Myocardial and haemodynamic responses to two fluid regimens in African children with severe malnutrition and hypovolaemic shock (AFRIM study)

The study was conducted between May 2013 and August 2014 at the Kilifi District Hospital (KDH), Kenya and the Mbale Regional Referral Hospital (MRRH), Uganda. Eligible patients included children aged 6–60 months with clinical signs of severe malnutrition (defined as any one of the following: mid-upper arm circumference (MUAC) <11.5 cm, weight-for-height Z score (WHZ) below –3 or oedema indicative of kwashiorkor plus acute hypovolaemic diarrhoea (>3 watery stools/24 h) with signs of severe dehydration (sunken eyes and/or decreased skin turgor) and shock. We defined shock as two or more signs of capillary refill time ≥3 seconds, temperature gradient (cooler extremities to warmer central body to touch) or rapid and weak pulse volume, instead of the WHO stringent shock definition (which includes all four features of impaired perfusion plus impaired consciousness) [1], which identifies very few children with extremely poor prognosis [8]. We excluded children with severe dermatitis of the groin (in whom catheterization for urine output monitoring was not possible), children with congenital heart disease and children whose guardians did not consent to study participation.

Written informed consent was obtained from the children’s guardians once the patients had been stabilised following initial verbal assent at the start of resuscitation [9]. Eligible and consenting participants were recruited sequentially. The bolus group (group 1) followed WHO standard protocol, which recommends a bolus of 15 ml/kg of RL over 1 h, with the option of repeating it once (15 ml/kg) if signs of shock persist [1], followed by intravenous half-strength Darrow’s/5% dextrose (HSD/5D) given at a rate of 4 ml/kg/h.

Following a pre-planned interim analysis of the clinical data after 10 patients were enrolled (which did not include detailed review of the echocardiographic and haemodynamic data) the study investigators were very concerned about the very high mortality rate. In light of the original findings and emerging published data from the FEAST trial [6, 10], the research team, concerned that the bolus fluid challenges could be leading to compromised myocardial and haemodynamic responses, opted to prospectively study a further 10 patients who would receive rehydration volume replacement without an initial fluid bolus. The rehydration-only group (group 2) received 10 ml/kg/h of RL over 5 h.

In both groups children were switched to oral rehydration (rehydration solution for malnutrition, ReSoMal) once they were able to tolerate oral intake or nasogastric fluids. Blood transfusion, if indicated by WHO guidelines (haemoglobin (Hb) <5 g/dl or shock unresponsive to crystalloid volume replacement), was administered as 10 ml/kg over 3 h once blood for transfusion was available. All the patients received routine standard treatments for severe malnutrition, as per the WHO guidelines, including antibiotics and therapeutic formula feeding once able to tolerate oral feeds [1].

Patients underwent the standard admission procedures including clinical examination and routine blood sampling. Blood was stored for retrospective analysis of biomarkers of myocardial injury (cardiac troponin I (cTnI)) or congestive heart failure (brain natriuretic peptide (BNP)) based on published literature [11‐15]. At each clinical review routine vital-status and physical examination parameters were systematically recorded and assessed for key endpoints (shock reversal, adverse events related to fluid overload including evidence of pulmonary oedema (bi-basal crepitations and worsening oxygen saturations), raised jugular venous pressure, gallop rhythm and increasing hepatomegaly). Electrocardiogram (ECG) and echocardiogram (ECHO) were used to assess cardiac function both pre and post resuscitation (immediately and at 24 h) and at one-month follow up.

All patients were studied using a Vivid.i portable ultrasound device (General Electric Medical Systems ®, USA) with simultaneous ECG display. A rapid haemodynamic assessment was conducted at admission and a comprehensive ECHO was performed when the patient was stabilised. Three echocardiographic recordings and readings were obtained during three consecutive cardiac cycles and an average of these was used to evaluate the baseline status and the response to fluid therapy. A standardized protocol was used for ECHO measurement of left ventricular (LV) size and function, which was in accordance with the guidelines published by the American Society of Echocardiography, updated in 2014 [16]. Additional file 1: Box 1 presents the details of the calculations.

A standard twelve-lead ECG (25 mm/sec) was recorded in all patients at admission, 24 h post resuscitation and during follow up (Cardiac Acquisition Module, CAM-14, GE Medical Systems and CardioSoft ™ software). Collected data included: duration of P wave, PR interval, duration of QRS complex, RR interval, PP interval and QT interval. Corrected QT interval (QTc) was calculated using Bazett’s formula [17, 18].

Data were analysed using STATA 12 software (STATA ® Corp, United States of America). We report descriptive statistics including median and interquartile ranges for pooled measures of cardiac function, haemodynamic response and blood parameters with comparison between groups 1 and 2. Cut-off values for normality were based on previously published data [19‐22]. In addition, we summarised the haemodynamic profiles of the children (heart rate, systolic and diastolic blood pressure) across the time point (0–48 h) using age-adjusted medians stratified by group and survival status. Kaplan-Meier analyses were used for comparison of time to mortality between the two groups. The exact 95% confidence intervals were evaluated using the log-rank test for equality of survivor functions. The significance level was 0.05 for all statistical tests conducted. No multivariate analyses were conducted owing to the small sample size.

A total of 174 severely malnourished patients were screened during the study period (Fig. 1). Of these, 107 did not have any signs of shock, 37 had signs of shock that was not secondary to gastroenteritis (septic shock), 5 had severe dermatitis in the groin, 1 had a congenital cardiac anomaly, 1 refused consent, 2 had already received fluid boluses and 1 presented outside the study daily recruitment time. A total of 20 patients with hypovolaemic shock and gastroenteritis were prospectively recruited, including 11 in group 1 (bolus followed by rehydration), and 9 in group 2 (rehydration-only fluid).

The overall median age in the two groups was 21.5 months (interquartile range 17, 25.5) with no difference between the cohorts (Table 1). More patients with oedematous or kwashiorkor phenotype (10/11; 91%) were enrolled in group 1 compared to group 2 (5/9; 56%). This was possibly due to the seasonal nature of this complication. Baseline clinical and haemodynamic parameters were similar in the two groups. Overall, respiratory signs were very prevalent including tachypnoea, respiratory distress and hypoxia (oxygen saturations <90% in room air on pulse oximetry). As expected features of shock, dehydration and impaired consciousness were common, but only 35% of patients (n = 7) fulfilled the strict WHO criteria for shock. Besides hypoglycaemia (<3 mmol/L), which was more common in group 2, there were no significant differences between the groups at baseline (Table 1). The large proportion of patients with malaria parasitaemia reflects the background endemic malaria in Eastern Uganda. Haemoglobin levels were low; but only two patients had severe anaemia (Hb <5 g/dl).

Owing to the very limited validation of the non-specific WHO clinical definitions of heart failure (see Table 2) we interpreted these incorporating clinical, echocardiographic and cardiac biomarkers definitions and applied the criteria retrospectively. At baseline they indicated that only 3/20 children (15%) fulfilled the WHO clinical criteria of putative congestive heart failure. However, one patient presented with pneumonic consolidation and two had respiratory distress (Kussmauls breathing) secondary to severe acidosis. None of the children had a gallop rhythm or raised jugular venous pressure (JVP).

In group 1, 9/11 patients (82%) received two boluses of 15 ml/kg/h RL to correct clinical signs of shock; 4 children (36%) received an additional 10 ml/kg blood transfusion for persistent clinical signs of shock after receiving the RL boluses (in accordance with the current WHO management guideline). One patient presenting with profound anaemia (Hb <2.2 g/dl) received only an initial intravenous RL bolus prior to receiving a blood transfusion. One patient went on to receive oral rehydration following improvement after a single RL bolus at 15 ml/kg/h.

In group 2, 3/9 patients (33%) received 10 ml/kg/h RL rehydration over a total of 5 h (i.e. 50 ml/kg), followed by blood transfusion due to persisting signs of shock (as per the WHO protocol). Four patients (44%) substantially improved after 3 h of intravenous rehydration (30 ml/kg) and were able to tolerate oral ReSoMal rehydration. Two patients died whilst receiving rehydration (at 2 h and 4 h after starting fluid therapy).

Tachypnoea, respiratory distress and the proportion of children with hypoxia did not worsen after fluid therapy in either of the groups. Although hepatomegaly was common at admission there was no evidence of increasing organomegaly complicating fluid resuscitation in any patient. Nor did we find evidence of our pre-specified criteria of fluid overload (Additional file 2: Table S1a and b). Median age-adjusted heart rate and blood pressure over time are presented in Additional file 3: Figure S1a-f. Severe hyponatraemia, present at admission in four patients in group 1 and five patients in group 2 (Table 1), changed little over time, with survivors to day 28 having persisting hyponatraemia at follow up (Additional file 4: Table S2 and Additional file 5: Figure S2). Similarly, median potassium was very low, particularly in group 1 at admission (2.4 mmol/L, IQR = 2.1–3.3 mmol/L) with little improvement over time. High serum creatinine (>80 μmol/L) was present in 4/11 patients (36%) and 3/9 patients (33%) in groups 1 and 2, respectively, which improved gradually over time. Reassuringly, we noted satisfactory urine output (>1 ml/kg/h) in 8/11 patients in group 1 and 7/9 patients in group 2 (Table 3).

Tables 4 and 3 present median (IQR) and percentage changes in the key cardiac and vascular performance indices and Additional file 6: Figure S3 presents box and whisker plots of echocardiography over time. On admission, median left ventricular fractional shortening (FS) in group 1 was within the normal range (28–44%) in 7/11 patients (64%), and improved modestly after fluid administration (Table 4). FS on admission was within the normal range in the same proportion of the children (6/9, 66%) in group 2 but changed little in the median FS after rehydration. In both groups the interquartile range of FS had reduced by 24 h compared to admission.

As shown in Table 3 in group 1 the stroke volume index (SVI) increased by >10% after fluid bolus in 5/11 patients (suggesting fluid responsiveness) but decreased by >10% in two patients and remained unchanged (<5% change) after the fluid challenge in 4 patients. In group 2 the SVI after rehydration (no fluid bolus) was more homogenous, with no children having a decrease in SVI: in five patients there was >10% increase, and it remained unchanged in two patients. Two children died before post-rehydration measurements. Figure 2 provides a summary of the SVI by group over time.

In order to evaluate possible echocardiographic signs of compromised heart function we related the fluid responsiveness (assessed by SVI) to a measure of compliance (assessed by the end-diastolic volume index (EDVI)). These are represented as Frank-Starling plots of SVI against EDVI in Fig. 3a and b. In four patients in group 1 the EDVI increased by >5%, which was associated with an increase in SVI (indicating fluid responsiveness) in three patients. However, in a fourth child (who survived) this was associated with a decrease in SVI (indicating non-fluid responsiveness). There was a very heterogeneous pattern of response in the other seven children, including either no change in EDVI after fluid administration or a decrease that was associated with an unchanged, increased or decreased SVI.

In group 2, EDVI increased in five patients after rehydration and this was associated with an increase in SVI in all patients, suggesting an “appropriate” myocardiac response to fluid therapy. SVI increased concomitant with a decrease in EDVI in two patients (one of these patients died). Two other patients died while undergoing initial rehydration prior to calculation of post-rehydration EDVI and SVI.

Myocardial deformation/strain analysis was performed in the radial, circumferential and longitudinal axes (Fig. 4). Overall, the percentage global radial strain (GRS %) in these children with severe malnutrition and hypovolaemic shock was significantly lower than published references (mean GRS 26.2%, SE 4.1, 95% CI 17.4–33.8%; p < 0.05) [23]. Pericardial effusion was present in 11 children (55%), all of whom had kwashiorkor. Notably, the percentage global radial strain (GRS %) decreased after bolus fluid in group 1 but increased after rehydration in group 2. However, this difference between the groups after receiving fluid was not statistically significant (F = 0.33; p =0.15) (Table 4).

At baseline, the inferior vena cava collapsibility index (IVCCI), a surrogate marker of intravascular volume status [24‐26], had a wide range of values across the two groups, but this narrowed after fluid administration, particularly in group 2. The interpretation of this is of questionable relevance, especially when many children had Kussmauls’ (acidotic) breathing. We noted that the median SVRI was generally high across the groups and despite initial reductions after fluid administration as expected, it rose again at 24 h and remained persistently high in the small number of survivors who were clinically well when reviewed at follow up on day 28.

With regard to cardiac biomarkers, the median brain natriuretic peptide (BNP) at admission was elevated (>300 pg/ml) in 6/10 patients (60%) in group 1 and 2/9 patients (22%) in group 2 (Additional file 7: Table S3). BNP remained elevated at 48 h in 4/7 patients (57%) in group 1 but not in any patients in group 2 (excluding two patients who died). Median troponin I was within the normal range in both groups at all the time points sampled. Electrophysiological recordings did not reveal any gross abnormalities or arrhythmias in any of the patients (Table 5).

By 48 h, 5/11 patients (46%) and 3/9 patients (33%) in groups 1 and 2, respectively, had died. Overall 28-day mortality was higher in group 1 (9/11 patients (81.8%)) than in group 2 (5/9 patients (55.6%)) but the difference was not statistically significant (relative risk 0.902, 95% CI 0.313, 2.599; p = 0.8484) (Additional file 8: Figure S4). Median age-adjusted heart rate and blood pressure over time were similar in the survivors and fatalities with the exception of median diastolic blood pressure, which was lower in patients in group 2 who died before 4 h, and was consistent with uncorrected hypovolaemia in the early deaths (Fig. 5a-c).

In group 1 the median IVCCI worsened (increasing from 29% to 38%) post fluid administration in the non-survivors, while in the survivors the IVCCI decreased as expected (from a mean of 34% to 30%) post fluid resuscitation. In group 2, the median IVCCI reduced from 39% to 24% in the survivors, suggesting improved intravascular filling, and this observation was supported by a concomitant median increase in SVI from 30 to 32 ml/m², whereas there was no clear trend in IVCCI in the non-survivors. As noted previously bradycardia at admission [8], present in 10% of patients (2/20) and low systolic blood pressure and persistent weak pulse after fluid administration in 55% of patients (11/20) were associated with early death, the majority occurring within 48 h of study admission.

We found no evidence of fluid overload adverse events in the study participants. This included children fulfilling the non-specific WHO criteria for CCF. That is, we found no evidence of increasing hepatomegaly or pulmonary oedema after fluid administration. None of the seven patients classified as having shock according to WHO guidelines at admission survived. Patients in whom BNP at 48 h remained persistently high (>300 pg/ml) died. HIV co-morbidity contributed to two deaths.

Fatal events can be classified into three time periods: early (0–24 h), intermediate (24–48 hour) and late (48 h to 15 days). In group 1 (receiving the WHO-recommended bolus, followed by maintenance) there were four early deaths (mainly within <8 h), which were all due to cardiovascular collapse secondary to hypovolaemia as a result of persistent high output (vomiting and stool). One child died at 48 h due to cardiovascular collapse (uncorrected acidosis/sepsis). The late deaths (n = 4) included two children who aspirated their nutritional feeds, and two children who died in the community. In group 2 (rehydration only) there were early deaths in three children who had cardiovascular collapse due to persistent hypovolaemic shock and another child had a fatal respiratory arrest at 20 h, which was associated with severe acidosis and hypoxaemia. Two late deaths occurred at day 4 (respiratory arrest associated with severe metabolic complications) and one child died in the community after 6 days. All fatal events were judged as unrelated to fluid challenges or the intravenous fluid regime, and largely were secondary to underlying co-morbidities (Additional file 9: Table S4 includes detailed clinical narratives and haemodynamic and echocardiographic findings).