Introduction
The question posed by the title of this chapter asks, “Does surgery accelerate dementia?” The short answer is that we should know, we don’t know, but we can know.
In their recent landmark article, Silbert et al. demonstrate that patients with cognitive impairments before surgery have an increased risk for postoperative cognitive dysfunction (POCD) compared to patients with no preexisting deficits at 3 (24.2% vs. 11.4%) and 12 months (9.4% vs. 1.1%) (1). The manuscript comprises many strengths including precise neuropsychological testing, use of 2 standard deviation thresholds, and ample power. Although broader implications of the authors’ findings are constrained by a lack of presurgical cognitive trajectories, exclusion of patients with overt dementia before surgery, a single year of follow-up, and a small number of no-surgery control participants, the data firmly link mild cognitive impairment (MCI) to increased risk for POCD. As the authors note, “subtle impairment of cognition is known to precede dementia” (1). From now on, investigators of either POCD or MCI must account for the incidence of both.
Whether POCD hastens the onset of MCI and dementia is the subject matter of the present contribution. To date, surgery is not considered a modifiable risk factor for dementia (2). Numerous, high-quality literature reviews of preclinical and clinical research published in the past 2 years arrive at the same conclusion, i.e., presently available evidence is inconclusive (3–9). Accordingly, it is not our aim here to provide another overview of primary sources one by one. Instead, and as a longer answer to the title’s query, we survey barriers to acquisition of the data that we and our patients require, suggest approaches to surmount these barriers in clinical research, and propose clinical care steps to consider in the face of partial knowledge and pressing need.
Surgery and Dementia
What Is Dementia?
Persons exhibiting signs of normal cognitive aging are aware of their own forgetfulness and word-finding difficulties, and are typically more concerned about the changes than family members. To the contrary, persons with MCI and dementia are often unaware of problems in thinking, whereas family members express concern. MCI describes cognitive performance that falls between the changes of normal aging, and those of early dementia in persons experiencing a decline, but who are able to function independently in their daily lives (10). Dementia is defined by cognitive impairment and poor performance on objective, psychometric evaluations that document a decline from the past in two or more cognitive domains (i.e., memory, reasoning, visuospatial skills, language, and personality) that is greater than expected for age. Additional evidence of difficulties in daily life that interfere with independence is required to support the diagnosis of dementia. Patients with early or mild dementia may function independently during simpler activities, with severe compromise of basic functions of daily living observed in patients with more advanced stages of dementia. Both MCI and dementia are syndromes that may have many causes including brain trauma, drugs, alcohol, hormone or vitamin imbalances, depression, vascular dementia, Lewy body disease, frontotemporal dementia, primary progressive aphasia, Creutzfeldt-Jakob disease, Pick disease, Parkinson’s disease, Huntington’s disease, normal pressure hydrocephalus, acquired immune deficiency syndrome (AIDS) dementia complex, Wernicke-Korsakoff syndrome, and Alzheimer’s disease (AD). Patients often meet diagnostic criteria for more than one type of dementia with, for example, concurrent vascular dementia and AD. As used here, postsurgical dementia refers to the accelerated onset or progression of dementia linked in phase with surgery conducted under general inhalational or intravenous anesthesia, spinal or epidural regional anesthesia, or a combination thereof, without denoting a specific cause or disease association.
The spectrum of dementia etiologies and stages of progression challenges researchers and clinicians seeking to disentangle risk factors, prevention strategies, and treatment options for a syndrome with an assortment of pathologies. Because postsurgical dementia is a diagnosis made by exclusion of other causes of cognitive decline (e.g., stroke, epilepsy, intoxication, delirium, major psychiatric disorders, and metabolic syndromes), surgeons and anesthesiologists acting alone or in concert lack expertise to ensure appropriate diagnosis and treatment. Collaboration with skilled neurologists, geriatricians, epidemiologists, and public health specialists, together with nonphysician caregivers including neuropsychologists and nurses, is critical to identify persistent postsurgical cognitive phenotypes and their impacts. Whether a unique postsurgical dementia exists in the absence of other dementia etiologies is uncertain, and is likely to remain so until the potential for surgery to influence the course of recognized dementias is resolved. Nor can it be expected that the effects of all surgical and anesthetic variables are equivalent in all variants of dementia. In particular, experimental designs are needed that distinguish an apparent link between surgery and the onset or progression of dementia that arises from a causal association rather than by chance alone.
What Is Alzheimer’s Disease?
Mutations in three genes, i.e., amyloid precursor protein (APP), presenilin 1 (PSN1), and presenilin 2 (PSN2), cause early-onset (<60 years of age) AD (EOAD) in 1%–5% of all patients diagnosed with AD. With a shared younger age of onset and mechanism of inheritance, overlapping pathogenesis, and parallel time courses, it is convenient to consider all variants of EOAD as a single disease despite underlying allelic and locus genetic heterogeneity, and despite the possibility that some patients with EOAD first exhibit signs and symptoms after 60.
Late-onset AD (LOAD), a clinical diagnosis made by exclusion of other forms of dementia, is thought to arise from a shared “pathophysiology of AD” (11). In a subset of patients, biomarker evidence of disordered brain amyloid-beta (Aβ) deposition (low cerebrospinal fluid [CSF] Aβ42, positive positron emission tomography [PET] amyloid imaging), and neuronal degeneration (elevated CSF tau, positive PET tau imaging, decreased PET 18fluorodeoxyglucose, structural atrophy on magnetic resonance imaging [MRI]) is available. Such evidence increases the probability that a patient’s dementia arises from a proposed AD pathophysiologic process that may be asymptomatic for many years to decades before its earliest clinical manifestations (12). Presence of the apolipoprotein E (APOE) ε4 allele increases the risk of AD in females more than in males, but is not sufficiently specific to serve as a causal or diagnostic LOAD genotype.
Unlike EOAD, it is not correct to regard LOAD as a single disease (13). To the contrary, LOAD is a multifactorial syndrome, or “spectrum,” of component diseases with varied pathophysiology (14). As demonstrated by genome-wide association studies (GWAS), numerous DNA sequence variations at multiple genetic loci contribute statistically significant but tiny risks for LOAD in diverse populations, with substantial unaccounted for heritability (15). Similarly, failure of more than 100 randomized controlled trials (RCTs) of LOAD drug therapy, and MCI observed in patients with suspected nonamyloid pathology (SNAP), casts the “amyloid hypothesis” of LOAD etiology into doubt (16,17). To complicate matters, patients with LOAD may first express signs of dementia before age 60.
Uncertainties in LOAD nosology and its hypothetical pathologic mechanisms have significant ramifications for investigators in the design of trials aimed at examining the relationship between surgery and dementia. A clear message is that the search must broaden beyond amyloid- and tau-based biochemical and imaging biomarkers. Greater resources and efforts must be dedicated to ensure that patients with targeted stages and etiologies of dementia are matched for risk factors before surgery, and that those with confounding conditions and mixed etiologies are controlled for or excluded. To this end, genotyping all research participants for EOAD causal alleles is a justified and expedient standard.
How Is the Date of Onset and Progression of Dementia Ascertained?
On its face, the date of dementia onset after surgery appears to be a straightforward and discrete primary outcome. It is not. As noted by Jack et al., dementia is preceded by slowly evolving signs and symptoms of MCI (12). Despite an inexorably progressive course after clinical onset, dementia fluctuates over shorter (hours) and longer (months) time intervals that may bracket serial observations. Many elements used to establish the onset of dementia may be delayed; for example, the first mention of dementia signs and symptoms in a medical record, or the date dementia was first evaluated by psychometric tests, biomarker levels, or neuroimaging, may be conditioned by how often a patient visits a caregiver. Other elements of LOAD onset are subjective and retrospective; for example, a patient’s or partner’s recollection of the date and nature of first deficits, either or both of which may be confounded by MCI or dementia. The onset and time course of co-existing disorders of cognition including cerebrovascular disease, drug and alcohol consumption, and other neurologic and psychiatric disorders, are also imprecise. Until serial cognitive testing is routinely documented in the primary care of elderly patients, reliable data upon which clinical care decisions may rest can only emerge from prospective observational and randomized controlled trials in which the date of dementia onset is meaningfully defined and accurately measured.
Is Anything Certain about the Relationship of Surgery and Dementia?
Literature that addresses the effects of surgery on dementia onset, progression, and severity is sparse, of limited quality, and controversial. The absence of case reports in individuals, families and subpopulations is remarkable, and suggests either that there is no association between surgery and dementia, or that perioperative caregivers and their patients have gone decades without recognizing the association. A single RCT reports that the progression of MCI to LOAD is faster in patients having surgery with inhaled general anesthesia compared to those receiving epidural or intravenous general anesthesia (18). However, the sample size of the investigation is small (n = 60/group), the duration of evaluation is limited (2 years), and the incidence of LOAD does not differ between the three groups.
Retrospective, case-control studies rarely are able to balance comorbidities and medications that are active in the central nervous system, do not compile surgical and anesthetic records, and are not matched for the type and length of surgery, or anesthetic regimen and doses. Retrospective cohort studies suffer from poorly defined LOAD outcomes, failure to account for dementia from other causes, and constraints of retrospective analysis of data collected for administrative or quality assurance rather than clinical management purposes (19). Sprung et al. found no association between surgery, general anesthesia, and dementia in patients older than 45 in a retrospective analysis of records dating to 1985 (20). More recently, two retrospective investigations of large population databases reported an increased risk of dementia after surgery compared to control groups matched for age and sex, although the data are prone to confounding with inconsistent diagnostic criteria, possible errors in coding, and ascertainment or other hidden biases (21,22). In virtually all retrospective investigations of surgery and dementia, background variables (e.g., comorbidities, medications) are heterogeneous, independent variables (e.g., surgery and anesthesia) are heterogeneous, and dependent variables (e.g., indices of cognitive performance, onset, and decline) are very heterogeneous. For at least these reasons, further retrospective investigations of administrative databases cannot be expected to decisively settle the relationship between surgery and dementia, if one exists.
Is There Biomarker Evidence of Postsurgical Dementia?
Neuroimaging and biochemical approaches to quantifying correlates of brain injury in patients with postsurgical dementia appear promising but are rudimentary at present. In an analysis of volumetric MRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), Kline et al. report increased rates of cortical gray matter atrophy, hippocampal atrophy, lateral ventricle enlargement, and reduced cognitive performance in patients 5–9 months after surgery compared to no-surgery controls (23). Surprisingly, differences between the surgical and control groups were not seen 12 months after surgery, suggesting neuronal plasticity and recovery from surgical injury in at least some participants (24). In an uncontrolled, “pilot proof-of-concept” investigation of 13 patients providing CSF samples before coronary artery bypass surgery, and at 1 week and 6 months afterward, Palotas et al. observed decreased Aβ, and increased tau and S-100β, consistent with changes seen in patients with AD (25). Corresponding cognitive decrements were statistically significant, but clinically insignificant. Other investigations describe perioperative changes in CSF biomarkers associated with AD, but are restricted to samples collected just before or just after surgery, and lack clinical tests of dementia over a relevant timeframe (26,27). No heritable, genomic indices of persistent postsurgical dementia lasting 3 months or longer after surgery have been identified to the present (28,29).
Will In Vitro or Animal Data Resolve Postsurgical Dementia Phenotypes?
No bench-top or in vivo models of human postsurgical MCI and dementia are available, or are ever likely to become available. Even the most advanced primate mind fails to satisfy psychometric and daily living skill thresholds of normal human cognitive performance. As the scope of noninvasive and minimally invasive methods of characterizing human dementia phenotypes broadens (see below), resources for animal investigations must be narrowly directed to efforts that answer a question of immediate correspondence to a human trait that cannot otherwise be answered in humans with co-administration of neuropsychological test panels. These principles are further offset by ethical concerns surrounding experimental surgery and anesthesia in animals with limited relevance to human disorders of higher cognition. Close scrutiny of “hypothesis generating” but indeterminate animal investigations aimed at modeling human cognition (e.g., learning, reasoning, language, and personality) is particularly warranted.
Are Cognitive Phenotypes after Surgery Related to One Another?
At least four phenotypes of disordered cognition may be observed after surgery: emergence delirium (hours); postoperative delirium (days); POCD (months); and postsurgical dementia (years). A reasonable and testable hypothesis is that the four phenotypes are a continuum with shared or overlapping etiologies in at least some patients. Patients with preoperative memory complaints and lower cognitive scores having cardiac procedures are at increased risk for postoperative delirium (30,31). Conversely, postoperative delirium is associated with an increase in the incidence of postsurgical dementia (32). Lundström et al. found that postoperative delirium in nondemented orthopedic patients is associated with the subsequent development of dementia, and a higher mortality rate up to 5 years after surgery (33). Steinmetz et al. observed that POCD was not associated with “registered dementia” over a median follow-up of 11 years after surgery (34). The single POCD protocol that targeted enrollment of patients with MCI reports a twofold increased risk for POCD in patients with presurgical cognitive deficits (1). Whether or not presurgical dementia increases the incidence of PND is untested because all published POCD protocols specifically exclude patients with dementia from participation. If, in fact, postsurgical cognitive disorders alter the risks for one another, failing to seek evidence of each phenotype in all participants is at best an inefficient and incomplete use of precious research assets, and at worst discounts predictors of scientific and clinical importance.
Current Research Priorities
We contend that in vitro laboratory investigations, in vivo animal studies, and retrospective inquiries of databases assembled for purposes other than research are ill-suited to confirming or refuting a putative relationship between surgery and dementia in humans. Absence of a clear cut and consensual postsurgical dementia phenotype renders the design and conduct of RCTs premature. As an alternative, we suggest that prospective, observational cohort experimental designs are a preferred option for acquisition of conclusive evidence configured to address whether or not such a trait exists. A first essential is that surgical and anesthetic independent variables are clearly defined, explicitly justified, and well documented. For example, a qualifying surgery for high-priority prospective investigations may be elective (i.e., to enable preoperative cognitive and biomarker investigations), of at least moderate duration (e.g., 1 hour or more), of at least moderate invasivity (e.g., body cavity or long bone), conducted under inhalational or intravenous general anesthesia, or spinal or epidural regional anesthesia, or a combination, in which full physical recovery is anticipated. A key advantage of cognitive research in the perioperative interval is the availability of rich and complete documentation of a patient’s evolving medical status including preoperative surgical and anesthetic appraisals, intraoperative anesthetic records, postanesthesia care unit (PACU) documents, the dictated surgeon’s summary, and the facility discharge record and instructions. Table 5.1 provides a working template for collating these data in preparation for entry into databases for correlative analysis. Seeking informed consent to obtain all medical records for at least 12 months before surgery, and for 3–5 years after surgery, may jeopardize enrollment, and potentially introduces bias (i.e., only participants who are otherwise unconcerned about the contents of their medical records may consent to enroll), but warrants strong consideration for reasons outlined above.
| Before surgery |
|
| Surgery |
|
| After surgery |
|
Note: Prepared from anesthesia pre-op record, intra-op record, PACU record, surgeon’s summary, and facility discharge summary.
Single “baseline” psychometric evaluations may be confounded by transient factors (e.g., emotional impacts of a surgical diagnosis and planned procedure, sleep deprivation, drug intoxication and hangover, and other temporary life stressors), and are not a proxy or substitute for a participant’s cognitive trajectory for comparison before and after surgery (35). A strongly encouraged option is to append prospective cohort studies of postsurgical dementia to any of the numerous ongoing longitudinal studies of cognitive aging and its disorders (36). Methods of designing, acquiring, and analyzing serial neuropsychological data panels must comport with the highest standards of the research community. In particular, test batteries flanking surgery should be constructed to target shared domains, but to not contaminate data acquisition of long-standing investigations in which they may be nested. Particular scrutiny of external data selected for normative comparisons is urged, and differences between a sample cohort and a normative standards cohort must be identified, controlled, and reported. Internal, prospective standards of stable and normal performance are a nonexclusive alternative to strict reliance on external standards alone (37). We further advocate analysis of repeated measures psychometric data before and after surgery as continuous variables, in addition to analysis by categorical thresholds and composite z-scores in which high-resolution data of potential clinical significance may be discarded.
Multiple noninvasive and minimally invasive biomarkers and predictors of altered cognition stand at the frontier of human postsurgical dementia research, and comparisons before and after operative procedures are timely. Contemporary dye-free MRI images display high-resolution brain anatomy, small and large vessel brain perfusion, and gray and white matter functional MRI (fMRI) connectomes in less than 1 hour of scanner time, with established validity in diverse settings of cognitive decline (38–40). PET tracers for tau aggregation and microglia activation have recently been approved for human use (41,42). Plasma proteomic and metabolomic profiles correlate with the progression from preclinical status to MCI, and from MCI to dementia (43–45). Damage-associated molecular patterns (DAMPs) released during surgery, necroptosis, and sterile inflammation activate adaptive autoimmune processes and mediate hyperinflammatory responses (46). High-resolution mass-spectrometry (MS)-based proteomics brings identification and quantification of DAMP mediators in fluids and tissues to the bedside and to the operating room (47). Patterns of methylation in genomic DNA of blood cells closely correlate with the chronologic age of the human source, and are shared with all proliferating and nonproliferating cells including those of the brain (48). The gap between chronologic age and biologic age is a heritable trait that varies with disease, but the effects of surgery and anesthesia are unknown. Neural-derived exosomes are recoverable from plasma, and share contents of their cells of origin. Levels of low-density lipoprotein receptor-related protein 6, heat shock factor-1, and repressor element 1-silencing transcription factor are significantly lower in exosomes from AD patients 2–10 years before their clinical diagnosis (49). It is now possible to generate induced pluripotent stem cells (iPSCs) from peripheral blood monocytes (PBMCs) that may be directed to glial and neuronal lineages (50). For at least these reasons, we recommend that whole blood and plasma samples be simultaneously collected at all psychometric and neuroimaging evaluations in all prospective investigations of postsurgical dementia and other postoperative cognitive phenotypes. In this fashion, biomarkers may enhance presurgical risk assessments, promote matching for background variables at the cellular and molecular level, and provide objective, postsurgical indices to test the effects of modifiable risk factors, preventive interventions, and therapies in future RCTs of postsurgical dementia.
In addition to psychometric and biomarker dependent variables, we counsel that all prospective investigations of postsurgical dementia include serial analyses of the effects of postoperative cognitive performance on measures of functional status, perceived health, and quality of life (51). Taking this step is the only way to assure that the nature and magnitude of differences that matter to investigators are aligned with those that matter to patients and their families. If identified, changes in the instrumental activities of daily living will provide compelling motivation for research funding, widespread dissemination of experimental results, and promote adoption of supported changes in care.
Beyond competence to provide informed consent (i.e., a nontrivial issue in longitudinal studies of cognition), psychometric methods must be validated within participant fluency. As well, criteria for inclusion in postsurgical dementia research must embody serial evaluations of visual and auditory acuity. The effects of surgery and anesthesia on sensory acuity in the elderly are not known with precision. Without these data deficits in perception may be confused with deficits in cognition. Most participants in prospective cohorts will be American Society of Anesthesiologists (ASA) physical status I, II, or well-compensated III enrollees by design, with sicker candidates reserved for enrollment in later trials. Nonsurgical exclusion and disenrollment criteria may comprise: no other central nervous system disorders in the year before or after elective surgery; no clinically significant, active, or unstable hepatic, renal, pulmonary, or endocrine disease, unstable angina, or decompensated heart failure; no history or treatment of drug or alcohol abuse in the preceding or succeeding 12 months; and no acute or chronic psychosis. Neurologic exams on enrollment, and at serial evaluation intervals, may be integrated to rule out intercurrent disorders of cognition that may be unrelated to surgery or anesthesia. Surgical exclusion and disenrollment criteria may comprise: no elective or emergent surgery or in the 12 months before a qualifying surgery, and no elective or emergent surgery in 12 months or more after a qualifying surgery. The potential for surgery to accelerate dementia after cardiac surgery, intracranial surgery, cerebrovascular surgery, other surgery known to alter cognition, and surgery with planned postoperative intensive care unit admission will most probably require designs that specifically target these procedures that may otherwise be excluded from qualifying surgeries not known to impair cognition. Similarly, experimental designs to test the effects of emergency surgery, subqualifying surgery (e.g., ophthalmologic surgery, colonoscopy), complicated surgery, early-life surgery, multiple surgeries and anesthetics, coincident traumatic brain injury, coincident postchemotherapy cognitive impairment, and nonsurgical critical illness on postsurgical dementia may be informed by the results of trials in which these events and disorders are excluded as possible confounders.
A significant and often underappreciated hurdle in the design of postsurgical dementia trials is the restricted number of volunteers eligible to serve as no-surgery control participants. We observe at least a 15% incidence of surgery with anesthesia each year in patients aged 60–70. This means that 60% or more of candidate no-surgery control participants will have surgery in the year before to 3 years after an index date matched to a participant having surgery. Matching no-surgery control participants with surgery participants for inclusion and exclusion criteria is essential, but takes a second bite out of the candidate no-surgery control pool. Matching for validated, unmodifiable predictors of cognitive decline including age, sex, educational attainment, and APOE ε4 further decreases pool size. Matching for all variables known or suspected to influence the incidence of MCI and dementia, and to influence performance on psychometric test batteries (e.g., comorbidities, hypertension, diabetes, hyperlipidemia, levels of vitamins B and D, centrally acting medications), and matching surgery participants with no-surgery controls receiving medical management for a shared condition, calls for multicenter designs to amply power comparisons between the surgery and no-surgery groups.
While prospective, observational cohort studies may provide the most efficient and definitive approach, other pathways to understanding the relation between surgery and dementia are also likely to be productive in the near and longer term. A first priority is professional and specialty advocacy for the introduction of validated measures of cognitive performance as a practice standard before and after surgery, and as a routine component of primary care. Consensual standards for definitions and nomenclature, experimental designs, sample size estimates, psychometric test panels, study intervals and durations, reporting standards for surgical and anesthetic variables, inclusion and exclusion criteria, and background variables may be provided as publication guidelines by a consortium of journal editors and reviewers, in compliance with The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies (52). Contacting directors and sponsors of ongoing observational and RCT investigations of dementia to encourage incorporation of surgical and anesthetic predictors into their databases is more likely to be met with success if pursued by a representative body of a surgical or anesthetic organization, than by unsolicited requests from single investigators. As the data of Silbert et al. (1) demonstrate, isolation of the dementia research community from the postsurgical dementia research community cannot be justified. The paucity of case reports of accelerated dementia after surgery in individuals, families, and larger groups indicates either that such patients do not exist, or that they have been dramatically underreported. Organizing a registry to parallel the successful malignant hyperthermia registry should warrant specialty support, perhaps with an eye kept on postsurgical dementia and other postoperative cognitive phenotypes in closed claims analysis. Extreme outliers in prospective observational studies and RCTs comprise a sample of “case report” patients who have done very well or very poorly, and may be consented at the time of enrollment for a more thorough exploration of patient-specific risk factors and comparisons.
Related posts:
Chapter 1 – Emergence Delirium
Chapter 2 – Postoperative Delirium
Chapter 4 – Postoperative Cognitive Improvement
Chapter 12 – Biomarkers of Postoperative Cognitive Dysfunction: Finding the Signal amid the Noise
Chapter 11 – Cognitive Testing for Perioperative Neurocognitive Disorder
Chapter 14 – Preoperative Testing to Identify Vulnerable Subgroups
Full access? Get Clinical Tree
