The relationship between ICU hypotension and in-hospital mortality and morbidity in septic patients
verfasst von:
Kamal Maheshwari, Brian H. Nathanson, Sibyl H. Munson, Victor Khangulov, Mitali Stevens, Hussain Badani, Ashish K. Khanna, Daniel I. Sessler
Current guidelines recommend maintaining a mean arterial pressure (MAP) ≥ 65 mmHg in septic patients. However, the relationship between hypotension and major complications in septic patients remains unclear. We, therefore, evaluated associations of MAPs below various thresholds and in-hospital mortality, acute kidney injury (AKI), and myocardial injury.
Methods
We conducted a retrospective analysis using electronic health records from 110 US hospitals. We evaluated septic adults with intensive care unit (ICU) stays ≥ 24 h from 2010 to 2016. Patients were excluded with inadequate blood pressure recordings, poorly documented potential confounding factors, or renal or myocardial histories documented within 6 months of ICU admission. Hypotension exposure was defined by time-weighted average mean arterial pressure (TWA-MAP) and cumulative time below 55, 65, 75, and 85 mmHg thresholds. Multivariable logistic regressions determined the associations between hypotension exposure and in-hospital mortality, AKI, and myocardial injury.
Results
In total, 8,782 patients met study criteria. For every one unit increase in TWA-MAP < 65 mmHg, the odds of in-hospital mortality increased 11.4% (95% CI 7.8%, 15.1%, p < 0.001); the odds of AKI increased 7.0% (4.7, 9.5%, p < 0.001); and the odds of myocardial injury increased 4.5% (0.4, 8.7%, p = 0.03). For mortality and AKI, odds progressively increased as thresholds decreased from 85 to 55 mmHg.
Conclusions
Risks for mortality, AKI, and myocardial injury were apparent at 85 mmHg, and for mortality and AKI risk progressively worsened at lower thresholds. Maintaining MAP well above 65 mmHg may be prudent in septic ICU patients.
In septic adults exposed to hypotension in the ICU, risks for in-hospital mortality, acute kidney injury (AKI) and myocardial injury were apparent by a mean arterial pressure of 85 mmHg, and for mortality and AKI risk progressively worsened at lower thresholds. Maintaining mean pressure well above 65 mmHg may be prudent in septic ICU patients.
Introduction
Sepsis affects approximately a million people each year in the United States, and many more globally; it is the leading cause of death in intensive care patients. In 2008, an estimated $14.6 billion was spent in the US on hospitalizations for septicemia [1]. The syndrome is caused by a dysregulated inflammatory response to bacterial infections [2, 3]. Among the major risks is end-organ damage consequent to hypoperfusion and cellular/metabolic dysfunction [2, 4, 5], especially renal and myocardial injury. Since hypotension worsens tissue perfusion, it seems likely that some organ injury can be prevented by maintaining a suitable arterial pressure.
Anzeige
Preventing hypotension is therefore a crucial component of sepsis management [5, 6]. The Society of Critical Care Medicine’s Surviving Sepsis Guidelines [3] suggest initially maintaining mean arterial pressure (MAP) > 65 mmHg (higher for older patients and those with cardiovascular morbidities), followed by monitoring via multiple hemodynamic parameters to an endpoint of tissue perfusion [3]. Systolic blood pressure of 100 mmHg or less is a component of the quick Sequential Organ Failure Assessment score (qSOFA) which helps identify adult patients with suspected infections who are more likely to have poor outcomes typical of sepsis [7]. Despite these guidelines, relationships between various durations and depths of hypotension and serious complications remain unclear. The evidence for clearly defining blood pressure targets in septic patients is currently contentious and weak [3]. We, therefore, evaluated associations between hypotension and in-hospital mortality, acute kidney, and myocardial injury in septic patients [8].
Methods
We analyzed in-patient data from the Cerner Health Facts electronic health records database (Kansas City, MO, USA), which includes clinical and administrative data from 720 US hospitals and health systems. Our analysis of de-identified data was determined to be exempt from local institutional review board (IRB) review in advance by Western IRB (Puyallup, WA, USA).
In-patients admitted and discharged between January 1, 2010 and November 30, 2016 were analyzed. The study included adults ≥ 18 years old with a sepsis diagnosis (primary or secondary, any priority) via International Classification of Disease (ICD) 9 and 10 codes as shown in Online Resource 1 and an ICU stay ≥ 24 h. We considered only the initial episode when patients had more than one qualifying hospitalization containing an ICU admission for sepsis within the database.
Patients were excluded when they lacked at least a 6-month database history before the septic hospital admission; had more than a single ICU stay during the index admission; lacked a baseline serum creatinine measurement within 6 months before ICU admission and at least one measurement during the ICU stay; had a history of acute kidney or myocardial injury within 6 months before ICU admission (based on ICD-9 or ICD-10 codes, Online Resource 2); received dialysis within 6 months before ICU admission through the first 24 h within the ICU (time from which outcomes were analyzed); or had fewer than five valid blood pressure readings during each ICU day in which exposure was analyzed. We also excluded patients whose records contained more than two 5-h gaps between MAP recordings or were missing age, sex, diagnosis codes, or medication records for the index hospitalization.
Anzeige
Exact ICU admission times are not recorded in Cerner Health Facts. Consequently, we defined admission time to be when the first laboratory test or medication order was noted in an ICU care setting. Discharge times were available for a proportion of patients, but when missing, we used the order location to estimate discharge time.
Hypotension exposure extended from ICU admission through the first of: ICU discharge, development of an outcome, or 7 days. We selected MAP as our global measure of blood pressure to be consistent with existing literature [6]. Hypotension exposure was characterized by: (1) Time-weighted average of MAP (TWA-MAP) below MAP thresholds of 55, 65, 75, or 85 mmHg. TWA-MAP was calculated as the area below the MAP threshold curve divided by the total time exposure was monitored; (2) cumulative time measured in minutes during which MAP was below absolute thresholds of 55, 65, 75, or 85 mmHg.
We used absolute thresholds because a previous study showed that absolute and relative thresholds were comparably predictive of myocardial injury and AKI [9]. To calculate relative thresholds, baseline MAP is required which could not be reliably determined in this analysis. We used recorded MAP values when available, or estimated MAP from systolic and diastolic pressures using the formula: [(2 × diastolic) + systolic]/3. MAP readings were deemed invalid and excluded if diastolic blood pressure (DBP) ≤ 5 mmHg, DBP ≥ 225 mmHg, or systolic blood pressure (SBP) ≤ DBP + 5 mmHg [9]. An average of 357 MAP readings were available per patient per ICU day.
Outcomes
The primary outcome was in-hospital mortality; secondary outcomes were acute kidney injury (AKI) and myocardial injury. Mortality was defined by a discharge status of “deceased” for the hospital visit. Secondary outcomes were determined from 24 h after ICU admission until the first of: ICU discharge, 7 days, death, or diagnosis of AKI or myocardial injury (Online Resource 3). Outcomes were largely limited to the ICU to maintain proximity to the hypotension exposure.
AKI was defined as stage 1 or higher based upon serum creatinine (SCr) readings according to the Kidney Disease Improving Global Outcomes 2012 guidelines (using criteria for SCr increase over baseline [defined as the lowest reading within 6 months prior, and closest to ICU admission] and with respect to SCr values within 48 h) [10]. Urine output was not used because there were insufficient data in the registry. Myocardial injury was defined by at least one elevated troponin value > 0.03 ng/mL of “Troponin I”, “Troponin T”, or “Troponin” before onset of AKI. Myocardial injury was not evaluated past the date upon which AKI was identified because renal dysfunction might have falsely elevated troponin concentrations.
Statistical analyses
Baseline patient characteristics were summarized via counts and percentages for binary or categorical variables and with means and standard deviations, or via medians and interquartile ranges for continuous variables. For univariate inferences, Chi square test or t test were used, as appropriate. Multivariable logistic regression quantified the relationship between hypotension exposure (TWA-MAP and cumulative time measured in minutes spent below MAP thresholds) and the primary and two secondary outcomes.
Specifically, we created individual logistic regression models, each with one hypotension exposure and one of the outcomes. We assessed the need for restricted cubic splines by plotting deciles and ventiles (20 equal-sized groups) of the hypotension exposure variable versus the mean proportion of the outcome for each exposure and outcome, and looked for any substantial non-linear trend. No substantial non-linear trend was evident so the exposures were modeled as linear predictors.
Table 1 lists all the covariates included in the models to reduce potential confounding. This includes the Acute Physiology and Chronic Health Evaluation (APACHE) III score used to adjust for patient acuity and the Elixhauser comorbidities used to adjust for chronic comorbidities. For the uncommon outcome of myocardial injury, an algorithm that used bootstrapping and stepwise regression was used to determine a parsed model containing a limited set of potential confounders with hypotension exposure and the outcome [11]. To adjust for a potential lack of independence among observations within hospitals, we derived robust (Huber–White) standard errors clustered at the hospital level for all regression models [12, 13]. We plotted the results of the logistic regression models as marginal probabilities of the outcome across the observed range of the hypotension exposure variable to facilitate interpretation of the results.
Table 1
Comprehensive list of potentially confounding variables for MAP threshold groups < 55 mmHg, < 65 mmHg, < 75 mmHg and < 85 mmHg
Variable group
Variable type
Min. MAP < 55 mmHg (n = 3308)
Min. MAP < 65 mmHg (n = 6310)
Min. MAP < 75 mmHg (n = 8039)
Min. MAP < 85 mmHg (n = 8609)
Mortality n (%)**
Mortality n (%)
Mortality n (%)
Mortality n (%)
Yes
No
p value
Yes
No
p value
Yes
No
p value
Yes
No
p value
Gender*
Male
329 (44%)
1079 (42%)
0.522
51 (47%)
2343 (45%)
0.302
592 (48%)
3183 (47%)
0.473
617 (48%)
3481 (48%)
0.521
Race
African American
77 (10%)
234 (10%)
0.243
120 (11%)
502 (10%)
0.240
137 (11%)
705 (10%)
0.380
142 (11%)
801 (11%)
0.592
Caucasian
585 (78%)
206 (81%)
858 (78%)
4164 (80%)
964 (78%)
5400 (79%)
994 (78%)
5777 (79%)
Other
74 (10%)
211 (9%)
97 (9%)
455 (9%)
107 (9%)
586 (8%)
110 (9%)
635 (9%)
Unknown
19 (3%)
48 (2%)
26 (2%)
88 (2%)
28 (2%)
112 (2%)
28 (2%)
122 (2%)
Age (years)*
Mean
70
65
< 0.001
69
63
< 0.001
69
62
< 0.001
69
63
< 0.001
Admission type
Elective
40 (5%)
119 (5%)
0.779
62 (6%)
213 (4%)
0.065
70 (6%)
274 (4%)
0.025
73 (6%)
291 (4%)
0.009
Emergency
671 (89%)
2276 (89%)
975 (89%)
4711 (90%)
1092 (88%)
6173 (91%)
1125 (88%)
6672 (91%)
Trauma Center
5 (0.7%)
12 (0.5%)
3 (0.3%)
10 (0.2%)
4 (0.3%)
11 (0.2%)
4 (0.3%)
11 (0.2%)
Urgent
2 (0.3%)
4 (0.2%)
53 (5%)
257 (5%)
62 (5%)
321 (5%)
64 (5%)
337 (5%)
Unknown
37 (5%)
142 (6%)
8 (0.7%)
18 (0.4%)
8 (0.7%)
24 (0.4%)
8 (0.6%)
24 (0.3%)
Discharge year
2010
3 (0.4%)
23 (0.9%)
0.009
6 (0.5%)
44 (0.8%)
0.009
9 (0.7%)
53 (0.8%)
0.001
9 (0.7%)
55 (0.8%)
< 0.001
2011
31 (4.1%)
79 (3%)
41 (4%)
172 (3%)
49 (4%)
220 (3%)
50 (4%)
232 (3%)
2012
82 (10.9%)
185 (7%)
115 (11%)
383 (7%)
130 (11%)
510 (8%)
134 (11%)
539 (7%)
2013
141 (19%)
434 (17%)
196 (18%)
848 (16%)
215 (17%)
1064 (16%)
223 (18%)
1125 (15%)
2014
183 (24%)
679 (27%)
270 (25%)
1320 (25%)
305 (25%)
1685 (25%)
313 (25%)
1818 (25%)
2015
165 (22%)
584 (23%)
236 (21%)
1202 (23%)
265 (21%)
1592 (23%)
275 (22%)
1733 (24%)
2016
150 (20%)
569 (22%)
237 (22%)
1240 (24%)
263 (21%)
1679 (25%)
270 (21%)
1833 (25%)
Census region
Midwest
93 (12%)
365 (14%)
< 0.001
141 (13%)
772 (15%)
< 0.001
161 (13%)
1054 (16%)
< 0.001
167 (13%)
1137 (16%)
< 0.001
Northeast
308 (41%)
793 (31%)
430 (39%)
1501 (29%)
477 (39%)
1922 (28%)
484 (38%)
2062 (28%)
South
213 (28%)
905 (36%)
336 (31%)
2003 (39%)
388 (31%)
2654 (39%)
404 (38%)
2873 (39%)
West
141 (19%)
490 (19%)
194 (18%)
933 (18%)
210 (17%)
1173 (17%)
219 (17%)
1263 (17%)
Hospital bed size
< 100
44 (6%)
208 (8%)
< 0.001
63 (6%)
443 (9%)
< 0.001
70 (6%)
603 (9%)
< 0.001
74 (6%)
665 (9%)
< 0.001
100–199
91 (12%)
403 (16%)
141 (13%)
871 (17%)
159 (13%)
1149 (17%)
162 (13%)
1232 (17%)
200–299
115 (15%)
496 (19%)
174 (16%)
994 (19%)
192 (16%)
1332 (20%)
197 (16%)
1457 (20%)
300–499
191 (25%)
558 (22%)
296 (27%)
1220 (23%)
344 (28%)
1612 (24%)
356 (28%)
1746 (24%)
500+
314 (42%)
888 (35%)
427 (39%)
1681 (32%)
471 (38%)
2107 (31%)
485 (38%)
2235 (31%)
ICU type
General ICU
472 (63%)
1722 (68%)
0.001
675 (61%)
3465 (67%)
0.001
765 (62%)
4549 (67%)
0.002
791 (62%)
4920 (67%)
< 0.001
Medical ICU
95 (13%)
278 (11%)
156 (14%)
668 (13%)
172 (14%)
880 (13%)
178 (14%)
947 (13%)
Surgical ICU
46 (6%)
181 (7%)
69 (6%)
309 (6%)
75 (6%)
382 (6%)
75 (6%)
407 (6%)
Cardiac ICU
70 (9%)
138 (5%)
99 (9%)
301 (6%)
103 (8%)
392 (6%)
109 (9%)
418 (6%)
Coronary care unit
72 (10%)
234 (9%)
102 (9%)
466 (9%)
121 (10%)
600 (9%)
121 (10%)
643 (9%)
Drugs received
Diuretics received
171 (23%)
341 (13%)
< 0.001
264 (24%)
653 (13%)
< 0.001
291 (23%)
838 (12%)
< 0.001
303 (24%)
897 (12%)
< 0.001
ACE inhibitors received
68 (9%)
232 (9%)
0.946
115 (11%)
453 (9%)
0.066
130 (11%)
594 (9%)
0.044
137 (11%)
648 (9%)
0.028
Beta blockers received
157 (21%)
378 (15%)
< 0.001
252 (23%)
773 (15%)
< 0.001
282 (23%)
1020 (15%)
< 0.001
298 (23%)
1105 (15%)
< 0.001
Calcium channel blockers received
82 (11%)
156 (6%)
< 0.001
136 (12%)
332 (6%)
< 0.001
156 (13%)
465 (9%)
< 0.0001
165 (13%)
519 (7%)
< 0.001
*Modified APACHE III score
Mean
76
65
< 0.001
74
62
< 0.001
73
60
< 0.001
72
59
< 0.001
Serum lactate
No reading available
335 (44%)
125 (49%)
< 0.001
505 (46%)
2745 (53%)
< 0.001
592 (48%)
3729 (55%)
< 0.001
614 (48%)
4077 (56%)
< 0.001
Normal < 2 mmol/L
183 (24%)
813 (32%)
279 25%)
1555 (30%)
304 (25%)
1934 (28%)
309 (24%)
2045 (28%)
Mild 2 to < 5 mmol/L
172 (23%)
406 (16%)
230 (21%)
780 (15%)
250 (20%)
983 (15%)
257 (20%)
1050 (14%)
Moderate 5 to < 8 mmol/L
40 (5%)
57 (2%)
58 (5%)
95 (2%)
61 (5%)
119 (2%)
63 (5%)
123 (2%)
Severe ≥ 8 mmol/L
25 (3%)
20 (0.8%)
29 (3%)
34 (0.7%)
29 (2%)
38 (0.6%)
31 (2%)
40 (0.6%)
Elixhauser index
Mean
18
13
< 0.001
18
13
< 0.001
18
13
< 0.001
18
12
< 0.001
Payer
Commercial
102 (14%)
403 (16%)
< 0.001
145 (13%)
883 (17%)
< 0.001
165 (13%)
1176 (17%)
< 0.001
168 (13%)
1293 (18%)
< 0.001
Medicaid
53 (7%)
334 (13%)
93 (8%)
696 (13%)
114 (9%)
917 (14%)
116 (9%)
989 (14%)
Medicare
494 (65%)
145 (57%)
700 (64%)
2839 (55%)
780 (63%)
3668 (54%)
807 (63%)
3907 (53%)
Other
48 (6%)
141 (6%)
70 (6%)
317 (6%)
74 (6%)
415 (6%)
75 (6%)
461 (6%)
Unknown
58 (7%)
225 (9%)
93 (9%)
474 (9%)
103 (8%)
627 (9%)
108 (9%)
685 (9%)
Teaching status
Yes
516 (68%)
1620 (64%)
0.014
723 (65.7%)
3161 (61%)
0.002
803 (65%)
4057 (60%)
< 0.001
824 (65%)
4342 (59%)
0.000
Urban/rural status
Urban
553 (73%)
2029 (80%)
< 0.001
270 (25%)
991 (19%)
< 0.001
930 (75%)
5513 (81%)
< 0.001
963 (76%)
5965 (81%)
< 0.001
Hospital acute status
Acute
755 (100%)
2553 (100%)
1101 (100%)
5209 (100%)
1235 (100%)
6803 (100%)
1273 (100%)
7335 (100%)
Elixhauser comorbidities
Congestive heart failure
316 (42%)
784 (31%)
< 0.001
438 (40%)
1477 (28%)
< 0.001
477 (39%)
1855 (27%)
< 0.001
494 (39%)
1953 (27%)
< 0.001
Valvular disease
163 (22%)
444 (17%)
0.009
234 (21%)
832 (16%)
< 0.001
259 (21%)
1051 (16%)
< 0.001
264 (21%)
1114 (15%)
< 0.001
Pulmonary circulation disease
139 (18%)
306 (12%)
< 0.001
190 (17%)
596 (11%)
< 0.001
211 (17%)
740 (11%)
< 0.001
218 (17%)
777 (11%)
< 0.001
Peripheral vascular disease
174 (23%)
490 (19%)
0.020
253 (23%)
950 (18%)
0.000
276 (22%)
1218 (18%)
< 0.001
285 (22%)
1302 (18%)
< 0.001
Paralysis
68 (9%)
301 (11%)
0.033
101 (9%)
588 (11%)
0.041
117 (10%)
726 (11%)
0.203
125 (10%)
766 (10%)
0.495
Other neurological disorders
250 (33%)
875 (34%)
0.554
368 (33%)
1702 (33%)
0.630
413 (33%)
2202 (32%)
0.470
422 (33%)
2352 (32%)
0.456
Chronic pulmonary disease
333 (44%)
1062 (42%)
0.220
510 (46%)
2159 (42%)
0.003
565 (46%)
2825 (42%)
0.006
589 (46%)
3019 (41%)
0.00
Diabetes w/o chronic complications
164 (22%)
604 (24%)
0.268
242 (22%)
1223 (24%)
0.285
273 (22%)
1581 (23%)
0.376
286 (23%)
1721 (24%)
0.430
Diabetes w/chronic complications
117 (16%)
341 (13%)
0.135
161 (15%)
695 (13%)
0.260
184 (15%)
950 (14%)
0.392
189 (15%)
1051 (14%)
0.635
Hypothyroidism
178 (24%)
602 (24%)
0.998
251 (23%)
1182 (23%)
0.939
281 (23%)
1469 (22%)
0.371
286 (23%)
1556 (21%)
0.321
Renal failure
210 (28%)
535 (21%)
< 0.001
304 (28%)
1007 (19%)
< 0.001
336 (27%)
1342 (20%)
< 0.001
344 (27%)
1440 (20%)
< 0.001
Liver disease
138 (18%)
279 (11%)
< 0.001
210 (19%)
587 (11%)
< 0.001
233 (19%)
790 (12%)
< 0.001
241 (19%)
852 (12%)
< 0.001
Peptic ulcer disease excl. bleeding
9 (1%)
28 (1%)
0.827
15 (1%)
53 (1%)
0.314
16 (1%)
64 (0.9%)
0.249
16 (1%)
69 (0.9%)
0.294
Acquired immune deficiency syndrome
6 (0.8%)
27 (1%)
0.523
10 (0.9%)
43 (0.8%)
0.785
11 (0.9%)
58 (0.9%)
0.896
12 (0.9%)
63 (0.9%)
0.769
Lymphoma
29 (4%)
72 (3%)
0.152
52 (5%)
160 (3%)
0.006
59 (5%)
188 (3%)
< 0.001
61 (5%)
200 (3%)
< 0.001
Metastatic cancer
112 (15%)
174 (7%)
< 0.001
154 (14%)
365 (7%)
< 0.001
184 (15%)
458 (7%)
< 0.001
192 (15%)
487 (7%)
< 0.001
Solid tumor without metastasis
73 (10%)
234 (9%)
0.676
110 (10%)
461 (9%)
0.231
119 (10%)
574 (8%)
0.170
122 (10%)
608 (8%)
0.128
Rheumatoid arthritis/collagen vascular diseases
46 (6%)
188 (7%)
0.231
71 (7%)
367 (7%)
0.479
83 (7%)
453 (7%)
0.942
84 (7%)
494 (7%)
0.852
Coagulopathy
285 (38%)
650 (26%)
< 0.001
410 (37%)
1244 (24%)
< 0.001
462 (37%)
1570 (23%)
< 0.001
474 (37%)
1654 (23%)
< 0.001
Obesity
137 (18%)
590 (23%)
0.004
201 (18%)
1186 (23%)
0.001
229 (19%)
1618 (24%)
< 0.001
237 (19%)
1767 (24%)
< 0.001
Weight loss
276 (37%)
730 (29%)
< 0.001
405 (37%)
1360 (26%)
<0.001
447 (36%)
1674 (25%)
< 0.001
458 (36%)
1752 (24%)
< 0.001
Fluid and electrolyte disorders
638 (85%)
1986 (78%)
<0.001
922 (84%)
4007 (77%)
< 0.001
1032 (84%)
5195 (76%)
<0.001
1063 (83%)
5575 (76%)
< 0.001
Chronic blood loss anemia
41 (5%)
122 (5%)
0.467
69 (6%)
227 (4%)
0.007
76 (6%)
285 (4%)
0.002
76 (6%)
303 (4%)
0.003
Deficiency anemias
448 (59%)
1399 (55%)
0.027
644 (59%)
2759 (53%)
0.001
723 (59%)
3516 (52%)
< 0.001
747 (59%)
3744 (51%)
< 0.001
Alcohol abuse
106 (14%)
294 (12%)
0.062
156 (14%)
579 (11%)
0.004
175 (15%)
785 (12%)
0.009
179 (14%)
869 (12%)
0.026
Drug abuse
65 (9%)
272 (11%)
0.103
90 (8%)
583 (11%)
0.003
100 (8%)
772 (11%)
0.001
105 (8%)
842 (12%)
0.001
Psychoses
88 (12%)
353 (14%)
0.123
13 (12%)
750 (14%)
0.044
150 (12%)
990 (15%)
0.025
156 (12%)
1066 (15%)
0.031
Depression
197 (26%)
722 (28%)
0.238
294 (27%)
1464 (28%)
0.346
325 (26%)
1876 (28%)
0.353
339 (27%)
2022 (28%)
0.480
Hypertension
556 (74%)
1782 (70%)
0.042
808 (73%)
3606 (69%)
0.006
914 (74%)
4741 (70%)
0.003
944 (74%)
5140 (71%)
0.004
Because of rounding, categories will not always add to 100%
APACHE III score includes physiology, chronic health investigation, and age variables [18]
*Variables included in the adjustment of regression models for myocardial injury
**Percentages for the variables types in this table are calculated by using a denominator which is the total number of patients within that particular variable group
Anzeige
We conservatively estimated that if the probability of in-hospital mortality was 13% with the hypotension exposure of interest at its mean and the probability of in-hospital mortality was 16% when hypotension exposure was one standard deviation above the mean, then the sample size would need to be 1766 to detect a difference as great as this or larger with 90% power and alpha = 0.05 [14‐17]. These power calculations further assumed a low correlation of 0.2 between the hypotension exposure and other predictors in the model. Consequently, we concluded that the sample size would be more than adequate to detect clinically significant associations with hypotension and the primary outcome of mortality. All statistical analyses were performed using Stata/MP 15.1 for Windows (StataCorp, College Station, TX, USA).
Results
We identified 8782 patients from 110 hospitals after applying all inclusion and exclusion criteria (Fig. 1). The mean (SD) age of the patients was 63 (18) years. Of these, 79% were self-identified as Caucasian and 48% were male. The mean (SD) APACHE III score was 61 (20) [18]. The unadjusted in-hospital mortality rate was 14.6% (n = 1283). Fifteen percent (n = 1315) experienced AKI and 0.7% (n = 63) experienced myocardial injury during the study period (AKI and myocardial injury rates appear low because patients who developed AKI or myocardial injury during the initial 24 ICU hours were excluded, Fig. 1). Table 1 and Online Resource 4 list all the covariates included in the regression models for the outcomes of in-hospital mortality and AKI, respectively. For myocardial injury, the regression models adjusted for hypotension along with age, sex, APACHE III score, and the Elixhauser comorbidities of congestive heart failure, diabetes with complications, and renal failure.
×
The odds ratios with 95% confidence intervals for the regression models with TWA-MAP are graphed in Fig. 2. The primary hypotension exposure of TWA-MAP < 65 mmHg was positively correlated with in-hospital mortality. The analysis indicates that for every one mmHg increase in TWA-MAP < 65 mmHg, the odds of in-hospital mortality increase by 11.4%; 95% CI (7.8, 15.1%); p < 0.001 (Fig. 2 and Online Resource 5). Sensitivity analyses show the odds ratios decreased as the MAP threshold increased from 55 to 85 mmHg. The predicted marginal probabilities of in-hospital mortality across TWA-MAP < 65 mmHg are shown Fig. 3a.
×
×
Cumulative time below a MAP threshold of 65 mmHg revealed that every 2 h (120 min) increased the odds of in-hospital mortality by 3.6%; 95% CI (2.5, 4.8%); p < 0.001 (Online Resource 6). The predicted marginal probabilities of in-hospital mortality for cumulative time of MAP < 65 mmHg showed similar trends to probabilities for TWA-MAP (Fig. 3). Predicted marginal probabilities are shown in Online Resource 7.
Anzeige
The relationship between TWA-MAP and AKI was similar to in-hospital mortality (Fig. 4a). For every one mmHg increase in TWA-MAP < 65 mmHg, the odds of developing AKI increase by 7.0%; 95% CI (4.7, 9.5%); p < 0.001 (Fig. 2). Patients who spent between 6 and 8 h in the ICU with MAP < 65 mmHg had odds of developing AKI 37% higher (95% CI 3, 82%; p = 0.031) compared to patients with no time below MAP of 65 mmHg (Online Resource 6). Although TWA-MAP below 55, 75, and 85 mmHg showed a positive correlation (p < 0.001) with developing AKI, we did not see similar trends for cumulative time below thresholds of 75 and 85 mmHg (Online Resource 6). Here, patients with the longest times below the thresholds tended to have fewer AKI events than the increasingly small number of patients with no time below these thresholds. This apparent contradiction is, in part, due to over-representation of the time below MAP threshold in the longest duration category, as these patient records had more and longer gaps between MAP readings.
×
For every one mmHg increase in TWA-MAP < 65 mmHg, the odds of developing myocardial injury increased by 3.7%; 95% CI (0.3, 7.3%), p = 0.03 (Fig. 2, Online Resource 5). We derived marginal probabilities of developing myocardial injury for TWA-MAP (Fig. 4b) and cumulative hours of MAP below 65 mmHg (Online Resource 7). Unlike the relationship for in-hospital mortality and AKI, there was no significant worsening of myocardial injury at lower MAP thresholds (Fig. 4b and Online Resource 8). A sensitivity analysis that repeated the regression modeling restricted to survivors found similar associations between hypotension at the 65 mmHg threshold and the outcomes of AKI and myocardial injury (data not shown).
Discussion
Given the complexity of defining hypotension exposure, we analyzed both time-weighted average (TWA) and the cumulative time under specific thresholds. TWA-MAP is a comprehensive measure of hypotension exposure because it measures both the degree and the duration below a threshold. Continuous time below a MAP threshold is intuitive but neglects severity of hypotension and is subject to two types of variation: the frequency of MAP readings, and the total time for MAP exposure calculation. Spurious extreme values can occur and distort the results due to (1) gaps between readings (carried forward) that can over-represent times below a given MAP threshold, especially as there were more and longer gaps in patients who spent long periods below various thresholds; and (2) patients with longer ICU stays may have been especially prone to hypotension. To minimize both sources of error, we excluded patients with fewer than five blood pressures recorded per each 24-h period, and those who had more than two gaps exceeding 5 h between readings.
We observed strong associations between the TWA-MAP below various thresholds and in-hospital mortality and kidney injury in septic patients. Substantial mortality risk was evident even among the higher thresholds, and the risk progressively increased as the MAP thresholds decreased from 85 to 55 mmHg. A similar relationship between in-hospital mortality and cumulative time < 65 mmHg was also observed. It is important to note that the TWA-MAP and cumulative time measures below a given threshold are nested and not mutually exclusive (e.g., patients with TWA-MAP < 55 mmHg are included in the analysis of TWA-MAP < 65 mmHg). We, therefore, cannot definitively determine an optimal threshold with this study design alone.
Anzeige
As with in-hospital mortality, we observed that odds of developing AKI was associated with hypotension characterized by TWA-MAP, with the odds of developing AKI being greatest for MAP readings < 55 mmHg and lowest for MAP < 85 mmHg. However, similar trends were not observed in the cumulative minutes of MAP below threshold groups. We theorize that is because we excluded AKI and myocardial injury before and within 24 h of ICU admission and restricted this outcome to the first 7 days of exposure. Also, myocardial injury may be undercounted when routine troponin monitoring is not performed.
Overall, our results are consistent with previous literature with notable exceptions. Prior research found that increasing the MAP from 65 to 85 mmHg with norepinephrine does not significantly affect systemic oxygen metabolism, skin microcirculatory blood flow, urine output, renal function, or splanchnic perfusion—although cardiac index increased [19, 20]. However, a prospective study of thirteen patients with septic shock found that increasing MAP to above 65 mmHg with norepinephrine increased cardiac output, improved microvascular function, and was associated with decreased blood lactate concentrations. The investigators noticed that microvascular responses varied considerably among patients, suggesting that individualization of blood pressure targets may be warranted [21].
Only limited evidence from randomized trials provides guidance on optimal thresholds. Asfar et al. [22] randomized 776 septic shock patients and reported that 28- and 90-day mortality did not differ significantly between those who were treated to reach a target MAP of 80–85 mmHg and those who were treated to reach a target of 65–70 mmHg [19]. However, even in the lower MAP target group the blood pressure was maintained at 70–75 mmHg and authors noticed a lower than expected death rate, which supports our findings using a much larger cohort and suggests a threshold above 65 mmHg may be more appropriate. This study also highlights the complexity of performing randomized trials in this population and the value of our analysis.
We observed that 14.6% of patients with sepsis died during hospitalization over the period from 2010 to 2016. Presumably patients in our cohort were sicker than all sepsis patients given a required ICU stay of at least 1 day. However, two recent European studies found sepsis mortality to range from 8 to 26%. [18, 19] This is in contrast to higher in-hospital mortality of 26% reported from analysis of a German patient population from 2007 to 2013 [23]. Furthermore, Freund and colleagues [24], based on 2016 European hospital data, observed 8% in-hospital mortality in patients with suspected sepsis. These results suggest that outcomes from sepsis may be improving over time.
In general, AKI in critically ill patients affects approximately 40% of the patients at some time during their stay and one third who develop renal injury die within 90 days [16]. In a study consistent with ours, hypotensive episodes of MAP < 73 mmHg were associated with progression of AKI in critically ill patients with severe sepsis [17].
Myocardial injury, measured by troponin elevation, may be as high as 15–25% in ICU patients, but is often unrecognized because routine troponin monitoring remains uncommon [16, 25]. When troponins are routinely monitored in septic ICU patients, only 7% of biomarker elevations happened within 24 h of ICU admission [26]. While our definition of myocardial injury artificially lowered the observed rate by excluding cases within 24 h of ICU admission and after AKI development), we still observed significant association between TWA-MAP < 65 mmHg and myocardial injury. However, it is possible that a raised troponin value is present in the absence of myocardial injury [27], although raised troponin values have been tied with myocardial injury within the septic population [28‐31]. Our results of myocardial injury analysis should be interpreted with caution due to lack of universal troponin screening and diverse troponin tests employed among various U.S. Hospitals.
We report associations between hypotension in ICU patients and both myocardial and kidney injury. However, we report associations which are surely at least to some degree confounded by unobserved baseline patient characteristics. Randomized trials will be required to confirm causal relationships that may benefit from intervention. Another study limitation is our inability to distinguish between untreated hypotension and hypotension that persisted despite treatment—and thus presumably indicated worse sepsis. To address this, we adjusted for medication use and other potential confounders. Nevertheless, unmeasured confounding remains likely. For example, septic patients are always given antibiotics, some of which are nephrotoxic. However, we did not attempt to link specific antibiotics to AKI. Hypotension identification and duration is dependent upon the frequency of recorded blood pressure readings. While some MAP data (up to two 5-h gaps per record) were missing, an average 357 MAP readings were available per ICU day which indicates that the exposure was generally well characterized.
We were also unable to distinguish hypotension that is a marker of severe sepsis from hypotension that directly contributed to organ dysfunction. The distinction is important because interventions to reduce hypotension will only improve the fraction of organ dysfunction that is causally related to blood pressure. Additionally, some treatments for hypotension can themselves provoke organ injury. For example, increased rates of atrial fibrillation are noted with higher vasopressor use [22]. Nonetheless, our results suggest that harm may begin to accrue well above the currently recommended initial threshold of 65 mmHg, and higher for older patients and those with cardiovascular comorbidities [3]. However, the definitive way to answer how the duration of hypotension impacts mortality and other outcomes in critically ill sepsis patients is via a prospective, randomized controlled trial that follows a standard protocol for vasopressor and intravenous fluid use. This study did not examine outcomes post ICU or hospital discharge; therefore the association with mid- to long-term outcomes are unknown. And finally, while our measure of ICU duration is based on the timing of laboratory and medication orders, given their frequency in critically ill patients, this limitation seems unlikely to bias our results substantially.
In summary, the Surviving Sepsis Guidelines suggest keeping mean arterial pressure initially above 65 mmHg, followed by individualized treatment to optimize tissue perfusion. In our analysis, risks for mortality, AKI and myocardial injury were apparent by 85 mmHg, and for mortality and AKI risk progressively worsened at lower thresholds. Until randomized trials show that the relationship between hypotension and serious complications is not causal, it would probably be prudent to keep mean arterial pressure well above 65 mmHg in septic ICU patients.
Acknowledgements
The authors wish to thank Dr. Seungyoung Hwang for his assistance with data analysis. This research was supported by Edwards Lifesciences, Irvine, CA.
Compliance with ethical standards
Conflicts of interest
Drs. Maheshwari and Sessler work as consultants for Edwards Lifesciences. Dr. Khanna consults for La Jolla pharmaceuticals. Drs. Khangulov, Munson and Badani work as consultants for Boston Strategic Partners, Inc. who received funds from Edwards Lifesciences to perform the research. Dr. Nathanson is an employee of OptiStatim, LLC, which received consulting fees from Boston Strategic Partners, Inc.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Mit e.Med Innere Medizin erhalten Sie Zugang zu CME-Fortbildungen des Fachgebietes Innere Medizin, den Premium-Inhalten der internistischen Fachzeitschriften, inklusive einer gedruckten internistischen Zeitschrift Ihrer Wahl.
Mit e.Med Anästhesiologie erhalten Sie Zugang zu CME-Fortbildungen des Fachgebietes AINS, den Premium-Inhalten der AINS-Fachzeitschriften, inklusive einer gedruckten AINS-Zeitschrift Ihrer Wahl.