During the study period, 2,333 patients underwent CABG. 1,945 patients were undergone isolated CABG with general anesthesia. We also excluded 506 patients because of the perioperative support of IABP or ECMO, or intraoperative use of CPB. Finally, 1,439 patients were included in the study. After data pre-processing, 815 patients in total were included. The flow chart for patient selection is shown in Figure 1.
The baseline and perioperative characteristics of the patients are shown in Table 1 (no delirium group vs. delirium group, 710 [87.1%] vs. 105 [12.9%] patients). The delirium group had a higher average age and C-reactive protein level, and lower hematocrit, eGFR, and albumin readings, and more underlying hypertension. The group also had longer ICU and hospital stays and more frequent postoperative acute kidney injury and new onset atrial fibrillation and reintubation. In-hospital deaths numbered 3 (2.9%) in the delirium group and 0 in the no delirium group, but this did not reach the level of statistical significance.
Table 1 Baseline and perioperative characteristics of patients with or without delirium
Characteristics
|
No delirium (n=710)
|
Delirium (n=105)
|
P-values
|
Patients characteristics
|
|
|
|
Age(year)
|
65.2±9.6
|
71.9±8.2
|
<0.001
|
Male sex
|
556(78.3%)
|
74(70.5%)
|
0.08
|
BMI (kg/m2)
|
24.6±3.3
|
24.1±3.1
|
0.13
|
ASA physical status
|
|
|
0.32
|
1
|
19(2.7%)
|
1(1.0%)
|
|
2
|
200(28.2%)
|
25(23.8%)
|
|
3
|
478(67.3%)
|
75(71.4%)
|
|
4
|
13(1.8%)
|
4(3.8%)
|
|
Hypertension
|
456(64.2%)
|
83(79.0%)
|
0.003
|
Diabetes mellitus
|
349(49.2%)
|
57(54.3%)
|
0.33
|
Dyslipidemia
|
268(37.7%)
|
35(33.3%)
|
0.38
|
Myocardial infarction
|
80(11.3%)
|
15(14.3%)
|
0.37
|
Atrial fibrillation
|
50(7.0%)
|
7(6.7%)
|
0.89
|
Chronic kidney disease
|
275(38.7%)
|
43(41.0%)
|
0.66
|
History of stroke
|
452(63.7%)
|
64(63.3%)
|
0.59
|
Left ventricle ejection fraction (%)
|
55.1±11.1
|
53.9±12.5
|
0.30
|
Hematocrit (%)
|
34.8±4.0
|
33.9±4.1
|
0.03
|
Creatinine (mg/dL)
|
1.4±1.7
|
1.5±1.7
|
0.3
|
Estimated GFR (ml/min/1.73m2)
|
73.9±27.3
|
63.5±26.7
|
<0.001
|
Albumin (g/dL)
|
4.0±0.4
|
3.8±0.4
|
<0.001
|
C-reactive protein (mg/dL)
|
0.7±1.4
|
1.1±2.3
|
0.006
|
Intraoperative variables
|
|
|
|
Operation duration (min)
|
362.2±53.4
|
362.6±61.7
|
0.95
|
Anesthesia duration (min)
|
438.1±54.6
|
436.1±68.2
|
0.77
|
Re-do operation
|
7(1.0%)
|
1(1.0%)
|
0.97
|
Emergency
|
76(10.7%)
|
12(11.4%)
|
0.82
|
Op year
|
|
|
0.602
|
2005-2009
|
54
|
6
|
|
2010-2014
|
433
|
69
|
|
2015-
|
223
|
30
|
|
Postoperative medical status
|
|
|
|
ICU length of stay (days)
|
2.3±1.7
|
5.8±7.1
|
<0.001
|
Hospital length of stay (days)
|
9.9±7.1
|
22.1±25.3
|
<0.001
|
Acute kidney injury
|
133(18.7%)
|
34(32.4%)
|
0.001
|
New onset atrial fibrillation
|
146(20.6%)
|
31(29.5%)
|
0.04
|
Reintubation
|
27(3.8%)
|
18(17.1%)
|
<0.001
|
In-hospital death
|
0
|
3(2.9%)
|
|
The values are expressed as mean ± standard deviation or number (%). ASA American Society of Anesthesiologists, BMI body mass index, GFR glomerular filtration rate
The duration and number of intraoperative rSO2 measurements below each cut-off are shown in Table 2. The duration of rSO2 reduction was significantly longer in patients with delirium for the cut-offs of <50% and 45% (for each group, mean duration(SD) of 138.7(202.7) and 64.6(141.5) vs. 100.9(159.6) and 39.3(100.6), p=0.031 and 0.027, respectively). There was a significantly higher proportion of patients with an rSO2 reduction <45% among those with delirium (for each group, number of patients (%) of 228(32.1) vs. 44(41.9), p=0.048). We also calculated number needed to treat (NNT) for the cut-off of 45%. According to the table 2, the control event rate (CER) for delirium is 44/(44+228) = 0.16 and the experimental event rate (EER) for delirium is 61/(61+482) = 0.11. The absolute risk reduction (ARR) = CER-EER = 0.16-0.11 = 0.05, therefore the NNT is 1/(ARR) = 1/0.05 = 20. That is, 20 patients need to be treated to reduce 1 episode of delirium.
Table 2 Comparison of intraoperative rSO2 between delirium and no delirium group
|
rSO2
|
No delirium (n=710)
|
Delirium (n=105)
|
P-values
|
Mean; %
|
55.5±6.8
|
54.8±7.74
|
0.32
|
Minimum; %
|
47.6±8.1
|
46.7±8.33
|
0.30
|
Mean duration of rSO2 reduction; min
|
|
|
|
<75%
|
451.0±141.7
|
468.0±175.7
|
0.27
|
<70%
|
442.0±147.9
|
459.9±182.1
|
0.28
|
<65%
|
402.3±167.4
|
418.9±195.4
|
0.36
|
<60%
|
318.2±193.4
|
341.9±230.9
|
0.25
|
<55%
|
204.1±196.1
|
231.0±230.3
|
0.20
|
<50%
|
100.9±159.6
|
138.7±202.7
|
0.03
|
<45%
|
39.3±100.6
|
64.6±141.5
|
0.03
|
<40%
|
11.7±49.2
|
18.3±82.1
|
0.26
|
<35%
|
4.0±28.9
|
7.1±50.7
|
0.38
|
Number of patients with rSO2 reduction
|
|
|
|
<70%
|
709(99.9%)
|
105(100%)
|
1
|
<65%
|
703(99.0%)
|
104(99.0%)
|
0.97
|
<60%
|
669(94.2%)
|
98(93.3%)
|
0.72
|
<55%
|
573(80.7%)
|
84(80.0%)
|
0.87
|
<50%
|
407(57.3%)
|
69(65.7%)
|
0.11
|
<45%
|
228(32.1%)
|
44(41.9%)
|
0.048
|
<40%
|
108(15.2%)
|
17(16.2%)
|
0.80
|
<35%
|
41(5.8%)
|
6(5.7%)
|
0.98
|
The values are expressed as mean ±standard deviation for mean, minimum rSO2 and mean duration of rSO2 reduction, number (%) for the incidence of rSO2 reduction. rSO2 regional cerebral oxygen saturation
|
Intraoperative hemodynamic variables are shown in Supplementary table 1 in Additional file 1. Based on the results of an ROC analysis for the mean values of each variable, the cut-off was determined as 68 mmHg, 2.2 L/min/m2, and 64% for MAP, C.I., and SvO2, respectively. The total durations of reduction below the cut-off and minimum values were calculated. For all three variables, the total duration of reduction below each cut-off was significantly longer in the delirium group than the no delirium group (p=0.001), and these cut-off values were selected for a multivariable analysis as categorical variables.
The odds ratio (OR), 95% confidence interval (CI), and p-values of rSO2 for each cut-off are shown in Table 3. The OR and 95% CI were calculated for every 5 min of rSO2 reduction below each cut-off value. Age, sex, hypertension, preoperative hematocrit, eGFR, serum albumin and C-reactive protein level, intraoperative MAP, C.I. and SvO2 reduction below each cut-off of ROC analysis were considered as covariables. There was no multicollinearity between the variables included in the analysis, especially between the intraoperative hemodynamic variables and rSO2 for the occurrence of postoperative delirium. Multivariable logistic regression analysis revealed that the duration of rSO2 below the 50% and 45% cut-offs was significantly associated with postoperative delirium (for every 5 min, adjusted OR 1.007 [95% CI 1.001-1.014] and 1.012 [1.003-1.021]; p=0.024 and 0.011, respectively). Each model showed good fitness (Hosmer-Lemeshow’s goodness-of-fit: p=0.729 and 0.962, respectively). The rSO2 values below 45% for at least one measurement were significantly associated with postoperative delirium, and the model fitness was good (adjusted OR 1.737, p=0.027; Hosmer-Lemeshow’s goodness-of-fit: p=0.923; Table 4). According to the analysis of variance between the two group for the duration of rSO2 reduction <45% or 50%, and delirium, the p-values of linearity were <0.05 and the p-values of deviation from linearity were >0.05, indicating that there is a linear relationship between the duration of rSO2 reduction below 45% or 50% and the probability of postoperative delirium. The duration of rSO2 below 50% and 45% was also associated with postoperative acute kidney injury, a longer ICU stay, and longer hospital stay (Supplementary table 2 in Additional file 1).
Table 3. Unadjusted and adjusted odds ratios of intraoperative reduction of rSO2 of each cut-offs for postoperative delirium
Intraoperative rSO2
|
Unadjusted OR
(95% CI)
|
P-values
|
Adjusted OR
(95% CI)
|
P-values
|
|
Mean
|
0.985(0.957 to 1.014)
|
0.32
|
0.976(0.942 to 1.011)
|
0.18
|
|
Minimum
|
0.987(0.962 to 1.012)
|
0.30
|
0.977(0.948 to 1.006)
|
0.12
|
|
Duration of rSO2 reduction (for every 5 min)
|
|
|
|
<75%
|
1.004(0.997 to 1.010)
|
0.27
|
1.006(0.999 to 1.013)
|
0.12
|
|
<70%
|
1.004(0.997 to 1.010)
|
0.28
|
1.005(0.998 to 1.012)
|
0.14
|
|
<65%
|
1.003(0.997 to 1.009)
|
0.36
|
1.004(0.997 to 1.011)
|
0.24
|
|
<60%
|
1.003(0.998 to 1.008)
|
0.25
|
1.004(0.998 to 1.010)
|
0.16
|
|
<55%
|
1.003(0.998 to 1.008)
|
0.20
|
1.004(0.999 to 0.010)
|
0.15
|
|
<50%
|
1.006(1.001 to 1.011)
|
0.03
|
1.007(1.001 to 1.014)
|
0.02
|
|
<45%
|
1.009(1.001 to 1.017)
|
0.03
|
1.012(1.003 to 1.021)
|
0.01
|
|
<40%
|
1.009(0.994 to 1.025)
|
0.26
|
1.013(0.995 to 1.030)
|
0.15
|
|
<35%
|
1.011(0.986 to 1.037)
|
0.38
|
1.021(0.990 to 1.053)
|
0.19
|
|
Occurrence of rSO2 reduction
|
|
|
|
<70%
|
.
|
1
|
.
|
1
|
|
<65%
|
1.036(0.126 to 8.502)
|
0.97
|
.
|
1
|
|
<60%
|
0.858(0.374 to 1.966)
|
0.72
|
1.460(0.423 to 5.044)
|
0.55
|
|
<55%
|
0.956(0.572 to 1.598)
|
0.87
|
0.935(0.492 to 1.777)
|
0.84
|
|
<50%
|
1.427(0.929 to 2.192)
|
0.11
|
1.599(0.965 to 2.649)
|
0.07
|
|
<45%
|
1.525(1.003 to 2.317)
|
0.048
|
1.737(1.064 to 2.836)
|
0.03
|
|
<40%
|
1.077(0.616 to 1.882)
|
0.80
|
1.236(0.657 to 2.326)
|
0.51
|
|
<35%
|
0.989(0.409 to 2.390)
|
0.98
|
0.839(0.306 to 2.299)
|
0.73
|
|
rSO2 regional cerebral oxygen saturation, OR odds ratio, CI confidence interval
|
|
Table 4. Odds ratios of predictors for postoperative delirium
Variables
|
Multivariable logistic regression
– OR (95% CI)
|
Univariable logistic regression
– OR (95% CI)
|
Age (year)
|
1.093(1.058 to 1.129)
|
1.097(1.066 to 1.128)
|
Sex (Female)
|
-
|
1.512(0.959 to 2.386)
|
Preoperative
|
|
|
Hypertension
|
1.908(1.062 to 3.428)
|
2.101(1.282 to 3.445)
|
Hematocrit (%)
|
-
|
0.943(0.896 to 0.993)
|
estimated GFR
(ml/min/1.73m2)
|
-
|
0.987(0.980 to 0.994)
|
Albumin (g/dL)
|
0.485(0.276 to 0.852)
|
0.384(0.244 to 0.605)
|
C-reactive protein (mg/dL)
|
-
|
1.163(1.044 to 1.295)
|
Intraoperative
|
|
|
MAP < 68mmHg
|
-
|
1.002(1.001 to 1.004)
|
C.I. < 2.2 L/min/m2
|
-
|
1.002(1.001 to 1.003)
|
SvO2 < 64%
|
-
|
1.003(1.001 to 1.005)
|
Occurrence of rSO2 < 45%
|
1.737(1.064 to 2.836)
|
1.525(1.003 to 2.317)
|
OR odds ratio, CI confidence interval, GFR glomerular filtration rate, MAP mean arterial pressure, C.I. cardiac index, SvO2 mixed venous oxygen saturation, rSO2 regional cerebral oxygen saturation
|
A post hoc power analysis was performed for the occurrence of rSO2 reduction <45% with chi-squared test based on the result described in table 2. The analysis revealed a power of this study above 0.95 for the occurrence of rSO2 reduction <45%.
Based on the ROC analysis, the cut-off age for postoperative delirium occurrence was 68. We conducted a subgroup analysis based on this cut-off. Among 815 patients, 398 (48.8%) were under age 68, and delirium occurred in 19 patients (4.8%). Baseline and perioperative characteristics, including intraoperative hemodynamic variables, are shown in Supplementary table 3 in Additional file 1. Based on a univariable analysis, preoperative EF, and albumin and C-reactive protein levels were selected for a multivariable analysis. Supplementary table 4 Additional file 1 shows the duration and number of intraoperative rSO2 values below each cut-off in patients under 68 years of age. The mean and minimum rSO2 values were significantly lower in the delirium group. The duration of rSO2 reduction was significantly longer in patients with delirium for the cut-offs of <55%, 50%, and 45%, and the proportion of patients with an rSO2 reduction below 50% and 45% was significantly higher among those with delirium. These cut-offs were higher than those of the overall group in Table 2. In the multivariable logistic regression analysis, the duration of rSO2 lower than 55%, 50%, and 45% was significantly associated with postoperative delirium (for every 5 min, adjusted OR 1.012, 1.015, and 1.015, p=0.035, 0.006, and 0.024, respectively), as shown in Table 5. However, the model fitness for the cut-off of 55% was not good (Hosmer-Lemeshow’s goodness-of-fit: p=0.022), whereas those for the other cut-offs were good. The area under receiver operating characteristic (AUROC) for prediction models for patients under 68 years of age are shown in Additional file 2. The AUROC for the model without rSO2 was 0.688 (95% CI 0.565–0.816, p=0.007), and improved with rSO2 measurement, up to 0.752 (95% CI 0.640–0.865, p<0.001) with the duration of rSO2 <50%.
Table 5. Odds ratios of intraoperative reduction of rSO2 of each cut-offs for postoperative delirium in patients under age 68
Intraoperative rSO2
|
Unadjusted OR (95% CI)
|
P-values
|
Adjusted OR (95% CI)
|
P-values
|
Mean
|
0.920(0.869 to 0.975)
|
0.004
|
0.927(0.874 to 0.984)
|
0.01
|
Minimum
|
0.934(0.886 to 0.984)
|
0.01
|
0.940(0.891 to 0.992)
|
0.03
|
Duration of rSO2 reduction (for every 5 min)
|
|
|
|
<75%
|
0.999(0.982 to 1.016)
|
0.87
|
0.997(0.979 to 1.016)
|
0.78
|
<70%
|
1.001(0.985 to 1.016)
|
0.95
|
0.999(0.982 to 1.016)
|
0.89
|
<65%
|
1.005(0.992 to 1.018)
|
0.48
|
1.003(0.988 to 1.017)
|
0.73
|
<60%
|
1.008(0.997 to 1.019)
|
0.16
|
1.005(0.993 to 1.018)
|
0.36
|
<55%
|
1.011(1.001 to 1.022)
|
0.03
|
1.012(1.001 to 1.022)
|
0.04
|
<50%
|
1.015(1.005 to 1.025)
|
0.004
|
1.015(1.004 to 1.025)
|
0.006
|
<45%
|
1.016(1.003 to 1.029)
|
0.02
|
1.015(1.002 to 1.029)
|
0.02
|
<40%
|
1.014(0.987 to 1.042)
|
0.3
|
1.010(0.982 to 1.039)
|
0.49
|
<35%
|
1.017(0.980 to 1.057)
|
0.37
|
1.011(0.972 to 1.052)
|
0.59
|
Occurrence of rSO2 reduction
|
|
|
|
<75%
|
.
|
1
|
.
|
1
|
<70%
|
.
|
1
|
.
|
1
|
<65%
|
.
|
1
|
.
|
1
|
<60%
|
.
|
1
|
.
|
1
|
<55%
|
4.970(0.654 to 37.782)
|
0.12
|
4.231(0.551 to 32.480)
|
0.17
|
<50%
|
4.156(1.191 to 14.503)
|
0.03
|
4.013(1.112 to 14.482)
|
0.03
|
<45%
|
2.634(1.034 to 6.709)
|
0.04
|
2.283(0.906 to 6.266)
|
0.08
|
<40%
|
2.662(0.971 to 7.295)
|
0.06
|
2.757(0.980 to 7.757)
|
0.06
|
<35%
|
1.114(0.141 to 8.817)
|
0.92
|
0.989(0.118 to 8.300)
|
0.99
|
OR odds ratio, CI confidence interval, rSO2 regional cerebral oxygen saturation
|
Among 417 patients over 68 years of age, the incidence of delirium was 20.6% (86/417). In the univariable analysis, older age, hypertension, and low preoperative eGFR were significantly associated with postoperative delirium in the old age group. However, there was no significant association between intraoperative reduction in rSO2 and postoperative delirium for all cut-offs in either the univariable or the multivariable logistic regression analysis.