Prevalence of Adverse Events in Children With Congenital Hyperinsulinism Treated With Diazoxide

Abstract

Context

Diazoxide, the only U.S. Food and Drug Administration–approved drug to treat hyperinsulinemic hypoglycemia, has been associated with several adverse events, which has raised concerns about the safety of this drug. Existing reports are limited to small studies and case reports.

Objective

To determine prevalence of and clinical factors associated with adverse events in infants and children treated with diazoxide.

Design

Retrospective cohort study of children with hyperinsulinism (HI) treated with diazoxide between 2003 and 2014.

Setting

The Congenital Hyperinsulinism Center at the Children’s Hospital of Philadelphia.

Patients

Children and infants with laboratory-confirmed diagnosis of HI.

Main Outcome Measures

Prevalence of pulmonary hypertension (PH), edema, neutropenia, thrombocytopenia, and hyperuricemia was determined. Tests of association and logistic regression were used to identify potential risk factors.

Results

A total of 295 patients (129 female) met inclusion criteria. The median age at diazoxide initiation was 29 days (interquartile range, 10 to 142 days; n = 226 available start dates); 2.4% of patients were diagnosed with PH after diazoxide initiation. Children with PH (P = 0.003) or edema (P = 0.002) were born at earlier gestational age and more frequently had potential PH risk factors, including respiratory failure and structural heart disease (P < 0.0001 and P = 0.005). Other adverse events included neutropenia (15.6%), thrombocytopenia (4.7%), and hyperuricemia (5.0%).

Conclusion

In this large cohort, PH occurred in infants with underlying risk factors, but no identifiable risk profile emerged for other adverse events. The relatively high prevalence of neutropenia, thrombocytopenia, and hyperuricemia suggests the value in proactively screening for these side effects in children treated with diazoxide.

This study evaluated the prevalence of and clinical factors associated with adverse events in a retrospective cohort of 295 children with HI treated with diazoxide.

Hyperinsulinism (HI) is the most common cause of persistent hypoglycemia in neonates, infants, and children. Genetic defects in 11 different genes encoding proteins important in the regulation of insulin secretion have been identified as associated with the monogenic forms of HI. The incidence of these genetic forms of HI is estimated to range from 1:50,000 in the general population to 1:2500 in populations with high consanguinity rates. Additionally, nongenetic forms of HI associated with perinatal stress are common, particularly among neonates born small for gestational age (1). Multiple studies have reported a high incidence of neurodevelopmental deficits in children with HI (2–6). Importantly, the poor developmental outcomes are not limited to children with the genetic forms of HI but have also been reported in children with transient forms (6). This highlights the importance of a prompt diagnosis and adequate treatment in all cases of HI (7).

Diazoxide is the first-line pharmacologic therapy for most patients with HI and the only therapy approved by the U.S. Food and Drug Administration. Diazoxide is an activator of the ATP-sensitive potassium channels. It suppresses insulin release from the pancreatic β cell by maintaining the plasma membrane as hyperpolarized. It also causes smooth muscle relaxation (8). Diazoxide was first described in the medical field in the 1930s as part of the study of antibacterial effects of sulfonamides. A similarity in the biochemical structure of diazoxide and chlorothiazide was discovered in the 1950s, with subsequent use of diazoxide as an antihypertensive. However, clinical studies were discontinued in the early 1960s after serious hyperglycemic events and edema were reported. In 1964, Drs. Drash and Wolff introduced the concept of diazoxide as a treatment of HI; it remains a key medical treatment of this condition (9).

Although hypertrichosis has been described as the most common side effect of diazoxide, the most clinically concerning effect of diazoxide is edema; this is related to its antidiuretic effect, which leads to a decrease in sodium and water excretion (10). Pulmonary hypertension (PH) has also been described in patients treated with diazoxide (11–13), and in September 2015, the Food and Drug Administration released a statement warning that PH had been reported in 11 cases of newborns and infants treated with diazoxide (14). Additional adverse effects reportedly associated with diazoxide include neutropenia, thrombocytopenia, and hyperuricemia. However, to our knowledge, no large, systematic evaluations of diazoxide’s potential adverse effects have been published.

The primary objectives of this study were to determine the prevalence of diazoxide-related adverse events and to assess associations with clinical factors, comorbidities, and diuretic use. Because no recommendations exist to guide prescribing or screening decisions for children treated with diazoxide, we sought to provide data to make evidence-based care possible.

Participants and Methods

Study population and outcome definitions

This observational cohort consisted of children with HI treated with diazoxide and cared for at the Children’s Hospital of Philadelphia Congenital Hyperinsulinism Center (Philadelphia, PA) between 2003 and 2014. HI was defined by using clinical criteria: having detectable insulin and/or suppressed β-hydroxybutyrate and/or suppressed free fatty acid plasma levels and/or positive glycemic response to glucagon at the time of hypoglycemia (plasma glucose <50 mg/dL) (15). Patients treated with diazoxide for insulinoma were excluded.

Diazoxide is the first-line pharmacologic treatment of HI. At the Children’s Hospital of Philadelphia Congenital Hyperinsulinism Center, a standard approach to diazoxide dosing is used. The initial diazoxide dose is typically 10 to 15 mg/kg/day divided into two oral doses. However, for suspected perinatal stress-induced cases of HI, lower starting doses (5 to 7.5 mg/kg/day) are used. Because the efficacy of diazoxide does not appear to increase for doses >15 mg/kg/day, this maximal dose is not exceeded. In line with previous studies (1, 16, 17), lower initial doses are used for patients suspected of having perinatal stress–induced HI and those with congenital heart disease, with gradual escalation to achieve euglycemia to minimize risk for fluid overload. In addition, a diuretic, typically chlorothiazide, is begun concurrently with diazoxide rather than reactively in cases of fluid retention, a practice supported by the 2017 Japanese Society for Pediatric Endocrinology Clinical Practice Guidelines (18). At our institution, to screen for potential adverse events associated with diazoxide, complete blood count and uric acid levels are requested approximately 2 to 3 months after initiation of diazoxide and every 6 months thereafter.

All outcomes of interest were considered as potentially related to diazoxide course if they were first noted at least 1 day after diazoxide initiation. For patients for whom the diazoxide start date was not available because of incomplete documentation, transfer records were reviewed to ensure each patient had been receiving diazoxide prior to transfer to our institution. Timing of the outcome in relation to diazoxide discontinuation was noted, and outcomes are reported as occurring during use of diazoxide, within 28 days of discontinuing diazoxide, or more than 28 days after discontinuation. Five outcomes of interest were assessed: PH, edema, neutropenia, thrombocytopenia, and hyperuricemia. The medical record was reviewed for history of PH. In cases where PH was documented, available echocardiograms were reviewed by a pediatric cardiologist with expertise in echocardiography and PH (C.M.A.) for confirmation of echocardiographic evidence of PH. PH was defined by at least one echocardiographic criterion: right ventricular pressure > half of the systemic systolic blood pressure estimated from tricuspid regurgitant jet or patent ductus arteriosus (when present) velocity by the modified Bernoulli equation (4 × velocity² + 5 mm Hg), bidirectional or right-to-left patent ductus arteriosus or intracardiac shunt, and/or flat or bowing ventricular septum at end-systole. Edema was defined clinically, as noted in the electronic medical record in the physical examination section of the daily progress notes. Neutropenia was defined as an absolute neutrophil count <1500/μL. Thrombocytopenia was defined as a platelet count <150,000/μL. Because neutropenia and thrombocytopenia may also result from infection, we assessed whether infection was clinically suspected or confirmed at the time of neutropenia or thrombocytopenia. Hyperuricemia was defined as serum uric acid greater than the reference range for age. We reviewed the medical record for cardiac or respiratory factors that could increase risk for PH. These included prematurity, intrauterine growth restriction, pulmonary hypoplasia, respiratory failure, bronchopulmonary dysplasia, genetic syndrome, cardiomyopathy, and structural heart disease (major cyanotic congenital heart disease, left heart obstruction, intracardiac shunt lesions, pulmonary vein stenosis, and other minor acyanotic congenital heart disease).

The Children’s Hospital of Philadelphia institutional review board approved the study. The study qualified for waiver of consent based on 45 CFR 46.116.

Statistical analysis

Cohort characteristics are summarized by using means and SDs for normally distributed continuous variables or median and interquartile range (IQR) for nonnormally distributed continuous variables. Our primary objective was to determine the prevalence of PH, edema, neutropenia, thrombocytopenia, and hyperuricemia in children treated with diazoxide. Our secondary objective was to identify clinical factors associated with the outcomes of interest. Wilcoxon rank-sum tests or Student t tests were used to compare continuous variables. Categorical variables, reported as proportions with 95% CIs, were assessed by using the Fisher exact test or logistic regression. Logistic regression was limited to the study of associated adverse effects (i.e., presence of edema and PH, or neutropenia and thrombocytopenia). Because of the small sample size, multivariable logistic regression was not performed. For all analyses, two-sided P values <0.05 were considered to indicate statistically significant differences. Statistical analyses were performed by using Stata 14 (Stata Corp., College Station, TX).

Results

Cohort characteristics

The cohort consisted of 295 children (129 female; 43.8%) with HI treated with diazoxide. Median age at diazoxide initiation was 29 days (n = 226 available; IQR, 10 to 142 days) and median duration of use was 48 days (IQR, 14 to 207 days) among those who discontinued diazoxide during follow-up (n = 137). Six (2%) patients died during follow-up; three of those were receiving diazoxide at the time of death. Cause of death was known for five of the six patients and included pericardial effusion with underlying congenital disorder of glycosylation (n = 2), multiorgan system failure (n = 1), necrotizing enterocolitis (n = 1), and cerebral edema (n = 1). None had documented PH. Cohort characteristics are listed in Table 1 and HI causes are listed in Table 2.

Table 1.

Cohort Characteristics (n = 295)

Characteristic	Data
Female	129 (43.8)
Median gestational age (IQR) (n = 241), wk	38 (35.7–39)
Premature (<37 wk) (n = 241), %	33.2
Median birthweight (IQR) (n = 226), kg	3.18 (2.35–3.7)
Race/ethnicity
White	130 (44.1)
Black	42 (14.2)
Hispanic	4 (1.4)
Asian/Pacific Islander	1 (0.3)
Other or unknown	118 (40.0)
Median age at HI diagnosis (IQR) (n = 235), da	24 (8–161)
Median age at diazoxide initiation (IQR) (n = 226), d	29 (10–142)
Median duration of diazoxide exposure (IQR) (n = 137), d	48 (14–207)
Cardiac comorbidity present at diazoxide initiation, n (%)	35 (11.9)
Pulmonary comorbidity present at diazoxide initiation, n (%)	28 (9.5)

Unless otherwise noted, values are number (percentage) of patients or median (IQR).

^aExcludes six patients diagnosed with HI within first month of life but missing date of diagnosis.

Table 2.

Distribution of Identified Causes of HI

Cause	Patients, n (%)
KATP-HI (ABCC8 and KCNJ11)	63 (21.3)
HNF1A-HI	7 (2.4)
GCK-HI	5 (1.7)
GLUD1-HI	10 (3.4)
HADH-HI	1 (0.3)
HNF4A-HI	1 (0.3)
Hypomethylation of 6q24	1 (0.3)
Beckwith-Wiedemann syndrome	1 (0.3)
None	116 (39.3)
Unknown	90 (30.5)

Factors associated with PH and edema

Fourteen patients (4.7%) had at least one echocardiogram with evidence of PH during follow-up. Seven of these patients had evidence of PH on echocardiogram at 8 days of life or earlier, before initiation of diazoxide. These patients were excluded from the main analysis because non–diazoxide-mediated PH could not be excluded, but patient characteristics are summarized in Table 3. Of those with PH before diazoxide initiation, PH is known to have persisted after diazoxide discontinuation in three of seven patients (42.9%). Six of 288 eligible patients (2.1%) had echocardiographic evidence of PH after initiation of diazoxide and within 28 days of discontinuation, and one additional subject had evidence of PH 78 days after discontinuation of diazoxide (total n = 7, or 2.4% of eligible patients). Of note, none of these seven subjects had echocardiograms before initiation of diazoxide available for review to confirm that PH was not present prior to diazoxide initiation. Time from diazoxide initiation to first echocardiogram consistent with PH ranged from 2 to 317 days (n = 5); timing was unknown in two cases. Two patients were briefly treated with inhaled nitric oxide. No other pulmonary vasodilators were required. PH resolution was documented in three of seven patients and unknown in the remainder. All three with resolution had discontinued diazoxide before repeat echocardiography.

Table 3.

Characteristics of Patients With PH That Was Diagnosed Before or After Diazoxide Initiation

Patient	Cardiac Anatomy	Respiratory Comorbidities	Age at Diazoxide Initiation (d)	Age at Echo With PH (d)	Diazoxide dose at PH Diagnosis, mg/kg/d	Echo Criteria for PH	Time from Diazoxide to Echo With PH, d	Age at Diazoxide Discontinuation, d	Age at Repeat Echo, d	PH Resolved
Patients with PH on echo after diazoxide initiation
1	Normal	Respiratory failure	6	23	16	RVPE > 1/2 systemic (by TRJV), flattened ventricular septum	17	23	71	Yes
2	Large PDA	Respiratory failure, prematurity (26 wk), severe IUGR	Unknown	65	15	RVPE > 1/2 systemic (by PDA), flattened ventricular septum	Unknown	62	100	Yes
3	Large PDA	Respiratory failure, prematurity (27 wk)	54	83	6.25	Flattened ventricular septum	29	83	88	Yes
4	ALCAPA		4	6	11.5	RVPE > 1/2 systemic (by TRJV), flattened ventricular septum	2	7	NA	Unknown
5	Conoventricular VSD	Respiratory failure, prematurity (30 wk)	19	336	15	flattened ventricular septum	317	Unknown	392	Unknown
6	PDA after ligation	Prematurity (30 wk)	142	145	10	RVPE > 1/2 systemic (by PDA)	3	1607	NA	Unknown
7	PDA after ligation	Prematurity (28 wk)	Unknown	246	10	flattened ventricular septum	unknown	168	NA	Unknown
Patients with PH on echo before diazoxide initiation
8	RCM	Respiratory failure, prematurity (27 wk)	44	6		Flattened ventricular septum	−38	65	128	No
9	Normal	Meconium aspiration, respiratory failure	6	1		–a	−5	239	NA	Unknown
10	CAVC, small PDA	Respiratory failure, influenza A, chylothorax	142	0		Flattened ventricular septum	−142	150	296	No
11	Large PDA	Respiratory failure	8	5		Suprasystemic RVPE (by TRJV)	−3	11	36	Yes
12	Conoventricular VSD, PDA, RLPV/LLPV stenosis, PAPVC– LUPV to innominate vein, isthmus hypoplasiab		24	8		Flattened ventricular septum	−16	37	187	Yes
13	Small PDA	Respiratory failure	1	1		–a	0	31	31	No
14	Pulmonary valve stenosis	OSAc	12	4		Flattened ventricular septum	−8	794	NA	Unknown

Abbreviations: ALCAPA, anomalous left coronary artery from the pulmonary artery; CAVC, complete atrioventricular canal; echo, echocardiography; IUGR, intrauterine growth restriction; LLPV, left lower pulmonary vein; LUPV, left upper pulmonary vein; NA, not available; OSA, obstructive sleep apnea; PAPVC, partial anomalous pulmonary venous connection; PDA, patent ductus arteriosus; RCM, restrictive cardiomyopathy; RLPV, right lower pulmonary vein; RVPE, right ventricular pressure estimate; TRJV, tricuspid regurgitant jet velocity; VSD, ventricular septal defect.

^aEchocardiography performed at outside institution and images unavailable for review.

^bTrisomy 21.

^cBeckwith-Wiedemann syndrome.

Patients with PH occurring while exposed to diazoxide or within 28 days of diazoxide discontinuation were more premature (median gestational age, 30 vs 38 weeks; P = 0.003) and had lower birthweight that was lower but not statistically significantly different (mean, 2.27 vs 3.03 kg; P = 0.1). The presence of at least one potential PH risk factor at the time of diazoxide initiation was more common among patients with PH than those without (100% vs 12.4%; P < 0.0001). Patients did not differ by sex or maximum dose of diazoxide (P > 0.05) (Table 4). Those with PH were significantly more likely to have had edema while receiving diazoxide or within 28 days of discontinuing the drug (OR, 5.6; 95% CI, 1.1 to 28.5; P = 0.04). Median time to chlorothiazide initiation after diazoxide initiation among those without PH was 0 days (IQR, 0 to 7 days; n = 138) vs 10 days (IQR, 3 to 253 days; n = 3) in those with PH (P = 0.03 by Wilcoxon rank-sum test).

Table 4.

Characteristics of Patients With and Without PH, Edema, Neutropenia, Thrombocytopenia, or Hyperuricemia Occurring During Exposure or Within 28 Days of Diazoxide Discontinuation

Condition and Presence During Exposure or Within 28 Days of Discontinuing Diazoxide	Patients (n)	Female, n (%)	Median Gestational Age (IQR) (wk)	Mean Birthweight (SD) (kg)	Presence of PH Risk Factors at Diazoxide Initiation (%)	Median Maximum Diazoxide Dose (IQR) (mg/kg/d)	Median Time to Chlorothiazide (IQR) (d)
PH
Yes	6	1 (16.7)	30 (27–37)	2.27 (1.86) (n = 5)	100.0	15 (12–15) (n = 5)	10 (3–253) (n = 3)
No	282	126 (45.0)	38 (36–39) (n = 228)	3.03 (1.05) (n = 214)	12.4	15 (10–15) (n = 224)	0 (0–7) (n = 139)
P value		0.2	0.003	0.1	<0.0001	0.6	0.03
Edema
Yes	49	19 (39)	37 (34–38)	2.97 (1.20) (n = 44)	30.6	15 (15–15)	3 (0–10) (n = 37)
No	246	110 (45)	38 (36–39.5) (n = 192)	3.01 (1.04) (n = 182)	13.0	15 (10–15) (n = 189)	0 (0–2) (n = 111)
P value		0.5	0.002	0.9	0.005	0.1	0.009
Neutropenia
Yes	42	16 (38.1)	38 (36–39)	2.99 (0.92) (n = 37)	11.9	15 (14–15)	0 (0–4) (n = 25)
No	253	113 (44.7)	38 (35.6–39) (n = 199)	3.01 (1.10) (n = 189)	16.6	15 (10–15) (n = 193)	0 (0–8) (n = 123)
P value		0.5	0.8	0.9	0.3	0.2	0.6
Thrombocytopenia
Yes	11	3 (27.3)	38 (35–39)	2.53 (1.33) (n = 9)	27.3	15 (10–15)	0 (0–8) (n = 6)
No	284	126 (44.4)	38 (36–39) (n = 230)	3.02 (1.06) (n = 217)	15.5	15 (10–15) (n = 224)	0 (0–7) (n = 142)
P value		0.4	0.4	0.18	0.4	0.8	1.0
Hyperuricemia
Yes	15	6 (40.0)	38 (38–39.4)	3.28 (1.17)	0.0	13.8 (11.6–15) (n = 14)	0 (0–3) (n = 6)
No	280	123 (43.9)	38 (35–39) (n = 226)	2.99 (1.07) (n = 211)	16.8	15 (10–15) (n = 221)	0 (0–8) (n = 142)
P value		0.5	0.3	0.6	0.1	1.0	0.3

Fifty-three (18.0%) patients were diagnosed with edema; 42 were diagnosed while they were receiving diazoxide, seven were diagnosed within 28 days of diazoxide discontinuation, and four were diagnosed more than 28 days after discontinuation. Diazoxide was discontinued in 27 of the 42 who were diagnosed with edema while receiving diazoxide. Median time to edema diagnosis after diazoxide initiation was 21 days (IQR, 7 to 150 days) in the 51 patients with diazoxide start dates available. Patients with edema had slightly lower gestational age (median, 37 vs 38 weeks; P = 0.002). Median time to chlorothiazide initiation was shorter among patients without documented edema compared with those with edema [0 (IQR, 0 to 2) vs 3 (IQR, 0 to 10) days; P = 0.009]. The presence of at least one potential PH risk factor at the start of diazoxide was more common among patients with edema (30.6% vs 13.0%; P = 0.005). Patients with and without edema did not differ significantly by sex, maximum dose of diazoxide, or birthweight.

Factors associated with neutropenia, thrombocytopenia, or hyperuricemia

Forty-six (15.6%) patients were diagnosed with neutropenia, 14 (4.7%) with thrombocytopenia, and 15 (5.0%) with hyperuricemia. Of those who developed neutropenia, 29 did so while receiving diazoxide, 13 did so within 28 days of discontinuing diazoxide, and four did so > 28 days after diazoxide discontinuation. Of note, 4 of the 42 (9.5%) patients with neutropenia within 28 days of diazoxide discontinuation had clinically suspected or confirmed infection at the time of neutropenia. Median neutrophil count at the time of first-documented neutropenia was 940/µL (IQR, 600 to 1320 /µL) but ranged from 188 to 1490/µL.

Of those with thrombocytopenia, seven did so while receiving diazoxide, four did so within 28 days of discontinuing diazoxide, and three did so >28 days after diazoxide discontinuation. One patient who developed thrombocytopenia while receiving diazoxide was also receiving divalproex sodium, which also has thrombocytopenia as a potential side effect (19). In addition, two of the 11 (18%) with thrombocytopenia within 28 days of diazoxide discontinuation had suspected or confirmed infection at the time of thrombocytopenia. Median platelet count at the time of first-documented thrombocytopenia was 106,000/µL (IQR, 73,000 to 137,000 /µL) but ranged from 46,000 to 145,000/µL.

In all 15 patients with hyperuricemia, it developed while patients were taking diazoxide, before discontinuation. Patients with neutropenia, thrombocytopenia, or hyperuricemia did not differ significantly from those without in any characteristic or condition compared, including gestational age, birthweight, PH risk factors, sex, diazoxide dose, or diuretic initiation (Table 4). Neutropenia and thrombocytopenia did not tend to cluster together and were not associated with PH (P > 0.05 in univariable logistic regression for each association). Laboratory values for absolute neutrophil count, platelet count, and uric acid level, as well as time to resolution in those who had repeated testing documenting laboratory value normalization, are included in Table 5.

Table 5.

Laboratory Values and Time to Resolution in Patients Who Had Repeated Testing Documenting Laboratory Value Normalization

Variable	Value
Neutropenia
Median absolute neutrophil count when neutropenia noted (IQR); range, cells/μL	940 (600–1320); 188–1490
Patients with repeat test available, n/n	29/42
Median time to resolution (IQR), d	10 (3–40)
Patients in whom diazoxide was discontinued before resolution, n/n	14/29
Thrombocytopenia
Median platelet count when thrombocytopenia noted (IQR); range, cells/μL	106,000 (73,000–137,000); 46,000–145,000
Patients with repeat test available, n/n	10/11
Median time to resolution (IQR), d	4 (2–8)
Patients in whom diazoxide was discontinued before resolution, n/n	3/11
Hyperuricemia
Median serum uric acid level when hyperuricemia noted, median (IQR); range, mg/dL	5.5 (5.1–5.9); 4.8–7.9
Patients with repeat test available, n/n	5/15
Median time to resolution (IQR), d	510 (434–811)
Patients in whom diazoxide was discontinued before resolution, n/n	0

Discussion

In this large retrospective cohort, we determined the prevalence of potential adverse effects of diazoxide among 295 children with HI treated with diazoxide. We found that PH, the most serious potential complication, was the least common among the conditions studied, while neutropenia was the most common laboratory abnormality. Our findings provide the first systematic documentation of adverse effects and associated clinical factors in a diazoxide-treated cohort.

In this cohort, seven patients (2.4%) developed echocardiographic evidence of PH after exposure to diazoxide, although only six cases occurred within 28 days of discontinuation of diazoxide. Patients with PH tended to be premature, and all had at least one potential PH risk factor at the time of diazoxide initiation. They were also more likely to have had edema. A smaller questionnaire-based study of Japanese infants with diazoxide-treated HI revealed high rates of circulatory complications (15 of 78, or 19.2%), but most (13 of 15) complications were edema, similar to the prevalence in our cohort, with the remainder reported as oliguria (4 of 15) or reopening of the ductus arteriosus (3 of 15) (20). Notably, none of the patients in the study by Yoshida et al. (20) had identified PH, and factors associated with the combined endpoint of any circulatory complication (edema, oliguria, reopening of ductus arteriosus) included lower gestational age and higher diazoxide dose. We did not find the same association with diazoxide dose in our study, but this may have been due to the lower variability in doses and maximal dose of 15 mg/kg/day used at our institution.

Patients who developed echocardiographic evidence of PH after diazoxide initiation may have had higher baseline risk for PH given underlying developmental or congenital abnormalities. This is suggested by higher rates of prematurity, respiratory failure, congenital heart disease, and other factors that can be associated with PH. The seven patients with PH documented before initiation of diazoxide had echocardiographic evidence of PH within the first 8 days of life. It is possible that this reflected normal neonatal physiology, with gradual fall in pulmonary artery pressures and pulmonary vascular resistance during the first week of life. As such, these patients were still started on diazoxide after the results of the echocardiography. However, because pre-existing PH could not be excluded, these patients were not included in our analysis of factors associated with PH after diazoxide exposure. Of note, PH is known to have persisted after diazoxide discontinuation in three of seven patients with evidence of early PH.

A clinically relevant finding of our study is that echocardiographic evidence of PH may be associated with delayed diuretic initiation. We found that patients with PH had a longer duration of diazoxide treatment without chlorothiazide use. However, this should be considered in the context of missing data; date of chlorothiazide initiation was unavailable in 50% of those with PH occurring within 28 days of diazoxide discontinuation. Regarding a potential mechanism for diazoxide-induced PH, it is notable that children with Cantu syndrome display clinical characteristics similar to diazoxide adverse effects, including PH, pericardial effusion, and hypertrichosis (21). Cantu syndrome is a complex disorder that results from mutations in ABCC9, which encodes SUR2 and results in K_ATP channel hyperactivity, although not in pancreatic β cells. We hypothesize that off-target effects of diazoxide on SUR2 may contribute to adverse effects of diazoxide, including PH.

Hematologic adverse effects were the most common finding in our cohort. Neutropenia, or an absolute neutrophil count <1500/μL, was noted in 46 patients (15.6%) after initiation of diazoxide. Although the overall incidence of drug-induced neutropenia has been reported to be between one and four cases per million population/year (22), these estimates reflect the general population and are not specific to a medically complex pediatric population. The pathophysiology of diazoxide-associated neutropenia is not well described. However, there does not appear to be increased susceptibility to infection, and neutropenia appears to resolve without discontinuation of diazoxide (10). Because of the retrospective nature of our study design, we were unable to assess causation regarding neutropenia and infection, as infection may also be the cause of neutropenia. However, because the median neutrophil count among patients with neutropenia was just below 1000/µL, this would not be expected to significantly increase risk for severe infection.

Thrombocytopenia is another hematologic condition that has been described as an adverse effect of diazoxide and may require discontinuation of the medication (10). In a systematic review of drug-induced thrombocytopenia in children, diazoxide was classified as probable or level 2, based on the level of evidence for association of the suspected substance and thrombocytopenia (23). In our study, thrombocytopenia (platelet count < 150,000/μL) was the fourth most common adverse event associated with diazoxide use. Notably, thiazide diuretics have also been associated with neutropenia (22) and thrombocytopenia (24). Because most of our patients were receiving diazoxide and chlorothiazide concurrently, we cannot infer a causal relationship between diazoxide and the hematologic adverse effects studied. As noted above, thrombocytopenia may also have been the result of infection rather than directly related to diazoxide. As with neutropenia, the degree of thrombocytopenia was relatively mild and would not be expected to contribute to significantly increased risk for bleeding.

Hyperuricemia secondary to a decreased excretion of uric acid has also been described as a side effect of diazoxide (10). In our cohort, 15 patients (5.0%) had hyperuricemia after diazoxide initiation. Notably, thiazide medications may potentiate hyperuricemia, and hyperuricemia has been found to be present in 44% of adult patients using thiazides (25). As with neutropenia and thrombocytopenia, hyperuricemia in our patients may be secondary not just to diazoxide alone but also to chlorothiazide.

Our study has several strengths. Ours is the largest single-center cohort of patients with congenital HI exposed to diazoxide published to date. In addition, our standard approach to dose determination, diuretic use, and routine screening for the conditions assessed here help to minimize unmeasured confounding and strengthen our ability to draw conclusions despite the retrospective study design. Our findings may reflect higher-end estimates of thrombocytopenia, neutropenia, and hyperuricemia due to the increased ascertainment possible through routine screening. In contrast, it is possible that mild PH without clinical manifestations was missed because of lack of screening echocardiography in all patients. However, any patient with a clinical concern would have undergone diagnostic echocardiography at our institution. In addition, because HI is a rare disorder and a substantial proportion of patients are transferred to our institution, the precise date of diazoxide initiation and dose were missing in some cases. We were also limited by lack of echocardiography reports or images from referring institutions for some patients diagnosed with PH before transfer, but this affected only two patients with reported evidence of PH before diazoxide initiation. In addition, we were unable to reliably evaluate fluid intake and output retrospectively, limiting our ability to strengthen the association of edema and PH. Lastly, not all patients have outpatient follow-up in our HI center after their initial admission, limiting our ability to assess long-term outcomes for all patients. Future case-control studies are recommended to compare the prevalence of the aforementioned adverse events to more precisely determine the impact of diazoxide.

In conclusion, diazoxide is typically a safe, effective therapy for patients with HI, but careful surveillance for more common adverse effects, including neutropenia, thrombocytopenia, and hyperuricemia, is warranted. PH may be more common among premature infants, as well as patients with baseline cardiac or pulmonary disease. To decrease the risk for fluid retention, chlorothiazide or an alternative diuretic should be started with diazoxide in all patients. In patients at greater risk for PH due to baseline comorbidities or birth history, alternative therapies for HI should be considered.

Glossary

Abbreviations:

HI

hyperinsulinism

IQR

interquartile range

PH

pulmonary hypertension