Propranolol vs Prednisolone for Symptomatic Proliferating Infantile Hemangiomas: A Randomized Clinical Trial

Importance  While propranolol is touted as superior to prednisolone for treating infantile hemangiomas (IH), a randomized clinical trial (RCT) comparing the outcome and tolerability of these medications for symptomatic, proliferating IH has not been reported.

Objectives  To determine if oral propranolol is more efficacious and better tolerated than prednisolone in treating symptomatic, proliferating IH and to determine the feasibility of conducting a multi-institutional, RCT comparing efficacy and tolerability of both medications.

Design, Setting, and Participants  Phase 2, investigator-blinded, multi-institutional RCT conducted in 3 academic vascular anomalies clinics on 19 of 44 eligible infants aged between 2 weeks and 6 months. All participating patients had symptomatic proliferating IH treated between September 1, 2010, and August 1, 2012.

Interventions  Treatment with oral propranolol vs prednisolone (2.0 mg/kg/d) until halted owing to toxic effects or clinical response.

Main Outcomes and Measures  Primary outcome was change in IH size after 4 months of therapy. Secondary outcomes were response rate and frequency and severity of adverse events (AEs).

Results  The primary outcome showed no difference in lesion size or affected skin area after 4 months of therapy: 41% and 1.32 mm² for prednisolone vs 64% and 0.55 mm² for propranolol (P = .12 for lesion size, and P = .56 for affected skin area). Longitudinal analyses showed a faster response in total lesion outer dimension with prednisolone (P = .03), but this advantage over time was not noted when central clearing and outer dimension were included in the analysis (P = .91). The overall frequency of AEs was similar (44 for prednisolone vs 32 for propranolol) (P = .84), but prednisolone-treated participants had more grade 3 severe AEs (11 vs 1) (P = .01), particularly growth retardation resulting in size and weight below the fifth percentile. Early study withdrawal owing to AEs occurred in 6 (75%) of 8 patients in the prednisolone group but 0 of 11 propranolol-treated participants. The mean duration of therapy was shorter for prednisolone (141 vs 265 days), reflecting the higher rate of early withdrawals.

Conclusions and Relevance  Both medications show similar efficacy for reducing the area of symptomatic, proliferating IH. Although prednisolone showed a faster response rate, propranolol was better tolerated with significantly fewer severe AEs. Propranolol should be the first line of therapy for symptomatic IH unless contraindicated or unless future studies demonstrate severe AEs from propranolol. Recruiting participants for a phase 3 RCT would be difficult owing to safety profiles measured here and emerging trends favoring propranolol.

Trial Registration  clinicaltrials.gov Identifier: NCT00967226

Infantile hemangiomas (IH) are the most common benign tumors of infancy, arising in 4% of live births.^1,2 While IH exhibit a characteristic growth pattern of proliferation and subsequent involution during the first decade of life, 12% of lesions require consideration of therapy during infancy for indications including ulceration, bleeding, impairment of sensory organ function, airway compromise, and risk of permanent deformity in cosmetically sensitive regions.^3,4

Subsequent to the recent serendipitous discovery of propranolol’s effects on symptomatic IH,⁵ the nonselective β-blocker may be supplanting systemic corticosteroids as the initial therapy of choice. Since there are no drugs for IH approved by the US Food and Drug Administration, both medications are used as off-label treatments, and their perceived efficacy and tolerability stems from observational studies and expert opinion. To our knowledge, the present multi-institutional, phase 2, randomized clinical trial (RCT) is the first to evaluate the performance of propranolol vs prednisolone in terms of (1) IH response, (2) medication tolerability, and (3) feasibility of conducting a phase 3 trial.

Written consent for study participation was obtained from all patients’ parents or guardians during the clinic evaluation. The study was reviewed by the Data Safety Monitoring Board (DSMB) quarterly, registered with clinicaltrials.gov (NCT00967226), and approved by institutional review boards of the Children’s National Medical Center (CNMC) (No. 0000467), University of Iowa (UIHC) (No. 201004770), and Johns Hopkins Hospital (JHH) (No. NA_00049604).

All consecutive eligible infants aged 2 weeks to 6 months with actively proliferating and symptomatic IH evaluated in the outpatient vascular anomalies clinics of the study institutions were invited to participate without sexual or racial restriction. Lesions were considered symptomatic if they impaired function (including breathing, eating, vision, and hearing), were ulcerated, caused pain, or were in cosmetically sensitive regions. Exclusion criteria included inadequate social support, nonproliferating IH, other treatment for IH, liver disease, abnormal blood glucose level (BG), hypertension, hypotension, reactive airway disease, cardiac anomalies, and PHACE (posterior fossa brain malformations, hemangiomas, arterial anomalies, coarctation of the aorta and cardiac defects, and eye abnormalities)^6,7 if significantly narrowed intracranial vasculature was present.

Intake history included demographic data, IH characteristics, and medical history. Participants self- reported race and ethnicity. A complete physical examination was completed, including evaluation of vital signs and nonfasting BG. Hypoglycemia was defined as BG lower than 70 mg/dL. Imaging studies were completed only when necessary to confirm diagnosis, determine IH extent, or evaluate for PHACE.^6,7 Participants with stridor or beard distribution of IH underwent airway endoscopy to exclude subglottic hemangiomas. Liver ultrasonography was performed for participants with 4 or more hemangiomas.

This multi-institution, single-blind, placebo-controlled RCT targeted enrolling 55 participants per group to achieve 80% power to detect a 30% absolute reduction in lesion size between groups assuming the smallest reduction was at least 50%. Consenting participants were randomized to prednisolone or propranolol using a CNMC institutional tamper-proof, pregenerated encrypted schedule, which allocated equal assignments to both treatments within institutions and within lesions affecting and not affecting vision. Medications were dispensed free of charge. No stipend was provided.

At enrollment, the blinded investigators assigned a number to each of the IH corresponding to 1 of 76 unique body segments⁸ and recorded the following established parameters of the IH⁸: type (superficial, deep, or mixed), presence of ulceration, total area (product of transverse and longitudinal measurements in square millimeters), degree of normal intervening skin (deep lesions, none; superficial and mixed lesions, 75%-100%, 50%-74%, 25%-49%, 10%-24%, or <10%). If subjects had multiple IH, the largest was analyzed. The same blinded investigator repeated the measurements of size and skin involvement monthly. Photographic documentation, blood pressure (BP), heart rate (HR), weight, height, and auscultation findings of heart and lungs were also recorded at each visit, and BG was measured at the first visit. Weight and height findings were plotted on infant growth sheets (http://www.cdc.gov/growthcharts). Involution was defined by reduction in size or percentage of skin involvement from prior visit.

The CNMC research pharmacy dispensed study drugs. Prednisolone was prescribed as 2.0 mg/kg/d (1.0 mg/kg orally twice daily in 15-mg/5-mL solution; Hi Tech Pharmacal). Unblinded caretakers received counseling and written instructions to administer the medication 15 minutes before meals, to avoid sudden cessation and to withhold live vaccines until 3 months after completion of therapy. A letter was sent to the primary care provider advising corticosteroid stress dose coverage under appropriate circumstances.

Propranolol was prescribed at 2.0 mg/kg/d (0.68 mg/kg orally 3 times daily, 20-mg/5-mL solution; Roxane Laboratories), and participants were admitted for cardiac monitoring for the first 3 doses after normal baseline electrocardiogram and vital signs were verified. The BP, HR, and BG were checked 1 hour after each dose. Dosing was reduced to 75% target for BG lower than 70 mg/dL or if BP or HR declined below the fifth percentile for adjusted age, weight, and height.^9,10 Caretakers were trained to recognize signs of hypoglycemia, hypotension, and bradycardia that would warrant withholding of medication. To reduce the hypoglycemia risk, caretakers were instructed to feed infants younger than 6 weeks at least every 4 hours; infants aged 6 weeks to 4 months at least every 5 hours; and infants older than 4 months at least every 6 hours and to allow a minimum of 6 hours between doses. Caretakers and primary care physicians were advised to withhold propranolol for babies whose oral intake declined for any reason. Medication administration was deemed compliant if less than 20% of dispensed medication volume was returned.

Caretakers were queried at each visit regarding interval AEs. This open-ended question was followed by specific inquiries regarding interval fever, use of medications, hospitalizations, visits to medical providers, ear infections, respiratory illnesses, rhinorrhea, cough, wheezing, and changes in eating, sleeping, or bowel habits. AEs were categorized by body system and severity per Common Terminology Criteria for Adverse Events (CTCAE) (http://evs.nci.nih.gov/ftp1/CTCAE/About.html): grade 1, mild; grade 2, moderate; grade 3, severe; grade 4, life-threatening; and grade 5, fatal.

Study participation ended and medications were weaned if (1) IH resolved, and at least 4 months of therapy was completed; (2) no measurable improvement was noted in the lesion at 2 sequential monthly evaluations; (3) severe, unresolving AEs occurred that were deemed related to therapy; (4) caretakers or referring physicians requested withdrawal; and/or (5) no clinical improvement was evident after 1 month of therapy (nonresponder). This last criterion identified treatment failures and protected participants with symptomatic IH from ineffective therapy. The study was not a crossover study.

The primary feasibility outcome was enrollment; secondary outcome, appointment and medication compliance. Feasibility was evaluated by contingency tables comparing target-to-actual enrollment and target-to-actual compliance for study visits and missed drug doses.

The a priori primary outcome was each of the IH’s size measured by the proportional change in the total surface area (TSA) based on the lesion’s outer margin dimensions at 4 months. The adjusted TSA (AdjTSA) was also compared for lesions with skin involvement to account for central clearing of the hemangioma within its outer margins that would not have been accounted for if measuring outer dimensions alone. The proportion of affected skin was defined as the inverse of the proportion of measured normal skin, with each category defined according to its central estimate (ie, the proportion of affected skin: 0.125, 0.38, 0.63, 0.83, or 0.95). The AdjTSA was defined as this proportion multiplied by the TSA.

Analyses were performed with an intention-to-treat approach. For normally distributed data, the t test was used, while the Wilcoxon rank sum test was used for skewed data. Longitudinal analyses evaluated the trajectory of change in lesion size over time accounting for multiple correlated measurements per participant. Cross-product terms, time × time and treatment group × time, were included if necessary to reflect nonlinearity as well as to allow for different trajectories per treatment group. Cross-product terms were retained in the model only if they achieved statistical significance or near significance. An adjusted longitudinal analysis was also performed, which controlled for baseline differences in size in both groups. These analyses adjusted variance estimates to account for multiple correlated assessments per participant. A sensitivity analysis was performed, which excluded noncompliant patients (n = 2). Statistical evaluation was performed using Stata 12.0 (StataCorp LP) using standard commands for tabular data, histograms, t test, Wilcoxon rank sum test, and panel data analysis.

Tolerability outcomes were the frequency, site, and severity of AEs. Contingency tables and χ² tests were used to evaluate the magnitude and statistical significance of differences by treatment group in the frequency of the most severe AE per person by type and severity. Mixed model logistic regression analysis was also used to compare the frequency by treatment group of all AEs, including multiple events per participant, by type and severity. The model allowed us to correct P values by adjusting variance estimates for the correlation between multiple events on the same participant.

The study was terminated prior to targeted enrollment at the DSMB’s recommendation owing to severe AEs described herein that prompted early withdrawal of 6 of the 8 the prednisolone participants (75%).

Of 44 consecutive eligible participants evaluated between September 1, 2010, and August 1, 2012, 19 were enrolled and randomly assigned to prednisolone (n = 8) or propranolol (n = 11) (Figure 1, Table 1). The rate of enrollment was 89% at CNMC (16 of 18), 20% at JHH (3 of 15) and 0% at UIHC (0 of 11). Fear of being randomized to prednisolone was the most commonly cited reason for study refusal, and this concern increased over the study period.

The female to male ratio was 4:1. There was no difference between groups in sex, race, ethnicity, mean enrollment age, presence of ulceration, morphologic characteristics, lesion depth at enrollment, baseline TSA, or baseline AdjTSA (Table 1). One participant had PHACE but without contraindication to propranolol; 3 underwent ultrasonographic imaging (2 to determine lesion extent and 1 to exclude liver hemangiomas); and 3 had subglottic hemangiomas excluded by direct laryngoscopy.

Cumulatively, 138 of 154 scheduled monthly visits were completed, for an attendance rate of 90%. Twelve participants missed no scheduled appointments; 3 missed 1 appointment; and 1 each missed 2, 4, and 5 appointments. Lack of transportation was the leading cause for multiple missed appointments. All participants were compliant with medication use with the exception of 2 participants who missed 1 week or 2 weeks of therapy because their prescriptions ran out at the time of a missed appointment.

In an intention-to-treat analysis, there was no difference in the proportion decrease of TSA or AdjTSA at 4 months, both when the data from 4 months alone was evaluated and when the data from 5 months was evaluated as a surrogate for subjects who missed their 4-month follow-up (P > .11) (Table 2).

The longitudinal analysis of the impact of treatment group on TSA showed that the rate of decline was statistically significantly faster in the prednisolone group (P = .04) (Figure 2A). When baseline TSA was taken into account (by adjusting for the larger baseline TSA in the propranolol group at the outset), results were similarly significant (P = .03) (Figure 2B). When individual month intervals were examined with the model as a predictor, no significant differences were noted at the individual month timeframes in both the univariate and the adjusted analysis.

The AdjTSA response over time was similar in both groups in both the unadjusted baseline (P = .20) and adjusted baseline analysis (P = .91) (Figure 3)

When excluding the data from the 2 noncompliant patients, there was still no difference between adjTSA (P = .12) or TSA (P = .57) in the primary analysis, and longitudinal unadjusted and adjusted analyses were consistent with the intention-to-treat analysis.

For participants concluding the study (2 [25%] of 8 in the prednisolone group and 9 [82%] of 11 in the propranolol group), the mean time in days from start to finish of therapy was 323 (95% CI, 223-423) for propranolol and 300 for prednisolone (95% CI, 0-643).

Regrowth after medication weaning was noted in 2 propranolol-treated participants and 1 prednisolone-treated participant, all of whom responded to resuming treatment with their respective medications.

There were more early withdrawals due to medication-related concerns in the prednisolone group than in the propranolol group (6:1) (P = .01). Adverse events were the cause of study withdrawal in none of the propranolol participants but in 5 of 6 of the prednisolone participants, either at the request of the primary care physician (owing to growth concerns; n = 4) or the request of the parents (owing to marked flushing and sweating that stopped with prednisolone treatment cessation; n = 1) (Table 3). Despite the absence of AEs, another prednisolone participant withdrew at the recommendations of his pediatrician and ophthalmologist, who counseled the parents that their infant should be treated with propranolol. A single propranolol-treated participant withdrew from the study, even though her lesion was regressing markedly after 5 months of therapy, because the parents requested laser treatment of telangiectasias. This necessitated study withdrawal because alternative treatment was a study exclusion criterion. Another propranolol participant expressed regrets about being in the study and withdrew immediately after enrollment but requested propranolol treatment outside the study.

The frequency of all AEs was similar in the prednisolone and propranolol groups (44 vs 32) (P > .84), but severe AEs (CTCAE grade 3) occurred more frequently in prednisolone participants (11 vs 1) (P = .01) (Table 3). Nine of the severe AEs in prednisolone participants were related to decline in linear growth and/or weight to less than 5% of the standardized growth curve measures and greater than 50% from participant’s pretherapy measures. The remaining severe AEs were due to adrenal crisis associated with viral gastroenteritis (n = 1) and severe dehydration warranting hospitalization (n = 1). The severe AE in the propranolol participant was dehydration warranting hospitalization. No propranolol-treated participant experienced significant decline in growth or weight.

Propranolol participants had significantly more pulmonary symptoms than prednisolone participants (14 vs 5) (P < .001), predominantly upper respiratory tract infections (URTIs) (Table 3). The number of vascular AEs was comparable but reflected BP increases in prednisolone participants and BP decreases in propranolol participants. All vascular changes were asymptomatic and resolved spontaneously. One propranolol participant developed asymptomatic hypoglycemia after the first dose of propranolol (64 mg/dL). Per protocol, the next dose of propranolol was skipped, and BG remained above 70 mg/dL for subsequent doses.

To our knowledge, the present study is the only phase 2 RCT to compare efficacy, in terms of reduction of IH size, as well as tolerability of propranolol vs prednisolone for treatment of symptomatic IH. Dramatic responses were seen with both medications (Figure 4). Study strengths include the rigorous adherence to enrolling only proliferating IH and the high rate of compliance, with fewer than 10% of missed monthly appointments. Since initiation of this study, propranolol has quickly become the first line of therapy for symptomatic IH,¹¹ even in the absence of data comparing it with the historic standard of prednisolone. Completing a similar study on a larger scale may no longer be feasible or desirable, given the findings of the present series showing fewer severe AEs with propranolol.

Propranolol’s rapid rise in popularity is remarkable.⁵ In September 2010, only 6 series described its use in 10 or more patients/series,¹² but by December 2011, the number of treated patients reported had grown to 1200.^13,14 Reports of propranolol’s efficacy quickly disseminated through the nonmedical community via the internet as well,¹² and parents now present to our vascular anomalies clinic and request propranolol, should IH therapy be indicated.

Enrollment of 19 participants instead of 55 was less than anticipated for several reasons. Patients often presented after the requisite period of active IH proliferation when prednisolone is considered most effective.¹⁵ Additionally, prior therapy was forbidden, so infants receiving any topical medications were ineligible. Most importantly, one of our collaborative institutions failed to recruit eligible participants during the entire first year of the study. It is now clear that the investigators from this institution had lost their point of equipoise and should not have agreed to participate in the study. The mounting evidence of propranolol’s efficacy made enrollment difficult for our second collaborative institution that joined the study 15 months after initiation.

Although the low enrollment reduced the trial’s power to detect differences as originally designed, the randomization schedule prevented bias and still allowed the study to detect larger than anticipated differences. The high rate of dropout of prednisolone-treated participants owing to AEs occurred beyond the first third of the study and beyond the period when efficacy effect was established. Despite the limited enrollment and high dropout rate, the study provides evaluation of the relative effectiveness of the treatments superior to that of the uncontrolled observational case reports currently existing in the literature.

Caretakers were not blinded because we could not justify unnecessarily admitting prednisolone participants to initiate therapy or unnecessarily withholding live vaccines from propranolol participants. Furthermore, caretakers needed to know the medication assignment to ensure participant safety. As an illustration, parents of a propranolol participant with poor oral intake owing to a gastrointestinal viral illness should withhold propranolol to prevent hypoglycemia, but parents of a prednisolone participant with the same symptom should not withhold prednisolone and should even possibly increase stress steroid coverage.

Both medications were efficacious in reducing IH TSA and AdjTSA; prednisolone showed a faster response rate than propranolol for TSA but not for AdjTSA. However, the need to stop treatment with prednisolone in 75% of participants owing to AEs offsets its potential advantage. It remains unclear if either medication offers an advantage in change in TSA or AdjTSA beyond 12 months.

The significantly higher rate of severe AEs in prednisolone participants prompting early withdrawal was remarkable. The overall number of AEs was similar, but those in propranolol subjects were predominantly grade 1 or 2, while the AEs in prednisolone subjects were grade 3 due to severe growth retardation. Although growth retardation is a predictable and temporary AE of prednisolone treatment in infants, the remarkable incidence necessitating study withdrawal was striking. After extensive deliberation, our DSMB concluded that the high rate of severe drug-related AEs warranting study withdrawal in prednisolone subjects was not offset by sufficient treatment advantages and that further enrollment should halt. It is likely that pediatricians caring for patients with IH share the concerns of our DSMB regarding the severe, albeit reversible, AEs associated with prolonged corticosteroid use.

The decision to halt enrollment based on a well-recognized AE of prednisolone was difficult, since it prevented the full comparison of the efficacy of these 2 medications. Knowledge of outcome efficacy may prove useful in the future should life-threatening cardiovascular events from propranolol occur or concerns of propranolol irreversibly affecting memory retention in children be demonstrated.¹⁶ Such findings might tilt the treatment scale away from propranolol, since severe growth restriction from prednisolone may be disturbing but is reversible.¹⁷ The favorable response profile of prednisolone is encouraging for patients with symptomatic IH and a contraindication to propranolol use.

Notably, the rate of respiratory AEs, particularly URTIs, was higher in the propranolol-treated group. Larger studies will be required to determine if propranolol increases the risk or severity of URTIs.

A major limitation of IH research has long been the lack of a validated objective method by which to measure growth and response to therapy. No such tool had been reported at the initiation of our study, but a recently described scale may prove useful for future IH studies.¹⁸

Clearly the most important remaining question of propranolol use is whether the medication merely hastens involution onset or actually improves the outcome of involution. This is particularly important because pharmacologic therapy with corticosteroids historically has been used only in symptomatic IH because of associated AEs, but the growing evidence of propranolol’s safety and efficacy is expanding its use to even small, superficial facial lesions that may leave unsightly facial scars once involuted.¹³ Whether early initiation of propranolol treatment when an IH first presents as a macular, vascular blush prevents its growth into a raised, cosmetically deforming lesion also needs to be determined.

The results of this RCT for treatment of symptomatic, proliferating IH show that prednisolone and propranolol showed no difference in the primary outcome of lesion area at 4 months but that prednisolone may result in a faster rate of involution of a lesion’s outer margin over time. Propranolol, however, is better tolerated than prednisolone, with significantly fewer severe AEs leading to drug-related withdrawals. Based on similar efficacy between medications but the higher rate of severe AEs with prednisolone therapy, propranolol should be considered the first line of therapy for symptomatic IH if contraindications to its use do not exist. Further studies are necessary to determine whether there are additional, undemonstrated long-term AEs associated with propranolol use and if propranolol merely hastens the involution process or changes ultimate outcome.

Corresponding Author: Nancy M. Bauman, MD, Department of Otolaryngology, Children’s National Medical Center, 111 Michigan Ave NW, Washington, DC 20010 (nbauman@childrensnational.org).

Submitted for Publication: September 20, 2013; final revision received November 8, 2013; accepted December 16, 2013.

Published Online: February 13, 2014. doi:10.1001/jamaoto.2013.6723.

Author Contributions: Drs Bauman and McCarter had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Bauman, McCarter, Guzzetta, Oh, Preciado, Puttgen.

Acquisition of data: Bauman, Guzzetta, Oh, Greene, Puttgen.

Analysis and interpretation of data: McCarter, Shin, Oh, Preciado, He, Greene, Puttgen

Drafting of the manuscript: Bauman, McCarter, Guzzetta, Shin, Puttgen.

Critical revision of the manuscript for important intellectual content: Bauman, McCarter, Guzzetta, Shin, Oh, Preciado, He, Greene, Puttgen.

Statistical analysis: Bauman, McCarter, Shin, He, Puttgen.

Obtained funding: Bauman.

Administrative, technical, and material support: Bauman, McCarter, Guzzetta, Oh, Preciado, Greene, Puttgen.

Study supervision: Bauman, Guzzetta, Greene, Puttgen.

Manuscript review: Guzzetta, Oh, Preciado, Greene.

Conflict of Interest Disclosures: Drs Bauman, McCarter, and Puttgen received modest salary support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, and from the National Center for Research Resources of the National Institutes of Health to complete the study. Dr Shin receives book royalties from Springer Publishing and Plural Publishing for Evidence-Based Otolaryngology and Otolaryngology Prep and Practice, respectively. No other disclosures were reported.

Funding/Support: Work for the research reported in this publication was supported by grant R21HD062959-02 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, and by grant UL1RR031988 from the National Center for Research Resources of the National Institutes of Health.

Role of the Sponsor: The financial support was used for design and conduct of the study, study medications, and collection, interpretation and analysis of the data, but the sponsors themselves had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Additional Contributions: The authors acknowledge the support of Ed Connor, MD, PriyaVaidyanathan, MD, and Marlene Lee, RN, who served on the DSMB; Jennifer Gode, RN, who served as clinical nurse coordinator for a portion of the study; Margie Brown, who assisted with manuscript preparation; and the study participants and their parents.

Correction: This article was corrected on March 11, 2014, to fix a value in Figure 1.