Oral Corticosteroid Use Is Effective for Cutaneous Hemangiomas: Title and subTitle BreakAn Evidence-Based Evaluation FREE

HEMANGIOMAS, the most common tumors of infancy and childhood,¹ do not require treatment unless they interfere with normal function (ie, breathing, vision, hearing, eating, voiding, stooling, and movement), produce serious disfigurement that is unlikely to resolve on its own, or are otherwise problematic (intractable bleeding, ulceration, infection, pain, coagulation defects, heart failure, etc).² When treatment is required, the mainstay of medical therapy is corticosteroids,³ but dosage recommendations, duration of treatment, recommendations for monitoring during and after treatment, and methods of tapering vary widely. No prospective study has been performed, to our knowledge, to determine the dose-response relationship of systemic corticosteroid treatment for hemangiomas of infancy, although many individual case reports and case series have been published.

Quantitative systematic review (meta-analysis) is a relatively new tool to help physicians make evidence-based decisions. This technique involves the systematic examination of available evidence on a given clinical topic and the summary of that evidence using statistical techniques that pool data from multiple studies to yield a single result.⁴

The purpose of this study was to use quantitative systematic review techniques to determine the efficacy of systemic corticosteroid therapy for enlarging, problematic, infantile cutaneous hemangiomas and to assess the relationship of dose to response and the adverse effect profile.

Two electronic databases, PubMed and MEDLINE, were used to search the medical literature for publications on the combined topics of "hemangiomas and corticosteroids." A careful review all of the titles and abstracts was done to locate all case series meeting the inclusion criteria: (1) at least 5 patients with cutaneous hemangioma(s) treated with systemic corticosteroids and (2) original studies. Articles were obtained and read in detail. A review of the bibliographies of these articles (hand search) was also done to determine other case series that might have been inadvertently missed using the electronic search strategy. This hand search produced 4 articles not found by the electronic search. Case series in foreign languages whose abstracts revealed compliance with inclusion criteria were obtained, translated into English, and read in detail. Unpublished data and "gray literature" were not useful for the purposes of this analysis.

The following exclusion criteria were then rigidly applied to the patients in each case series to eliminate inappropriate patients and produce a homogeneous population amenable to systematic review:

Data collected from each study for each appropriate patient included age and weight at the initiation of corticosteroid treatment, treatment regimen and duration, response, time to response, whether rebound occurred, adverse effects, and a description of nonresponders. We pooled the data from all patients for further analysis. Our operational definition of response was cessation of growth of a previously growing hemangioma or a reduction in size (involution) of the hemangioma coincident with initiation of systemic corticosteroid therapy. Rebound was operationally defined as increased growth of a hemangioma after corticosteroid treatment ceased or was reduced. To allow comparison of results with different corticosteroid regimens, conversion was made to a prednisone equivalent daily dose (in milligrams per kilogram) based on corticosteroid dosing used (relative potency of corticosteroids⁵ : hydrocortisone, 1.0; prednisone, 4.0; prednisolone, 4.0; and methylprednisolone, 5.0) and mean 50th percentile weight based on patient age.⁶ For alternate-day dosing, the dose was divided in half to obtain a "daily dose."

Elementary statistical analysis was then performed using SAS statistical software (SAS Institute Inc, Cary, NC). Univariate and multivariate analyses were performed to assess the relationship between clinical response and corticosteroid dose, patient age, duration of treatment, and whether adverse effects were experienced. Univariate methods included a Satterthwaite t test of unequal variance. For multivariate analysis, we used a generalized estimating equation procedure contained within SAS Proc Genmod.

Review of the literature per the methods described in the previous section revealed 200 related references in 7 languages. Careful review of all titles and abstracts of these references revealed 24 original case series with a minimum of 5 patients having infantile hemangiomas treated with systemic corticosteroids that met inclusion criteria. After eliminating patients meeting the exclusion criteria (see the "Materials and Methods" section), a total of 10 case series representing 184 appropriate patients remained for systematic analysis (Table 1).^{7 - 16}

The largest case series was reported in the Chinese literature¹⁷ and included 411 patients with hemangiomas treated with systemic corticosteroids for 0.5 to 3 years. These patients had to be excluded because the major outcome measure for successful treatment in this study was not whether corticosteroid therapy induced stabilization or involution but rather whether the hemangioma ultimately resolved. Also, the patients were treated significantly longer than those in other studies. The long duration of therapy made it difficult to determine any effect the corticosteroid therapy had on the hemangiomas vs spontaneous involution.

Two other case series, with 10¹⁸ and 31¹⁹ otherwise appropriate patients, were excluded because of inability to determine the dose of corticosteroids used.

The mean age of patients at initiation of systemic corticosteroid administration was 4.5 months (95% confidence interval [CI], 3.8-5.2 months; range, 0.25-24.0 months). Statistical analysis revealed a mean prednisone equivalent daily dose of 2.9 mg/kg (95% CI, 2.7-3.1 mg/kg; range, 1.0-4.5 mg/kg) given over a mean of 1.8 months (95% CI, 1.5-2.2 months; range, 0.5-5.4 months) before tapering, if applicable, was begun. The mean response rate (stabilization or involution coincident with initiation of corticosteroid use) was 84% (95% CI, 78%-89%; range, 60%-100%). The mean rate of rebound was 36% (95% CI, 29%-44%; range, 0%-65%).

In univariate analysis, a t test showed a significant difference between the mean prednisone equivalent daily dose administered to responders (3.0 mg/kg; 95% CI, 2.8-3.3 mg/kg) vs nonresponders (2.1 mg/kg; 95% CI, 1.7-2.6 mg/kg) (P<.001). There was no significant difference in duration of treatment (1.7 months [95% CI, 1.4-2.1 months] vs 2.3 months [95% CI, 1.4-3.3 months]; P = .3) or the occurrence of adverse effects (34% [95% CI, 25%-43%] vs 47% [95% CI, 21%-74%]; P = .3) in responders vs nonresponders, respectively. A t test confirmed a dose-response relationship (P<.001). This relationship is also depicted in tabular format (Table 2).

Adverse effects were reported in 35% (95% CI, 27%-44%) of patients. This figure excludes the study by Pongprasit¹⁴ wherein some patients experienced cushingoid features, infections, or both but exact numbers were not given. No catastrophic adverse effects were seen in any reported patients, although Enjolras et al¹³ indicated serious temporary growth retardation in 4 patients. The most commonly reported adverse effects were behavior changes and irritability, cushingoid appearance, and transient growth delay. Other adverse effects reported included altered appetite and gastrointestinal tract upset, and 1 patient reportedly had osteoporosis, but no information was given as to degree of severity.

Multivariate analysis showed that the prednisone equivalent dose was the only significant factor associated with clinical response (P = .04). Patient age and duration of treatment were not associated with response (P>.5).

Corticosteroids are the treatment of choice for hemangiomas of infancy that interfere with normal function, threaten permanent disfigurement, or are otherwise problematic.²⁰ No consensus exists, however, as to the optimal treatment regimen or expected response rate. The initial dosing of prednisone at 1 to 3 mg/kg per day probably arose from existing treatment regimens for conditions such as nephrotic syndrome and asthma in infants and children.⁶ In more recent years, some clinicians have used higher doses^{14 - 15} with good efficacy, as illustrated by at least one study¹⁵ that demonstrated a greater response rate with 5 mg/kg per day than with 3 mg/kg per day. However, because no controlled trials exist, the optimal treatment regimen is unknown. We performed a quantitative systematic review to provide an evidence-based answer to this question.

A key step in performing a meta-analysis is study selection using inclusion and exclusion criteria. There is a trade-off between generalizability and precision.²¹ We included originally reported patients with cutaneous hemangiomas treated with systemic corticosteroids. Exclusion criteria were chosen to provide generalizable data about cutaneous hemangiomas likely to respond to corticosteroid treatment. Patients with Kasabach-Merritt syndrome were excluded because the syndrome does not occur in true hemangiomas.²² Hepatic and parotid hemangiomas were excluded because of their poorer known response.^{13 ,23 - 27} Laryngeal and subglottic lesions were excluded because they are managed differently than cutaneous lesions, since minor growth might require tracheostomy, laser surgery, or intubation.²⁸ The upper age limit of 2 years and the exclusion of stable or involuting hemangiomas were intended to eliminate lesions that were no longer proliferating.²⁹ Patients simultaneously treated with other modalities, insufficiently documented, or lost to follow-up were also eliminated. Because of feasibility issues, we excluded individual case reports and used case series with at least 5 appropriate patients; only a single series (2 appropriate patients) was eliminated using this criterion.³⁰ The large Chinese case series¹⁷ was eliminated because its end point was resolution, not stabilization or involution, of hemangiomas and patients were treated for much longer periods.

The results of this meta-analysis suggest that most infants with growing cutaneous hemangiomas will respond to systemic corticosteroid administration. A mean prednisone equivalent daily dose of 2.9 mg/kg given over a mean of 1.8 months before tapering had a response of 84%. Lower response rates in some series were due to inclusion of stable lesions^{8 ,12} and lesions other than hemangiomas (ie, venous malformations)^{11 ,14 - 15} or considering rebound growth as failure.¹⁸

The overall response of 84% is less than that reported by some authors and greater than that reported by others. Zarem and Edgerton⁷ and Brown et al⁹ each reported a 100% response in 7 and 9 patients, respectively. Sadan and Wolach¹⁵ used the highest corticosteroid dose (mean, 4.5 mg/kg per day) for a short duration and reported a 93% response. Pongprasit,¹⁴ Fost and Esterly,⁸ and Femenia and Sarinana¹⁰ reported 86%, 83%, and 53% response, respectively, but these increased to 100% with elimination of patients with inappropriate lesions (eg, Kasabach Merritt or Klippel-Trénaunay syndrome, stable lesions, and age >2 years). Sasaki et al¹² had a 39% response that increased to 60% after elimination of patients aged 6 to 46 years (suspect vascular malformations) and those with stable or involuting lesions. Bartoshesky et al¹¹ and Enjolras et al¹³ reported a 30% response. Bartoshesky et al¹¹ considered stable lesions and those treated for "cosmetic reasons" (were not growing) as failures, and all of the patients of Enjolras et al¹³ had "severe" hemangiomas. More than half of the cases were noncutaneous and were thus eliminated by our criteria, leaving a 69% adjusted response.¹³ These issues considered, our calculated response of 84% is probably a good estimate for true cutaneous hemangiomas of infancy treated with standard doses of oral corticosteroids.

The case series of Haik et al¹⁸ and Sharma and Dalal¹⁹ were not included in the meta-analysis because of failure to provide sufficient information to determine prednisone dosing. Haik et al¹⁸ also considered all rebounders as failures. The large excluded Chinese study¹⁷ (different response criteria) reported a 100% response.

The optimal dose of corticosteroid for treatment of hemangiomas has never been studied systematically. In this analysis, administration of prednisone, more than 2 to 3 mg/kg per day, resulted in a 75% response (Table 2) (>3 mg/kg per day resulted in a 94% response but greater adverse effects). Lesser dosing (≤2 mg/kg per day) resulted in fewer responses and adverse effects reported, but rebound occurred in 70% of patients. Randomized controlled studies are needed to confirm the apparent dose-response relationship.

The optimal regimen for tapering is also not known. Most of the case series reviewed herein did not mention if or how tapering was accomplished. In earlier series, corticosteroid administration was sometimes stopped abruptly; in latter series, tapering varied and methods often were not specified.

No life-threatening adverse effects were observed; in fact, almost half of the authors reported no adverse effects. The failure to report major adverse effects is probably an accurate observation, but varying thresholds for monitoring and reporting minor adverse effects and the retrospective nature of the reviews might have contributed to underreporting of minor adverse effects. Two potential but usually occult adverse effects not reported in any of the included patients are hypothalamic-pituitary-adrenal axis suppression and hypertension.³¹ Without specific monitoring, these would usually go unnoticed.

This review demonstrates the need for prospective studies (preferably randomized and controlled) to determine optimal dosing, response, toxic effects, and alternative means of giving corticosteroids (eg, alternate-day therapy, short bursts with rapid taper, and use of dexamethasone). This analysis and our own experience indicate that administration of oral prednisone or prednisolone (2-3 mg/kg per day) is effective in stabilizing or shrinking most growing cutaneous hemangiomas. Treatment should be maintained until cessation of growth or shrinkage of the hemangioma is accomplished. Then the dose can be reduced, but the pace of tapering depends on several factors (ie, the age of the infant, the indication for treatment, any toxic effects, and any rebound growth). The initial corticosteroid dose was maintained a mean duration of 1.8 months in this analysis. Because nearly 40% of patients reported rebound with tapering, brief courses of 2 to 3 weeks' duration are probably inadequate.

Several limitations of this review are evident. Strict exclusion criteria reduced the number of included patients. This decreased the power of the study but improved the generalizability of results. Extracutaneous hemangiomas were excluded; therefore, our results would not be a good predictor of their response rate. Because controlled studies are not available on this topic, studies included were retrospective case series; data were not always complete; treatment regimens varied within series; and studies were not randomized, masked, or controlled. Also, there was no way to eliminate the possibility of spontaneous stabilization or involution (although lesions were not stable at initiation of therapy).

A review of this type is never able to overcome the limitations of the primary studies. Biases such as selection bias are impossible to eliminate. This occurs when patients are not consecutively selected or entry criteria or criteria to assess treatment effect are not clear, rigorously applied, or uniform among studies. Other such problems include failure to address reliability of outcome criteria or to provide full information on dropouts. Publication bias is also impossible to eliminate (includes a tendency to report studies with positive results).

Some assumptions were used to calculate prednisone equivalent daily dose. Weight, for example, was a calculated value based on patient age (50% percentile weight was determined by normogram) when not provided, and the alternate-day dose divided by 2 was used to determine the equivalent daily dose. These assumptions could have resulted in slightly erroneous results. Finally, we recognize the limitations of quantitative systematic reviews: such analyses are not devoid of subjectivity and cannot eliminate all sources of variability among studies or the need for sound reasoning and individualizing care on a case-by-case basis.

In conclusion, corticosteroid treatment is effective for problematic hemangiomas of infancy. The benefit-risk ratio is acceptable, and at least 3 of every 4 appropriately selected and treated patients respond. This meta-analysis provides evidence-based justification for the use of corticosteroids for hemangiomas of infancy.

A cooperative effort of the Clinical Epidemiology Unit of the Istituto Dermopatico dell' Immacolata–Istituto di Ricovero e Cura a Carattere Scientifico (IDI-IRCCS) and the Archives of Dermatology.

Accepted for publication May 22, 2001.

We acknowledge and thank Jean Hu, MD, Jose Catacora, MD, and Dong Fang, MD, for their excellent assistance in translating foreign-language articles.

Corresponding author and reprints: Ilona J. Frieden, MD, Departments of Dermatology and Pediatrics, University of California–San Francisco, Campus Box 0316, San Francisco, CA 94143-0316 (e-mail: ijfrieden@orca.ucsf.edu).