How should clinicians interpret these results?

In this issue of the EJA, Pennings et al.1 present results that will undoubtedly raise significant questions among professional readers. Suppose the content of this article was shared with patient groups. In that case, they might ask, ‘Doctor, does this mean that because of general anaesthesia itself—rather than the surgery and anaesthesia—I have a risk of cognitive decline?’ It is crucial from the very beginning to address this ambiguity: ‘general anaesthesia’ versus ‘surgery requiring general anaesthesia’. The title is: ‘Anaesthesia as a Risk Factor for Long-Term Cognitive Decline’. Throughout the manuscript, this is rightly ‘downplayed’ to ‘exposure to surgery under general anaesthesia’, which is slightly different, but the distinction is conceptually very important. The potential role of anaesthesia, specifically general anaesthesia, as a contributor to postinterventional cognitive decline has been a concern for many years.2,3 Addressing these concerns in the context of the article published by Pennings et al.1 requires a rigorous methodological approach. Understanding the history and the context of the study is essential. The Maastricht Ageing Study (MAAS) followed 1823 cognitively healthy participants aged 24 to 82 years and assessed cognitive function at baseline, 6 years and 12 years.1,4 Researchers analysed verbal memory, executive function and information processing speed in participants with pre-existing cardiovascular disease (CVD), newly diagnosed CVD and those without CVD. The study found that individuals with CVD at baseline exhibited greater cognitive decline compared with healthy controls. Those who developed CVD during the survey also showed cognitive decline, particularly in the later follow-up period (6 to 12 years). Notably, these effects were more pronounced in individuals younger than 65 years, and the authors concluded that both pre-existing and newly developed CVD predict cognitive decline. These observations of the MAAS study are consistent with the accepted paradigm that the ‘health trajectory’ over several years (lifetime) alters cognitive performance.2 The design of the MAAS facilitated the investigation of the statistical relationship between exposure to surgery under general anaesthesia and long-term cognitive decline. However, the occurrence of surgery was self-reported, and there were no details on intra-operative or postoperative complications, which could have had significant long-term effects on cognition. Using data from the MAAS, at baseline, researchers estimated previous exposure to surgery under general anaesthesia and monitored cognitive function over 6 to 12 years. Interestingly, the authors only considered the estimated cumulative duration of exposure to surgery under general anaesthesia before the baseline (see their Table 2). Although procedures under general anaesthesia were noted during the follow-up period, data from these were not used in calculation of the estimates. This suggests that the estimation of the statistical relationship between exposure to surgery under general anaesthesia and long-term POCD is incomplete, and may potentially underestimate the effects of anaesthesia. The authors employed a variety of validated cognitive tests assessing learning, verbal memory, executive function and complex attention. Importantly, when the MAAS study was planned, cognitive function was accurately evaluated through cognitive tests, but self-assessment was not used, as recommended in 2018.5 Pennings et al.1 concluded that increased exposure to surgery under general anaesthesia independently contributes to long-term cognitive decline. Readers should interpret these findings by considering several issues: (1) Differences (but also possible temporal causality) between short-term and long-term postoperative cognitive dysfunction Short-term cognitive dysfunction such as postoperative delirium (POD) and postoperative cognitive dysfunction (POCD), which starts in the hours/days following surgery/anaesthesia, has been widely studied.6,7 Randomised trials suggest that general anaesthesia does not leave an independent ‘statistical signature’ when compared with regional anaesthesia without sedation.8 For the same surgical procedure (e.g. surgery for hip fracture) when the control for ‘general anaesthesia’ is ‘spinal anaesthesia without sedation’, several studies and meta-analyses have shown that exposure to general anaesthesia is NOT associated with an increased risk of short-term POCD9: Joo and Whitlock discuss this in more detail.2 Why is the relationship between short-term and long-term cognitive alterations important? It is important because several publications identified a statistical relationship between early (< 30 days) cognitive decline that included but was not limited to POD developing less than 3 days following intervention/anaesthesia and cognitive decline at 6 to 12 months.7 The study by Pennings et al.1 extends this to 6 to 12 years. However, long-term POCD remains more difficult to study than short-term POCD. There are reasons for these difficulties and specifically the lack of appropriate control groups.2 Ideally, control groups would include patients undergoing similar procedures without general anaesthesia; patients with the same medical conditions (e.g. arterial hypertension and diabetes that contribute to cognitive decline10 and stroke11) undergoing nonsurgical interventions such as minimally invasive nonsurgical procedures; patients with similar health conditions who received no intervention at all but were hospitalised; patients with similar diseases who were never hospitalised; and age/sex-matched healthy controls. In the absence of such control groups, although Pennings et al.1 adjusted for covariables, several covariables known to be associated with long-term cognitive decline were not accounted for: sleep quality,12 physical activity13 and lifestyle.14 Hence, it is perfectly conceivable that with different covariates, the statistical relationship between exposure to ‘surgery under general anaesthesia’ and 6- and 12-year cognitive performance would have been different. (2)The strength of the statistical association between exposure to surgery under general anaesthesia and long-term cognitive decline There is an important detail that readers should take into account when analysing the results of the linear mixed models (see their Tables 2 and 3). First, the relationship between exposure to surgery under general anaesthesia is not significantly associated with the occurrence of POCD (see their Tables 2 and 3). What is statistically associated with the occurrence of POCD is the interaction term (exposure to surgery under general anaesthesia versus the three periods of evaluation of POCD). The clinical relevance of this interaction term is difficult to analyse otherwise than by the fact that the two terms of the interaction evolve differently across time. With these difficulties in interpreting the interaction term, the values of the estimator (both for the nonsignificant association between exposure to surgery under general anaesthesia and for the significant association of the interaction term), the values of the estimates (whatever the P values) are much lower for ‘total anaesthetic time at baseline x time (minutes)’ as compared with the value of the estimator for ‘educational level’. Similarly, very low values of the estimators for exposure to surgery under general anaesthesia are reported for all the other cognitive tests (see their Tables 2 and 3). This implies that when compared to other covariables in the mixed linear model the contribution of ‘exposure to surgery under general anaesthesia’ is minimal. The lower the value of the estimator, the more susceptible it is to the presence of other covariates in the linear mixed model. The minimal (or absent) role of exposure to anaesthetic drugs seems to be the most reproducible finding in the literature.2 (3)The existence of pathophysiological models that could explain the alleged impact of general anaesthesia drugs on cognitive function The existence of pathophysiological (experimental) models that could explain the statistically independent impact of general anaesthesia drugs on early cognitive decline in experimental models should be taken into consideration.15 Given the evidence in humans of a relationship between short-term and long-term cognitive dysfunction, such a short-term impact from general anaesthesia could be extrapolated to a longer-term impact.7 There are credible experimental arguments that surgery and general anaesthesia may affect brain functions through mechanisms such as neuro-inflammation and neurodegeneration, and these may explain observed changes in cognitive performance.16 Furthermore, there is credible evidence that, experimentally, neuro-inflammation can be modified,17,18 and this would confer a nearly causal role to these pathophysiologic mechanisms. Finally, in humans, there are structural arguments, such as changes in brain structures detected by MRI,19–21 linking pre-operative brain alterations, surgery/anaesthesia and POCD. Nevertheless, although experimental evidence supports the mechanisms in short-term POCD, their role in long-term POCD requires experimental verification. (4)Defining ‘anaesthesia’ and its possible relationship with cognitive function/decline Anaesthesia is not merely administration of anaesthetic drugs, but it also involves the management/preservation of homeostasis, including anaemia,22 oxygenation, CO2 regulation23,24 and haemodynamic stability: these are all conditions associated with short-term POCD. These components of anaesthesia could not be taken into account in the study by Pennings et al.1 It must also be understood that the surgical/anaesthetic procedures Pennings et al.1 took into account were performed before the baseline cognitive evaluation (i.e. 1993 to 1996), an era during which optimisation of anaesthesia was not yet the object of recommendations. Indeed, in 2010, the European Society of Anaesthesiology (ESA) recommended that ‘one should maintain homeostasis’.25 It was only in 2014 that ESA recommended avoidance of arterial hypotension and avoidance of anaesthetic drugs overdosing, as estimated by BIS values less than 45 for cumulative durations of 15 to 30 min.26 If surgery under general anaesthesia has a statistically independent effect on long-term cognitive decline, this raises the question: was anaesthetic drug management optimised in these patients in 1993 to 1996? The answer is probably ‘NO’. Current anaesthetic practice aims to optimise anaesthetic drug doses using numerical EEG-based monitoring (e.g. BIS values of 40/45 to 60) and to avoid anaesthetic drugs overdosing and an isoelectric EEG.27 However, this approach does not provide a truly individualised assessment of anaesthetic depth and is rarely, if ever, able to avoid (but can correct) anaesthetic drugs overdose. Future tools may refine anaesthesia optimisation by preventing rather than correcting overdosing and thus improve outcomes.28,29 The authors of this editorial conclude that anaesthesia practice before the ‘baseline’ evaluation (1993 to 1996) may not align with current practices. Conclusion The study by Pennings et al.1 raises an important and valid question. However, the robustness of the statistical association between ‘exposure to surgery under general anaesthesia’ and long-term POCD is constrained by the limitations of the study design, particularly the absence of appropriate control groups. The authors’ title, ‘Anaesthesia as a risk factor for long-term cognitive decline’ may be overstated, both statistically and mechanistically: the authors explore ‘surgery under general anaesthesia’ and not ‘general anaesthesia’. The causal link between the exposure to anaesthetic drugs in the absence of surgery and cognitive decline remains uncertain: it is more than likely that the main culprit is the disease/surgery. Although it is impossible to rule out the role of ‘general anaesthesia’ per se (consider the issue of self-evaluation of POCD2), current evidence from short-term POCD human data as well as the majority of the studies on long-term POCD does not strongly support this hypothesis. So, if patients ask, ‘Will undergoing general anaesthesia increase my risk of postoperative cognitive dysfunction?’ a reasonable response would be: ‘On average, it is highly unlikely, but not entirely impossible. Furthermore, for reasons that the medical community does not really understand, some patients may experience a long-term alteration of cognitive function. Meanwhile, the anaesthesia and critical care community continues to refine techniques, optimise anaesthetic management, and develop sophisticated tools to assess and mitigate potential cognitive effects of anaesthesia’.