To:
ADC Fetal and Neonatal Edition Letters and ADC Education and Practice Letters
Electronic Letters to:
|
|
Electronic letters published:
-
![[Read eLetter]](/icons/misc/sectionDOWN.gif)
Concerns with the linear regression model
- Hassib Narchi, United Arab Emirates University, Al Ain, U.A.E. (8 January 2007)
-
Authors' response to concerns over the linear regression model.
- Richard G Feltbower, Joanna Allen, John Livingston and Colin Ferrie. (12 January 2007)
|
|
|||
|
Hassib Narchi, Associate Professor Faculty of Medicine & Health Sciences, United Arab Emirates University, Al Ain, U.A.E.
Send letter to journal:
hassib.narchi{at}uaeu.ac.ae Hassib Narchi, et al.
|
Dear Editor, I would like to thank the authors of the paper entitled “Recovery of consciousness after epileptic seizures in children” for their hard work trying to answer a very important and practical question facing clinicians in front of a postictal child 1. I would like however to raise some concerns regarding their analysis using the linear regression model they described, which might have effected their findings and thus their recommendations. First, the 2 explanatory variables in the model, “duration of seizure” and “use of emergency drugs”, may not be truly independent, thereby violating one of the assumptions for using this model. This is because emergency anticonvulsant therapy is administered to children with a prolonged convulsion, as advised in most guidelines, while short seizures usually require no therapy. The second point is that “acute symptomatic”, idiopathic” and “remote symptomatic” explanatory variables in the model are not truly separate explanatory variables, but constitute (like “febrile convulsion”, which was not considered) some of the possible values which a single explanatory variable (seizure aetiology) might take. I hope these concerns did not affect the results of the study nor the conclusions and recommendations of the authors for this important clinical question. Dr. Hassib Narchi
Associate Professor Email: hassib.narchi@uaeu.ac.ae References: 1. Allen J.E., Ferrie C. D., Livingston J.H., Feltbower R.G Recovery of consciousness after epileptic seizures in children. Arch Dis Child 2007; 92: 39-42. |
|||
|
|
|||
|
Richard G Feltbower, Senior Research Fellow University of Leeds, Joanna Allen, John Livingston and Colin Ferrie.
Send letter to journal:
R.G.Feltbower{at}leeds.ac.uk Richard G Feltbower, et al.
|
Dear Editor, We thank Dr Narchi for raising a couple of important statistical issues. The first point suggesting that emergency drug use may not be independent of duration of seizure is certainly valid statistically. Median duration time was significantly higher in those patients who were administered emergency anticonvulsant therapy compared to those who did not receive this treatment. However, when this variable was excluded from the regression model, the parameter estimates for the remaining variables were virtually identical to the results presented in Table 2; only the p- values differed slightly for duration (p=0.01, previously p=0.10) and idiopathic aetiology (p=0.04. previously p=0.08). From a clinical point of view, we felt that the importance of adjusting for emergency drug use in examining seizure recovery time outweighed the potential statistical problem outlined above and that the paper would be criticised if this variable was omitted. The robustness of the results to the addition and exclusion of the 'drug use' variable would seem to indicate that this approach was justified. The second point arises from some confusion over the inclusion of 'seizure aetiology' as a categorical variable in the regression model. We can confirm that this was exactly the approach we took in the analysis and acknowledge that this may not have been clear in the methods. We compared the relative effects on recovery time for acute symptomatic, idiopathic and remote symptomatic seizures compared to febrile seizures, the latter category being used as the baseline group. |
|||
