Introduction
Chronic rhinosinusitis is a persistent inflammatory disease affecting the nasal cavity and paranasal sinuses for longer than 12 weeks. Traditionally, chronic rhinosinusitis was classified into two subtypes primarily based on the absence or presence of nasal polyps: chronic rhinosinusitis without nasal polyps and chronic rhinosinusitis with nasal polyps.Reference Fokkens, Lund, Mullol, Bachert, Alobid and Baroody1 Furthermore, it can be classified as eosinophilic chronic rhinosinusitis and non-eosinophilic chronic rhinosinusitis based on the eosinophilic infiltration level in the nasal mucosa or polyps.Reference Ishitoya, Sakuma and Tsukuda2 The latest European Rhinologic Society Guidelines propose a classification method based on the type of associated inflammation (type 2 or non-type 2 inflammation).Reference Fokkens, Lund, Hopkins, Hellings, Kern and Reitsma3 Eosinophilic chronic rhinosinusitis is a type 2 inflammatory disease characterised by good steroid responsiveness, worse olfactory dysfunction and high recurrence rate after surgery. In contrast, non-eosinophilic chronic rhinosinusitis responds well to medical or surgical interventions and exhibits the features of lower post-operative recurrence rate.Reference Ahn, Lee, Ha, Hwang, Yoon and Kim4–Reference Wang, Zheng, Liu and Guo6 Thus, it is important to discriminate the patients’ endotypes and formulate a personalised treatment strategy for chronic rhinosinusitis with nasal polyps patients. It is therefore necessary to find a simple classification method that is applicable pre-operatively.
Recently, examination of peripheral blood eosinophil has been used as a predictor for the identification of eosinophilic chronic rhinosinusitis.Reference Chen, Wang, Xie, Xie and Wang7–Reference Zhou, Fan, Qin, Liu, Gao and Lv29 Unfortunately, the results of these studies are not consistent. Some authors showed that examination of peripheral blood eosinophil may be a useful method for the differential diagnosis of eosinophilic chronic rhinosinusitis,Reference Chen, Wang, Xie, Xie and Wang7,Reference Han, Wang, Zhu, Wang, Chen and Wang9–Reference Liu, Yan, Qi, Zhang, Zhang and Wang13,Reference Xu, Du, Zheng, Duan, Jia and Chen16,Reference Zhong, Yuan, Du, Tan, Yang and Liu19,Reference Zuo, Guo, Chen, Xu, Xu and Shi23–Reference Li, Zhang, Wang, Yan, Jiang and Yu25 whereas others suggested that serum eosinophilia was not a good marker of tissue eosinophilia.Reference DU, Kong, Yu, Xia, Zhang and Guo8,Reference Lv, Liu, Xiang, Zhang, Chen and Kong14,Reference Sivrice, Okur, Yasan, Tüz, Kumbul and Akın15,Reference Zhang, Ma, Liu, Xie, Cao and Zhang17,Reference Zhang, Zhu, Chen, Xia, Yu and Gao18,Reference Zhou, Zhang, Jin, Ma, Xian and He20–Reference Zhu, He, Li, Wang, Wen and Wen22,Reference Ma, Hsieh, Huang, Li, Tsou and Lin26–Reference Zhou, Fan, Qin, Liu, Gao and Lv29 The associations between peripheral blood eosinophils and tissue histopathology are yet to be defined. This study aimed to perform a systematic review and meta-analysis to evaluate the predictive value of peripheral blood eosinophil levels for eosinophilic chronic rhinosinusitis.
Methods
This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Material 1).Reference Liberati, Altman, Tetzlaff, Mulrow, Gøtzsche and Ioannidis30 The protocol used in this article has been registered on PROSPERO and the registry number for this study is CRD42023402824.
Search strategy
A literature search was performed on PubMed, Embase and the Cochrane Library up to 8 April 2024 by two independent reviewers. The search terms were: (((((((chronic rhinosinusitis) OR (nasal polyps)) OR (CRSwNP)) OR (eosinophilic chronic rhinosinusitis)) OR (ECRS)) OR (ECRSwNP)) AND ((((blood eosinophil count) OR (blood eosinophil percentage)) OR (eosinophil ratio)) OR (blood eosinophilia))) AND ((((((sensitivity) OR (specificity)) OR (area under the curve)) OR (AUC)) OR (ROC curves)) OR (receiver operating characteristic curves)). No restrictions regarding the publication language were applied. The titles and abstracts of each retrieved study were screened to determine records that should be further evaluated for eligibility. Full texts of the eligible studies were retrieved for further assessment. In addition, the reference lists of the relevant articles were also scanned to identify other potentially eligible studies.
Selection criteria
The inclusion criteria were follows: (1) studies that evaluated the predictive value of peripheral blood eosinophils for the diagnosis of eosinophilic chronic rhinosinusitis; (2) studies that reported complete data on the predictive value of blood eosinophil examination or in which the number of true–positive, false–positive, true–negative and false–negative outcomes could be extracted; and (3) the ‘gold standard’ for diagnosis of eosinophilic chronic rhinosinusitis came from histopathological examination. The exclusion criteria were as follows: (1) case reports, reviews, comments, thesis, conference abstracts, editorials and letters; (2) studies for which data could not be fully extracted; and (3) repeated publications (the research with the largest sample size was selected).
Data extraction
Two reviewers extracted the following data independently: first author, publication date, nationality, number of patients, diagnostic criteria of eosinophilic chronic rhinosinusitis, blood eosinophil related predictors, cut-off values of predictors for the prediction of eosinophilic chronic rhinosinusitis. The outcomes of true–positive, true–negative, false–positive and false–negative were extracted and cross-checked.
Qualitative assessment
The risk of bias of each included study was assessed based on the Quality Assessment of Diagnostic Accuracy Studies–2 tool. This tool comprises four domains: patient selection, index test, reference standard, and flow and timing. Each domain contains a set of signalling questions and is scored high, low or unclear by two independent reviewers. Review Manager 5.4 was used for the evaluation of methodological quality in this meta-analysis.
Any discrepancies in the process of article selection, data extraction and quality assessment were resolved through discussions or elucidated by a third party.
Statistical analysis
Statistical analysis of the data was performed using Meta-Disc 1.4 and Stata 16.0. The Q test and I Reference Ishitoya, Sakuma and Tsukuda2 statistic were used to evaluate heterogeneity among the outcomes of included studies. Significant heterogeneity was indicated by p less than 0.05 in the Q tests and I Reference Ishitoya, Sakuma and Tsukuda2 greater than 50 per cent. Pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and diagnostic odds ratio with 95 per cent confidence interval (CI) were calculated for each blood eosinophil predictor. We also developed a symmetric receiver operator characteristic curve and calculated the area under the curve. Meta-regression, sub-group analyses and sensitivity analyses were performed to explore the sources of heterogeneity. Deek's funnel plot asymmetry test was used to assess publication bias and p less than 0.05 was considered statistically significant.
Results
Eligible studies
We identified 635 potentially relevant articles after initial electronic searching. One additional article was identified through a review of reference lists of the relevant articles. Fifty articles were left for further selection after removal of duplicates and a review of the titles and abstracts. After full-text screening, 3 studies were excluded because the diagnostic criterion of eosinophilic chronic rhinosinusitis was not pathological examination, 2 studies were excluded because the research subjects overlapped with another study, 21 studies were excluded because the essential data could not be extracted and 1 study was excluded because it was a conference abstract. Finally, 23 studies fulfilled the inclusion criteria and underwent data extraction. Study selection and screening proceeded based on the strategy outlined in the standard PRISMA statement. The article selection process is given in Figure 1. The characteristics of the included studies are summarised in Table 1.
Figure 1. Flow diagram of the systematic literature search and study selection process.
Table 1. Characteristics of the studies included for the meta-analysis
ECRS = eosinophilic chronic rhinosinusitis; BEP = blood eosinophil percentage; BEC = blood eosinophil count; HPF = high power field; NR = not reported
Quality assessment
Figure 2 shows the results of the quality assessment of the included studies according to Quality Assessment of Diagnostic Accuracy Studies–2 criteria. Of the 23 included studies, 1 study fulfilled 6 items, 8 studies fulfilled 5 items and 14 studies met 4 items. Among these, 22 studies were labelled as unclear in the patient selection domain because the author did not report whether the patients were consecutive or randomly selected. Overall, 22 studies were classified as high risk in the index test domain because the threshold used was not pre-specified. The reference standard domain of 15 studies was evaluated as unclear risk because authors did not report whether reference standard results were interpreted without knowledge of the results of the index tests.
Figure 2. Methodological quality assessment of included studies according to Quality Assessment of Diagnostic Accuracy Studies–2 (QUADAS-2) criteria.
Predictive value of peripheral blood eosinophils for eosinophilic chronic rhinosinusitis
Of the 23 included articles, 15 reported the predictive value of peripheral blood eosinophil percentage for identification of eosinophilic chronic rhinosinusitis. The Spearman correlation coefficient of the blood eosinophil percentage was 0.35 (p = 0.20), suggesting that there was no threshold effect. The pooled sensitivity and specificity were 0.77 (95 per cent CI = 0.69–0.83) and 0.74 (95 per cent CI = 0.68–0.80), respectively (Figure 3A). The overall positive and negative likelihood ratios (positive likelihood ratio and negative likelihood ratio) were 2.97 (95 per cent CI = 2.38–3.72) and 0.31 (95 per cent CI = 0.24–0.42), respectively. The pooled diagnostic odds ratio (DOR) was 9.47 (95 per cent CI = 6.31–14.22). The area under the summary receiver operating characteristic (SROC) curve was 0.82 (95 per cent CI = 0.78–0.85) (Figure 4A; Supplementary material, Table 2).
Figure 3. Forest plot of the sensitivity and specificity of the blood eosinophil percentage (A) and blood eosinophil count (B) for ECRS. CI = confidence interval; df = degrees of freedom
Figure 4. Summary receiver operating characteristic curve for blood eosinophil percentage (A), and blood eosinophil count (B). SROC = summary receiver operating characteristic; SENS = sensitivity; SPEC = specificity; AUC = area under the curve
Seventeen studies reported the predictive values of peripheral blood eosinophil count for eosinophilic chronic rhinosinusitis. The Spearman correlation analysis revealed a coefficient of −0.10 for blood eosinophil count (p = 0.71), indicating the absence of a threshold effect. The following results were obtained: pooled sensitivity, 0.78 (95 per cent CI = 0.73–0.82); pooled specificity, 0.73 (95 per cent CI = 0.69–0.77) (Figure 3B); pooled positive likelihood ratio, 2.93 (95 per cent CI = 2.45–3.50); pooled negative likelihood ratio, 0.30 (95 per cent CI = 0.24–0.37); pooled diagnostic odds ratio, 9.74 (95 per cent CI = 6.73–14.08). The area under the SROC (summary receiver operating characteristic) curve was 0.82 (95 per cent CI = 0.78–0.85) (Figure 4B; Supplementary material, Table 3).
Meta-regression and sub-group analysis
Because of the high heterogeneity in the included studies, we performed meta-regression including patient number (sample size) and the diagnostic criteria of eosinophilic chronic rhinosinusitis. The meta-regression of the blood eosinophil percentage (Figure 5A) showed that diagnostic criteria had an influence on the heterogeneity of sensitivity and specificity, and patient number led to a difference of sensitivity. The meta-regression of blood eosinophil count (Figure 5B) showed that both patient number and diagnostic criteria are likely to be the sources of heterogeneity.
Figure 5. Meta-regression for blood eosinophil percentage (A), and blood eosinophil count (B) in the identification of ECRS.
Sub-group analyses were conducted based on the number of patients (≥100 or <100) and the diagnostic criteria of eosinophilic chronic rhinosinusitis (proportion of tissue eosinophils to the total number of inflammation cells or tissue eosinophils per high-power field) to explore the effects of various study characteristics on the predictive value of blood eosinophil percentage and blood eosinophil count for eosinophilic chronic rhinosinusitis. The results of blood eosinophil percentage (Figure 5A) indicated that the sub-group with at least 100 patients exhibited lower sensitivity compared with the sub-group with fewer than 100 patients (p < 0.01). In addition, the sub-group using the diagnostic criteria of the proportion of tissue eosinophils to the total number of inflammation cells was less sensitive (p < 0.05) and specific (p < 0.01) compared with the sub-group using tissue eosinophils per high-power field. Regarding the results of blood eosinophil count (Figure 5B), the sub-group with at least 100 patients exhibited lower sensitivity (p < 0.001) but higher specificity (p < 0.01), while the sub-group using the diagnostic criteria of the proportion of tissue eosinophils to the total number of inflammation cells had lower sensitivity (p < 0.001) and specificity (p < 0.001).
Sensitivity analysis
We performed a sensitivity analysis by excluding the included studies one by one. The pooled effect size of blood eosinophil percentage (Figure 6A) showed no significant change, indicating the findings were relatively robust. The pooled effect size of blood eosinophil count (Figure 6B) decreased significantly after excluding the study by Li (2024–2) and increased after excluding the study by Sivrice (2020), indicating that the findings of blood eosinophil count are not so robust.
Figure 6. Sensitivity analysis of blood eosinophil percentage (A), and blood eosinophil count (B) in the prediction of ECRS; CI indicates confidence interval.
Publication bias analysis
Publication bias was assessed by Deeks’ funnel plot asymmetry test. The results (Figure 7) showed that publication bias was not statistically significant for studies regarding blood eosinophil percentage (p = 0.17) and blood eosinophil count (p = 0.30).
Figure 7. Deek's funnel plot asymmetry test of blood eosinophil percentage (A), and blood eosinophil count (B). ESS = effective sample size
Discussion
Principal findings
Peripheral blood is an easily accessible biological sample capable of reflecting the inflammatory state of the body. Some authors recently attempted to examine blood eosinophil percentage and blood eosinophil count to evaluate their predictive value for eosinophilic chronic rhinosinusitis patients.Reference Chen, Wang, Xie, Xie and Wang7–Reference Zhou, Fan, Qin, Liu, Gao and Lv29 The systematic review and meta-analysis by Kim et al. compared the differences in blood eosinophil percentage and blood eosinophil count between eosinophilic chronic rhinosinusitis and non-eosinophilic chronic rhinosinusitis patients. The results showed that both blood eosinophil percentage and blood eosinophil count were significantly higher in the eosinophilic chronic rhinosinusitis sub-group than in the non-eosinophilic chronic rhinosinusitis sub-group, suggesting that the blood eosinophil parameter may be used as a simple indicator for subclassification of eosinophilic chronic rhinosinusitis and non-eosinophilic chronic rhinosinusitis. However, they did not analyse the predictive ability of the blood eosinophil percentage and blood eosinophil count for eosinophilic chronic rhinosinusitis.Reference Kim, Kim, Basurrah and Hwang31 In the present study, we conducted a systematic review and meta-analysis of the included 23 clinical studies to evaluate the predictive efficiency of blood eosinophil percentage and blood eosinophil count for the identification of eosinophilic chronic rhinosinusitis. The results showed that the pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and diagnostic odds ratio of the blood eosinophil percentage were 0.77, 0.74, 2.97, 0.31 and 9.47 respectively. The blood eosinophil count exhibited similar values for the identification of eosinophilic chronic rhinosinusitis with a pooled sensitivity of 0.78, pooled specificity of 0.73, pooled positive likelihood ratio of 2.93, pooled negative likelihood ratio 0.30 and diagnostic odds ratio of 9.74.
• The results of previous systematic reviews or meta-analyses showed that both peripheral blood eosinophil percentage and peripheral blood eosinophil count were significantly higher in eosinophilic chronic rhinosinusitis patients than in non-eosinophilic chronic rhinosinusitis patients
• Examination of peripheral blood eosinophil has been used as a predictor for the identification of eosinophilic chronic rhinosinusitis
• This systematic review and meta-analysis of 23 studies that fulfilled the inclusion criteria evaluated the predictive value of peripheral blood eosinophil levels for the identification of eosinophilic chronic rhinosinusitis
• The results of this meta-analysis suggest that there is not sufficient evidence to support peripheral eosinophilia as a good predictor of eosinophilic chronic rhinosinusitis
Clinical implications
Receiver operating characteristic curve analysis has been widely used to evaluate diagnostic predictors in many fields. The predictive efficiency of a predictor was evaluated by the area under the receiver operating characteristic curve.Reference Eusebi32 An area under the receiver operating characteristic curve greater than 0.9 indicates high accuracy, while an area under the curves of 0.7–0.9 or 0.5–0.7 is considered to have moderate or low accuracy, respectively.Reference Fischer, Bachmann and Jaeschke33 The results of this study showed that the area under the curve of blood eosinophil percentage and blood eosinophil count was 0.82, indicating moderate diagnostic accuracy for eosinophilic chronic rhinosinusitis. In clinical practice, it is important to know how a particular test result predicts the risk of abnormality. The likelihood ratio is a comprehensive index calculated by pooling sensitivity and specificity. It can be used to calculate the probability of abnormality and thus might be more helpful than sensitivity and specificity. According to literature reports, positive likelihood ratio greater than 10 and negative likelihood ratio less than 0.1 suggests excellent accuracy, and positive likelihood ratio greater than 5 and negative likelihood ratio less than 0.2 indicates strong predictive power.Reference Deeks34 The results of this study showed that the positive likelihood ratio of blood eosinophil percentage and blood eosinophil count are 2.97 and 2.93, while the negative likelihood ratio is 0.31 and 0.30, indicating neither blood eosinophil percentage nor blood eosinophil count exhibited sufficient predictive value for eosinophilic chronic rhinosinusitis.
Limitations
This study is limited by several factors. Firstly, studies included in this meta-analysis have a certain degree of risk of bias. For instance, the cut-off values of blood eosinophil percentage and blood eosinophil count for diagnosis of eosinophilic chronic rhinosinusitis were not preset, but the optimal cut-off value determined based on the receiver operating characteristic curve, which might improve their diagnostic value. Secondly, moderate heterogeneity was found among the included studies. We used meta-regression and sub-group analysis to explore the heterogeneity of data. The results of meta-regression and sub-group analysis indicated that the sample size and diagnostic criteria of eosinophilic chronic rhinosinusitis can explain the heterogeneity of included studies. However, in the included studies in the present study, different histological criteria were applied to define eosinophilic chronic rhinosinusitis, including 16 criteria for the percentage of eosinophils and/or inflammatory cells (>10 per cent, >27 per cent, >50 per cent), 7 criteria for absolute eosinophil count (>5, >10, >55 eosinophils per high power field). Thirdly, the vast majority of included studies were performed in the Chinese population. Given the difference in the prevalence of T Helper Cell Type 2 (TH2) inflammation and eosinophilic disease between eastern and western countries,Reference Jankowski, Bouchoua, Coffinet and Vignaud35–Reference Kim, Lee, Kim, Cho, Park and Shin38 the results of this meta-analysis may not be applicable to other populations.
Conclusion
This meta-analysis indicates that there is not convincing evidence to support peripheral eosinophilia as a good predictor of eosinophilic chronic rhinosinusitis. Given the aforementioned limitations, the conclusion of this meta-analysis should be interpreted with caution.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022215124001208.
