Association between anaesthesia-related factors and postoperative neurocognitive disorder: a post-hoc analysis

This study is a post-hoc analysis of the PICNIC study (PostoperatIve Cognitive dysfunctioN In elderly Cancer patients [PICNIC], trial number NL31486.042.10), a prospective observational study conducted at the University Medical Center Groningen (UMCG, Groningen, NL) [14‐19]. The PICNIC study was approved by the Medical Ethical Committee of the UMCG and data collection was conducted in accordance with the Declaration of Helsinki. All subjects provided written informed consent.

The cohort of the PICNIC study consisted of patients aged ≥ 65 years, scheduled for surgery for a solid malignant tumour at the UMCG between 2010 and 2014 (see Fig. 1). Patients were excluded from the PICNIC study if they had a physical condition potentially impeding compliance with the study procedures, for instance severe visual or auditory impairment, recent stroke or preoperative cognitive deficits. Furthermore, patients were excluded if their understanding of the Dutch language was insufficient.

Prospectively, the following data were collected: age, sex, BMI, tumour type, disease stage, and surgical characteristics. The Groningen Frailty Index (GFI) was used preoperatively to assess patients’ condition with regards to the physical, cognitive, social and psychological domains [20]. Furthermore, Charlson Comorbidity score (CCS) and Mini-Mental State Examination (MMSE) were used preoperatively to classify comorbidities and measure cognitive impairment, respectively [21, 22]. Anaesthesia and procedure related relevant intraoperative data were recorded perioperatively. Details of complications occurring up to 30 days postoperatively were recorded and classified according to the Clavien-Dindo Classification (POD excluded).

To analyse the association between anaesthesia and cognitive function at 3 months postoperatively, several factors were taken into account. First, the association between type of anaesthesia (general versus combined general and neuraxial anaesthesia) and pNCD was analysed. Second, the association between depth of anaesthesia quantified as the bispectral index (BIS) and pNCD was assessed. We analysed the effect of mean BIS between induction and end of surgery, the total time BIS < 45, the percentage time BIS < 45, the total time BIS > 60 as well as the percentage time BIS > 60. A BIS < 45 was chosen as a cut-off as it is a commonly used definition for deep anaesthesia and has been shown to increase the risk of adverse postoperative outcomes [23, 24]. A BIS > 60 was chosen as an upper limit, because it is usually recommended to avoid values higher than 60 [25]. Furthermore, the risk of having pNCD was compared between patients with a mean intraoperative BIS < 45 and patients with a mean intraoperative BIS > 45. Additionally, the effects of different sedatives and analgesics used during induction and maintenance of anaesthesia were examined. Other factors that were taken into account included the duration of anaesthesia, the total time mean arterial pressure (MAP) < 70, the area under the curve (AUC) MAP < 70, and if the patient received additional vasoactive medication during anaesthesia. Blood pressure and BIS were only included in the analysis if data was available for at least 85% of the total duration of surgery.

Cognitive function was evaluated at baseline (2 weeks before surgery) and at 3 months postoperatively using different neuropsychological tests. Three cognitive domains were assessed: memory, executive functioning and information processing speed. Memory function was assessed using the Dutch versions of the Rey’s Auditory Verbal Learning Test (RAVLT) immediate recall and the RAVLT delayed recall [26]. The outcome of the RAVLT was defined by the total number of words (maximum of 75) correctly remembered during the five immediate recall trials (lowest score, 0; highest score, 75) and the total number of words (maximum of 15) remembered at the delayed recall trial (lowest score, 0; highest score, 15). For assessment of executive function, the Ruff’s Figural Fluency Test (RFFT) and the Trail Making Test part B (TMT-B) were used [27, 28]. Performance on the RFFT was expressed by the total number of unique designs drawn in parts 1 to 5 (lowest score: 0; highest score: 175). For the TMT-B, the number of seconds it took to complete the test was used as the outcome variable (lowest score: 0; highest score: 480). Performance with regards to information processing speed was determined using the Trail Making Test part A (TMT-A) [27]. The number of seconds it took to complete the TMT-A (lowest score, 0; highest score, 480) was used as the outcome variable.

Postoperative NCD was defined as a decline ≥ 25%, relative to the baseline score, in the scores of ≥ 2 of the aforementioned cognitive tests at 3 months postoperatively. In addition, change in score on postoperative cognitive tests as compared to preoperative tests was expressed as a continuous variable by means of z-scores, with a positive z-score indicating an overall increased outcome on cognitive tests and a negative score indicating a decline. This method takes into account a general deterioration in all tests as well as a substantial deterioration in only one or some tests and enables the analysis of the overall change in scores on the postoperative cognitive tests as compared to the preoperative tests.

As the present study is a secondary analysis of a pre-existing database, no power calculation was conducted for this analysis and the sample size was based on available data. Data with regards to patient characteristics and anaesthesia-related factors are presented as frequencies with percentages for categorical and ordinal variables. Median and interquartile range (IQR) are used for continuous, not normally distributed data and mean and standard deviation (SD) are used for continuous, normally distributed data. Univariate and multivariate logistic regression analysis were performed to analyse the association between patient characteristics and pNCD as well as between anaesthesia-related factors and pNCD. Odds ratios (ORs) and their corresponding 95% confidence intervals (95% Cis) are reported. Variables with P values of < 0.15 on univariate analyses were included in the multivariate analysis. Hosmer-Lemeshow goodness-of-fit test was used to evaluate model fit. Furthermore, change in performance on each cognitive test was assessed as a continuous variable by using a z-score. For this purpose, a change score (postoperative minus baseline score) was calculated for each separate test and each patient. For the Trail Making Test the sign was corrected, so that a higher z-score indicated cognitive improvement. The z-score for each patient and each test was subsequently calculated by dividing the change score by the standard deviation of the group’s mean change score for the corresponding test. Lastly, per patient, a combined z-score was derived by adding the z-scores of the individual tests, enabling cognitive function to be expressed as a continuous variable with a positive value indicating a postoperative improvement in cognitive tests. The association between anaesthesia-related factors and change in cognitive function as expressed by the z-scores was analysed in a univariate analysis using Spearman’s rank correlation coefficient for continuous explanatory variables and using independent samples t test and ANOVA tests for nominal explanatory variables with 2 or > 2 categories respectively. A p-value of < 0.05 was considered statistically significant. All analyses were performed using the Statistical Package for Social Sciences: SPSS version 20.0.0 (SPSS Inc., Chicago, DL USA).