Beware the rebound effect: regional anesthesia increases opioid utilization after humerus fracture surgery

Abstract

Introduction

Regional anesthesia (RA) is used reduce pain in proximal humerus and humeral shaft fracture surgery. The study hypothesis was that RA would decrease opioid demand in patients undergoing fracture surgery.

Materials and methods

Opioid demand was recorded in all patients ages 18 and older undergoing proximal humerus or humeral shaft fracture surgery at a single, Level I trauma center from 7/2013 – 7/2018 (n = 380 patients). Inpatient opioid consumption from 0–24, 24–48, and 48–72 h and outpatient opioid demand from 1-month pre-operative to 90-days post-operative were converted to oxycodone 5-mg equivalents (OE's). Unadjusted and adjusted models were constructed to evaluate the impact of RA and other factors on opioid utilization.

Results

Adjusted models demonstrated increases in inpatient opioid consumption in patients with RA (6.8 estimated OE's without RA vs 8.8 estimated OE's with RA from 0–24 h post-op; 10 vs 13.7 from 24–48 h post-op; and 8.7 vs 11.6 from 48–72 h post-op; all p < 0.05). Estimated cumulative outpatient opioid demand was significantly higher in patients with RA at all timepoints.

Discussion

In proximal humerus and humeral shaft fracture surgery, RA was associated with increased inpatient and outpatient opioid demand after adjusting for baseline patient and treatment characteristics.

Introduction

The incidence of surgically treated humerus fractures is increasing. ¹ Similarly, proximal humerus fractures have been increasing in frequency for more than three decades and are the third most common fracture in the elderly.^2,3 Subsequent postoperative pain management is typically achieved with opioid analgesia. Orthopaedic providers are among the highest in opioid prescription rates with 98.6% of patients receiving these medications after undergoing orthopaedic procedures.^4–7 However, with opioid misuse and addiction continuing to rise, alternative methods to achieve effective post-procedural analgesia are being explored. Regional anesthesia is one such method that has been demonstrated to reduce pain in the acute period after orthopaedic surgery.^8–10 However, some studies have demonstrated a “rebound pain” phenomenon where patients who receive regional anesthesia develop hyperalgesia and counterproductively utilize increased amounts of opioids. ¹¹

Few studies have explored the impact that regional anesthesia can have on opioid demand after humerus fracture surgery. It remains unclear how regional anesthesia influences early postoperative demand in these fractures and the potential impact on chronic pain and long-term opioid use. ¹² The purpose of this study is to evaluate the impact of these modalities on inpatient opioid consumption and outpatient opioid demand in patients undergoing proximal humerus and humeral shaft fracture surgery. The study hypothesis is that RA will be associated with a decrease in inpatient opioid consumption but have little impact on outpatient opioid demand.

Materials and methods

Study design

This is a retrospective, observational study of inpatient opioid consumption and outpatient opioid demand in all patients age 18 years and older undergoing proximal humerus and humeral shaft fracture surgery at a single institution from 7/1/2013 – 7/1/2018. This study is designed and reported in accordance with the STROBE statement on reporting observational studies. ¹³

Variables and data sources

Inpatient opioid utilization (0–24 h, 24–48 h, 48–72 h post-operative) and outpatient opioid demand (1-month pre-op to 2-weeks, 6-weeks, and 90-days post-operative) were recorded in all patients ages 18 and older undergoing proximal humerus or humeral shaft fracture fixation (Current Procedural Terminology codes 23615, 23616, 23630, 23670, 23680, 24515, and 24516) at a single, Level I trauma center from 7/2013 – 7/2018. Opioid utilization was converted to oxycodone 5-mg equivalents using conversion factors recommended by the CDC. ¹⁴ Additional baseline and treatment factors including RA usage, age, sex, race, body mass index (BMI), smoking status, American Society of Anesthesiologists (ASA) score, injury mechanism, additional injuries, open fracture, and additional surgery. All patients having undergone fracture fixation as identified above were included in the study regardless of supplemental analgesics since many of these would be non-prescription. General 90-day post-operative complications were also recorded through chart review including mortality, surgical site infection, compartment syndrome, loss of fixation, deep vein thrombosis (DVT), pulmonary embolism (PE), falls, delirium, and ileus.

As shown in Table 1, patients without RA tended to be younger, male, non-Caucasian and had higher rates of smoking, high-energy injury, additional injury, and open fracture. Continuous (182 of 276, 65.9%) blocks were more common than single shot blocks (92 of 276, 33.3%). One patient (0.4%) had multiple routes (continuous and single shot blocks) and one patient (0.4%) had a route that was not documented. Interscalene (218 of 276, 79.0%), supraclavicular (42 of 276, 15.2%), and infraclavicular (15 of 276, 5.4%) peripheral nerve blocks were the most common anatomic locations. Suprascapular (3 of 276, 1.1%), brachial plexus (2 of 276, 0.7%), and not documented (1 of 276, 0.4%) were uncommon locations. Patients without RA stayed for a median of 5 (2, 13) days (quartile 1, quartile 3) while patients with RA stayed for 1.3 (1, 2.3) days. These differences are likely related to the previously-noted differences between cohorts, with a higher proportion of patients not receiving RA having sustained polytrauma.

Table 1.

Baseline patient, injury, and treatment characteristics for patients with and without RA.

Factors	Without RA (n = 104)	With RA (n = 276)	p-value
Age (years)	43.5 (27, 61.6)	61.8 (51.1, 69.1)	<0.001
Female sex	47 / 104 (45.2%)	176 / 276 (63.8%)	0.002
Caucasian race	62 / 104 (59.6%)	232 / 276 (84.1%)	<0.001
BMI (kg/m2)	27 (23.7, 32.2)	27.3 (23.7, 32.2)	0.66
Smoking	25 / 104 (24%)	39 / 276 (14.1%)	0.031
Pre-operative opioid usage	12 / 104 (11.5%)	36 / 276 (13%)	0.86
ASA score
ASA 3 or greater	49 / 104 (47.1%)	130 / 276 (47.1%)	1
ASA 1	16 / 104 (15.4%)	21 / 276 (7.6%)	0.032
ASA 2	39 / 104 (37.5%)	125 / 276 (45.3%)	0.2
ASA 3	43 / 104 (41.3%)	110 / 276 (39.9%)	0.82
ASA 4	6 / 104 (5.8%)	20 / 276 (7.2%)	0.82
Injury mechanism
High energy mechanism	67 / 104 (64.4%)	80 / 276 (29%)	<0.001
High energy mechanism
Crush injury	1 / 104 (1%)	0 / 276 (0%)	0.27
Fall from height	7 / 104 (6.7%)	38 / 276 (13.8%)	0.074
GSW	6 / 104 (5.8%)	3 / 276 (1.1%)	0.015
MVC	42 / 104 (40.4%)	37 / 276 (13.4%)	<0.001
MVC vs ped	11 / 104 (10.6%)	2 / 276 (0.7%)	<0.001
Low energy mechanism
Assault	0 / 104 (0%)	3 / 276 (1.1%)	0.57
Ground level fall	30 / 104 (28.8%)	161 / 276 (58.3%)	<0.001
Pathologic	5 / 104 (4.8%)	19 / 276 (6.9%)	0.64
Sporting injury	2 / 104 (1.9%)	13 / 276 (4.7%)	0.37
Additional injury	61 / 104 (58.7%)	46 / 276 (16.7%)	<0.001
Open fracture	22 / 104 (21.2%)	5 / 276 (1.8%)	<0.001
Fracture location
Proximal humerus	42 / 104 (40.4%)	194 / 276 (70.3%)	<0.001
Humeral shaft	62 / 104 (59.6%)	82 / 276 (29.7%)	<0.001
Additional surgery within 7-days	11 / 104 (10.6%)	3 / 276 (1.1%)	<0.001
Additional surgery within 90-days	16 / 104 (15.4%)	5 / 276 (1.8%)	<0.001
LRA route none	104 / 104 (100%)	0 / 276 (0%)	n/a
LRA route continuous	0 / 104 (0%)	182 / 276 (65.9%)	n/a
LRA route multiple	0 / 104 (0%)	1 / 276 (0.4%)	n/a
LRA route not documented	0 / 104 (0%)	1 / 276 (0.4%)	n/a
LRA route single shot	0 / 104 (0%)	92 / 276 (33.3%)	n/a
LRA (number)	0 (0, 0)	1 (1, 1)	n/a
Length of stay (days)	5 (2, 13)	1.3 (1, 2.3)	<0.001

Proportions (percentages) and medians (Q1, Q3) displayed. p-values from Fisher's exact test or Wilcoxon rank-sum. Red coloring highlights statistical significance. BMI: body mass index; ASA: American Society of Anesthesiologists score; GSW: gunshot wound; MVC: motor vehicle crash; MVC vs ped: motor vehicle vs pedestrian crash; LRA: local regional anesthesia.

Pain protocol

Our institution's multimodal pain regimen generally includes oral opioids administered according to a visual analog scale (VAS) pain scale (generally 5–15 mg oxycodone every 4 h as needed for pain), intravenous (IV) opioids (typically hydromorphone) for breakthrough pain, and scheduled acetaminophen. Adjunctive oral and IV non-steroidal anti-inflammatory pain medications (NSAID's) are not frequently used after proximal humerus and humeral shaft fracture surgery at our institution. Patients are considered for RA by their treating anesthesiologist Surgeons do not typically infiltrate wound edges with local anesthetic, though there was no standardization of this in this retrospective study. Discharge pain medications are prescribed by the treating team, and opioids are commonly prescribed. While the decision to discharge patients with opioids is made on a case-by-case basis by the primary team, 320 of 380 (84.2%) of patients in this series received a discharge opioid prescription. Outpatient opioid fills and refills were made at the discretion of the treating clinician and was not standardized in this retrospective study.

Missing data

There were 13 of 393 (3.3%) of patients whose BMI could not be determined. Overall results were evaluated with and without these patients and found to be similar. In order to adjust for the potential impact of this characteristic, these patients were excluded from the multivariable analyses leaving 380 patients for analysis.

Statistical analysis

Medians with quartiles and proportions with percentages were used to display descriptive statistics. Wilcoxon rank-sum and Fisher's exact test were used to compare baseline and outcome differences between patients with and without RA. Adjusted models were then planned. Histograms of opioid outcomes demonstrated positive skew, so generalized linear modeling with log link function and negative binomial distribution was planned. Propensity score weighting was performed since treatment (RA vs no RA) was not randomly assigned, including age, sex, race, BMI, smoking, pre-op opioid usage, ASA score (binarized to 1 to 2 vs 3 or more), injury energy (binarized to high vs low energy), presence of additional injuries, open injury, fracture location (proximal humerus vs humeral shaft), and additional surgery within 7-days post-fracture surgery as model covariates. Generalized linear modeling was then carried out incorporating the propensity score weights, allowing for a “doubly robust” analysis.^15,16 Incident rate ratios from adjusted modeling were displayed, but treatment effect was then simulated for each patient within the dataset using model parameters to provide a median and 95% confidence interval for RA vs no RA. R and R Studio (R: A Language and Environment for Statistical Computing, R Core Team, R Foundation for Statistical Computing, Vienna, Austria, 2020) were used for statistical calculations. p-values less than 0.05 were considered significant.

Funding

There were no sources of funding for this study.

Results

After adjustment for baseline patient and treatment factors, RA was associated with significant increases in inpatient opioid consumption from 0–72 h post-op (Table 2 and Figure 1) of 2–3.7 oxycodone 5-mg equivalents (OE's). Similarly, RA was associated with increased outpatient opioid demand at all timepoints after adjustment for baseline patient and treatment factors (Table 3 and Figure 2) with patients receiving RA being prescribed 55.4 more OE's by 90-days than patients without RA. Patients receiving perioperative RA were noted to have significantly higher rates of opioid fill and refill rates in adjusted analyses (Table 4). Lastly, as shown in Table 5, general 90-day outcomes did not differ significantly between groups with exception of a higher rate of PE in patients without RA.

Table 2.

Adjusted inpatient oxycodone 5-mg equivalents consumed in patients with and without RA.

Timeframe	Oxycodone without RA (95% CI)	Oxycodone with RA (95% CI)	Incident rate ratios (95% CI, p-value)
0–24 h post-op	6.8 (5.2, 8.9)	8.8 (6.7, 11.4)	1.29 (1.1, 1.51; p = 0.002)
24–48 h post-op	10 (6.6, 16.2)	13.7 (9, 21.9)	1.37 (1.07, 1.75; p = 0.013)
48–72 h post-op	8.7 (5, 14.8)	11.6 (6.8, 19.4)	1.33 (1.00, 1.77; p = 0.048)

Red coloring highlights statistical significance. Simulated estimates from multivariable model (95% CI) displayed. Incident rate ratios and p-values from multivariable model.

Figure 1.

Predicted inpatient opioid consumption histogram in patients with and without RA. Vertical bars represent mean consumption.

Table 3.

Adjusted outpatient oxycodone 5-mg equivalents prescribed in patients with and without RA.

Timeframe	Oxycodone without RA (95% CI)	Oxycodone with RA (95% CI)	Incident rate ratios (95% CI, p-value)
1 month pre-op to 2 weeks	112.9 (88.6, 144.8)	137.3 (108, 175.9)	1.22 (1.05, 1.41; p = 0.008)
1 month pre-op to 6 weeks	126.6 (98.9, 161.8)	168.4 (131.4, 215)	1.33 (1.15, 1.54; p = <0.001)
1 month pre-op to 90 days	139.5 (108, 179.5)	194.9 (150.5, 251)	1.4 (1.2, 1.63; p = <0.001)

Red coloring highlights statistical significance. Simulated estimates from multivariable model (95% CI) displayed. Incident rate ratios and p-values from multivariable model.

Figure 2.

Predicted outpatient opioid prescription histogram in patients with and without RA. Vertical bars represent mean prescription.

Table 4.

Adjusted odds of opioid fill and refill. Odds ratio (95% CI) displayed.

Factors	Discharge to two week opioid refill	Two week to six week opioid fill	Six week to ninety day opioid fill
LRA	1.65 (1.19, 2.28; p = 0.002)	1.54 (1.1, 2.17; p = 0.011)	1.93 (1.3, 2.88; p = 0.001)

p-values from multivariable modeling. Complete model displayed in Appendix.

Table 5.

General 90-day perioperative complications.

Outcomes	All subjects (n = 380)	Without RA (n = 104)	With RA (n = 276)	p-value
Non-opioid outcomes
Mortality	5 / 380 (1.3%)	1 / 104 (1%)	4 / 276 (1.4%)	1
SSI	6 / 380 (1.6%)	1 / 104 (1%)	5 / 276 (1.8%)	1
Mechanical failure	5 / 380 (1.3%)	0 / 104 (0%)	5 / 276 (1.8%)	0.33
DVT	1 / 380 (0.3%)	1 / 104 (1%)	0 / 276 (0%)	0.27
PE	3 / 380 (0.8%)	3 / 104 (2.9%)	0 / 276 (0%)	0.02
ACS	1 / 380 (0.3%)	1 / 104 (1%)	0 / 276 (0%)	0.27
Falls	10 / 380 (2.6%)	4 / 104 (3.8%)	6 / 276 (2.2%)	0.47
Delirium	6 / 380 (1.6%)	2 / 104 (1.9%)	4 / 276 (1.4%)	0.67
Ileus	0 / 380 (0%)	0 / 104 (0%)	0 / 276 (0%)	n/c

Proportions (percentages displayed). Unadjusted p-values from Fisher's exact test “N/c”: not calculable due to low event rate; SSI: surgical site infection; DVT: deep venous thrombosis; PE: pulmonary embolism; ACS: acute compartment syndrome.

Appendix Tables 1–3 display complete results of multivariable modeling. For inpatient opioid consumption, adjusted models demonstrated significant decreases with increased age, increased BMI, high-energy injury (0–24 h post-op), additional surgery (0–24 h post-op), and proximal humerus fracture location (0–48 h post-operative) and significant increases with female sex, Caucasian race, smoking, pre-operative opioid usage, elevated ASA, additional injury, open fracture, and RA (Appendix Table 1). For outpatient opioid prescribing, adjusted models demonstrated significant decreases with increased age, female sex, and additional injury and significant increases with pre-operative opioid usage, elevated ASA, high-energy injury, and RA (Appendix Table 2). Odds of opioid fill or refill were significantly decreased with increased age, Caucasian race, and pre-operative opioid usage (discharge to 2-weeks post-op) while it was significantly increased with elevated BMI, smoking, pre-operative opioid usage (2-weeks to 90-days post-op), elevated ASA, high-energy mechanism, and RA.

Appendix Tables 4–6 highlight the importance of adjustment on the study results. Unadjusted results suggested an early (0–24 h) decrease in opioid demand in patients with RA and no significant difference after that time. Adjusted results demonstrated that other factors such as age, injury energy, and additional surgery significantly impacted the results to a greater degree than RA. Further, outpatient opioid demand did not differ significantly in unadjusted analyses, but, as mentioned previously, after adjustment, RA was associated with increased outpatient opioid demand at multiple time points. Thus, even when ignoring important baseline patient and treatment factors, RA did not improve opioid demand after 24-h post-operative, and the unadjusted improvement in opioid demand is likely due to baseline differences between cohorts that was negated with model adjustment.

Discussion

In this study of perioperative opioid demand in patients undergoing fixation of proximal humerus and humeral shaft fractures with and without RA, there were increases in inpatient opioid consumption and outpatient opioid demand at all timepoints after adjusting for baseline patient and treatment characteristics. Age, sex, race, BMI, smoking, pre-operative opioid usage, ASA score, injury energy, additional injury, and additional surgery were significant drivers of inpatient and outpatient opioid demand metrics.

There is limited research on the impact of regional anesthesia on acute opioid consumption when considering humeral shaft and proximal humerus fractures. A 2019 systematic review and meta-analysis by Iliaens et al. found only two studies that focused on the proximal humerus. Supporting this study's findings, they found that a single shot interscalene block was associated with increased pain 16–24 h after surgery. ¹² While minimal research has focused solely on humeral shaft or proximal humerus fractures, a number of studies have examined opioid consumption after regional anesthesia for shoulder surgery. A 2015 randomized control trial by Bjornholdt et al. found that regional anesthesia led to a reduction in opioid consumption in the first 24 h after shoulder arthroplasty but no difference from hours 24–96. ¹⁷ A 2015 meta-analysis determined that regional anesthesia reduced opioid consumption in the first 12 h after shoulder surgery, but no reduction from hours 12–24 or when looking at the first 24 h as a whole. ¹⁸ Multiple other studies have found similar results.^19,20 Our unadjusted results demonstrated similar findings, but, after adjustment for baseline patient and treatment factors, there was no significant improvement in inpatient opioid demand and a significant late inpatient increase in opioid demand. Since the prior results were relevant to elective shoulder surgery, these results expand our understanding of the impact of RA on patients undergoing fracture treatment.

With regards to outpatient post-operative opioid consumption, this study found that individuals that received regional anesthesia were more likely to receive larger volume of opioids at all time points up to 90 days post-op. To our knowledge no study has previously examined outpatient opioid use after humeral shaft and proximal humerus fracture fixation. However, two retrospective studies of claims data, Sun et al. in 2018 and Mueller et al. in 2017, found no difference in opioid consumption long-term after regional anesthesia for total knee arthroplasty and shoulder arthroplasty respectively.^21,22 Another review of health care claims after abdominal surgery also demonstrated no effect after abdominal surgery. ²³ While our study findings may be influenced by the fact that we included opioids received within 1 month prior to surgery, these opioids contribute to the overall burden of opioids that patients were prescribed and are likely reflective of opioids received at the time of the patient's initial injury. Importantly, we re-analyzed study data and excluded the 1-month pre-operative opioid contribution. Patients with RA still had a significantly greater opioid demand by 6-weeks and 90-days post-operative. Even when not including adjustment in analyses of outpatient opioid prescribing results, there were no significant differences between patients with and without RA, which suggests that, at best, RA does not decrease outpatient opioid prescribing.

Compared to previous research which has largely demonstrated some acute benefit to regional anesthesia with largely no difference in opioid consumption after the initial hours after surgery, our study demonstrated a negative effect of regional anesthesia on both inpatient and outpatient opioid consumption. Many of the existing studies included in the discussion focused solely on a single block location and technique and compare block efficacy at different anatomic locations in patients undergoing elective shoulder surgeries. Our review was not specific to a particular block location though most patients received interscalene, supraclavicular, or infraclavicular nerve blocks and compares patients both with and without blocks. While a 2017 study by Hussain et al. found no difference in efficacy between interscalene and suprascapular block locations, more research is needed to determine if certain block locations are more likely to reduce acute pain, decrease opioid consumption or potentially even increase the risk of rebound pain.^18,24,25 Suprascapular blocks were rarely used in this study (3 of 276). Our study found the largest difference in opioid consumption at 24–48 h post-operative. This difference may be related to rebound pain, which refers to pain from previously-masked post-surgical pain as a block wears off. Continuous nerve blocks have been suggested as a means to prevent rebound pain. Kim et al. demonstrated that continuous interscalene blocks had lower pain initially and did not report rebound pain in the first day. ²⁶ 65% of patients in our study had continuous nerve blocks, and the benefit of RA was still not definitively seen. Further research should determine whether continuous blocks simply result in delayed rebound pain at the discontinuation of the block as well as determine whether these longer blocks could impact opioid consumption long-term. ¹²

Uncontrolled acute pain, which could include hyperalgesia after a block, has been associated with an increased risk of chronic pain and opioid use.^27,28 Therefore, it is conceivable that rebound pain not only leads to increased consumption after the immediate postoperative period but also affects long-term outpatient use as well. Furthermore, regional anesthesia has potential negative clinical impact that should be considered when indicating patients. These procedures often add time to the perioperative experience for the patient, surgeon, and anesthesiologist, incur additional cost to the healthcare system, and are an additional procedure for the patient with potential risk for neurovascular injury. Some surgeons have a negative view of regional anesthesia due to perceived delays in surgical care and unpredictable clinical results as reported in a 2004 study of orthopaedic surgeons. ²⁹

This study has a few limitations to be considered. First, we are unable to evaluate non-opioid post-operative analgesia since this information was not collected prospectively. We also could not study pain, since this was not collected at routine time intervals. Despite this limitation, the studies mentioned in the discussion did not find any benefit to regional anesthesia after 24 h post-operative, and it is, therefore, unlikely that we would have demonstrated improvements in pain in this retrospective study. Thirdly, we provide data on outpatient opioid demand in terms of opioid prescription volume, fills, and refills. While we can evaluate inpatient opioid consumption as this data is available in our medical records, we cannot measure outpatient opioid consumption in this retrospective study as this data was not prospectively collected. Further, opioid prescribing was not standardized as this was a retrospective study, so prescription volumes and rates were at the discretion of any clinician treating the patient. Additionally, there were baseline differences between patients with and without RA. While we included propensity score weighting and adjustment for these characteristics, it is not possible to remove all bias in these measured co-variates, and there may be additional unmeasured factors that influence opioid demand. Lastly, we evaluated patients that received RA through a variety of techniques, which reflects real-world practice. Because of the heterogeneity, we were unable to adjust for each specific block type and combination within multivariable analyses. However, we did perform limited non-parametric inferential statistics which demonstrated minimal changes between block types while inpatient and a small increase in outpatient opioid prescribing in patients with continuous interscalene blocks compared to single-shot supraclavicular or interscalene blocks. However, these results could not be adjusted for potential effect modifiers, so this data is not displayed. While heterogeneity may decrease the specificity of our results to one particular technique, we believe that analyzing the data in this fashion broadens the clinical interpretation since it more closely matches the scenario encountered in clinical practice, where the choice of block is often performed by the treating anesthesiologist Strengths of this manuscript include the large sample size of patients with proximal humerus and humeral shaft fractures. Additionally, the statistical analysis allowed investigation of the impact of RA on perioperative opioid consumption within the bounds of a retrospective study. This study also evaluated the impact of these RA on outpatient opioid demand, which has not previously been reported.

In conclusion, perioperative RA in humerus fracture surgery did not reduce inpatient opioid consumption or outpatient opioid demand. Additionally, it may be associated with a rebound pain effect impact on patients leading to increased opioid demand. This information should give clinicians pause in the utilization of this analgesia modality and adds much needed depth to patient-physician and physician-physician discussions when determining the optimal treatment for humerus fracture patients.