This study was part of a secondary analysis of data collected from 899 patients between November 2017 and April 2019 for the PAWEL study (Patient safety, cost-effectiveness and quality of life: reduction of delirium risk and postoperative cognitive dysfunction after elective procedures in older adults). The study protocol of this randomized, prospective, multicenter study was published previously [15].

For this study, baseline data from patients undergoing surgical procedures including cardiac, vascular, orthopedic and general surgery at five tertiary academic medical centers in Germany, were analyzed excluding the intervention group. Patients aged 70 years or older and scheduled for elective surgical procedures with an expected duration of at least 60 min were eligible for inclusion. Exclusion criteria were limited life expectancy of less than 15 months, insufficient German language skills and a recently diagnosed severe dementia without a health care proxy. To avoid bias, we included the baseline cohort of the PAWEL study, containing only untreated participants in terms of multimodal interventions to avoid a POD, and patients not treated for OSA.

Patients’ demographic and clinical data were recorded at PAWEL-study admission. These included demographic data, preoperative medical history and comorbidities, preoperative physical status as per classification of the American Society of Anesthesiologists (ASA), and risk of OSA assessed using the STOP BANG score. Moreover, perioperative data including premedication, type of surgical procedure and anesthesia, and the cut-to-suture duration time were charted. The latter data were obtained from anesthesia and surgery records.

An internationally established and validated screening tool, the STOP BANG questionnaire, was used to preoperatively detect patients at increased risk of sleep apnea [16]. The assessment focuses on questions about the occurrence of snoring, fatigue and observed nocturnal cessation of breathing, and also considers data on blood pressure, body mass index, age, neck circumference, and gender. Thus, the resulting composite score has a maximum of eight points, with a higher score-point value indicating a higher risk of OSA. Chung and co-workers published an alternative scoring model and introduced an improved STOP BANG questionnaire by specific constellations of predictive factors. Three escalation levels were described, which define a low risk at 0 to 2 points, an intermediate risk at 3 to 4 points and a high risk at 5 or more points [17]. In our study, we adopted this classification using a three-group comparison. Because, the correlation between an increasing AHI and an increasing number of risk factors in the STOP BANG screening test with regard to OSA severity has been well investigated in the literature [17]. The screening process was carried out during the PAWEL-study admission and was scheduled not later than 3 weeks before surgery.

S-100B and NSE were obtained as blood-based cerebral ischemia biomarkers. Both blood biomarkers are proteins released in response to injury to central neurons and glial cells. Serum concentrations of NSE and S-100B indicate to which extent hypoxic-ischemic brain injury, for example following cardiac arrest, occurs [18]. Both markers were obtained as preoperative baseline measurements on the day of PAWEL-study admission. The second blood sample was scheduled for the 9th postoperative day. Sample preparation was performed according to standard operating procedures: After standing vertically at room temperature for 10 to 30 min the collection tubes were centrifuged at 4 °C and 2,000 g for 10 min. The serum was aliquoted and stored at -80 °C within 2 h until analysis. NSE and S100B were measured with Elecsys® electrochemiluminescence immunoassays (ECLIA) for quantitative in-vitro detection of biomarkers on a cobas e411 System in Goettingen, Germany according to the manufacturers’ instructions. Lower level of detection for NSE was 0.05 ng/ml, for S100B 0.005 µg/l.

Cognitive function was assessed using two different neuropsychological tests. The first test was the Montreal Cognitive Assessment (MoCA) Nasredinne 2003, which was carried out as a baseline measurement on the day of PAWEL-study admission as well as on the day of discharge, but at the latest on the 10th postoperative day. The MoCA is a brief cognitive screening test for assessing cognitive impairment among older people. The test focuses on multiple cognitive domains including visuospatial ability, executive functions, memory, attention, language, abstraction, and orientation, well-established to detect patients with mild cognitive impairment and discriminate them from cognitively unhindered patients [19]. The test is available in three parallel versions to avoid learning effects.

As a second test, the Nursing Delirium Screening Scale (NuDESC) was carried out on the second and on the 6th postoperative day. The NuDESC is a five-item scale based on observations of the nurse staff assessing the functions disorientation, inappropriate behavior and communication, hallucinations, and psychomotor retardation over a 24-h period [20].

In addition, the Confusion Assessment Method (CAM) assessment was carried out on each postoperative day consecutively from the first until the eight postoperative day. With one positive-CAM assessment during any of these postoperative days, the patient was considered as delirious.

Primary outcome was the occurrence of a new cognitive impairment in terms of postoperative cognitive decline (POCD) or POD after elective surgery based on the MoCA, the NuDESC and the CAM assessment. In addition, blood-based cerebral ischemia biomarkers were obtained and correlated with OSA risk, to find potential biomarkers indicating cognitive decline in the postoperative phase.

Low, intermediate, and high-risk patients for OSA were compared regarding the incidence of postoperative cognitive impairment. Moreover, the results were linked to blood-based cerebral ischemia biomarkers. If more than 10% of the data per patient was missing, the patient was not included in the data analysis.

To avoid bias arising from individual baseline differences in cognitive performance, we calculated the difference between first and second measurement for each patient. The individual performance was compared between groups to take into account intellectual baseline differences between patients.

Statistical analysis was performed using IBM SPSS Statistics version 27.0 for macOS platforms (IBM SPSS Inc., Chicago, IL, United States). The Chi-square test and univariate ANOVA were used to examine differences in demographic data between the three OSA-risk groups.

The univariate ANOVA was used to compare the mean scores of the three OSA-risk groups regarding biomarker levels, NuDESC, MoCA and CAM assessment. To investigate the mean differences between the individual groups, a post-hoc test was carried out using the Tukey test with variance equality. To avoid alpha error inflation due to the high number of tests, the level of significance was set at p < 0.05 for all statistical tests.

A power analysis for the ANOVA test was performed. With a medium effect size with an alpha error of 0.05 and a power of 0.85, a sample size of 122 was calculated.