PEDIATRICS Vol. 117 No. 1 January 2006, pp. 254-255 (doi:10.1542/peds.2005-2615)
LETTER TO THE EDITOR |
Niflumic Acid and Mucocutaneous Reactions: In Reply
Miriam Sturkenboom, PhDDepartments of Epidemiology and Biostatistics and Medical Informatics
Erasmus University Medical Center
3000 CA, Rotterdam, Netherlands
International Pharmacoepidemiology and Pharmacoeconomics Research Center
20033 Desio, Italy
Alfredo Nicolosi, PhD
Department of Epidemiology and Medical Informatics
Institute of Biomedical Technologies
National Research Council
20090 Segrate (Milano), Italy
Gertrude H. Sergievsky Center
School of Public Health
Columbia University
New York, NY 10032
Salvatore Mannino, MD
Azienda Ospedaliera "Istituti Ospitalieri"
26100 Cremona, Italy
Carlo Giaquinto, MD
Department of Pediatrics
University of Padova
35100 Padova, Italy
Luigi Cantarutti, MD
Societa' Servizi Telematici
35138 Padova, Italy
We recently published our population-based cohort study on the incidence of mucocutaneous reactions in children treated with niflumic acid, other nonsteroidal antiinflammatory drugs (NSAIDs), or nonopioid analgesics.1 This study followed an ad hoc, hospital-based, case-control study by Menniti-Ippolito et al,2 which raised a safety issue for pediatric use of niflumic acid. The 2 studies have contrasting findings, as occurs frequently in pharmacoepidemiology. Whereas Menniti-Ippolito et al found a 4.9-fold increased risk of mucocutaneous disorders with niflumic-acid use, we found no association. In light of these contrasting findings, we speculated in our discussion about potential reasons for the discrepancies.
The Naples study was initiated to study various adverse drug reactions in children, and all children who presented to the emergency unit with conditions that are often drug-related (ie, gastroduodenal ulcers, neurologic disorders, noninfectious mucocutaneous disorders, and thrombocytopenia) were included in the study.1 The association between use of niflumic acid and mucocutaneous disorders was assessed with reference to nonuse of NSAIDs, paracetamol, and antibiotics. Information on drug and vaccination use was gathered by interviewing the parents.
Menniti-Ippolito et al discussed our explanations for the contrasting findings, which were the choice of reference group, residual confounding, and protopathic bias. We would like to respond to their counterarguments.
We speculated that the choice of reference group by Menniti-Ippolito et al might have resulted in confounding by indication or residual confounding, because the reference group of children not using any drug (niflumic acid, paracetamol, and antibiotics) was certainly healthier than the group of those who were using niflumic acid and/or paracetamol and/or antibiotics. In our study, we compared niflumic-acid users with users of other NSAIDs or paracetamol to avoid confounding by indication. Menniti-Ippolito et al defended their choice of reference group by saying that niflumic acid and paracetamol are compared with the same nonuser reference group. Although this is true, it does not address the question of confounding by indication. To address the question, characteristics of the persons using niflumic acid or paracetamol and not using any drug should be compared. Their data underline the fact that confounding is strong: the risk estimate for niflumic acid changes from 8.2 to 4.9 on adjustment for other drugs, but it is not mentioned to which drugs this refers. The authors stated that additional adjustments were not possible because of the small sample size. In light of the strong confounding that was observed, we are still not convinced that residual confounding cannot explain at least a part of their results, especially because many niflumic-acid–exposed subjects in the study by Menniti-Ippolito et al used a combination of niflumic acid, antibiotics, and/or paracetamol. It is puzzling how independent effects of these drugs can be estimated if there are almost no children on monotherapy. Was niflumic acid associated with mucocutaneous disorders in those not using antibiotics? What happened when mutually exclusive categories were created of mono- and combination therapy? None of these issues were addressed in their study.
Without information on the exact mode-of-exposure assessment, we thought it was not possible to exclude protopathic bias in the study by Menniti-Ippolito et al. Protopathic bias (spurious association occurs because the drug is used to treat early symptoms of the disease/event) can only be avoided if the exact onset of disease is known and the exposure is assessed before that onset. It was not mentioned in their article, but if all of this was done, we agree that protopathic bias may not explain the results. To demonstrate the absence of protopathic bias, sensitivity analyses with various lag times should be presented.
Menniti-Ippolito et al suggest that our negative results may be explained by general misclassification. This was motivated by the fact that no association was found between antibiotics and mucocutaneous reactions in our study. However, their argument is based on a misunderstanding and/or incomplete reading of the results. Among NSAID or analgesic users, antibiotics were not associated with all grouped mild mucocutaneous reactions (see our Table 3 in ref 1). However, in our discussion we state that use of antibiotics during NSAID/analgesic use is associated with a significant 40% to 60% increased risk of erythema/rash and urticaria (the reactions they are known to cause), and the association was strongest for penicillins.
Each observational study has its limitations. The best way to inspect the robustness of the results is to conduct various sensitivity analyses. We are quite confident with our conclusion of no association, because our conclusions were not sensitive to a whole range of sensitivity and subgroup analyses such as in (or excluding) antibiotic users, changing exposure definition, and comparing against paracetamol or other NSAIDs. Contrasting findings should stimulate additional research. The study by Menniti-Ippolito et al raised a safety issue, and we reacted to that by trying another design and a database with prospective rather than retrospective data collection. We look forward to other studies that can add evidence.
REFERENCES
- Sturkenboom M, Nicolosi A, Cantarutti L, et al. Incidence of mucocutaneous reactions in children treated with niflumic acid, other nonsteroidal antiinflammatory drugs, or nonopioid analgesics.
Pediatrics. 2005;116
(1). Available at: www.pediatrics.org/cgi/content/full/116/1/e26
- Menniti-Ippolito F, Sagliocca L, Da Cas R, Saggiomo G, Di Nardo R, Traversa G. Niflumic acid and cutaneous reactions in children. Arch Dis Child. 2001;84 :430 –431
- Menniti-Ippolito F, Sagliocca L, Da Cas R, Saggiomo G, Di Nardo R, Traversa G. Niflumic acid and cutaneous reactions in children. Arch Dis Child. 2001;84 :430 –431
PEDIATRICS (ISSN 1098-4275). ©2006 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
Related articles in Pediatrics:
- Niflumic Acid and Mucocutaneous Reactions
- Francesca Menniti-Ippolito, Giuseppe Traversa, Roberto Da Cas, Roberto Leone, and Achille Caputi
Pediatrics 2006 117: 253-254.[Extract] [Full Text]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
