Send letter to journal:
Re: Niflumic acid and cutaneous reactions: a controversial association.

DEAR EDITOR,

We read the Short Report by Menniti-Ippolito et al[1] with interest. Their conclusions assert that the administration of niflumic acid to paediatric patients increases the risk of mucocutaneous reactions with statistical significance, and thus it would not seem opportune to prescribe this drug given the availability of other safer analgesics and antipyretics. We believe that their results cannot justify those conclusions given that their research presents a series of critical aspects that we list below.

1. The hospital surveillance was organised as a case control study. In this type of study, the cases must represent patients afflicted by the mucocutaneous pathology in the population, whereas the controls must represent the "non-cases" in the population. The controls can be found amongst subjects that have a pathology of origin that is not mucocutaneous, and that could be treated with niflumic acid. Menniti- Ippolito et al. chose their 162 controls amongst patients admitted for three different pathological conditions that were not mucocutaneous, but that nevertheless formed a sub-group not necessarily representative of the entire population of controls. In this way they may have introduced selection bias which would compromise the correct estimate of the odds for exposure of the controls, and would definitely create a bias in the odds ratio.

2. Menniti-Ippolito et al. assert that 79 cases and 162 controls were considered. This is not actually the case considering the analyses reported in Table 2. The calculation of the odds ratio in relation to the niflumic acid (as with the analyses in relation to the three other drug categories) includes only the "non users" amongst those not exposed. This, in fact, is the only way that one can derive a raw odds ratio of 8.2 -- if patients who had been administered another drug as well were also included amongst those exposed, the odds ratio would be 5.2. In the end, the results in relation to niflumic acid are based on a total of 27 cases and 53 controls. This choice goes against the original idea of the study, which was presented as a case control study. A case control study requires that the sample selection of cases and controls be done to observe the frequency of the exposition to the factor under examination (in this case, niflumic acid) for the two groups. Here, however, the choice of cases and controls is further conditioned by a specific type of non-exposure (note in Table 1 that 13 of the 15 exposed cases were also treated with other drugs). This way, a likely interaction between the niflumic acid and the other drugs in potentially provoking adverse reactions modifies the odds ratio and reduces its interpretability.

3. If one bears in mind how the non-exposed to the drug under consideration were actually selected, one can foresee a referral bias for the category of "non-users" cases in comparison to the cases that used niflumic acid. In fact, in the former, the tendency to go to an emergency room could be lower than for the latter who observed a reaction appear immediately after taking the drug and who already had an original pathology which would make one worry more.

4. Given the low sample size, adjusting the odds ratio for the concomitant use of other drugs is problematic. The methodology used for this adjustment should be specified. In any case, as regards what has been stated above, it might be more important to make adjustments for the type of initial pathology.

5. Cases and controls should be described by characteristics such as age, gender, and pathology of origin. In particular, since the pathology of origin can induce the mucocutaneous reaction (note that 12 of 58 "non users" are affected), a possible unbalance of cases and controls by pathology of origin could seriously alter the odds ratio. In other words, a defined pathology could confound the association here examined in that it could enhance both the mucocutaneous pathology and the use of niflumic acid. If this were the case, then the analyses could show an association between niflumic acid and mucocutaneous pathology, when in reality the association is between the pathology of origin and niflumic acid.

6. The cases are defined as patients affected by the mucocutaneous pathology from which the safety of the drug is judged in relation to other similar drugs. Safety should be evaluated comparatively based on all possible potential adverse reactions. The limited results of this study do not allow for asserting the greater safety of the other drugs for which a comprehensive comparison has not been done.

References

(1) 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.
(2) Schlesselman JJ. Case-Control Studies - Design, Conduct, Analysis. New York: Oxford University Press 1982:80.

Francesco Grigoletto, ScD
Department of Environmental Medicine and Public Health
University of Padua - School of Medicine
Padova, Italy

Send letter to journal:
Re: Use of niflumic acid in children

Dear Editor

The recent report of serious mucocutaneous reactions in fifteen Italian children who took niflumic acid in general practice raises concerns about the rational use of drugs in children everywhere [1]. Niflumic acid (and its beta-morpholinoethyl ester: morniflumate) is a non-steroidal anti-inflammatory drug (NSAID) that has been marketed in Italy since 1981 and also available for use in children in France and Spain. In a few other European countries the drug is registered only for use in adults, whereas it has not been marketed in UK. Although children seem less prone to side-effects of NSAIDs than adults,[2] previous French and Spanish reports documented both severe cutaneous reactions and renal impairments in children taking niflumic acid.[3] More specifically, since 1994 2 cases of Lyell's syndrome and 10 cases of acute renal failure in 6 month-14 year old children have been already reported. Despite a search of the Medline and Embase databases, and manual search of bibliographic citations, no robust evidence on which to base the rational use of this drug in children was found. Controlled clinical trials involving niflumic acid were scant, had methodological limitations, had small enrolled population sizes, and their results published mainly on national journals, most of whom were not indexed by the bibliographic databases.[3] Nevertheless, niflumic acid is widely used in Italian children. A recent survey of 27,777 children cared for by 35 general practitioners, showed that the prescription rate of niflumic acid was 57 per 1000 children, compared to 125.7 per 1000 for acetaminophen.[4] The most frequent therapeutic indications related to niflumic acid use, alone or with an antibiotic, were pharyngotonsillitis, common cold and acute otitis media. In an attempt to clarify the observed drug's profile we searched and reviewed twelve international therapeutic guidelines for these conditions.[3] Consistent agreement was found suggesting acetaminophen as the first line choice, followed by ibuprofen for the relief of mild to moderate pain, inflammation and fever in children with ear, nose and throat diseases. No other NSAID were suggested. In a context such as this one, characterised by frequent utilisation of certain compounds with a scantly documented risk/benefit profile and a lack of compliance with therapeutic guidelines, the prescribing attitudes in children can be considered largely 'evidence unbased'. Despite progress towards harmonising licensing processes in Europe and the attempts to setting up an efficient international postmarketing drug surveillance system, a guarantee of a rational use of drugs, particularly in children, must be achieved.[5] The case of niflumic acid here described is not the only case of a drug used in an irrational way in children and is not exclusive to Italy. It is only one of the distortions that a systematic and continuous monitoring of drug use on the national level can identify. One of the many transnational challenges could be an initiative capable of providing continuous information and effective education concerning (also) the appropriate use of drugs, for health professionals, children and their parents.

References
(1) Minniti-Ippolito F, Sagliocca L, Da Cas R, Saggiomo G, Di Nardo R, Traversa G and the Santoro Study Group for Adverse Drug Reactions in Children. Niflumic acid and cutaneous reactions in children. Arch Dis Chil 2001;84:430-431.
(2) RCPCH (1999). Medicines for Children. RCPCH Publications Limited, London.
(3) References available on request.
(4) Cazzato T, Pandolfini C, Campi R, Bonati M and the ACP Puglia- Basilicata Working Group. Drug prescribing in out-patient children in Southern Italy. Eur J Clin Pharmacol; in press.
(5) Bonati M, Impicciatore P, Pandolfini C. Medicines for Children in Europe at the Beginning of the New Millennium. Paediatr Perinatal Drug Ther 2001;4:82-84.

Send letter to journal:
Re: Niflumic acid and cutaneous reactions in children: a reply

Dear Editor,

We saw the letter by Grigoletto and Viola [1], in response to our article [2]and we would like to reply to the specific comments that were put forward. The first remark relates to the possibility that we may have introduced a bias in the choice of the conditions that were selected as controls. The selection of the control group is always a critical aspect in the design of a case-control study. In order to reduce biases, two principles should guide the selection process: controls should represent the source population of cases (in our study, those people that had developed a cutaneous disease of interest would have been included as cases); the selection of controls should not be influenced by the exposure status. In our opinion, both aims were attained. With regard to the source population of children hospitalized with cutaneous diseases, children hospitalized with the three other acute conditions, which represented the control group (neurological disorders, endoscopically confirmed gastroduodenal lesions, and thrombocytopenia), were referred to the same hospital and were hospitalized through the same emergency department. Most presumably, had any of the children included in the control group developed a cutaneous disorder he or she would have used the same hospital and would have been included in the case series.

Grigoletto and Viola also suggest the possibility that “non users” cases (i.e. children who did not receive any drug during the 3 weeks preceding the onset of symptoms) may have a lower “tendency to go to the emergency room” than cases that used niflumic acid. It needs to be emphasized that only children that were hospitalized through the emergency department were included in our study. Given the severity of the conditions, it is highly unlikely that the exposure status (of niflumic acid as of any other drug) could have affected the decision to hospitalize a child (it has also to be considered that drug exposure was ascertained during the hospitalization and not at the presentation at the emergency department). Moreover, there is no reason to expect that the exposure status may have affected differentially cutaneous cases and the three other acute conditions included in the study as controls.

Grigoletto and Viola take the view that the decision to calculate the crude odds ratios using as reference the “non-users” category is inconsistent with the case-control design of the study. We disagree. To clarify the point, simply consider the case of a case-control study aimed at estimating the effect of cigarette smoking on the risk of lung cancer. If the reference category were not limited to the never-smokers, but also included, say, ex-smokers, and smokers of less than 2 packs per day, the result would be an underestimate of the true OR. The mistake can go as far as to reverse a true association. For instance, if light smokers were compared with a category that mix together non smokers and heavy smokers, it might even be that light smoking appears to protect against the occurrence of lung cancer. The same apply in the case of drug exposure. The difference between the crude OR we calculated using the non-users as reference category (OR=8.2) and the OR that would be obtained by including in the reference category a mix of non-users and of users of other drugs (OR=5.2) can be interpreted as the estimate of the bias that would be introduced by using this erroneous approach. Furthermore, only by using the non-user category as reference it is possible to compare ORs of different drugs.

We were also well aware of the fact that cases and controls were frequently users of various categories of drugs, and consequently, for each of the considered categories of exposure, we adjusted the odds ratios for concomitant use of other drugs (we used the logistic model, BMDP software). As expected, the adjusted odds ratios were lower than the crude ones, which clearly support the need for the adjustment. For instance, the odds ratio for niflumic acid decreased from 8.2 to 4.9, with 95% confidence intervals still highly significant (1.9-12.8). Incidentally, the fact that the adjusted odds ratio is statistically significant suggests that the sample size was large enough to allow for adjustment. It has to be noted that for paracetamol there was a corresponding decrease in the odds ratios, from a crude estimate of 2.4 to 1.2 after adjusting for concomitant use of other drugs.

A further remark relates to the possibility that the increase in the risk associated to the use of niflumic acid might in fact be attributable to the clinical conditions for which the drug was indicated. We consider this possibility remote. The distribution of these conditions, mainly represented by upper respiratory tract infections, was relatively similar within the four study groups. Moreover, to act as a confounder, the condition that motivated the use of drugs would need to be selectively treated by the physicians with niflumic acid. The question of confounding has to be assessed on realistic terms: as far as we know there is no known condition that is both an important determinant of severe cutaneous diseases and for which niflumic acid is specifically indicated.

Grigoletto and Viola are correct in affirming that, in this study, we did not investigate the entire risk profile of niflumic acid in comparison with similar drugs. However, as reported in the international literature, paracetamol is considered the drug of choice in children as antipyretic and analgesic. Our study provides further evidence of the relative safety of paracetamol: the odds ratio of serious mucocutaneous reactions was 4.9 among users of niflumic acid and 1.2 among users of paracetamol.

The final point we would like to add is that almost all the comments made by Grigoletto and Viola appear as the literal translation of a two page annex of a report submitted to the Italian Ministry of Health, in response to our study, from one of the companies that markets niflumic acid in Italy. Moreover, in that report, the comments were specifically attributed to the statistical unit of the company. We obviously do not know who really contributed those comments. Nonetheless, in our opinion, in submitting their letter to the Archives the authors should have acknowledged this fact.

References

(1) Grigoletto F. Niflumic acid and cutaneous reactions: a controversial association. Arch Dis Child 2001, eLetter.
(2) 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.

Francesca Menniti-Ippolito, Giuseppe Traversa, Roberto Da Cas
Department of Epidemiology and Biostatistics,
Istituto Superiore di Sanità,
Viale Regina Elena 299, 00161 Rome, Italy
Luciano Sagliocca, Giovanni Saggiomo, Rocco Di Nardo
Santobono Hospital,
Via Mario Fiore 6,
80129 Naples, Italy