Influence of the postoperative inflammatory response on cognitive decline in elderly patients undergoing on-pump cardiac surgery: a controlled, prospective observational study

Patients who were aged ≥ 60 years and scheduled for elective on-pump cardiac surgery were enrolled in the cardiac surgery group. To create an age-matched, normative control group, subjects aged ≥ 60 years were recruited from primary care practice. The exclusion criteria in the cardiac surgery group, as well as in the normative controls, were significant dementia (Mini Mental State Examination score < 24), a history of cerebrovascular disease, intracranial pathology or psychiatric disease, regular treatment with benzodiazepines or anti-inflammatory drugs (e.g., steroids or non-steroids), elevated baseline PCT or CRP levels, severe left ventricle dysfunction (ejection fraction < 35%) or any kind of organ failure. Surgery or hospitalization within the last 12 months was an additional exclusion criterion in the control group.

Anaesthesia was performed using a midazolam bolus 0.05 mg/kg intravenously (IV), sufentanyl bolus 0.5 μg/kg IV, propofol bolus 1 mg/kg IV, or atracurium bolus 0.5 mg/kg IV for induction and then propofol 3–5 mg/kg/h continuous IV and sufentanyl boluses for maintenance of anaesthesia, including the cardiopulmonary bypass (CPB) period. Intraoperative monitoring of patients was based on anaesthesia standards extended with arterial blood pressure, central venous pressure, nasopharyngeal temperature and a bispectral index (BIS, Covidien LLC, Mansfield, MA USA) monitor. The depth of anaesthesia was controlled in range 45–60 of BIS for the entire surgery. CPB was provided by a roller-pump (MAQUET HL 20, MAQUET GmbH & Co. KG, Rastatt, Germany) and membrane oxygenator (MAQUET Quadrox, MAQUET GmbH & Co. KG, Rastatt, Germany). The components of CPB prime were 1200 mL of Ringer lactate, 100 mL of mannitol, and 60 mL of sodium bicarbonate 8.4%. The non-pulsatile flow rate of CPB was maintained in the range of 2.2–2.4 L/min/m². The mean arterial blood pressure (MAP) was controlled with noradrenaline or glyceryl trinitrate to retain the target MAP of 60–80 mmHg during CPB. Clinical management of anaesthesia and CPB was based on institutional standards, including the temperature, metabolic targets (α-stat acid-base management) and transfusion triggers.

Continuous propofol IV infusion was administered as postoperative sedation during the mechanical ventilation period in the intensive care unit (ICU). Postoperative analgesia consisted of morphine sulphate IV boluses adjusted to the patients’ requirements and 1 g of paracetamol IV infusion (every 6 h as needed). The treatment of study patients in the ICU and cardiothoracic surgical ward did not involve benzodiazepine.

The assessment of neurocognitive functions was performed on the day before surgery and the seventh postoperative day in the cardiac surgery group by the following test battery: Mini Mental State Examination (MMSE; dementia screening); Trail Making Tests A and B (TMA and TMB, respectively; executive functions: organized visual search, planning, attention, set shifting, cognitive flexibility, and divided attention); Digit Symbol Test (DS; attention, psychomotor speed, coding task, and visual short-term memory); Stroop Colour and Word Test (cognitive flexibility and control, as well as resistance to interference). The Beck Depression Inventory (BDI; validated in native language [27]) and State-Trait Anxiety Inventory (STAI; validated in native language [28]) were used to examine mood states and anxiety levels at the same time points as the neurocognitive assessment. BDI was only performed preoperatively. Subjects in the normative control group were examined and retested after a time interval of 7 days with the same test battery and protocol. All tests were performed in a specified room separated from the cardiothoracic surgical ward and evaluated by one clinical psychologist who was blinded to the inflammatory status of the patients.

After the collection of test-retest data for each individual, the within-subject change in the performance on neurocognitive tests was measured using the Reliable Change Index modified for practice (RCIp) [29]: where r _xx is test-retest reliability coefficient, SD₁ is the standard deviation of the baseline score (normative controls), SE_m is the standard error of measurement (normative controls), SE_diff is the standard error of the difference (normative controls) [29], X₂ is the postoperative test score and X₁ is the preoperative test score (cardiac surgery patients). The practice effect was computed by changes in the mean scores over the test-retest time interval (normative controls). A significant change was considered for an RCIp Z score ≥ ± 1.96 (α = 0.05) in all neurocognitive tests, including the MMSE. POCD was defined by a significant decline in ≥ two neurocognitive tests [10]. The criterion of ≥ 2 scores difference in the MMSE was used to specify the threshold of significance in the comparison of baseline cognitions in the three groups [30, 31]. MMSE scores between 28 and 30 were considered normal cognition, and the range of 24–27 was considered mild cognitive impairment [32].

Venous blood was collected to measure the PCT and CRP levels at the following six pre-specified time points: before the operation and then every 24 h during the first five postoperative days. Blood samples were analysed using the electrochemiluminescence immunoassay (Elecsys BRAHMS PCT, Roche Diagnostics GmbH, Mannheim, Germany) and particle enhanced turbidimetric assay (COBAS INTEGRA C-Reactive Protein Latex, Roche Diagnostics GmbH, Mannheim, Germany) techniques to quantify the PCT and CRP levels, respectively. Concentrations greater than 0.5 μg/L PCT and 5.0 mg/L CRP were considered elevated levels according to their normal values. The inflammatory response was defined as “low” for PCT ≤ 0.5 μg/L or “high” for PCT > 0.5 μg/L measured on the first postoperative day (POD).

Continuous variables were analysed with the Shapiro-Wilk test for normality. Descriptive statistics are presented as the mean ± standard deviation for normally distributed data and the median (interquartile range) for non-normally distributed data. The unpaired t test and Mann-Whitney U test were used for comparisons of group means or medians. The differences in observed frequencies were determined by the χ² test and Fischer’s exact test. Relationships between variables were assessed using the Spearman correlation test. To justify the sample size of this study, we calculated the statistical power of the difference between the two inflammatory responses post hoc, which was 0.7. Statistical significance was defined at p < 0.05 by all tests. Analysis was performed with IBM® SPSS Statistics version 23.0 (IBM® Armonk, NY, USA).

Cardiac surgery is frequently accompanied by non-infective systemic inflammatory response syndrome (SIRS) induced by surgical trauma, CPB, organ ischaemia-reperfusion injuries, a change in the body temperature and endotoxin release [33, 34]. This genetically influenced, complex inflammatory process – depending on its magnitude – can result in multi-organ dysfunction, increasing the risk of mortality in the postoperative period [33, 35, 36]. Numerous inflammatory mediators have been investigated to characterize the inflammatory responses related to cardiac surgery with the purpose of using them as prognostic markers of the complications and outcome [37]. PCT is one of the accepted and widely used inflammatory markers, and it is also applied for follow-up in the cardiac surgery setting [24, 25]. Its use is based on the assumption that proinflammatory cytokines can contribute to PCT induction after surgery [24, 26]. In the case of non-infective SIRS, the PCT concentration increases in the first 24–48 h after cardiac surgery up to its maximum level and returns to the baseline range in the subsequent days [24, 26]. The peak level of PCT seems to be associated with the extent of surgical procedure, duration of aortic cross-clamp and CPB, and the length of operation [24]. Constantly elevated PCT levels in the postoperative period suggest ongoing inflammation due to a possible septic process [20, 24, 38].

In our study, we followed up PCT as a primary and CRP as a secondary inflammatory marker to distinguish the grade of the inflammatory response in the perioperative period. We observed significant differences in the magnitudes of PCT changes in the individuals who were clearly isolated to the LIR and HIR groups. While the main characteristics of the PCT time course in the HIR group were consistent with previous reports as it reached a peak value at 7.71 μg/L on POD1 [35, 38‐40], PCT did not exceed the cut-off value of 0.5 μg/L at any time-point in the LIR group (Fig. 1a). Despite the markedly divergent PCT kinetics in the LIR and HIR groups, they did not differ in the perioperative parameters that are supposed to influence the inflammatory response and course, such as the preoperative statin use, duration of aortic cross-clamp and CPB, length of operation, propofol requirements, number of transfusions and postoperative infection [33, 41‐44] (Table 2). We also investigated the possible factors of immune priming. Coexisting chronic diseases, such as diabetes mellitus (DM) or peripheral vascular disease (PVD), are frequently accompanied by a pro-inflammatory state, which possibly amplifies the inflammatory response in the postoperative period [2]. However, our comparative analysis did not show differences in DM and PVD between the LIR and HIR groups. These results are similar to the findings of previous studies [36, 45], supporting that an evolving complex inflammatory response to stimuli of CPB surgery is primarily determined by the individual reactivity of cytokines and the proinflammatory-anti-inflammatory balance. In the present study, we found a completely different behaviour of the CRP level compared to PCT. The results published elsewhere show that CRP increases after cardiac surgery irrespective of the extent of the surgery or presence of SIRS [24, 26, 39]. This fact makes the interpretation of the CRP time course uncertain. Our results agree with these findings because the peak CRP did not correlate with peak PCT, and it was markedly elevated in all subjects of the cohort and did not vary between the LIR and HIR groups during the postoperative period. Interestingly, the WBC count did correlate with PCT rather than with CRP in the HIR group. Hence, our data confirm that PCT follow-up is appropriate to discriminate the grade of the non-infective inflammatory response related to cardiac surgery.

The occurrence of POCD was 35.7% in this study based on RCIp analysis [29] of neurocognitive tests and the definition of decline in at least two tests. Instead of frequently used fixed cut-off methods (i.e., 20% or 1–2 SD change) [8, 29], we applied RCIp involving the age-matched healthy non-surgical control group to determine the incidence of POCD. RCIp employs statistical change criteria –corrected for measurement error and mean practice effect– to estimate the valid change of performance on a neurocognitive test [8, 29, 46]. There are only a few comparable publications in the literature that apply statistical change criteria methods in a cardiac surgery setting [29, 47‐49]. The observed POCD incidence in our elderly group of patients is in line with their findings (36 vs 33–43%, respectively), which supports the conclusion of a recent investigation of Raymond et al. [29] Analyses that use statistical change criteria can considerably contribute to valid estimation of POCD as they minimize the risk of both overestimation and underestimation of decline in the test performance [8, 29].

The main result of this study was that a direct relationship has not been revealed between the degree of PCT elevation and decline in any neurocognitive test or early POCD. The secondary analysis of our data validated this result as perioperative mood states, and predisposing factors of POCD [2, 6, 47, 50] were similar in the two inflammatory response groups defined by the PCT levels (Tables 2 and 4). Furthermore, we did not observe postoperative complications that affect cognitive function. Numerous investigations have focused on the link between proinflammatory cytokines and cognitive dysfunction related to cardiac surgery in the last two decades [2, 17]. Using animal models, Cibelli et al. [51] and Terrando et al. [52] described a potential link between systemic and hippocampal inflammation through the TNF-α, IL-1β and NF-κB pathways, and the impairment of memory as a consequence of the former processes. Jungwirth et al. also confirmed significant cerebral expression of NF-κB in the hippocampus after CPB surgery in a high-quality randomized controlled animal study; however, it was not associated with the neurocognitive outcome [53]. Clinical trials applying arbitrary cut-off criteria for POCD definition in either the cardiac or non-cardiac surgery setting concluded conflicting results on the relationship between pro-inflammatory cytokines and POCD [54‐56]. In a recent investigation of elective coronary artery bypass patients, Hudetz et al. demonstrated that short- and medium-term cognitive dysfunction was related to elevated postoperative IL-6 and CRP levels [57]. Their results were based on the POCD definition involving data from the normative population and Z score [57]. By contrast, most recently published large randomized clinical trials have conflicting results on the incidence of POCD or postoperative delirium when they used pharmacological anti-inflammatory treatment (i.e., dexamethasone or methylprednisolone) during non-cardiac and cardiac surgery [19, 58, 59]. This result might strengthen earlier assumptions that factors other than the grade of the inflammatory response play key role in the pathogenesis of POCD [53, 56]. Our results support this concept because we could clearly demonstrate that the incidence of POCD measured in this study does not depend on the magnitude of the inflammatory response.

Our investigation has several limitations. First, the study was conducted in a single centre, which influenced the sample size over the study period. The description of the postoperative inflammatory response was based on the PCT, CRP and WBC count measurements, and we did not involve pro-inflammatory cytokines in the analysis for further specifications. Our study did not strictly adhere to the Statement of Consensus on Assessment of Neurobehavioral Outcomes after Cardiac Surgery [60] in terms of the applied neurocognitive tests. This study was designed for the short interval outcome measure that aimed to assess the early changes in cognitive function after cardiac surgery, which restricts the interpretation of our results. Considering the post hoc statistical power of this observational study, our presented results are preliminary results, while these data are not confirmed by further investigations.