Impact of Intraoperative Remifentanil on Postoperative Pain and Opioid Use in Thyroid Surgery

Abstract

Background and Objectives:

Remifentanil infusion is used as an intraoperative anesthetic for thyroidectomy, but has been associated with acute opioid tolerance and hyperalgesia. A national shortage of remifentanil provided an opportunity to study postoperative pain in patients undergoing thyroidectomy.

Methods:

Retrospective review of prospectively collected data from an outpatient surgery center. Primary analysis compared patients treated before and after remifentanil shortage.

Results:

Median postoperative opioid consumption was 20 MMEs among those treated in the high-dose period and 15 MMEs in the low-dose period. Remifentanil/weight received was a significant predictor of requiring a postoperative narcotic (p=0.006). Total non-remifentanil narcotics administered were equivalent but patients in the low dose period received higher amounts of intraoperative long acting narcotics.

Conclusions:

Remifentanil infusion for thyroid surgery is associated with higher postoperative pain and postoperative narcotics requirement. While a hyperalgesia state is possible, shifting of longer acting narcotics from intraoperative to postoperatively is also supported.

Introduction

Thyroid cancer is the most common cancer in young women, and conservative estimates report that 100,000 thyroid surgeries are performed annually in the United States.^1,2 Despite excellent oncologic outcomes, five percent of patients may develop opioid addiction and abuse following thyroid surgery.³ This finding calls for analyses of best practices in perioperative management in thyroid surgery to reduce opioid use while providing adequate intraoperative anesthesia and postoperative pain control.

Remifentanil is a μ-opioid agonist, used as a continuous infusion in thyroid surgery because of its rapid onset, rapid recovery, and predictable metabolism.⁴ However, the use of remifentanil has been associated with acute opioid tolerance and postoperative hyperalgesia.^5,6 Importantly, increased severity of postoperative pain is associated with the development of chronic pain ^5,7, which may further contribute to long term narcotic use.

At the Josie Robertson Surgery Center at Memorial Sloan Kettering Cancer Center, it was noticed that thyroidectomy patients that received remifentanil also received higher postoperative total morphine milligram equivalents (MME) of opioid analgesics than patients undergoing comparable ambulatory procedures. Critical review also identified that remifentanil infusions (RI) were used more frequently in the intraoperative anesthetic management of patients undergoing thyroidectomy than other procedures. A national shortage of the drug in 2017 created a natural experiment, allowing for rigorous study of the effect of intraoperative remifentanil infusion on postoperative pain and narcotic consumption in our high-volume thyroid surgery center. The purpose of this study was to examine if use of intraoperative RI in thyroid surgery leads to increased use of postoperative opioids. Specifically, the study was done to improve and standardize perioperative management of patients with thyroid disease.

Materials and Methods

We performed a retrospective review of prospectively collected data from our outpatient surgery center from April 2016 to July 2018. In December 2017 remifentanil was removed from our formulary due to a shortage. From April 2016 through February 2018 the rate of remifentanil usage in thyroidectomies was 88%, abruptly decreasing to 1.2% from February through July 2018 after supplies on hand were consumed. A total of 497 patients underwent 500 thyroid lobectomies or total thyroidectomies at our hospital from April 1^st, 2016 to July 9^th, 2018. We excluded one patient who received intraoperative ketorolac and secondary procedures, leaving 496 for analysis. Accordingly, the primary analyses compared those treated from April 1st, 2016 through February 1st, 2018 versus those after, in an intent to treat approach. A secondary analysis was planned using the continuous total amount of intraoperative remifentanil received divided by patient’s weight (remifentanil/weight), irrespective of study period. The units of opioid consumption were standardized to MME.

The distribution of the amount of postoperative narcotics received was skewed right, since a significant fraction of patients received no postoperative narcotics. To address the non-normality of this outcome we investigated an outcome of receipt of any postoperative narcotics using multivariable logistic regression, after adjusting for age, gender, operative time, BMI, ASA score (1-2 vs 3), surgical procedure (lobectomy versus total thyroidectomy), intraoperative acetaminophen dose, anesthesiologist-administered lidocaine dose, and intraoperative narcotics (spline terms for intraoperative MMEs were allowed where significant). Additionally, we categorized the amount of postoperative narcotics received based on the quintiles of postoperative narcotics to define none, low, moderate, high, and very high amounts of postoperative narcotics and tested its association with remifentanil using ordinal logistic regression adjusting for the aforementioned covariates.

We estimated the association between the use of remifentanil and narcotics (total and intraoperative) after adjusting for age, gender, operative time, BMI, ASA score (1-2 vs 3), surgical procedure (lobectomy versus total thyroidectomy), intraoperative acetaminophen dose, and anesthesiologist-administered lidocaine using separate multivariable linear regression models. We also tested the association between remifentanil and postoperative narcotics to assess whether remifentanil use was associated with a hyperalgesic state.

We also tested for any association between remifentanil and maximum postoperative pain score using multivariable linear regression adjusting for the aforementioned covariates in addition to total intraoperative MMEs (spline terms for intraoperative MMEs were allowed where significant). We used multivariable logistic regression to test the association between remifentanil and the need for a postoperative nausea and vomiting rescue medication, adjusting for Apfel score, age, operative time, and surgical procedure (lobectomy versus total thyroidectomy). We assessed whether the use of remifentanil was associated with rapidity of recovery by testing the association between remifentanil and discharge time using multivariable linear regression adjusted for surgical start time.

Results

Patient demographics and total opioid consumption (MMEs)

Remifentanil was used for anesthesia in 88% of cases in the remifentanil high-use period, and 1.2% in the remifentanil low-use period. Patient characteristics are displayed in Table 1. Eighty three percent of the patients studied had surgery in the high use period. Patients treated in the low-use period had lower ASA scores, and a higher percentage of patients underwent lobectomy (p-values=0.027, and 0.0006, respectively; Table 1). Mean total (non-remifentanil) opioid consumption was 54 MMEs (95% CI 51, 56) among those treated in the high-dose period and 58 MMEs (95% CI 52, 64) in the low-dose period; the association between study period and total opioids received was not statistically significant (p=0.3; Table 2).

Table 1:

Patient characteristics by study period. Values are displayed as median (quartiles) or frequency (percentage).

	April 1, 2016-February 1, 2018 (N=410; 83%)	February 2, 2018-July 9, 2018 (N=86; 17%)	p-value*
Received Remifentanil
No	49 (12%)	85 (99%)
Yes	361 (88%)	1 (1.2%)
Age	46 (36, 57)	46 (36, 55)	0.7
Male	111 (27%)	23 (27%)	1
BMI	27 (23, 32)	27 (24, 31)	0.5
Preoperative Weight (kg)	75 (63, 91)	75 (63, 87)	0.5
ASA Score
1	19 (4.6%)	7 (8.1%)	0.027
2	249 (61%)	61 (71%)
3	142 (35%)	18 (21%)
Surgery Type
Partial	162 (40%)	48 (56%)	0.006
Total	248 (60%)	38 (44%)
Operative Time (mins)	115 (94, 139)	117 (99, 153)	0.089
Intraoperative Acetaminophen Dose (mg)
None	22 (5.4%)	4 (4.7%)	0.5
1-999 mg	4 (1.0%)	2 (2.3%)
1000mg	384 (94%)	80 (93%)
Anesthesiologist-administered Lidocaine
0	108 (26%)	69 (80%)	<0.0001
1-100	27 (6.6%)	2 (2.3%)
101-200	275 (67%)	15 (17%)

^*p-values determined by Wilcoxon Rank-sum for continuous variables and Fisher’s exact test for categorical variables

Table 2:

Means (95% CI) of total narcotics (not including remifentanil) received (MMEs), intraoperative narcotics (not including remifentanil) received (MMEs), maximum postoperative pain and discharge time by study period and the mean difference (95% CI) for the average patient.

	April 1, 2016-February 1, 2018 (N=412)	February 2, 2018-July 9, 2018 (N=87)	Adjusted Difference	95% CI	p-value
Total Narcotics (MMEs)	54 (26)	58 (27)	3.9	−2.8,11	0.3
Intraoperative narcotics (MMEs)	31 (15)	39 (15)	7.5	3.8, 11	<0.0001
Maximum Postoperative Pain Score	5.9 (2.0)	5.8 (2.1)	−0.14	−0.66, 0.38	0.6
Discharge time (hours)	10 (1.4)	10 (1.4)	−0.07	−0.42, 0.27	0.7

Intraoperative narcotics and maximum postoperative pain mean differences were adjusted for age, gender, BMI, ASA, surgical approach, amount of intraoperative acetaminophen, amount of anesthesiologist administered lidocaine, and OR time. Maximum postoperative pain was also adjusted for intraoperative narcotics. Estimates of discharge time were adjusted for surgery start time. Differences represent the increase when a patient was treated in the second study period. 95% Confidence intervals and p-values obtained using the delta method.

Intraoperative opioid consumption (MMEs)

Patients treated in the remifentanil high-use period received, on average, 7.5 less intraoperative MMEs than those treated in the low-use period (p<0.0001; Table 2). The distributions of opioids used intraoperatively for both remifentanil and non-remifentanil based anesthesia are shown in Figure 1. The association between amount of remifentanil dose by weight and intraoperative total MMEs was significant and is displayed in Figure 2 (p<0.0001). As remifentanil/weight increased, the total intraoperative MMEs decreased.

Figure 1:

Histogram of narcotics (intraoperative and postoperative) received. The black shaded region represents the distribution for those who did not receive remifentanil while the grey shaded region represents those who received remifentanil.

**From intraoperative: one patient with 110.5 MME intraoperative narcotics and who received remifentanil was excluded. From postoperative: two patients were excluded who received 100.5 and 119 MME postoperative narcotics and both received remifentanil.

Figure 2:

Predicted intraoperative narcotics (solid line) with 95% confidence interval (dashed lines) by intraoperative remifentanil/weight corresponding to the average patient (p<0.0001). The model allowed for splines for remifentanil/weight and adjusted for age, gender, BMI, ASA, surgical procedure, amount of intraoperative acetaminophen, amount of anesthesiologist administered lidocaine, and operative time. The histogram represents the distribution of remifentanil/weight.

Postoperative opioid consumption (MMEs) and pain scores

The median postoperative opioid consumption was 20 MMEs (Q1 5.0, Q3 33) among those treated in the high-dose period and 15 MMEs (Q1 3.8, Q3 30) in the low-dose period. The distribution of opioid consumption was non-normal, and 20% of the study population did not require any postoperative narcotics (Figure 1). The estimated decrease in risk of requiring a postoperative narcotic associated with the low-dose period was 12.2% (95% CI −0.1%, 24.5%; Table 3). We also established that remifentanil/weight was a significant predictor of requiring a postoperative narcotic (p=0.0062; Figure 3).

Table 3:

Estimated probability (95% CI) of requiring a postoperative narcotic and requiring rescue medication for postoperative nausea or vomiting by study period and risk difference (95% CI) for the average patient.

	April 1, 2016-February 1, 2018 (N=410)	February 2, 2018-July 9, 2018 (N=87)	Risk Difference (95% CI)	p-value
Postoperative narcotic	82% (78%, 85%)	69% (58%, 81%)	12.2% (−0.14%, 25%)	0.053
Rescue medication for nausea or vomiting	38% (34%, 43%)	38% (28%, 48%)	0.01% (−11%, 11%)	1

Requiring a postoperative narcotic was adjusted for age, gender, BMI, ASA, surgical approach, amount of intraoperative acetaminophen, amount anesthesiologist administered lidocaine, and amount of intraoperative narcotics (MME). Requiring rescue medication for postoperative nausea or vomiting was adjusted for Apfel score, age, OR time, and surgical approach. Difference represents the decrease when a patient was treated in the second study period. 95% Confidence intervals and p-values obtained using the delta method.

Figure 3:

Predicted probability of receiving postoperative narcotics (solid line) with 95% confidence interval (dashed lines) by intraoperative remifentanil/weight corresponding to the average patient (p=0.0062). The model allowed for splines for remifentanil/weight and adjusted for age, gender, BMI, ASA, surgical procedure, amount of intraoperative acetaminophen, amount of anesthesiologist administered lidocaine, operative time, and amount of intraoperative narcotics. The histogram represents the distribution of remifentanil/weight.

On multivariable ordinal analysis we found patients in the low-use period were less likely to require higher quintiles of postoperative opioid use, with an odds ratio of 0.62 (95% CI 0.39, 1.00; p=0.050). The estimated probability being in the highest quintile of postoperative narcotics (37.5 MMEs or more) for the average patient was 19.6% and 15.5% in those treated in the high-use and low-use periods, respectively. For each increase of 0.1 mcg/kg of remifentanil/weight the odds of requiring a higher quintile of postoperative narcotics was 1.30 (95% CI 1.08, 1.56; p=0.005). For the average patient who received 0, 0.1, 0.2, and 0.3 mcg/kg of remifentanil/weight, the estimated probability that they will require the highest quintile of postoperative narcotics was 14%, 18%, 22%, and 27%, respectively. The maximum postoperative pain score was significantly associated with remifentanil dose (p=0.025; Figure 4), but we did not find a significant difference between study periods (p=0.6, Table 2).

Figure 4:

Predicted maximum postoperative pain score (solid line) with 95% confidence interval (dashed lines) by intraoperative remifentanil/weight corresponding to the average patient (p=0.025). The model allowed for splines for remifentanil/weight and adjusted for age, gender, BMI, ASA, surgical procedure, amount of intraoperative acetaminophen, amount of anesthesiologist administered lidocaine, and operative time. The histogram represents the distribution of remifentanil/weight.

Postoperative nausea and discharge time

The probability of requiring rescue medication for postoperative nausea or vomiting was not significantly associated with study period or remifentanil/weight (p=0.99 and 0.8, respectively). We found no evidence that study period was associated with discharge time (p=0.7; Table 2). While the association between remifentanil/weight and discharge time was significant (p=0.033), the estimated increase in discharge time going from remifentanil/weight of 0 to 0.4 mcg/kg did not result in a meaningful increase (35 minutes, 95% CI 3 minutes, 67 minutes).

Discussion

It has been demonstrated that intraoperative RI use may lead to acute opioid tolerance and a hyperalgesic state with increased postoperative pain and analgesic requirements.^6,8,9 We did observe that thyroidectomy patients who received RI intraoperatively also require higher postoperative total MMEs when compared to patients undergoing other ambulatory procedures. A national shortage of remifentanil created an abrupt change in clinical practice and allowed for analysis of this surgical population in an unbiased setting. Utilizing an intention to treat, pre- and post-shortage approach, we demonstrated that intraoperative RI was associated with increased postoperative opioid consumption. Consistent with this finding, we also found that maximum pain scores were higher with increasing doses of remifentanil (mcg/kg) in a dose-dependent manner. However, when we considered the total dose of long-acting opioids administered, both intra- and postoperatively, there was no significant difference between the groups.

Others have demonstrated similar effects of intraoperative remifentanil in both thyroidectomy and other surgical procedures. When comparing volatile anesthesia to intraoperative RI, increased pain scores were observed in the RI population. ¹⁰ This was found to be more common in patients receiving “high dose” RI.^11,12 This finding is consistent with our data showing for each increase of 0.1 mcg/kg of remifentanil, the odds of requiring a higher quintile of postoperative narcotics was significantly increased. In patients undergoing orthopedic surgery, remifentanil was associated with a significant increase in postoperative sore throat as measured by pain scores. ¹³ This finding is particularly relevant in thyroid surgery given that sore throat is reported in more than 60% of patients secondary to the proximity of the endotracheal tube and surgical field.¹⁴ Wound hyperalgesia, the size of the area of hypersensitivity surrounding an incision, has also been associated with high dose remifentanil.¹⁵ In the present cohort, increased sore throat and incisional pain with RI may have contributed to our observation of increase maximum pain scores.

Increased post procedure pain and narcotic use in patients receiving remifentanil has been attributed to acute opioid tolerance and hyperalgesia^5,16. An alternative hypothesis is that patients who receive RI intraoperatively, as opposed to those receiving conventional opioids as part of their anesthetic, may leave the OR with a long acting analgesic deficit and therefore require a “catch up” period during which time pain scores are higher and more narcotics are required. This explanation is supported by our finding that the total non-remifentanil opioids given was equivalent between groups, irrespective of remifentanil administration. During the “low remifentanil use” period, patients received higher intraoperative MMEs of longer acting opioids, such as fentanyl and dilaudid, followed by lower doses postoperatively. While the total MMEs given are comparable, the temporal distribution of opioid use across the perioperative stages is different. Increased opioid use is shifted to the postoperative period in patients that receive remifentanil intraoperatively, while the patients in the “low dose“period receive more opioids intraoperatively. Since strategies targeting a remifentanil hyperalgesia-induced mechanism, including concurrent naloxone ¹² and magnesium sulfate ¹¹, did not decrease postoperative pain or analgesic consumption, reduction of postoperative pain after RI may simply require more aggressive supplementation with longer acting opioids prior to emergence.

Independent of mechanism, patients who received remifentanil intraoperatively also have worse pain control beyond the acute perioperative period.^15,17 One year postoperatively, intraoperative remifentanil use was predictive for chronic pain in a dose-dependent manner after sternotomy.¹⁷ Others have reported that wound hyperalgesia corresponds to an increased risk of chronic pain at 3 and 6 months.¹⁵ Unsatisfactory postoperative pain management can lead to chronic pain. ^5,7 Considering the excellent prognosis of thyroid cancer, long-term disability related to opioid use reported in up to five percent of patients³ is a significant concern. Limiting perioperative opioid consumption with adjunctive dexmedetomidine ¹⁸ and gabapentin ¹⁹ are two promising approaches.

Limitations exist in the present work. The data presented represents pain scores and opioid consumption during the short period of time the patient remains in the hospital postoperatively. It is unclear if the increased postoperative opioid consumption associated with remifentanil results in increased opioid consumption after discharge. Additionally, during the period of the remifentanil shortage, a higher percentage of patients underwent thyroid lobectomy rather than total thyroidectomy. However, importantly, there was no difference in MMEs or pain scores in patients who had a total thyroidectomy versus lobectomy. Finally, practice patterns across providers, including need for rescue medications, use of inhalation agents, and nonopioid pain control administration, vary. The intention to treat approach was used to reduce this bias.

Conclusions

Patients who receive intraoperative remifentanil infusion for thyroid surgery have more pain, and subsequently also have a higher likelihood of requiring postoperative narcotics. While acute opioid tolerance and a hyperalgesia state is one possible mechanism, the present data suggest that the decreased use of longer acting narcotics intraoperatively with RI shifts MME requirements to the postoperative period, secondary to an opioid “catch up” period. Continual review of clinical practices provides opportunities to improve perioperative management. With the goal of reducing postoperative pain and potentially long-term opioid use, alternative non-opioid based analgesics should be explored for thyroidectomy, consequential to these observations.

Synopsis:

Hyperalgesia and increased pain is a potential side effect of intraoperative remifentanil infusion. In this retrospective review of prospective data, we compared patients that received remifentanil versus other analgesics during thyroid surgery. We found that patients that received remifentanil had higher postoperative pain scores and a higher postoperative narcotic requirement compared to patients that did not receive remifentanil, though the total amount of narcotics (intraoperative and postoperative) received was not different between groups.