Determinants of post-operative cognitive decline in elderly people

: Background : Surgery and anesthesia can result in temporary or permanent deterioration of the cognitive functions, for which causes remain unclear. Objectives: In this pilot study, we analyzed the determinants of cognitive decline following a non-emergency elective prosthesis implantation surgery for hip or knee. Design: Prospective single-center study investigating psychomotor response time and changes in MoCA scores between the day before (D-1) and 2 days after (D+2) following surgery in Lariboisière Hospital (Paris, France). Participants: 60 patients (71.9±7.1-year-old, 72% women) were included. Measurements: Collected data consisted in sociodemographic data, treatment, comorbidities and type of anesthesia (local, general or both). Furthermore, we evaluated pain and well-being before as well as after the surgery using point scales. Results: Post-operative (D+2) MoCA scores were significantly lower than pre-operative ones (D-1) with a difference of median[IQR]=2[4]pts (p<0.001), we found no significant difference between locoregional and general anesthesia. Pre-operative benzodiazepine or anticholinergic treatment were also associated to a drop in MoCA scores (p=0.006). Finally, the use of ketamine during anesthesia (p=0.043) and the well-being (p=0.006) evaluated before intervention, were both linked to a reduced cognitive impact. Conclusion: In this pilot study, we observed a post-operative short-term cognitive decline following a lower limb surgery. We also identified pre and perioperative independent factors linked to cognitive decline following surgery. In a next stage, a larger cohort should be used to confirm the impact of these factors on cognitive decline.


INTRODUCTION
Post-Operative Cognitive Decline (POCD) is a major cause of mortality and morbidity costing over $150 billion dollars yearly in health care expenses in the United States (Braunwald, et al. 2001).POCD includes postoperative delirium, NeuroCognitive Disorder (NCD) and delayed NeuroCognitive Recovery (Mahanna-Gabrielli, et al. 2019).NCD definition from the DSM-V consists in a significant cognitive decline from a previous level of performance, diagnosed at least 30 days after the surgery and assessed by standardized neuropsychological testing.Delayed NeuroCognitive Recovery are tested using the same criteria as for NCD except that the diagnosis window must be less than 30 days after the intervention (Mahanna-Gabrielli, et al. 2019).
However, clear negative consequences of anesthesia on POCD are still debatable (Rasmussen, et al. 1999, Dokkedal, et al. 2016).Some papers speculate that POCD in the elderly is attributable to per-operative use of benzodiazepines (BDZ) and anticholinergic (AC) drugs (Pandharipande, et al. 2006, Pratico, et al. 2005), while others suggest that POCD would result from a pre-existing cerebral fragility (Wildes, et al. 2019, Dokkedal, et al. 2016).Taking into account the ageing of the population linked to a growing demand for surgery, the identification of factors increasing the risk of POCD or more severe cognitive dysfunction is needed.
Diagnosing NCD and delayed NeuroCognitive Recovery requires preoperative cognitive status along with additional stages into the routine care.Indeed, a minimal setup depends upon establishing a baseline of cognitive functions prior the surgery, then on the administration of at least a second neuropsychometric test in the post-operative period (Tsai, Sands and Leung 2010).Tools for quick assessment of POCD are for example the Confusion Assessment Method (CAM), Mini Mental State (MMS) or the Montreal Cognitive Assessment (MoCA) (Fritz, et al. 2016, Nasreddine, et al. 2005).
In particular, the MoCA test evaluates several cognitive functions which impairment is a common mechanism shared by most of POCD, including memory, concentration and verbal abstraction among others.Identifying POCD risk factors (drugs, comorbidity, surgical or anesthesia setups) might help targeting patients that should be followed-up, leading to a personalized care.This prospective pilot study aims to investigate the cognitive decline following a non-emergency elective hip or knee replacement, and to identify independent risk factors associated with it.Additionally, we search for factors linked with persistent cognitive decline evaluated six weeks after the surgery.

Study population
Between February and September 2017, patients from the orthopedic department of Lariboisière Hospital, programmed for a non-emergency hip or knee replacement and who provided consent, were included in this monocentric, prospective study based on the daily clinical practice.We did not include non-French-speaking patients and those refusing to participate to the study.Based on the routine clinical practice, all patients underwent a standardized clinical examination, including medical history and physical examination, laboratory tests were performed in all subjects including chemistry panel and complete blood count.Patients underwent either a general anesthesia (GA), a locoregional anesthesia (LRA) or LRA+ propofol (LRA+P).

Cognitive assessment
To evaluate the cognitive diminution, we administered the MoCA one day before, then two days and six weeks after the surgery.The same practitioner administered MoCA tests.Additionally, the same patient never performed twice the same version of the test.

Statistical analysis
Numerical variables were expressed by the median and interquartile range or mean and standard deviation, categorical variables were expressed as the count and percentage.For all statistical test, we chose α=0.05 as significance level.Patient characteristic data, duration of surgery, and MoCA scores were compared using chi2, Student's t-test or Wilcoxon test as appropriate.When t-test was used, the distribution normality was assessed with the two-tailed Shapiro-Wilk and Lilliefors tests.Sensitivity of variable of interest were computed from confusion matrices.
Comparison of MoCA (D-1) scores distribution and medians between GA, LRA, and LRA+P categories were analyzed with Kolmogorov-Smirnov and Mann-Whitney tests, both two-tailed.
For binary analysis, we associated for each patient the dummy variable 1 for the loss of at least one MoCA point between (D-1) and (D+2), 0 otherwise.We compared this response variable to binary variable and non-binary variables.To identify factors associated to a post-operative decline, we first proceeded with a univariate analysis (Chi2 for dummy variables and Logistic Regression (LR) for non-binary variable).We computed associated factors unless there were less than 5 occurrences in the entire population.Time duration of MoCA administration were compared using a two tailed paired t-test, after log-transform and normality evaluation (Shapiro-Wilk).
To investigate the impact of AC and BDZ treatment, we constructed two dummies variables for 'AC' AND 'BDZ' and for 'AC' OR 'BDZ' treatments.We searched for an association between these variables and cognitive decline (ΔMoCA ≥ 1) between D-1 and D+2.
We used a multivariate class weight corrected logistic regression model to produce a risk model for loss of at least one MoCA point between D-1 and D+2.Variables with a p-value ≤0.05 in the univariate analysis were introduced in the multivariate model.In the case of incomplete data, patients were excluded from the multivariate analysis.We performed statistical analysis using R-studio software.

RESULTS
Sixty patients (71.9±7.1 yr., 72% women) above 60 year-old were prospectively included in this study (Fig. 1, Tab. 1).Among them, 27 (resp.33) underwent a non-emergency elective prosthesis hip (resp.knee) implantation surgery.The distributions of age, time of surgery, MoCA, pain and well-being scores were not different between patient from the hip and the knee groups (Kolmogorov-Smirnov test, unsignificant difference, data not shown).Therefore, for the statistical analysis we decided to consider patients that had a hip and knee surgery as a single study group.
Similarly, we found that, persistent cognitive decline was associated with BDZ (p-value = <0.001) and perioperative complications (p-value = 0.013).Additionally, we found that post-operative confusion was also associated with a W+6 cognitive decline (p-value = 0.002).Results are shown in Tab.2B.

DISCUSSION
In this study, MoCA scores were significantly impacted by the surgery, with a loss of 2.25 points in average.We described that POCD were independently associated with pre-operative well-being, per-operative use of ketamine, peri-operative complications and pre-existing treatment based on either BDZ or AC drugs.In particular, well-being and ketamine had a protective effect, while complications and BDZ/AC treatments were linked to cognitive decay.Here, the type of anesthesia was not linked to cognitive decline and, persistent cognitive decline was linked to pre-existing BDZ medication and per-operative complications.Finally, MoCA evaluation duration was unchanged according the kind of anesthesia and the occurrence of POCD, however, the duration at W+6 was significantly shorter than D-1 and D+2 reflecting the training effect (Fig. 3B).These results suggest insight to improve perioperative cognitive protection and in particular to identify upstream of the intervention patients that could be affected by surgery or anesthesia.
Ketamine is a non-barbiturate anesthetic inhibiting acetylcholine that could lead to an increased number of delirium (Sleigh, et al. 2014).Per-operative use of ketamine is still debated, with on one side studies defending its neuroprotective and anti-inflammatory actions (Dale, et al. 2012), and on the other side studies arguing for no beneficial effects (Avidan, et al. 2017).Interestingly ketamine was shown to speed up recovery to consciousness after GA, yet possible impact on POCD was unclear (Hambrecht-Wiedbusch, Li and Mashour 2017).In the present study, per-operative ketamine administration was associated to fewer POCD, this support previous findings suggesting that administration of ketamine during a surgery reduces occurrences of POCD.
Chaiwat et al 2019 (Chaiwat, et al. 2019) reported more POCD among patients sedated with propofol, which was also suspected to decrease mean arterial pressure with a risk of brain hypoxia leading to cognitive decline (Wild, et al. 2008).However, ketamine could preserve the hemodynamic stability during propofol-based anesthesia.In our study, we compared three groups with and without propofol.We did not find any impact of the kind of anesthesia on the cognitive decline suggesting that propofol did not impact cognition in this cohort.
Furthermore, patients who received ketamine also received propofol supporting the hypothesis of a benefic effect of the ketamine during propofol infusion.However, even if several studies are in favor of ketamine protective effect, further prospective randomized controlled studies would be needed to confirm this hypothesis.
BDZ and AC drugs have long been suspected to be involved in drug-induced cognitive decline (Starr andWhalley 1994, Bartus, et al. 1982).On one hand, BDZ act as positive allosteric modulators of GABA-A (-aminobutyric acid) receptors, which results in an overall increased cortical inhibition, the latter possibly responsible for mental deterioration (Griffin III, et al. 2013).On the other hand, delirium pathway arises from a dopamine excess and acetylcholine decrease, that could be due or accentuate directly by AC drugs (Maldonado 2013).Furthermore, consequences of BDZ or AC treatment on cognition have been extensively described for speed-up age-related cognitive decay and Alzheimer's disease, not for POCD (Gage, et al. 2014, Grande, et al. 2018).In the present study, using a homogenous prospective non-emergent population, we observed detrimental effects of BDZ and AC drugs on cognitive functions.This finding supports previous results and hypothesis, however larger prospective and multicentric studies will be needed to confirm these results, allowing recommendations regarding the use of BDZ and AC drugs in non-emergent surgery.Clarifying a possible mechanisms underpinning BDZ/AC deleterious effects on sedated patients will benefit the prevention of temporary delayed NeuroCognitive Recovery as well as permanent POCD (NCD).We already know that the GABA-A agonist role of propofol is potentiated by BDZ, that could in turn promote POCD (Orser and Miller 2001).However, among our 54 patients, 15 had no propofol, yet for this subpopulation the cognitive decline was statistically undistinguishable from the 39 other patients.A regular intake of BDZ was also associated to a reduced cognitive reserve (Gage, et al. 2014), the latter linked to more POCD (Feinkohl, Winterer and Pischon 2017).Nevertheless, we found no correlation between MoCA (D-1) scores or sociocultural levels, and intake of BDZ/AC drugs.
Per-operative complications were associated with ΔMoCA both two days and six weeks after the surgery.These findings are consistent with previous works (Greene, et al. 2009), and even more with recent works about complications following total hip arthroplasties (Aziz, et al. 2018).This suggests that an extended period under anesthesia does not affect post-operative cognition while a physiological response to a surgical complication does.Surgery-related inflammatory response has been associated to POCD (Alam, et al. 2018).Although we did not collect inflammatory-related biomarkers, we can note that among patient with peri-operative complications, only two had NSAIDs and three had ketamine.
Finally, depressed mood, in particular among the elderly, was associated with cognitive impairment (Brommelhoff, et al. 2009).Several studies outlined the link between POCD and elderly mood deterioration, assessed using Geriatric Depression Scale for patients receiving a cardiac as well as a non-cardiac surgery (Rudolph andMarcantonio 2011, Greene, et al. 2009).In our study, we found no link between cognitive decline and patient treated for depression, which agrees with these studies.Interestingly, we found that well being measured before and after the surgery were not linked to pain scores, in fact pre-operative well-being was here an independent factor associated with POCD, which was not the case for pain.These results might be explained by the fact that pain was carefully managed after the surgery (av.score ≈ 4), which does mitigate POCD (Inouye, et al. 2015).In the end, it emerges from this study that well-being measured before the surgery is a better indicator of cognitive decline than post-operative pain, which suggests that the patient's state before surgery affects post-operative cognitive trajectory.

Limitations and perspectives
This study has some limitations.The patients included were very homogeneous but the inclusion criteria, the prospective methodology during a limited period led to a small number of patients.Larger multicentric studies would be needed to confirm or not these results and to propose recommendation regarding non-emergency arthroplasty.
In some group the number of patients was too low to provide useful data and/or multivariate analysis.However, the prospective methodology and our statistical approach allowed us to underline interesting insights about POCD.This article is the first result of a pilot study suggesting that cognitive decline occurring after a surgery could be anticipated based on pre-operative factors, yet only partially.We are performing a larger study in order to identify patients who are at risk of both delayed neurocognitive recovery and permanent cognitive decline.

CONCLUSION
To conclude, we reported a significant cognitive decline two days following a non-emergency lower limb surgery.
BDZ or AC drugs were associated with a poor cognitive trajectory, both in short-term, but also in long-term in the case of BDZ drugs.At the same time, use of per-operative ketamine appeared to have a cognitive protective effect.Pre-operative well-being and perioperative complications were the two non-drug factors associated with positive and negative outcome respectively.Findings in the present study suggest that patient's medical prescriptions and the surgery/anesthesia smooth progress can both independently affect POCD.Nevertheless, further research, based on a larger cohort, should investigate separately factors discussed above, in particular drugs for which neurophysiological mechanisms remain unclear and the role in cognitive decline/protection is still a divisive topic.

STATEMENT OF INFORMED CONSENT
We obtained an authorization by the French data privacy administrative body 'Commission Nationale de l'Informatique et des Libertés' (CNIL), under the reference number z2c2680420w.In agreement with the ethics committee for this non-interventional study an oral agreement was obtained from each patient during their first medical visit.

Figure 3 :
Figure 3: Distribution of MoCA score (A) and evaluation time (B) for three different periods: one day before (D-1), two days (D+2) and six weeks (w+6) following the surgery and anesthesia.

Table 1 :
Comparison of population characteristic data between the patient with Δ ≥ 1pt and patients with no point loss.Binary data are number and percentage, while non binary data are shown in median and interquartile range (IQR).NSAIDs ; Nonsteroidal anti-inflammatory drugs, † includes the following complications: failed LRA -modification of the surgery -Hemorrhage -incisors defect (see Supplementary).

Table A1 :
MoCA D-1/D+2 drop statistics for the three types of anesthesia

Table A2 :
List of peroperative complications