Rescuing the Aging Blood Vessel: Old Arthritis Drugs and Viagra-Class Compounds Show Preclinical Promise, But the Evidence Base Is Thin

News
#1

BloodVessels
BloodVessels2752×1536 425 KB

A Dutch team at Erasmus MC has pooled every usable animal study on drugs that target the nitric oxide to cyclic GMP (NO-cGMP) signalling axis in aged or prematurely aged animals. Thirteen studies survived screening out of more than 18,000 records. Three drug classes were tested: anti-inflammatory antirheumatic drugs (DMARDs), phosphodiesterase (PDE) inhibitors, and soluble guanylate cyclase (sGC) modulators. DMARDs produced a very large improvement in endothelium-dependent vessel relaxation. PDE inhibitors and sGC modulators did not improve endothelium-dependent relaxation but did improve smooth-muscle-dependent relaxation and microvascular blood flow. The overall picture is that upstream and downstream halves of the same pathway respond to different drugs, arguing for combination approaches.

Blood vessels age badly. They stiffen, they stop dilating properly, and they leak inflammation. Most of the clinical effort so far has gone into scraping cholesterol plaque out of the equation with statins, but a large share of cardiovascular risk in older people has nothing to do with plaque. It is the vessel wall itself losing the ability to relax on command.

The command signal in question is nitric oxide. The endothelial lining makes it, it activates an enzyme called soluble guanylate cyclase in the muscle layer beneath, that enzyme makes cyclic GMP, and cyclic GMP tells the muscle to let go. Aging degrades this chain at every link: less nitric oxide made, a damaged and unresponsive receptor enzyme, and more phosphodiesterase enzyme chewing up the cyclic GMP before it can act.

Jouabadi and colleagues asked a simple question of the animal literature. If you pharmacologically attack any one of those three links in an old animal, does the vessel work better? They ran three separate systematic searches, all preregistered, and found thirteen studies.

The results split along a mechanistic fault line that is the most interesting thing in the paper. Old anti-inflammatory drugs used for rheumatoid arthritis and lupus, methotrexate, hydroxychloroquine, gold and penicillamine, produced the single largest effect on endothelium-dependent relaxation, the standard readout of endothelial health. They work upstream, by lowering the inflammatory and oxidative burden that cripples nitric oxide production in the first place.

Drugs acting downstream did the opposite. Sildenafil and its relatives, and sGC activators, failed to improve endothelium-dependent relaxation, which makes sense because they bypass the endothelium entirely. But they did improve microvascular perfusion, actual blood flow through small vessels, and smooth-muscle relaxation.

So the aging vessel appears to have two separable lesions: a broken signal source and a leaky signal amplifier. Fixing one does not fix the other. The authors suggest that combining anti-inflammatory upstream treatment with downstream cyclic nucleotide support may be additive, though no included study actually tested that combination.

The honest verdict is that this is a mapping exercise, not a demonstration of efficacy. Thirteen studies is very few. Only two used naturally aged animals, both rats. The rest used progeroid or disease models. No study measured arterial stiffness by pulse wave velocity, the measure that actually predicts human outcomes.

Insights

First, the size of the reported benefits. The effect sizes are given as standardized mean differences, which express the gap between treated and control groups in units of the natural spread within each group. The DMARD figure of 2.55 means that if you picked one treated aged animal and one untreated aged animal at random, the treated one would have better vessel relaxation about 96 times out of 100. For microvascular perfusion pooled across all three drug classes, the figure of 0.89 corresponds to winning about 74 times out of 100. Those are the honest translations. They are also almost certainly inflated by tiny group sizes.

Second, the mechanistic lesson is more portable than the drugs. Downstream drugs did not repair endothelial function. If endothelial nitric oxide output is your target, upstream inflammation control is where the signal was, not cyclic GMP support.

Third, of the compounds tested, PDE5 inhibitors are the only class where the animal doses translate to roughly standard human clinical doses, and they have a long safety record.

Context and Source

  • Open Access Paper: Targeting the NO-cGMP axis in age-related vascular dysfunction: a systematic review and meta-analysis of preclinical animal studies
  • Institutions: Division of Pharmacology and Vascular Diseases, Department of Internal Medicine, Erasmus MC University Medical Center, Rotterdam, Netherlands; Erasmus MC Medical Library, Rotterdam; Department of Anesthesiology, Pain and Palliative Care, Radboud University Medical Centre, Nijmegen, Netherlands
  • Country: The Netherlands
  • Journal: GeroScience (Springer, official journal of the American Aging Association)
  • Impact evaluation: The impact score of this journal is 6.0 (2025 Journal Impact Factor, up from 5.7 the prior year; CiteScore approximately 8.3), evaluated against a typical high-end range of 0 to 60+ for top general science, therefore this is a Medium impact journal.

Related Reading:

#2

Eleven distinct molecules were actually tested across the 13 studies.

DMARDs — 4 molecules, 5 studies, 6 arms

Drug Study Model Dose Duration
Methotrexate (gold nanoparticle conjugate) Rafik 2022 Rat, adjuvant arthritis 0.04 mg/kg/d 4 wk
Gold (nanoparticles) Rafik 2022, separate arm Rat, adjuvant arthritis 0.04 mg/kg/d 4 wk
Methotrexate (lipid core nanoparticles) Guido 2022 Mouse, Marfan 0.15 mg/kg/d 12 wk
Hydroxychloroquine Virdis 2015 Mouse, SLE 3 mg/kg/d 18 wk
Hydroxychloroquine Gómez-Guzmán 2014 Mouse, SLE 10 mg/kg/d 5 wk
D-penicillamine Shukla 2005 Rabbit, hyperhomocysteinemia 10 mg/kg/d 4 wk

Worth noting: the two big positive arms come from one experiment, and the gold is nanoparticulate gold, not conventional gold salt therapy. The one study that used a non-inflammatory model with a structural endpoint (Guido, Marfan, aortic dilation) was the null result, and it was excluded from the pooled estimate.

PDE inhibitors — 3 molecules across 3 isoform subclasses, 5 studies

Drug Isoform Study Model Dose
Sildenafil PDE5 Golshiri 2020 Ercc1 mouse 10 mg/kg/d, 8 wk
Sildenafil PDE5 Zhang 2005 Naturally aged rat, 18 mo 2 mg/kg/d, 1 wk
Lenrispodun (ITI-214) PDE1 Golshiri 2021 Ercc1 mouse 100 mg/kg/d, 8 wk
Lenrispodun (ITI-214) PDE1 Golshiri 2022 SMC-specific Ercc1 mouse 40 mg/kg/d, 12 wk
Cilostazol PDE3 Moreira 2018 Naturally aged rat, 15 mo 100 mg/kg/d, 8 wk

Only 3 of the 11 known PDE subfamilies have ever been tested here.

sGC modulators — 3 molecules, 3 studies, 4 arms

Drug Mechanism Study Model Dose
BAY 54-6544 Activator, heme-independent Ataei Ataabadi 2022 Ercc1 mouse 200 mg/kg/d in chow, 8–9 wk
BAY 54-6544 Activator Potoka 2018, arm 1 BERK sickle cell mouse 17 and 30 mg/kg/d, 4 wk
BAY 41-8543 Stimulator, heme-dependent Potoka 2018, arm 2 BERK sickle cell mouse as above
Praliciguat (IW-1973) Stimulator, heme-dependent Foussard 2023 Lepr db/db mouse 10 mg/kg/d, 4 wk

One data-integrity problem I found while compiling this. Table 1 and the page-5 text name Ataei’s compound as BAY 58-2667 (cinaciguat). That is wrong, and the review contradicts itself: its own page-9 discussion says Ataei used BAY 54-6544, the primary source is literally titled after BAY 54-6544, and the 200 mg/kg/day dose matches that paper exactly. So cinaciguat was almost certainly never tested in any included study, yet it appears twice in the review as though it was. I’d verify the Potoka compound assignments in the primary source before relying on them.

Two incidental findings from that verification. The Ataei paper reports 33 percent mortality in vehicle-treated Ercc1 mice versus 0 of 12 on BAY 54-6544 (p < 0.001), which is the only mortality data anywhere in this evidence base and the meta-analysis did not extract it. And the exclusion criteria removed any study where the drug was used in combination with another treatment, which means the paper’s central recommendation to combine upstream and downstream targeting was structurally excluded from its own evidence base by design.

#3

Biomarker Data (Effect Size Extraction)

How to read these numbers. A standardized mean difference of 1.0 means the average treated animal sits one full standard deviation above the average control animal. Two useful plain-language conversions follow from it. “Probability of superiority” is the chance that a randomly picked treated animal beats a randomly picked control. “Percentile” is where the average treated animal would rank inside the control distribution. Both are given below.

Drug class Outcome SMD [95% CI] n studies I squared Probability of superiority Average treated animal ranks above this % of controls Statistically significant
DMARDs ACh-induced vasorelaxation 2.55 [1.68, 3.42] 5 8.8% 96% 99.5% Yes
PDE inhibitors ACh-induced vasorelaxation 1.54 [-0.80, 3.88] 5 81.1% 86% 94% No
PDE inhibitors SNP-induced vasorelaxation 0.50 [0.04, 0.96] 5 20.6% 64% 69% Yes, marginally
PDE inhibitors Microvascular perfusion 0.76 [0.28, 1.25] 4 0% 70% 78% Yes
sGC modulators ACh-induced vasorelaxation 0.45 [-0.23, 1.13] 3 1.7% 62% 67% No
sGC modulators Microvascular perfusion 1.12 [0.46, 1.78] 2 0% 79% 87% Yes
All three pooled ACh-induced vasorelaxation 1.73 [0.63, 2.82] 13 76.1% 89% 96% Yes
All three pooled Microvascular function 0.89 [0.49, 1.28] 6 0% 74% 81% Yes

Individual study effect sizes worth flagging, from the forest plots:

  • DMARDs on ACh relaxation: Shukla 2005 penicillamine 3.44 [1.43, 5.44]; Rafik 2022 gold 3.13 [0.79, 5.48]; Virdis 2015 HCQ 2.98 [1.44, 4.51]; Rafik 2022 methotrexate 2.91 [0.68, 5.15]; Guzman 2014 HCQ 1.67 [0.72, 2.63]. Direction is consistent across all five, which is the strongest internal-consistency signal in the paper.
  • PDE inhibitors on ACh relaxation: Moreira 2018 cilostazol 7.78 [4.41, 11.14] versus Golshiri 2022 lenrispodun -0.01 [-0.85, 0.83]. The cilostazol result is an extreme outlier and is single-handedly responsible for both the large point estimate and the 81% heterogeneity. Remove it and the class effect on endothelium-dependent relaxation is essentially nil.
  • sGC modulators on ACh relaxation: Potoka 2018 sGC stimulator -0.73 [-2.55, 1.09], directionally harmful, though the interval spans zero widely.

Absolute magnitude interpretation and a warning. The paper reports no absolute values, no percentage-point change in maximum relaxation, no millimetres of mercury, no perfusion units. This matters. An SMD of 2.55 on a vasorelaxation curve, given that ex vivo wire myography in aged rodents typically shows within-group standard deviations of roughly 8 to 15 percentage points of maximal relaxation, would imply an absolute improvement on the order of 20 to 38 percentage points of relaxation capacity. That is close to full functional restoration of an aged vessel by an old antirheumatic drug, which is biologically implausible at face value. The more likely explanation is unstable standard deviation estimates from group sizes near 5, which inflate SMD.