Eurohyp.org

Cooling for Acute Ischemic Brain Damage
(COOL AID)
A feasibility trial of endovascular cooling
M.A. De Georgia, MD; D.W. Krieger, MD; A. Abou-Chebl, MD; T.G. Devlin, MD, PhD; M. Jauss, MD; S.M. Davis, MD, FRACP; W.J. Koroshetz, MD; G. Rordorf, MD; and S. Warach, MD, PhD Abstract—Objective: To report results of a randomized pilot clinical feasibility trial of endovascular cooling in patients
with ischemic stroke. Methods: Forty patients with ischemic stroke presenting within 12 hours of symptom onset were
enrolled in the study. An endovascular cooling device was inserted into the inferior vena cava of those randomized to
hypothermia. A core body temperature of 33 °C was targeted for 24 hours. All patients underwent clinical assessment and
MRI initially, at days 3 to 5 and days 30 to 37. Results: Eighteen patients were randomized to hypothermia and 22 to
receive standard medical management. Thirteen patients reached target temperature in a mean of 77 Ϯ 44 minutes. Most
tolerated hypothermia well. Clinical outcomes were similar in both groups. Mean diffusion-weighted imaging (DWI) lesion
growth in the hypothermia group (n ϭ 12) was 90.0 Ϯ 83.5% compared with 108.4 Ϯ 142.4% in the control group (n ϭ 11)
(NS). Mean DWI lesion growth in patients who cooled well (n ϭ 8) was 72.9 Ϯ 95.2% (NS). Conclusions: Induced moderate
hypothermia is feasible using an endovascular cooling device in most patients with acute ischemic stroke. Further studies
are needed to determine if hypothermia improves outcome.
Reperfusion reduces infarct size during acute ische- Clinic Foundation, Erlanger Medical Center, Massachusetts Gen- mic stroke. Several challenges exist, however, in- eral Hospital), Germany (Justus Liebing University, Giessen), andAustralia (Royal Melbourne Hospital). The study protocol was cluding the short therapeutic time window,1 the risk approved by each institutional review board, and informed con- of reperfusion injury,2 and hemorrhage.3 In animal sent was obtained from all patients or designated surrogates. Ran- models of middle cerebral artery (MCA) occlusion, domization was done by opening sealed randomization envelopes.
hypothermia consistently decreases infarct volume.4-6 Patients were older than 18 years, had anterior circulation terri-tory ischemic stroke, had an NIH Stroke Scale (NIHSS) score of Studies of postischemic hypothermia have also Ն8 and Յ25, and presented within 12 hours of symptom onset.
shown a benefit in the immediate postischemic peri- Exclusion criteria included recent sepsis; contraindication to MRI; od.4,7 Recently, hypothermia has been shown in two platelet count of Ͻ75,000/mm3; known coagulopathy; significant randomized trials to improve outcomes after cardiac ventricular cardiac dysrhythmias or QTc interval of Ͼ450 milli-seconds; pregnancy; intolerance to buspirone or meperidine; treat- arrest.8,9 Despite the promise of hypothermia, several ment with monoamine oxidase inhibitors; life expectancy of Ͻ6 problems have limited its use in acute stroke. Sur- months; baseline modified Rankin Scale (mRS) score of Ͼ2; rap- face cooling is slow and cumbersome and requires idly improving symptoms; intracerebral hemorrhage, mass, or an- general anesthesia to counteract shivering. Endovas- eurysm by brain CT scan; hypersensitivity to hypothermia; heightof Ͻ1.5 m; inferior vena cava filter in place; and renal insuffi- cular cooling using a counter-current heat exchange ciency. Patients returned for a clinical assessment at 30 ϩ 7 days catheter is a new method to induce hypothermia and including mRS and NIHSS scores (the NIHSS score was corrected precisely control temperature. Recently, the combina- for side of stroke by adding 5 to the NIHSS score in patients with tion of meperidine, buspirone, and surface warming right hemispheric strokes).12 Adverse events were predefined,monitored through 30 ϩ 7 days, and reviewed by an independent was found to reduce shivering during endovascular Clinical Events Committee (Harvard Clinical Research Institute, cooling.10,11 We tested whether endovascular cooling Boston, MA). A Data Safety Monitoring Board was also convened combined with meperidine, buspirone, and surface warming could achieve hypothermia rapidly in pa- ing the Reprieve Endovascular Temperature Management System (Radiant Medical, Redwood City, CA), which consists of a propri-etary triple-lobed, helically wound, heat-exchange balloon cathe- Materials and methods.
ter that is placed in the inferior vena cava via the femoral vein patients with acute ischemic stroke were randomized to receive and a microprocessor-driven controller. The catheter is inserted endovascular cooling or standard medical management. Patients through a 10F femoral introducer sheath until the distal tip is at were enrolled at five centers in the United States (The Cleveland the level of the diaphragm and is connected to a cassette with a From The Cleveland Clinic Foundation (Drs. De Georgia, Krieger, and Abou-Chebl), OH, Erlanger Medical Center (Dr. Devlin), Chattanooga, TN, JustusLiebing University (Dr. Jauss), Giessen, Germany, Royal Melbourne Hospital (Dr. Davis), Australia, Massachusetts General Hospital (Drs. Koroshetz andRordorf), Boston, and NIH (Dr. Warach), Bethesda, MD.
Dr. De Georgia has received in excess of $10,000 in grant support and Dr. Davis has received honoraria from Radiant Medical, Inc.
Received September 19, 2003. Accepted in final form March 10, 2004.
Address correspondence and reprint requests to Dr. M.A. De Georgia, Department of Neurology, Cleveland Clinic Foundation, 9500 Euclid Ave., Cleveland,OH 44195; e-mail: degeorm@ccf.org Copyright 2004 by AAN Enterprises, Inc.
Table 1 Reasons for exclusion from study*
Transferred to outside hospital/no bed available Figure 1. The Reprieve Endovascular Temperature Man- * Nine patients were excluded for miscellaneous reasons.
NIHSS ϭ NIH Stroke Scale; mRS ϭ modified Rankin Scale.
groups: 72% men in the hypothermia group and 71% pump that circulates cold or warm saline (figure 1). The targetcore body temperature was 33 °C (monitored with an esophageal women in the control group (p ϭ 0.02). In the hypothermia probe). Cooling was maintained for 24 hours. Controlled rewarm- group, 61% had right hemispheric infarcts vs 43% in the ing was performed using the same catheter by setting the target control group (NS). The corrected NIHSS score (mean Ϯ temperature to 36.5 °C. Rewarming was performed at a rate of 0.2 SD) in the hypothermia group was 18.2 Ϯ 4.4 and in the °C/h. Shivering was suppressed using a forced-air warming blan-ket (BairHugger; Augustine Medical, Eden Prairie, MN), oral bu- control group 16.7 Ϯ 5.5 (NS).12 Cardioembolism was the spirone (60 mg), and IV meperidine (50- to 75-mg loading dose most common mechanism of stroke (72% in the hypother- followed by an IV infusion at 25 to 35 mg/h).
mia group and 67% in control group). Large artery disease All patients received standard medical treatment, including was found in one hypothermia patient (5.6%) and dissec- thrombolytic therapy if indicated. Hematologic variables and elec- tion in two (11%). Large artery disease was found in four trolytes were measured every 8 hours in the hypothermia groupand daily in the control group. Esophageal temperatures were control patients (19%). The cause was undetermined in two monitored every 15 minutes during active cooling or warming and hypothermia patients (11%) and three control patients every 30 minutes during temperature maintenance in the hypo- (14%). Rates of thrombolytic therapy were similar between thermia group. Bladder or rectal temperatures were monitored the two groups. The mean admission oral temperatures After randomization but before cooling, all were also similar (36.3 Ϯ 0.6 °C in the hypothermia group and 36.3 Ϯ 0.5 °C in the control group).
perfusion-weighted imaging (PWI), and MR angiography (MRA) performed. Those who received thrombolytic therapy underwent a to presentation was slightly shorter in the hypothermia brain CT scan after 24 hours. DWI, PWI, and MRA were repeated group (3 hours 27 minutes Ϯ 2 hours 46 minutes) com- on days 3 to 5. Follow-up T2-weighted and fluid-attenuated inver-sion recovery (FLAIR) MR images were obtained to measure final pared with the control group (4 hours 16 minutes Ϯ 1 hour infarct volumes at day 30. In patients who died before day 30, the 58 minutes) (NS). The mean time to IV thrombolytic ther- DWI from days 3 to 5 was used to measure infarct volumes. DWI apy was 130 Ϯ 43 minutes in the hypothermia group and lesion growth was calculated by comparing the baseline DWI with 150 Ϯ 23 minutes in the control group. The mean time the day 3 to 5 DWI. All imaging analysis was performed by Per-ceptive Informatics (Bethesda, MD), which was blinded to treat- from stroke onset to initiation of cooling was 8 hours 59 minutes Ϯ 2 hours 52 minutes. The main reason for delaywas the time needed to obtain baseline MRI. Catheter Results.
placement was well tolerated in all patients. Cooling was tients with acute ischemic stroke were screened, and 40 also generally well tolerated, with no case being discontin- were randomized. Table 1 lists the reasons for exclusion.
ued because of patient discomfort. Significant shivering Eighteen patients were assigned to the hypothermia group occurred in two patients (Patients H3 and H13). Of the 18 and 22 to the control group. One patient assigned to the patients in the hypothermia group, 13 cooled with a mean control group had a normal DWI after randomization and time to target temperature of 77 Ϯ 44 minutes, the most was diagnosed with a TIA; this patient was excluded from rapid reaching target in 15 minutes. These 13 patients achieved a core temperature of at least 35.0 °C in 37 Ϯ 26 Baseline patient characteristics are summarized in ta- minutes, with 9 of them achieving at least 35.0 °C in 22 Ϯ ble 2. The distribution of gender differed between the two 11 minutes. There was no overshoot. Cooling was less effi- July (2 of 2) 2004
Table 2 Baseline patient characteristics
Figure 2. Examples of temperature curves of patients un- dergoing endovascular cooling. Patient H3: Time to targettemperature ϭ 20 minutes; Patient H4: time to target tem- was considered by the Data Safety Monitoring Board to be related to the device or to hypothermia. In the hypother- mia group, three of the five patients died from complete MCA territory strokes, brain swelling, and herniation. One Values in parentheses are percentages.
patient died from a large stroke with hemorrhagic trans-formation, and another died from multiorgan system fail- ure. This man (Patient H2), with a history of ischemiccardiomyopathy, hypertension, and chronic renal failure, NIHSS ϭ NIH Stroke Scale; rt-PA ϭ recombinant tissue plas-minogen activator.
developed cardiogenic shock 24 hours after rewarming. ADNR/comfort care order was written following a long in-tensive care unit stay, and he died on day 30. The mean cient in five patients; target temperature was not reached NIHSS score of hypothermia patients who died was 19.2 Ϯ in four and reached after several hours in one. Reasons 1.9. In the control group, all four deaths occurred in pa- included incorrect catheter placement (the catheter was tients with massive strokes. One woman (Patient C14) not advanced far enough into the inferior vena cava) in with ischemic cardiomyopathy developed cardiogenic shock three patients (one of whom also shivered), kinking of the on day 4 and died of a cardiac arrest on day 24. The mean catheter outside the body in one patient, and shivering in NIHSS score of control patients who died was 20.3 Ϯ 4.5.
one patient. Examples of temperature curves of patients undergoing endovascular cooling are shown (figure 2).
group developed symptomatic hemorrhagic transforma- Endovascular cooling was generally well toler- tion. Neither received thrombolysis, and one had a hemor- ated. There were no significant differences in hemody- rhage present on the baseline MRI but not seen on CT namic variables or laboratory values between the two scan. This patient (H7) was very hypertensive, had recana- groups. There was a slight increase in mean systolic blood lized at the time of angiography, and was loaded with pressure and heart rate during induction of cooling and a clopidogrel and begun on aspirin. Both patients had rela- slight decrease in respiratory rates. During the mainte- tively late initiation of cooling (13 hours 29 minutes after nance phase, systolic blood pressures came back to base- stroke onset in one and 10 hours after stroke onset in line and heart and respiratory rates dropped slightly below another). One patient in the control group had a second baseline (figure 3). There were 16 complications that oc- ischemic stroke 4 days after the index stroke following curred in 11 patients in the hypothermia group and 14 complications in 10 patients in the control group (table 3).
In general, patients with complications had higher NIHSS (H2), as mentioned, developed cardiogenic shock 24 hours scores (median 17.0 Ϯ 4.6) than those without complica- after rewarming and wide complex tachycardia later dur- tions (median 10.5 Ϯ 4.7) (p ϭ 0.009).
ing the hospitalization. One patient developed new-onset There were five deaths in the hypothermia atrial fibrillation thought to be the source of his stroke. In group and four in the control group. None of the deaths the control group, one patient developed cardiogenic shock July (2 of 2) 2004
Figure 3. Hemodynamic data. SBP ϭ systolic blood pressure; RR ϭ respiration rate; HR ϭ heart rate. on day 4, another developed new-onset angina pectoris, nonintubated patients in the control group (excluding 3 and a third developed wide complex tachycardia.
who were DNR) was 11 Ϯ 5.1. In the hypothermia group, the mean total meperidine dose was similar for intubated patients in the hypothermia group and three in the control (12.5 Ϯ 4.0 mg/kg) and nonintubated (12.2 Ϯ 7.1 mg/kg) group. Two of the five hypothermia patients developed patients, but the mean duration of meperidine infusion pneumonia, and three developed pulmonary edema. One was longer in patients who were intubated (33 hours 51 developed radiographic signs of pulmonary edema, though minutes Ϯ 7 hours 9 minutes) than in those who were clinically was asymptomatic. A second patient developed not intubated (26 hours 40 minutes Ϯ 10 hours 14 min- pulmonary edema and a pleural effusion following a retro- peritoneal hemorrhage. A third patient (H10) developed pulmonary edema on day 1 that resolved with diuresis. On in four patients in the hypothermia group and two in the day 13, he developed bilateral infiltrates vs pulmonary control group. One patient in the hypothermia group de- edema and was treated with antibiotics and diuresis. On veloped a retroperitoneal hemorrhage on the side of the day 27, he was intubated for an adult respiratory distress femoral vein introducer sheath on day 6 while fully antico- syndrome-like pattern and to facilitate bronchoscopy. In agulated with heparin. She had just received intra-arterial the control group, two patients developed pneumonia and thrombolysis for an MCA occlusion and was on aspirin and clopidogrel at the time of the initial femoral venous punc- Eight patients in the hypothermia group required intu- ture. She was transfused 2 units of packed red blood cells bation during their hospitalization (p ϭ 0.002). Four of the and remained hemodynamically stable. Three patients had eight were patients with massive strokes who were intu- lower extremity deep vein thromboses (DVTs) detected by bated for airway protection. Only one was intubated dur- routine Doppler examinations. This surveillance Doppler ing the maintenance phase of hypothermia, and the other monitoring was performed only in treatment patients and three were intubated 48 to 72 hours after stroke onset. The not control subjects. In one patient, the DVT was detected other four patients were intubated for cardiopulmonary 9 days after enrollment and 3 days after a venogram reasons at a mean of 9.5 Ϯ 10 days after stroke onset.
showed no thrombosis. The DVT occurred at the previous These reasons included cardiogenic shock (Patient H2 with site of access for both the device and the venogram. In a an ejection fraction of 20%), pulmonary edema/pleural ef- second patient, the DVT was detected 13 days after enroll- fusion (Patient H4 following retroperitoneal hemorrhage), ment and 1 day after a central line was placed at the same worsening aspiration pneumonia present on admission site (and was still in place when the routine Doppler ultra- (Patient H5), and pulmonary edema (Patient H10). The sound was performed). In a third patient, the DVT was median age (71.0 Ϯ 11.1 years) and NIHSS score (18.5 Ϯ detected 2 days after enrollment; the catheter and intro- 3.5) of those requiring intubation were greater than the ducer had just been removed, and manual pressure and a median age (54.5 Ϯ 11.2 years; p ϭ 0.05) and NIHSS score sandbag had been used at the site within hours of the (14.5 Ϯ 4.0; p ϭ 0.08) of those who were not intubated.
examination. Routine surveillance Doppler monitoring was Only one patient in the control group was intubated: a performed only in treatment patients. In the control group, woman with an NIHSS score of 23 who developed cardio- one patient had a pulmonary embolism and another had a genic shock. The median NIHSS score of the remaining 18 July (2 of 2) 2004
Table 3 Complications
HT ϭ hemorrhagic transformation; DVT ϭ deep vein thrombosis; UTI ϭ urinary tract infection.
In our previous surface cooling pilot trial, we were one patient developed a urinary tract infection and an- able to reach target temperature (32 °C) on average other had a positive blood culture that probably was a 3.5 hours after induction with considerable effort.
contaminant (one of two bottles). In the control group, four We experienced no overshoot of target temperature patients developed urinary tract infections. Finally, in the with endovascular cooling in contrast to surface cool- control group, one patient developed hematuria on heparin ing.14 In a few patients, the target temperature of 33 that resolved once the heparin was stopped.
°C was not achieved, mainly because of mechanical NIHSS and mRS scores at days 30 to 37 were reasons. Only one patient with a massive stroke was similar between the two groups. Comparative MRI data were intubated during active cooling. All other 17 patients available in only 23 of the 40 patients. Comparing the initial (94.4%) remained awake and responsive during image with the day 3 to 5 image, the DWI lesion growth inthe hypothermia group was 90.0 Ϯ 83.5% (n ϭ 12) and in the control group 108.4 Ϯ 142.4% (n ϭ 11) (NS). The mean DWI The safety of mild to moderate hypothermia has lesion growth in patients who cooled well (n ϭ 8) was 72.9 Ϯ been demonstrated in several clinical settings, in- 95.2%. The mean DWI lesion growth in patients who cooled cluding traumatic brain injury,13 cardiac arrest,8,9 and poorly (n ϭ 4) was 124.4 Ϯ 45.5% (NS).
myocardial infarction.11 Although deep hypothermiamay cause ventricular arrhythmia, coagulopathy, or Discussion.
immunosuppression, mild to moderate hypothermia vascular cooling was feasible in patients with moder- has not been associated with these complications. All ate to severe anterior circulation territory ischemic patients remained hemodynamically stable throughout stroke. The heat-exchange catheter could be inserted hypothermia. The frequency of complications was sim- quickly and easily, and we achieved a core body tem- ilar in the hypothermia and control groups. In both perature of 33 °C in 14 of 18 patients (78%). Reduc- groups, patients with complications were older and had tion in core temperature was rapid in 13 patients more severe strokes. Endovascular cooling is invasive, with a mean time to target temperature of 77 Ϯ 44 however, and complications can occur. The one retro- minutes. In recent surface cooling trials, Ͼ4 hours peritoneal hemorrhage was directly related to the ini- was required to cool patients with severe brain inju- tial femoral puncture when inserting the venous ry13 and 8 hours to cool patients after cardiac arrest.9 sheath. This occurred in a patient who had received July (2 of 2) 2004
intra-arterial thrombolysis, followed by aspirin, clopi- dation, OH); Thomas G. Devlin, MD, PhD (Erlanger Medical Center, Chat-tanooga, TN); Marek Jauss, MD (Justus Liebing University, Giessen, dogrel, and heparin. Three patients had DVTs in the Germany); Stephen M. Davis, MD, FRACP (Royal Melbourne Hospital, treatment group that may have been related to the Australia); Walter J. Koroshetz, MD, Guy Rordorf, MD (Massachusetts endovascular cooling device, although there were sev- General Hospital, Boston); Steven Warach, MD, PhD (NIH, Bethesda, MD);Gregory W. Albers, MD (Stanford Medical Center, Palo Alto, CA); Werner eral other confounding factors in these patients. Other Hacke, MD (University of Heidelbeg Kopf Klinik, Heidelberg, Germany); studies using the Reprieve11,15 and other endovascular Stephan A. Mayer, MD (Neurological Institute, New York); Wade S. Smith,MD, PhD )UCSF Medical Center, San Francisco, CA).
cooling devices16 have not demonstrated an increasedincidence of DVT. Four patients in the hypothermiagroup with massive strokes were intubated for airway Acknowledgment
protection (only one during active cooling), and another The authors thank the patients and their families as well as the four were intubated later for cardiopulmonary reasons.
residents and nursing staff that made this study possible. Theyalso thank Patricia McMahon, RN, BSN, study coordinator, for Intubated patients were older and had more severe assistance with data collection and analysis.
strokes than nonintubated patients. They also hadlonger durations of meperidine infusions, however, andit is possible that this contributed to the need for intu- References
bation. As anticipated, given the small size, there were 1. Marler JR, Tilley BC, Lu M, et al. Early stroke treatment associated no differences in clinical outcomes. Infarct volume with better outcome: the NINDS rt-PA Stroke Study. Neurology 2000;55:1649 –1655.
growth was less in the hypothermia group, but this 2. Aronowski J, Strong R, Grotta JC. Reperfusion injury: demonstration of brain damage produced by reperfusion after transient focal ischemia inrats. J Cereb Blood Flow Metab 1997;17:1048 –1056.
We sought to evaluate the feasibility of induced 3. Hamann GF, del Zoppo GJ, von Kummer R. Hemorrhagic transforma- moderate hypothermia in patients with acute ische- tion of cerebral infarction—possible mechanisms. Thromb Haemost1999;82(suppl 1):92–94.
mic stroke using an endovascular cooling device cou- 4. Maier CM, Sun GH, Kunis D, Yenari MA, Steinberg GK. Delayed in- pled with meperidine, buspirone, and surface duction and long-term effects of mild hypothermia in a focal model oftransient cerebral ischemia: neurological outcome and infarct size.
warming to suppress shivering. The results suggest that this approach is feasible; moderate hypothermia 5. Chen H, Chopp M, Zhang ZG, Garcia JH. The effect of hypothermia on can be induced in patients with ischemic stroke transient middle cerebral artery occlusion in the rat. J Cereb BloodFlow Metab 1992;12:621– 628.
quickly and effectively and maintained for 24 hours 6. Yanamoto H, Nagata I, Niitsu Y, et al. Prolonged mild hypothermia in most patients. Endovascular cooling is generally therapy protects the brain against permanent focal ischemia. Stroke2001;32:232–239.
safe and well tolerated in most patients. Some el- 7. Yanamoto H, Hong SC, Soleau S, et al. Mild postischemic hypothermia derly patients with severe strokes and those with limits cerebral injury following transient focal ischemia in rat neocor-tex. Brain Res 1996;718:207–211.
comorbidities treated for Ͼ24 hours developed pul- 8. Bernard SA, Gray TW, Buist MD, et al. Treatment of comatose survi- monary compromise, however. The medications used vors of out-of-hospital cardiac arrest with induced hypothermia. N Engl to suppress shivering induce sedation and need to be 9. Hypothermia After Cardiac Arrest Study G. Mild therapeutic hypother- titrated carefully. This may be especially important mia to improve the neurologic outcome after cardiac arrest. N Engl in the elderly stroke patient with compromised air- 10. Mokhtarani M, Mahgoub AN, Morioka N, et al. Buspirone and meperi- way control or with propensity toward sedative- dine synergistically reduce the shivering threshold. Anesth Analg 2001; induced hypoventilation and atelectasis. We believe 11. Dixon SR, Whitbourn RJ, Dae MW, et al. Induction of mild systemic that the best target population for induced hypother- hypothermia with endovascular cooling during primary percutaneous mia includes younger patients and those with moder- coronary intervention for acute myocardial infarction. J Am Coll Car-diol 2002;40:1928 –1934.
ate strokes. Although the optimal duration is not 12. Woo D, Broderick JP, Kothari RU, et al. Does the National Institutes of clear, treatment for Յ24 hours may reduce the risks.
Health Stroke Scale favor left hemisphere strokes? Stroke 1999;30: Larger clinical trials are needed to determine 13. Clifton GL, Miller ER, Choi SC, et al. Lack of effect of induction of whether endovascular cooling improves clinical out- hypothermia after acute brain injury. N Engl J Med 2001;344:556 –563.
comes in patients with acute ischemic stroke.
14. Krieger DW, De Georgia MA, Abou-Chebl A, et al. Cooling for Acute Ischemic Brain Damage (COOL AID): an open pilot study of inducedhypothermia for acute ischemic stroke. Stroke 2001;32:1847–1854.
Appendix
15. O’Neill WW. A prospective randomized trial of mild systemic hypother- mia during PCI treatment of ST elevation MI—the COOL MI Trial.
Transcatheter Cardiovascular Therapeutics Meeting, September 15 to ical Center, Boston, MA); Jennifer J. Gassman, PhD (Cleveland Clinic 16. Georgiadis D, Schwartz S, Kollmar R, Schwab S. Endovascular cooling Michael A. De Georgia, MD, Derk W. Krieger, for moderate hypothermia in patients with acute stroke. Stroke 2001; MD, Alex Abou-Chebl, MD, Anthony J. Furlan, MD (Cleveland Clinic Foun- July (2 of 2) 2004

Source: http://eurohyp.org/COOL.pdf