Abstract
Rationale: Children with severe asthma experience persistent symptoms despite maximal conventional treatment. Fraction of exhaled nitric oxide (FeNO) and sputum eosinophils are used as markers of airway inflammation to guide treatment with steroids, but no data are available on how reliable they are in predicting airway eosinophilia assessed bronchoscopically in these children.
Objectives: To determine how FeNO and sputum eosinophils predict airway eosinophilia measured in both bronchoalveolar lavage (BAL) and endobronchial biopsy.
Methods: Twenty-seven children with moderate to severe persistent asthma attempted measurement of FeNO and sputum eosinophils, followed by bronchoscopy, BAL, and endobronchial biopsy within 24 h.
Main Results: Significant correlations were found between eosinophils in sputum and both BAL eosinophils (n = 20, r = 0.45, p = 0.045) and FeNO (n = 23, r = 0.42, p = 0.049). The relationship between FeNO and BAL eosinophils was also significant with a stronger correlation (n = 24, r = 0.54, p = 0.006). The positive predictive value (PPV) for increased sputum eosinophil percentage (> 2.5%) to detect elevated eosinophils in BAL (> 1.19%) was 75%; the negative predictive value (NPV) was 63%. All patients with both increased sputum eosinophils and an elevated FeNO value (> 23 ppb) had elevated eosinophils in BAL (PPV, 100%); the NPV of these two markers was 65%. Eight of nine patients without any sputum eosinophils had normal subepithelial eosinophil numbers (< 1.2%; NPV, 89%). However, the PPV of any sputum eosinophils for increased subepithelial eosinophilia was only 36.4%.
Conclusions: There was moderate agreement between both FeNO and sputum eosinophils and BAL eosinophils. There was good NPV, but only poor PPV for these markers for mucosal eosinophilia.
Asthma is characterized by eosinophilia in sputum, in bronchoalveolar lavage, and in the bronchial wall, but the relationship between the three compartments is not known.
We show that, whereas bronchoalveolar lavage and sputum eosinophil counts are very similar, they do not relate to airway wall eosinophilia; it is unclear which is clinically important.
Endobronchial biopsy and bronchoalveolar lavage (BAL) have traditionally been used for assessment of airway inflammation. However, bronchoscopy is not practical for most children or most centers. Even in specialist centers, bronchoscopic assessment of airway inflammation cannot be performed repeatedly. Longitudinal studies that have investigated the role of markers of eosinophilic airway inflammation to guide therapy have therefore used noninvasive methods, such as FeNO and induced sputum. Previously, a number of research groups have demonstrated an association between FeNO and airway eosinophils, measured bronchoscopically either using BAL or endobronchial biopsy, in children and adolescents with asthma (4–6). However, to our knowledge, there are no complete data in this age group investigating the relationship between noninvasive markers (induced sputum eosinophils, FeNO) and eosinophils in either BAL or endobronchial biopsy. This is important because if the aim of treatment is to prevent bronchial inflammation, incorrect use of noninvasive markers could lead to the wrong doses of inhaled steroids being prescribed. The hypothesis of this study is that FeNO and sputum eosinophils are closely related to BAL and mucosal eosinophilia.
The aim of this study was therefore to investigate the relationships between FeNO and eosinophils in induced sputum, BAL, and the bronchial subepithelium in a group of children and adolescents with asthma undergoing bronchoscopy as part of their clinical evaluation. This involves assessment of lung function, FeNO, and sputum cytology, followed by a bronchoscopy, BAL, and endobronchial biopsy. Most children follow our difficult asthma protocol, in which measurements are done after a course of systemic corticosteroids, usually prednisolone 40 mg daily, including noninvasive assessments of inflammation at the beginning of this course. The full details are discussed elsewhere (7), and given in more detail in the online supplement. We specifically aimed to determine how often FeNO or sputum eosinophil values miscall BAL or endobronchial eosinophils. Some of the results of this study have been previously reported in the form of an abstract (8).
We recruited 27 children and adolescents with moderate to severe persistent asthma who underwent bronchoscopy, including endobronchial biopsy, as part of their clinical assessment (7, 9). Asthma was diagnosed according to American Thoracic Society (ATS) guidelines (10). Moderate to severe persistent asthma was characterized by “daily symptoms despite > 400 μg of inhaled budesonide (or equivalent)” (11). Patient characteristics are shown in Table 1.
Number | 27 |
| Age, yr, mean (range) | 11.4 (5–15) |
| Male | 14 |
| Atopic* | 20 |
| Baseline FEV1, % predicted† | 73 (65–87) |
| Post-bronchodilator FEV1, % predicted | 84 (79–91) |
| Treatment in previous 2 wk | |
| Inhaled steroid dosage (budesonide equivalent), μg/d | 2,000 (1,000–3,000) |
| Budesonide, n | 8 |
| Fluticasone, n | 19 |
| Oral or intramuscular steroids | 23 |
| 2-wk course of prednisolone, 40 mg/d | 10 |
| Intramuscular triamcinolone, 80 mg | 9 |
| Short course of prednisolone because of worsening of symptoms | 2 |
| Maintenance oral prednisolone | 2 |
| Long-acting bronchodilator | 27 |
| Montelukast | 16 |
| Theophylline | 2 |
| Use of rescue bronchodilator in previous 2 wk* | |
| None | 3 |
| 1–2 times/wk | 2 |
| 3–6 times /wk | 4 |
| ∼ Daily | 16 |
Subjects underwent measurement of FeNO, spirometry before and after bronchodilator, and sputum induction, followed by bronchoscopy, BAL, and endobronchial biopsy within 24 h. A questionnaire was completed about recent asthma symptoms, use of rescue bronchodilator, and maintenance treatment. Atopy was diagnosed if serum specific IgE was raised (> 0.34 kU/L) or skin-prick tests were positive (wheal > 2 mm larger than negative control) to one or more antigens (Dermatophagoides pteronyssinus, cat, dog, grass pollen, Aspergillus fumigatus). The study was approved by the local ethics committee, and written, informed consent was obtained from all parents and children recruited into the study.
FeNO was measured using a chemiluminescence analyzer (NIOX; Aerocrine, Stockholm, Sweden) according to European Respiratory Society/ATS guidelines (12). The upper limit of normal for FeNO was taken as 23.0 ppb (13).
Spirometry before and after bronchodilator and sputum induction and processing was performed as previously described (14). For subjects with a post-bronchodilator FEV1 less than 65% predicted, sputum induction was performed with 0.9% instead of 3.5% saline. A normal sputum eosinophil percentage was defined as approximately 2.5% (14).
Bronchoscopy was performed as previously described (9). For the BAL, three aliquots of normal saline (1 ml/kg, maximum 40 ml) were instilled in the right middle lobe through the bronchoscope and the fluid retrieved by mechanical suction. Up to four endobronchial biopsies were taken from the subsegmental carinae of the right lower lobe using cupped forceps. Biopsies were initially fixed in formal saline.
BAL processing is described in the online supplement. A normal BAL eosinophil percentage was defined as less than 1.19% (15).
Endobronchial biopsies were processed into paraffin blocks. Five-micrometer sections were cut and stained with hematoxylin and eosin to assess morphology. Suitable sections (those with visible epithelium, reticular basement membrane, and subepithelium) were stained for eosinophils using EG2 antibodies as previously described (16). Positively stained cells were quantified using the validated stereologic technique of point counting as previously described (16), and expressed as volume density (%). For each subject, one section from between one to four biopsies was analyzed. A normal eosinophil volume density was defined as less than 1.2%, based on previous data from nonasthmatic control subjects (5, 16). All measurements were made independently and investigators were blinded to other results.
All analyses were performed using the Statistical Package of the Social Sciences (SPSS for Windows, release 12.0; SPSS, Inc., Chicago, IL). Correlations were measured using Spearman rank correlation. A p value less than 0.05 was considered statistically significant. Sensitivity, specificity, and positive (PPV) and negative (NPV) predictive values were calculated, and receiver operator characteristic (ROC) curves constructed according to Altman and Bland (17–19).
Details of the subjects are shown in Table 1. Seventeen children were approximately 12 yr old; eight of these children were male and nine were female. Nineteen patients were included in our difficult asthma protocol, receiving either a course of oral corticosteroids or a single intramuscular injection of triamcinolone 2 wk preceding the study (7). There were no significant adverse events during sputum induction and bronchoscopy. We found anatomic abnormalities in 3 of 27 (11%) patients: a short right main bronchus with an abnormal branching pattern in one patient and tracheomalacia in two patients. Six patients had positive BAL cultures (Staphylococcus aureus, n = 4; Haemophilus influenzae and Streptococcus pneumoniae, n = 1 each).
FeNO could be measured in all subjects. The median (interquartile range [IQR]/absolute range [AP]) value was 20.3 (9.6–46.3/4.2–100.1) ppb. Eleven of 27 subjects had elevated values. Sputum induction was successful in 23 of 27 (85%) subjects. The median (IQR/AP) eosinophil percentage was 0.5% (0–2.5/0–23.25%). Detectable eosinophils were found in 12 subjects, of whom 5 had elevated values.
A satisfactory BAL cytospin was obtained in 24 of 27 (90%) subjects (median [IQR/AP] eosinophils: 1.0% [0.3–2.5/0–14.0%]). Ten subjects had BAL eosinophils above the normal range. Subepithelial eosinophils could be measured in 22 subjects. Five had values above the normal range.
There was a significant negative correlation between age and pre- and post-FEV1 (r = −0.64, p = 0.001, and r = −0.61, p = 0.002, respectively) but not between age and FeNO, sputum eosinophils, or BAL or biopsy eosinophils.
Pre- or post-bronchodilator FEV1 alone did not correlate with any of the other parameters (FeNO, or sputum, BAL, or biopsy eosinophils). However, bronchodilator reversibility correlated significantly with the percentage of sputum eosinophils (r = 0.473, p = 0.035), but not with FeNO or BAL or biopsy eosinophils. There was no significant difference in pre- or post-bronchodilator FEV1, FeNO, or sputum, BAL, or biopsy eosinophils whether patients had been using rescue bronchodilator at least daily or less than daily.
There was 61% concordance (n = 14) between the two markers. Thirty percent of patients (n = 7) had a raised FeNO but no elevation in sputum eosinophils; 9% (n = 2) had raised sputum eosinophils but normal FeNO (Table E1 of the online supplement). There was a positive significant correlation between sputum eosinophils and FeNO (n = 23, r = 0.42, p = 0.049; Figure 1).
Figure 1. Spearman's correlations (r) between sputum eosinophils and FeNO. Open symbols indicate patients without systemic steroids. The upper limit of normal is shown by the dashed line.
[More] [Minimize]There was agreement between the two techniques in only 58% of children (n = 11). BAL eosinophilia with no biopsy eosinophilia was present in 32% (n = 6), whereas a positive biopsy with negative BAL was found in 10% (n = 2). There was no significant correlation between biopsy and BAL eosinophils (n = 19, r = 0.36, p = 0.10; Figure 2).
Figure 2. Spearman's correlations (r) between biopsy and bronchoalveolar lavage (BAL) eosinophils. Open symbols indicate patients without systemic steroids. The upper limit of normal is shown by the dashed line.
[More] [Minimize]The concordance between any noninvasive marker and any invasive marker ranged between 59 and 75% (Table E1). There was no consistent pattern whereby one marker consistently over- or underpredicted the other.
There was a significant relationship between sputum and BAL eosinophils (n = 20, r = 0.45, p = 0.045; Figure 3). There was also a positive relationship between FeNO and BAL eosinophils and the correlation between both markers was even stronger (n = 24, r = 0.54, p = 0.006; Figure 3). We analyzed how well the presence or absence of elevated sputum eosinophil percentages or increased or normal FeNO values predicted the presence or absence of eosinophils in BAL (Table 2). The PPVs were 75 and 64% and the NPVs were 63 and 77%, respectively. We also analyzed the predictive values for combining elevated sputum eosinophils and raised FeNO values to predict increased BAL eosinophils. Because all children with both elevated FeNO values and increased sputum eosinophil percentages had increased BAL eosinophils, the PPV was 100%. However, the NPV was 65% for this combination. Figure 3 also shows the ROC curves of FeNO and sputum eosinophils for prediction of BAL eosinophilia at various cutoff points. The area under the curve (AUC) was 0.850 for FeNO (p = 0.004) and 0.747 for sputum eosinophils (p = 0.063).
Figure 3. (A) Spearman's correlation between BAL eosinophils and FeNO (for A–B, open symbols indicate patients without systemic steroids; the upper limit of normal is shown by the dashed line). (B) Spearman's correlation between BAL eosinophils and sputum eosinophils. (C) Receiver operator characteristics (ROC) curve for FeNO for predicting BAL eosinophilia. (D) ROC curve for sputum eosinophils for predicting BAL eosinophilia.
Sensitivity (%) | Specificity (%) | Positive Predictive Value (%) | Negative Predictive Value (%) | |
|---|---|---|---|---|
| Sputum eosinophils | ||||
| > 0% | 78 | 64 | 64 | 78 |
| > 2.5% | 33 | 91 | 75 | 63 |
| FeNO > 23 ppb | 70 | 79 | 64 | 77 |
| Sputum eosinophils > 0% and FeNO > 23 ppb | 44 | 91 | 80 | 67 |
| Sputum eosinophils > 2.5% and FeNO > 23 ppb | 33 | 100 | 100 | 65 |
There were no significant correlations between biopsy eosinophils and any of the other two parameters (sputum eosinophils [n = 20, r = 0.36, p = 0.12] and FeNO [n = 22, r = 0.07, p = 0.76]; Figure 4). Eight of nine patients without any sputum eosinophils had subepithelial eosinophils within the normal range (NPV, 89%). However, only 4 of 11 patients who had detectable sputum eosinophils had elevated biopsy eosinophils (PPV, 36.4%). The PPV and NPV for FeNO were similar to those of sputum (see Table 3). Figure 4 also shows the ROC curves for FeNO and sputum eosinophils at various cutoff points. The AUC was 0.624 for FeNO (p = 0.411) and 0.600 for sputum eosinophils (p = 0.513).
Figure 4. (A) Spearman's correlation between subepithelial eosinophils and FeNO (for A–B, open symbols indicate patients without systemic steroids; the upper limit of normal is shown by the dashed line). (B) Spearman's correlation between subepithelial eosinophils and sputum eosinophils. (C) ROC curve for FeNO for predicting subepithelial eosinophilia. (D) ROC curve for sputum eosinophils for predicting subepithelial eosinophilia.
[More] [Minimize]Sensitivity (%) | Specificity (%) | Positive Predictive Value (%) | Negative Predictive Value (%) | |
|---|---|---|---|---|
| Sputum eosinophils | ||||
| > 0% | 80 | 53 | 36 | 89 |
| > 2.5% | 20 | 73 | 20 | 73 |
| FeNO > 23 ppb | 60 | 59 | 30 | 83 |
| Sputum eosinophils > 0% and FeNO > 23 ppb | 40 | 80 | 40 | 80 |
| Sputum eosinophils > 2.5% and FeNO > 23 ppb | 20 | 80 | 0 | 71 |
To our knowledge, this is the first pediatric study to investigate predictive values of both FeNO and sputum eosinophils for airway eosinophils assessed bronchoscopically, using BAL and endobronchial biopsy. The principal findings of this study are as follows: first, FeNO is frequently elevated when sputum eosinophils are normal; second, there is moderate agreement between both FeNO and sputum eosinophils and BAL eosinophils; third, normal eosinophils on endobronchial biopsy were predictable from either FeNO or sputum eosinophils. The data suggest that an increased sputum eosinophil percentage is a reasonable predictor of the presence of airway eosinophils in BAL (PPV, 75%), but a normal sputum eosinophil percentage is less good at ruling out the presence of BAL eosinophils (NPV, 63%). Considering elevated sputum eosinophils and increased FeNO together improves the PPV to 100%, but not the NPV, although the numbers are small. The ability to predict the presence of biopsy eosinophilia from sputum eosinophilia or FeNO is poor (PPV, 20 and 30%, respectively); in contrast, biopsy eosinophilia is unlikely when there are no sputum eosinophils or a normal FeNO value (NPV, 89 and 83%, respectively).
The subjects recruited to this study represent a highly selected subgroup of children and adolescents with problematic asthma, and the number of subjects studied is relatively small. It would have been valuable to try to correlate our findings with asthma severity or treatment, but the homogeneous severity of the group, and the difficulty of making comparisons between patients on prednisolone and triamcinalone, precluded us from performing this analysis. The findings cannot therefore be generalized to all patients with asthma. However, in practice, patients attending a tertiary center are those who will likely derive most benefit from (and have access to) regular monitoring of sputum eosinophils and FeNO. The data are therefore of particular importance for clinicians working in specialist centers.
The majority of subjects underwent assessment of eosinophilic airway inflammation after treatment with high-dose systemic corticosteroids. This means that the relationships described here may have been altered by steroid therapy, and may not be relevant in steroid-naive patients with asthma, or those inhaling only a low dose of corticosteroids. Furthermore, the overall level of airway eosinophilia, for all methods used, was low and the range was narrow. In particular, over half the subjects had no detectable eosinophils present in their biopsy. Assessment of correlation between biopsy eosinophils and the other parameters is therefore difficult to interpret. However, the lack of any significant positive correlation between biopsy eosinophils and sputum eosinophils is in keeping with data from adult studies (20–22).
Of more practical relevance, and the major strength of this study, is that the data shed light on the relationships between eosinophilic airway inflammation assessed noninvasively and bronchoscopically. The combination of sputum eosinophils and FeNO may help to identify, with confidence, patients with evidence of airway eosinophilia in BAL. However, it was not possible to predict the presence of biopsy eosinophilia—only its absence. The presence or absence of eosinophilia is of practical importance. Absence of eosinophils leads to consideration of corticosteroid reduction. Persistent eosinophilia with symptoms leads to intensification of corticosteroid therapy, or the addition of a steroid-sparing agent such as cyclosporine. It should be stressed that we do not know, nor does this study elucidate, which measure of eosinophilia is most useful in guiding therapy. However, it should be noted that this study was cross-sectional, and thus the significance of the bronchoscopic findings could not be assessed.
It appears from the current study that the relationship between sputum and either BAL or biopsy is less in children than in adults with asthma (20–22). The percentage of sputum eosinophils may be due to the smaller pediatric sample sizes, which may be less representative for the complete airways. However, it is clear from the current study and from previous studies involving adults (20–22) that the different techniques described sample different airway compartments. The relationship between eosinophils in sputum and BAL is consistently stronger than that between sputum and biopsy. A possible explanation for this is the fact that both sputum and BAL sample inflammatory cells from a much larger area of the airways than is possible with biopsy. Another explanation may be that eosinophils traverse from the peripheral blood into the airway lumen far distal from the major airways that are sampled by biopsy, and thus are not seen in the proximal airways.
The data from the current study cannot be used to determine which is the most appropriate method for measuring eosinophilic airway inflammation in children and adolescents with difficult asthma. We do not know whether intraluminal or intramural eosinophils are more important in causing or monitoring this disease, and this study cannot answer that question. It is possible that clearing infiltrating eosinophils from within the airway wall, as assessed by biopsy, might even require lower steroid doses or a shorter treatment period than that used in this study, as we had fewer patients with biopsy eosinophilia than patients with BAL eosinophilia in our study. In clinical practice, noninvasive methods are more acceptable for monitoring of inflammation over time, with invasive methods more suited to one-off investigations of airway pathology in children with loss of symptom control despite high doses of inhaled and/or systemic steroids. The data from the current study help to highlight the similarities and differences between the methods described. However, if, as seems likely, targeting asthma therapy to reduce airway eosinophilia is desirable (1), then we need better markers than FeNO and induced sputum if this is to be optimally achieved. For example, sputum cytokine pattern, perhaps in combination with FeNO and sputum eosinophils, may be shown to be useful in future. Until then, children with difficult asthma who are on high doses of steroids may benefit from endobronchial biopsy to prevent overtreatment. However, it must be stressed that serial endobronchial biopsy is unlikely ever to be acceptable. Furthermore, the significance of proximal airway mucosal eosinophilia has been called into question by a recent study demonstrating that patients with asthma in apparent complete remission had ongoing airway wall eosinophilia to a similar degree to that of symptomatic patients with asthma (23).
In summary, these data show that, in a group of highly selected children on high-dose inhaled and systemic steroids, FeNO and induced sputum do not completely reflect but are a pointer to airway eosinophilia as measured by BAL or endobronchial biopsy. Furthermore, BAL and endobronchial biopsy themselves seem to measure inflammation in different compartments. However, the study shows a relationship between FeNO, sputum eosinophils and BAL eosinophils. A combination of elevated sputum eosinophils and elevated FeNO levels may be used as a predictor for BAL eosinophilia. An absence of sputum eosinophils or a normal FeNO level may help to exclude mucosal eosinophilia measured in endobronchial biopsy. However, better noninvasive markers are needed to guide antiinflammatory therapy, at least in this difficult group of patients, if the goal of treatment is to reduce airway wall eosinophilia, rather than merely treat symptoms.
| 1. | Green RH, Brightling CE, Hargadon B, Parker D, Bradding P, Wardlaw AJ, Pavard ID. Asthma exacerbations and sputum eosinophil counts: a randomised controlled trial. Lancet 2002;360:1715–1721. |
| 2. | Pijnenburg MW, Bakker EM, Hop WC, De Jongste JC. Titrating steroids on exhaled nitric oxide in children with asthma: a randomized controlled trial. Am J Respir Crit Care Med 2005;172:831–836. |
| 3. | Zacharasiewicz A, Wilson N, Lex C, Erin EM, Li AM, Hansel T, Khan M, Bush A. Clinical use of noninvasive measurements of airway inflammation in steroid reduction in children. Am J Respir Crit Care Med 2005;171:1077–1082. |
| 4. | Warke TJ, Fitch PS, Brown V, Taylor R, Lyons JD, Ennis M, Shields MD. Exhaled nitric oxide correlates with airway eosinophils in childhood asthma. Thorax 2002;57:383–387. |
| 5. | Payne DN, Adcock IM, Wilson NM, Oates T, Scallan M, Bush A. Relationship between exhaled nitric oxide and mucosal eosinophilic inflammation in children with difficult asthma, after treatment with oral prednisolone. Am J Respir Crit Care Med 2001;164:1376–1381. |
| 6. | van den Toorn LM, Overbeek SE, de Jongste JC, Leman K, Hoogsteden HC, Prins JB. Airway inflammation is present during clinical remission of atopic asthma. Am J Respir Crit Care Med 2001;164:2107–2113. |
| 7. | Payne D, Bush A. Phenotype-specific treatment of difficult asthma in children. Paediatr Respir Rev 2004;5:116–123. |
| 8. | Lex C, Ferreira F, Payne DNR, Zacharasiewicz A, Nicholson A, Haslam P, Wilson NM, Hansel TT, Bush A. Comparison of noninvasive and invasive methods for assessment of airway inflammation in children with asthma. Thorax 2003;58SIII:iii61. |
| 9. | Payne D, McKenzie SA, Stacey S, Misra D, Haxby E, Bush A. Safety and ethics of bronchoscopy and endobronchial biopsy in difficult asthma. Arch Dis Child 2001;84:423–426. |
| 10. | American Thoracic Society. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease (COPD) and asthma. Am Rev Respir Dis 1987;136:225–244. |
| 11. | National Institutes of Health; National Heart, Lung, and Blood Institute. Expert panel report 2: guidelines for the diagnosis and management of asthma. Washington, DC: U.S. Government Printing Office; 1997. NIH Publication No. 97-4053. |
| 12. | Baraldi E, de Jongste JC; European Respiratory Society; American Thoracic Society. Measurement of exhaled nitric oxide in children, 2001. Eur Respir J 2002;20:223–237. |
| 13. | Warke TJ, Fitch PS, Brown V, Taylor R, Lyons JD, Ennis M, Shields MD. Exhaled nitric oxide correlates with airway eosinophils in childhood asthma. Thorax 2002;57:383–387. |
| 14. | Lex C, Payne DN, Li AM, Zacharasiewicz A, Wilson NM, Hansel T, Bush A. Sputum induction in children with difficult asthma: safety, feasibility, and inflammatory cell pattern. Pediatr Pulmonol 2005;39:318–324. |
| 15. | Fitch PS, Brown V, Schock BC, Taylor R, Ennis M, Shields MD. Chronic cough in children: bronchoalveolar lavage findings. Eur Respir J 2000;16:1109–1114. |
| 16. | Payne DN, Qiu Y, Zhu J, Peachey L, Scallan M, Bush A, Jeffery PK. Airway inflammation in children with difficult asthma: relationships with persistent symptoms and airflow limitation. Thorax 2004;59:862–869. |
| 17. | Altman DG, Bland JM. Diagnostic tests. 1: sensitivity and specificity. BMJ 1994;308:1552. |
| 18. | Altman DG, Bland JM. Diagnostic tests. 2: predictive values. BMJ 1994;309:102. |
| 19. | Altman DG, Bland JM. Statistics notes: diagnostic tests 3: receiver operating characteristic plots. BMJ 1994;309:188. |
| 20. | Maestrelli P, Saetta M, Di Stefano A, Calcagni PG, Turato G, Ruggieri MP, Roggeri A, Mapp CE, Fabbri LM. Comparison of leukocyte counts in sputum, bronchial biopsies, and bronchoalveolar lavage. Am J Respir Crit Care Med 1995;152:1926–1931. |
| 21. | Silkoff PE, Trudeau JB, Gibbs R, Wenzel S. The relationship of induced inflammatory cells to BAL and biopsy. Chest 2003;123:S371–S372. |
| 22. | Grootendorst DC, Sont JK, Willems LN, Kluin-Nelemans JC, Van Krieken JH, Veselic-Charvat M, Sterk PJ. Comparison of inflammatory cell counts in asthma: induced sputum vs bronchoalveolar lavage and bronchial biopsies. Clin Exp Allergy 1997;27:769–779. |
| 23. | van den Toorn LM, Overbeek SE, de Jongste JC, Leman K, Hoogsteden HC, Prins JB. Airway inflammation is present during clinical remission of atopic asthma. Am J Respir Crit Care Med 2001;164:2107–2113. |
