| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
University of São Paulo Medical School, São Paulo, Brazil University of Washington, Seattle, WA
Correspondence to: Rogério Souza, MD, PhD, Pulmonary Department, University of São Paulo Medical School, R. Afonso de Freitas 451, No. 112, São Paulo, Brazil 04006-052; e-mail: rogerio.souza{at}incor.usp.br
To the Editor:
In an article recently published in CHEST (May 2006), Fijalkowska et al1 provided important information on the value of N-terminal pro-type B natriuretic peptide (NT-proBNP) as a prognostic marker in patients with pulmonary hypertension (PH). There is growing interest in the role of natriuretic peptides in understanding the natural history of PH.23 The authors demonstrated that NT-proBNP is correlated with echocardiographic variables of right ventricular function and confirmed previous findings about the correlation of NT-proBNP with invasive hemodynamic variables. Unfortunately, the prognostic utility of NT-proBNP is limited by several factors that were not acknowledged in their article.
First, they did not present 95% confidence intervals around the estimates of sensitivity and specificity for predicting death. Since there were only 16 deaths in this cohort, the confidence intervals were quite wide, making the utility of the proposed NT-proBNP cutoff problematic. Although the sensitivity of this test was reported as 88%, the lower bound of the 95% confidence interval was 63%. The specificity of this test was reported as 53%, but the lower bound of the 95% confidence interval was 38%. Second, the authors used stepwise regression to build their survival model and screened several variables for inclusion. Multivariate models built using stepwise regression techniques are unreliable when there are < 10 outcomes per screened variable.4 Finally, once the presented model is prognostic rather than explanatory it is important to use appropriate statistical shrinkage techniques to account for the fact that the authors used the same data set to derive and validate this model, and to avoid overfitting.5
The study adds further details about the role of natriuretic peptides in the evaluation of PH patients, but its provocative findings about the prognostic value of NT-proBNP remain to be confirmed.
Footnotes
None of the authors presents any conflict of interest with this letter.
The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.
References
National Research Institute of TB and Lung Diseases, Warsaw, Poland
Correspondence to: Anna Fijalkowska, MD, National Research Institute of TB and Lung Diseases, Chest Medicine, Plocka 26, Warsaw 01-138, Poland; e-mail: a.fijalkowska{at}igichp.edu.pl
To the Editor:
We thank very much Dr. Souza et al for their interest in our article (May 2006)1 on the potential clinical role of N-terminal-pro-brain natriuretic peptide (NT-proBNP) estimation in patients with pulmonary arterial hypertension (PAH). We also agree with their concerns regarding the universal validity of the prognostic cutoff values that were reported in our article. Clearly, with a study group of slightly > 50 patients we would not impose or recommend such values for universal prognostic use. Furthermore, we do not believe that even the most sophisticated statistical analysis could extract more objective truth from such a database. On the other hand, by looking at the receiving operating characteristic analysis and trying to identify prognostic thresholds we wanted to avoid the common practice of reporting the prognostic significance of median values. While highly dependent on the initial characteristics of the study group of the index trial, such median values later often become surprisingly "magic numbers" and are extrapolated to other populations with even less evidence. This was the case both for a 6-min walk test distance of 332 m or an NT-proBNP concentration of 150 pg/L as a result of the reports of Nagaya et al2 and Miyamoto et al.3 Therefore, rather than defending our 1,400 and 3,400 pg/L thresholds, even though they were derived from a receiving operating characteristic analysis, we would reply by issuing a call for pulling together data on PAH patients whose baseline clinical characteristics and NT-proBNP concentrations were tested with a method similar to the one used in our study. This may be true, for instance, for the study population of a trial recently published by Souza et al.4 We have recently witnessed a great step forward in the understanding of the efficacy of the new drugs developed for the treatment of PAH thanks to global collaboration between expert centers. It is time to try to start building networks of collaboration to learn more about prognostic indexes and clinical end points. We would be available for this type of collaboration.
Regarding other, technical remarks, the 95% confidence intervals were not given for prognostic thresholds in order to make the presentation of the results clearer. However, all data needed for the calculation of those intervals were available in the published version of the article. The stepwise analysis was not performed mechanistically on the whole set of 10 variables. Instead, smaller sets of variables were selected for separate analysis, based on clinical needs and common sense. The models were compared with the 
2 test. The best performing model was then selected and presented in the published article.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
