Abstract
Serum precipitins have a controversial diagnostic value in hypersensitivity pneumonitis (HP). The present authors’ objective was to assess their diagnostic value by developing scores from a panel of specific antigens tested by two techniques (electrosyneresis and double diffusion) to discriminate active HP from other interstitial lung diseases.
Consecutive patients presenting with a condition for which HP was considered in the differential diagnosis were included in the study. All patients underwent the same standardised diagnostic procedure, including precipitin tests performed in routine conditions. Clinical manifestations, bronchoalveolar lavage and high-resolution computed tomography defined the presence or absence of HP. Receiver-operating characteristic curves and logistic regression were used to develop the serological scores.
A total of 122 patients (including 31 cases of HP) were included in the study. Five antigens from the panel were selected for the serological scores (Absidia corymbifera, Eurotium amstelodami, Wallemia sebi, Saccharopolyspora rectivirgula and mesophilic Streptomyces sp.). Electrosyneresis was more discriminative than the double-diffusion technique. Predictive negative values varied 81–88% and predictive positive values varied 71–75% for prevalence of HP 20–35%.
In conclusion, serological scores using a panel of relevant antigens may guide both biological and clinical practice in areas of high prevalence of hypersensitivity pneumonitis.
Interstitial lung diseases (ILDs) represent a large group of diseases that includes hypersensitivity pneumonitis (HP). HP, also called extrinsic allergic alveolitis, is a syndrome that results from an immunologically induced inflammation of the lung parenchyma in response to inhalation of a large variety of antigens to which subjects have been previously sensitised. Despite its apparently low prevalence, the impact of HP in individuals of all ages throughout the world continues to be a major concern 1. The most common forms are farmers' lung (FL) and bird breeders'/fanciers' lung (BBL). Their prevalence rates and distribution vary widely among countries and geographical locations due to local customs and occupational and climatic conditions 2. Differential diagnosis between HP and other ILDs is difficult and relies on an array of signs or criteria 1, 3–5. Although nonconsensual, these diagnostic indicators generally include the following: 1) clinical symptoms and signs developed in an appropriate environment; 2) the presence of precipitating antibodies against offending antigens; 3) consistent chest radiograph or high-resolution computed tomography (HRCT); 4) bronchoalveolar lavage (BAL) lymphocytosis; 5) decreased carbon monoxide diffusion capacity; and/or 6) a granulomatous reaction on lung biopsies; and/or 7) a positive inhalation challenge. Among the above-mentioned indicators, the diagnostic value of serum precipitins has long been controversial 6, particularly because of their lack of sensitivity and, especially, of specificity 7. These limitations may be due to the use of poorly purified antigens, the lack of the specific inciting antigen in the test panel, inappropriate techniques 8 or, in epidemiological studies, the use of inappropriate control groups 5, 9, 10. Despite the pitfalls presented above, a recent multicentre study by Lacasse et al. 4 identified serum precipitins as a significant predictor of HP, regardless of exposure (odds ratio 5.3 (95% confidence interval (CI) 2.7–10.4)). However, the choice of antigens tested and the use and interpretation of tests were left to each investigating centre according to their usual local practice; thus, the diagnostic value of precipitins could not be assessed reliably. With the objective of developing serological scores to discriminate active HP from other ILDs, the present authors conducted a prospective cohort study of patients presenting in their department with a suspected diagnosis of HP. Subjects were tested by two serological techniques (electrosyneresis and double diffusion) for a panel of seven antigens. The choice of both immunological methods and the antigenic panel was based on previous studies aimed at identifying the best immunological set for diagnosing HP in the present study’s region 11, 12.
METHODS
Patients
From January 1, 1999, consecutive patients were recruited aged ≥18 yrs presenting in the Dept of Respiratory Diseases (University Hospital of Besançon, Besançon, France) with a pulmonary syndrome for which HP was considered in a differential diagnosis (patients diagnosed in 1999 were also included in the study by Lacasse et al. 4). The proportion of HP and other ILDs was unspecified a priori.
All patients underwent the same standardised procedure (discussed later). Only HP cases presumably due to moulds (fungi and thermophilic actinomycetes) were included, since BBL, chemical HP and hot-tub lung are due to antigens of a nature different from those of other HP caused by microorganisms.
Diagnostic criteria
Since patients were to be classified as either HP or non-HP (other ILD), and in the absence of a recognised gold standard, the following diagnostic criteria were used for HP according to those of a previous HP study 4. 1) Compatible clinical manifestations; 2) BAL lymphocytosis (≥30% for non- and ex-smokers and ≥20% for current smokers); and 3) bilateral ground glass or poorly defined centrilobular nodular opacities. When the association of BAL and HRCT did not yield a reliable diagnosis of HP, the presence of a positive specific provocation test 13, an improvement after avoidance of supposed antigens, or suggestive features on transbronchial biopsies was required. Patients underwent surgical lung biopsy when the previously described diagnostic procedure failed to yield a diagnosis. Classical histopathological descriptions were used 14. Patients considered to have residual HP (a chronic and inactive form in which BAL lymphocytosis has disappeared) were classified in the non-HP group. When HP was diagnosed, the patient's exposure to antigenic sources of HP was then specifically considered in order to determine the nature of HP, e.g. FL, other mould-induced occupational HP, home HP caused by moulds, etc.
When a complete diagnostic procedure set did not lead to a reliable classification (for example, difficulties in interpreting HRCT), cases were reassessed by an independent ILD expert (J-F. Cordier), who considered, in particular, BAL and HRCT data but also data on clinical history, physical examination, spirometry and diffusing capacity of the lung for carbon monoxide (or the results of a 6-min walk test) to reach a diagnosis. If diagnosis remained inconclusive, patients were excluded.
Precipitins tests
The choice of antigenic panel was based on the recognised causes of FL and HP due to moulds at an international 15, 16 and local level 11. The antigenic panel tested included Absidia corymbifera, Eurotium amstelodami, Wallemia sebi, Aspergillus fumigatus, Saccharopolyspora rectivirgula, mesophilic Streptomyces sp. and hay extract. Antigen extracts were produced and tested as described in a previous article 11. The immunological methods used were agar gel double diffusion and electrosyneresis on cellulose acetate, which, as shown by a recent study, are more relevant than ELISA and immunoblots in discriminating FL from healthy exposed dairy farmers in the present study’s region 12. Briefly, electrosyneresis was carried out as follows. Cellulose acetate sheets (Sartorius, Göttingen, Germany) were placed in the electrophoresis vat filled with buffered Tris glycin solution at pH 8.8; 15 µL of each serum was placed on three spots on the anode side and a 15-µL line of antigen was placed on the cathode side. A current (110 V) was then applied for 75 min and the sheets were stained with Coomassie blue. Serological tests were performed and interpreted (by counting the number of arcs) under routine conditions by four different experienced technicians and a supervisor blinded to the patients’ diagnoses and exposure.
Statistical analysis
Characteristics of HP and non-HP patients were compared using the two-tailed Fisher's exact test for dichotomous variables and the unpaired t-test for continuous variables.
The serological results for each antigen were interpreted as a number (xa) of precipitating arcs (discrete quantitative variable). They were used, either as discrete (xa) or dichotomous variables (xd), to develop serological scores. The selection of antigens for multivariate analysis was based on receiver-operating characteristic (ROC) curves to each antigen 17. Efficiency was used to determine the positivity threshold of each antigen. Serological scores were then constructed for electrosyneresis and double diffusion using multivariate logistic regression (full model) and stepwise logistic regression (predictive model). Coefficients (β) were estimated with their 95% CIs by multivariate analysis. Scores (Sa and Sd) with antigenic results as discrete or as dichotomous variables (Xd: positive or negative) were expressed respectively, as follows.
Sa = β0+∑βaXa, with Xa = number of precipitating arcs for each antigen tested (1)
Sd = β0+∑βdXd, with Xd = 0 (negative) or 1 (positive) for each antigen tested (2)
The stability of the models was studied using the bootstrap method 18. In total, 1,000 random bootstrapped samples (size = study sample size) were generated, and a robust calculation of 95% CIs and coefficient of variation (CV) of estimated parameters (β) was obtained.
ROC curves were constructed and areas under the curves (AUC) were compared for electrosyneresis and double diffusion in order to determine the best immunological method.
Thresholds of serological score positivity for different HP prevalence were determined on ROC curves 17, and their associated predictive values were calculated. The serological score results, expressed as either positive or negative, were then studied in association with patients’ exposure and smoking status.
Estimated accuracy of serological scores
With the hypothesis of sensitivity (Se) and specificity (Sp) values between 70–90% and an HP expected prevalence of 35% (in the present authors’ centre), 120 patients needed to be included in the study in order to obtain an accuracy of ∼10% for Se and Sp.
RESULTS
Patients
Between January 1999 and April 2005, 162 patients were enrolled in the present study. Of these, 40 were excluded for the following reasons: BBL (n = 3), hot-tub lung (n = 2); unavailable or uninterpretable BAL (n = 3); serological results unavailable (n = 30) and indeterminate diagnosis (n = 2). The independent investigator reassessed eight cases, three of which were classified in the HP group and four in the non-HP group; one was inconclusive. Thus, 122 patients were included, comprising 31 HP (22 FL, three HP due to nonprofessional farming activities (sheep and horses), two home HP due to moulds, one HP due to nonprofessional sawing (mould from sawdust), one cheese-workers' lung, and two HP with no identified exposure) and 91 non-HP ILDs (41 idiopathic interstitial pneumonia (33 idiopathic pulmonary fibrosis, three cryptogenic organising pneumonia, two acute interstitial pneumonia, two respiratory bronchiolitis ILD, one desquamative interstitial pneumonia), 18 sarcoidosis, seven drug-induced interstitial pneumonia, four pneumoconiosis, four eosinophilic interstitial pneumonia, three collagen vascular disease, three left heart failure, two histiocytosis, two neoplasic interstitial pneumonia, two infectious interstitial pneumonia, two residual form of HP and three undetermined ILD).
Patients from the HP group were significantly younger, less frequently smokers and more frequently exposed than the patients from the non-HP group (table 1⇓).
Selection of candidate antigens for serological scores
ROC curves for A. fumigatus and hay extract were respectively nondiscriminative (AUC values 0.54–0.55) and uninterpretable. Therefore, threshold values for these antigens were not determined and they were not used in multivariate analysis to develop serological scores. Positivity thresholds for each antigen for the two immunological methods are indicated in table 2⇓.
Development of serological scores
ROC curves associated with electrosyneresis scores were more discriminative than those associated with double diffusion, regardless of the type of variables (fig. 1⇓). Similarly, full models appeared to be more relevant than predictive models. Only results concerning electrosyneresis with the full model are presented. Estimated coefficients used to calculate serological scores and their CIs are indicated in table 3⇓. Coefficients of variation from bootstrap re-sampling showed a moderate stability of estimated parameters (table 3⇓).
Performances of serological scores
The positivity thresholds of serological scores according to different prevalence rates (20, 25 and 35%) of HP and their diagnostic values for electrosyneresis are presented in table 4⇓. Performances of scores with discrete variables were equivalent to those with dichotomous variables for negative predictive values (NPV; 81%–88%) but better in terms of positive predictive values (PPV). In full models, PPV varied 71–75% with discrete variables and 57–64% with dichotomous variables. If serological tests were considered as positive when at least one of two scores was positive, Sp did not change but Se was higher. For a HP prevalence of 35%, Se was 76% (95% CI 0.60–0.92), Sp 82% (95% CI 0.74–0.90), PPV 69.5 and NPV 86.4.
After adjustment by disease group, a positive serological result, with discrete variables, was positively associated with exposure (p = 0.032). However, 20% of subjects with precipitins had no identified exposure and 25% of subjects without precipitins had exposure. The presence of precipitins tended to be negatively associated with tobacco consumption (p = 0.08; table 5⇓).
Eventually, eight false positives (8.8%) and 11 false negatives (33.3%) were obtained. The false positives were as follows. Three idiopathic pulmonary fibrosis (nonexposed), one eosinophilic interstitial pneumonia (exposed farmer), one silicosis (exposed), one left heart failure (nonexposed) and two residual FL (both exposed farmers). The false negatives comprised eight FL, one home HP due to moulds in the walls, one HP due to nonprofessional farming activities (horses) and one HP without identified exposure. The home HP subject was tested for the antigens corresponding to the actual exposure, identified by microbiological sampling in the subject’s home. This additional investigation showed at least two arcs of precipitation with electrosyneresis for Fusarium verticilloides, Cladosporium sp. and Trichoderma sp. The same investigations were carried out for the nonprofessional farming HP female and Saccharomonospora viridis was identified as the probable responsible antigen (four arcs of precipitation). Six out of the eight false-negative FL results had at least two arcs for hay extracts and a significant number of arcs with electrosyneresis for at least one of the present study’s panel of antigens (without reaching a positive score), and the presence of 1–3 arcs of precipitation for at least two of the following antigens that were not in the present study’s panel: Aspergillus nidulans, Streptomyces griseoflavus, Thermoactinomyces vulgaris, S. viridis, Rhodotorula rubra, Fusarium solani and Penicillium sp.
DISCUSSION
In the absence of standardised and validated diagnostic criteria, diagnosis of HP is difficult 1. Several proposals for diagnostic criteria have been published 19–22 but their diagnostic accuracy has not been tested. Among these criteria, serology has long been considered to be either a minor criterion or a simple marker of exposure 23–26. However, with typical FL patients, the use of antigens isolated from patients' environments has lead to convincing results 11, 27. In addition, a recent study by Lacasse et al. 4 identified serum precipitins as a significant predictive factor of HP, regardless of exposure. However, the diversity of antigens and immunological techniques used in this multicentre study did not allow evaluation of the real diagnostic value of precipitins. The HP cohort in the study of Lacasse et al. 4 consisted mainly of BBL disease and subjects were generally tested by ELISA. In the present study, which concerns HP due to moulds, the sera of patients were tested by electrosyneresis and double diffusion using a panel of antigens that included A. corymbifera, E. amstelodami, W. sebi, three antigens responsible for FL in the region of the present study and rarely considered in other studies. The use of local antigens limits the feasibility of a multicentre study, and it is for this reason that the number of cases in the present study is limited.
The results of the present study showed that it is possible to use a serological score in the diagnosis of HP caused by moulds for patients suffering from an ILD, when an appropriate panel of antigens is used. The results also showed electrosyneresis to be more appropriate than double diffusion for differential diagnosis of HP. However, the sensitivity of the serological test was somewhat disappointing, since about one-third of HP had negative results. However, in eight of the 11 false-negative results for the present study’s serological score, the responsibility of antigens not included in the tested panel was either suggested or demonstrated by further antigenic testing. The lack of sensitivity in scores can therefore be explained by the limited panel tested with respect to the great diversity of antigens that cause HP. The use of a larger panel of antigens would improve the sensitivity of the methodological approach.
The most discriminative antigens within the tested panel were S. rectivirgula, A. corymbifera, E. amstelodami and W. sebi; the first is recognised as the major aetiological antigen of FL 27 and is responsible for other professional HP due to actinomycetes. The three others are likely to be aetiological factors of FL in Eastern France 11; however, A. corymbifera and W. sebi have also been shown to be aetiological agents in FL in Scandinavia 28, 29. Conversely, the present study, as well as the Scandinavian studies 28, 29, have not demonstrated the ability of A. fumigatus to discriminate between HP and non-HP. This antigen has been evoked as an aetiological factor of HP for more than four decades 30 but its actual responsibility in HP is probably rare.
One concern of the present authors is the applicability of the results of the current study. Both the panel of antigens used and the score are not strictly applicable worldwide, mainly because the current ILD cohort includes a high proportion of HP cases that do not represent the usual aetiologies of HP. However, the approach used, which consisted of developing a serological score through use of a combination of local and international antigens, can be applied worldwide. In addition, it would be useful to test the current panel in other parts of the world. Indeed, this panel probably allows detection of immunological responses induced by a larger number of fungal species and thermophilic actinomycetes, given the known cross-reactivity between different fungal species. In this respect, the present study partially answers the need to develop panels of standardised antigens, highlighted in the recent recommendations of the National Heart, Lung and Blood Institute/Office of Rare Diseases workshop 1.
Conclusion
Serological tests using a panel of relevant antigens and performed by electrosyneresis help to diagnose mould-induced hypersensitivity pneumonitis by providing precise estimates of the probability of active disease in a region where hypersensitivity pneumonitis, especially farmers' lung, is frequent.
Acknowledgments
The authors thank Y. Lacasse for his comments on the study protocol, N. Richardson-Peuteuil for proofreading and G. Rival for his assistance in collecting data.
- Received January 4, 2006.
- Accepted December 12, 2006.
- © ERS Journals Ltd