Lifetime Surgical Exposure, Episodic Memory, and Forniceal Microstructure in Older Adults.

Abstract

Introduction:

Aging is associated with heterogeneous cognitive trajectories. There is considerable interest in identifying risk factors for pathological aging, with recent studies demonstrating a link between surgical procedures and proximal cognitive decline; however, the role of lifetime exposure to surgical procedures and cognitive function has been relatively unexplored. This pilot study aimed to evaluate the association between total lifetime surgical procedures and memory function in older adults.

Methods:

A cohort of 62 older adults underwent a neuropsychological evaluation and health history assessment. Self-reported lifetime surgical history was categorized as ‘cardiac’ or ‘non-cardiac.’ General linear models were fit with demographics as nuisance covariates, and the total number of non-cardiac surgeries as our predictor of interest. Total scores on measures of episodic memory, language, working memory, fluency, and visuospatial function were separate outcome variables. In a secondary analysis, vascular risk factors were included as covariates. Diffusion tensor imaging was obtained for exploratory analyses of selected regions of interest.

Results:

The mean age of participants was 70, and 0–13 lifetime non-cardiac surgical procedures were reported. Higher numbers of lifetime non-cardiac surgical procedures were associated with worse verbal learning and memory (p = .04). The negative association between lifetime non-cardiac procedures and cognition was specific to memory. Exploratory analyses showed that higher number of lifetime non-cardiac procedures was related to lower FA in the fornix body (p = .02).

Conclusions:

These results of this pilot study suggest that greater lifetime exposure to surgery may be associated with worse verbal learning and memory in healthy older adults. These findings add to a growing body of literature suggesting that cumulative medical events may be risk factors for negative cognitive outcomes.

Introduction

Normal cognitive aging is characterized by heterogeneous cognitive trajectories, with some individuals displaying decline in memory, executive functions, and processing speed, while others experience a stable course. The presence of Alzheimer’s pathology and the APOE e4 allele are both predictors of more substantial cognitive decline (Hayden et al., 2011; Mungas et al., 2010); however, it is also estimated that up to 35% of dementia may be due to potentially modifiable risk factors present in early-, mid-, and later-life (Livingston et al., 2017). As such, there has been increasing interest in understanding modifiable risk factors that may presage cognitive decline.

Surgery is a potential exposure which has been linked to cognitive dysfunction. The majority of this literature has focused on proximal changes to cognition immediately following exposure to surgery, referred to as post-operative cognitive decline (POCD; Evered and Silbert, 2018; Price, Garvan, & Monk, 2008). Although this remains an area of great interest, the role of cumulative non-cardiac surgical exposure temporally distant from cognitive outcomes is much more challenging to study, and literature is sparse. Moreover, while there is a comprehensive literature linking cardiac procedures and associated vascular risk factors with late life cognitive decline (Newman et al., 2001; Roach et al., 1996; Selnes et al., 2012; Selnes et al., 2009; Selnes et al., 2001), few studies have examined the cumulative relationship between lifetime non-cardiac surgical procedures and cognitive function in older adults. A single community-based cohort study of residents in Vienna, Austria examined the effect of combined exposure to surgery and general anesthesia (GA; i.e., minimum of 10 minutes of GA) in 606 subjects, irrespective of the surgery type. They found no association between cognitive outcomes and lifetime exposure to combined surgery and anesthesia (Fischer et al., 2007). Study inclusion criteria required an exposure to GA, which likely restricted the type and range of medical events that would necessitate such a procedure. Type of anesthesia does not appear to play a significant role in the development of cognitive difficulties, however, as patients undergoing cardiac procedures with conscious sedation and general anesthesia had similar rates of cognitive dysfunction (Evered, Scott, Silbert, & Maruff, 2011).

Although a dearth of human studies have addressed the relationship between cumulative non-cardiac surgical events and cognitive outcomes, there are important mechanistic reasons to suggest that lifetime exposure to procedures may induce risk for later cognitive decline. In preclinical studies, peripheral surgeries (akin to orthopedic surgeries) have been associated with a systemic inflammatory cascade that directly impairs hippocampal-related memory functions (Cibelli et al., 2010; Rosczyk, Sparkman, & Johnson, 2008; Terrando, Monaco, et al., 2010). This suggests that mere exposure to a non-cardiac procedure may be sufficient to induce immune-related changes in brain structure and memory consolidation. In humans, surgery has also been shown to result in systemic inflammation that has effects on the CNS (Saxena and Maze, 2018); moreover, systemic inflammation has been associated with diminished memory performance in older adults (Bettcher et al., 2016; Bettcher et al., 2019). Taken together, this suggests that repeated exposure to surgery over the life course may have an effect on memory performance and the neuroanatomic substrate of memory functions.

In the current pilot study, we examined the association between lifetime, non-cardiac surgical procedures and cognitive function in a group of clinically healthy, community dwelling older adults. In order to limit the confound of cerebrovascular events on surgical procedures, we elected to focus our analyses on non-cardiac procedures. Given the sensitivity of episodic memory function to cognitive aging, inflammation, and Alzheimer’s disease (Ewers et al., 2012; Kramer et al., 2003), as well as its link to animal models of non-cardiac surgical exposure on cognitive outcomes (Cibelli, et al., 2010; Rosczyk, et al., 2008), our primary analyses were focused on associations with verbal memory performance; however, we also examined associations with non-memory domains to elucidate possible changes that were not included in our initial hypotheses. We further examined whether associations were driven by vascular risk factors. In exploratory analyses, we investigated whether non-cardiac procedures were deleteriously related to white matter integrity, with a specific focus on white matter regions associated with verbal memory performance (i.e., fornix), as forniceal damage has been shown to result in episodic memory impairment (Squire and Zola-Morgan, 1991), and atrophy of the fornix is observed in patients with mild cognitive impairment and Alzheimer’s disease (Douaud et al., 2013; Ringman et al., 2007). We elected to focus on white matter integrity due to substantial evidence supporting a role of white matter pathology in the pathogenesis of dementia including Alzheimer’s disease (Bettcher et al., 2015; Brickman, 2013; Filley and Fields, 2016; Filley, Franklin, Heaton, & Rosenberg, 1988) and the high number of microglia, which are the brain’s resident immune cells, in white matter (Mittelbronn, Dietz, Schluesener, & Meyermann, 2001).

Methods

Participants

A sample of 65 clinically healthy, non-demented, community dwelling older adult participants was selected from the Rocky Mountain Alzheimer’s Disease Center’s database (see Table 1). All participants were enrolled in the Bio-AD study, which is a longitudinal cohort study that entails comprehensive cognitive testing, health history assessment, neurological and physical examination, informant interview, and brain MRI scans of all participants. Inclusion as a “clinically healthy” participant in this study was based on several criteria, including a Clinical Dementia Rating score of 0, no informant report of significant cognitive decline during the previous year, and consensus conference adjudication as a normal control. Participants were excluded if they had a major psychiatric disorder, current neurological condition known to affect cognition (e.g., Parkinson’s disease; large vessel infarct; Multiple Sclerosis), current evidence or history in the past 2 years of a focal brain lesion, dementia or mild cognitive impairment, current substance abuse, significant systemic medical illness or active neoplastic disease (e.g., current cancer), significant sensory or motor deficits that would interfere with cognitive testing, or traumatic brain injury with loss of consciousness greater than 5 minutes.

Table 1.

Participant (N = 62) Demographics and Characteristics

	Mean (SD)	Range
Demographics
Age	70 (6.7)	53 to 83
Sex (% Female)	68%
Education (Years)	17 (2.2)	12 to 20
Health History
Total Lifetime Non-Cardiac Procedures	3.5 (2.7)	0 to 13
History of Cardiac Procedures (% Present)	6%
Body Mass Index (BMI)	26 (3.5)	19 to 35
History of Hypertension (% Present)	28%
History of Hyperchoesterolemial (%)	63%
Cognitive Assessment
MoCA (Total Score)	27 (2.0)	22 to 30
SENAS Verbal Memory Composite	0.98 (0.60)	−0.78 to 2.45
SENAS Working Memory Composite	0.71 (0.61)	−0.45 to 2.85
SENAS Spatial Composite	1.04 (0.69)	−0.70 to 2.44
SENAS Phonemic Fluency Composite	0.58 (0.69)	−1.19 to 2.03
SENAS Language/Semantic Composite	1.86 (0.65)	0.56 to 3.21

Legend: SENAS= Spanish and English Neuropsychological Assessment Scales

SENAS scores represent IRT unadjusted standard scores based on a demographically diverse normative sample

For the purposes of the current study, participants were selected based on the availability of memory testing and surgical history data. In addition to measures of episodic memory and global cognitive function, all participants also underwent cognitive testing that included measures of executive functions, language, and visuospatial function. These measures were included in secondary analyses (see below) to evaluate the specificity of our findings.

All participants were reviewed at a case conference with a board-certified neuropsychologist (BMB), neurologist, and clinical research coordinator. A subset of cognitive measures from the research protocol were reviewed in a consensus conference and used in the adjudication of diagnosis as a clinically normal older adult; however, to reduce circularity in our methodological approach, cognitive measures that were reviewed in the consensus conference for adjudication of diagnosis were separate from those used as primary outcomes in our cognitive research. Of note, one- to two-year diagnostic follow-up data were available for the majority of individuals in the study; three individuals in the participant sample were subsequently diagnosed with Mild Cognitive Impairment in either the following year or two years later. These individuals were excluded from the current study in order to ensure that our participant sample was clinically normal (total sample size=62).

The study was approved by the Colorado Multiple Institutional Review Board (COMIRB), and all subjects provided written, COMIRB-approved informed consent before participating.

Cognitive Assessment

All participants were administered the Montreal Cognitive Assessment (MoCA; Nasreddine et al., 2005), which is a measure of global cognitive function. In addition, participants underwent a comprehensive research neuropsychological assessment battery as part of the Bio-AD study. The cognitive measures utilized in this study were tests of episodic memory, language/semantic knowledge, attention/working memory, phonemic fluency, and spatial ability from the Spanish and English Neuropsychological Assessment Scales (SENAS). Notably, construction of the SENAS battery was based on item response theory (IRT), and psychometrically matched measures were created across different scales, and thus assuring reliability across the full ability continuum (Mungas, Reed, Crane, Haan, & Gonzalez, 2004; Mungas, Reed, Haan, & Gonzalez, 2005; Mungas, Reed, Marshall, & Gonzalez, 2000; Mungas, Reed, Tomaszewski Farias, & DeCarli, 2005; Mungas, Widaman, Reed, & Tomaszewski Farias, 2011). For the purposes of this study, IRT composite scores were used for each of the domains described below. These IRT scores do not have floor or ceiling effects and are normally distributed. IRT scores may be interpreted as unadjusted standard scores (Mean=0; SD=1) based on a demographically diverse sample of older adults aged 60+ (Mungas, et al., 2004; Mungas, et al., 2011).

Our primary cognitive outcome was verbal episodic memory and was selected due to its associations with negative cognitive aging trajectories, inflammation, and early stages of Alzheimer’s disease (Ewers, et al., 2012; Kramer, et al., 2003). The Verbal Memory IRT composite score served as the primary variable, and was assessed with a multi-trial list-learning measure from the SENAS (5 learning trials; 15 items), incorporating both learning trials and delayed recall.

Secondary cognitive measures from the SENAS included the language/semantic knowledge IRT composite, which is based on scores from a nonverbal picture association measure and verbal object naming task. To assess executive functions, we included SENAS measures of working memory (Working Memory IRT Composite: digit span backward, visual span backward, list sorting) and fluency (i.e., Phonemic Fluency IRT Composite: letter fluency measures, “F” and “L”). Finally, spatial ability was measured using the SENAS Spatial IRT composite (i.e., Spatial Localization) scale, which evaluates the ability to perceive and reproduce two-dimensional spatial relationships. Stimuli includes a page, with the top half showing a dog burying a bone, next to a doghouse and tree; the bottom half of the page is the target, and the participant is required to point to the perceived buried bone, based on location of other elements). Administration procedures, measure development, and psychometric characteristics of the SENAS battery are described in detail elsewhere (Mungas, et al., 2004).

Health History Assessment

All participants underwent a neurological and physical examination. Body mass index (BMI) was calculated using the formula [weight (kg)/height (m)²], and analyzed as a continuous variable. All available information regarding medical and health history was garnered via participant self-report. Specifically, to assess vascular risk factors, all participants were queried about their history of hypertension and hypercholesterolemia (i.e., presence or absence). With respect to surgical history, all participants were queried in a short, open-ended survey form about their history of surgical procedures and asked to provide the following information: history of any (i.e., presence or absence) surgical procedures; description of procedure; and date of procedure. Answers were then categorized as ‘cardiac’ or ‘non-cardiac’, based strictly on whether the surgery involved a cardiac procedure (e.g., aortic valve replacement, cardiac stent placement; see Supplemental Table). In order to parse out proximal versus distal procedures, an additional variable of ‘time since procedure’ in years was created.

Neuroimaging (Exploratory Analyses)

Whole brain MRI scans were obtained on a 3.0 Tesla Siemens (Iselin, NJ) Skyra scanner equipped with a 20-channel head coil. Diffusion imaging data were acquired via a spin-echo, echo planar imaging sequence with 56 slices 2.2 mm thick (TR/TE = 8400/105 ms, matrix = 112×112) in monopolar series (multi-shell; B2500: 64 diffusion directions, B0 [9 averages]).

Diffusion tensor imaging (DTI) data were preprocessed and analyzed using FMRIB’s Diffusion Toolbox (FDT) (Smith et al., 2006). Raw data were corrected for head movement and eddy current distortions using FDT. Brain extraction and binary brain mask creation used the b0 mean image through the FMRIB Software Library (FSL) Brain Extraction Tool (BET). Fractional anisotropy (FA) maps were created using FSL DTIFIT. All subjects’ FA data were registered using the nonlinear registration FNIRT to the IXI Aging DTI Template (Zhang, Yushkevich, Rueckert, & Gee) masked by a study-specific average image. For region of interest (ROI) analyses, we employed the Johns Hopkins University ICBM-DTI-81 white matter labels (Mori et al., 2008) to label and mask areas of the white matter ROIs. Mean FA values for each white matter region was calculated using the FSL utility fslstats.

Given our interest in the association between cumulative surgical history and episodic memory, we focused primarily on a temporal lobe (i.e., fornix body) ROI. We also included a control region as a means of comparison; for data reduction purposes, we calculated a mean value for tracts in the brain stem, which included an average of the following: corticospinal tract, cerebellar peduncles (inferior, middle, and superior), medial lemniscus, and pontine fiber tracts (Brickman, 2012). Brain stem tracts were selected as the control region given their limited association with episodic memory and higher level cognition. Finally, we also examined whole brain FA to explore more diffuse white matter associations with lifetime surgical history.

Statistical Analyses

Primary analyses: All analyses were adjusted for demographic variables (i.e., age, sex, and education). To address the initial study goal (i.e., association between cumulative non-cardiac surgical procedures and episodic memory), we fit general linear models with demographics as covariates and the total number of non-cardiac surgical procedures as the primary predictor of interest. The SENAS Verbal Memory IRT composite served as the outcome variable. We appraised whether global cognitive function impacted the association by including performance on the MoCA as a covariate.

In a follow-up analysis, we included prior cardiac procedures as a covariate. In order to assess whether the association between cumulative non-cardiac procedures and episodic memory was independent of canonical vascular risk factors, we repeated the regression analyses with a history of hypertension (presence vs. absence), history of hypercholesterolemia (presence vs. absence), and current BMI as covariates. To assess whether the association between cumulative non-cardiac surgical procedures and cognition extended beyond the domain of episodic memory, we repeated all regression analyses with the language/semantic memory IRT composite, working memory IRT composite, phonemic fluency IRT composite, and visuospatial function IRT composite as separate outcome variables.

Neuroimaging Exploratory Analyses

For exploratory analysis, we conducted general linear models with fornix body as the primary outcome variable, demographics and global cognitive function (MoCA) as nuisance covariates, and lifetime non-cardiac surgical history as the primary predictor. We repeated the analysis with our control white matter FA ROI (i.e., brain stem average), as well as whole brain FA.

Results

Overall, our sample consisted of well-educated, predominantly female older adults (mean age = 70; see Table 1). The majority of participants had undergone more than 1 non-cardiac surgical procedure, with a lifetime range of 0 to 13 procedures. Very few participants had undergone a cardiac surgical procedure (mean <1); as such, we dichotomized this variable to reflect the presence or absence of a history of cardiac procedures.

Question 1: Are total lifetime non-cardiac surgeries related to memory?

Controlling for demographic variables, greater number of lifetime non-cardiac surgical procedures was related to worse memory performance on the SENAS (unstandardized estimate of total non-cardiac procedures: −0.06; 95% confidence interval (CI) [−0.115, 0.002]; standard error (SE) = 0.03; t = −2.09; p = .04).

To determine whether these findings were driven by a proximal history of non-cardiac procedures, we examined the number of surgical procedures that occurred within the past 5 years. Of note, 51% of the sample did not have a surgical procedure in the prior 5 years. This datum suggests that our analyses were not driven solely by proximal events.

We also examined whether the association between higher total non-cardiac procedures and verbal memory was independent of global cognitive function. Upon controlling for global cognitive function (i.e., MoCA total score), non-significant trends were noted (p = .14). Of note, despite average unadjusted standard scores for memory performance amongst our sample, memory performance and total MoCA score were moderately correlated, raising concerns for the interpretation of our regression analyses. As such, we further assessed whether associations between total non-cardiac procedures and verbal memory remained when controlling for non-memory SENAS composite scores. Upon controlling for the four non-memory composites (i.e., language/semantic; phonemic fluency; working memory; spatial), significant associations between higher total non-cardiac procedures and worse memory performance remained (unstandardized estimate = −0.08; SE = 0.03; 95% CI [−0.132, −0.025]; t = −2.96; p = .005).

Question 2: Are total lifetime non-cardiac surgeries associated with memory, independent of a history of cardiac procedures (See Table 2)?

Table 2.

Linear Regression Results: Higher Total Lifetime Non-Cardiac Surgical Procedures are Related to Worse Memory (N = 62)

Effect	Estimate	SE	95% CI	T-Value	P-Value
Intercept	1.645	0.975	−0.865 to 3.226	1.69	0.097
Age	−0.012	0.011	−0.035 to 0.010	−1.09	0.282
Female Sex	0.465	0.156	0.153 to 0.777	2.98	0.004
Education	0.032	0.033	−0.034 to 0.099	1.23	0.336
Cardiac Procedures (Presence of)	0.369	0.300	−0.233 to 0.971	1.23	0.224
Total Non-Cardiac Procedures	−0.065	0.028	−0.122 to −0.008	−2.29	0.026

Adjusting for demographic variables and the presence or absence of a history of cardiac procedures did not substantively impact the results of the analyses (unstandardized estimate of total non-cardiac procedures: −0.07; SE = 0.03; 95% CI [−0.122, −0.008]; t = −2.29; p = .03).

Question 3: Does this association remain after controlling for vascular risk factors?

To determine whether vascular risk factors influenced the association between non-cardiac surgical procedures and memory function, we included canonical vascular measures as covariates in our analyses. Results remained significant when including these variables (unstandardized estimate of total non-cardiac procedures: −0.074; SE = 0.03; 95% CI = [−0.133, −0.016]; t = −2.54; p = .014).

Question 4: Is the deleterious cognitive effect of higher non-cardiac surgical procedures specific to memory?

We also investigated whether a lifetime history of these procedures was related to worse performance in other cognitive domains. No significant associations were noted between total non-cardiac procedures and IRT composites of working memory (p = .81), phonemic fluency (p = .23), spatial localization (p = .13), or semantic knowledge (p = .64).

Exploratory Analyses: Are lifetime non-cardiac surgical procedures associated with white matter integrity (Figure 1)?

Figure 1.

Association between total non-cardiac procedures and fornix fractional anisotropy (FA). The y-axis displays the fornix FA, residualized for demographic variables and MoCA total score.

Based on our primary analyses demonstrating an association between higher lifetime non-cardiac surgical procedures and verbal memory, we examined associations between total non-cardiac surgical procedures and a medial temporal lobe region of interest (i.e., fornix body) in exploratory analyses. Controlling for demographic variables, higher lifetime non-cardiac procedures were related to lower FA in the fornix body (t = −2.50, p = .02; see Figure 1), and the association remained when adding a history of cardiac procedures as a predictor (t = −2.33, p = .02). A significant interaction between sex and non-cardiac procedures was noted, such that the association between higher non-cardiac procedures and lower FA in the fornix was primarily driven by women (interaction: p = .03; see Figure 2)

Figure 2.

Displays the association between total non-cardiac surgical procedures and mean Fornix FA (residualized for demographics and MoCA), as a function of sex. 0 represents males, 1 represents females.

No associations between lifetime non-cardiac procedures and our control white matter region (i.e., mean brain stem tract FA) were noted (p = .47). However, a trend was observed for higher lifetime non-cardiac surgical procedures and lower mean whole brain FA (p = .07)

Discussion

Results from this pilot study suggest that in a cohort of healthy, non-demented, typical aging subjects, there is a negative association between the number of lifetime non-cardiac surgical exposures and learning and memory performance. This negative association was independent of a history of cardiac surgery and common vascular risk factors. Finally, we demonstrated that this association was specific to memory and was not noted in non-memory cognitive domains. In an exploratory analysis we examined potential structural explanations for our findings by examining white matter microstructure using DTI. We found that higher lifetime non-cardiac procedure exposure was related to lower FA in the body of the fornix, an association that remained present when controlling for a history of cardiac procedures. There was a significant interaction between sex and non-cardiac procedures whereby the association between higher non-cardiac procedures and lower FA in the fornix was driven primarily by women.

Our results showed some specificity to a single cognitive domain, with higher numbers of lifetime non-cardiac surgical procedures related solely to verbal episodic memory. Of note, while it was not fully independent of global cognitive function (i.e., MoCA total score; non-significant trends), the negative relationship between total non-cardiac procedures and verbal memory remained when controlling for non-memory cognitive measures (i.e., language, spatial function, working memory, and phonemic fluency). As noted in the results, we believe that the discrepancy is likely due to the strong association between memory performance and MoCA score. Memory complaints are common in older adults without dementia (Cargin, Maruff, Collie, Shafiq-Antonacci, & Masters, 2007) and are sensitive to the presence of Alzheimer’s disease (Weintraub, Wicklund, & Salmon, 2012). It is therefore important to understand the earliest signs of memory changes and contributions to trajectories of decline. Our study participants were all clinically healthy and their cognitive test performance (see Table 1) was within normal limits based on a reference group; yet, lower performance on a verbal memory test – even when in the normal range - was associated with the lifetime non-cardiac surgical exposure.

To ensure that our findings were not due strictly to proximal surgical exposure, we examined whether procedures in the past 5 years were driving the results. Of note, 51% of the sample had not undergone a surgical procedure in the past 5 years. Given the low rate of proximal events, we did not have the range to evaluate effects of ‘recent’ procedures. Nonetheless, our data tentatively suggest that the results were not driven by recent surgical exposures. The literature is varied on the meaning of more proximal cognitive dysfunction after surgery; some contend that POCD is a direct causal effect of exposure to operative conditions and associated physiologic changes (Skvarc et al., 2018), while others have shown that the strongest predictors of POCD are age, education, and pre-operative cognitive dysfunction (Avidan et al., 2009; Monk et al., 2008; Silbert et al., 2015). In combination with studies calling into question the presence of long-term cognitive dysfunction after surgery (Avidan and Evers, 2016; Dokkedal et al., 2018; Hughes et al., 2017), it may be that temporary physiologic changes produce transient dysfunction that leads to subtle cognitive changes, such as those we found, only after cumulative exposure.

To assess whether our major cognitive finding was not simply due to a spurious correlation with vascular risk factors, we controlled for hypertension, hypercholesterolemia, and BMI, and found a similar result with comparable effects. The relative specificity to episodic memory, in tandem with the lack of association with executive functions, also points away from a purely vascular etiology of cognitive dysfunction, which most often produces a frontal-subcortical profile rather than isolated episodic memory changes. Notably, peripheral insults and elevated peripheral inflammatory markers have been associated with a reduction in hippocampal volume (Marsland, Gianaros, Abramowitch, Manuck, & Hariri, 2008), a finding that would be expected to result in changes to episodic memory. Executive dysfunction suggests a different site of involvement, and thus the modality specificity is important in suggesting neuroanatomic specificity.

In exploratory analyses, the total number of lifetime non-cardiac surgeries was associated with lower FA in the body of the fornix. The fornix is a critical white matter pathway that serves as the primary efferent pathway from the hippocampus and is part of the Papez circuit (Papez, 1937) that is critical for learning and episodic memory (Douet and Chang, 2014). Indeed, injury to the fornix has been associated with impaired episodic memory (Squire and Zola-Morgan, 1991), and atrophy of the fornix has been observed in mild cognitive impairment and Alzheimer’s disease (Douaud, et al., 2013; Ringman, et al., 2007). Studies across the healthy aging lifespan have found worse episodic memory in subjects with lower volume or lower FA in the fornix (Douet et al., 2014; Fletcher et al., 2013; Rudebeck et al., 2009). The negative association between lifetime surgical procedures and episodic memory prompted us to evaluate the structural integrity of the fornix in exploratory analyses, and we found that a higher number of non-cardiac surgical procedures was associated with lower FA values in this region, with no effects found in our control region (i.e., mean brain stem tracts). Although these results provide preliminary support for temporal lobe white matter anatomy underlying the deleterious association between lifetime non-cardiac surgical history and episodic memory, we also noted a trend with whole brain FA, raising questions regarding specificity. Given the pilot nature of the study, future work should examine whether more diffuse white or gray matter alterations are associated with lifetime surgical procedures, and whether these changes mediate associations with memory.

In interpreting the results, it is important to consider possible pathogenic mechanisms by which lifetime exposure to surgical procedures might be related to verbal memory. The results of this pilot study demonstrated associations with verbal memory and no associations with non-memory cognitive domains, suggesting some degree of neuroanatomic specificity. Prior animal studies have noted an association between peripheral insults and memory performance (Cibelli, et al., 2010; Rosczyk, et al., 2008; Terrando, Rei Fidalgo, et al., 2010). Specifically, surgeries have been shown to induce an inflammatory cascade that crosses the blood-brain barrier in both animals (Cibelli, et al., 2010; Rosczyk, et al., 2008; Terrando, Monaco, et al., 2010) and humans (Buvanendran et al., 2006; Reinsfelt et al., 2012; Tang et al., 2011). In humans, higher levels of systemic inflammation are associated with worse white matter microstructure (Bettcher et al., 2013; Wersching et al., 2010), and a longitudinal decline in inflammatory markers has likewise been associated with better markers of white matter microstructure (Bettcher, et al., 2015). In this regard, it is of considerable interest that microglia, the innate immune cells of the brain, are far more abundant in white than gray matter (Mittelbronn, et al., 2001). POCD appears to share overlapping mechanisms and our findings may represent a subclinical cumulative effect of past episodes of POCD. However, due to the lack of routine neuropsychological testing after surgery, we do not know how many of our subjects ever experienced transient POCD. Taken together, these observations thus provide a potential inflammatory mechanism for our findings, which are further supported by prior work demonstrating negative associations between circulating cytokines/chemokines and memory performance in healthy individuals (Bettcher, et al., 2016; Bettcher, et al., 2019). Indeed, given the increasingly recognized role of white matter pathology in dementia (Bettcher, et al., 2015; Brickman, 2013; Filley and Fields, 2016; Filley, et al., 1988), neuroinflammation in the fornix and other tracts may be a critical feature early in the pathogenesis of Alzheimer’s Disease.

We found a significant interaction between sex and non-cardiac procedures, with the association between higher non-cardiac procedures and lower forniceal FA driven primarily by women. Notably, women have approximately twice the risk of developing Alzheimer’s disease compared to men (Seshadri et al., 1997). The reason for this increased risk is not entirely clear, but possible explanations include estrogen variation and immune dysregulation, both of which are higher in women (Hall et al., 2013; Podcasy and Epperson, 2016). Differences in mitochondrial and white matter proteins have also been described in women with Alzheimer’s disease (Gallart-Palau et al., 2016). These potential biological vulnerabilities may help explain why our finding of increased FA was driven by females; however, the reason(s) for the sex differences we found cannot be determined based on our cross-sectional pilot study. Future studies should examine dynamic, sex-specific associations between surgical history, inflammation, and cognitive function.

There are several important strengths to our study. First the appraisal of vascular risk factors and global cognitive function in our statistical models enabled more detailed analysis. In addition, our study employed measures of cognition (i.e., selected measures from the SENAS) with strong psychometric properties. The construction of the SENAS battery was based on item response theory, and psychometrically matched composite measures were created across the different scales. The use of IRT composite scores in our study thus allows for more robust, head-to-head comparisons across cognitive domains. Perhaps more importantly, the cognitive measures used as outcome variables in this study were separate from the measures reviewed in the interdisciplinary consensus conference for each participant. This approach limits the likelihood of circularity in our findings (i.e., cognitive measures used in diagnosis adjudication as a healthy control were different than our outcome variables).

There are also several weaknesses that warrant further consideration. First, this is a pilot study with a relatively small sample size; although the associations between total lifetime non-cardiac procedures and episodic memory are remarkable (particularly after accounting for global cognitive function and vascular risk factors), it is possible that other associations with non-memory domains were not adequately captured due to limited statistical power to detect small or medium effects. Although we controlled for the type of procedures (e.g., non-cardiac, controlling for presence or absence of a prior cardiac procedure), we did not control for the specific types of non-cardiac surgeries (e.g., orthopedic; ophthalmologic) or whether general anesthesia, conscious sedation, or regional anesthesia was used. Although the latter point should be further addressed in future studies, it is important to note that prior work in the context of POCD has not demonstrated an association between type of anesthesia and cognitive outcomes (Evered, et al., 2011; Hlatky et al., 1997; Silbert, Evered, & Scott, 2014; Williams-Russo, Sharrock, Mattis, Szatrowski, & Charlson, 1995). Nonetheless, it cannot be assumed that mechanisms which appear irrelevant to POCD are also unimportant with respect to cumulative surgical exposure, since the longitudinal pathophysiology may differ substantially. Studies of waste clearance in the brain via the glymphatic pathway have shown that general anesthesia impairs clearance (Gakuba et al., 2018), although there appear to be variations based on anesthetic type and dose (Benveniste et al., 2017). As such, future studies should consider not only the broad category of general anesthesia, but the specific agents and the doses at which they are delivered. Finally, the ICBM-81-DTI white matter atlas used has several problems to consider. The participants in the atlas were younger than our cohort, the stria terminalis fibers of the hippocampus are labelled as part of the fornix (due to resolution limitations at the time of development), and the forniceal sections are not continuous (Brown et al., 2017; Mori, et al., 2008).

In summary, in a cohort of community-dwelling, healthy aging adults, we found an association between higher number of lifetime non-cardiac surgical procedures and worse performance on a robust measure of episodic memory. This finding was specific to episodic memory and remained present when we controlled for vascular risk factors. In an exploratory analysis, we found lower white matter microstructural integrity in the body of the fornix, a structure critical for episodic memory. Results suggest that exposure to surgery may trigger cumulative, lasting effects that could be mediated through immunologic changes associated with surgery as a stressor. A broadened framework of allostatic load may help situate surgery alongside other life stressors, and inform thinking about a multifaceted approach to understanding life events in the pathogenesis of Alzheimer’s disease. Future work will attempt to further characterize the immunologic status of this cohort to better elucidate the mechanisms through which cognitive changes may occur.