Z-score is a possible predictor of the risk of coronary artery lesion development in patients with Kawasaki disease in Japan

Risk factors for coronary artery lesion (CAL) development in patients with Kawasaki disease (KD) include male sex, age < 12 months, intravenous immunoglobulin (IVIG) resistance, and delayed diagnosis.. We aimed to explore the relationship between CAL development and Z-score. We enrolled 281 patients with KD who were treated with our protocol. Echocardiography was performed in three phases: pre-treatment (P1), post-treatment (P2), and 4 weeks after onset (P3). The highest Z-score of the right, left main, left anterior descending, and left circumflex coronary arteries was expressed as Zmax at each phase. P3-Zmax ≥ 2.5 represented CAL development. Clinical parameters, such as laboratory data and Z-scores, were retrospectively compared between patients with and without CAL development. Sixty-seven patients (23.8%) showed a P1-Zmax ≥ 2.0, and CAL development occurred in 21 patients (7.5%). Independent risk factors associated with CAL development were P1-Zmax, a ΔZmax (P2-Zmax − P1-Zmax) ≥ 1, male sex, < 12 months of age, and resistant to the first intravenous immunoglobulin (IVIG) administration (adjusted odds ratio [95% confidence interval]: 198 [1.01–3.92], 4.04 [1.11–14.7], 6.62 [1.33–33.04], 4.71 [1.51–14.68], 5.26 [1.62–17.13], respectively). Using receiver operating characteristic curve analysis, a P1-Zmax ≥ 1.43 detected CAL development with an area under the curve of 0.64 (sensitivity = 81.0%; specificity = 48.1%). Conclusion: Our results suggest that P1-Zmax and a ΔZmax (P2-Zmax − P1-Zmax) ≥ 1 may predict CAL development. What is Known: • KD is an acute vasculitis predominantly affecting the coronary artery of young children. • Although P1 Z-max ≥ 2.0 has been a predictor of CAL development, it has not yet been shown in Japan. What is New: • P1-Zmax and a ΔZmax ≥ 1 are presumably associated with CAL development. • In the ROC curve analysis, P1-Zmax ≥ 1.43 detected CAL development, a sensitivity (81%) and a specificity (48%). We need to consider intensified initial therapy for patients with these risk factors. What is Known: • KD is an acute vasculitis predominantly affecting the coronary artery of young children. • Although P1 Z-max ≥ 2.0 has been a predictor of CAL development, it has not yet been shown in Japan. What is New: • P1-Zmax and a ΔZmax ≥ 1 are presumably associated with CAL development. • In the ROC curve analysis, P1-Zmax ≥ 1.43 detected CAL development, a sensitivity (81%) and a specificity (48%). We need to consider intensified initial therapy for patients with these risk factors.


Introduction
Kawasaki disease (KD) is an acute systemic vasculitis occurring in infants and children, especially those younger than 5 years of age [1]. The most serious complication of KD is coronary arteritis, resulting in coronary arterial dilatation, aneurysm, and stenosis. Risk factors for coronary artery lesion (CAL) development in patients with Kawasaki disease (KD) include male sex, age < 12 months, intravenous immunoglobulin (IVIG) resistance, and delayed diagnosis [2][3][4][5][6]. Therefore, coronary artery lesion (CAL) is now the leading cause of acquired heart disease in children in developed countries.
As the etiology of KD is still unknown, radical treatments to prevent CAL development in patients with KD are not available. Recently, intensified initial standard steroid [7] and cyclosporin A (CsA) [8] therapy have contributed to a decrease in the incidence of CAL. However, the actual number of patients with CAL has not decreased because the incidence of KD continues to increase in Japan [9]. Therefore, it is vital to identify predictive markers of CAL development before treatment.
In recent times, Z-score has been used to evaluate CAL in patients with KD [10,11]. American Heart Association (AHA) guidelines [12] recommend intensified initial treatment for patients considered to be at a high risk for CAL. Several studies [13][14][15][16] have shown that CAL development may be predicted using pre-treatmentZ-score; however, this has not yet been shown in Japan.
In the present study, we aimed to predict CAL development using Z-scores in patients with early acute phase of KD to enable early intervention with intensified therapy, and we serially examined coronary artery Z-scores in patients with KD at different time points from admission (pre-treatment) to 1 month after onset. We analyzed how Z-scores in the early acute phase of KD may correlate with Z-scores 1 month after KD onset; that is, the coronary arterial sequalae of KD.

Study population
A total of 324 patients who fulfilled the diagnostic criteria for KD and who were admitted to our hospital to undergo our treatment protocol from March 2009 to March 2018 were enrolled. The diagnosis of KD was made in accordance with the criteria from the 5th Diagnostic Guidelines established by the Japan Kawasaki Disease Research Committee [17]. Day 1 of KD was defined as the first day of fever (axillar temperature ≥ 37.5°7). The study was approved by the ethics committee of Wakayama Medical University (No. 2801). Patients who fulfilled at least one of the following criteria were excluded: (1) no IVIG treatment or not following our treatment protocol; (2) recurrent KD; (3)follow-up echocardiographic measurements not available; (4) patients who might have an effect on Zscores, such as a high body mass index (≥ 25), a history of cardiovascular disease, systemic arterial hypertension, and other syndromes, such as Noonan/LEOPARD syndrome. All enrolled patients underwent our treatment protocol: IVIG (2 g/kg/day for 24 h) + ASA (30 mg/kg/day) (first-line therapy). If patients were resistant to IVIG after the first administration (persistent or recrudescent fever ((axillar temperature ≥ 37.5°C ) 24 h after the first IVIG administration), they were treated with additional IVIG (2 g/kg/day for 24 h) as a second-line therapy. In addition, patients who were resistant to the second IVIG administration (persistent or recrudescent fever (axillar temperature ≥ 37.5°C) after completion of IVIG) were treated with oral CsA (5 mg/kg/day) as a third-line therapy. The medical records of all enrolled patients were retrospectively reviewed, including demographics, clinical characteristics, medications, serial echocardiographic findings, and laboratory tests on admission, such as white blood cell count (WBC), percentage of neutrophils, hematocrit (Ht), platelet count (Plt), aspartate aminotransferase (AST), serum albumin (Alb), serum sodium (Na), and C-reactive protein (CRP). Laboratory data were converted into categorical variables based on the previous scoring system for KD [18][19][20][21].
Evaluation of coronary artery Z-score Echocardiography was performed using the Philips EPIQ7 or iE33 echocardiographic machine (Philips, Eindhoven, Netherlands) and a sector probe with a frequency of 8 MHz. Three pediatric cardiologists measured the inner diameter of the right coronary artery (RCA), left main coronary artery (LMCA), left anterior descending artery (LAD), and left circumflex artery (LCX), and the measurements were performed in three phases: pre-treatment (P1; at the time of KD diagnosis or before IVIG treatment), post-treatment (P2; after first IVIG treatment and around 10 days from KD onset), and the convalescent phase (P3; around 4 weeks from KD onset). We measured internal coronary artery diameter twice at the end of the T wave as the maximum diameter [22] and calculated an average value. Z-scores were retrospectively calculated for each four coronary artery dimensions using the Z-score calculator (Version 4.0 Full, LMS_Z_Score) [11,23]. Body surface area (BSA) was calculated using the Haycock formula. As the four coronary artery segments need to be evaluated in patients with KD, the highest Z-score of these four segments was expressed as Zmax in each phase. As multiple echocardiograms were performed in the P2 phase, we used the highest Z-score (P2-Zmax) during this phase. In addition, we calculated the difference between P2-Zmax and P1-Zmax to examine the change in Z-score (ΔZmax) after IVIG treatment. A P3-Zmax ≥ 2.5 denoted CAL development.

Statistical analysis
Continuous variables are presented as median (interquartile range: IQR), and categorical variables are presented as frequency and percentage. All statistical analyses were performed using JMP Pro version 14 (SAS Institute Japan Ltd., Tokyo, Japan). Associations between two groups were analyzed using Fisher's exact test for categorical variables and the Mann-WhitneyU test for continuous variables. Multivariate logistic regression analysis was used to assess the independent predictors of CAL development. Moreover, the odds ratio (OR) and 95% confidence interval (CI) were calculated. We also confirmed no significant multicollinearity between the explanatory variables. A predictor of P1-Zmax for CAL development was determined by receiver operating characteristic (ROC) analysis. Differences with a two-tailedp value of < 0.05 were considered statistically significant.

Patient characteristics
A total of 324 patients were diagnosed with KD at our hospital during the study period. Based on the exclusion criteria, we excluded patients who did not follow our treatment protocol (n = 16), patients with recurrent KD (n = 12), and patients without follow-up echocardiographic measurements (n = 15). There were not patients who might have an effect on Zscores. A total of 281 patients were enrolled in the study (Fig.  1). All patients were treated within 9 days of KD onset. In terms of treatment response, 177 patients responded to the first IVIG administration, 45 patients responded to additional IVIG administration, and 59 patients (21.0%) did not response to either initial or additional IVIG administration. The latter cohort of patients received CsA as a third-line therapy. Characteristics, clinical data, treatment data, and echocardiographic data of all enrolled patients stratified by the presence or absence of CAL are shown in Table 1. There was a significantly greater number of male children in the group with CAL compared with the group without CAL ( Table 1). The age range of patients with KD was 1 month to 10 years (median, 23 months). There was a significantly greater number of children < 12 months of age in the group with CAL compared with the group without CAL. As for laboratory data, serum Alb concentration (Alb ≤ 3.5 g/dL) was significantly lower during the pre-treatment phase in the group with CAL compared with the group without CAL. As for the number of illness days before initiating the first IVIG treatment, there was no significant difference between the two groups. However, a significantly greater number of patients were resistant to the first IVIG administration and additional IVIG administration in the group with CAL compared with the group without CAL.

Echocardiographic data
All four coronary artery dimensions were measured in 281 cases, with the exception of LAD and LCX artery dimensions in P1, which were measured in 265 cases and 243 cases, respectively. The P1-Zmax in the group with CAL was significantly higher compared with the group without CAL. In addition, there was a significantly greater number of patients with a ΔZmax ≥ 1 in the group with CAL compared with the group without CAL (Table 1).
Serial changes in the Z-score of the four segments are shown in Fig. 2. The contributions of each segment and each phase to Zmax in the group with CAL are indicated by full triangle and full circle. The full triangle symbol indicates the Zmax of patients with CAL who did not respond to two IVIG administrations and who were thus treated with CsA. The full circle symbol shows patients with CAL who responded to either the first or additional IVIG treatments. In all phases, either the RCA or the LMCA mainly contributed to Zmax in patients with CAL. The contributions of the RCA and the LMCA to Zmax in each phase were as shown by the following ratios: P1: 17/21 (81.0%); P2: 16/21 (76.2%); P3: 17/21 (81.0%). In addition, serial changes in P1-Zmax, P2-Zmax, and P3-Zmax in all patients are shown in Fig. 3. The median P1-Zmax, P2-Zmax, and P3-Zmax were 1.54 (IQR: − 0.48 to + 4.51), 1.84 (IQR: + 0.52 to + 7.86), and 1.42 (IQR: + 0.02 to + 6.42), respectively. Sixty-seven patients (23.8%) had a P1-Zmax ≥ 2.0. The relationships between P1-Zmax and ΔZmax in patients with and without CAL are shown in Fig. 4. A P1-Zmax ≥ 2 was not a predictor of CAL, but a ΔZmax ≥ 1 was a strong predictor of CAL development (Table 1). Moreover, only 5 of the 67 patients with a P1-Zmax ≥ 2 had a ΔZmax ≥ 1.

Independent risk factors for CAL development and ROC analysis
We identified six potential risk factors associated with CAL development with a univariate analysis: male sex (p = 0.002), age < 12 months (p = 0.022), serum Alb concentration ≤ 3.5 g/ dL (p = 0.031), resistance to initial IVIG administration (p < 0.001), P1-Zmax (p = 0.035), and a ΔZmax ≥ 1.0 (p = 0.007) ( Table 1). Multivariate logistic regression was subsequently conducted to assess the independent effects of the six factors of CAL development. P1-Zmax (p = 0.047), ΔZmax ≥ 1.0 (p = 0.034), male sex (p = 0.021), < 12 months of age (p = 0.007), and resistance to initial IVIG administration (p = 0.005) were independent risk factors for CAL development (Table 2). Furthermore, in the ROC curve analysis, the area under the ROC curve of P1-Zmax to predict CAL development was 0.64. The optimal cutoff value for P1-Zmax was 1.43 with a sensitivity of 81.0% and a specificity of 48.1% (Fig. 5).

Discussion
Although there are three predictive scores for IVIG resistance in patients with KD in Japan [18][19][20], there are currently no indicators to predict CAL before IVIG treatment. In this study, we identified five independent predictive factors for CAL development in patients with KD: P1-Zmax, a ΔZmax ≥ 1, male sex, < 12 months of age, and resistance to the first IVIG administration. Although male sex, < 12 months of age, and resistance to initial IVIG administration as risk factors for CAL development have been previously reported [2][3][4][5][6], P1 Z-score ≥ 1.43 and a ΔZmax ≥ 1.0 are novel predictors of CAL development in Japan. Evaluating serial changes in Zscore from pre-treatment to post-treatment provides important    CAL in patients with KD is associated with pathological changes in the coronary artery wall. Based on an analysis of autopsy cases, edematous changes in the media are seen in the early stages of KD [24], and coronary artery dilatation may also occur in the early stages of KD onset. As the evaluation of Z-score using echocardiography can consecutively recognize changes in coronary artery dimensions, it is possible that echocardiography may be more accurate than pathological observations for the prediction of CAL development in patients with KD.
The Z-score is an index of coronary artery dimensions that is standardized to sex and BSA. Therefore, the Z-score may be useful for direct and sensitive detection of early coronary artery dilatation compared with clinical findings, such as fever and serum CRP concentration. Fuse et al. reported that 23.4% of patients showed coronary artery dilatation (Z-score ≥ 2.0) after 5 days of illness during the pre-treatment phase [25], which is consistent with our findings (67/281, 23.8%).
Several studies [13][14][15][16] have shown that CAL development may be predicted using the P1 Z-score; however, this has not yet been shown in Japan. These studies reported that a P1-Zmax ≥ 2.0 is better than that of IVIG resistance for predicting CAL development. In addition, Dionne et al.  Odds ratio (OR) and 95% confidence interval (CI) for coronary artery lesion (CAL); median [interquartile range (IQR)]; IVIG, intravenous immunoglobulin; P1-Zmax, pretreatment maximum coronary artery Z-score; P2-Zmax, post-treatment (around 10 days from KD onset) maximum coronary artery Z-score * p < 0.05 reported that intensified initial therapy with steroids or infliximab in patients with a P1 Z-score ≥ 2.0 might prevent progression of coronary arterial dilatation [15]. Thus, if the P1 Z-score can predict CAL development and progression, it will be useful for pediatricians when selecting initial therapies, including intensified therapies. However, many patients in our study with a P1-Zmax ≥ 2.0 did not develop CAL (Fig.  4). In other words, a P1-Zmax ≥ 2.0 had a low positive predictive value for predicting CAL (7/60, 11.7%). There are two possible reasons why our results are different compared with those of other studies. First, the number of days of illness at which first IVIG was administrated in other studies was greater than in our study. Although the median number of illness days at which first IVIG was administrated in a previous report was approximately 7 [13][14][15][16], the median in our cohort was 4 (IQR: [4][5]. Therefore, there is an interval of approximately 3 days in terms of the number of days of illness at which P1 Z-score was measured in our cohort versus other studies. In the Nationwide Survey in Japan [4], the median number of illness days at which initial IVIG was administrated was also [5][6]. Patients with Kawasaki disease in Japan tend to start treatment earlier than patients included in other studies. Fuse et al. reported that 42.9% of patients showed coronary artery dilatation (Z-score ≥ 2.0) after 7 days of illness during the pre-treatment phase, but only 11% of patients showed coronary artery dilatation (Z-score ≥ 2.0) after 4 days of illness according to the cumulative probability curve [25]. Thus, it may not be appropriate to predict CALs at a P1-Zmax ≥ 2.0 in cases where treatment is started earlier. Second, the treatment protocol is a very important factor for coronary outcomes and CAL development. Our treatment protocol included IVIG (first line), IVIG (second line), and CsA (third line), without using the predictive score of IVIG resistance. A total of 60 of 67 patients with a P1-Zmax ≥ 2 showed a ΔZmax < 1 (Fig. 4). It is very interesting that most patients with coronary artery dilatation, in spite of the number of early illness days, did not develop CAL. It is possible that our protocol may be effective for patients with severe KD and dilatation in the early acute phase. From the above, it is clear that a different predictor with a P1-Zmax ≥ 2 is needed. A total of 46 of the 214 patients (21.5%) with a P1-Zmax < 2 showed a ΔZmax ≥ 1 (Fig. 4) and ΔZmax ≥ 1, even after treatment, which were strong predictors of CAL development (p = 0.007, Table 1). These data suggest another important point; Some patients with a P1-Zmax < 2 may be treated with intensified initial therapy. We attempted to examine the cutoff of P1-Zmax for CAL development using ROC curve analysis. A P1-Zmax ≥ 1.43 detected CAL development with an area under the ROC curve of 0.64 (sensitivity = 81.0%; specificity = 48.1%). Although the specificity of a P1-Zmax ≥ 1.43 was only 48.1% for predicting CAL development, the sensitivity was as high as 81.0%. A Z-score < 2.0 is usually in the normal range, but Fuse et al. reported that a P1-Zmax ≥ 1.5 was present in 30.1% of patients by 5 days of illness, 58.4% by 7 days of illness, and 79.0% by 10 days of illness [25]. According to the statistical normal distribution, a Z-score ≥ 1.5 is considered normal in 6.7%. Thus, Fuse et al. also calculated the positive predictive value of P1-Zmax ≥ 1.5 was 0.777 after 5 days of illness [25]. This may indicate that some patients with acute KD who have been treated by 5 days of illness and have a P1-Zmax ≥ 1.5 are at risk of CAL development.
Based on the results of the present study, we propose that patients with a P1-Zmax ≥ 1.43 should be treated with intensified initial therapy using CsA or steroids due to its high sensitivity to predict CAL development. In addition, patients with a ΔZmax ≥ 1 should also be treated with an intensified protocol when one is identified. It is possible that these two steps in the treatment strategy could reduce the incidence of CAL.

Study limitations
There are several limitations of the present study that should be highlighted. First, this study adopted a retrospective design and was carried out at a single institution. Moreover, the sample size was small because many patients were excluded to simplify the inclusion of different conditions. Thus, further multi-center studies should be performed to confirm the findings. Second, the treatment protocol used in this study was our protocol only, and other treatment protocols may demonstrate different rates of CAL development. Third, we could not detect all coronary artery segments in all phases. Fourth, the Zscore formula used in this report is different from the previous report. Thus, comparisons should be made with caution. Fifth, Fig. 5 Receiver operating characteristic (ROC) curve of pretreatment maximum coronary artery Z-score (P1-Zmax) for predicting coronary artery lesion (CAL). The optimal cutoff value for P1-Zmax was 1.43 with a sensitivity of 81.0% and a specificity of 48.1%. AUC, area under the curve four coronary artery branches were measured in accordance with Japanese guidelines [17]. The AHA recommends not to rely on LMCA values, but to use RCA and LAD values instead. In the dominant left coronary artery situs or the dominant right coronary artery situs, patients can be considered as having a P1-Zmax > 1.43.

Conclusions
Independent risk factors associated with CAL development were P1-Zmax, a ΔZmax ≥ 1, male sex, < 12 months of age, and resistance to the first IVIG administration. In addition, using a ROC curve analysis, a P1-Zmax ≥ 1.43 detected CAL development with an area under the ROC curve of 0.64 (sensitivity = 81.0%; specificity = 48.1%). For patients with these risk factors, we need to consider the best approach for intensified initial treatment. Further investigations will be needed in patients receiving alternative initial treatments.
Abbreviations CAL, Coronary artery lesion; CsA, Cyclosporin A; IVIG, Intravenous immunoglobulin; KD, Kawasaki disease; LAD, Left anterior descending artery; LCX, Left circumflex artery; LMCA, Left main coronary artery; P1, Pre-treatment phase (at the time of KD diagnosis or before IVIG treatment); P2, Post-treatment phase (after first IVIG treatment and around 10 days from KD onset); P3, The convalescent phase (around 4 weeks from KD onset); RCA, Right coronary artery; Zmax, The highest Z-score of four segments of coronary artery; ΔZmax, Difference between P2-Zmax and P1-Zmax Authors' contributions Takayuki Suzuki collected data, designed the study, and drafted initial analyses and manuscript. Nobuyuki Kakimoto, Tomohiro Suenaga, and Takashi Takeuchi were responsible for echocardiography examination and data collecting. Tomoya Tsuchihashi, Shoichi Shibuta, Naomi Kitano, and Hiroyuki Suzuki contributed to the study design and data analysis and revised the manuscript. All authors read and approved the final manuscript Funding This work was supported by Grants-in-Aid for Scientific Research (KAKENHI 20K08189) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT).

Declarations
Ethics approval This study was approved by the ethics committee of Wakayama Medical University (No: 2801).
Consent to participate Informed consent was waived, because all data were de-identified.

Conflict of interest
The authors declare no competing interests.