Feasibility of optical coherence tomography angiography to assess changes in retinal microcirculation in ovine haemorrhagic shock

Blood flow in the microcirculation (vessels smaller than 100 μm) plays an important role in the delivery of oxygen and nutrients to cells. Maintaining adequate blood flow to the microcirculation is a prerequisite for normal organ perfusion and function. In critically ill patients, capillary blood flow is often impaired and changes in capillary perfusion are associated with outcome [1‐4].

In acute haemorrhagic shock, excessive blood loss results in depressed cardiac output, which leads to decreased tissue perfusion with tissue hypoxia, organ dysfunction, and finally death. However, routine clinical parameters, which often serve as surrogates for monitoring of tissue perfusion in daily clinical practice (e.g. cardiac output, arterial blood lactate, heart rate, or urine output), do not always adequately reflect the level of tissue perfusion, and optimizing these parameters does not necessarily ensure adequate tissue perfusion [5‐9]. In consequence, there is a need to introduce direct parameters of tissue perfusion (i.e. microcirculatory parameters) to support therapy decisions and to guide therapy in the future [5].

Using various novel techniques, it is possible to monitor patients’ microcirculation directly at the bedside. Research using these new techniques has demonstrated the important role of the microcirculation in critical diseases [3, 4]. The microcirculation in critically ill patients has been evaluated in the splanchnic region, skin, muscles, and the sublingual area [3]. In contrast, to our knowledge only a few studies evaluated changes in microvascular perfusion of the retina in critically ill patients, for example in sepsis [10].

The retina is an embryological projection of the forebrain. Impaired retinal perfusion might indicate reduced cerebral blood flow, since the ophthalmic artery arises from the internal carotid artery. Previous studies have shown a close relationship between cerebral events and qualitative measurements of retinal microvascular abnormalities [11‐15].

Optical coherence tomography angiography (OCT-A) is a relatively novel technology, which provides high-resolution images of the retinal and choroidal vascularization. This method is non-invasive, reproducible, and can be quickly performed at the bedside. In consequence, this technology has attracted increasing interest over the last 2 years. It has been described in healthy subjects, in a range of ocular and systemic diseases in humans, and in various animal models [16‐24].

The purpose of the current study is to evaluate the feasibility of OCT-A to assess the changes in retinal microcirculation in shock and fluid resuscitation in an experimental model of haemorrhagic shock in sheep.

At the end of the third step of blood withdrawal (haemorrhagic shock), a median 27 (24; 30) ml·kg^− 1 blood was shed from the animals (animal 1, 27 ml·kg^− 1; animal 2, 30 ml·kg^− 1; animal 3, 30 ml·kg^− 1; animal 4, 21 ml·kg^− 1; animal 5, 27 ml·kg^− 1). The MAP decreased significantly following blood loss (baseline 117 (108; 122) mmHg vs haemorrhagic shock 33 (30; 38) mmHg; P = 0.001) and increased almost to baseline levels after resuscitation (93 (76; 100) mmHg vs haemorrhagic shock; P = 0.164). The CI tended to decrease in haemorrhagic shock (1.1 (0.8; 1.2) L∙min^− 1∙m^− 2 vs baseline 2.5 (2.2; 2.7) L∙min^− 1∙m^− 2; P = 0.051) and increased significantly after resuscitation (3.0 (2.7; 3.8) L∙min^− 1∙m^− 2 vs haemorrhagic shock; P = 0.001; Fig. 2).

Further haemodynamic variables at baseline, during progressive haemorrhage, and after resuscitation are summarized in Table 1.

The central venous oxygen saturation decreased significantly at the time of haemorrhagic shock compared to baseline (baseline 86.5% (84.0; 88.5) vs haemorrhagic shock 38.8% (29.4; 46.1); P = 0.003) and increased to baseline levels after resuscitation (resuscitation 85.2% (79.7; 89.2) vs haemorrhagic shock; P = 0.027). Urine output and parameters of blood gas analyses at baseline, during progressive haemorrhage, and after resuscitation are summarized in Table 2.

With OCT-A imaging it was possible to visualize the optic disc and the retinal vasculature with very good image quality (Additional file 1). It was also possible to perform imaging of the same retinal area for several hours throughout the experimental procedure (Fig. 1). Figure 1 exemplifies the retinal vessels at baseline, during haemorrhagic shock, and after resuscitation. Flow density data of the same retinal area in a 3 × 3 mm² scan were also evaluated throughout the experimental procedure. Flow density_WF in the superficial OCT angiogram of the retina decreased significantly after shock induction (baseline 44.7% (40.3; 50.5) vs haemorrhagic shock 34.5 (32.8; 40.4) %; P = 0.027) and recovered to baseline values after resuscitation (resuscitation 46.9% (41.7; 50.7) vs haemorrhagic shock; P = 0.027; Table 1 and Fig. 1).

The conjunctival microcirculation results are summarized in Table 3. The PPV decreased significantly at haemorrhagic shock compared to baseline values (baseline 100.0% (98.0; 100.0) vs haemorrhagic shock 72.0% (57.4; 76.3); P = 0.013) and increased substantially after resuscitation (resuscitation 98.7% (97.3; 99.1); P = 0.173 vs haemorrhagic shock). The MFI decreased in haemorrhagic shock (baseline 3.1 (3.0; 3.3) vs haemorrhagic shock 1.9 (1.9; 2.1); P = 0.034) and tended to increase after resuscitation (3.0 (3.0; 3.3) vs haemorrhagic shock; P = 0.081). Representative videos of the conjunctival microcirculation at baseline, in haemorrhagic shock, and after resuscitation are available as supplemental digital content (Additional files 2, 3 and 4).

Flow density_WF in the superficial retinal OCT angiogram correlated significantly with parameters of the conjunctival microcirculation (PVD: Spearman’s rank correlation coefficient ρ = 0.750, P = 0.001; PPV: Spearman’s rank correlation coefficient ρ = 0.620, P = 0.014; MFI: Spearman’s rank correlation coefficient ρ = 0.679, P = 0.005; see Fig. 3) and haemodynamic parameters (CI: Spearman’s rank correlation coefficient ρ = 0.693, P < 0.001; MAP: Spearman’s rank correlation coefficient ρ = 0.594, P = 0.002; SVI: Spearman’s rank correlation coefficient ρ = 0.639, P = 0.002).

The present study shows for the first time that OCT-A has potential as a novel technology for non-invasive and contactless monitoring of the retinal microcirculation during progressive haemorrhagic shock and resuscitation. Flow density_WF measured by OCT-A decreased significantly in haemorrhagic shock and recovered after resuscitation. The changes in Flow density_WF measured with OCT-A correlated with parameters of systemic haemodynamics and conjunctival microcirculation. The latter was analysed using IDF video microscopy and is the current gold standard for microcirculatory analyses.

In the current study, induction of haemorrhagic shock by removal of a large amount of blood resulted in typical changes in systemic haemodynamics, with reductions in cardiac index and a decrease in perfusion pressure. The drop in central venous oxygen saturation and increasing lactate concentrations may be interpreted as signs of tissue hypoperfusion and cellular oxygen deficit. To compensate, sheep reacted with an increase in heart rate and by recruiting fluid from the interstitium, as indicated by haemodilution (reflected by decreases in haematocrit and haemoglobin levels). These changes were accompanied by a reduced microvascular flow (MFI) and a drop in capillary perfusion (PPV) of the conjunctival microcirculation. These observations confirmed those obtained in previous studies on sheep in haemorrhagic shock [25, 26]. We now show that these changes seen in systemic and microcirculatory variables during haemorrhagic shock and after resuscitation are closely correlated to OCT-A data of the retina. As the ophthalmic artery arises from the internal carotid artery, altered retinal perfusion may indicate impaired cerebral microcirculation. Many previous studies have shown a close relationship between cerebral events and qualitative measurements of retinal microvasculature [11‐13, 15]. In this context, it should be mentioned that the brain and retina are controlled by a local autoregulatory mechanism in which the blood flow is regulated according to metabolic needs and oxygen consumption demands despite moderate variations in perfusion pressure [34]. In the current study, the relative maintenance of the retinal Flow density_WF after blood loss of 10 ml∙kg^− 1 body weight in all sheep (Fig. 2), while MAP was reduced by approximately one third, may be interpreted as a MAP within the retinal autoregulatory range, while further blood loss undercut the retinal autoregulatory threshold, which was reflected by a drop in Flow density_WF in the current study. In this context, Park et al. [35] presented preliminary results of OCT-A analyses during haemorrhagic shock in rats and reported a constant retinal blood flow despite a drop in MAP. Although exact values were not reported, it may be hypothesized that the MAP remained above the threshold of autoregulation since they removed 40% of rat blood volume [35], as compared to about 60% in the present study. Against this background, Ono et al. [36] monitored the cerebral autoregulation response during cardiac surgery using near-infrared spectroscopy and showed that blood pressure excursions below the cerebral autoregulation threshold were associated with acute kidney injury. Therefore, further research may investigate the association between retina perfusion, assessed by OCT-A, and organ (dys)function.

Future clinical use of OCT-A to quantify tissue perfusion could be valuable in situations in which classical surrogates for monitoring of tissue perfusion, such as haemodynamic parameters or urine output, do not always adequately reflect the level of tissue perfusion (e.g. prolonged haemorrhagic or septic shock). Optimizing macro-haemodynamic parameters in this setting does not necessarily ensure adequate tissue perfusion [5‐9]. However, the utility of OCT-A for these important issues needs to be evaluated in further experimental and clinical studies.

The present feasibility study demonstrates for the first time that, as a novel, non-invasive, and contactless technology, OCT-A can monitor microcirculatory perfusion during shock states in large animals. It may also have the potential to overcome some critical issues in monitoring microvascular perfusion, which impede the clinical use of methods to monitor the microcirculation [37]. In this context, a quantitative evaluation of the retinal microcirculation can be automatically performed with OCT-A without the need for intravenous dye injection [10]. The reproducibility of flow density measurements using OCT-A was evaluated in healthy subjects, in patients with different ocular and systemic diseases, and in different animal models [20, 21, 24, 28, 38]. In addition, by taking simple precautions, such as using a lid speculum, imaging in mydriasis, and lubrication of the cornea, blood flow can be analysed in the same retinal vessels of the same retinal area over the entire experimental period, which is not possible with most other methods of microvascular monitoring.

The fact that retinal perfusion, conjunctival microcirculation, and systemic haemodynamics showed a very close correlation is probably owed to the relatively homogeneous impairment of tissue perfusion in the early phase of haemorrhagic shock. It must be expected that results would be more heterogeneous in septic or other types of distributive shock. Heterogeneous changes in the microcirculation have been described previously [39, 40].

The utility of the innovative technology presented here in an intensive care setting remains to be evaluated in further experimental and clinical studies. In particular, the OCT-A instrument used in the current study needs a horizontal orientation of the ocular bulb, which is not feasible in critically ill humans. In this regard, Heidelberg Engineering, for example, recently presented the OCT FLEX device (with OCT-A module), which enables imaging on recumbent patients. However, there are further problems to be solved before OCT-A can be used routinely in critically ill patients as, for example, imaging in miosis (stimulant monitoring of the pupil reaction) or imaging through different media opacities (especially cataract in elderly patients).

The current study shows for the first time that OCT-A is feasible to monitor retinal perfusion in sheep during experimental haemorrhagic shock and fluid resuscitation. We show that changes in retinal perfusion examined using OCT-A correlate with changes in systemic haemodynamic parameters and with variables of conjunctival microcirculation. This makes OCT-A a promising tool for in-vivo monitoring of the central microcirculation in critical illness. Further studies are needed to evaluate the utility of this technology in clinical practice.