2.1 Ethics and Patients
This is a prospective monocenter parallel randomized study conducted in St Antoine University Hospital (Assistance-Publique. Hôpitaux de Paris). All methods were carried out in accordance with relevant guidelines and regulations and with CONSORT recommendations . Ethical committee approval for this study (Ethical committee n° 13887-P120702) was provided by the Ethical Committee: CPP Comite de Protection des Personnes Ile de France V, 184 rue du Fbg St Antoine, Paris, France (Chaiperson Prof JJ Boffa) on 2 April 2013. The study was registered in ClinicalTrials.gov (Clinical Trials identifier: NCT 02036736).
Patients less than 75 years old, undergoing hip or knee arthroplasty under general anesthesia were included in the study. Patients with preoperative dementia (defined as a Mini Mental State evaluation (MMS)  of 24 or less), unable to perform the cognitive tests, or who received preoperative psychotropic agents, as well as obese patients (BMI >35kg.m-2), patients with chronic alcoholism or addiction were not included.
Definitive inclusion was decided by the anesthesiologist in charge of the patient on the pre-anesthetic visit the day before surgery. The information was given and the consent form was signed at that time.
The randomisation sequence was generated electronically with nQuery (version 6.01). Enrolment was done by clinicians at the operating room. After enrolment, treatment assignment was done with a secure study website (Cleanweb, Telemedicine Technologies, Boulogne- Billancourt, France) after verification of eligibility and consent status. The anesthesiologist, responsible for enrollment and care at the operating room, was the only one knowing the allocation arm of the treatment. He wasn’t involved in judgment criteria measurement thereafter. Access profiles to the e-CRF have been limited depending on the function of the investigator (evaluator vs anesthesiologist).
Depending on the randomization, the anesthesia maintenance was provided either by Desflurane (Group D) or Propofol in TIVA mode (Group P).
2.3. Anesthetic protocol
No anxiolytic premedication was given to the patient before surgery. Anesthetic induction was performed with Propofol + Sufentanil + Atracurium. Patients had standard monitoring including depth of anesthesia using the Bispectral (BIS®) index. Hypothermia was prevented by using warming blankets.
All patients were intubated and ventilated with a mixture of O2/N2O: 50/50%. Fluid loading was achieved with crystalloids and/or colloids depending on requirements.
According to randomization, patients were allocated to receive either Desflurane (Groupe D) or Propofol (Groupe P) for anesthesia maintenance.
Induction with a bolus of Propofol 2-3 mg/kg
Maintenance with a closed circuit of Desflurane with minimal alveolar concentration adapted to maintain a BIS value between 40 and 60.
Target controlled administration of Propofol at 2 and 4 µg/ml to be adjusted in accordance with 40<BIS<60
Supplemental boluses of Sufentanil and Atracurium were given as required. At the end of surgery (T0), the patient was transferred to the post-anesthesia care unit (PACU). Tracheal extubation was carried out when the patient was conscious, with a respiratory rate above 12.min-1, a core temperature > 36°C, and without residual muscle weakness (residual curarization was assessed with Double-Burst Stimulation and antagonized if necessary).
Post-operative pain intensity at rest was evaluated using the Numerical Rating Scale (NRS) with 0=no pain and 10=maximal imaginable pain intensity. Post-operative analgesia was multimodal. The use of locoregional techniques for post-operative analgesia was encouraged (nerve block, trunk block +/- placement of a perineural catheter +/- wound infiltration). During the stay in PACU, if NRS > 3, morphine was administered by titration (bolus of 1 mg IV repeated every 5 minutes until NRS at rest < 3).
After arrival in the PACU, the Aldrete score was checked every 5 minutes. Once the score of ≥9/10 had been attained, the cognitive tests were carried out for a second time.
The data from these tests was collected by the same investigator as the day before surgery in the case report form. NRS measurement of the pain was done at the same time.
2.4. Cognitive assessment
Preoperatively, patients had a measurement of their anxiety level using the Amsterdam Preoperative Anxiety Information Scale . Educational status was registered.
Cognitive tests were performed by a blinded anesthesiologist. The same anesthesiologist made the preoperative and postoperative assessments. Cognitive tests were chosen on the basis of experimental validation and feasibility criteria. Because the process of cognition is multidimensional, it is mandatory to have several different tests exploring multiple distinct components [17,18]. In this perspective, it was chosen to use, the Digit Symbol Substitution Test (DSST) , the Stroop colour word interference test [20,21], and the Visual Verbal Learning Test . The DSST was derived from the Wechsler adult intelligence scale: On a sheet of paper with a code indicating 9 letters corresponding to 9 digits, the patient must fill out horizontal rows with letters associated with empty cells in 90 sec. In the word and colour interference test (Stroop colour word interference test): the patient reads a list of words indicating colours (task 1), then gives the name of the colors in a list of colored rectangles (task 2). Finally, the patient must read words indicating one colour with the word printed in a different colour (task 3). Patients have 45 seconds to complete each task. The number of correct words was counted. The Visual Verbal Learning Test is a memory test that explores the immediate and long-term recall of a list of 10 words. All tests were affected in the same way by cognitive dysfunction.
In accordance with guidelines on how to conduct a multidimensional cognitive evaluation, an overall score that takes into account inter-individual variability and learning effect, in relation to the standard deviation of the population was calculated (Z score) .
For any test, the average performance of a population is diminished by the pre-operative control value and divided by the standard deviation for the variation in the population, thereby giving a measurement of the magnitude of the deviation from the reference with appropriate sign. Signs were adjusted to assure that deterioration corresponds to a negative score for all tests.
The Z-scores for all tests can be summarized, calculating a combined Z-score that is calculated as the sum of all Z-scores divided by the standard deviation for the sum Z-scores. In our case, cognitive dysfunction was defined as a combined Z-score <-2, or at least 2 Z-scores for single test parameters <-2 .
2.5. Criteria of evaluation
The main judgement criterion was Z combined scores at the time Aldrete score ≥9.
As secondary criteria, each test was analyzed separately in order to evaluate its sensitivity in screening for post-operative cognitive and psychomotor dysfunction. These analysis were performed using Z test for each individual test but also by calculating the difference between pre- and postoperative assessments. Moreover, intraoperative parameters, such as sufentanil consumption and BIS value was registered, as well as time interval between end of surgery and tracheal extubation, time interval between tracheal extubation and Aldrete score ≥ 9, pain intensity and opiate consumption in PACU. Finally, patient’s satisfaction was assessed with a 5 points categorical scale.
2.6. Statistical analysis
Statistical analysis was performed according to published guidelines by the International Study of Postoperative Cognitive Dysfunction (ISPOCD) group .
Zscores between the two groups were compared by a t-test. In each group, Stroop, DSST and VLT scores between D0 and D1 were compared by paired t-tests or conditional logistic regressions when the assumption about symmetric distribution failed.
Calculating the required sample size was complex as it depends on the tests selected and the way in which they are processed. To date, no published works have used the same battery of tests as in our evaluation context. Given the pilot data obtained in our department, the difference in combined Z-score was 1.1. It seemed that a total number of 30 patients per group would make it possible to highlight a difference of 20% (i.e. 0.9) in the Z score with an α risk of 5% and a β risk of 10%. Enrolled patients that didn't participate further in the study, or that couldn't respond appropriately to postoperative tests were excluded for final analysis. Results are presented in medians [interquartils] or mean±SD. The threshold for statistical significance was set at P<0.05.