Perioperative outcomes of patients with restless legs syndrome: a single-center retrospective review

Abstract

Study Objectives:

There are multiple stressors in the perioperative period for patients with restless legs syndrome (RLS) that may be implicated in the worsening of symptoms. Our primary objective was to compare the perioperative course of patients with RLS to patients without the diagnosis.

Methods:

This was a single-center, matched-cohort, retrospective chart review of patients with RLS undergoing inpatient procedures from 2015–2019 matched 1:1 with patients without the diagnosis.

Results:

Patients with RLS had a higher comorbidity burden; specifically, pulmonary, renal, diabetes mellitus, and congestive heart failure. The perioperative course was notable for higher maximum pain scores for patients with RLS in the postanesthesia care unit (odds ratio, 1.29; 95% confidence interval, 1.19–1.40; P < .001). Postoperative patients with RLS also had higher maximum pain scores on postoperative days 0, 1, and 2. The odds of rapid-response calls were higher in patients with RLS (odds ratio, 1.43; 95% confidence interval, 1.18–1.73; P < .001). There were no other significant differences in postoperative complications. The odds of using RLS-triggering medications were lower in the RLS group (odds ratio, 0.85; 95% confidence interval, 0.78–0.92; P < .001).

Conclusions:

Our single-center retrospective review found that patients with RLS had higher pain scores in the postanesthesia care unit and on the first few postoperative days. Rapid-response team calls were more common in patients with RLS. RLS-triggering medications were significantly less likely to be used in patients with RLS. There were no significant differences in other postoperative events.

Citation:

Gali B, Silber MH, Hanson AC, Portner E, Gay P. Perioperative outcomes of patients with restless legs syndrome: a single-center retrospective review. J Clin Sleep Med. 2022:18(7):1841–1846.

BRIEF SUMMARY

Current Knowledge/Study Rationale: There is limited information about the perioperative course of patients with restless legs syndrome (RLS) despite the RLS triggers, including medications, stress, and anemia. The purpose was to review the perioperative course of patients with RLS in comparison to those without the diagnosis.

Study Impact: Pain scores were higher in patients with RLS, as were rapid-response calls. Future work to identify how best to manage pain in patients with RLS and understand the etiology of increased rapid-response calls is needed.

INTRODUCTION

Restless legs syndrome (RLS) is a neurologic, sensory motor, sleep-related movement disorder that is characterized by an urge to move accompanied by unpleasant limb sensations.¹ Typically, symptoms occur during wakefulness, worsening in the evenings, and are relieved by movement.² The typical management regimen includes alpha-2-delta ligands, dopaminergic agents, opioid agonists, and iron supplementation.³

The perioperative period can increase stress in multiple ways, including the anxiety of having surgery, the anesthetic course, surgical intervention, alterations in physiology, and medication changes. Many routinely utilized perioperative medications, including neuroleptics, antihistamines, and antidopaminergic agents, are implicated in the worsening of RLS symptoms.⁴ Major surgery may lead to anemia with iron reduction, limitation of movement, and sleep alterations, which are changes that may exacerbate RLS.⁵ Sleep-disordered breathing is known to worsen postoperatively and has been associated with postoperative morbidity.^6–8 Because of the impact of the increased stress, surgical blood loss, possible administration of RLS-triggering medications that are commonly utilized perioperatively, and omission of standard RLS treatment, patients with RLS may be at increased risk of complications during the perioperative period. There is limited literature on the impact of RLS on perioperative outcomes, mainly case reports and reviews of RLS with recommendations for management.^4,9–12

Our primary aim was to compare the perioperative course of patients with RLS with patients without the diagnosis. Our secondary aim was to determine whether home RLS medications (standard medications prescribed in the treatment of patients diagnosed with RLS) were continued or resumed postoperatively. We hypothesized that RLS would have a negative impact on the perioperative course, including postanesthesia care unit (PACU) events, postoperative complications, and hospital length-of-stay (LOS).

METHODS

This was a single-center, matched-cohort, retrospective chart review study. After institutional review board approval, we included those patients who had consented to have their data utilized for research. Patients with a documented diagnosis of RLS (based on intensive care unit [ICU] 9 and 10 codes from International Classification of Diseases, Tenth Revision codes) who underwent inpatient surgery at our institution from 2015–2019 were matched 1:1 with patients without a diagnosis of RLS.

Data regarding demography, chronic comorbidities, home medications that are typically utilized in the treatment of RLS, American Society of Anesthesiology (ASA) classification scores, and opioid use were collected. The ASA physical status classification incorporates a patient’s medical comorbidities to help identify those at risk during the perioperative period.¹³ The ASA ranges from 1 (healthy patient) to V (very ill patient not expected to survive without the operation), with VI as an organ donor who is brain-dead.

Perioperative events reviewed included type of surgical procedure, intraoperative time, estimated blood loss, fluids administered, and opioids administered. Perioperative refers to the immediate preoperative period, the intraoperative period, and the immediate postoperative period. PACU events documented included LOS, respiratory-specific events (apnea, desaturations, hypoventilation, pain/sedation mismatch), pain scores, and opioids administered. Postoperative events reviewed included pain scores on the first 3 days, ICU admission, rapid-response team calls, code team calls, acute myocardial infarction, thromboembolic events, use of known RLS-triggering medications commonly utilized in the perioperative period (haloperidol, quetiapine, droperidol, diphenhydramine, promethazine, scopolamine), and hospital LOS. Metoclopramide, olanzapine, and risperidone are not part of the postoperative order sets and are not commonly utilized during the postoperative period. Postoperative medications reviewed included opioids and medications utilized in the standard treatment of RLS (including iron, pramipexole, ropinirole, rotigotine, pregabalin, and gabapentin).

Statistical methods

Patients were matched by variable optimal matching on age (± 20 years), sex, date of surgery (± 4 years), ASA physical status classification score (exact according to 6 groups: 1–2, 3, 4, 5, 6, and unknown), type of surgical procedure (based on 46 categories observed in patients with RLS, eg, orthopedic lower body, general, orthopedic surgery, urology, orthopedic upper body), and emergent status.¹⁴

Patient characteristics (Table 1) and PACU and postoperative outcomes (Table 2) are presented according to RLS diagnosis as median (25th percentile, 75th percentile) for continuous variables (or ordinal variables with ≥ 6 distinct levels) and number (percentage) for categorical variables (or ordinal variables with < 6 distinct levels). Postoperative time was defined as immediately following surgery to time of discharge from the hospital. Absolute standardized differences in patient characteristics between RLS diagnosis groups are presented.

Table 1.

Patient characteristics, preoperative laboratory tests, and home medications (n = 11,584).

Variable	No RLS Diagnosis (n = 5,792)	RLS Diagnosis (n = 5,792)	Absolute SD
Age, y*	65 (55, 74)	65 (55, 74)	0.000
Sex*
Male	2,081 (36%)	2,081 (36%)	0.000
Female	3,711 (64%)	3,711 (64%)	0.000
Body mass index, kg/m2 (n = 11,582)	29.2 (25.0, 34.3)	30.9 (26.5, 36.3)	0.207
Charlson score	4 (3, 6)	5 (3, 7)	0.207
ASA physical status score*
1–2	1,816 (31%)	1,816 (31%)	0.000
3–4	3,778 (65%)	3,778 (65%)	0.000
5–6	19 (0%)	19 (0%)	0.000
Unknown	179 (3%)	179 (3%)	0.000
Emergent status*	436 (8%)	436 (8%)	0.000
Preoperative hemoglobin, g/dL (n = 11,065)	14.3 (13.4, 15.3)	14.6 (13.7, 15.5)	0.178
Preoperative ferritin, µg/L (n = 5,391)	94 (37, 224)	92 (49, 201)	0.067
Surgery duration, min	123 (79, 209)	117 (76, 193)	0.052
Chronic comorbidities
Cancer	1,342 (23%)	1,359 (23%)	0.007
Pulmonary disease	813 (14%)	1,403 (24%)	0.261
Renal disease	606 (10%)	946 (16%)	0.173
Asthma	505 (9%)	936 (16%)	0.227
Chronic obstructive pulmonary disease	365 (6%)	664 (11%)	0.182
Diabetes with complications	320 (6%)	698 (12%)	0.232
Congestive heart failure	369 (6%)	621 (11%)	0.156
Cerebrovascular accident	375 (6%)	519 (9%)	0.093
Myocardial infarction	299 (5%)	472 (8%)	0.120
Connective tissue disease	227 (4%)	330 (6%)	0.083
Home medications
Any RLS home medication	1,980 (34%)	3,690 (64%)	0.618
Gabapentin	609 (11%)	1,415 (24%)	0.373
Oxycodone/extended-release oxycodone	682 (12%)	965 (17%)	0.140
Iron	469 (8%)	907 (16%)	0.235
Tramadol	533 (9%)	840 (15%)	0.165
Pramipexole	18 (0%)	1,099 (19%)	0.666
Ropinirole	20 (0%)	735 (13%)	0.517
Hydrocodone	288 (5%)	413 (7%)	0.091
Pregabalin	118 (2%)	256 (4%)	0.135
Hydromorphone	88 (2%)	243 (4%)	0.161
Quetiapine	65 (1%)	163 (3%)	0.108
Codeine	70 (1%)	132 (2%)	0.069
Fentanyl	60 (1%)	129 (2%)	0.076
Morphine/extended-release morphine	56 (1%)	78 (1%)	0.024

Values number (percentage) for categorical variables and median (25th percentile, 75th percentile) for continuous variables. Absolute standardized differences are presented. *Matching variable. ASA = American Society of Anesthesiology, RLS = restless legs syndrome.

Table 2.

Estimated association between prior RLS diagnosis and postoperative outcomes (n = 11,584).

Outcome	No RLS Diagnosis	RLS Diagnosis	Estimate (95% CI)	P Value
PACU outcomes
PACU length of stay, min (n = 9,180)†	91 (67, 124)	91 (67, 125)	1.01 (0.99, 1.03)	.197
Any PACU event (n = 9,180)§**	189 (4%)	225 (5%)	1.21 (0.99, 1.49)	.069
Maximum PACU pain score (n = 8,926)‡	4 (0, 7)	5 (0, 8)	1.29 (1.19, 1.40)	<.001
Postoperative hospital outcomes
Maximum day 0 pain score (n = 9,136)‡	6 (4, 8)	7 (5, 8)	1.43 (1.32, 1.54)	<.001
Maximum day 1 pain score (n = 9,086)‡	6 (4, 8)	7 (5, 8)	1.56 (1.44, 1.68)	<.001
Maximum day 2 pain score (n = 6,042)‡	6 (4, 8)	7 (5, 8)	1.58 (1.43, 1.75)	<.001
ICU admission§	807 (14%)	774 (13%)	0.94 (0.83, 1.06)	.336
Delirium§	278 (5%)	316 (5%)	1.15 (0.97, 1.37)	.115
Rapid-response team call	246 (4%)	344 (6%)	1.43 (1.18, 1.73)	<.001
Medical emergency/code††	34 (1%)	35 (1%)	1.03 (0.64, 1.67)	.905
Acute myocardial infarction/DVT/PE§	517 (9%)	576 (10%)	1.13 (0.99, 1.29)	.068
Triggering medications used∞	2,118 (37%)	1,933 (33%)	0.85 (0.78, 0.92)	<.001
Postoperative hospital length-of-stay, d†	3 (1, 5)	3 (1, 5)	1.00 (0.98, 1.03)	.725

Values are number (percentage) for categorical outcomes and median (25th percentile, 75th percentile) for continuous and ordinal (pain score) outcomes. All analyses account for multiple observations per patient by using robust variance estimates (generalized estimating equations). Analyses are adjusted for date of surgery, age, sex, ASA physical status score, emergent status, and procedure type (matching variables) unless otherwise noted. Not all patients went to the PACU after surgery. Analyses are limited to matched sets where both patients had observed outcomes. The number of observations included in each analysis is presented unless all observations are included. †PACU and hospital LOS outcomes are modeled on the log scale with identity link assuming normal error distribution to satisfy distributional assumptions (linear regression model). Estimates are for the multiplicative increase in geometric mean associated with prior RLS diagnosis. §Categorical endpoints are modeled using logistic regression. Estimates are ORs and represent the multiplicative increase in the odds of an event associated with prior RLS diagnosis. ‡Ordinal outcomes are modeled using the proportional odds model. Estimates are ORs and represent the multiplicative increase in the odds of a higher pain score associated with prior RLS diagnosis. **PACU events include episodes of apnea, desaturations, hypoventilation, and pain/sedation mismatch. ††Because of a low event count, not all covariates were included in this model. The model was adjusted for age, sex, ASA physical status score, and emergent status. ^∞Triggering medications include haloperidol, quetiapine, droperidol, diphenhydramine, promethazine, scopolamine. ASA = American Society of Anesthesiology, CI = confidence interval, DVT = deep vein thrombosis, ICU = intensive care unit, LOS = length-of-stay, OR = odds ratio, PACU = postanesthesia care unit, PE = pulmonary embolism; RLS = restless legs syndrome.

Associations between RLS diagnosis and outcomes were assessed using multivariable linear regression models for continuous outcomes, logistic regression models for categorical outcomes, and proportional odds models for ordinal outcomes. All analyses accounted for potential correlation between observations of the same patient using generalized estimating equations. The estimated effect of RLS was summarized with corresponding 95% confidence limits and P values. To reflect the matched-set design, all analyses were adjusted for variables included in the matching algorithm, including age, sex, date of surgery, ASA physical status score, and procedure type except for the medical emergency/code event. Because of the low number of events, the model for the medical emergency/code event was adjusted for age, sex, ASA physical status score, and emergent surgery. PACU and postoperative hospital LOS were modeled on the log scale to satisfy distributional assumptions, and estimates corresponded to the multiplicative increase in the geometric mean associated with RLS. Pain scores were modeled using the proportional odds model, and estimates presented corresponded to odds ratios (ORs) and represented the multiplicative increase in the odds of a higher pain score associated with RLS. Categorical outcomes were modeled using logistic regression with estimates for the multiplicative increase in the odds of an event (ie, ORs) associated with RLS presented.

The proportion of patients with an RLS diagnosis who resumed home medications before discharge along with the proportion discharged without having resumed home medications was presented with 95% binomial confidence limits. Estimates at the start of day 3 were summarized (ie, following patients for the initial day of surgery plus 2 full days after the initial day of surgery). For each medication endpoint, patients were only included in the summary if they were indicated to be taking the given medication at home before surgery.

No formal power analysis was done a priori. In general, when a continuous outcome between 2 groups is compared, equal sample sizes of n = 5,700 in each group will provide a statistical power of > 99% to detect a difference between groups of 0.1 standard deviations (2-tailed alpha = 0.05). When binary endpoints are compared, this same sample size will have 99% power to detect a difference of ≥ 4 percentage points.

P values < .05 were considered statistically significant. All analyses were done using SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

Demographics

Of 5,849 procedures in 4,317 patients, 5,792 (99%) procedures in 4,291 patients with RLS were matched to the same number of procedures in 5,688 patients without RLS (Table 1). All variables in the matching algorithm were exactly equal between patients with and without RLS except for age (97% within 3 years), date of surgery (99% within 1 year), and ASA physical status score (10 patients were not matched exactly). When comparing basic demographics not included in the matching algorithm, we found that patients with RLS had a slightly higher body mass index and Charlson score (predicts 1 year mortality of patients based on a summary of 17 medical conditions) by approximately 0.2 standard deviations.^13–15 Both groups had similar preoperative hemoglobin and ferritin levels, both differing by < 0.2 standard deviations.

Chronic comorbidities differed between groups, with patients with RLS having an increased comorbidity burden including a higher percentage of pulmonary disease (24% vs 14%), renal disease (16% vs 10%), asthma (16% vs 9%), chronic obstructive pulmonary disease (11% vs 6%), diabetes mellitus (12% vs 6%), and congestive heart failure (11% vs 6%). Patients with RLS paients more often used at least 1 RLS home medication in the preoperative period before hospitalization (64% vs 34%). Patients with RLS used opioids more often compared to patients without RLS; however, the magnitudes of the differences tended to be small, with the exception of tramadol (15% vs 9%).

Perioperative

Surgical duration was similar between the RLS group and non-RLS group (median, 117 minutes vs 123 minutes). PACU LOS and PACU respiratory-specific events were similar for the 2 groups (Table 2). In the adjusted analysis, patients with an RLS diagnosis were more likely to have a higher maximum pain score in the PACU (OR, 1.29; 95% confidence interval (CI), 1.19–1.40; P < .001).

Postoperative

RLS diagnosis was associated with an increased odds of higher maximum pain scores on postoperative days 0 (OR, 1.43; 95% CI, 1.32–1.54; P < .001), 1 (OR, 1.56; 95% CI, 1.44–1.68; P < .001), and 2 (OR, 1.58; 95% CI, 1.43–1.75; P < .001) (Table 2). The odds of rapid-response calls were increased in patients with RLS (OR, 1.43; 95% CI, 1.18–1.73; P < .001). Other postoperative outcomes including ICU admission, delirium, acute myocardial infarction, and thromboembolic events were not associated with RLS diagnosis in the adjusted analysis (Table 2). The odds of RLS-triggering medication use were lower in the RLS group (OR, 0.85; 95% CI, 0.78–0.92; P < .001).

The timing of the resumption of medications for RLS management was different according to medication (Table 3). Before surgery, 1,415 (24%) patients with RLS were indicated to be using gabapentin at home; of those, (83%, 95% CI, 81%–85%) had resumed use before hospital day 3. Codeine was the least likely medication to be resumed before hospital day 3 (2%; 95% CI, 1%–6%). Both hydrocodone and iron were unlikely to be resumed before hospital day 3 (7%; 95% CI, 5%–10% and 18%; 95% CI, 15%–20%, respectively).

Table 3.

Cumulative proportion of patients with RLS restarting medication before discharge or leaving the hospital without restarting medications at hospital day 3.

Medication	Restarted Before Hospital Discharge, Estimate (95% CI)	Discharged From Hospital Without Restarting, Estimate (95% CI)
Gabapentin (n = 1,415)	0.83 (0.81, 0.85)	0.07 (0.06, 0.08)
Oxycodone/extended-release oxycodone (n = 965)	0.83 (0.80, 0.85)	0.07 (0.06, 0.09)
Iron (n = 907)	0.18 (0.15, 0.20)	0.33 (0.30, 0.36)
Tramadol (n = 840)	0.55 (0.51, 0.58)	0.21 (0.18, 0.23)
Pramipexole (n = 1,099)	0.83 (0.80, 0.85)	0.10 (0.08, 0.11)
Ropinirole (n = 735)	0.84 (0.81, 0.86)	0.06 (0.05, 0.08)
Hydrocodone (n = 413)	0.07 (0.05, 0.10)	0.35 (0.31, 0.40)
Pregabalin (n = 256)	0.68 (0.62, 0.73)	0.16 (0.12, 0.21)
Hydromorphone (n = 243)	0.71 (0.65, 0.76)	0.13 (0.09, 0.18)
Quetiapine (n = 163)	0.32 (0.25, 0.39)	0.10 (0.06, 0.15)
Codeine (n = 132)	0.02 (0.01, 0.06)	0.42 (0.33, 0.50)
Fentanyl (n = 129)	0.64 (0.55, 0.71)	0.09 (0.05, 0.14)
Morphine/extended-release morphine (n = 78)	0.44 (0.32, 0.54)	0.12 (0.06, 0.20)

Estimates are the cumulative incidence estimates of either restarting medication before discharge or leaving hospital without restarting medications by hospital day 3 (ie, day of surgery plus 2). Each analysis is limited to patients with prior RLS diagnosis and preoperative home use of the given medication. CI = confidence interval, RLS = restless legs syndrome.

DISCUSSION

Our single-center retrospective review found that patients with RLS had higher pain scores both in the PACU and on the first few postoperative days than patients without RLS. Rapid-response team calls for patients with RLS were significantly more common than for patients without RLS. Despite these findings, overall LOS, acute myocardial infarction, thromboembolic events, ICU admission, and code team activation were not associated with RLS diagnosis. RLS-triggering medications were significantly less likely to be used with patients with RLS than with patients without RLS, although 33% of patients with RLS received triggering medications.

There has been limited literature on the perioperative course of patients with RLS. Patients with other sleep disorders, including obstructive sleep apnea, have increased postoperative risk.^6–8 Increased postoperative risk of morbidity has also been suggested by the existing literature in patients with narcolepsy, insomnia, and treatment-emergent central sleep apnea.^16–19

Case reports and case series constitute much of the perioperative documentation of RLS symptoms and outcomes.^4,9,20,21 Because of the medication classes that are utilized to manage RLS, the stress of the perioperative period including routinely utilized medications, and the risk of blood loss, we hypothesized that patients with RLS would have an increased risk of complications compared to patients without RLS.

Our single-center, retrospective, matched-cohort study found increased pain scores and increased risk of rapid-response calls. The increased pain scores could have been because of opioid tolerance in patients receiving chronic opioid therapy for RLS or because of patients not differentiating between pain and RLS symptoms. This finding indicates that careful attention to differentiating pain from RLS and optimal management of both conditions may improve patient comfort. Because our study was retrospective, we were unable to determine the specific nature of the rapid-response calls but note that there was not an increased likelihood of ICU admissions or code team activation. We noted that 33% of patients with RLS received potentially triggering medications. This number of patients was lower than the number of patients without RLS who received these medications, but it is unclear whether this lower frequency was coincidental or was observed because care providers were aware of the potential RLS-triggering effect of these drugs.

Our study is limited by its single-center patient population and retrospective nature. We are unable to clearly ascertain what caused the rapid-response activations, and we are also unable to determine whether triggering medications were knowingly or inadvertently administered. In addition, the higher burden of chronic comorbidities in the patients with RLS compared to the that in the matched patients may have influenced the greater number of rapid-response calls in the RLS group. This information would be helpful in determining how best to manage patients with RLS in the perioperative period.

CONCLUSIONS

Our single-center, retrospective, matched-cohort study found that patients with RLS are more likely to have higher pain scores that matched those of patients without RLS and are more likely to require rapid-response team activation postoperatively. We did not find an increased risk of postoperative myocardial infarctions, thromboembolic events, ICU admission, code team activation, or increased LOS. Future work could help determine how best to manage pain and RLS with a proactive approach and identify the need for rapid-response intervention. In addition, an understanding of why and when RLS-triggering medications are administered to patients with RLS would be helpful in educating perioperative care teams.

ABBREVIATIONS

ASA

American Society of Anesthesiology

ICU

intensive care unit

LOS

length of stay

PACU

postanesthesia care unit

RLS

restless legs syndrome

DISCLOSURE STATEMENT

All authors have reviewed and approved the manuscript. This study was funded by the Research Subcommittee, Critical Care Independent Multidisciplinary Practice, Mayo Clinic, Rochester, Minnesota. The authors report no conflicts of interest.