Assessing the efficacy and safety of magnesium sulfate for management of autonomic nervous system dysregulation in Vietnamese children with severe hand foot and mouth disease

This was a randomized, double-blind, placebo-controlled trial of intravenous MgSO₄ versus placebo in Vietnamese children with a clinical diagnosis of Grade 3 HFMD (according to guidelines issued by the Vietnamese MoH [22]) and signs of ANS dysregulation including systemic hypertension. The study was approved by the ethics committees of HTD and the Vietnamese MoH, as well as the Oxford University Tropical Research Ethics Committee, and was carried out in strict compliance with all ICH-GCP guidelines and recommendations. The trial was registered with ClinicalTrials.​gov (NCT01940250), and the full protocol has been published [23].

Essential details of the general study methodology are presented in Additional file 1: Appendix A, and the Vietnamese grading system for HFMD is summarised in Additional file 2: Table S1. In brief, patients aged 6 months to 15 years admitted to PICU at HTD with a clinical diagnosis of Grade 3 HFMD were eligible for enrolment if the BP, measured invasively, was sustained above the cut-off for Stage 1 hypertension (i.e. the 95th centile for age, gender and length [24]) for at least 30 min while the child was not distressed, and the individual exhibited at least one other criterion of ANS dysregulation. Children presenting with Stage 2 hypertension were also eligible for enrolment provided they had no evidence of acute target organ damage, with the proviso that study treatment must commence within 30 min of admission and that milrinone should be added within a further 30 min if there was no improvement – ie the total time to commencing milrinone should be 1 h if the child did not improve (see Additional file 1: Appendix A for further information). An explanation of the Vietnamese guidelines for treatment of severe HFMD was given to the parent/guardian, together with specific information on the proposal to intervene with either MgSO4 or placebo when their child would not usually receive specific therapy for hypertension and details of the systems in place to commence specific treatment with milrinone within 1 h should this be necessary. Following written informed consent by the parent/guardian, study participants were randomly allocated in a 1:1 ratio to receive a loading dose of 50 mg/kg of either 10% MgSO₄ (Fresenius Kabi, Germany) or visually matched placebo. The study drug was given as a continuous infusion over 20 min followed by a maintenance infusion of 30–50 mg/kg/hr. for 72 h according to response, aiming for plasma total Mg levels between 1.8 and 2.5 mmol/l in the treatment arm. Recognizing that having a child in ICU is extremely stressful, detailed information on the study protocol was presented to the family a second time within 12–24 h of enrolment, in order to be sure that they had fully understood the information presented earlier. At this time the option to withdraw from the study was again clearly articulated to the parent/guardian.

All staff involved in clinical care were blind to the treatment allocation, and Mg levels were monitored and adjusted by independent doctors from another clinical facility; these doctors focused particularly on safety, especially given that children in the active arm were receiving an antihypertensive agent at an earlier stage than usual. Detailed clinical assessments were performed daily, with vital signs documented at least hourly for the first 72 h and full biochemical profiles and ECGs performed at least once daily. Catecholamine levels, specifically adrenaline and noradrenaline, were measured on plasma and urine specimens collected daily for 3 days after study enrolment between 8 and 10 am. All clinical and laboratory adverse events were graded following the CTCAE Version 4.03 guidelines [25], modified for children (Additional file 1: Appendix A.7, A.8).

The primary endpoint was a composite of disease progression defined as occurrence of any of the following within 72 h of commencing the study drug: pre-specified BP criteria necessitating addition of milrinone in accordance with MoH guidelines for HFMD with Stage 2 hypertension; ventilation; shock; or death. Secondary outcomes included the time to requirement for milrinone, the area under the curve (AUC) for heart rate (HR), systolic blood pressure (SBP), and mean arterial pressure (MAP) above the Stage 1 hypertension level during the first 72 h, the duration of hospitalization, and neurodevelopmental status assessed 6 months after discharge [26].

Using HTD’s electronic patient database, we identified the hospital files of all children with HFMD who received milrinone between January 2011 and December 2015. From these files we identified cases where the BP remained above the Stage 2 hypertension level despite high dose milrinone (0.6–0.7 μg/kg/minute) but without cardiovascular decompensation at that time (see Additional file 1: Appendix B for details). We wished to compare responses in terms of hemodynamic stability and severe outcomes in patients who fulfilled these criteria and received MgSO₄, with similar cases who achieved the same basic cardiovascular severity level but for whom MgSO₄ was not used. All available data on clinical progress, management and outcomes were extracted from the hospital files to a special case report form (CRF) and entered into an electronic database. The MgSO₄ regimen used in the exposed patient group was nominally the same as that used in the trial. The primary outcome was a composite including development of shock, respiratory compromise requiring ventilation, or death, occurring within the first 72 h from time zero – i.e. the time when MgSO₄ was actually introduced (exposed group), or might theoretically have been introduced based on the individual’s cardiovascular status (control group). Secondary outcomes included the AUC for SBP and MAP above the Stage 1 hypertension level over 24 h from time zero, (since subsequent blood pressure monitoring was less detailed than in the trial), and the duration of hospitalization.

Summary statistics are absolute count and percentage for categorical variables and median (range) for continuous data. For the trial, in view of the small number of patients enrolled we only performed the intention-to-treat analysis. For both the trial and the observational study, between-group comparisons for the primary outcome of disease progression were based on log-binomial regression models (generalized linear models for binomial outcome and using log link function) with adjustment for age and day of illness at study entry/time zero.

For the trial secondary outcomes, the time from commencement of study drug to addition of milrinone was compared using the log-rank test and a Cox proportional hazards model with adjustment for baseline SBP and day of illness at study entry. Patients who never received milrinone were censored at 72 h. Between-group comparisons of the AUCs for the various cardiovascular parameters were based on linear regression models with adjustment for the baseline value of the variable of interest. As AUCs for SBP and MAP were skewed to the right, we log10-transformed these variables before applying regression analysis. For the neurodevelopmental assessments, Z scores for each domain were derived for each individual using data from healthy Vietnamese children matched by age group [26]. We then calculated mean differences in these Z scores between the intervention groups using linear regression and adjusting for age, sex and maternal education level. For comparison of the catecholamine profiles, we used a linear mixed effects regression model that accounted for the evolution of the responses in the two groups.

In the observational study, the group of children who did not receive MgSO₄ were generally hospitalised during the early period of the outbreak, thus forming a historical control group for those who received MgSO₄ later_. For the exposed group, decision-making regarding when to commence MgSO₄ appeared to be influenced by the patient’s age, the rapidity of the deterioration, and the experience of the clinician, in addition to the specific BP criteria outlined above and in Additional file 1: Appendix B. We therefore developed a logistic regression model for initiation of MgSO₄ for each time point at which SBP was measured in the exposed group. We then used this model to impute a time at which MgSO₄ could have been initiated for each set of observations for the control patients. We produced 20 imputed datasets and performed comparisons similar to those described above for the clinical trial on each dataset, finally combining the results using Rubin’s rule [27] and a modified rule that accounted for differences in size between imputed datasets. Detailed information describing these procedures is provided in Additional file 1: Appendix B, Additional file 3: Tables S2, Additional file 4: Table S3 and Additional file 5: Table S4.

Between June 2014 and September 2016, 59 children with Grade 3 HFMD and signs of ANS dysregulation underwent screening for potential inclusion in the trial, of whom 26 (MgSO₄ = 14, Placebo = 12) were enrolled following written consent by a parent or guardian (Fig. 1, Panel a). Compliance with all aspects of the study protocol was excellent, with only a small number of minor protocol deviations. All trial participants received the designated study drug according to the randomization code, and 23/26 cases completed the full 72-h infusion as per protocol. The study drug was unblinded in 3 cases. One child in the placebo group developed emergency hypertension (Additional file 1: Appendix A.3); after unblinding showed that the child was allocated to the placebo arm, open label MgSO₄ was commenced, following which the BP settled without additional therapy. The two other children developed severe respiratory distress requiring ventilation; one child was in the placebo group, with Mg/total calcium (Ca) levels within the normal range, while the other was in the active intervention arm, with plasma Mg and Ca levels of 2.3 and 1.9 mmol/l respectively.

Clinical and laboratory characteristics were similar at baseline between the MgSO₄ and placebo groups (Table 1), except that the proportion of children with Stage 2 hypertension was considerably greater in the placebo group (8/12, 67%) than in the MgSO₄ group (3/14, 21%). Enteroviruses were detected by generic RT-PCR [28] in 18/26 (70%) of the cases overall, which is within the same range as previous reports for clinically diagnosed HFMD [29, 30], with EV-A71 confirmed by specific RT-PCR [31] in 9/14 (64%) MgSO₄ recipients and 4/12 (33%) placebo recipients.

No patient died or developed shock, while 6/12 (50%) cases in the placebo group required milrinone compared to 6/14 (43%) cases in the MgSO₄ group. Two children who received milrinone also required ventilation, one in each treatment group. There was no significant difference in the composite primary outcome between the two groups, adjusted relative risk (95%CI) 0.84 (0.37, 1.92), p = 0.682 (Table 2). There was also no significant difference in the time from study drug initiation to addition of milrinone between the treatment groups (Table 2, Panel A and Fig. 2). Evaluation of the AUCs for the various cardiovascular parameters showed each AUC to be slightly lower in the MgSO₄ group, but there were no statistically significant differences (Table 3).

Plasma catecholamine levels were generally elevated or at the upper end of the expected normal ranges for children at study enrolment (Additional file 6: Figure S1), falling progressively over the 3 days period of observation. However, there were no significant differences in the overall profiles between the study arms (p = 0.609 and p = 0.997, respectively, for plasma adrenaline and noradrenaline levels). Urine catecholamine levels were also increased at enrolment, fell progressively over time, and were generally lower in the MgSO₄ group compared to the placebo group at each timepoint, but with no statistically significant differences overall (p = 0.545 and p = 0.589, respectively, for urine adrenaline and noradrenaline levels).

A total of 21 clinical adverse events (AEs) occurred in 10 patients in the placebo group (83%) compared to 22 events in 10 children in the MgSO₄ group (71%) (Table 4). Serious clinical AEs occurred in 5 patients overall. Respiratory distress developed in 2 patients in the placebo group and 3 in the MgSO₄ group, with one patient in each group requiring ventilation. In no case were the AEs or SAEs associated with Mg levels in excess of 3 mmol/l (maximum recorded value, 2.72 mmol/l), the usual threshold of concern for respiratory or CNS depressant effects of hypermagnesaemia. Laboratory abnormalities were common but were observed with similar frequencies in the two study arms.

All participants were well enough for discharge by study day 14, although 3 children had persisting clinically apparent neurological problems: two with generalized weakness and one with 9th and 11th cranial nerve palsies. However, all three had recovered by the 6-month follow-up visit, at which time all study participants had normal neurological and neurodevelopmental assessments (see Additional file 3: Table S2).

From January 2011 to December 2015, 165 children with severe HFMD were treated with milrinone (Fig. 1, Panel b). Among the 103 children who received high dose milrinone for Stage 2 hypertension, 45 cases where the BP remained poorly controlled were selected for detailed study; 33 received MgSO₄ (the exposed group) while 12 did not (the control group). The baseline characteristics were similar between the two groups, except that the children in the MgSO₄ group were significantly older than the controls and EV-A71 was identified less frequently in the MgSO₄ recipients, (23/33 (70%) compared to 12/12 (100%)) (Table 5). Of note, 11 of the 12 children in the control group were admitted in the early stage of the outbreak when EV-A71 was known to be prevalent, whereas later the pattern of serotypes in circulation became more diverse [32].

Ten deaths occurred within the specified time period, nine in 2011 and one in 2012. Six patients developed severe hypertension and deteriorated very rapidly before reaching the maximum dose of milrinone, and none of the 10 patients fulfilled the selection criteria for this analysis. In the selected study population, no child died or progressed to shock, although three children, two controls and one MgSO₄ recipient, required ventilation. There was no significant difference in the primary outcome between the MgSO₄ and control groups (relative risk (95%CI) of 0.16 (0.01, 1.79), p = 0.132) (Table 2, Panel B). Similarly, we found no difference in the AUCs for SBP and MAP above the Stage 1 hypertension level between the groups (Table 3).

In both patient groups that received MgSO₄, plasma Mg levels increased within a few hours. Almost all trial participants achieved levels in the desired therapeutic range (1.8 to 2.5 mmol/l), within the first 12 h (Fig. 3). However, Mg levels were generally lower in the observational cohort; although the treatment regimen was nominally the same in both groups, the peak infusion rate (median, IQR) achieved for the trial participants was 50 (40, 50) mg/kg/hr. compared to 40 (30, 50) mg/kg/hr. in the observational cohort. A total of 7/130 (5%) Mg levels measured in the MgSO₄ recipients in the trial were in the 2.5–3 mmol/l range, resulting in a recommendation by the independent safety monitoring doctor to reduce the dose in 6 cases. In the observational cohort, 2 Mg levels were documented in the 2.5–3 mmol/l range, plus a single high value of 3.13 mmol/l. However, this child had no respiratory problems, the ECG was normal, and the Mg level dropped promptly to the therapeutic range following dose reduction.

The median Ca levels in the MgSO₄ group decreased from 2.28 mmol/l at baseline to a minimum of 1.85 mmol/l on Day 1 then increased again to 2.13 mmol/l on Day 3, demonstrating a reciprocal relationship with the Mg levels (Fig. 3). In total 32 Ca values, including 20 in the observational study MgSO₄ group, 11 in the trial MgSO₄ group and 1 in the placebo group, were documented below the level of 1.8 mmol/l (equivalent to an ionised Ca²⁺ level of 0.9 mmol/l, i.e. the threshold for potential concern), with the lowest recorded value of 1.55 mmol/l. No clinical AEs linked to low Ca levels occurred and no interventions were required.