Introduction

Invasive pulmonary aspergillosis (IPA) is a common opportunistic fungal infection, contributing to high mortality in immunocompromised patients. Early diagnosis of IPA in patients at high risk is essential. It is estimated that more than 200,000 cases of IPA occur every year. The mortality rate of IPA reaches over 50% even if patients are treated with antifungal therapy. Once the diagnosis is delayed or missed, the mortality can be nearly 100% [1]. Traditional diagnostic methods, including histology, cytology, and culture, are time-consuming and have low sensitivity [2, 3].

The molecular diagnostic tool polymerase chain reaction (PCR) is one of the most valuable methods used in diagnosis of respiratory pathogens such as virus and mycoplasma [4, 5]. It is also considered as a rapidly expanding technology for fast detection and accurate identification of fungi [6]. Several groups have investigated the performance of PCR from blood or serum in the diagnosis of IPA. A recent meta-analysis performed by Cruciani et al. [7] has assessed the quality of serum PCR for diagnosing IPA from 29 primary studies and concluded that serum PCR showed moderate diagnostic accuracy when used as a diagnostic test. Mengoli et al. [8] has summarized that the diagnostic sensitivity and specificity of blood or serum PCR were 0.75 and 0.87. Arvanitis et al. [9] also finds the similar results.

Bronchoalveolar lavage (BAL) fluid is likely to be more sensitive in early diagnosis [10]. PCR from BAL fluid is recommended for screening the diagnosis of IPA in latest clinical practice guideline from official American Thoracic Society (ATS) and revised European Organization for Research and Treatment of Cancer/Mycoses Study Group (EORTC/MSG) definition [11, 12]. However, both guidelines mainly aim at the diagnosis of IPA in patients with hematological malignancies (HM) and hematopoietic stem cell/solid organ transplantation (HSCT/SOT). The value of BAL fluid PCR in IPA diagnosis among patients with other diseases, such as chronic obstructive pulmonary disease (COPD), solid tumor, pulmonary fibrosis, liver cirrhosis, diabetes, autoimmune disease with prolonged use of corticosteroids (at therapeutic doses ≥ 0.3 mg/kg for ≥ 3 weeks within the past 60 days), treatment with T-cell or B-cell immunosuppressants, ICU admission, etc., remains unclear. Due to lack of standardization, PCR assays vary with respect to DNA extraction protocols, gene targets and amplification platforms, leading to the uncertainty of diagnostic accuracy. Furthermore, several randomized controlled trials (RCT) about BAL fluid PCR in IPA diagnosis have been published, and no updated meta-analysis has been done since 2012. Therefore, we performed a systematic meta-analysis of clinical trials to evaluate the accuracy of BAL fluid PCR assay for the diagnosis of IPA among high-risk patients.

Methods

Search strategy

Two investigators independently searched for relevant articles published in the PubMed, Web of Science, Embase and Cochrane Central Register of Controlled Trials databases up to June 2022. Search terms contained “aspergil*”, “PCR”, “bronchoalveolar lavage”, “respiratory” and “sputum”. The syntax was as follows: ((bronchoalveolar lavage) OR (sputum) OR (respiratory)) AND (PCR) AND (aspergil*). The references of included review articles or identified trials were also checked. Searches were restricted to English language literature on human subjects. Duplicate articles identified in mentioned databases were manually deleted.

Inclusion criteria and definitions

Full-text publications using PCR on BAL fluid were included if (1) they used EORTC/MSG criteria [2, 3, 12] or similar criteria if studies were published before the publication in 2002 for the diagnosis of IPA. (2) they provided data about true-positive, false-positive, false-negative and true-negative results, and (3) the studies included immunocompromised or at-risk patients. Based on EORTC/MSG or similar criteria, patients were classified into four groups on IPA diagnosis: proven, probable, possible and no IPA. We defined the true positive cases if they were classified as proven or probable IPA. Possible IPA cases were excluded because they were considered not reliable enough in the clinical management [13].

Data collection and risk of bias assessment

The first selection was carried out on the basis of the title and abstract by two investigators. The full paper of each potentially eligible study was then obtained. Two investigators independently assessed eligible studies for inclusion. The relevant information was collected from each selected studies including: first author, year of publication, country, sample size, mean age, prevalence of IPA, percentage of patients with HM and HSCT/SOT in the study population, study design, reference standard, DNA extraction, PCR technology, primers and antifungal therapy. Disagreements between authors were resolved by consensus through group discussion. We assessed the quality of studies by the QUADAS-2 (Quality Assessment of studies of Diagnostic Accuracy included in Systematic reviews) checklist [14] to test potential bias in all studies.

Outcome of interests

The primary outcomes of interest were the summary sensitivity and specificity of Aspergillus PCR in BAL fluid for high-risk patients. Secondly, we aimed to evaluate the effect on heterogeneity among different studies about several critical parameters, such as proportion of patients with HM and HSCT/SOT in the study population, the use of antifungal treatment at time of BAL, method of DNA extraction, PCR technology, etc.

Data analysis

The sensitivity and specificity of all studies were calculated by constructing two-by-two tables (proven or probable IPA versus possible or no IPA, proven IPA versus probable, possible or no IPA). All the tables included true-positive, false-positive, false-negative and true-negative results of Aspergillus PCR assay. We calculated the pooled sensitivity, specificity, likelihood ratios and diagnostic odds ratio (DOR) by random-effect model [15]. Summary receiver operating characteristic curves (SROC) was constructed and diagnostic accuracy (area under the curve, AUC) was estimated [16]. Publication bias was assessed by using the Deeks’ regression test for asymmetry [17]. We assessed statistically heterogeneity by the employment of I2 statistic [18]. Potential heterogeneity was estimated by meta-regression and subgroup analyses [19] for sensitivity and specificity. The Pearson's r was applied to measure the linear correlation between the proportion of patients with HM and HSCT/SOT in the study population and the application of antifungal treatment. All analyses were performed using STATA software version 15 (Stata Corp, College Station, Texas) with the program “midas”. P values of < 0.05 denoting statistical significance.

Results

Eligible study characteristics

The study identified 2551 references by the initial search. Eighty-two were selected based on abstract and title search. Forty-one studies, including 5668 patients, satisfied our inclusion criteria and were in the final analysis [20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60] (Fig. 1).

Fig. 1
figure 1

Flow diagram. PCR Polymerase chain reaction; IPA Invasive pulmonary aspergillosis

Full size image

Table 1 summarized the main characteristics of all included studies. Forty-one studies comprised 17 prospective and 24 retrospective ones, of which 6 were case–control studies. Among all studies, the prevalence of proven/probable IPA ranged from 5 to 51%. Most patients suffered HM. Twelve studies involved 1147 patients mainly with COPD, solid tumor, autoimmune disease with prolonged use of corticosteroids, treatment with T-cell or B-cell immunosuppressants, etc., rather than HM or HSCT/SOT. Fourteen studies containing 2061 patients provided data about proven IPA only. Antifungal intervention against IPA before BAL was definitely described in 20 studies. Details of PCR techniques were summarized in Table 2. Two or three different PCR methods were used in 2 studies respectively. Quality assessment is shown according to the QUADAS-2 (Additional file 1: Fig. S1).

Table 1 Main characteristics of studies included
Full size table
Table 2 Technical details of the PCR methods used in the studies included
Full size table

Pooled diagnostic performance for proven/probable IPA in all patients

The summary pretest probability of disease was 20%. The sensitivity and specificity with 95% confidence interval for proven/probable IPA (Additional file 1: Fig. S2) were 0.75 (0.67–0.81) and 0.94 (0.90–0.96). Positive and negative likelihood ratios (PLR and NLR) were 11.8 (7.7–18.1) and 0.27 (0.20–0.36). DOR was 44 (25–77). I2 was more than 50%, which indicated significant heterogeneity was present. The AUC was 0.92 (0.90–0.94) (Fig. 2). The post-test probability indicated that when pretest probability was 20%, PCR method increased the probability to 75% for IPA when the results were positive and decrease to 6% when negative (Fig. 3).

Fig. 2
figure 2

Summary receiver operating characteristic curve of PCR from BAL fluid. The smaller regions (confidence contours) contain possible combinations of sensitivity and specificity means. Broader regions (prediction contours) indicate more uncertainty as to where the likely values of sensitivity and specificity for individual studies might arise

Full size image
Fig. 3
figure 3

Fagan’s nomogram for calculating post-test probabilities in all studies. Straight edges were used to relate the pretest probability of IPA to post-test probabilities by crossing the likelihood ratio line at the point describing the obtained outcome. Solid lines extend from prevalence to PLR and dashed lines extend from prevalence to NLR

Full size image

Pooled diagnostic performance for proven IPA only

Fourteen studies containing 2061 patients described sufficient data for two-by-two table about proven IPA. The pooled sensitivity, specificity (Additional file 1: Fig. S3), PLR, NLR, DOR, AUC with 95% confidence interval were 0.91 (0.68–0.98), 0.80 (0.74–0.85), 4.6 (3.4–6.1), 0.11 (0.03–0.47), 41 (9–193) and 0.89 (0.86–0.91), respectively. The sensitivity for proven IPA was better than that of proven/probable IPAs, whereas the specificity was lower.

Heterogeneity and publication bias

Univariable meta-regression and subgroup analyses were estimated for investigating the heterogeneity in all studies (Fig. 4). Subgroup analyses showed that the underlying diseases and the use of antifungal treatment had a significant impact on the diagnostic sensitivity of BAL fluid PCR. Twenty-nine studies involving patients mostly with HM and HSCT/SOT were enrolled. The summary estimates of BAL fluid for proven/probable IPA were as follows (Table 3): sensitivity 0.68 (0.58–0.76), specificity 0.94 (0.89–0.97), PLR 11.3 (6.3–20.3), NLR 0.34 (0.25–0.45), DOR 33 (16–69), AUC 0.89 (0.86–0.92). For patients with disease such as COPD, solid tumor, autoimmune disease with prolonged use of corticosteroids, treatment with T-cell or B-cell immunosuppressants, etc., the pooled sensitivity, specificity, PLR, NLR, DOR, AUC were 0.88 (0.75–0.95), 0.92 (0.83–0.96), 11.0 (5.4–22.7), 0.13 (0.05–0.29), 88 (33–237), 0.96 (0.94–0.97). The diagnostic sensitivity of BAL fluid PCR was much higher in patients without HM and HSCT/SOT (P < 0.001). Moreover, use of antifungal agents notably lower the sensitivity of PCR. The antifungal treatment had a strong correlation with the underlying diseases. Pearson correlation coefficient between the proportion of patients with HM and HSCT/SOT in the study population and the application of antifungal treatment was 0.76 (P < 0.001).

Fig. 4
figure 4

Forest plot of meta-regression and subgroup analyses for sensitivity and specificity. Ssize100: sample size > 100; Prevalence: prevalence above 15%; Country: European countries; Disease: percentage of patients with hematological malignancy above 50%; Criteria: revised EORTC/MSG criteria in 2008; Walldisruption: proteinase for cell wall disruption; DNAisolation: phenol–chloroform for DNA isolation protocol; PCR: real-time PCR; Primer: 18S rRNA primer

Full size image
Table 3 Results of subgroup analyses
Full size table

Besides, we found some covariates such as study types, group size, prevalence, criteria, DNA extraction protocols, PCR method and primers affected the sensitivity and/or specificity. It was shown sensitivity was lower in prospective, cohort, small group studies and those using revised EORTC/MSG criteria [3]. When DNA isolation kit was used for cell wall disruption, the specificity was a little higher and real-time PCR presented higher specificity whereas the primer 18S rRNA presented lower sensitivity than others. There was significant publication bias in all studies (Additional file 1: Fig. S4).

Discussion

The article evaluated the value of BAL fluid PCR in diagnosis of IPA. To the best of our knowledge, this is the first meta-analysis focusing on the diagnostic value of BAL fluid PCR for IPA patients with disease such as COPD, solid tumor, autoimmune disease with prolonged use of corticosteroids, treatment with T-cell or B-cell immunosuppressants, etc., rather than HM and HSCT/SOT. In our meta-analysis, the overall sensitivity and specificity was 0.75 and 0.94, respectively in all 41 studies. For proven IPA only, sensitivity was higher, which increased to 0.91, but specificity decreased to 0.80. Twelve studies included patients without HM and HSCT/SOT. The pooled sensitivity of BAL fluid PCR for proven/probable IPA for these 12 studies was 0.88, significantly higher than that among patients with HM or HSCT/SOT (0.68), which was related to the high frequency use of antifungal agents. Different factors containing study types, group size, prevalence, percentage of patients with HM and HSCT/SOT in the study population, use of antifungal agents, DNA extraction, PCR methods and primers were responsible for heterogeneity of included studies. Quality items had no significant influence on diagnostic characteristics. However, there was significant publication bias in our study.

Overall, our analysis showed that BAL fluid PCR was an effective test for IPA diagnosis, especially for patients without HM and HSCT/SOT. In patients with HM and HSCT/SOT, the results of specificity, DOR and AUC of PCR for proven/probable IPA diagnosis were reliable, indicating the test had a good discriminative ability. However, the sensitivity and NLR results were not satisfactory, which could increase the rate of missed diagnosis. But when patients with HM and HSCT/SOT were mainly excluded, sensitivity became higher.

To better explore the diagnostic capacity of PCR in BAL fluid, we made analysis for proven IPA separately. In contrast to the diagnosis of probable IPA depending on host, clinical and mycological factors, the criteria for proven IPA involves histopathologic, cytopathologic, direct microscopic examination or culture from sterile material [2, 3, 12], which could be considered as “gold standard” for IPA diagnosis. Consequently, the analysis for proven IPA only seems to be more objective and we gained higher sensitivity for BAL fluid PCR from proven IPA only, increasing the reliability of the diagnostic test. In addition, the decreased specificity of PCR for proven IPA may be due to the incorrect exclusion of those probable cases who were actually infected.

Significant heterogeneity was present in our analysis. Therefore, we implemented subgroup analyses and meta-regression to search for reasons behind these inconsistencies. It was found that studies comprising patients without HM and HSCT/SOT had higher sensitivity. Patients with HM and HSCT/SOT were tend to be treated with azoles prophylaxis more frequently, especially among HSCT/SOT and neutropenic patients [61], which could explain why the sensitivity was lower in studies whose percentage of patients with HM and HSCT/SOT was high. In our analysis, Pearson correlation coefficient was applied to evaluation the relationship between the proportion of patients with HM and HSCT/SOT and the application of antifungal treatment. We found a strong correlation between antifungal therapy and underlying disease. Use of empiric antifungal agents could influence the summary evaluation, which was confirmed to impact the release of Aspergillus DNA, thus lower the residual fungal burden in lung tissue and therefore diminished the sensitivity of PCR assay [62].

Fungal DNA extraction methodology was considered to be the major cause for heterogeneity [63]. Different kinds of DNA extraction methods have been applied in the studies included. The efficiency and the overall performance of wall disruption played a significant role in our analysis as the cell walls of fungi could impede the efficient lysis and liberation of DNA, generating false-negative PCR results [64]. It seemed that commercial nucleic acid extraction methods were more efficient. Besides, the use of DNA isolation mattered as well. Based on these causes, the optimal DNA extraction protocol required verification.

Aspergillus species PCR assay may be another reason for variable test performance. Many various PCR amplification protocols have been published, all of whom remain heterogeneous for they lack of standardization. The different methods led to diverse sensitivity or specificity, and we found studies using real-time PCR had better sensitivity and specificity in our analysis. However, according to the European Aspergillus PCR initiative (EAPCRI), PCR amplification was not limited and most amplification methods would provide acceptable analytical performance in combination with commercial extraction [65]. What’s more, Aspergillus gene targets varied in 18S ribosomal RNA (rRNA), 28S rRNA, the intervening internal transcribed spacer (ITS)-5.8S region, mitochondrial DNA or species-specific primers. A single species, several related species or pan-fungal amplification indicated diverse sensitivity and specificity [63]. In general, because there was so much uncertainty for the use of PCR, the application calls for more standardization.

Compared with BAL galactomannan (GM), PCR from BAL fluid seemed to have superiority in the diagnosis of IPA. Though GM test in plasma, serum or BAL fluid has been incorporated into the EORTC/MSG criteria as one of the clinical diagnostic basis, GM still has its limitations. Affolter et al. reported a moderate diagnostic value of GM in BAL fluid with 50% sensitivity and 73% specificity for proven/probable IPA [66]. For proven IPA only, the specificity decreased, whereas the sensitivity was similar. A meta-analysis conducted by Heng et al. [67] found excellent sensitivity and specificity of GM in BAL fluid for proven/probable IPA. However, when it was estimated for proven IPA only, the specificity decreased to 72%. Similar findings have been reported by Guo et al. [68]. The declining specificity may due to a consequence of false-positive of GM test, increasing the classification of probable IPA using the EORTC/MSG definitions. Besides, the diagnostic accuracy of GM from BAL fluid for patients without HM and HSCT/SOT has not been defined.

There were several limitations in this study. First, the total number of patients without HM and HSCT/SOT included for analysis was relatively small. Second, there was significant heterogeneity in summary estimates. The heterogeneity could be partially explained by study types, group size, the use of antifungal agents, percentage of patients with HM and HSCT/SOT in the study population, DNA extraction, PCR methods and primers. Third, there was significant publication bias in the study. Lastly, misclassification bias could occur when the clinical criteria were used because the EORTC/MSG definition is not the “gold standard” for probable IPA. To identify the accuracy of BAL fluid PCR in IPA diagnosis and its superiority over other methods such as BAL or serum GM test, multicenter RCT study designed with standard criteria is needed.

Conclusion

In conclusion, PCR from BAL fluid is an effective test in IPA diagnosis, indicating the infection of Aspergillus when the result is positive. And the technique seems to be more valuable in the diagnosis of high-risk patients without HM and HSCT/SOT. To improve the accuracy of the test, standardization of DNA extraction and PCR methods is needed for clinical diagnosis.