Cognitive decline after surgery and anaesthesia: correlation does not mean causation

Patients are keen to understand the risks and benefits of surgery. The focus has previously been largely on mortality risk and physical morbidity, but there is increasing concern about potential impact on cognitive ability in both the academic 1 and popular press 2, especially as increasing numbers of older people have surgery. Patients over 75 years already account for 23% of surgical procedures performed in the UK, and this is likely to increase due to increasing life expectancy and advances in surgical and anaesthetic techniques 3. Dementia – chronic cognitive decline with impaired functional ability - is the most feared aspect of growing old, and it is important to consider if surgery or anaesthesia may precipitate or accelerate cognitive decline. In the UK, approximately 7% of people aged ≥ 65 years, and 17% of people age > 80 years have a diagnosis of dementia 4, 5. In hospital, one-third of patients have dementia, and up to half of these patients have not been previously diagnosed 6. It is also important to recognise people with cognitive impairment: impaired performance on cognitive tests but with no impact on functional ability. The benefits of diagnosing this are controversial, as typically fewer than half of individuals with mild cognitive impairment will progress to dementia 7. For anaesthetists, it is also essential to consider the roles of delirium and postoperative cognitive dysfunction. These are separate entities, whose relationship is not yet fully understood 8, 9. Delirium is a well-characterised, acutely fluctuating, neuropsychiatric syndrome resulting mostly from peripheral conditions that precipitate acute brain dysfunction 10. Delirium can be transient or prolonged, and strongly predicts future new-onset dementia (there is an eight fold increased risk), and may accelerate existing dementia 10, 11. Postoperative cognitive dysfunction is a transient alteration in cognitive function, including memory, information processing, and executive function, lasting a few days to as long as a few weeks after surgery 12. There is little evidence that it persists into the long-term 13, 14. Peri-operatively, causes common to both include hypoxia, metabolic abnormalities, drug effects (especially sedatives like benzodiazepines), postoperative pain and systemic inflammation. In this edition of Anaesthesia, Patel et al. have published results from a longitudinal observational cohort study (OPTIMA), looking at the association of surgery and anaesthesia with cognitive decline in older people 15. Longitudinal studies (observational studies of cohorts followed over time) provide excellent opportunities to identify risk factors and mechanisms for the onset and progression of diseases and conditions. Repeated measurements of individuals over time enable analysis of both intra- and inter-variability, increasing the power of the study to detect real changes. The Oxford Project to Investigate Memory and Ageing (OPTIMA) study was established in 1988 to recruit both cognitively normal and cognitively impaired older people to understand the causes, treatment and prevention of dementia 16, and has included over 1100 people. Participants underwent repeated neuropsychological assessments, as well as brain scans, blood and cerebrospinal fluid sampling, and physical examination. This has already enabled investigators to identify biochemical and genetic factors that increase the risk of Alzheimer's disease 17-19. The careful phenotyping, and longitudinal nature of this cohort provides an excellent opportunity to identify the impact of further relevant risk factors. Patel et al. explore the impact that an episode of surgery and anaesthesia has on cognitive decline, by comparing the trajectory of decline in people with two or more cognitive measures (n = 394), of whom 109 had surgery. Of these, 36 patients were diagnosed with mild cognitive impairment, possible or probable Alzheimer's disease, during follow-up. The findings that cognitive decline appears to accelerate after surgery in elderly patients diagnosed with cognitive impairment, but not other elderly patients, now need to be replicated in larger studies and different populations, and used as an impetus to explore underlying mechanisms and potential treatments. We would like to discuss some issues that should be considered in interpreting the study, and in designing future studies. This is a secondary analysis of the OPTIMA cohort data, which is a useful research method. It reduces research waste 20, and maximises the use of data provided by research participants. As public funding often funds the original research, this provides a better return for the tax payer. It is important that such analyses are pre-specified before any data analysis if possible, and that there is clarity in how the sample relates to the whole cohort and the wider population. It would have been useful to have a descriptive table of the baseline and follow-up data of participants included in this analysis to enable practitioners to compare this sample to their own patients. Epidemiological studies often take place in noisy, uncontrolled environments, and we need to look very carefully at potential confounders and bias before concluding there is a causal relationship. This was illustrated in March's issue of Anaesthesia, where Choi explained that the correlation between higher chocolate consumption and a lower risk of cardiovascular events does not mean that chocolate prevents heart attacks 21. It is possible that people who like to eat chocolate do something else that offers cardiac protection, like eat a wide variety of healthy foods, and this confounds the relationship between chocolate and heart attacks. Bradford Hill proposed a framework of nine criteria to assist with the assessment of causality in observational studies: temporality; strength of the relationship; biological plausibility; specificity; dose-response relationship; consistency; experiment; coherence; and analogy 22. It is logically necessary for the cause to precede the effect in time, i.e. does the episode of surgery and anaesthesia precede cognitive decline? In Patel et al.'s paper, there is the potential for reverse causation. This cohort is comprised of participants who underwent cognitive testing at least twice. In 36 patients who were diagnosed with mild cognitive impairment before surgery, there is a statistically significant reduction in the CAMCOG score in 16 of these patients after surgery. However, there is no information on the type of surgery or anaesthesia; the risks of unscheduled major surgery (e.g. for fractured neck of femur) are different from elective moderate surgery (e.g. laparoscopic cholecystectomy). It is possible that a patient with a diagnosis of mild cognitive impairment is more likely to be frail, and therefore at higher risk of falls, and subsequent surgery for, as an example, fractured neck of femur. This requirement for surgery may therefore be a reflection of the accelerated trajectory of cognitive decline that these 16 patients are on, rather than the cause. This is particularly relevant, as even non-surgical hospitalisation of older patients for acute care may be associated with a higher likelihood of incident dementia 23. With limited information regarding the time between operation and cognitive assessment, it is also possible that delirium or postoperative cognitive dysfunction may have contributed to a diagnosis of mild cognitive impairment. In assessing cognitive function, it is also important to recognise individual differences in cognition, and that cognition is a life-long trait. Absolute CAMCOG scores are not equivalent between individuals: a score of 90 may be a large decline for one individual, but unchanged from earlier life in another. Patients with dementia typically score below 80 on the CAMCOG 24. The concept of ‘cognitive reserve’ suggests that people with higher IQ and better education can have more resilience: ‘further to fall’ before any insults have an impact on their performance 25. The statistics included in the study have clearly been performed by an expert statistician, and may be challenging for many clinicians to interpret. Patients who underwent surgery after a diagnosis of mild cognitive impairment appear to fall by five to ten CAMCOG points over two years, compared with a smaller drop in patients who did not undergo surgery (Patel et al., Fig. 4). Statistical significance does not necessarily relate to clinical significance; it would be helpful to explain what changes in function might be expected with this fall in score. Patel et al. identified “episodes of surgery that could be regarded as moderate or major in severity, and therefore highly likely to be associated with general or regional anaesthesia or both” 15. As is implied by this definition, moderate to major surgery is not performed without anaesthesia, and in turn, anaesthesia is not performed unless there is a need for surgery. This means we are unable to assess separately the effects of surgery from those of anaesthesia (general or regional) on the increased rate of cognitive decline. Both will also be partly confounded by the overall health of the patient, of which cognition will be one aspect. Inhalational anaesthetic agents have been associated with increased formation of Alzheimer's disease precursors including beta-amyloid plaques and neurofibrillary tangles in animal models and in vitro studies 26, 27. On the other hand, propofol has been associated with a neuroprotective effect against inflammation 28. Surgical patients have high levels of circulating pro-inflammatory cytokines, prostaglandins and damage-associated molecular patterns (DAMPs), causing an endothelial pro-inflammatory response which in turn activates neurons and glial cells, causing subsequent neuro-inflammation. This inflammation is generally effectively cleared by glial cells and growth factors 29, but the abnormal persistence of this inflammatory state has been associated with cognitive decline after surgery 30. This is supported by evidence that critical illness, without surgery, is associated with cognitive dysfunction 23. Biomarkers of brain damage have been shown to be transiently suppressed with general anaesthesia on its own but were increased when anaesthesia was given in combination with surgery 31. Future studies should try to separate the effects of surgery, and different types of anaesthesia. To assume causality, the effect of the predictive variable should be specific to the outcome in question. This is challenging in diseases with multifactorial aetiology, and it is important to consider all the other potential factors around the peri-operative period that may impact on postoperative cognitive dysfunction. Delirium and postoperative cognitive dysfunction are important potential confounders. Hypoxaemia, hypotension, embolism, medications and postoperative infections have all been implicated in the development of postoperative cognitive dysfunction and delirium 30, and could contribute to cognitive decline. Postoperative cognitive dysfunction and delirium can be reduced by multicomponent interventions 32. Patel et al. state that, in patients with pre-existing cognitive impairment, a second surgery reverses the cognitive decline seen after first surgery 15. It is difficult to hypothesise how a first surgery and anaesthetic causes accelerated cognitive decline, but a second exposure slows this. Other studies that report a positive association between surgery/anaesthesia (or critical illness) and cognitive decline have found a dose-response relationship, such that repeated episodes, or increased duration, were associated with greater cognitive decline 23. This would suggest a potential selection bias of patients in this cohort who were considered for second surgeries, and that second surgery is not protective, but that fitter people are selected for second surgery: a ‘survivor effect’. Previous observational studies have also investigated the association between surgery/anaesthesia and cognitive decline, with some finding an increased risk 33, and others no increased risk 34-37. The potential for permanent and increased cognitive decline has significant implications for how we view peri-operative risk in the elderly population. Patel et al. review this literature, and include these findings in their conclusions, but the evidence still appears to be inconsistent and is ripe for a systematic review and meta-analysis of the field to establish if robust conclusions can be drawn by formally synthesising the existing studies. It will clearly never be ethical to consent patients to surgery without anaesthesia, or to anaesthesia without surgery. There has been one randomised controlled trial which randomised patients to regional versus general anaesthesia for non-cardiac surgery, and found no difference in the incidence of cognitive dysfunction three months after surgery 13. However, this was a heterogeneous group of surgeries, and the general anaesthesia given was determined by the anaesthetist, which may have introduced bias. Furthermore, they did not look at long-term cognitive decline. Patel et al.'s paper raises the important issue of whether surgery and anaesthesia impact on cognitive decline. It is important that this question is addressed to allow surgeons and anaesthetists to help patients and their relatives make informed decisions, both for people already diagnosed with cognitive impairment or dementia, or those who are concerned about the impact surgery may have. Although there is biological plausibility that surgery and/or anaesthesia may cause increased cognitive decline, particularly in susceptible people, the Bradford Hill criteria have not been fulfilled in this study, or other observational studies to date, and we would urge caution in drawing any general conclusions due to the methodological concerns about this study outlined above. There is no temporal or dose-response relationship, little idea of the strength of the relationship, or any form of consistency in the published literature. Studies have been unable to control for delirium or postoperative cognitive dysfunction, or to assess the specificity of anaesthesia versus surgery as the prime cause. With so much uncertainty, it is now essential that there is a comprehensive systematic review of all (published and unpublished) literature, with critical appraisal and assessment of risk of bias. This will inform a clearly-defined research question in a high quality randomised controlled trial of patients (including those with pre-existing cognitive impairment), by randomising to a specified surgery with different types of protocolised anaesthesia. This will remove the majority of the confounding that has limited the conclusions of these many observational studies. Patel et al.'s paper is a welcome impetus to encourage discussion about cognitive risk in the consent process between surgeons, anaesthetists and patients, but we hope that it does not directly affect decision-making that could discriminate against people with, or concerned about, cognitive impairment. It highlights the need for increased awareness of the cognitive consequences of surgery and anaesthesia, and the need for future high quality research to answer this important question. AD is the trainee fellow with the editorial board for Anaesthesia and this editorial has therefore undergone additional external review. No other conflicts of interest declared.