PC₂₀ ADENOSINE 5'-MONOPHOSPHATE IS MORE CLOSELY ASSOCIATED WITH AIRWAY INFLAMMATION IN ASTHMA THAN PC₂₀ METHACHOLINE

To the Editor :

Van Den Berge and colleagues compared the PC₂₀ methacholine and the PC₂₀ adenosine 5'-monophosphate (AMP) in two subsequent papers (1, 2). The conclusions of the papers were that the PC₂₀ AMP better reflects (1) the airway inflammation and (2) the steroid-induced improvement. Their conclusions are based on the differences in the correlation coefficients between the percentage of eosinophils in induced sputum and the PC₂₀ AMP (r = 0.49) or PC₂₀ methacholine (r = 0.28) (1) or the change in these percentages (r = 0.43 and r = 0.28) (2).

A correlation between two parameters is always a weak basis to claim a true relation between the two. The association needs to be strong, specific, and repeatable. In this case, the correlation cannot be considered as strong: a r value of 0.49 means that only 24% of the variance is explained by the independent parameter.

In both papers, the noted differences remain untested: the authors only compare the point estimates of the correlation coefficients. They seem not to realize that every point estimate has an uncertainty reflected by the sampling variance, and that a conclusion as to whether or not two estimates truly differ must be based on that variance. The authors state that the PC₂₀ methacholine and the PC₂₀ AMP describe different inflammatory mechanisms and are therefore independent: in several statistical text books, comparative tests are described (3) and the outcomes indicate that none of the differences between the correlation coefficients differ statistically (p = 0.0615 [1] and p = 0.21 [2]).

To add to this, any regression and correlation approach can be severely hampered by the presence of outliers. These points can influence a regression line and the correlation coefficients. Especially in the first paper, in Figure 2 many outlying datapoints seem to be present (1). Statistical software packages like SPSS offer a wide range of diagnostics to detect and correct for influential datapoints. The authors should state in what way outliers influenced their results, especially because these outliers seem to be those patients who did not show a 20% FEV₁ decline at the highest methacholine or AMP dose possible. In these cases, that highest AMP or methacholine dose was used as a PC₂₀ value.

The conclusion that the PC₂₀ AMP better reflects the airway inflammation is in need of more and proper evaluation before this can truly be accepted.

Universitair Medisch Centrum, Utrecht, The Netherlands

1. Van Den Berge M, Meijer RJ, Kerstjens HAM, de Reus DM, Koëter GH, Kauffman HF, Postma DS. PC₂₀ adenosine 5'-monophosphate is more closely associated with airway inflammation in asthma than PC₂₀ methacholine. Am J Respir Crit Care Med 2001; 163: 1546-1550 [Abstract/Free Full Text].

2. Van Den Berge M, Kerstjens HAM, Meijer RJ, De Reus DM, Koëter GH, Kauffman HF, Postma DS. Corticosteroid-induced improvement in the PC₂₀ of adenosine monophosphate is more closely associated with reduction in airway inflammation than improvement in the PC₂₀ of methacholine. Am J Respir Crit Care Med 2001; 164: 1127-1132 [Abstract/Free Full Text].

3. Dunn OJ, Clark VA. Applied statistics: analysis of variance and regression analysis, 2nd ed. New York: John Wiley & Sons; 1984.

From the Authors:

We agree with Dr. Zanen that a monovariate correlation between two parameters is a weak basis to claim a relation between two variables. For this reason, we clearly stated in both papers that the conclusions were based on the outcomes of a multivariate regression analysis, which is the appropriate method to determine the dependency of a variable on a series of independent variables.

In our first paper, we investigated the association between the severity of hyperresponsiveness to AMP or methacholine with FEV₁ and inflammatory markers in sputum, blood, and exhaled air in a large group of 120 subjects with asthma (1). We found a dichotomy in the factors explaining the level of PC₂₀ methacholine and PC₂₀ AMP. The FEV₁ was the most important explanatory variable for the variation in PC₂₀ methacholine (explained variance [ev] = 18%) with the number of peripheral blood monocytes being a weak additional predictor (total ev = 23%). In contrast, the level of PC₂₀ AMP was predominantly predicted by the percentage of sputum eosinophils (ev = 25%), whereas FEV₁ was only an additional independent predictor.

In our second paper, we assessed the relationship between changes in both PC₂₀ methacholine and PC₂₀ AMP after two weeks of treatment with corticosteroids and the concomitant change in FEV₁ and airway inflammation (2). Improvement in PC₂₀ AMP was solely associated with reduction in airway inflammation in the multivariate analysis, whereas improvement in PC₂₀ methacholine was associated with both reduction in airway inflammation and increase in FEV₁. Moreover, the total explained variance of the model accounting for improvement in hyperresponsiveness was larger for AMP than for methacholine (36% versus 22%). On the basis of these analyses, we concluded that (1) the PC₂₀ AMP better reflects airway inflammation and (2) the corticosteroid-induced improvement in PC₂₀ AMP better reflects the concomitant reduction in airway inflammation (1, 2).

Another point of criticism of Zanen was that the correlation coefficients presented in the first paper in Figure 2 could have been influenced by outlying data of patients who were unresponsive to AMP (2) (in our study those patients were assigned a value of twice the highest concentration of AMP applied). However, the presented correlation coefficients were derived from a nonparametric analysis, and therefore, the presented regression coefficients would not change from any other handling of the truncated data except leaving them out all together. In summary, we feel that the conclusions in both our papers were properly founded.

University Hospital Groningen, Groningen, The Netherlands

1. Van Den Berge M, Meijer RJ, Kerstjens HAM, de Reus DM, Koëter GH, Kauffman HF, Postma DS. PC₂₀ adenosine 5'-monophosphate is more closely associated with airway inflammation in asthma than PC₂₀ methacholine. Am J Respir Crit Care Med 2001; 163: 1546-1550 .

2. Van Den Berge M, Kerstjens HAM, Meijer RJ, De Reus DM, Koëter GH, Kauffman HF, Postma DS. Corticosteroid-induced improvement in the PC₂₀ of adenosine monophosphate is more closely associated with reduction in airway inflammation than improvement in the PC₂₀ of methacholine. Am J Respir Crit Care Med 2001; 164: 1127-1132 .