Safety of Oral Propranolol for Infantile Hemangioma

The source was the Système National Inter-Regime de l’Assurance Maladie (SNIIRAM) database,^15,–17 the French national medico-administrative exhaustive database linked to the French hospital discharge database (Programme de Médicalisation des Systèmes d’information), covering of 98.8% of the 66 million people in the French population. Insurance and hospital databases are unambiguously linked (SNIIRAM database) through a single anonymous civil registration. The study protocol was authorized and approved by the ANSM (decision 2014S0121, October 18, 2016) by the National Institutional Review Board (Institut des Données de Santé) and by the Commission Nationale de l’Informatique et des Libertés (French data protection authority). The SNIIRAM database contains anonymous and individual data on demographic characteristics (sex, date of birth, and date of death) and all medical reimbursements, including drugs and their date of delivery and their quantity dispensed by month, laboratory tests, outpatient medical care and consultations, and hospitalizations (public hospital and private clinic), with primary, related, and associated diagnoses based on the International Classification of Diseases, 10th Revision (ICD-10) code. There is no information on the results of clinical examination, the results of laboratory tests and radiology, the diagnoses not linked to a hospitalization, or the drug daily doses. We selected a sample for the reference population, unexposed to Hemangiol, from a fraction of the complete database named Echantillon Généraliste des Bénéficiaires (EGB). This fraction is a 1/97th random and dynamic representative sample of the SNIIRAM database, has the same structure, and contains the same variables as the mother database.

The selection criteria for the patients exposed to Hemangiol included the following: all infants <3 years of age covered by the French National Health Insurance, with at least 1 delivery of Hemangiol between July 28, 2014, and June 30, 2016, and identified through the SNIIRAM database. We focused on pediatric use at the age of IH according to the PUMA and on patients who had received propranolol after the July 28, 2014, date of the availability of Hemangiol in French hospital pharmacies after the PUMA was obtained. The CUP period was not considered and the patients with a switch to Hemangiol propranolol under CUP were excluded (n = 192).

The selection criteria for the unexposed patients (reference population) included the following: all infants <3 years old; covered by the French National Health Insurance; identified through the EGB database; born after January 1, 2010, and before September 30, 2015; and free of any delivery of Hemangiol.

We performed a nationwide, reimbursement-based observational study conducted on preexisting data. Because some underlying diseases could be a risk factor associated with the outcomes screened for or an indication for receiving Hemangiol in patients without IH, we analyzed 2 populations: a population of “healthy” children without any of the prespecified underlying diseases, and a population of “nonhealthy” children (ie, children with 1 of the prespecified underlying diseases). Underlying cardiovascular, respiratory, and metabolic diseases were defined through primary or linked hospitalization or associated diagnosis ICD-10 codes during the study period and are presented in Supplemental Tables 4 through 9.

Included patients had their data collected from birth to September 30, 2016. The date of the first delivery of Hemangiol (propranolol) was considered to be the beginning of exposure.

Hemangiol (propranolol 3.75 mg/mL [vial of 120 mL]) deliveries were identified by using the anatomic, therapeutics, and chemical classification code C07AA05 between July 28, 2014, and June 30, 2016. A Hemangiol course was defined as consecutive deliveries interspaced by <3 months. A new course of Hemangiol was assumed if the time between 2 deliveries was >3 months. Exposure to Hemangiol was considered from the first delivery of Hemangiol course up to 3 months after the last Hemangiol delivery in the same course.

For each distinct, identified risk (cardiovascular, respiratory, and metabolic risks), we defined 3 composite cardiovascular, respiratory, and metabolic outcomes.

The cardiovascular composite outcome was any hospitalization with an ICD-10 diagnostic code from a prespecified list aimed at identifying conduction disorders, bradycardia, hypotension, and cardiac arrest; the respiratory composite outcome was any hospitalization with an ICD-10 diagnostic code from a prespecified list aimed at identifying bronchitis, bronchiolitis, and bronchospasm; and the metabolic composite outcome was any hospitalization with an ICD-10 diagnostic code from a prespecified list aimed at identifying hypoglycemia or related seizure and hyperkalemia. The date of admission was used as the date of the event. ICD-10 codes are presented in Supplemental Tables 4 through 9.

For cerebrovascular outcomes, any hospitalization with an ICD-10 diagnostic code from a prespecified list aimed at identifying cerebrovascular diseases and episodic and paroxysmal cerebrovascular disorders during the exposure period was analyzed separately.

All hospitalizations for cardiovascular, respiratory, or metabolic events were taken into account for the descriptive analysis, and patients could have several hospitalizations linked to the same or different events. In the description of the population, all outcomes occurring under Hemangiol were considered. Premature children (ICD-10 code P07) defined 1 stratum.

In the comparative analysis for each of the cardiovascular, respiratory, and metabolic outcomes, only the first event during the first course of Hemangiol was considered. The numbers of events in the exposed and unexposed reference populations were compared by using indirect standardization. For each event (cardiovascular, respiratory, or metabolic), the ratio between the number of observed events and the number of expected events was calculated within the strata defined by sex, age (3-month periods), and prematurity. “Observed over expected” ratios (standardized morbidity ratios [SMRs]) were calculated with their 95% confidence interval (CI). P values < .05 were considered to be statistically significant. The duration of follow-up is presented in person-months in Supplemental Table 10.

In all, 1934 children <3 years of age had at least 1 delivery of Hemangiol between July 28, 2014, and June 30, 2016. Among these 1934 children, 181 have had a single delivery of Hemangiol (Fig 1). Among the 1753 exposed patients (>1 delivery of Hemangiol; 14 523 person-months), 420 patients (23.9%) had a history of premature birth. Before Hemangiol introduction, 109 (6.2%) patients have had at least 1 delivery of timolol ophthalmic solution, and 284 (16%) have had at least 1 delivery of systemic corticosteroids. No patient had a delivery of vincristine or interferon α 2a. The mean age of patients at Hemangiol introduction was 5.7 months (±5; range: 0–35) and the median age was 4 months (interquartile range [IQR]: 3–7). Most patients received 1 course of Hemangiol, 198 (11%) patients received at least 2 courses, 15 (0.8%) patients received at least 3 courses, and 1 patient (0.06%) received 4 courses. The median duration of first course was 5.4 months (IQR: 3.2–7.5; range: 1–21.4 [Table 1]). The distribution of Hemangiol initiations according to age in months and per calendar month are given in Figs 2 and 3.

Because of the risk of confounding by the presence of an underlying disease (ie, a disease that could influence both the outcome and the probability of prescribing Hemangiol), we isolated a healthy population of 1484 children and conducted a distinct analysis for the same outcomes in a nonhealthy population of 269 children with an underlying cardiovascular, respiratory, or metabolic disease (Fig 1).

The cohort of healthy children exposed to Hemangiol was composed of 1484 individuals, representing a total of 12 320 person-months. SMRs were used to compare incidences of various outcomes within this cohort with incidences in a population of matched healthy children who had never been exposed to Hemangiol.

During the exposure period, 2 healthy children had cardiovascular events. No observed risk for cardiovascular outcomes was identified in healthy children exposed to Hemangiol (SMR = 2.8; 95% CI: 0–6.7; Table 2). One patient had hypotension 1 day after Hemangiol initiation and Hemangiol treatment was continued for 3 months without any further cardiovascular, respiratory, or metabolic events. One patient with a history of premature birth developed bradycardia 38 days after Hemangiol introduction, and Hemangiol treatment was continued for 7 months without any further cardiovascular, respiratory, or metabolic events (Supplemental Table 11).

In all, we observed 51 acute respiratory events, which was more than the number to be expected (SMR = 1.7; 95%CI: 1.2–2.1; Table 2). More specifically, a significant risk for hospitalized bronchiolitis was observed in healthy children (n = 38) treated with Hemangiol (SMR = 1.6; 95% CI: 1.1–2.1). In a sensitivity analysis, we took account of the epidemic period and the results were similar (Supplemental Table 12). No observed risk was observed for bronchospasm and acute bronchitis, but the number of events was much smaller (n = 2 for each of them). Respiratory events occurred after a median time of 67 days (IQR: 40–137) after Hemangiol initiation (Supplemental Table 11).

Only 3 metabolic events (hypoglycemia) were observed in the healthy children exposed to Hemangiol, with no identified related risk (SMR = 5.1; 95% CI: 0–10.9; Table 2). Metabolic events occurred after a median time of 141 days (IQR: 62–212). None had an intercurrent illness during these hospitalizations.

The cohort of nonhealthy children was composed of 269 individuals, representing a total of 2203 person-months (Table 3).

Underlying cardiovascular diseases were identified in 133 children. In this population, 11 cardiovascular events for 1025 person-months were observed during Hemangiol treatment (including 9 conduction disturbances; details given in the Supplemental Information) in a median time after exposure of 104 days (IQR: 53–308), compared with 1.8 expected in the reference population (51 events for 18 563 person-months), giving an SMR of 6 (95% CI: 2.5–9.6; Table 2 and Supplemental Table 10). The risk of conduction disturbances, including blocks, was significantly higher among children exposed to Hemangiol (SMR = 19; 95% CI: 6.6–31.5; Table 2). No observed risk for the cardiovascular outcomes (hypotension and bradycardia) was identified. No patient was hospitalized for a cerebrovascular event.

Underlying respiratory diseases were identified in 49 children. In this population, 11 respiratory events for 366 person-months were observed during Hemangiol treatment in a median time after exposure of 72 days (IQR: 39–141; details given in the Supplemental Information), compared with 1.1 expected in the reference population (86 events for 3186 person-months), giving an SMR of 1.1 (95% CI: 0.5–1.8; Table 2 and Supplemental Table 10).

During the exposure period, we did not observe any patients with an underlying metabolic disease experience a metabolic event.

A population of 181 children with a single delivery of Hemangiol was analyzed separately because the reason for discontinuation was unknown. It could be related to failure to take the treatment or conversely to early intolerance. We observed 3 patients with a cardiovascular event in a mean time after exposure of 49 days (±49) and all of them had an underlying cardiovascular disease. There were 12 patients with a respiratory event (7 patients with bronchiolitis, 4 patients with bronchospasm, and 1 patient with bronchitis) with a mean time after exposure of 47 days (±30), and none of them had an underlying respiratory disease. No child was hospitalized for metabolic or cerebrovascular events. The population with a single issue of Hemangiol is detailed in Supplemental Table 13.

This nationwide claim-based cohort study of 1934 patients treated with Hemangiol (propranolol) is the first analysis conducted at the population level since the PUMA was granted. Compared with the general population, the risks under Hemangiol were contrasted according to the type of outcome and the absence or presence of an underlying disease. In healthy children, we observed no association between Hemangiol and cardiovascular events, but a higher risk for bronchiolitis under Hemangiol was identified. In children with an underlying cardiovascular disease (mainly represented by congenital heart diseases), we observed a higher risk of conduction disturbances under Hemangiol. The number of initiations was stable over the study period from mid-2014 to mid-2016.

Our main objective was to assess the safety among children who were given Hemangiol for their IH. Although IH accounts for the vast majority of indications for oral propranolol, other indications have emerged in recent years. For instance, it is well known that oral propranolol has been given to patients with a tetralogy of Fallot because it was the only pediatric-specific oral propranolol solution available, even before the ANSM issued a recommendation for a temporary use in such patients in February of 2017. Although conduction disturbances are not a direct manifestation of tetralogy of Fallot, the context points to complex interactions between comorbid diseases, side effects of multiple drugs, and the outcomes screened for here. In the absence of information about the precise indication of Hemangiol, a global analysis would have been prone to both confounding by an underlying disease (which could influence the probability of prescribing Hemangiol and the occurrence of outcomes) and to misclassification (Hemangiol given for indications other than IH).

The absence of risk for cardiovascular outcomes in the healthy population is reassuring. However, much concern has been raised about cardiovascular effects and, more precisely, conduction disturbances, bradycardia, and hypotension. In a recent literature review including data on >5000 children treated with Hemangiol for IH, no case of conduction disturbances was reported. Only 2 cases of conduction disturbances were identified in the CUP and the manufacturer’s clinical trials: 1 patient with second-degree Mobitz type 1 atrioventricular block followed by the identification of an underlying cardiovascular disease and 1 patient with type third-degree atrioventricular block occurring during sclerotherapy of esophageal varicose veins due to biliary atresia.¹³ An excess of cardiovascular events (conduction disturbances) under Hemangiol seems restricted to the population with an underlying cardiovascular disease. In our study, in the population of children with underlying cardiovascular diseases, the risk of conduction disturbances was higher in children with an underlying cardiovascular disease and under Hemangiol than in the general population with an underlying cardiovascular disease. The context of cardiovascular disease (tetralogy of Fallot) is likely to have been be the indication for the treatment with Hemangiol. Hemangiol had a recommendation for a temporary use for the treatment of pediatric patients with high blood pressure, heart failure and cardiomyopathy, cardiac rhythm disorders, tetralogy of Fallot, congenital long QT syndrome, and Marfan syndrome, which justify a treatment with β-blockers; however, we cannot exclude the case of patients treated with Hemangiol for IH that also have a cardiovascular disease. In our study, 1 healthy child was hospitalized for a primary diagnosis of bradycardia 38 days after Hemangiol introduction without treatment, discontinuation, or new hospitalization for a cardiovascular event. During the CUP and the manufacturer’s clinical trials, 3 and 2 cases, respectively, of bradycardia were observed; in our recent literature review on the safety of propranolol given for IH, we identified 53 children with bradycardia. This event was considered nonserious, transient, and manageable. In our study, 1 healthy child had a hypotension 1 day after Hemangiol introduction without discontinuation. Among the identified cardiovascular risks, hypotension was the most frequent, including 7 patients in the manufacturer’s clinical trials, 4 patients during the CUP, and 118 patients in the literature; the majority of them were asymptomatic. Lastly, no hospitalization for a stroke was observed under Hemangiol. The question of risk of arterial ischemic stroke under propranolol and in case of PHACES syndrome is still in debate.^11,12 We cannot know whether Hemangiol was given to patients with PHACES syndrome. The researchers of a large, prospective study initiated by the Hemangioma Investigator Group found that among 1096 children with hemangiomas, 25 had extracutaneous anomalies associated with PHACES syndrome.¹⁰ We can hypothesize that among the patients exposed, some of them had PHACES syndrome.

The higher risk of acute respiratory disease, mainly bronchiolitis, in healthy children is worthy of note. This is a well-identified risk in children who are given propranolol because it increases airway resistance. Because this is a well-monitored risk, we cannot exclude the possibility that the higher rate of hospitalization for bronchiolitis in children under Hemangiol is related to better monitoring and parental education on this risk and that hospitalization is more a consequence of better monitoring than an indication of more severe disease. The absence of risk for acute respiratory disease identified in children with an underlying respiratory disease seems paradoxical. However, indications for Hemangiol were carefully weighted in this population, and it is probable that people with more severe diseases are negatively selected.

Regarding the risk of hospitalization for metabolic events, 4 healthy patients had hypoglycemia under Hemangiol, with treatment discontinuation for 3 children. During the CUP and the manufacturer’s clinical trials, 2 and 9 cases, respectively, of hypoglycemia were observed, and in the recent literature review on the safety of propranolol given for IH, 26 children with hypoglycemia were identified, but details on these events are not complete. For the cases in the CUP and the manufacturer’s clinical trials, hypoglycemia occurred during concomitant infection and reduced nutritional intake and could be minimized with appropriate education on the administration of Hemangiol.

Our study has several strengths. First, the study was based on 1 of the largest continuous nationwide health reimbursement databases, with a size comparable to the Sentinel in the United States or the Canadian Network for Observational Drug Effect Studies, providing comprehensive data and covering 98.8% of the French population, which is >66 million patients.^15,17,18 This database provides exhaustive information on the medical prescription of Hemangiol in France, with virtually no selection bias toward the population included and no attrition bias. Secondly, this is the first safety study on Hemangiol in a pharmacoepidemiological approach on a large claims database since the PUMA was obtained in which patients are less closely monitored than in the CUP and the manufacturer’s clinical trials, providing useful information for prescribers. Third, the comparative study included an appropriate control group drawn as a random sample of the general population of children <3 years old from the SNIIRAM database, without any exposure to Hemangiol, and with the same definitions for cardiovascular, respiratory, and metabolic events and underlying diseases.

Some limitations should be discussed. First, cardiovascular, respiratory, metabolic, and cerebrovascular events were based on information encoded in the Programme de Médicalisation des Systèmes d’information, and some misclassifications may have occurred. Second, we had access to the deliveries of drugs, but we cannot ascertain that the treatment was actually taken. This limitation is of special importance for children who had a single issue, and results from this subgroup are presented separately. However, repeated issues can confidently be assumed to reflect actual drug consumption. Third, we had no access to clinical events not linked to hospitalization, and this is a bias toward more severe side effects. Without clinical data from the patients’ health records, it was not possible to analyze precisely the indications for treatment nor to monitor the side effects in an exhaustive manner. Fourth, although we examined the totality of deliveries at a nationwide level, a lack of power is a limitation for rare events.

In this study on a large continuous nationwide database based on health reimbursements, we provide data on the safety and use of Hemangiol after the PUMA was granted. The data confirm the overall good safety profile of oral propranolol in children with IH, in particular for serious side effects with hospitalization, and the data show a stability of prescription levels for oral propranolol in children since the PUMA was granted. Our results do not support a reinforcement of Hemangiol monitoring in infants. It is currently limited to a clinical monitoring, and avoidance is advised in case of lower respiratory infection or low food intake; during the titration phase, a cardiovascular monitoring based on clinical parameters is recommended.

ANSM

Agence Nationale de Sécurité du Médicament

CI

confidence interval

CUP

compassionate use protocol

EGB

Echantillon Généraliste des Bénéficiaires

ICD-10

International Classification of Diseases, 10th Revision

IH

infantile hemangioma

IQR

interquartile range

PHACES

posterior fossa brain malformations, hemangioma, arterial abnormalities of the blood vessels in the neck or head, cardiac abnormalities, eye abnormalities, and sternal anomalies

PUMA

pediatric use marketing authorization

SMR

standardized morbidity ratio

SNIIRAM

Système National Inter-Régime de l’Assurance Maladie