Cognitive and Cerebrospinal Fluid Alzheimer's Disease-related Biomarker Trajectories in Older Surgical Patients and Matched Nonsurgical Controls

Abstract

Background: Anesthesia and/or surgery accelerate Alzheimer's disease pathology and cause memory deficits in animal models, yet there is a lack of prospective data comparing cerebrospinal fluid (CSF) Alzheimer's disease-related biomarker and cognitive trajectories in older adults who underwent surgery versus those who have not. Thus, the objective here was to better understand whether anesthesia and/or surgery contribute to cognitive decline or an acceleration of Alzheimer's disease-related pathology in older adults.

Methods: The authors enrolled 140 patients 60 yr or older undergoing major nonneurologic surgery and 51 nonsurgical controls via strata-based matching on age, sex, and years of education. CSF amyloid β (Aβ) 42, tau, and p-tau-181p levels and cognitive function were measured before and after surgery, and at the same time intervals in controls.

Results: The groups were well matched on 25 of 31 baseline characteristics. There was no effect of group or interaction of group by time for baseline to 24-hr or 6-week postoperative changes in CSF Aβ, tau, or p-tau levels, or tau/Aβ or p-tau/Aβ ratios (Bonferroni P > 0.05 for all) and no difference between groups in these CSF markers at 1 yr (P > 0.05 for all). Nonsurgical controls did not differ from surgical patients in baseline cognition (mean difference, 0.19 [95% CI, -0.06 to 0.43]; P = 0.132), yet had greater cognitive decline than the surgical patients 1 yr later (β, -0.31 [95% CI, -0.45 to -0.17]; P < 0.001) even when controlling for baseline differences between groups. However, there was no difference between nonsurgical and surgical groups in 1-yr postoperative cognitive change in models that used imputation or inverse probability weighting for cognitive data to account for loss to follow up.

Conclusions: During a 1-yr time period, as compared to matched nonsurgical controls, the study found no evidence that older patients who underwent anesthesia and noncardiac, nonneurologic surgery had accelerated CSF Alzheimer's disease-related biomarker (tau, p-tau, and Aβ) changes or greater cognitive decline.

Figure 1.

Participant consort diagram. Surgical patients (left) and nonsurgical controls (right).

Figure 2.

CSF levels of (a) Aβ, (b) tau, (c) p-tau-181p, (d) tau/Aβ, and (e) p-tau-181p/Aβ in surgical patients and nonsurgical controls. The first column represents baseline, 24-hour, and 6-week data from the AlzBio3 assay platform. Error bars represent the 25^th and 75^th percentiles of the data. The second column represents 1-year CSF AD biomarker levels in surgical patients (red) and nonsurgical controls (blue) from the Fujirebio Lumipulse assay platform; 1-year data was log-transformed to reduce skew. Each dot represents data from an individual patient at a single time point; the width of the colored area indicates the data distribution. Within the boxplots, the middle line shows the median of the data, and the upper and lower edges show the interquartile range. There were no significant group differences (see the main text for analysis details). Missingness: 98 surgical patients had CSF data at baseline, 90 at 24-hours, 94 at 6 weeks, and 48 at 1 year. 1 additional surgical patient was missing tau data at 1 year due to assay artifact. 46 nonsurgical controls had CSF data at baseline, 42 at 24-hours, 36 at 6 weeks, and 32 at 1 year. 1 additional control was missing tau data at 24-hours; 2 other controls were missing Aβ and tau data at 1 year, respectively, due to assay artifact.

Figure 3.

Cognitive function by domains and overall CCI (the average of the 4 domain scores) over time, in surgical patients (red) and nonsurgical controls (turquoise). Each dot represents data from an individual patient at a single time point; the width of the colored area indicates the data distribution. Within the boxplots, the middle line shows the median of the data, and the upper and lower edges show the interquartile range (see Table 2 for statistical comparisons). P-values are Bonferroni corrected for the 5 cognitive models.