Is a hyperosmolar pump prime for cardiopulmonary bypass a risk factor for postoperative delirium? A double blinded randomised controlled trial

Abstract

Objective: Postoperative delirium (POD) is common after cardiac surgery. We have previously identified plasma sodium concentration and the volume of infused fluids during surgery as possible risk factors. Both are linked to the selection and composition of the pump prime used for cardiopulmonary bypass (CPB). Present study aims to examine whether hyperosmolality increases the risk for POD. Design: Patients ≥65 years (n = 195) scheduled for cardiac surgery were prospectively enrolled into this double blinded randomised clinical trial. Study group received a pump prime containing mannitol and ringer-acetate (966 mOsmol) (n = 98) vs. ringer-acetate (388 mOsmol) (n = 97) in the control group. Postoperative delirium was defined according to DSM-5 criteria based on a test-battery pre- and postoperatively (days 1–3). Plasma osmolality was measured on five occasions and coordinated with the POD assessments. The primary outcome was the POD incidence related to hyperosmolality as the secondary outcome. Results: The incidence of POD was 36% in the study group and 34% in the control group, without intergroup difference (p=.59). The plasma osmolality was significantly higher in the study group, both on days 1 and 3 and after CPB (p<.001). Post hoc analysis indicated that high osmolality levels increased the risk for delirium on day 1 by 9% (odds ratio (OR) 1.09, 95% CI 1.03–1.15) and by 10% on day 3 (OR 1.10, 95% CI 1.04–1.16). Conclusions: Use of a prime solution with high osmolality did not increase the incidence of POD. However, the influence of hyperosmolality as a risk factor for POD warrants further investigation.

Keywords:

• Cardiac surgery
• cardiopulmonary bypass
• osmolality
• delirium
• mannitol

Introduction

The reported incidence of postoperative delirium (POD) after cardiac surgery ranges between 4.1 and 54.9% [1]. The etiology refers to a series of different patient and treatment associated risk factors [2,3]. Preventive measures are therefore vitally important to minimise the negative effects associated with these risk factors [4,5].

We have experienced a high POD incidence over a 20-year period [2,6]. Most cardiac centres in Sweden use mannitol as a pump prime additive. The use of mannitol internationally is probably less frequent. In a recent survey, the reported usage was below 30% [7]. The effects of a hyperosmolar prime solution from a more general perspective are however scarcely investigated. Research initiatives detected a series of POD related risk factors as the volume of infused ringer solution and changes of the serum sodium level postoperatively [2,5,8,9]. It made us suspect a possible link to the pump prime used for cardiopulmonary bypass (CPB) and the high incidence of POD at our cardiac centre. The inclusion of mannitol supplemented with sodium chloride made the plasma osmolality increase well above 300 mOsmol/kg at commence of CPB [10], which is associated with significant changes of the circulatory plasma volume and possibly also the concentration of electrolytes [2].

We therefore initiated a randomised controlled trial hypothesising that hyperosmolality increases the risk for POD.

Materials and methods

Patient enrolment

Patients were enrolled after individual signed informed consent. The Swedish Ethical Review Authority (2018-500-32M), The Swedish Medical Products Agency (5.1-2018-88647), The European Union Drug Regulating Authorities Clinical Trials Database (2018-002385-39) and The Clinical Trial Unit at Umeå University Hospital approved the study protocol.

Patients aged 65 years or older listed for cardiac surgery at the Heart Centre of Umeå University Hospital were consecutively screened for possible inclusion. Exclusion criteria comprised acute surgery within 24 h, cases requiring profound hypothermia, documented psychiatric disease or dementia, severe visual or hearing problems, documented allergic reactions and finally, not speaking Swedish.

Supervision and safety

The Clinical Trial Unit at Umeå University verified consistency of collected data, study protocol adherence, randomisation and safety aspects. The attending physician reported serious and non-serious adverse events detected within 24 h postoperatively to the European Union Regulating Authorities Clinical Trials (EudraCT) Database (2018-002385-39).

Study design and randomisation

This is a prospective intension to treat double blinded controlled phase IV clinical trial. Patients were assigned either to a study or control group (1:1) using simple random sampling. The pump prime composition in the study group comprised: Ringer’s acetate 1000 mL + mannitol 400 mL (60 g)+sodium chloride 40 mL (160 mmol) and heparin 2 mL (10,000 IU) osmolality = 966 mOsmol vs. Ringer’s acetate solution 1400 mL + heparin 2 mL (10,000 IU), osmolality = 388 mOsmol in the control group. A staff member not assigned to the case completed both randomisation and priming of the heart-lung machine. This disguised the pump prime content from staff members treating the patient.

Intraoperative management

Patient monitoring included radial artery and central venous blood pressure, 5-lead electrocardiography and transcutaneous oxygen saturation. General anaesthesia combined propofol (6–10 mg/kg), fentanyl (1–7 µg/kg), rocuronium (0.6–1.0 mg) and sevoflurane (1 MAC). Patients were normo-ventilated. Systemic blood pressure was controlled by norepinephrine and phenylephrine. Surgery followed case specific standardised methods using St. Thomas II blood cardioplegia for cardiac protection. Heparin was used for anticoagulation monitored by the activated clotting time >480 s. Use of roller pumps and membrane oxygenation for CPB aimed to preserve mean arterial pressure (MAP) >50 mmHg and the mixed venous saturation (S_vO₂) >75% by systemic blood flow adjustments. Target body temperature was 34 °C. Shed blood was retransfused continuously.

Data collection

Patient data used for analyses were collected from different hospital registries. Psychometric test results were logged and thereafter registered digitally.

Analyses of osmolality

Osmolality in plasma was analysed on five occasions: before anaesthesia induction and repeated at commence and termination of CPB and on the 1st and 3rd day postoperatively.

Assessment of postoperative delirium

Assessments were performed preoperatively and repeated after extubation on day 1 (+1) and day 3 (±1) postoperatively. Five persons blinded to group assignment were after formal training assigned to administer the test instruments. The test battery included: (1) The Mini Mental State Examination Second Edition Standard Version (MMSE-2 SV) to assess cognition [11]. (2) The Organic Brain Syndrome Scale (OBS) to assess disturbances of awareness and orientation and fluctuations of cognition and degree of emotional reactions and psychotic symptoms [12]. (3) The Nursing Delirium Screening Scale (Nu-DESC) to assess disorientation, inappropriate behaviour, inappropriate communication, illusions or hallucinations and psychomotor retardation [13]. This is a routine procedure repeated three times every day from admittance to ICU until discharge from hospital. (4) Richmond Agitation Sedation Scale (RASS) to assess degree of agitation or sedation [14]. (5) Glasgow Coma Scale (GCS) to assess level of consciousness [15]. (6) Geriatric Depression Scale (GDS-15) to assess depressive symptoms [16]. (7) Activities of Daily Living (ADL) to assess functional ability based on the Katz [17] and Barthel index [18].

The diagnosis of POD was set according to the DSM-5 criteria [19] by YG, BO and HLC blinded to group assignment. All test results were reviewed independently by YG and BO as an initial step, followed by a concluding joint agreement to finalise the POD diagnosis. In addition, the clinical motor profiles, such as hyper-hypo-mixed delirium, duration and the severity of delirium were assessed. The severity was based on assessments based on Nu-DESC, OBS and MMSE-2 SV scores, rated as mild (n = 1), moderate (n = 2) or severe (n = 3). Since participants were assessed twice postoperatively, the total severity score amounted to 6.

Statistical methods

The hypothesis was tested by comparing the intergroup postoperative incidence of delirium defining our primary endpoint. The secondary endpoint was defined by the risk of developing POD in relation to changes in plasma osmolality.

The POD incidence was analysed in contingency tables comparing the frequency of observed vs. expected frequencies using either the Chi-squared test or Fisher’s exact test, as appropriate. The risk of developing POD related to hyperosmolality was tested using univariate logistic regression presented as odds ratio (OR) as an ad hoc analysis in relation to the original study protocol. Distribution patterns for continuous variables were analysed using histograms, Q–Q plots, measures of skewness and kurtosis, combined with Shapiro–Wilk’s test. Independent group differences were analysed using Student’s t-test for normally distributed variables, elsewise the Mann–Whitney U-test. Categorical data were presented in percent or number of observations; continuous variables with a normal distribution as means ± standard deviation (SD); elsewise as median values with associated interquartile range (IQR). A conservative Bonferroni correction was implemented to compensate for multiplicity. Based on the registered POD incidence in our department, a sample of 200 patients (100 + 100) would be required to verify a POD incidence reduction from 26% to 10% to obtain a power >80% and 2 alpha = 0.05. The null hypothesis was rejected for two-tailed p<.05. The IBM SPSS statistical package version 25 (Armonk, NY) was used for all analyses.

Results

Figure 1 shows a schematic illustration of the patient enrolment. A total of 195 patients were included between April 2019 and June 2020: 98 patients in the study group and 97 patients in the control group. Data collection was uneventful. Eight (n = 8) non-serious adverse events were reported to the EudraCT database: six (n = 6) in the study and two (n = 2) in the control group. None of these deemed to be protocol related. No serious adverse events were recorded.

Figure 1. Schematic diagram of patient enrolment.

Intra- and postoperative observations

Randomisation established two uniform groups presented in Table 1.

Table 1. Preoperative variables.

	Study (n = 98)	Control (n = 97)
Age (years)	73 ± 4	73 ± 4
Female sex (%)	30	25
Body mass index (kg/m^2)	27.6 ± 4	27.1 ± 4
EuroSCORE II	1.7 (1.7)	2.0 (2.1)
Haemoglobin (g/L)	136 ± 14	135 ± 14
Erythrocyte volume fraction	0.40 (0.05)	0.40 (0.05)
Leucocytes (×10^9/L)	6.8 (2.3)	7.1 (2.1)
Platelets (×10^9/L)	220 (67)	228 (68)
NT-p-BNP (pg/mL)	561 (1127)	351 (815)
Sodium (mmol/L)	140 (3)	140 (2)
Potassium (mmol/L)	4.1 (0.5)	4.1 (0.4)
Creatinine (µmol/L)	83 (28)	81 (22)
Hypertension (%)	84.7	78.4
Left systolic ventricular function (%)
Normal	71	67
Mild dysfunction	18	17
Moderate dysfunction	10	12
Severe dysfunction	0	4
Diabetes mellitus (%)
Diet treatment	4	1
Oral medication	10	8
Insulin	3	6
Insulin + oral medication	5	6
Atrial fibrillation (%)	22	17
Cerebrovascular disorder (%)	5	12
Chronic obstructive pulmonary disease (%)	4	3
Education (%)
Elementary school	35	38
Upper secondary school	43	41
University college	22	21
Alcohol use (%)	68	66
Alcohol intake interval (%)
Monthly	30	26
Weekly	35	40
Daily	5	2
Smoking (%)
Non-smoker	94	98
More than 20 cigarettes per year	6	2
Sleep disorder (%)	24	23
MMSE-score	28 (3)	28 (3)
Katz ADL Index = 0 (%)	95	98
Barthel ADL Index = 20 (%)	96	90

MMSE: mini mental state examination.

Central tendency: mean ± SD or median (IQR).

The intra- and postoperative phase showed few intergroup dissimilarities (Table 2). However, the urine output was significantly higher in the study group: 573 (340) vs. 330 (245) mL (p<.001), combined with a lower net positive fluid balance intraoperatively: 1496 (738) vs. 1911 (887) mL (p<.001) and haematocrit level during CPB 30.2 ± 4.0 vs. 32.1 ± 3.7 (p=.001), despite no difference in the preoperative haemoglobin level. Blood component transfusions showed no significant intergroup differences. Study group patients had a slightly higher sodium level on the first postoperative day 139 (3) vs. 137 (3) (p<.001) and lower potassium level 4.1 (0.4) vs. 4.2 (0.4) (p=.02) on the third postoperative day (Table 3). Thirty-day mortality was nil.

Table 2. Intra- and postoperative variables.

	Study (n = 98)	Control (n = 97)	p Value
Surgical procedures (%)			.26
Coronary artery bypass grafting	41	38
Aortic valve replacement	43	36
Coronary artery bypass grafting + aortic valve	11	22
Miscellaneous	5	4
Surgery (min)	179 ± 47	186 ± 45	.29
Cardiopulmonary bypass (min)	87 (43)	92 (44)	.22
Aorta occlusion (min)	60 (33)	66 (36)	.37
Net positive fluid balance intraoperatively (mL)	1496 (738)	1911 (887)	<.001
Mean arterial pressure cardiopulmonary bypass (mmHg)	63 (7)	63 (8)	.52
Mixed venous oxygen saturation CPB (%)	80 ± 3.4	79 ± 3.8	.61
Haematocrit cardiopulmonary bypass (%)	30.2 ± 4.0	32.1 ± 3.7	.001
Activated clotting time (s)	566 (166)	585 (219)	.90
Urine output intraoperatively (mL)	573 (340)	330 (245)	<.001
Urine output (12 h) (mL)	1248 (573)	1155 (448)	.03
ICU 24 h chest drain volume (mL)	640 (495)	650 (410)	.85
Transfusion blood products (%)	43	39	.60
Stroke (%)	0	2	.25
Red packed cells (%)			.94
1–2 units	27	24
3–5 units	10	10
>6 units	4	3
Plasma (%)			.42
1–2 units	9	4
3–5 units	4	2
>6 units	2	2
Platelets (%)			.73
1–2 units	9	6
3–5 units	2	2
Ventilator time (h)	6 (3)	6 (3)	.34
ICU stay (h)	23 (6)	22 (4)	.21
Hospital stay (days)	6 (2)	6 (2)	.93

Central tendency: mean ± SD or median (IQR).

Table 3. Postoperative biochemical analyses.

	Study (n = 98)	Control (n = 97)	p Value
Day 1
Haemoglobin (g/L)	102 ± 14	101 ± 13	.93
Erythrocyte volume fraction	0.31 ± 0.04	0.30 ± 0.04	.51
Leucocytes (×10^9/L)	11.0 (4.5)	10.8 (3.7)	.22
Platelets (×10^9/L)	154 (57)	156 (54)	.53
Sodium (mmol/L)	139 (3)	137 (3)	<.001
Potassium (mmol/L)	4.4 (0.4)	4.5 (0.4)	.08
Creatinine (µmol/L)	76 (32)	78 (23)	.48
Day 3
Haemoglobin (g/L)	93 (15)	90 (15)	.26
Erythrocyte volume fraction	0.28 (0.05)	0.28 (0.04)	.22
Leucocytes (×10^9/L)	9.6 (4)	9.7 (2.9)	.74
Platelets (×10^9/L)	135 (60)	143 (66)	.38
Sodium (mmol/L)	137 (5)	137 (3)	.61
Potassium (mmol/L)	4.1 (0.4)	4.2 (0.4)	.02
Creatinine (µmol/L)	79 (34)	76 (26)	.62

Central tendency: mean ± SD or median (IQR).

Perioperative pharmacological treatment

Almost all patients received paracetamol and oxycodone as a routine premedication and for postoperative analgesia. Medication prescribed preoperatively or other remedies sensitive for interacting with development or treatment of delirium were identified, however, without intergroup deviance. Analyses of administered medication for the perioperative period showed no intergroup differences. Use of inotropes appeared less frequently in the study group 33% vs. 46% (p=.05). Pharmacological treatment of POD was required in 29% of the patients, without any intergroup difference (p=.14).

Postoperative delirium

The overall incidence of POD was 36%, without any intergroup difference (p=.59). Degree of POD severity (graded 1–6) was 2 (3) in the study group vs. 2 (2) in the control group (p = .40). The incidence of POD was generally higher in the study group, however without reaching statistical significance. Psychometric test results are presented in Tables 4 and 5.

Table 4. Psychometric test results.

	Study (n = 98)	Control (n = 97)	p Value
POD (%, [n])	38 [37]	34 [33]	1.00
POD severity (1–6)	2 (3)	2 (2)	.40
POD on day 1 (%, [n])	29 [28]	22 [21]	.27
POD on day 3 (%, [n])	29 [28]	25 [24]	.55
POD requiring pharmacological treatment (%, [n])	14 [14]	10 [10]	.40
Organic Brain Syndrome Scale (%, [n])
Duration day 1–3 (%, [n])			.38
Day 0	37 [14]	38 [12]
Day 1	16 [6]	28 [9]
Day 3	47 [18]	34 [11]
Day 1 (%, [n])
Hyperactive	32 [9]	10 [2]	.09
Hypoactive	68 [19]	81 [17]	.35
Hyper – hypoactive	14 [4]	5 [1]	.38
Emotional	39 [11]	24 [5]	.25
Psychotic	32 [9]	24 [5]	.52
Mixed emotional – psychotic	21 [6]	14 [21]	.71
Degree of severity			.11
Mild	57 [16]	76 [16]
Medium	25 [7]	24 [5]
Severe	18 [5]	0 [0]
Day 3 (%, [n])
Hyperactive	21 [6]	17 [4]	.74
Hypoactive	54 [15]	54 [13]	.97
Hyper – hypoactive	21 [6]	0 [0]	.03
Emotional	18 [5]	25 [6]	.53
Psychotic	46 [13]	46 [11]	.97
Mixed emotional – psychotic	11 [3]	13 [3]	1.00
Degree of severity			.35
Mild	59 [16]	48 [11]
Medium	30 [8]	48 [11]
Severe	11 [3]	4 [1]

Postoperative delirium (POD) ordinal data: median (IQR). Nominal data: percent [n].

Table 5. Psychometrics to verify development of postoperative delirium.

	Study (n = 98)	Control (n = 97)	p Value
Nursing Delirium Screening Scale [0–10] (%)
Score ≥2 day 0	0	0
Score ≥2 day 1	19	13	.31
Score ≥2 day 3	15	18	.68
Mini mental state examination-2 SV [0–30]
Score day 0	28 (3)	28 (3)	.64
Change day 1	0 (4)	–1 (3)	.93
Change day 3	0 (2)	–1 (3)	.97
Geriatric depression scale [0–15]
Score day 0	1 (2)	1 (2)	.48
Change day 1	0 (1)	0 (1)	.68
Change day 3	0 (2)	0 (1)	.48
Richmond Agitation Sedation Scale [–5 to +4] (%)
Day 0			.12
	0	3
Day 1			.98
Score <0	12	11
Score >0	1	1
Day 3			.08
Score <0	4	10
Score >0	3	0
Glasgow Coma Scale [3–15] (%)
Score <15 day 0	0	0
Score <15 day 1	13	9	.37
Score <15 day 3	6	10	.38

Day 0 refers to the preoperative examination. Ordinal data: median (IQR). Nominal data: percent. Reference range within brackets.

POD and plasma osmolality

The risk for POD increases by 9% in the study group (OR 1.09, 95% CI 1.03–1.15) on day 1 and by 10% (OR 1.10, 95% CI 1.04–1.16) on day 3 postoperatively. Patients diagnosed with POD had higher plasma osmolality levels, both on days 1 and 3 after surgery. The plasma osmolality increased to 323 (14) mOsmol/kg in the study group vs. 295 (5) mOsmol/kg (p < .001) in the control group at the start of CPB. These differences persisted at weaning: 309 (8) mOsmol/kg vs. 297 (5) mOsmol/kg (p < .001) and on day 1 postoperatively: 294 (9) mOsmol/kg vs. 288 (7) mOsmol/kg (p < .001). Plasma osmolality returned to normal on day 3: 290 (7) mOsmol/kg. All patients in the study group had plasma osmolality levels above 295 mOsmol/kg during CPB. Of note, this was also detected among 63% of the control patients at weaning from bypass. Refer to Figure 2 for a summary.

Figure 2. Intergroup plasma osmolality levels pre- during and after cardiopulmonary bypass (panel A). Intergroup proportion of patients with osmolality >295 mOsmol/kg (panel B). Osmolality levels for patients with and without postoperative delirium (panel C). Bar chart: median (IQ3). POD: postoperative delirium. **p<.04, Bonferroni corrected, ***p<.001.

Discussion

Use of a hyperosmolar prime solution for CPB did not increase the incidence of POD. Higher osmolality was at the same time more common among POD patients. The associated risk was however low and for that reason unable to outweigh the influence of other risk factors [1]. It may explain, why the incidence did not differ between the two randomised groups.

We therefore believe that a hyperosmolar pump prime should not be used on a regular basis, even if the associated risk for POD seems low. The main reasons are that pump prime alternatives are readily available [7,20]. Further research may change this recommendation.

The hyperosmolar pump prime produced a lower net volume fluid balance, which is believed to protect against POD [2,21,22]. The inclusion of mannitol increased the urine output, which is one explanation for the more favourable fluid balance. The other is related to the osmotic effects generated by the combination of mannitol and sodium chloride. From a physiological viewpoint, this is explained by the shift of water from the interstitial and intracellular compartments into the plasma volume. As a result, it will lower the requirement of extra fluids during CPB to maintain a normal circulatory blood volume. This is indicated by the significantly lower haematocrit level measured during CPB. The negative consequences of using body water as plasma volume expander are the effects it might have on organ function, especially the brain. While commencing CPB, the volume of 1400 mL is introduced intra-arterially in close connection of the cerebral circulation over a time-period less than 3 min, with an osmolality threefold higher than normal. Of note is that hyperosmolality is seldom suspected to cause POD, but should be considered. In the worst scenario, it may lead to osmotic demyelination [23]. It may also shrink glia cells [24], especially if the blood–brain barrier is disrupted, which is common after cardiac surgery [25].

The reported POD incidence for cardiac surgical patients is high [1], with old age as the dominating risk factor [26]. The incidence at 36% encountered in this investigation confirms this and our ambition to reduce the incidence failed. The overwhelming problem is the multifactorial origin associated with POD [1,26–28], which makes it difficult to foresee and identify in the individual case. Validated test procedures are therefore required [6,8,29,30]. One challenge is to detect POD within the hypoactive domain, which often dominates among high risk patients [29,31]. The use of cognitive testing, astute clinical observations and training of staff members are therefore essential. In this study, we strived to adjust our methods by conducting repeated measurements, including a preoperative evaluation to obtain an individual reference and thereafter repeated twice in addition to 24/7 Nu-DESC surveillance [29].

It should be mentioned that we once again observed significant postoperative intergroup deviations of both potassium and sodium chloride levels [2]. It seems likely that these deviations are linked to the relative plasma hyperosmolality observed in the study group.

This randomised clinical trial has several limitations: the study was not powered to statistically verify the marginal POD incidence difference observed between groups. A type II error cannot be excluded. We did not assess cognitive performance prior to patient enrolment, why participants predisposed for delirium may have been included. Results were obtained from a single centre. Age cut-off at 65 years reduces generalisation of findings.

In conclusion, use of a hyperosmolar pump prime did not increase the incidence of POD. To quantify the identified risk of hyperosmolality in relation to POD, further investigations are warranted.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Financial support was provided through a regional agreement between Umeå University and Västerbotten County Council (ALF). This work was supported by Medical Research Foundation at Umeå University, Sweden; Heart Foundation of Northern Sweden; Capio Research Foundation, Sweden; Erik and Anne-Marie Deltofts Foundation, Sweden; PG and Ragnhild Lundgrens Foundation Sweden; Anna Cederbergs Foundation for Medical Research, Sweden; and Strategic Research Area Health Care Science (SFO-V), Sweden.