Establishment and validation of a nomogram for predicting intravenous immunoglobulin resistance and coronary artery lesions involvement in Kawasaki disease: A retrospective study

doi:10.21203/rs.3.rs-4617455/v1

Download PDF

Article

Establishment and validation of a nomogram for predicting intravenous immunoglobulin resistance and coronary artery lesions involvement in Kawasaki disease: A retrospective study

https://doi.org/10.21203/rs.3.rs-4617455/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Objective

We aimed to develop a useful nomogram for early identification of Kawasaki disease (KD) children at a high risk of intravenous immunoglobulin (IVIG) resistance and coronary artery lesions (CAL) complications to improve KD management. Methods: Clinical data from 400 patients treated at our hospital between January 1, 2016, and December 31, 2023, were collected. Lasso regression was utilized to screen risk factors for IVIG resistance and CAL involvement. Subsequently, a Logistic regression model incorporating parameters screened by Lasso regression was established and visualized as a nomogram. The discrimination, calibration, clinical applicability, and universality of the model were evaluated using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and internal validation.

Results

NEU%, HDL-C and MHR were identified as predictors of IVIG resistance by Lasso regression, with C-index of the Logistic model being 0.886 for the training set and 0.855 for the validation set. For predicting CAL development, sex, fever date before the first IVIG administration, KD type, the level of HDL-C and MHR were the optimal variables, yielding C-index of 0.915 and 0.866 for the training and validation set, respectively. Calibration curves for both validation sets performed well, indicating strong predictive abilities of the models.

Conclusions

We established two nomograms for predicting IVIG resistance and CAL complications in KD patients, based on the Lasso-Logistic regression model. These nomograms were of guiding significance for screening KD children at high risk of developing IVIG resistance and CAL complications, thereby improving prognosis.

Health sciences/Diseases/Cardiovascular diseases

Health sciences/Diseases/Immunological disorders

Health sciences/Health care/Paediatrics

Lasso regression

nomogram

Kawasaki disease

Intravenous immunoglobulin resistance

Coronary artery lesions

prediction model

Kawasaki disease (KD), also known as mucocutaneous lymph node syndrome, is an acute febrile vasculitis that primarily affects children under the age of five. This multisystem illness shows a predilection for medium and small-sized arteries, particularly the coronary arteries. KD has been considered as the leading cause of acquired cardiac disease in children in developed countries ¹. Appropriate treatment including intravenous immunoglobulin (IVIG) and aspirin, has significantly reduced the incidence of coronary artery lesions (CAL) ². Nevertheless, up to 20% of KD patients fail to respond to the initial standard dose of IVIG treatment, termed IVIG resistance ³, and are at a high risk of CAL and poor prognosis, necessitating co-treatment with glucocorticoids or other immunosuppressants ⁴. Therefore, it is essential to develop a useful predictive score model to identify KD children at high risk of IVIG resistance and CAL complications early for more effectively management of the disease.

To date, numerous studies have introduced risk scoring systems combining various demographic and clinical parameters ^5–7. Among these, Kobayashi ⁸, Egami ⁹ and Sano ¹⁰ scores are most frequently employed to predict IVIG resistance. Nonetheless, the efficacy of these established prediction models varies significantly in clinical practice across different countries or geographical population. Additionally, the majority of data evaluation methods rely on univariate and multivariate analyses, which have inherent limitations in addressing multicollinearity among variables. The least absolute shrinkage and selection operator (Lasso) regression offers a solution by constructing penalty functions to screen optimal predictive factors, thereby developing a more precise model. This method minimizes the correlation among variables and prevents overfitting in the resulting nomograms ¹¹. In this study, we combined Lasso regression and Logistic regression: the former effectively screens variables, while the latter enables modeling and dynamic visualization for more intuitive interpretation.

Therefore, the aim of this study is to establish and validate a new predictive model for IVIG resistance and CAL involvement based on Lasso-Logistic regression method, thereby enhancing clinical assessment and treatment in KD children through a retrospective study. In addition, we compared this new predictive model with previously reported models.

Patients enrolled

The study included a total of 400 patients who attended Qilu Hospital of Shandong University from January 2016 to December 2023 and were diagnosed with complete or incomplete KD according to the AHA guidance published in 2017 ¹². Patients were excluded if they met any of the following criteria: (1) received glucocorticoids or other immunosuppressants during or before the initial IVIG treatment; (2) did not receive IVIG treatment at Qilu Hospital of Shandong University; (3) had incomplete clinical data. The patients were randomly assigned to a training set (N1 = 200) and a validation set (N2 = 200) in a 1:1 ratio in order to construct a more stable and reliable model. The characteristics of the patients between the two cohorts were contrasted. This study was approved by the Ethics Committee of Qilu Hospital of Shandong University, and all methods were carried out in accordance with relevant guidelines and regulations. All clinical data were obtained from the electronic medical records. Because the data was retrospective, a waiver of the informed consent requirement was granted by the Ethics Committee of Qilu Hospital of Shandong University.

All patients received the initial standard treatment, consisting of IVIG (2g/kg, administered as a single intravenous infusion over 10–12 hours) and aspirin (30–50 mg/kg/day, given orally three times a day) ¹². After 48–72 hours of fever resolution, the aspirin dosage was reduced to 3–5 mg/kg/day and maintained depended on the degree of coronary artery dilatation ³. IVIG resistance was defined as the recurrence or persistence of fever at least 36 hours after the completion of IVIG infusion, and CAL was defined as Z-score ≥ 2 ¹².

Data collection

Demographic and clinical data prior to the initial IVIG treatment were collected. There were 35 variables: age, sex, fever date before the first IVIG administration (day), clinical type (complete KD or incomplete KD), white blood count (WBC), neutrophil count (NEU), neutrophil percentage (NEU%), lymphocyte count (LYM), lymphocyte percentage (LYM%), monocyte count (MON), monocyte percentage (MON%), eosinophil count (EOS), eosinophil percentage (EOS%), basophil count (BAS), basophil percentage (BAS%), hemoglobin (HB), mean corpusular volume (MCV), hematocrit (HCT), mean platelet volume (MPV), platelet count (PLT), C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), neutrophil to lymphocyte ratio (NLR), platelet count to lymphocyte ratio (PLR), lymphocyte to monocyte ratio (LMR), alanine aminotransferase (ALT), aspartate aminotransferase (AST), Gamma-glutamyl transferase (GGT), total bilirubin (TB), albumin (ALB), CRP to ALB ratio (CAR), high-density lipoprotein cholesterol (HDL-C), MON to HDL-C ratio (MHR), lactate dehydrogenase (LDH), LDH to ALB ratio (LAR). Coronary diameter and Z-score were obtained from cardiac ultrasound at the acute stage of KD (usually no more than 10 days).

Statistical analysis

All independent variables were compared between the training and validation sets. The training set was used to build model, whereas the validation set was used to evaluate the effectiveness of the trained model. Categorical variables were reported as frequency and percentage and compared using chi-square tests. Mann–Whitney U-test was employed to perform the difference comparison of continuous variables, which were expressed as median and Inter-Quartile Range (IQR). Two-tailed p-values less than 0.05 were considered statistically significant. Lasso regression was used to select the best predictive risk factors for IVIG resistance and CAL development, respectively. Two risk prediction models, incorporating characteristics screened by Lasso regression, were established based on Logistic regression. Then, optimal predictors were selected to construct nomograms for IVIG resistance and CAL involvement. The performance of the models was evaluated according to their C-index by calculating the area under the curve (AUC) of the receiver operating characteristic (ROC) curve and the calibration curve using a 1000-bootstrap resampling procedure, reflecting the discrimination and consistency of the models, respectively. Decision curve analysis (DCA) was performed to assess the clinical applicability by examining the net benefits at varying threshold probabilities. Finally, the performance of nomogram on the validation set was completed to assess generalization. All statistical analyses were performed with R software v4.3.2.

Comparison of baseline data

A total of 400 patients were randomly divided into training set and validation set in a ratio of 1:1. There was no significant difference in all variables between the two groups

(p > 0.05), indicating the random and reasonable nature of the data grouping (Table 1).

Table 1. Baseline characteristics in training set and validation set.

Variables	Training set (N1 = 200)	Validation set (N2 = 200)	P
CAL involvement, n (%)
no	116 (58)	115 (57.5)	1
yes	84 (42)	85 (42.5)
IVIG resistance, n (%)
no	174 (87)	177 (88)	0.76
yes	26 (13)	23 (12)
sex, n (%)
female	72 (36)	73 (36)	1
male	128 (64)	127 (64)
Age (years), median [IQR]	2.29 [1.25, 4.17]	2.25 [1.08, 3.92]	0.751
Fever date, median [IQR]	6 [5, 8]	6 [5, 7]	0.337
type, n [%]
Incomplete KD	96 [48]	94 [47]	0.92
Complete KD	104 [52]	106 [53]
WBC, median [IQR]	13.11 [8.79, 17.98]	12.99 [9.25, 16.67]	0.741
NEU, median [IQR]	7.74 [4.54, 11.95]	7.65 [4.78, 11.24]	0.775
NEU%, median [IQR]	62.15 [46.3, 75.05]	61.3 [48.1, 74.8]	0.98
LYM, median [IQR]	3.23 [2.27, 4.7]	3.41 [2.23, 4.59]	0.872
LYM%, median [IQR]	27.3 [17, 40.15]	27.4 [18.28, 38.95]	0.903
MON, median [IQR]	0.84 [0.58, 1.23]	0.88 [0.64, 1.27]	0.488
MON%, median [IQR]	6.8 [4.9, 8.72]	6.95 [5, 9]	0.381
EOS, median [IQR]	0.21 [0.03, 0.48]	0.2 [0.06, 0.46]	0.582
EOS%, median [IQR]	1.7 [0.2, 4.1]	1.65 [0.48, 4]	0.566
BAS, median [IQR]	0.02 [0.01, 0.04]	0.02 [0.01, 0.04]	0.935
BSA%, median [IQR]	0.2 [0.1, 0.3]	0.2 [0.1, 0.3]	0.8
HB, median [IQR]	113.5 [106.75, 121]	113 [103.75, 118]	0.155
MCV, median [IQR]	82.25 [79.3, 85]	81.9 [78.88, 84.43]	0.386
HCT, median [IQR]	34.05 [30.8, 36.4]	33.2 [30.67, 35.42]	0.248
MPV, median [IQR]	8.7 [8.1, 9.6]	8.9 [8.1, 9.7]	0.279
PLT, median [IQR]	400 [300.25, 513.5]	396.5 [318.75, 497.75]	0.81
CRP, median [IQR]	44.28 [20.43, 84.9]	42.74 [22.14, 88.59]	0.41
ESR, median [IQR]	72 [49, 91]	79 [55.75, 98]	0.096
NLR, median [IQR]	2.33 [1.14, 4.35]	2.21 [1.25, 4.12]	0.93
PLR, median [IQR]	117.93 [86.49, 189.7]	123.1 [85.7, 170.77]	0.744
CAR, median [IQR]	1.23 [0.57, 2.33]	1.22 [0.59, 2.45]	0.351
LMR, median [IQR]	4.15 [2.63, 5.54]	3.97 [2.76, 5.24]	0.331
ALT, median [IQR]	20 [11, 42.25]	18 [11, 45.25]	0.911
AST, median [IQR]	23 [20, 29.25]	25 [20, 33]	0.126
GGT, median [IQR]	22 [13, 57]	19 [11.75, 65.5]	0.669
TB, median [IQR]	4.25 [3.1, 6.03]	4.7 [3.3, 7.53]	0.079
ALB, median [IQR]	36.7 [33.08, 39.82]	36.05 [33.3, 39.6]	0.616
HDL-C, median [IQR]	0.82 [0.57, 1.23]	0.83 [0.62, 1.23]	0.587
MHR, median [IQR]	1.01 [0.69, 1.43]	0.98 [0.74, 1.41]	0.957
LDH, median [IQR]	282.5 [241.75, 337.25]	291.5 [253.75, 346.25]	0.378
LAR, median [IQR]	7.78 [6.61, 9.36]	8.07 [6.71, 9.71]	0.289

Screening of characteristic factors for risk of IVIG resistance

Lasso regression was conducted to screen parameters. The coefficient variation features of these variables are shown in Fig. 1a and the Lasso model of parameters iterative analysis using 20-fold cross-validation method is depicted in Fig. 1b. The results showed that the 35 variables were reduced to 22 (lambda with minimum mean square error = 0.004) and then to 4 (lambda with one standard error of the minimum distance = 0.042). The 4 variables screened when lambda was 0.042 included NEU%, MON%, HDL-C and MHR. The 4 characteristic factors were used to construct the Logistic regression model in training set, and ultimately, only NEU%, HDL-C and MHR were determined as significant predictors (p < 0.05) of IVIG resistance for the final model, as shown in Table 2.

Development of the predictive nomogram of IVIG resistance

For more convenient clinical application, we constructed a nomogram for predicting IVIG resistance in KD using the screened three optimal predictor factors, as illustrated in Fig. 1c. In this nomogram, the varying values for the three variables correspond to their respective scores, and the total score is calculated by adding up the individual scores of the three variables. The bottom of Fig. 1c provides the prediction probabilities of IVIG resistance corresponding to different total scores. The higher the total score, the greater the probability of occurring IVIG resistance. For example, if a KD child has clinical data showing NEU% at 70, HDL-C at 0.8 mmol/L and MHR at 3, the corresponding score of each variable would be 25, 75 and 30, respectively, resulting in a total score of 130. Thus, the child would have approximately a 50% probability of experiencing IVIG resistance.

Table 2. Multivariate logistic regression analysis for IVIG resistance.

Variable	R	SE	Z	Pr(>\|Z\|)
NEU%	0.0404	0.0193	2.09	0.0363
MON%	-0.0804	0.1489	-0.54	0.5894
HDL-C	-4.2897	1.3819	-3.1	0.0019
MHR	1.2088	0.5899	2.05	0.0405
Intercept	-2.5077	1.7048	-1.47	0.1413

R, regression coefficient; SE, Standard error.

Performance and internal validation of the nomogram

The C-index of the nomogram in the training set and validation set was 0.886 and 0.855, respectively. The sensitivity and specificity of the training set were 0.846 and 0.770, respectively (Fig. 2a). The DCA showed exceedingly better net benefits in the predictive model (Fig. 2b). Furthermore, the calibration curves for both sets displayed bias-correction and an apparent curve that closely resembled the ideal line (Fig. 2c and Fig. 2d), indicating that the observed and predicted levels of IVIG resistance were in good agreement. These data indicate that our nomogram for predicting IVIG resistance in KD demonstrated good discrimination, calibration, clinical applicability, and generalization, making it a valuable tool for clinical decision-making.

Screening of characteristic factors for risk of CAL involvement

We used Lasso regression to screen the parameters. Fig. 3a displayed the coefficient variation features of these variables, and Fig. 3b displayed the Lasso model of parameters iterative analysis using the 20-fold cross-validation method. The total number of variables were reduced from 35 to 16 (lambda with minimum mean square error = 0.008) and then to 6 (lambda with one standard error of the minimum distance = 0.040). The 6 variables obtained with a lambda of 0.040 included sex, fever date before the first IVIG administration, clinical type, HCT, HDL-C and MHR. Excluding HCT due to p>0.05, we used the remaining 5 characteristic factors to construct the Logistic regression model in training set, as shown in Table 3.

Development of the predictive nomogram of CAL involvement

By the screened 5 optimal predictors, we conducted a nomogram for predicting CAL involvement in KD for more convenient clinical practice, as illustrated in Fig. 3c. Each option of each variable corresponds to a specific point, and the total point is obtained by summing the points of all 5 variables. Different predictive probabilities corresponding to various total points are given at the bottom of the figure. For each child, higher total points indicate a higher risk of developing CAL.

Table 3. Multivariate logistic regression analysis for CAL involvement.

variable	R	SE	Z	Pr(>\|Z\|)
sex	1.0268	0.4809	2.14	0.0327
fever date	0.21	0.0747	2.84	0.0045
type	-1.2942	0.4482	-2.89	0.0039
HCT	-0.1068	0.0549	-1.94	0.0518
HDL-C	-2.0128	0.566	-3.56	0.0004
MHR	3.1828	0.5705	5.58	<0.0001
Intercept	0.0229	2.0878	0.01	0.9912

Performance and internal validation of the nomogram

The C-indexes of the nomogram were 0.915 and 0.866 for the training set and validation set, respectively. In the training set, the sensitivity and specificity were 0.786 and 0.931, respectively (Fig. 4a). In the predicted model, the DCA demonstrated greater net benefits (Fig. 4b). The observed and expected levels of CAL involvement also coincided well, as shown by the calibration curves in both sets (Fig. 4c and Fig. 4d). Above data indicate that our nomogram has substantial potential for clinical decision-making.

Comparison with existed models

We assessed the performance of three existed risk scores in predicting IVIG resistance (Table 4A) and CAL involvement (Table 4B) for each patient in validation set, respectively. The numbers of children in the high-risk and low-risk categories, along with the sensitivity, specificity, positive predictive value (PPV), negative predictive value (PPV), AUC and 95% confidence interval (95%CI) of each model were calculated. Among IVIG resistance scoring systems, the Sano scoring system ¹⁰ had the highest sensitivity, but its specificity was extremely low. The Egami ⁹ and Wu ⁷ scoring systems exhibited relatively higher specificity but lower sensitivity. The AUCs of these three scoring systems for predicting IVIG-resistance were all low, ranging from 0.482 to 0.627. For CAL involvement scoring systems ^6,9,13, the sensitivity was below 0.5 for all, with the Egami score ⁹ having the highest specificity, while the specificity of the other two scores was comparable. The AUCs of the three scoring systems were similar, ranging from 0.508 to 0.561. These results above indicate that the existing scoring systems showed poor performance in our population.

Table 4a Prediction score models in patients with IVIG resistance

Prediction models	Category	Response to IVIG		Sensitivity	Specificity	PPV	NPV	AUC	95%CI
		Resistant	Responsive
Sano score	High risk	21	168	0.913	0.051	0.111	0.818	0.482	0.4209-0.543
	Low risk	2	9
Egami score	High risk	10	32	0.435	0.819	0.238	0.918	0.627	0.5196-0.7344
	Low risk	13	145
Wu score	High risk	10	43	0.435	0.757	0.189	0.912	0.596	0.4876-0.7042
	Low risk	13	134

Table 4b Prediction score models in patients with CAL involvement

Prediction models	Category	CAL involvement		Sensitivity	Specificity	PPV	NPV	AUC	95%CI
		CAL+	CAL-
Tawai score	High risk	28	36	0.329	0.687	0.438	0.581	0.508	0.4423-0.574
	Low risk	57	79
Egami score	High risk	19	23	0.224	0.8	0.452	0.582	0.512	0.454-0.5695
	Low risk	66	92
Zhe jiang score	High risk	40	40	0.471	0.652	0.5	0.625	0.561	0.4924-0.6304
	Low risk	45	75

For patients with KD, we developed an intuitive prediction model for IVIG resistance and CAL involvement. In previous studies, many researchers have developed predictive models of IVIG resistance in KD patients, employing techniques such as Logistic regression, Lasso regression, Random Forest, Extreme Gradient Boosting, Support Vector Machine, and Decision Tree models ^14–17. Many of them have demonstrated good predictive performance in their respective populations. However, due to the opaque nature of machine learning algorithms, interpreting and understanding the implications of these models can be challenging ¹⁸, limiting their applicability in healthcare. Consequently, the nomogram technique has emerged as a valuable tool for visually and interpretable understanding of predictive results, leading to enhanced accuracy ¹⁹. While several nomograms for IVIG resistance in KD have been developed ^14,20, their application in predicting CAL remains relatively limited. In this study, we provided a precise approach by constructing two nomograms to predict the risk of IVIG resistance and CAL development in KD patients, respectively, which makes a contribution to identify the children at a risk of IVIG resistance or CAL to improve prognosis. Furthermore, our internal validation has demonstrated that the nomograms we created exhibit good calibration and discriminating power. Specifically, the internal validation yielded a high C-index value, indicating that the constructed nomograms can be widely and accurately utilized in larger sample sizes.

In our population, 12.25% (49) of KD patients experienced IVIG resistance, consistent with reported findings from China in recent years ranging from 5.1–15.4% ^21–23. Interestingly, we observed a higher incidence of developing CAL, with 42.25% (169) of patients affected, compared to previous statistics with 10.2–21.3% ^23,24. This study identified NEU%, HDL-C and MHR as optimal predictors for KD patients occurring IVIG resistance, necessitating the second IVIG infusion or additional treatment, such as glucocorticoids and other immunosuppressants. According to our nomogram, boy, longer fever date before the first IVIG administration, incomplete KD diagnosis, lower HDL-C level, and higher MHR level were identified as the critical risk factors for the development of CAL in KD cases.

Many studies have integrated neutrophil-related parameters into predictors for determining IVIG response and vascular damage in KD, including NEU, NEU% and NLR ^20,25,26. During the acute phase of KD, neutrophils, as the critical mediators of host defense, undergo proliferation in response to infection and inflammation, leading to markedly increased production of oxygen intermediate, neutrophil elastase, and myeloperoxidase, which have the potential to cause tissue self-destruction ^27,28, thus reflecting the severity of disease. Monocytes also play a pivotal role in the primary innate immune response. During the acute phase, the absolute monocyte counts in the peripheral blood increases ²⁹, resulting in elevated serum concentrations of interleukin-1, interleukin-6 and tumor necrosis factor-α in inflammatory and immunological reactions ^30–32. The treatment of IVIG has been shown to be effective for KD patients by reducing peripheral blood monocytes, and the count of monocytes is an important parameter for assessing the severity of vascular damage ²⁹. HDL-C is a crucial lipid class, and dyslipidemia has been shown to be associated with systemic inflammation ^33–35. The concentration of HDL-C has been reported to be lower in KD patients experiencing IVIG resistance and CAL ²³, which is consistent with our findings. The level of HDL-C is positively related to the activity of lipoprotein lipase ³⁶, and endothelial cell dysfunction in KD may attribute to the diminished activity of lipoprotein lipase, resulting in the decreased level of HDL-C ^37,38. Low HDL-C concentration limits its anti-inflammatory effect ³⁹. MHR, the ratio of MON to HDL-C, has been reported to have a predictive value in cardiovascular diseases ^40–42. We observed that elevated MHR level had predictive potential in IVIG resistance and CAL involvement, and its participation contributed to satisfactory predictive performance.

In comparison of groups with or without CAL development, we found that a longer fever duration before the first IVIG infusion, male gender, and atypical clinical manifestations were risk predictors for CAL occurrence. Previous investigations have reported that fever duration ≥ 7 days ⁴³, longer illness days prior to initial IVIG therapy ⁴⁴, or delayed diagnosis contribute to the development of CAL, aligning with our findings. Many scholars have associated atypical clinical manifestations with CAL involvement ⁴⁵. Clinical diagnosis poses challenges in children with incomplete KD type as they do not present typical symptoms, and delayed treatment can elevate the risk of coronary artery injury.

Certain limitations of our study warrant consideration. Firstly, our investigation was a retrospective design, and data were collected from a single medical center, potentially introducing selection bias of the subjects and limiting the generalizability of our findings to all KD patients. Secondly, although our established nomogram demonstrated robust performance in internal validation, as confirmed by bootstrapping, external validation was not conducted. Therefore, further research involving larger sample sizes and diverse geographic regions is necessary to validate the generalizability of the nomogram.

The present study constructed two nomograms for predicting IVIG resistance and CAL development in KD patients based on the Lasso-Logistic regression model. These nomograms hold considerable guiding significance in early identification of KD children at a high risk of IVIG resistance and CAL complications, thereby facilitating more effectively management of KD.

Author contribution The study was designed by C. Z and Y. G. Data was collected by Y. G and J. L. Data analysis and interpretation was completed by Y. G and L. P. Y. G, L. P and J. L prepared the manuscript. C. Z revised the manuscript. All authors read and approved the final manuscript.

Acknowledgements This study was supported in part by the grants from the Shandong Province Science and Technology project (2019GSF108186).

Ethics approval The ethics committee of Qilu Hospital of Shandong University approved this research.

Consent to Participate Not necessary, in accordance with institutional ethic review board, due to retrospective study design.

Consent for publication Not applicable.

Conflict of interest The authors declare no competing interests.

Data availability The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Uehara, R. & Belay, E. D. Epidemiology of kawasaki disease in Asia, Europe, and the United States. Journal of epidemiology. 22, 79-85. https://doi.org/10.2188/jea.je20110131 (2012).
Furusho, K. et al. High-dose intravenous gammaglobulin for Kawasaki disease. The Lancet. 324, 1055-1058. https://doi.org/10.1016/s0140-6736(84)91504-6 (1984).
Fukazawa, R. et al. JCS/JSCS 2020 guideline on diagnosis and management of cardiovascular sequelae in Kawasaki disease. Circulation Journal. 84, 1348-1407. https://doi.org/10.1253/circj.CJ-19-1094 (2020).
Gorelik, M. et al. 2021 American College of Rheumatology/Vasculitis Foundation guideline for the management of Kawasaki disease. Arthritis care & research. 74, 538-548. https://doi.org/10.1002/art.42041 (2022).
Lin, M.-T. et al. Risk factors and derived formosa score for intravenous immunoglobulin unresponsiveness in Taiwanese children with Kawasaki disease. Journal of the Formosan Medical Association. 115, 350-355. https://doi.org/10.1016/j.jfma.2015.03.012 (2016).
Hua, W. et al. A new scoring system to predict Kawasaki disease with coronary artery lesions. Clinical rheumatology. 38, 1099-1107. https://doi.org/10.1007/s10067-018-4393-7 (2019).
Wu, S. et al. Prediction of intravenous immunoglobulin resistance in Kawasaki disease in children. World Journal of Pediatrics. 16, 607-613. https://doi.org/10.1007/s12519-020-00348-2 (2020).
Kobayashi, T. et al. Prediction of intravenous immunoglobulin unresponsiveness in patients with Kawasaki disease. Circulation. 113, 2606-2612. https://doi.org/10.1161/CIRCULATIONAHA.105.592865 (2006).
Egami, K. et al. Prediction of resistance to intravenous immunoglobulin treatment in patients with Kawasaki disease. The Journal of pediatrics. 149, 237-240. https://doi.org/10.1016/j.jpeds.2006.03.050 (2006).
Sano, T. et al. Prediction of non-responsiveness to standard high-dose gamma-globulin therapy in patients with acute Kawasaki disease before starting initial treatment. European journal of pediatrics. 166, 131-137. https://doi.org/10.1007/s00431-006-0223-z (2007).
Friedman, J., Hastie, T. & Tibshirani, R. Regularization paths for generalized linear models via coordinate descent. Journal of statistical software. 33, 1. (2010).
McCrindle, B. W. et al. Diagnosis, treatment, and long-term management of Kawasaki disease: a scientific statement for health professionals from the American Heart Association. circulation. 135, e927-e999. https://doi.org/10.1161/CIR.0000000000000484 (2017).
Chang, L.-S. et al. Neutrophil-to-lymphocyte ratio and scoring system for predicting coronary artery lesions of Kawasaki disease. BMC pediatrics. 20, 1-7. https://doi.org/10.1186/s12887-020-02285-5 (2020).
Huang, H. et al. Nomogram to predict risk of resistance to intravenous immunoglobulin in children hospitalized with Kawasaki disease in Eastern China. Annals of Medicine. 54, 442-453. https://doi.org/10.1080/07853890.2022.2031273 (2022).
Chen, X. et al. Risk factors and predictive models for intravenous immunoglobulin resistance in children with recurrent Kawasaki disease. Journal of Inflammation Research, 2877-2889. https://doi.org/10.2147/JIR.S360802 (2022).
Takeuchi, M. et al. Novel risk assessment tool for immunoglobulin resistance in Kawasaki disease: application using a random forest classifier. The Pediatric infectious disease journal. 36, 821-826. https://doi.org/10.1097/INF.0000000000001621 (2017).
Joung, J. et al. A decision tree model for predicting intravenous immunoglobulin resistance and coronary artery involvement in Kawasaki disease. BMC pediatrics. 22, 474. https://doi.org/10.1186/s12887-022-03533-6 (2022).
Stiglic, G. et al. Interpretability of machine learning‐based prediction models in healthcare. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery. 10, e1379. (2020).
Balachandran, V. P., Gonen, M., Smith, J. J. & DeMatteo, R. P. Nomograms in oncology: more than meets the eye. The lancet oncology. 16, e173-e180. https://doi.org/10.1016/S1470-2045(14)71116-7 (2015).
Wang, J., Huang, X. & Guo, D. Predictors and a novel predictive model for intravascular immunoglobulin resistance in Kawasaki disease. Italian Journal of Pediatrics. 49, 126. https://doi.org/10.1186/s13052-023-01531-7 (2023).
Tang, Y. et al. Prediction of intravenous immunoglobulin resistance in Kawasaki disease in an East China population. Clinical rheumatology. 35, 2771-2776. https://doi.org/10.1007/s10067-016-3370-2 (2016).
Yang, S., Song, R., Zhang, J., Li, X. & Li, C. Predictive tool for intravenous immunoglobulin resistance of Kawasaki disease in Beijing. Archives of disease in childhood. 104, 262-267. https://doi.org/10.1136/archdischild-2017-314512 (2019).
Shao, S. et al. Predictive value of serum lipid for intravenous immunoglobulin resistance and coronary artery lesion in Kawasaki disease. The Journal of Clinical Endocrinology & Metabolism. 106, e4210-e4220. https://doi.org/10.1210/clinem/dgab230 (2021).
Tang, Y. et al. Coronary artery aneurysm regression after Kawasaki disease and associated risk factors: a 3-year follow-up study in East China. Clinical rheumatology. 37, 1945-1951. https://doi.org/10.1007/s10067-018-3977-6 (2018).
Muto, T. White blood cell and neutrophil counts and response to intravenous immunoglobulin in Kawasaki disease. Glob. Pediat. Health. 6. https://doi.org/10.1177/2333794X19884826 (2019).
Kanai, T. et al. The combination of the neutrophil-to-lymphocyte and platelet-to-lymphocyte ratios as a novel predictor of intravenous immunoglobulin resistance in patients with Kawasaki disease: a multicenter study. Heart and Vessels. 35, 1463-1472. https://doi.org/10.1007/s00380-020-01622-z (2020).
Takeshita, S. et al. The role of bacterial lipopolysaccharide-bound neutrophils in the pathogenesis of Kawasaki disease. The Journal of infectious diseases. 179, 508-512. https://doi.org/10.1086/314600 (1999).
Niwa, Y. & Sohmiya, K. Enhanced neutrophilic functions in mucocutaneous lymph node syndrome, with special reference to the possible role of increased oxygen intermediate generation in the pathogenesis of coronary thromboarteritis. The Journal of pediatrics. 104, 56-60. https://doi.org/10.1016/s0022-3476(84)80589-2 (1984).
Furukawa, S., Matsubara, T. & Yabuta, K. Mononuclear cell subsets and coronary artery lesions in Kawasaki disease. Archives of disease in childhood. 67, 706-708. https://doi.org/10.1136/adc.67.6.706 (1992).
Leung, D. M. et al. Endothelial cell activation and high interleukin-1 secretion in the pathogenesis of acute Kawasaki disease. The Lancet. 334, 1298-1302. https://doi.org/10.1016/s0140-6736(89)91910-7 (1989).
Matsubara, T., Furukawa, S. & Yabuta, K. Serum levels of tumor necrosis factor, interleukin 2 receptor, and interferon-γ in Kawasaki disease involved coronary-artery lesions. Clinical immunology and immunopathology. 56, 29-36. https://doi.org/10.1016/0090-1229(90)90166-n (1990).
Furukawa, S. et al. Kawasaki disease differs from anaphylactoid purpura and measles with regard to tumour necrosis factor-α and interleukin 6 in serum. European journal of pediatrics. 151, 44-47. https://doi.org/10.1007/BF02073890 (1992).
Shenoy, C., Shenoy, M. M. & Rao, G. K. Dyslipidemia in dermatological disorders. North American journal of medical sciences. 7, 421. (2015).
Evereklioglu, C., Er, H., Türköz, Y. & Cekmen, M. Serum levels of TNF-α, sIL-2R, IL-6, and IL-8 are increased and associated with elevated lipid peroxidation in patients with Behçet's disease. Mediators of inflammation. 11, 87-93. https://doi.org/10.1080/09629350220131935 (2002).
Nepomuceno, R. et al. Dyslipidemia rather than type 2 diabetes mellitus or chronic periodontitis affects the systemic expression of pro-and anti-inflammatory genes. Mediators of Inflammation. 2017. https://doi.org/10.1155/2017/1491405 (2017).
Henderson, H. E. et al. Lipoprotein lipase activity is decreased in a large cohort of patients with coronary artery disease and is associated with changes in lipids and lipoproteins. Journal of lipid research. 40, 735-743. (1999).
Beigneux, A. P. et al. Glycosylphosphatidylinositol-anchored high-density lipoprotein-binding protein 1 plays a critical role in the lipolytic processing of chylomicrons. Cell metabolism. 5, 279-291. https://doi.org/10.1016/j.cmet.2007.02.002 (2007).
Routhu, S. K. et al. Assessment of endothelial dysfunction in acute and convalescent phases of Kawasaki disease using automated edge detection software: a preliminary study from North India. JCR: Journal of Clinical Rheumatology. 27, 143-149. https://doi.org/10.1097/RHU.0000000000001233 (2021).
Murphy, A., Chin-Dusting, J., Sviridov, D. & Woollard, K. J. The anti inflammatory effects of high density lipoproteins. Current medicinal chemistry. 16, 667-675. https://doi.org/10.2174/092986709787458425 (2009).
Zhang, D.-P. et al. An elevated monocyte-to-high-density lipoprotein–cholesterol ratio is associated with mortality in patients with coronary artery disease who have undergone PCI. Bioscience reports. 40, BSR20201108. https://doi.org/10.1042/BSR20201108 (2020).
Ganjali, S. et al. Monocyte‐to‐HDL‐cholesterol ratio as a prognostic marker in cardiovascular diseases. Journal of cellular physiology. 233, 9237-9246. https://doi.org/10.1002/jcp.27028 (2018).
Aydin, E., Ates, I., Fettah Arikan, M., Yilmaz, N. & Dede, F. The ratio of monocyte frequency to HDL cholesterol level as a predictor of asymptomatic organ damage in patients with primary hypertension. Hypertension Research. 40, 758-764. https://doi.org/10.1038/hr.2017.36 (2017).
Kim, J.-J. et al. Assessment of risk factors for Korean children with Kawasaki disease. Pediatric Cardiology. 33, 513-520. https://doi.org/10.1007/s00246-011-0143-1 (2012).
Miura, M. et al. Coronary risk factors in Kawasaki disease treated with additional gammaglobulin. Archives of disease in childhood. 89, 776-780. https://doi.org/10.1136/adc.2003.032748 (2004).
Wang, L., Zeng, X. & Chen, B. Clinical manifestations and risk factors of coronary artery lesions in children with Kawasaki disease. Medicine. 102, e34939. https://doi.org/10.1097/MD.0000000000034939 (2023).

No competing interests reported.

Download PDF

Editor invited by journal
27 Jun, 2024
Submission checks completed at journal
27 Jun, 2024
First submitted to journal
21 Jun, 2024

You are reading this latest preprint version

Establishment and validation of a nomogram for predicting intravenous immunoglobulin resistance and coronary artery lesions involvement in Kawasaki disease: A retrospective study

Status:

Version 1

Abstract

Objective

Results

Conclusions

Figures

Introduction

Materials and methods

Patients enrolled

Data collection

Statistical analysis

Results

Discussion

Conclusion

Declarations

References

Additional Declarations

Status:

Version 1