COMPARISON OF TWO METHODS OF PROCESSING INDUCED SPUTUM: SELECTED VERSUS ENTIRE SPUTUM

To the Editor:

We have enjoyed seeing this comparison of methods of examination of sputum selected from the expectorate of sputum plus saliva (SS) versus the entire expectorate (ES) (1). We disagree, however, with the display and interpretation of eosinophil cell counts and levels of fluid phase ECP in SS compared with ES, and we suggest an alternative method of displaying and interpreting data on the discriminative properties of a diagnostic test.

The authors have compared the two methods as suggested by Bland and Altman (2). However, they have done this differently from what was suggested. The mean of eosinophils and ECP by ES and SS have been plotted against the ratio of the two and not against the differences. The scale of the units are different on both axes (logarithmic on x axis and linear on y axis). When the data from Table 3 is displayed correctly as in the figure, it provides additional information. Although the individual values are within the limits of agreement, the new figure shows that the differences in eosinophils and ECP with the two methods tend to be greater when the mean values are higher. This tendency can be predicted by linear regression equations for eosinophils (r² = 0.35, p = 0.002, difference = 1.6-0.5 × mean) and for ECP (r² = 0.90, p = 0.001, difference = 36.5- 1.5 × mean). This additional information can be used to compute the values of eosinophil counts and ECP for one method when the results from the other method are available.

We agree that there is a difference in neutrophil counts between ES and SS in this study (around 12%, Table 2). The lower percentage of neutrophils in SS might be explained by the higher percentage of eosinophils. In our experience this is unlikely to obscure asthma exacerbations caused by infections because airway infections are usually associated with high total and neutrophil differential cell counts, which are obvious (3). Moreover, the demonstration of a higher eosinophil count may be more relevant for the control, follow-up, and treatment of asthma than underestimation of neutrophil counts.

The cells were counted by two observers in this study. The interobserver agreement for the two methods is not presented. When the squamous cell contamination is high, as is often the case in ES, it interferes with cell identification and the interobserver agreement is lower (4).

Finally, the discriminative properties of eosinophil counts and ECP in distinguishing between healthy and asthmatic people have been established in this paper by comparing the mean of the counts and of ECP. However, this is better achieved by expressing the ratio between the variability in the measurements between groups (signal) and the total variability in the measurements (noise) as intraclass correlation coefficient (ICC) (5). From data in Table 3 the ICC for eosinophil counts and ECP in SS are 17% and 30%, and in ES are 12% and 21%, respectively. This demonstrates that although the mean values are similar, both have poor disease discriminative properties, with a slight advantage for SS.

Asthma Research Group
St. Joseph's Hospital and McMaster University
Hamilton, Ontario, Canada

View larger version (16K): [in this window] [in a new window]   Figure 1.   Bland and Altman plot of sputum eosinophil percentages and ECP levels obtained by the two methods (entire and selected sputum). Open circles are healthy subjects and closed circles are asthmatic subjects (in the upper panel, all the healthy subjects are clustered together). Horizontal continuous and dashed lines represent mean difference and 2 SD of mean difference, respectively. The slanted dashed line represents the regression line.

1. Spanevello, A., B. Beghe, G. B. Migliori, M. Ambrosetti, M. Neri, and P. W. Ind. 1998. Comparison of two methods of processing induced sputum: selected versus entire sputum. Am. J. Respir. Crit. Care Med. 157: 665-668 [Abstract/Free Full Text].

2. Bland, J. M., and D. G. Altman. 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet i: 307-310 .

3. Berlyne, G. S., A. Efthimiadis, P. Hussack, D. O'Shaughnessy, D. Groves, J. Dolovich, and F. E. Hargreave. 1997. Macroscopic versus microscopic sputum analysis in asthma. Eur. Respir. J. 10: 184s .

4. Pizzichini, E., M. M. M. Pizzichini, A. Efthimiadis, F. E. Hargreave, and J. Dolovich. 1996. Measurement of inflammatory indices in induced sputum: effects of selection of sputum to minimize salivary contamination. Eur. Respir. J. 9: 1174-1180 [Abstract].

5. Guyatt, G. H., B. Kirshner, and R. Jaeschke. 1992. Measuring health status: what are the necessary measurement properties? J. Clin. Epidemiol. 45: 1341-1354 [Medline].

	From the Authors:

We would like to thank Dr. Hargreave and coworkers for their comments. The mean sputum eosinophils and ECP by ES and SS was plotted against the ratio, as in our previous manuscript (1), to eliminate the tendency of the difference to be increased proportionally with the increase of the mean of the two values. Clearly, as Hargreave and coworkers write, if the mean of the two values is plotted against the difference, then the difference in eosinophils and ECP between the two methods tend to be greater when mean values are higher. This information is found in Table 3. The relationship very often holds as a mathematical consequence. We thank Dr. Hargreave for displaying (Figure 1, their letter) our data, as strictly suggested by Bland and Altman (2). However, as they mention, the individual values are within the limits of agreement showing concordance of the two methods.

According to data recently published by Hargreave and colleagues (3), we agree that the higher eosinophil count (in SS) is unlikely to mask the situations usually associated with a high total and neutrophil differential count. However, Turner and coworkers have previously shown a lower percentage of neutrophils (around 11%) in subjects with exacerbation of asthma with sputum eosinophilia with respect to those subjects with exacerbation without sputum eosinophilia (4). In addition Saetta and coworkers showed an increase of eosinophils in the sputum of subjects with chronic bronchitis during exacerbation. There was also a decrease of neutrophils (more than 10%) during exercerbations with respect to the value at baseline (5). We also agree that the demonstration of a higher eosinophil count may be more relevant for the control, follow-up, and treatment of asthma than underestimation of neutrophils counts; in fact we conclude that selected sputum is a better method of analyzing induced sputum.

In our study, the interobserver agreement was very high (R = reliability coefficient > 0.8 for macrophages, neutrophils, eosinophils, lymphocytes, and epithelial cells) in both ES and SS (6). We found that counting the entire sputum sample can be as reliable as counting selected sputum, but is more time consuming. The poorer agreement between observers found by Hargreave might be due to the analysis of the residual part of the sputum (after selection) (7). Residual part samples of expectorate presented a higher squamous contamination (median: 70%, range: 14-97), compared to ours of ES (median: 55%, range: 17-79). The main aim of our paper was to compare the techniques rather than to compare two groups of subjects. We are not familiar with the use of the intraclass correlation coefficient for comparing between groups and have relied on local and general statistical advice for methodology (8). Incidentally, we have used the same method as Hargreave's group in previous publications validating sputum induction (7, 9, 10). Clearly, very large groups of subjects are necessary to really examine variability in the measurement and to truly assess the discriminative value of a test (1).

In conclusion, we appreciate the careful consideration of our manuscript by Dr. Hargreave and colleagues, and we are happy to see that we do not disagree on any major conclusion. Selected sputum is a better method for analysing induced sputum, providing more viable cells, more eosinophilis, and higher concentration of ECP in asthmatics, but in our hands, the entire sputum method was also effective at distinguishing asthmatics from healthy subjects.

Fondazione Maugeri Tradate, Italy

BIANCA BEGHÉ

University of Ferrara Ferrara, Italy

PHILIP W. IND

Imperial College Hammersmith Hospital London, United Kingdom

1. Spanevello, A., G. B. Migliori, A. M. Sharara, L. Ballerdini, P. Bridge, P. Pisatl, M. Neri, and P. W. Ind. 1997. Induced sputum to assess airway inflammation: a study of reproducibility. Clin. Exp. Allergy 27: 1138-1144 [Medline].

2. Bland, J. M., and D. G. Altman. 1986. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet i: 307-310 .

3. Berlyne, G. S., A. Efthimiadis, P. Hussack, D. O'Shaughnessy, D. Groves, J. Dolovich, and F. E. Hargreave. 1997. Macroscopic versus microscopic sputum analysis in asthma. Eur. Respir. J. 10: 184s .

4. Turner, M. O., P. Hussack, M. R. Sears, J. Dolovich, and F. E. Hargreave. 1995. Exacerbations of asthma without sputum eosinophilia. Thorax 50: 1057-1061 [Abstract/Free Full Text].

5. Saetta, M., A. Di Stefano, P. Maestrelli, G. Tureto, M. P. Ruggieri, A. Roggeri, P. Calcagni, C. E. Mapp, A. Ciaccia, and L. M. Fabbri. 1994. Airway eosinophilia in chronic bronchitis during exacerbations. Am. J. Respir. Crit. Care Med. 150: 1646-1652 [Abstract].

6. Fleiss, J. L. 1986. Designed and analysis of clinical experiments. John Wiley and Sons, New York. 8-13.

7. Pizzichini, E., M. M. M. Pizzichini, A. Efthimiadis, F. E. Hargreave, and J. Dolovich. 1996. Measurement of inflammatory indices in induced sputum: effects of selection of sputum to minimize salivary contamination. Eur. Respir. J. 9: 1174-1180 .

8. Wonnacott, T. H., and R. J. Wonnacott. 1991. Introductory Statistics. John Wiley and Sons, New York.

9. Pin, I., A. P. Freitag, P. M. O'Byrne, A. Girgis-Gabardo, R. M. Watson, J. Dolovich, J. A. Denburg, and F. E. Hargreave. 1992. Changes in the cellular profile of induced sputum after allergen induced asthmatic responses. Am. Rev. Respir. Dis. 145: 1265-1269 [Medline].

10. Kidney, J. C., A. G. Wong, A. Efthimiadis, M. M. Morris, M. R. Sears, J. Dolovich, and F. E. Hargreave. 1996. Elevated B cells in sputum of asthmatics: close correlation with eosinophils. Am. J. Respir. Crit. Care Med. 154: 540-544 .