Both letters comment on the apparently high percentage of lymphocytes in the preradiotherapy group [1]. These lymphocyte percentages are due to important methodologic variables and not to another underlying abnormality. In our study, the pre-radiotherapy total cell number (22.4 ± 18.9 million) and proportion of lymphocytes (34.5% ± 13.4%) before radiotherapy did not differ significantly from the values for normal persons (34.7 ± 19 million and 29.1% ± 12.9%, respectively) [2], which we have determined in other studies [2, 3].

These letters do highlight an important common misconception about lymphocyte counts in bronchoalveolar lavage samples that is based on data from the inaccurate method for cell quantification, cytocentrifuge preparation, which is widely used because of its simplicity, because of the large body of data gathered using this method, and because of the absence of a need to justify the use of alternate nonstandard methods. Although the cytocentrifuge method appears to be relatively accurate for the enumeration of neutrophils, Saltini and colleagues [4] have shown that it markedly underestimates lymphocyte numbers. Even worse, the error is not predictable, obviating the use of any correction factor [4]. These factors have led to the commonly held but inaccurate view that normal bronchoalveolar lavage fluid contains very low lymphocyte numbers.

Our experience [1-3] is consistent with that of Saltini and coworkers [4], and we use their suggested millipore filter method. Further, findings from flow cytometric evaluation, which is unlikely to distort relative cell numbers, are consistent with the percentage of lymphocytes present using the millipore filter (unpublished data). A recent article [5] concerning the distribution of cytocentrifuge-determined lavage cellularity data in large numbers of normal volunteers reinforces this point. Merchant and colleagues [5] found that the proportion of neutrophils and eosinophils, cells that are not subjected to the same errors during the cytocentrifugation procedure, follow a Poisson distribution. Lymphocyte and the inversely related macrophage proportions, however, failed to show either a normal or Poisson distribution. A random error of large magnitude generated by the cytocentrifuge counting method would certainly account for this otherwise unexpected and unexplained failure.

We strongly suggest that alternate methods to the cytocentrifuge should be used for determining the proportion of lymphocytes and macrophages in bronchoalveolar lavage fluid.