hydrolysis of RTV's carbamate, urea and amide bonds. The intramolecular nature of most of these attacks (Figure 10) is the authors' reason why intermolecular interactions with a polymer could shield RTV. For the protic branch, the authors point to the substituents: HPMC, the worst polymer, has the highest methoxy (28–30%) and hydroxypropoxy (7–12%) substitution, and the higher succinoyl content of HPMCAS LF "rationalizes the greater degradation level, compared to the HPMCAS MF". In the conclusions this becomes a general statement: "polymers with protic substituents such as cellulose derivatives are prone to catalyze degradation reactions when compared to aprotic polymers, such as PVP and PVP/VA".

Card check: the order "P4 first, then hydrolysis" was not measured (no time course of individual degradants and no water measurement), and Figure 10 draws every product, P4 included, as formed directly from RTV with H₃O⁺ (note X8). The acid-release argument fits HPMCAS but not HPMC, which carries no acetyl or succinoyl groups, and no free acid, acidity or polymer degradation product was measured (note X9).

Why the TGA overpredicts

The authors note that harsher conditions in the TGA than in HME were also seen in earlier studies (refs 3, 26). They suggest two reasons. The TGA heats the sample from room temperature to the hold temperature and cools it afterwards, which lengthens the effective exposure. And the nitrogen purge, unlike the air in an extruder, "may have eliminated volatile compounds produced during degradation", pushing the reactions forward by Le Chatelier's principle. Their practical conclusion is two-sided. Assay-based isothermal TGA is useful to rank-order drug–polymer compatibility and to study mechanisms, but its degradation extent is overpredicted, and a large apparent loss "may improperly suggest" that a formulation is unsuitable for thermal processing; mass-loss TGA is worse, being "underpredictive/incorrect". "It is therefore essential that such experiments be coupled with actual HME experimentation." A convection oven "may be a more appropriate material-sparing approach" because its results were more consistent with HME.

Note by the AI writer (not discussed in the paper): the extruder residence time of 5 min is counted from the beginning of feeding to the opening of the valve, so material fed later spends less than 5 min in the barrel, while TGA and oven samples are held for the full time. This difference in exposure may also contribute to the lower degradation in HME (note X14).

What it suggests for formulation

Stated by the authors: aprotic PVP and PVP/VA gave RTV better thermal stability than the protic cellulose polymers; PVP K-30 protected best but needs higher extrusion temperatures because of its high $T_g$ and viscosity, so PVP/VA 64 was the extrudable choice; isothermal TGA with assay can rank polymers but must be confirmed by HME; and an oven may be a better small-scale screen.

The AI writer's reading (not stated in the paper): for a drug that degrades by hydrolysis, the polymer family mattered far more than the polymer grade or the drug loading. Across the four PVP grades the dispersion $T_g$ spanned 86.1–135.4 °C while the assay at 180 °C stayed between about 78 and 83%; RTV-PVP/VA 37 and RTV-HPMC have almost the same $T_g$ (54.6 and 54.3 °C), yet at 180 °C about half of the RTV survived in the first and none in the second. HPMCAS and HPMC are often chosen for crystallization inhibition; for ritonavir at extrusion temperatures they carried a chemical-stability cost, so polymer family and process temperature need to be screened together.

Limitations

Limits the authors state. The polymers may themselves degrade during heating (SI Figure S9), but "analysis of their degradation extent and mechanisms was not studied herein". $T_g$ provides "only a crude indicator of molecular mobility". The FT-IR evidence for interactions is "somewhat tenuous". The conclusion speaks of "Limited evidence for stabilization mechanisms based on molecular mobility and intermolecular interactions".

The AI writer's own judgement.

• The central result, the polymer ranking, is solid: it is a direct assay at seven temperatures and recurs in the oven and in the extruder. The mass-loss comparison is equally direct.
• The mechanism offered for the ranking rests on product identities, not on measurements of the steps: no time course of P4 against the other products, no water or acid measurement, and no study of polymer degradation. HPMC, the most destabilizing polymer, does not carry the acid-forming groups the explanation relies on, so its effect remains unexplained.
• The structures of the new products P2–P4 rest on exact mass and fragmentation alone; no product was isolated, compared with a standard or studied by NMR (note X17).
• "No effect of drug loading or grade" rests on ANOVAs that pool all temperatures; at 180 °C the plotted assays do spread with loading, and PVP/VA 37 was not tested. Within PVP/VA, the grade with the most vinyl acetate degraded most; the authors read this as protection by the PVP portion, but vinyl acetate is also an acetate ester, the kind of group they link to acid release in HPMCAS, and the paper does not decide between the readings (notes X6, X7, X10).
• Several text and figure values disagree: the $T_g$ ranges quoted for the PVP-grade dispersions, the RTV-PVP K-30 assays in different figures, the degradant labels in SI Figure S12 (P1–P3 there are P4–P6 in the text), and the discussion's statement that $T_g$ shows no correlation across polymer types, which is stronger than the results section (notes X1, X2, X4, X5).
• The study uses one drug, mostly one loading, a 2 mL extruder and three extruded polymers; the TGA assays carry no error bars, and the solid form of the heated neat RTV is not stated.

Conclusion

The paper makes three points. First, an isothermal TGA hold followed by HPLC assay ranks polymers by the thermal degradation of RTV: PVP K-30 and PVP/VA 64 kept more RTV than the neat drug from 160 °C upward, while HPMCAS MF, HPMCAS LF and HPMC destroyed it, almost completely at 180 °C. Mass loss does not give this ranking. Second, varying $T_g$ through polymer grade or drug loading did not change degradation significantly, so the authors look to chemistry: RTV forms seven degradation products, led by the dehydration product P4, and the authors propose that the water it releases drives hydrolysis that protic polymers accelerate. Third, HME at 130–160 °C keeps the same rank order with far less degradation, and a convection oven tracks HME better than the TGA. The authors' own conclusion is that "the thermal stability mechanisms are related to polymer substituent groups". The parts that remain to be shown are the sequence of the reactions, the role of acid released by the polymers, and why HPMC, without acid-forming groups, is the most damaging polymer of all.