Background
The microcirculation is a vast network of small vessels (terminal arterioles, capillaries and venules < 100 µm diameter) in which the exchange of oxygen and nutrients with tissues takes place [
1]. Its derangement, defined as “microcirculatory shock” [
2], is recognised as an important cause of organ dysfunction in critically ill patients, affected by various disease states, such as sepsis, severe trauma, haemorrhagic shock and post-cardiac arrest [
2‐
5]. Furthermore, microcirculatory abnormalities and its persistence despite adequate macro-hemodynamic resuscitation were independently associated with morbidity and mortality in many critical conditions [
6‐
12].
Today, the development of new technologies of in vivo video microscopy and its integration in easy-to-handle microscopes as in Sidestream Dark Field (SDF) imaging allow us to assess the (sublingual) microcirculation at the bedside, in a non-invasive way [
13]. However, until 2015, data on microcirculatory alterations in the Intensive Care Unit (ICU) were restricted to small sample-sized studies in high-risk patients [
7,
8,
14].
The MicroSOAP study by Vellinga and colleagues [
15] gave a first insight in the
prevalence of microcirculatory alterations in a large number of ICU patients. However, due to its design with a single time-point observation, the
incidence in a time-dependent manner remains to be elucidated. Primary aim of the study was to detect a difference in the incidence of microvascular flow abnormalities between ICU survivors and non-survivors. Secondary outcomes were long-term mortality (in-hospital mortality and 90-day mortality) and development of organ dysfunction (described by sequential organ failure assessment, SOFA).
Methods
Patients enrolment and data collection
The MicroDAIMON (Microcirculation DAIly MONitoring in critically ill patients) is a single-centre prospective observational study (clinicaltrials.gov, NCT 02649088 registered on 23 December; retrospectively registered). The recruiting phase was performed in a 9-month period in 2013 (from 1 April to 31 December) in a 12-bed mixed ICU of an Italian teaching hospital with a mean number of yearly-admitted patients of 400. The ICU was structured in three subunits of four beds each, caring for respiratory, traumatology and medical critically ill patients, respectively. For the study purpose, each subunit was subsequently included and monitored during a 3-month period for patients’ screening and the recruitment: from 1 April to 30 June 2013, the medical subunit, from 1 July to 30 September 2013, the traumatology subunit and from 1st October to 31st December 2013, the respiratory subunit.
Patients were screened and included in the study within the first 12 h from ICU admission. Exclusion criteria were age < 18 years, lack of informed consent and pathophysiological conditions that may interfere with the sublingual microcirculation videos acquisition (maxillofacial traumas/surgery, oral bleeding, mucositis, etc.). In context to the microcirculatory assessments, demographic, laboratory, microbiologic, hemodynamic and other clinical data were recorded. All patients were followed up for 90 days after the ICU admission.
The study protocol was approved by the Local Ethics Committee and conducted in respect of the principles of Helsinki declaration (last revision, Edinburgh 2000). A written informed consent was obtained from all the included subjects or their next of kin in compliance with national applicable laws.
Microcirculation assessment
The sublingual microcirculation was evaluated at the moment of the inclusion and every 24 h until discharge/death with SDF imaging (Microscan
®, Microvision Medical, Amsterdam, The Netherlands) [
13].
The video acquisition technique is extensively described in previous papers [
16]. For every session, videos from at least five different sites were registered trying to obtain a good video quality and to avoid artefacts that may affect flow or vessels density variables [
16].
The three best videos were chosen from each session, in compliance with recommendations from Massey et al. [
17] and blindly analysed offline with a dedicated software (Automated Vascular Analysis, AVA Software 3.0, MicroVision Medical, Amsterdam, The Netherlands) by a restricted group of four experienced investigators. Inter-observer variability was calculated, based on the simultaneous analysis of ten randomly selected SDF videos by all the investigators. Variables of flow (Microvascular Flow Index, MFI and proportion of perfused vessels, PPV), as well as capillary density (total vessel density, TVD, perfused vessel density, PVD) and flow distribution (Heterogeneity Index, HI) were calculated according to international criteria [
18,
19]. Flow was scored per quadrant as 0 (no flow), 1 (intermittent flow), 2 (sluggish flow) and 3 (continuous flow). The MFI is the average over 4 quadrants × 3 areas of interest. Total vessel density (TVD, mm/mm
2) was calculated as the total length of vessels divided by the total area of the image. The percentage of perfused vessels (PPV) was estimated as follows: 100 × [(total number of grid crossings − [no flow + intermittent flow])/total number of grid crossings] and expressed as percentage. The perfused vessel density (PVD, mm/mm
2) was estimated by multiplying TVD by PPV as estimated with the De Backer method. The Flow Heterogeneity Index (FHI, arbitrary units) was calculated as the highest MFI minus the lowest MFI, divided by the mean MFI of all sublingual sites [
18].
Analogous to previous data, a threshold for the MFI < 2.6 was a priori established to define an abnormal microcirculation [
3,
8,
15,
20].
Statistical analysis
Data analysis was conducted with SPSS Software 17.0 (IBM, New York, NY) and GraphPad Prism 6 (GraphPad Software, La Jolla, CA). All data are presented as mean ± standard deviation (SD) or median [interquartile range, IQR].
Descriptive statistics were performed to obtain patients’ baseline characteristics. Quantitative variables distribution was tested with Kolmogorov–Smirnov normality test. Parametric (Student’s t test with Welch’s correction) and nonparametric tests (Mann–Whitney U test) were applied to describe the differences between groups for the variables of interest as appropriate. Fisher’s exact test was performed for comparisons between categorical variables, and the results are presented as percentage, odds ratio (OR) and 95% confidence interval (CI). Kaplan–Meier 90-day survival curves with Tarone–Ware test for the comparison of the hazard ratio between groups were built for the survival analysis.
Binary logistic regression analysis was performed with a forward stepwise entry method. A p value of less than 0.05 was considered statistically significant.
Discussion
This MicroDAIMON study is currently the largest prospective longitudinal observational study to describe the incidence of microcirculatory derangements among a mixed group of critically ill patients, offering a day-by-day follow-up. The incidence of baseline microcirculatory flow abnormalities was 20.6%, and more than half (55.7%) of the patients displayed an abnormal MFI in at least one observation during ICU stay. The main finding of this study is that in this mixed ICU population, an abnormal baseline MFI is independently associated with unfavourable outcome in terms of ICU, in-hospital and 90-day mortality. In addition, the contemporary presence of tachycardia showed an additive predictive power towards mortality in the survival analysis. However, the change of MFI over time was not associated with outcome, in terms of both organ failure (SOFA) and mortality. In contrast to an abnormal MFI on day 1, we could not associate an abnormal MFI after day 1 with unfavourable outcome. No associations were found between the other microvascular variables and outcome.
In 2015, the MicroSOAP study provided the first and largest database on the prevalence and the significance of the microcirculatory alterations in a heterogeneous ICU population, with a time-point observation across 36 ICUs worldwide [
15]. The authors reported a prevalence of MFI abnormalities of 17%, using the same predefined cut-off value [
15,
26]. This difference in reported percentage of MFI abnormalities can be explained by the difference in study design (longitudinal vs. point prevalence). Our data confirm previous observations, showing an important prognostic role of the microcirculation in various subsets of critically ill patients [
4‐
9,
14]. In contrast to the existing literature, these findings extend the predictive value of early microcirculatory alterations towards 90-day mortality. Patients with an abnormal MFI at baseline showed an absolute risk of non-survival almost three times higher in comparison with patients with a normal MFI.
However, in the present study, routine day-by-day microcirculatory monitoring does not confirm previous observations. In 2004, Sakr et al. [
11] introduced for the first time the concept of serial observations of microcirculation in a cohort of 49 patients with septic shock. In this highly selected group of patients, the persistence of microcirculatory alterations was associated with persistence of shock, development of multiple organ failure and mortality.
Conversely, ICU survivors showed early improvement of microcirculation. These data were confirmed by others [
21]. Duranteau and colleagues observed in another selected cohort of 18 patients with traumatic haemorrhagic shock, early derangements of microcirculatory flow and vessel density, as well as its persistence, were able to predict a worse SOFA score after 96 h from ICU admission [
5]. A possible explanation for this discrepancy may lie in the heterogeneous composition of our study population and in considerable differences in microcirculatory baseline abnormalities. Alternatively, microvascular alterations represent differences in underlying pathology between study populations. Careful selection of patients at risk may contribute to the prognostic power of microcirculatory observation. As of now, our data indicate that routine daily monitoring of the microcirculation in an unselected group of ICU patients is of limited prognostic value.
This study has several limitations. Although this study contains the largest reported database on day-by-day monitoring of the microcirculation in critically ill, it appears to have insufficient sample size to correlate differences in the evolution of microcirculatory conditions over time with clinically relevant endpoints (SOFA, mortality) also due to the considerable number of patients lost to follow-up, due to death/discharge. And although the independent predictive value of an abnormal MFI on day 1 was established, the integration of such variable in a model with more commonly used hemodynamic variables was clearly limited by the sample size as well. Further research is needed to establish the additive value of microcirculatory imaging on top of the existing hemodynamic variables. In addition, it is conceivable that other microvascular variables and different cut-off yield different results. We did not found any significant association between the other microcirculatory variables (TVD, PVD, PPV) and the outcome either on day 1 or in the following days. This could be explained by the fact that the MFI, especially if used as a dichotomous variable based on an a priori cut-off of 2,6, could have been the most sensitive variable to detect an association with the outcome in a such heterogeneous population which is expected to cause a “dilution effect” on the microcirculatory alterations. It is also possible that a vessel-by-vessel MFI calculation could have been more precise and provide different results depending on a more accurate evaluation of the capillary blood flow, especially in the presence of marked heterogeneity. In this respect, the burden of time-consuming offline analysis remains a major practical limitation for the study population sample size until the time of the development and full validation of automated analysis software. Real-time “eyeballing” the microcirculation by bedside assessment of MFI is a major advantage in the development of a bedside tool and showed good agreement with the gold standard offline analysis [
27]. Post hoc analysis of our data confirmed 2.6 as the optimal cut-off for the discrimination between survivors and non-survivors. Finally, this was a pure observational study: patients were treated following the international guidelines and principles of good clinical practice, and clinicians had no information about the microcirculation during the study. Therefore, our study design is insufficient to draw conclusions on the applicability of microcirculatory monitoring as a tool to guide resuscitation. Even in the setting where there is an absence of additional prognostic information, derived from microcirculatory monitoring, the observation itself may contain valuable information about the underlying pathophysiologic mechanisms. For example, an increased lactate may adequately predict outcome, but does not reveal its underlying mechanism. Under these conditions, additional assessment of microvascular blood flow may not be useful to predict outcome, but may be helpful for the clinician to select the appropriate resuscitation strategy. Further research is needed to address this topic. Careful selection of subgroups and adequate timing remain of the essence in this process.
Authors’ contributions
CS, AD, ED, RR, EA designed the study, contributed to the interpretation of the results and critically revised the manuscript. CS and ED performed the statistical analysis, drafted the manuscript and interpreted the data. ED, CS, RD, AC, SP, ST, VM, SC, MR, BM made a substantial contributions to the acquisition of the data and the analysis of SDF videos and revised the manuscript for important intellectual content. CB made a substantial contribution in drafting the manuscript and interpreting the results. CI made a substantial contribution in the study design and critically revised the manuscript for important intellectual content. All authors had full access to the data, take responsibility for the integrity of the data and the accuracy of the analysis. All authors read and approved the final manuscript.