Relevance of microvascular flow assessments in critically ill neonates and children: A systematic review

Abstract

Objective:

Resolution of impaired microvascular flow may lag the normalization of macrocirculatory parameters. The significance of microcirculatory dysfunction in critically ill children and neonates is unknown, but microcirculatory variables can be measured using doppler or videomicroscopy imaging techniques. We outline the current understanding of the role of the microcirculation in critical illness, review methods for its assessment and perform a systematic review of how it has been monitored in critically ill neonates and children.

Design:

Systematic review (PROSPERO CRD42019117993)

Setting:

N/A

Subjects:

N/A

Interventions:

None

Methods:

We systematically searched MEDLINE, EMBASE, PubMed and Web of Science. We included studies of critically ill patients 0 to 18 years old investigating microcirculatory blood flow. Two reviewers analyzed abstracts and articles. Results were qualitatively analyzed due to study heterogeneity.

Results:

A total of 2559 abstracts met search criteria, of which 94 underwent full text review. Of those, 36 met inclusion criteria. Seven studies investigated microcirculatory changes in critically ill children. Twenty studies investigated the microcirculatory changes in neonates with variable diagnoses compared to a diverse set of clinical end points. Nine studies assessed the effects of age, sex and birth weight on microvascular flow in neonates. Across all studies, microcirculatory dysfunction was associated with poor outcomes and may not correlate with observed macrovascular function.

Conclusions:

Assessment of microvascular flow in critically ill children and neonates is possible, though significant challenges remain. In many such patients, microvascular blood flow is disrupted despite medical management targeting normalized macrovascular parameters. Future studies are needed to define normal pediatric microvascular flow parameters and to assess the impact of patient and treatment factors on its function.

Introduction

The microcirculation is a critically important and potentially underappreciated component of the cardiovascular system. This dynamic vascular segment resides between the arterioles and venules and is comprised of pre-capillary arterioles, capillaries and post-capillary venules(1,2). These vessels are less than 100 μm in diameter, smaller than a human hair. The microcirculation consists of endothelial cells and associated supporting cells. Pre-capillary arterioles are surrounded by smooth muscle cells (SMCs), which act to regulate blood flow to tissue. Capillaries and post-capillary venules are supported by pericytes and less frequent SMCs(3, 4). Capillaries, the smallest of these vessels range between 3 and 6 μm in diameter, are the most numerous and dynamic part of the microcirculation(5, 6). These tiny vessels are responsible for the delivery of nutrients and signaling molecules, removal of waste products, flux of intravascular fluid and heat exchange at the level of individual cells(7, 8). These processes require intimate contact, and consequently a single capillary may be responsible for supporting a layer of only two or three parenchymal cells(9, 10).

All blood flow is dictated by pressure gradients. In the microcirculation more specifically, the difference between pre-capillary hydrostatic pressure (P_cap) and post-capillary venular pressure (mean systemic filling pressure, P_msf) drives flow. Under normal conditions, P_cap is dictated by precapillary SMC sphincter tone, subject to tissue specific autoregulation which is, in part, regulated by capillary pericytes(11). Regulation of P_msf is dictated by venous volume and compliance of the larger venules and veins(12). Normal microcirculatory flow is characterized by homogenous red blood cell (RBC) density and velocity throughout the length of the capillary, with pre-capillary oxygen tension (PO₂) of 90 mmHg and post-capillary PO₂ of 45 mmHg. The transit time for a RBC in capillaries is highly variable across different organs(13).

In times of physiologic stress, the microcirculation adapts to meet increased tissue metabolism by three mechanisms. First, oxygen extraction is increased throughout the length of the capillary. This adaptation is passive and results in decreased central venous oxygen saturation. Second, greater cardiac output augments microvascular flow decreasing RBC transit time and boosting the total number of RBCs crossing the capillary per unit time. Lastly, most tissues have excess capillaries to augment perfusion during times of greater need. The relative perfusion of specific capillary networks is controlled by autonomic nervous system activity (neurogenic), small molecule autocrine signaling (metabolic) and pressure dependent autoregulation (myogenic) of the pre-capillary SMCs and pericytes(3). Microvascular pericytes in particular, are essential to regulating capillary diameter(14), barrier function(15), basement membrane properties(16) and endothelial sprouting(17), all essential functions to maintain adequate flow to support parenchymal cells. Increased capillary recruitment reduces the distance between perfused capillaries and lessens the distance for O₂ to diffuse. These compensatory mechanisms vary by organ. For example, the heart and diaphragm have no recruitable capillaries, whereas this is the predominate adaptive method in skeletal muscle(18).

In times of pathophysiologic stress, such as shock, microvascular flow may be disrupted with a number of deleterious consequences. Although confirmatory evidence is limited, there are several proposed mechanisms by which the microcirculation may become altered(19). Since the microvasculature is responsible for oxygen and nutrient delivery, perturbations in capillary flow result in organ hypoperfusion. In late stages of shock, cellular hypoxia is evidenced by increased markers of perfusion inadequacy, such as increasing serum lactate or organ-specific enzymes. Blockage of capillaries results in an inappropriately heterogeneous microvascular flow (Figure 1, #1), with flow diverted through pathologically dilated capillaries, leading to microvascular shunt(20, 21). Hemodilution, which may result from volume resuscitation, reduces the number of RBCs in each capillary decreasing oxygen content (Figure 1, #2). In the setting of venous congestion, such as occurs in cardiogenic shock, P_msf rises and impairs flow and oxygen delivery (Figure 1, #3). Finally, tissue edema, in the setting of capillary leak, expands the distance that oxygen must diffuse to reach cells, producing cellular dysoxia and dysfunction (Figure 1, #4). Tissue-specific capillaries have different responses to the above pathophysiologic processes. This heterogeneity of function makes it impossible to extrapolate findings from a single capillary to the system as a whole and thus undermines our understanding of global microvascular function(22).

Figure 1:

Microcirculatory blood flow and its pathologic alterations. Normal microvascular blood flow is determined by pre-capillary smooth muscle cell tone. Physiologic flow, illustrated by the block arrows, provides oxygen tension sufficient for cellular homeostasis along the entire capillary, as shown in the top 2 capillaries. Pathophysiologic microvascular flow, resulting in inadequate cellular oxygen delivery, may occur through 4 proposed mechanisms. (1) Capillaries may be blocked by platelet microthrombi, leukocyte plugs, non-deformable RBC or pre-capillary SMC dysfunction. Obstruction produces blockage to flow, resulting in cellular hypoxia distal to the obstruction and increased flow heterogeneity, as blood flow will be redirected to open capillaries, creating tissue-level shunting. (2) Hemodilution as seen after cardiopulmonary bypass or massive volume resuscitation, may result in cellular hypoxia despite appropriate flow. (3) Increased venous pressures, from rising CVP, or post-capillary tamponade from increased tissue hydrostatic pressure. This produces and increase in mean systemic filling pressure (P_msf) with little change in the capillary hydrostatic pressure (P_cap) thereby decreasing driving pressure for capillary flow (P_cap- P_msf). This results in rapid decrease is O₂ tension and cellular hypoxia. (4) Tissue edema, from capillary leak or massive volume resuscitation, increases the distance for O₂ to diffuse resulting in cellular hypoxia despite adequate flow. Multiple problems may be present in a single capillary network and their effects may overlap, exacerbating organ dysfunction.

Monitoring of microvascular function is challenging since the common macrovascular variables followed (e.g. blood pressure (BP), heart rate (HR), central venous pressure (CVP) and cardiac index) may have limited relevance. Hemodynamic coherence is a term that describes the extent to which normal macrovascular function correlates with appropriate microvascular function(23). Multiple studies in adults have demonstrated that hemodynamic coherence may be lost in critically ill humans(23, 24). Furthermore, persistent microvascular dysfunction has been associated with organ dysfunction and death in adults(25). Nonetheless, goal directed resuscitation of children in shock typically targets macrovascular hemodynamic goals, assuming that hemodynamic coherence is intact. Yet, specific treatment strategies, such as fluid resuscitation or vasopressor therapy, may improve macrovascular variables (such as CVP) to the detriment of microvascular flow (by decreasing P_cap-P_msf)(26).

Microcirculatory function can be directly assessed in several ways. Videomicroscopy permits the observation of the number of RBCs and their velocity through individual capillaries (Table 1). These devices utilize dark-field microscopy where polarized light at 550 nm is preferentially scattered by the hemoglobin in RBCs. The scattered light is captured for analysis and produces short video clips of illuminated capillary networks on dark backgrounds. Recent advances in optics have resulted in more precise instruments, from orthogonal polarized spectral (OPS), side-stream dark field (SDF), and incident dark field (IDF) imaging. These devices, notably IDF, are state of the art for current research. Other technologies include laser doppler flowmetry (LDF) which measures the relative velocity of RBC flow through sections of tissue from the doppler shift of laser focused light (700 to 1000 nM). Finally, intravital microscopy and capillaroscopy utilize light microscopy to visualize flow velocity through capillary loops, although in humans, this technology is usually limited to the nailbed.

Table 1:

Types of microvascular imaging devices.

Assessment technique and reference	Application and site of assessment	Variables measured	Limitations to use	Examples of commercial available devices
Videomicroscopy: OPS (27), SDF (28), IDF (29)	Direct viewing of buccal or sublingual microcirculation	TVD, FCD, PPV, PSPV, MFI, MHI	Image acquisition affected by operator skill Image analysis performed offline	MicroScan CapiScope CytoCam
Doppler: LDF (30), LDPM (31)	Microcirculatory functional integrity assessment of superficial tissue	Relative flow, Microvascular Hgb content, Microvascular integrity	Does not differentiate between microvascular segments (i.e. arteriole, capillary and venule)	Blood FlowMeter Vein Finders
Microscopy: IVM (32)	Nailbed microscopy	Capillary density	Nailfold area only, unsuitable for critically ill patients	Capiscope

OPS: Orthogonal polarization spectral imaging, SDF: Side-stream dark field imaging, IDF: Incident dark field imaging, LDF: Laser doppler flowmetry, LDPM: Laser doppler perfusion measurement, IVM: Intravital microscopy, TVD: Total vessel density, FCD: Functional capillary density, PPV: Proportion of perfused vessels, PSPV: Proportion of small perfused vessels, MFI: microcirculatory flow index, MHI: microcirculatory heterogeneity index, Hgb: Hemoglobin

Since the ultimate aim of resuscitation is to restore adequate cellular oxygen delivery, monitoring the microcirculation may provide the most pertinent measures (Table 2)(24). However, microcirculatory function is difficult to assess in critically ill children. A major challenge is clinical feasibility. OPS requires an external light source and is prone to blurring, blockage or difficulty positioning the device. SDF performs better but still requires manual intervention and scoring systems. Both of these methods are sensitive to motion artifacts. Some of the devices are quite bulky, and all require training and expertise to obtain and interpret measurements. Newer devices are smaller and more intuitive with the promise of automated data analysis that may be more amenable to pediatric studies. A more significant concern is the sampling location. Monitoring oral microcirculation may not provide information on perfusion of central, more essential organs(33). Obstructed or heterogeneous flow may be present in one organ and not in another. Another major impediment is the lack of normal pediatric values, complicating interpretation of image analysis in children and infants. Indirect assessments, such as near infrared spectroscopy (NIRS) or arterio-venous PCO₂ differences, can provide insight into the global function of the microvasculature relative to metabolism.

Table 2:

Parameters that can be measured to describe the microcirculation. In general, the parameters targeted are assessments of vessel recruitment (FCD, PVD, PPV) or the character of flow through those vessels (MFI, HI, MFV). Not all devices can measure all parameters, and no pediatric normal values have been established.

Microcirculatory variables	Units	Significance
Vessel diameter	μm	Small <10μm, Medium 10μm to 20μm, Large >20μm
Total vessel density (TVD)	mm/mm2 (1/mm)	Measure of all vessels in a field of view. Total length of vessels (small, medium or large) divided by total area of the field of view.
Perfused vessel density (PVD)	mm/mm2 (1/mm)	Proportion of perfused vessels multiplied by total vessel density.
Proportion of perfused vessels (PPV)	% vessels perfused	Percentage of perfused vessels in a given image quadrant obtained. Calculated by total length of perfused small vessels divided by total length of small vessels.
Microvascular flow index (MFI)	Continuous (normal), sluggish (slow), intermittent or absent (no flow)	Qualitative assessment of flow over quadrants. An average of images obtained.
Heterogeneity index (HI)	Arbitrary unit	Measure of flow heterogeneity. Maximum MFI quadrant minus Minimum MFI quadrant divided by mean MFI.
De Backer score	Vessels/mm	Alternate for total vessel density, using standard grid format with 3 equally spaced vertical and horizontal lines. Calculated as number of vessels crossing grid lines divided by the total length of the lines.
SO2	% hemoglobin saturation	Tissue oxygenation for a specific organ studied, i.e. kidneys.
Microvascular flow velocity (MFV)	μm/sec	Can differentiate speeds between small, medium and large vessels

FCD: functional capillary density, PVD: Perfused vessel density, PPV: proportion of perfused vessels, MFI: microcirculatory flow index, HI: microcirculatory heterogeneity index, MFV: microvascular flow velocity.

A better understanding of microcirculatory responses may result in more effective treatments for common conditions such as shock. The extent to which microvascular flow and hemodynamic coherence are disrupted by critical illness in children is much less well defined than in adults. The purpose of this review it to gain a deeper understanding of this issue. Therefore, we performed a systematic review to assess the current state of research into microvascular dysfunction in critically ill children and neonates.

Methods

Search Strategy.

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses(34). Our study protocol is registered in the PROSPERO database of systematic reviews, CRD42019117993 (Supplementary Table 1). We began our search with the Yale MeSH Analyzer(35), using key articles to refine the search strategy for the concepts critical illness, microcirculation, and children (Supplementary Table 2). In each database, we performed search queries and used an iterative process to translate and refine the searches. All searches were limited to the English language. Searches in Embase and MEDLINE were limited using the human filter. Additional articles were identified by examining other systematic reviews, reference lists, bibliographies, and pre-identified websites such as NIH reporter, conference abstracts (Web of Science), and publicly available internet searches (Google Scholar).

On December 6, 2018, the authors searched PubMed (conception through December 06, 2018), MEDLINE (Ovid MEDLINE ALL 1946 to December 06, 2018), Embase (Ovid Embase 1974 to 2018 December 07), and Web of Science (conception through December 06, 2018). On March 14, 2019, a second search was completed in Ovid MEDLINE ALL, Ovid Embase, PubMed, and Web of Science Core Collection. The search repeated the controlled vocabulary terms and free text terms. Manuscripts identified from these searches were de-duplicated in EndNote X8 (Clarivate Analytics) and uploaded to Covidence(36).

Study selection.

Two independent reviewers (LM, EN) completed title, abstract, and full text screens (Figure 2). The senior author (RP) mediated consensus meetings to resolve discrepancies. Studies were screened for meeting the inclusion criteria: (1) Observational case report, cohort, case control, or clinical trial studies; (2) Studies involving critically ill pediatric patients; and (3) Studies assessing microcirculatory flow. Since a continuum of disease states exist under the definition of critical illness, there is not one single condition that satisfies our query. Our search strategy (Supplemental Table 2) includes many terms associated with critical illness. For including or excluding studies in ambiguous cases, we relied on the primary authors definition or assessment of the study participants’ “critically ill status.” Studies investigating animal or human cellular models as well as those that did not specifically assess microcirculatory flow were excluded.

Figure 2:

Study flow diagram and selection of eligible articles.

Assessment of the evidence and data abstraction.

All collected studies were reviewed using the QUIPS tool to assess risk of bias in prognostic factors studies from the Cochrane Review Group(37). A quantitative meta-analysis was not performed for several reasons. Relative estimates of effect could not be calculated for case series and studies that lacked control groups. Methodological heterogeneity, such as the type of assessment, the timing of assessment relative to disease onset, and the wide range of patient and treatment factors further prohibited quantification of results. Therefore, all studies were qualitatively analyzed.

Results

A total of 4178 citations were retrieved, pooled and de-duplicated to 2559. Of these, 94 articles met requirements for full text review, after which 36 studies were included in the investigation for evaluation of microcirculatory flow: 7 assessing critically ill children and 29 involving neonates.

Overall, the quality of the evidence for using microvascular assessment to predict outcomes is low. Of the 27 studies included, 8 rated as low, 18 as medium and 1 as high quality (Supplemental Table 3). One limitation of feasibility and generalizability was study design, as only a single randomized controlled trial involving 12 total neonates was identified(55). All other studies were observational trials or case reports. Studies assessed had low numbers of subjects, and most lacked control groups. Only 20 of the 36 studies included control patients, and 28 of the 36 studies involved 40 or less subjects. In all cases, primary data could not be completely reviewed. High subject dropout rate, lack of standardized of normal values, timing and method of assessment, parameters measured and comparators further downgraded the quality of the available evidence.

We identified 7 studies of microcirculation in critically ill children (Table 3). Of these, three investigated microcirculatory flow in sepsis and found decreased MFI and reduced functional capillary density (FCD) and proportion of perfused vessels (PPV), using OPS or SDF with sublingual or buccal mucosa sampling(38–40, 43). These variables correlated with increased inflammatory markers and improved over time with the patient condition. Two studies investigated the effects of cardiopulmonary bypass (CPB) on microcirculatory flow. CPB appeared to minimally alter baseline microvascular parameters during the procedure(42), but did negatively affect induced hyperemic response, while measuring microvascular vasodilation with cutaneous vascular conductance, intraoperatively(44). A single study investigated children after cardiac arrest and demonstrated decreased flow and perfused vessel density (PVD), which improved as patient condition recovered(41). Finally, a study using SDF sublingual sampling and investigating all-comers to the PICU found decreased MFI, PPV and PVD that correlated with low BP and mixed venous saturation and improved after macrocirculatory normalization(32).

Table 3:

Studies of microvascular function in critically ill children.

Study	Design	Critically ill patients	Control patients	Method of assessment	Site of assessment	Quality Assessment	Major findings
Top, et al, 2011(38)	POC	15 Survivors 3 Non-survivors	0	OPS	Buccal mucosa	Medium	Children who died from septic shock had initially increased then decreased FCD (3.2 cm/cm3, 0.8-3.8, to 1.9 cm/cm3, 1.0-2.1) compared to controls (1.7 cm/cm3, 0.8-3.4, to 4.3 cm/cm3, 2.1-6.9). FCD and MFI remained low despite correction of heart rate and blood pressure.
Paize, et al 2012(39)	POC	20	40	SDF	Sublingual mucosa	Medium	Children with severe meningococcal disease have significant reduction in MFI, PPV, & PVD that correlated with elevated markers of endothelial origin and predicted ventilatory support requirements. MFI and PPV normalized pre-extubation.
da Luz Caixeta, et al, 2013(40)	Case report	2	0	SDF	Sublingual mucosa	Low	Children with Dengue shock were found to have low capillary flow velocity, PVD, PPV and MFI. Patient 2 had increased PPV on day 2 but low flow resulting in persistent tissue hypoperfusion.
Buijs, et al, 2014(41)	POC	11 Survivors 9 Non-survivors	20	SDF	Buccal mucosa	Medium	Children receiving hypothermia after cardiac arrest had decreased PVD NS, PPV NS, MFI NS, and MFI S that were associated with mortality risk. Perfusion improves after re-warming. Improvement in microcirculatory parameters coincided with improvements in arterial lactate and O2 tension.
Scolletta, et al, 2016(42)	POC	17 Acyanotic 7 Cyanotic	0	SDF	Sublingual mucosa	Medium	Children undergoing congenital heart surgery with cardiopulmonary bypass had generally stable TVD, PVD, and PPV for acyanotic lesions and lower TVD, PVD and PPV for cyanotic lesions. There was weak inverse correlation between red blood cell storage time and MFI.
Gonzalez, et al, 2017(43)	POC	18	0	SDF	Sublingual mucosa	Medium	In children admitted to the intensive care unit for any reason, MFI, HI, PPV, & PVD correlate positively with systolic arterial pressure, arterial O2 tension, temperature and central venous saturation and negatively with central venous pressure and serum lactate.
Ugenti, et al, 2018(44)	Cross sectional observational study	61 Cyanotic 39 Acyanotic	0	LDPM	Forehead skin	Low	Microvascular skin flow is similar in patients with cyanotic and acyanotic heart disease. Local thermal hyperemia response was blunted while patients were on cardiopulmonary bypass.

POC: prospective observational cohort study, OPS: Orthogonal polarization spectral imaging, MFI: microvascular flow index, FCD: functional capillary density, SDF: Side-stream dark field imaging, CD: capillary density, PPV: proportion of perfused vessels, PVD: perfused vessel density, NS: Non-small vessels (Ø 11–100 μm), S: Small vessels (Ø ≤ 10 μm), HI: heterogeneity index, LDPM: laser Doppler perfusion monitoring.

See supplemental table 3 for details of quality assessment.

We identified, 29 studies involving neonates, with 20 studies relating microcirculatory flow to critical illness (Table 4) and 9 studies relating microcirculatory flow to gender or post-natal age (Supplementary Table 4). The 20 studies investigating microcirculatory changes in neonates included 300 unique patients of which 94 were full term, 82 were preterm and 124 had no gestational age reported. Included studies related flow to a variety of etiologies of critical illness, including acute respiratory distress syndrome (ARDS), sepsis and hypoxic ischemic encephalopathy (HIE, 4 studies each), cardiopulmonary bypass (3 studies) and congenital diaphragmatic hernia, anemia, hypotension and patent ductus arteriosus (a single study each).

Table 4:

Studies of microvascular function in critically ill neonates.

Study	Design	Critically ill patients	Control patients	Method of assessment	Site of assessment	Quality assessment	Major findings
Pöschl, et al, 1994(45)	POC	12	21 Healthy 7 Non-septic	LDF	Back, thigh, heel skin	Medium	Infants with sepsis have reactive hyperemia compared to controls and changes occur before C-reactive protein or leukocyte count increase.
Martin, et al., 2001(46)	POC	12	20	LDF	Dorsal hand skin	Low	Infants with sepsis have reactive hyperemia (by post-occlusive perfusion measurement) that correlates with increased IL-6, IL-8 and TNF compared to controls.
Genzel-Boroviczény, et al., 2004(47)	POC	13	0	OPS	Upper arm skin	Low	Blood transfusion improved FVD in critically ill preterm anemic neonates but did not affect vessel diameter, RBC velocity or flow.
Top, et al., 2009(48)	POC	14	10	OPS	Buccal mucosa	Medium	FCD is significantly lower in neonates with ARDS before ECMO compared with control patients. FCD increased significantly after ECMO decannulation.
Weidlich, et al., 2009(49)	POC	17	4	OPS	Upper inner arm skin	Medium	FSVD declined in infants with infection before lab findings.
Hiedl, et al., 2010(50)	POC	13	12	SDF	Inner upper arm skin	Medium	FVD was lower in infants with hemodynamically significant PDA but same as control group after PDA closure.
Top, et al., 2011(51)	POC	8	0	OPS	Buccal mucosa	Low	Inhaled nitric oxide improves FCD in the buccal mucosa of infants with hypoxemic ARDS.
Top, et al, 2012(52)	POC	21	7	OPS	Buccal mucosa	Medium	FCD increased slightly but MFI and HI did not change after initiation of veno-arterial ECMO in neonates with cardiopulmonary failure in comparison to ventilated controls.
Ergenekon, et al., 2013(53)	POC	7	7	SDF	Axilla skin	Medium	Infants treated with TH had lower MFI and decreased flow velocity that normalized with re-warming.
Schwepcke, et al., 2013(54)	POC	10	11	SDF	Right arm skin	Low	Within the first 6 hours of life, hypotensive infants have higher FVD. FVD did not respond to medical management and normalized by 12 hours after birth.
Buijs, et al., 2014(55)	POC	21	7	SDF	Buccal mucosa skin	Low	Infants with CDH requiring dopamine had lower PPV and MFI. All microcirculatory variables in hypotensive neonates were lower than normal newborn controls. Inotropes corrected macrovascular but not microvascular indices.
Dehaes, et al., 2014(56)	POC	10	17	FDNIRS, DCS	Left, middle, and right frontal regions	High	Infants with HIE undergoing TH had a lower cerebral blood flow in comparison to control infants or post-TH.
Ishiguro, et al., 2014(57)	Case report	1	0	LDF	Forehead and foot skin	Low	A neonate in septic shock had decreased skin blood flow and MAP despite inotropes but flow and MAP did increase with blood transfusion.
Nussbaum, et al., 2015(58)	POC	40	15	SDF	Ear conch skin	Medium	Infants undergoing cardiopulmonary bypass had decreased MFI, PVD and vessel glycocalyx width. Values were similar to control infants within 24 hours.
Fredly, et al., 2016(59)	POC	28	0	LDPM, CAVM, DRS	Chest skin	Medium	Neonates with HIE and high C-reactive protein have lower FVD compared to infants with HIE and low C-reactive protein. Flow and heterogeneity were not significantly different.
Fredly, et al., 2016(60)	POC	28	25	LDPM, CAVM, DRS	Chest skin	Medium	Infants with HIE undergoing TH had a lower capillary flow velocity and increased FCD, HI and O2 extraction. Values were similar to control infants after re-warming.
Neunhoeffer, et al., 2016(61)	POC	20	30	LDF, tissue spectrometry	Placed above both kidneys	Medium	Infants undergoing cardiopulmonary bypass that developed AKI had decreased renal oxygen saturation and higher renal blood flow and renal resistive indices.
Troiani, et al., 2017(62)	POC	12 ARDS 7 Cardiac	28	LDF	Back of forearm skin	Medium	Infants with ARDS had decreased skin microcirculatory flow velocity and reserve compared to infants with cardiac abnormalities or those without ARDS.
Puchwein-Schwepcke, et al, 2018(63)	RCT	6	6	SDF	Inner right arm skin	Medium	Infants randomized to high CO2 tension group had reduced FVD and increased medium/large versus small vessel distribution.
Neunhoeffer, et al., 2018(64)	POC	14 HLHS 14 TGA	0	LDF, tissue spectrometry	Right forehead skin	Medium	Relative cerebral blood flow did not change in infants before and after heart surgery requiring cardiopulmonary bypass despite differences in cerebral oxygen extraction.

POC: prospective observational cohort study, LDF: Laser doppler flowmetry, TNF: Tumor necrosis factor, OPS: Orthogonal polarization spectral imaging, FVD: Function vessel density, SDF: side-stream dark-field imaging, PDA: Patent ductus arteriosus, FCD: Functional capillary density, PARDS: Pediatric acute respiratory distress syndrome, ECMO: Extra-corporeal membranous oxygenation, FSVD: Functional small vessel density, MFI: microvascular flow index, HI: heterogeneity index, CDH: Congenital diaphragmatic hernia, FDNIRS: Frequency-domain near infrared spectroscopy, DCS: Diffuse correlation spectroscopy, HIE: hypoxic ischemic encephalopathy, TH: therapeutic hypothermia, LDPM: Laser doppler perfusion measurement, CAVM: computer-assisted video microscopy, DRS: Diffuse reflectance spectroscopy, AKI: Acute kidney injury, CHD: congenital heart disease, RCT: randomized clinical trial, HLHS: Hypoplastic left heart syndrome, TGA: Transposition of the great arteries.

See supplemental table 3 for details of quality assessment.

Neonates with sepsis demonstrated a decreased hyperemia response(45, 46), skin flow(57) and functional vessel density(49). Reactive hyperemia response refers to skin perfusion time, measuring velocity of red blood cells with laser doppler, pre and post-arterial occlusion. Vessels in neonates with sepsis took longer to establish perfusion in the venules post-occlusive stimulus as opposed to healthy newborns. Similarly, neonates with ARDS demonstrated decreased functional vessel density (FVD)(48, 51, 63) as evidenced by OPS and SDF imaging on skin or mucosal surfaces. Reduced flow and vessel density parameters were detected in neonates with congenital heart disease(50) and those undergoing CPB(58, 61), although the cerebral circulation was not affected allowing oxygen delivery to the brain to be preserved(64). Neonates undergoing therapeutic hypothermia for HIE demonstrated decreased flow parameters that improved with rewarming(53, 56, 60). Additionally, decreased vessel density was associated with a rise in systemic inflammatory markers(59). Among a general population of critically ill neonates, FVD improved after blood transfusion(47). However, neither flow nor vessel density was improved upon initiation of veno-arterial ECMO(52). Neonates who are hypotensive immediately after birth have altered vessel density that does not respond to resuscitation but improves within 12 hours(54). Finally, neonates with congenital diaphragmatic hernia had altered flow and vessel density that did not respond to inotrope therapy(55).

The 9 studies relating microcirculatory flow to gender or post-natal age are not discussed because they did not separate cohorts based on diagnosis or disease severity, rendering analysis on those parameters impossible(65–73).

Discussion

The vascular system is responsible for the delivery of O₂ and nutrients to cells, a task so critical it has been dubbed “the organ of the intensivist”(74). Despite this significance, our review has revealed a paucity of studies on microcirculatory blood flow in critically ill children, with only 502 patients assessed across 27 studies. The settings of these studies and disease severity of their participants, demonstrate that assessment of microcirculatory function in critically ill children and neonates may be difficult but is possible. Analysis of the collected literature is challenging due to great variability in the patient populations, diagnoses, treatment factors and assessment timepoints. Despite these limitations, several important conclusions may be drawn on the nature and implications of microcirculatory changes in severely ill children and neonates.

The prevalence of microcirculatory dysfunction in children remains unknown. In adults, more than 15% of intensive care patients will have dysfunctional microcirculatory flow, and that its presence correlates with mortality(75, 76). Based on the studies we reviewed, microcirculatory irregularities appear to be more prevalent among critically ill pediatric patients despite a broad range of ages and conditions. Gonzalez and colleagues conducted the most comprehensive study of microvascular dysfunction in general PICUs(43). However, definitive conclusions from this study are difficult to draw due to low enrollment of only 18 eligible children, including the exclusion of multiple critically ill as well as generally well patients. The exact prevalence of microvascular dysfunction and its consequences is not known because studies lacked control groups, as well as the lack of pediatric normative data for microcirculatory variables and low PICU mortality rates(77). More worrisome, multiple studies demonstrated that microcirculatory dysfunction correlated with biochemical markers of tissue perfusion inadequacy, such as lactate and interleukins, but not with macrovascular parameters, implying loss of hemodynamic coherence(41, 43, 78). Across similar disease states, several studies of demonstrate more significant microcirculatory dysfunction in children and neonates compared to adults(38, 43), suggesting this may be a more significant problem in younger patients.

The effects of common therapies on microcirculatory dysfunction are more clear. Six studies in children and neonates demonstrated that interventions to normalize macrovascular parameters did not correct microvascular flow in the setting of inotropic therapy(38, 57), therapeutic hypothermia(41, 53) and ECMO(52). These findings are especially concerning since patients are considered for such therapies due to inadequate DO_2, yet their application may not correct the underlying microvascular defects. Such differences are likely explained by variations in the device used and timing and location of assessments. In instances of rapid clinical decline, it is difficult to tease out the effects of multiple simultaneous interventions. Many of the studies in neonates used LDF instead of videomicroscopy, making direct comparisons between patient populations impossible, as in the comparison of inotrope use and blood transfusion with sampling via LDF versus SDF (47,49). The variable parameters assessed and outcome measures render more general conclusions untenable. These results are especially significant given the finding of the recent ANDROMEDA-SHOCK trial showing improved mortality in peripheral perfusion rather than lactate directed resuscitation in adults with septic shock(79). The collected literature indicate that microvascular dysfunction is common and likely underappreciated in the NICU and PICU and may have significant clinical consequences and treatment implications.

Few studies provide some insight into the effects of various therapies on the microcirculation. In neonates with respiratory distress syndrome, the application of inhaled nitric oxide improved perfused vessel density(51). Well known for its pulmonary vasodilatory effects, nitric oxide has multiple extra-pulmonary actions(80). In the only randomized control trial identified, permissive hypercapnia, a therapeutic approach commonly used in neonatal RDS, was shown to decrease vessel density(63). Similarly, neonates with congenital diaphragmatic hernia, treated with dopamine to correct HR and BP, continued to have reduced flow and vessel density(55). These findings have important clinical relevance since inotropic infusions commonly applied to treat shock may, in some instances, compromise end organ tissue perfusion. Defining how the microcirculation changes respond to common therapies and the ultimate impact on tissue perfusion and outcomes is an urgent research priority.

There are limitations to this investigation. As we have noted, there are multiple methods to assess the microcirculation that are not directly comparable. Standardization of techniques and measurements, as well an agreement on age-specific normal values, are essential to advancing our understanding of microvascular function. Without standard normal ranges for specific ages, body sizes and genders, the work done to assess the microcirculatory flow at the capillary level is not generalizable to the neonatal and pediatric ICU settings. More trials with greater subject numbers are needed to validate a method, timing and location of assessment. The limited number of studies, low quality of evidence and diagnostic and methodologic heterogeneity precluded a quantitative meta-analysis. Our review may further be limited by the potential impact of publication bias, as studies not finding a correlation between critical illness and disruption of microvascular flow are less likely to be published. Although we attempted to identify all relevant studies, it is possible that qualifying studies were inadvertently omitted from this systematic review. Despite these challenges, we felt it was important to try to understand the range of clinical settings in which microvascular monitoring is feasible and potentially informative. In fact, technology has sufficiently advanced to allow clinical researchers to begin to explore how patient and treatment factors alter the microcirculation of our patients. Multiple ongoing studies aim to address some of these concerns.

Conclusion

Adequate microvascular blood flow is essential for the maintenance of cells and organ function. Homeostasis relies on hemodynamic coherence, where microcirculatory flow is mirrored by macrovascular parameters. Despite logistic challenges, microvascular flow may be measured with laser Doppler flowmetry or videomicroscopy. In critically ill children, microvascular blood flow may be disrupted, and hemodynamic coherence may be lost. The timing, severity and clinical consequences of these changes are not well defined. More research is required to better understand how microvascular variables can be monitored in critically ill children, define age appropriate normal values, assess responses to common interventions and ultimately, define its impact on PICU morbidity and mortality.