From the authors:
We deeply appreciate L. Bertoletti and M. Righini's interest and important comments about our paper published in the June issue of the European Respiratory Journal 1.
L. Bertoletti and his colleague stated in their letter that application of the Geneva clinical prediction rule for pulmonary emboli (PE) is not appropriate in hospitalised patients because it was originally derived from a database of emergency ward patients. We completely agree with this comment that probability of PE development after hospitalisation should not be assessed using the Geneva criteria. In our study population, in the majority of patients, evaluation of prediction rules for PE was done within a few hours of admission to the hospital, either in the emergency department or in outpatient clinics, before transferring the patients to the intensive care unit or respiratory ward. In other words, all necessary data for the Geneva and the Wells clinical prediction rules had already been obtained during their initial evaluation at the emergency department or outpatient clinics, and the probability scoring for each patient was based upon that admission data, not later data. So we do not think that our approach was a clear violation of application of the Geneva model in this study. As mentioned in our article 1, every chronic obstructive pulmonary disease (COPD) patient on an apparently severe exacerbation could be considered as an at-risk patient for venous thromboemboli (VTE) due to the presence of some strong common risk factors, such as being elderly, immobile or having cardiac problems, or infection of varying severity. Hence, application of the Geneva prediction rule should not yet be denied in this specific group of patients, unless new data from further studies will suggest otherwise.
An item in the Wells scoring system requires the clinician's judgment of whether an alternative diagnosis is less likely than PE or not. As mentioned by L. Bertoletti and M. Righini, as well as other authors 2, 3, several times, inclusion of this item in the Wells model seems to increase the subjectivity of the scoring and to decrease the external validity and reproducibility of any study related to PE prediction. Although the relevant item has long been considered as the weak point of this Canadian scoring system, findings from the available studies did not support this opinion clearly. In general terms, the Wells criteria were found to have moderate to substantial interobserver agreement 4–6.
Our study was designed and initiated in 2005 and 2006. At that time, the revised version of the Geneva criteria was neither available nor tested in an outcome study. Moreover, the main advantage of the new version of the Geneva model is its simplicity, by exclusion of arterial blood gas analysis and chest radiography. Although we also agree that the revised version suggests an easier alternative tool, it does not mean that it is superior to the original one regarding their PE prediction powers. No doubt, new studies will utilise the revised version of the Geneva model. As for the Wells score used in our study, we think that neither dichotomised nor trichotomised scores make too much sense if they are combined with a d-dimer test. It must be noted that positive d-dimer test allows the shifting of some patients from the “unlikely” PE arm of the dichotomised Wells score back to the “likely” arm. Since only few COPD studies related to the Wells scoring system exist in the literature, we preferred including all COPD patients for evaluation who are considered naturally at risk for PE, as mentioned above, to excluding some unlikely PE patients with negative d-dimer test results. We believe this approach would be necessary for the further validation of the Wells system in this specific group of patients.
As a complicating or triggering factor, presence of VTE in COPD patients on exacerbation is an important issue. 1-yr mortality was found doubled in VTE cases in our study. We believe that clinical prediction rules developed for PE will help us to manage COPD. As L. Bertoletti and M. Righini also underline, in order to have better prediction powers in cases with severe underlying specific diseases, we might need some modifications in the current prediction models, or some new disease-specific models should be developed in future.
Footnotes
Statement of Interest
None declared.
- ©ERS 2011