The use of induced sputum in clinical trials
- Leader of the Working Group:,
- J.C. Kips 1 ,
- Members of the Working Group:,
- M.D. Inman 2 ,
- L. Jayaram 3 ,
- E.H. Bel 4 ,
- K. Parameswaran 5 ,
- M.M.M. Pizzichini 6 ,
- I.D. Pavord 7 ,
- R. Djukanović 8 ,
- F.E. Hargreave 2 and
- P.J. Sterk 4
- 1University Hospital Ghent, Dept of Respiratory Diseases, Ghent, Belgium. 2Firestone Institute for Respiratory Health, McMaster University, Hamilton, Ontario, Canada. 3Respiratory and Sleep Disorders Unit, Repatriation General Hospital, Daw Park, South Australia, Australia. 4Dept of Pulmonology, Leiden University Medical Centre, Leiden, the Netherlands. 5FRCAU, St. Joseph's Hospital, Hamilton, Ontario, Canada. 6Asthma Research Centre, University Hospital, Federal University of Santa Catarina, Florianópolis, Santa Catarina, Brazil. 7Institute for Lung Health, Dept of Respiratory Medicine and Thoracic Surgery, Glenfield Hospital, Leicester, UK.8Southampton University General Hospital, Southampton, UK
- J. Kips, University Hospital Ghent, Dept of Medical Genetics, De Pintelaan 185, 9000, Ghent, Belgium. Fax: 32 9 2402341. E‐mail: johan.kips@rug.ac.be
As asthma is considered to be an inflammatory disorder of the airways, it seems logical to include an assessment of this inflammatory process as an outcome measure in clinical trials. Biopsy studies illustrate that clinical or lung function characteristics such as symptoms, peak flow variability or degree of airway responsiveness do not consistently correlate with histological alterations. Therefore, these clinical indices cannot be seen as accurate markers of airway inflammation 1–3. Conversely, repeated bronchoscopic sampling is not feasible in largescale clinical studies. Hence there is interest in a relatively noninvasive but direct marker of airway inflammation.
Analysis of induced sputum seems to meet these criteria. Provided proper precautions are taken, induction of sputum is safe, even in patients with more severe asthma 4, 5. In addition, sputum cell counts, particularly eosinophil counts, have been well validated in terms of responsiveness to intervention. It has been argued that, in comparison with other noninvasive markers of inflammation, induced sputum offers the most balanced assessment of the degree of inflammation, being more responsive to intervention than serum eosinophil cationic protein, yet not as oversensitive as exhaled nitric oxide 6–8.
As for any outcome measure, when including induced sputum in a clinical trial, specific features of sputum analysis need to be taken into account when designing the study: 1) origin of sputum; 2) methodological aspects; 3) selection of subjects; and 4) power calculations.
Origin of sputum
The induced sputum technique samples the inflammatory cells and soluble markers present in the airway lumen of the bronchial tree, which, although reflective of, does not represent an identical situation to the local inflammatory process in the mucosa. This probably explains the poor correlation between the cellular composition of biopsy samples and sputum, bronchial wash or bronchoalveolar lavage 9–11. Therefore, although induced sputum can provide information regarding the ongoing overall inflammation in asthma, it might not be the ideal substrate for studying the exact pathophysiological events that occur within the airway wall.
Methodological aspects
Regarding the use of induced sputum in clinical trials, a few methodological issues should be reemphasised. Consecutive inductions within a short time interval can cause an increase in the percentage of neutrophils in sputum 12, 13. It has also been reported that the composition of sputum can change somewhat throughout the duration of the induction procedure and that standardising this variable would seem advisable 14. Current recommendations are to process sputum samples within 2 h after induction, although a recent study has shown that this can be prolonged to 9 h (A. Efthimiadis, Firestone Institute for Respiratory Health, Hamilton, Ontario, Canada, personal communication). This can be important when deciding on the number and timing of inductions to be performed.
Selection of subjects
It should be remembered that, to date, analysis of induced sputum has mainly been validated with regard to the percentage of eosinophils in the cell pellet. The responsiveness to intervention of other possible outcome measures, be they cells or soluble mediators, has been far less thoroughly established.
The use of eosinophil counts as the main outcome measure of the induced sputum technique requires specific consideration, in part related to the overall aim of the study planned, to be given to the screening of patients for study inclusion. If the aim of the study is to evaluate the biological activity of a given compound on eosinophils, it is acceptable to select subjects based on the number of eosinophils in their sputum samples. However, this does not apply if the aim of the study is to conduct a clinical trial in asthma as a disease entity. Although sputum eosinophil counts in samples obtained from the same patient on different occasions are repeatable, the interindividual variability is large, even in samples obtained from patients with very similar clinical characteristics 15, 16.
Therefore, including only those patients with a specific degree of sputum eosinophilia leads to a selection bias that hampers a proper interpretation and generalisation of the results.
Another point that must be considered is that, when subjects are randomised based on lung function or clinical criteria, the wide variability in sputum eosinophilia can cause unexpected differences between the groups, at baseline. An approach to preventing this is to stratify patients at randomisation to ensure that the full range of baseline eosinophilia is equally represented in each study group 17. It can only be used in proof of concept studies aimed at reducing numbers of sputum eosinophils.
Power calculations
An important consideration in the design of most clinical trials is estimation of the appropriate sample size required to draw adequate conclusions from the study. This depends on the set probability of making a type I (α level) or type II (β level) error, the design of the study, the reliability of the outcome measurement and the desired effect size. The α level or probability of falsely rejecting the null hypothesis is usually set at 5%. The β level represents the probability of falsely accepting the null hypothesis, usually set at 10–20%. Hence, 1‐β, the power of the study, expresses the probability of avoiding a type II error. The power of the study relates intimately to the variability in the outcome measurement (either inter or intrasubject, depending on study design) and the desired effect size. The present authors strongly recommend that researchers determine sample size requirements based on the variability and reproducibility of sputum eosinophil counts in their own population. A less favoured option is to rely on data from the literature, taking into account the induction and processing techniques used. As an example, duplicate measurements were examined from 84 volunteer subjects from multiple centres (table 1⇓). All of these subjects had been diagnosed with asthma and were not currently receiving antiinflammatory treatment. No intervention took place between the two sputum measurements, which were separated by ∼1 week. The formula used to determine sample size requirement is (T1‐α/2+T1‐β)ϕ/LCID, where T is score of the paired t‐test corresponding to the desired α and power (1‐β), ϕ is the standard deviation of the outcome of interest and LCID is the least clinically important difference. When designing a parallel study, ϕ relates to the interindividual variability in eosinophil count, whereas, in a crossover study, the variable of interest is the intrasubject reproducibility. An alternative consists of feeding these data into power tables. This allows the determination of the sample size required to draw conclusions with various levels of confidence regarding the significance of a predefined difference in eosinophil count over the study period within or between groups. An example is given in figure 1⇓. These data are based on the reproducibility of changes in sputum eosinophil count following allergen challenge. From these data, it would appear that, when performing repeated measures analysis of variance on sputum samples obtained at 7 and 24 h after allergen challenge, the estimated sample size required to observe a 50% attenuation of allergeninduced increase in the percentage of eosinophils in a crossover study design is only five subjects. As total cell counts in sputum show poorer reproducibility, using total counts instead of percentage of eosinophils increases the estimated sample size required from five to 21, for the same level of power, desired effect size and study design 18.
A more difficult element is how to determine the desired effect size. A first possibility is to aim at normalisation of sputum eosinophil counts. Increasing evidence based on studies in healthy volunteers indicates that a normal range of sputum eosinophils in adults and children is <2.5% 19–21. However, any form of treatment rarely achieves full normalisation of all outcome measures in asthma. To date, it is uncertain whether reducing eosinophil counts, in addition to treating symptoms or lung function characteristics, improves the longterm clinical outcome of the disease. As a consequence, it is equally unclear what constitutes a “least clinically important difference” (LCID) as opposed to a statistically significant change in sputum eosinophil count. What has been shown is that an allergen inhalation challenge that causes a late asthmatic response is associated with an approximately fivefold increase in sputum eosinophil numbers 18. Conversely, treatment with steroids reduces sputum eosinophil counts. In the various studies reported to date, oral or inhaled steroids were given at different doses for varying time periods. It is therefore difficult to compare these various studies, but, overall, it would seem that inhaled steroids from low doses onwards, offer a ≥60% reduction in median sputum eosinophil percentage (table 2⇓). These data indicate that sputum eosinophilia changes substantially in response to intervention. It could therefore be proposed that the LCID should be a ≥50% change in sputum eosinophil count.
Of interest is that, from the limited amount of data available, it would seem that, as for clinical outcome measures such as symptom score or peak flow, the dose/response curve for the effect of inhaled steroids on sputum eosinophil counts is rather flat 8. Importantly, however, the change in eosinophil count does not always correlate with the degree of clinical improvement 24, 25, 28, 31, thus further suggesting that measurement of sputum eosinophil number could offer complementary information to mere clinical followup.
Key points
1) Analysis of induced sputum has only been validated for the percentage of eosinophils in the cell pellet; 2) the interindividual variability in sputum eosinophilia is large even in subjects with similar clinical characteristics; and 3) the least clinically important difference in sputum eosinophil counts remains to be established.
Outstanding questions
1) The least clinically important difference in sputum eosinophil counts remains to be established; and 2) the prognostic significance of reductions in sputum eosinophil count on the longterm clinical outcome in asthma needs to be investigated.
- Received April 4, 2002.
- Accepted April 16, 2002.
- © ERS Journals Ltd