Inhibition of Neointimal Proliferation With Low-Dose Irradiation From a β-Particle–Emitting Stent

Restenosis occurs in approximately 30% to 50% of cases after successful PTCA and is the major factor limiting the long-term effectiveness of this procedure.¹²³⁴⁵ There is clinical and experimental evidence to suggest that restenosis results from a complex interaction between early elastic recoil, arterial remodeling, and neointimal formation. Neointimal formation is an important contributor to this process, particularly after intracoronary stent placement, and results from the migration and rapid proliferation of a subset of predominantly medially derived SMCs. These cells proliferate as a response to catheter-induced mechanical injury and stimulation by a variety of growth factors and cytokines.⁶⁷⁸⁹ Experimental models of arterial injury have demonstrated that these cells begin proliferating within the first week after injury and return to baseline levels of proliferation within 1 to 2 months.^10111213 This early proliferation is followed by the elaboration of an extracellular matrix that further contributes to neointimal formation and ultimately results in narrowing of the arterial lumen.

Numerous trials in humans with a variety of pharmacological agents have failed to demonstrate a reduction in the incidence of restenosis.⁸ The use of radiation for the treatment of malignant and benign proliferative conditions has been well established. Low-dose external beam irradiation has been shown to inhibit keloid formation after surgery^14151617 and heterotopic ossification after total hip arthroplasty.¹⁸¹⁹²⁰ This safe and effective use of radiation therapy for benign proliferative conditions has led several investigators to test whether local irradiation might inhibit SMC proliferation after arterial injury. A single dose of external beam irradiation has been tested and was shown to be effective in two different experimental models of restenosis.²¹²² Several investigators have demonstrated that gamma irradiation by use of an endovascular after-loading technique can successfully inhibit neointimal proliferation in a porcine coronary injury model.²³²⁴²⁵ These investigators used a ¹⁹²Ir source to deliver an endovascular dose from 350 to 2500 cGy either immediately before, after, or within a few days of balloon injury. A dose-response relationship was demonstrated, with the greatest effect seen at higher doses (≥1400 cGy). Intraarterial beta irradiation at a dose of 1800 cGy has also been shown to be effective in a model of hypercholesterolemic rabbit restenosis.²⁶ Hehrlein et al²⁷²⁸ showed that a radioactive stent can inhibit neointimal proliferation after implantation in nondiseased rabbit iliac arteries. They used a stainless steel stent (Palmaz-Schatz, Johnson & Johnson Interventional Systems) made radioactive by proton bombardment in a cyclotron with a resulting array of gamma- and β-particle–emitting radionuclides ^55,56,57Co, ⁵²Mg, and ⁵⁵Fe. These radionuclides have a half-life between 17.5 hours and 2.7 years. At 4 and 12 weeks after stent implantation, there was almost complete inhibition of neointimal proliferation in this model.

The previously described techniques suffer somewhat from impracticality as well as questions of safety regarding the use of gamma irradiation and radionuclides of such long half-life. β-Particle irradiation with ³²P offers the advantages of no associated gamma radiation, a short half-life (14.3 days), and limited range of the β-particles in tissue (3 to 4 mm). This should essentially eliminate the risk to catheterization laboratory personnel and minimize the exposure of surrounding cardiac and pulmonary tissue to ionizing radiation. The objective of the current study was to investigate whether low-dose β-particle irradiation from a ³²P-impregnated stent could inhibit neointimal proliferation after implantation in porcine iliac arteries.

Methods

Stent Preparation

The stents used in the present study are a modification of the Strecker Stent (Boston Scientific Corp). The design consists of a single filament of polished titanium wire, 0.125 mm in diameter, woven into a flexible mesh tube. The stents were 20 mm in length with an expanded diameter of 5 mm. After stent fabrication, the nonradioactive element ³¹P was ion-implanted beneath the outer surface of the titanium wire (Spire Corp). Ion implantation was accomplished by placing the ³¹P into a special vacuum apparatus, then vaporizing, ionizing, and accelerating the ions with a high voltage so that the ³¹P atoms became buried beneath the surface of the titanium wire.

The stents were then made radioactive by placement in the core of an experimental nuclear reactor at the Massachusetts Institute of Technology. They were exposed for several hours to a flux of slow neutrons. Neutron irradiation caused approximately 1 of every 100 000 ³¹P atoms to be converted into ³²P, the pure β-particle emitter with a maximum energy of 1.709 MeV, an average energy of 0.695 MeV, and a half-life of 14.3 days. The reaction is given by the simple equation ³¹P+N=³²P, where N indicates neutron. This technique results in even distribution of ³²P within the stent wire, which ensures uniform delivery of β-particle irradiation into the vessel wall. After a period of approximately 10 days, the radiation from other short-lived isotopes created by neutron bombardment of the titanium base metal becomes undetectably low. Spectroscopic measurements indicated that no other long-lived radionucleotides were created. The radioactive stents were mounted onto balloon catheters (Boston Scientific Corp). The assembly was then packaged and gamma-ray sterilized in the conventional manner. The control stents were fabricated in a manner identical to the radioisotope stents except that they were not placed in a reactor, so that no atoms of ³¹P were converted to ³²P.

After each radioactive stent was placed over the deflated balloon of the balloon angioplasty catheter, a cylindrical acrylic resin radiation shield with an OD of 16 mm, an ID of 2.2 mm, and a wall thickness of 6.9 mm was placed over the stent. Radiation detectors placed outside each shield indicated no increase in radiation activity above background levels. The shield was designed so that its distal end could be advanced into the elastomer gland of a Tuohy-Borst “Y” adapter. With the shield in place, the stent could be advanced through the Tuohy-Borst adapter into the guiding catheter without exposing the operator to any radiation from the stent (Fig 1).

After removal from the reactor, each stent had an activity level of approximately 0.4 μCi. The activity level of each stent was measured with a gas-filled proportional-counting device. This detection system has been well described previously.²⁹ The stents were implanted 20 days after removal from the reactor when the radiation level had decreased to an average value of 0.14 μCi for all the stents. The radiation levels at implantation were determined by calculations that used the known half-life for ³²P (14.3 days) and the following equation: A_t=A₀×2^−t/τ, where A_t is the activity level at the time t (μCi), A₀ is the initial activity level (μCi), t is time (days), and τ is the half-life of the radioisotope (days). The use of a 5-mm-diameter stent with an activity level of 0.14 μCi resulted in a calculated total radiation dose at the surface of the stent wires (over the lifetime of the isotope) of approximately 300 cGy. The radiation dose delivered over the 28-day study period was equivalent to three fourths of the total radiation dose. The dose to tissue versus radial distance outward from the stent wires (both total and 28-day dose) is plotted in Fig 2. The calculations of dose distribution were performed by use of the dose-point-kernel method, which has been described previously.³⁰ This method uses several assumptions, including the assumption that the source is a uniform cylinder. A calculated dose was compared with a directly measured dose for an unexpanded, 2.1-μCi, 2.2-mm-diameter, 2-cm-long titanium-mesh stent (Fig 3). There was good correlation between the directly measured and calculated radiation doses.

Animal Preparation

The animal work was approved by the institutional scientific review committee and conformed with the position of the American Heart Association on animal research. Fourteen stents (seven control, seven ³²P-impregnated) were implanted in the right and/or left external iliac arteries of nine NIH miniature swine (weight, 20 to 30 kg) fed a normal chow diet. The animals were medicated with aspirin (650 mg) and extended-release nifedipine (30 mg) (Pfizer Laboratories Division) by mouth the evening before stent placement. Ketamine (20 mg/kg IM) and xylazine (4 mg/kg IM) were used for induction of anesthesia. General anesthesia was maintained by use of an intravenous fentanyl infusion (Elkins-Sinn) at a rate of 40 μg/kg per hour. A 9F sheath was placed retrograde in the right carotid artery. A bolus of heparin (150 U/kg) was administered intraarterially. The activated clotting time (Hemochron, Intl Technidyne) was measured to ensure a minimal value of 300 seconds for stent placement. Distal aortic and iliac angiography with 66% meglumine diatrizoate and 10% sodium diatrizoate (Hypaque-76, Sanofi Winthrop) was performed with a 9F multipurpose guiding catheter placed at the aortic bifurcation. The proximal iliac artery was engaged with the guiding catheter. A 0.014-in high-torque floppy guide wire (Advanced Cardiovascular Systems) was advanced to the femoral artery under fluoroscopy. The guiding catheter was used as a reference for stent sizing in an attempt to achieve a balloon (stent) to artery ratio just >1.0. Placement of the titanium-mesh stent (5.0 mm diameter, 20 mm length) was completed, after advancing the prosthesis over the guide wire to the desired site, with two balloon inflations at 8 atm for 30 seconds. This procedure was then repeated for stent placement in the opposite iliac artery. Angiography was completed after placement to confirm patency. The animals were allowed to recover and were returned to care facilities. All animals remained on a normal laboratory diet and received aspirin (81 mg) daily by mouth. The animals were returned to the research catheterization laboratory for angiography 28 days after implantation. After completion of follow-up angiography, the animals were euthanatized with a lethal dose of a barbiturate.

Pathological Evaluation

The iliac arteries were isolated and perfusion fixed with an ethanol-based fixative (Histochoice; Amresco) at 60 to 80 mm Hg for 30 minutes via a catheter positioned in the distal aorta. The distal aorta and iliac vessels were then carefully removed and placed in fixative. The stented arterial segments were sectioned transversely into 2-mm segments. Sections from the proximal, mid, and distal portion of the stent were embedded in paraffin and cut with a carbide knife at 4 to 5 μ. The sections were stained with hematoxylin-eosin and Movat pentachrome stains. The cross-sectional area of each section was measured with digital morphometry to determine area within the external elastic lamina, the IEL, the media, and the residual vessel lumen. The area within the IEL was considered the normal lumen area. The percent area stenosis was then defined as: [(IEL area−residual lumen area)/IEL area]×100. Neointimal area was determined by subtracting the area of the residual lumen from the area within the IEL. A mean value for each area within the stented arterial segment was derived from the proximal, mid, and distal measurements.

The vessel injury score was determined by the method used by Schwartz et al.³¹ In brief, the degree of injury at each wire site was assessed as follows: grade 0, IEL intact with media compressed; grade 1, IEL lacerated with media compressed; grade 2, IEL and media lacerated with the external elastic lamina intact; and grade 3, external elastic lamina lacerated. Neointimal thickness was measured at each wire site. The mean injury score for each arterial segment was calculated by dividing the sum of injury scores at each wire site by the total number of wires from the midstent sections. In addition to the mean injury score, the degree of stent oversizing (stent to artery ratio) was determined with quantitative angiography.

Quantification of SMC Proliferation

Stent wires were carefully removed from adjacent tissue sections of each arterial segment. The segments were dehydrated in graded series of alcohol and embedded in paraffin, and multiple sections were cut at 4 to 5 μ. The sections were stained with monoclonal antibodies to α-smooth muscle actin and PCNA. In brief, slides were deparaffinized and pretreated with antigen pretreatment solution (BioGenex Labs) for 5 minutes. The slides were then incubated with 10% normal horse serum followed by incubation with mouse anti-human PCNA antibody (Dako), 1:40 dilution, for 60 minutes. Biotinylated horse anti-mouse IgG was used as the secondary antibody, and the detection system used was the Vectastain Elite ABC kit (Vector Labs, Inc). Slides were counterstained lightly with hematoxylin. A section of tonsil served as a positive control for each series of immunohistochemical staining for PCNA. The total number of cells and the number of PCNA-positive cells per high-power field were counted from three randomly selected regions of each section. Only cells with distinct nuclear PCNA staining were considered positive. At least 50 cells per high-power field were manually counted for a minimum total of 150 cells per section to derive a PCNA index (percentage of PCNA positive cells=number of PCNA positive cells/total number of cells×100%). Actin staining followed the same procedure as above with mouse anti-human α-smooth muscle actin antibody (Sigma Chem Co) in a 1:5000 dilution.

Scanning Electron Microscopy

Segments were selected from four control and four β-particle–stented arteries, sectioned longitudinally, rinsed in three changes of sodium phosphate for 15 minutes each, then fixed in 1% osmium tetroxide for 1 hour and rinsed in distilled water. The specimens were then dehydrated in a series of graded ethanols (50% to 100%) for 15 minutes. After critical point drying, the specimens were mounted, placed in a vacuum coater, and coated with a 30- to 40-nm layer of gold for scanning electron microscopy.

Quantitative Angiography

The angiograms were analyzed with electronic digital calipers with use of the guiding catheter as a reference standard. The baseline and postimplant artery diameters (data expressed in millimeters) were measured at midstent from nonoverlapped and nonforeshortened views. The acute stent to artery ratio (midstent lumen diameter/baseline lumen diameter) was calculated from these data for each vessel to provide an estimate of stent oversizing as a measure of arterial injury.

Statistical Analysis

For each of the 14 arterial segments in the study, the stent to artery ratio, mean injury score, neointimal thickness, neointimal area, medial area, and percent area stenosis were determined. All data are expressed as mean±SD. Lesion morphology and PCNA index scores were compared for the control and radioactive stents by use of ANOVA with Scheffé’s F test for multiple comparisons. A value of P<.05 was considered significant. All statistics were calculated with Statview 4.0 (Abacus).

Results

Fourteen titanium stents were placed in the external iliac arteries of nine pigs (Table 1). There were 7 control stents and 7 treatment stents. All 14 stents were deployed without complication and each of the animals survived to 28 days. Follow-up angiography demonstrated 100% patency of the stents. The radioactive stents had an activity of 0.14 μCi at the time of implantation, with a calculated total dose over the lifetime of the isotope of approximately 300 cGy delivered at the surface of the stent wires. The dose delivered during the 28-day study period was equivalent to three fourths of the total lifetime dose.

Histopathology

As previously described, attempts were made to size the stents just >1:1 to the treatment vessel. There was no difference in the stent to artery ratio between the treatment and control stents (1.01±0.03 versus 1.02±0.08, P=.86). Review of the histological sections revealed no difference between the treatment and control groups with regard to vessel injury (grade 0 for all sections). The IEL was intact in each case.

There was a statistically significant reduction in neointimal area (1.76±0.37 mm² versus 2.81±1.22 mm², P=.05) and percent area stenosis (24.6±2.9% versus 36.0±10.7%, P=.02) in the β-particle emitting–stents compared with control stents. There was a trend toward an increase in lumen area in the treatment group (5.38±0.75 mm² versus 4.68±0.55 mm², P=.07) (Table 2). Neointimal thickness was assessed at each wire site, and the mean neointimal thickness was also significantly less within the β-particle–emitting stents compared with the control stents (0.26±0.04 mm versus 0.38±0.10 mm, P=.01). There was no difference in the size of the vessels as measured by the area within the external elastic lamina and IEL (Fig 4).

Review of the histological sections from both the treatment and control stents demonstrated neointimal proliferation with a predominant population of spindle-shaped cells staining positive for α-actin. These SMCs were haphazardly arranged near the stent wires with abundant extracellular matrix. At the lumen surface, the SMCs were more circumferentially organized and had a higher cellular density compared with cells adjacent to the stent wires (Fig 5). The IEL was intact in all sections, indicating an absence of deep vessel wall injury. There was mild compression of the media with minimal medial SMC necrosis. Scattered inflammatory cells were present adjacent to stent wires (Fig 5B and 5D). The degree of inflammatory reaction was minimal and similar at the site of β-particle and control stent wires. There was no excess fibrosis in the media, adventitia, or perivascular space of the vessels treated with a β-particle–emitting stent compared with control vessels.

The percentage of PCNA-positive cells was measured in a midstent section from each stent at 28 days. The mean cell proliferation index was similar for the treatment and control stents (2.15±0.91% versus 3.95±4.21%, P=.44). Cell density was also measured within the same sections and was similar between the two groups (362±42 versus 349±45 cells per high-power field, P=.46).

Scanning electron microscopy demonstrated complete endothelialization of all control and treatment stents except at the proximal and distal ends of the stents over protruding wire loops. The stents were covered with a population of immature, cobblestone-shaped endothelial cells (Fig 6).

Discussion

The rationale for use of irradiation to inhibit neointimal proliferation is based on the enhanced sensitivity of actively proliferating cells to the lethal effects of ionizing radiation. When the cell nucleus is exposed to ionizing radiation, DNA mutations and chromosome aberrations can occur. The ability of the cells to replicate is impaired, and ultimately cell death results. Radiation has been shown to inhibit benign “hyperplastic” responses such as keloid formation after surgery, fibroblast-mediated obliteration of filtering blebs after glaucoma surgery, and heterotopic ossification after total hip arthroplasty.^{1415161718192032} More recently, several investigators showed that external beam x-irradiation and endovascular gamma irradiation with an after-loading technique can inhibit neointimal proliferation in experimental models of restenosis.^2122232425 There was also one clinical study³³ that demonstrated a beneficial effect of endovascular irradiation on restenosis in the peripheral circulation after stenting. Only one experimental study³⁴ failed to show a beneficial effect of radiation on the restenotic process. Schwartz et al³⁴ found that x-irradiation actually modestly increased the neointimal proliferation seen after oversized stent placement in porcine coronary arteries. This negative result possibly was due to methodological differences, which included the use of relatively low radiation doses in the setting of more extensive vessel injury.

In the current study, we have shown that low-dose irradiation from a β-particle–emitting stent inhibits neointimal proliferation after stent placement in porcine iliac arteries. There was a 37% reduction in neointimal area and a 32% reduction in percent area stenosis for β-particle–emitting stents compared with control stents at 28 days in this model.

Rationale for Use of a β-Particle–Emitting Stent

The use of a β-particle–emitting stent to deliver low-dose irradiation to the vessel wall offers several advantages over previously described techniques for endovascular irradiation. Use of a radioactive stent couples the radiation therapy to a platform with a known benefit for the prevention of restenosis.³⁵³⁶ Mechanical factors such as ineffective dilatation, early elastic recoil, and chronic remodeling contribute to restenosis and would not likely be affected by the radiation therapy alone. Endovascular after-loading represents an alternative method for intravascular irradiation. This technique requires that a special radioactive catheter or wire be advanced across the lesion before PTCA, immediately after PTCA, or within a few days of the procedure to deliver the ¹⁹²IR source. This might result in significant ischemia if performed before PTCA and in dissection or vessel trauma if performed sometime after the dilatation. Because the delivery catheter may not always be centered within the vessel lumen, there will be nonuniformity of dose delivery into the vessel wall. This may range from 0.6× to 2.9× the prescribed dose.²⁴ In addition, these after-loading devices are costly, require handling of highly radioactive sources (0.2 to 10 Ci), and may require isolation of patients when gamma radiation is used.

β-particles have a limited effective range in tissue (≈95% of the particles are absorbed within 3 to 4 mm) and thus minimize the risk of radiation exposure to surrounding cardiac or pulmonary tissue. The radiation exposure to the interventionist implanting such a device and to technicians in the catheterization laboratory will be essentially zero. The total dose of radiation required from a ³²P source to inhibit neointimal proliferation appears to be small. Previous in vitro work has shown that β-particle irradiation from a ³²P-impregnated stent wire with an activity as low as 0.006 μCi/cm of wire results in complete inhibition of SMC proliferation within 6.0 mm of the wire in cell culture.²⁹ In the current study, stents with an activity of 0.14 μCi, resulting in a 28-day tissue dose of approximately 210 cGy at the stent surface, proved effective in inhibiting neointimal proliferation. On the basis of prior published in vivo studies, it is likely that higher doses of radiation from a β-particle–emitting stent could provide further inhibition of neointimal proliferation.²⁴

Endovascular Radiation to Inhibit Neointimal Formation

Various factors affect the radiation damage in a cell and hence the cell survival curve. The total dose, dose rate, linear energy transfer of the radiation, presence of chemical molecules, and stage of the cell cycle all affect the survival curve.³⁷ The total radiation dose delivered in this model was low relative to previous models and in a range that might not be expected to result in cell death. Also, the dose rate was very low. Why then was this low-dose β-particle irradiation effective in decreasing neointima formation? The radiation was delivered constantly over a period of time during which it is known that SMCs in the intima and media are undergoing rapid proliferation. Thus, although a lethal dose of radiation is not given, this constant, low-dose radiation therapy may impair cell proliferation (static effect) without producing cell death (lethal effect) in all SMCs. In addition, this low-dose radiation may have other, unknown effects. Fischell et al²⁹ demonstrated that β-particle irradiation from ³²P-impregnated wire resulted in a zone of inhibition of up to 10.6 mm when placed in a cell culture of proliferating SMCs. Since very few β-particles could reach a range of 8 to 10 mm in cell culture, it is conceivable that very low (sublethal) doses of β-particle irradiation may also inhibit SMC migration.

An important question regarding the use of radiation to inhibit neointimal hyperplasia is whether reendothelialization of the stents will also be inhibited. A delay in reendothelialization could, theoretically, adversely affect the risk of subsequent stent thrombosis. In cell culture, proliferating bovine endothelial cells are less sensitive to the effects of β-particle irradiation than are rat or human SMCs.²⁹ In the current study, endothelialization of the radioactive stents was evident by scanning electron microscopy 4 weeks after stent implantation. There was no noticeable difference with regard to endothelial cell growth within the treatment stents compared with control stents. Importantly, there was no stent thrombosis in the present study. Our data are consistent with preliminary work by Hehrlein et al,²⁷ who showed nearly complete inhibition of neointimal hyperplasia, endothelial regrowth at 4 weeks, and no increase in thrombotic complications after implantation of radioactive stents.

The use of radiation to inhibit neointimal proliferation does not appear to only postpone the restenotic process. Both Wiederman et al³⁸ and Waksman et al²⁴ demonstrated that the beneficial effect of endovascular irradiation was well maintained at 6-month follow-up in the porcine coronary model. C. Hehrlein, MD, (personal communication) has obtained follow-up at 1 year after use of radioactive stents in rabbit iliac arteries; the initial beneficial effect of radiation therapy was still present at 1 year. Because the degree of vessel injury and the type and method of radiation delivery differ in the current study, further study is required to assess the long-term effect of irradiation with this model. The use of fractionated dosing, with a total tissue dose less than that given as a single dose with the after-loading technique, would be expected to result in improved long-term safety.³⁷

In the current study, we used the PCNA index to quantify SMC proliferation at 28 days after stent implantation. There was no significant difference in PCNA index between the treatment and control groups. Although this is a single measurement at a single point in time, it demonstrates that there is no delayed (rebound) SMC proliferation at 28 days after implantation of a radioactive stent with a short-lived radioisotope.

Biocompatibility of Stents Modified to Deliver Radiation

Modification of a metallic stent surface may alter the mechanical performance of the device or its blood and tissue biocompatibility. Polymer-coated stents designed to reduce thrombogenicity and provide a local method for drug delivery have been limited by problems with chronic tissue compatibility. Severe neointimal formation has been seen after implantation of these devices in porcine coronary arteries.³⁹⁴⁰ The ion implantation technique to render the stent radioactive is highly desirable in that it does not require the use of a potentially biologically active substance, such as a biodegradable polymer, to deliver the endovascular radiation. The absence of a significant foreign body reaction or evidence of a greater degree of vascular injury by the β-particle–emitting stent indicates acceptable tissue compatibility. The present stent modification techniques may alter the surface charge of the metal, thus affecting the risk for thrombosis. Although stent thrombosis did not occur in this study, further analysis of stent surface characteristics is required.

Study Limitations

The present study is limited by the use of a single radiation dose and the small number of stents. Although the dose selected for this study was based on prior in vitro analysis and represented an effective dose to inhibit SMC proliferation without preventing endothelial cell growth, additional study is required to define an optimal in vivo dose. As previously suggested, higher doses could provide more effective suppression of neointimal growth.²⁴ These stents were implanted in normal arteries. In addition, stent placement was not associated with any significant vessel injury. This model differs from other models in which a proliferative response is invoked by more severe damage to the vessel wall. The dose and duration of endovascular radiation required to reduce neointimal formation in atherosclerotic vessels or vessels with more severe, deep injury may be different. The contribution of thrombus to the neointimal response in this model is also unknown. Larger studies with long-term follow-up will be required to confirm the persistence of this inhibitory effect on neointimal formation and to ensure the long-term safety of this technique. Further studies of lesion formation and SMC proliferation after coronary intervention are also necessary to enable selection of the appropriate radioisotope and dose for clinical studies.

In summary, the present study demonstrates that a locally applied continuous source of low-dose endovascular radiation via a β-particle–emitting stent inhibits neointimal formation after stent implantation in porcine arteries. This novel therapy may have a significant impact on stent restenosis and warrants further investigation.

Selected Abbreviations and Acronyms

The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the Department of the Army or the Department of Defense.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Footnotes