Prediction nomogram for coronary artery aneurysms at one month in Kawasaki disease

Kawasaki disease (KD) is one of the most common pediatric vasculitis all over the world. It is responsible for most acquired heart diseases due to its predilection for coronary arteries [1]. The incidence of coronary artery lesions (CALs) was reported to be approximately 25% in untreated patients [1]. Fortunately, nowadays these patients benefit greatly from the standard treatment regimen of intravenous immunoglobulin (IVIG) and aspirin in the acute stage. Most dilations were reported to resolve within 4 to 8 weeks of disease onset [2‐4]. However, some coronary artery aneurysms (CAA), especially large or giant CAA, will persist for a long time, leading to a higher cardiac event risk in later life [3, 5, 6].

It was reported baseline CAA was associated with persistent CAA in Western countries whereas recent studies found that more than half of the coronary artery dilations recovered one month after disease onset, drawing a conclusion that CAA at one month after disease onset was of great significance and could act as a guide for late coronary outcomes [3, 4, 7‐9]. Previously, Son and his colleagues established a risk model for CAA at 2 to 8 weeks in North America, which was identified to have a poor performance in the Japanese population [10, 11]. In China, the existing scoring systems for CAA were predictive of the occurrence of CAA in acute KD when the predictive model for CAA at one month after disease onset was still lacking [12, 13]. Therefore, the present study was designed to: (1) create a risk model for CAA prediction at one month with development and validation cohorts in an East China population. (2) identify the predictive utility of the Son score using data from this area. We aimed to facilitate the prediction of CAA at one month in clinical practice and further alleviate the unnecessary worries of the parents of patients with CAA which could possibly resolve over time.

In the present study, we developed a model for predicting CAA at one month in KD. The prediction model included age, sex, and CALs in the acute stage, which exhibited good performance in both the DC and VC. Besides, we evaluated the performance of the Son score, and found that although its sensitivity was quite high, it was no better than our prediction model in our cohort. These findings shed light on an early estimation of CAA at one month after KD onset.

Arteritis in KD model experienced three pathological processes, of which one month after disease onset fell into the subacute stage [1]. Pathologically, the coronary arteries were infiltrated with lymphocytes, plasma cells, and eosinophils without luminal myofibroblastic proliferation [1]. Clinically, patients were in the convalescent stage and routine follow-up was performed during this period [16]. Also, the Japanese Circulation Society guidelines of KD classified the severity of CALs based on the echocardiographic or coronary angiographic findings at one month from the onset [17]. Taken together, the coronary artery status at one month had a practical significance and an outsize role in clinical practice.

In previous multicenter studies in Japan and North America, an increased baseline Z score was recognized as a major risk factor for CALs at 1–2 months [8, 18, 19]. We also found that coronary artery status in the acute stage had a prominent predictive role in our model. When we classified acute CALs into five categories based on Z score, we found that dilations could possibly resolve and giant CAA would possibly persist, when the small and medium CAA were intermediate. Our results were in consistence with Ryusuke Ae’s [4]. He and his colleagues found that four-fifths of patients with dilatations improved at one month when aneurysms were less likely to improve over time. Indeed, larger aneurysms were prone to have a relatively poor prognosis, suffering from a series of cardiovascular events in later life [3, 5, 6].

Younger males were considered to be more vulnerable to KD. Moreover, younger age and males had been widely reported as risk factors for CALs in the acute stage, as well as at 1–2 months in a number of previous studies [8, 18‐20]. Although the specific mechanism remained unclear, we speculated it might be related to the vulnerability of the immature coronary arteries in male infants caused by inflammation. Thus, pediatricians should pay more attention to these younger males with regard to a relatively unsatisfactory outcome. Unlike the results in the Post RAISE studies, IVIG resistance was not considered a risk factor in the present study [18, 19]. The AUC of IVIG resistance was 0.549 with a sensitivity of 15.6% and a specificity of 94.3%. The underlying reasons might be the divergence in different definitions of CALs, when we used Dallaire’s Z score equation and the Post RAISE studies used either Kobayashi’s Z score equation or the absolute coronary diameters to define CALs. Otherwise, there was no difference in CAAs between patients with and without IVIG resistance in our study, as was reported in another study which was carried out in Anhui, China [13].

When we compared our model with the Son score, we found that the latter was not more predictive than our model, although the Son score had a better performance in our cohort than in a Japanese one [11]. The reasons might lie in the different ethnic backgrounds of the patients enrolled and the unsatisfactory performance of CRP with an AUC of 0.557 (sensitivity: 27.1%, specificity: 84.3%) in our study. Actually, CRP was recognized as an independent risk factor of IVIG resistance rather than CALs in the acute stage in most studies.

The present study has several limitations. First, we excluded 255 patients who were lost in the one-month follow-up, which could lead to bias because most of the lost patients had less severe CALs in the acute stage and wouldn’t follow the doctor's advice after discharge. Second, the patients enrolled in our study were all Chinese, which would not guarantee the performance of the model in other nations. Moreover, a multicenter study is warranted in the future.

Our prediction model for CAA at one month after disease onset in KD had an excellent predictive utility.