Postoperative pain and opioid use are associated with postoperative delirium. We designed a single-center, randomized, placebo-controlled, parallel-arm, double-blinded trial to determine whether perioperative administration of gabapentin reduced postoperative delirium after noncardiac surgery.
Patients were randomly assigned to receive placebo (N = 347) or gabapentin 900 mg (N = 350) administered preoperatively and for the first 3 postoperative days. The primary outcome was postoperative delirium as measured by the Confusion Assessment Method. Secondary outcomes were postoperative pain, opioid use, and length of hospital stay.
Data for 697 patients were included, with a mean ± SD age of 72 ± 6 yr. The overall incidence of postoperative delirium in any of the first 3 days was 22.4% (24.0% in the gabapentin and 20.8% in the placebo groups; the difference was 3.20%; 95% CI, 3.22% to 9.72%; P = 0.30). The incidence of delirium did not differ between the two groups when stratified by surgery type, anesthesia type, or preoperative risk status. Gabapentin was shown to be opioid sparing, with lower doses for the intervention group versus the control group. For example, the morphine equivalents for the gabapentin-treated group, median 6.7 mg (25th, 75th quartiles: 1.3, 20.0 mg), versus control group, median 6.7 mg (25th, 75th quartiles: 2.7, 24.8 mg), differed on the first postoperative day (P = 0.04).
Although postoperative opioid use was reduced, perioperative administration of gabapentin did not result in a reduction of postoperative delirium or hospital length of stay.
Pain perception is a major risk factor for the development of postoperative delirium after major surgery. Patients who develop delirium often receive more opioids.
In patients undergoing major surgery, the adjunctive administration of gabapentin was evaluated for its efficacy in reducing pain, opioid use, and delirium.
Preoperative and postoperative administration of gabapentin reduced postoperative opioid use.
However, gabapentin did not reduce the incidence of delirium after major surgery.
DELIRIUM is a major challenge facing geriatric practice due to its prevalence, complex etiology, and potentially severe impact on patients and their families. One setting in which high rates of delirium are found is after major surgery. Postoperative delirium is associated with longer hospital stays, poor functional outcomes, and higher healthcare costs.1 Despite the prevalence and clinical importance of postoperative delirium, an effective therapy to prevent its occurrence has not been identified.
Patient risk for the development of delirium is determined by predisposing baseline vulnerabilities and exposure to factors that precipitate poor patient outcomes (e.g., pain or new medications associated with surgery). We and others have identified pain after surgery as an independent predictor of postoperative delirium2 and therefore a potentially important and modifiable precipitating factor for adverse cognitive outcomes. Opioids are another potential risk factor, because patients with postoperative delirium also received more intravenous opioids postoperatively than those without delirium.2
Based on results from a pilot study, we found a promising intervention involving the use of an adjunctive nonopioid therapy to reduce postoperative pain and the consumption of opioids, which ultimately resulted in a reduction of the incidence of postoperative delirium.3 Our main objective was to test the hypothesis that rates of delirium could be reduced through intensive supplementary pain management in addition to standard opioid analgesics after surgery. We conducted a double-blind, placebo-controlled study using gabapentin as an additional agent in the treatment of postoperative pain in older patients undergoing major noncardiac surgery.
Our specific aims were as follows: (1) to assess whether the administration of gabapentin was associated with a decreased occurrence of delirium; (2) to determine the extent to which gabapentin-associated reductions in pain and/or opiate use reduced the occurrence of delirium; and (3) to determine whether the administration of gabapentin was associated with shorter hospital stays. We hypothesized that postoperative intensive pain management using an adjuvant agent, gabapentin, would lead to a decrease in the amount of opioids received and a decrease in postoperative pain experienced, thereby resulting in a decrease in the incidence of postoperative delirium.
Materials and Methods
This was a double-blind, randomized, placebo-controlled study of 750 patients 65 yr of age or older undergoing spine surgery or joint replacement surgery at the University of California San Francisco Medical Center (San Francisco, California). The study received approval from the institutional review board, and all of the patients provided written informed consent. The trial was registered with clinicaltrials.gov (updated in April 2017 to clarify primary outcome; Identifier NCT00221338) and conducted in accordance with the original protocol. We formed a data and safety monitoring board to monitor participant safety and data quality and to evaluate the progress of the study ( appendix 2).
Potential subjects were recruited within 1 week before the planned surgical procedure. The inclusion criteria included patients 65 yr of age or older who were undergoing surgery involving the spine or arthroplasty of hips or knees who were fluent in English and with an anticipated length of hospital stay of at least 3 days after surgery. These types of patients were selected because they have substantial preoperative and postoperative pain and had a high incidence of postoperative delirium.2
Exclusion criteria included patients with known sensitivity to gabapentin; use of preoperative gabapentin, pregabalin, or other antiepileptics, spinal surgery that was two staged involving more than one surgical procedure to be performed within the same hospitalization period; emergency surgery; preoperative renal dialysis; or opioid tolerance (total daily dose of an opioid at or more than 30 mg morphine equivalent for more than 1 month within the past year; source: Institutional Chronic Pain Management Center).
A simple randomization method was used for this trial. Randomization into placebo or the gabapentin groups was created by a computerized random number generation method by the study statistician using a 1:1 randomization ratio. Randomization occurred after consent for study participation was obtained during the preoperative interview.
The randomization schedule was blinded from the investigators and treating clinicians because it was kept and administered by the central research pharmacy. The assignment of gabapentin versus placebo was made on the day of surgery, and the study drug was delivered by the research pharmacists directly to the preoperative holding area to be administered by clinical nurses to the study patients.
A balanced anesthetic was administered for study patients who underwent spinal surgery, which included a volatile anesthetic agent and intravenous agents such as propofol and fentanyl. Preoperatively, a femoral nerve block was placed for patients undergoing knee arthroplasty, and a lumber plexus block was placed for patients undergoing hip arthroplasty. Ropivacaine was used for both blocks. In addition to the blocks, the patients undergoing arthroplasty typically received either spinal anesthesia or general anesthesia. Postoperatively, all of the patients who had undergone spine surgery received on-demand patient-controlled analgesia with intravenous hydromorphone. For patients who underwent arthroplasty, postoperative analgesia was administered via the femoral nerve block or the lumbar plexus block for the first 2 postoperative days. In the case of additional analgesia for patients with incomplete analgesia from regional analgesia (less than 10% of cases), typically intravenous hydromorphone was administered via patient-controlled analgesia, and oral hydrocodone/acetaminophen, oxycodone/acetaminophen, or oxycodone was administered on demand by nurse administration.
Gabapentin Dosing Regimen
We administered either gabapentin 900 mg (or placebo) orally 1 to 2 h before surgery and anesthesia. This dose continued postoperatively for the first 3 days (300 mg three times per day). We adjusted the dose of gabapentin based on patient preoperative and postoperative renal function, as described previously.4 The rationale for choosing a clinical dose of 900 mg was based on a previous study, which demonstrated that this dose was well tolerated by older patients with herpes zoster and was effective in reducing the median pain level from baseline by more than 50%.5 Larger doses used in previous studies targeted primarily relatively healthy and younger surgical patients.6–15
Measurement of Cognitive Status
Trained research assistants who were blinded to the study drug assignment conducted cognitive tests preoperatively to determine the presence of delirium and to determine baseline cognitive function. The cognitive testing occurred in the preoperative clinic or ward and was repeated again daily for 3 days after surgery. Preoperative cognitive status was measured by the Telephone Interview for Cognitive Status (TICS) test,16 which was adapted from the Mini-Mental State Examination for use either in person or over the telephone. To minimize patient test burden, we used the nine-item word list test in lieu of the word naming in the TICS test during the preoperative testing.1 (The following cognitive tests were administered: the Word List Learning, the Digit Symbol Test, and the Controlled Verbal Fluency Test. Results are not included in this article.)
The primary outcome was postoperative delirium as measured by the Confusion Assessment Method (CAM). Secondary outcomes included postoperative pain, opioid use, and the length of hospital stay.
Measurement of the Primary Outcome: Postoperative Delirium
For the occurrence of delirium, we used the CAM rating scale,17 which was developed as a screening instrument based on operationalization of Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Text Revision, criteria for use by nonpsychiatric clinicians in high-risk settings. CAM has a sensitivity of 94% to 100%, a specificity of 90% to 95%, a high interobserver reliability,17 and a convergent agreement with four other cognitive status tests. Identifying delirium requires the presence of acute onset and fluctuating course, inattention, and either disorganized thinking and/or altered level of consciousness as measured by the CAM rating scale. Training of the research assistants in the use of the CAM was described in our previous publication.17
At approximately 24 h after surgery, the patient was rated on the Richmond Agitation and Sedation Scale (RASS).18 If a patient was too sedated to be interviewed (RASS score of –4 or –5), delirium status would be considered unevaluable. The severity of delirium was measured using the Memorial Delirium Assessment Scale (MDAS),19 an instrument that contains 10 items using information from the Mini-Mental State Examination and structured interview to rate delirium severity.
Measurement of Secondary Outcomes
During each assessment of cognitive status and delirium, patients rated their pain using the 11-point verbal version of the visual analog scale (0 = no pain and 10 = the worst pain imaginable). Postoperative intravenous opioid use was measured for the first 3 postoperative days. We converted all opioids to morphine equivalents as follows: hydromorphone and fentanyl doses were converted to morphine equivalents using the conversion formula: 1.5 mg hydromorphone = 10.0 mg morphine equivalents, 0.1 mg fentanyl = 10 mg morphine equivalents.20–22 Detailed conversion for all opioids are shown in appendix 3. Postoperative length of stay was measured and compared between interventional and control groups.
Measurement of Other Covariates
Preoperative risk was measured using the American Society of Anesthesiologists physical classification23 and the Charlson comorbidity index.24 Mood was measured using the standard screening tool for geriatric depression, the 15-question Geriatric Depression Scale.25 Other covariates included functional status including Activities of Daily Living (ADL) and Instrumental Activities of Daily Living (IADL). Independence in ADL and IADL was determined by asking the subjects if they needed the help of another person to do the activity.
For subgroup analysis, which was preplanned, we stratified patients by preoperative risk status: low risk was defined as patients with risk scores of three or less and high risk with risk score of more than 3 based on our previous risk prediction index where one point was assigned each to female sex, history of central nervous system disorder, high surgical risk, and age greater than 75 yr. A TICS score between 30 to 35 was assigned one point, and a TICS score less than 30 was assigned two points.26
We controlled for the severity of the surgical procedures such as duration and blood loss statistically (see statistical analysis section for details). Briefly, surgical risk was estimated by taking into consideration the type and duration of surgery and intraoperative blood loss.27
Measurement of In-hospital Drug-related Side Effects and Complications
Sample Size Calculation
The sample size was calculated based on the ability to detect a significant difference in rates of delirium between interventional versus placebo groups with an absolute difference in the delirium rate of 10% (25% vs. 15%) with 90% power. The level of significance was set at two-sided α = 0.05 to support the hypothesis that the delirium rates in the gabapentin group were different than that in the placebo group. The rates of delirium described above were determined using a combination of our earlier published pilot data of gabapentin delirium3 and rates of delirium among more than 500 subjects enrolled in our prospective observational study.2
All of the primary and secondary outcomes were analyzed according to the intention-to-treat paradigm. For the primary outcomes, to compare the postoperative delirium rates between gabapentin and placebo groups, we performed a chi-square test to determine the association between gabapentin administration and delirium rate. For the secondary outcomes, subjectively reported pain scores by the visual analog scale were stratified into low (1 to 3), medium (4 to 6), or high (7 to10) for each postoperative day. The difference in pain levels was measured by chi-square between gabapentin and placebo groups. Opioid use was defined as low versus high. Cutoff value for opioid dose use was based on the top third quartile (75th percentile) on 3 postoperative days, respectively. Specifically, a daily use of more than 22 mg of morphine equivalents in a 24-h period was considered to be the top 75th percentile of opioid doses, a high dose. Low opioid use was defined as patients who used 22 mg or less of morphine equivalents in a 24-h period. The justification of stratifying opioids dose into high- versus low-dose for analysis was based on our previous work on a model of prediction of postoperative delirium.26 The difference in morphine equivalent dosing on postoperative day 1 between the gabapentin and placebo groups was determined by the Mann–Whitney U test. Hospital lengths of stay between groups were compared using the unpaired t test.
In subgroup analyses, we conducted poststudy stratification of clinical characteristics relevant to translation of results. Postoperative delirium rates were stratified by surgery type, anesthesia type, dose of postoperative opiates and pain, and preoperative risk, and reported P values were adjusted using Bonferroni correction as needed.
In addition, logistic regression was performed to analyze the effect of gabapentin on postoperative delirium with sex and ADL as covariates. For other outcome variables of interest, including the MDAS, P values were calculated based on the chi-square test if the variables were categorical; otherwise P values were based on independent t tests or Mann–Whitney U test for data that were not normally distributed. To compare delirium-free days between the two treatment groups, we performed the Mantel–Haenszel test to take into account the ordinal distribution of delirium-free days. All of the data were reported as mean ± SD. Median values (25th, 75th quartiles) were included if the data were not normally distributed.
The study began in January 2006 and ended in January 2014. The patient recruitment scheme is depicted in figure 1. Overall, 697 patients were included in this intention-to-treat algorithm. A total of 198 patients had total hip arthroplasty, 183 had knee arthroplasty, and 316 underwent spine surgery. The demographic variables of the patients who received gabapentin versus placebo are shown in table 1. Overall, there were more women (55.1% vs. 45.5%) and more patients who were dependent in one or more activities of daily living (34.4% vs. 25.7%) in the gabapentin compared with the placebo group.
Completion of Study Drugs
The compliance of study drugs received by patients was similar between the gabapentin- versus the placebo-treated patients. All of the patients received the preoperative study drugs. For the first postoperative day, 88.6% of patients in the gabapentin-treated group completed the assigned dosing versus 92.8% of the patients who received placebo; for the second postoperative day, 80.9% of patients in the gabapentin completed the assigned dosing versus 80.6% in the placebo group; and for the third postoperative day, 43.1% of patients in the gabapentin completed the assigned dosing versus 39.6% in the placebo group. The lower rate of receiving study drugs on the third postoperative day was in part due to earlier unanticipated discharge (84% of patients who did not receive the study drug or placebo were discharged earlier than anticipated).
Study Outcomes Measurement: Primary and Secondary Outcomes
No patient had preoperative delirium. The overall incidence of postoperative delirium in any of the first 3 days for the entire cohort was 22.4% (95% CI, 19.3% to 25.5%; 24.0% in the gabapentin group; 95% CI 19.2% to 28.8%; and 20.8% in the placebo group; 95% CI 16.2% to 25.4%). The difference of 3.2% (95% CI, –3.2% to 9.7%) was not statistically significant (P = 0.30). When stratifying by surgery type (table 2) or anesthesia type (table 3), the incidence of postoperative delirium was also not significantly different between the gabapentin versus the placebo group.
Pain scores for the first 3 postoperative days are shown in table 4. Overall, patients who experienced high postoperative pain levels had higher rates of postoperative delirium compared with those with lower pain levels (19.5%; 95% CI, 14.9% to 24.1%; vs 9.1%; 95% CI 6.3% to 11.9%; P = 0.0001). However, the delirium rates were not significantly different between the gabapentin and placebo groups when stratified by pain levels: delirium in low pain level group was 9.2% (95% CI, 5.5% to 14.8% gabapentin) versus 6.9% (95% CI, 3.5% to 12.7% placebo; P = 0.16); in the medium pain level group it was 11.6% (95% CI, 6.2% to 20.2% gabapentin) versus 15.6% (95% CI, 9.8% to 23.5% placebo; P = 0.52); and in the high pain level group it was 33.4% (95% CI, 22.0% to 44.6% gabapentin) versus 17.1% (95% CI, 9.7% to 27.8% placebo; P = 0.05). Data for postoperative days 2 and 3 were similar (table 5).
Despite the lack of effect on postoperative delirium, gabapentin was shown to be opioid sparing, particularly in the first postoperative day. For example, morphine equivalents are lower for the gabapentin-treated group, with a median (25th, 75th quartile) of 6.7 mg (1.3, 20.0 mg) when compared with the control group of 6.7 mg (2.7, 24.8 mg) for the first postoperative day (P = 0.04 based on the Mann–Whitney U test). Note that this test compares the rank sum between the two groups. Boxplots of the opioid use for the first 3 days are shown in figure 2. The amount of postoperative opioid use and pain level were not associated with postoperative delirium (table 5).
When we stratified patients with respect to their preoperative risk of developing postoperative delirium, high-risk patients had significantly higher rates of postoperative delirium than low-risk patients. However, the rates of delirium on any of the postoperative days were not significantly different between the gabapentin and placebo groups with preoperative risk stratification. The incidence of postoperative delirium in any of the first 3 postoperative days was 21% in the low-risk gabapentin group versus 18.8% in the low-risk placebo group (P = 0.56) and 47.5% in the high-risk gabapentin group versus 39.4% in the high-risk placebo group (P = 0.65).
We also compared the severity of delirium using the MDAS. Again, comparison of the MDAS scores between the gabapentin- and placebo-treated groups was not different for each of the 3 postoperative days (day 1, 5.2 ± 2.8 vs. 5.2 ± 2.5, P = 0.85; day 2, 4.6 ± 2.7 vs. 4.9 ± 2.8, P = 0.35; day 3, 3.8 ± 2.4 vs. 4.1 ± 2.1, P = 0.37).
All of the study patients had delirium data for 1 or more of the first 3 postoperative days. For those with missing delirium data for 1 or 2 of the 3 postoperative days (n = 102), we determined whether missing data might bias the results. Overall, patients with missing delirium data compared with those with no missing data were younger, more likely to be men, had higher level of education, a higher incidence of alcohol use, a lower incidence of a history of central nervous system disorders, were less likely to depend on one or more activities of IADL, and had lower mean Charlson comorbidity scores. Excluding the 102 patients with incomplete delirium assessments, the rates of delirium between gabapentin and placebo groups were 28.0% versus 24.4% (95% CI of the difference, –3.8% to 11.0%; P = 0.67). This comparison suggests that those patients with missing delirium data did not have covariates that were associated with an increased risk of postoperative delirium.
We additionally evaluated whether patients who were treated with gabapentin had a difference in delirium-free days for the first 3 postoperative days when compared with placebo-treated patients. This analysis included all of the intention-to-treat patients with a hospital length of stay of 3 days or longer, and patients with missing delirium data were excluded. Again, the comparison did not show any difference between the two groups (table 6).
Regarding the secondary outcome, the length of hospital stay in patients with postoperative delirium was significantly longer than those without delirium (5.5 ± 3.1 days, 95% CI, 5.2 to 5.8 days vs. 3.9 ± 2.8 days, 95% CI, 3.6 to 4.2 days; P < 0.0001). However, there was no difference in length of hospital stay between patients treated with gabapentin versus placebo (4.4 ± 3.4 days, 95% CI, 4.0 to 4.7 days vs. 4.1 ± 2.3 days, 95% CI, 3.9 to 4.3 days; P = 0.26).
Safety Evaluation of Gabapentin Administration
Regarding the safety of perioperative gabapentin administration, we measured postoperative clinically significant oversedation as determined by the RASS scores and also postoperative adverse events. Overall, the incidence of serious oversedation rates (RASS scores of –4 or –5) were not different on any of the postoperative days between the gabapentin or placebo groups (day 1, 2/333 = 0.6% vs. 1/329 = 0.3%, P = 0.61; day 2, 4/321 = 1.3% vs. 1/328 = 0.3%, P = 0.37; and day 3, 0/289 = 0% vs. 1/284 = 0.4%, P = 0.60). Detailed comparison of the RASS scores is shown in table 7. We also compared other potential drug-related side effect, such as dizziness, and no significant difference was found between study groups, including 10 of 345 patients (2.9%) in the gabapentin group versus 5 of 340 patients (1.5%) in the placebo group (P = 0.30). No patient reported nystagmus or ataxia in either study group. The incidence of adverse postoperative events relating to the cardiovascular, pulmonary, renal, or neurologic systems and infection and thrombotic events also was not significantly different between gabapentin- versus placebo-treated groups (8.9% vs. 12.7%; P = 0.13).
This large prospectively conducted randomized clinical trial revealed no difference in rates of postoperative delirium when gabapentin was administered perioperatively to older surgical patients when compared with placebo, despite its opioid-sparing effects.
Comparison with Previous Studies
Aside from our previous pilot study,30 no previous study has investigated the use of perioperative gabapentin as a means to reduce postoperative delirium. However, there have been a number of other pharmacologic interventional trials aimed at delirium reduction in surgical patients but with mixed results. Most studies found no effects of pharmacologic treatments with antipsychotics or anticholinesterase agents on delirium reduction.31–33 Although several small studies have suggested that antipsychotics may reduce the risk of delirium, these finding were not supported by meta-analyses.34,35 Moreover, the prophylactic administration of both conventional and atypical antipsychotics to older patients is potentially hazardous, with cardiac and metabolic side effects reported because of age-related changes in pharmacokinetics and pharmacodynamics, as well as potential adverse drug interactions with other medications.36 Hence, the evidence to date does not support the use of antipsychotics for prevention of postoperative delirium.
Other types of intervention reported involved the evaluation of sedatives or anesthetic agents, such as dexmedetomidine or ketamine.37,38 However, these clinical trials produced mixed results, and definitive therapies based on trials with adequate sample size have yet to be developed. A recent large trial in postoperative patients recovering in the intensive care unit reported that intravenous infusion of dexmedetomidine immediately after surgery reduced the occurrence of postoperative delirium when compared with placebo.39 Whether these results can be generalized to nonintensive care unit patients remains to be determined. In contrast to pharmacologic prophylactic treatment, nonpharmacologic intervention, such as fast-track surgery,40 specialized postoperative geriatric wards,41 and proactive geriatric consultation,42 reported more success in delirium reduction. Thus, a recent systematic assessment conducted by the American Geriatrics Society concluded that only nonpharmacologic interventions were proven to be efficacious and should be widely practiced. Recently, it has been proposed that deep anesthetic depth contributes to an increased rate of postoperative delirium.43–46 However, the mechanism of this deep anesthesia effect has not been completely elucidated despite a recent report that burst suppression on electroencephalogram indicative of deep anesthesia may have been the etiologic factor.46
Our results did not support the second hypothesis that gabapentin-associated reductions in pain and/or opiate use reduced the occurrence of delirium. Although we did find that gabapentin was opioid sparing, we did not demonstrate that the opioid-sparing effect resulted in a reduction of postoperative delirium. It is likely that the opioid-reducing effect of gabapentin was attenuated by the concomitant use of postoperative analgesia, such as femoral nerve block or lumbar plexus block in some of our patients who underwent arthroplasty. This explanation is in part supported by previous studies, which demonstrated that reducing opioid exposure may be achieved with regional analgesia, such as femoral nerve block or fascia iliaca block47,48 ; both of these techniques have been shown to be associated with lower risk of postoperative delirium, but definitive large-scale trials are lacking.
A recent meta-analysis examined the effect of preoperative gabapentin in reducing postoperative opioid consumption.49 In the 17 randomized trials that were examined, the dosages of gabapentin ranged from 300 to 1,200 mg. Our study chose the 900-mg preoperative dose, which is within the range identified in this review. Of note, meta-regression analyses identified a statistical association between reduced postoperative opioid consumption and gabapentin dosage.49
Potential Study Limitations
First, despite a computerized randomization of recruited patients, we observed some unbalance across treatment groups with respect to preoperative patient characteristics. However, inclusion of the covariates that were different between groups did not affect results of the outcome measurements. Second, we studied patients with three types of surgery, and the methods of intraoperative anesthetics and postoperative management were different between groups. However, inclusion of the type of surgery and anesthetics as covariates did not affect the rates of postoperative delirium between the gabapentin-treated and placebo groups. Third, because of changing perioperative practice patterns during the duration of the study, the inclusion of multimodal oral analgesics, such as acetaminophen and nonsteroidal anti-inflammatory agents administered to the placebo patients who underwent arthroplasty surgery perioperatively might have resulted in lower rates of postoperative delirium in that group when compared with historical control subjects. Lastly, we did not specifically measure other opioid-related side effects, such as pruritus, nausea, vomiting, and whether gabapentin through its opioid-sparing action produced salutary effects will need to be determined by additional investigations.
Results from this large, randomized, double-blind, placebo-controlled trial showed that perioperative administration of gabapentin did not result in a lower rate of postoperative delirium in older patients undergoing major spine and arthroplasty surgery, despite its opioid-sparing effects. Our results suggest that the prophylactic use of gabapentin as a means to reduce postoperative delirium is not indicated.
Supported in part by National Institutes of Health grants AG031795-05 and 5R21AG048456-02 (Bethesda, Maryland; to Dr. Leung).
The authors declare no competing interests.
Appendix 1: Perioperative Medicine Research Group
The principal investigator is Jacqueline M. Leung, M.D., M.P.H. Research associates Stacey Chang, B.A., Gabriela Meckler, B.A., Stacey Newman, B.A., Tiffany Tsai, M.D., Vanessa Voss, M.D., and Emily Youngblom, B.A., participated in patient recruitment, cognitive assessments, data entry, and data management.
Appendix 2: Data Safety Monitoring
Data Safety Monitoring Board
Before the commencement of patient recruitment, we formed a data and safety monitoring board to monitor participant safety and data quality and to evaluate the progress of the study. The data and safety monitoring board focused on performance (subject recruitment, retention, and follow-up; flow of data forms; protocol adherence; and quality of data) and safety (magnitude and frequency of adverse events were measured).
Grading Method and Attribution for Adverse Event Reporting
Adverse events (AEs) were graded by the principal investigator using the 0- to 5-point scale where 0 = no AE or within normal limits or not clinically significant; 1 = mild AE, did not require treatment; 2 = moderate AE, resolved with treatment; 3 = severe AE, resulted in inability to carry on normal activities and required professional medical attention; 4 = life-threatening or disabling AE; and 5 = fatal AE. The principal investigator, who was blinded to the study group assignments, determined the relationship of AEs to the test study drug as one of the following: not related, possibly related, and definitely related.
Description of Anticipated Adverse Events
Sedation (evaluated by the Richmond Agitation and Sedation Scale), dizziness (patient self-report), ataxia, and nystagmus were adverse outcomes reported previously for patients treated with gabapentin.
Safety Data Evaluated
The data that were evaluated included but were not limited to subject interview, vital signs, physical examination results, clinical test results such as creatinine, and postoperative analgesic dosages.
Adverse Event Reporting
All serious adverse events (both anticipated and unanticipated) were reported to the University of California San Francisco Committee on Human Research, and National Institutes of Health (San Francisco, California).
Events that May Cause Termination/Dropout of a Participant from the Study
Adverse events (anticipated or unanticipated), subject’s unwillingness to continue with the study, or treating physician’s request were included.
Data and Safety Monitoring Board Meetings
A total of three data and safety monitoring board meetings were conducted throughout the study period to evaluate safety (first meeting), efficacy (second meeting, results blinded to the investigators), and final report (third meeting).
Early stopping rule was based on the development of prohibitive toxicity by the treatment. Reduction in postoperative delirium alone was not used as the sole early stopping criterion, because the standard treatment (typically opioids) was not considered to be an unsatisfactory option at present. Other reasons for termination of that study included poor accrual, significant negative effect of the treatment on the primary outcome, and excessive loss to follow-up.
Interim analyses focused on whether the death and adverse event rates for patients in this study exceeded the current in-hospital rates for this surgical population at our institution. Evidence of overwhelming efficacy was determined by statistically comparing the rate of delirium between the placebo and treatment groups. We used the O’Brien–Fleming guidelines for stopping due to overwhelming evidence of efficacy,42 in which more stringent P values were used to determine stopping earlier in the trial compared with later in the trial.
Statistical results were not the sole basis for the decision to stop or continue the trial.43 Additional factors used in consideration of termination of the intervention included the need for evaluation of this medication in the surgical setting given its pervasive off-label use in surgical patients in the postoperative period. Currently no large-scale experimental evidence exists regarding a variety of patient outcomes associated with use of this drug in the surgical setting. Therefore, the interim analyses included other important clinical outcomes in addition to the effect of the intervention on postoperative delirium, such as sedation and other possible drug-related side effects. Another factor to be considered included the effectiveness in lowering pain when compared with the placebo-treated patients.
Appendix 3: Opioid Conversion to Morphine Equivalents
Appendix 4: Definitions of Nonfatal Postoperative Complications
In-hospital course was followed daily until discharge for the new occurrence of postoperative outcomes, which included ischemic cardiac complications (new occurrence of chest pain, electrocardiogram changes, or cardiac enzyme changes), clinically diagnosed myocardial infarction, dysrhythmias, and heart failure or clinically significant respiratory complications (pulmonary edema, tracheal reintubation, pulmonary consolidation on chest x-ray, pneumothorax, or pleural effusion). Renal insufficiency was defined as a new requirement of dialysis postoperatively or elevation of serum creatinine. Neurologic event was defined as new occurrence of transient ischemic attack or stroke, delirium, or confusion. Infection required documentation by a positive culture. Gastrointestinal event was defined as bowel ischemia, perforation, bleeding, cholecystitis or pancreatitis, or elevated postoperative liver enzymes with or without postoperative jaundice. Thromboembolic event was defined as deep venous thrombosis or pulmonary embolism. Other postoperative outcomes measured included death, surgical complications, and reoperation during the same hospitalization.31