PEDIATRICS Vol. 107 No. 5 May 2001, pp. 1108-1115

Invasive Group A Streptococcal Infection and Nonsteroidal Antiinflammatory Drug Use Among Children With Primary Varicella

Samuel M. Lesko, MD, MPH^*, Katherine L. O'Brien, MD, MPH^{, §,} , Benjamin Schwartz, MD^{, ¶}, Richard Vezina, MPH^*, and Allen A. Mitchell, MD^*

From the ^* Slone Epidemiology Unit, School of Public Health, Boston University School of Medicine, Brookline, Massachusetts;  Respiratory Diseases Branch, Division of Bacterial and Mycotic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia; ^§ Epidemic Intelligence Service, Epidemiology Program Office, Centers for Disease Control and Prevention, Atlanta, Georgia;  Center for American Indian and Alaskan Native Health, Johns Hopkins University School of Public Health, Baltimore, Maryland; and the ^¶ National Immunization Program, Centers for Disease Control and Prevention, Atlanta, Georgia.

	ABSTRACT

Top Abstract Methods Results Discussion References

Objectives.  To test the hypothesis that nonsteroidal antiinflammatory drug use increases the risk of necrotizing soft tissue infections and, secondarily, all invasive group A streptococcal (GAS) infections in children with primary varicella infection.

Methods.  We conducted a prospective, multicenter case-control study among children <19 years old. Cases were children hospitalized with primary varicella complicated by invasive GAS infection or necrotizing soft tissue infection identified by a network of 45 pediatric infectious disease specialists located throughout the United States. Controls were children with uncomplicated primary varicella residing in the same communities as the cases. Data on medical history, clinical features of the varicella infection, signs and symptoms of infectious complications, and medication use were collected by structured telephone interviews. Univariate and multivariate matched odds ratios were calculated using conditional logistic regression.

Results.  Between June 1996 and September 1998, 52 cases of invasive GAS infection, including 21 with necrotizing soft tissue infection, and 172 controls with uncomplicated primary varicella were enrolled. Risk of invasive GAS infection was increased among children who were nonwhite (multivariate odds ratio [OR] 3.8, 95% confidence interval [CI]: 1.4-11), living in low-income households (OR 5.1, 95% CI: 1.7-15), exposed to varicella at home (OR 6.4, 95% CI: 2.6-16), or had a persistent high fever (OR 9.6, 95% CI: 2.8-33). Antipyretic regimen was associated with several measures of varicella illness severity among the controls. The risk of necrotizing soft tissue infection was not associated with the use of ibuprofen before the development of signs or symptoms of this complication (OR 1.3, 95% CI: 0.33-5.3). Risk of any invasive GAS infection was increased among children who had received ibuprofen (OR 3.9, 95% CI: 1.3-12), but not acetaminophen (OR 1.2, 95% CI: 0.50-3.0). However, there was no evidence of increasing risk with increasing duration of ibuprofen use. Subgroup analyses revealed that the risk of invasive GAS infection was increased only among children who had received both acetaminophen and ibuprofen.

Conclusions.  These data do not support the hypothesis that nonsteroidal antiinflammatory drugs, or ibuprofen in particular, increase the risk of necrotizing GAS infections. A statistically significant association was observed between nonnecrotizing invasive GAS infection and ibuprofen use; however, because of potential confounding, the meaning of this unexpected result is unclear. Nonetheless, these data suggest that parents use ibuprofen or ibuprofen together with acetaminophen to treat high fever and severe illness, which seems to identify children at high risk for invasive GAS infection.  Key words:  children, confounding, group A streptococcus, ibuprofen, necrotizing infection, NSAIDs, varicella.

Invasive group A streptococcal (GAS) infection, a potentially life-threatening illness, has become more common during the past decade in part because of the reappearance of virulent M-serotype strains, which may produce pyrogenic exotoxins.^1,2 Infection can result in a variety of clinical syndromes including bacteremia, pneumonia, myositis, and necrotizing fasciitis (NF). NF may require extensive surgical debridement or amputation, may be complicated by streptococcal toxic shock syndrome (STSS),³ and can result in prolonged hospitalization or death. Although other organisms can cause NF, the majority of cases are attributable to GAS. This is particularly true among children with varicella, which is a known risk factor for NF,^4-6 presumably because varicella's numerous skin lesions provide a portal of entry for GAS. In recent years, clusters of NF cases among children with primary varicella infection have been reported from California, Washington, Massachusetts, and Ontario, Canada.^4,7-9

Coincident with the increased frequency of severe, invasive GAS infections, numerous case reports have linked NF with the use of nonsteroidal antiinflammatory drugs (NSAIDs).^7,10-22 These reports, along with in vitro studies demonstrating decreases in neutrophil function after exposure to NSAIDs,^23-25 have resulted in the suggestion that NSAIDs may play a role in the cause of varicella-associated NF. Therefore, we conducted a prospective multicenter, case-control study to test the hypothesis that NSAID use increases the risk of invasive GAS infection, with a primary interest in necrotizing infections, in children with varicella.

	METHODS

Top Abstract Methods Results Discussion References

This study, designed by investigators from Boston University and the Centers for Disease Control and Prevention, was conducted between June 1996 and September 1998 with the guidance of an independent Advisory Committee (see "Acknowledgments"). The study design and consent documents were approved by the Institutional Review Board for Human Research at Boston University Medical Center.

Case Identification

Cases were defined as children <19 years old who were hospitalized with necrotizing soft tissue infections (NSTI) or other invasive GAS infection within 2 weeks of onset of primary varicella. Because the presence of fasciitis may not always be documented histologically, we use the more inclusive term NSTI to refer to patients with evidence (clinical or pathologic) of soft tissue necrosis, including those with documented NF. Cases were identified prospectively by a network of 45 pediatric infectious disease specialists at hospitals throughout the continental United States. Patients were judged to have an invasive GAS infection if GAS was isolated from a specimen obtained from a normally sterile site (eg, blood, joint, or pleural space) or exudate from a severe soft tissue infection. One of the investigators (K.O.B.) classified cases with regard to necrotizing complications by reviewing inpatient medical records after all references to medication exposures had been deleted. Cases were considered to have NSTI if there was any mention of gross or histologic evidence of soft tissue necrosis in pathology reports or surgical or medical notes, regardless of culture results. Patients with a history of immunodeficiency, chronic systemic steroid use, or contraindication to treatment with acetaminophen or any NSAID were not eligible for enrollment.

Collaborating pediatric infectious disease specialists identified eligible cases at their facilities by reviewing microbiology logs and by communicating with clinical colleagues in intensive care, emergency medicine, surgical specialties, and infection control. (Because of the unusual nature of the illness, collaborating physicians saw most of these children in consultation.) Study staff contacted all participating physicians every 2 weeks to ensure complete reporting of eligible patients. After a potential case was identified, informed consent was obtained from the parent/guardian ("parent"), the study-coordinating center was notified, and the child's parent was interviewed by telephone as soon as practical after diagnosis.

Control Selection

Controls were children with primary varicella infection who did not develop invasive GAS infection but came from the same base population as the cases. The primary care physician for each case was asked to identify 1 or more colleagues with comparable practices in the same community; if the primary care physician could not provide a referral, pediatricians or family physicians from the same community were selected at random from the American Medical Association physician directory. These control physicians were asked to identify patients in their practices who were <19 years old and had a recent primary varicella infection. After providing informed consent, the parents of potential controls were interviewed by telephone. Control interviews were scheduled so that the interval between onset of varicella and the interview was comparable to the corresponding interval for the matched case. We attempted to enroll 4 controls for each case.

Data Collection Procedures

Data were collected by a trained nurse-interviewer using the same computer-assisted, structured telephone interview for cases and controls. The data gathered included demographic and medical history information, a summary of symptoms occurring during the 7 days before the onset of varicella, detailed information on the timing and severity of symptoms of the primary varicella infection, and clinical signs and symptoms of any secondary complication (eg, rash, fever, pain, swelling). The route used to measure each temperature was recorded, and all measurements were converted to the equivalent oral temperature. The data included summary information for analgesic/antipyretic medications used in the 7 days before the onset of varicella and detailed information concerning all medications used from the onset of varicella until 14 days after onset or hospitalization (whichever occurred first). The medication information included the timing of each dose of any analgesic/antipyretic received during the varicella illness. To identify factors associated with the decision to treat with an antipyretic and the specific medication chosen, we inquired about the reason parents chose one medication over the other in the past and the relation between clinical features of the current varicella illness and the antipyretic regimen used. To facilitate the correct identification of analgesic/antipyretic medications, we provided the parent a booklet containing color images of the most common nonprescription pediatric analgesic/antipyretic products.

To test the hypothesis of interest, we necessarily restricted our analysis to medications used before the occurrence of the invasive GAS infection. Therefore, it was important to establish the timing of the onset of the invasive GAS infection as accurately as possible. This required distinguishing varicella-associated fever from fever associated with a secondary infection. To do this, two investigators (K.O.B., S.M.L.) independently reviewed the portion of each case interview detailing the chronology of symptoms. The reviewers were aware of the date of hospital admission and site of invasive GAS infection but were blinded to all medication use. The time of onset of a complication, the index hour, was defined for each case by the first report of any of the following:

1. A fever (temperature [T] 37.8°C) occurring 4 or more days after the onset of the varicella rash in a child who had been afebrile for at least 24 hours;
2. A fever (T 37.8°C) occurring during the first 3 days of the varicella illness in a child who had been afebrile for at least 24 hours if followed within 8 hours by local signs or symptoms compatible with invasive GAS infection or NSTI;
3. One or more local signs or symptoms including pain, swelling, skin discoloration, exudate, guarding, limping or reduced movement of an involved limb, cough, or difficulty breathing (the last 2 were used for cases with pneumonia or empyema); or
4. Two or more of the following systemic signs or symptoms occurring within an 8 hour period (used only for cases with isolated GAS bacteremia): chills, myalgias, parent's description of the illness as "severe', or need for bed rest; the time of the first symptom was used to determine the index hour.

Each control was assigned an index hour such that the interval between the onset of varicella and this hour equaled the corresponding interval for the matched case.

The exposure period of interest included medications used during the 7 days before the index hour; exposures more remote from the complication were judged unlikely to affect the risk of invasive GAS infection. Because of potential misclassification of the index hour and timing of medication use, we excluded from our analysis medication exposures occurring within the 12 hours immediately preceding the index hour. Therefore, we defined the exposure window as the period beginning 7 days before the index day and ending 12 hours before the index hour. As an example: for a case whose index hour is 8:00 AM on March 10, the exposure window would be from midnight on March 3 until 8:00 PM on March 9.

Analysis

We used the ² test to compare proportions and the Wilcoxon rank-sum test to compare continuous variables. The exposure odds ratio (OR) was used as the measure of association between medication use and the risk of invasive GAS infection and NSTI. Because cases and controls may have been given both acetaminophen and ibuprofen during the exposure window, we evaluated the distribution of mutually exclusive categories of exposure (acetaminophen only, ibuprofen only, and both medications). We calculated univariate and multivariate matched ORs using conditional logistic regression. Ninety-five percent confidence limits were calculated using the test-based method.²⁶ We created a multivariate model to identify clinical and demographic risk factors for invasive GAS infection and used this model to control for confounding in analyses involving medication exposures. We first tested a model including indicator terms for race, ethnicity, sex, age of the child, respondent's education level, household income, household crowding (persons/bedroom), duration of high fever (T >39.4°C), duration of rash, number of varicella lesions, mucosal involvement, and location of exposure to varicella (ie, home vs school or other location). We reduced this model to include only significant predictors of risk and retested each of the excluded terms by individually adding them back into the model to determine if any were significant; none was. The final model included terms for race, household income, duration of high fever, and location of exposure to varicella. We used this model to control for confounding in the relationship between medication use and the risk of invasive GAS infection and NSTI.

	RESULTS

Top Abstract Methods Results Discussion References

Collaborators at 25 tertiary care pediatric hospitals from across the United States identified 97 children hospitalized with possible invasive GAS infection or NSTI. Of these, 16 were ineligible (a diagnosis of invasive GAS infection or NSTI could not be confirmed by hospital records for 12, the complication was first recognized >14 days after varicella for 2, and 1 child each was allergic to acetaminophen and ibuprofen), 5 refused to participate in the study, 7 were lost to follow-up or could not be interviewed after 10 or more attempts, 1 did not complete a satisfactory interview, and no eligible controls could be found for 16. Most of the 16 unmatched cases occurred late in the second varicella season, so it was not possible to identify controls with uncomplicated varicella before the end of the study. The remaining 52 cases are included in the present analysis; 33 (63%) were diagnosed during the first varicella season (before October 15, 1997). Of the 52 cases, 21 had evidence of NSTI (20 confirmed by histology; 19 had cultures positive for GAS), 18 had severe soft tissue infection without clinical or histologic evidence of soft tissue necrosis (all had cultures positive for GAS), and 13 children had other culture-positive invasive GAS infections (pneumonia [n = 5], isolated bacteremia [n = 4], septic arthritis [n = 3], osteomyelitis [n = 1]). Most cases (45/52; 87%), were classified as having severe disease, defined by a need for surgery, treatment in an intensive care unit, receipt of vasopressors, occurrence of STSS (n = 7), a hospital stay exceeding 9 days, or death (1 child). Parents were interviewed a median of 27 days after the onset of varicella. Whether defined as invasive GAS infection or NSTI, all cases were analyzed together as the invasive GAS infection group.

Primary care physicians in the case communities identified 226 children with primary varicella infection who represented potential control participants. Of these, 1 was ineligible because of chronic antibiotic therapy, 2 refused to participate, 5 could not be interviewed within 8 weeks of onset of varicella, 25 were lost to follow-up or could not be interviewed after 10 or more attempts, 1 did not complete a satisfactory interview, and 20 were excluded because the matched case was excluded. None of the remaining 172 children suffered an infectious complication of varicella, and all were included as controls. Of these, 103 (60%) developed varicella before October 15, 1997.

The mean age of the cases was 58 months; for the controls, it was 62 months. Descriptive characteristics of the cases and controls are shown in Table 1; clinical features of the varicella illness among these children are shown in Table 2. (The demographic and clinical features of the 16 cases excluded for lack of an eligible control were similar to the cases studied13% were nonwhite, 31% Hispanic, 50% male, 19% lived in households with incomes of <$15 000, 38% had >100 pox lesions, 44% were exposed to varicella at home, and 19% had NSTI.) For 71% of cases studied (37/52), the index hour was determined by the onset of a local sign or symptom, 15% (8/52) by new fever, and 14% (7/52) by other symptoms (eg, cough, chest pain, chills, and myalgias). The median interval between onset of the varicella rash and the index hour was 3 days (range: 0-11 days). Varicella occurred in the controls a median of 79 days (interquartile range: 42-129 days) after varicella onset in cases. Compared with the cases, controls were more likely to be white, live in households with higher income and fewer persons per bedroom, were less likely to have been exposed to varicella in the home, had a lower mean maximum temperature, and had fewer days with fever >39.4°C (102.9^oF).

A multivariate model describing the independent relationship between demographic factors, clinical features of the primary varicella, and risk of invasive GAS infection is shown in Table 3. Factors not significantly associated with risk of invasive GAS infection are not included in this model. Risk was independently increased among nonwhites, children in low-income households, those exposed to varicella at home (ie, a secondary case in the same family), and those with a long duration of high fever.

Of the 224 children in this study, the parents of 123 reported that they had used acetaminophen and had also used ibuprofen (not necessarily in combination with acetaminophen) to treat their children. Effectiveness and recommendation of a health care provider were the reasons cited most often for acetaminophen, whereas high fever was the most commonly cited reason for ibuprofen use. Similar patterns were seen for case and control families.

Antipyretic use during the 7 days before and after the development of the varicella rash was examined among the controls. None received aspirin during this interval. Before the onset of varicella, use of ibuprofen and acetaminophen was uncommon, and none of these children used both medications on the same day. Medication use peaked on day 3 of varicella, when 38% were treated with acetaminophen, 12% with ibuprofen, and 6% received both medications. The cumulative exposure to any antipyretic medication during the 8 days beginning with the onset of rash was 81%; 54%, 16%, and 11% of controls received acetaminophen only, ibuprofen only, or both medications (not necessarily on the same day), respectively.

To test the hypothesis that medication use increases the risk of invasive GAS infection, we considered patients to be exposed only if they received the medication during the 7-day exposure window as defined above. During this period, 66% of controls received an antipyretic; 45%, 13%, and 8% received acetaminophen only, ibuprofen only, or both medications, respectively. Among the controls who received both acetaminophen and ibuprofen, 9 (69%) received both medications on the same day, and 6 of these received alternating doses of the 2 medications for at least some period. Clinical characteristics of the varicella illness according to antipyretic use among the controls are shown in Table 4. Choice of antipyretic regimen was significantly associated with a number of indicators of varicella severity (eg, fever, chills, somnolence, confinement to bed, and parental description of the illness as severe).

During the exposure window, 32 cases (62%) and 91 controls (53%) received at least 1 dose of acetaminophen. (Among the 16 cases excluded because we could not find matching controls, 63% had used acetaminophen.) The matched OR for acetaminophen use was 1.4 (95% confidence interval [CI]: 0.69-2.9); the multivariate OR, taking into account race, household income, source of varicella exposure, and duration of fever >39.4°C, was 1.2 (95% CI: 0.50-3.0). The matched OR for the subgroup of cases with severe invasive GAS infection (27 exposed, 18 unexposed) was 1.1 (95% CI: 0.55-2.4). For the subgroup of cases with NSTI (16 exposed, 5 not exposed) and their matched controls (31 exposed, 30 not exposed), the matched OR was 3.8 (95% CI: 0.92-16). The corresponding OR for nonNSTI cases (16 exposed, 15 unexposed) and their controls (60 exposed, 51 unexposed) was 0.91 (95% CI: 0.41-2.0).

During the same exposure time window, 18 cases (35%) and 36 controls (21%) received at least 1 dose of ibuprofen. (Among the 16 cases excluded because we could not find matching controls, only 6% had used ibuprofen.) The matched OR for ibuprofen use was 2.9 (95% CI: 2-6.9); the multivariate OR was 3.9 (95% CI: 1.3-12). For cases with severe invasive GAS infection (15 exposed, 30 unexposed), the matched OR was 2.9 (95% CI: 1.1-7.5). The magnitude of the association did not increase with increasing duration (in days) of ibuprofen use or number of doses received. Among patients receiving 1 to 2 doses of ibuprofen (9 cases, 14 controls), the matched OR was 3.4 (95% CI: 1.2-10); among those receiving 3 to 4 doses (6 cases, 10 controls), it was 3.4 (95% CI: 1.0-11); and among those receiving 5 or more doses (3 cases, 12 controls), it was 1.7 (95% CI: 0.40-7.1). For the subgroup of NSTI cases (6 exposed, 15 unexposed) and their controls (18 exposed, 43 unexposed), the matched OR was 1.3 (0.33-5.3). The corresponding OR for the nonNSTI cases (12 exposed, 19 unexposed) and controls (18 exposed, 93 unexposed) was 4.7 (1.5-14).

The distribution of mutually exclusive categories of antipyretic use during the exposure window among cases and controls is shown in Table 5. (Among the 16 cases excluded because we could not find matching controls, none was exposed to both acetaminophen and ibuprofen.) Among cases exposed to both medications, 10 (77%) received both on the same day, 6 in an alternating dose schedule. The matched ORs for exposure to acetaminophen only, ibuprofen only, and both medications were 0.96, 1.5, and 5.0, respectively. When all potential confounders were added to the model, the ORs for treatment with ibuprofen only and both medications were somewhat larger than, but not significantly different from, the corresponding matched ORs. Only the ORs for treatment with both medications were significantly >1.0. Of note, when the proportion of days with fever >39.4°C was not included in the model (ie, when confounding by duration of high fever was not controlled), the OR for exposure to both medications was 9.4 (95% CI: 2.3-38). For cases of invasive GAS infection with severe disease, the matched ORs were similar to the matched ORs for all cases. There were too few exposed cases to permit examination of mutually exclusive exposure categories stratified by the presence of NSTI among the cases.

	DISCUSSION

Top Abstract Methods Results Discussion References

In this relatively large study, we set out to test the hypothesis that among children with primary varicella the use of NSAIDs increases the risk of NF and, secondarily, whether NSAIDs increase the risk for all invasive GAS infections. We observed no association between the use of ibuprofen, the only NSAID used in these children, and NSTI. We did observe an association between ibuprofen use and invasive GAS infection overall. However, there was no evidence of a dose-response relationship, and in an analysis of clinically distinct subgroups, ibuprofen use was only associated with nonnecrotizing invasive GAS infection. Furthermore, the association between ibuprofen and invasive GAS infection was confined to children who had received both acetaminophen and ibuprofen in the 7 days preceding this complication, and this association was attenuated when duration of high fever was added to the multivariate model.

Three mechanisms have been proposed by which NSAIDs might influence the incidence or severity of GAS infections.²⁷ In patients with established invasive GAS infection, NSAIDs could increase the risk of systemic complications (eg, NF complicated by STSS or renal failure) either by a direct pharmacologic effect of the drug or because it masks symptoms and contributes to a delay in diagnosis and treatment. Secondly, among patients with localized GAS infection, NSAIDs might increase the risk of dissemination or invasion (eg, GAS cellulitis progressing to NF or bacteremia). Lastly, among patients without infection, NSAIDs could increase the risk of invasive GAS infection. The first hypothesis may be best evaluated using a case-case analysis in which patients with invasive GAS infection with complications are compared with those without the same complications. This hypothesis was not the subject of this study (we had too few cases with less-severe disease to allow us to test whether NSAID use was associated with disease severity) but has been examined by others. Investigators conducting surveillance of invasive GAS infection in Ontario, Canada found no association between premorbid NSAID use and the severity of infection, as defined by NF, STSS, or death.²⁸ Furthermore, using a rabbit model, Guibal et al²⁹ observed that exposure to diclofenac after the onset of infection significantly limited extent of NF. The present study was not designed to test the hypotheses that NSAID use might increase the risk of a superficial GAS infection becoming invasive. To study this hypothesis, one would need to examine medication exposures that occurred subsequent to the onset of the superficial infection but prior the onset of NF. Because localized soft tissue GAS infection and NF are 2 ends of a continuum and it is not possible to reliably identify the onset of NF, we believe this hypothesis is not amenable to epidemiologic study. We designed the current study to test the hypothesis that NSAID use increases the risk of developing invasive GAS infection, including NSTI.

There are a number of study design and analysis issues that must be considered regarding an observational study to test this hypothesis. Chief among these are 1) confounding by indication or the severity of the varicella illness; and 2) misclassification of the temporal relation between NSAID exposure and the onset of invasive GAS infection.

Confounding by indication occurs when patients who receive different treatments differ in their risk of an adverse outcome, independent of the treatment received.³⁰ Because the reported use of acetaminophen and ibuprofen differs according to varicella severityespecially the height and duration of fever (Table 4)and because illness severity is related to the risk of invasive GAS infection (Table 3), confounding by indication may account for the observed association between ibuprofen use and the risk of invasive GAS infection. Although we attempted to control for severity of the varicella illness by including relevant terms in the multivariate models, it remains possible that confounding was incompletely controlled. The clinical measures we used to control for confounding are not direct measures of illness severity, but are only surrogates (eg, duration of high fever, number of pox lesions, parental report of illness severity). Furthermore, antipyretic use alters the course of fever and the 2 are inextricably related; therefore, it is likely that controlling for the proportion of days with high fever only partially controlled for confounding in the present analysis.

The observation that risk of invasive GAS infection was increased only among children who received both acetaminophen and ibuprofen is compatible with a pharmacologic interaction between these medications, but no such interaction has been described. Rather, our data indicate that choice of antipyretic regimen is not random but is related to the parents' assessment of the severity of their children's illness. Compared with children who did not receive an antipyretic, those who received either medication alone were more severely affected by varicella, and those who received both medications were the most severely affected. This latter choice of antipyretic regimen likely reflects the clinical practice of administering alternating doses of acetaminophen and ibuprofen every 2 hours to control refractory fever in children; such use may identify the sickest children, who are at greatest risk of invasive GAS infection.

In these data, our attempt to control for confounding by duration of high fever produced a substantial decrease in the OR. It is possible that if confounding by severity of the varicella illness had been completely controlled, the observed association might have been reduced to the null value of 1. We believe that residual confounding may be the most likely explanation for the observed association.

The second issue relates to the potential for confusing cause and effect in the relation between symptoms of invasive GAS infection and NSAID exposure. NSAIDs are frequently used to treat the symptoms of what proves to be early invasive GAS infection (ie, pain and fever), but for NSAIDs to increase risk, they must have been used before the onset of the invasive GAS infection (ie, before symptoms have developed). This study was specifically designed to avoid confusing the effects of medications that might increase the risk of invasive GAS infection with exposures used to treat the early signs or symptoms of this complication of varicella. To this end, we made special efforts to accurately determine the temporal relationship between medication exposures and the onset of the invasive GAS infection. We collected detailed information concerning the onset of clinical features compatible with early invasive GAS infection and used these data to define an index hour for each case (blind with respect to medication use) and restricted our analyses to exposures that occurred before that time. Possibly, the invasive GAS infection may have started some time before its signs or symptoms were recognized by parents (ie, the index hour), or they may have inaccurately recalled the timing of these signs and symptoms and/or the timing of medication doses. However, by studying only exposures that occurred at least 12 hours before the index hour, we believe the majority of medication exposures actually occurred before the invasive GAS infection developed.

The present study has additional strengths that include its multicenter design, the relatively large number of cases of invasive GAS infection (including 21 with NSTI), and the availability of information on potential confounders. We minimized selection bias by maintaining frequent contact with collaborating pediatric infectious disease specialists to ensure prospective identification of all potential cases, regardless of previous antipyretic exposure, and by enrolling a high proportion of eligible cases. We sought to minimize recall bias by collecting data symmetrically from cases and controls and by scheduling control interviews so that the interval between interview and the onset of varicella was similar to their matched cases. We controlled for confounding by matching on location, type of medical practice, and time (controls were enrolled as soon as possible after a case interview was completed) and by including race, household income, and markers of illness severity in our multivariate models. These data are limited by the relatively large number of cases who could not be matched with controls and the small number of cases exposed to only ibuprofen, which produced wide confidence intervals around the point estimate for this exposure. The infrequent use of ibuprofen among the cases without matching controls suggests that their exclusion did not obscure a stronger association between invasive GAS infection and exposure to ibuprofen or both drugs.

In adults, NF has been linked to NSAID use by a number of case reports.^10-22 However, it is not possible to determine causality from such reports, and few controlled studies of invasive GAS infection in adults have been reported. Compared with unselected hospital patients, Barnham³¹ reported that NSAID use was higher in patients with streptococcal bacteremia. However, insufficient data were included in this brief report to determine if most NSAID exposures preceded or followed the first signs of infection.

The hypothesis that NSAID use might increase the risk of NF in children with varicella was first suggested by a series of 14 cases of NF complicating primary varicella, described by Brogan et al.⁷ Five received ibuprofen before hospitalization, but the timing of this exposure in relation to NF symptoms was not clear, and there was no control group for comparison. Two subsequent studies have attempted to document an association between NSAID use and invasive GAS infections. In the first, a case-control study of invasive GAS infection complicating varicella in children, Peterson et al⁸ observed a nonsignificant elevation in risk associated with the use of ibuprofen during the first 5 days of varicella. Information on the indications for or the specific timing of NSAID use relative to the onset of symptoms of the invasive GAS infection was not provided, and chance and bias cannot be ruled out as explanations for the observed risk. A more recent report by Zerr et al³² compared 19 children with varicella complicated by NF to 29 control children hospitalized with varicella complicated by cutaneous infection without evidence of necrosis. Although cases were significantly more likely than controls to have received NSAIDs before hospitalization (OR 10.2), all but 2 of the cases were given the medication after signs of NF had developed. This latter observation is incompatible with the hypothesis that NSAIDs cause NF. Potential explanations include NSAIDs being used to treat the symptoms of NF that had not yet been recognized clinically or that NSAIDs may have contributed to the progression of an already established GAS infection. Also, 14 of the 19 cases were included in the original case series described by Brogan et al,⁷ and it is unclear what steps the authors took to avoid the bias inherent in testing a hypothesis in the same data that were used to generate the hypothesis. Taken together, previously published reports do not provide evidence for a causal association between NSAID use and the initiation of invasive GAS infections overall or necrotizing infections in particular.

The present findings do not support the hypothesis that NSAID use increases the risk of necrotizing GAS infections. However, a statistically significant association between ibuprofen use and nonnecrotizing invasive GAS infection was observed. Because of potential confounding, the meaning of this unexpected association is unclear. Whether this association is causal or reflects residual confounding cannot be resolved despite these extensive data. Because of the difficulty controlling for confounding and accurately determining the exposure-to-disease relationship in an epidemiologic study, and because the incidence of varicella has been decreased by vaccination, additional investigation of NSAIDs and GAS infection may be best conducted with an animal model. What is clear from these data are that parents use ibuprofen, and particularly ibuprofen together with acetaminophen, to treat more severe varicella illness, and that the severity of the predisposing varicella illness may be a risk factor for subsequent invasive GAS infection. As such, it may be important to closely monitor such severely ill children for early signs of invasive GAS infection regardless of the use of antipyretics.

	ACKNOWLEDGMENTS

Major support for this research was provided by McNeil Consumer Healthcare, Fort Washington, Pennsylvania, with additional support provided by Whitehall-Robins Healthcare, Madison, New Jersey.

We thank the study advisory committee for their valuable advice and guidance throughout the study: Alan Leviton, MD (Chair), Children's Hospital and Harvard Medical School, Boston, Massachusetts; Sarah Long, MD, St. Christopher's Hospital for Children and MCP-Hahnemann University School of Medicine, Philadelphia, Pennsylvania; Richard Monson, MD, Harvard School of Public Health; Craig Rubens, MD, PhD, Children's Hospital and Medical Center and University of Washington School of Medicine, Seattle, Washington; Eugene Shapiro, MD, Yale-New Haven Hospital and Yale University School of Medicine; and Stanford Shulman, MD, Children's Memorial Hospital and Northwestern University School of Medicine, Chicago, Illinois.

Liaisons to the advisory committee were Anthony R. Temple, MD, Medical Affairs, McNeil Consumer Health Care; and Lisa M. Ford, MD, Clinical Research, Whitehall-Robins Health Care.

We also wish to thank Sandra Hatfield for coordinating data collection and the case locator network; Mary Kreiger, who conducted the telephone interviews; and Leonard Gaetano, for programming and data management assistance. We are especially grateful to the collaborating clinicians and institutions that composed our case locator network: Marsha Anderson and James Conway, Children's Hospital of Denver, Denver, Colorado; Ann Arvin, Stanford University Medical Center, Stanford, California; Elia Ayoub, Shands Hospital at the University of Florida, Gainesville, Florida; Parvin Azimi, Children's Hospital of Oakland, Oakland, California; William Barson, Columbus Children's Hospital/Ohio State University, Columbus, Ohio; Dale Bergamo, Henry Ford Hospital, Detroit, Michigan; Nasteen Bhumbra, Medical College of Ohio, Toledo, Ohio; Steven Black, Kaiser Hospital, Oakland, California; Subash Chaudhary, Southern Illinois University School of Medical, Springfield, Illinois; Marilyn Crain, University of Alabama at Birmingham, Birmingham, Alabama; Penelope Dennehy, Cheri Sarizen, and Mary O'Brien, Rhode Island Hospital, Providence, Rhode Island; Filip Dubovsky, Myron Levine, and Margaret Rennels, University of Maryland School of Medicine, Baltimore, Maryland; Kathryn Edwards, Vanderbilt University Hospital, Nashville, Tennesee; Arthur Frank and Daisy Mangat, Columbia Michael Reese Hospital and Medical Center, and University of Illinois at Chicago, Chicago, Illinois; Johanna Goldfarb and Carla Saracusa, Cleveland Clinic Foundation, Cleveland, Ohio; Carolyn Cleary and Jill Halterman, University of Rochester Medical Center/Strong Memorial Hospital, Rochester, New York; Samantha Hauger, Austin Diagnostic Children's Clinic, Austin, Texas; Ann Marie Higgins and Leonard Weiner, University Hospital-Syracuse, Syracuse, New York; David Hodes, Mount Sinai Medical Center, New York, New York; Jill Hoffman and Susana Aragon, Children's Hospital of Los Angeles, Los Angeles, California; Claudia Hoyen, Rainbow Babies and Children's Hospital, Cleveland, Ohio; Cecelia Hutto, University of Miami, Miami, Florida; Hal Jenson, University of Texas Health Science Center, San Antonio, Texas; George Johnson and Ronald Turner, Medical University of South Carolina, Charleston, South Carolina; Michael Ledwith, Jeff Kahn, Peter Hortez and Eugene Shapiro, Yale University School of Medicine, New Haven, Connecticut; Jerome Klein and Elizabeth Barnett, Boston Medical Center, Boston, Massachusetts; William Kock, Medical College of Virginia, Richmond, Virginia; Sarah Long, St. Christopher's Medical Center, Philadelphia, Pennsylvania; Kristine Macartney, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania; Judy Martin and Ellen Wald, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania; Maryann Mathis and Blaise Congeni, Children's Hospital Medical Center, Akron, Ohio; Edward O'Rourke, Children's Hospital of Boston, Boston, Massachusetts; Cindy Olsen-Burgess and Mary Anne Jackson, Children's Mercy Hospital, Kansas City, Missouri; Gary Overturf, University of New Mexico Hospital, Albuquerque, New Mexico; April Palmer, University of Mississippi Medical Center, Jackson, Mississippi; Larry Pickering, Children's Hospital of the King's Daughter, Norfolk, Virginia; William Raszka, Jr, University of Vermont College of Medicine, Burlington, Vermont; Lorry Rubin, Long Island Jewish Medical Center, New Hyde Park, New York; Gordon Schutze and Richard Jacobs, University of Arkansas, Little Rock, Arkansas; Stanford Shulman, Children's Memorial Hospital, Chicago, Illinois; Barbara Stechenberg, BayState Medical Center, Springfield, Massachusetts; Bradley Sullivan, Marshfield Clinic, Marshfield, Wisconsin; Russell Van Dyke, Tulane University Hospital, New Orleans, Louisiana; John Waldhausen, University of Washington School of Medicine, Seattle, Washington; and Rod Willoughby, Brijit Reis, and Beth Rockcress, Johns Hopkins School of Medicine, Baltimore, Maryland.

	FOOTNOTES

Dr Lesko and Dr Mitchell have previously conducted research sponsored by and served as consultants to McNeil Consumer Healthcare.

The opinions expressed in this publication are those of the authors and do not necessarily reflect those of the study's sponsors, the Centers for Disease Control and Prevention, or the US Public Health Service, nor does mention of trade names, commercial products, or organizations imply endorsement by the Public Health Service or the US government.

Received for publication Jun 29, 2000; accepted Oct 5, 2000.

Address correspondence to Samuel M. Lesko, MD, MPH, Slone Epidemiology Unit, 1371 Beacon St, Brookline, MA 02446. E-mail: leskos{at}bu.edu

	ABBREVIATIONS

GAS, group A streptococcal; NF, necrotizing fasciitis; STSS, streptococcal toxic shock syndrome; NSAID, nonsteroidal antiinflammatory drug; NSTI, necrotizing soft tissue infections; T, temperature; OR, odds ratio; CI, confidence interval.

	REFERENCES

Top Abstract Methods Results Discussion References

1. Schwartz B, Facklam AR, Breiman RF Changing epidemiology of group A streptococcal infection in the USA. Lancet 1990; 336:1167-1171 [CrossRef][Medline]
2. O'Brien KL, Levine OS, Schwartz B The changing epidemiology of group A streptococcus infections. Semin Pediatr Infect Dis 1997; 8:10-16
3. Centers for Disease Control Defining the group A streptococcal toxic shock syndrome. JAMA 1993; 269:390-391 [Abstract/Free Full Text]
4. Davies HD, McGeer A, Schwartz B, Invasive group A streptococcal infections in Ontario, Canada. N Engl J Med 1996; 335:547-554 [Abstract/Free Full Text]
5. Falcone PA, Pricolo VE, Edstrom LE Necrotizing fasciitis as a complication of chickenpox. Clin Pediatr 1988; 27:339-343
6. Wilson GJ, Talkington DF, Gruber W, Edwards K, Dermody TS Group A streptococcal necrotizing fasciitis following varicella in children: case reports and review. Clin Infect Dis 1995; 20:1333-1338 [Medline]
7. Brogan TV, Nizet V, Waldhausen JHT, Rubens CE, Clarke WR Group A streptococcal necrotizing fasciitis complicating primary varicella: a series of fourteen patients. Pediatr Infect Dis J 1995; 14:588-594 [Medline]
8. Peterson CL, Vugia DJ, Meyers HB, Risk factors for invasive group A streptococcal infections in children with varicella: a case-control study. Pediatr Infect Dis J 1996; 15:151-156 [CrossRef][Medline]
9. Centers for Disease Control and Prevention Outbreak of invasive group A streptococcus associated with varicella in a childcare centerBoston, Massachusetts, 1997. MMWR Morb Mortal Wkly Rep 1997; 46:944-948 [Medline]
10. Brun-Buisson DJL, Saada M, Trunet P, Rapin M, Roujeau J-C, Revuz J Haemolytic streptococcal gangrene and non-steroidal anti-inflammatory drugs. BMJ 1985; 290:1786
11. Krige JEJ, Spence RAJ, Potter PC, Terblanche J Necrotising fasciitis after diflunisal for minor injury. Lancet 1985; 2:1432-1433 [Medline]
12. Rimailho A, Bruno R, Richard C, Auzepy P Fulminant necrotizing fasciitis and nonsteroidal anti-inflammatory drugs. J Infect Dis 1987; 155:143-146 [Medline]
13. Frankish PD, Mason GH, Allen PR, Milsom FP, Christmas TI Acute streptococcal necrotising fasciitis. N Z Med J 1988; 101:625-626 [Medline]
14. Wojno K, Spitz WU Necrotizing fasciitis: a fatal outcome following minor trauma. Am J Forensic Med Pathol 1989; 10:239-241 [Medline]
15. Smith RJ, Berk SL Necrotizing fasciitis and nonsteroidal anti-inflammatory drugs. South Med J 1991; 84:785-787 [Medline]
16. Van Ammers PM, Moore PJ, Sacho H Necrotising fasciitis after caesarean section-association with non-steroidal anti-inflammatory drugs. S Afr Med J 1991; 80:203-204 [Medline]
17. Demey HE, Goovaerts GC, Pattyn SR, Bossaert LL Streptococcal myositis: a report of two cases. Acta Clinica Belgica 1991; 46:82-88 [Medline]
18. Chosidow O, Saiag P, Pinquier L, Nonsteroidal anti-inflammatory drugs in cellulitis: a cautionary mote. Arch Dermatol 1991; 127:1845-1846 [Abstract/Free Full Text]
19. Ricketts D, Teale G, Roberts J, O'Malley R Streptococcal gangrene presenting as hip pain: two case reports and review of the literature. Injury 1992; 23:134-135 [CrossRef][Medline]
20. Guy C, Gonthier R, Balique JG, Bertrand JC, Ollangnier M Fasciite necrosante fulminante apres injection intramusculaire d'un corticoide. (English translation: Fulminating necrotizing fasciitis after intramuscular injection of a corticoid. Contributing role of nonsteroidal antiinflammatory agents.) Therapie 1993; 48:489-490 [Medline]
21. Hird B, Byrne K Gangrenous streptococcal myositis: case report. J Trauma 1994; 36:589-591 [Medline]
22. Esperson GT, Larsen K NSAID and fasciitis necroticans. Ugeskr Laeger 1987; 149:3475-3477 [Medline]
23. Dong Y-L, Fleming D, Yan T, Herdon DN, Waymack JP Effect of ibuprofen on the inflammatory response to surgical wounds. J Trauma 1993; 35:340-343 [Medline]
24. Kaplan HB, Edelson HS, Korchak HM, Given WP, Abramson S, Weissmann G Effects of non-steroidal anti-inflammatory agents on human neutrophil functions in vitro and in vivo. Biochem Pharmacol 1984; 33:371-378 [CrossRef][Medline]
25. Venezio RR, DiVincenzo C, Pearlman F, Phair JP Effects of the newer nonsteroidal anti-inflammatory agents, ibuprofen, fenoprofen and sulindac on neutrophil adherence. J Infect Dis 1985; 152:690-694 [Medline]
26. Miettinen OS Estimability and estimation in case-referent studies. Am J Epidemiol 1976; 103:226-235 [Abstract/Free Full Text]
27. Stevens DL Could nonsteroidal antiinflammatory drugs (NSAIDs) enhance the progression of bacterial infections to toxic shock syndrome? Clin Infect Dis 1995; 21:977-979 [Medline]
28. McGeer A, Green K, Schwartz B, Kaul R, Low DE. Do non-steroidal anti-inflammatory drugs predispose to severe disease due to group A streptococci? [abstract]. Presented at: 35th Interscience Conference on Antimicrobial Agents and Chemotherapeutics, American Society of Microbiology; September 17-20, 1995; San Francisco, CA
29. Guibal F, Muffat-Joly M, Terris B, Garry L, Morel P, Carbon C Effects of diclofenac on experimental streptococcal necrotizing fasciitis (NF) in rabbit. Arch Dermatol Res 1998; 290:628-633 [CrossRef][Medline]
30. Slone D, Shapiro S, Miettinen OS, Drug evaluation after marketing. Ann Intern Med 1979; 90:257-261
31. Barnham M Nonsteroidal antiinflammatory drugs: concurrent or causative drugs in serious infection? [letter]. Clin Infect Dis 1997; 25:1272-1273 [Medline]
32. Zerr DM, Alexander ER, Duchin JS, Koutsky LA, Rubens CE A case-control study of necrotizing fasciitis during primary varicella. Pediatrics 1999; 103:783-790 [Abstract/Free Full Text]