|
|
| |||||||||||||||||||||||||||||||||||||||||
Background The ability to identify infants with sickle cell anemia who are likely to have severe complications later in life would permit accurate prognostication and tailoring of therapy to match disease-related risks and facilitate planning of clinical trials. We attempted to define the features of such babies by following the clinical course of 392 children with sickle cell disease from infancy to about the age of 10 years.
Methods We analyzed the records of 392 infants who received the diagnosis of homozygous sickle cell anemia or sickle cell–ß0-thalassemia before the age of six months and for whom comprehensive clinical and laboratory data were recorded prospectively; data were available for a mean (±SD) of 10.0±4.8 years. Results obtained before the age of two years were evaluated to determine whether they predicted the outcome later in life.
Results Of the 392 infants in the cohort, 70 (18 percent) subsequently had an adverse outcome, defined as death (18 patients [26 percent]), stroke (25 [36 percent]), frequent pain (17 [24 percent]), or recurrent acute chest syndrome (10 [14 percent]). Using multivariate analysis, we found three statistically significant predictors of an adverse outcome: an episode of dactylitis (defined as pain and tenderness in the hands or feet) before the age of one year (relative risk of an adverse outcome, 2.55; 95 percent confidence interval, 1.39 to 4.67), a hemoglobin level of less than 7 g per deciliter (relative risk, 2.47; 95 percent confidence interval, 1.14 to 5.33), and leukocytosis in the absence of infection (relative risk, 1.80; 95 percent confidence interval, 1.05 to 3.09).
Conclusions Three easily identifiable manifestations of sickle cell disease that may appear in the first two years of life (dactylitis, severe anemia, and leukocytosis) can help to predict the possibility of severe sickle cell disease later in life.
The Cooperative Study of Sickle Cell Disease15,16 included a cohort of infants with sickle cell anemia or sickle cell–ß0-thalassemia who were enrolled before six months of age and monitored for complications. Laboratory tests were also performed during a mean (±SD) follow-up of 10.0±4.8 years.17 We used these results to determine whether the features of the disease early in life could be used to predict its severity later in life.
Methods
Patients
Recruitment procedures for the study have been described elsewhere.15,16 Consent for participation in the study was obtained from the patients' parents or guardians in accordance with the requirements and guidelines of the human subjects committee at each participating clinical center. From October 1978 through October 1988, 414 infants with homozygous sickle cell anemia or sickle cell–ß0-thalassemia were enrolled in the study before they were six months old. Most of the infants were identified as newborns by screening programs. Of these 414 infants, 22 were excluded because they were followed for less than one year, leaving 380 patients with homozygous sickle cell anemia and 12 with sickle cell–ß0-thalassemia. The mean age at entry into the study was 3.0±1.5 months, and the mean duration of follow-up was 10.0± 4.8 years. Follow-up ended on August 31, 1998, and the data base was closed in August 1999.
An additional group of 105 patients with sickle cell anemia, plus 6 with sickle cell–ß0-thalassemia, termed the validation cohort, were enrolled in the study from 1979 through 1981. These 111 children were enrolled at six months to two years of age and had complete clinical and laboratory data available. Their mean age at enrollment was 14.9±5.4 months, and the mean duration of follow-up was 5.8±1.3 years. Data on these patients were collected prospectively only after enrollment, but information was obtained at study entry about acute events that occurred earlier.
Laboratory Methods
Hemoglobin typing was performed at the Centers for Disease Control and Prevention with the use of cellulose acetate and citrate agar electrophoresis,18 column chromatography (for the measurement of hemoglobin A2), and alkali denaturation (for the measurement of fetal hemoglobin).19 The blot hybridization method was used to map the 
-globin gene20 in 321 patients (82 percent) and identify the ßs-globin gene haplotypes21 in 304 patients (78 percent). Leukocyte counts were not corrected for the presence of nucleated red cells.22 The percentage of pocked (vesiculated) red cells (an indicator of splenic function) was calculated with the use of interference phase-contrast microscopy.23 Hematologic data collected in two children after splenectomy were not included.
Definitions of Clinical Events
Stroke was defined as an acute neurologic syndrome due to vascular occlusion or hemorrhage in which neurologic symptoms or signs lasted more than 24 hours. Transient ischemic attacks and silent infarctions24 were not included. All diagnoses of stroke were confirmed by three pediatric neurologists. A painful event was defined as pain in the arms and legs, back, abdomen, chest, or head that lasted at least two hours, led to a request for medical care, and was attributable to sickle cell disease. Dactylitis, the acute chest syndrome, the right upper quadrant syndrome, osteomyelitis, and appendicitis were not counted as painful events. The acute chest syndrome was defined as the presence of a new pulmonary infiltrate on a chest x-ray film, a defect on radionuclide imaging of the chest, or both, in association with an acute respiratory tract illness.
Dactylitis was defined as pain and tenderness, with or without swelling, in the hands, feet, or both. Acute splenic sequestration was defined as a decrease from base line in the hemoglobin level or hematocrit of at least 20 percent plus a simultaneous increase in the size of the spleen to at least 2 cm below the left costal margin.
Adverse Events
Four adverse events served as proxies for severe sickle cell disease: death known or believed to be related to sickle cell disease; stroke; an average of at least two painful events per year for the entire follow-up period, with at least two events per year for three consecutive years (referred to as frequent pain); and an average of at least one episode of acute chest syndrome per year for the entire follow-up period, with at least one episode per year for three consecutive years (referred to as recurrent acute chest syndrome). Pain and the acute chest syndrome were chosen not only because of the morbidity associated with them, but also because of the association of the acute chest syndrome with chronic lung disease25 and of both complications with death during adulthood.3 Rates were calculated from enrollment through the end of follow-up or until the initiation of regular treatment with transfusions (in the case of 39 patients) or hydroxyurea therapy (in the case of 9 patients).
Statistical Analysis
The Cox proportional-hazards model was used to determine the age at which a severe outcome occurred.26 All covariates were based on data collected up to the age of two years, the occurrence of an adverse event, or the initiation of regular transfusions, whichever occurred first.
Well-child visits were required every three months in the second year of life. We analyzed laboratory data collected during these routine visits from the ages of 11 to 25 months to determine steady-state values. The steady-state hemoglobin level and leukocyte count were based on measurements obtained during a median of four visits, the steady-state percentage of reticulocytes was based on measurements obtained during a median of three visits, and the steady-state platelet count was based on measurements obtained during a median of two visits. The steady-state fetal hemoglobin level was based on measurements obtained during a median of one visit made between the ages of 17 and 25 months. These mean laboratory values were first examined as continuous covariates to explore potential linear and nonlinear relations. In order to identify the patients at highest risk for severe disease, we created indicators based on the mean laboratory values for the 75th, 90th, and 95th percentiles for the leukocyte count, hemoglobin level, and percentage of reticulocytes and the 5th, 10th, and 25th percentiles for total hemoglobin and fetal hemoglobin levels. On the basis of prior research,23 we defined an early elevated pocked red-cell value as a value of at least 3.5 percent on at least one visit between birth and the age of 12 months; measurements were obtained during a median of four visits. Indicators for early dactylitis and splenic sequestration were created, with "early" defined as occurrences before one and before two years of age, respectively.
Variables with a P value of 0.20 or less on univariate analyses were evaluated in a multivariate model. The assumption used for the proportional-hazards model was found to be valid for the multivariate model.
After we developed a prognostic model based on findings in the original cohort of patients, we applied the model to data from the validation cohort to evaluate the predictive power of these findings. We fit a logistic-regression model using factors from the multivariate Cox model along with a covariate denoting the duration of follow-up for each patient. The sensitivity and specificity of the model were estimated for each cohort. We used the jackknife procedure27 in the analysis of the original cohort to reduce the possibility of bias in the estimates of sensitivity and specificity.
Results
Characteristics of the Patients
Of the 392 patients enrolled in the study, 70 (18 percent) had adverse outcomes that qualified their disease as severe (Table 1). In this group the mean duration of follow-up was 9.0±4.9 years, and in the remaining 322 patients, it was 10.2±4.7 years. Half the patients in the group with severe disease were boys. The most common adverse event that resulted in the classification of disease as severe was stroke (36 percent). The causes of death in this group of 70 patients were infection (56 percent), the acute chest syndrome (6 percent), splenic sequestration (6 percent), and other causes thought to be related to the disease (33 percent). The mean age at which infants were classified as having severe disease due to the acute chest syndrome was 3.5 years. Frequent attacks of the acute chest syndrome, however, did not predict the occurrence of other severe complications (stroke or pain) or death. For patients classified as having severe disease because of frequent pain, the mean number of painful events during the entire follow-up period was 41.7±10.0, and for patients classified as having severe disease because of recurrent acute chest syndrome, the mean number of episodes was 16.9±9.2.
|
The mean steady-state hemoglobin level for the entire cohort was 9.0±1.3 g per deciliter (Table 2). The steady-state hemoglobin level was the value obtained during a routine visit during the second year of life, when the child had no acute medical problems. The steady-state hemoglobin levels correlated linearly and inversely with the severity of disease (P<0.001) (Table 2). The patients with severe anemia (defined as a hemoglobin level of less than 7 g per deciliter) were 2.64 times as likely to have severe disease as patients with hemoglobin levels of at least 7 g per deciliter (P=0.01).
|
Clinical Data
As compared with patients who did not have dactylitis before the age of one year, patients who had dactylitis before their first birthday were 2.67 times as likely to have severe disease (P<0.001). Of the 41 infants with early dactylitis, 80 percent had swelling of both hands and both feet; only 2 infants had no documented swelling. There was no association between the risk of severe disease and the occurrence of splenic sequestration before the age of one year (P=0.18).
Genetic Factors
One third of the tested children had -thalassemia (two or three -globin genes), and this trait was not significantly associated with the risk of severe disease (relative risk, 0.72; 95 percent confidence interval, 0.40 to 1.30; P=0.27). None of the most common ßs-globin gene haplotypes, BEN/BEN (35.5 percent), BEN/CAR (23.0 percent), BEN/SEN (10.2 percent), or CAR/CAR (5.9 percent), were significantly associated with the risk of severe disease, nor were indicators for the presence of any single haplotype (BEN, CAR, SEN, or CAM).
Multivariate Analysis
The multivariate model included three variables that were significantly associated with an adverse clinical course: early dactylitis (P=0.002), leukocytosis (P= 0.03), and a hemoglobin level of less than 7 g per deciliter (P=0.02) (Table 3). The percentage of reticulocytes (P=0.08), the occurrence of early splenic sequestration (P=0.68), and the percentage of fetal hemoglobin (P=0.62) did not significantly affect the multivariate model. There were no significant interactions among any of the variables in the final model.
|
|
Evaluation of the Model
A receiver-operating-characteristic curve was constructed on the basis of the final model in order to evaluate the usefulness of different classifications of severity (Figure 2). Adopting a classification that was associated with a low false positive rate was of prime importance in order to minimize the likelihood that a patient who did not have severe disease would be referred for a high-risk therapy. The use of a predicted probability of severe disease of at least 36 percent, or twice the observed rate in our original cohort, as the criterion for severe disease was associated with a sensitivity of 23 percent and a specificity of 91 percent (i.e., a 9 percent false positive rate).
|
Fifteen members (13.5 percent) of the validation cohort had an adverse outcome (five died, five had a stroke, four had recurrent painful events, and one had recurrent acute chest syndrome), a rate similar to that of the original cohort (P=0.27). The validation cohort had rates of early dactylitis (P=0.86) and early splenic sequestration (P=0.74) that were similar to those of the original cohort. Perhaps because hematologic measurements were obtained at older ages on average in the validation cohort than in the original cohort (although within the first and second years of life), the mean hemoglobin level (8.7 vs. 9.0 g per deciliter, P=0.04) and percentage of fetal hemoglobin (14.8 percent vs. 16.9 percent, P=0.03) were lower in this group, and the leukocyte count (15.0 x 103 vs. 13.7 x 103 per cubic millimeter, P=0.04) was higher. These differences may in part explain why our logistic-regression model did not result in estimates of relative risk that were similar to those obtained in the original cohort.
By using a predicted probability of 36 percent that a child would have a severe course (Figure 2), we predicted that four members of the validation cohort would have a severe course, whereas three actually did so (positive predictive value, 75 percent). The negative predictive value — that is, the proportion of patients who did not have severe disease among those estimated not to have severe disease — was 89 percent (95 of 107). The use of this classification would result in the identification of only 20 percent of patients with severe disease (3 of 15); however, the rate of false positive results (1 of 96) would be low (1 percent).
Discussion
In this observational study of children with sickle cell anemia, we found that dactylitis, a steady-state hemoglobin level of less than 7 g per deciliter, and leukocytosis in the absence of infection at an early age correlated significantly with adverse outcomes later in childhood. Early loss of splenic function, as indicated by an increase in the percentage of pocked red cells in the blood, may also be prognostically important. The relevant laboratory tests for predicting outcome — measurement of hemoglobin levels and white-cell count — are readily available,30 but steady-state values must be obtained at a clinic visit when there are no acute complications or other illnesses.
A low hemoglobin level has previously been shown to correlate with an increased risk of death in childhood4 or adulthood3 and of stroke.8 However, higher hemoglobin levels correlate with increased risks of the acute chest syndrome9 and painful crisis.10 Although almost 40 percent of the children with severe disease in our study were so classified because of frequent episodes of pain and the acute chest syndrome, a low hemoglobin level was associated with the composite end point of severe disease. Perhaps low hemoglobin levels in infancy are determinants of severity in children, whereas the deleterious effects of an elevated hemoglobin level are cumulative.
An elevated leukocyte count was an independent predictor of the severity of disease in our analysis and was associated with the risk of stroke in a cohort of Jamaican patients.31 Whether leukocytes contribute to the pathogenesis of sickle cell disease, perhaps by releasing cytotoxic enzymes,32 is unknown. The adverse effect of neutrophils on vascular endothelium,33 which is related in part to abnormal adhesion,34 is of particular interest with regard to stroke and cerebrovascular disease in patients with sickle cell anemia.35
A low fetal hemoglobin level was not independently associated with the risk of severe disease. This may be due to its correlation with total hemoglobin, or to a low statistical power of the study, since over half the cohort did not have values for fetal hemoglobin recorded at 17 to 25 months of age. Accelerated reduction in the fetal hemoglobin level during infancy has been associated with loss of splenic function23 (reflected by a pocked-red-cell value of 3.5 percent or more), and the occurrence of this event before one year of age may be strongly associated with the risk of severe disease. However, the importance of this finding is reduced by the known variation between laboratories in the performance of this test36 and by the fact that this test is not usually performed during routine visits. Neither the ßs-globin gene haplotype nor the presence of -thalassemia37 was associated with the risk of severe disease.
How should clinicians use this index? We must emphasize that there may be other factors that we did not examine or recognize that, if added to our model, would enhance our ability to predict a severe outcome. To minimize the false positive rate, we suggest that children whose probability of severe disease is 36 percent or greater (i.e., at least twice that in the general population of children with sickle cell disease) be classified as having a high risk of severe disease. Although this conservative approach will not identify all children who are destined to have a severe course, its use will make it extremely unlikely that a child with mild disease will be considered for a potentially dangerous intervention, such as stem-cell transplantation.
The best use of this prognostic model is to estimate the level of risk, rather than strictly to classify a patient as having or not having severe disease. This approach may be useful in designing therapeutic clinical trials. Excluding children with relatively mild manifestations of disease will allow a larger treatment effect to be easily observed, thus reducing the number of subjects required in the study, and prevent low-risk children from receiving high-risk therapy.
Our model has been validated only retrospectively, and therefore it should also be validated by other means. We urge any who use the model as a means for justifying therapeutic intervention to do so only in controlled clinical trials, with thorough discussions among physicians, patients, and their families of the risks and benefits of both watchful waiting and the proposed intervention.
Supported in part by a contract (N01-HB-47110) with the National Heart, Lung, and Blood Institute.
We are indebted to Dr. Clarice Reid for her support of the investigators, patients, and families over the past two decades; to Dianne Gallagher and Drs. Samuel Charache, William Mentzer, Wendell Rosse, Myron Waclawiw, and Duane Bonds for their helpful review of the manuscript; to Dr. Stephen Embury at the University of California, San Francisco, for
-globin gene mapping and ßs-globin gene haplotyping; to Dr. Howard Pearson at Yale University for performing pocked-red-cell counts; and to the Centers for Disease Control and Prevention, Atlanta, for phenotyping.
Source Information
From the State University of New York–Downstate Medical Center, Brooklyn (S.T.M.); New England Research Institutes, Watertown, Mass. (L.A.S., L.E.E., S.J.W.); the University of Miami, Miami (C.H.P.); St. Jude Children's Research Hospital, Memphis, Tenn. (W.C.W.); St. Luke's–Roosevelt Medical Center, New York (D.L.W.); Harlem Hospital Center, New York (J.S.); and Duke University Medical Center, Durham, N.C. (T.R.K.).
Address reprint requests to Dr. Miller at the State University of New York–Downstate Medical Center, 450 Clarkson Ave., Box 49, Brooklyn, NY 11203, or at stmseelig{at}aol.com.
References
The following institutions and principal investigators participated in the study: L. McMahon (Boston Medical Center, Boston); O. Platt (Children's Hospital, Boston); K. Ohene-Frempong, F. Gill (Children's Hospital of Philadelphia, Philadelphia); E. Vichinsky, B. Lubin (Children's Hospital Medical Center of Northern California, Oakland); S. Leiken, J. Kelleher, G. Bray (Children's National Medical Center, Washington, D.C.); S. Piomelli (Columbia–Presbyterian Medical Center, New York); T. Kinney, R. Ware (Duke University Medical Center, Durham, N.C.); J. Smith, Y. Khakoo (Harlem Hospital, New York); R. Scott, O. Castro, C. Reindorf, C. Uy-Lee (Howard University, Washington, D.C.); R. Bellevue, L. Guarini (Interfaith Medical Center, Brooklyn, N.Y.); W. Mentzer, W. Lande (San Francisco General Hospital, San Francisco); W. Wang, J. Wilimas (St. Jude Children's Research Hospital, Memphis, Tenn.); D. Wethers, R. Grover (St. Luke's–Roosevelt Medical Center, New York); A. Brown, S. Miller (State University of New York Health Science Center at Brooklyn, Brooklyn); N. Talischy-Zahed (University of Illinois, Chicago); C. Pegelow (University of Miami, Miami); H. Zarkowsky, A. Schwartz, M. DeBaun (Washington University, St. Louis); R. Chilcote, J. Moohr, C. Dampier, U. Subramanian (Wyler Children's Hospital, Chicago); H. Pearson, A. Ritchey (Yale University, New Haven, Conn.); statistical coordinating centers — E. Chen, P. Levy, M. West, M. Espeland, D. Gallagher (University of Illinois School of Public Health, Chicago) and S. McKinlay, D. Gallagher, L. Sleeper, E. Wright, D. Brambilla (New England Research Institutes, Watertown, Mass.); program administrators for the Division of Blood Diseases and Resources, National Heart, Lung, and Blood Institute, Bethesda, Md. — C. Reid, M. Gaston, D. Bonds, J. Verter, M. Waclawiw.
| |||||||||||||||||||||||||||||||||||||||||
Related Letters:
Prediction of Adverse Outcomes in Children with Sickle Cell Disease
Horan J., Lerner N., Miller S. T., Sleeper L. A., Enos L. E.
Extract |
Full Text
N Engl J Med 2000;
342:1612-1613, May 25, 2000.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
