How far can thermogravimetric analysis be trusted to set a hot melt extrusion temperature?

Background and motivation

Hot melt extrusion (HME) makes an amorphous solid dispersion (ASD) by melting a crystalline drug, or dissolving it into a molten polymer, while both are held at high temperature and sheared. The drug and the excipients must come out of this process with their identity, strength, purity and potency intact, so the formulator needs to know how hot, and for how long, a given drug–polymer mixture can be processed. The amorphous form makes the question harder in two ways. Amorphous drugs are known to react faster than their crystals, which generally do not degrade much until they melt; but an amorphous drug can also recrystallize, so its own degradation is difficult to measure at all (Byrn 2001; Carstensen and Morris 1993; Pikal 1977; Baird 2010) C1·C15.

Thermogravimetric analysis (TGA) is the usual screen. It weighs a sample while heating it, and the sample loses mass when water or solvent leaves and when degradation produces volatile products C2. The authors reviewed recent HME studies that used TGA to judge whether a polymer or a crystalline drug was suitable: all 30 used a nonisothermal ramp, 10 °C/min was the most common rate, and only 5 (17%) also held the sample at one temperature, for 8 to 60 min R1. The degradation onset temperature, written $T_{\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{e}\mathrm{t}}$ (or $T_{\mathrm{d}\mathrm{e}\mathrm{g}}$), is often assigned by eye from the curve, with little information on how much degradation has already happened at that temperature and little attempt to compare it with real extrusion; the authors warn that the method "may" therefore give misleading information C3. They also take up an argument from DiNunzio and co-workers that "cumulative exposure", time and temperature together, must be considered when designing an HME process C4. (The paper counts 30 studies but cites 27 references for the survey; the percentages follow the stated count X13.)

The three model drugs illustrate the confusion. Posaconazole (PCZ) is sold as a hot-melt-extruded ASD with HPMCAS; one patent reports significant degradation above 160 °C and processing below its melting point, while a TGA study by an unspecified method found no degradation below 325 °C R2. Indomethacin (IDM) has been reported to start losing mass at 250 °C or at 293 °C, and bicalutamide (BCL) to show no volatile degradation below 250 °C at 10 °C/min and less than 0.5% mass loss after 60 min at 160 °C R2.

The paper therefore sets out three tasks A1–A8. First, to test nonisothermal and isothermal TGA on the three crystalline drugs (and on their amorphous forms where possible) and on six common HME polymers, using differential scanning calorimetry (DSC) to find out whether the amorphous drugs crystallize during an isothermal test, and ASDs to keep them amorphous. Second, to compare the polymer results with literature assessments and describe each polymer's processing window. Third, to measure by HPLC the degradation after real extrusion and compare it with isothermal TGA, to judge what TGA can and cannot say about drug suitability, polymer choice and process design.

Background concepts

Two ways of running TGA. In a nonisothermal run the furnace temperature rises at a constant rate (here 1, 2, 5 or 10 °C/min from ambient to 300 °C or higher) and mass is recorded against temperature (Figure 2a) P8. In an isothermal run the sample is brought quickly to one temperature (here at 50 °C/min) and held there, and mass is recorded against time (Figure 2b; here 4 h at 130–200 °C) P9. A ramp mixes temperature and time: at 1 °C/min the sample spends ten times longer at each temperature than at 10 °C/min.

Reading an onset from a ramp. The paper compares two readings of the same curve P8. The tangent intersection is where the extrapolated flat baseline crosses a tangent drawn to the steep part of the mass-loss step; it is the default output of the instrument software and lies part-way down the step. A weight-loss threshold is the temperature at which a fixed fraction of mass, here 0.1, 0.5 or 1.0%, has been lost. Because water and residual solvent also leave on heating, the curve is first renormalized: to 100% at 140 °C for polymers and ASDs (mass lost below 140 °C is taken as water or solvent) and to 100% "just above the melting point" for the crystalline drugs P8·C8.

Undercooled liquid and glass-forming ability. A melt cooled below its melting point $T_m$ without crystallizing is an undercooled liquid; below the glass transition temperature $T_g$ it becomes a glass. Between $T_g$ and $T_m$ it is mobile enough to nucleate and grow crystals, which is exactly the range in which HME operates. The three drugs were chosen because they are class III molecules, compounds with high glass-forming ability that do not crystallize easily on cooling from the melt (Alhalaweh 2014; Baird 2010; Wyttenbach and Kuentz 2017), so that both crystalline and amorphous states could in principle be tested C5·M1–M3. Table I of the paper gives $T_m$ = 167, 160 and 193 °C and $T_g$ = 60, 45 and 54 °C for PCZ, IDM and BCL M1–M3.

Mass loss, assay and potency. TGA sees only what leaves the pan: water, solvent and volatile degradation products. An HPLC assay measures how much intact drug remains, whatever the degradants are, so potency loss is 100% minus the assay. The two numbers answer different questions, and a degradation route that produces only non-volatile products is invisible to TGA C14·C53.

Cumulative exposure and specific mechanical energy. Degradation during extrusion depends on how long the melt spends at each temperature, not only on the set temperature; the paper uses the isothermal mass-loss map (temperature against time) as a direct picture of this cumulative exposure C4·C28. In an extruder the screws also put mechanical energy into the melt, expressed per unit mass as the specific mechanical energy; at larger scale this, and local heating, can add to the thermal load C50.

Methods in detail

The study runs the same materials through two TGA modes, checks with DSC whether the amorphous drugs survive an isothermal hold, and finally compares TGA with HPLC on samples exposed in the TGA and in an extruder. The flow below summarizes it.

Materials and samples

Crystalline IDM (γ polymorph) and BCL (form I) came from ChemShuttle and PCZ (form I) from Dr. Reddy's Laboratories M1–M3. The six polymers were PVPVA (Kollidon VA64), PVP K-30 and Soluplus from BASF, HPMCAS (Aqoat MF) and HPMC (Affinisol HME 100LV) from Shin Etsu, and Eudragit EPO from Evonik M4–M9. Their Table I glass transitions span 55 °C (Eudragit EPO) to 161 °C (PVP K-30) M4–M9. Four ASDs were made by rotary evaporation at 60 °C, dried under vacuum and cryomilled: 25/75 PCZ/HPMCAS and 25/75 PCZ/PVPVA from methylene chloride–methanol 1:1, 50/50 IDM/PVPVA from methylene chloride–methanol 2:1, and 30/70 BCL/PVPVA from ethanol, with $T_g$ = 97, 99, 67 and 98 °C P2. Physical mixtures were tumble-blended, BCL with micronized (D50 3.5 µm) or large (D50 93 µm) crystals P3·M3. A neat amorphous drug was made in the TGA or DSC pan itself by melting it about 10 °C above $T_m$ for 5 min and cooling to the test temperature P1. PXRD confirmed the identity and solid form of the starting materials (data not shown) P5.

The two TGA protocols

Nonisothermal runs used a TA Discovery TGA 5500 under nitrogen at 1, 2, 5 and 10 °C/min; each run gave four onsets (tangent and three thresholds), so each material has sixteen onset values P8. The paper does not state how many replicate runs lie behind the error bars of these onsets X15. Isothermal runs heated the sample at 50 °C/min to the hold temperature and held it for 4 h, in duplicate; the mass loss was then drawn as a two-dimensional contour map against hold temperature and time (Figures 9 and 12) P9. Figure 2b shows hold temperatures of 130, 145, 155, 165, 180 and 200 °C, while the IDM experiment of Figure 7 used 130–155 °C; the maps interpolate between the hold temperatures actually run P9.

Testing whether the amorphous drug survives the hold

Each crystalline drug (3–5 mg) was melted in a TA Q2000 DSC at about 10 °C above $T_m$ for 2 min, cooled at 20 °C/min to a hold temperature below $T_m$, held for 1 h (PCZ, BCL) or 4 h (IDM), cooled to 25 °C and reheated at 20 °C/min P7. A melting peak on reheating was taken as evidence that the drug had nucleated during the hold C7. The hold temperatures were 130–160 °C for PCZ, 130–150 °C for IDM and 130–180 °C for BCL P7.

Extrusion and assay

Extrusion used a small Xplore Pharma melt extruder with a 5 mL barrel, co-rotating conveying screws and a recirculation channel; the set temperature was adjusted to reach the target product melt temperature read by an in-line thermocouple, and residence time was counted from the moment the recirculation valve was opened P4. PCZ ASDs (25/75 with HPMCAS and PVPVA) were extruded at melt temperatures of 125–175 °C, and 30/70 BCL/PVPVA was extruded from the micronized-crystal mixture at 165 and 180 °C for 10 min at 50 rpm; the authors describe the BCL batches as having minimal mechanical input because of the extruder design and the low melt viscosity P4·C6. BCL was assayed by reversed-phase HPLC at 270 nm (method of Nageswara Rao and co-workers, at a higher wavelength than the original to avoid PVPVA interference), in triplicate, after TGA holds of 10, 30, 60 and 240 min and after 10 min of extrusion; the summed area of the degradant peaks was also compared P12·P13. PCZ and IDM extrudates were assayed by their own HPLC methods P10·P11.

Experimental analysis

1. The onset depends on the heating rate and on how the curve is read

Figure 3 shows the ramp curves of the crystalline drugs and their sixteen onsets each. Crystalline PCZ lost no mass through the usual HME range of 130–200 °C and began to lose mass only far above its melting point; IDM began to lose mass at low levels just above its $T_m$ of 160 °C; BCL lost mass significantly from near its $T_m$ of 193 °C R3·C9. For every drug the curve moved to lower temperature when the ramp was slowed R3. Table A gives the onsets read from the bar charts.

Table A. Onset temperatures (°C) of the crystalline drugs at the slowest and fastest heating rate, read from the bars of Figure 3 (approximate); the plain values are the range ends printed in the text R4.a·R4.d·R4.e·R4.h·R4.i·R4.l·R5.a–c.

The spread between the lowest onset (1 °C/min, 0.1% loss) and the highest (10 °C/min, tangent) was 91 °C for PCZ (285–376 °C), 95 °C for IDM (165–260 °C) and 62 °C for BCL (193–255 °C) R5.a–c. The authors argue that the gap to the literature values "can largely be accounted for" by the methods used, and that the rate dependence "clearly indicates that exposure time plays a role in thermal degradation", which justifies an isothermal approach C10·C11. Their practical reading is that the usual fast ramp read by the tangent or by eye is "likely to overestimate acceptable processing temperatures" C12.

The polymers (Figure 8) and the ASDs (Figure 11) behaved the same way, with even wider spreads: PVPVA 166–299 °C, PVP K-30 147–199 °C, Soluplus 168–295 °C, HPMCAS 149–251 °C, HPMC 160–330 °C and Eudragit EPO 156–264 °C, and for the ASDs 158–257 °C (PCZ/HPMCAS), 174–301 °C (PCZ/PVPVA), 175–273 °C (IDM/PVPVA) and 163–227 °C (BCL/PVPVA) R5.d–m. Figure 4 collects these spreads, from 52 °C for PVP K-30 to 170 °C for HPMC R5. The abstract compresses this into an onset temperature "that ranged from 52-170°C"; those numbers are the widths of the ranges, not onset temperatures C56·X14. The authors conclude that nonisothermal heating "does not provide suitable guide" for choosing HME conditions C22.

Note by the AI writer (not discussed in the paper): the rule "low end at 1 °C/min and 0.1% loss, high end at 10 °C/min and tangent" has exceptions in the figures. HPMC's 0.1% onset falls as the heating rate rises (about 186, 170, 160 and 160 °C at 1, 2, 5 and 10 °C/min), the PVP K-30 high end (about 198 °C) comes from the 1.0% criterion, and the low end of PCZ/PVPVA (about 173 °C) from 2 °C/min X5. More important, the 0.1% bars of IDM/PVPVA in Figure 11c reach down to about 153 °C at 1 °C/min, while the text and Figure 4 give 175 °C as the low end X1·R14.i–l.

2. At the tangent onset, several percent of the mass is already gone

Figure 5 answers the question the tangent method hides: how much mass has the sample lost by the time the tangent says degradation "starts"? Table B gives the values at the slowest and fastest rate.

Table B. Mass already lost (%) at the tangent-intersection onset, read from Figure 5 (approximate) R6.

For the three drugs the tangent onset sat at roughly 10–21% mass loss, which the authors call "well above any pharmaceutically acceptable level", stating more than 10% at every rate; the PCZ bar at 1 °C/min reads just under 10% C13·R6.a–c·X3. For the polymers the value was small and, except for HPMC, nearly independent of heating rate even though the onset itself moved C25·R6.d–i. The abstract concludes that the tangent method "significantly overestimated" the onset of acceptable levels of degradation, while weight-loss thresholds "were more suitable" C12. PVP K-30 is a partial exception: its tangent onset lies below its 1.0% onset at every rate, because its only mass loss in this range is a small step X4·R10.e–h.

Two further points complete the nonisothermal picture. The ramp ignores the greater reactivity of the amorphous form, and because it sees only volatile degradation it "may falsely predict" that a material is stable where non-volatile degradation occurs C14. And for the drug inside an ASD, a crystalline physical mixture is a poor stand-in: in 30/70 BCL/PVPVA, a mixture made with micronized BCL lost mass like the ASD, but one made with 93 µm crystals lagged by about 10 °C at every heating rate (Figure 10) R13·C34. The authors explain this by the time needed for crystals to dissolve into the viscous polymer melt (a rate set by particle size, as in the Noyes–Whitney relation) and for the drug to diffuse through it, citing a diffusion layer of about 8 µm around IDM crystals in a polymer melt; physical mixtures are therefore "likely unsuitable", and the ASD is the "worst-case scenario" for drug degradation C33·C35·R21·E1.

3. A neat glass crystallizes, so the isothermal test needs an ASD

Isothermal heating comes closer to what an extruder does, provided shear or local heating does not accelerate degradation C15. But a neat amorphous drug held between $T_g$ and $T_m$ may crystallize before it degrades. Figure 6 shows the DSC reheating scans after the holds. PCZ held 1 h at 130, 140, 150 or 160 °C showed a melting endotherm near 172 °C every time, largest after the 130 °C hold; BCL held 1 h at 130–180 °C showed melting near 190–196 °C after every hold, with a crystallization exotherm near 125–155 °C on reheating after the 130 and 160 °C holds; IDM held 4 h at 130, 140 or 150 °C showed only its glass transition R7. The authors conclude that a polymer additive "would be necessary" to keep PCZ and BCL amorphous for isothermal degradation tests below $T_m$, whereas IDM "could be suitable" for testing in its neat amorphous form C17·C18.

IDM was therefore held in the TGA both as crystals and as a glass (Figure 7). After 240 min the crystal kept about 0.999, 0.998 and 0.997 of its mass at 130, 145 and 150 °C and about 0.980 at 155 °C; the glass kept about 0.9975, 0.996, 0.994, 0.990, 0.985 and 0.972 at 130, 135, 140, 145, 150 and 155 °C R8.a–f. The authors state that "at all temperatures, the amorphous form degrades at a faster rate", in line with Carstensen and Morris, and read the different curve shapes (a nearly straight decline for the crystal, a steepening one for the glass) as "a change in degradation mechanism" C19·C20.

Note by the AI writer (not discussed in the paper): at 155 °C, 5 °C below the melting point, the crystal actually lost slightly more mass than the glass during the first 90 min or so (about 0.32 against 0.26% at 30 min, about 0.59 against 0.55% at 60 min) before the glass overtook it (about 2.8 against 2.0% at 240 min) X9. The DSC test of IDM went only to 150 °C, so the state of the glass at 155 °C is not documented X17.

4. Polymers: HPMCAS and Eudragit EPO set the tightest limits

Figure 9 maps the isothermal mass loss of the six polymers over 130–200 °C and 4 h. The authors read it as "a guide for suitable HME processing conditions" for neat polymers C28. PVPVA, PVP K-30, Soluplus and HPMC stayed below 2% everywhere in this range; HPMCAS and Eudragit EPO lost more at high temperature and long hold, although suitable conditions "can be identified" for both C29. Table C gives readings from the maps.

Table C. Isothermal mass loss of the neat polymers, read from the contour maps of Figure 9 (approximate) R12.a–f.

HPMCAS is known to release acetic and succinic acid under harsh extrusion, with yellower extrudates (Sarode 2014) C30. The authors add a caveat of their own: mass loss "may not be" a complete indicator of polymer suitability, since processing also changes molecular weight, polydispersity, rheology and dissolution, and the link between TGA results and product performance is "worthy of future investigation" C31. Two side observations from the ramps belong here. PVP K-30 lost only about 2% between about 150 and 200 °C and nothing more of note up to 300 °C, so the literature onset of 171 °C (Gupta 2014) "likely does not reflect" its main degradation, which starts near 380 °C (Jablonski 2008) R9·R11·C26·C27. And at 5 °C/min the tangent onsets of the polymers (about 291, 175, 286, 213, 330 and 252 °C for PVPVA, PVP K-30, Soluplus, HPMCAS, HPMC and Eudragit EPO) were in "approximate agreement" with the Serajuddin series, except HPMC, whose literature ramps stopped at 300 °C R11·C24.

5. Inside the ASDs: whose mass is lost?

In the ramps (Figure 11) the ASD onsets also moved with rate and criterion C36. The authors write that PCZ ASD onsets were about 30 °C lower than those of crystalline PCZ and differed with the polymer, that IDM/PVPVA onsets were slightly higher than those of crystalline IDM, and that BCL/PVPVA onsets were about 30 °C lower than those of crystalline BCL C36. They explain the lower PCZ and BCL onsets by two effects: the polymer degrades on its own (lowest onsets 149 °C for HPMCAS and 166 °C for PVPVA) and the amorphous drug can be expected to degrade at lower temperature than its crystal C37. They suggest that PVPVA "may be more protective" to PCZ than HPMCAS and that PVPVA gives IDM "some protection", as reported for PVP with IDM by Vyazovkin and colleagues, where PVP could stabilize or destabilize depending on interactions and drug–polymer ratio C38·C39.

Note by the AI writer (not discussed in the paper): the printed ranges do not support a 30 °C drop for PCZ. Crystalline PCZ spans 285–376 °C, PCZ/HPMCAS 158–257 °C and PCZ/PVPVA 174–301 °C, drops of 75–127 °C; only BCL (193–255 against 163–227 °C) fits "about 30 °C" X2. The IDM comparison depends on the IDM/PVPVA low end, which Figure 11c puts near 153 °C rather than 175 °C X1. Crystalline-drug thresholds are referenced to a point just above $T_m$, ASD thresholds to 140 °C and to a sample that is mostly polymer, so the two kinds of onset are not on the same footing X6. Finally, the PCZ onset difference between the two polymers mirrors the neat polymers' own onsets, which is what the authors themselves later use to attribute the PCZ ASD mass loss to the polymer X18.

The isothermal maps of the ASDs (Figure 12) are the paper's main tool for this question. PCZ/HPMCAS lost much more than PCZ/PVPVA, which stayed below about 1% over the whole range C40. Comparing Figure 12a with Figure 9d and Figure 12b with Figure 9a, the authors write that "it is apparent that the level of degradation experienced by the formulation is driven by that of the polymer itself", so amorphous PCZ stabilized by a polymer is thermally stable over a wide range of exposures C41. HPLC of PCZ extrudates made at melt temperatures of 125–175 °C gave 100–105% assay with no impurity peaks beyond those of the standards, which the authors take as confirmation R16·C42. The IDM/PVPVA and BCL/PVPVA maps do not resemble that of PVPVA, so "their degradation must be primarily driven by the amorphous API" C43. IDM/PVPVA stayed below 1% at short times up to 200 °C and for extended times at 160 °C and below C44. BCL/PVPVA was the fastest system: below 1% only for hold times under 10 min at 180 °C, 1% after 40 min at 165 °C and after more than 100 min at 155 °C, so for this drug time and temperature "must be chosen carefully" C45·C46.

Table D. Isothermal mass loss of the ASDs, read from the contour maps of Figure 12 (approximate) R15.a–d.

Note by the AI writer (not discussed in the paper): the attribution of mass loss to polymer or drug rests on the shape of the maps alone; no drug assay after TGA holds was reported for the PCZ or IDM ASDs, and no solid-state check after the 4 h holds shows that the drugs stayed amorphous X11·X10. The IDM/PVPVA 1% contour lies at about 157 °C at 240 min, so at 160 °C about 1% is reached within the 4 h hold X19.

6. Mass loss and assay: the central comparison

BCL/PVPVA, the system with the most severe mass loss, was chosen to test TGA against HPLC. Figure 13 plots the mass remaining during the TGA holds at 180 and 165 °C together with the assay of samples taken out of the TGA after 10, 30, 60 and 240 min P13. Table E lists the values.

Table E. 30/70 BCL/PVPVA: TGA mass remaining and HPLC assay (mean ± SD of triplicate preparations where printed); other values read from Figure 13 (approximate) R17·R18.

At every time point the assay fell far below the mass remaining. After 10 min at 180 °C the sample had lost under 1% of its mass but its assay was 88.5%; after 240 min it had lost about 11% of its mass and about 42% of its drug R17·R18.a·R18.d. The authors conclude that "loss of volatiles does not predict the complete degradation profile of bicalutamide", as Huang and co-workers had found for gliclazide C47·R20.

Note by the AI writer (not discussed in the paper): mass loss is a percent of the whole sample, 70% of which is PVPVA, whereas assay is a percent of the drug, so the paper's single axis compares different bases. Even if every lost milligram were bicalutamide, the 10 min loss at 180 °C would correspond to only about 3% of the drug against 11.5% by assay, so the conclusion stands X7. The paper does not define the basis of the assay (nominal or initial content) X7.

7. The TGA pan was harsher than the extruder

Extrudates made at 180 and 165 °C for 10 min both assayed 95.0% (± 1.3 and ± 0.4%), against 88.5 and 92.63% for 10 min holds in the TGA R18.a·R18.e·R18.f·R18.j. The chromatograms (Figure 14) show the same degradant peaks, at about 4.9, 5.2, 5.9, 8.8 and 12.4 min, in both kinds of sample and none in the standard or the physical mixture; their total area was about 5-fold larger after the TGA hold (1.5 ± 0.3% against 0.3 ± 0.1% at 180 °C, 0.34 ± 0.10% against 0.07 ± 0.01% at 165 °C) R19. The authors call the extrudates "slightly less degraded" C48. They offer two reasons, both introduced with "may": the 50 °C/min ramp to the hold and the air cooling afterwards could make the time at high temperature "up to twice as long" as the hold, and the nitrogen purge could remove volatile products quickly and so accelerate the reaction, consistent with less mass loss under air than under nitrogen for gliclazide C49·R20. Because the extruder here put little mechanical energy into the melt, temperature was the main driver in both settings, and the authors find it "even more surprising and significant" that the two correlated so poorly C51.

Note by the AI writer (not discussed in the paper): the total degradant peak area (0.07–1.5%) is far smaller than the assay loss after 10 min (5.0–11.5%), and the paper does not discuss this mass balance; non-absorbing products at 270 nm, volatile products or insoluble residues would all fit X8. The TGA samples were solvent-evaporated ASDs while the extrudates started as a micronized-crystal mixture, the thermal histories differ, and one extrusion was run per temperature X12.

Discussion

What a TGA number can and cannot say. The paper's argument runs as a chain. Onsets move with heating rate, so time at temperature matters and the isothermal hold is the relevant test C11·C15. A neat glass may crystallize during that hold, so the drug is tested inside an ASD, which is also the worst case compared with a crystalline mixture C17·C33·C35·C55. Within the ASD, mass loss may come from the polymer or from the drug, and the paper reads the PCZ ASDs as polymer-driven and the IDM and BCL ASDs as drug-driven C41·C43. Even then, TGA records only the products that leave the pan, and for BCL most of the drug loss did not show up as mass loss C47·C53. The flow below summarizes how the reading of a TGA result splits.

On this basis the authors draw their main recommendation: TGA "can provide a guide" to which temperatures and formulations are susceptible to degradation, but "acceptable processing windows in terms of temperature range should be defined by assay experiments" C54. The abstract adds that TGA methods "should be selected to generate highly sensitive outcomes" and that "caution should be applied" when extrapolating them C54. If a system were also degraded by mechanical input, predictions from TGA mass loss alone "would be far from sufficient"; in larger continuous extruders, higher specific mechanical energy and local heating can add to the thermal load C50·C52.

Scope. The comparison of mass loss with assay covers one drug and one polymer, 30/70 BCL/PVPVA, at two temperatures; the PCZ and IDM extrudates were assayed but not compared with TGA holds R16·R18. The extruder was a 5 mL recirculating lab device run at 50 rpm with minimal mechanical input P4·C6. All TGA runs were under nitrogen P8·P9.

What it suggests for formulation (the AI writer's reading, not stated in the paper). A mass-loss limit applied to an ASD should at least be converted to a drug basis by dividing by the drug fraction, and even then treated as an underestimate: for 30/70 BCL/PVPVA, under 1% mass loss went with 11.5% drug loss R17·R18.a·X7. The onset spread is wide enough that the same material can look safe or unsafe at a given extrusion temperature depending only on the test settings, so an onset quoted without its heating rate and criterion says little R5. In a product with 75% HPMCAS, the polymer itself (onset as low as 149 °C, above 5% mass loss after about 40 min at 200 °C) may limit the window before the drug does R5.g·R12.d. An early HPLC check at the intended melt temperature and residence time, with a mass balance, is the direct answer the paper points to C54.

Limitations

Stated by the authors. Mass loss of volatiles "is not equivalent to formation of degradants", which "significantly impedes" TGA as a consistent indicator of degradation C53. Isothermal heating translates to HME only "assuming degradation is not accelerated due to the shear inputs or localized heating" C15. Polymer mass loss "may not be a comprehensive indication" of suitability, and its relation to molecular weight, polydispersity and product performance needs further work C31. The mechanism by which polymers stabilize or destabilize drugs is "complex" C39. The explanations for the harsher TGA exposure are offered with "may" and were not tested C49.

The AI writer's own judgement. The method dependence of TGA onsets is shown directly for 13 materials and is robust R4·R10·R14. The central warning, that mass loss underestimates drug loss, is large and reproducible at two temperatures and four times, but rests on one drug–polymer pair, compares quantities on different bases and leaves most of the lost drug unaccounted for by the degradant peaks R18·X7·X8. The attribution of ASD mass loss to polymer or drug is qualitative, and the solid state of the ASDs after the holds was not checked X11·X10. Several summary statements in the text do not match the figures: the IDM/PVPVA low onset (175 °C printed, about 153 °C in Figure 11c), the "about 30 °C" drop for the PCZ ASDs (75–127 °C in the printed ranges), the more than 10% mass loss at the tangent for every drug (about 9.7% for PCZ at 1 °C/min), and the abstract's onset range of 52–170 °C, which is a range of spreads X1·X2·X3·X14. The number of replicates behind the nonisothermal onsets is not stated, and the supporting information with the PCZ and IDM extrusion data was not available for this reading X15·R16.

Conclusion

The paper takes apart a routine step of HME development. A nonisothermal TGA onset is not a property of the material: for three drugs, six polymers and four ASDs it moved by 52–170 °C with heating rate and criterion, and the tangent onset that instrument software reports already meant about 10–21% mass loss for the crystalline drugs R5·R6. Isothermal TGA is the better test, provided the drug is held amorphous in a polymer, and it ranks the polymers sensibly: PVPVA, PVP K-30, Soluplus and HPMC lose under 2% over 4 h up to 200 °C, HPMCAS and Eudragit EPO more than 5% C55·R12. But for bicalutamide in PVPVA, a 10 min hold at 180 °C that cost under 1% of the mass cost 11.5% of the drug, and the TGA pan degraded the drug more than a 10 min extrusion R17·R18·R19. The authors' conclusion follows: consider TGA mass loss "only an indicator of degradation" and define the processing window by assay C53·C54.