IMPLICATIONS OF CALCIFICATION IN PERCUTANEOUS CORONARY INTERVENTIONS

Vascular calcification is a process closely associated with atherosclerosis. It can occur in the media (in peripheral arteries mainly) or intima layers (in coronary arteries). In the context of coronary atherosclerosis it debuts in intermediate or advanced stages in plaque evolution due to conversion of smooth muscle cells into osteoblastic phenotypes and infiltration of atheromatous plaque due to macrophages that clear out apoptotic smooth muscle cells and contain calcified vesicles.1 Atheromatous plaque calcification can take different shapes that probably correspond to different stages of the same disease like microcalcifications (< 15 μm), punctiform calcifications (circumferential arc < 90º), leaves or thin calcium layers (circumferential arc > 90º or > 3 mm in length), and calcium nodules.1

The main risk factors associated with coronary artery calcification are age, Caucasian race, diabetes mellitus, and chronic kidney disease.1

The prevalence of coronary artery calcification if variable based on the population studied and the diagnostic method used.2 The traditional angiographic definition of moderate calcification described radiopacities seen during cardiac motion while severe calcification is described as radiopacities seen without cardiac motion, usually on both sides of the arterial lumen. The prevalence of moderate or severe calcification is between 18% and 60%.3,4

Calcification complicates percutaneous coronary interventions (PCI) for various reasons: a) resistance to the advance of different devices especially in the presence of tortuosity (eventually, “non-crossable” lesions); b) reduced plaque compliance that will eventually require higher pressure in dilatation balloons or plaque modification devices (“non-dilatable” lesions); and c) difficulties advancing the stent and expanding it.5 Other issues would be malapposition and polymer damage that can lead to a non-homogeneous release of antiproliferative drugs. Everything combined makes calcification one of the major determinants of the SYNTAX score,6 and associated with worse PCI outcomes and higher rates of adverse events at follow-up including mortality in patients with extremely calcified coronary artery lesions.7 In addition, it increases the rate of procedural complications associated with calcification per se and with the tools necessary for treatment: coronary artery dissection, loss of side branches, PCI material entrapment, stent distortion or even stent loss, and the dreaded coronary artery perforation that is particularly severe since it is very difficult to advance any kind of sealing materials.8

To stop these issues and their prognostic implications from happening numerous plaque modification devices have been developed. The appropriate use of these devices is essential to perform safe and effective PCIs on calcified coronary artery lesions.

This position paper has been promoted by the Interventional Cardiology Association of the Spanish Society of Cardiology (ACI-SEC) with contributions from different expert professionals in this setting. It describes the plaque modification techniques currently available in our field and proposes an algorithm for the management of calcified coronary artery lesions.

INTRACORONARY IMAGING MODALITIES FOR CALCIFIED LESION ASSESSMENT

Intracoronary imaging modalities play a key role in the assessment of calcified coronary artery lesions. The use of optical coherence tomography (OCT) or intravascular ultrasound (IVUS) can be very useful to improve the detection and assesment of coronary artery calcium, select the plaque modification technique, and optimize results especially in association with stent expansion.

Calcification detection and assessment

Angiography is a limited sensitivity tool to detect coronary artery calcium. Unlike angiography both the IVUS and the OCT have higher sensitivity and specificity to assess the characteristics and degree of calcification, which are basic aspects to determine the therapeutic options.2,9 Table 1 shows the differences of these 2 intracoronary imaging modalities regarding calcium detection. The main difference between the 2 is that, since calcium creates posterior acoustic shadowing on the IVUS, calcium thickness cannot be properly assessed. As alternative marker, the presence of reverberations on the IVUS has been associated with the presence of thinner calcium layers (< 0.5 mm). On the OCT, parietal calcium does not create posterior acoustic shadowing and, therefore, its thickness can be assessed accurately. Nodular calcium, however, creates a shadow in both the IVUS and the OCT (figure 1).

Table 1. Intracoronary imaging modalities for calcified coronary artery lesion calcification

Figure 1. Coronary artery calcium assessment with IVUS and OCT. A: Parietal calcium on the OCT, low reflectivity structure with demarcated borders (asterisk). B: Parietal calcium on the IVUS, hyperechogenic structure with posterior shadowing. C: Calcium nodule on the OCT, structure protruding into the lumen with posterior shadowing. D: Calcium nodule on the IVUS, structure protruding into the lumen with posterior shadowing.

Different scoring systems have been developed for both intracoronary imaging modalities (table 2) including the characteristics of calcification that have been associated with stent underexpansion. The first OCT suitable scale ever developed includes 3 different parameters: calcium arc > 180º (score = 2), length > 5 mm (score = 1), and thickness > 0.5 mm (score = 1). Lesions with scores > 2 have a higher risk of stent underexpansion if proper plaque preparation is lacking.5 A similar scale has been developed for IVUS using 4 different criteria: calcium arc > 270º with > 5 mm in length (score = 1), calcium arc > 360º (score = 1), presence of calcified nodule (score = 1) and adjacent vessel < 3.5 mm (score = 1). Scores ≥ 2 are indicative of the need for plaque modification prior to stenting.10

Table 2. Intracoronary calcium scores based on optical coherence tomography and intravascular ultrasound

Selection of plaque modification techniques under intracoronary imaging modality guidance

The characteristics of calcium as seen on the intracoronary imaging modalities can contribute to the selection of the most adequate plaque modification technique. There is in depth information on this aspect in the last section of the document but, overall, lesions where calcium does not have underexpansion risk criteria can be treated with high-pressure or modified balloons (scoring, cutting). However, if these criteria exist it will be necessary to use more advanced plaque modification techniques. Added to these criteria, we should also mention calcium depth since some imaging modalities only act on the superficial—not deep—layer of the plaque.

Optimization of stenting under intracoronary imaging modality guidance

Both the IVUS and the OCT allow us to determine whether proper stent expansion has been achieved. This is especially relevant in calcified coronary artery lesions that happen to be the ones that are most associated with stent underexpansion, the parameter most strongly associated with stent failure.11 Proper apposition and lack of dissection or significant border hematoma, as well as proper lesion coverage are other optimization parameters under intracoronary imaging modality guidance that should also be assessed after stenting.12

BALLOON-FREE TECHNIQUES

Rotational atherectomy

The rotational atherectomy (RA) technique uses a metal olive-shaped burr covered with diamond crystals in its distal third that rotates at high speed and performs a differential cut when advancing through the vessel (figure 2A) while pulverizing calcified tissue and preserving the adjacent elastic tissue.13

Figure 2. Plaque modification devices. A: Rotational atherectomy device. B: Orbital atherectomy device. C: Coronary laser device with the 2 existing console models. D: Intracoronary lithotripsy device. Modified with permission from Cubero-Gallego et al.13

It appeared over 30 years ago to facilitate the management of coronary artery lesions by reducing plaque burden. Early enthusiasm turned into an elevated use of RA in different settings without the proper scientific back-up. This triggered suboptimal results14 that reduced its use to highly selected cases only. Through all these years, RA has evolved with technological improvements, and also of the technique itself, as well as the selection of patients.

Currently, the ROTAPRO system (Boston Scientific, United States) is available. It has made the technique easier because it has replaed the pedal of the early version for a button placed on top of the olive-shaped burr advancer. There is another button on the side of the advancer to change to the Dynaglide mode (rotation at low revolutions is advised to introduce and remove the burr). Console is smaller and comes with a digital screen. Size of the burrs is between 1.25 mm and 2.5 mm, and they are compatible with 6-Fr-to-8-Fr catheters based on the size of the olive-shaped burr that advances on a 0.009 in specific guidewire (0.014 in the radiopaque side) of which 2 different versions exist (RotaWire Floppy and RotaWire Extra-Support) that should be used depending on the characteristics of the plaque and support needed.13

The main indication is to treat extremely calcified non-crossable or non-dilatable coronary artery lesions with balloon. Probably, the optimal scenario is a concentric calcified lesion with a smaller luminal area compared to the olive-shaped burr. Eccentric angulated lesions are less favorable since they are associated with a higher risk of complications.13,15 It can be used as a primary or a bailout strategy after “balloon failure”. The primary strategy has been associated with shorter procedures, less radiation and contrast, and probably lower cost regarding the material used.15

The target of RA has also changed from the old idea of removing as much plaque as possible (debulking) to the modern approach of modifying plaque to “facilitate” the PCI. Technical recommendations to perform RA have evolved too. Current recommendations are shown on table 3.16

Table 3. Recommendations for a safe use of rotational atherectomy

The most common complication is slow/no-flow although its rate has dropped down to 2.6%.17 It is due to debris embolization towards microcirculation and there is higher risk in long lesions where multiple and prolonged passes with large olive-shaped burrs are performed without proper pauses among them and in the presence of a poor distal vessel. The management of dominant right coronary or left circumflex coronary arteries can be associated with transient conduction disorders. Severe complications like burr entrapment, perforation, and coronary dissection are rare.13 Factors predisposing burr entrapment are lesion severity, steep angulations, and the use of very small burrs. Tortuosity and the lack of guide catheter coaxiality in the management of ostial lesions can trigger dissections and coronary perforations.

Although RA has demonstrated that it facilitates PCI with higher rates of success compared to balloon angioplasty, No clinical benefit has been yet confirmed.18-21

To analyze its results we should mention that RA has been used in patients of higher clinical risk with more complex coronary artery lesions.22 Another aspect we should take into consideration is the high percentage of cases where this technique was used as a bailout strategy (12% to 50%)20,21,23 meaning that without RA these cases would not have been performed or had had worse results. Although ongoing trials are studying the advantages of elective or bailout RA, proper patient and lesion assessment should lean towards increasing its elective or earlier use with a potential beneficial impact on clinical outcomes.24

In conclusion, when performed under the current recommendations RA is a safe and effective procedure. It should become part of our therapeutic arsenal in our cath labs with trained personnel for proper use.

Orbital atherectomy

The Diamonback-360 (OAS) device (Cardiovascular Systems, United States) is a diamond-coated bidirectional orbital crown that uses a combination of centrifugal force (by creating elliptical orbits) and friction to the surface to modify the calcified plaque and increase compliance (figure 2B). Also, with the pulsatile impact of the crown after speeding up, microfractures can occur that eventually modify deep calcium layers (figure 2B and figure 3). That is why a single 1.25 mm crown can treat vessels from 2.5 mm up to 4 mm.

Figure 3. Characteristics of orbital atherectomy catheter and its effects. (Modified with permission from Cardiovascular Systems.25)

Compared to the remaining plaque modification techniques, this orbital atherectomy (OA) has arrived late to our country and our experience is still scarce.

Its main indication is to treat no-dilatable calcified coronary artery lesions.26

Preparation is very similar to RA, but here a specific guidewire is needed, the Viper-Wire. Crown advances with the Glide-Assist system (rotation at low revolutions) until coming close to the lesion. Another distinctive feature of this device is its antegrade and retrograde ablation functionalities. Unlike RA, the speed at which the device moves forward needs to be very slow (between 1 mm and 3 mm per second) to guarantee good procedural results and reduce complications.17,26 The mechanism of action of OA consists of the crown elliptical rotation that achieves a gradual increase of orbital diameter as rotation speed increases from 80 000 rpm up to 120 000 rpm. Cycles ≤ 30 seconds are advised (it comes with a sound warning signal to end the cycle) with pauses between 20 and 30 seconds among them that can duplicate in cases of poor hemodynamic tolerance.26 The continuous infusion of a lubricant solution is necessary to minimize thermal lesions during OA. Also, 18 mL/min are administered to cool the device down and get rid of debris, thus reducing the chances of ischemia and distal embolization.13,26,27

Complications are similar to those of RA. However, the possibility of retrograde application reduces the chances of crown entrapment and the risk of dissection or perforation in angulated or ostial lesions. The rate of perforation is between 0.7% and 2%.28,29 Theoretically speaking, the debris created by OA is smaller compared to the debris created by RA. This added to the fact that the crown does not stop coronary flow during atherectomy reduces the risk of slow/no-reflow and endothelial thermal lesion.27 However, transient conduction disorders are not rare when dominant right coronary or left circumflex arteries are treated.

Current evidence available is based on the ORBIT I30 and ORBIT II28 clinical trials where OA obtained good results regarding procedural success (94% and 89%, respectively) with higher rates of major adverse cardiovascular events (MACE), and target lesion failure of 23.5% and 7.8%, respectively, at 3 years.31 Afterwards, the COAST trial29 was conducted where the new MicroCrown system was used. A total of 100 patients were included with rates of procedural success and MACE of 85% and 22.2%, respectively, at 1-year follow-up. We are waiting to see the results from the ECLIPSE trial that will randomize a total of 2000 patients with severe calcifications to receive OA or balloon predilatation prior to by drug-eluting stent implantation.

In conclusion, OA is another calcium plaque modification technique with potential technical advantages like having only 1 size of crown compatible with 6-Fr to treat all lesions and with pull-back capabilities. Although there are insufficient data from comparative studies, choosing this technique will depend on the profile of the patient and the lesion to be treated, the intracoronary being an essential aspect.

Excimer laser

Excimer laser coronary angioplasty (ELCA) is based on a xenon chloride laser that generates short ultraviolet pulses of 308 mm that only penetrate 50 µm in depth, which makes it safer compared to old continuous-wave-near-infrared lasers. It modifies the plaque through a triple mechanism: photochemical (by breaking molecular binds), photothermal (through tissue vaporization), and photokinetic (through the expansion and collapse of the bubble of the catheter tip as it moves forward). Fragments released are < 10 μm, which minimizes microvascular damage after being absorbed by the reticuloendothelial system.

The system currently used is the CVX-300 Laser System (Philips) although there is already a new generation one, the LAS-100 Laser System (Philips) that will be replacing it shortly (figure 2C). There are different sizes of catheters available (0.9 mm, 1.4 mm, 1.7 mm, and 2.0 mm) (table 4). The selection of the catheter depends on the type of lesion and size of the vessel (catheter to vessel diameter ratio, 0.5-0.6) being the 0.9 mm catheter the most widely used for its lower profile and because it can reach higher fluency (80 mJ/mm²), pulse repetition rate (80 Hz), and longer application durations (10 seconds with 5-second rests), which increases the chances of success in fibrous calcific plaques.32,33

Table 4. Characteristics of Excimer laser coronary angioplasty catheters

Before being used, it is necessary to calibrate the console and then the catheter. In both cases, health professionals and patients should use protective glasses to prevent eye damage. Afterwards, a 0.014 in intracoronary guidewire is inserted until it reaches the lesion. There is a monorail system to facilitate moving forward. Energy is released through the catheter distal border as it slowly advances (0.5 mm/second) to modify the plaque. Catheter withdrawing can also be applied. It is important to optimize support to ensure that the catheter advances. There is no limit in the number of pulses that can be administered since the more it is used, the stronger the effect. However, there is also a higher risk of complications involved. Some authors suggest a maximum of 12 applications.33 The state of the vessel should be assessed after each application. Regarding the selection of parameters, traditionally it started at 45 mJ/mm², and 25 Hz. However, more and more operators prefer higher energies and initial frequencies especially to treat resistant or calcified lesions.33

Before and during the applications, the blood vessel should be washed, and contrast administered through the infusion of a physiological saline solution (1 mL/s to 3 mL/s). In resistant lesions with severe calcification or underexpanded stents, more energy may be needed. This can be reached by not washing the blood with the physiological saline solution or even administering contrast during applications (the so-called “explosion technique”). This technique reaches maximum power although it increases the chances of complications. Some authors33 recommend it as the first option to treat non-thrombotic lesions, although it seems reasonable to spare it for ELCA-resistant lesions with saline infusion.

Traditionally, the indications for ELCA have been categorized into 2 different groups: “thrombotic” (not discussed in this document) and “calcified” lesions (non-thrombotic like in-stent restenosis, chronic total coronary occlusions, calcified lesions, etc.). The latter can be categorized into non-crossable or non-dilatable lesions:

Non-crossable lesions

The laser main advantage is that it is compatible with all intracoronary guidewires. Therefore, non-crossable lesions with balloon/microcatheter are its main indication.17 In a multicenter registry of non-crossable lesions, the rate of procedural success was 87.3% with 0.8% of dissections showing an impaired flow and no perforations.34 Severe calcification has been associated with a higher probability of technique failure34 since ablation is primarily performed in the tissues between calcium.35 However, the use of ELCA with contrast can increase its chances of success in these lesions.33

Non-dilatable lesions

Although the success of ELCA in non-dilatable lesions is high,36 it has never been the first-line therapy. Among these lesions, an interesting scenario is in-stent lesions (restenosis or underexpansion). In acute underexpansion, ELCA could be the treatment of choice. It allows the modification of resistant tissue located behind the stent without changing its architecture. Its use with contrast can be safer thanks to the stent “protective” effect. Isolated cases and small case series with success rates > 95% and few complications have been published.37

It is a safe technique when the recommendations given are observed. Coronary artery dissection is the most common complication (5), although it is rarely flow-limiting (1%). The rate of coronary artery perforation is < 1%,38 and distal embolizations and ventricular arrhythmias are exceptional.39

In conclusion, ELCA is especially useful to treat non-crossable lesions thanks to its compatibility with all kinds of angioplasty guidewires. It has also proven effective to treat non-dilatable lesions including in-stent lesions. However, there is still scarce information on its efficacy in calcified coronary artery lesions.

BALLOON-BASED TECHNIQUES

Cutting and scoring balloons

Cutting balloons (CB) are plaque modification devices that appeared as an alternative to old coronary angioplasty balloons.40 Their objective is to achieve controlled ruptures of the plqeu (through incisions in fibrocalcific tissue) (figure 4), thus facilitating balloon expansion, minimizing damage to the intima layer, and reducing stenosis.18,41

Figure 4. Rotacutting technique. Rotational atherectomy (RA) effect and cutting balloon (Rotacutting) with greater plaque modification and minimum lumen area.

There are 2 different types: CB and scoring balloon (BS). Their use has been described in different settings like in-stent restenosis, aorto-ostial lesions, bifurcations, and small vessels associated with the use of drug-eluting stents.42

The main limitations of CBs are their worst navigability and crossing profile compared to conventional balloons. However, over the past few years, both aspects have improved. SBs are associated with better navigability compared to old CBs.

The most dreaded complication is the rupture of coronary artery, although it has significantly increase following its use.

The main difference among the different devices lays in their different external atherotomy elements as described herein (figure 5).

Figure 5. Characteristics of modified balloons. A: Cutting balloon (Boston Scientific, United States). B: WOLVERINE (Boston Scientific, United States). C: AngioSculpt (Spectranetics, United States). D: Scoreflex (OrbusNeich, Hong Kong). E: Grip (Acrostak, Switzerland). F: NSE-Alpha (B.Braun, Germany). G: Naviscore (iVascular, Spain).

Cutting Balloon Flextome

Cutting Balloon Flextome (Boston Scientific, United States) consists of a noncompliant (NC) balloon with 3 microrazors longitudinally mounted on the surface. Its superiority over conventional balloons in A/B lesions has not been confirmed yet, which is why, so far, its use is limited to complex17 and calcified lesions only.43

WOLVERINE

Wolverine (Boston Scientific, United States) is an evolution of the former one with a better crossing profile, greater flexibility, and a more visible tip.

AngioSculpt

AngioSculpt (Spectranetics, United States) is a semicompliant balloon with low crossing profile surrounded by 3 nitinol filaments arranged in a helical cage to secure balloon anchorage. There is a lower risk of dissection or perforation associated with this device.17 It provides more flexibility and better navigability compared to old CBs,44 as well as good results compared to dilatation with semicompliant balloons.45

Scoreflex

Scoreflex (OrbusNeich, Hong Kong) is a NC consisting of a NC balloon with a nitinol dual-wire system to facilitate the controlled modification of the plaque at low pressures. It has a low profile and a combination of hydrophilic and hydrophobic coating that minimizes friction during lesion crossing.

Grip

Grip (Acrostak, Switzerland) is a high-pressure balloon with 4 rows of 3 or 4 knobs in each row. It allows dilatations of up to 22 atm. It comes with a cone-shaped tip in 2 different versions: Grip with a short 2 mm tip, and Grip TT with a long 4 mm tip for greater navigability in tortuous anatomies. It comes with a hydrolubricated coating on its tip and the catheter (not on the balloon), which facilitates both its anchorage to the lesion and navigability across lesions.

NSE Alpha

NSE Alpha (B. Braun, Germany) is a SB with 3 nylon scoring elements and 1 triangular cutting section connected in both borders of the balloon and arranged in a 120º disposition. We should mention its flexibility and navigability with good results in de novo lesions and in-stent restenosis.18

NaviscoreTM (iVascular, Spain)

NaviscoreTM (iVascular, Spain) is a SB with a design that combines the benefits of SB plus CB. It consists of a high-pressure balloon with 125 μm nitinol filaments. These have an axial orientation for greater crossing abilities and flexibility, and plaque modification in a 90º angle, which is associated with lower chances of perforation. The catheter hydrophilic coating improves its navigability.

In conclusion CBs and SBs are useful plaque modification devices to treat non-dilatable lesions when calcification is not very severe. Their main advantage is how easy they are to use since it is a balloon-based technique compatible with conventional angioplasty guidewires.

Very high-pressure balloons

The NC very high-pressure balloon (VHPB) OPN (SIS medical, Switzerland) is a double-layer balloon for homogeneous expansion at extremely high pressures without increasing its diameter (from 2 mm to 4 mm) with a rated burst pressure of 35 atm, although the manufacturer’s testing rated burst pressure limit is 45 atm (table 5).46

Table 5. Compliance of the NC very high-pressure OPN balloon

The VHPB has been used for over 10 years now and it has proven safe and effective in up to 40 atm in extremely calcified coronary artery lesions where other devices have failed or in stent underexpansion. Success rates are as high as 75% to 100% without evidence of dissection, perforation or balloon bursts in small case series.47 Compared to conventional NC balloon, it can achieve minimum luminal diameters and major acute gains with less residual stenosis in non-dilatable lesions.48

The largest registry ever conducted to this date included 326 patients with non-dilatable lesions treated with VHPB after failed NC balloon. Patients were categorized into 2 groups: those who responded to pressures between 30 atm and 40 atm, and those who responded with pressures > 40 atm. Procedural success was reached in up to 96.6% of the patients. A total of 53% of the patients responded to pressures between 30 atm and 40 atm while the remaining 47% did so to pressures > 40 atm. A total of 180 patients were treated with intracoronary imaging modalities and 106 of these showed calcifications > 270º. In this subgroup of patients, the pressured needed for optimal expansion was > 40 atm in 78.3% of the cases. Three patients (0.9%) showed coronary artery ruptures that resolved with prolonged inflation or covered stent implantation. In the 3 cases, the ruptures occurred during predilatation and were associated with balloon bursts with pressures between 30 atm and 40 atm. This is suggestive that perforations don’t seem to be associated with inflation pressure but with the characteristics of the plaque or the vessel size estimate that was angiographic in the 3 cases reported.49

The ISAR-CAL trial50 was published back in 2021. It randomized 70 patients with extremely calcified coronary artery lesions and failed predilatation with NC balloon to receive a SB or a VHPB. The study primary endpoint was to compare stent expansion on the OCT. No differences were reported in the percentage of stent expansion. However, differences were seen in angiographic secondary endpoints like improved minimum luminal diameters and residual stenoses favorable to VHPB.50

Finally, chronic total coronary occlusions are the pinnacle of calcified complex lesions. In the PLACCTON trial the use of the VHPB both alone and with other plaque modification techniques was safe and effective in selected cases with chronic total coronary occlusions.51

In conclusion, the VHPB is a safe and effective alternative to treat non-dilatable calcified coronary artery lesions. Randomized clinical trials better defining this device strategy of use and the remaining plaque modification techniques are lacking.

Intracoronary lithotripsy

Intracoronary lithotripsy (ICL) system consists of a specific balloon catheter (Shockwave Medical, United States) connected to a rechargeable portable generator (figure 2D). The generator produces energy pulses that are transmitted to emitters placed inside the balloon. Pulses are emitted at a frequency of 1 per second up to a maximum of 10 pulses per application. Each balloon catheter can administer a maximum of 80 pulses. The catheter consists of a rapid-exchange semicompliant balloon with a 0.042 in crossing profile compatible with any 0.014 guidewires and 6-Fr guide catheters.

Its main indication is to treat calcified non-dilatable coronary artery lesions.

A 1:1: ratio between the vessel and balloon diameters is advised. Once it has been placed inside the lesion, the balloon inflates at 4 atm to secure proper contact between the balloon surface and the vascular wall to allow energy transfer. The balloon includes 2 emitters that receive an electric discharge from the generator that vaporizes the fluid inside generating sound waves that have a local effect. Each pulse releases the equivalent of 50 atm.

These waves run across soft tissues causing selective calcium microfractures at intima and media layer level. After pulse emission and the corresponding modification of calcium, the balloon inflates up to 6 atm to maximize luminal gain. Balloon catheter is only available at a length of 12 mm and comes in diameters of 2.5 mm, 30 mm, 3.5 mm, and 4.0 mm.52

The greatest evidence available comes from the Disrupt-CAD III trial, a prospective registry that assessed the efficacy and safety profile of ICL in 431 patients with calcified lesions. The 30-day rate of MACE (death, infarction or target lesion revascularization) was 7.8% while the rate of effectiveness (procedural success with in-stent stenosis < 50%) was 92.4%. No patients with acute myocardial infarction or complex lesions were included in this study.53 Recently, 12-month follow-up results have been published confirming rates of MACE and stent thrombosis of 13.8% and 1.1%, respectively.54

Controlled break down of coronary calcium is the basis of treatment of ICL balloons. In a OCT substudy of the Disrupt-CAD II trial after ICL calcium fractures were seen in 79% of the lesions55 compared to 67% of the lesions of the Disrupt-CAD III trial.53

Although the use of ICL balloons has become very popular worldwide, information on its safety and efficacy profile regarding its use in complex settings is still scarce (acute coronary syndrome, chronic total coronary occlusions, bifurcations or aorto-ostial lesions). As a matter of fact, its use is often limited to isolated cases or short series.52 The main limitations of this system are its reduced crossing profile in extremely calcified or tortuous stenoses and difficult use in diffuse or multivessel lesions (due to the limited number of pulses per catheter and the different caliber of target vessels).

A recent trial assessed the use of underexpanded stents due to severe coronary artery calcification and confirmed angiographic success rates of up to 73%, which is lower compared to the 75% seen in native lesions56 probably because it is more difficult to expand a calcified lesion when the stent has already been deployed. Therefore, regardless of the technique used, stenting is ill-advised until the lesion has been properly prepared. Also, the application of lithotripsy in this context, especially on freshly implanted stents, can cause structural damage to the polymer.57 Another multicenter registry proved the device was successful 92.3% of the times in this type of lesions.58 Mid- and short-term data on the safety profile of this technique are still lacking.

The combined use of ICL balloon and other plaque modification devices like RA,59 OA60 or ELCA61 has been described, and it seems like a very attractive strategy in cases where the ICL balloon cannot reach the target lesion.

In conclusion, ICL has grown exponentially in the management of non-dilatable calcified coronary artery lesions thanks to its safety and efficacy profile, and short learning curve. However, information on its use in complex scenarios and comparative results with other plaque modification techniques are still lacking.

COMBINED TECHNIQUES

There is not much evidence on the combination of devices or plaque modification techniques in extremely calcified coronary artery lesions.

The use of RA followed by CB (RotaCutting) (figure 4) or lithotripsy (RotaTripsy) (figure 6) has been described as an additional, safe, and effective technique.62-64 In both cases the concept is similar. Primarily, RA damages superficial calcium, but not the deepest calcium layers, and there are times when it is not enough for proper plaque preparation. On the other hand, CB or lithotripsy can complement the plaque modification provided by RA. However, when calcified lesions progress into very severe aortic stenosis, the target lesion can be difficult to reach with these balloons. In a combined use, RA modifies superficial calcium by creating a tunnel that the CB or lithotripsy balloon can use to move forward and, when in position, complete plaque modification. One of the differences between both techniques is that CB can contribute to breaking down the calcium layer in the absence of very severe calcification. The RotaTripsy technique59,63 can be more effective to treat extremely calcified coronary artery lesions with thick calcium layers. However, its cost is also higher. Based on a similar concept, the combination of OA plus lithotripsy has been recently described with good results.60

Figure 6. Rotatripsy technique. A: Extremely calcified stenosis in left anterior descending coronary artery. Intravascular ultrasound (IVUS) with 360º calcification. B: Rotational atherectomy (RA) (1.5 mm olive-shaped burr). C: 3 mm noncompliant balloon underexpansion after RA. D: Intracoronary lithotripsy with 3 mm balloon and proper expansion at 6 atm after 50 pulses. E-G: IVUS and optical coherence tomography showing the combined effect of RA plus lithotripsy with multiple zones of intimal “sanding/dissection” caused by the RA (asterisks), and deep and intimal fractures caused by lithotripsy. H: Final outcomes after stenting.

RA has also teamed up with ELCA (the RASER technique).65 Laser can be the only option in truly non-crossable lesions to facilitate the advancement of a microcatheter, perform the RotaWire exchange, and complete the PCI. This can also be used similarly by combining laser plus OA.

The combination of ELCA plus lithotripsy (the ELCATripsy technique) has been described for cases where RA or OA are ill-advised like nearby lesions or at freshly implanted stent level. In these cases, laser can create a tunnel through which the lithotripsy balloon can advance without the risk or damaging the freshly implanted stent.61

ALGORITHM FOR THE OPTIMAL MANAGEMENT OF CALCIFIED CORONARY ARTERY LESIONS

To select the most suitable plaque modification technique we need to become familiar with the characteristics of the different techniques available, their indications, and risks (table 6). Also, the patient’s clinical profile should be taken into consideration, as well as the characteristics of the lesion, the resources available, and the operator’s skills. In some complex cases, it can be reasonable to perform an ad hoc PCI for proper planning and even an angioplasty between 2 expert operators.

Table 6. Comparison of the different plaque modification techniques available

Current evidence available from comparative or clinical trials allowing us to select among the different plaque modification techniques available is very limited66,67 (table 7). Therefore, although several algorithms have been proposed on the type of calcium and the plaque modification technique that should be used,67 there are no clear indications in the routine clinical practice guidelines. Currently ongoing studies may bring us more in-depth information in the future.

Table 7. Main clinical trials on plaque modification techniques

In cases of mild angiographic calcification and proper balloon expansion, further plaque preparation prior to stenting may not be required. However, when angiographic calcification is moderate or severe, the use of intracoronary imaging modalities is advised for their great utility to plan the procedure and optimize results (figure 7).

Figure 7. Central figure. Calcified plaque modification algorithm. ELCA, Excimer laser coronary angioplasty; NC, noncompliant; OA, orbital atherectomy; RA, rotational atherectomy; SC, semicompliant; VHPB, very high-pressure balloon.
0 Predilatation with low-profile balloons can be attempted. At times, this allows early assessment with intravascular imaging tools.
1 Of choice if microcatheter is unable to cross.
2 If lithotripsy balloon is unable to cross, predilatation can be attempted with balloon or combination of other techniques (Rotatripsy, Elcatripsy, Orbital-tripsy).
3 Currently, lithotripsy is preferred in the presence of acute stent underexpansion.
4 In addition to NC balloon final angiographic expansion, intracoronary imaging are useful to confirm the effect of the techniques used on plaque modification.

Overall, it is useful to apply the “rule of 5”: lesions where calcium occupies < 50% of arc circumference (180º), does not extend longitudinally > 5 mm, and thickness is not > 0.5 mm can be properly treated with high-pressure or modified balloons (CB or SB).

If these criteria are met or calcium nodules are spotted further advanced plaque modification techniques should be used. In addition to circumferential and longitudinal spread, and thickness, calcium depth is important as well since some techniques like RA act basically on the superficial—and not on the deep—portion of calcium plaque.

Lesions with extremely severe calcifications so stenotic that cannot be crossed with the IVUS or OCT probe probably need RA/OA or laser (that can be of choice if the lesion is non-crossable not even with a microcatheter to allow specific RA/OA guidewire exchange). Another alternative is to try predilatation with low-profile balloons that often allow early assessments with intravascular coronary imaging to guide the decision-making process as already described.

Balloon expansion after using these techniques will guide us on proper plaque preparation. Also, intracoronary images are very useful to confirm proper calcium modification to allow stent expansion. The effects of different techniques like the presence of fractures (with balloon or lithotripsy), superficial calcium sanding (with RA) or both effects (with OA)70 can be visible when intracoronary imaging modalities are used (figure 6). After the use of ELCA, superficial and deep fractures have been described. However, effects may not be visible on the OCT and, same as it happens with ICL, that does not mean that the plaque has not been modified.

Based on the type of lesion and effects caused by these techniques, the combination of 1 or more of these techniques can be necessary to secure optimal stenting and favorable clinical outcomes.

CONCLUSIONS

Coronary artery calcification is probably the greatest determinant of poor PCI outcomes and incomplete percutaneous revascularizations, and is associated with higher rates of adverse events. Intracoronary imaging modalities play a key role in the understanding of calcified coronary artery lesions, help us select the plaque modification technique we’ll eventually use, and optimize the PCI results. Knowing the different plaque modification techniques available is essential for the optimal management of calcified coronary artery lesions. Until comparative trials among techniques are conducted, it seems reasonable to combine them depending on the type of lesion. In addition, there are situations in which techniques should be combined to secure optimal stenting and the most favorable clinical outcomes.

FUNDING

None whatsoever.

AUTHORS’ CONTRIBUTIONS

Manuscript drafting: A. Jurado-Román, A. Gómez-Menchero, N. Gonzalo, J. Martín-Moreiras, R. Ocaranza, Soledad Ojeda, J. Palazuelos, O. Rodríguez-Leor, P. Salinas, B. Vaquerizo, X. Freixa, and A.B. Cid-Álvarez. Design, coordination, review process of the final version, and manuscript submission: A. Jurado-Román.

CONFLICTS OF INTEREST

S. Ojeda is an associate editor of REC: Interventional Cardiology; the journal’s editorial procedure to ensure impartial handling of the manuscript has been followed. A. Jurado-Román, X. Freixa, and A. B. Cid-Álvarez are members of the ACI-SEC board of directors.

REFERENCES

INTRODUCTION: MATHETMATICS, PROBABILITY, AND STATISTICS

In the words of world famous astrophysicist Stephen Hawking, the goal of science is «nothing less than a complete description of the universe we live in».1 Male and female scientists alike pursue this objective by building theories and assessing their predictions. It is the very essence of the scientific method.

Statistics is a scientific discipline that designs experiments and learns from data. It formalizes the process of learning through observations and guides the use the knowledge accumulated in decision-making processes. Concepts like chance, uncertainty, and luck are almost as old as mankind, and reducing uncertainty has always been a common goal for most human civilizations. Probability is the mathematical language used to quantify uncertainty and is at the core of statistical learning that represents—in probabilistic terms—both the study populations and the random samples that come from such populations.

There is not such a thing as a single statistical methodology. The most widely known and used ones are, by far, frequentist statistics, and Bayesian statistics. Both share common goals and use probability as the language of statistical learning. However, both understand the concept probability different. As a matter of fact, it is the element on which they largely disagree. According to the frequentist concept, it is only legitimate to assign probabilities to random phenomena that can be defined through experiments that can be repeated multiple times, and only under identical and independent conditions.

The Bayesian concept of probability is a much wider idea because it allows us to assign probabilities to all elements with uncertainty regardless of their nature. Bayesian probability applies to the occurrence of random events, both those that can be repeated under the conditions required by frequentist probability and those that don’t (chances that Arnau, a 60-year-old male who lives alone will recover from a heart attack). The differences between both methodologies grow even larger because Bayesian probability assigns probabilities to different parameters (like the prevalence of people between the ages of 45 and 65 who have suffered a heart attack), statistical hypotheses (the efficacy of a new treatment for diabetic patients with heart failure is greater compared to conventional treatment), probabilistic models or even to missing data generated by non-randomized losses to follow-up (eg, ignoring the information of patients with losses to follow-up in a survival study on a given end-stage process would introduce biased information to the study).

The second distinctive element between both statistical methodologies is the use of Bayes’ theorem. For Bayesian statistics it is an essential tool to sequentially update the relevant information that comes from a study. Therefore, after an initial analytical phase, the knowledge generated will be used to start a new process of learning that will be providing new information on the problem at stake.

Both the frequentist and Bayesian concepts of probability share the same axiomatic system, and the same probabilistic properties. This common niche makes them share a common mathematical language too.

The map of basic Bayesian concepts and their different associations is not easy to explain without falling into a plethora of technicalities. And this is even more evident in real-world studies in the cardiovascular research setting. Therefore, in this article we will be working on very clear cases we believe are powerful examples regarding conceptual terms that are, nonetheless, simple, and devoid of technical complexities.

This article includes 6 different sections. The first one, this introduction, refers to the general wisdom regarding Bayesian statistics and its association with mathematics, probability, and statistics. The second section includes brief historical references on Bayesian methodology. The next section is about Bayes’ theorem in its most innocent version regarding the occurrence of random events. Afterwards, we’ll be dealing with the concepts and basic protocol of Bayesian statistics: previous distribution, function of verisimilitude, posterior distribution, and predictive distribution to predict experimental results. Also, we’ll include a brief explanation on the computational problems associated with the practical application of Bayesian methods and their power to generate inferences on relevant derived quantities. Hypothesis testing—the P value in particular—will also be dealt with later on as well as the Bayesian hypothesis testing proposal. The article will end with a small comment on the use of prior distributions.

IT ALL STARTED WITH BAYES, PRICE, AND LAPLACE

Knowing a little bit of Bayesian history is important because it allows us to put it into a temporal and social perspective that illuminates and boosts its learning. We’ll give a few relevant hints on this history now. McGrayne2 gives us an easy-to-understand and rigorous big picture on Bayesian history.

The very first time anybody heard of Bayes’ theorem was in Great Britain halfway into the 18^th century through Reverend Thomas Bayes while trying to prove the existence of God through mathematics. He would never dare to publish his findings. Prior to his death, he bequeathed all his savings to his friend Richard Price who—if okay with it—was supposed to spend this money to publish the findings, which is something he eventually did. However, these results went totally unnoticed.

We’re still in the 18^th century, but now we’ll have to travel to France to meet Pierre-Simon Laplace, one of the most prominent mathematicians in history. He discovered, independently of Bayes and Price, Bayes’ theorem in the format that we know today. Also, he developed the Bayesian concept of probability. After his death, his work fell into oblivion, under attack too because it was not in tune with the ruling idea of objectivity so embedded in the scientific world at the time.

Back to Great Britain now. Bletchley Park was a 19^th century mansion in Northern London turned into a working center to break the secret messages of the German army during the Second World War (1939-1945). Here Alan Turing and his team—that included the Bayesian statistician Jack Godd—played a key role in the history of Bayesian statistics: Bayes’ theorem was tremendously useful to decipher the code of the Enigma machines the Germans were using to code and decode messages. After the war, the British government classified all the information that had anything to do with Turing, mathematics, statistics, and decoding as top secret. Bayes’ theorem became a useful tool for just a handful of scientists, and an anathema (or worse) for most of them. As a truly revealing anecdote, McGrayne2 tells the story when Good presented the details of the method that Turing and his team had used to decipher the Nazi codes to members of Britain’s Royal Statistical Society. This is what the next speaker had to say about the whole thing: «After that nonsense […]».

During the second half of the 20^th century, the future of Bayesian statistics looked grim: support from the English-speaking academic world grew thin, and the rest of the scientific community knew very little about Bayesian statistics. Also, there were many computational difficulties to implement it to real-world studies with data. But what seemed to be destined to happened never did. We’re now back to the Second World War to Los Alamos National Laboratory in the state of New Mexico, United States. This center was created with absolute secrecy during Second World War to investigate the construction of nuclear weapons under the umbrella of the so-called Manhattan Project, led by the United States with the participation of Great Britain and Canada. It is in this context where the early Monte Carlo simulation methods were discovered back in 1946 by Polish mathematician Stanislaw Ulam while playing solitaire. Also, at that time, Metropolis et al.3 publish the first Markov chain Monte Carlo (MCMC) simulation algorithm while conducting his investigations on the H-bomb.

Several years go by without any direct links whatsoever between Bayesian statistics and MCMC methods. However, some studies are published—especially on image recognition—combining both elements.4 Encouraged by the technological advances made, especially in computing, Alan Gelfand, an American, and Adrian Smith, a British, collect former studies on MCMC methods and make a direct connection with Bayesian statistics.5 This will mark the beginning of the great Bayesian revolution that starts in the field of applications to, little by little, move on to the academic world. Bayesian inference is now recognized, accepted, and validated by the scientific community as a useful statistical methodology for scientific and social development.

BAYES’ THEOREM

The most widely known format of Bayes’ theorem is presented for the occurrence of random events. If A and B are random events, then

being p(A) the probability of event A, p(A|B) the associated probability A conditioned by the information that B occurred, and comparably p(B)> 0 and p(B|A). It is important to distinguish between probabilities p(A) and p(A|B). Both quantify the occurrence of A, but p(A) does so in absolute terms while p(B|A) does so in relative terms and conditioned by the information picked up in B. For example, nobody would question that the chances that a person may suffer from angina pectoris are higher if this person is hypertensive as opposed to not having that information available, p (Angina pectoris|Hypertension) > p (Angina pectoris).

Example I: Infections and tests

The prevalence that a certain infection affects any given population is .004. There is a test to detect its presence with a 94% sensitivity and a 97% specificity. We want to assess the chances that a person from such population is really infected if he tested positive for an infection.

We will use V and V^c to define a success that describes whether a person is infected or not, respectively. Therefore, p(V) = .004 and p(V^c) = .996. We’ll use (+) and (–) to describe a positive and negative test result for infection, respectively. In probabilistic terms, if a person is infected, he will test positive with a .94 probability, and negative with a .06 probability, p(+|V) = .94 and p(–|V) = .06 (false negative). If not infected, the person will test negative with a .97 probability and positive with a .03 probability, p(–|V^c) = .97 and p(+|V^c) = .03 (false positive).

According to Bayes’ theorem, the chances that a person is infected with a positive test result are

being p(+) = p(+|V) p(V) + p(+|V^c) p(V^c) = .0336 after implementing the total probability theorem (figure 1).

Figure 1. Posterior probability of infection with a positive test result p(V|+) (upper chart) and a negative test result p(V|–) (upper chart) in relation to the prior probability, p(V), of infection.

In principle, it is disconcerting that such a reliable test and with a positive test result for infection generates a small posterior probability, .112 favorable to the infection. However, if we take into consideration that the initial probability of being infected is p(V) = .004 and that, after the positive test result, probability is p(V|+) = .112, we’ll see that it has gone up from 4 to 112 by a thousand—has multiplied by a factor of 28—and we believe that the influence of the test result in such posterior probability is more relevant. Anyways, we would, at least, need a second test to increase the evidence for or against the infection.

Figure 2 shows 2 charts. The upper curve is the posterior probability of infection when the test is positive, p(V|+). The lower curve is the probability of infection too, but with a negative test result, p(V|–). In both cases, such posterior probabilities are represented in terms of prior probability, p(V), of being infected. When p(V) is close to 0, as it is the case with this example, the probability p(V|+) goes up a lot although, in absolute terms, it remains very low. On the contrary, when p(V) is close to 1, the probability p(V|–) will still be high despite evidence against a very reliable negative test result. The main element to understand this situation is that posterior probability, p(V|+) = .112, combines a very small probability of having an infection with a very high probability of testing positive when infected.

Figure 2. Approximate number of people infected, not infected, true positives, true negatives, false positives, and false negatives in a population of 100 000 inhabitants with an early prevalence for infection of .004 and a 94% sensitivity test and a 97% specificity test.

We’ll move on now to assess our results. The prevalence of infection, p(V) = .004 indicates that in a population of 100 000 people we should be expecting around 400 people infected and, approximately, 99 600 people not infected (figure 2). If the entire population were tested, we would expect to see that around 376 of the 400 people infected would test positive (true positives), as opposed to 24 (false negatives). In the group of healthy people, around 96 612 people would test negative (true negatives), but nearly 2988 people would test positive (false positives). If we looked at the number of people who tested positive, we’d have around 376 true positives, and 2988 false positives. Therefore, most people with a positive test (nearly 89%) would not actually be infected.

A second test with a positive result too would provide further evidence favorable to the infection. Its probability should be updated including the positive result of the second test as new information. If now (+1) and (+2) represent a positive test result for the first and second tests, the relevant probability would be p(V|+1,+2). The sequential use of Bayes’ theorem allows us to estimate such probability considering p(V|+1) = .112 as prior probability. The result obtained, p(V|+1,+2) = .798, is meaningful evidence favorable to an infection after 2 positive test results.

PARAMETER ESTIMATE

The true protagonists of basic Bayesian studies are probabilistic models governed by unknown parameters. These are at the core of Bayesian inferential machinery. We should mention that a parameter is a characteristic of a statistical population under study. Examples of parameters are the percentage of effectiveness of any given drug, the 5-year survival rate of soft tissue sarcoma, the basic reproduction factor R₀ of an infection, etc. Parameters are estimated using partial information of the study population from data samples obtained through random procedures that guarantee their representativity and small population condition.

The most widely known probabilistic model is normal distribution with a dome-shaped symmetrical density function and defined by 2 different parameters, mean, µ, and standard deviation (σ). Mean is the center of gravity of distribution and corresponds to the peak of the dome. Standard deviation is a dispersion measurement that determines the width of the dome: in all normal distributions, the interval (µ − 3σ, µ + 3σ) includes 99.7% of the values of distribution. Therefore, the interval-related probability (−3, 3) in a normal distribution with mean = 0 and standard deviation = 1 would be the same as the one associated with the interval (−6, +6) of a normal distribution with mean = 0 and standard deviation = 2 (figure 3).

Figure 3. Chart of normal density with mean of 0, and standard deviation of 1 (green color), and normal density with mean of 0, and standard deviation of 2 (gray color).

Mean and standard deviation are unknown parameters in most studies based on normal data. In our case, and to avoid any technical complications, we’ll assume that the standard deviation is known. Therefore, the statistical process will only have eyes from the mean µ. Bayes’ theorem adapts itself to the territory of probability distributions with focus on the population mean symbol µ as parameter of interest according to the following formula

being p(μ) the previous distribution (or prior distribution) of μ that quantifies, in probabilistic terms, the initial information available on μ and p(μ | data), the posterior distribution of µ that contains the information on µ available when the initial information is added to data. The term p(data | μ) is the verisimilitude function of μ, a measurement that assesses the compatibility of data with the possible μ values. The element p(data) is the previous predictive distribution (also evidence in the automatic and machine learning setting), and assesses the plausibility of the data obtained.

Example II: the heart of boys and girls with spinal muscular atrophy

Falsaperla et al.6 present the results of an observational study on the heart electrical conduction system disorder that causes bradycardia or electrocardiographic abnormalities in boys and girls with spinal muscular atrophy type 1 and 2 (SMA1, and SMA2, respectively). We gained our inspiration from this study to build a simulated database. Therefore, the results from this example do not come from Falsaperla et al.6 and should not be compared to those from the original study.

We simulated the data on the PR interval length that extends from the origin of atrial depolarization until the origin of ventricular depolarization from 14 children with SMA2. We assumed a normal model with unknown measurement and known standard deviation. Out statistical goal was to estimate the mean.

We’ll go on now with the Bayesian protocol. First, we need a previous distribution p(μ) to express our information of such parameter. Afterwards, we consider a scenario without new information on μ except for the information provided by data and use Jeffreys Prior to treat all possible μ values the same way.7 The posterior distribution of μ, p(μ | data), is a normal distribution with a mean of .13, and a standard deviation of .03/√∙14 seconds that we can graphically see on figure 4. We estimate that μ is .13 seconds, and directly assess the accuracy of such estimate through a credibility interval that tells us that the posterior probability of μ will be taking values between .114 and .146 seconds is .95. We give it a very low probability of .05 that μ will be > .146 or < .114.

Figure 4. Posterior distribution of the PR interval mean length in children with spinal muscular atrophy type 2.

A frequentist analysis of this data would never allow direct probabilistic assessments of μ. A frequentist 95% confidence interval for μ would provide the same numerical results compared to the Bayesian interval. However, it should be interpreted in a completely different way. The frequentist 95% confidence interval is on the capacity of the interval to include μ true value, and not on the possible μ values. The interval built, (.114, .146), has a .95 probability of capturing μ true value, but also a .05 probability of not doing so. We should remember that the frequentist concept of probability prevents allocating probabilities to parameters and establishing direct probabilistic assessments of μ.

PREDICTING NEW OBSERVATIONS

Prediction and estimation are fundamental statistical concepts. We estimate parameters, but we predict data and experimental results always through distributions of probability.

The posterior predictive distribution of the results of a future experiment is built by combining the probabilistic model that correlates future data and parameters with posterior distribution. A significant aspect of predictive process regarding estimation is its greater uncertainty. Overall, the precision of estimations improves with larger samples, which means that in the hypothetical case of having all the data available, our estimation would be accurate. The process of prediction does not have such a feature. Although the precision of predictions increases parallel to the size of the sample, in the hypothetical case of having access to all the information from a population, error-free predictions would still be impossible.

Example II: the heart of boys and girls with muscular atrophy (continues)

In the estimation stage we studied the PR interval mean length in girls and boys with SMA2, learning based on a sample of 14 simulated data. Now we’ll be dealing with totally different situation. We have a kid with SMA2 who did not participate in the study. Our objective is to predict his PR interval length. The goal now is not to estimate population means with SMA2, but to predict the value of the PR interval length in a particular kid.

Figure 5 shows the posterior predictive distribution of the PR interval length of a new boy with SMA2. Although this prediction is based on information from 14 children from the sample, it is about a new boy with SMA2. The anticipated value of this new child’s PR interval length is .13 seconds. The accuracy of prediction is quantified through prediction intervals. In this case with a .95 probability, the anticipated value will be between .069 and .191 seconds.

Figure 5. Posterior predictive distribution of the PR interval length of a new child with spinal muscular atrophy type 2.

SIMULATION AND GROUP COMPARISON

The Bayesian protocol with the 3 basic elements, previous distribution, verisimilitude function, and posterior distribution is common to almost all kinds of settings, both the basic ones including some parameters, as well as the complex ones with many sources of uncertainty with complex hierarchical structures. It is a robust and easy-to-use protocol, and a powerful and appealing idea.

Difficulties appear if we want to extrapolate this protocol to real studies of certain complexity. It is in just a couple of these cases that an analytical expression for the posterior distribution of parameters can be achieved. Impossible, nonetheless, for posterior distributions associated with derived quantities of interest. In most studies, mathematics becomes complicated, and posterior distributions are difficult to obtain. In these cases, the MCMC methods come to the rescue of Bayesian analysis. They can simulate our estimates of the relevant posterior distribution and from them generate the inferences or predictions required by the study.

In the following example we’ll be showing the most basic situation we have discussed: starting with a target posterior (analytical) distribution we’ll be simulating posterior distributions of relevant, non-analytical quantities of interest.

Example III: acute myocardial infarction and stents

The following example has been inspired by Iglesias et al.8. This is a study of 1300 patients with acute myocardial infarction treated with percutaneous coronary intervention. Each patient was randomized to sirolimus-eluting stent implantation with degradable polymer (group S) or everolimus-eluting stent implantation with durable polymer (group E).

We compared both treatments in relation to the rate of deaths 12 months after treatment. A total of 35 out of the 649 patients from group S stopped treatment or were lost within first year of follow-up, and 24 died. In group E, initially with 651 patients, 25 were lost or stopped treatment, and 22 died. The presence of missing data due to losses to follow-up is an important issue that should be dealt with carefully. In this case we’ll omit it because our goal is to illustrate Bayesian procedures using the least possible technicalities.

We’ll start by analyzing the risk of death θ_s and θ_E in groups S and E, respectively 1 year into treatment. Since anybody from either one of the 2 groups can die, or not, within the first year of treatment, in each group, the probabilistic model is binomial distribution that will be describing the number of deaths reported. The risk of death from each group is a rate with values that range between 0 and 1. For each rate we’ll be selecting beta distribution because it’s the proper probabilistic model to use with rates and doesn’t pose any estimate difficulties. β distribution (that we’ll representi as Be(α, β)) has 2 different parameters, α > 0 and β > 0, that determine the way of the distribution as well as its mean and variance. It is a flexible distribution that can be symmetrical or asymmetrical, positive or negative (figure 6).

Figure 6. Chart of β densities: Be(.5, .5) in green color, Be(2, 2) in gray color, Be(3, 22) in maroon color, Be(12, 2) in blue color, and Be(12, 12) in pumpkin color.

The prior β distribution that best describes the lack of information is Be(.5, .5). Its justification only responds to theoretical criteria. In each group, the posterior distribution of the risk of death will also be β whose updated parameters can be obtained by adding the number of deaths and the number of people alive in the study to the 2 values 0.5 and 0.5 of the prior beta distribution:

p(θ_s) = Be(.5, .5); p(θ_s | data) = Be(24.5, 590.5),

p(θ_E) = Be(.5, .5); p(θ_E | data) = Be(22.5, 604.5).

The estimation of the risk of death in patients from groups S and E is the mean of its posterior distribution, .040 and .036, respectively. Also, with a .95 probability the risk of death of group S is between .026 and .057, and between .023 and .052 in group E. These results indicate that the rate of death from both groups is small, although slightly higher in group S. The 95% credibility interval—both of θ_s and θ_Eis very informative (figure 7).

Figure 7. Posterior distribution of the mortality rate reported in patients from group S (green color) and group E (gray color).

We assume that our goal is to compare the 1-year mortality risk in both groups. Although the tool we could think of first is hypothesis testing (that we’ll introduce later on), the epidemiological and statistical literature on this regard is abundant, and 2 groups are often compared through relative risk (RR) or absolute risk (AR).9 The 1-year RR of mortality in patients with type S stents vs patients with type E stents is RR = θ_s/θ_E, a ratio between 2 rates. RR values < 1 are indicative that the mortality rate from group S is lower compared to group E while values > 1 are indicative of precisely the opposite. Since RR is defined through θ_s and θ_E, the information from both rates, expressed through its posterior distribution, can propagate to RR as a posterior probability distribution, p(RR | data) (figure 8). This distribution is not analytical, but it can come close to it through a Monte Carlo simulation from the 2 posterior distributions p(θ_s | data) and p(θ_E | data). Using this approximation, the posterior RR mean is 1.160, its standard deviation, .346, and the posterior probability of RR being >1 is .641. An analogue frequentist analysis would be more complicated from the mathematical standpoint and would not provide a direct probabilistic assessment of RR.

Figure 8. Posterior distribution of the relative risk of mortality at 1 year in patients with S-type stents vs patients with E-type stents. The shadow region is the probability, .641, that the mortality rate of group S is higher compared to group E.

If we compare both groups through the AR of mortality at 1 year, our goal would be AR = θ_s-θ_E. This is a difference between 2 rates that could take values between −1 and 1. Negative values will be indicative that the mortality rate of group E is higher compared to groups S while positive values will indicate just the opposite. figure 9 shows the AR posterior distribution.

Figure 9. Posterior distribution of the absolute risk of mortality at 1 year in patients with S-type stents vs patients with E-type stents. The shadow region is the probability, .641, that the mortality rate of group S is higher compared to group E.

The posterior AR mean is .004, its standard deviation, .011, and with a .641 probability, the AR will be > 0.

HYPOTHESIS TESTING: FREQUENTIST P VALUES

Hypothesis testing is the topic that generates the most irreconcilable differences between the Bayesian and the frequentist scientific communities because it is here where the consequences of their different concepts of probability become more evident. Testing hypothesis means testing new theories. Most of the latest theories have a quiet appearance among the scientific community, but little by little they start accumulating evidence in their favor until the sitting evidence is debunked.

The most widely known and used concept in frequentist statistics is the P value, as well as its .05 value that in some studies appears as the magical number used to reject or accept hypotheses or scientific theories. P value is another tool in the frequentist inference armamentarium that simply does not exist in the Bayesian one regarding the testing of 2 different hypotheses: the null hypothesis H₀ (that often represents the sitting scientific theory), and the alternative hypothesis H₁ (the new theory). P values are always associated with data because without the latter there are no P values. Under these conditions, P value is the probability that a certain theoretical summary of data will be equal to the one observed or more incompatible with the null hypothesis supposing that such hypothesis holds true. Such compatibility is often represented by the P = .05 threshold. P values ≥ .05 keep confidence in the null hypothesis; P values < .05, however, are favorable to the alternative one.

The excessive, and sometimes, inappropriate use of P values in scientific studies is still under discussion in the statistical community. It started in small scientific circles, but the use of the P value soon became a somehow «magical» element rather than a scientific tool. Back in 2014, the American Statistical Association, one of the world’s leading statistical societies, approached this issue and drafted a document that has become the go-to guideline on this topic.10 The following ones are some of the conclusions on significant P values for the management of biometrical data:

1. They are a probabilistic measurement of the compatibility of data with the null hypothesis. Smaller P values are associated with more data incompatibility with such hypothesis.
2. Do not assess the probability that a hypothesis will hold true or not.
3. The conclusions of a study should not only be based on whether a given P value exceeds this or that threshold. The use of the expression «statistically significant» (P < .05) to establish conclusions distorts all scientific procedures.
4. Do not measure the size of an effect or the significance of a given result. All small effects can produce small P values when the size of the sample or the accuracy of measurements is big, and all big effects can generate big P values with small samples or imprecise observations.

The P value has been given an unfair treatment because it has been attributed fantastical and surreal properties that have turned against it. Controversy has shattered the scientific debate and encouraged criticism in scientific disciplines that use data to generate knowledge. The huge interest in today’s scientific reproducibility topics owes volumes to this debate.11^-15

ACCUMULATING EVIDENCE FOR THE PROBABILISTIC ASSESSMENT OF NEW THEORIES

The Bayesian concept of probability is the key element to put hypotheses and theories to the test because it allows us to assign direct probabilities to both hypotheses and theories, both prior, p(theory holds true) and posterior, p(theory holds true|data ).16

Frequentist statistics is based on hypothesis testing using p-type probabilities (data|theory holds true) while Bayesian statistics is based on p-type probabilities (theory holds true|data). The p-type(data|theory holds true) frequentist probability assumes that the theory tested holds true, and based on that assumption assesses the concordance of data with such hypothesis. The p-type(theory holds true|data) Bayesian probability probabilistically assesses the certainty of the theory being tested in association with the data obtained.

The fundamental tool of Bayesian statistics to choose between hypotheses

H₀: theory #1 holds true,

H₁: theory #2 holds true,

based on a dataset is Bayes factor,17 the ratio between the probabilities associated with data according to both theories. It can also be expressed as the ratio between posterior odds (p(theory #1 holds true|data )/p(theory #2 holds true|data)) favorable to the certainty of theory #1 compared to theory #2 and the corresponding prior odds (p(theory #1 holds true)/p(theory #2 holds true). Like this:

Bayes factor (B) holds evidence favorable to the certainty of theory #1 (compared to theory #2) provided by data: it turns prior probabilities into posterior probabilities. In logarithmic scale, log (B), the Bayes factor is also known as «weight of evidence», a term coined by Turing back in Bletchley Park during Second World War. Small Bayes factor values give little support to H₀ vs H₁; however, big Bayes factor values provide extensive support to H₀.

Example I: Infections and tests (continues)

Let’s go back to the data from example I: Vallivana needs to be diagnosed on an infection with 2 positive test results. This problem can be faced as 2 hypotheses being tested:

H₀: Vallibana has an infection

H₁: Vallibana doesn’t have an infection

We assume that Vallibana does not have any particular characteristics that give her a probability of infection different from the remainder of the population. Therefore, we know that p(Vallibana has an infection) = .004, and p(Vallibana doesn’t have an infection) = .996. Vallibana’s prior odds favorrable to the infection compared to non-infection are:

Vallibana takes the test, and it turns out positive (+₁). She decides to retake it and tests positive again (+₂.). Vallibana’s posterior odds favorable to the infection compared to non-infection are:

The Bayes factor in favor of Vallibana being infected, that is, the ratio between the posterior odds and the prior odds is 987.75. Indeed, this value provides strong evidence in favor of Vallibana being infected (+).

Example II: the heart of boys and girls with spinal muscular atrophy (continues)

Now let’s go back to the study conducted by Falsaperla et al.6 from example II that aimed to compare the PR interval mean length in girls and boys with SMA1 and SMA2—that we’ll refer to as μ₁ and μ₂, respectively—through hypothesis testing:

H₀: μ₂ ≤ μ₁

H₁: μ₂ > μ₁

where the null hypothesis, H₀, claims that the PR interval mean length in girls and boys with SMA2 is shorter or equal to that reported in children with SMA1. The alternative hypothesis H₁ says otherwise. Here we’ll be working with simulated normal data in both groups: n₁ = 14 observations in the SMA1 group with sample mean and standard deviation values of .10 and .02 seconds, respectively, and n₂ = 14 observations in the SMA2 group with sample mean and standard deviation values of.13 and .03 seconds, respectively.

What we´ll do is build an inferential process for the mean of each group separately. In both cases, we’ll be considering a neutral previous distribution that gives all prominence to data. figure 10 shows the posterior distribution of the mean of each group. Both distributions are rather separate from one another, which means that the posterior probabilities associated with each hypothesis will be very different as well: .002 for H₀, and .998 for H₁.

Figure 10. Posterior distribution of the PR interval mean length in girls and boys with spinal muscular atrophy type 1 (green color), and type 2 (gray color).

p(H₀|data) = p(μ₂ ≤ μ₁|data) = .002

p(H₁|data) = p(μ₂ > μ₁|data) = .998

On a roughly basis, it is 500 times more likely that H₁ will hold true than not. In light of such an overwhelming piece of evidence the wise decision would be to choose H₁. The frequentist treatment of this testing is based on the P value. In our case, we’d obtain a P value of .002, which would imply rejecting the null hypothesis in favor of the alternative one. Both methodologies propose the same decision and provide the same numerical results: probabilities of .002. However, both probabilities are conceptually different. Bayesian probability tests the null hypothesis based on the data reported. Frequentist probability assesses the data observed with the assumption that the null hypothesis will hold true.

Still following in the footsteps of Falsaperla et al.6 we wish to mention that our examples are not based on original cases. They are merely illustrative of Bayesian procedures. We’ll now be working with the P-wave on the electrocardiogram. We wish to compare the P-wave mean length in children with SMA1 and SMA2. We’ll be simulating 14 observations of the P-wave length in the group of children with SMA1 and compared it to children with SMAs. The sample mean and standard deviation is .09 and .05 seconds in group SMA1, respectively, and .07 and .03 seconds in group SMA2. We’ll be comparing the means of both groups through hypothesis testing:

H₀: mean P-wave in SMA1 = mean P-wave in SMA2

H₁: mean P-wave in SMA1 > mean P-wave in SMA2

based on Bayesian inferential process like the one from the previous example. figure 11 shows the posterior distribution of the P-wave mean length in both groups. There are fewer data from the group of girls and boys with SMA2 compared to the group with SMA1. The posterior probability associated with each hypothesis is:

p(H₀ | data) = p(mean P-wave in SMA1 ≤ mean P-wave in SMA2 | data)

p(H₁ | data) = p(mean P-wave in SMA1 > mean P-wave in SMA2 | data)

These results provide a significant piece of evidence favorable to the alternative hypothesis that is almost 8 times more likely than H₀. From the frequentist standpoint, the P value associated with data would be .107 (>.05), which is why we would conclude that data does not provide enough evidence to reject H₀. Bayesian decision could perfectly be the same. However, the Bayesian analysis provides a direct assessment on the certainty of both hypotheses. The findings from the Bayesian analysis could be used as previous information in future studies with more data. Therefore, the posterior distributions obtained (figure 11) would be prior distributions in this new study. Bayes’ theorem allows us to generate knowledge sequentially searching for evidence for or against different hypothesis.

Figure 11. Posterior distribution of the P-wave mean length in girls and boys with spinal muscular atrophy type 1 (green color), and type 2 (gray color).

CARDIOLOGY POSES SOME OF DOUBTS ON THE IMPLEMENTATION OF METHODOLOGY IN CLINICAL TRIALS

One of the most controversial topics in Bayesian methodology is the selection of previous distributions. A Bayesian analysis will always allow us to avoid using any information on the amount of interest not provided by data. In this case, it works with previous distributions that play a neutral role in the learning process and that are useful only as the starting point of the Bayesian inferential protocol.

Previous informative distributions contain information that adds to the one provided by data like expert knowledge18^-20 or results from previous studies.21^,22 It is a highly valuable Bayesian characteristic in studies on which data is scarce like studies on rare diseases and orphan drugs. We should mention that inferential processes based on previous informative distributions should include sensitivity analyses of the results obtained regarding previously used distribution or distributions. Similarly, it has become popular to consider communities of previous distributions with diverse previous distributions—and a certain degree of skepticism or enthusiasm—with the effect under test because they provide a scientific framework of reference.

On many occasions, in clinical trials, the use of previous informative distributions reduces the sizes of frequentist samples based on preassigned values of the test power, and previous parameter estimates.22 An example of this situation is the BIOSTEMI trial.8 The 1300 patient-sample was estimated using Bayesian methods through a previous robust distribution as a mixture that included—in equal proportion—historical information of 407 patients from the BIOSCIENCE trial,23 and a practically non-informative distribution. The flexibility of Bayesian sequential learning is a key element of the so-called adaptative Bayesian designs24 that allow us to include additional information in different phases of the trial without damaging the consistency and reliability of results.

FUNDING

This study was partially funded by Biotronik Spain S. A, and by project PID2019-106341GB-I00 from the Spanish Ministry of Science and Innovation, and Universities of the Spanish Government.

AUTHORS’ CONTRIBUTIONS

C. Armero was involved in the structure, content, and drafting of this manuscript; P. Rodríguez, and J.M. de la Torre Hernández were actively involved in the review process of the manuscript final version.

CONFLICTS OF INTEREST

C. Armero, P. Rodríguez, and José M. de la Torre Hernández declared no conflicts of interest regarding the content, authorship, and publication of this manuscript. J.M. de la Torre Hernández is the editor-in-chief of REC: Interventional Cardiology. The journal’s editorial procedure to ensure impartial handling of the manuscript has been followed.

REFERENCES

1. Hawking S. A Brief History of Time: The Origin and Fate of the Universe. New York: Bantam; 1988.

2. McGrayne SB. La teoría que nunca murió: De cómo la regla de Bayes permitió descifrar el código Enigma, perseguir los submarinos rusos y emerger triunfante de dos siglos de controversia. Barcelona: Crítica; 2012.

3. Metropolis N, Rosenbluth A, Rosenbluth M, Teller A, Teller E. Equations of state calculations by fast computing machines. J Chem Phys. 1953;21:
1087-1092.

4. Robert CP, Casella G. A Short History of Markov Chain Monte Carlo: Subjective Recollections from Incomplete Data. Stat Sci. 2011;26:102-115.

5. Gelfand AE, Smith AFM. Sampling-based approaches to calculating marginal densities. J Am Stat Assoc. 1990;85:398-409.

6. Falsaperla R, Vitaliti G, Collotta AD, et al. Electrocardiographic Evaluation in Patients With Spinal Muscular Atrophy: A Case-Control Study. J Child Neurol. 2018;33:487-492.

7. Robert CP, Chopin N, Rousseau J. Harold Jeffreys’s Theory of Probability Revisited. Stat Sci. 2009;24:141-172.

8. Iglesias JF, Muller O, Heg D, et al. Biodegradable polymer sirolimus-eluting stents versus durable polymer everolimus-eluting stents in patients with ST-segment elevation myocardial infarction (BIOSTEMI): a single-blind, prospective, randomised superiority trial. Lancet. 2019;394:1243-1253.

9. Christensen R, Johnson W, Branscum A, Hanson TE. Bayesian Ideas and Data Analysis: An Introduction for Scientists and Statisticians. Boca Raton: CRC Press; 2011.

10. Wasserstein RL, Lazar NA. The ASA Statement on p-Values: Context, Process, and Purpose. Am Stat. 2016;70:129-133.

11. Goodman SN. Toward Evidence-Based Medical Statistics. 1: The P Value Fallacy. Ann Intern Med. 1999;130:995-1004.

12. Greenland S, Senn SJ, Rothman KJ, et al. Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. Eur J Epidemiol. 2016;31:337-350.

13. Halsey LG, Curran-Everett D, Vowler SL, Drummond GB. The fickle P value generates irreproducible results. Nat Methods. 2015;12:179-185.

14. Ioannidis JPA. Why Most Published Research Findings Are False. PLoS Med. 2005;2:e124.

15. Ioannidis JPA. The Proposal to Lower P Value Thresholds to .005. J Am Med Assoc. 2018;319:1429-1430.

16. Kruschke JK. Doing Bayesian data analysis: A Tutorial with R, JAGS, and Stan. 2nd ed. Amsterdam: Academic Press/Elsevier; 2015.

17. Kass RE, Raftery AE. Bayes Factors. J Am Stat Assoc. 1995;90:773-795.

18. Hampson LV, Whitehead J, Eleftheriou D, et al. Elicitation of Expert Prior Opinion: Application to the MYPAN Trial in Childhood Polyarteritis Nodosa. PLoS One. 2015;10:e0120981.

19. Mason AJ, Gomes M, Grieve R, Ulug P, Powell JT, Carpenter J. Development of a practical approach to expert elicitation for randomised controlled trials with missing health outcomes: Application to the IMPROVE trial. Clin Trials. 2017;14:357-367.

20. Jansen JO, Wang H, Holcomb JB, et al. Elicitation of prior probability distributions for a proposed Bayesian randomized clinical trial of whole blood for trauma resuscitation. Transfusion. 2020;60:498-506.

21. Grant RL. The uptake of Bayesian methods in biomedical meta-analyses: A scoping review (2005–2016). J Evidence-Based Med. 2019;12:69-75.

22. Spiegelhalter DJ, Abrams KR, Myles JP. Bayesian Approaches to Clinical Trials and Health-Care Evaluation. Chichester: Wiley; 2004.

23. Pilgrim T, Heg D, Roffi M, et al. Ultrathin strut biodegradable polymer sirolimus-eluting stent versus durable polymer everolimus-eluting stent for percutaneous coronary revascularisation (BIOSCIENCE): a randomised, single-blind, non- inferiority trial. Lancet. 2014;384:2111-2122.

24. Schmidli H, Gsteiger S, Roychoudhury A, O’Hagan A, Spiegelhalter D, Neuenschwander B. Robust meta-analytic-predictive priors in clinical trials with historical control information. Biometrics. 2014;70:1023-1032.

INTRODUCTION

The role of the sympathetic nervous system in the pathophysiology of hypertension (HTN) is well known. In 2007, to address the unmet need of patients with resistant HTN (R-HTN), the first percutaneous renal sympathetic denervation (RSD) procedures were performed. First observational studies showed positive results, and the use of RSD started in select centers around the world.1,2 However, in 2014, the publication of a study which did not demonstrate a greater efficacy of RSD vs a sham-control group to control blood pressure (BP)3 dramatically reduced the interest of the scientific community in this procedure, as well as in its clinical application. An increased knowledge of renal anatomy combined with the development of second-generation devices has led to new studies, in which the efficacy of RSD vs a sham-control group has been demonstrated.4-7 Although the road ahead is long, the new evidence provides a clear role for RSD in the management of patients with HTN.

The clinical practice guidelines on the management of hypertension published by the European Society of Cardiology and European Society of Hypertension (ESC/ESH) back in 2018 outlined the role of device-based approaches for the management of HTN in the context of clinical trials only.8 The practical effect of this is that it discouraged the use of RSD. Despite the short time that has gone by since the publication of these guidelines, the data provided by the new clinical trials would justify treating selected patients with RSD.

This document reviews the evidence available on RSD for the management of HTN, analyzes possible indications, and suggests strategies to identify potentially eligible patients, formulated from the opinion of a panel of experts selected by the Spanish Society of Arterial Hypertension-Spanish League for the fight against Arterial Hypertension (SEH-LELHA), and the Interventional Cardiology Association of the Spanish Society of Cardiology (ACI-SEC). The writing of the document was carried out by professionals proposed by these scientific societies undersigning this work following their experience in the management of patients treated with RSD. After the first draft, other experts with and without previous experience in RSD carried out a critical review of the document, and agreed on the changes that were deemed appropriate.

CLINICAL EVIDENCE ON THE ROLE OF RENAL SYMPATHIC DENERVATION IN THE MANAGEMENT OF HYPERTENSION

The supplementary data shows the epidemiological characteristics of HTN (section 1), as well as the role of sympathetic nervous system in the management of HTN (section 2), thus improving our understanding of clinical trials. Table 1 of the supplementary data shows the main studies in which the efficacy of RSD has been assessed.

Table 1. Studies prior to renal sympathetic denervation in patients with uncontrolled hypertension

Back in 2009, the first study on RSD in patients with R-HTN, the SYMPLICITY HTN-1 trial, was published. This study suggested a high efficacy profile of RSD, and a lower office systolic blood pressure (SBP) down to 27 mmHg at 12 months, without any significant complications being reported.1

The SYMPLICITY HTN-3 trial was the first one to include a control group with a sham procedure and a 24-hour BP endpoint. No differences between the 2 treatment groups in terms of the efficacy profile of BP control in patients with R-HTN were reported at the 6-month follow-up.3 The disagreement between the results of this study and the previous ones, as well as the identification of several confounding factors9 brings up the need for designing new studies specifically aimed at solving these questions.

Definitive evidence on the efficacy of RSD has come from the SPYRAL HTN and RADIANCE-HTN studies. The SPYRAL HTN-ON MED trial enrolled patients with uncontrolled HTN treated with 1 to 3 antihypertensive drugs, randomized to receive RSD or a sham procedure. The 24-hour ambulatory SBP and diastolic BP (DBP) levels, as well as the office SBP and DBP levels dropped significantly in the RSD group compared to the sham control at the 6-month follow-up.4 With a similar design, the SPYRAL HTN-OFF MED Pivotal trial enrolled uncontrolled hypertensive patients with office SBP levels between 150 mmHg and 180 mmHg in the absence of antihypertensive treatment. The 24-hour SBP and office SBP levels were reported at the 3-month follow-up.5 The RADIANCE-HTN SOLO trial enrolled patients with HTN and ambulatory blood pressure monitoring (ABPM) levels ≥ 135/85 mmHg and ≤ 170/105 mmHg without pharmacological treatment. The 24-hour ambulatory SBP and DBP levels as well as the office SBP and DBP levels dropped significantly in the RSD group compared to the sham control at the 2-month follow-up.6 The RADIANCE-HTN TRIO trial enrolled patients with R-HTN on a fixed-dose, single-pill combination of a calcium channel blocker, an angiotensin receptor blocker, plus a thiazide diuretic, randomized to receive ultrasound catheter-based RSD or a sham procedure. The 24-hour ambulatory SBP and DBP levels, as well as office SBP and DBP levels dropped significantly in the RSD group compared to the sham control at the 2-month follow-up.7

Real-life registries have enrolled more than 3500 patients treated with RSD showing lower office BP and ABPM levels. Some registries have demonstrated that the reduction of BP is not associated with the medication burden or with an increased number of antihypertensive drugs. RSD has proven to be safe and has a low rate of complications associated with the procedure.10 The GLOBAL SYMPLICITY registry, with over 2900 patients is the largest and longest duration analysis to this date of renal sympathetic denervation to show the efficacy and safety profile of RSD in a real-life scenario.10 Table 2 of the supplementary data shows a summary of different registries on RSD.

Table 2. Causes of secondary hypertension

RSD has been confirmed as a safe intervention. The incidence rate of both immediate complications associated with the procedure and renal and vascular complications in the short- and mid-term (6-12 months) is very low and is mainly associated with local problems at the puncture site; serious renal complications (renal artery dissection or stenosis) are anecdotal. Table 3 of the supplementary data summarizes the safety data from the main randomized clinical trials that often have a short-term clinical follow-up.

Table 3. Precautions and contraindications to renal sympathetic denervation

POSSIBLE INDICATIONS FOR RENAL SYMPATHETIC DENERVATION WITH DATA FROM THE LATEST CLINICAL TRIALS

Data from both randomized clinical trials and registries prove that the RSD procedure is safe and effective reducing BP, which is consistent across different populations including high-risk subgroups, and with different devices. Section 3 of the supplementary data reviews various consensus documents and recommendations previously published by different scientific societies prior to the publication of the SPYRAL HTN and RANDIANCE-HTN clinical trials.

RSD can be considered in patients with resistant HTN (BP > 140/90 mmHg despite lifestyle changes treated with ≥ 3 antihypertensive drugs at optimal doses, one of them being a diuretic or HTN controlled with ≥ 4 drugs),8 and also in patients with uncontrolled HTN (BP > 140/90 mmHg in patients with poor therapeutic compliance), and high cardiovascular risk.

Renal sympathetic denervation in patients with resistant hypertension

Patients with R-HTN were the first group in whom the role of RSD was assessed. The SYMPLICITY HTN-3 trial failed to demonstrate the increased efficacy of RSD vs sham control in patients with R-HTN.3 However, subsequent analysis revealed design and execution limitations that cast doubts on the reliability of the results.9 In the recently published RADIANCE-HTN TRIO trial, patients with R-HTN treated with a standardized triple combination pill experienced a drop in their BP levels 2 months after RSD compared to a sham procedure.7 If the BP lowering effect and the safety of RSD are maintained in the long term, RSD might be an alternative to the addition of more antihypertensive medications in patients with R-HTN.

Renal sympathetic denervation in patients with uncontrolled hypertension

The new evidence available introduces a paradigm shift for a technique that was initially conceived for the management of R-HTN when all other therapeutic options fail, and is currently an option that should be taken into consideration in patients with persistent BP > 140/90 mmHg despite drug treatment.

The concept of uncontrolled HTN includes a high percentage of hypertensive patients (maybe even > 60%) with highly heterogenous clinical characteristics and cardiovascular risk. Given the invasive nature of the RSD procedure, and until more information becomes available on the reduction of cardiovascular events in more specific subgroups of patients, there are some high-risk situations in which BP control is essential to reduce the risk of cardiovascular events:

a) Patients with frequent hypertensive crises. Hypertensive crises with SBP levels > 180 mmHg and/or DBP levels > 110 mmHg can cause brain, cardiac or microvascular damage. Emergency visits for hypertensive crises exceed 4% of all visits to the emergency room.11 Even in the absence of hypertension-mediated organ damage (HMOD), episodes of hypertensive crisis can have long-term implications to the extent that these patients may have a 50% higher risk of suffering cardiovascular events compared to controlled hypertensive patients. Nonetheless, outside the crisis setting they show similar BP levels.12

b) Patients with low compliance to pharmacological treatment. Pharmacological treatment of HTN is generally a long-term option and in most cases, for life. Poor compliance is a common problem to the extent that almost one third of all hypertensive patients do not start a new prescription of antihypertensive drugs,13 and around 50% become non-compliant within the first year after starting treatment.13 In the SPYRAL HTN trials, the 24-hour ABPM levels showed decreased BP levels throughout the entire 24-hour period in patients treated with RSD compared to no changes in the control group in the absence of drugs or incomplete control in the presence of drugs.4,5 Furthermore, in the SPYRAL HTN OFF-MED trial, the treatment group experienced an average reduction of 9.2 mmHg in office SBP levels.5 A meta-analysis of 123 studies including 613 815 patients showed that a drop of office SBP levels of 10 mmHg was associated with a significantly lower risk of cardiovascular events.14 Poor compliance is a serious problem of public health since these patients in whom an adequate BP control is not achieved, even due to poor therapeutic compliance, have a high cardiovascular risk.15 However, we should stress that RSD alone cannot bring BP levels down enough to achieve BP control in most patients. In the RADIANCE-HTN SOLO trial, the 24-h ABPM only 25% of the patients treated with RSD reached values < 130/80 mmHg.6 In these non-compliant patients, the main strength of RSD is the “always on” effect regardless of pharmacokinetics and compliance to drugs.

c) Patients with hypertension-mediated organ damage. The presence of HMOD identifies a group of patients with high cardiovascular risk in whom conventional treatment has failed to prevent the progression of the disease.16 Achieving the BP levels recommended is especially important in these patients because, in the early stages of the disease, some types of HMOD can be reversed; in more advanced stages, HMOD is irreversible despite adequate BP control. But this is important since it slows its progression while reducing the cardiovascular risk of these high-risk patients.17 A meta-analysis including 698 patients treated with RSD revealed an independent effect of RSD on HMOD, which advocates for the use of RSD in this group of high-risk patients.18

d) Patients at high cardiovascular risk. The European guidelines on the management of HTN establish the factors that influence cardiovascular risk in hypertensive patients including clinical characteristics, analytical characteristics, presence of HMOD or established cardiovascular or kidney disease. All these factors establish a 10-year cardiovascular risk that is categorized into 4 groups: low, moderate, high or very high risk to the extent that, for example, in high-risk patients the estimated cardiovascular mortality is 5% and in very high-risk patients > 10%.8 The assessment of cardiovascular risk should play an important role in the decision-making process to the extent that the higher the risk, the greater the benefits expected with better BP control. Therefore patients at high or very high-risk would be eligible for RSD whenever BP control is not adequate.

Empowering the hypertensive patient in the setting of a shared decision-making process

Over the last few years, shared decision-making process has emerged as the go-to model in the management of different conditions. In the field of RSD, a recent survey revealed that 38% of hypertensive patients who still don’t take antihypertensive medication would prefer RSD to lifelong drug therapy even knowing that it would probably not replace medication in many cases. Just this already reduces BP significantly.19 With the evidence provided in recent trials, RSD could be a valid treatment option in patients with uncontrolled HTN and high to very-high cardiovascular risk in whom, in a shared decision-making process context, consensus with the patient can be reached. In any case, we should mention that the treatment of HTN always requires the adoption of healthy lifestyle habits, and the recommendation to patients should include drug treatment as the first option.

STUDY PRIOR TO RENAL SYMPATHETIC DENERVATION

Patients should be examined in a unit specialized in HTN and vascular risk 3 months prior to the procedure in a center with proven experience.20 Table 1 summarizes the studies to be conducted in patients eligible for RSD.

Uncontrolled HTN should be confirmed through 24-hour ABPM.21 After confirming the presence of uncontrolled HTN, the clinical situations that increase BP levels such as obesity or obstructive sleep apnea should be identified and corrected. Also, substances such as salt or certain drugs that may also lead to HTN should be suspended or minimized. Non-compliance to treatment, which is very common and not always identified by the patient, if not rigorously investigated, should be ruled out.22 It is essential to rule out secondary HTN (table 2) or, if diagnosed, treat it effectively. Still, it is not an absolute contraindication to RSD.23

RENAL SYMPATHETIC DENERVATION PROCEDURE WITH RADIOFREQUENCY DEVICES

Section 4 of the supplementary data shows more in-depth technical aspects of RSD. Figure 1 of the supplementary data summarizes the RSD procedure.

A better knowledge of the anatomy of renal nerves24 and the development of new ablation devices have optimized the treatment technique,5,6 which is based on 3 main objectives:

Figure 1. Detail of renal sympathetic innervation. Renal nerves are usually arranged in large bundles and only form a true plexus when they are close to entering the kidney. Some nerves bypass the main renal artery and join distally to the different arterial divisions of the main renal artery (late arrival nerves). In this case, a late arrival nerve is seen joining the proximal third of the anterior division of the main renal artery (blue asterisk). It can also be seen how the proximal main renal artery is occupied by fused ganglia of the solar plexus (GM), and by the lumbar sympathetic chain (LSC). Both provide innervation to the kidneys, but also to other abdominal and pelvic organs, which can be accidentally denervated if the proximal third of the main renal artery is treated. The image also shows that the maximum proximity of nerve fibers to the arterial wall mainly occurs at branch level, but also at main trunk level. This is the target area of treatment, always avoiding the application of radiofrequency at renal pelvis level. AG, adrenal gland; CT, celiac trunk; GM, ganglionic mass made up of the aorticorenal and celiac ganglia; LSC, lumbar sympathetic chain; RK, right kidney; SMA, superior mesenteric artery; blue asterisk, late arrival nerve. In red, arterial structures. In yellow, nervous tissue.

Management of the renal artery main trunk and branches

It is common for the renal nerves to reach the kidney after bypassing the main renal artery.24 In animal models, it has also been confirmed that the application of combined radiofrequency in the renal artery main trunk and branches reduced the content of norepinephrine in the renal tissue even more, and in the cortical axonal density, both associated with the response to RSD.25

In patients treated with RSD, the presence of untreated accessory arteries leads to a lower hypotensive response.26 Their identification and treatment is essential and, if they are amenable to treatment thanks to their diameter (minimum diameters of 3 mm), the treatment of accessory arteries is advised.

Last but not least, the perivascular space around the ostium and the proximal third of the main renal artery is often occupied by ganglia of solar plexus and by the lumbar sympathetic chain (figure 1). Both carry innervation to the kidneys, but also to other abdominal and pelvic organs, and they could be accidentally denervated if the treatment is applied to the ostium and proximal third of the main renal artery. Therefore, until more information becomes available, it seems reasonable to be cautious when treating the most ostial portion of renal arteries.24

Treatment of the 4 quadrants of the renal artery

The distribution of nerve fibers around the renal artery follows a variable pattern across different individuals.27 Preclinical studies in a porcine model have shown that the application of radiofrequency in one point produces effects on approximately 25% of the arterial circumference,27 and procedures that use multiple helically staggered ablations in the 4 quadrants are more effective reducing the norepinephrine content into the renal tissue.28

Application of the maximum possible number of ablation points

A post-hoc analysis of the SYMPLICITY HTN-3 trial confirmed that patients with a greater number of radiofrequency applications reduced their BP levels even more without any associated adverse events.9 We recommend applying the maximum number of ablation points possible, always respecting a distance of 5 mm among them with a 4-quadrant distribution.

Section 4 of the supplementary data shows how to perform a RSD procedure using a tetrapolar radiofrequency catheter. Table 3 shows the precautions and contraindications regarding RSD.

Care after renal sympathetic denervation procedure

Once the procedure is finished, it is important to ensure adequate hemostasis in the femoral puncture. Usually, in the absence of complications, patients can be discharged after 24 to 48 hours with the same antihypertensive treatment they had before the procedure or with treatment adjustments in cases that show an immediate response, but still with adjustment appointments within 5 to 7 days. Of note, the effects of the intervention can take weeks to materialize.29

CLINICAL MANAGEMENT AFTER RENAL SYMPATHETIC DENERVATION

The main objective of the follow-up should be to confirm the safety of the intervention and the absence of complications in the short-, mid-, and long-term follow-up, as well as to monitor the evolution of BP levels and the adjustment of drug treatment.

At the clinical follow-up, it is important to maintain a multidisciplinary team same as during the selection of candidates. Table 4 shows the clinical management after RSD.

Table 4. Clinical management after renal sympathetic denervation*

REQUIREMENTS OF A RENAL SYMPATHETIC DENERVATION PROGRAM

The success of a RSD program is based on the existence of a multidisciplinary team that performs a comprehensive assessment of the patient from the selection of candidates through their assessment prior to the intervention, the RSD procedure, and subsequent follow-up. This process should be carried out at specific units specialized in the management of HTN in collaboration with interventional cardiology units. Figure 2 shows the selection process of eligible patients.

Figure 2. Identification process, patient selection and decision on RDN. Patients with uncontrolled HTN (BP > 140/90 mmHg despite treatment) should be evaluated in an HTN unit. The lack of control should be confirmed by ABPM, assess adherence/intolerance to drugs, rule out secondary causes and cardiovascular risk. If, after optimizing the treatment, the lack of control persists, in patients at high or very high risk, and in a shared-decision process with the patient, RDN may be indicated. Adherence is defined as the extent to which a person’s behavior —taking medication, following a diet, and/or executing lifestyle changes— corresponds with the agreed recommendations from a healthcare provider. Drug intolerance refers to an inability to tolerate the adverse effects of a medication, generally at therapeutic or subtherapeutic doses. Treatment optimization refers to lifestyle changes and pharmacological recommendations, including target doses, recommended by clinical practice guidelines.8 ABPM, ambulatory blood pressure monitoring; BP, blood pressure; RSD: renal sympathetic denervation.

We strongly discourage isolated procedures outside this controlled environment. RSD should not be performed in centers with volumes < 10 cases/year. Centers without a structured RSD program, but with eligible patients, should refer them to an experienced center rather than performing isolated procedures.

RSD procedures should be performed by operators experienced in the management of endovascular treatment. The SYMPLICITY HTN-3 trial post-hoc analysis showed the importance of an experienced interventional specialist given one of the factors influencing the results of the study was the operator’s lack of experience.9 Therefore, we recommend that procedures should be performed at centers with proven experience only and that, in centers that lack this experience, the possibility of monitoring should be available including assistance during the patient selection process and supervision of the procedure until enough experience is gained to ensure optimal results.

CONCLUSIONS

This expert consensus document has reviewed the information available regarding RSD in the management of patients with HTN. Also, it has established, for the first time, the indication for RSD in cases of uncontrolled HTN, especially in patients at high cardiovascular risk with HMOD or cardiovascular disease while taking the patient’s opinion into consideration as part of a shared decision-making process, and as long as it is evaluated by a multidisciplinary team and performed by experienced operators.

FUNDING

None whatsoever.

AUTHORS’ CONTRIBUTIONS

Study concept and design: O. Rodríguez-Leor, and J.A. García-Donaire; manuscript writing: O. Rodríguez-Leor, F. Jaén-Águila, J. Segura, I. J. Nuñez-Gil, A. García-Touchard, E. Rubio, M. Troya, J. Diego-Mediavilla, and J.A. García-Donaire; critical review: O. Rodríguez-Leor, Á. Cequier, R. Moreno, N. Martell, P. Beltrán, and E. Molina.

CONFLICTS OF INTEREST

O. Rodríguez-Leor, and J. A. García-Donaire have received personal fees from Medtronic, outside the submitted work. A. García-Touchard reports having received grants from Medtronic, and personal fees from Medtronic, also outside the submitted work. Á. Cequier reports having received grants and personal fees from Abbott Vascular, grants, and personal fees from Biosensors, grants from Boston Scientific, grants, and personal fees from Medtronic, grants from Biomenco, Cordis, Orbus Neich, and from the Spain Society of Cardiology, personal fees from Ferrer International, Terumo, Astra Zeneca, and from Biotronik outside the submitted work. R. Moreno is associate editor of REC: Interventional Cardiology. The journal’s editorial procedure to ensure impartial handling of the manuscript has been followed. F. Jaén-Águila, J. Segura, I. J. Núñez-Gil, E. Rubio, M. Troya, J. Diego-Mediavilla, R. Moreno, N. Martell, P. Beltrán, and E. Molina declared no relationship whatsoever relevant to the contents of this paper worthy of disclosure.

SUPPLEMENTARY DATA

REFERENCES

INTRODUCTION

Coronary vasoreactivity testing performed by intracoronary infusion of acetylcholine is basically used with 2 goals in mind: for endothelial function assessment and as a vasospasm provocation test in clinically suspicious cases. Although these tests have been known and used for decades, its use is not yet fully consolidated in our setting. This is mainly due to a scarce suspicion of myocardial ischemia due to micro or vasomotor disorders that has eventually lowered the demand for these tests. In addition, the lack of test standardization and training, the off-label use of acetylcholine, and the doubts surrounding these tests safety profile have not encouraged their widespread use in the routine clinical practice.

Over the last few years, this scenario has changed dramatically thanks to the growing evidence on the importance of diagnosing the causes of myocardial ischemia not directly related to fixed stenoses. Currently, invasive coronary spasm provocation tests are formally recommended by the European Society of Cardiology in its clinical practice guidelines on the management of chronic coronary syndromes, non-ST-segment elevation acute coronary syndromes, and ST-segment elevation acute coronary syndromes.1-3 The most common indications are to treat patients with angina or ischemia but without non-obstructive coronary lesions (ANOCA, INOCA—in this document, both coined under the term INOCA—), myocardial infarction with non-obstructive coronary arteries (MINOCA), persistent angina after coronary revascularization, obstructive coronary artery disease with clinical suspicion of associated angina of microvascular origin, and finally, patients with recovered sudden death of undetermined causes.1-3 Table 1 summarizes all clinical indications and level of recommendation to perform vasospasm provocation testing. Although this paper focuses on coronary vasoreactivity testing, we should remember that its use is often recommended simultaneously with other coronary functional testing performed using pressure guidewires like coronary flow reserve and microcirculation resistance measurements.1-5 The specific diagnosis of functional damage to coronary arteries and its targeted therapies have both improved the quality of life of patients with INOCA.6 The treatment recommended for coronary vasospasm is calcium channel blockers, nitrates, and nicorandil.6,7

Table 1. Clinical indications for the coronary vasospasm provocation test performed by intracoronary infusion of acetylcholine

Based on the current clinical practice guidelines, the objective of the Working Group on Intracoronary Diagnostic Techniques of the Interventional Cardiology Association of the Spanish Society of Cardiology (ACI-SEC) is to facilitate and standardize the use of coronary vasoreactivity testing. Thus, this paper has been drafted to expose all technical steps in a practical way to encourage the performance and interpretation of these tests in our setting.

NORMAL ENDOTHELIAL FUNCTION OF CORONARY ARTERIES

The modulatory function of vascular endothelium in the blood flow towards the myocardium is intrinsically associated with its metabolic characteristics. Compared to the skeletal muscle, the heart has pretty high oxygen needs (some 20 times higher). The way this oxygen supply is achieved is through very high tissue extraction at baseline: at rest the myocardium extracts approximately 70% to 80% of the oxygen transported by hemoglobin compared to 30% of the skeletal muscle. This explains why, unlike other organs, the mechanism through which the heart regulates oxygen supply to the myocardium changing metabolic needs is fast regulation and constant blood flow into the coronary system.8

Microcirculation (arteries and arterioles < 400 µm) is basically responsible for regulating coronary blood flow. Although regulation is complex and includes metabolites, hormones, neurotransmitters, and other factors, the main protagonist is the vascular endothelium that produces nitric oxide—a powerful vasodilator—in response to different stimuli. Also, other vasodilator factors—like the hyperpolarizing endothelial factor—and vasoconstrictor factors like endothelin. Endothelium-dependent vasodilation can be stimulated through different ways, but the most commonly used one is the infusion of acetylcholine.

In normal conditions, an artery with a healthy endothelium responds to acetylcholine by releasing nitric oxide that translates into vasodilation. In the presence of artery denudation from the endothelium or if the action of the nitric-oxide synthase enzyme is blocked, the artery responds to acetylcholine with vasoconstriction due to the stimulation of smooth muscle muscarinic receptors not counteracted by the nitric oxide of endothelial origin. Therefore, the infusion of acetylcholine can be used to assess endothelial function: if normal, vasodilation becomes evident. If not, vasoconstriction kicks in. The macrovascular compartment endothelial function (epicardial) can be assessed on an angiography. However, to assess the endothelium-dependent response in microcirculation, blood flow should be measured using a Doppler guidewire or thermodilution. From the macrovascular point of view, visually evident epicardial vessel vasoconstriction in response to acetylcholine is considered endothelial dysfunction. Figure 1 shows examples of vasodilation (physiological) and vasoconstriction responses (suggestive of endothelial dysfunction) to the administration of acetylcholine. From the microvascular point of view, flow reductions or increases < 50% in response to the administration of acetylcholine are considered anomalous.9 Figure 2 shows examples of microvascular function assessment using the Doppler technique or intracoronary thermodilution.

Figure 1. Possible results of the vasospasm provocation test with acetylcholine. Case 1: physiological response (vasodilator response to acetylcholine). Case 2: endothelial dysfunction with vasoconstriction with respect to baseline that does not meet the criteria for micro or macrovascular spasm. Case 3: macrovascular spasm with significant vasoconstriction of the left coronary tree. Case 4: microvascular spasm due to moderate vasoconstriction of the left tree meeting the clinical and ECG criteria for ischemia. ACh, acetylcholine; DS, diameter stenosis; NTG, nitroglycerin.

Figure 2. Combined assessment of macro and microvascular function. A: pressure-Doppler guidewire study (Combowire, Philips, The Netherlands). After the angiography and measurement of baseline flow velocity, growing doses of acetylcholine are injected. After the second dose, moderate vasoconstriction of the left anterior descending coronary artery occurs followed by an occlusive spasm at circumflex artery level plus a slower flow velocity in the left anterior descending coronary artery indicative of microvascular vasoconstriction. The spasm is solved with intracoronary nitroglycerin followed by the adenosine non-endothelium-dependent assessment of microvascular function. The patient shows macro and microvascular endothelial dysfunction and a normal non-endothelium-dependent microvascular function. B: macrovascular endothelial function assessment appears normal; adenosine non-endothelium-dependent assessment with thermodilution guidewire (Pressurewire X, Abbott, United States). Coronary flow reserve is 5.9, and the index of microcirculatory resistance is 11, which is suggestive of a normal microvascular function. In conclusion, physiological examination without any relevant findings.

CORONARY VASOSPASM PROVOCATION TESTING

Different stimuli can be used to provoque epicardial or microvascular coronary spasm. Non-pharmacological stimuli like hyperventilation or coming into contact with cold are associated with an excessive number of false negatives for clinical use. Non-invasive coronary vasospasm assessment (based on changes on the ECG or the echocardiography through the IV administration of ergonovine) is associated with a risk of causing nitrate-resistant flow-limiting coronary spasm.4 For this reason, to this date, invasive studies based on the intracoronary administration of drugs are considered the single most sensitive and safe method. Actually, to this date, it is the method recommended by European guidelines and consensus docments.2,7 The direct administration of drugs allows us to use lower doses and establish a time correlation between the development of coronary spasm followed by symptom onset and changes on the ECG. Also, it facilitates immediate treatment through the direct administration of nitrates.4,10 The use of acetylcholine vs ergonovine is advised too since the former acts on a specific pathway (by stimulating cholinergic receptors only), and its safety profile is good because its half-life is shorter. Also, because it responds faster to nitrates in case of vasoconstriction.11 Also, acetylcholine allows us to assess the vascular endothelial response specifically, which is an additional advantage. The studies that compared the results of vasospasm provocation testing with acetylcholine vs ergonovine found similar sensitivity and high matching (94%) between the two. Therefore, in the presence of a negative acetylcholine test no additional studies with different drugs are advised.12

ACETYLCHOLINE

Acetylcholine is a neurotransmitter largely found in the nervous system (central, autonomous, and peripheral). It is used in the neuromuscular junction, in all synapses of the parasympathetic autonomous system, and in the first synapsis of the sympathetic nervous system. The muscarinic receptor of acetylcholine has 5 different subtypes; among them, subtype M2 is largely found in the myocardium where it reduces the heart rate and the cardiac conduction system; subtype M3 is found in the coronary arteries both in the endothelium and the smooth muscle. In the coronary arteries, the M3 receptor stimulates the contraction of the vascular smooth muscle (vasoconstriction). Also, it stimulates the endothelial production of nitric oxide that spreads into the smooth muscle reducing the concentration of calcium, and causing relaxation (vasodilation).13,14 Acetylcholine is rapidly hydrolyzed in both the neuromuscular junction and blood by the action of cholinesterases. When infused intracoronary in the doses described here, no systemic effects occur, and its cardiac effects only last a few minutes.

ACETYLCHOLINE-INDUCED ENDOTHELIAL DYSFUNCTION AND CORONARY VASOSPASM

Endothelial dysfunction is associated with the number of cardiovascular risk factors and is a well-known precursor of atherosclerosis.15 Also, the presence of endothelial dysfunction has been associated with the appearance of ischemia in the ischemia exercise test, heavier calcification and presence of necrotic and lipidic content in the vascular wall, and more cardiovascular adverse events in the long run.16-18 The prevalence of a vasoconstrictor response to the intracoronary infusion of acetylcholine, therefore, similar to an epicardial endothelial dysfunction, is variable depending on the characteristics of the patients being more common among males.19 In the studies conducted in patients with INOCA, the prevalence of endothelial dysfunction is somewhere between 45% and 75%.19,20

Although, to this point, no cut-off value has been universally accepted, it has been confirmed that moderate degrees of vasoconstriction (20% to 50%) with respect to the artery baseline diameter after the intracoronary infusion of acetylcholine have an important prognostic impact.18,21,22 Quantitative coronary angiography studies consider the variability of the technique when measuring changes in the mean luminal diameter of a segment with respect to the different doses of acetylcholine (usually the dose with the highest vasoconstriction with respect to the baseline one). Small imaging variations due to respiratory movements in every cine coronary arteriography, different limits of the study segment in the different measures taken, the analysis of diameters at different times of the cardiac cycle between the baseline image and maximum vasoconstriction, and the operator’s variability are the reasons why vasoconstriction can only be confirmed after variability is excluded from this measuring process. Several studies have established this variability (2 times the standard deviation of the percent difference) somewhere between 3% and 6%. Therfore, endothelial dysfunction is defined as a vasoconstrictor response that is greater than this variability.23,24

The pathophysiological factors of vasospastic angina, both in their macro and microvascular manifestations are less known and, also, probably multifactorial. Vasospastic angina has been associated with the presence of coronary plaques, vascular smooth muscle cell hyperreactivity, a high baseline vagal tone, hyperreactivity to sympathetic stimulation, and finally, to a significant degree of endothelial dysfunction.10 Vasospastic angina, both macro and microvascular, is more common among women.4 The traditional criteria to define vasospastic angina of macrovascular origin have been described by the Coronary Vasomotion Disorders International Study Group (COVADIS).5 In their document they describe the diagnostic criteria of this disease that go beyond the traditional definition of variant angina described by Prinzmetal et al.25. We should mention that, unlike the definition of endothelial function where baseline angiography is used as the reference, to define macrovascular spasm the COVADIS group recommends assessing the coronary spasp in the segment with the greatest constriction of all after the administration of acetylcholine and then compare it with the diameter of the same segment after the infusion of nitroglycerin.4 Also, this group recommends the use of drug provocation testing performed by intracoronary infusion of acetylcholine given its high sensitivity and specificity values (90% and 99%, respectively).26 Based on former studies and the traditional definition, the prevalence of epicardial coronary artery vasospasm, whether associated with microvascular spasm or not, occurs in 30% to 40% of the patients with INOCA.6,27

Fewer consensus documents have been published on the definition and diagnosis of microvascular spasm.28,29 Over the last few years, the appearance of thoracic pain and changes on the ECG suggestive of ischemia in response to acetylcholine and in the absence of macrovascular spasm have been accepted for the diagnosis of microvascular spasm (located in the arterioles). By this definition, 25% of the patients with INOCA meet the microvascular spasm criteria.27

TEST PERFORMED BY INTRACORONARY INFUSION OF ACETYLCHOLINE

Preparing the patient

The best way to prepare patients eligible for the coronary vasoreactivity test with acetylcholine is still under discussion. Historically, these procedures used to be performed in a dedicated procedure while avoiding and withdrawing all kinds of vasodilator drugs (like calcium channel blockers and nitrates) for, at least, 18 hours before the infusion of acetylcholine.27,30,31 However, after the publication of the randomized clinical trial CorMicA and the consensus document of the European Association of Percutaneous Coronary Interventions (EAPCI) on the study of patients with INOCA, conventional wisdom has changed.6,7 Currently, the use of intracoronary functional testing is recommended including the vasoreactivity test to acetylcholine within the same diagnostic procedure where the coronary angiography is performed.

This brings greater comfort to the patient, uses cath lab resources more efficiently, and alleviates the pressure of the hospital agenda. In any case, this procedure should be fully adapted to the needs and possibilities of every cath lab; in polymedicated patients with vasodilators or in inexperienced centers using this test the scheduled procedure should be used.

If radial access is used in patients eligible for a coronary vasoreactivity test the administration of calcium channel blockers to prevent radial spasm is ill-advised. In these cases, the administration of low doses of nitroglycerin through the introducer sheath (100 µg to 200 µg) can be considered. However, its effect will probably mostly be gone by the time acetylcholine is infused. Also, the coronary vasoreactivity test can be performed after studying microvascular function with a pressure guidewire (with the corresponding administration of intracoronary nitroglycerin before advancing the guidewire).6,7 In this case, a 2- to 3-min washout period should be observed before the infusion of acetylcholine.6,7

Finally, a specific informed consent should be obtained before running any vasoreactivity tests. Also, 12-lead ECG monitoring is required to assess the results. The use of radiotransparent wiring and electrodes is advised here to avoid interfering with the cine-fluoroscopy images obtained for each of the doses infused during coronary angiography. Figure 3 shows a schematic sample of how to prepare a patient before running a coronary vasoreactivity test with intracoronary acetylcholine.

Figure 3. Preparation of the patient before running any vasoreactivity tests. ACh, acetylcholine; Ca, calcium; Cine, cine coronary arteriography; IC, informed consent; NTG, nitroglycerin.

Regarding the use of beta-blockers, certain groups also recommend their cessation before running the test to avoid any possible vasoconstrictor effects. Until more scientific data become available, the opinion of this group is that beta-blockers do not affect the results of the test significantly and in no way give false negative results.

Preparing the acetylcholine

The acetylcholine available in Spain is a preparation for the intraocular injection of 20 mg of acetylcholine chloride (powder) for its dilution in a 2 mL vial of physiological saline solution. After the solution has been prepared, the drug is still unstable, which is why the best thing to do is to prepare it right before the test; if several consecutive tests are going to be performed, the same preparation can be used. Figure 4 summarizes the way to prepare the acetylcholine solutions suggested for the test. An important safety tip is to identify correctly every solution of acetylcholine we will be using; saline solution systems and color syringes for every dose can be useful.

Figure 4. Preparation of growing doses of acetylcholine. ACh, acetylcholine; DS, diameter stenosis; PSS, physiological saline solution.

Intracoronary infusion protocol

Over the last 30 years, different protocols on the administration and doses of intracoronary acetylcholine have been used. Table 2 shows the different protocols used in landmark studies.^{6,10,19,29,30,32-34} There are differences regarding the routes of administration (manual infusion through guide catheter or controlled selective infusion into an artery through an infusion pump and microcatheter), the number of doses infused (from 2 to 4), the amount of acetylcholine used (from 0.3 µg to 200 µg), and the infusion time (from 20 seconds to 3 minutes). Below we will be seeing the most widely accepted protocols based on the objective pursued (endothelial function assessment or vasospasm provocation) followed by a proposal according to the last consensus documents published to this date.

Table 2. Comparison of the different protocols of coronary vasoreactivity to acetylcholine

Endothelial function assessment

Growing doses of acetylcholine are used for endothelial function assessment. If this procedure is performed by selective drug infusion into 1 of the main coronary vessels with a microcatheter (usually the left anterior descending coronary artery), the concentrations used are 10 moL/L to 6 moL/L, 10 moL/L to 5 moL/L, and 10 moL/L to 4 moL/L. Considering the left anterior descending coronary artery flow (some 80 mL/min), it is estimated that the drug reaches concentrations that are 100 lower in coronary microcirculation. Using the microcatheter these dilutions are injected the into the proximal left anterior descending coronary artery or into the artery to be interrogated at a rate of 1 mL/min for 3 minutes or 2 mL/min for 2 minutes through an infusion pump.24,35 Infusion starts with the least concentrated dilution and, if no complications or overt vasospasm are reported the next infusion should start 2 to 3 minutes later. In practice, this method injects 0.5 µg, 5 µg, and 50 µg of acetylcholine in each of the doses. As already mentioned, in the presence of a non-dysfunctional vascular endothelium, the physiological response is the vasodilation of major epicardial vessels.

The procedure described, although widely used in clinical trials, is somehow complicated and expensive, which is why easier and more practical alternatives have been developed for macrovascular endothelial function assessment. The most important one that has already become the standard may be the one used in the ENCORE trials (Evaluation of nifedipine and cerivastatin on recovery of coronary endothelial function)36,37 consisting of the infusion of growing doses of 2 µg, 20 µg, and 100 µg directly into the left main coronary artery for 3 minutes each followed by the performance of an angiography after every dose. Also, it consists of the assessment of the arterial diameter compared to the one measured on the baseline angiography. If macrovascular endothelial function needs to be assessed, the recommendation is to follow this infusion pattern. As we will be seeing, this protocol has already been widely adopted in recent publications and, with minor changes, has become the go-to protocol for the diagnosis of coronary vasospasm although with a faster infusion of the doses.

Microvascular endothelial function can also be assessed using dedicated guidewires for the simultaneous measurement of coronary flow. In general, this procedure is performed using a Doppler guidewire (Combowire, Philips, The Netherlands),9 although the assessment can also be performed through thermodilution with thermistor-based temperature measuring guidewires (Pressurewire, Abbott, United States).38,39 Figure 2 shows 2 examples of this procedure.

Coronary spasm provocation testing

Although there are different protocols on doses and infusion times, the protocol recommended here for vasospasm provocation has been widely accepted by the most experienced groups. In addition, there are data available on its safety profile in many patients and is the protocol backed by the EAPCI in its recent consensus document.7

Three doses of 2 µg, 20 µg, and 100 µg are used in the left coronary artery, and 3 doses of 2 µg, 20 µg, and 50 µg in the right coronary artery. If the test is negative or inconclusive and the previous doses are well-tolerated, a 200 µg or a 80 µg dose can be used in the left or right coronary artery, respectively, if suspicion runs high.

Regarding the infusion time, a slow 20 second-bolus can be administered, although this is based on clinical tolerability. The highest doses, especially in the right coronary artery, often require infusions at a slower rate to avoid sinus arrest-induced bradycardia or atrioventricular block. It is important to carefully and slowly wash the guide catheter with a saline solution to avoid the sudden injection of the remaining drug into the catheter by the time the cine-fluoroscopy imaging is acquired. After every dose both the symptoms and the repolarization and angiographic changes should be assessed while paying special attention to the appearance of epicardial spasms or significant reductions of coronary flow velocity. At the end of the test, intracoronary nitroglycerin is infused (200 µg to 300 µg) and spasm is solved within a few seconds.

Safety and complications

Before indicating the test, the presence of factors that may be correlated with a risk of complications associated with the intracoronary infusion of acetylcholine should be discarded. The test should be carefully performed in patients with a past medical history of asthma or bronchospasm and serious disorders of automaticity and cardiac conduction.

Although safe in experienced hands, coronary vasoreactivity tests to acetylcholine are not stranger to potentially serious complications. These tests should always be performed paying extra care by trained personnel and ready to face the possible complications that may arise. In a metanalysis of different studies with over 6000 procedures, the rates of major (eg, ventricular arrhythmias, need for cardiopulmonary resuscitation or infarction), and minor complications (symptomatic bradycardia, transient atrioventricular block, appearance of ventricular arrhythmias or air embolism) were 1% and 6%, respectively.40 We should mention that no death was reported in this metanalysis.40 Table 3 shows the most common complications and their corresponding treatments.

Table 3. Complications associated with the intracoronary infusion of acetylcholine

During the infusion of acetylcholine, sinus bradycardia, sinus arrests or episodes of atrioventricular block are common. This is often associated with too fast infusions, especially in the right coronary artery. If these complications occur, infusion should stop for a few seconds and restarted at a slower velocity. Atrial fibrillation can sometimes occur, but it often solves spontaneously; the most persistent cases often solve after the administration of amiodarone or other antiarrhythmic drugs. If bradyarrhythmias make a comeback, the test should stop immediately or be performed with a transient pacemaker in very selected cases where the test is considered indispensable.

An unwanted effect of the test is flow-limiting vasospasm that is not well-tolerated. In general, the consequences depend on the time elapsed between the occurrence of the vasospasm and the infusion of intracoronary nitrates to reverse it. The ischemia originated can cause hypotension and ventricular fibrillation that should be treated with nitroglycerin and immediate defibrillation. To stop this from going unnoticed, the patient’s blood pressure should be checked halfway into the infusion of acetylcholine, especially after the highest doses have been infused and when injected into a dominant left coronary branch. Under no circumstance a growing dose of acetylcholine should be infused if a significant or flow-limiting spasm or any other important complication have been spotted after the infusion of lower doses. Also, we should remember that, at the end of the infusion, the guide catheter still contains 2 mL of acetylcholine dilution that should be slowly pushed with a saline solution to stop it from entering the bolus with the injection of contrast. Same as it happens with any other invasive coronary procedures, and especially in this test, preloaded nitroglycerin should be available and ready to be infused. In most of the cases its infusion causes vasodilation, and fast flow recovery without needing further doses. On the other hand, atropine is a cholinergic receptor antagonist that can be used as an antidote when necessary.

Some operators perform the vasoreactivity test using the pressure guidewire inside the coronary artery as a safety measure. This brings more stability to the catheter, provides better selective infusion of dilutions, and allows us to monitor distal pressure (that can decrease in the case of flow-limiting spasm). Also, it controls the velocity of manual infusion in case of a long infusion without a pump (temperature or velocity changes are indicative that infusion is happening too fast). However, we should remember that the passage of the guidewire itself can cause vasospasm. Actually, it can simulate pseudo-spasms in tortuous arteries due to curve rectification.

INTERPRETING THE CORONARY VASOSPASM PROVOCATION TESTING

General concepts

Interpreting this test rests on 3 basic pillars:

DS = 100 – [(MLD_Ach / RVD_NTG) × 100]

In practice, it is better to use the quantitative coronary angiography on the proximal segments of major arteries than in more distal segments where the reference diameter is often small and, according to the formula described above, could underestimate the DS.

Possible test results

Figure 1 shows the 4 results that can be obtained from a vasospasm provocation test performed by the infusion of intracoronary acetylcholine:

LEGAL ASPECTS PERTAINING TO THE USE OF INTRACORONARY ACETYLCHOLINE

The use of drugs in off-label indications different from those approved in their instructions for use and outside the clinical trial setting like intracoronary acetylcholine for diagnostic purposes requires the approval of local pharmaceutical committees. Because the Spanish Agency of Medicines and Medical Devices accepts the use of these drugs under very particular circumstances no general consensus has been achieved and local approvals are still required.

Also, in observance of Royal Decree 1015/2009 of 19 June,41 the use of a drug through a route of administration different from the one described in the drug labeling requires the provision of information as well as the patient’s written informed consent prior to its administration. For that reason, some centers also require the signing of a specific informed consent to be able to use intracoronary acetylcholine. These documents are available on the center intranet or the hospital pharmacy.

CONCLUSIONS

Figure 5 summarizes the key takeaways of this document. In conclusion, vasoreactivity testing with acetylcholine is an essential part of the assessment of patients with non-obstructive coronary artery disease and symptoms or ischemia. The result of this assessment allows us to target specific therapies and has proven effective in the routine clinical practice. Cath labs should be prepared to perform this kind of tests, and computers should be ready to use and interpret them.

Figure 5. Key takeaways of this document. ECG, electrocardiogram; INOCA, ischemia with non-obstructive coronary arteries; MINOCA, myocardial infarction with non-obstructive coronary arteries.

FUNDING

This document had no funding whatsoever.

AUTHORS’ CONTRIBUTIONS

E. Gutiérrez and J. Gómez-Lara equally contributed to the manuscript first draft, and to the figures, and tables. The remaining authors performed a thorough revision of the paper and made comments and changes to its content and form.

CONFLICTS OF INTEREST

None reported.

REFERENCES