Reply to A Letter Concerning A Role for Blood-brain Barrier Dysfunction in Delirium following Non-Cardiac Surgery in Older Adults

We thank Drs. Le Guennec, Balcerac, and Weiss for their interest in our study of postoperative blood-brain barrier dysfunction and postoperative delirium.¹ The authors raise several questions, which we clarify here.

We agree that cerebrospinal fluid-to-plasma albumin ratio (CPAR) is an imperfect marker for blood-brain barrier (BBB) dysfunction, though it is the most widely accepted molecular biomarker for overall blood-brain barrier integrity/function.^2,3 Indeed, few if any human biomolecular markers (including CPAR) distinguish between blood-cerebrospinal fluid barrier (BCSFB) and BBB permeability. The BBB and BCSFB are two anatomically distinct central nervous system barriers: the BBB is located at the level of endothelial cells within brain microvessels, while the ependymal cells of the choroid plexus form the BCSFB. To distinguish between BCSFB and BBB permeability, direct barrier examination using neuroimaging⁴ or post-mortem brain tissue immunostaining is required.⁵ Although these immunostaining techniques are not possible to use in living patients, rodent surgical studies have found postoperative decreases in brain aquaporin-4 and vascular markers (ie, CD31) and increases in dextran diffusion into the brain, both of which suggest increased postoperative BBB permeability.^6,7 In a small study of older cardiac surgery patients, the efflux rate of gadolinium into the brain measured with “k-trans” MRI was increased 24-hours postoperatively, suggesting that this postoperative BBB dysfunction occurs in humans.⁸ Thus, although we did not study the BBB directly (such as via immunostaining or neuroimaging), the studies mentioned above suggest that the postoperative CPAR increases we observed in older surgical patients were due (at least in part) to postoperative BBB dysfunction. Nonetheless, we agree that formally differentiating between BBB vs BCSFB permeability changes following surgery is an important question for future studies on delirium and perioperative neurocognitive disorders in older adults.

We also agree that changes in BBB permeability to molecules smaller than albumin could occur after surgery, although we did not assess BBB permeability to smaller molecules such as urea. However, prior studies have shown that the CSF-to-plasma urea ratio is not as strongly associated with inflammation (ie, CSF cytokines) than the CPAR,⁹ suggesting that BBB permeability to larger molecules (rather than to smaller molecules such as urea) is more likely to be involved in inflammation-related mechanisms in delirium. Nonetheless, we agree that understanding postoperative changes in BBB (and BCSFB) permeability to molecules of various sizes and their clinical significance is an important goal for future studies.

Drs. Le Guennec et al also raise the potential concern that serial lumbar punctures could cause a local intrathecal inflammatory response because a prior study demonstrated increased inflammatory CSF biomarkers (YKL-40) 24-hours after lumbar punctures that were performed with 22 g needles. However, our study used 25 g needles for serial lumbar punctures, which may cause less inflammation and trauma given their smaller size. To evaluate this, we analyzed CPAR data from 18 similarly-aged (median age 67; interquartile range 64, 75) non-surgical controls that underwent two lumbar punctures at the same time intervals as in our surgical patients (i.e. 24 hrs apart), using the same lumbar puncture protocol we used in our surgical cohort (ie with 25 g pencil point spinal needles). In these non-surgical controls, we found no significant 24-hour change in CPAR (median change −0.14, interquartile range −1.22, 0.96, p = 0.80), in contrast to the pre- to 24-hour postoperative CPAR increase we observed in surgical patients (median, 0.28; interquartile range −0.48, 1.24; p = 0.001).¹ Thus, it is unlikely that serial lumbar punctures with 25 g needles significantly influenced postoperative CPAR levels in our study.

Our study did not include any patients undergoing neurosurgical procedures with intra-dural manipulation, which eliminates the possibility that direct surgical manipulation of the spinal cord (or brain) could have caused the observed alterations in CPAR. The orthopedic surgeries included in our study did not include any spine surgeries, and mostly consisted of total hip, knee, ankle and shoulder arthroplasties.

Drs. Le Guennec et al were concerned that our 2-hit model for delirium “lacks data on peri-operative drugs”. We found no major differences in perioperative medications between those patients who developed (versus those who did not develop) postoperative delirium (see Table 1). Thus, the 2-hit theory is still supported by our data, which showed that, in addition to the increased delirium risk due to impaired baseline cognition, there was also a dose-dependent association of increased pre- to 24-hour postoperative CPAR change with increased delirium rates. We also note that the 2-hit theory is a conceptual model, not a complete description of the molecular and/or systems neuroscience mechanisms of delirium. Other, additional risk factors and mechanisms likely play key roles in postoperative delirium; thus, an important goal for future studies is to better predict delirium incidence by incorporating these additional factors and measures in addition to baseline cognitive measures and perioperative CPAR values.

Table 1.

Perioperative Medications in the Older Non-Cardiac Surgical Patient Cohort¹

	Overall (N=207)	No Delirium (N=181)	Yes Delirium (N=26)	Standardized Difference
Inhaled Anesthetics Used	123 (59.4%)	105 (58.0%)	18 (69.2%)	0.235
Propofol (mg)	220 [150, 862]	240 [150, 873]	200 [130, 656]	0.105
Midazolam Received	80 (38.6%)	73 (40.3%)	7 (26.9%)	0.189
Intraoperative MME	17.62 (12.87)	17.08 (12.29)	21.40 (16.14)	0.301
Phenylephrine (mcg/kg/min)	0.06 [0.00, 0.22]	0.05 [0.00, 0.22]	0.10 [0.02, 0.28]	0.039
Any Paralytics Used	133 (64.3%)	113 (62.4%)	20 (76.9%)	0.319
Any Other Vasopressor	28 (13.5%)	24 (13.3%)	4 (15.4%)	0.061
Any Ephedrine	77 (37.2%)	67 (37.0%)	10 (38.5%)	0.030
Any Dexmedetomidine	37 (17.9%)	33 (18.2%)	4 (15.4%)	0.076

MME, morphine milligram equivalent. Other vasopressors included norepinephrine, epinephrine, and vasopressin

In conclusion, our data point to a significant role for BBB dysfunction (as measured by CPAR increases) in postoperative delirium. Indeed, future studies would benefit from a multimodal assessment of BBB (and BCSFB) function that includes neuroimaging and a broader set of biomarkers. Given that blood biomarkers poorly reflect BBB function, the addition of other CSF biomarkers such as soluble platelet-derived growth factor receptor β,¹⁰ a marker of pericyte dysfunction, could help determine the etiology of postoperative BBB dysfunction. Understanding the cellular and molecular changes that occurs at the BBB could be accomplished using BBB-on-chip technologies,^7,11,12 that use microfluidics and primary human cells to model the human BBB, although the relevance of the BBB-on-chip to in vivo postoperative BBB changes in humans is currently unknown.

Neuroimaging would enable better anatomic localization of BBB dysfunction, but as Drs. Le Guennec et al noted, 24-hour postoperative magnetic resonance imaging would be challenging to conduct amidst other postoperative care activities in older surgical patients (especially in those with delirium). Further, the “k-trans” MRI scans require gadolinium contrast, which the FDA has recently warned can be retained in patient’s bodies and thus are now requiring patient-education of these risks.¹³ CT-perfusion imaging may be more feasible because it is faster than MRI, but would have risks of iodinated contrast toxicity that need to be considered in study design and discussed during informed consent.

In closing we note that our study was the first large-scale study to examine the relationship between the most widely accepted molecular biomarker of BBB dysfunction and delirium, and demonstrated a strong independent association between CPAR increases and delirium. Based on these findings, future studies on this topic are warranted, including those using neuroimaging and other molecular biomarkers to yield insights into localization and etiology of dysfunction in barriers between the central nervous system and periphery in postoperative delirium.