| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Division of Cardiovascular-Interventional Radiology, Stanford University Medical Center, Stanford Vascular Center, H-3647, 300 Pasteur Dr, Stanford, CA 94304-5105. Received November 30, 1998; revision requested January 21, 1999; revision received March 19; accepted May 6. Address reprint requests to M.D.D. (e-mail: mddake@leland.stanford.edu).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Phantoms with an aortic arch, true and false lumina with abdominal branch vessels, and a distal bifurcation were used to model a Stanford type B aortic dissection. The effects of anatomic factors (entry-tear size, branch-vessel flow distribution, fenestrations, distal reentry communication) and physiologic factors (peripheral resistance in the branch vessels, pump output and rate, vascular compliance) on true-lumen collapse were investigated. The morphology of the true lumen was observed. Branch pressures and flow rates were measured.
RESULTS: True-lumen collapse was induced and was exacerbated by an increase in the size of the entry tear, a decrease in the false-lumen outflow caused by occluding the false-lumen branch vessels, and an increase in the true-lumen outflow caused by lowering the peripheral resistance in true-lumen branch vessels. Two kinds of true-lumen collapse depended on pump output. With low pump output and low outflow resistance from the true lumen, the true lumen collapsed. With high pump output and low inflow resistance in the false lumen, the true lumen was compressed. Distal reentry communication between the true and false limbs was more effective than aortic fenestrations in preventing true-lumen collapse.
CONCLUSION: True-lumen collapse in this dissection model strongly depends on the difference in the ratios of inflow capacity to outflow capacity in the true and false lumina. Both anatomic and physiologic factors can affect true-lumen collapse.
Index terms: Aorta, dissection, 942.4124, 943.743, 981.743 • Aorta, flow dynamics • Aorta, stenosis or obstruction, 943.743, 981.743 • Aorta, US, 943.1298 • Phantoms, 942.412, 943.743, 981.743
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
In the past, the direct propagation of a dissection flap into an aortic branch with the resultant compromise or obstruction of the true lumen was considered to be the basic mechanism for ischemic complications in the peripheral vasculature. This understanding was based on observations of cross-clamped or decompressed aortas without flow and on findings at necropsy.
Recently, collapse or obliteration of the true lumen was proposed as another important mechanism for compromise of the aortic branch in aortic dissection (4,5). This is based on antemortem cross-sectional imaging studies, including those performed with intravascular ultrasonography (US), that facilitate an appreciation of the effects of flow on the anatomic relationships between the flap, aortic lumina, and branch vessels (4,5). In this setting, the plane of the dissection flap spares the branch vessel. Instead, the flap is positioned in a curtainlike fashion across the origin of the vessel, which causes dynamic obstruction of the branch artery (5). According to the report by Williams et al (6), dynamic obstruction due to true-lumen collapse was the cause of the infradiaphragmatic organ or limb ischemia in 20 of 24 patients. Among the 20 patients, 14 had ischemia in multiple organs that involved the mesenteric, renal, and lower-limb circulations.
Theoretically, if it is possible to identify the factors that induce true-lumen collapse, it may be possible to develop effective treatments. The pathophysiologic mechanism of true-lumen collapse in aortic dissection is, however, still unclear.
The purpose of this study was to investigate the causative factors for true-lumen collapse in aortic dissection by using phantoms with pulsatile flow. The implications for effective treatment methods used to relieve true-lumen collapse are discussed in part II.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Pulsatile Mock-Flow Loop
Figure 1 shows the components of the pulsatile-flow loop that was filled with water at 28°C. The system was capable of providing average volumetric flow rates of 6 L/min, with physiologic pressures and flow velocities in the models. A centrifugal pump (model 3E-12N; Little Giant Pump, Oklahoma City, Okla) continuously pumped water from the storage tank through the supply control valve to the source tank. The pulsatile pump was a ventricular assist device (Thoratec Laboratories, Berkeley, Calif) and had disk valves. It was pneumatically driven by a controller that was supplied with compressed air from a gas cylinder. The following parameters were controlled: pump rate, systolic pressure, diastolic pressure, and systolic duration. The pump rates used in the experiments were 30, 60, 90, and 120 beats per minute.
|
|
|
|
|
|
The exit branches and limbs of the model contained resistive, compliant, and measurement elements. The compliant and resistive elements were tuned to provide physiologic pressure and volumetric flow waveforms in the model. The 10-mm–ID PTFE arch and branch vessels were sequentially connected to rigid pressure taps, compliant sections, and polyvinyl chloride stopcock valves (Ryan Herco Products, Burbank, Calif), which served as resistive elements. The compliant elements consisted of 10-cm lengths of rubber tubing made from a 69 x 3-cm (27 x 1.25-inch) bicycle tire inner tube. The nominal ID of these compliant sections was 19 mm at a pressure of 70 mm Hg. In the parts distal to these elements, the flow traveled through 5.5 m of 6.4-mm–ID polyvinyl tubing before it reached the exit tank (Fig 1). The resistance of the exit branches was also controlled by changing the height of the tubing in the exit tank.
Each outflow tube from the bifurcation was sequentially connected to a reducing adapter, a compliant element, a polyvinyl chloride reducing tee, and an acetal wye. The compliant elements consisted of 10-cm lengths of rubber tubing made from a 66 x 5-cm (26 x 2-inch) bicycle tire inner tube. The nominal ID of the compliant section was 32 mm at a pressure of 70 mm Hg. The acetal wye created two lumina, which allowed an exit path for the flow and an access to each lumen for catheterization. The access side of the wye was connected to a tapered adapter. The end of the adapter with the Luer-Lok hub was connected to a Y-arm adapter, which allowed sealable access for catheters with diameters up to 11 F. A pressure tap and a distal stopcock valve were connected to the side of the wye with flow. Each limb was completed with 5.5 m of 6.4-mm–ID polyvinyl tubing.
Between the reducing tees for each limb, a reentry branch was added to allow distal communication between the true and false lumina. The reentry branch contained a resistive element in the form of a stopcock valve and a flow rate sensor. The stopcock valve was used with 69 cm of 6.4-mm–ID polyvinyl tubing; this was the small reentry branch. The large reentry branch was 10 cm of 12.7-mm–ID polyvinyl tubing without the stopcock valve and had a resistance much lower than that of the small reentry branch.
Rigid, Transparent Model
The rigid, transparent model shown in Figures 5 and 6 was made to allow observation of flap movement and to test the effects of the fenestrations. Its arch was composed of a compliant 90° bend similar to that used in the compliant model, a compliant bulb with a 10-mm–ID PTFE branch vessel, and a transparent acrylic 90° bend. The compliant bulb smoothed the pressure changes and provided capacitance that maintained aortic flow through diastole. The bulb was 10 cm long, with a nominal 38-mm ID at 70 mm Hg. It was constructed from an inner tube from a motor scooter tire. An arch branch vessel that was glued to the compliant bulb allowed access for catheters with diameters up to 22 F.
|
|
As in the compliant model, 10-mm–ID PTFE graft segments were glued to the body to create branch vessels that were contiguous with only the directly contributory aortic lumen. In this model, the two branches from the false lumen were diametrically opposed to the two distal branches from the true lumen. At the distal part, the true lumen extended 11.4 cm through the bifurcation into a 20-mm–ID acrylic tube, where it was sealed around the inside circumference. The flow from the false lumen exited through the exit tube of the false limb with a 25-mm ID.
Other features were incorporated into this model. The rigid nature of the model allowed us to install pressure taps to monitor the pressure in the lumina proximal and distal to the exit branches. Fenestration-branch loops were also incorporated into the model. These loops allowed fluid exchange between the true and false lumina in much the same way natural and artificial fenestrations do in human dissections. These loops consisted of two 3.8-cm–long pieces of 10-mm–ID PTFE that were glued to the aortic tube at diametrically opposed positions. Glued to each of these was a 23-cm–long piece of 7.9-mm–ID tubing (Tygon R-3603; Norton Performance Plastics, Akron, Ohio). The loops were completed by joining these tubes to a double-tapered tubing connector with a 6.3-mm ID. Dura clamps (Ryan Herco Products) were used to open and close these fenestration branches. The exit branches and limbs and the pulsatile-flow loop were the same as those used in the compliant model.
Instrumentation and Measurement Systems
A model 486 66-MHz personal computer (Adisys, Sunnyvale, Calif) with a 12-bit Lab PC+ data acquisition card (National Instruments, Austin, Tex) was used to sample up to eight pressure and flow signals. Data were continuously sampled at 50 hertz per channel for up to 2 minutes. The software for data acquisition was written in the LabView programming language (National Instruments).
Pressure was measured by using two instruments. Initially, a single Mikro-tip catheter (model SPC-350MR; Millar Instruments, Houston, Tex) was used. As it became important to monitor pressure in several locations simultaneously, pressure transducers from pressure monitoring kits (Sorenson Transpac IV; Abbott Laboratories, North Chicago, Ill) were used. This pressure measurement system had a resolution of 0.5 mm Hg. The accuracy of the Sorenson transducers was ±2 mm Hg due to our calibration technique and due to the fact that the calibration sometimes drifted ±1 mm Hg in the 10-minute measurement periods between calibrations. This accuracy was sufficient for a comparison of the pressure waveforms in different operating conditions. In some cases, we were interested in the pressure differences in the lumina that would move the intimal flap. In these cases, all of the transducers were calibrated precisely, and data were collected immediately thereafter. The uncertainty in the relative pressure measurements was only a fraction of 1 mm Hg.
The volumetric flow rate was measured in two ways. To measure the average volume flow rates through the exit branches, a bucket and a stopwatch were used. The measurements proved to be highly repeatable to within 5%. As an alternative, a 6.4-mm In-Line Flowprobe (model 6N; Transonic Systems, Ithaca, NY) and a T101 ultrasonic blood flow meter (Transonic Systems) were used to measure the time-dependent volumetric flow rate through the branches, the direction of flow through the reentry branch, and the relationship of the flow rate to the pressure waveforms.
Temperature affected the viscosity of the water, the compliance of the PTFE graft material, and the rubber components of the model. The water temperature in the source tank was monitored with a type K thermocouple probe and a digital thermometer (2166A; Fluke, Everett, Wash). The temperature was maintained at 28°C ± 1; this amount of drift did not affect the dynamics we observed.
The compliant model described previously was opaque, so intravascular US with a 10-F, 10-MHz intracardiac catheter (Cardiovascular Imaging Systems, Sunnyvale, Calif) was used to monitor flap movement primarily in the branch-vessel region of the model. Because the PTFE was exceedingly echoic and because the rubber tube was relatively anechoic, the intravascular US catheter was placed in only the false lumen for imaging.
Investigation of Causative Factors
With these two models, we observed the morphology of the true lumen and the presence or absence of branch-flow compromise for a large number of pump settings and model configurations. True-lumen collapse was defined as the obliteration of the true lumen with associated compromise of flow in the true-lumen branch. True-lumen obliteration was assessed by means of visual inspection (rigid, transparent model) or intravascular US (compliant model). Compromise of the flow in the true-lumen branch was evaluated as an evident decrease in the branch outflow at the exit tank and as a decreased intraluminal pressure and flow rate, as measured at the branch.
The investigated variables were pump parameters (rate and output), patterns of flow distribution in the branch vessels, peripheral resistance of the branch vessels, compliance of the phantom, entry-tear size, and communications between true and false lumina through the distal reentry and fenestration branches. The effects of the flow distribution in the branch vessels, peripheral resistance of the branch vessels, compliance, and distal reentry communication were tested with the compliant model; the effects entry-tear size were tested in both models; and the effects of the fenestrations were tested with the rigid, transparent model.
Five branch-vessel flow distributions similar to those seen in clinical situations were investigated (Fig 7). Lower peripheral resistance in abdominal branch vessels that originated from the true lumen was simulated by lowering the height of the tubing in the exit tank by 30 cm; this was equivalent to a 22 mm Hg decrease in pressure. The effects of the compliance in the model were tested with the compliant model by strapping rigid 32-mm–ID acrylic tubing around the aorta and compliant elements in the true and false limbs. The effects entry-tear size were observed by gradually increasing its transverse length from 10 to 30 mm. The ability of the natural fenestrations to prevent and to contribute to true-lumen collapse was investigated by opening the fenestration-branch loops to allow fluid exchange between the true and false lumina.
|
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
|
The data in Table 1 show that the total volumetric flow rate increased with decreasing pump settings for diastolic pressure. Lower settings caused the pump sac to fill, which increased the stroke volume. Thus, high-output conditions existed for low settings for diastolic pressure; low-output conditions existed for high settings for diastolic pressure. In many cases, the distinction between high output and low output was important because the stroke volume had a marked effect on flap movement.
Table 2 summarizes the observations of the compliant model with a 30-mm entry tear and with the same peripheral resistance in the true- and false-lumen branches. Note that true-lumen collapse developed only in the cases with the most restricted outflow from the false lumen (cases 3–5). In case 1, with equal outflow from both lumina, there was no true-lumen collapse. Case 5, with the false lumen with a blind sac, exhibited the most severe true-lumen collapse.
|
However, in case 5, the false-lumen pressure was usually higher than the true-lumen pressure. In this case, an increase in the pump output preferentially increased the false-lumen pressure and reversed the direction of the reentry flow. As a result, true-lumen collapse occurred with most conditions in case 5. Interestingly, there were small differences in the status of the true lumen that corresponded to the pump rate and output. For pump rates of 60 and 90 beats per minute, there was a window between the high- and low-output states in which the true-lumen collapse did not occur.
The changes in the true lumen given in Table 3 indicate that lowering the peripheral resistance in the abdominal branches from the true lumen promoted true-lumen collapse and caused low-output collapse in every pattern of flow distribution from the branch vessels. In the cases with true-lumen collapse, the mean true-lumen pressure was usually lower than the mean false-lumen pressure.
|
|
|
|
|
Patent fenestration-branch loops between the true and false lumina in the body of the aorta were not effective in preventing true-lumen collapse (Table 8). Opening the first and second fenestration-branch loops, proximal to the branch vessels, slightly aggravated the true-lumen collapse. Opening the third, fourth, or all fenestration loops slightly alleviated the true-lumen collapse. The third and fourth fenestrations loops were distal to the branch vessels. Compared with the fenestration-branch loops, the distal reentry communication between the true and false limbs was more effective in preventing true-lumen collapse.
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
On the basis of the observations, it is hypothesized that the difference in the ratios of the inflow capacity to the outflow capacity between the true and false lumina generated a transmural pressure gradient across the dissection flap and that this pressure gradient moved the flap. According to this hypothesis, true-lumen collapse was the result of the higher ratio of inflow capacity to outflow capacity for the false lumen. Therefore, to relieve true-lumen collapse, it is conceptually desirable to decrease the false-lumen inflow capacity and to increase the false-lumen outflow capacity. The placement of a stent-graft over the entry tear and the creation of outflow branches are two possible treatment options.
Pump output was an important factor in true-lumen collapse in both models. Pump output seemed to have different effects according to the ratio of inflow capacity to outflow capacity in the true and false lumina. For conditions in which the false-lumen inflow capacity was restricted compared with its outflow capacity, an increase in the pump output preferentially increased the true-lumen pressure, and true-lumen collapse never developed. However, for conditions in which the false-lumen outflow capacity was restricted compared with its inflow capacity, the effect of the increase in the pump output was complicated.
For example, in case 5 with a pump rate of 90 beats per minute (Table 2), there was a true-lumen collapse in the low-output states. The increase in the pump output initially prevented true-lumen collapse. However, a further increase in pump output caused a preferential increase in the false-lumen pressure and induced a severe true-lumen collapse. The presence of a safe window between the zones with true-lumen collapse suggested there were two mechanisms: low-output collapse and high-output compression. This safe-window phenomenon was more prominent in the compliant model. Low-output collapse was exaggerated when the peripheral resistance in the true-lumen branches was lower than that of the false-lumen branches.
Low peripheral resistance in the true-lumen abdominal branches can contribute to low-output collapse. Williams et al (7) presented data from a patient in whom sudden restoration of true-lumen outflow resulted in a true-lumen collapse and in a profound systolic-pressure deficit. That clinical experience strongly supports the observations that an increase in the true-lumen outflow accelerates true-lumen collapse. Williams et al (7) also suggest that if false-lumen pressure exceeds true-lumen pressure, an element of true-lumen compression may be added to the intrinsic true-lumen collapse. The present experiment proved the presence of true-lumen compression in the high-output states and confirmed a higher false-lumen pressure in these situations.
When we compared the observations of steady-state flow (Table 4) with those of pulsatile flow (Tables 2, 3), pulsatile flow seemed to have prevented true-lumen collapse in many conditions (eg, the medium to high output results for case 1 (Table 3). It is believed that these safe windows for pump operation are related to wave propagation within the models and that they are functions of model compliance and geometry. Although they are not exactly the same, there are similar waves in the pulsatile flow within the human vascular system. In light of this, the observations made in this experiment suggest that the regulation of heart rate and systemic pressure may provide a feasible treatment for the prevention of true-lumen collapse in patients with dissection.
In both steady-state and pulsatile flow, true-lumen compression started from the lower-limb vessels and propagated proximally to the abdominal vessels (Fig 9). Therefore, in our models, isolated lower-limb ischemia was indicative of a form of true-lumen collapse that was milder than that associated with mesenteric and/or renal ischemia. However, this conclusion is based on a relatively specific set of anatomic parameters that does not fully represent the complexity typically observed in clinical cases.
An evaluation of the effects of the communication channels between the true and false lumina (ie, fenestrations and reentry tears) showed that a distal communication was better at preventing true-lumen collapse. The beneficial effect of the fenestration-branch loops in the aorta was unexpectedly limited or absent (Table 8). In fact, the fenestration-branch loops proximal to the abdominal branch vessels slightly aggravated the true-lumen collapse; this suggests that they may have acted as additional entry tears.
The distal reentry communication between the true and false limbs showed variable effectiveness in preventing true-lumen collapse, depending on its size. Although the distal reentry communication in our models simulated a distal reentry tear, strictly speaking, it was close to a femorofemoral bypass. Therefore, radiologic fenestrations just above the iliac bifurcation or femorofemoral bypass may mitigate the effects of true-lumen collapse.
In conclusion, true-lumen collapse in aortic dissection was demonstrated in phantoms with pulsatile flow; it strongly depended on the difference between the ratios of the inflow capacity to the outflow capacity for the true and false lumina. True-lumen collapse occurred in low- and high-output states. Low peripheral resistance in the true-lumen branches contributed to the low-output collapse of the true lumen. High-output compression of the true lumen developed in the conditions in which the false-lumen outflow was restricted compared with the inflow. Fenestrations in the dissection flap proximal to the aortic bifurcation had little effect on true-lumen collapse, whereas distal reentry communications showed a considerable effect in preventing true-lumen collapse.Practical application: Both models simulated the gross morphology of the human aorta. However, there were many differences in the compliance of the vascular wall, the hemodynamic characteristics of the pulsatile flow, the viscosity of the working fluid, and the detailed anatomy of the dissection. Therefore, one should be cautious when applying the results of this experiment to human aortic dissection. The purpose of this study was to give insight into the pathophysiology and treatment of true-lumen collapse in aortic dissection and to motivate further experimental and clinical studies.
|
Footnotes |
|---|
See also the article by Chung et al (pp 99–106 ) in this issue.
Abbreviations: ID = inner diameter PTFE = polytetrafluoroethylene
Author contributions: Guarantor of integrity of entire study, M.D.D.; study concepts, J.W.C., C.E., T.S., M.D.D.; study design, J.W.C., C.E., T.S., N.K., M.D.D.; definition of intellectual content, J.W.C., C.E., M.D.D.; literature research, J.W.C.; experimental studies, J.W.C., C.E., T.S., N.K., T.V.; data acquisition, J.W.C., C.E., T.S., N.K., T.V.; data analysis, J.W.C., C.E., M.D.D.; manuscript preparation, J.W.C., C.E., M.D.D., C.P.S., S.M.S.; manuscript editing and review, C.E., M.D.D., C.P.S., S.M.S.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Related Article
This article has been cited by other articles:
|
|
 |
|
  T. T. Tsai, A. Evangelista, C. A. Nienaber, T. Myrmel, G. Meinhardt, J. V. Cooper, D. E. Smith, T. Suzuki, R. Fattori, A. Llovet, et al. Partial Thrombosis of the False Lumen in Patients with Acute Type B Aortic Dissection N. Engl. J. Med., July 26, 2007; 357(4): 349 - 359. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. C. Moon, J. Pablo Morales, and R. K. Greenberg Complicated Acute Type B Dissection and Endovascular Repair: Indications and Pitfalls Perspectives in Vascular Surgery and Endovascular Therapy, June 1, 2007; 19(2): 146 - 159. [Abstract] [PDF]
|
|
|
|
|
|
S. L. Kahn and M. D. Dake Stent Graft Management of Stable, Uncomplicated Type B Aortic Dissection Perspectives in Vascular Surgery and Endovascular Therapy, June 1, 2007; 19(2): 162 - 169. [Abstract] [PDF]
|
|
|
|
 |
|
 D. Bockler, H. Schumacher, M. Ganten, H. von Tengg-Kobligk, M. Schwarzbach, C. Fink, H.-U. Kauczor, H. Bardenheuer, and J.-R. Allenberg Complications after endovascular repair of acute symptomatic and chronic expanding Stanford type B aortic dissections. J. Thorac. Cardiovasc. Surg., August 1, 2006; 132(2): 361 - 368.e4. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Garcia, J. Ferreiros, M. Santamaria, A. Bustos, J. L. Abades, and N. Santamaria MR Angiographic Evaluation of Complications in Surgically Treated Type A Aortic Dissection. RadioGraphics, July 1, 2006; 26(4): 981 - 992. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Uchida, H. Ishihara, H. Shibamura, Y. Kyo, and M. Ozawa Midterm results of extensive primary repair of the thoracic aorta by means of total arch replacement with open stent graft placement for an acute type A aortic dissection J. Thorac. Cardiovasc. Surg., April 1, 2006; 131(4): 862 - 867. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Saijo, H. Akimoto, Y. Saiki, K. Tabayashi, T. Horinouchi, T. Kobayashi, and S.-i. Nitta Proximal (entry) tear of dissecting aortic aneurysm visualized by three-dimensional echocardiography J. Thorac. Cardiovasc. Surg., December 1, 2002; 124(6): 1245 - 1246. [Full Text]
|
|
|
|
|
|
N. Kato, T. Shimono, T. Hirano, T. Suzuki, M. Ishida, H. Sakuma, I. Yada, and K. Takeda Midterm results of stent-graft repair of acute and chronic aortic dissection with descending tear: The complication-specific approach J. Thorac. Cardiovasc. Surg., August 1, 2002; 124(2): 306 - 312. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Sailer, P. Peloschek, T. Rand, M. Grabenwoger, S. Thurnher, and J. Lammer Endovascular Treatment of Aortic Type B Dissection and Penetrating Ulcer Using Commercially Available Stent-Grafts Am. J. Roentgenol., December 1, 2001; 177(6): 1365 - 1369. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
