INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although several recent studies have reported on the safety of sputum induction (1), on the reproducibility of measures of inflammation in induced sputum from asthmatic subjects (2, 6), and although one trial has reported on the use of induced sputum in a multicenter trial (11), there have been no published studies of reproducibility or safety in a multicenter study. The Asthma Clinical Research Network (ACRN) of the National Heart, Lung, and Blood Institute, an established network of six clinical research centers and a Data Coordinating Center, is using sputum induction and analysis of induced sputum as an outcome indicator of airway inflammation in a clinical trial of asthma treatments in subjects with moderate to severe asthma. Because of uncertainty about the safety and reproducibility of sputum induction and analysis of induced sputum in this setting, we examined these issues in asthmatic subjects with moderate to severe disease as part of a pilot study for a clinical trial. The specific research questions of our study were to determine the safety and reproducibility of sputum induction and analysis of induced sputum in subjects with moderate and severe asthma in a multicenter clinical trial. Because asthma is an intrinsically variable condition, we compared the reproducibility of sputum markers of inflammation with the reproducibility of the FEV₁ and the provocative concentration of methacholine causing a 20% reduction of FEV₁ (PC₂₀) in the same subjects.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and Protocol

Seventy-nine subjects with asthma, with no history of an asthma exacerbation in the preceding 6 wk and who had a baseline FEV₁ > 40% predicted were recruited at six centers (Table 1), and enrolled in a two-visit study, 2 to 7 d apart. Procedures on the first visit included medical and asthma history, physical examination, allergen skin testing, and baseline spirometry. Next they underwent methacholine challenge, reversal of methacholine-induced bronchoconstriction with albuterol 360 µg by metered-dose inhaler (MDI), repeat spirometry 20 min later, followed immediately by sputum induction. Procedures on the second visit were spirometry, methacholine challenge, and sputum induction. Subjects withheld short-acting bronchodilators for 8 h before spirometry and PC₂₀ measurements and salmeterol, theophylline, and oral -agonists were withheld for 48 h, 12 to 24 h (short-acting versus long-acting theophylline), and 1 wk, respectively, before visit 1. To qualify for methacholine challenge subjects needed to have a baseline prebronchodilator FEV₁  55% predicted; for sputum induction subjects needed to have an FEV₁ after methacholine challenge and treatment with albuterol of  60% predicted. The protocol was approved by the institutional review boards of the participating centers, and written informed consent was obtained from each subject.

Standardization of Procedures and Quality Assurance

Procedures for spirometry and methacholine challenge were performed and standardized across participating centers (12, 13). Research assistants from each center were trained in sputum induction and in processing of induced sputum during a two-day workshop. Individual centers performed total and differential cell counts in induced sputum (see Figure 1); sputum supernatant analyses were performed at the San Francisco center. The San Francisco center also overread all differential cell counts and provided feedback on slide quality and cell differential accuracy to individual centers (Table 2). The primary data for cell differentials were the data from individual centers.

View larger version (96K): [in this window] [in a new window]   Figure 1.   Photomicrographs of cytospin preparations of induced sputum samples stained with the Leukostat stain from six different asthmatic subjects. The figure demonstrates the morphologic criteria used for differential cell analysis in this study, which is based on descriptions by Chodosh (21) and by Linder and Rennard (22). Seven different cell types are identifiable in the six panels. Macrophages (#1) are identified principally by their vacuolated cytoplasm, round shape, and nuclei which are often eccentric; cell size varies, and the cell may have two nuclei (all panels). Neutrophils (#2) are identifiable by their segmented nuclei and by their indistinct cytoplasm and indistinct cytoplasmic borders (Panels A, B, C, D, F  ). Epithelial cells (#3) are identifiable by their columnar shape, by the presence of a terminal bar, or by the presence of cilia; cilia may be absent, and the size and shape of these cells varies considerably (Panels A, E, F  ). Eosinophils (#4) are identifiable by their eosinophilic granulated cytoplasm and by a nucleus that is usually bilobed but can have three or more lobes (Panels B, D). Lymphocytes (#5) are identifiable by their high nuclear-to-cytoplasmic ratio (scant cytoplasm) and by their small size in relation to macrophages (Panels C, D, E ). Squamous cells (#6) are identifiable by their large size, high cytoplasmic-to-nuclear ratio and by their bland cytoplasmic staining (bacteria are often seen in association with these cells) (Panel D). Monocytes (#7) are identifiable by their kidney shaped nucleus (Panel F  ); these cells were included in the differential cell count for macrophages. This figure was made available to all centers as part of the sputum manual of procedures.

Sputum Induction

Sputum induction was performed, as previously described (8). Briefly, all subjects were pretreated with 360 µg albuterol, and spirometry was repeated 10 min later to ensure that the postalbuterol FEV₁ was  60% of predicted. A 12-min sputum induction was then performed during which peak flow was monitored every 4 min. Subjects were instructed to spit saliva into one cup before coughing sputum into another (saliva was later discarded). An inadequate induced sputum sample was defined by the following criteria: sputum induction tolerated for less than 4 min, induced sputum volume < 1 ml, or squamous cell percentage > 80%. The choice of a cutoff for 80% squamous cells was based on practical issues. Typically, cytocentrifuged slides have approximately 500 cells. If 80% of the cells are squamous cells then 100 cells will be nonsquamous cells, and five slides will be needed to read at least 500 nonsquamous cells. Five slides was considered the maximum number of slides that could feasibly be prepared and stained for the purposes of this multicenter study. If a subject tolerated sputum induction for more than 4 min but less than 12 min at visit 1 then the duration of sputum induction at visit 2 was kept the same as at visit 1. Eosinophil cationic protein (ECP) and tryptase concentrations in induced sputum were measured as previously described (14).

Statistical Considerations and Analysis

The data are presented as mean and standard deviation or as median and interquartile range. Tryptase was less than the detectable range of the assay (2.0 IU/L) in 33% of the visit 1 samples and 34% of the visit 2 samples, and a random value between 0 and 2.0 was imputed for these samples. None of the ECP values was below the detectable range. To determine whether subject characteristics (e.g., age, sex, baseline FEV₁) were each independently associated with a decrease in FEV₁ during sputum induction, a logistic regression model was used. The outcome variable in the model was whether or not the subject's FEV₁ decreased by  20% during sputum induction, and a separate model was fit to each subject characteristic. The model results are summarized as odds ratios and 95% confidence intervals (CI). For example, the odds ratio for gender is interpreted as the odds of a decrease in FEV₁ of  20% for a female versus the odds of a decrease for a male. Reproducibility was estimated in two waysthe concordance correlation coefficient (15) and the Bland-Altman correlation coefficient (16). The baseline prebronchodilator FEV₁ value was used for calculation of reproducibility of FEV₁. The distribution of the difference of two dependent concordance correlations is not known. A bootstrap method was applied to test the hypothesis that the difference was zero (17). We decided that a clinically important level of reproducibility, measured by a concordance correlation, was 0.75. The study had 80% power to detect a concordance correlation of 0.75 with a possible loss of precision of 25% with a sample size of 54 subjects.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Enrollment and Demographics

The majority of the subjects had moderate or severe asthma as evidenced by moderate to severe airflow obstruction (Table 1) (89% had an FEV₁ < 80% predicted, and 21% had an FEV₁ < 60% predicted despite the use of inhaled corticosteroids by 67% of the subjects). Eighteen of the 79 subjects (23%) had a history of salmeterol use. Seventy-nine subjects were enrolled at visit 1, and 63 subjects returned for visit 2. Of the 16 subjects who did not return for visit 2, seven had produced an inadequate induced sputum sample on visit 1 and so were ineligible. The reasons for the inadequate induced sputum sample were as follows: in one subject the duration of sputum induction was < 4 min (because of bronchospasm), and the sample produced was < 1 ml; in another subject the sample volume after 12 min was < 1 ml; and in five subjects the percentage of squamous cells was > 80). Nine subjects had adequate induced sputum on visit 1 but did not have a second sputum induction because five withdrew consent, two could not be scheduled within the protocol time window, one suffered a broken rib between the two study visits (unrelated to sputum induction), and one developed dizziness during methacholine challenge on the second visit and so did not proceed to sputum induction.

Safety

On visit 1, 76 of the 79 subjects had FEV₁ data before and after sputum induction. The FEV₁ values before methacholine challenge, after albuterol reversal of methacholine-induced bronchoconstriction, and after sputum induction are summarized in Table 2. Eleven of the 76 subjects (14%) had a decline in FEV₁ from the postalbuterol baseline of 20% or greater (Figure 2). All 11 subjects who had a decrease in FEV₁ of > 20% from postalbuterol baseline values recovered within 1 h to within 12% of their baseline prealbuterol premethacholine FEV₁ value after treatment with additional albuterol (360 µg albuterol by MDI in nine subjects; two required additional albuterol 2.5 mg by nebulizer). No subject developed refractory bronchoconstriction requiring emergency room treatment or hospitalization. The largest decrease in FEV₁ during sputum induction was 43%. This subject had a prealbuterol baseline FEV₁ of 1.6 L (50% predicted) which increased to 2.1 L (65%) with albuterol treatment. After sputum induction the FEV₁ was 1.2 L, and increased to 2.0 L after nebulized albuterol treatment. Four of 16 subjects (25%) with a baseline prealbuterol FEV₁ between 40 and 60% predicted had a fall in FEV₁ of > 20% during sputum induction at visit 1, compared with seven of the 51 subjects (14%) whose FEV₁ was between 60% and 80% predicted; none of the nine whose FEV₁ was greater than 80% predicted had a decrease in FEV₁ greater than 20%. However, in a logistic regression model, none of the subjects' characteristics before sputum induction, including demographic data (age, sex, race), physiological data (baseline prebronchodilator FEV₁ [% predicted], or PC₂₀ methacholine significantly predicted a decrease in FEV₁ during sputum induction (all odds ratios were not significantly different from 1.0; all p values > 0.15). In addition, we found no significant relationship between dose of inhaled corticosteroid and fall in FEV₁ at visit 1 (p = 0.54). Of the 11 subjects who had a decrease in FEV₁ of  20% during sputum induction on visit 1, four developed this fall at 4 min, three at 6 min, none after 8 min, four at 10 min, and none at 12 min. The percent change in FEV₁ during the first sputum induction was generally predictive of the change in FEV₁ during the second sputum induction (concordance correlation coefficient: 0.73 [95% CI: 0.59, 0.83]).

View larger version (17K): [in this window] [in a new window]   Figure 2.   Change in FEV1 during sputum induction in 79 subjects with moderate and severe asthma. The figure shows the percentage of subjects who had an increase or a decrease in their FEV1 from the post bronchodilator baseline during sputum induction.

Markers of Inflammation in Induced Sputum

Seven of the 79 subjects had an inadequate induced sputum sample on visit 1. Samples from five subjects had > 80% squamous cells, and two subjects produced < 1 ml of sputum. The cellular and biochemical characteristics of the induced sputum from the 72 subjects with adequate induced sputum on visit 1 are presented in Table 3. In a linear regression analysis, the sputum eosinophil percentage on visit 1 was significantly associated with the percent decrease in FEV₁ during sputum induction (p = 0.03); none of the other sputum markers, including sputum ECP or tryptase, was significantly associated with the percent fall in FEV₁ during sputum induction. We examined the correlation between the readings for sputum eosinophil percentage at visit 1 at individual centers and the overreading data for sputum eosinophil percentage visit 1 obtained by a single reader at the San Francisco center. The concordance correlation coefficient for the log transformed data (n = 69 [three slides damaged during shipping]) was 0.82 (95% CI: 0.72, 0.88).

Reproducibility

Fifty-nine subjects produced adequate induced sputum on visit 1 and visit 2, and data on cell differentials were available for 59 pairs. Data on ECP were available for 54 pairs (one lost sample, four samples with insufficient volume). Data on tryptase were available for 56 pairs (one lost sample, two samples with insufficient volume). The percentage of eosinophils in induced sputum on visit 1 was significantly and positively correlated with the sputum eosinophil percentage on visit 2 (Table 4, Figure 3). In addition, the concentrations of ECP in induced sputum on visit 1 and on visit 2 were positively and significantly correlated (Table 4, Figure 3). Log transformed data were used for these correlations. The reproducibility of the eosinophil percentage was similar to that of the ECP concentration (Table 4, Figure 3). In addition, the reproducibility of the eosinophil percentage, ECP, and tryptase measurements was similar to the reproducibility of the PC₂₀ for methacholine (Table 4, Figure 3). None of these measurements was as reproducible as the FEV₁ (Table 4, Figure 3). The reproducibility of the pulmonary function measurements was compared formally with the reproducibility of the measurements of inflammation in induced sputum by examining differences in the concordance correlation estimates (Table 5). We found that the measurements of FEV₁ were more reproducible than the measurements of inflammation in induced sputum. The reproducibility of the methacholine PC₂₀ was not significantly better than the reproducibility of the induced sputum measurements. Sample size calculations trials based on these data for one and two clinical samples are presented in Table 6.

View larger version (38K): [in this window] [in a new window]   Figure 3.   Bland-Altman summaries for the reproducibility of eosinophil% and ECP in induced sputum compared to the reproducibility of FEV1 and methacholine PC20 in the same subjects. The Bland-Altman summaries plot the difference in the measure between two visits versus the average value of the measure for the two visits for each subject. The overall average difference (solid line) ± 2 standard deviations (dashed line) is also shown on the plot. The figures show the various measures to be reproducible, because the individual points on the Bland-Altman plot are randomly scattered around the overall average difference, and most points fall within 2 standard deviations of the overall average difference. A nonrandom scatter (e.g., most points above the overall average difference) within the 2 standard deviations might reflect a strong correlation, but a shift in magnitude from one visit to the next. A random scatter with several points outside the 2 standard deviations would indicate a weak correlation.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, we examined the safety and reproducibility of sputum induction in asthmatic subjects with moderate to severe disease enrolled in a multicenter study. Our principal findings are that the risks of sputum induction are minimal and tolerable in this carefully controlled setting and that the reproducibility of measures of inflammation in induced sputum compares favorably with the reproducibility of other commonly used outcome measures of asthma control.

We found that 14% of the subjects in this study had a decrease in FEV₁ of greater than 20% during sputum induction. No subject required emergency room treatment or hospitalization as a result of sputum induction. The subjects with the lowest baseline prebronchodilator FEV₁ were at greatest risk for bronchospasm. For example, 25% of the 16 subjects with a baseline FEV₁ between 40% and 60% of predicted had a decrease in FEV₁ of greater than 20% compared with none of the nine subjects whose baseline FEV₁ was greater than 80% of predicted. Bronchospasm occurred as early as 4 min into sputum induction, indicating that some subjects remain very sensitive to hypertonic saline even after pretreatment with 360 µg albuterol. Care needs to be taken in extrapolating the safety results of our study to studies that use a different protocol for sputum induction. Protocols that pretreat with lower doses of -agonist, that use higher concentrations of hypertonic saline for longer periods, and that use nebulizers with a higher output may have a different incidence of bronchospasm. In addition, many of the subjects enrolled in our study were taking inhaled corticosteroids, which are known to attenuate hypertonic saline-induced bronchoconstriction (18). We speculate that sputum induction protocols in which subjects are enrolled who are not taking inhaled corticosteroids might be associated with a higher incidence of bronchoconstriction, especially in subjects with a low baseline FEV₁. Pretreatment with albuterol and measurement of peak flow every 4 min were measures included to reduce the risk of excessive bronchoconstriction, but these measures did not eliminate the risk. Because of this, we have empirically modified our sputum induction protocol further to begin peak flow monitoring after 2 min of hypertonic saline inhalation and at 2-min intervals thereafter.

It is possible that the preceding methacholine challenge modified the airway response to hypertonic saline or the composition of the induced sputum. However, we believe that any effect of methacholine challenge would have led to an overestimation rather than an underestimation of the frequency of bronchoconstriction during sputum induction, and although a preceding methacholine may increase the percentage of neutrophils in induced sputum, it has little effect on sputum eosinophils (19, 20). We found that the reproducibility of the sputum eosinophil percentage, the sputum ECP, the sputum tryptase, and the PC₂₀ for methacholine were similar. This finding is reassuring, because the reproducibility of PC₂₀ for methacholine has proven to be adequate for its use as an outcome indicator in clinical trials in asthma, and our data demonstrate that sputum markers may be equally useful.

In summary, we found that sputum induction in subjects with moderate to severe asthma studied in a multicenter setting has acceptable risks and that markers of inflammation in induced sputum are as reproducible as the methacholine PC₂₀. Despite the apparent safety of sputum induction in this setting, sputum induction has a predictable risk of bronchoconstriction and should only be undertaken in carefully monitored conditions with rigorous safeguards to identify and treat bronchoconstriction.