Evaluation of the Double Agar Gel Immunodiffusion Test and of the Enzyme-Linked Immunosorbent Assay in the Diagnosis and Follow-Up of Patients with Chronic Pulmonary Aspergillosis

The diagnosis of chronic pulmonary aspergillosis (CPA) depends on the radiologic image and the identification of specific antibodies. The present study aimed to evaluate accuracy parameters of enzyme-linked immunosorbent assay (ELISA) and of the determination of serum galactomannan level in the diagnosis of patients with CPA, comparing these results with the double agar gel immunodiffusion (DID) test. In addition, the prevalence of cross-reactivity and the serological progression after treatment were evaluated by comparing DID and ELISA. Six study groups were formed: G1: 22 patients with CPA, 17 of whom had Aspergillus fungus ball, one chronic cavitary pulmonary aspergillosis (CCPA) and four chronic fibrosing pulmonary aspergillosis (CFPA); G2: 28 patients with pulmonary tuberculosis (TB); G3: 23 patients with histoplasmosis (HST); G4: 50 patients with paracoccidioidomycosis (PCM); G5: 20 patients with cryptococcosis (CRC); and G6: 200 healthy controls. Serum antibodies were measured by DID and ELISA, with two antigen preparations—Aspergillus fumigatus (DID1, ELISA1) and a pool of A. fumigatus, A. flavus and A. niger antigens (DID2, ELISA2). The Platélia Aspergillus Enzyme Immunoassay (EIA) kit was used to measure galactomannan. The cut-off points of ELISA were determined for each antigen preparation and for the 95% and 99% confidence intervals. Despite the low sensitivity, DID was the technique of choice due to its specificity, positive and negative predictive values and positive likelihood ratio–especially with the antigen pool and due to the low frequency of cross-reactivity. ELISA1 and a 0.090 cut-off showed high sensitivity, specificity and negative predictive value, but a high frequency of cross-reactivity with CRC. The best degree of agreement was observed between ELISA1 and ELISA2. The detection of serum galactomannan showed high sensitivity, comparable to ELISA2. The immunodiffusion test showed an excellent relationship with the progression after treatment, which made it the reaction of choice for patient follow-up.

galactomannan showed high sensitivity, comparable to ELISA 2 . The immunodiffusion test showed an excellent relationship with the progression after treatment, which made it the reaction of choice for patient follow-up.
(Universidade Estadual Paulista-UNESP), where patients with other systemic mycoses are also treated.
Patients with CPA. Patients with CPA (G1) in different clinical forms were evaluated: aspergilloma, CCPA and CFPA. Based on specifications by Denning et al. [3,21] and Camuset et al. [22], the case definitions used in the present study are presented below.
Aspergilloma. Cases of aspergilloma exhibited a clinical picture and radiography and/or planigraphy and/or chest computed tomography (CT) consistent with lung cavitation and solid rounded mass inside it, which is suggestive of fungus ball.
CCPA. Cases of CCPA exhibited a clinical picture and radiography and/or planigraphy and/or chest CT scan consistent with lung cavitation and no solid rounded mass inside it.
CFPA. Cases of CFPA exhibited clinical picture, and radiography, and/or planigraphy and/or chest CT scan consistent with pulmonary fibrosis and no cavitation.
All aspergilloma, CCPA and CFPA cases were classified as confirmed, probable and possible cases, as follows: Confirmed case. Characterized by the presence of specific serum antibodies determined by the DID test or by the identification of Aspergillus sp. in bronchoalveolar lavage.
Probable case. Characterized by the presence of hyaline hyphae consistent with Aspergillus sp. in the sputum.
Possible case. Characterized by negative results on specific anti-Aspergillus serum antibody detection by DID test, as well as the absence of hyaline hyphae consistent with Aspergillus sp. in the sputum.
The inclusion and exclusion criteria were the same for the different clinical forms of CPA. Inclusion criteria. Confirmed and probable aspergilloma, CCPA and CFPA patients were included in the study.
Exclusion criteria. Patients with AIDS or extra-pulmonary lesion caused by systemic diseases of inflammatory or neoplastic origin as comorbidity, pregnancy and lactation were excluded from the study.

Patients with pulmonary tuberculosis
In total, 28 patients who had pulmonary TB confirmed (G2) by the identification of acid-fast bacilli (AFB) in the sputum, in a histopathological examination of a lung fragment stained by the Ziehl-Neelsen method and/or in a culture of these clinical samples in Lowenstein-Jensen medium were included in the study. These patients presented no underlying diseases and responded to the classic treatment regimen for Mycobacterium tuberculosis.

Patients with histoplasmosis, paracoccidioidomycosis and cryptococcosis
In total, 23 patients with histoplasmosis-HST (G3), 50 with paracoccidioidomycosis-PCM (G4) and 20 with cryptococcosis-CRC (G5) presenting clinical condition consistent with each fungal disease and direct mycological and / or histopathological examination of clinical samples (confirmed cases), or without the identification of the fungus but with positive detection of specific serum antibodies by DID test (HST and PCM), or of the specific antigen by the latex agglutination test (probable cases) were included in the study [23].
Exclusion criteria. The exclusion criteria were the same for patients with pulmonary TBC and PCM. Patients with other infectious, inflammatory or neoplastic systemic diseases as comorbidities, pregnancy and lactation. Patients with AIDS were excluded from the study. The exclusion criteria for patients with HST and CRC were AIDS, pregnancy and lactation.

Healthy individuals
The control group consisted of 200 healthy individuals (G6) who were blood donors of the Blood Center of Botucatu.

Study design
The study was designed to compare the accuracy parameters of two methods in the detection of serum antibodies (DID and ELISA) using two types of antigen preparations (A. fumigatus antigen and Aspergillus spp. antigen pool) and serum measurement of GM. The rate of crossreactivity with sera from patients with other infectious diseases that are part of the differential diagnosis for CPA was also determined. All evaluations were performed with sera collected when the patients were admitted.
In addition, the serological follow-up of patients with CPA under treatment was evaluated by comparing two serological methods (DID and ELISA) and two antigen preparations (A. fumigatus antigen and Aspergillus spp. antigen pool). Serum samples obtained before the treatment started and at three time points after it started were used in the study.

Score of symptoms and signs
Symptoms and signs were periodically evaluated. Every complaint was scored by the same physician (RSC) on a single 0-4 system, based on De Beule et al. (1988) specifications, with some modifications: 0-absent, 1-mild, 2-moderate, 3-intense, and 4-very intense. Cough, expectoration, dyspnea, and thoracic pain were scored using this schedule. The sputum was scored as without blood (score 0), blood-streaked sputum (score 1), gross blood sputum (score 2), a coffee cup of blood-up to 50 mL a day (score 3), and more than 50 mL of blood a day (score 4). Weight-loss was classified in relationship to the usual body weight as absent / score 0, mild / score 1 (<5%), moderate / score 2 (5-10%), intense /score 3 (between 11 and 20%), and very intense (>20%). Fever was scored in relationship to the axillary temperature: absent / score 0 (<37,0°C), mild / score 1 (37°-38°C), moderate / score 2 (38°C-39°C), intense / score 3 (39°C-40°C or >38°C more than 3 times per week), and very intense score 4 (>40°C or >39°C daily). The global score was the sum of the scores given to every complaint. The scores given at admission served as a baseline with which the subsequent evaluations were compared. A careful evaluation of the scores was performed to standardize the inter-patients analysis [24].

Antigens used
The following antigens, obtained by culture filtrate, were used: a) A. fumigatus (DID), pro- This study was carried out with serum samples from our serum bank at -80°C. DID was the immunodiffusion test performed with A. fumigatus antigen when each patient was clinically evaluated. Thus, different antigen batches were used along the years. When this study was performed, we used two different antigen preparations recently prepared; all the reactions were carried out with the same batch antigen. The comparison between the DID and DID 1 showed difference of intensity no higher than one dilution.

Determination of serum antibodies and GM level
Agar gel DID. The serum levels of anti-Aspergillus antibodies (A. fumigatus and antigen pool) were determined by DID, according to the specifications of Restrepo et al. [8,9], at the Tropical Diseases Research Laboratory of the School of Medicine of Botucatu-UNESP. The tests were performed upon the admission of each patient (DID) and when the present study was conducted (DID 1 and DID 2 ).
The tests were performed with undiluted serum followed by two-fold dilutions starting with 1:2. For each test, a positive and a negative control serum were included.
Detection of antibodies by the enzyme immunoassay (ELISA)-indirect method. The serum levels of anti-Aspergillus antibodies were determined by ELISA [25,26], according to the standard protocol of the Tropical Diseases Research Laboratory of the School of Medicine of Botucatu-UNESP. All tests were performed at the same place, and the samples were processed in duplicate.
The plates (NUNC-MaxiSorp) were pre-sensitized with 10 μL of A. fumigatus antigen and Aspergillus spp. antigen pool (A. fumigatus, A. flavus and A. niger) and incubated for 2 hours at 37°C and then for 18 hours in the refrigerator. Subsequently, five washes were performed with 300 μL of phosphate-buffered saline with 0.1% Tween (0.1% PBS-T: blocking buffer) in a plate washer (Biotek-Elx 50). The wells were then filled with 100 μL of PBS-T 0,1% with bovine serum albumin (BSA-Sigma), and maintained at room temperature for 1 hour. Next, the serum samples were diluted to 1:100, and 100 μL were added to the microplate and incubated for 1 hour at 37°C. Then, five washes were performed with 300 μL of the blocking buffer solution and goat anti-human immunoglobulin G (IgG) marked with peroxidase (Sigma) diluted to 1:3000; next, the plates were incubated for 1 hour at 37°C. Subsequently, five washes were performed with 300 μL of the sodium phosphate buffer solution with 0.1% Tween and 0.1% PBS-T, and the developing solution was prepared with 200 μL of tetramethylbenzidine (TMB) and 2 μL of 30% H 2 O 2 in 10 mL of citrate-acetate buffer at pH 6.0. A volume of 100 μL was added to each well, and the reaction was allowed to proceed for 30 minutes at 37°C in an oven. The reaction was interrupted with 50 μL of 4 N sulfuric acid solution. The spectrophotometric reading (Bio-Rad, Mark Microplate Reader) was performed at 450 nm. For each test, a positive and a negative serum were included.
GM detection by sandwich ELISA. Serum samples stored at -80°C were processed for sandwich ELISA (Bio-Rad Platélia Aspergillus EIA-62796) according to the manufacturer's instructions.
Serological follow-up of patients treated with itraconazole. The serological follow-up of 10 patients with CPA was evaluated using ELISA and DID. The clinical and serological evaluations were performed before the treatment started, which was defined as the 0 time (T 0 ), and periodically until the patients were clinically cured, called time 3 (T 3 ). Clinical cure was defined as the disappearance of the symptoms previously exhibited by the patient. Between these two times, the patient was evaluated at two other times, which were defined in relation to the number of weeks of treatment: T 1 : 4 to 6; T 2 : 7 to 10.
Determination of the cut-off point of the ELISA test. The cut-off point of the ELISA test was defined by the construction of the receiver operator characteristic (ROC) curve, using confidence intervals of 95% and 99%, according to the specifications of Fletcher & Fletcher [27] and Frei et al. [28].

Accuracy parameters
The accuracy parameters were determined in 22 patients with CPA and in 200 healthy individuals (controls) to determine the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy (Ac), positive likelihood ratio (PLR) and negative likelihood ratio (NLR), also called verisimilitude ratio, according to the specifications of Fletcher & Fletcher [27].
When calculating the likelihood ratios, when one of the frequencies was zero, this value was replaced by 1.0 and the likelihood ratio was called corrected positive or negative likelihood ratio (CPLR or CNLR) [29].

Cross-reactivity with sera from patients with other granulomatous diseases
The sera of 20 patients with CRC, 23 patients with HST, 28 patients with TB and 50 patients with PCM were used to test the cross-reactivity ratio with diseases that can be considered part of the differential diagnosis of CPA.

Ethics statement
The project was approved by the Institutional Ethics Research Committee (Process no. 210.781-CEP). The informed consent form was signed by the patients included in the prospective study.

Statistical analysis of results
The likelihoods in dependent populations were compared by Cochran's Q test, according to the specifications of Curi [30], followed by McNemar's test, according to Siegel [31]. The degree of agreement between both the two tests was evaluated using the kappa coefficient, according to the specifications of Landis et al. [32]. The kappa coefficient was interpreted as follows: (a) poor: when below 0.00; (b) slight: between 0.00 and 0.20; (c) fair: between 0.21 and 0.40; (d) moderate: between 0.41 and 0.60; (e) substancial: between 0.61 and 0.80; and (d) almost perfect: between 0.81 and 1.00 [32]. For each statistical test, the differences were considered significant when p 0.05.

Results
The survey of the 25 patients with CPA showed the presence of 20 cases of aspergilloma (confirmed-16; probable-1; possible-3), one with CCPA (probable) and four CFPA (confirmed -3; probable-1). The predominant clinical manifestations were cough (77.3%), weight loss (72.7%) and hemoptysis (63.6%). The CPA data were analysed according to the procedures performed, time between the underlying disease and CPA, concomitant diseases and use of antifungal compounds. In the present study, confirmed and probable cases were included, totalling 22 patients.
The procedure most frequently performed in these patients was transbronchial biopsy. The identification of Aspergillus spp. and pulmonary fibrosis were the most common histopathological findings. The time between the underlying disease and the diagnosis of CPA was highly variable, with a median of 5 years (range, 0-45 years). TB was the most prevalent underlying disease. Itraconazole was the antifungal compound used by most patients. Finally, there was a predominance of satisfactory responses to treatment ( Table 1).

Determination of the cut-off point
The cut-off points were an OD of 0.120 and 0.130 when the A. fumigatus antigen (A) was used and an OD of 0.090 and 0.100 with the Aspergillus spp. antigen pool (B) and confidence intervals of 95% and 99%, respectively (Fig 1).

Accuracy parameters
The use of an antigen pool led to a trend of increasing DID sensitivity of 14%, while the other accuracy parameters remained unchanged ( Table 2). A positive DID test increased by 90.9 to 118.2 times the likelihood of the pre-test diagnosis being aspergillosis, depending on the antigen used ( Table 2). The DID sensitivity in patients with aspergilloma (76.5%) was greater than that of patients with CPA, assessed as a whole. In the ELISA, using a higher cut-off point led to decreased sensitivity for both types of antigens but also to increased PPV and PLR (Table 2). In addition, the use of the antigen pool also led to increased PPV and PLR but decreased sensitivity when the cut-off was higher ( Table 2). The ELISA showed higher sensitivity than DID. Regardless of the antigen preparation used, a positive DID test increased much more the chance of pre-test diagnosis than did a positive ELISA test, as indicated by the PLRs ( Table 2). The indices, determined as the ratio between the optical density observed in every sample and the cut-off, were 0.173 and 0.420 in the negative samples and ranged from 0.63 to 5.72 for the positive ones (median = 2.08). Finally, using the higher cut-off, there was decreased sensitivity, which approached the one presented by DID, but with a slight increase in specificity.

Cross-reactivity
The detection of anti-Aspergillus antibodies by DID test in sera from patients with other granulomatous diseases was negative in almost all of them ( Table 3). The ELISA test showed a greater prevalence of cross-reactivity, which was lower when the highest cut-off point was used. Cross-reactivity was more frequent in sera from patients with PCM and almost absent in those with CRC ( Table 3).

Comparison of the sensitivity of the tests
The comparison of the sensitivity of the different diagnostic tests indicated higher frequencies for ELISA using both types of antigen and the lowest frequency for DID 1 with A. fumigatus antigen. GM and DID 2 showed intermediate frequency, which did not differ from the others ( Table 4). The agreement between the positivity of the GM test and the detection of antibodies was always small, regardless of the method and the antigen used. The strongest agreement was observed between the ELISA 1 and ELISA 2 . The agreement between the DID tests with different antigens and DID with ELISA was appreciable overall (Table 5).

Evaluation of the serological follow-up of patients under treatment with itraconazole
The regression curves obtained in the follow-up of patients with positive initial serology according to DID 1 and DID 2 exhibited the same pattern as a function of treatment time, with decreased serum levels (p<0.01). The curves obtained from ELISA were different from those observed with DID because they showed a slight increase in serum antibody levels, more evident when the A. fumigatus antigen was used (p<0.00001), as shown in Fig 2. The characterization of the 10 patients evaluated and treated showed clinical cure or improvement in all cases, but with persistence of aspergilloma in three patients ( Table 6). The mortality rate was 30%; two patients showed persistence of aspergilloma and the third, fibrotic scars (Table 6).

Clinical progress. Comparison with double agar gel immunodiffusion test
Clinical follow-up, evaluated by using complaints scores, showed evident clinical improvement after introduction of antifungal treatment, as it was observed with the decreasing antibodies serum levels, determined by DID test. However, the clinical curve was different from the DID test ones, taken together (Fig 3).  Discussion CPA has several clinical manifestations: aspergilloma, also called aspergilloma, CCPA and CFPA. The pulmonary sequelae caused by TB are an important antecedent of CPA [3,33]. Worldwide, more than 36 million people have been cured of TB from 1995 to 2008, and 9 million new cases per year have been diagnosed. The British Thoracic and Tuberculosis Association noted that 6% of patients with open-scar tuberculous cavities developed aspergillomas in three years, with a mortality of 6% per year [34][35][36]. In another study, the mortality rate was 31% in 5 years and 56% in 10 years; of the 27 patients who died, 3 of them died due to hemoptysis, 7 due to surgical complications, 6 due to chronic respiratory failure, 6 due to acute pneumonia and 5 due to chronic suppurative pneumonia [34]. In Brazil, in 2003, the incidence of TB was of 44.4 new cases/100,000 inhabitants. In 2013, 71,123 new cases were reported, with an incidence of 35.4 cases/100,000 inhabitants [35]. All these data indicate the importance of CPA for public health and the high number of patients with underlying diseases that favor its appearance. These data also suggest that the prevalence of CPA in our environment must be much higher than the number of cases referred to university hospitals and therefore that many of them have not been diagnosed. However, it should be noted that simple identification in the sputum of fungi that colonize the bronchial tree is not confirmation of the etiologic diagnosis of lung injury, which occurs with Cryptococcus and Aspergillus spp. The solution to this problem depends on surgical procedures, such as lung biopsy or resection of lung injury and serological tests. Surgical procedures are often aggressive and sometimes contraindicated. Serological tests can be performed using several methods and different antigen preparations. In addition, the presence of GM and IgG in the serum of these patients can be evaluated.
Thus, the present study aimed to compare the accuracy parameters of two serological tests (DID and ELISA), using two antigen preparations (A. fumigatus antigen and A. fumigatus, A. flavus and A. niger antigen pool) and qualitative detection of serum GM in patients with CPA. Moreover, we evaluated the serological progression of the patients after introduction of the antifungal treatment.
Several serological methods, such as complement fixation reactions, gel precipitation reactions, latex particle agglutination, electrophoretic tests and enzymatic immunoassays have been evaluated for the diagnosis of aspergillosis [10,37]. The immunodiffusion reaction in agar gel is the most used method due to the simplicity of execution, reproducibility [38], low prevalence of cross reactions and a high positive likelihood ratio, which are the reasons that led to its standardization for the detection of anti-Aspergillus antibodies in most clinical laboratories. However, the low sensitivity and the long period necessary to the reading of the slides are drawbacks of this method. However, the antigen concentration required to detect precipitins in the DID is a critical factor of this reaction [39], which can only be read 96 hours after the slides are prepared. With the advent of enzyme immunoassays, the introduction of ELISA test has been encouraged in routine laboratories, given its sensitivity, speed of execution and direct measurement of antibody levels [26,[37][38][39][40].
In the present study, the prevalence of patients with aspergilloma among those with CPA, the males, and those with cough, weight loss and hemoptysis as clinical manifestations of pulmonary TB as pre-existing disease confirms the findings of other authors [20,[34][35][36]. The use of serum diluted to 1/100 aimed to better use the serum of each patient, given that the results did not differ from those observed at 1/25 and 1/50 dilutions. The sensitivity of DID test observed in patients with aspergilloma in the present study (76.5%) was lower than that found by other authors (91.0 to 98.0%) [10,[41][42][43]. The finding of high specificity and PPV confirms those same studies [10,[41][42][43]. The PLR or verisimilitude ratio, which was not evaluated by other authors, was very high, which makes the post-test likelihood almost a certainty. This fact justifies defining a confirmed case in patients who have positive DID test with suggestive clinical and radiological manifestations.
The sensitivity of ELISA test was higher than that of DID test, although without reaching 95%, confirming the findings of other authors [10,[41][42][43]. The higher sensitivity of ELISA test can be explained by the identification of antibodies that do not precipitate on agar or agarose gel but are immobilized on polystyrene plates [37]. The PLR was high, although lower than that observed with DID test, thus strengthening the CPA pre-test hypothesis.
The antigen preparation used is essential to evaluate the accuracy of serological tests [10]. Because different preparations were used in different studies, comparisons cannot be direct. Nevertheless, the overall set of results can be considered harmonious. The use of an Aspergillus spp. antigen pool is well indicated because cases of aspergillosis by A. fumigatus, A. flavus, A. niger, A. nidulans and A. terreus have been described, though with a predominance of the first [10,[41][42][44][45].
The cross-reactivity measured in sera from patients with PCM, HST, TB and CRC were infrequent with DID test, present only with HST and PCM. In contrast, ELISA test showed false-positive results with HST, TB and PCM, and ELISA 2 test showed false positives with the four granulomatous diseases. These findings can be explained by the production of GM by Paracoccidioides brasiliensis, Histoplasma capsulatum and Cryptococcus neoformans [46][47][48][49][50][51][52]. Cross-reactivity with sera from patients with active TB must be carefully evaluated because TB is often the underlying disease of CPA and anti-Mycobacterium tuberculosis antibodies may persist for a long time.
The serum GM detection showed the same sensitivity as ELISA test. Considering that immunodiffusion tests present high specificity but low sensitivity, and the ELISA tests present higher sensitivity but an elevated percentage of cross-reactions with paracoccidioidomycosis and tuberculosis. Considering that tuberculosis is usually an underlying disease for chronic pulmonary aspergillosis, we decided to include galactomannan in this study. The results were very good, suggesting the determination of galactomannan in new studies. However, this finding must be carefully interpreted due to the possibility of cross-reactivity with other fungal diseases that also affect the lungs [46][47][48][49][50][51][52]. Interestingly, Aspergillus spp. produce GM, a complex polysaccharide composed of D-galactose and D-mannose, and C. neoformans also contains a capsular polysaccharide complex, composed of glucuronoxylomannan (GXM) and GMs. GXM is found in body fluids, and its identification allows the diagnosis of CRC [53]. Despite the apparent similarity between these polysaccharides, the cross-reactivity in sera from patients with CRC was much less frequent than in patients with HST or PCM.
The progression of the serum antibody levels evaluated by DID test was different from that of ELISA test, although both produced polynomial curves. While the curve shown in the DID evaluation was descending and with no difference depending on the antigen preparation, the curve produced by ELISA test was slightly ascending and behaved differently according to the antigen used. Considering that (a) DID test measures only precipitating antibodies in agar/agarose gel and ELISA test identifies non-precipitating antibodies that are immobilized on polystyrene plates [37] and that (b) patients with aspergilloma have increased serum IgG as determined by ELISA test [26], which continue to increase after the introduction of the antifungal agent, the behavior of the two serological progression curves after introduction of the treatment can be explained. This hypothesis can be tested by assessing the progression of the serological curve of anti-Aspergillus IgG, as currently there is a kit available for this determination, which will be performed soon. The decrease in serum antibody levels, determined by DID test, after the introduction of an effective treatment, was also observed by other authors [5,43,54,55]. Thus, based on the results available, the follow-up of patients must be performed using DID.
The present study has some limitations: a small sample, as a result of the small number of referred patients with the CPA hypothesis; the Aspergillus species was rarely identified; and, finally, the anti-Aspergillus IgG was not measured in the serum of these patients.