Postoperative delirium (POD) is among the most common complications after major surgery in older adults, and the acute neurocognitive syndrome of delirium has a bidirectional relationship with chronic cognitive impairment. For instance, even categorical mild cognitive impairment is associated with greater than five times the odds of POD relative to cognitively unimpaired patients (Oldham etal., Reference Oldham2015). Conversely, in the SAGES study, which evaluated cognition up to 3 years after major surgery, POD was associated with subsequent cognitive impairment (Inouye etal., Reference Inouye2016). Whereas the clinical associations between delirium and chronic neurocognitive impairment are well established, the mechanisms underlying these associations need further investigation.
In this issue of International Psychogeriatrics, Wang and colleagues present a systematic review and meta-analysis on the relationship between POD and disease-state biomarkers of Alzheimer’s disease and related dementias (ADRD) (Wang etal., Reference Wang2022). This publication builds upon their prior review (Wang etal., Reference Wang2020) by focusing on the subset of studies specific to POD and evaluating the relationship of each biomarker with POD through meta-analysis. The authors use the National Institute on Aging – Alzheimer’s Association (NIA-AA) AT(N) Diagnostic Framework for Alzheimer’s disease (AD), which categorizes biomarkers as amyloid (A), tau (T), and neurodegeneration (N) while allowing additional biomarkers such as vascular or synuclein to be designated by an X (Jack etal., Reference Jack2018). It is important to clarify that the AT(N) framework was designed for AD alone and that the authors are using X as an extension to cover ADRD’s “related dementias.” This distinction notwithstanding, research on the link between POD and this more inclusive approach to dementia’s varied neuropathological substrates is warranted as it allows for a broader appreciation of disease mechanisms (Khachaturian etal., Reference Khachaturian2020).
As expected, the authors found that both inflammatory (interleukin 6 and C-reactive protein) and neuronal injury biomarkers (serum S100B and neurofilament light chain) are associated with POD. Elevated levels of serum S100B in POD could also point to changes in blood–brain barrier integrity (Taylor etal., Reference Taylor2022). These findings reflect the current inflammatory and neurovascular models of POD (Yang etal., Reference Yang, Velagapudi and Terrando2020) and appear to implicate cerebrovascular disease as a potential link between delirium and dementia.
On the other hand, the authors identified only three studies that investigated the two primary AD biomarkers of amyloid and tau. Interestingly, only two of the three amyloid studies investigated amyloid biomarkers included in the NIA-AA Diagnostic Framework, both in hip fracture surgery cohorts. In the first of these, POD was associated with CSF Aβ42 levels obtained during spinal anesthesia, but only among those without dementia (Idland etal., Reference Idland2017); the absence of association between CSF Aβ42 levels and POD among those with dementia in this study, however, is inconsistent with a dose-dependent relationship that one might anticipate if the amyloid burden were to confer delirium risk. Even more striking, the second amyloid study found no amyloidosis on postoperative PET scans in the five subjects who developed POD, yet amyloidosis was present in 6/11 (55%) of those without POD (Rolandi etal., Reference Rolandi2018). The third amyloid study, in older adults undergoing surgery for oral cancer, evaluated serum Aβ40 and found no differences between those with and without POD, both immediately before and after surgery; of unclear significance, the degree of Aβ40 elevation 12 h after surgery was associated with POD status (Sun etal., Reference Sun2016). The only study evaluating phosphorylated tau in relation to POD failed to find a statistical association (Idland etal., Reference Idland2017). Taken together, these studies raise considerable doubts as to whether amyloid or tau pathology increases the risk of POD.
This, however, is only a part of the equation, perhaps even the lesser part. What these studies do not address is whether the pathophysiology underlying POD triggers permanent neuropathology (e.g. by way of inflammatory or neurovascular insults) or leads to an acceleration of ADRD-related neuropathology including but not limited to amyloid and tau and, if so, on what timescale (see Figure 2 [Wang etal., Reference Wang2022]). Indeed, the issue of whether certain pathophysiological subtypes of delirium, such as POD, lead to permanent neuropathology remains among the most important gaps in the delirium literature to date. This type of mechanistic information is critical to developing interventions to improve delirium’s cognitive outcomes.
An important distinction for context is the largely unexplored relationship between ADRD and the developing model of perioperative neurocognitive disorders (Evered etal., Reference Evered2018), which includes POD, delayed neurocognitive recovery (up to 30 days post-surgery), and either mild or major neurocognitive disorder postoperative, based on DSM-5 criteria (from 30 days to 12 months post-surgery). The model of perioperative neurocognitive disorders is intended to harmonize efforts in the field, to explore the effects of perisurgical events such as anesthesia, tissue injury, inflammatory response, and perioperative cerebral insults on cognitive outcomes. Notably, the word “perioperative” in these terms refers only to the temporal association, not etiology. Further research will be required to align this perioperative model with ADRD to investigate putative mechanisms. (Evered etal., Reference Evered2018).
Despite the limitations in the literature, this review by Wang and colleagues emphasizes the importance of brain mechanisms linking delirium and ADRD. It also reveals a worryingly high heterogeneity of results for nearly every biomarker studied; the authors’ Table 2 reports I 2 values ranging from 30.8% to 96.4% with only one outlier, one with no relationship with POD (i.e. interleukin-10, SMD 0.12, 95% CI [−0.14, 0.37]). Further, considering these studies from the perspective of the recently proposed best practice methods for delirium biomarkers (Amgarth-Duff etal., Reference Amgarth-Duff, Hosie, Caplan and Agar2020), one finds several sources of potential variance in terms of timing alone. These include collecting data at different times of the day (introducing circadian variation), variable timing of data collection relative to delirium onset and peak (which is unspecified in many studies), and collecting samples at different times relative to surgery (i.e. different postoperative hours and days). Additionally, for studies with data available at multiple time points, this review’s meta-analysis includes biomarker data with significant results or where available in the largest sample size to maximize statistical power, but even this pragmatic decision, though understandable, introduces further variance into the results.
While interpreting these results, it is also important to keep in mind the distinction between pathophysiology and neuropathology. This is a review of biomarkers reflecting ADRD neuropathology, yet how the neuropathology associated with these biomarkers is mechanistically associated with the acute physiological disturbance of brain network function in delirium – or the pathophysiology specific to POD – remains unclear. Clinicopathological concordance is an imperfect science, and even in AD it is challenging to connect simple morphometry and neuropsychiatric phenotypes (Siafarikas etal., Reference Siafarikas2021), fully separate from the physiological events that occur in the perioperative setting.
However, even if POD and specific ADRD biomarkers assayed in the perioperative setting are not as closely aligned as one might suspect, the findings of this review and meta-analysis call for study on longitudinal outcomes. Moreover, the fields of delirium and ADRD have much to gain from one another. Such a collaboration represents an intuitive point of connection between acute and chronic expressions of neurocognitive impairment.
First, consider the benefits to the delirium field. The ADRD literature is far more robust than the delirium literature in terms of biological and phenotypic subtyping, both of which are directions that the delirium field would do well to adopt. ADRD is categorized based on underlying neuropathology; by analogy, future delirium models should explore biological subtypes because not all delirium is pathophysiologically the same (Maldonado, Reference Maldonado2018). Specifically, regarding POD, it is also unclear to what extent the pathophysiology of POD overlaps with that of other commonly encountered deliria, such as delirium due to sepsis or hepatic encephalopathy. Also, the neuropsychiatric symptoms of ADRD are especially distressing and functionally impairing, with several discrete syndromes described (Gerlach and Kales, Reference Gerlach and Kales2020). The delirium field should also consider adapting this approach to define the behavioral and psychological symptoms of delirium (Oldham, Reference Oldham2022).
Second, consider the benefits of collaboration to the ADRD field. Understanding the factors that contribute to disease progression and identifying means of preserving cognitive and functional status are foundational to memory care. Yet, the ADRD field might gain crucial mechanistic insights on neuropathological progression by studying the role of the acute neurophysiological insults that underlie different deliria. POD typically represents a nadir of cognitive function after surgery with resultant recovery over time, though not all patients return to their presurgical baseline. The postoperative model of delirium offers a unique opportunity to characterize a sample before anticipated physiological events, record relevant biological factors surrounding the event, and evaluate longitudinally. Emerging blood biomarkers for ADRD such as phosphorylated tau 181 (Karikari etal., Reference Karikari2020) could also be useful to integrate information from the perioperative setting with longitudinal data before and after surgery. Provocatively, if certain nonpathogenic types of delirium are identified (e.g. certain medication-induced states), one wonders whether a cognitive stress test may be feasible, as an analog to cardiac stress tests (Khachaturian etal., Reference Khachaturian2020).
A subtle element of advocacy also comes across in this review, even if largely by implication. The segregation of delirium and dementia research is all too common. In fact, it is reminiscent of that between the fields of delirium and encephalopathy (Slooter etal., Reference Slooter2020), but it is likely no less impactful. The pressing call is for greater integration of delirium and dementia research and integrated memory care. Granted, delirium is important on its own for the considerable distress and danger it poses to patients and those around them, but one of its most dire consequences is that it often predicts dementia in a dose-dependent fashion (Richardson etal., Reference Richardson2021).
As this review highlights, there is much we do not yet know about this relationship or the potential mechanisms underlying it. Whereas many questions remain, the importance of this relationship is unquestioned. Systematic delirium recognition is a means of enhancing detection of not only current cognitive impairment but also of many who will develop dementia over the ensuing months and years. Memory care should be seen as essential in both inpatient and outpatient settings: from routine screening to comprehensive care delivery. Such care should also be integrated across levels of care because acute cognitive impairment in the hospital often has permanent implications after discharge.
You have
Access
Postoperative delirium (POD) is among the most common complications after major surgery in older adults, and the acute neurocognitive syndrome of delirium has a bidirectional relationship with chronic cognitive impairment. For instance, even categorical mild cognitive impairment is associated with greater than five times the odds of POD relative to cognitively unimpaired patients (Oldham etal., Reference Oldham2015). Conversely, in the SAGES study, which evaluated cognition up to 3 years after major surgery, POD was associated with subsequent cognitive impairment (Inouye etal., Reference Inouye2016). Whereas the clinical associations between delirium and chronic neurocognitive impairment are well established, the mechanisms underlying these associations need further investigation.
In this issue of International Psychogeriatrics, Wang and colleagues present a systematic review and meta-analysis on the relationship between POD and disease-state biomarkers of Alzheimer’s disease and related dementias (ADRD) (Wang etal., Reference Wang2022). This publication builds upon their prior review (Wang etal., Reference Wang2020) by focusing on the subset of studies specific to POD and evaluating the relationship of each biomarker with POD through meta-analysis. The authors use the National Institute on Aging – Alzheimer’s Association (NIA-AA) AT(N) Diagnostic Framework for Alzheimer’s disease (AD), which categorizes biomarkers as amyloid (A), tau (T), and neurodegeneration (N) while allowing additional biomarkers such as vascular or synuclein to be designated by an X (Jack etal., Reference Jack2018). It is important to clarify that the AT(N) framework was designed for AD alone and that the authors are using X as an extension to cover ADRD’s “related dementias.” This distinction notwithstanding, research on the link between POD and this more inclusive approach to dementia’s varied neuropathological substrates is warranted as it allows for a broader appreciation of disease mechanisms (Khachaturian etal., Reference Khachaturian2020).
As expected, the authors found that both inflammatory (interleukin 6 and C-reactive protein) and neuronal injury biomarkers (serum S100B and neurofilament light chain) are associated with POD. Elevated levels of serum S100B in POD could also point to changes in blood–brain barrier integrity (Taylor etal., Reference Taylor2022). These findings reflect the current inflammatory and neurovascular models of POD (Yang etal., Reference Yang, Velagapudi and Terrando2020) and appear to implicate cerebrovascular disease as a potential link between delirium and dementia.
On the other hand, the authors identified only three studies that investigated the two primary AD biomarkers of amyloid and tau. Interestingly, only two of the three amyloid studies investigated amyloid biomarkers included in the NIA-AA Diagnostic Framework, both in hip fracture surgery cohorts. In the first of these, POD was associated with CSF Aβ42 levels obtained during spinal anesthesia, but only among those without dementia (Idland etal., Reference Idland2017); the absence of association between CSF Aβ42 levels and POD among those with dementia in this study, however, is inconsistent with a dose-dependent relationship that one might anticipate if the amyloid burden were to confer delirium risk. Even more striking, the second amyloid study found no amyloidosis on postoperative PET scans in the five subjects who developed POD, yet amyloidosis was present in 6/11 (55%) of those without POD (Rolandi etal., Reference Rolandi2018). The third amyloid study, in older adults undergoing surgery for oral cancer, evaluated serum Aβ40 and found no differences between those with and without POD, both immediately before and after surgery; of unclear significance, the degree of Aβ40 elevation 12 h after surgery was associated with POD status (Sun etal., Reference Sun2016). The only study evaluating phosphorylated tau in relation to POD failed to find a statistical association (Idland etal., Reference Idland2017). Taken together, these studies raise considerable doubts as to whether amyloid or tau pathology increases the risk of POD.
This, however, is only a part of the equation, perhaps even the lesser part. What these studies do not address is whether the pathophysiology underlying POD triggers permanent neuropathology (e.g. by way of inflammatory or neurovascular insults) or leads to an acceleration of ADRD-related neuropathology including but not limited to amyloid and tau and, if so, on what timescale (see Figure 2 [Wang etal., Reference Wang2022]). Indeed, the issue of whether certain pathophysiological subtypes of delirium, such as POD, lead to permanent neuropathology remains among the most important gaps in the delirium literature to date. This type of mechanistic information is critical to developing interventions to improve delirium’s cognitive outcomes.
An important distinction for context is the largely unexplored relationship between ADRD and the developing model of perioperative neurocognitive disorders (Evered etal., Reference Evered2018), which includes POD, delayed neurocognitive recovery (up to 30 days post-surgery), and either mild or major neurocognitive disorder postoperative, based on DSM-5 criteria (from 30 days to 12 months post-surgery). The model of perioperative neurocognitive disorders is intended to harmonize efforts in the field, to explore the effects of perisurgical events such as anesthesia, tissue injury, inflammatory response, and perioperative cerebral insults on cognitive outcomes. Notably, the word “perioperative” in these terms refers only to the temporal association, not etiology. Further research will be required to align this perioperative model with ADRD to investigate putative mechanisms. (Evered etal., Reference Evered2018).
Despite the limitations in the literature, this review by Wang and colleagues emphasizes the importance of brain mechanisms linking delirium and ADRD. It also reveals a worryingly high heterogeneity of results for nearly every biomarker studied; the authors’ Table 2 reports I 2 values ranging from 30.8% to 96.4% with only one outlier, one with no relationship with POD (i.e. interleukin-10, SMD 0.12, 95% CI [−0.14, 0.37]). Further, considering these studies from the perspective of the recently proposed best practice methods for delirium biomarkers (Amgarth-Duff etal., Reference Amgarth-Duff, Hosie, Caplan and Agar2020), one finds several sources of potential variance in terms of timing alone. These include collecting data at different times of the day (introducing circadian variation), variable timing of data collection relative to delirium onset and peak (which is unspecified in many studies), and collecting samples at different times relative to surgery (i.e. different postoperative hours and days). Additionally, for studies with data available at multiple time points, this review’s meta-analysis includes biomarker data with significant results or where available in the largest sample size to maximize statistical power, but even this pragmatic decision, though understandable, introduces further variance into the results.
While interpreting these results, it is also important to keep in mind the distinction between pathophysiology and neuropathology. This is a review of biomarkers reflecting ADRD neuropathology, yet how the neuropathology associated with these biomarkers is mechanistically associated with the acute physiological disturbance of brain network function in delirium – or the pathophysiology specific to POD – remains unclear. Clinicopathological concordance is an imperfect science, and even in AD it is challenging to connect simple morphometry and neuropsychiatric phenotypes (Siafarikas etal., Reference Siafarikas2021), fully separate from the physiological events that occur in the perioperative setting.
However, even if POD and specific ADRD biomarkers assayed in the perioperative setting are not as closely aligned as one might suspect, the findings of this review and meta-analysis call for study on longitudinal outcomes. Moreover, the fields of delirium and ADRD have much to gain from one another. Such a collaboration represents an intuitive point of connection between acute and chronic expressions of neurocognitive impairment.
First, consider the benefits to the delirium field. The ADRD literature is far more robust than the delirium literature in terms of biological and phenotypic subtyping, both of which are directions that the delirium field would do well to adopt. ADRD is categorized based on underlying neuropathology; by analogy, future delirium models should explore biological subtypes because not all delirium is pathophysiologically the same (Maldonado, Reference Maldonado2018). Specifically, regarding POD, it is also unclear to what extent the pathophysiology of POD overlaps with that of other commonly encountered deliria, such as delirium due to sepsis or hepatic encephalopathy. Also, the neuropsychiatric symptoms of ADRD are especially distressing and functionally impairing, with several discrete syndromes described (Gerlach and Kales, Reference Gerlach and Kales2020). The delirium field should also consider adapting this approach to define the behavioral and psychological symptoms of delirium (Oldham, Reference Oldham2022).
Second, consider the benefits of collaboration to the ADRD field. Understanding the factors that contribute to disease progression and identifying means of preserving cognitive and functional status are foundational to memory care. Yet, the ADRD field might gain crucial mechanistic insights on neuropathological progression by studying the role of the acute neurophysiological insults that underlie different deliria. POD typically represents a nadir of cognitive function after surgery with resultant recovery over time, though not all patients return to their presurgical baseline. The postoperative model of delirium offers a unique opportunity to characterize a sample before anticipated physiological events, record relevant biological factors surrounding the event, and evaluate longitudinally. Emerging blood biomarkers for ADRD such as phosphorylated tau 181 (Karikari etal., Reference Karikari2020) could also be useful to integrate information from the perioperative setting with longitudinal data before and after surgery. Provocatively, if certain nonpathogenic types of delirium are identified (e.g. certain medication-induced states), one wonders whether a cognitive stress test may be feasible, as an analog to cardiac stress tests (Khachaturian etal., Reference Khachaturian2020).
A subtle element of advocacy also comes across in this review, even if largely by implication. The segregation of delirium and dementia research is all too common. In fact, it is reminiscent of that between the fields of delirium and encephalopathy (Slooter etal., Reference Slooter2020), but it is likely no less impactful. The pressing call is for greater integration of delirium and dementia research and integrated memory care. Granted, delirium is important on its own for the considerable distress and danger it poses to patients and those around them, but one of its most dire consequences is that it often predicts dementia in a dose-dependent fashion (Richardson etal., Reference Richardson2021).
As this review highlights, there is much we do not yet know about this relationship or the potential mechanisms underlying it. Whereas many questions remain, the importance of this relationship is unquestioned. Systematic delirium recognition is a means of enhancing detection of not only current cognitive impairment but also of many who will develop dementia over the ensuing months and years. Memory care should be seen as essential in both inpatient and outpatient settings: from routine screening to comprehensive care delivery. Such care should also be integrated across levels of care because acute cognitive impairment in the hospital often has permanent implications after discharge.
Conflict of interest
None.