© 2004 American Thoracic Society
Design-based CountingTo the Editor:With the exception of a few groups that have recognized the significance of design-based methods in counting three-dimensional particles (organelles, cells, or alveoli) instead of two-dimensional section profiles (1, 2), the enormous progress made in design-based stereology during the last decade has largely been ignored by the scientific community in respiratory research. This is especially surprising because in the past respiratory researchers have pioneered the field of quantitative histology and histopathology (3). Stereology has now moved an additional important step forward. Ochs and coworkers (2) developed a new tool to estimate the total number of alveoli of an individual lung. There appear to be multiple advantages of this design-based technique. Because there is no need for any assumption, it is unbiased with regard to alveolar size, size distribution, shape, and orientation. This is an important prerequisite for studies of lung development or emphysema formation in which researchers are confronted with marked heterogeneities in size, shape, and/or distribution of alveoli. Furthermore, the authors claim that their tool is highly efficient with regard to the time needed to obtain a valid estimate, although they do not tell us how much time is needed to set up the sampling and counting design. Ochs and coworkers (2) make a firm statement for design-based stereology and ask for a clear policy from respiratory biology journals to follow; for example, the Journal of the American Society of Nephrologists (the first of the 47 journals listed by the Journal Citation Report) requests "that appropriate stereologic methods be used to quantify structures in tissue sections in all manuscripts submitted to the Journal" (4). However, we are left with the open questions of whether we can still rely on data obtained by the "old" assumption-based methods, and whether alternative methods, for example, the indirect calculation of alveolar numbers from total alveolar volume and mean alveolar volume (5), may still be used. Design-based versus assumption-based methods have been vividly discussed in journals focusing on neurosciences, for example, in Trends in Neurosciences (the fourth of 197 journals listed) (6). To assure the high scientific quality standard of AJRCCM, we cannot refrain from entering the discussion about a common standard that studies submitted to this journal must meet when reporting quantitative structural data. Progress in science depends both on innovative concepts of thinking and on the development (as well as the usage) of pioneering methodologic tools.
Philipps-University of Marburg Marburg, Germany FOOTNOTES Conflict of Interest Statement: H.F. has no declared conflict of interest. Dr. Ochs was given an opportunity to respond to this letter but declined to do so. REFERENCES
Conflict of Interest Statement: P.K.J. has been reimbursed by GlaxoSmithKline (GSK), Astra-Zeneca (A-Z), and Merck, Sharpe & Dohme (Merck) for attending many conferences, and has participated as a paid speaker in scientific meetings or courses organized and financed by various pharmaceutical companies (such as GSK, A-Z, Merck, and Boehringer Ingelheim); has served as a consultant to GSK & Novartis; has received research grants from several pharmaceutical companies over many years and currently holds research grants from GSK, Merck, and A-Z, the first of which includes a grant for a multicenter clinical trial; and his institution has received unrestricted grants from a wide variety of pharmaceutical companies.
From the Authors: We thank Dr. Fehrenbach for his comments concerning the appropriateness of quantitative tools in respiratory research. He cites an article by Ochs and colleagues (1) in which a stereologic approach is applied to provide an unbiased estimate of human alveolar number. Of course, the focus of our own supplement (2) was not lung parenchyma but rather the airway mucosa, sampled by endobronchial biopsy, for the purpose of counting inflammatory cells of distinct phenotype (and size) in asthma and chronic obstructive pulmonary disease. The number of clinical respiratory research centers with biopsy expertise has increased. The lack of precise description of the different sampling procedures, quantitative methods, and statistical analyses has made comparisons among laboratories difficult. This has likely compromised interpretation and confirmation of previously published data. Our workshop (2) discussed these issues in the hope that it would help in the future to reduce data variability (from all sources), improve the capacity to detect disease-related differences, and especially to enhance statistical power to assess the value of various classes of therapeutic intervention. Although stereologic methods are often applied to relatively large tissue samples in histology, there is little awareness as to the relative strengths and weaknesses of two-dimensional (area profile) and three-dimensional methods for counting inflammatory cells in endobronchial biopsies. In contrast, this issue has been debated (inter alia) by neuroanatomists (3) and nephrologists (4). Theoretically, design-based stereology provides statistically unbiased count estimates regardless of the size, shape, orientation, and spatial distribution of objects (5). Although the case for its application to the objective assessment of endobronchial biopsies is persuasive, further debate would be helpful to test whether theoretic assumptions of stereology fit well with the limitations of endobronchial biopsy. Biopsy usually only samples relatively large conducting airways, the number of biopsies is small, and there are limits imposed by their relatively small (1–2 mm) size. We need more information as to the sources of variation and we should compare the distributions of data derived from both area profile and point counts. The latter may provide less biased and more normally distributed data than the non-normally distributed data currently obtained from area profile counts. As a result, these data may be amenable to parametric statistical analyses, which intrinsically have greater power to detect real differences among study populations. We join with Dr. Fehrenbach in encouraging a debate on the application of these stereologic tools to endobronchial biopsies for the purposes of both generating and testing hypotheses.
Imperial College London at the Royal Brompton Hospital London, United Kingdom REFERENCES
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
