Impact of Methylprednisolone on Postoperative Quality of Recovery and Delirium in the Steroids in Cardiac Surgery Trial: A Randomized, Double-blind, Placebo-controlled Substudy

CARDIAC surgery with cardiopulmonary bypass is common worldwide¹  and is associated with an ubiquitous inflammatory response.²  This sterile inflammatory response may be associated with adverse outcomes after surgery. Suppressing or modulating the inflammatory response may improve outcomes. High-dose steroids suppress inflammation after cardiopulmonary bypass,³  but two large randomized controlled trials in cardiac surgery failed to show improvement in mortality or major morbidity,^4,5  postoperative cognitive dysfunction (POCD),⁶  delirium,⁷  or persistent pain.⁸  The largest and most definitive relevant study was the Steroid In caRdiac Surgery (SIRS) trial⁵ ; the study we present here is a substudy of SIRS.

Animal studies have shown associations of cognitive decline with surgical inflammation⁹  and anesthesia drugs,¹⁰  and a clinical study has shown differences in early but not late POCD with desflurane versus propofol anesthesia.¹¹  However, there are few data examining the impact of intraoperative high-dose steroids on the postoperative quality of recovery in the days to months after surgery. Further, there are few data identifying whether reduction of inflammation may be of additional benefit in patients with poor presurgery cognition (who could represent a more fragile population) on the incidence of delirium or poor quality of recovery.

Delirium is relatively common after cardiac surgery and is associated with delayed recovery and worse long-term outcomes.^12,13  Both major trials of steroids for cardiac surgery failed to show a difference in the incidence of delirium with steroids (range, 8.0 to 11.7%).^4,5  However, the impact of delirium on multiple domains of recovery over time remains poorly characterized.

Our goal was to determine whether high-dose methylprednisolone improves the quality of postoperative recovery and reduces the incidence of delirium after cardiac surgery involving cardiopulmonary bypass. Specifically, we tested the primary hypothesis that high-dose methylprednisolone improves the quality of recovery compared to placebo. Secondarily, we tested the hypothesis that high-dose methylprednisolone reduces the incidence of delirium during the initial three postoperative days.

Three centers participated in this substudy of the SIRS trial (ClinicalTrials.gov identifier: NCT00427388): the Royal Melbourne Hospital (Melbourne, Victoria, Australia), the Cleveland Clinic (Cleveland, Ohio), and the Hamilton Health Sciences Center at McMaster University (Hamilton, Ontario, Canada). Written and informed consent was obtained from participants for the substudy. We evaluated the quality of recovery and delirium without altering the underlying SIRS study. Detailed description of the SIRS design including participants, intervention, randomization, blinding, and allocation has been previously published.⁵  Briefly, the SIRS study was a double-blinded, placebo-controlled, multicenter, randomized controlled trial. The intervention was 250 mg methylprednisolone given at anesthetic induction and again just before starting cardiopulmonary bypass. Amendments to the SIRS ethics approvals and local governance approvals were obtained at each center participating in this substudy.

Participants were included if they were enrolled in the SIRS study and could speak sufficient English to complete the Postoperative Quality of Recovery Scale (PostopQRS)¹⁴  and did not have known cognitive impairment or psychiatric illness. Inclusion criteria for the SIRS study included age more than 18 yr, European System for Cardiac Risk Evaluation (EuroSCORE) more than or equal to 6, and being able to provide written consent.

Our primary outcome was the quality of recovery over a 6-month period, measured using PostopQRS. The secondary outcomes included the cumulative incidence of delirium measured using the Confusion Assessment Method Intensive Care Unit (CAM ICU) scale,¹⁵  quality of recovery in individual recovery domains, and subdomains of the CAM ICU.

The PostopQRS¹⁴  was used to measure the quality of recovery after surgery. Details of the construct and validation of the PostopQRS have been previously published.^14,16–18  In brief, the quality of recovery is measured using a verbal survey tool that assesses recovery in multiple domains (physiologic, emotive, nociceptive, functional [activity of daily living {ADL}], and cognitive).

Baseline measurements are acquired before surgery. Recovery is a dichotomized outcome defined by a return to at least baseline values or better at each of the postoperative measurement time points. Overall recovery requires recovery in all domains being assessed, and failure in any domain results in failure of overall recovery. The tool is designed for repeated measurements and can be administered either face to face or via the telephone.¹⁶ 

A description of the items within each domain is shown in table 1. The physiologic domain is assessed whilst the participant is in hospital (days 1 to 3 in this study) and consists of seven items applicable to the postemergence period. The nociceptive domain consists of pain and nausea subscales, the emotive domain consists of depression and anxiety subscales, and the ADL domain consists of four activities of daily living (ability to stand, dress, eat, and walk). The cognitive domain consists of five verbal tests, and domain recovery requires recovery in all five tests. Variance in cognitive performance is a normal event, and accordingly, the definition of recovery in cognitive tests was modified to include a tolerance factor to account for normal variability.¹⁶  Participants are allowed to perform a little worse than their baseline performance and still be scored as recovered, so that recovery = score − (baseline score − tolerance factor) for each question. Participants with baseline cognitive scores that are equal to or less than the tolerance factor are not included in the cognitive domain at subsequent time points as the scoring rules would automatically score them as recovered. Therefore, to be included in the cognitive domain, participants must score at least 1 more than the tolerance factor for each question of the domain (table 1). Further, they are excluded from scoring overall recovery as they cannot be evaluated in the cognitive domain. However, the caveat is that if they fail in any other domain, then they will be scored as failed in overall recovery. That is, failure in any domain results in failure in all domains. These participants are not excluded from other recovery domains. The incidence of low baseline scores differs among populations, ranging from near 0 in young volunteers,¹⁶  to 5 to 15% in orthopedic,¹⁸  and around 25 to 35% in cardiac surgery (C. Royse, M.B.B.S., M.D., unpublished data—audit data, 2013 to 2015).

The PostopQRS was conducted within 2 weeks before surgery, at 1, 2, and 3 days after surgery, and 1 and 6 months after surgery. The physiologic domain was only measured on days 1 to 3. Overall recovery is defined as recovery in all of the domains that are measured at each time point, so that overall recovery assessment does not include the physiologic domain after day 3.

Delirium was measured for all patients in the morning of days 1 to 3 after surgery using the CAM ICU¹⁵  and scored according to the published criteria. Patients in whom the CAM ICU had not been performed on all 3 days were included if CAM ICU had been completed on at least one of the 3 days. The Richmond Agitation and Sedation Scale¹⁹  evaluates the sedation level in every patient before the CAM ICU assessment. Patients with a Richmond Agitation and Sedation Scale score of −4 or −5 were excluded from further evaluations.

Three centers participated in this substudy of SIRS, and each included as many patients as practical. For the postoperative quality of recovery, we estimated sample size from the PostopQRS validation study.¹⁴  We used a Cochran–Mantel–Haenszel design for multiple measurements up to 3 months and a two-sided difference with 80% power at the 0.05 significance level. For an effect size of odds ratio (OR) of 2.0, 225 participants in each group were required.

Participants were excluded post hoc from quality of recovery analysis if they had incomplete baseline data. Participants with low baseline cognitive scores were excluded from cognitive analysis and from overall recovery (unless they failed to recover in a different domain). Participants were excluded from delirium analysis if they had insufficient data to be able to score the CAM ICU at all time points but were included if there was at least 1 time point with complete CAM ICU data. We considered patients to have experienced postoperative delirium when CAM ICU testing scored positive at any one assessment.²⁰ 

Differences in the quality of recovery between groups over time were assessed using the Cochrane-Mantel-Haenszel test. The incidence of delirium was assessed using Fisher exact test. P < 0.05 defined significance. A Bonferroni correction would be used if the difference between groups for either primary outcome was P < 0.05. Analysis was performed using SPPS version 22 (IBM, USA). For secondary outcomes, P < 0.01 was used to define significance in order to reduce the risk of type I error due to multiple analyses or for subgroup interactions.

Two subanalyses were conducted. First, we analyzed whether participants who suffered delirium had poorer quality of recovery, which was a planned analysis. The second, which was decided post hoc due to the high incidence of low cognitive baseline scores in this cohort of patients, was to examine whether participants with low baseline cognitive scores recovered more slowly or suffered more delirium than participants with normal cognitive baseline scores.

Five hundred and fifty-five participants were enrolled into the substudy. The participant flowchart is shown in figure 1 for both the PostopQRS and delirium analyses. Participants were excluded from analysis for the PostopQRS if they had incomplete baseline data (methylprednisolone, 28; control, 20) or no postsurgery data (methylprednisolone, 13; control, 12). Participants were excluded from analysis for delirium if they had no complete CAM ICU data on any day (methylprednisolone, 14; control, 18) or no CAM ICU data on any day (methylprednisolone, 13; control, 12). The preoperative and operative data for both groups are shown in table 2.

The overall quality of recovery and domain level recovery for all included participants are shown in figure 2. Recovery improved over time, but steroids did not alter the primary endpoint of overall recovery (P = 0.943) nor any of the individual recovery domains (all P > 0.05). The incidence of delirium for the control group was 10%, which was similar to that in patients given methylprednisolone (8%; OR, 0.74; 95% CI, 0.40 to 1.37; P = 0.357; table 3). There were also no differences in any of the delirium domains (all P > 0.05).

The quality of recovery for participants with low versus normal or high cognitive baseline scores is shown in figure 3. There were no significant differences in recovery domains (all P > 0.05). A comparison of groups for those with low baseline scores is shown in figure 4, showing no difference in the recovery domains. A comparison of groups for participants with normal cognitive baseline scores is shown in figure 5. There were no differences in overall recovery or any individual recovery domain (all P > 0.05).

The incidence of delirium for participants with low baseline cognitive scores was 11% versus 7% in patients with normal baseline scores (OR, 0.63; 95% CI, 0.33 to 1.21; P = 1.91). The incidence of delirium for each group and for the CAM ICU components is shown in table 3. There was no difference in the incidence of delirium or of CAM ICU components between groups. Participants with low baseline cognitive scores had a higher incidence of inattention (23% vs. 11%; P = 0.001) but not more disorganized thinking (10% vs. 4%; P = 0.011). In patients with normal baseline cognitive scores, there were no significant differences between the methylprednisolone and control groups in delirium (7% vs. 10%; OR, 0.51; 95% CI, 0.57 to 4.22; P = 0.455) or components. For patients with low baseline cognitive scores, the incidence of delirium was 7% in the methylprednisolone group versus 14% in the control group (OR, 0.456; 95% CI, 0.86 to 5.14; P = 0.13). There were no significant differences in the incidence of inattention (17% vs. 29%; P = 0.037) on the CAM ICU scale.

The comparison of the quality of recovery for participants with and without delirium is shown in figure 6. Participants suffering delirium in the first 3 days after surgery had similar recovery at 6 months although not significantly so. Poorer recovery occurred in the cognitive domain (P = 0.004) and physiologic domain (P < 0.001) but not in the other recovery domains.

The effect size expressed as OR with 95% CIs for the above comparisons is shown in table 4 for each measurement and at each time point.

Intraoperative high-dose methylprednisolone before cardiopulmonary bypass did not improve the postoperative quality of recovery nor did steroids reduce the incidence of delirium. These findings are consistent with the SIRS and Dexamethasone for Cardiac Surgery (DECS) studies that showed that high-dose steroids did not improve mortality, morbidity, delirium, POCD, or persistent pain.^4–8 

The effect of steroids on delirium on day 3 after surgery was evaluated in the DECS trial, with delirium defined by the need for treatment with postoperative neuroleptic drugs.⁴  Delirium was similar in patients given dexamethasone (9.2% vs. 11.7%; P = 0.06). In a substudy of 768 patients in whom delirium was measured over 4 days with the CAM ICU, the incidence of delirium and the duration did not differ between the groups (dexamethasone, 14.2% vs. 14.9%; adjusted OR, 0.85; 95% CI, 0.55 to 1.31; mean duration [interquartile range], 2 [1 to 3] vs. 2 [1 to 2] days, respectively; P = 0.45).⁷  In the SIRS study, the incidence of delirium on day 3 was 8% in both groups (P = 0.80). The incidence of delirium in our substudy is consistent with the DECS trial as was the lack of steroid efficacy.

Participants in the SIRS trial all had a EuroSCORE I of 6 or greater, and thus, all had considerable comorbidities and/or were scheduled for especially major surgery. The incidence of low baseline scores approaches 0 in young volunteers,¹⁶  5 to 15% in orthopedic surgery,¹⁸  and approximately 25 to 35% in the general cardiac surgery population (C. Royse, M.B.B.S., M.D., unpublished data—audit data, 2013 to 2015). In our current cohort, almost half of the patients had low baseline scores, indicating that baseline cognition was poor. It is possible that these patients were frail and generally unwell, and poor cognition is part of that medical state (cognitive frailty). Thus, they were more likely to suffer delirium or poor recovery in other domains. A theoretical consideration is that steroids might especially benefit such patients; however, there was no evidence that methylprednisolone was effective in this cohort. Our data do not support high-dose methylprednisolone in cognitively frail patients.

The concept of cognitive frailty and poor postoperative outcomes was explored by Silbert et al.^21,22  Using a seven-component neuropsychologic test battery, they compared a cohort of patients undergoing coronary artery surgery against nonoperative controls to identify the incidence of baseline cognitive impairment. The cardiac surgery group performed worse than controls on all tests other than the Grooved Pegboard test. The incidence of baseline cognitive impairment, defined by impairment in more than or equal to two of seven tests compared to nonoperative controls, was 35%, which is consistent with our finding of a high proportion of patients with low baseline cognitive scores. In a study of patients having hip arthroplasty, Silbert et al.²¹  identified baseline cognitive impairment in 32% of patients. In their cohort, there was an increased incidence of POCD and cognitive decline compared to those with normal cognition preoperatively. Oldham et al.²³  assessed cognitive and functional status preoperatively and showed association with cognitive impairment after coronary artery surgery. Our study did not specifically investigate cognitive decline or dementia, and we, therefore, cannot determine whether patients with low baseline cognitive scores had a greater risk of developing postoperative dementia.

Our second subanalysis was to identify the effect of delirium on the quality of recovery after surgery. Participants with delirium had worse overall quality of recovery, most evident at 6 months after surgery, although not statistically so. Within individual domains, cognitive recovery and physiologic recovery were worse. Our results are consistent with other studies showing worse long-term clinical outcomes in participants who develop delirium. Brown et al.²⁴  showed that delirium after cardiac surgical procedures was independently associated with increased intensive care length of stay and greater hospital charges. Mangusan et al.¹²  found that patients with delirium after cardiac surgery stayed longer in the hospital, suffered more falls, and were more likely to be discharged to a nursing facility or need home medical services. In a prospective, observational, follow-up study 1 to 1.5 yr after surgery, Koster et al.¹³  showed that patients with delirium after cardiac surgery had higher mortality, more readmissions to hospital, and poorer cognitive and functional outcomes. There is, thus, a strong consensus that patients with delirium do worse in many respects.

Although our study is large, it is not large enough to provide robust power, especially for identifying subgroups that could be clinically important. For example, it would have been interesting to evaluate a subset of especially frail patients. Approximately 15% of participants were excluded after randomization because of incomplete or missing data, with roughly comparable numbers excluded in each randomized group. CAM ICU testing was performed in the morning and only once a day, for logistic reasons. More frequent assessments might capture a higher incidence of delirium but potentially increase the risk of test fatigue and refusal to participate, especially in predelirious patients. We previously observed this attrition bias with patients experiencing postoperative delirium being more likely to refuse delirium assessment.²⁵  We did not record the use of major tranquillizers or other medications used to treat delirium nor any medical interventions, which may affect the incidence of delirium. Six months is a reasonable follow-up period; however, it remains possible that further recovery occurs between 6 and 12 months—although it seems unlikely that recovery so long after surgery would differ in the treatment groups. Postdischarge follow-up occurred via the telephone, but the telephone use of PostopQRS evaluations has been validated.¹⁶  No attempt was made to assess the impact of site heterogeneity or to make adjustments for site effects.

In summary, high-dose methylprednisolone did not reduce delirium nor improve the quality of recovery in a high-risk cohort of patients recovering from cardiac surgery. Previous work indicates that steroid administration also does not improve other major outcomes in cardiac surgical patients, and patients having cardiac surgery should not routinely be given steroids.

The authors thank Mrs. Zelda Williams, M.Chur., Research Nurse, Department of Surgery, University of Melbourne, Melbourne, Victoria, Australia, and the professional research staff at the other participating centers for contribution to patient recruitment and data collation.

This study was a substudy to the Steriods In caRdiac Surgery (SIRS) study and not separately funded. SIRS was funded by the Canadian Institutes of Health Research (Ottawa, Ontario, Canada).

Dr. Saager was supported by a Mentored Research Training Grant for Health Services Research from the Foundation for Anesthesia Education and Research (Schaumburg, Illinois) and the Anesthesia Quality Institute (Schaumburg, Illinois). The other authors declare no competing interests.

Full protocol available from Dr. Whitlock: richard.whitlock@phri.ca. Raw data available from Dr. Whitlock: richard.whitlock@phri.ca.