- Email: [email protected]
A Predictive Instrument for Coronary Artery Aneurysms in Kawasaki Disease 1
A Predictive Instrument for Coronary Artery Aneurysms in Kawasaki Disease Alexa S. Beiser, PhD, Masato Takahashi, MD, Annette L. Baker, RN, MSN, PNP, Robert P. Sundel, MD, and Jane W. Newburger, MD, MPH, for the US Multicenter Kawasaki Disease Study Group* To construct a predictive instrument for developing coronary artery abnormalities in patients with acute Kawasaki disease treated with aspirin and intravenous gamma globulin within the first 10 days of illness, data available from a multicenter database of patients with acute Kawasaki disease were analyzed. A development data set (n 5 212) was used to construct a sequential risk classification instrument based on easily measured baseline laboratory test results and temperature. The instrument was then validated in 3 test data sets (n 5 192, 264, and 92, respectively). Risk factors used in the sequential classification instrument included baseline neutrophil and band counts, hemoglobin concentration, platelet count, and temperature on the day after infusion of intravenous gamma globulin. In the development
data set, the instrument classified 123 of 212 patients (58%) as low risk; none developed coronary artery abnormalities. Among 89 patients classified as high risk, 3 of 36 female (8.3%) and 9 of 53 male patients (17.0%) developed coronary artery abnormalities. The instrument performed similarly in the 3 test data sets; no patient in any data set classified as low risk developed coronary artery abnormalities. This simple instrument allows the clinician to identify within 1 day of treatment low-risk children in whom extensive and frequent cardiac testing may be unnecessary, as well as high-risk children who require closer monitoring and may be candidates for additional therapies. Q1998 by Excerpta Medica, Inc. (Am J Cardiol 1998;81:1116 –1120)
awasaki disease is an acute illness of childhood characterized by fever, rash, conjunctivitis, inK flammation of the mucous membranes, swollen, ery-
who were treated with aspirin alone rather than with aspirin plus gamma globulin. The present study constructed a predictive instrument for the development of coronary artery lesions among patients treated with high-dose IVGG within the first 10 days of illness onset.
thematous hands and feet, and cervical adenopathy. Histopathologic features of vasculitis involving arterioles, capillaries, and venules occur in the earliest phase of the disease. Subsequently, the walls of the coronary arteries and other medium-size muscular arteries may show focal segmental destruction, with coronary artery aneurysms or ectasia developing in approximately 15% to 25% of affected children.1–3 High-dose intravenous gamma globulin (IVGG) therapy, together with aspirin, has been demonstrated to be safe and effective in reducing the prevalence of coronary artery abnormalities when administered early in the course of Kawasaki disease.4 –7 Even when treated with high-dose IVGG regimens, however, approximately 5% of children with Kawasaki disease develop at least transient coronary artery dilation and 1% develop giant aneurysms.8 Several scoring systems have been developed to identify children at highest risk for formation of coronary artery abnormalities.9 –12 These scoring systems are based on patients From the Department of Epidemiology and Biostatistics, Boston University School of Public Health, Boston, Massachusetts; the Department of Pediatrics, Children’s Hospital of Los Angeles, Los Angeles, California; and the Departments of Cardiology and Medicine, Division of Immunology (Rheumatology), Children’s Hospital, Boston, Massachusetts. This study was supported, in part, by grants HL34545 and HL48606 from the National Institutes of Health, Bethesda, Maryland, and by the Kobren Fund, Boston, Massachusetts. Manuscript received November 13, 1997; revised manuscript received and accepted January 28, 1998. Address for reprints: Jane W. Newburger, MD, Department of Cardiology, Children’s Hospital, 300 Longwood Avenue, Boston, Massachusetts 02115. *See Appendix for members of the study group.
1116
©1998 by Excerpta Medica, Inc. All rights reserved.
METHODS
Patients: Data were collected on patients with acute Kawasaki disease who were enrolled in 1 of 2 therapeutic trials of IVGG or during the interim phase between the trials.6,13 We enrolled patients at 7 clinical centers in the US from February 1984 through November 1989. Among patients enrolled in trial I, half were randomized to receive IVGG 400 mg/kg/ day, administered over 2 hours, for 4 consecutive days. The remaining half did not receive IVGG and were excluded from these analyses. In trial II, patients were randomly assigned to either the 4-infusion IVGG regimen or a single infusion of 2 g/kg given over approximately 10 hours. During the interim between trials I and II, patients received the same 4-infusion IVGG regimen as in trial I. Aspirin was administered to all patients at a dose of 100 mg (range, 80 to 120) per kilogram per day through day 14 of the illness and then 3 to 5 mg/kg/day through the seventh week of illness. In addition, data were gathered prospectively on patients with Kawasaki disease evaluated and treated with IVGG (2 g/kg over 8 to 10 hours) at the Boston Children’s Hospital after the termination of trial II from November 1989 through November 1992. Enrolled patients had at least 5 of the following: fever; nonexudative conjunctival infection; changes in the oropharynx, including mucosal erythema, dry fissured lips, and ‘‘strawberry tongue’’; changes in the 0002-9149/98/$19.00 PII S0002-9149(98)00116-7
extremities, including erythema of the palms and soles, edema of the hands and feet, or periungual desquamation in the subacute phase of the disease; rash; and cervical adenopathy ($1 node at least 1.5 cm in diameter). In addition, patients were enrolled within 10 days of the onset of illness and could have no clinical or laboratory evidence of any other disease known to mimic Kawasaki syndrome. We excluded from this analysis patients with coronary artery abnormalities before treatment with IVGG. Assessment of coronary arteries (echocardiography):
Two-dimensional echocardiography was performed at the time of diagnosis at participating clinical centers. These tests were repeated approximately 2 and 7 weeks later. The primary outcome variable was the development of a coronary artery abnormality in at least 1 arterial segment on either of the echocardiograms performed 2 or 7 weeks after diagnosis. Coronary artery abnormalities were defined as follows: (1) lumen diameter at least 3 mm in a child ,5 years old or at least 4 mm in a child $5 years; (2) internal diameter of a segment at least 1.5 times as large as that of an adjacent segment; or (3) clearly irregular lumen. During the trial period, echocardiograms were interpreted independently by 2 echocardiographers who were blinded to treatment assignment. A third reading was performed if necessary to resolve discordant interpretations. In patients studied beyond the trial period, echocardiograms were interpreted by a single reader without clinical knowledge of the patient. Data sets: We randomly sampled half of the patients who received the 4-infusion regimen (during trial I, the interim phase, or trial II) and had normal echocardiograms at diagnosis to create a development data set. To ensure representative outcome prevalences in the 2 groups, we performed the sampling separately for those who did and did not develop coronary artery abnormalities. We validated models on 3 test data sets: the remaining half of the patients who received the 4-infusion regimen (test data set 1), patients who received a single infusion of IVGG during trial II (test data set 2), and patients seen at Boston Children’s Hospital since trial II enrollment ended (test data set 3). Statistical analysis: Risk classification instruments were developed using a sequential classification approach. The following variables, measured in the 24 hours before treatment with IVGG, were considered as potential risk factors: age (years), sex, race, temperature (°C), white blood cell count, differential white blood cell count, absolute band count, absolute lymphocyte count, hemoglobin concentration, and platelet count. We also included temperature measured the day after the initiation of treatment. We excluded sparse variables (at least 20% missing data) from consideration in model development; these included serum immunoglobin (Ig)G, IgA, and IgM; quantitative C-reactive protein; hematocrit; a1-antitrypsin; albumin; and blood urea nitrogen. We performed univariate analyses on individual potential risk factors and on combination variables
based on interactions between hematologic parameters (e.g., low neutrophils and low bands as a percentage of neutrophils). We used chi-square tests for categoric variables and univariate logistic regression for continuous variables. Risk factors for the development of coronary abnormalities identified during this process were considered for multivariate analysis. Multivariate logistic regression techniques were not used, because maximum likelihood parameter estimation was not possible for models including empty cells. Patients were assigned to risk groups at each step of the sequential classification process according to their values for the risk factor under consideration during that step. We developed many such processes, each using a different combination of risk factors. Each instrument produced predicted outcomes of high or low risk. Instrument performance was assessed by sensitivity, specificity, predictive values (positive and negative), and percentage of patients classified as low risk. Instruments that performed well on the development data set were validated using each of the 3 test data sets.
RESULTS
Data set creation: We randomly divided the 404 patients who had normal echocardiograms at diagnosis, received the 4-infusion regimen of IVGG, and had follow-up echocardiograms at 2 or 7 weeks after treatment into a development data set and test data set 1. Twelve of the 212 patients (5.7%) in the development data set developed coronary artery abnormalities; 10 of the 192 patients (5.2%) in test data set 1 developed such abnormalities. Patients enrolled during or after trial II who had normal echocardiograms at diagnosis, received a single infusion of IVGG, and had follow-up echocardiograms at 2 or 7 weeks after treatment formed 2 other test data sets. Seven of the 264 patients (2.7%) in test data set 2 and 4 of the 92 patients (4.3%) in test data set 3 developed coronary artery abnormalities. Sequential classification approach: Numerous schemes were developed using the sequential classification approach. Among these, the predictive instrument depicted in Figure 1 was the most accurate in predicting risk of coronary artery aneurysms. First, patients were classified according to baseline band and neutrophil counts. At this first step, the low-risk group consisted of patients whose total leukocytes were composed of either (1) ,50% neutrophils, of which ,50% were bands, or (2) ,75% neutrophils, of which ,10% were bands. Of the 212 patients, 57 (26.9%) were in this low-risk group; none had subsequent coronary artery abnormalities. Second, a high-risk group was selected from among the remaining 155 patients according to baseline hemoglobin concentration and platelet count. Those with a baseline hemoglobin concentration ,10 mg/dl or a baseline platelet count ,350,000/mm3 formed the high-risk group. Of the original 212 patients, 62 (29.2%) were in this high-risk group; coronary lesions developed in 8 (12.9%). The remaining 93 patients (43.9%) formed a middle-risk group; cor-
CORONARY ARTERY DISEASE/PREDICTING ANEURYSMS IN KAWASAKI DISEASE
1117
scheme in the development data set were 100% and 63%, respectively, for females, and 100% and 60%, respectively, for males. This final predictive instrument was subsequently validated on each of the 3 test data sets (Table I). The instrument classified as low risk 52%, 64%, and 58% of patients in test data sets 1, 2, and 3, respectively; none of these patients developed coronary artery abnormalities (i.e., the predictive value negative was 100%). The predictive value positives ranged from 2.3% to 6.7% for females and from 11.8% to 13.3% for males. An infusion regimen of 1,600 mg/kg over 4 days was used in patients in the development data set and test data set 1, whereas a single large infusion dose (2 g/kg) was administered to those in test data sets 2 and 3. The predictive instrument performed similarly for patients on either regimen. The instrument classified as low risk 52.3% and 65.3% of males treated with the 4-dose and single-dose regimens, respectively. Among females treated with the 4-dose and singledose regimens, 58.7% and 57.9%, respectively, were classified as low risk. No patient classified as low risk developed coronary artery abnormalities. When all data sets were combined, males and females classified as high risk had a 15.0% and 7.2% chance, respectively, of developing coronary abnormalities if treated with the 4-day regimen of 400 mg/kg/day and a 12.0% and 3.4% risk, respectively, if treated with the singledose regimen of 2 g/kg. FIGURE 1. Diagram depicting the performance of our sequential classification scheme in the development data set.
DISCUSSION
onary abnormalities developed in 4 (4.3%) of these patients. Third, we used temperature measured on the day after initiation of IVGG treatment (day 2) to further classify the middle-risk group as low or high risk. Patients whose temperatures were at least 38°C were classified as high risk, with the remaining patients classified as low risk. Among the 27 febrile patients, coronary abnormalities developed in 4 (14.8%). No coronary abnormalities developed in the remaining 66 patients, who were either afebrile or in whom temperature was not recorded. Fourth, the groups of patients classified as high or low risk in all preceding steps were merged. The low-risk group contained the 57 patients selected in the first step and the 66 patients selected in the third step. Among these 123 low-risk patients (58% of the original 212), none developed coronary artery abnormalities. The 62 patients selected in the second step and the 27 patients selected in the third step constituted the high-risk group. Coronary artery abnormalities developed in 12 (13.5%) of these 89 patients. Finally, we divided children in the high-risk group into males and females. Among males classified as high-risk, 9 of 53 (17.0%) developed coronary artery abnormalities; among high-risk females, 3 of 36 (8.3%) developed such abnormalities. The sensitivity and specificity for this classification
Despite early treatment with high-dose IVGG, some children with Kawasaki disease develop progressive abnormalities of the coronary arteries. Separation of patients who will develop aneurysms from those who will remain unaffected early in the course of the disease would be useful in patient management. Using the database of the Multicenter Kawasaki Study Group, we developed an instrument to predict the development of coronary artery lesions among patients treated within the first 10 days of illness with high-dose IVGG. Models based on fever and baseline laboratory data were developed and then applied prospectively using a sequential classification approach. In the combined data sets, no patient classified as low risk developed coronary artery abnormalities. The frequency of development of coronary abnormalities in males classified as high risk was 13.8%, and in highrisk females, 5.5%. Several scoring systems have been developed previously to identify children at highest risk for formation of coronary artery abnormalities. The first such scoring system was reported by Asai in 198310 and was useful at a time in Japan before 2-dimensional echocardiography was widely used for routine monitoring of all patients with Kawasaki disease. Risk factors included male sex, age ,1 year, fever for at least 14 days or recrudescent fever, hemoglobin concentration ,10 g/dl, white blood cell count .30,000 3 1026/L, erythrocyte sedimentation rate .101 mm/hour (Westergren), persistence of elevation
1118 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 81
MAY 1, 1998
TABLE I Instrument Performance on Development and Test Data Sets
Classified low risk Classified high risk CAA among high-risk females CAA among high-risk males Sensitivity Specificity
Development Data Set (n 5 212)
Test Data Set 1 (n 5 192)
Test Data Set 2 (n 5 264)
Test Data Set 3 (n 5 92)
123 (58%) 89 (42%) 3/36 (8.3%) 9/53 (17.0%) 100% 62%
99 (52%) 93 (48%) 2/33 (6.1%) 8/60 (13.3%) 100% 54%
169 (64%) 95 (36%) 1/44 (2.3%) 6/51 (11.8%) 100% 66%
53 (58%) 39 (42%) 1/15 (6.7%) 3/24 (12.5%) 100% 50%
CAA 5 coronary artery aneurysms.
of C-reactive protein or sedimentation abnormality for .30 days, and various cardiac findings consistent with ischemia or myocardial infarction. Duration of fever, presumably reflecting the severity of ongoing vasculitis, has been confirmed as a powerful predictor of coronary artery aneurysms in other studies14 –16 and indeed was the only factor that discriminated between those with and those without aneurysms in a study by Koren et al.14 However, duration of fever cannot be used for risk stratification in the era of almost universal acceptance of early IVGG therapy. The scoring systems by Nakano et al9 and Iwasa et al12 have not been verified using multicenter data. These scoring systems also require repeated laboratory tests over several days before treatment. The Harada risk index11 is used by some Japanese physicians to decide whether IVGG therapy should be administered; risk factors include high white blood cell count, low platelet count, low hematocrit, high C-reactive protein levels, low albumin levels, age ,12 months, and male sex. Age .5 years recently has been found to be associated with a greater incidence of coronary abnormalities.17 In our analysis, lower hemoglobin at presentation was a strong predictor of coronary artery outcome, perhaps indirectly reflecting severity of illness. The anemia of Kawasaki disease is poorly understood. The mean hematocrit of affected children is 2 SDs below the expected mean for age at the time of acute presentation, before the effects of decreased reticulocytosis should be clinically evident. The anemia is normochromic and normocytic, without direct evidence of hemolysis.18 These findings suggest that suppression of bone marrow production of red blood cells may begin well before clinical expression of Kawasaki disease. A similar pattern of anemia has been found with parvovirus and in other acute viral diseases.19 Additional laboratory parameters used in our predictive instrument included a left shift in the white blood cell count and a low platelet count; as noted above, both have been found to be risk factors in previous analyses in aspirin-treated patients. Fever and laboratory parameters of anemia, left shift, and low platelet count secondary to increased platelet turnover20 presumably reflect a greater release of inflammatory mediators and worse vascular inflammation.21–24 Inferences from this data set must be qualified for several reasons. Our instrument was constructed using
data accrued in patients treated with a total IVGG dose of either 1.6 or 2.0 g. Therefore, its performance may not be reliable in patients treated with lower doses of IVGG or with aspirin alone. Similarly, the instrument may be inaccurate in patients who present after the tenth day of illness or who have atypical disease. Temperature data were missing for many patients in the development data set and test data set 1, whereas the data for those in test data sets 2 and 3 were more complete. Febrile patients in whom temperature data were not recorded on the day after IVGG treatment might have been misclassified into the low-risk group. This would cause overestimation of the specificity of the instrument in the development data set and test data set 1 but would not affect sensitivity, as none of these patients developed coronary abnormalities. We used the largest high-quality database currently available: enrollment was based on strictly defined criteria, laboratory data were measured in a standardized fashion, and echocardiograms were interpreted by at least 2 blinded pediatric echocardiographers.
APPENDIX The following were members of the US Multicenter Kawasaki Disease Study Group (in alphabetical order by city and study center): Boston: Department of Pediatrics and Cardiology, Children’s Hospital and Harvard Medical School (Steven D. Colan, MD); Department of Epidemiology and Biostatistics, Boston University School of Public Health (Theodore Colton, ScD); and New England Medical Center and Tufts University School of Medicine, Boston (David R. Fulton, MD, H. Cody Meissner, MD); Chicago: Department of Pediatrics, Children’s Memorial Hospital and Northwestern University School of Medicine (C. Elise Duffy, MD, Anne H. Rowley, MD, Stanford T. Shulman, MD); Denver: Children’s Hospital, University of Colorado Health Sciences Center, and University of Colorado School of Medicine (Mary P. Glode, MD, James W. Wiggins, MD); Hawaii: Kapiolani Children’s Medical Center and John A. Burns School of Medicine (Marian E. Melish, MD, Venudhar Reddy, MD); Los Angeles: Children’s Hospital of Los Angeles and the University of Southern California, School of Medicine (Wilbert H. Mason, MD); San Diego: Children’s Hospital, the University of California Medical Center, and the University of California School of Medicine (John Bastian, MD, Jane C. Burns, MD, Kyung Ja Chung, MD).
1. Kato H, Ichinose E, Yoshioka F, Takechi T, Matsunaga S, Suzuki K, Rikitake
N. Fate of coronary aneurysms in Kawasaki disease: serial coronary angiography and long-term follow-up study. Am J Cardiol 1982;49:1758 –1766. 2. Kato H, Ichinose E, Kawasaki T. Myocardial infarction in Kawasaki disease: clinical analyses in 195 cases. J Pediatr 1986;108:923–927. 3. Suzuki A, Kamiya T, Kuwahara N, Ono Y, Kohata T, Kimura K, Takamiya M. Coronary arterial lesions of Kawasaki disease: cardiac catheterization findings of 1100 cases. Pediatr Cardiol 1986;7:3–9. 4. Furusho K, Kamiya T, Nakano H, Kiyosawa N, Shinomiya K, Hayashidera T, Tamura T, Hirose O, Manabe Y, Yokoyama T. Japanese gamma globulin trials for Kawasaki disease. Prog Clin Biol Res 1987;250:425– 432. 5. Nagashima M, Matsushima M, Matsuoka H, Ogawa A, Okumura N. High-dose gammaglobulin therapy for Kawasaki disease. J Pediatr 1987;110:710 –712. 6. Newburger JW, Takahashi M, Burns JC, Beiser AS, Duffy CE, Glode MP,
CORONARY ARTERY DISEASE/PREDICTING ANEURYSMS IN KAWASAKI DISEASE
1119
Mason WH, Reddy V, Sanders SP. The treatment of Kawasaki syndrome with intravenous gamma globulin. N Engl J Med 1986;315:341–347. 7. Furusho K, Kamiya T, Nakano H, Kiyosawa N, Shinomiya K, Hayashidera T, Tamura T, Hirose O, Manabe Y, Yokoyama T. High-dose intravenous gammaglobulin for Kawasaki disease. Lancet 1984;2:1055–1058. 8. Dajani AS, Taubert KA, Takahashi M, Bierman FZ, Freed MD, Ferrieri P, Gerber M, Shulman ST, Karchmer AW, Wilson W, Peter G, Durack DT, Rahimtoola SH. Guidelines for long-term management of patients with Kawasaki disease. Report from the Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, American Heart Association. Circulation 1994;89:916 –922. 9. Nakano H, Ueda K, Saito A, Tsuchitani Y, Kawamori J, Yoshida T. Scoring method for identifying patients with Kawasaki disease at high risk of coronary artery aneurysms. Am J Cardiol 1986;58:739 –742. 10. Asai T. Evaluation method for the degree of seriousness in Kawasaki disease. Acta Paediatr Jpn 1983;25:170 –175. 11. Harada K, Yamaguchi H, Kato H, Nishibayashi Y, Seki I, Okazaki T, Sato Y, Furusho K, Okawa S, Kawasaki T. Indication for intravenous gamma globulin for Kawasaki disease. In: Takahashi M, Taubert K, eds. Proceedings of the Fourth International Symposium on Kawasaki Disease. Dallas: American Heart Association, 1993:459 – 462. 12. Iwasa M, Sugiyama K, Ando T, Nomura H, Katoh T, Wada Y. Selection of high-risk children for immunoglobulin therapy in Kawasaki disease. Prog Clin Biol Res 1987;250:543–544. 13. Newburger JW, Takahashi M, Beiser AS, Burns JC, Bastian J, Chung KJ, Colan SD, Duffy CE, Fulton DR, Glode MP, et al. A single intravenous infusion of gamma globulin as compared with four infusions in the treatment of acute Kawasaki syndrome. N Engl J Med 1991;324:1633–1639. 14. Koren G, Lavi S, Rose V, Rowe R. Kawasaki disease: review of risk factors for coronary aneurysms. J Pediatr 1986;108:388 –392.
1120 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 81
15. Ichida F, Fatica NS, Engle MA, OLoughlin JE, Klein AA, Snyder MS,
Ehlers KH, Levin AR. Coronary artery involvement in Kawasaki syndrome in Manhattan, New York: risk factors and role of aspirin. Pediatrics 1987;80: 828 – 835. 16. Daniels SR, Specker B, Capannari TE, Schwartz DC, Kaplan S. Correlates of coronary artery aneurysm formation in patients with Kawasaki disease. Am J Dis Child 1987;141:205–207. 17. Yanagawa H, Yashiro M, Nakamura Y, Kawasaki T, Kato H. Epidemiologic pictures of Kawasaki disease in Japan: from the nationwide incidence survey in 1991 and 1992. Pediatrics 1995;95:475– 479. 18. Melish ME. Kawasaki syndrome: a 1986 perspective. Rheum Dis Clin North Am 1987;13:7–17. 19. Jansson LT, Kling S, Dallman PR. Anemia in children with acute infections seen in a primary care pediatric outpatient clinic. Pediatr Infect Dis 1986;5:424 – 427. 20. Burns JC, Glode MP, Clarke SH, Wiggins J Jr. Coagulopathy and platelet activation in Kawasaki syndrome: identification of patients at high risk for development of artery aneurysms. J Pediatr 1984;105:206 –211. 21. Lang BA, Silverman ED, Laxer RM, Rose V, Nelson DI, Rubin LA. Serum-soluble interleukin-2 receptor levels in Kawasaki disease. J Pediatr 1990; 116:592–596. 22. Lin CY, Lin CC, Hwang B, Chiang BN. The changes of interleukin-2, tumor necrotic factor and gamma-interferon production among patients with Kawasaki disease. Eur J Pediatr 1991;150:179 –182. 23. Leung DY. New developments in Kawasaki disease. Curr Opin Rheumatol 1991;3:46 –55. 24. Lin CY, Lin CC, Hwang B, Chiang B. Serial changes of serum interleukin-6, interleukin-8, and tumor necrosis factor alpha among patients with Kawasaki disease. J Pediatr 1992;121:924 –926.
MAY 1, 1998
data set, the instrument classified 123 of 212 patients (58%) as low risk; none developed coronary artery abnormalities. Among 89 patients classified as high risk, 3 of 36 female (8.3%) and 9 of 53 male patients (17.0%) developed coronary artery abnormalities. The instrument performed similarly in the 3 test data sets; no patient in any data set classified as low risk developed coronary artery abnormalities. This simple instrument allows the clinician to identify within 1 day of treatment low-risk children in whom extensive and frequent cardiac testing may be unnecessary, as well as high-risk children who require closer monitoring and may be candidates for additional therapies. Q1998 by Excerpta Medica, Inc. (Am J Cardiol 1998;81:1116 –1120)
awasaki disease is an acute illness of childhood characterized by fever, rash, conjunctivitis, inK flammation of the mucous membranes, swollen, ery-
who were treated with aspirin alone rather than with aspirin plus gamma globulin. The present study constructed a predictive instrument for the development of coronary artery lesions among patients treated with high-dose IVGG within the first 10 days of illness onset.
thematous hands and feet, and cervical adenopathy. Histopathologic features of vasculitis involving arterioles, capillaries, and venules occur in the earliest phase of the disease. Subsequently, the walls of the coronary arteries and other medium-size muscular arteries may show focal segmental destruction, with coronary artery aneurysms or ectasia developing in approximately 15% to 25% of affected children.1–3 High-dose intravenous gamma globulin (IVGG) therapy, together with aspirin, has been demonstrated to be safe and effective in reducing the prevalence of coronary artery abnormalities when administered early in the course of Kawasaki disease.4 –7 Even when treated with high-dose IVGG regimens, however, approximately 5% of children with Kawasaki disease develop at least transient coronary artery dilation and 1% develop giant aneurysms.8 Several scoring systems have been developed to identify children at highest risk for formation of coronary artery abnormalities.9 –12 These scoring systems are based on patients From the Department of Epidemiology and Biostatistics, Boston University School of Public Health, Boston, Massachusetts; the Department of Pediatrics, Children’s Hospital of Los Angeles, Los Angeles, California; and the Departments of Cardiology and Medicine, Division of Immunology (Rheumatology), Children’s Hospital, Boston, Massachusetts. This study was supported, in part, by grants HL34545 and HL48606 from the National Institutes of Health, Bethesda, Maryland, and by the Kobren Fund, Boston, Massachusetts. Manuscript received November 13, 1997; revised manuscript received and accepted January 28, 1998. Address for reprints: Jane W. Newburger, MD, Department of Cardiology, Children’s Hospital, 300 Longwood Avenue, Boston, Massachusetts 02115. *See Appendix for members of the study group.
1116
©1998 by Excerpta Medica, Inc. All rights reserved.
METHODS
Patients: Data were collected on patients with acute Kawasaki disease who were enrolled in 1 of 2 therapeutic trials of IVGG or during the interim phase between the trials.6,13 We enrolled patients at 7 clinical centers in the US from February 1984 through November 1989. Among patients enrolled in trial I, half were randomized to receive IVGG 400 mg/kg/ day, administered over 2 hours, for 4 consecutive days. The remaining half did not receive IVGG and were excluded from these analyses. In trial II, patients were randomly assigned to either the 4-infusion IVGG regimen or a single infusion of 2 g/kg given over approximately 10 hours. During the interim between trials I and II, patients received the same 4-infusion IVGG regimen as in trial I. Aspirin was administered to all patients at a dose of 100 mg (range, 80 to 120) per kilogram per day through day 14 of the illness and then 3 to 5 mg/kg/day through the seventh week of illness. In addition, data were gathered prospectively on patients with Kawasaki disease evaluated and treated with IVGG (2 g/kg over 8 to 10 hours) at the Boston Children’s Hospital after the termination of trial II from November 1989 through November 1992. Enrolled patients had at least 5 of the following: fever; nonexudative conjunctival infection; changes in the oropharynx, including mucosal erythema, dry fissured lips, and ‘‘strawberry tongue’’; changes in the 0002-9149/98/$19.00 PII S0002-9149(98)00116-7
extremities, including erythema of the palms and soles, edema of the hands and feet, or periungual desquamation in the subacute phase of the disease; rash; and cervical adenopathy ($1 node at least 1.5 cm in diameter). In addition, patients were enrolled within 10 days of the onset of illness and could have no clinical or laboratory evidence of any other disease known to mimic Kawasaki syndrome. We excluded from this analysis patients with coronary artery abnormalities before treatment with IVGG. Assessment of coronary arteries (echocardiography):
Two-dimensional echocardiography was performed at the time of diagnosis at participating clinical centers. These tests were repeated approximately 2 and 7 weeks later. The primary outcome variable was the development of a coronary artery abnormality in at least 1 arterial segment on either of the echocardiograms performed 2 or 7 weeks after diagnosis. Coronary artery abnormalities were defined as follows: (1) lumen diameter at least 3 mm in a child ,5 years old or at least 4 mm in a child $5 years; (2) internal diameter of a segment at least 1.5 times as large as that of an adjacent segment; or (3) clearly irregular lumen. During the trial period, echocardiograms were interpreted independently by 2 echocardiographers who were blinded to treatment assignment. A third reading was performed if necessary to resolve discordant interpretations. In patients studied beyond the trial period, echocardiograms were interpreted by a single reader without clinical knowledge of the patient. Data sets: We randomly sampled half of the patients who received the 4-infusion regimen (during trial I, the interim phase, or trial II) and had normal echocardiograms at diagnosis to create a development data set. To ensure representative outcome prevalences in the 2 groups, we performed the sampling separately for those who did and did not develop coronary artery abnormalities. We validated models on 3 test data sets: the remaining half of the patients who received the 4-infusion regimen (test data set 1), patients who received a single infusion of IVGG during trial II (test data set 2), and patients seen at Boston Children’s Hospital since trial II enrollment ended (test data set 3). Statistical analysis: Risk classification instruments were developed using a sequential classification approach. The following variables, measured in the 24 hours before treatment with IVGG, were considered as potential risk factors: age (years), sex, race, temperature (°C), white blood cell count, differential white blood cell count, absolute band count, absolute lymphocyte count, hemoglobin concentration, and platelet count. We also included temperature measured the day after the initiation of treatment. We excluded sparse variables (at least 20% missing data) from consideration in model development; these included serum immunoglobin (Ig)G, IgA, and IgM; quantitative C-reactive protein; hematocrit; a1-antitrypsin; albumin; and blood urea nitrogen. We performed univariate analyses on individual potential risk factors and on combination variables
based on interactions between hematologic parameters (e.g., low neutrophils and low bands as a percentage of neutrophils). We used chi-square tests for categoric variables and univariate logistic regression for continuous variables. Risk factors for the development of coronary abnormalities identified during this process were considered for multivariate analysis. Multivariate logistic regression techniques were not used, because maximum likelihood parameter estimation was not possible for models including empty cells. Patients were assigned to risk groups at each step of the sequential classification process according to their values for the risk factor under consideration during that step. We developed many such processes, each using a different combination of risk factors. Each instrument produced predicted outcomes of high or low risk. Instrument performance was assessed by sensitivity, specificity, predictive values (positive and negative), and percentage of patients classified as low risk. Instruments that performed well on the development data set were validated using each of the 3 test data sets.
RESULTS
Data set creation: We randomly divided the 404 patients who had normal echocardiograms at diagnosis, received the 4-infusion regimen of IVGG, and had follow-up echocardiograms at 2 or 7 weeks after treatment into a development data set and test data set 1. Twelve of the 212 patients (5.7%) in the development data set developed coronary artery abnormalities; 10 of the 192 patients (5.2%) in test data set 1 developed such abnormalities. Patients enrolled during or after trial II who had normal echocardiograms at diagnosis, received a single infusion of IVGG, and had follow-up echocardiograms at 2 or 7 weeks after treatment formed 2 other test data sets. Seven of the 264 patients (2.7%) in test data set 2 and 4 of the 92 patients (4.3%) in test data set 3 developed coronary artery abnormalities. Sequential classification approach: Numerous schemes were developed using the sequential classification approach. Among these, the predictive instrument depicted in Figure 1 was the most accurate in predicting risk of coronary artery aneurysms. First, patients were classified according to baseline band and neutrophil counts. At this first step, the low-risk group consisted of patients whose total leukocytes were composed of either (1) ,50% neutrophils, of which ,50% were bands, or (2) ,75% neutrophils, of which ,10% were bands. Of the 212 patients, 57 (26.9%) were in this low-risk group; none had subsequent coronary artery abnormalities. Second, a high-risk group was selected from among the remaining 155 patients according to baseline hemoglobin concentration and platelet count. Those with a baseline hemoglobin concentration ,10 mg/dl or a baseline platelet count ,350,000/mm3 formed the high-risk group. Of the original 212 patients, 62 (29.2%) were in this high-risk group; coronary lesions developed in 8 (12.9%). The remaining 93 patients (43.9%) formed a middle-risk group; cor-
CORONARY ARTERY DISEASE/PREDICTING ANEURYSMS IN KAWASAKI DISEASE
1117
scheme in the development data set were 100% and 63%, respectively, for females, and 100% and 60%, respectively, for males. This final predictive instrument was subsequently validated on each of the 3 test data sets (Table I). The instrument classified as low risk 52%, 64%, and 58% of patients in test data sets 1, 2, and 3, respectively; none of these patients developed coronary artery abnormalities (i.e., the predictive value negative was 100%). The predictive value positives ranged from 2.3% to 6.7% for females and from 11.8% to 13.3% for males. An infusion regimen of 1,600 mg/kg over 4 days was used in patients in the development data set and test data set 1, whereas a single large infusion dose (2 g/kg) was administered to those in test data sets 2 and 3. The predictive instrument performed similarly for patients on either regimen. The instrument classified as low risk 52.3% and 65.3% of males treated with the 4-dose and single-dose regimens, respectively. Among females treated with the 4-dose and singledose regimens, 58.7% and 57.9%, respectively, were classified as low risk. No patient classified as low risk developed coronary artery abnormalities. When all data sets were combined, males and females classified as high risk had a 15.0% and 7.2% chance, respectively, of developing coronary abnormalities if treated with the 4-day regimen of 400 mg/kg/day and a 12.0% and 3.4% risk, respectively, if treated with the singledose regimen of 2 g/kg. FIGURE 1. Diagram depicting the performance of our sequential classification scheme in the development data set.
DISCUSSION
onary abnormalities developed in 4 (4.3%) of these patients. Third, we used temperature measured on the day after initiation of IVGG treatment (day 2) to further classify the middle-risk group as low or high risk. Patients whose temperatures were at least 38°C were classified as high risk, with the remaining patients classified as low risk. Among the 27 febrile patients, coronary abnormalities developed in 4 (14.8%). No coronary abnormalities developed in the remaining 66 patients, who were either afebrile or in whom temperature was not recorded. Fourth, the groups of patients classified as high or low risk in all preceding steps were merged. The low-risk group contained the 57 patients selected in the first step and the 66 patients selected in the third step. Among these 123 low-risk patients (58% of the original 212), none developed coronary artery abnormalities. The 62 patients selected in the second step and the 27 patients selected in the third step constituted the high-risk group. Coronary artery abnormalities developed in 12 (13.5%) of these 89 patients. Finally, we divided children in the high-risk group into males and females. Among males classified as high-risk, 9 of 53 (17.0%) developed coronary artery abnormalities; among high-risk females, 3 of 36 (8.3%) developed such abnormalities. The sensitivity and specificity for this classification
Despite early treatment with high-dose IVGG, some children with Kawasaki disease develop progressive abnormalities of the coronary arteries. Separation of patients who will develop aneurysms from those who will remain unaffected early in the course of the disease would be useful in patient management. Using the database of the Multicenter Kawasaki Study Group, we developed an instrument to predict the development of coronary artery lesions among patients treated within the first 10 days of illness with high-dose IVGG. Models based on fever and baseline laboratory data were developed and then applied prospectively using a sequential classification approach. In the combined data sets, no patient classified as low risk developed coronary artery abnormalities. The frequency of development of coronary abnormalities in males classified as high risk was 13.8%, and in highrisk females, 5.5%. Several scoring systems have been developed previously to identify children at highest risk for formation of coronary artery abnormalities. The first such scoring system was reported by Asai in 198310 and was useful at a time in Japan before 2-dimensional echocardiography was widely used for routine monitoring of all patients with Kawasaki disease. Risk factors included male sex, age ,1 year, fever for at least 14 days or recrudescent fever, hemoglobin concentration ,10 g/dl, white blood cell count .30,000 3 1026/L, erythrocyte sedimentation rate .101 mm/hour (Westergren), persistence of elevation
1118 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 81
MAY 1, 1998
TABLE I Instrument Performance on Development and Test Data Sets
Classified low risk Classified high risk CAA among high-risk females CAA among high-risk males Sensitivity Specificity
Development Data Set (n 5 212)
Test Data Set 1 (n 5 192)
Test Data Set 2 (n 5 264)
Test Data Set 3 (n 5 92)
123 (58%) 89 (42%) 3/36 (8.3%) 9/53 (17.0%) 100% 62%
99 (52%) 93 (48%) 2/33 (6.1%) 8/60 (13.3%) 100% 54%
169 (64%) 95 (36%) 1/44 (2.3%) 6/51 (11.8%) 100% 66%
53 (58%) 39 (42%) 1/15 (6.7%) 3/24 (12.5%) 100% 50%
CAA 5 coronary artery aneurysms.
of C-reactive protein or sedimentation abnormality for .30 days, and various cardiac findings consistent with ischemia or myocardial infarction. Duration of fever, presumably reflecting the severity of ongoing vasculitis, has been confirmed as a powerful predictor of coronary artery aneurysms in other studies14 –16 and indeed was the only factor that discriminated between those with and those without aneurysms in a study by Koren et al.14 However, duration of fever cannot be used for risk stratification in the era of almost universal acceptance of early IVGG therapy. The scoring systems by Nakano et al9 and Iwasa et al12 have not been verified using multicenter data. These scoring systems also require repeated laboratory tests over several days before treatment. The Harada risk index11 is used by some Japanese physicians to decide whether IVGG therapy should be administered; risk factors include high white blood cell count, low platelet count, low hematocrit, high C-reactive protein levels, low albumin levels, age ,12 months, and male sex. Age .5 years recently has been found to be associated with a greater incidence of coronary abnormalities.17 In our analysis, lower hemoglobin at presentation was a strong predictor of coronary artery outcome, perhaps indirectly reflecting severity of illness. The anemia of Kawasaki disease is poorly understood. The mean hematocrit of affected children is 2 SDs below the expected mean for age at the time of acute presentation, before the effects of decreased reticulocytosis should be clinically evident. The anemia is normochromic and normocytic, without direct evidence of hemolysis.18 These findings suggest that suppression of bone marrow production of red blood cells may begin well before clinical expression of Kawasaki disease. A similar pattern of anemia has been found with parvovirus and in other acute viral diseases.19 Additional laboratory parameters used in our predictive instrument included a left shift in the white blood cell count and a low platelet count; as noted above, both have been found to be risk factors in previous analyses in aspirin-treated patients. Fever and laboratory parameters of anemia, left shift, and low platelet count secondary to increased platelet turnover20 presumably reflect a greater release of inflammatory mediators and worse vascular inflammation.21–24 Inferences from this data set must be qualified for several reasons. Our instrument was constructed using
data accrued in patients treated with a total IVGG dose of either 1.6 or 2.0 g. Therefore, its performance may not be reliable in patients treated with lower doses of IVGG or with aspirin alone. Similarly, the instrument may be inaccurate in patients who present after the tenth day of illness or who have atypical disease. Temperature data were missing for many patients in the development data set and test data set 1, whereas the data for those in test data sets 2 and 3 were more complete. Febrile patients in whom temperature data were not recorded on the day after IVGG treatment might have been misclassified into the low-risk group. This would cause overestimation of the specificity of the instrument in the development data set and test data set 1 but would not affect sensitivity, as none of these patients developed coronary abnormalities. We used the largest high-quality database currently available: enrollment was based on strictly defined criteria, laboratory data were measured in a standardized fashion, and echocardiograms were interpreted by at least 2 blinded pediatric echocardiographers.
APPENDIX The following were members of the US Multicenter Kawasaki Disease Study Group (in alphabetical order by city and study center): Boston: Department of Pediatrics and Cardiology, Children’s Hospital and Harvard Medical School (Steven D. Colan, MD); Department of Epidemiology and Biostatistics, Boston University School of Public Health (Theodore Colton, ScD); and New England Medical Center and Tufts University School of Medicine, Boston (David R. Fulton, MD, H. Cody Meissner, MD); Chicago: Department of Pediatrics, Children’s Memorial Hospital and Northwestern University School of Medicine (C. Elise Duffy, MD, Anne H. Rowley, MD, Stanford T. Shulman, MD); Denver: Children’s Hospital, University of Colorado Health Sciences Center, and University of Colorado School of Medicine (Mary P. Glode, MD, James W. Wiggins, MD); Hawaii: Kapiolani Children’s Medical Center and John A. Burns School of Medicine (Marian E. Melish, MD, Venudhar Reddy, MD); Los Angeles: Children’s Hospital of Los Angeles and the University of Southern California, School of Medicine (Wilbert H. Mason, MD); San Diego: Children’s Hospital, the University of California Medical Center, and the University of California School of Medicine (John Bastian, MD, Jane C. Burns, MD, Kyung Ja Chung, MD).
1. Kato H, Ichinose E, Yoshioka F, Takechi T, Matsunaga S, Suzuki K, Rikitake
N. Fate of coronary aneurysms in Kawasaki disease: serial coronary angiography and long-term follow-up study. Am J Cardiol 1982;49:1758 –1766. 2. Kato H, Ichinose E, Kawasaki T. Myocardial infarction in Kawasaki disease: clinical analyses in 195 cases. J Pediatr 1986;108:923–927. 3. Suzuki A, Kamiya T, Kuwahara N, Ono Y, Kohata T, Kimura K, Takamiya M. Coronary arterial lesions of Kawasaki disease: cardiac catheterization findings of 1100 cases. Pediatr Cardiol 1986;7:3–9. 4. Furusho K, Kamiya T, Nakano H, Kiyosawa N, Shinomiya K, Hayashidera T, Tamura T, Hirose O, Manabe Y, Yokoyama T. Japanese gamma globulin trials for Kawasaki disease. Prog Clin Biol Res 1987;250:425– 432. 5. Nagashima M, Matsushima M, Matsuoka H, Ogawa A, Okumura N. High-dose gammaglobulin therapy for Kawasaki disease. J Pediatr 1987;110:710 –712. 6. Newburger JW, Takahashi M, Burns JC, Beiser AS, Duffy CE, Glode MP,
CORONARY ARTERY DISEASE/PREDICTING ANEURYSMS IN KAWASAKI DISEASE
1119
Mason WH, Reddy V, Sanders SP. The treatment of Kawasaki syndrome with intravenous gamma globulin. N Engl J Med 1986;315:341–347. 7. Furusho K, Kamiya T, Nakano H, Kiyosawa N, Shinomiya K, Hayashidera T, Tamura T, Hirose O, Manabe Y, Yokoyama T. High-dose intravenous gammaglobulin for Kawasaki disease. Lancet 1984;2:1055–1058. 8. Dajani AS, Taubert KA, Takahashi M, Bierman FZ, Freed MD, Ferrieri P, Gerber M, Shulman ST, Karchmer AW, Wilson W, Peter G, Durack DT, Rahimtoola SH. Guidelines for long-term management of patients with Kawasaki disease. Report from the Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, American Heart Association. Circulation 1994;89:916 –922. 9. Nakano H, Ueda K, Saito A, Tsuchitani Y, Kawamori J, Yoshida T. Scoring method for identifying patients with Kawasaki disease at high risk of coronary artery aneurysms. Am J Cardiol 1986;58:739 –742. 10. Asai T. Evaluation method for the degree of seriousness in Kawasaki disease. Acta Paediatr Jpn 1983;25:170 –175. 11. Harada K, Yamaguchi H, Kato H, Nishibayashi Y, Seki I, Okazaki T, Sato Y, Furusho K, Okawa S, Kawasaki T. Indication for intravenous gamma globulin for Kawasaki disease. In: Takahashi M, Taubert K, eds. Proceedings of the Fourth International Symposium on Kawasaki Disease. Dallas: American Heart Association, 1993:459 – 462. 12. Iwasa M, Sugiyama K, Ando T, Nomura H, Katoh T, Wada Y. Selection of high-risk children for immunoglobulin therapy in Kawasaki disease. Prog Clin Biol Res 1987;250:543–544. 13. Newburger JW, Takahashi M, Beiser AS, Burns JC, Bastian J, Chung KJ, Colan SD, Duffy CE, Fulton DR, Glode MP, et al. A single intravenous infusion of gamma globulin as compared with four infusions in the treatment of acute Kawasaki syndrome. N Engl J Med 1991;324:1633–1639. 14. Koren G, Lavi S, Rose V, Rowe R. Kawasaki disease: review of risk factors for coronary aneurysms. J Pediatr 1986;108:388 –392.
1120 THE AMERICAN JOURNAL OF CARDIOLOGYT
VOL. 81
15. Ichida F, Fatica NS, Engle MA, OLoughlin JE, Klein AA, Snyder MS,
Ehlers KH, Levin AR. Coronary artery involvement in Kawasaki syndrome in Manhattan, New York: risk factors and role of aspirin. Pediatrics 1987;80: 828 – 835. 16. Daniels SR, Specker B, Capannari TE, Schwartz DC, Kaplan S. Correlates of coronary artery aneurysm formation in patients with Kawasaki disease. Am J Dis Child 1987;141:205–207. 17. Yanagawa H, Yashiro M, Nakamura Y, Kawasaki T, Kato H. Epidemiologic pictures of Kawasaki disease in Japan: from the nationwide incidence survey in 1991 and 1992. Pediatrics 1995;95:475– 479. 18. Melish ME. Kawasaki syndrome: a 1986 perspective. Rheum Dis Clin North Am 1987;13:7–17. 19. Jansson LT, Kling S, Dallman PR. Anemia in children with acute infections seen in a primary care pediatric outpatient clinic. Pediatr Infect Dis 1986;5:424 – 427. 20. Burns JC, Glode MP, Clarke SH, Wiggins J Jr. Coagulopathy and platelet activation in Kawasaki syndrome: identification of patients at high risk for development of artery aneurysms. J Pediatr 1984;105:206 –211. 21. Lang BA, Silverman ED, Laxer RM, Rose V, Nelson DI, Rubin LA. Serum-soluble interleukin-2 receptor levels in Kawasaki disease. J Pediatr 1990; 116:592–596. 22. Lin CY, Lin CC, Hwang B, Chiang BN. The changes of interleukin-2, tumor necrotic factor and gamma-interferon production among patients with Kawasaki disease. Eur J Pediatr 1991;150:179 –182. 23. Leung DY. New developments in Kawasaki disease. Curr Opin Rheumatol 1991;3:46 –55. 24. Lin CY, Lin CC, Hwang B, Chiang B. Serial changes of serum interleukin-6, interleukin-8, and tumor necrosis factor alpha among patients with Kawasaki disease. J Pediatr 1992;121:924 –926.
MAY 1, 1998