Authors: Runjin Cai, Yifei Yang, Huan Ge, Xuemei Chen, Xinyue Hu, Yuanyuan Jiang, Lisha Luo, Shuanglinzi Deng, Jiale Tang, Chendong Wu, Huan Tang, Xiaoxiao Gong, Xiaozhao Li, Juntao Feng
Categories: Correspondence
Source: Chinese Medical Journal
To the Editor: Allergic bronchopulmonary aspergillosis (ABPA) is an allergic lung disease caused by hypersensitivity to antigens from Aspergillus, and it most commonly occurs in patients with asthma or cystic fibrosis.^[1]^ The main clinical characteristics of the disease are those related to elevated serum immunoglubulin (Ig) E level, positive Aspergillus fumigatus (A. fumigatus) specific IgE/IgG, repeated pulmonary infiltration, and bronchiectasis. However, due to the lack of unified diagnostic criteria and screening methods, ABPA is always underdiagnosed. In 2021, new diagnostic criteria were proposed for allergic bronchopulmonary mycosis (ABPM) in the Japanese population. The criteria were verified by Asano et al^[2]^ for the first time (so-called Asano criteria), and showed that compared with the existing diagnostic criteria (Rosenberg–Patterson criteria, International Society for Human and Animal Mycology [ISHAM] criteria), the Asano criteria showed better sensitivity and specificity for the diagnosis of ABPM. In this retrospective study, we modified the Asano criteria to verify its effectiveness for diagnosing ABPA. In recent years, latent class analysis (LCA) has been used to evaluate the performance of diagnostic criteria for diseases without diagnostic gold standards.^[3]^ LCA is a form of finite mixture modeling which can integrate multiple variables to define a subset when the true outcome is unknown.^[3]^ In diagnostic tests, LCA can be used to define diagnostic and non-diagnostic groups. Previous studies also used “multidisciplinary team (MDT) diagnosis” or “physician diagnosis”^[2]^ as the reference standard to evaluate the diagnostic efficiency of the criteria. We compared the performance of the modified Asano criteria [Supplementary Table 1, http://links.lww.com/CM9/B940] with existing criteria (Rosenberg–Patterson criteria [Supplementary Table 2, http://links.lww.com/CM9/B940], ISHAM criteria [Supplementary Table 3, http://links.lww.com/CM9/B940], and modified ISHAM criteria [Supplementary Table 4, http://links.lww.com/CM9/B940]^[3]^) using LCA and MDT diagnosis of ABPA as the reference standard, respectively.
This retrospective study was approved by the Institutional Review Board of Xiangya Hospital of Central South University (No. 202210218) and was implemented in compliance with the Declaration of Helsinki. The requirement to obtain the informed consent was waived.
Inpatients and outpatients with persistent asthma (according to the guidelines of the Global Initiative for Asthma) or asthma-like symptoms who were admitted to the Department of Respiratory Medicine, Xiangya Hospital of Central South University from January 2017 to December 2021 were considered for inclusion. Additionally, patients also needed to meet one of the following inclusion increased eosinophil numbers (≥500 cells/mm^3^), increased serum IgE levels (≥417 IU/mL), or pulmonary imaging findings of bronchiectasis or non-fixed pulmonary infiltration. Patients younger than 18 years of age, receiving oral steroids for more than 3 weeks recently or with incomplete data were excluded.^[4]^
Detailed demographic information, including age, sex, peripheral blood eosinophil counts, and total serum IgE levels of the patients, was collected. The results of high-resolution computed tomography (HRCT), bronchoscopy, fungal culture of bronchoalveolar lavage fluid, sputum culture, and specific antibodies against A. fumigatus were also collected. We modified the items 4–7 of the Asano criteria, mainly changing the test for fungi to a test for A. fumigatus [Supplementary Table 1, http://links.lww.com/CM9/B940].
Data were analyzed using SPSS Version 22.0 software (IBM SPSS Inc, Armonk, NY, USA). We analyzed the differences for continuous variables using the independent samples t test, and the categorical data were analyzed by chi-squared, likelihood ratio chi-squared, or Fisher’s exact test. R (version 4.3.2, https://www.r-project.org/) was used for LCA. GraphPad Prism (version 8.0; GraphPad Software Inc, La Jolla, California, USA) was used to generate statistical charts. The statistical difference of area under curves (AUCs) was analyzed by Z test.
A total of 118 patients were included in the study. Of the 118 patients, 29 (24.6%) were diagnosed with definite ABPA, and 6 (5.1%) were diagnosed with probable ABPA according to the modified Asano criteria. According to the Rosenberg–Patterson criteria, 14 patients (11.9%) were diagnosed with definite ABPA, and 8 (6.8%) were diagnosed with probable ABPA. In addition, 25 patients (21.2%) met the ISHAM criteria for definite ABPA, and 2 (1.7%) for probable ABPA. As for modified ISHAM criteria, 25 patients (21.2%) were diagnosed with definite ABPA [Table 1].
First, dynamic disease characteristics of the ABPA patients were re-analyzed and MDT discussion was performed (the discussion group includes more than three physicians with lung allergy and asthma and ABPA research experience). Through investigation and MDT discussion, a total of 40 cases of ABPA were diagnosed. The difference in demographic data between MDT-diagnosed ABPA cases and non-ABPA cases is shown in Supplementary Table 5, http://links.lww.com/CM9/B940. Taking “MDT-diagnosed ABPA” as the reference standard, the performance of the diagnostic criteria was evaluated again. Similarly, the modified Asano criteria had better sensitivity (87.5%) for diagnosing ABPA than the Rosenberg–Patterson criteria (55.0%), ISHAM criteria (67.5%), and modified ISHAM criteria (62.5%), but it had lower specificity (80.8%) [Table 1].
Then, the performance of the four sets of diagnostic criteria was evaluated using LCA, and the goodness-of-fit test of conditional independence (P <0.005) showed that the model was not fit. The diagnostic sensitivity of the modified Asano criteria was the highest (97.1%). In line with Saxena et al study,^[3]^ the sensitivity of the modified ISHAM criteria (83.1%) was higher than that of ISHAM criteria (65.6%), while the sensitivity of Rosenberg–Patterson criteria was the lowest (62.2%). Except for the modified Asano criteria (82.4%), the specificity of the other criteria was 100.0% [Table 1].
Taking “MDT-diagnosed ABPA” as the reference, we generated a receiver operating characteristics (ROC) curve using the data from 118 patients. The AUC of the modified Asano criteria was 0.922, and the best cutoff value was 5 or 6. This means that the Youden index was the largest when 5 or 6 components were met, with a sensitivity of 70.0% and a specificity of 98.7%. This is consistent with the components required for the diagnosis of probable/definite ABPA in the criteria. The AUC values of the Rosenberg–Patterson criteria, ISHAM criteria, and the modified ISHAM criteria were 0.898, 0.929, and 0.913, respectively [Supplementary Figure 1, http://links.lww.com/CM9/B940]. There was no statistically significant difference among the AUCs of the four sets of diagnostic criteria (Z test, P = 0.573 [Rosenberg–Patterson vs. modified ISHAM], P = 0.456 [Rosenberg–Patterson vs. modified Asano], P = 0.713 [modified ISHAM vs. modified Asano], P = 0.711 [ISHAM vs. modified Asano], P = 0.257 [Rosenberg–Patterson vs. ISHAM]).
To evaluate these results in a more realistic context, the item “Presence of A. fumigatus hyphae in bronchial mucus plugs” was removed from the modified Asano criteria (pathological diagnosis is needed, and relevant data are rarely obtained). The AUC was 0.902, the best critical value was 4 or 5 when the sensitivity and specificity were 77.5% and 87.2%, respectively [Supplementary Figure 1, http://links.lww.com/CM9/B940].
In patients in the non-ABPA group, we screened the related diseases (such as pulmonary aspergillosis^[5]^) that are easily confused with ABPA to verify the differential ability of each criteria for ABPA. Compared with the existing criteria, modified Asano criteria has a slightly higher misdiagnosis rate [Supplementary Table 6, http://links.lww.com/CM9/B940].
Our results show that the diagnostic sensitivity of the modified Asano criteria was indeed increased compared to that of the Rosenberg–Patterson criteria, ISHAM criteria, and modified ISHAM criteria whether using LCA or taking “MDT-diagnosed ABPA” as the reference standard, but the specificity did not increase as suggested in the study of Asano et al.^[2]^
We analyzed the reasons for the high diagnostic sensitivity of the modified Asano criteria. First of all, the critical value of serum IgE is still set at 417 IU/mL. A previous study showed that the difference threshold of total IgE level affects the specificity and sensitivity of ABPA total IgE level ≥1000 IU/mL has high specificity (99.5%) but poor sensitivity (55.6%); total IgE level ≥417 IU/mL has higher sensitivity (97.1%), but the specificity is lower (70.7%).^[4]^
Compared with ISHAM criteria, the most obvious change of modified ISHAM criteria is the serum IgE level is changed to 500 IU/mL. Saxena et al^[3]^ set the serum IgE level to 417 IU/mL, 1000 IU/mL, and 500 IU/mL in ISHAM criteria to test their diagnostic efficiency for ABPA, and it was found that the sensitivity and specificity are both 100% with IgE level ≥500 IU/mL as a component.^[2]^ In this study, we also found that the diagnostic sensitivity of modified ISHAM criteria is higher than that of ISHAM criteria on LCA. Then, as a part of the criteria for diagnosing ABPA, it may be more appropriate to set the threshold value of IgE level below 1000 IU/mL.
Second, compared with the three other sets of criteria, the modified Asano criteria include additional etiological components, such as “the growth of A. fumigatus in sputum culture or bronchial lavage fluid” and “the presence of A. fumigatus hyphae in bronchial mucus plugs”. While the latter needs pathological diagnosis, it was found that approximately 40–60% of ABPA patients had sputum cultures that were positive for A. fumigatus, but the specificity was poor.^[6]^ Thus, it can be a secondary criterion for diagnosis of ABPA. Additionally, mucus plugs were also added to the Asano criteria. In earlier research, it was found that 18.7% of patients diagnosed with ABPA present with high attenuation mucus in the bronchi on CT scans.^[6]^
The diagnostic conditions of the modified Asano criteria were more liberal than those of the other three sets of criteria and therefore diagnostic specificity was low in our study. We found that in the modified Asano criteria and Rosenberg–Patterson criteria, the same weight is given to each component. However, in practical research, some components obviously occupy a more important proportion. Therefore, the ISHAM criteria and modified ISHAM criteria provide both necessary and non-necessary conditions.
There are limitations in our study. First, this study verified the validity of the modified Asano criteria for the diagnosis of ABPA in the Chinese population, and whether it is truly applied to the diagnosis of ABPA still needs more support from additional patients in different races, especially non-Asian population. In addition, the goodness-of-fit test indicates that our model is not fit. That means conditional independence cannot be satisfied, so the diagnostic performance calculated by LCA may be inflated. Therefore, it is necessary to evaluate the sensitivity and specificity again using the MDT diagnosis as the reference standard.
In conclusion, our findings are in good agreement with those of Asano et al.^[2]^ The modified Asano criteria improve the diagnostic sensitivity for ABPA, but the diagnostic specificity is lower. Compared with the three existing diagnostic criteria, we think that the modified Asano criteria are beneficial to the early screening of ABPA, which reduce the frequency of missed diagnosis of ABPA, and reduce complications in later stages caused by ABPA from underdiagnoses or undertreatment. However, due to the low specificity of modified Asano criteria in the diagnosis of ABPA, the follow-up diagnosis needs the support of other criteria.
This study was funded by the National Natural Science Foundation of China (Nos. 82270033 and 81873407) and the Natural Science Foundation of Hunan province (No. 2022JJ30924).
None.