Comparison of predictive ability of macrocirculation and microcirculation markers on outcomes of patients with septic shock during initial �uid resuscitation: A prospective observational study

This study was conducted to compare the predictive ability of the macrocirculation (mean arterial pressure, heart rate and central venous pressure), microcirculation markers (capillary re�ll time and peripheral perfusion index) and also �uid balance, lactate and clearance lactate on outcomes of patients with septic shock during initial resuscitation. This was a single-center observational study of 100 adult patients with septic shock who were admitted to the intensive care unit (ICU) at a university a�liated hospital in northwest of Iran. The area under the curve (AUC) of receiver operating characteristic (ROC) curve and multivariate logistic regression analyses were used. The AUCs for the macrocirculation and microcirculation markers for predicting mortality ranged from 0.517 to 0.770 and 0.695 to 0.993, respectively. Among the macrocirculation markers, the best predictive values for mortality were related to the central venous pressure and mean arterial pressure markers at baseline, with AUCs of 0.770 and 0.753, respectively. The microcirculation markers had a better predictive accuracy value for mortality in patients with septic shock than macrocirculation markers which the PPI is the best one. The combination of markers had a greater AUC with higher sensitivity and speci�city than the marker alone to predict outcomes.


Introduction
Septic shock is a subset of sepsis, one of the common serious conditions in critically ill patients, which leads to profound circulatory, cellular, and metabolic abnormalities 1 .It is associated with a greater risk of mortality, prolonged hospital stays, high duration of mechanical ventilation (MV), need to use vasopressors, renal replacement therapy (RRT), increased costs for patients, and increased use of health care resources 2,3 .Despite the intensive research for the management of septic shock, it still remains unacceptably high in critically ill patients 4 .As septic shock is a time-dependent disease, early diagnosis and treatment, as well as risk strati cation, have led to identifying many prognostic markers that could help to implement a more aggressive and effective management [5][6][7] .Hemodynamic changes such as heart rate (HR), mean arterial pressure (MAP), central venous pressure (CVP) in patients with septic shock are complex and various 8 .Many patients with septic shock would manifest an increased level of blood lactate 9,10 .Ensuring and monitoring adequate perfusion is essential for recovery and better prognosis in this population.
Although hyperlactatemia is still considered a key parameter to disclose tissue hypoxia and trigger resuscitation 6, 8 , other factors such as sustained hyperadrenergia and impaired hepatic clearance than hypoperfusion may contribute to increased serum lactate levels and measurements of lactate levels may not be universally available [11][12][13] .In addition, the kinetics of lactate recovery is relatively slow, which makes it a suboptimal target for uid resuscitation 14,15 .Therefore, resuscitation based only on persisting hyperlactatemia might be inappropriate.According to evidence, many indicators have been proven reliable in re ecting perfusions, such as peripheral perfusion index (PPI) 16,17 , capillary re ll time (CRT) 18,19 , and mottling score 20 .ANDROMEDA-SHOCK trial and its post hoc analysis suggested that a microcirculation-guided strategy based on CRT might limit organ failure and lower mortality compared to a lactate-targeted 18 .CRT could be used as a target for uid resuscitation in septic shock because of the simplicity of its assessment, bedside applicability, rapid recovery following uid resuscitation, and availability in a resource-limited setting 21 .Moreover, mottling score as a noninvasive marker is easy to assess and could be a parameter to guide treatments that speci cally target organ perfusion and microcirculation 22 .The results of a study by Dumas et al. 23 supported the high prognostic value of mottling score for 14-day mortality in septic patients.
While MAP and lactate serum levels are commonly used markers in resuscitation, clinicians may be able to improve their resuscitation by broadening their assessment of the microcirculation markers, which more adequately re ects tissue perfusion.Thus, this study was conducted to compare the predictive ability of the macrocirculation and microcirculation markers on outcomes of patients with septic shock.Addition to macro and microcirculation markers, the prognostic values of three factors include uid balance, lactate and clearance lactate on outcomes were evaluated, separately.In order to evaluate the cumulative effect of the markers, we combined the microcirculation and macrocirculation markers separately in three different time intervals to predict the consequences of septic shock.

Baseline characteristics and clinical outcomes of participants
Baseline characteristics and clinical outcomes of the patients are shown in Table 1.One hundred patients with septic shock, including 55 men and 45 women, were enrolled in this study.The median (IQR) age, APACHE IV and SOFA scores of the patients were 66 (59-70), 27 (24-30) and 14 (12-15), respectively.All participants used vasopressors and underwent MV.The most common comorbidities in the patients were HTN (26.5%) and DM (25%).In terms of outcomes, the mean ± SD of ICU LOS and MV duration of the patients were 13.5 ± 2.3 and 10.8 ± 2.5 days, respectively.Seventy percent of the patients need only one vasopressor and 27% of the patients' needs RRT.The incidence of mortality was 34%.In terms of multivariate analysis, the male gender was associated with an increased use of more than one vasopressor (OR: 2.858, 95% CI: 1.075-7.599,P = 0.035) (Additional File 1, Table 1).

Time trend changes of markers
The time trend of macrocirculation, microcirculation and independent markers are presented in Table 2.
According to the results, the trend changes of all markers based on the time effect were signi cant (P < 0.05).Parameters including hemodynamic (HR, MAP, and CVP), PPI, uid balance and clearance lactate signi cantly increased after 6h and 24h of resuscitation.However, serum lactate concentration and CRT signi cantly decreased after 6h and 24h of resuscitation (P < 0.001).

Logistic regression analysis
Univariate and multivariate logistic regression analysis to determine the effect of microcirculation markers and independent factors ( uid balance, lactate and clearance lactate) on outcomes are presented in Tables 3 and 4.

Predicting outcomes by markers
The AUCs for the macrocirculation and microcirculation markers for predicting mortality ranged from 0.517 to 0.770 and 0.695 to 0.993, respectively.Best performing predictive value for mortality in the macrocirculation markers was related to the CVP at baseline with (AUC: 0.770, 95% CI: 0.676-0.849,P < 0.001), and the best cutoff value (> 13) had a value sensitivity 61.76%, speci city 90.91%, (LR+) 6.79, (LR-) 0.42, PPV 77.8%, NPV 82.2%, and 65% of accuracy.All microcirculation markers had signi cantly excellent predictive value for mortality (AUC between 0.837 and 0.993), except CRT at baseline with (AUC: 0.695, 95% CI: 0.594-0.783,P < 0.001).Best performing predictive value for mortality in the microcirculation markers was related to the PPI at 24h with (AUC: 0.993, 95% CI: 0.983-1.000,P < 0.001), and the best cut-off value (< 1.55) had a value sensitivity 94.1%, speci city 93.9%, (LR+) 15.53, (LR-) 0.06, PPV 88.89%, NPV 96.88%, and 65% of accuracy.The microcirculation markers had a better predictive accuracy value for mortality in patients with septic shock than macrocirculation markers.Nevertheless, most markers in both of macrocirculation and microcirculation had poor predictive value for the other outcomes with an AUC between 0.408 and 0.705.In macrocirculation markers, MAP (at baseline) for ICU length of stay, and HR (at 24h), MAP (at 6h) for MV duration had AUC values signi cantly greater than 0.6.Results for macrocirculation and microcirculation markers with signi cant AUC values for other outcomes (RRT, vasopressors, ICU length of stay, and MV duration) are shown in Tables 5 and   6.
The cumulative effect of macrocirculation marker at 6h was comparable as predictors of mortality with HR (AUC: 0.726 vs. 0.615, P = 0.0433) (Fig. 1B).However, there was no signi cant difference between the AUCs of combined marker and the macrocirculation markers at 2h for predicting mortality (P > 0.05) (Fig. 1C).
Among the uid balance, lactate and clearance lactate at 6h, the AUC of lactate was comparable as predictor of mortality with uid balance (0.897 vs. 0.738, P = 0.0028) and clearance lactate (0.897 vs. 0.799, P = 0.0119) (Fig. 1G).However, among these factors at 24 h, the AUC of uid balance was signi cantly lower than lactate (0.714 vs. 0.984, P < 0.0001) and clearance lactate (0.714 vs. 0.961, P < 0.0001) markers for predictive mortality (Fig. 1H).The rest of the results based on ROC curves to predict other outcomes with pairwise comparison between markers are presented in Supplementary le, Fig. 2 to 31.

Discussion
In this prospective observational study, we compare the predictive ability of the macrocirculations (HR, MAP and CVP), microcirculations (CRT, PPI and malting score) and uid balance, lactate and clearance lactate on the outcomes (mortality, RRT, vasopressors, ICU LOS and MV duration) of patients with septic shock during initial resuscitation.In addition, to have the cumulative effect of the markers, we combined the microcirculation and macrocirculation markers separately at three different time intervals to predict the consequences of septic shock.Our ndings showed that the accuracy of the microcirculation markers to predict outcomes were comparable and higher than that of the macrocirculation markers.The three most important predictors of mortality identi ed in our study were PPI with (AUC: 0.993), CRT with (AUC: 0.990) and lactate with (AUC: 0.984) at 24 h.
Of the microcirculation markers we evaluated, PPI provided the most bene cial prognostic information about mortality in adult patients with septic shock.PPI is an indicator mirroring inadequate perfusion in critical patients 16 , and to compare with blood lactate testing, PPI is real-time and non-invasive 24 .The plotted ROC curves of PPI and CRT at baseline for predicting mortality showed the AUC 0.837 and 0.695, respectively.These results showed that PPI predicts mortality earlier with better sensitivity and speci city.In the early stages of hypoperfusion, peripheral blood vessels constrict to return enough blood to the heart 25,26 .At this stage, macro vital signs such as HR and blood pressure (BP) are normal.Since the general circulation is stable, blood lactate levels do not rise.However, PPI, an indicator of regional perfusion, decreases as vasoconstriction reduces blood ow in that area.This means that PPI is superior to other micro parameters such as lactate in alerting physicians to hypoperfusion.However, one of the reasons limiting the use of PPI is that it can be easily affected by other factors such as low temperature, vascular diseases, and the position of patients 27,28 .
In this study, there was no signi cant difference in the accuracy of CRT, and the lactate concentration, and both markers had excellent predictive value for mortality.The ability of the CRT to predict mortality was similar to that of the serum lactate concentration; therefore, measurement of the CRT may be an alternative for invasive measurement of the blood lactate concentration in evaluating patients with septic shock.However, CRT assessment is susceptible to factors that can have a signi cant impact on results, such as ambient temperature, skin temperature, core temperature, age, ambient light conditions, time of pressure, level, and site of pressure application 29 .Some of these conditions can be controlled and the measurements standardized to reduce ambient-related and technical variability 30 .Van Genderen et al. 24 investigated the reliability of observers in CRT assessments between different health care workers and showed good overall consensus.
Another study showed a good correlation with objective variables of peripheral perfusion 29 .Ait-Oufella et al. 19 showed that CRTs were highly reproducible in a prospective cohort of patients with septic shock and showed excellent interracial concordance [16].However, to optimize CRT reproducibility, teaching, standardizing, and reducing the impact of environmental factors are effective.
Our observational study shows that the fair predictive value (AUC: 0.738 and 0.714) of uid balance for mortality at 6h and 24h, respectively.These results are consistent with previous studies that showed the higher positive uid balance in resuscitation over the rst two days was associated with increased risk of mortality in septic shock patients 31,32 .Our results also showed a good predictive accuracy value of serum lactate and lactate clearance for mortality and number of vasopressors and also a poor signi cant predictive value for RRT.
Similar to our results, a retrospective study by Ryoo et al. 33 showed that the lactate and lactate clearance levels at 6-hour are more effective tools for the prognosis of septic shock patients who were treated with protocoldriven resuscitation bundle therapy.A study by Mahmoodpoor et al. 34 , showed that the serial measurements of serum lactate with special emphasis on its concentration at 24hour after admission remains the most predictive of short-term mortality in the ICU.Another study by Marty et al. 35 reported that lactate clearance was the best parameter associated with the 28-day mortality rate in septic patients during the rst 24hr in the ICU.Moreover, a similar study showed that combining lactate levels and its clearance is a reliable predictor of mortality in sepsis 36 .
We also found that the mottling score as a non-invasive marker was associated with an increased risk of mortality and RRT in septic patients.Our results con rmed previous studies which reported a strong relation between 14-day mortality and mottling score 22,23 .However, several factors could modify the clinical evaluation of mottling, such as vasopressors dosage 20 .An important observation of the present study is the combined models for both macrocirculation and microcirculation markers had a greater AUC and signi cance with higher accuracy, SN, SP, LR+, PPV, and NPV, and lower values of LR-as compared to the models created with each marker alone; since each of these indicators alone can be in uenced by factors that play a role in their accuracy to predict outcomes.While when they are combined, they certainly cover the negative points and highlight the positive points.Therefore, the combination of markers seems to be the most reliable predictor of the septic shock outcome.It seems that the most important variables were microcirculatory variables at 24h, suggesting that clinicians could rely on using microcirculatory parameters to evaluate the adequacy of resuscitation.To support the importance of these variables it would then be appropriate to use the cutoffs identi ed by the ROC analyses in a multivariable analysis, adjusted for appropriate confounders to identify prognostic value.
Our study has several limitations.It is a single-center study, and results need to be con rmed in a larger population.Our participants were critically ill patients with various pre-existing comorbidities which acute dysfunction precludes a correct interpretation of results.We excluded patients with advanced liver dysfunction and ARDS but cannot rule out some degree of subclinical dysfunction.CRT assessment might be subjected to inter-observer variability, but we used a standardized technique that decreases the likelihood of bias.Nevertheless, while the size of this preliminary study was not very large, it was su cient to highlight signi cant results.
In conclusion, microcirculation markers to predict mortality was superior to macrocirculation markers.The three most important predictors of mortality identi ed in our study were PPI, CRT and serum lactate at 24 h.An important observation of the present study is the combined models for both macrocirculation and microcirculation markers at baseline which had a greater AUCs with higher sensitivity and speci city than the marker alone to predict morbidity and mortality in patients with septic shock.

Study design and setting
This was a single-center, prospective, observational study of adult patients with septic shock who were admitted to the intensive care unit (ICU) of a university a liated hospital in northwest of Iran, between December 2019 and September 2021.This study was conducted to compare the predictive ability of the macrocirculation and microcirculation markers on outcomes of patients with septic shock.

Participants
All consecutive adult patients with septic shock admitted at ICU subjected were considered as eligible for this study.Septic shock was diagnosed at ICU admission according to the Sepsis-3 Consensus Conference 6 .It was de ned as suspected or con rmed infection by a serum lactate > 2 mmol/liter and requirements of norepinephrine (NE) to maintain a MAP ≥ 65 mmHg after an intravenous uid load of at least 20 ml/kg over 60 minutes 6 .Patients in whom CRT could not be evaluated (Raynaud syndrome, severe hypothermia), pregnant patients, patients with active bleeding, patients with liver cirrhosis, patients with severe acute respiratory distress syndrome (ARDS), patients who required a peremptory surgical procedure during the rst 6h after potential inclusion, or had a Do-Not-Resuscitate (DNR) order, were excluded from the study.

Management of patients
All participants with septic shock were treated according to the standard protocol recommended by the Surviving Sepsis Campaign (SSC) guidelines for the rst 3 hours includes obtaining blood culture before antibiotics, obtaining lactate level, administering broad-spectrum antibiotics, administering 30 mL/kg of crystalloid uid in 2h for hypotension (de ned as mean arterial pressure (MAP) < 65) or lactate (> 4), and standardized assessment of the circulation including serum lactate levels before and after this initial FR 12 .
After an initial uid bolus, a uid challenge technique was implemented as long as hemodynamic improvement was witnessed.Norepinephrine up to a maximum dose of 1.5 µg/Kg/min was started as the rst choice of vasopressor to maintain mean arterial pressure (MAP) > 65 mmHg., and sequential organ failure assessment (SOFA) scores 40 .For this research protocol, we collected data that re ected macrocirculation and organ perfusion at time zero (0h), as well as at 6 and 24 hours (6h and 24h) after ICU admission.Macrocirculation was assessed using heart rate (HR), mean arterial pressure (MAP), and central venous pressure (CVP).Microcirculatory dysfunction and organ perfusion were assessed through capillary re ll time (CRT), peripheral perfusion index (PPI), and mottling score.In addition, uid balance (at 6h and 24h), serum lactate levels (at baseline, 6h and 24h) and clearance lactate levels (at 6h and 24h), as three independent factors were assessed for all participants.The study period was 24 hours that at zero time, 6h and 24h, all markers were recorded for all patients.After this, attending intensivists continued to treat patients according to standard protocol care.All patients were followed until ICU discharge.

Micro/macro circulation markers
Heart rate was monitored by electrocardiography and mean arterial pressure was measured through noninvasive blood pressure technique.HR, MAP, and CVP were recorded at baseline (0h), 6h, and 24h.Normal range was de ned for MAP greater than 65 (mmHg), CVP between 8 and 12 (mmHg), and HR for adults from 60 to 100 beats per minute (bpm).
CVP monitoring is prescribed every 3 h by the medical staff.After checking the absence of spontaneous breathing of the patient on the screen of the ventilator, CVP was carefully measured by the physician or by a trained nurse.Measurements in patients in whom a cardiac arrhythmia occurred were not performed or excluded.The zero reference was carefully checked and calibrated as required.The pressure line was connected to the monitoring system.The research staff (physician or research nurse) froze the screen allowing the measurement of pressure at end expiratory time at the base of the "c" wave.
Capillary re ll time (CRT) was measured by applying rm pressure to the distal phalanx of the index nger for 10 seconds.The pressure applied was just enough to remove the blood at the tip of the physician's nail, illustrated by the appearance of a thin white distal crescent (blanching) under the nail.The time for the return of the normal skin color will be registered with a chronometer, and > 3 seconds is de ned as abnormal.During the prospective observational study, CRT was measured three times (0h, 6h, and 24h), and the mean value ± standard division was recorded.
PPI was derived from the photoelectric plethysmographic signal of pulse oximetry which re ects changes in peripheral perfusion that can be measured continuously and noninvasively.A PPI value < 1.4% indicates the presence of poor peripheral perfusion in critically ill patients.
Mottling score was de ned as patchy skin discoloration, is a common sign of cutaneous hypoperfusion Blood ow reduction may be due to local vasoconstriction and endothelial dysfunction 20,41 .The mottling score provided a semi-quantitative evaluation of mottling based on skin area extension on knee: Score 0 no mottling, score 1 small mottling area (coin size) localized to the center of the knee, score 2 mottling area that does not exceed the superior edge of the knee cap, score 3 mottling area that does not exceed the middle thigh, score 4 mottling area that does not exceed the fold of the groin, and score 5 severe mottling that extends beyond the groin.
Fluid balance was calculated as follows: total uid input minus total uid output within the rst 72 h of ICU stay.As uid accumulation during 72 h is an independent marker for mortality in septic patients we measured uid balance at 72 h after initiation of initial resuscitation.Lactate and lactate clearance levels outcomes: Serum was sampled at baseline and after 6h and 24h of resuscitation from arterial.A normal serum lactate value is de ned as less than 2 mmol/l.Lactate will be assessed with the techniques, including arterial serum levels point-of-care or artery blood gas analysis at the central laboratory.Moreover, we calculated lactate clearance using the serum lactate level at 6h and 24h after admission using the following formula: lactate clearance = (lactate baseline − lactate at 6h or 24h) / lactate baseline 42 .

Outcomes
Outcomes including ICU length of stay (LOS), need of renal replacement therapy (RRT), need and number of vasopressors, duration of mechanical ventilator (MV), and mortality rate were registered for all participants.We also de ned a binary outcome based on median of quantitative variables and for qualitative variables according to yes or no; that included ICU LOS (≥ 13 days vs. <13 days), MV duration (≥ 10 days vs. <10 days), need for renal replacement therapy (yes vs. no), number of vasopressors administered (1 vs. >1) and ICU mortality (yes vs. no).

Statistical analysis
All quantitative variables were tested for normal distribution with Kolmogorov-Smirnov test.Patient characteristics are reported as mean ± standard deviation (SD), median (IQR) for non-normal distributions, and percentages as appropriate.Changes of macrocirculation and microcirculation markers in participants according to time trends (baseline, 6h and 24h) were assessed based on two-way analysis of variance with repeated measures (RMANOVA), adjusted and non-adjusted models.Adjusted RMANOVA model was done according to the baseline characteristics, including age, gender, comorbidities, and APACHE IV and SOFA scores.The results of the mottling score were reported as median (IQR), and P-value was calculated based on the Friedman test.The effect of patients' characteristics, macrocirculation and microcirculation markers on outcomes (ICU length of stay, need for renal replacement therapy (RRT), need and number of vasopressors, duration of MV, and mortality rate) were determined by univariate and multivariate logistic regression, and the results were expressed as odds ratio (OR) with 95% con dence interval (CI).Receiver operating characteristic curve (ROC) was generated, and area under the curve (AUC) gures were calculated alongside sensitivity (SN), speci city (SP), positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), negative likelihood ratio (LR-), for each macrocirculation and microcirculation markers at baseline, 6h, and 24h.
In addition, we combined the microcirculation and macrocirculation markers separately in three different time intervals to create new markers to predict outcomes using binary logistic regression to get the probability and then run a ROC curve using the probability as the test markers, and AUCs were compared using the DeLong test.

Table 2
Evaluation markers in patients with septic shock according to time trends *Non-adjusted time main effect based on two-way analysis of variance with repeated measures (RMANOVA), ** Adjusted RMANOVA according to the baseline characteristics including age, gender, comorbidities, APACHE IV and SOFA score; Abbreviations: Heart Rate (HR), Mean Arterial Pressure (MAP), Central venous pressure (CVP), Capillary Re ll Time (CRT), peripheral perfusion index (PPI).*The results of mottling score was reported as median (IQR) and P-value was calculated based on Friedman test.

Table 3
Univariate and multivariate logistic regression analysis to determine the effect of microcirculation markers on outcomes *P < 0.05 considered as signi cant, Abbreviations: Odds Ratio (OR) *P < 0.05 considered as signi cant, Abbreviations: Odds Ratio (OR)
37e protocol study was reviewed and approved by Research Ethics Committees of Islamic Azad University-Tabriz Branch (IR.IAU.TABRIZ.REC.1400.044), in accordance with the Declaration of Helsinki of the World Medical Association37.Written informed consent was obtained from the patients or from their legally accepted representatives.This observational study was conducted and reported in accordance with the recommendations of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement 38 .
General characteristics of the patients were recorded at baseline (0h) includes demographic data (age and gender), comorbidities (congestive heart failure [CHF], cerebral vascular accident [CVA], diabetes mellitus [DM], end-stage renal disease [ESRD], hypertension [HTN], and ischemic heart disease [IHD]), severity of illness which evaluated by the acute physiology and chronic health evaluation (APACHE IV)