Outpatient Fluoroquinolone Use in Children, 2000–2018

Abstract

Background

While fluoroquinolones are commonly used in adults, the use in children has been low. Since 2000, there were 3 US Food and Drug Administration (FDA) Boxed warnings regarding fluoroquinolones (2008, 2013, and 2016). Our objective was to describe the use of fluoroquinolones in children and assess the impact of 3 recent FDA warnings on fluoroquinolone use.

Methods

From 2000 to 2018, we assessed claims for all outpatient prescription fills to measure the use of systemic fluoroquinolones and other broad-spectrum antibiotics in children  less than 18 years old in the MarketScan Commercial Claims and Encounters database. We describe demographics, indication for antibiotic, and clinical characteristics. To assess the impact of FDA warnings on fill rates, we conducted an interrupted time-series analysis.

Results

The cohort included 34.6 million unique beneficiaries less than 18 years old with 441 062 fluoroquinolone fills (5.5 fills per 1000 person-years). The fluoroquinolone fill rate was highest among children > 11 years old. Urinary tract infection was the most common associated diagnosis (21.8%). Since 2008, the fluoroquinolone fill rate has declined. By the end of the study period in December 2018, in the (counterfactual) absence of the FDA warnings, fluoroquinolone fill rate would have been 7.5 (95% confidence interval [CI]: 5.2-9.7); however, the corresponding rate in observed data was 2.8 (95% CI: 1.7-3.9).

Conclusions

Fluoroquinolone use was low compared with other common broad-spectrum antibiotics and declining trends over time were associated with FDA warnings, even though these warnings were not pediatric specific. Future work should assess the adverse events at issue in these warnings in children.

Fluoroquinolones are commonly used in adults but rarely in children [1–5]. In the 1970s, studies in juvenile animals identified potential musculoskeletal adverse effects, and pediatric research on these antibiotics was halted [6]. Fluoroquinolones have 3 US Food and Drug Administration (FDA)-approved pediatric indications: complicated urinary tract infection (UTI) and pyelonephritis, inhalation anthrax postexposure, and plague [7]. However, off-label medication use in children is common [8], and the broad-spectrum coverage, high tolerability and bio-availability, and easy oral dosing make fluoroquinolones attractive and a critical target for antimicrobial stewardship efforts [9]. The American Academy of Pediatrics (AAP) statements about fluoroquinolones [6, 10, 11] addressed many potential uses, including respiratory tract infections, gastrointestinal infections, and prophylaxis/empiric therapy for neutropenic patients. The 2011 statement stated that “fluoroquinolones are reasonably safe in children,” but should be reserved for “treatment of infections for which no safe and effective alternative exists” [10].

In 2008, the FDA issued a Boxed warning regarding the increased risk of tendinitis and tendon rupture associated with fluoroquinolone exposure [12, 13]. Another FDA Boxed warning followed in 2013 regarding the risk of peripheral neuropathy [14]. These warnings did not specifically mention children and data concerning these adverse events in children are limited. In 2016, the FDA updated the Boxed warning stating that “serious side effects associated with fluoroquinolone antibacterial drugs generally outweigh the benefits for patients . . . who have other treatment options” [15, 16]. There have been conflicting reports of recent trends of pediatric fluoroquinolone use [5, 17], and the impact of these FDA warnings has not been assessed. Perceptions of potential harms in children predate these warnings and have likely contributed to historically low use in children [4]. Better understanding the epidemiology of fluoroquinolone use in children will improve the design and interpretation of comparative effectiveness and safety studies and helps contextualize this antibiotic class as a potential target for outpatient antibiotic stewardship.

Our objective was to describe the use of systemic, outpatient fluoroquinolone antibiotics in children in the United States from 2000 to 2018 and to examine the impacts of the FDA warnings on temporal trends of fluoroquinolone use in children.

PATIENTS AND METHODS

Data Source, Study Setting, and Population

Data were obtained from the IBM Watson Health MarketScan Commercial Claims and Encounters database (2020 IBM Watson Health). These data contain de-identified, individual-level adjudicated claims data on inpatient and outpatient diagnoses and procedures, and outpatient pharmacy dispensing for beneficiaries across the United States with employer-sponsored private insurance sponsored by approximately 350 large employers. Our study included beneficiaries less than 18 years of age with fee-for-service and prescription drug coverage between January 1, 2000, and December 31, 2018.

Dispensing of Systemic Antibiotics

We assessed dispensed outpatient prescription claims (henceforth, “fills”) for systemic fluoroquinolones (ciprofloxacin, levofloxacin, moxifloxacin, gatifloxacin, ofloxacin, gemifloxacin, norfloxacin, lomefloxacin, and sparfloxacin) and comparator broad-spectrum systemic antibiotics (amoxicillin-clavulanate, cefixime, cefdinir, nitrofurantoin, and sulfamethoxazole-trimethoprim [SMX-TMP]). These comparators, treatment alternatives to fluoroquinolones for UTIs, were selected to provide context for patterns of use and temporal trends. To classify fills, we identified National Drug Codes by searching for the generic names in National Drug Data File Plus (First Databank, www.firstdatabank.com/) and Red Book (IBM Watson Health). Multiple claims for a specific antibiotic (ie, generic name) on the same date were counted as a single fill, and we retained the maximum days supplied among the observed claims as the days supplied for that fill.

We estimated the rate of antibiotic fills as the number of fills per 1000 person-years with prescription drug coverage. We calculated standard errors assuming a Poisson distribution (Rothman et al [18], p. 267).

Indication for Antibiotic Fills

To describe indication, we assessed International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9-CM and ICD-10-CM) diagnosis codes from outpatient and inpatient records on the day of the antibiotic fill (index fill) or in the 3 days before. Codes were grouped into 3 hierarchically defined tiers: tier 1, antibiotics almost always indicated; tier 2, antibiotics may be indicated; and tier 3, antibiotics not indicated [19]. An index fill qualified as tier 1 if any tier 1 diagnosis code was present; if none were present, then tier 2 codes were considered, and subsequently tier 3.

Clinical Characteristics

We assessed markers of underlying health and healthcare-seeking characteristics before the index fill. We assessed fills for any systemic antibiotic (Anatomical Therapeutic Chemical Group J01) [20] within 14 days before the index fill to identify potential occurrences of initial treatment failure, and 90 days before the index fill as a proxy for recurrent infection history and general healthcare-seeking behaviors. During 180 days before the index fill, we assessed healthcare utilization and complex chronic conditions (CCCs) [21, 22]. CCCs were defined using diagnosis codes, ICD-9-CM procedure codes, and ICD-10 Procedure Coding System codes. To analyze healthcare utilization and CCCs, we required continuous prescription drug coverage for 90 days and continuous insurance coverage for 180 days before the index fill, respectively.

Analysis of Time Trends

To describe trends in the rate of antibiotic fills, we calculated monthly rates standardized by age, sex, and region (Supplementary Table 1), using 2013 as the standard year given demographic similarities that year between MarketScan and the US pediatric population with employer-based insurance [23]. Sex-stratified trend analyses were standardized by age and region. We excluded beneficiaries with unknown region from trend analyses (1.1% of cohort).

We originally aimed to assess the impact of the July 2008 FDA Boxed warning (tendon safety) on fluoroquinolone fills as this was the first fluoroquinolone warning. We implemented interrupted time-series analysis to compare slopes (change in rate over time) before and after the warning. We fit a segmented linear regression model for the standardized monthly rates, allowing for a change in intercept and slope at July 2008. We allowed a change in the intercept to increase model flexibility but explicitly did not interpret the intercept change. After visually inspecting the fitted trends, we concluded that allowing for a single change in July 2008 did not adequately reflect the observed changes in prescribing after the 2008 warning. Therefore, we allowed for a change in intercept and slope at 2 additional time points: August 2013 FDA warning (peripheral neuropathy) and May 2016 FDA warning (risk-benefit tradeoff). We implemented an autoregressive error model with errors assumed to follow a second-order process (chosen after visual inspection of partial autocorrelation function plots). From this model, we estimated the trend (annual change in the fill rate) with 95% confidence intervals (CI), before and after each FDA warning. We calculated an annual percentage change by dividing the annual change in the rate by the fitted rate immediately prior to the warning. To estimate what the fill rate would have been in the absence of any FDA warnings (counterfactual), we extrapolated the estimated slope from the period before the 2008 warning (January 2000-June 2008) into the period after the warning. We implemented interrupted time-series analyses for fluoroquinolone fills overall and stratified by sex. For comparison, we also analyzed nitrofurantoin as the fill rate was similar to the fluoroquinolone fill rate. None of the described FDA warnings pertained to nitrofurantoin. Inclusion of indicators for season in the autoregressive model did not impact estimates (data not shown) and thus were not included.

Analyses were performed using SAS, v9.3 (SAS Institute Inc., Cary, NC). This study was reviewed and exempted by the University of North Carolina at Chapel Hill institutional review board (#19-2483).

RESULTS

The cohort included 34.6 million unique beneficiaries less than 18 years old who had nearly 80 million person-years of prescription drug coverage. Table 1 summarizes cohort characteristics and fill rates by demographics. There were 441 062 outpatient fluoroquinolone fills (rate 5.5 fills per 1000 person-years) and 20 707 892 outpatient fills for the other (non-fluoroquinolone) broad-spectrum antibiotics (rate 259 per 1000 person-years). Ciprofloxacin was the most common fluoroquinolone, accounting for 69% of all fluoroquinolone fills (Supplementary Table 2). Levofloxacin was the next most common (25%). Amoxicillin-clavulanate and cefdinir were the most common other broad-spectrum antibiotics (42% and 36%).

Table 1.

Description of Cohort and Outpatient Fills of Fluoroquinolone and Other Broad-Spectrum Antibiotics Among Patients Aged 0–17 years, 2000–2018

			Outpatient Prescription Fills
	Cohort		Fluoroquinolones			Other Broad-Spectrum Antibioticsa
	Covered Yearsb	%	No.	%	Ratec (SE)d	no.	%	Ratec (SE)d
Overall	79 977 858		441 062		5.5 (0.008)	20 707 892		259 (0.06)
Regione
Northeast	11 672 739	14.6	58 853	13.3	5.0 (0.02)	2 599 097	12.6	223 (0.14)
Midwest	20 758 317	26.0	108 000	24.5	5.2 (0.02)	4 782 335	23.1	230 (0.11)
South	33 688 715	42.1	209 800	47.6	6.2 (0.01)	10 907 007	52.7	324 (0.10)
West	12 986 393	16.2	59 393	13.5	4.6 (0.02)	2 188 682	10.6	169 (0.11)
Other/unknown	871 694	1.1	5016	1.1	5.8 (0.08)	230 771	1.1	265 (0.55)
Sex
Male	40 859 595	51.1	158 079	35.8	3.9 (0.01)	10 094 867	48.7	247 (0.08)
Female	39 118 262	48.9	282 983	64.2	7.2 (0.01)	10 613 025	51.3	271 (0.08)
Age
Median (IQR)				16	(14, 17)		5	(2, 11)
<1	3 491 892	4.4	2010	0.5	0.6 (0.01)	1 729 159	8.4	495 (0.38)
1-5	19 377 818	24.2	18 323	4.2	0.9 (0.01)	8 911 571	43.0	460 (0.15)
6-11	26 859 564	33.6	40 456	9.2	1.5 (0.01)	5 238 345	25.3	195 (0.09)
12-17	30 248 584	37.8	380 273	86.2	12.6 (0.02)	4 828 817	23.3	160 (0.07)

Abbreviations: IQR, interquartile range; SE, standard error.

^aOther broad-spectrum antibiotics: amoxicillin-clavulanate, cefdinir, cefixime, nitrofurantoin, and sulfamethoxazole-trimethoprim.

^bPerson-years with prescription drug coverage.

^cRate of prescription fills per 1000 person-years of prescription drug coverage.

^dStandard errors assuming a Poisson distribution, as events^1/2/person-time.

^eNortheast: Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont; Midwest: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin; South: Alabama, Arkansas, Delaware, District of Columbia, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia; West: Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming.

The rate of fluoroquinolone fills among females (7.2 per 1000 person-years) was nearly twice the rate among males (3.9) (Table 1). For the other broad-spectrum antibiotics, rates were similar between females and males (271 vs 247 per 1000 person-years). Fills were more common in females across all the fluoroquinolone antibiotics (Supplementary Table 2). Among the other broad-spectrum antibiotics, nitrofurantoin (91%), SMX-TMP (61%), and cefixime (57%) were more commonly filled by females compared with males.

Figure 1 depicts fill rates by age and sex. The rate of fluoroquinolone fills was lowest (0.89 per 1000 person-years) among children less than 6 years old, whereas the fill rate for other broad-spectrum antibiotics was highest in this age group (465 per 1000 person-years) (Table 1). Between children aged 10 and 17 years, the fluoroquinolone fill rate increased (Figure 1) and was 12.6 per 1000 person-years for children aged 12–17 years (Table 1). The fluoroquinolone fill rate continued increasing into adulthood (aged 18-24, rate 60 per 1000 person-years; aged 25-40, 95 per 1000 person-years; data not shown). The median age did not vary by specific fluoroquinolone (Supplementary Table 2). Conversely, there was variation in the median age by specific other broad-spectrum antibiotics; only nitrofurantoin had a median age greater than 10 (median 13, interquartile range 6-16).

Figure 1.

Sex and age-stratified rates of prescription fills for fluoroquinolone and other broad-spectrum antibiotics per 1000 person-years, 2000-2018. Rates are displayed separately for fluoroquinolone antibiotics (solid lines) and other broad-spectrum antibiotics (dashed lines; amoxicillin-clavulanate, cefdinir, cefixime, nitrofurantoin, and sulfamethoxazole-trimethoprim) and for females (black) and males (gray).

Indication for Antibiotic Fills

As presented in Table 2, a smaller proportion of fluoroquinolone fills had a diagnosis code for which antibiotics may be indicated (tier 1 or tier 2 diagnosis) compared with the other broad-spectrum antibiotics (51% vs 65%). Over 30% of fluoroquinolone fills did not have a diagnosis code related to the antibiotic fill compared with 18% of other broad-spectrum fills. Among antibiotic fills with a diagnosis code present, antibiotics were not indicated (tier 3) for 25% of fluoroquinolone fills and 21% of other broad-spectrum fills.

Table 2.

Prescription Fills in Each Diagnostic Category by Antibiotic Group Among Patients Aged 0–17 Years, 2000–2018

	Fluoroquinolones		Other Broad-Spectrum Antibioticsa
	(N = 441 062)		(N = 20 707 892)
	No.	%	No.	%
Diagnosis likely related to antibiotic identifiedb
Tier 1—antibiotics almost always indicated	119 224	27.0	1 803 220	8.7
Tier 2—antibiotics may be indicated	105 735	24.0	11 574 282	55.9
Tier 3—antibiotics not indicated	76 534	17.4	3 508 027	16.9
Diagnosis not identified
No recent codesc	107 291	24.3	3 104 874	15.0
Only codes unrelated to antibioticd	32 178	7.3	717 489	3.5
Specific diagnoses related to antibiotice
Tier 1—antibiotics almost always indicated
Pneumonia	12 066	2.7	410 733	2.0
Urinary tract infection	96 020	21.8	1 024 301	4.9
Miscellaneous bacterial infections	16 747	3.8	427 333	2.1
Tier 2—antibiotics may be indicated
Acne	2870	0.7	135 896	0.7
Gastrointestinal infections	29 137	6.6	321 746	1.6
Pharyngitis	14 644	3.3	2 518 209	12.2
Sinusitis	35 453	8.0	3 091 067	14.9
Skin, cutaneous, and mucosal infections	23 325	5.3	1 482 000	7.2
Suppurative otitis media	9593	2.2	5 264 997	25.4
Acute exacerbation of chronic bronchitis	6	<0.1	161	<0.1
Tier 3—antibiotics not indicated
Asthma and allergy	7895	1.8	382 144	1.8
Bronchitis and bronchiolitis	9992	2.3	398 835	1.9
Fever	2559	0.6	207 286	1.0
Influenza	689	0.2	39 979	0.2
Nonsuppurative otitis media	1757	0.4	689 748	3.3
Other gastrointestinal conditions	10 529	2.4	124 399	0.6
Other skin, cutaneous, and mucosal conditions	15 463	3.5	857 349	4.1
Other genitourinary conditions	22 851	5.2	299 398	1.4
Viral pneumonia	78	<0.1	4910	<0.1
Viral upper respiratory infection	11 200	2.5	937 822	4.5
Other respiratory conditions	6831	1.5	382 008	1.8
Miscellaneous other infections	1040	0.2	38 225	0.2

^aOther broad-spectrum antibiotics: amoxicillin-clavulanate, cefdinir, cefixime, nitrofurantoin, and sulfamethoxazole-trimethoprim.

^bCodes obtained from any encounter on the day of antibiotic fill, in 3 days before antibiotic fill, or, if the fill occurred during an inpatient admission, within 2 days after the fill. Codes classified using diagnostic categories developed by Fleming-Dutra et al [19]. Beneficiaries only counted in priority category (priority order: tier 1, tier 2, and tier 3).

^cNo recent codes: beneficiary did not have any healthcare encounters on the day of the antibiotic fill or in the 3 days before.

^dOther diagnosis unrelated to antibiotic: beneficiary had a healthcare encounter but the diagnosis codes provided were unrelated to receiving an antibiotic (eg, a diagnosis of depression) or were nonspecific (E-codes, codes for symptoms, and signs or ill-defined health conditions). This category includes all other diagnosis codes that were not included in tier 3.

eA single fill may be included in more than one diagnosis within a single tier.

The most common indications for fluoroquinolones were UTI (22%), sinusitis (8%), and gastrointestinal infections (7%). For the other broad-spectrum antibiotics, suppurative otitis media (25%), sinusitis (15%), and pharyngitis (12%) were the most common. There was no proximal healthcare encounter for 24% of fluoroquinolone fills and 15% of other broad-spectrum fills.

Clinical Characteristics

Table 3 presents the clinical characteristics. Twenty-one percent of fluoroquinolone fills were preceded by another antibiotic in the prior 14 days compared with 14% for other broad-spectrum fills; however, in the prior 90 days, there was no difference (both 53%). Fluoroquinolone fills were more often preceded by a recent inpatient admission than other broad-spectrum fills (prior 180 days: 11% vs 3%). A CCC was more common among fluoroquinolone fills (18% vs 7%). Among fluoroquinolone fills, 44% had an antibiotic or an inpatient admission in the prior 30 days, or a CCC; among other broad-spectrum antibiotic fills, 33% did.

Table 3.

Clinical Characteristics of Patients aged 0–17 Years Receiving Fluoroquinolones and Other Broad-Spectrum Antibiotics, 2000 to 2018

	Outpatient Prescription Fills
	Fluoroquinolones		Other Broad-Spectrum Antibioticsa
	No.	%	No.	%
Among children with ≥90 days continuous prescription drug coverage prior to index fill	402 277		18 499 890
Recent outpatient antibioticb
Prior 14 days	85 773	21.3	2 617 706	14.2
Prior 30 days	134 641	33.5	5 278 239	28.6
Prior 90 days	211 648	52.6	9 844 642	53.4
Among children with ≥ 180 days continuous insurance coverage prior to index fill	369 271		16 630 039
Healthcare utilization
No. of healthcare encounters (median, IQR)c
Prior 30 days	2	(1, 3)	1	(1, 2)
Prior 180 days	5	(2, 10)	5	(3, 8)
Inpatient admissions
Any in prior 30 days	23 776	6.4	186 016	1.1
Any in prior 180 days	40 992	11.1	489 085	2.9
Comorbid conditions
Any CCCd	65 768	17.8	1 095 426	6.6
Specific CCCs
Neurological/neuromuscular	12 062	3.3	205 157	1.2
Cardiovascular	10 759	2.9	194 122	1.2
Respiratory	16 101	4.4	82 533	0.5
Renal/urologic	10 000	2.7	165 648	1.0
Gastrointestinal	17 088	4.6	106 691	0.6
Hematologic/immunologic	8879	2.4	153 005	0.9
Metabolic	8662	2.3	137 015	0.8
Other congenital/genetic defect	11 298	3.1	256 896	1.5
Malignancy	11 773	3.2	212 225	1.3
Premature/neonatal	1129	0.3	25 872	0.2
Transplantation	3295	0.9	31 406	0.2
Device	14 163	3.8	129 075	0.8

Abbreviations: IQR, interquartile range; CCC, complex chronic condition.

^aOther antibiotics: amoxicillin-clavulanate, cefdinir, cefixime, nitrofurantoin, and sulfamethoxazole-trimethoprim.

^bAny outpatient pharmacy prescription fill for systemic antibiotic (ATC: J01).

^cCount of unique service/discharge dates.

^dFeudtner et al [21] and Feudtner et al [22].

Time Trends

Over the study period, the rate of fluoroquinolone fills was much lower than the rates of amoxicillin-clavulanate and SMX-TMP (Figure 2A). The cefdinir fill rate, although similar to fluoroquinolones at the start of the study period, had a steep increase before plateauing. The rate of nitrofurantoin fills was lower than the rate of fluoroquinolone fills until they crossed in June 2016 (Figure 2B). The cefixime fill rate was low but highly variable.

Figure 2.

Standardized rates of prescription fills for fluoroquinolone and other broad-spectrum antibiotics per 1000 person-years, 2000–2018, aged 0-17 years. (A) fluoroquinolones, amoxicillin-clavulanate, cefdinir, and sulfamethoxazole-trimethoprim. (B) fluoroquinolones, cefixime, and nitrofurantoin. Trend lines are smoothed by locally estimated scatterplot smoothing (LOESS).

Figure 3 depicts standardized fluoroquinolone (panel A) and nitrofurantoin (panel B) fill rates with fitted trends before and after each FDA warning. Both antibiotics had a similar increasing trend before the 2008 FDA warning (tendon safety) (Figure 3, Supplementary Table 3). After the 2008 warning, the fluoroquinolone rate decreased (−3.2% per year) as did the nitrofurantoin rate (−3.8% per year). The trends in fluoroquinolone and nitrofurantoin fills remained parallel until the 2016 warning (risk-benefit tradeoff), after which the fluoroquinolone rate decreased (−13.2% per year); however, the nitrofurantoin rate increased (+3.0% per year). In the (counterfactual) absence of the FDA warnings, if the trend prior to the FDA warnings had continued, the predicted fluoroquinolone fill rate by December 2018 (end of the study period) would have been 7.5 per 1000 person-years (95% CI: 5.2-9.7); the rate estimated from observed data was 2.8 (95% CI: 1.7-3.9). For nitrofurantoin, the December 2018 counterfactual prediction was 5.9 (95% CI: 5.1-6.7) compared with the observed estimate of 4.9 (95% CI: 4.3-4.5). After 2008, the fill rate decreased for both ciprofloxacin and levofloxacin; however, the decline was greater for ciprofloxacin (Supplementary Figure 1).

Figure 3.

Standardized rate of fills (2000-2018) allowing for changes at each FDA Boxed warning (July 2008 [13], August 2013 [14], and May 2016 [15]), aged 0-17 years. (A) fluoroquinolones and (B) nitrofurantoin. Rates standardized by sex, age, and region. Shaded areas are 95% confidence bands of trend. To estimate the December 2018 counterfactual rate (the rate that would have been observed in the absence of any FDA warnings), we extrapolated the estimated slope from the time period before the 2008 warning (January 2000-June 2008) into the period after the warning. The December 2018 observed rate was estimate from the segment regression fit on the observed data. Abbreviations: FDA, US Food and Drug Administration.

The trend in fluoroquinolone fills varied by sex (Figure 4, Supplementary Table 3). Before the 2008 warning (tendon safety), the rate among females increased, while the rate among males was flat. After the 2016 warning (risk-benefit tradeoff), the decline in fills was more pronounced among females (−13.7% per year) than among males (−11.8% per year). For females, the December 2018 counterfactual prediction was 10.4 fills per 1000 person-years (95% CI: 8.1-12.7), and the rate estimated from observed data was 3.6 (95% CI: 2.3-4.9). For males, these estimates were 4.7 (95% CI: 2.3-7.1) and 2.0 (95% CI: 0.9-3.2), respectively.

Figure 4.

Standardized rate of fluoroquinolone fills (2000-2018) allowing for changes at each US Food and Drug Administration Boxed warning (July 2008 [13], August 2013 [14], and May 2016 [15]), aged 0-17 years, stratified by sex. Rates standardized by age and region.

DISCUSSION

We described outpatient fluoroquinolone use in a large population of privately insured children over 2 decades. The rate of fluoroquinolone use was substantially lower than other broad-spectrum antibiotics used for similar indications. The majority of fluoroquinolone use was for children aged above 11 years, while other broad-spectrum antibiotics were predominantly prescribed to younger children.

Given the rarity of anthrax exposure and plague, complicated UTI and pyelonephritis are the only pertinent FDA-approved pediatric indication for fluoroquinolones [7]. UTI was the most common indication; however, this accounted for only one-fifth of fluoroquinolone fills. Therefore, the majority of fluoroquinolone use was off-label. Other common indications that are off-label but discussed in AAP statements or Infectious Diseases Society of America pediatric treatment guidelines included respiratory tract infections; gastrointestinal infections; and skin, cutaneous, and mucosal infections [6, 10, 11, 24–26]. Just half of fluoroquinolone fills were associated with diagnoses for which antibiotics may be indicated, while 25% of fills had diagnoses for which an antibiotic is not indicated. These numbers indicate a continued need for antibiotic stewardship and improved diagnostic coding. Approximately, 15% of fills had no recent healthcare encounters, and this differed by antibiotic type. This finding aligns with previous research [27]. These fills may represent nonvisit-based prescribing, visits not captured in claims data, such as retail clinic, or visits paid out of pocket, or refills [27].

The declining fluoroquinolone fill rate was temporally associated with FDA warnings in 2008 (tendon safety) and 2016 (risk-benefit tradeoff). Although research in adult populations has observed an association of fluoroquinolone use with the 2008 FDA warning [28], this has not previously been investigated in children [4, 5]. Fluoroquinolones have been plagued with pediatric-specific safety concerns since the 1970s, and this perceived potential harm has greatly limited their use [4, 6]. Given these preexisting concerns, the decline associated with the FDA warnings, which did not specifically note pediatric risks, is compelling and may reflect that limited pediatric safety data have left clinicians struggling to make informed prescribing decisions for children. The steepest decline was observed after the 2016 warning, which discussed the risk-benefit trade-off when oral alternatives are available. The other warnings discussed specific adverse effects that are rare in children. The impact of FDA warnings on this rate shed light on what levers might change prescribing, which could inform future public health efforts. Regardless of the rationale, lower fluoroquinolone use is important to reduce adverse drug effects and antibiotic resistance pressure, particularly for the emergence of drug-resistant pneumococcus. However, oral fluoroquinolones can be an important alternative when only intravenous treatment options are available. As perceived potential harms of fluoroquinolones increase, it is possible that a provider may select an intravenous option over an oral fluoroquinolone. Given that musculoskeletal toxicity has not been observed in recent pediatric trials with long-term follow-up [29, 30] and FDA warnings do not cite pediatric-specific risks, it is unlikely that the risk of adverse effects of oral fluoroquinolones outweighs the risk of adverse effects of intravenous treatment. Future studies should evaluate this question.

Other factors may have contributed to the observed decline in fluoroquinolone rates. Antimicrobial stewardship efforts have increased over time, particularly in the outpatient setting, and have had notable success in reducing prescribing in children and spreading awareness about the appropriate use of antibiotics [31]. Additionally, declines may be partially due to the promotion of “watchful waiting” for respiratory infections [32, 33] and initiatives against antibiotics for asymptomatic bacteriuria [34].

The greater decline in ciprofloxacin compared with levofloxacin and greater reduction among female beneficiaries suggest that UTI prescribing likely contributed to the observed fluoroquinolone decline. However, it does not appear that either nitrofurantoin or SMX-TMP, likely alternatives [35], completely replaced fluoroquinolone use. Given the wide range of observed indications, it is unlikely that fluoroquinolones were replaced by a single substitute. Additionally, fluoroquinolones may have been withheld in lieu of no treatment.

This study has several strengths. We provide new population-based estimates of prescription fill rates for fluoroquinolone and other broad-spectrum antibiotics in children with commercial prescription drug coverage. The data used include accurate individual-level data on insurance enrollment, diagnosis and procedure codes associated with insurance reimbursement, and medication-specific outpatient prescription claims. Our interrupted time-series analysis examined a 19-year period of changes in fluoroquinolone guidelines and increasing stewardship.

Our work has limitations. The use of antibiotics in this population of children with private insurance may not be generalizable to non-privately insured children, since there are differences in antibiotic use and healthcare utilization across racial and socioeconomic groups. Antibiotic fills not billed to insurance were not captured. A fill does not indicate the actual use of the medication. For example, a “wait-and-see” prescription for otitis media may result in an antibiotic fill that is ultimately not ingested [32]. Additionally, indication is subject to misclassification since fills are not linked to specific diagnosis codes in claims data; however, our use of hierarchical diagnosis code gathering may have mitigated this potential measurement error. Lastly, our time-series analysis lacks formal causal interpretation because we were unable to compare trend changes in our population to a “control” population because the FDA warnings are released nationwide and also gather attention globally; however, we compare trends across several antibiotics to provide context.

CONCLUSIONS

Fluoroquinolone use in children is low overall compared with other common broad-spectrum antibiotics. Declining trends over time were associated with FDA warnings discussing potential adverse events, even though these warnings were not pediatric specific. Future work should assess these adverse events in children.

Notes

Acknowledgment. None.

Financial support. The database infrastructure used for this work was funded by the Pharmacoepidemiology Gillings Innovation Lab (PEGIL) for the Population-Based Evaluation of Drug Benefits and Harms in Older US Adults (grant number GIL200811.0010); the Center for Pharmacoepidemiology, Department of Epidemiology, University of North Center for Pharmacoepidemiology, Department of Epidemiology, University of North Carolina Gillings School of Global Public Health; the CER Strategic Initiative of UNC’s Clinical and Translational Science Award from the National Center for Advancing Translational Sciences (UL1TR002489); the Cecil G. Sheps Center for Health Services Research, UNC; and the UNC School of Medicine. This work was also supported by the Antimicrobial Stewardship Fellowship Award from the Pediatric Infectious Diseases Society.

Potential conflicts of interest. The authors have no conflicts of interest.

All authors have submitted the ICMJE Form for Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.