Escrever um comentário

Summary and contribution

A bioresorbable intra-arterial device (Mg microwire core, PGS drug reservoir) in helical and linear geometries, loaded with dexamethasone or cisplatin by post-cure soaking or co-curing. In vitro: 30-day release (373 µg DEX with 73% released on day 1; 65 µg CIS with no burst), Mg²⁺ efflux near media baseline, HUVEC viability >90%, released CIS active against F98 glioma. In vivo (rat, day 7): devices in the left renal or right common carotid artery gave far higher target-organ than serum DEX, with NODL values of 29 and 109 (kidney cohorts 1, 2) and 68 (brain), and arterial H&E showing patent lumens.

The contribution is genuine and well framed: clinical IA delivery is bounded by catheter dwell time, so regional dosing is intrinsically acute. A resorbable upstream depot decouples regional delivery from the procedure and needs no retrieval, and the concept held in two organs with very different extraction and barrier properties. The gap is between that concept and the evidence offered for it.

Major issues

• Cumulative release standard deviations shrink over time, which is impossible. Fig. 3A shows the DEX cumulative error bar spanning roughly ±92 µg at day 1 and collapsing to ±1.41 µg by day 30. A running sum accumulates variance; its SD cannot fall below the SD of any constituent day across independent devices. CIS at ±0.06 µg (0.09% CV) is likewise not credible for five separately fabricated devices. Either the dispersion is computed on replicate injections of pooled media rather than n=5 devices, or the values are mislabeled. This must be resolved first — it governs every release claim and the batch-consistency assertion in §3.1.

• Three of four characterization methods do not actually confirm drug loading. In Fig. 2A the DEX-IADD, CIS-IADD, and blank IADD TGA traces are superimposed; the text concedes DEX is indistinguishable, and the claimed 3% versus 0% residue for CIS is within trace noise. FTIR likewise shows no resolvable difference for DEX. That leaves EDS, where F (0.6 at.%), Pt (0.1 at.%), and Cl (0.2 at.%) sit at or below the practical EDS detection floor — and inferring stoichiometric Pt:Cl from two values at that floor is over-interpretation. The abstract's claim that these methods confirmed incorporation and stability is not supportable as written.

• The 109-fold headline is imputed, and the cohort comparison it anchors is invalid. Serum was below LLOD and 0.97 ng/mL was substituted into the NODL denominator (Eq. 2). It is then compared against cohort 1's measured 29 to infer relief of renal uptake saturation — but a smaller device drives serum below detection almost by construction, and absolute target delivery fell four-fold (606 → 154 ng/mL). Compounding this, the cohorts differ simultaneously in geometry, fabrication method, and payload, and their oral-control levels differ ~5.8-fold under a nominally identical regimen (kidney 11.88 vs 68.39; serum 8.08 vs 47.02 ng/mL). Remove the saturation mechanism and do not lead with 109-fold.

• Helical device hemodynamics confound cohort 1. A 0.45 mm lumen inside a 1.1 mm artery is an ~83% reduction in flow cross-section (~35× Poiseuille resistance). Since Dedrick's framework — which §4.3 invokes as confirmation — holds that low regional flow amplifies regional advantage, much of cohort 1's ratio may be iatrogenic stenosis rather than delivery performance. No perfusion, renal function, or downstream parenchymal data are presented.

• No efficacy endpoint, and no target-organ histology anywhere. Fig. 9 shows arterial sections only; kidney and brain parenchyma are never examined, despite embolic infarction being the dominant hazard of the concept. There is also no sham-surgery or empty-device group, so route is confounded with laparotomy, clamping, 15–20 min of arterial occlusion, and arteriotomy. "Proved the safety and efficacy" should be replaced with language describing focal biodistribution.

• Statistics do not match the design, and one claim is contradicted by the figures. §3.5 states all three cohorts showed significantly higher target-organ levels, but Fig. 6B carries no significance markers and n=2 cannot support inference. Shapiro–Wilk at n=2–3 has no power, so the parametric-assumption statement is vacuous. Comparing multiple compartments from the same animals by one-way ANOVA (Fig. 7B) treats within-animal measurements as independent — pseudoreplication requiring a mixed model. No power analysis, randomization, or blinding; ARRIVE 2.0 not met.

• Fig. 10J undermines the retention claim it is cited for. Fluorine appears at 0.3–0.5 at.% in the non-implanted control regions and 0.4 at.% in explanted ones — indistinguishable. The text reads this as retention of DEX, but equal signal in both means EDS cannot resolve loss, so the panel supports neither retention nor release. These are also single-spot regional analyses with no replicates or statistics. Combined with the absence of any direct assay of loaded or residual drug, delivered dose, loading efficiency, and mass balance remain unknown, so nothing in the paper is dose-normalized.

• A single day-7 terminal timepoint cannot support a 30-day sustained-release claim in vivo. With 73% of the DEX payload gone on day 1 in vitro, the day-7 snapshot is equally consistent with a largely depleted device and a decaying tail. No Cmax, no AUC, no time course.

• Serum below LLOD alongside a contralateral kidney at 38× LLOD is internally contradictory — the only route to the right kidney is systemic. Separately, tissue values are reported as ng/mL from 10% w/v homogenate supernatant while serum is a true volumetric concentration; the ~10× convention gap is never reconciled, so the direct fold-changes (42-fold, 11-fold) are not interpretable as stated. NODL, being a ratio of ratios, is robust to this; the individual claims are not.

• The glioma result is normalized against the wrong control, and the control itself contradicts the degradation narrative. In Fig. 5B the blank-device media drive viability from ~110% to ~180% over 30 days. The 38% reduction is computed against untreated cells rather than the same-day vehicle, which understates the CIS effect; more importantly, a monotonically rising control implies progressively increasing PGS degradation-product release, which conflicts with the flat Mg²⁺ profile used as evidence of stable erosion. That Mg²⁺ assay is in any case insensitive by design — a small increment measured against a 31.6–36 mg/L media baseline — so "no burst degradation" is an underpowered null, not evidence of stability. Relatedly, §4.2 attributes post-burst release to steady surface erosion while the same data show essentially no measurable erosion for 23 days; diffusion from the matrix is the more parsimonious mechanism.

• Two miscitations, one substantive. Refs [46] and [47] are cited in §4.3 to support intra-arterial restriction of drug distribution, but both concern intra-articular injection for osteoarthritis — [47] is a bibliometric analysis. Ref [18] is cited for Mg²⁺ enhancing blood flow and inhibiting platelet activation in vascular stents, but is a study of oral magnesium supplementation in hypertensive patients and does not address either claim.

• The endovascular premise was never demonstrated. Every implantation was open arteriotomy (laparotomy or neck cutdown), yet minimally invasive fluoroscopy-guided placement is asserted in the Introduction, §4.1, and Conclusion. Carotid retention is attributed to inherent tissue adhesion with no adhesion, pull-out, migration, or imaging data — and Fig. 9D notes the device dislodged during processing. Both gaps belong in the text as limitations, not as capabilities.

• CMC gaps for what is functionally a combination product. No content uniformity, coating-thickness distribution, or batch-to-batch data despite the consistency claim in §3.1; sterilization deferred entirely to the supplement with no drug-stability data under it; no endotoxin. DMF used for CIS loading raises an ICH Q3C Class 2 residual-solvent issue that a DMF-soaked-device viability assay does not address — headspace GC would close it. Finally, CIS has no in vivo data at all, so the two-drug platform framing should be softened.

Minor issues

• Fig. 4 panels A and B are swapped relative to both the legend and §3.3: panel A shows ethanol/DMF (%v/v), panel B shows DEX/CIS (µg/mL).

• Fig. 6A labels a 0.4 mm annular thickness, inconsistent with the stated 0.45 mm lumen in a 1 mm device (which implies 0.275 mm).

• Fixation is described two ways: 4% paraformaldehyde in PBS (§2.8.1–2.8.2) versus 10% neutral buffered formalin (§2.8.3). The OCT-embedding-then-resinous-mount workflow is also internally inconsistent, and cryosectioning likely explains the dislodged carotid device.

• Oral dose units conflict: mg/mL in §2.8.1, §2.8.2, and §3.5.1 versus mg/L in §3.5. Water intake was not measured, so consumed dose is unknown.

• §2.7 cross-references itself for the staining method; should point to §2.6. §2.6 gives no separate n for DEX- versus CIS-loaded devices.

• §3.1 abbreviates chlorine as "C" in a sentence where C denotes carbon.

• Fig. 3A legend reads "cumulative … per day," which is self-contradictory.

• Adrenal suppression is asserted from refs [41–43] but never measured. Serum corticosterone in both arms would substantiate reduced systemic exposure pharmacodynamically — a low-cost, high-value addition.

• Define NODL in the abstract and state explicitly that it is a ratio of ratios; report tissue concentrations as ng/g with the homogenate correction stated.

• Computing the theoretical Dedrick advantage (Rd = 1 + CL/Q) for kidney and brain and comparing it against observed values would let readers judge how much of the achievable benefit was captured — a substantial strengthening of §4.3.

• Typo "renal update" (§3.5.1); IADD/I-ADD inconsistent between text, Limitations, and Figs. 3C/3D and 5C.

• Broken and dual right-hand axes in Figs. 1E and 5C impede reading.

• No data availability statement. Refs [4], [7], [9], [28], [38], [40], and [46] lack volume or page ranges.

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.