Cortisol in sickle cell disease: a systematic review and meta-analysis Open Access

Cortisol is a glucocorticoid hormone that regulates multiple physiological processes, including cardiovascular, immune, and metabolic functions, and therefore plays a critical role in the biological link between psychosocial stress and health outcomes.^1,2 Regulated by the hypothalamic-pituitary-adrenal (HPA) axis, cortisol production follows a diurnal pattern that rises in the morning hours and slowly declines throughout the course of the day.³ This profile is interrupted when an individual faces a stressor, and the hypothalamus produces corticotropin-releasing hormones,³ which stimulates the pituitary gland to release adrenocorticotropic hormones (ACTHs). ACTHs then signal the adrenal gland to produce cortisol.³ This stress response is often momentarily adaptive. However, chronic or recurrent acute stressors can result in HPA dysregulation,³ causing a prolonged physiological stress response with elevated cortisol secretion, resulting in several pathophysiological consequences related to increased allostatic load.⁴ Alternatively, HPA dysregulation may cause a disrupted or blunted cortisol response to stress, which can trigger compensatory inflammatory responses, affecting immune function, chronic pain, fatigue, cardiovascular and metabolic health, as well as neurocognition.^5,6 

Individuals with sickle cell disease (SCD), a genetic disease characterized by chronic hemolytic anemia and acute episodes of vaso-occlusion that cause pain and tissue ischemia, experience significant disease-related and psychosocial stressors throughout their lives.⁷ Acute and chronic pain are hallmark symptoms of SCD,⁸ and pain has been investigated as both a stressor and an outcome in the stress-pain cycle in this population.^9,10 For example, an acute vaso-occlusive pain crisis (VOC) can result in missing school or work^11,12; however, psychosocial stressors can increase pain severity and duration.¹³ Additional SCD-related stressors include hospital admissions, treatment side effects, emotional distress, communication barriers, social challenges, and concerns about the future.¹⁴ Alongside disease-related stressors, individuals with SCD face social-ecological stressors related to systemic inequities. SCD is primarily prevalent among individuals of African descent, and, because of social and racial disparities across race, African American individuals with SCD are particularly vulnerable to economic hardship,¹⁵ structural racism,^16,17 and disease-related stigma.^18-20 Although these stressors are not unique to individuals with SCD, they may lead to more severe physiological consequences that negatively affect health outcomes.

It is critically important to understand the biopsychosocial influences of cortisol among individuals living with SCD. Figure 1 proposes a biopsychosocial framework for SCD that posits cyclical biological and behavioral pathways through which psychosocial stressors influence SCD morbidities. The objective of this systematic review was to evaluate the existing evidence of cortisol in SCD, identify current consensus, and highlight gaps in biopsychosocial investigations in this population. Specifically, we aimed to (1) examine cortisol measurement and methodology; and (2) synthesize findings of biopsychosocial factors that have been investigated in association with cortisol among individuals with SCD. A meta-analytic comparison of cortisol among individuals with SCD relative to healthy controls was also conducted.

This systematic review was conducted according to the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines.²¹ Studies were included if they (1) involved patients with SCD, (2) examined and reported a measure of cortisol, and (3) included ≥5 participants. Case reports of <5 participants, reviews, guidelines, viewpoints, editorials, letters to the editor, animal studies, and studies of laboratory investigations were excluded. Conference abstracts without an associated peer-reviewed articles were included to reduce risk of publication bias.²² 

The literature search was conducted in the following databases: PubMed (MEDLINE), EMBASE, and psycINFO. Database searches included articles indexed as of 15 June 2023, without publication date restriction. The search terms used were “sickle cell AND cortisol.” The search was limited to human data, without initial language restrictions. We also scanned reference lists of included studies for additional articles. Two independent reviewers screened titles and abstracts, full-text reviews, and quality assessments. Disagreements were resolved by discussion. Data extraction for studies meeting inclusion criteria was conducted using a standardized form.

Criteria from the National Institutes of Health quality assessment tool for observational cohort and cross-sectional studies were adapted for this review,²³ excluding items that were irrelevant to the aims and inclusion/exclusion criteria, as an examination of each study’s internal validity and risk for bias. Studies were double coded and assigned 1 point per each criterion met, which were summed for a total quality score of 0 to 6 (0 indicating lowest quality and 6 highest quality).

The following information was extracted from each study when available: (1) country in which the study was conducted; (2) descriptive statistics of sample age and sex; (3) hemoglobin (Hb) genotype of SCD sample and healthy controls; (4) study design; (5) cortisol collection protocol; and (6) key findings for cortisol in SCD, including correlates investigated and comparisons across subgroups. Given the variations across plasma and serum in protein binding and the potential for cortisol degradation or alteration during the clotting process, findings for plasma and serum cortisol were evaluated separately.²⁴ After data extraction, results for correlates and comparison groups were categorized into the following categories: (1) comparison with a control group, (2) comparison with an adrenal standard, (3) response to ACTH stimulation, (4) comparison across VOC status, (5) correlation with disease severity, (6) correlation with medication, (7) correlation with behavioral factors, and (8) correlation with additional hormonal and laboratory values.

Summary statistics of cortisol across groups were extracted from each study, as available. Meta-analytic effect sizes were computed for comparisons that had ≥5 independent studies with complete data required for analysis. An aggregate effect size was computed per study when >1 effect size was reported (eg, separate effects across sex). Effect sizes in meta-analyses based on a very small number of studies are subject to problems in data synthesis²⁵; therefore, a minimum of 5 studies (k = 5) was set to estimate effect sizes. Corresponding authors with ≥2 publications included in the quantitative synthesis were contacted to ensure that samples were unique. Analyses and forest plots were conducted/created using R studio using random effects models within “dmetar” and “metafor” packages.

Weighted mean effect sizes (Hedges g) and 95% confidence intervals (CIs) were calculated. Sample size across studies differ greatly; therefore, Hedges g was used as opposed to the Cohen d for comparisons across groups. By convention, Hedges g of magnitudes of 0.2, 0.5, and 0.8 are considered small, medium, and large, respectively.²⁶ The 95% CI on effect size represents the range in which the mean effect size falls in 95% of cases. A mean effect is considered significant when the CI does not include 0.

Figure 2 shows the PRISMA flow diagram detailing study selection. Database searches resulted in 155 articles, and 135 articles remained after duplicates were removed. Of these, 102 were excluded during abstract screening, with 90.9% reliability across raters. Studies were further excluded for having a duplicate sample of another published study (eg, published abstract with full article included in synthesis; k = 12), being published in a non-English language (k = 1),²⁷ not reporting cortisol findings in a SCD-only sample (k = 1), not reporting cortisol findings (k = 6), or being a review paper or chapter (k = 3). Full-text review reliability across raters was 97.0%. Twenty articles met all inclusion criteria and were included in this systematic review. Of these, 12 studies provided minimum data to be included in the quantitative meta-analysis of cortisol in SCD relative to a control group.

Study quality ratings are provided in Table 1. Study quality ranged from 1 to 5 (mean, 3.05; standard deviation, 0.94). Reliability across 2 coders was 91.7%.

A total of 710 individuals with SCD (209 females, 220 males, and 281 unreported biological sex) and 454 healthy controls are represented across the 20 included studies. Participant age ranged from 1 to 77 years, with means ranging from 7.9 to 39.14 years. Three studies included exclusively children (aged 1-10 years)^32,37,38; 4 included children and adolescents (aged 1-17 years)^33,36,42,44; 3 included adolescents and adults (aged 11-77 years)^29,34,40; 8 included exclusively adult participants (aged ≥18 years)^{28,30,31,35,39,45-47}; and 1 included participants across the life span (aged 4-50 years).⁴¹ One study did not specify participant age range.⁴³ 

Study characteristics are summarized in Table 2. Studies were conducted in the United States (k = 5),^{28,35,39,43,45} Nigeria (k = 3),^30,31,44 Saudi Arabia (k = 3),^33,41,42 England (k = 2),^29,40 Oman (k = 2),^37,38 Egypt (k = 1),³⁶ Bahrain (k = 1),³⁴ Jordan (k = 1),³² Brazil (k = 1),⁴⁶ and Cameroon (k = 1).⁴⁷ Cortisol was measured via hair (k = 1),²⁸ plasma (k = 8),^40-47 saliva (k = 1),²⁹ and serum (k = 10).^30-39 Plasma cortisol for patients with SCD ranged from 137.93 nmol/L to 862 nmol/L, and serum cortisol for patients with SCD ranged from 78.89 nmol/L to 841 nmol/L. Assay type included radioimmunoassay (k = 8),^{32,35,37,41,42,44-46} enzyme-linked immunosorbent assay (k = 3),^30,31,33 enzyme immunoassay (k = 2),^28,29 chemiluminescence immunoassay (k = 2),^34,47 and electrochemiluminescence immunoassay (k = 1).³⁶ Four studies did not report assay type.^38-40,43 Eleven studies limited their SCD samples to sickle cell anemia genotypes (ie, HbSS and/or HbSβ⁰),^{29-34,36,4142,44,46} 4 studies included multiple SCD genotypes,^28,40,43,45 and 5 studies did not specify SCD genotype.^35,37-39,47 All studies were cross-sectional and observational in nature.

A summary of study findings is described in Table 3. Included studies reported findings for cortisol in SCD in comparison with a control group (k = 14)^{29,30,32-34,36-38,41,42,44-47}; comparison with an adrenal standard (k = 5)^{34,35,39,40,44}; response to ACTH stimulation (k = 7)^36-39,45-47; response to insulin stimulation (k = 1)⁴⁵; correlation with SCD severity (k = 3)^28,31,41; correlation with medication use (k = 2)^33,40; and additional behavioral,²⁹ hormonal,^34,36 and other laboratory values.^33,42,43 

Overall, 14 studies examined cortisol in SCD relative to a matched control group without SCD.^{29,30,32-34,36-38,41,42,44-47} Of these, 4 studies included an HbAA control group,^30,32,42,47 1 study included separate HbAA and HbAS control groups,⁴⁴ 1 study included a mixed control group including both HbAA and HbAS,²⁹ and 8 studies did not specify the Hb type of the participants in the control group.^{33,34,36-38,41,45,46} Six studies explicitly reported using age-matched controls,^{29,33,34,37,38,47} whereas 2 included sex-matched controls.^34,47 Five studies reported that plasma and serum cortisol were significantly lower among individuals with SCD relative to the control group (2 HbAA, 3 unspecified Hb type),^{32,34,36,41,42} whereas 6 studies did not find a significant difference in cortisol across groups (1 HbAA, 5 unspecified Hb type).^{29,33,37,38,46,47} The remaining 3 studies reported conditional effects. For example, Osifo et al found that plasma cortisol was significantly lower in participants with SCD relative to healthy controls with HbAA, yet plasma cortisol was not significantly different in participants with SCD relative to participants with sickle cell trait (HbAS).⁴⁴ Akinlade et al found that cortisol was significantly lower among individuals with SCD in steady state relative to the control group, but cortisol from individuals with SCD in VOC was not significantly different from the control group.³⁰ However, Rosenbloom et al reported the opposite conditional effect, such that steady state cortisol was not significantly different across SCD and control groups, and cortisol from individuals with SCD in VOC was significantly lower than cortisol in the control group.⁴⁵ 

Twelve studies were included across 2 meta-analytic assessments of effect size (g) for cortisol in SCD relative to healthy controls (Figure 3).^{30,32-34,36-38,41,42,44,46,47} Findings show that the standardized mean difference in serum cortisol (g = −1.51; 95% CI, −3.99 to 0.97) and plasma cortisol (g = −0.72; 95% CI, −1.56 to 0.13) was not significantly different from participants without SCD. Funnel plots portraying publication bias are shown in the supplemental Figures 2 and 3.

Five studies examined cortisol relative to a standard measurement of adrenal functioning to determine the proportion of participants with low or high values.^{34,35,39,40,44} Garadah et al reported that 4 of 82 participants with SCD (4.8%) displayed low cortisol, defined as morning cortisol that was <190 nmol/L.³⁴ Osifo et al found that 9 of 36 participants (25%) displayed plasma cortisol that was lower than the normal range (165-825 nmol/L).⁴⁴ Solomon et al found that 24 of 38 participants with SCD (63%) displayed low serum cortisol (<193.2 nmol/L) on at least 1 occasion.³⁹ However, when only morning cortisol values were considered (4 AM to 8 PM), 13 of 20 (65%) participants with valid samples displayed low serum cortisol.³⁹ Finally, Adams et al found that 7 of 36 participants with SCD (20%) displayed suppressed plasma cortisol (<100 IU/L).⁴⁰ Only 1 study reported findings for elevated cortisol. Groom et al found that 1 of 9 participants with SCD (11%) displayed elevated serum cortisol level (defined as cortisol range of 44-1030 nmol/L).³⁵ 

Meta-analytic analyses of effect sizes could not be conducted because of an insufficient number of studies reporting rate of low cortisol and variable cutoffs for criteria.

Seven studies examined ACTH-stimulated cortisol response.^36-39,45-47 Synthetic ACTH stimulants included cortrosyn (cosyntropin),^36,39,45 synacthen (tetracosactide),^36,38,47 corticotropin,⁴⁶ and glucagon.³⁷ One study used both cosyntropin and synacthen with different patients within the same protocol.³⁶ Of 7 studies examining ACTH stimulation, 5 studies examined this response relative to a control group.^36-38,46,47 Findings showed that the ACTH-stimulated cortisol value was not significantly different in SCD relative to the control group, either 30 minutes (k = 1),⁴⁶ 60 minutes (k = 4),^36,38,46,47 or 120 minutes after stimulation (k = 1),⁴⁶ and peak ACTH-stimulated cortisol was not significant different across groups during a 180-minute test (k = 1).³⁷ However, in Sobngwi et al, the value of the change in cortisol from baseline to 60 minutes after stimulation was significantly lower in participants with SCD relative to those in the control.⁴⁷ Furthermore, 2 studies reported on the proportion of participants that reached “optimal adrenal stimulation.” Sobngwi et al found that 5 of 10 of participants with SCD (50%) did not achieve optimal functioning, defined as reaching a twofold increase of basal cortisol 1 hour after ACTH stimulation.⁴⁷ Solomon et al only assessed optimal ACTH stimulation among participants with SCD, displaying low cortisol in relation to an adrenal standard (ie, cortisol of <193.2 nmol/L).³⁹ Findings from this study showed that 10 of 16 participants with adrenal sufficiency (63%) showed optimal adrenal stimulation, defined as cortisol increases to ≥497 nmol/L 1 hour after stimulation.³⁹ Finally, Rosenbloom et al investigated how VOC status affected ACTH response in a small sample of individuals with SCD, with findings showing no significant difference in ACTH-stimulated cortisol in patients in VOC (n = 6) relative to the crisis-free group (n = 5) 60 minutes after stimulation.⁴⁵ 

These data were insufficient for meta-analytic synthesis (k <5 across comparisons at each poststimulation time point).

Insulin tolerance tests are also used to investigate HPA functioning.^48,49 Rosenbloom et al investigated how VOC status affected cortisol response to insulin stimulation in a small sample of individuals with SCD (n = 5-7) relative to health controls (n = 4).⁴⁵ Insulin-stimulated plasma cortisol among participants with SCD during VOC (n = 7) and a subset of the same participants in steady state (n = 5) was not significantly different from controls 30 and 60 minutes after stimulation; however, at 90 minutes after stimulation, individuals with SCD in VOC displayed lower insulin-stimulated cortisol relative to the controls.⁴⁵ 

Three studies examined the difference in basal cortisol levels among participants with SCD during VOC relative to participants with SCD in steady state.^30,44,45 Of these, 2 studies found that cortisol levels were significantly higher during VOC relative to steady state.^30,44 Conversely, 1 study found that plasma cortisol was not significantly different during VOC relative to repeated measures plasma cortisol among the same participants in steady state (n = 5).⁴⁵ 

Three studies examined cortisol in relation to broad disease severity indices.^28,31,41 Hollister et al found that hair cortisol content was not associated with a cumulative disease severity index comprising both acute and chronic complications including pain, infection, acute chest syndrome, retinopathy, leg ulcers, renal failure, cardiopulmonary dysfunction, and endocrine abnormalities.²⁸ el-Hazmi et al also found that plasma cortisol was not related to a disease severity index comprising total number of symptoms, transfusions, and hospitalizations in the previous year.⁴¹ However, Akinlade et al found that serum cortisol was significantly higher in participants with severe sickle cell anemia (ie, greater number of blood transfusions, VOCs, presence of anemia, and organ complications) relative to participants with milder disease, as indicated by the index used.³¹ 

Two studies examined the association of cortisol with medication use.^33,40 El-Sonbaty et al investigated the relation between hydroxyurea use, the most-used disease-modifying treatment for SCD, with cortisol levels. Plasma cortisol was not significantly different among participants with SCD receiving hydroxyurea relative to participants with SCD not receiving hydroxyurea.³³ Adams et al examined cortisol level relative to prescription opioid dose, showing that plasma cortisol was not correlated with opioid dose.⁴⁰ 

Behavioral correlated were only assessed in 1 study.²⁹ Kölbel et al examined salivary cortisol relative to sleep and cognitive profiles. Morning saliva cortisol was negatively correlated with morning wake time. Participants with a flat saliva cortisol profile had a later wake time compared with participants with a typical cortisol profile. Cortisol profiles were not related to the remaining sleep variables assessed or cognitive profiles.

Two studies investigated ferritin as a correlate of plasma cortisol, with mixed findings.^34,36 Garadah et al found no association between ferritin and plasma cortisol,³⁴ whereas Hagag et al found an inverse correlation such that lower morning plasma cortisol was related to higher ferritin.³⁶ This study also found an inverse correlation with serum iron, and a positive correlation with total iron binding capacity.

El-Sonbaty et al found that plasma cortisol was not significantly related to age, gender, melatonin level, weight, height, or body mass index.³³ Findings from el-Hazmi et al show that plasma cortisol was not significantly correlated with gonadotropic hormones, including luteinizing hormone, follicle stimulating, and testosterone.⁴² Finally, Grewen et al found a positive correlation between oxytocin and plasma cortisol.⁴³ 

Findings from this systematic review and meta-analysis elucidate how cortisol has been investigated among individuals living with SCD, and they reveal gaps in current knowledge. Importantly, inconsistencies in results identified in the reviewed studies highlight opportunities for future research.

One area that we believe will be of interest in the design of future studies is the choice of methodology to measure cortisol in patients with SCD. Given the diurnal pattern of cortisol, it is imperative to ascertain repeated samples over the course of a day and to obtain samples over several days to determine a typical diurnal profile. Thus, saliva has been deemed the gold-standard method of cortisol measurement because of the noninvasive methodology of multiple assessment.⁵⁰ Yet, only 1 study in this systematic review used salivary cortisol and investigated a diurnal pattern.²⁹ Most identified studies measured cortisol in plasma and serum, likely because of the ease of assessment of samples stored in institutional biobanks or obtained with routine blood draws. However, the HPA axis is highly sensitive to environmental stressors,^1-4 and the sampling environment can influence cortisol measurements.^51,52 Therefore, measurements within a hospital or laboratory setting may lead to an overestimation of basal levels in the SCD population that is often faced with substantial medical trauma.⁵³ Moreover, plasma measurements cannot distinguish between free cortisol, which is biologically active, and bound cortisol (attached to proteins such as cortisol-binding globulin, which is biologically inert).⁵⁴ In addition, only 1 study used hair measurements,²⁸ which is a particularly useful method for assessing the impact of chronic stressors over weeks to several months.⁵⁵ Although work in SCD is limited, there is a literature on using hair collection in Black and African American populations that show moderate acceptability and feasibility of this protocol.^56,57 Thus, the literature on cortisol in SCD could be strengthened by future work using salivary and hair measures of cortisol to assess diurnal patterns and cumulative stress, respectively.

Adrenal insufficiency can occur because of dysfunction of the adrenal gland (ie, primary insufficiency), pituitary gland (ie, secondary insufficiency), or hypothalamus (tertiary insufficiency).^58-60 Although comparatively low cortisol values relative to an adrenal standard or in response to ACTH stimulation may indicate adrenal insufficiency, diagnosis typically requires both clinical symptoms and laboratory findings. Nevertheless, individuals with blunted cortisol responses that do not yet meet criteria for adrenal insufficiency continue to show increased morbidities in mental and physical health.⁶¹ Qualitative findings on cortisol levels relative to an adrenal standard in this review were inconsistent, with ∼5% to 65% of participants with SCD displaying low cortisol on at least 1 occasion. However, these studies used several benchmarks to calculate the incidence rate, despite previously published guidelines suggesting that morning serum cortisol of <165 nmol/L is the standard cutoff for primary adrenal insufficiency, and serum cortisol of <100 nmol/L is the cutoff for secondary adrenal insufficiency.⁶² Newer research now suggest assay-specific cutoffs.⁶³ Finally, findings from this review suggest that ACTH-stimulated cortisol does not differ from that of healthy controls. Future research investigating ACTH response in SCD should consider using more sensitive methodologies and newer guidelines related to adrenal insufficiency diagnoses. In addition, it will be important to clarify SCD-specific mechanisms of adrenal response patterns to determine if, and how, hemolysis and ischemia affect the HPA function.

Findings for comparisons of cortisol related to a control group without SCD are further complicated because of the limited use of age- and sex-matched participants. Because cortisol levels and patterns are significantly affected by age and sex,^64,65 these demographic factors should be carefully considered in study design and analyses. Moreover, several studies included in the review feature wide age ranges that span multiple developmental stages, further complicating comparison of cortisol levels. Future research should account for the evolving stress response system when assessing cortisol in children and adolescents with SCD. In addition, given the highest rates of morbidity and mortality occur during young adulthood,⁶⁶ understanding how environmental and psychosocial stressors affect individuals in this critical developmental period is particularly important.

Findings also suggest that cortisol increases during VOC, which aligns with current research on cortisol during acute pain events.⁶⁷ Findings from this systematic review also suggest that cortisol is not related to SCD severity composites; however, these results are obscured by the broad measures used within cumulative risk variables and the lack of a validated measure of disease severity. These findings are further limited by their cross-sectional nature, because no study used prospective design to truly determine whether cortisol is a predictor of VOC and SCD outcomes or an outcome of these changes. The literature would benefit from longitudinal research to determine how cortisol is related to SCD mechanisms.

Among the most important findings from this systematic review is that, shockingly, no study investigated specific environmental or patient-reported stressors in relation to cortisol in SCD. Given that cortisol plays a critical mediating and moderating role in the association between social and environmental exposures and health outcomes,^1-4,68-70 it is critical to investigate these pathways among individuals living SCD who experience additional biopsychosocial risks because of their disease.^66,71-75 Although biopsychosocial frameworks, such as the 1 described in Figure 1, describe broad heuristic associations for the role of stress response in the association between stressors and disease, it appears that the current literature in SCD has only begun to elucidate a single physiological pathway between cortisol and SCD morbidities without taking psychosocial or social-ecological stressors or cyclical behavioral pathways into account. There is a fundamental need to further investigate biobehavioral pathways to then identify intervention targets and mechanisms of change to improve health outcomes for individuals living with SCD.

This systematic review has several strengths,⁷⁶ including rigorous methodology, search strategy without limitations on year of publication, inclusion of published abstracts, review process by 2 independent reviewers, and inclusion of meta-analytic assessment when available. Limitations of the review and included studies include limited interpretation of findings because of the inclusion of studies with small sample size, low study quality because of inconsistency in collection timing, low or unreported participation rate, inconsistent cortisol level cutoff ranges, varied conceptual definitions (eg, disease severity), and broad age ranges with a lack of consideration of how developmental stages influence cortisol measurement and patterns.

Findings from this systematic review and meta-analysis show that cortisol in SCD is (1) primarily measured using plasma and serum methods, (2) is not significantly lower relative to healthy controls, and (3) is elevated during VOC. Furthermore, findings show that there is no research in relation to self-reported stressors, and this patient population would benefit from more research aimed at understanding the relationships between cortisol values, social-ecological stressors, and adverse health outcomes. Understanding cortisol in the context of biopsychosocial investigations offer several benefits to individuals living with SCD, including improved diagnostic accuracy of HPA-related complications and targeted interventions to reduce stressors and maladaptive biobehavioral outcomes.

This project was funded by the Children's Hospital of Philadelphia Research Institute and the Bridge to Faculty Program.

Contribution: K.V.P. designed the review protocol; K.V.P. and M.M.B. carried out the screening processes for inclusion and exclusion and extracted and analyzed the quantitative data; K.V.P., M.M.B., K.G., R.J.W., and A.A.T. interpreted the data; K.V.P. and M.M.B. drafted the manuscript; K.G., R.J.W., and A.A.T. critically revised the manuscript; and all authors approved the submitted final version of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Kemar V. Prussien, Children’s Hospital of Philadelphia, 3500 Civic Center Blvd, Philadelphia, PA 19104; email: prussienkv@chop.edu.