Category: 4. Technology

---

• Microsoft has revealed how it protects Azure with an Integrated HSM chip
• Azure security stack includes Azure Boost, Hydra BMC and Caliptra 2.0
• Cybercrime reportedly worth $10.2 trillion annually, making it the world’s third-largest economy in 2025

---

Microsoft has revealed more on the custom-built security chip it deploys across every Azure server, aiming to counter what it calls a cybercrime “pandemic” now costing $10 trillion annually.

The Azure Integrated HSM, which was first announced in late 2024, is the centerpiece of a wider security architecture the company outlined at the recent Hot Chips 2025 event.

A slide Microsoft showed there claims the global cost of cybercrime is currently $10.2 trillion – meaning it now ranks as the equivalent of the third-largest economy in the world.

Azure security measures

VPN users are overwhelmed with choice, and there are as many bad options out there as there are good ones. Luckily, NordVPN sits in the latter category, and right now Nord is offering discounted plans across its various tiers. If you take out a two-year NordVPN Plus plan (the company’s most popular plan) it’ll cost you $108 for the duration of the contract, with Nord throwing in three extra months at no extra cost. That’s 73 percent off the usual rate.

As well as Nord’s VPN service, a Plus plan also includes the Threat Protection Pro anti-malware tool, password management and an ad- and tracker-blocker. A Prime plan additionally comes with encrypted cloud storage or NordProtect, which insures you against identity theft and monitors dark web activity. That’s also on sale — down to $189 on the same two-year commitment with those three additional months thrown in, which works out to a 77 percent savings on the regular price.

When Engadget’s Sam Chapman reviewed NordVPN earlier this year, he praised its excellent download speeds, exclusive features and extensive server network. Less impressive is its clunky interface and inconsistent design when jumping between different platforms running a NordVPN app. While it doesn’t quite make the cut in our guide to the best VPNs available right now, it generally performed well in speed tests and Threat Protection Pro is really worth having.

Follow @EngadgetDeals on X for the latest tech deals and buying advice.

Elyse Bennett

Apple made it official – the iPhone 17 series’ announcement, among is coming on September 9. At the event, Apple will unveil the iPhone 17, iPhone 17 Air, iPhone 17 Pro, and iPhone 17 Pro Max. The tagline Apple chose for this event is “awe-dropping” – get it?

The Galaxy S26 Ultra’s battery capacity was tipped in a certification filing, and it’s the same as the Galaxy S25 Ultra’s – a rated capacity of 4,855mAh, which translates to a typical capacity of 5,000mAh. At least we could still hope for faster 65W charging.

Realme showed us two concepts during its 828 Fan Festival in China. The first is a smartphone with a 15,000mAh battery, advertised to offer up to five days of usage on a single charge. It gets a whopping 1,200 Wh/L energy density, and it helps achieve some unheard-of running times.

The second concept has a 10,000mAh battery and comes with a built-in cooling fan and thermoelectric cooler, which Realme claims can reduce temperatures by up to 6°C.

The Galaxy S25 FE is coming on September 4 and we know the specs, and the price. It has a 6.7-inch 1080×2340 Dynamic AMOLED 2X screen with 120Hz refresh rate and Gorilla Glass Victus+ on top, the Exynos 2400 chipset at the helm, 8GB of RAM, and 128GB of storage.

It has a 50 MP main camera with OIS, a 12 MP ultrawide, an 8 MP telephoto with 3x optical zoom and OIS, a 12 MP selfie camera, an ultrasonic in-display fingerprint sensor, and a 4,900 mAh battery with support for 45W wired and 15W wireless charging. It measures 161.3 x 76.6 x 7.4 mm and weighs 190g.

The rumored price of the Galaxy S25 FE is $649.99 with 128GB of storage, and $709.99 with 256GB

Xiaomi unveiled HyperOS 3, based on Android 16. It brings a slight visual revamp with new icons, a new status bar, and the addition of Super Island. Xiaomi shared a list of devices eligible for the HyperOS 3.0 Beta program. The stable HyperOS 3.0 version is expected to arrive sometime in the fourth quarter of this year.

The Galaxy A07 4G is also official with a Helio G99 and 6 years of OS updates. The phone is up for sale in Indonesia, starting at IDR 1,400,000 for the base 4/64GB model.

If there’s one thing I’ve learned after building a self-hosted network stack, it’s that DIY routers tend to pack extra features, better web UIs, and superior security provisions than the ones shipped by ISPs. Heck, I’d been using a Raspberry Pi-powered OpenWRT setup for a long time before making the transition to an OPNsense router built from a mini-PC last year. Speaking of OPNsense, the router OS recently received a major update, but judging from the community posts, it has some quirks that need ironing out.

But just moth to a flame, I couldn’t resist the urge to try setting up the newest version of OPNsense. Since I didn’t want to put my primary DIY OPNsense router in harm’s way, I figured it’d be a good idea to install the newest update on an old PC and resurrect it as a makeshift router. What I didn’t anticipate were the sheer compatibility issues with using a Marvell AQtian Ethernet adapter with OPNsense and an old accessory that would help bring this project to fruition when all hope seemed lost.

Setting up OPNsense

It went smoothly until I got betrayed by my PCIe NIC

I’ve configured OPNsense a couple of times already, so I began this project expecting it wouldn’t take more than an hour. After going through the usual procedure of downloading the image, writing its contents to a flash drive using Balena Etcher, and switching the BIOS on my old PC to boot from the USB drive, I plugged my PCIe Ethernet card into the system.

For the uninitiated, OPNsense requires two Ethernet interfaces: a WAN port that connects to a modem or an ISP device, and a LAN interface that lets my devices access the Internet. My B450 motherboard already includes a built-in 1GbE port, so all I needed was another Ethernet controller. Since I had a spare TP-Link TX401 Ethernet card lying around, I figured I could leverage it as the LAN port and connect it to my 10GbE switch – which turned out to be a big mistake.

Now, I’m fully aware that the Marvell AQtian controller isn’t ideal for OPNsense, but I didn’t have a single Mellanox card on hand. Plus, seeing how I’ve had zero problems using the same TX401 card with Proxmox, TrueNAS, and other Linux platforms, I thought it would work with OPNsense (or rather, FreeBSD) just as well – and any driver issues would be resolved after an update.

Unfortunately, my hopes were dashed when OPNsense failed to detect the TX401 NIC. Since I only had the built-in Ethernet controller (which I planned to use as the WAN interface), I couldn’t just configure everything from my PC over an SSH connection. Well, I could technically run some commands using the OPNsense shell, but I didn’t want to spend minutes typing multi-line scripts and commands manually.

My USB-to-Ethernet adapter came in clutch

Since the PCIe network card was borderline useless in its current state, I decided to put it on the backburner until after I finished installing OPNsense. That’s when I remembered I had a spare USB-to-Ethernet adapter that could serve as the temporary LAN interface.

Interestingly, the TP-Link USB Ethernet dongle was detected by OPNsense immediately. So, I quickly configured both the WAN and LAN interfaces and ran the installer command inside the live boot version of the distro. Afterward, I picked the Keymap settings and Installation directory before waiting for the wizard to perform its magic. Soon, OPNsense was installed on my old PC, and it was time to give my NIC another shot.

Configuring OPNsense using the web UI

All my attempts at installing the NIC drivers were in vain

With the OPNsense web interface finally accessible, I put on my troubleshooting hat. Since I wanted to run some terminal commands, I enabled Secure Shell Login and Root Login inside the Administration Settings of the System tab. I know it’s far from secure, but I wanted to avoid running into privilege issues when configuring OPNsense to detect my TX401 Ethernet card.

The first step was installing TX401 drivers, though it was a lot easier said than done. The driver package I wanted to install was aquantia-atlantic-kmod, and every repository link to it was broken – be it the outdated 0.0.5_2 version or its (comparatively) new 0.0.5_3 variant. In the end, I ditched FreeBSD repos and ran the fetch https://pkg.fr.ghostbsd.org/stable/FreeBSD:13:amd64/latest/All/aquantia-atlantic-kmod-0.0.5_2.pkg command to pull the necessary package from GhostBSD’s repo before installing it with pkg install aquantia-atlantic-kmod-0.0.5_2.pkg.

I also installed the nano editor, used it to open the /boot/loader.conf.local file, and added the if_atlantic_load=”YES” line before restarting the router. Unfortunately, the driver refused to work even after all my efforts, and it was time to put the Ethernet card back into my cupboard, where it belonged.

Nevertheless, I managed to get my custom-built router up and running

With the PCIe card out of the way, I had no choice but to use the USB-to-Ethernet adapter. So, I added the usual traffic rules to the Firewall tab, set up a new IPv4 gateway, and configured its built-in IPS. While I plan to buy a Mellanox card simply because I’d rather rely on a dedicated PCIe NIC, the USB adapter isn’t bad per se. I’d used a similar setup on my OpenWRT Raspberry Pi firewall for months, and it didn’t give me too much trouble.

Preparation of SnFe₂O₄/SiO₂/PM-Pd

The synthesis of SnFe₂O₄@SiO₂ began with the production of SnFe₂O₄ following well-established protocols. Subsequently, 1 g of SnFe₂O₄@SiO₂ was dispersed in 50 mL of toluene through 30 min of sonication. To this dispersion, 3 mL of 3-aminopropyltrimethoxysilane (APTES) was added, and the mixture underwent reflux for 24 h. The resulting SnFe₂O₄/SiO₂/PM nanoparticles were purified by washing four times with ethanol, separated via magnetic decantation, and dried at 50 °C. In the next phase, 2 g of SnFe₂O₄/SiO₂/PM was dispersed in 40 mL of toluene using a similar 30-minute sonication process, followed by the addition of 3 mmol of dicyclohexylcarbodiimide (PM). The mixture was then stirred under reflux conditions for another 24 h. The obtained nanoparticles were similarly washed multiple times with ethanol, magnetically separated, and dried at 60 °C. To graft Pd onto the heterogenized ligand, 1 g of SnFe₂O₄/SiO₂/PM was reacted with 1.5 mmol of Pd(OAc)₂ in CH₃CN at room temperature for 24 h. The SnFe₂O₄/SiO₂/PM-Pd nanoparticles were isolated by filtration, thoroughly cleaned with hot water and hot ethanol to remove any residual Pd, and subsequently dried at 60 °C. The overall procedure is represented schematically in Fig. 1.

General procedure for the suzuki reactions

A combination of aryl halide (1.0 mmol), phenylboronic acid (1.0 mmol), K₂CO₃ (1.5 mmol), and 30 mg of SnFe₂O₄/SiO₂/PM-Pd catalyst was stirred in 3 mL of PEG-400 at 100 °C. The reaction progress was monitored using TLC, and upon completion of the cross-coupling reaction, hot distilled water was added to the mixture. The catalyst was then separated by filtration, and the product was further purified. Extraction of the desired product was carried out using a 1:1 mixture of water and ethyl acetate (Fig. 2).

Selected NMR data

4-methyl-1,1’-biphenyl (Table 2, Entry 3) (Figure S1)¹:H NMR (400 MHz, DMSO): δ_H = 7.4 (m, 4 H), 7.1 (m, 5 H), 2.0 (s, 3 H), ppm.

3-methyl-1,1’-biphenyl (Table 2, Entry 9) (Figure S2)¹:H NMR (400 MHz, DMSO): δ_H = 7.7 (m, 3 H), 7.5 (m, 2 H), 7.4 (m, 4 H), 1.9 (s, 3 H), ppm.

4-nitro-1,1’-biphenyl (Table 2, Entry 11) (Figure S3)¹:H NMR (400 MHz, DMSO): δ_H = 8.2 (s, 1 H), 7.5 (s, 1 H), 7.3 (s, 3 H), 7.1 (s, 3 H), 7.0 (s, 1 H), ppm.

4-nitro-1,1’-biphenyl (Table 2, Entry 5) (Figure S4)¹:H NMR (400 MHz, DMSO): δ_H = 8.1 (s, 2 H), 7.3 (m, 7 H), ppm.

1,1’-biphenyl (Table 2, Entry 12) (Figure S5)¹:H NMR (400 MHz, DMSO): δ_H = 7.7 (s, 3 H), 7.4 (s, 4 H), 7.2 (s, 3 H), ppm.

1,1’-biphenyl (Table 2, Entry 10) (Figure S6)¹:H NMR (400 MHz, DMSO): δ_H = 7.2 (m, 7 H), 6.9 (s, 3 H), ppm.

4-methoxy-1,1’-biphenyl (Table 2, Entry 13) (Figure S7)¹:H NMR (400 MHz, DMSO): δ_H = 7.0 (m, 1 H), 6.9 (s, 3 H), 6.5 (m, 5 H), 4.0 (s, 3 H), ppm.

4-methoxy-1,1’-biphenyl (Table 2, Entry 4) (Figure S8)¹:H NMR (400 MHz, DMSO): δ_H = 8.3 (m, 1 H), 7.9 (m, 1 H), 7.6 (s, 3 H), 7.1 (m, 4 H), 4.3 (s, 3 H), ppm.

3-methoxy-1,1’-biphenyl (Table 2, Entry 6) (Figure S9)¹:H NMR (400 MHz, DMSO): δ_H = 8.2 (m, 2 H), 7.8 (m, 1 H), 7.6 (m, 3 H), 7.3 (m, 3 H), 3.7 (s, 3 H), ppm.

1,1’-Biphenyl (Table 2, Entry 1) (Figure S10)¹:H NMR (400 MHz, DMSO): δ_H = 7.4 (m, 4 H), 7.0 (m, 6 H) ppm.

3-Nitro-1,1’-biphenyl (Table 2, Entry 2) (Figure S11)¹:H NMR (400 MHz, DMSO): δ_H = 7.5 (m, 1 H), 7.4 (m, 2 H), 7.2 (m, 3 H), 7.1 (m, 3 H), ppm.

1-Phenylanthracene (Table 2, Entry 7) (Figure S12)¹:H NMR (400 MHz, DMSO): δ_H = 7.6 (m, 2 H), 7.4 (m, 2 H), 7.0 (m, 3 H), 6.8 (m, 7 H), ppm.

1-([1,1’-Biphenyl]−4-yl)ethan-1-one (Table 2, Entry 8) (Figure S13)¹:H NMR (400 MHz, DMSO): δ_H = 7.0 (m, 5 H), 6.7 (m, 4 H), ppm.

Catalyst characterizations

To verify the successful functionalization of SnFe₂O₄/SiO₂/PM-Pd magnetic nanoparticles (MNPs), FT-IR spectroscopy was conducted using the KBr pellet method (Fig. 3). In Fig. 3(a), vibrational bands at 740 and 580 cm⁻¹, observed across all FT-IR spectra, correspond to the stretching vibrations of the Sn-O bond. Additionally, peaks near 3490 cm⁻¹ are attributed to hydroxyl groups located on the surface of the magnetite. Figures 3(b–c) reveal peaks at 1064 cm⁻¹ and within the range of 2919–3029 cm⁻¹, which can be attributed to the ν(Si─O) and ν(C─H) vibrational modes, respectively; these specific bands are absent in the magnetite spectrum. The presence of these vibrations indicates the successful incorporation of SiO₂ and APTES onto the SnFe₂O₄ nanoparticles. The immobilization of the PM group onto SnFe₂O₄/SiO₂ is confirmed by the appearance of a C═N vibrational band at approximately 1630 cm⁻¹, along with a C─H vibrational band near 1418 cm⁻¹ (Fig. 3d). Furthermore, in the spectrum of SnFe₂O₄/SiO₂/PM-Pd (Fig. 3e), the downward shift of the C═N vibrational band to a lower frequency at approximately 1466 cm⁻¹ provides evidence for the successful formation of the Pd complex on the surface of the functionalized SnFe₂O₄ nanoparticles^37,38,39,40.

The crystalline phase of SnFe₂O₄/SiO₂/PM-Pd MNPs was analyzed using XRD. As depicted in Fig. 2, the material exhibited seven distinct and well-defined peaks at 2θ values of 18.3°, 30.4°, 36.7°, 37.4°, 43.1°, 54.8°, 56.4°, 63.5°, 72.5°, and 75.3°. These peaks corresponded to the (111), (220), (311), (222), (400), (422), (511), (200), (620), and (533) planes, respectively, and aligned closely with previously reported XRD patterns for SnFe₂O₄ MNPs. This confirms that the tubular structure of SnFe₂O₄ remains intact following its functionalization and stabilization within the silica sulfuric acid shell. Additionally, the analysis highlighted a noisy background caused by the amorphous dried PM-Pd shells, as illustrated in Fig. 4^41,42..

The TGA curve of SnFe₂O₄/SiO₂/PM-Pd demonstrates three distinct stages of mass loss. The initial stage, involving a 7% mass reduction at temperatures below 200 °C, is attributed to the evaporation of adsorbed organic solvents. As illustrated in Fig. 5, the second stage occurs between 200 and 600 °C, during which the removal of organic components results in a 20% mass loss. Based on this data, it can be concluded that the synthesized catalyst remains thermally stable up to 600 °C without degradation. Lastly, a third stage of mass loss, approximately 3.27%, is observed at temperatures exceeding 600 °C. This is associated with the condensation of silanol groups, leading to the loss of OH groups. These findings confirm the successful integration of the PM-Pd complex into the SnFe₂O₄ framework.

To confirm the presence of palladium metal on the surface of functionalized boehmite, the EDS technique was employed. The resulting EDS spectrum of SnFe₂O₄/SiO₂/PM-Pd nanoparticles is depicted in Fig. 6. The spectrum clearly indicates the presence of Sn, Si, C, N, O, and Fe, along with Pd species in the SnFe₂O₄/SiO₂/PM-Pd composition. These results from the EDS analysis enhance the understanding of the catalyst’s structural makeup and offer potential insights into its catalytic performance. Further exploration of the relationship between elemental distribution and catalytic properties could provide critical guidance for refining catalyst design and improving performance in future research endeavors.

A scanning electron microscope (SEM) was employed to analyze the size and morphology of SnFe₂O₄/SiO₂/PM-Pd particles (Fig. 7). The SEM analysis indicated that these particles possess uniform spherical shapes, with diameters ranging from 20 to 60 nm. Furthermore, the images revealed particle agglomeration, likely caused by the magnetic properties of the nanoparticles.

Figure 8 presents the XPS spectrum of the synthesized SnFe₂O₄/SiO₂/PM-Pd catalyst, showcasing distinct peaks corresponding to Si, Sn, O, C, N, Fe, and Pd. These elemental contributions are further detailed in Fig. 8a. An in-depth XPS analysis of palladium is provided in Fig. 8b, which identifies its oxidation state through characteristic binding energy peaks at 331.2 eV and 346.4 eV, attributed to Pd 3d_5/2 and Pd 3d_3/2, respectively. This thorough examination confirms the structural composition of the SnFe₂O₄/SiO₂/PM-Pd catalyst^{43,44,45,46,47,48}..

Figure 9 presents TEM images of the SnFe₂O₄/SiO₂/PM-Pd magnetic nanoparticles (MNPs). The images reveal nearly spherical layers, with darker regions likely corresponding to the SnFe₂O₄ nanoparticles forming the core. Encasing these core nanoparticles, a brighter layer is visible, representing a distinct coating. This outer layer is likely evidence of the successful immobilization of the palladium (Pd) complex on the surface of the titanium ferrite nanoparticles. The contrast between the darker core and the brighter coating underscores the composite structure of the material, clearly delineating its core and functionalized outer layer as shown in Fig. 9.

The ICP method was employed to determine Palladium concentrations in the original catalyst and to evaluate Pd leaching following recycling. The findings revealed that the Pd contents in the fresh and recycled catalysts were 2.3 × 10⁻³ and 2.2 × 10⁻³ mol g⁻¹, respectively. This demonstrates negligible Pd leaching from the SnFe₂O₄/SiO₂/PM-Pd structure.

The magnetization of SnFe₂O₄ and SnFe₂O₄/SiO₂/PM-Pd was examined at room temperature using a VSM, as illustrated in Fig. 10. The resulting magnetization curves indicated saturation magnetization values of 59 emu/g for SnFe₂O₄ and 41 emu/g for SnFe₂O₄/SiO₂/PM-Pd. The decrease in saturation magnetization observed in the prepared catalyst is linked to the stabilization of the PM-Pd complex on the surface of SnFe₂O₄.

Catalytic study

After characterizing SnFe₂O₄/SiO₂/PM-Pd, its catalytic efficiency for synthesizing biphenyl byproducts was analyzed using the Suzuki–Miyaura coupling reaction. Reaction parameters were initially optimized by varying the solvent, base, and catalyst concentrations in a model reaction between phenylboronic acid and iodobenzene, as outlined in Table 1. A blank experiment conducted without the catalyst in PEG solvent at 100 °C showed no reaction progress, even after an extended duration (Table 1, entry 1). Additionally, when the catalyst was removed midway through the reaction at 50% conversion, no further transformation occurred, reinforcing its critical role in driving the reaction forward. This underscores the importance of the SnFe₂O₄/SiO₂/PM-Pd catalyst in facilitating the Suzuki–Miyaura coupling process. Table 1 (entry 4) demonstrates that 30 mg of SnFe₂O₄/SiO₂/PM-Pd is sufficient to complete the reaction within 20 min with 1 mmol of iodobenzene in PEG at 100 °C. The next phase involved evaluating the effect of various solvents, including H₂O, EtOH, PEG-400, DMF, and toluene, under identical conditions with 30 mg of catalyst. These tests revealed that environmentally friendly PEG was superior to all other solvents. To determine the impact of different bases on the reaction rate, accessible inorganic bases such as K₂CO₃, NaOH, and KOH were tested. Among these, K₂CO₃ emerged as the most effective base (Table 1), likely due to its higher solubility in PEG-400 medium. From both economic and reaction optimization perspectives, K₂CO₃ was validated as the most suitable choice. Further optimization involved varying the molar ratio of K₂CO₃. It was observed that reducing the amount to 1 mmol led to a slower reaction rate, whereas increasing it to 1.7 mmol provided no additional benefit (Table 1). These findings highlight the importance of fine-tuning reaction parameters for optimal efficiency.

After optimizing all reaction conditions, the coupling of phenylboronic acid with various aryl halides (I, Br, and Cl) was carried out using 30 mg of SnFe₂O₄/SiO₂/PM-Pd under ultrasonication at the established optimal conditions. The results, summarized in Table 2, highlight the efficiency of the process in facilitating reactions between phenyl chlorides, bromides, and iodides with phenylboronic acid to produce the desired products. As shown in Table 2, the catalyzed reactions involving aryl halides with either electron-donating or electron-withdrawing functional groups consistently achieved high yields of the corresponding products. It is worth noting that while the process delivers high yields across all evaluated reactions, the coupling of aryl chlorides with phenylboronic acid requires a longer reaction time to achieve moderate product yields compared to the coupling reactions of aryl bromides and iodides.

A proposed mechanism for the anthrax reaction involves a series of key steps. First, aryl halides engage in oxidative addition to the Pd complex, resulting in the formation of intermediate 1. This is followed by intermediate 2 undergoing transmetallation, which leads to the production of intermediate 3. Lastly, reductive elimination takes place, restoring the SnFe₂O₄/SiO₂/PM-Pd species and yielding the target product as outlined in Fig. 11.

Hot filtration

The hot filtration and leaching tests were employed to verify the heterogeneous nature of the synthesized material, independent of whether any catalyst particles were present in the filtrate solution. During the Suzuki–Miyaura cross-coupling reaction, the nanocatalyst was utilized for 10 min under optimal conditions, after which the reaction mixture was split into two portions. The catalyst was extracted from one portion using a magnetic field, and both portions were then allowed to react for an additional 10 min. It was observed that no conversion occurred in the catalyst-free environment, while the reaction in the other portion proceeded to completion, as detailed in the Supporting Information. These results strongly indicate that minimal to no Pd leaching occurred within the reaction mixture, thus confirming its genuine heterogeneity.

Type 2 diabetes mellitus (T2DM) is a chronic metabolic disease associated with a significantly increased risk of cardiovascular (CV) disease, including coronary artery disease (CAD) and heart failure (HF) [1]. Despite advancements in glycemic control therapies, reducing CV morbidity and mortality in T2DM patients remains a critical challenge [2]. Although microvascular complications are closely associated with CV morbidity, and glycemic control effectively reduces microvascular damage, to date, no trial has conclusively demonstrated that lowering HbA1c alone significantly reduces major adverse cardiovascular events (MACE) in patients with established microvascular or macrovascular disease, i.e., in the context of secondary prevention [3]. However, some therapies have shown cardioprotective effects, reducing MACE independently of glycemic control [4]. Sodium-glucose cotransporter-2 inhibitors (SGLT2i) represent a breakthrough in the management of T2DM, with extensive cardiovascular outcome trials (CVOTs) showing reductions in CV mortality, HF hospitalizations, and renal disease progression [5,6,7,8,9].

Coronary flow reserve (CFR), a key marker of cardiovascular health, particularly in patients with type 2 diabetes mellitus (T2DM) [10], is a fundamental indicator of coronary circulation, reflecting the ability of coronary arteries to increase blood flow during stress compared to rest (CFR = myocardial blood flow [MBF] at stress/MBF at rest) [11]. In T2DM, CFR is frequently reduced due to microvascular dysfunction, [12, 13] which contributes to a higher risk of ischemic events [6, 14, 15]. A potential cause of microvascular dysfunction may be the loss of the cardio protective effect of epicardial adipose tissue (EAT) [16, 17]. In type 2 diabetes, this distinct brown adipose tissue depot has been reported to undergo morphological and functional alterations shifting toward white adipose tissue characteristics along with increased thickening and inflammation¹⁶. Increased inflammation of EAT, indicated by the whitening of brown adipose tissue, has been widely documented in the literature [18,19,20,21]. Chronic inflammation contributes to microvascular dysfunction, which is detectable through a reduction in CFR [22, 23]. Due to its close anatomical proximity to the heart and its potential to regain its anti-inflammatory molecule secreting activity, thus improving CFR, EAT is now recognized as an important, independent, and modifiable cardiovascular risk factor [24,25,26].

Another potential consequence of microvascular dysfunction is the reduction in myocardial mechano-energetic efficiency (MEEi) [27], a measure of cardiac performance, which assesses the heart’s efficiency in converting metabolic energy into mechanical work [28]. MEEi, which is typically reduced in T2DM, due to cardiac insulin resistance and altered substrate utilization [28], provides valuable insights into the functional impact of therapeutic interventions on myocardial efficiency [15, 29]. Preserving or improving MEEi is essential for maintaining cardiac energy efficiency in this high-risk population.

In the DAPAHEART trial, a recent single-center, 4-week, randomized, double-blind, controlled study (1:1 dapagliflozin 10 mg vs. placebo), we demonstrated that treatment with dapagliflozin induced a 30% increase in CFR in patients with T2DM and stable CAD [30, 31]. The improvement in CFR was associated with a selective 19% reduction in EAT thickness and glucose uptake, suggesting a restoration of the protective effect of this fat depot. Indeed, the reestablishment of the protective function of EAT may have effects on microvascular dysfunction, potentially accounting for the observed improvements in CFR , as previously mentioned [22, 23, 32].

However, whether these benefits are maintained over longer periods remains unexplored. This follow-up study aimed to address this knowledge gap by evaluating the long-term effects of dapagliflozin on CFR, EAT and MEEi over a 4-year period, providing insights into its sustained cardioprotective potential.