Discussion

The findings of this study utilising data from a national inpatient database demonstrate that isolated diastolic dysfunction is present in 0.5% of paediatric admissions in the neonatal ICU, paediatric ICU, or paediatric cardiac ICU. Inpatient mortality is approximately 5% in such admissions. Several comorbidities are associated with increased inpatient mortality. Several pharmacologic interventions are also associated with increased risk of inpatient mortality although milrinone and angiotensin converting enzyme inhibitors are associated with decreased risk of inpatient mortality. When only admissions with survival to discharge were considered, several factors were associated with increased length of stay while only older age and diuretics were associated with decreased length of stay in survivors.

There are very limited data regarding paediatric diastolic heart failure in general. Much of the data that exist are focused on echocardiographic evaluation of diastolic heart failure. ^{Reference Recher, Botte, Soquet, Baudelet, Godart and Leteurtre6,Reference Dragulescu, Mertens and Friedberg7} Medical management of isolated paediatric diastolic left heart failure is largely anecdotal, based on theory as well as adult data. The issue with this approach is that theory does not always translate into reality. Furthermore, children and adults have physiologic differences which may lead to different effects of interventions. ^{Reference Das5}

What is currently present in the adult literature regarding diastolic heart failure, commonly referred to as heart failure with preserved ejection fraction consists of few large trials. The J-DHF and SENIORS trials demonstrated no benefit of angiotensin converting enzyme inhibitors with all-cause mortality. ^{Reference Yamamoto, Origasa, Hori and Investigators8,Reference Flather, Shibata and Coats9} With respect to angiotensin receptor blockers, the CHARM-preserved and I-PRESERVED trials have demonstrated that these agents are not associated with decreased hospitalisation or mortality in those with heart failure with preserved ejection fraction. ^{Reference Yusuf, Pfeffer and Swedberg10} The PARAGON heart failure trial did not find benefit of a combined angiotensin receptor and neprolysin receptor blocker in the entire study cohort with respect to heart failure hospitalisation or cardiovascular mortality. ^{Reference Solomon, McMurray and Anand11} The TOPCAT trial found no benefit of spironolactone on a composite outcome of heart failure hospitalisation, cardiac arrest, or cardiovascular death. ^{Reference Pitt, Pfeffer and Assmann12} The SENIORS and J-DHF trials did not find benefit of beta-blockers with respect to heart failure hospitalisation, cardiovascular death, or all-cause mortality although significant was noted in a subset of patients with higher dose carvedilol. ^{Reference Yamamoto, Origasa, Hori and Investigators8,Reference Flather, Shibata and Coats9} The trials DILATE-1, SOCRATES-PRESERVED, and RELAX did not find benefit of pulmonary vasodilators in mean pulmonary artery pressure, left atrial volume, n-terminal pro-beta-natriuretic peptide levels, or exercise capacity. ^{Reference Bonderman, Pretsch and Steringer-Mascherbauer13– Reference Hussain, Mohammed and Forfia15} As previously stated, due to difference in children and adults all the way down to the cellular level, this adult data should not directly be applied to the management of children although it may offer a starting point to help design paediatric trials.

No data on treatment of isolated paediatric diastolic left heart failure truly exist. This is for both the inpatient and outpatient settings. A previous study looked at the impact of inpatient vasoactive medications in those with heart failure after cardiac surgery and found that epinephrine, norepinephrine, dopamine, dobutamine, and vasopressin were associated with increased inpatient mortality. Only milrinone was associated with decreased inpatient mortality. ^{Reference Loomba and Flores16}

The current study presents the most robust data regarding inpatient paediatric admissions with isolated diastolic heart failure. This study is able to characterise the cohort in general but also identifies factors associated with inpatient mortality. Such data may be helpful in risk-stratifying patients, particularly those at high risk for mortality. The current study also characterises the effect of pharmacologic interventions on inpatient mortality as well as length of stay in those who survive to discharge.

The current study demonstrates that acute kidney injury, malignancy, and cardiac arrest are associated with increased mortality in those with diastolic left heart failure. Malignancy is likely to be a contributing factor to the development of diastolic left heart failure while acute kidney injury and cardiac arrest are likely to be, at least in part, due to the diastolic heart failure. Previous studies have demonstrated that chemotherapy agents utilised for the treatment of malignancy, specifically anthracycline medications, are associated with the development of both diastolic and systolic left heart failure. Acute kidney injury and cardiac arrest, however, are likely a result of impaired oxygen delivery as a result of diastolic left heart failure. It has been well demonstrated that impaired oxygen delivery, as can be clinically monitored by venous saturations, or near-infrared spectroscopy, is associated with increased risk for acute kidney injury and cardiac arrest. ^{Reference Loomba, Rausa and Sheikholeslami17–Reference Sheikholeslami, Dyson and Villarreal24}

In regard to mechanical interventions, the need for mechanical ventilation and extracorporeal membrane oxygenation is noted to be associated with increased mortality in diastolic left heart failure. These likely represent a cohort of patients with greater severity of illness which may be due to the diastolic left heart failure itself or other comorbidities. The association of need for mechanical ventilation and need for extracorporeal membrane oxygenation with increased mortality is previously described.

With regards to vasoactive medications, epinephrine, norepinephrine, dopamine, and vasopressin were all associated with increased mortality while milrinone was the only agent that was associated with decreased mortality. No significant association of dobutamine with mortality is noted by the current study. The association of most vasoactive medications with increased mortality, even after adjusting for comorbidities and other pharmacologic interventions, likely represents some underlying effect of the vasoactive medications themselves but also likely is also modulated by severity of illness as sicker patients are more likely to require vasoactive medications. This is not the first study to have noted that vasoactive medications may be associated with increased inpatient mortality. This has been demonstrated in both paediatric and adult patients, in those with and without heart failure. ^{Reference Loomba and Flores16,Reference Abraham, Adams and Fonarow25}

Systemic oxygen delivery is the product of oxygen content and cardiac output. Oxygen content can further be broken down into the components of arterial saturation and haemoglobin while cardiac output can further be broken down into the components of preload, afterload, and contractility. Many vasoactive agents increase myocardial oxygen consumption which may come at the expense of systemic oxygen delivery. Many vasoactive agents also have an arterial constrictor effect. Increasing left ventricular afterload may further impact stroke volume and systemic oxygen delivery. It is important to note that simply augmenting mean arterial blood pressure does not imply augmentation of stroke volume and that increased mean arterial blood pressure may be associated with decreased stroke volume if systemic vascular resistance increases a substantial amount. Milrinone can decrease afterload and improve contractility, both of which can help augment cardiac output. Furthermore, milrinone’s leusotropic properties can help improve ventricular filling, which in the setting of diastolic left heart failure may be of obvious benefit. This may explain the beneficial effect milrinone on survival noted in this study. Previous studies have also demonstrated the beneficial effects of milrinone on systemic oxygen delivery and outcomes, specifically in prevention of low cardiac output syndrome after cardiac surgery. ^{Reference Hoffman, Wernovsky and Atz26}

With regards to enteral medications, digoxin and beta-blockers are noted to have a significant association with mortality. Angiotensin converting enzyme inhibitors are noted to be associated with decreased mortality. This may be due to the afterload reduction effects of these agents.

Pulmonary vasodilators are noted to be associated with increased mortality in the current study. This differs from adult studies which demonstrate no change.

With respect to length of stay in survivors, older age and diuretics tended to be associated with decreased length of stay. The association of younger age with increased mortality in paediatric heart failure has been previously described. The association of diuretics with decreased length of stay is likely secondary to symptomatic management and underlying decrease in left ventricular filling pressures associated with diuretics.

While these data are additive to the current literature, specifically for a physiology for which much data do not exist, these data are not without their limitations. First, the definition of paediatric diastolic heart failure is not clearly defined, so its diagnosis and timing likely have physician and institutional variability. Second, the timing of onset of diastolic left heart failure and the comorbidities is not available. Thus, temporal associations cannot be characterised objectively. The temporal associations are discussed elsewhere in this manuscript in the setting of previously published data. Third, the dose and duration of mechanical and pharmacologic interventions cannot be characterised. Thus, dose-dependent effects of medications are not able to be characterised. Fourth, comorbidities are captured by coding. How specific comorbidities are defined may differ at the physician and institution levels. Thus, what one individual or institution defines as acute kidney injury or liver failure, for instance, may be different. Fifth, laboratory data such as N-terminal-pro-brain-natriuretic peptide and serum creatinine are not available in the database. Sixth, imaging data are also not available in the database. Specifically, with respect to the specific aims of this study, echocardiographic data such as tissue Doppler around the mitral valve, mitral valve inflow Doppler, or pulmonary venous Doppler data to characterise left heart diastolic function are not available. Seventh, as with any data that are sourced from multiple centres, individual and institutional-level variability is present. Eight, the overall frequency of use of the combined angiotensin receptor blocker and neprolysin inhibitor combination was too low to allow for regression analysis. Finally, these findings are representative of the average in this cohort, conclusions cannot be drawn for outlying subsets.

To overcome the issue of severity of illness which is exacerbated by the lack of laboratory, haemodynamic, and imaging data this study used various comorbidities that are associated with severity of illness. Examples of this include acute kidney injury, liver failure, and respiratory failure. It is well demonstrated that the number of impacted organ systems correlates well with severity of illness and mortality in critically ill children. Medications also serve as a surrogate marker, specifically vasoactive medications. The vasoinotrope score is based on this very principle and is demonstrated to correlate with severity of illness and mortality. Being able to include a large number of comorbidities and pharmacologic agents in the regression analyses allows for an indirect adjustment for severity of illness. To overcome the issue of institutional variability in clinical practice, centre was entered as an independent variable into the regression analyses to adjust for some of the institutional-level variability.

It must also be noted that the data are specific to outcomes of inpatient admissions. Chronic effects, outpatient symptom effects, and readmission effects cannot be assessed by these data.

On the most general level with regard to any inpatient admission the most general goal is survival. More specific goals for inpatient admissions then become minimising morbidity to maximise quality of life, minimising length of stay for survivors, and optimising systemic and regional oxygen delivery. The nature of the database allowed for modelling survival and length of stay. More granular analyses were not possible from this database. Nonetheless, these data offer objective insight into inpatient management of this unique patient population. These data can be used to help guide acute clinical management in an ICU setting but, more importantly, can be used as pilot data to help with design of future studies.