Associations between urinary 3-indoxyl sulfate, a gut microbiome-derived biomarker, and patient outcomes after intensive care unit admission

Abstract

Purpose:

3-indoxyl sulfate (3-IS) is an indole metabolism byproduct produced by commensal gut bacteria and excreted in the urine; low urinary 3-IS has been associated with increased mortality in bone marrow transplant recipients. This study investigated urinary 3-IS and patient outcomes in the ICU.

Materials and Methods:

Prospective study that collected urine samples, rectal swabs, and clinical data on 78 adult ICU patients at admission and again 72 hours later. Urine was analyzed for 3-IS by mass spectrometry.

Results:

Median urinary 3-IS levels were 17.1 μmol/mmol creatinine (IQR 9.5 to 26.2) at admission and 15.6 (IQR 4.2 to 30.7) 72 hours later. 22% of patients had low 3-IS (≤6.9 μmol/mmol) on ICU admission and 28% after 72 hours. Low 3-IS at 72 hours was associated with fewer ICU-free days (22.5 low versus 26 high, p=0.03) and with death during one year of follow-up (36% low versus 9% high 3-IS, p<0.01); there was no detectable difference in 30-day mortality (18% low versus 5% high, p=0.07).

Conclusions:

Low urinary 3-IS level 72 hours after ICU admission was associated with fewer ICU-free days and with increased one-year but not 30-day mortality. Further studies should investigate urinary 3-IS as an ICU biomarker.

INTRODUCTION

In patients who are critically ill, the normal microbiome is disrupted, leading to loss of gut colonization resistance and expansion of disease-promoting pathogens [1, 2]. This “dysbiosis” has been shown to be a risk factor for nosocomial infections, sepsis, multiple organ dysfunction, and increased mortality [3–5]. Fecal microbial diversity has been used as a surrogate for colonization resistance and low fecal biodiversity has been associated with increased risk for death and infection, most notably in bone marrow transplant recipients [3, 6].

In the intensive care unit (ICU), fecal microbial diversity is low at admission, and the microbiome becomes dominated by pathogens including Enterococcus and Gram negative bacteria [2, 7]. Downstream, ICU patients who are colonized with enteric pathogens have increased risk for death or infection [8]. It would be important to rapidly identify patients with dysbiosis in order to develop clinical strategies to help augment commensal microbes, restore diverse flora, and prevent loss of colonization resistance. Currently, a major challenge facing such strategies is that it is difficult or impossible to rapidly identify the patients at highest risk of dysbiosis within the diverse ICU population [4]. There are no tests for clinical features that clearly identify patients likely to benefit from microbiome-based interventions.

3-indoxyl sulfate (3-IS) is an indole metabolism byproduct produced by bacteria expressing tryptophanase and has been shown to be an important intercellular signal in microbial communities [9] and may support the survival of mixed microbial communities [10]. In the bone marrow transplant population, low levels of urinary 3-indoxyl sulfate were associated with higher transplant-related mortality and worse overall survival [11]. Since urinary 3-IS levels can be rapidly measured, it may be useful as a biomarker for gut dysbiosis in other high-mortality populations. This study sought to evaluate the utility of urinary 3-IS levels as a gut-specific biomarker and to investigate its ability to predict outcomes in a diverse ICU population.

MATERIAL AND METHODS

Study population

This was a prospective, single-center study that included consecutive adults ≥18 years old who were newly admitted to any one of the 5 medical or surgical ICUs and could be reached within 4 hours of ICU admission and 72 hours later. Assessments at both ICU admission and 72 hours later were required in order to examine the dynamic changes in 3-IS over time. Patients were included if a urine specimen was obtained at the 72-hour assessment, even if no urine was available during the initial assessment. In order to limit the population to ICU patients with a gut microbiome that was relatively preserved at the time of admission, patients were excluded if they had prior Clostridioides difficile infection within 90 days, bacterial bloodstream infections within 30 days, or ICU admissions within 30 days. Informed consent was obtained from all subjects or their appropriate surrogates when the patients lacked capacity. The default consent was written, but verbal surrogate consent was accepted when subjects lacked capacity and the appropriate surrogate was not available in person [12]. The institutional review board of Columbia University Medical Center approved this study.

Sample and data collection

Patients were enrolled from February 2015 to August 2018. Rectal swab and urine specimens were collected within 4 hours of ICU admission with a second set of specimens collected 72 hours later (±4 hours). At ICU admission and 72 hours later, 8 mL of urine was taken via the collection port of an indwelling catheter that had previously been placed. A deep rectal flocked nylon swab (Copan Diagnostics, Murrieta, CA) was collected at each time point. Demographic information, laboratory data and data regarding the interventions performed in between sample collections was extracted from the electronic medical record. Data was collected as the worst clinical and laboratory values during the preceding 24 hours of each sample collection time, in order to calculate the sequential organ failure assessment (SOFA) score for each subject at each time point [13]. Information for interventions performed between sample collection time points included antibiotic use (any dose or duration), proton pump inhibitor use, mechanical ventilation, hemodialysis (intermittent or continuous), immunosuppression use, opioid use, and enteral feeding (by mouth or by naso-enteric tube).

Urinary 3-indoxyl sulfate

Urinary levels of 3-IS were analyzed as both a categorical and continuous variable. Based on the cutoffs used in a prior study analyzing urinary 3-IS in the bone marrow transplant population, patients were classified as low 3-IS at levels ≤6.9 μmol/mmol creatinine and as high 3-IS if levels were >6.9 μmol/mmol creatinine [11]. Urinary 3-IS was measured on an Agilent 1200 SL HPLC system with a 4000 QTrap mass spectrometer and a TurboV electrospray ion source using stable isotope-labeled internal standards and stock solutions as previously described [14]. Within-individual change was calculated as 3-IS after 72 hours divided by 3-IS at ICU admission.

16S rRNA gene sequencing

Swabs were sequenced at two centers because the study evolved slowly, over more than two years. For the first batch of samples, polymerase chain reaction was performed targeting the V4 hypervariable region of the 16S ribosomal RNA gene with primers from the human microbiome project. For the second batch, V3-V4 sequencing was performed using standard Illumina primers. To combine datasets for downstream analyses, paired-ended V3-V4 reads were trimmed using BBMap to isolate the V4 region, and all sequences were then imported into QIIME2 [15]. The single-ended DADA2 pipeline was used to cluster sequences into sequence variants (SVs) which were assigned taxonomic classifications based on a naïve Bayesian approach trained on the GreenGenes 97% clustered database. A minimum threshold of 10,000 counts per sample was set for inclusion in our analysis. Exported BIOM and reconstructed phylogenetic tree files from QIIME2 were imported into R using the phyloseq package to calculate α-diversity metrics (Shannon diversity index) [16]. These metrics focused on the Shannon index, which was calculated as ${\sum }_{i=1}^R{p}_i\log p_i$ where p is the relative abundance of each operational taxonomic unit based on 16S sequencing. Using phyloseq, an unsupervised analysis was performed to assess for any differences in the microbiome, comparing those with low versus those with high urinary 3-IS. False discovery adjusted p-value of <0.05 was considered statistically significant in this analysis.

Statistical approach

Urinary 3-IS was measured at both ICU admission and at 72 hours, but the analyses focused on 72 hour 3-IS based on the rationale that the trajectory of 3-IS would be more important than its value at the moment of ICU admission. The primary outcome was mortality after ICU admission, assessed as 30-day mortality and one-year mortality gathered through hospital records, which interface with the social security death index. Other outcomes investigated were ICU-free days and culture proven infections within 30 days of ICU admission, which were determined as previously described [8]. ICU-free days were calculated as 30 days minus the number of days in the ICU, and patients were given a score of 0 if they died while in the ICU [17]. Comparisons of clinical and laboratory data were performed using Wilcoxon ranksum tests and χ² tests. For multivariable analysis, logistic regression modeling was used with one-year mortality as the primary outcome. A multivariable model was built to assess the independent relationship between 3-IS and mortality after adjusting for the presence of sepsis and the SOFA score, with those co-variables selected a priori as the most likely predictors of death. All statistical analysis was performed using Stata 15 (StataCorp, College Station, TX) at the alpha 0.05 level of significance.

RESULTS

Urinary 3-indoxyl sulfate

A total of 78 patients met study criteria and were analyzed (Figure 1). Median urinary 3-IS level at ICU admission was 17.1 μmol/mmol (IQR 9.5 to 26.2) and 72 hours after ICU admission was 15.6 μmol/mmol (IQR 4.2 to 30.7). Thirteen patients were included in the analysis who had a 72 hour urine sample, but for whom no admission urine sample was available.

Figure 1. Enrollment flowchart.

Clinical characteristics

Baseline characteristics at ICU admission are shown in Table 1 and are stratified by urinary 3-IS levels at 72 hours after admission. There were 22 patients (28%) that had low urinary 3-IS levels compared to 56 patients (72%) that had high levels at the second time point. The majority of patients were admitted to the ICU for sepsis (37%), post-surgical procedure (27%), or after a cardiac event or procedure (18%). Among those with low urinary 3-IS levels, there was a larger proportion of females (59% low 3-IS versus 30% high, p=0.02). Sepsis was a more common reason for admission among patients with low urinary 3-IS levels at 72 hours (77% low 3-IS versus 21% high, p<0.01). At 72 hours after ICU admission, patients with low urinary 3-IS levels had a median SOFA score of 6 (IQR 2 to 12) and those with high 3-IS levels had a median SOFA score of 3 (IQR 1 to 7, p<0.01). SOFA components are detailed in Table 2. Patients with low 3-IS levels at 72 hours, had significantly lower values of platelets and higher values of bilirubin compared to those with high 3-IS levels.

Table 1.

Baseline patient characteristics, stratified by urinary 3-IS level 72 hours after ICU admission.

Characteristics at ICU admission	Low 3-IS at 72h (N=22)	High 3-IS at 72h (N=56)	p-value
Age in years (tertiles)			0.64
30–59	6 (27%)	20 (36%)
60–70	9 (41%)	17 (30%)
70–90	7 (32%)	19 (34%)
Gender			0.02
Male	9 (41%)	39 (70%)
Female	13 (59%)	17 (30%)
Reasons for admission			<0.01
Sepsis	17 (77%)	12 (21%)
Surgical procedure	3 (14%)	18 (32%)
Cardiac	0 (0%)	14 (25%)
Gl bleed	2 (9%)	3 (5%)
Neurologic	0 (0%)	5 (9%)
Other	0 (0%)	4 (7%)
Charlson Comorbiditiesa			0.25
0–1 points	4 (18%)	21 (37%)
2–3 points	10 (45%)	21 (37%)
4+ points	8 (36%)	14 (25%)
SOFA score			0.50
0–4 points	9 (41%)	23 (41%)
5–8 points	8 (36%)	14 (25%)
≥9 points	5 (23%)	19 (34%)
Hematocrit at admission	28.8 (24.3–36.7)	34.5 (28.6–40)	0.09
Albumin at admission	3.1 (2.5–3.8)	3.6 (2.9–3.9)	0.22

3-IS: 3-indoxyl sulfate; ICU: intensive care unit; SOFA: Sequential Organ Failure Assessment score.

^aModified Charlson's Comorbidity Index

Table 2.

Sequential Organ Failure Assessment score characteristics 72 hours after ICU admission, organized by urinary 3-IS levels 72 hours after ICU admission.

SOFA characteristics 72 hours after ICU admission	Low 3-IS at 72h (N = 22)	High 3-IS at 72h (N = 56)	p-value
Median SOFA score (IQR)	6 (2–12)	3 (1–7)	p<0.01
Total SOFA score			0.21
0–2 points	6 (27%)	26 (46%)
3–6 points	6 (27%)	15 (27%)
7–24 points	10 (45%)	15 (27%)
PaO2/FiO2			0.10
≥400	12 (55%)	40 (71%)
300–399	3 (14%)	3 (5%)
200–299	4 (18%)	12 (21%)
100–199 and mechanically ventilated	2 (9%)	0 (0%)
<100 and mechanically ventilated	1 (5%)	1 (2%)
Platelets, ×103/μL			0.03
≥150	10 (45%)	34 (61%)
100–150	1 (5%)	13 (23%)
50–99	5 (23%)	5 (9%)
20–49	5 (23%)	3 (5%)
<20	1 (5%)	1 (2%)
Glasgow Coma Scale			0.06
15 points	10 (45%)	36 (64%)
13–14 points	0 (0%)	4 (7%)
10–12 points	4 (18%)	2 (4%)
6–9 points	2 (9%)	7 (13%)
<6 points	6 (27%)	7 (13%)
Bilirubin, mg/dL			<0.01
<1.2	14 (64%)	42 (75%)
1.2–1.9	1 (5%)	10 (18%)
2.0–5.9	6 (27%)	1 (2%)
6.0–11.9	0 (0%)	1 (2%)
≥12.0	1 (5%)	2 (4%)
Mean arterial pressure or vasoactive agents			0.51
No hypotension	10 (45%)	35 (63%)
MAP <70 mmHg	5 (23%)	11 (20%)
Dopamine ≤5 or phenylephrine <0.3	0 (0%)	1 (2%)
Dopamine >5 or norepinephrine ≤0.1 or phenylephrine >0.3	5 (23%)	7 (13%)
Dopamine >15 or norepinephrine >0.1	2 (9%)	2 (4%)
Creatinine, mg/dL or urine output			0.27
<1.2	7 (32%)	31 (55%)
1.2–1.9	6 (27%)	13 (23%)
2.0–3.4	3 (14%)	5 (9%)
3.5–4.9 or UOP <500 mL/day	2 (9%)	4 (7%)
≥5.0 or UOP <200 mL/day	4 (18%)	3 (5%)

IQR: interquartile range; MAP: mean arterial pressure; UOP: urine output

Relationship with microbiome features

The Shannon index was also analyzed for both time points and used as a surrogate for fecal microbial diversity. At ICU admission, median Shannon Index was 3.75 (IQR 3.4 to 3.9) and 72 hours after was 3.71 (IQR 3.1 to 4.0). Those classified as low 3-IS levels at the second time point also had a lower Shannon Index at 72 hours compared to those with high 3-IS levels (2.78 versus 3.79, p=0.02). At both time points, there was a weak positive correlation between log urinary 3-IS and Shannon Index, which was statistically significant at the second time point (r = 0.42 p<0.01) (Figure 2A and 2B). Patients with high urinary 3-IS at 72 hours had rectal swabs significantly enriched in sequences assigned to E.coli and Parabacteroides (both log2-fold change>20 p<0.01).

Figure 2. Correlations between log urinary 3-IS and Shannon Index.

Pearson correlations between log urinary 3-IS and Shannon Index at ICU admission (A) and 72 hours later (B). 60 out of 78 patients had adequate data for both urinary 3-IS and Shannon Index at ICU admission, whereas 69 out of 78 were available 72 hours later.

Impact of ICU interventions on urinary 3-indoxyl sulfate

The impact of ICU interventions on urinary 3-IS levels was studied by analyzing within-individual change in urinary 3-IS (Table 3). Antibiotic use, both before ICU admission and between the two time points was not associated with a significant difference in levels of urinary 3-IS. Urinary 3-IS levels were significantly lower in those that were exposed to immunosuppression during the first 72 hours in the ICU compared to those that were not (1.08 fold change versus 0.62 fold change, p=0.04). There was no difference in change within urinary 3-IS when patients were exposed to proton pump inhibitors, opioids, feeding or hemodialysis between the two time points. No patients received digestive decontamination or regional anesthesia during the 72 hours prior to ICU admission.

Table 3.

Within-individual change in 3-IS, organized by ICU interventions.

Intervention Type	Without intervention Median (IQR)	With intervention Median (IQR)	p-value
Antibioticsa
Before (N=49)	0.82 (0.39–1.16)	1.05 (0.52–1.60)	0.54
After (N=56)	0.61 (0.4–01.00)	1.04 (0.52–1.54)	0.45
Any (N=58)	0.61 (0.39–1.00)	1.04 (0.52–1.60)	0.39
Immunosuppressionb	1.08 (0.57–1.67)	0.62 (0.26–1.16)	0.04
Proton pump inhibitors	0.87 (0.39–1.23)	1.06 (0.55–1.97)	0.31
Opioids	0.69 (0.36–1.88)	1.03 (0.55–1.44)	0.48
Feedingc	0.94 (0.27–1.59)	1.01 (0.42–1.44)	0.86
Hemodialysis	1.01 (0.39–1.46)	0.81 (0.65–2.24)	0.42

^aAntibiotics received before ICU admission, after ICU admission for the next 72 hours and any antibiotics received prior or after ICU admission

^bImmunosuppression was defined as chemotherapy or prednisone (>5 mg/day dose), cellcept, tacrolimus/everolimus/sirolimus, azathioprine, infliximab or adalimumab.

^cIncludes both tube feeds and ad-lib diets.

Relationship between urinary 3-indoxyl sulfate and mortality

At 30 days after ICU admission, there was no detectable difference in mortality (18% low versus 5% high, p=0.07). Low urinary 3-IS levels at 72 hours after admission was associated with a statistically significant increase in mortality at one year (36% mortality for low 3-IS versus 9% for high, p<0.01). Patients with low 3-IS levels had 22.5 ICU free days compared to 26 ICU free days in patients with high 3-IS levels (p=0.03). The median value of urinary 3-IS at ICU admission was 17.2 among those who survived one year and was 7.5 among those who died (p=0.09). The median value of urinary 3-IS after 72 hours was 18.5 among those who survived one year and 2.06 among those who did not (p<0.01). This is shown as a violin plot in Figure 3. Examining within-individual change in 3-IS, there was a decrease among those who died during follow-up but not among those who survived, although this was not statistically significant (0.63 IQR (0.24–1.63) for those who died versus 1.03, IQR (0.52–1.46) for those who survived, p=0.27). Kaplan-Meier curves show the survival estimates between low and high urinary 3-IS at both follow-up times of 30 days (Figure 4A) and one year (Figure 4B). Since indole metabolism involves the liver [18] and indoxyl-sulfate is secreted by the kidneys [19], the disruption of these organs has the potential to affect urinary 3-IS levels. Therefore, we re-ran the primary analysis excluding those with the highest SOFA categories of both bilirubin and creatinine, and found similar associations with mortality (33% mortality for low 3-IS versus 6% for high 3-IS; p<0.01). After adjusting for the presence of sepsis and the SOFA score in a multivariable model, low urinary 3-IS did not remain statistically significant with mortality (adjusted odds ratio for low 3-IS 2.81, 95% CI 0.61–12.9). This finding was substantively unchanged after stratifying for sex and for sepsis as the admitting diagnosis. A receiver-operator curve comparing SOFA score and urinary 3-IS levels at 72 hours showed similar predictability of one-year survival with similar areas under the curves (Figure 5). The number of culture-proven infections or patients with colonization of MDR organisms was not different between those with high versus low urinary 3-IS levels.

Figure 3. Violin plot showing urinary 3-IS change.

Log urinary 3-IS levels at time of ICU admission and 72 hours later, stratified by survival outcome at one year.

Figure 4. Kaplan-Meier survival estimates.

Kaplan-Meier curves show that low urinary 3-IS levels at 72 hours after ICU admission have a trend towards increased mortality at 30 days (A) and significantly increased mortality at one year (B).

Figure 5. Receiver-operator curves.

The area under the curve of urinary 3-IS levels and SOFA scores after 72 hours as predictors of survival at one year.

DISCUSSION

In this prospective study, low levels of urinary 3-IS measured at 72 hours after ICU admission was associated with a trend towards increased 30-day mortality and with significantly increased mortality within the year. This was true despite relatively small sample size and limited ability to detect anything other than large differences in mortality. Levels of urinary 3-IS measured at ICU admission did not correlate with mortality at 30 days or one year after ICU admission. Those that died had a median decrease in urinary 3-IS levels, suggesting that the dynamics of urinary 3-IS may be more relevant to mortality in this cohort than the value ascertained immediately at the time of ICU admission. Urinary 3-IS at 72 hours was a reasonably good predictor of mortality at one year compared to the well-established SOFA score.

Further studies will need to reexamine the relationship between urinary 3-IS and mortality at both the 30-day and one-year time points. An intact microbiome may confer long-term protection against poor clinical outcomes by providing colonization resistance and preventing infections weeks or even months after ICU discharge. On the other hand, the association between 3-IS and long term mortality suggests that 3-IS may represent baseline patient factors, such as frailty, but this study was not designed to assess frailty measures.

Disruption of the gut microbiome has been connected to critical illness as both a consequence and as a central driver of sepsis and multi-organ dysfunction [4, 20]. There are clinical studies investigating strategies to augment the microbiome in critical illness with probiotics, prebiotics, synbiotics, or even fecal transplant, but none of these interventions have proven benefit [4, 21]. Currently, the typical sequencing approach used to determine microbiome diversity is not rapid enough to determine in real time whether patients are good candidates for microbiome-based therapies. The ability to quickly assess and then reassess microbiome “health” may be especially important in the ICU population as critically ill patients undergo rapid changes to their microbiome [1]. Based on the findings of our study and future studies needed to confirm our results, we hope that measuring urinary 3-IS will be useful as a risk assessment biomarker [22]. Furthermore, for clinical studies to be successful, it would be critical to have a marker that can rapidly identify patients that are at high risk for gut dysbiosis and then triage these patients into future studies aimed at modifying the microbiome.

Weber et al. first developed and studied urinary 3-IS as a marker for gut microbiota disruption in patients undergoing bone marrow transplant. They showed that low levels of urinary 3-IS within the first 10 days after bone marrow transplant was associated with higher transplant-related mortality and, similar to our study, low urinary 3-IS was also associated with worse overall one-year survival [11]. In contrast to our study, they showed that high urinary 3-IS was associated with Lachnospiraceae and Ruminococcaceae of the Clostridia class, and low levels of urinary 3-IS were associated with the Bacilli class [11]. Similarly to our study, they showed only a moderate correlation to alpha diversity (r = 0.25 with the Simpson Index). Tryptophanase is common and 3-IS and its metabolites are associated with a wide array of microbes [9]. Prior 3-IS studies have shown that prolonged use of systemic antibiotic treatment was associated with lower urinary 3-IS levels [11, 23], which was not demonstrated in this ICU cohort, where immunosuppression was the only intervention that was associated with significant change of urinary 3-IS levels. This may reflect receipt of antibiotics in the recent past among our patients, before collection of the baseline sample.

A variety of biomarkers in sepsis and critical care have been evaluated, including cytokines, cell markers, vascular endothelial damage markers, and acute phase reactants, but none of them have been sensitive or specific enough to become routinely used in clinical practice [24, 25]. The most widely studied marker is probably procalcitonin, which has been shown to be associated with ICU mortality in multiple studies [26, 27] with an AUC ranging from 0.67 to 0.71 [28], which is comparable to the AUC of urinary 3-IS shown in our study. Despite this association, procalcitonin was unable to successfully guide the escalation of diagnostic and therapeutic management of sepsis in a large randomized controlled trial [29]. The difficulty in finding effective biomarkers for sepsis and critical illness may be due to the complex pathophysiology of sepsis and the inherent heterogeneity of ICU patients. Therefore, a combination of organ-specific markers or those used to guide therapies, such as the use of procalcitonin-guided antibiotic de-escalation [30, 31], may be where biomarkers eventually serve a larger role in clinical practice. There are a few biomarkers specific to the gut, such as serum citrulline and intestinal fatty acid-binding protein, which have both been independently associated with mortality in critically ill patients [32]. However, these are markers of enterocyte damage and, unlike urine 3-IS, would be unlikely to indicate if a patient would benefit from microbiome-directed interventions.

There are some strengths to our study. To our knowledge, this is the first study to investigate a potential biomarker specific to the microbiome in an ICU population, which includes a broad spectrum of illness types and severities. It included a detailed overview of patient characteristics associated with low urinary 3-IS levels and compared 3-IS to the well-established SOFA score. This study also has several limitations that are recognized. If urinary 3-IS is to be used as a marker to triage patients into future studies, ideally the initial level at admission could be used, but our study showed that only 3-IS level at 72 hours was related to mortality. Similarly to procalcitonin, it may be the kinetics of urinary 3-IS that best guides management and timing of microbiome interventions in the ICU. Admission urine samples were not available for all patients in the study. Last, although this study included a variety of ICUs, it was relatively small and should be reproduced to confirm generalizability [33].

CONCLUSIONS

In summary, we found that low levels of urinary 3-indoxyl sulfate at 72 hours after ICU admission was associated with a fewer ICU-free days and with one-year but not 30-day mortality. Future, larger studies should test whether urinary 3-IS can identify ICU patients that will benefit from targeted microbiome interventions.

Highlights:

• The normal gut microbiome is disrupted in patients who are critically ill

• 3-indoxyl sulfate is a gut bacteria byproduct that is excreted in the urine

• Evaluate urine 3-indoxyl sulfate as a gut-specific biomarker in the ICU population

Abbreviations:

3-IS

3-indoxyl sulfate

BMT

bone marrow transplant

ICU

intensive care unit

SOFA

sequential organ failure assessment