Association between Use of Angiotensin-converting Enzyme Inhibitors or Angiotensin Receptor Blockers and Postoperative Delirium

In this single-site retrospective cohort analysis, we considered patients who had at least one Confusion Assessment Method—ICU assessment while in the surgical ICU of the Cleveland Clinic main campus after noncardiac procedures between January 2013 and June 2018. Patients with preexisting dementia, Alzheimer’s disease, or other cognitive decline, and patients undergoing neurosurgical procedures were excluded from the study. With Institutional Review Board approval from the Cleveland Clinic (Cleveland, Ohio) and waived consent, data were extracted from our electronic health record system and the Perioperative Health Documentation System registry of the Anesthesia Institute (Cleveland, Ohio). The analysis plan was developed a priori and approved by the Institutional Review Board.

Patients who were treated with angiotensin-converting enzyme inhibitors and/or angiotensin receptor blockers preoperatively were compared to patients who used neither. Angiotensin-converting enzyme inhibitors included benazepril, captopril, fosinopril, quinapril, ramipril, enalapril, and lisinopril. Angiotensin receptor blockers included irbesartan, losartan, telmisartan, olmesartan, and valsartan. We considered patients to be angiotensin-converting inhibitor and angiotensin receptor blocker users when the medications were taken within 1 month before the date of surgery. Normally, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers are held the morning of surgery. Postoperative angiotensin-converting enzyme inhibitors and angiotensin receptor blockers use was determined as angiotensin-converting enzyme inhibitors and angiotensin receptor blockers use before the first positive Confusion Assessment Method—ICU test for those with postoperative delirium and as angiotensin-converting enzyme inhibitors and angiotensin receptor blockers use before the last surgical ICU discharge and within 7 days postsurgery for the group without postoperative delirium. Postoperative benzodiazepines use was similarly defined as any of midazolam, lorazepam, or diazepam.

Delirium was assessed using the Confusion Assessment Method. Normally, critical care patients at the Cleveland Clinic are evaluated at 12-h intervals by trained nurses. We considered patients to have experienced postoperative delirium when their Confusion Assessment Method—ICU tests were positive at any assessment during their surgical ICU stay within 12 h to 7 days after surgery. The Confusion Assessment Method—ICU evaluates four features: (1) acute onset and fluctuating course; (2) inattention; (3) disorganized thinking; and (4) altered consciousness. A positive Confusion Assessment Method test was defined by the presence of features 1 and 2, and either 3 or 4.^17–20 

For the primary analysis, the association between preoperative use of angiotensin-converting enzyme inhibitors and angiotensin receptor blockers and the incidence of postoperative delirium was assessed through a conditional logistic regression model, with baseline variables in table 1 adjusted for through propensity score matching. Specifically, the probability (i.e., the propensity score) of being treated with angiotensin-converting enzyme inhibitors and angiotensin receptor blockers preoperatively (vs. neither) was estimated in a logistic regression model using demographic and baseline variables, emergency surgery, surgery type, and surgery length (as a proxy of surgery complexity) as predictors. American Society of Anesthesiologists status was treated as a continuous variable in the model. Patients treated with angiotensin-converting enzyme inhibitor or angiotensin receptor blocker were then matched with controls at a 1:2 treatment–control ratio with a matching calliper of within 0.2 SDs of the logit of the propensity score. Treatment and control patients were required to have the same surgery type, and surgery within a year of each other. A conditional logistic regression model was then used for comparing the matched pairs on the outcome, adjusting for imbalanced confounding variables (absolute standardized difference greater than 0.10) after matching.

The association between postoperative use of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and delirium was similarly assessed. All of the variables in table 1 (details in tables A1 and A2), as well as preoperative use of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and intraoperative use of midazolam, were considered as confounding factors and used in the propensity score model. To reduce potential bias arising from the fact that early delirium patients are less likely to have had the postoperative exposure, each exposed patient was matched to controls who had not yet had delirium at the time when the exposed patient started or resumed angiotensin-converting enzyme inhibitor or used angiotensin receptor blocker.

Two sensitivity analyses were performed for the association between preoperative use of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and delirium. In a preplanned sensitivity analysis, we compared the two groups on in-hospital mortality (instead of delirium) using a conditional logistic regression model in the matched patients. In the other sensitivity analysis, we assessed the association between preoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use and the detection of delirium using a logistic regression model, adjusting for all of the potentially confounding factors directly in the model instead of using propensity score matching.

Three sensitivity analysis were performed for the association between postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use and delirium. First, because postoperative use of benzodiazepines (midazolam, lorazepam, or diazepam) is strongly associated with delirium, we included postoperative use of benzodiazepines in the propensity score model when assessing the association between postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use and outcome. Because angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and benzodiazepines could be used concurrently postoperatively, the interaction between angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and benzodiazepines was also tested, at a significance level of 0.10. Second, to correct for potential bias arising from the timing of postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, we constructed a time-to-delirium survival model using all patients and considering the use of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker as a time-varying covariate, such that patients were designated as “exposed” for the remaining follow-up period once they started/resumed angiotensin-converting enzyme inhibitor or angiotensin receptor blocker. Patients without delirium were censored at 7 days postoperatively, and baseline variables were adjusted for directly in the model. Finally, to reduce bias induced by early delirium detection, in a third sensitivity analysis we compared the angiotensin-converting enzyme inhibitor or angiotensin receptor blocker and control groups on the incidence of delirium after 24 h, adjusting for baseline variables through propensity score matching.

In secondary analyses, we assessed the association between detection of postoperative delirium and preoperative use of angiotensin-converting enzyme inhibitor only, angiotensin receptor blocker only, or neither, in a logistic regression model while adjusting for the same confounding factors as in the main analysis. An analogous model was constructed for postoperative use of angiotensin-converting enzyme inhibitor only, angiotensin receptor blocker only, or neither.

Two-sided tests were used for all analyses. Based on a post hoc power calculation, we had more than 90% power at the 0.05 significance level to detect an odds ratio of 0.80 or smaller for preoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use versus delirium detection, and an odds ratio of 0.70 or smaller for postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, with 4,989 patients who met the enrolment criteria and a delirium incidence of 38% in the control group. Power calculations were conducted post hoc using SAS software version 9.4 (SAS Institute, USA), and R and SAS software version 9.4 (SAS Institute, USA) was used for all statistical analysis.

We identified 4,864 patients who met our inclusion and exclusion criteria, including 1,396 who were given angiotensin-converting enzyme inhibitors or angiotensin receptor blockers in the preoperative period and 440 who were given angiotensin-converting enzyme inhibitors or angiotensin receptor blockers postoperatively before delirium or intensive care unit discharge. Confusion Assessment Method—ICU score was assessed a mean ± SD of 3 ± 1 times per day for the first 7 postoperative days in both preoperative angiotensin-converting enzyme inhibitors or angiotensin receptor blockers and control groups, and in both postoperative angiotensin-converting enzyme inhibitors or angiotensin receptor blockers and control groups (fig. 1), and timing of the assessments was similar in each group. Complete baseline data were available for 5,421 patients. Patient characteristics are detailed in table A1; opioid, propofol, and other perioperative medications are summarized in table 2 and detailed in tables A3 and A4. In addition, Blood pressure and heart rate were summarized by 4-h intervals postoperatively, and based on a visual check, hemodynamic data were similar between patients who used angiotensin-converting enzyme inhibitors or angiotensin receptor blockers postoperatively versus not (fig. 2).

Delirium was detected in 39% (551 of 1,396) of patients who were treated with angiotensin-converting enzyme inhibitors or angiotensin receptor blockers in the preoperative period and in 39% (1,344 of 3,468) of patients who were not (unadjusted P value = 0.643). We matched 1,151 patients treated with angiotensin-converting enzyme inhibitor or angiotensin receptor blocker preoperatively to 1,627 patients who were not. All of the baseline variables were adequately balanced between the treatment and the control groups after matching (absolute standardized difference less than 0.10) except for body mass index and diabetes, which were adjusted for directly in the model.

In the propensity-matched patients, preoperative treatment with angiotensin-converting enzyme inhibitor or angiotensin receptor blocker was not associated with postoperative delirium (P = 0.700), with an estimated odds ratio (95% CI) of 0.98 (0.86 to 1.10) for angiotensin-converting enzyme inhibitor or angiotensin receptor blocker compared to control (fig. 3). Consistent results were obtained from the sensitivity analysis where confounders were included directly in the model with outcome, with an estimated odds ratio of 0.91 (95% CI, 0.79 to 1.05; P = 0.202). Mortality was 4% (43 of 1,151) in the matched angiotensin-converting enzyme inhibitor or angiotensin receptor blocker group, and 5% (84 of 1,627) in the matched controls. In-hospital mortality did not differ in patients who did and did not use angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, with an estimated odds ratio of 0.76 (95% CI, 0.51 to 1.12; P = 0.169).

Delirium was detected in 23% (100 of 440) of patients who received angiotensin-converting enzyme inhibitors or angiotensin receptor blockers postoperatively before the first incidence of delirium or final ICU discharge (as described in Materials and Methods), compared to 41% (1,795 of 4,424) of patients who did not receive angiotensin-converting enzyme inhibitor or angiotensin receptor blocker before delirium or ICU discharge. After matching 321 patients with postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use to 577 control patients, we observed a confounder-adjusted odds ratio for delirium of 0.55 (95% CI, 0.43 to 0.72; P < 0.001) comparing postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker users to control patients. Baseline and intraoperative confounders were balanced after matching, except for preoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, which was adjusted for directly in the model. No interaction was detected between preoperative and postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use (P = 0.55) and delirium.

In the sensitivity analysis in which postoperative benzodiazepines was included in the propensity score model, we also observed a significant association between postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use and delirium, with an estimated odds ratio of 0.54 (95% CI, 0.40 to 0.71; P < 0.001). The groups were well balanced with respect to postoperative benzodiazepines use, making confounding due to this factor very unlikely even though it is associated with delirium. The interaction between postoperative use of benzodiazepines and postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker was not significant (P = 0.359).

Among patients with delirium, the time to onset of delirium (median [first quartile, third quartile]) was 57 [30, 110] h in those treated with angiotensin-converting enzyme inhibitor or angiotensin receptor blocker postoperatively, and 19 [14, 46] h in those who were not. The hazard ratio of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker versus control was an estimated 0.73 (95% CI, 0.60 to 0.89; P = 0.002) when we took into account the time at which angiotensin-converting enzyme inhibitor or angiotensin receptor blocker were started/resumed postoperatively in a time-varying covariate analysis.

Considering all patients, the incidence of delirium after 24 h was 14% (93 of 653) in patients using angiotensin-converting enzyme inhibitor or angiotensin receptor blocker postoperatively and 33% (1,408 of 4,211) in control patients. After matching to control for confounding, the odds ratio comparing 423 postoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker users matched with 745 controls was an estimated 0.46 (95% CI, 0.35 to 0.59; P < 0.001), a result consistent with the main analysis.

In the secondary analysis, we compared patients who preoperatively used angiotensin-converting enzyme inhibitor only, angiotensin receptor blocker only, and neither. We did not observe significant differences in the detection of delirium between the three groups after adjusting for the same variables as in the primary analysis above (overall P = 0.774). When comparing postoperative use of angiotensin-converting enzyme inhibitor only, angiotensin receptor blocker only, and neither, we found that patients who used angiotensin-converting enzyme inhibitor only or angiotensin receptor blocker only had a significantly lower odds of having delirium compared to patients who used neither (P < 0.001). No significant difference was observed between angiotensin-converting enzyme inhibitor only and angiotensin receptor blocker only patients (fig. 2).

Preoperative use of angiotensin-converting enzyme inhibitors and angiotensin receptor blockers was not independently associated with detection of postoperative delirium in surgical ICU patients. The most likely reason is simply that most formulations are relatively short-acting. Since our routine is to hold angiotensin-converting enzyme inhibitors and angiotensin receptor blockers on the morning of surgery, it is unlikely that these drugs were biologically active even by the evening of surgery, much less over days of critical care.

In contrast to the lack of preoperative association between angiotensin-converting enzyme inhibitors and angiotensin receptor blockers and delirium, patients who used angiotensin-converting enzyme inhibitors and angiotensin receptor blockers postoperatively had about half the odds of developing delirium, after adjusting for baseline and intraoperative confounding factors. Furthermore, patients in the postoperative control group not only had more delirium, but had a much earlier delirium onset at 19 [14, 48] h versus to 64 [35, 112] h. Our results therefore suggest that angiotensin-converting enzyme inhibitors and angiotensin receptor blockers must be present in biologically relevant concentrations to prevent delirium.

The observed suggestion of a protective effect of angiotensin-converting enzyme inhibitors and angiotensin receptor blockers against delirium may be explained by new discoveries in the classic renin angiotensin system. In particular, angiotensin II has neurotoxic effects mediated by its actions on angiotensin type 1 receptor. In contrast, there are neuroprotective effects of alternate renin angiotensin system that are mediated by angiotensin (1-7), angiotensin III, and angiotensin IV.²¹  Angiotensin-converting enzyme inhibitors increase brain substance P, which is normally degraded by angiotensin-converting enzyme, which in turn enhances the activity of neprilysin,²²  a recognized amyloid β degrading enzyme.²³  Additionally, angiotensin-converting enzyme inhibitors enhance production of angiotensin (1-7), which has neuroprotective, antiinflammatory, and cerebral vasodilation effects.²¹  The injurious effects of angiotensin II on the brain are mainly mediated by its action on angiotensin receptor 1A subtype and include hypertension, inflammation, increased oxidative stress, blood–brain barrier disruption, and neurotoxicity.

Angiotensin II also induces nitric oxide production, neurite outgrowth, and brain development through angiotensin receptor 2 activation.^21,24  Use of angiotensin receptor 1 blockers enhances the stimulation of neuroprotective angiotensin receptor 2 by angiotensin II. In addition to their blocking actions on angiotensin receptor 1, angiotensin receptor blockers also stimulate the peroxisome proliferator–activated receptor-γ, which is a nuclear receptor. Peroxisome proliferator–activated receptor-γ induces the polarization of the microglia into type M2. Type M2 microglia has antiinflammatory and neuroprotective effects.^25–27 

The beneficial effect of stimulating peroxisome proliferator–activated receptor-γ by candesartan was shown in a subgroup analysis of the Study on Cognition and Prognosis in the Elderly. Use of candesartan was associated with less decline in the Mini Mental State Examination, especially in patients with slightly low cognitive function at baseline (Mini Mental State Examination score 24 to 28).^28,29  Losartan treatment compared with atenolol in elderly hypertensive patients significantly improves cognitive function, in particular immediate and delayed memory cognition domains.³⁰  Additionally, use of losartan in hypertensive patients improves cognitive function including memory, attention/concentration, comprehension, anxiety/depression, and interpersonal relationships.³¹  Available evidence thus suggests that inhibition of classic renin angiotensin system pathway and simultaneous stimulation of alternate renin angiotensin system pathways by angiotensin-converting enzyme inhibitors and angiotensin receptor blockers has neuroprotective and antiinflammatory effects that might explain reduced postoperative delirium.

The standard delirium diagnostic criteria for delirium are specified in the fifth edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders.¹⁷  However, the most widely used instrument is the Confusion Assessment Method.¹⁹  We used the intensive care version, Confusion Assessment Method—ICU, which was developed to accurately diagnose delirium in intensive care unit patients, who are often nonverbal due to mechanical ventilation.¹⁸  The Confusion Assessment Method—ICU test has a sensitivity of 95% and specificity of 89%.²⁰  It therefore seems unlikely that our negative preoperative results are consequent to an insensitive measurement tool. Furthermore, the median number of Confusion Assessment Method—ICU assessments was about 10, which should have been quite sufficient to detect clinically important delirium.

We performed a comprehensive confounder-adjusted analysis to assess independent associations between the intraoperative use of angiotensin-converting enzyme inhibitors and angiotensin receptor blockers and postoperative delirium, with year of surgery and type of surgery well-adjusted through propensity score matching. However, the association between postoperative angiotensin-converting enzyme inhibitors or angiotensin receptor blockers use (exposure) and delirium was subject to immortal time bias³²  because the definition of exposure required that patients had not already developed the outcome (delirium) until after resumption of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker. In other words, it would be difficult for patients with early delirium detection to have an exposure in our study design, and patients who never restarted (the controls) had a longer period to develop delirium. To reduce such survival bias we performed the following steps: (1) matched exposed patients to controls who had not yet developed delirium in the main postoperative exposure analysis; (2) conducted a sensitivity analysis using time-varying covariate survival analysis to account for the timing of the exposures; and (3) added another analysis limiting delirium outcomes to after 24 h. While these sensitivity analyses provided results consistent with the main analysis, none completely eliminates the risk of immortal time bias.

The postoperative exposure analyses surely suffer from some unmeasured confounding bias. While the confounder adjustment was comprehensive for preoperative variables, we did not thoroughly adjust for postoperative confounding variables co-occurring with the exposure and outcome. For example, the appendix table 4 showed that some postoperative medications (e.g., propofol) remained imbalanced after matching. Additionally, variations in the duration of action of the drugs added heterogeneity to the inference. For example, captopril has no active metabolite and a half-life of only 1.7 h, whereas ramipril has an active metabolite and a half-life up to 50 h. Among angiotensin receptor blockers, there was similar variation: losartan has the shortest half-life of only 2 h, whereas telmisartan has a half-life of 24 h. Angiotensin receptor blockers also have varying degrees of lipid solubility and therefore a different ability to cross the blood–brain barrier.^27,33 

In summary, preoperative use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers was not associated with lower incidence of postoperative delirium in critical care patients—presumably because most drugs were no longer active during the relevant period. In contrast, postoperative use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers was associated with about half the incidence of delirium. Given the limitations inherent to our study design, especially overlapping exposure and outcome, our results should be considered exploratory rather than a basis for practice change. However, our results provide strong support for randomized trials to assess whether resumption of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers can prevent delirium in critical care patients.

Support was provided solely from institutional and/or departmental sources.

The authors declare no competing interests.