Big Strokes in Small Persons FREE

In 1910, American physician James Herrick, MD, published the first case in the Western medical literature on sickle cell disease (SCD) in a Grenadian dental student living in Chicago, Illinois. A blood smear from this student showed “peculiar elongated cells.”¹ Pauling et al drew attention to SCD as a “molecular disease.”² Much about SCD is understood. There are approximately 80 000 patients with the most severe form of the disease in the United States, and more than 15 000 articles have been cited in PubMed on the topic since 1949.

Well before the major biological features of this genetic disorder were elucidated, stroke had been noted as an associated complication of SCD.³ Early brain pathologic studies described many abnormalities of the brain and its vasculature, and in a few cases, large cerebral infarctions were reported.⁴ Why would a blood disorder, characterized by severe hemolytic anemia and known to engender pathophysiologic features on a microscopic scale,⁵ lead to stroke at all, much less large brain infarctions?

Most of the neurologic reviews published before 1976 emphasized the presumed basis for circulatory problems, including stroke arising from occlusion of small vessels by the sickled erythrocytes that give the disease its name.⁶ These were known to be formed, at first reversibly, then finally into a state of permanent distortion, when cells containing sickle hemoglobin and insufficient other hemoglobins (such as fetal hemoglobin, which retards sickling) become deoxygenated.⁷ The cellular distortion is caused by sickle cell hemoglobin forming intracellular polymers. Although sickle cell hemoglobin carries oxygen in a similar manner as hemoglobin A in solution (although the oxygen dissociation curve is shifted to the right compared with normal), its presence in cells in the deoxygenated state leads to ischemia by reducing the ability of the erythrocyte to traverse the microcirculation. These cells are also subject to premature destruction, which leads to severe anemia and also releases toxic elements (eg, plasma free hemoglobin) into the circulation.

The fundamental paradigm for the most common clinical feature, the painful “crisis,” is now understood as a sequence of events that begins with attachment of large immature red blood cells (RBCs) (reticulocytes) to the postcapillary venule and propagation backward until there is delayed transit time of RBCs, causing further deoxygenation and sickling. These events take place in vessels the size of RBCs that measure 5 to 10 μm.⁵ Recent data have also implicated RBC-leukocyte and leukocyte-endothelial interactions, which may precede the RBC-endothelial attachment and do not take place in mice deficient in E selectin and P selectin.⁸

Although the microcirculatory events have been studied ex vivo, as yet there is no comparable large-vessel animal model that would allow a better understanding of the events leading to occlusion of arteries such as the middle cerebral or internal carotid. How the genetic disorder of sickle cell anemia creates large arterial vasculopathy and leads to stroke remains somewhat of a mystery, but the schema proposed by Platt⁹ is a good approximation of what is believed to happen in large brain arteries before stroke (Figure 1).

The publication in 1972 by Stockman et al¹⁰ of cerebral angiography in 7 patients with SCD and neurologic complications drew the first real attention to the fact that large intracerebral arteries were sometimes involved in a stenotic and obliterative process¹⁰ preferentially located just beyond the origin of the ophthalmic artery and involving part or all of the anterior circle of Willis.

The example of advanced cerebrovascular disease shown in Figure 2 is based on studies performed on an 8-year-old boy who had a stroke and was studied using angiography and transcranial Doppler ultrasonography (TCD) early in our project at the Medical College of Georgia. It shows several important and typical features. On the side with stroke is occlusion of the internal carotid artery just distal to the origin of the ophthalmic artery. The TCD shows a very low and blunted waveform. On the opposite side, as yet without stroke but at risk, is severe stenosis that, while still allowing flow, has the characteristic high-velocity TCD “signature” that has become the basis for presymptomatic screening and institution of preventive therapy before brain infarction. In this case, the velocity is approximately 230 cm/s, well above what would later become the threshold for prophylactic treatment of 200 cm/s.

After the study by Stockman et al,¹⁰ there followed further confirmation that large-artery disease of the internal carotid and middle cerebral arteries is typically found in many but not all children with brain infarction and in approximately half of those with intracranial hemorrhage.¹¹ Angiography in the other cases of hemorrhage showed only diffusely dilated arteries or aneurysms that caused subarachnoid hemorrhage.

We began working on this problem in 1985 at the Medical College of Georgia when a young girl with SCD had a massive infarction after internal carotid artery occlusion. Russell et al¹² had just published an article with extensive angiographic documentation of what Stockman et al and others had reported and also suggested that regular transfusion might, if not reverse the process, at least prevent arterial worsening. Their study, while crucial, was not a clinical trial, and in fact no trial for secondary stroke prevention in SCD has yet been published. However, regular blood transfusions soon became an accepted therapy to prevent recurrent stroke in SCD based on comparison with historically high recurrence rates in children with SCD.

In 1982, Aaslid et al¹³ published an article on TCD for the detection of subarachnoid hemorrhage–related vasospasm. In 1985, the Medical College of Georgia SCD Cohort Study began using TCD in patients along with magnetic resonance imaging (MRI). Although both of the new techniques have much to offer in the study of stroke and patients at risk for stroke, it was to be TCD that proved more useful for primary stroke prevention in these patients, largely because it was so much easier to use.

In children with homozygous SCD, a yearly first stroke risk of approximately 0.5% had been established by the Cooperative Study of Sickle Cell Disease based on a large population study¹⁴ that predated TCD and MRI use (Table 1). Although some risk factors for stroke were identified (slightly lower total hemoglobin level, transient ischemic attack, and acute chest syndrome), a predictive model sufficient to plan a clinical prevention trial was not available. This stroke rate, which is very high for children, was still too low to execute a practical randomized controlled trial given the limited number of patients available. Even assuming that transfusion was completely effective in preventing the first stroke, without a way to select the children at greatest risk, one would have to transfuse approximately 200 children each year to prevent 1 stroke. Primary stroke prevention would depend on finding an acceptable way to identify individuals at highest risk to make the number needed to treat more attractive and manageable.

It seemed reasonable to focus on the large intracranial arteries rather than the brain itself for 2 reasons. First, selecting patients on the basis of infarction that had already taken place meant that we would be “following” the process of brain injury rather than preventing it firsthand. Second, the collective angiographic information, derived from children who had already had a stroke, had shown that major arteries opposite the hemispheres that had evident brain infarction often showed early or sometimes extensive arterial narrowing that had not yet become symptomatic. This suggested that the arterial process leading to large strokes might develop at a slow enough rate to create a “window of intervention” before the brain is affected in children developing vasculopathy and on this basis the high-risk state for stroke. Using angiography to screen large numbers of children who were asymptomatic did not seem practical. El Gammal et al²² and Pavlakis et al²³ at Columbia began using MRI, but MR angiography (MRA) was not yet available.

The TCD seemed to be a possible method, but it was not clear whether it would work in this application. Use of Doppler ultrasonography had become an important way to identify extracranial carotid stenosis based on a derivation of the Bernoulli principle, which is that in the area where the vessel is narrowed, the velocity of blood flowing in that vessel is increased. However, cervical Doppler ultrasonography would at best reflect flow changes secondary to intracranial stenosis because the internal carotid artery in SCD is rarely affected by vasculopathy; TCD seemed a more direct diagnostic approach. The TCD was also attractive because it was harmless, painless, and generally well tolerated, even by children, and did not require sedation. It was also portable and low cost compared with other methods.

However, in SCD there were specific challenges in using TCD. Would these young children with so many other medical issues tolerate the examination? What velocity standards should be used? It was known that cerebral blood flow and flow velocity as estimated by TCD are increased in the anemic condition owing to the lowered oxygen-carrying capacity of the blood. It was soon evident from work by Brass et al²⁴ at Columbia and Adams et al^25- 26 that the evolving velocity standards of TCD for stenosis in adults would not apply in SCD owing to the impact of young age and anemia. Would there be sufficient separation between children with elevated TCD velocities due to anemia with no vessel disease and average risk and those with anemia who were developing stenosis and increased risk?

To determine whether TCD would predict stroke, a series of studies were performed between 1985 and 1992. The first study was to establish the expected middle cerebral artery velocity in children of the target age, with²⁵ and without²⁶ SCD and without stroke. Transcranial Doppler ultrasonography was also performed on children with stroke at the time of cerebral angiography, providing an opportunity to correlate velocity to stenosis visualized using that method.¹⁵ In parallel, a large cohort of children with SCD but no stroke history and not undergoing regular transfusion were recruited, studied using TCD, and followed up prospectively for stroke outcome.^16,27 The portability and ease of use of TCD was instrumental because many of these TCD studies were performed in outreach clinics across Georgia during regular clinic visits. This feature of TCD greatly enhanced enrollment and follow-up.

The early use of TCD in adults had shown that the normal velocity (time-averaged mean of maximum blood flow as opposed to systolic or diastolic) was approximately 60 cm/s (Table 2). It was determined that the expected middle cerebral artery velocity in healthy children was approximately 90 cm/s, and in children with SCD without overt stroke it was approximately 130 cm/s, elevated on the basis of young age and severe anemia. When the angiogram showed severe stenosis, the velocity was always greater than 190 cm/s and often much higher except in cases of severe stenosis, in which very low velocities were recorded (<70 cm/s).¹⁵ These data provided key cross-sectional information, but what was needed to make primary stroke prevention possible was to demonstrate that TCD predicted risk of future stroke.

That elevated TCD velocity was associated with a high risk of future stroke was demonstrated in the Medical College of Georgia SCD Cohort study. The long-term outcome of 315 children aged 3 to 18 years at the time of first screening and free of stroke when enrolled was evident from the stroke-free survival curves. Arbitrary velocity cutoff points (there are no evident “inflection points” in the risk relationship, but cutoff points were needed to define risk strata) were used to stratify risk: less than 170 cm/s represented “normal” or average risk, 170 to 199 cm/s was called “conditional” and was associated with moderate risk, and 200 cm/s or greater was called “abnormal” or high risk. A risk of stroke during the next 36 months of 13% per year was observed.^16,27 Although the time from abnormal TCD velocity to stroke was variable, children with the highest velocity seemed to have more proximate risk (5 children with TCD velocity >240 cm/s had a stroke within 9 months of TCD).

Armed with this information, a proposal was made to the National Heart, Lung, and Blood Institute to fund the first randomized controlled trial of stroke prevention for any indication in SCD and the first primary stroke prevention trial in children with any disease. This study, STOP, was conducted between 1995 and 2000 at 14 sites in the United States and Canada.²⁸ A unique feature was the selection of participants based on TCD. Almost 2000 children aged 2 to 16 years with no history of stroke were screened using TCD, and those with 2 TCDs showing a middle cerebral artery or internal carotid artery velocity of 200 cm/s or higher were approached for randomization. Rigorous operator training and standardization of TCD protocol and equipment were used to ensure uniform testing. Screening for high-risk cases in STOP showed the same prevalence of TCD findings at all 14 sites (and was similar to that found in the Medical College of Georgia cohort): approximately 10% had 1 TCD, and 85% of those with 1 abnormal TCD were confirmed to have abnormal results on a second study. The final high-risk rate was approximately 9%; approximately 15% fell into a “conditional” range of 170 to 199 cm/s, and 70% had velocities less than 170 cm/s, falling into a lower-risk group. The others had inadequate studies that had to be repeated. The TCDs were recorded as digital files, which facilitated file transfer between the research sites and the data and reading centers, study masking, and blinded reading. Neither MRI nor MRA was performed until after randomization, and imaging for “silent stroke” was evaluated as a secondary end point.^28- 29

The trial was halted 16 months early when 11 of the 67 children randomized to receive standard care (episodic transfusions only depending on symptoms such as pain) had a stroke compared with a single child undergoing long-term transfusion.²¹ The observed stroke rate without transfusion was 10% per year across 2 years, confirming the reliability and validity of TCD in a multicenter application and the dramatic (>90%) reduction in stroke with regular transfusion (Figure 3). One surprising finding was that MRAs of cerebral vessels in children with velocities greater than 200 cm/s but less than 250 cm/s did not usually show severe stenosis,³⁰ suggesting that TCD indicates risk at an earlier (and probably more reversible) stage than MRA.

The STOP results were announced in 1997. Despite the clear results and recommendations based on STOP from the National Heart, Lung, and Blood Institute³¹ and the American Stroke Association (also endorsed by the American Academy of Neurology)²⁰ for widespread screening and prophylactic transfusion in high-risk cases, the long-term problems with transfusion, and the fact that a stopping point for transfusion was not clear, some physicians remained hesitant to adopt this strategy.

In STOP, many patients undergoing transfusion saw their TCD velocities revert to apparent low risk (<170 cm/s; approximately 53%) or intermediate risk (170-199 cm/s; approximately 17%), especially if the velocity at treatment initiation was in the low abnormal range (200-230 cm/s) and the MRA findings were relatively normal (see Figure 4 and Figure 5 for an example of change with regular transfusion). Approximately 30% of children who were compliant with regular transfusion still had abnormal TCD velocities even after years of transfusion. A second study was then planned to determine whether continued treatment was still needed in the subset that normalized with prolonged transfusion. The STOP II³² was designed to determine whether transfusion could be safely withdrawn after a defined treatment period with acceptable stroke risk and rates of return to transfusion. This trial was planned to enroll 100 children, all with originally high-risk TCD velocities that reverted to less than 170 cm/s after at least 30 months of transfusion and who had an MRA that did not show moderate or severe stenosis or occlusion. The design called for half the children to be randomized to continued transfusion and half to have transfusion stopped. All the participants received close follow-up and TCD at least every 12 weeks.

This trial was also stopped after 74 patients were enrolled when it was clear that stopping transfusion was associated with rapid reversion of TCD velocities to high-risk levels only in children who stopped transfusion (Figure 6). Two strokes occurred in children who had reversion to abnormal TCD velocities and before the reinstitution of transfusion.³² Even in this low-risk subset (those with severe stenosis on MRA and whose TCD velocity did not normalize after ≥30 months of transfusion were not included), by the end of 1 year, more than half of those who discontinued transfusion had restarted it. Although the long-term problems of transfusion, especially iron overload, can be predicted and managed, new approaches that limit the long-term use of this powerful but intensive therapy are needed.

Although there are theories (see Figure 1 proposed by Platt⁹) as to how circle of Willis vasculopathy develops, there are no animal models. Much attention has been given to the abnormal tenacity with which RBCs (especially immature cells that contain sickle hemoglobin) adhere to the endothelium in cultured cell preparations as a cause of vascular occlusion generally in SCD. White blood cells and platelets are probably involved to some extent as well. How RBC adherence might initiate or promote vascular injury and eventual stenosis of large arteries, such as those of the circle of Willis, requires further study.

A connection between SCD and nitric oxide has become a subject of increasing interest in the past several years.³³ It has long been known that plasma free hemoglobin inactivates nitric oxide. The hemolytic anemia of SCD releases significant amounts of free hemoglobin that comes into contact with the endothelial surface. This is believed to cause abnormalities in vascular function, which could include impairment of vasodilation, one of the functions that nitric oxide plays in systemic circulation. Although it is relatively easy to see how consumption of nitric oxide and a secondary feature of hemolysis, an increase in arginase activity in the serum, which further reduces the available nitric oxide by depleting the substrain, lead to pulmonary hypertension by reduction in vasodilation, it is less easy to see how aberrations in nitric oxide might lead to cerebral vasculopathy. Nitric oxide has other functions, such as reduction in inflammatory mediators, and this may contribute to cerebral vasculopathy. An animal model that mimics the development of abnormally high cerebral blood flow rates and the development of large-vessel vasculopathy in the circle of Willis is a critical need to understand how nitric oxide and other metabolic factors contribute to stroke risk in SCD.

Finally, there is an encouraging indication that the approach proved in STOP may be working to reduce stroke. Fullerton et al³⁴ evaluated administrative data in California comparing the rates of hospital admission for first stroke in children with SCD between the early 1990s (before STOP) and from 1998 to 2000 (after STOP) and found a sharp reduction in first stroke admissions, whereas overall SCD and SCD pain admissions did not decline. Such data do not establish a long-term trend or prove that the change is due to use of the STOP primary prevention strategy, but they are encouraging. While we try to learn more about why these large strokes afflict these small patients, we do have a way to make a positive impact now. Early and repeated screening should result in reduction in the prevalence of severe arterial disease and stroke in SCD, albeit at the price of extensive use of transfusion³⁵ until other therapies are developed. Considering the reports from the 2 STOP trials³⁶ still to be published and 2 currently ongoing clinical trials in children with SCD, one testing other approaches to screening (Silent Cerebral Infarct Multi-Center Clinical Trial)³⁷ and the other testing hydroxyurea compared with transfusion for secondary stroke prevention (Stroke With Transfusions Changing to Hydroxyurea trial),³⁸ we can look forward to extensive data from 4 randomized controlled trials aimed at stroke prevention in children with SCD.

Correspondence: Robert J. Adams, MS, MD, South Carolina Center of Economic Excellence, Medical University of South Carolina Stroke Center, 96 Jonathan Lucas St, Charleston, SC 29425 (adamsrj@musc.edu).

Accepted for Publication: January 18, 2007.

Financial Disclosure: Dr Adams has given lectures sponsored by Boehringer Ingelheim, Bristol-Myers Squibb, Sanofi-Aventis, and Novartis; has served as a consultant for Boehringer Ingelheim, Bristol-Myers Squibb, Sanofi-Aventis, and Merck Pharmaceuticals; has served on advisory boards for Boehringer Ingelheim, Bristol-Myers Squibb, and Sanofi-Aventis; has received small, unrestricted educational grants or support for training courses from Boehringer Ingelheim, Bristol-Myers Squibb, and Nicolet; and receives honoraria from Novartis for speaking about SCD prevention and the management of iron overload.

Funding/Support: This study was funded by the National Institutes of Health.

Additional Contributions: I thank the study participants from the Medical College of Georgia SCD Cohort Study, STOP, and STOP II and the many investigators and staff for their work on these studies; Thomas Robert Swift, MD, for his enduring encouragement throughout my career, and several colleagues, including Virgil McKie, MD, and Kathy McKie, MD, the very dedicated pediatricians who early on invested their trust and their patients' well-being in this research, thereby making it possible; Fenwick Nichols, MD, for his expert assistance with TCD and his knowledge of the field; Don Brambilla, PhD, for serving as the statistical expert and co-investigator, and Dianne Gallagher, MS, Suzanne Granger, MS, and many others on the staff at New England Research Institutes; Robert Zimmerman, MD, for leading the imaging review core; Steve Roach, MD, for leading the event adjudication core; the many dedicated physicians, research nurses, and coordinators for carrying out the day-to-day work; my staff (Betsy Carl Rhode, Nadine Odo, and Judi Schweitzer, to name only a few) for their long service and assistance; the research support staff at the Medical College of Georgia; Anne Jones, RN, for volunteering for more than 15 years to advance this work; Mike Jensen of the Medical College of Georgia, for preparing the figures for this article; Judy Luden, for performing and reading more than 10 000 TCDs to help reduce childhood stroke; Ann Sapp and Judi Schweitzer, for helping with the manuscript; my wife, Gaye Adams, MD, for her faithful support; my son Chris, for volunteering as a TCD training subject many times in the early days; and my family for their encouragement. We remember Charles Pegelow, MD, David Ode, MD, and Katie Allen, RN, who contributed to this work; the several patients who died during STOP and STOP II; and Larry Brass, MD, who worked in this field early in his career and who died in 2006.