Epidemiology and Management of Orbital Cellulitis in Children

Abstract

Background

The epidemiology of orbital cellulitis likely has evolved due to the emergence of methicillin-resistant Staphylococcus aureus (MRSA) and the adoption of pneumococcal conjugate vaccination. In the absence of published guidelines, management is highly variable. We characterized epidemiology and management over an 11-year period.

Methods

A retrospective cohort study of children 0 to 21 years of age with orbital cellulitis +/− subperiosteal orbital abscess hospitalized at a large quaternary children’s hospital from January 2008 to June 2018. We reviewed charts for demographic characteristics, clinical features, management, and outcomes. Using multivariable logistic regression, we evaluated predictors of surgical intervention and assessed whether corticosteroid use or antibiotic duration was related to clinical outcomes.

Results

Among 220 patients, methicillin-susceptible S. aureus was the most common organism (26.3%), with MRSA found in only 5.0%. Rates of vancomycin use fluctuated annually from 40.9% to 84.6%. Surgery was performed in 39.5% of the patients. Corticosteroids, used in 70 patients (32.1%), were unrelated to treatment failure (n = 9), defined as persistent signs and symptoms or initial clinical improvement followed by worsening (P = .137). The median antibiotic duration was 17 days (interquartile range 14-26). After controlling for age, gender, proptosis, eye pain with movement, eyelid swelling, neutrophil count, and corticosteroid use, treatment failure was not significantly associated with receipt of ≥ 3 weeks of antibiotic therapy (8/84, 9.5%) compared with > 2 but < 3 weeks (0/51, 0.0%) or ≤ 2 weeks (1/85, 1.2%) (adjusted odds ratio = 5.83 for ≥ 3 vs ≤2 weeks; 95% confidence interval: 0.58, 59.0).

Conclusions

Although MRSA was rare, empiric vancomycin use was high. Treatment failure was uncommon in patients who received ≤ 2 weeks of therapy, suggesting that shorter durations are adequate in some patients.

In a large 11-year cohort of children with orbital cellulitis, outcomes were excellent despite surgical intervention in only 39.5% and a median antibiotic duration of 17 days. Isolation of methicillin-resistant Staphylococcus aureuswas low, but vancomycin use was consistently high.

Orbital cellulitis is a serious invasive bacterial infection of the postseptal tissues of the eye [1]. It is often preceded by acute or chronic sinusitis, in contrast with periorbital (preseptal) cellulitis, which typically occurs as a result of trauma or bacteremia [2]. Chandler’s 5-stage classification is commonly used to describe complications of acute sinusitis based on localized anatomic findings, ranging from least to most severe: (I) preseptal cellulitis, (II) orbital cellulitis, (III) subperiosteal orbital abscess (SPOA), (IV) orbital abscess, and (V) cavernous sinus thrombosis [3]. Left untreated, orbital cellulitis can lead to significant complications, including vision loss, meningitis, and intracranial abscesses [4]. Though this condition has become increasingly rare in the vaccine era, early recognition and prompt treatment remain an essential component of management. Historically, typical management has included frequent surgical intervention and antibiotic courses of 3 weeks’ duration [1, 5, 6].

Because of the pathophysiology of orbital cellulitis, the microbiology of this disease often reflects organisms associated with upper respiratory tract or sinus infections, such as Haemophilus influenzae type B, Streptococcus pneumoniae, other streptococcal species, Staphylococcus aureus, and anaerobes [1, 4, 7, 8]. With the emergence of methicillin-resistant S. aureus (MRSA) and the adoption of pneumococcal conjugate vaccination, we hypothesized that the causative organisms of orbital cellulitis likely had evolved over the past decade.

Despite substantial literature on orbital cellulitis, to date, there remain no standardized clinical practice guidelines for this condition. The Infectious Disease Society of America guidelines for skin and soft-tissue infections provide a comprehensive algorithm for pathogen-specific cellulitis and/or wound infections, but orbital cellulitis is not specifically addressed [9]. In the absence of published guidelines, there has been wide heterogeneity in the treatment of this condition both within and across hospitals [10]. Aspects of management that are uncertain—and thus tend to vary among clinicians—include criteria for surgical intervention, the role of systemic corticosteroids for reducing inflammation, optimal empiric antibiotic selection, and the duration of therapy that is sufficient to eradicate infection while avoiding harms of unnecessarily prolonged antibiotic exposure.

Using the largest US cohort of pediatric patients reported to date, we investigated the epidemiology and management of orbital cellulitis over an 11-year period. We determined rates and predictors of surgical intervention and characterized corticosteroid use, causative organisms, and antibiotic selection and duration. To help inform practice guideline development, we also assessed whether corticosteroid use or antibiotic duration was related to clinical outcomes.

METHODS

Study Population and Setting

The study cohort consisted of patients ≤ 21 years of age with a primary diagnosis of orbital cellulitis with or without SPOA admitted to Boston Children’s Hospital (BCH), a quaternary care, 404-bed freestanding children’s hospital in Boston, Massachusetts, from January 2008 to June 2018. Patients with potential orbital cellulitis were identified using the International Classification of Diseases, Ninth Revision (ICD-9) and International Classification of Diseases and Related Health Problems, Tenth Revision, Clinical Modification (ICD-10-CM) codes H05.0X and 376.0X, respectively, for “acute inflammation of the orbit,” which encompass orbital (also referred to as postseptal) cellulitis, periorbital cellulitis, orbital osteomyelitis, and SPOA. The diagnosis of orbital cellulitis was then verified via a review in the electronic health record of Ophthalmology consultation notes and results of orbital/facial computed tomography (CT) (performed in 98.2% of all cases). Findings of postseptal edema, inflammation of extraocular muscles, diffuse fat stranding/infiltration, anterior displacement of the globe, SPOA, or orbital abscess on CT were used to confirm a diagnosis of orbital cellulitis [11–14]. Patients with underlying immunodeficiency, malignancy, recent trauma, or recent facial/orbital surgery (defined as within 4 weeks of presentation) were excluded. The BCH Institutional Review Board approved the study with a waiver of informed consent.

Predictors, Exposures, and Outcome Measures

We determined demographic characteristics (age, gender, race, and ethnicity), presenting clinical features, underlying medical comorbidities, prior use of antibiotics, laboratory results, operative culture results, and imaging findings via electronic chart review. We also collected data on treatment exposures, including surgical intervention, systemic corticosteroid use, choice of antimicrobial therapy and route of administration both in the hospital and upon discharge, and duration of antimicrobial therapy. We assessed the clinical outcome of treatment failure, defined as persistent signs and symptoms of orbital cellulitis or clinical worsening after an initial period of improvement, for a follow-up period of 30 days from the last day of antibiotic therapy. Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at BCH. REDCap is a secure, web-based application designed to support data capture for research studies [15].

Statistical Analysis

We calculated rates of surgical intervention and systemic corticosteroid use. We also determined frequencies of bacterial pathogens and rates of rates of antibiotic use by drug. Categorical variables were summarized using proportions and analyzed using the Chi-square (χ²) test or Fisher exact test. Continuous variables were summarized using mean with standard deviation (SD) or median with interquartile range (IQR) as appropriate according to normality. A 2-tailed P-value < .05 was considered statistically significant.

We used multivariable logistic regression to identify predictors of surgical intervention and to investigate the relationship between antibiotic duration and treatment failure. Covariates selected for potential inclusion in the regression models were pre-determined based on previous studies and clinical practice. Variables related to the outcome of interest with a P-value of < .10 on univariable analysis were assessed in multivariable analysis. The relationship between age and treatment failure was evaluated with a Lowess smoother to examine for a possible nonlinear relationship. Based on this smoother, we then fit a linear spline with a single knot at age 5 to be used for the multivariable model. All statistical analyses were performed using Stata 16.0 (StataCorp, College Station, TX).

RESULTS

Patient Characteristics and Clinical Presentation

Of 486 potential patients identified using ICD-9 and ICD-10-CM codes, we confirmed via chart review that 220 ultimately were diagnosed with orbital cellulitis and met study inclusion criteria. Please see Supplementary eFigure 1 for details of excluded cases. The mean age (SD) was 8.3 (4.4) years (range: 2 months to 21 years) (Table 1). Sixty-one percent of patients were male; 14.1% were Black and 58.6% were White, with 9.1% identifying as Hispanic/Latino. Eyelid swelling was the most common presenting clinical feature (95.9%), followed by eyelid erythema (77.7%), fever (70.9%), and eye pain with extraocular movement (55.0%). An SPOA was identified on CT in 116 (52.7%) patients, with an average size of 20.9 mm at the longest dimension. Because measurements in 3 dimensions were not consistently included in CT reports, the calculation of SPOA volume was not feasible for many patients.

Table 1.

Patient Characteristics, Clinical Presentation, and Management

	Total N = 220
Age (years)a	8.3 ± 4.4
Genderb
Male	135 (61.3%)
Racial/ethnic backgroundb
White	129 (58.6%)
Black	31 (14.1%)
Hispanic/Latino	20 (9.1%)
Other	40 (18.2%)
Presenting clinical symptomsb
Fever	156 (70.9%)
Proptosis	74 (33.6%)
Eye pain with extraocular movements	121 (55.0%)
Diplopia	30 (13.6%)
Eyelid erythema	171 (77.7%)
Eyelid swelling	211 (95.9%)
Nasal congestion	96 (43.6%)
Diminished visual acuity	33 (15.0%)
Impaired color discrimination	2 (0.91%)
Hospitalizationa
Length of stay (days)	5.4 ± 4.1
Laboratory valuesa
CRP on admission, mg/dL	7.7 ± 8.0
ANC on admission, K/μL	10.1 ± 5.2
Radiographic findingsa
Frontal sinusitis on CT scan	15 (6.8%)
SPOA on CT scan	116 (52.7%)
Treatment
Use of antibiotics prior to admissionb	87 (39.7%)
IV antibiotic durationc	4 [IQR:3-6.75]
Total IV + PO antibiotic durationc	17 [IQR:14-16]
Systemic corticosteroidsb	70 (32.1%)
Any surgical interventionb	127 (57.7%)

Abbreviations: ANC, absolute neutrophil count; CRP, C-reactive protein; CT, computed tomography; IV, intravenous; PO, per os (by mouth); MRI, magnetic resonance imaging; SPOA, subperiosteal abscess.

Results expressed as

Mean ± SD,

N (%),

Median [interquartile range].

Surgical Management

Functional endoscopic sinus surgery was performed in 87 (39.5%) patients overall, of whom 64 also underwent drainage of an SPOA. The annual rate of surgical intervention ranged from 27.8% to 62.5% over the study period (Figure 1). The operative intervention occurred on hospital day (HD) 0 in 12.3% (27/220) of patients; these patients had a median age of 9.0 years [IQR: 5-13] and, for those who had an SPOA, median abscess size of 21 mm at the longest dimension [IQR: 5-31.5]. Patients who underwent surgery on HD 1 comprised 13.6% (30/220) of patients and had a median age of 10.5 years and median abscess size of 26 mm [IQR: 20.5-30]. For the remaining 15.9% (35/220) of patients who underwent surgical intervention, the procedure(s) occurred on HD 2 or later, and their median age was 9.0 years [IQR: 4-13] and median abscess size was 19 mm [IQR: 7-22]. Adjusting for gender, eye pain with movement, ophthalmoplegia, eyelid swelling, and absolute neutrophil count, patients were more likely to undergo surgery, if they were > 5 years of age (adjusted odds ratio [aOR] = 1.1; 95% confidence interval [CI]: 1.04, 1.28) or had proptosis (aOR = 2.2; 95% CI: 1.06, 4.62), diplopia (aOR = 3.6; 95% CI: 1.28, 10.04), or an SPOA ≥ 20 mm (aOR = 14.5 compared with no SPOA; 95%CI: 5.19, 40.49) (Table 2).

Figure 1.

Rates of surgical intervention.

Table 2.

Predictors of Surgical Intervention on Multivariable Analysis

	Adjusted Odds Ratio (aOR) (95% Confidence Interval)	P-value
Agea
<5 years	0.7 (0.54, 1.04)	.086
>5 years	1.1 (1.04, 1.28)	.007
Female gender (reference = male)	0.8 (0.40, 1.58)	.513
Clinical features
Proptosis	2.2 (1.06, 4.62)	.035
Eye pain with extraocular movements	0.7 (0.33, 1.44)	.323
Diplopia	3.6 (1.28, 10.04)	.015
Ophthalmoplegia	1.4 (0.59, 3.17)	.464
Eyelid swelling	1.6 (0.24, 10.95)	.621
Absolute neutrophil count (K/µL)	0.6 (0.24, 1.67)	.357
SPOA
None	Reference	—
≤10 mm	1.6 (0.64, 4.08)	.309
>10 mm, <20 mm	1.9 (0.72, 5.03)	.197
≥20 mm	14.5 (5.19, 40.49)	<.0001

Abbreviation: SPOA, subperiosteal orbital abscess.

Age was modeled using a linear spline with a single knot at age 5. The aORs for age express the change in odds for each 1-year increase in age.

Corticosteroids

Overall, 70 patients (32.1%) received adjunctive systemic corticosteroids; their average age was 9.6 years (SD 4.6) (Supplementary eTable 1). The annual rate of corticosteroid use varied from 20.0% to 61.5% (Supplementary eFigure 2) with no apparent trend. The use of this intervention was not related to treatment failure on multivariable analysis (P = .137, Table 3).

Table 3.

Relationship Between Antibiotic Duration and Treatment Failure on Multivariable Analysis

	Adjusted Odds Ratio (aOR) (95% Confidence Interval)	P-value
Agea
<5 years	3.45 (0.53, 22.3)	.194
>5 years	1.04 (0.81, 1.34)	.738
Female gender (reference = male)	4.82 (0.86, 27.1)	.074
Clinical features
Proptosis	0.07 (0.004, 1.2)	.067
Eye pain with extraocular movement	0.34 (0.058, 2.0)	.227
Eyelid swelling	0.18 (0.008, 3.8)	.266
Absolute neutrophilcount (K/µL)	1.00 (1.0, 1.0)	.119
Antibiotic duration
≤ 2 weeks	Reference	—
>2 weeks to < 3 weeksb	—	—
≥3 weeks	5.83 (0.58, 59.0)	.135
Systemic corticosteroids	3.96 (0.65, 24.3)	.137

Age was modeled using a linear spline with a single knot at age 5. The aORs for age express the change in odds for each 1-year increase in age.

Patients who received antibiotic treatment for > 2 weeks to < 3 weeks had no failures; this category was, therefore, automatically dropped from the model.

Microbiology and Antibiotic Management

Thirty-nine percent of patients had previous (outpatient) antibiotic use upon initial presentation with no documentation of the duration and/or regimen used, possibly affecting the sensitivity of operative cultures. Among the 99 patients who had cultures (see Supplementary eFigure 3 for annual rates of positive cultures during the study period), methicillin-susceptible S. aureus was the most common organism isolated (26.3%), followed by Streptococcus intermedius (21.2%), Cutibacterium (formerly Propionibacterium) acnes (15.1%), and Streptococcus pyogenes (13.1%). Without additional information on how the specimens were collected, it is unclear whether the isolates of C. acnes, an unusual cause of orbital cellulitis, represented true pathogens vs skin contaminants. Streptococcus pneumoniae was present in only 7.1% of cases, and MRSA was isolated in only 5.0% of cases (Supplementary eTable 2). More than one organism was isolated from cultures for 36 of the 99 (36.0%) patients.

Fifty-six (25.4%) patients received a single parenteral antibiotic during the course of their hospitalization; 80 (36.5%) received 2 parenteral antibiotics either sequentially or in combination, while 54 (24.5%) received 3. On HD 0, ampicillin/sulbactam was the most commonly prescribed regimen (175/220, 79.5%), followed by clindamycin (33/220, 15.0%) and ceftriaxone/vancomycin/metronidazole (28/220, 12.7%). At the time of discharge, the most common regimens were amoxicillin/clavulanate, clindamycin, and trimethoprim/sulfamethoxazole plus metronidazole. Vancomycin was used in 60.9% of cases, with annual rates ranging from 40.9% to 84.6% over the study period (Supplementary eFigure 4). A total of 42 patients (19.1%) were discharged on intravenous antibiotics administered via a percutaneously inserted central catheter, all of whom had intracranial involvement (eg, subdural abscess, epidural abscess, intraparenchymal abscess, or cavernous sinus thrombosis). The remaining patients were discharged on oral therapy. The median total duration of antibiotics (intravenous and oral) was 17 days (IQR:14-26) (Table 1).

By 30 days from the end of treatment, 211 (95.9%) patients had experienced clinical improvement. On univariable analysis, treatment failure was more common in patients who received ≥ 3 weeks of antibiotic therapy (9.5%) than in those who received > 2 but < 3 weeks (0%) or ≤ 2 weeks (1.2%) (P = .005) (Supplementary eTable 3). (Please see Supplementary eTable 4 for individual characteristics of patients who experienced treatment failure.) However, after adjusting for age, gender, proptosis, eye pain with movement, eyelid swelling, absolute neutrophil count, and corticosteroid use on multivariable analysis, patients who received ≥ 3 weeks of treatment did not differ significantly in their odds of treatment failure compared with those who were treated for ≤ 2 weeks (aOR 5.83; 95%CI: 0.58, 59.0) (Table 3).

DISCUSSION

In a large cohort of children with orbital cellulitis over an 11-year period, we found that nearly all (95.9%) experienced resolution of disease within 30 days of completing antibiotic therapy. This was true despite the use of medical management alone in about 60% of patients and of antibiotic durations < 3 weeks in about 60% of patients. In keeping with the marked reduction in invasive pneumococcal disease following the adoption of pneumococcal conjugate vaccination, S. pneumoniae was isolated from operative cultures in only 7.1% of cases, notably lower than the pre-pneumococcal-vaccine rate of 22.4% reported in a study of orbital cellulitis by Peña et al [16–18]. MRSA was isolated in only 5.0% of cases, yet vancomycin use remained high throughout the study period.

Before the 1980s, evidence of SPOA was an indication for immediate surgical intervention. Since then, several studies have endorsed a trial of antibiotics before surgical intervention. Current practices have shifted to using medical management for many patients, with the initiation of empiric intravenous antibiotics without surgery and close observation for 24 to 48 hours for patients with no evidence of impaired visual acuity or increased intraocular pressure [19–23]. The high rate of treatment success in our cohort offers further general support for this approach although our study suggests that certain factors predict ultimately requiring surgical intervention and thus could be helpful in identifying patients who warrant especially close clinical monitoring if medically managed. We found, for example, that children > 5 years of age were more likely to undergo surgery, as were those who presented with proptosis or diplopia or had a large SPOA (≥20 mm). These results are consistent with risk factors outlined by Chandler’s classification, which in conjunction with CT remains the cornerstone for assessing an indication for surgical intervention [24]. Younger age (<9 years) has been associated with greater success with conservative management [25, 26]. Severe proptosis, worsening visual acuity, elevated intraocular pressure, color indiscrimination, intracranial involvement, inability to perform a reliable serial ophthalmologic examination, and poor response to a trial of intravenous antibiotics for 24 to 48 hours are all common indications for surgery [3, 19, 27]. In addition, the size of SPOA on CT has emerged as a significant prognostic factor in determining which patients require surgical intervention [28]. SPOAs with diameter > 10 mm and more recently volumes ≥ 500 mm³ are thought typically to require surgical intervention, with the remainder going to surgery only in the event of clinical deterioration, lack of clinical improvement after 48 hours of antibiotic therapy, or radiographic findings indicative of worsening abscess [11, 29–32]. Given the retrospective design of this study, there was no uniformity in reporting the size of SPOA radiographically as CT measurements were often not reported in 3 dimensions, preventing the calculation of volume for several patients. Further studies are needed to assess volume calculations as another diagnostic predictor of those who may need surgical intervention.

The role of systemic corticosteroids in orbital cellulitis remains controversial. Treatment with intravenous corticosteroids could decrease inflammation and mucosal edema, potentially helping to facilitate sinus drainage [33, 34]. At the same time, corticosteroid use has been met with caution, particularly in the setting of an active infectious process. One study reported a reduction in required duration of intravenous antibiotics in patients who received corticosteroids as well as a reduction in mean length of stay [22, 34]. However, we found no association between corticosteroid therapy and the key outcome of treatment failure, and the use of this intervention varied widely from year to year in our study cohort. This is an aspect of orbital cellulitis management for which investigation in multicenter randomized clinical trials would be valuable.

We found that MRSA was rarely identified as a causative pathogen of orbital cellulitis. This is in keeping with surveillance by the Centers for Disease Control and Prevention, which demonstrated a decrease in the overall incidence of community-acquired MRSA from 2009 to 2016 with a slight increase in 2017 [35]. In contrast, a Boston-based study demonstrated that MRSA was the etiology of 21% of ophthalmological and otolaryngological infections over a 1-year period but did not specify the contribution of MRSA to orbital cellulitis cases [36]. Despite the low frequency of MRSA in our orbital cellulitis cases, empiric vancomycin was used at high rates throughout the study period, without an appreciable decline even in more recent years. These findings perhaps reflect ingrained practices lagging behind the evolving microbiology. In institutions with low rates of MRSA, ampicillin/sulbactam followed by amoxicillin/clavulanate is a reasonable first-line regimen to target MSSA, streptococcal species, and anaerobes. In the setting of a penicillin allergy, ceftriaxone is a good empiric alternative. For patients who cannot receive ampicillin/sulbactam but warrant anaerobic coverage, such as when intracranial extension has not yet been assessed or orbital cellulitis is associated with chronic sinusitis, adding metronidazole with ceftriaxone offers comparable coverage. For cases complicated by intracranial abscesses, a regimen with adequate central nervous system penetration and broader coverage, such as vancomycin, ceftriaxone, and metronidazole, is reasonable [1, 23, 37–39]. While not all institutions have readily available comprehensive data on their MRSA rates for orbital cellulitis, local antibiograms provide useful guidance for empiric antibiotic choices. As with many other conditions with a bacterial source, empiric coverage should be influenced by both local epidemiology and appropriate risk assessment. We recommend the use of an anti-MRSA agent in the following clinical scenarios: (1) suspected or confirmed evidence of intracranial extension, (2) severe illness or toxicity upon initial presentation, or (3) known personal prior history or household contact history of MRSA infection.

Despite the potential for serious complications in orbital cellulitis, management using medical therapy alone for a duration of < 3 weeks appeared to be effective in most of our study cohort, suggesting that 2 to 3 weeks is an adequate length of therapy for uncomplicated cases. Among those who received a longer treatment duration, we postulate that increased failure rates may reflect greater disease severity, either at initial presentation or manifesting as complications during hospitalization, for which our multivariable model was able to adjust. Per the American Academy of Ophthalmology, empiric intravenous antibiotic therapy should be targeted against the most common sinus pathogens followed by an additional 1 to 3 weeks of oral therapy [40]. A consensus statement by the American Academy of Otolaryngology—Head and Neck Surgery Foundation indicated that clinical response may be better with 20 vs 10 days of treatment [41]. Multiple preceding studies support the use of a brief initial parenteral antibiotic course and relatively short total courses [16, 42, 43]. Courdert et al evaluated 121 children who received intravenous metronidazole and a third-generation cephalosporin for a mean duration of 4 days, followed by oral treatment with amoxicillin-clavulanate for about 8.5 days; all were cured without sequelae [8, 23, 44]. Hurley and Harris also used a median of 4 days of intravenous therapy for nonsurgical cases and reported that there was no increase in recurrences and readmissions [45]. Georgakopoulous et al found that a mean duration of 8.6 ± 5.5 days of parenteral therapy, with a transition to oral therapy upon clinical improvement, was effective in most cases among a cohort of 83 children with periorbital or orbital cellulitis although 6% required surgical intervention; the most common parenteral antibiotics used in this study were ceftriaxone and clindamycin (54%) and amoxicillin-clavulanate (24%) [46, 47].

Our study had multiple limitations. While the use of ICD-9 and ICD-10-CM diagnosis codes facilitates the rapid identification of potentially eligible patients, this approach is limited by the accuracy and completeness of coding practices. To mitigate this issue, we confirmed the diagnosis of orbital cellulitis via a manual review of ophthalmologic exam documentation and radiological findings. Because this was a single-center study, findings may not be generalizable to other hospitals with different patient populations or local microbiology. Data on antibiotic use prior to hospitalization, which could have affected microbiology results, were not available. Lastly, the retrospective observational design limits the ability to infer causal relationships or to control completely for confounding factors.

CONCLUSION

Our findings suggest opportunities for antibiotic stewardship in orbital cellulitis management.

Although MRSA was rare, empiric vancomycin use was high. Treatment failure was uncommon in patients who received ≤ 2 weeks of therapy, suggesting that shorter durations are adequate in at least some patients.

Notes

Acknowledgment

We thank Dr Lakshmi Ganapathi for her thoughtful input on the study design and analysis.

Financial support. This work was supported by the National Institute of Child Health and Human Development of the National Institutes of Health (grant number T32HD055148 to B. I. A.). The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Potential conflicts of interest. B. I. A., V. G., and M. M. N. have no associations that might pose a conflict 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.