Opioid Analgesia in Mechanically Ventilated Children: Results from the multicenter MOTIF study

Abstract

Objective

To examine the clinical factors associated with increased opioid dose among mechanically ventilated children in the Pediatric Intensive Care Unit (PICU).

Design

Prospective, observational study with 100% accrual of eligible patients.

Setting

Seven PICUs from tertiary-care children’s hospitals in the Collaborative Pediatric Critical Care Research Network.

Patients

419 children treated with morphine or fentanyl infusions.

Interventions

None

Measurements and Main Results

Data on opioid use, concomitant therapy, demographic and explanatory variables were collected. Significant variability occurred in clinical practices, with up to 100-fold differences in baseline opioid doses, average daily or total doses, or peak infusion rates. Opioid exposure for 7 or 14 days required doubling of the daily opioid dose in 16% patients (95%CI: 12–19%) and 20% patients (95%CI: 16–24%) respectively. Among patients receiving opioids for longer than 3 days (n=225), this occurred in 28% (95%CI 22–33%) and 35% (95%CI 29–41%) by 7 or 14 days respectively. Doubling of the opioid dose was more likely to occur following opioid infusions for 7 days or longer (OR 7.9, 95%CI 4.3–14.3; p<0.001) or co-therapy with midazolam (OR 5.6, 95%CI 2.4–12.9; p<0.001), and it was less likely to occur if morphine was used as the primary opioid (vs. fentanyl) (OR 0.48, 95%CI 0.25–0.92; p=0.03), for patients receiving higher initial doses (OR 0.96, 95%CI 0.95–0.98; p<0.001), or if patients had prior PICU admissions (OR 0.37, 95%CI 0.15–0.89, p=0.03).

Conclusions

Mechanically ventilated children require increasing opioid doses, often associated with prolonged opioid exposure or the need for additional sedation. Efforts to reduce prolonged opioid exposure and clinical practice variation may prevent the complications of opioid therapy.

Children requiring mechanical ventilation and invasive monitoring routinely receive opioids for analgesia/sedation in the Pediatric Intensive Care Unit (PICU) (1, 2). Use of opioid analgesics and other sedatives helps to reduce pain, anxiety, or agitation, facilitate mechanical ventilation, prevent physiological stress responses, and avoid secondary complications (3–5). Opioid therapy may lead to opioid-induced hyperalgesia or opioid tolerance, dependence, and withdrawal (6). These effects occur more commonly in children than in adults, because of developmental changes in metabolism, excretion, drug efficacy, receptor subtypes, signal transduction, receptor induction, or cellular regulatory pathways (6–11).

Surveys of analgesia/sedation practices in PICUs demonstrated wide variability in clinical practices (1, 12). The use of several drug classes, multiple agents, large variations in the doses, frequency, and routes of administration, off-label use of analgesic drugs, or untested drug combinations occurs routinely, often driven by individual preferences or local culture (1, 6, 13). Consensus guidelines for sedation, analgesia, or neuromuscular blockade in PICU patients have highlighted the paucity of high-quality evidence and called for more randomized trials in this area (14, 15).

Given this variability, it is difficult to define best practices, develop guidelines, or launch scientific efforts to investigate the key hypotheses in this area. Randomized trials in this population await descriptive observational studies to determine associations between clinical practices and patient outcomes, needed for generating hypotheses that can be tested formally. The Collaborative Pediatric Critical Care Research Network (CPCCRN) designed this prospective, observational study to characterize the exposure to opioid analgesia among mechanically ventilated children in the PICU, in order to prepare for a randomized trial comparing alternative analgesic strategies.

Methods

Study Design

The MOTIF (Measuring Opioid Tolerance Induced by Fentanyl (or morphine)) Study was conducted in the seven PICUs participating in CPCCRN. Institutional Review Boards (IRBs) of all clinical centers and the Data Coordinating Center (DCC) approved the protocol. No interventions were performed and a waiver of informed consent was obtained in order to allow for the accrual of all eligible patients.

Participants

Patients were eligible for enrollment if they were: 1) over 37 weeks post-conceptual age and less than 18 years; 2) receiving ventilatory assistance via endotracheal tube or tracheostomy; and 3) receiving morphine or fentanyl infusions for analgesia. Patients were excluded if they: 1) had a history of drug abuse or alcohol dependence, 2) if the patient was less than 3 months old and the patient’s mother had a history of drug abuse during pregnancy, 3) were likely to be placed on extra-corporeal support soon after admission, 4) were receiving continuous drug infusions for epidural analgesia, nerve blocks, or plexus blocks, 5) were receiving opioid therapy via patient-controlled analgesia (PCA) pumps, 6) were receiving chronic opioid therapy before PICU admission, 7) were previously enrolled in this study, or 8) if parents were considering withdrawal of care or do-not-resuscitate orders.

Outcomes

Our primary outcome was defined as the need for doubling the initial opioid doses in order to achieve the same pharmacological effects as those seen at initiation of therapy. Patients met the primary outcome if the total daily dose required to maintain adequate analgesia was doubled as compared to opioid doses required in the first 24 hours of therapy. Any subjects who died, were discharged from the PICU, or exited the study prior to Day 14 (e.g., for initiation of ECMO) were included in the denominator for the primary endpoint. Secondary outcomes included the average daily opioid dose (μg/kg/day, in fentanyl equivalent doses, as defined below), peak opioid infusion rate (μg/kg/hour), total cumulative opioid dose (μg/kg) and the duration of opioid exposure (hours).

Data collection

We also collected data on demographic and clinical variables, history of prior opioid exposure and the use of sedation with benzodiazepines, to identify potential risk factors or mitigating factors for increased opioid dosage in children. Daily doses of opioids and benzodiazepines were calculated using the total infusion-based dosing as well as the pro re nata (PRN) or bolus doses required each day. Data were collected daily by research coordinators at the CPCCRN sites and entered into an electronic data system. Data collection began with initiation of opioid therapy in the PICU (Day 1) and continued until Day 14 unless one of the study exit criteria were met. Study exit criteria were defined as death, PICU discharge, discontinuation of opioid therapy, or initiation of ECMO. Since some agents are more lipophilic than others, we calculated the BMI (body mass index) and evaluated its impact using the BMI percentile for age. Morphine doses were converted to fentanyl equivalents using well-known opioid potency ratios (1:80) (6, 16), whereas midazolam and lorazepam doses were combined assuming a potency ratio of 1:1 (17, 18).

Statistical analyses

Demographic and clinical characteristics were summarized for all subjects and by site, using counts and percentages for categorical data, and the median and interquartile range (IQR: 25^th and 75^th percentiles) for continuous data. Differences across sites were evaluated using Chi-square or Fisher’s exact test for categorical variables and Kruskal-Wallis test for continuous variables.

Our primary outcome (defined as doubling of the daily opioid dose) after 7 and 14 days of opioid therapy was described using proportions and 95% confidence intervals (95%CI). Univariable associations of demographic and clinical factors with increased opioid use by 14 days were evaluated using the Chi-square test for categorical variables and the Wilcoxon rank-sum test for continuous variables. Multivariable logistic regression analyses were performed for all patients and separately for postoperative and non-postoperative patients. All models included a priori factors, such as age, sex, baseline opioid dose, and primary opioid used (morphine vs. fentanyl). Additional factors were selected for inclusion in the final model based on backward variable selection at a significance level of 0.05. Adjusted odds ratios and corresponding 95%CI were reported. A Kaplan-Meier “freedom from event” curve was generated for the outcome of time to doubling of the initial opioid dose. Patients not achieving this outcome were censored at the time of study exit.

Analyses were performed in SAS v9.2 (SAS Institute Inc., North Carolina), using a significance level of 0.05 for all analyses. In analyses examining the associations between patient characteristics and opioid use, or for comparisons between clinical sites we made no adjustment for multiple comparisons because these investigations are descriptive and hypothesis-generating in nature.

Results

Total 571 subjects were screened, 500 were enrolled, and 419 subjects were eligible for study analyses (Figure 1). Baseline characteristics of study subjects (N=419) are listed in Table 1. Patients had a median age of 16 months (IQR 4, 80), a Pediatric Risk of Mortality (PRISM III) score of 6 (IQR 3, 11), height of 75 cm (IQR 58, 112), weight of 10.0 kg (IQR 5.5, 21.3), and BMI of 16.6 (IQR 14.7, 18.9). Factors possibly associated with previous opioid exposure included a history of NICU admission (n=96, 24%), PICU admission (n=81, 20%) or surgical operation (n=161, 39%) prior to this PICU admission.

Figure 1.

CONSORT (Consolidated Standards of Reporting Trials) Diagram showing subjects screened, enrolled, and included in this analysis.

Table 1.

Patient Characteristics

Variable	n (%) Yes
Female	177 (42)
Male	242 (58)
Age groups:
Newborn (< 1 mo)	42 (10)
Infant (1 mo – < 1 yr)	148 (35)
Pre-school (1 yr – < 5 yrs)	109 (26)
Pre-adolescent (6 yrs – < 10 yrs)	44 (11)
Adolescent (10 yrs–< 18 yrs)	76 (18)
Ethnicity:
Hispanic or Latino	78 (19)
Not Hispanic or Latino	259 (62)
Unknown	82 (20)
Race:
Black or African American	95 (23)
White	236 (56)
Other/Unknown	88 (21)
Primary Diagnosis:
Cardiac	131 (31)
Neurological	14 (3)
Respiratory	131 (31)
Shock	32 (8)
Surgical/Trauma	54 (13)
Transplant	21 (5)
Other	36 (9)
Post Operative	210 (50)
Types of Surgery (N = 210):
Cardiac surgery	97 (46)
Neurosurgery	13 (6)
Orthopedic surgery	6 (3)
Transplant	21 (10)
Trauma	7 (3)
Other	66 (31)
Opioid therapy initiated with:
Fentanyl	267 (64)
Morphine	152 (36)

Therapeutic variables related to opioid and benzodiazepine use are listed in Table 2 and adjuvant therapies are listed in Table 3. Striking variability occurred in the opioids administered to individual patients, with more than 100-fold differences in the initial opioid doses (0–24 hrs), average daily or total cumulative doses, and peak infusion rates. Doubling of the opioid dose from the initial baseline occurred in 16% patients by Day 7 (95%CI 12–19%) and 20% patients by Day 14 (95%CI 16–24%). Days on which individual patients achieved the primary outcome are presented in Figure 2 and cumulative estimates are presented in Figure 3. Among patients receiving opioids for at least 96 hours (n=225), doubling of the opioid dose occurred in 28% by Day 7 (95%CI 22–33%) and 35% by Day 14 (95%CI 29–41%). Reasons for study exit prior to Day 14 included discontinuation of opioid therapy (73%), PICU discharge (8%), initiation of ECMO (2%), or death (2%). Opioid doses had doubled among 43% (95%CI 30–55%) of the 61 patients still receiving opioids on Day 14.

Table 2.

Opioid and Benzodiazepine Usein All Patients

Opioids, N=419	Minimum, Maximum	Median (Interquartile Range)
Baseline opioid dose (μg/kg/day)	1.2, 161.5	34.9 (20.9, 60.2)
Average daily dose (μg/kg/day)	1.1, 185.5	34.0 (15.4, 53.8)
Total cumulative dose (μg/kg)	3.4, 2431	179.7 (72.5, 436.1)
Peak infusion rate (μg/kg/hr)	0.1, 16.0	2.5 (1.0, 4.0)
Duration of opioid infusion (hrs)	5.3, 336.0	104.1 (57.7, 213.0)
Duration on study (hrs)	33.0, 336.0	133.2 (74.5, 233.5)
Interval between unit admission and opioid infusion (hrs)	0, 865.7	3.3 (1.1, 8.8)
	Rate	95% Confidence Interval
Doubling of opioid dose at 7 days	0.16 (66/419)	0.12, 0.19
Doubling of opioid dose at 14 days	0.20 (83/419)	0.16, 0.24
Benzodiazepines, N=387a	Minimum, Maximum	Median (Interquartile Range)
Days receiving benzodiazepines	1, 14	5 (3, 9)
Average daily dose (mg/kg)	0.02, 23.4	1.1 (0.2, 2.9)
Total dose (mg/kg)	0.02, 227	4.1 (0.7, 18.1)

IQR = interquartile range (25th, 75th percentile). Morphine doses were converted to fentanyl equivalents using well-definedopioid potency ratios (1:80).

^*387/419 (92%) of patients received midazolam or lorazepam during the study period.

Average daily dose is basedonly on the days when any benzodiazepines were received.

Table 3.

Adjuvant Sedatives and Analgesics Administered

Sedatives	n (%)	Analgesics	n (%)
Ketamine	73 (17)	Acetaminophen	317 (76)
Propofol	60 (14)	Dexmedetomidine	90 (21)
Chloral hydrate	62 (15)	Methadone	73 (17)
Pentobarbital	26 (6)	Ibuprofen	51 (12)
Phenobarbital	24 (6)	Oxycodone	32 (8)
Clonazepam	9 (2)	Hydromorphone	30 (7)
Diphenhydramine	9 (2)	Ketorolac	25 (6)
Diazepam	5 (1)	Acetylsalicylic acid	25 (6)
Hydroxyzine	5 (1)	Clonidine	8 (2)
Isoflurane	4 (1)	Meperidine	7 (2)
Sevoflurane	3 (1)	Nalbuphine	3 (1)
Haloperidol	3 (1)	Hydrocodone	2 (0)
Promethazine	3 (1)	Remifentanil	1 (0)
Thiopental	2 (0)	Codeine	1 (0)
Etomidate	2 (0)
Droperidol	1 (0)

Figure 2.

Number of patients who required doubling of their opioid doses after initiation of opioid therapy.

Figure 3.

Kaplan-Meier “freedom from event” curve for patients meeting the criterion for the primary outcome (time to doubling of the initial opioid dose). Patients not achieving this outcome were censored at the time of study exit.

Clinical and demographic factors included in univariable and multivariable analyses are listed in Table 4. Age (p=0.50), sex (p=0.82), race (p=0.08), ethnicity (p=0.13), BMI (p=0.33), severity of illness (PRISM III scores, p=0.47), previous NICU admission (p=0.54) or previous surgical operations (p=0.81) were not associated with this outcome. Doubling of the initial opioid dose occurred less frequently in patients admitted just after surgery (p<0.001) and those with previous PICU admission(s) (p=0.01). Subjects receiving lower opioid doses at baseline (p=0.003) and those requiring surgery while on the study (p=0.001) had greater likelihood of doubling their daily opioid doses. The median baseline (0–24 hr) fentanyl doses were 29 μg/kg/day (IQR 20–46) vs. 41 μg/kg/day (IQR 21–66) respectively for those who did or did not require doubling of their initial opioid dose. Children who required doubling of the opioid dose were more likely to receive additional opioids (other than morphine or fentanyl; 49% vs. 27%, p<0.001) and benzodiazepines (99% vs. 91%, p=0.02), require multiple adjuvant therapies (median 5 (IQR 3–7) vs. 3 (IQR 2–4), p<0.001) and to receive methadone therapy (39% vs. 12%, p<0.001).

Table 4.

Factors associated with Doubling of the DailyOpioidDose

Variable	n (%)	p
Sex:		0.82
Female	36 (20)
Male	47 (19)
Age group:		0.50
Newborn (< 1 mo)	8 (19)
Infant (1 mo – < 1 yr)	23 (16)
Pre-school (1 yr – < 5 yrs)	24 (22)
Pre-adolescent (6 yrs – < 10 yrs)	9 (20)
Adolescent (10 yrs – < 18 yrs)	19 (25)
Postoperative Status:		< 0.001
Postoperative	26 (12)
Not postoperative	57 (27)
Surgery Required While On Study:		0.001
Yes	31 (31)
No	52 (16)
NICU Admission(s) in Previous Year:		0.54
Yes	17 (18)
No	63 (21)
PICU Admission(s) in Previous Year:		0.01
Yes	8 (10)
No	73 (23)
Surgical Operation(s) in Previous Year:		0.81
Yes	31 (19)
No	52 (20)
Primary Opioid Used:		0.30
Fentanyl	47 (18)
Morphine	36 (22)

Multivariable logistic regression analyses are presented in Table 5. Factors associated with our primary outcome included opioid infusions for 7 days or longer (OR=7.9) and concomitant infusions of midazolam (OR=5.6). This risk was mitigated in patients receiving primarily morphine vs. fentanyl (OR=0.48), those given higher initial opioid doses (OR=0.96), or those with prior PICU admissions (OR=0.37). Logistic regression analyses limited to postoperative patients identified similar factors, but females were more likely than males (OR=2.8) to require doubling of their initial opioid dose, as were those receiving co-therapy with midazolam (OR=6.3) or lorazepam (OR=5.2). Using morphine primarily appeared to reduce the risk of doubling opioid doses in the surgical patients (OR=0.27), but not in the medical or non-postoperative patients (OR=0.60).

Table 5.

Factors associated with Doubling of the Daily Opioid Dose

Variables	All Patients (N=419) Adjusted OR (95% Confidence Intervals)	P-values	Postoperative (N=210) Adjusted OR (95% Confidence Intervals)	P-values	Medical (N = 209) Adjusted OR (95% Confidence Intervals)	P-values
Age groups:		0.50		0.53		0.90
Infant (< 1 yr)	0.81 (0.37, 1.79)		0.47 (0.12, 1.79)		0.95 (0.37, 2.50)
Child (1 yr – < 10 yrs)	1.21 (0.55, 2.65)		0.69 (0.17, 2.81)		1.15 (0.45, 2.95)
Adolescent (10 yrs–< 18 yrs)	Reference		Reference		Reference
Female (vs. male)	1.10 (0.61, 1.96)	0.75	2.79 (0.99, 7.87)	0.052	0.68 (0.33, 1.40)	0.29
Baseline opioid dose (1 unit mcg/kg)	0.96 (0.95, 0.98)	<0.001	0.96 (0.94, 0.98)	0.001	0.97 (0.95, 0.99)	0.004
Primary opioid morphine (vs. fentanyl)	0.48 (0.25, 0.92)	0.03	0.27 (0.08, 0.88)	0.03	0.60 (0.27, 1.33)	0.21
Midazolam used	5.57 (2.41, 12.9)	<0.001	6.31 (1.73, 23.0)	0.005	6.91 (2.21, 21.7)	<0.001
Lorazepam used	N.S.	--	5.21 (1.49, 18.2)	0.01	N.S.	--
Opioid infusion ≥ 7 days	7.85 (4.32, 14.3)	<0.001	5.86 (2.10, 16.3)	<0.001	7.06 (3.33, 15.0)	<0.001
History of previous PICU admission	0.37 (0.15, 0.89)	0.03	N.S.	--	N.S.	--

Note that age group, sex, baseline opioid dose, and primary opioid used were included a priori in all models. Other variables were selected based on backward variable selection with a significance level of 0.05. The variable selection procedure was performed separatelyfor each of the three analyses above. The adolescent age group was used as a reference group because it is closest to adult patients. OR = odds ratio, NS = not selected in the logistic model because p>0.05.

Opioid and benzodiazepine usage by clinical site is shown in Table 6. Significant differences occurred in the baseline opioid doses used, the average daily doses, the cumulative total doses, peak opioid infusion rates, and duration of opioid infusions between the 7 clinical sites (p<0.001). Four sites used fentanyl as the primary opioid, two sites preferred morphine, whereas one site used both drugs equally. Variations in median daily opioid doses for all clinical sites are displayed in Figure 4. Site E enrolled a relatively high percentage of cardiac surgical patients and used high initial opioid doses (fentanyl at 3–5 mcg/kg/hr). No patients at this site required doubling of the opioid dose during this study. Site F was the only institution that followed an explicit sedation protocol at the time of this study. Minimal variation occurred in their median daily doses and very few patients at this site (3%) required doubling of their initial opioid dose. Benzodiazepines were used for sedation for most patients at each clinical site (83–98%), although the drugs used, the average daily benzodiazepine doses, and the total cumulative doses differed significantly between the clinical sites (p<0.001).

Table 6.

Variability of Opioid and Benzodiazepine Use by Clinical Site, N=419

Variable	A	B	C	D	E	F	G
Opioids	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	p
Baseline opioid dose (μg/kg)	48 (29, 58)	54 (35, 87)	41 (29, 59)	24 (15, 30)	77 (57, 97)	14 (9, 19)	37 (28, 70)	< .001
Average daily dose (μg/kg)	42 (30, 54)	49 (32, 76)	37 (27, 65)	26 (15, 41)	44 (20, 68)	9 (6, 13)	46 (14, 70)	< .001
Total cumulative dose (μg/kg)	281 (124, 531)	320 (129, 682)	227 (102, 458)	169 (64, 463)	202 (102, 394)	39 (18, 100)	291 (71, 634)	< .001
Peak infusion rate (μg/kg/hr)	3.0 (2.0, 4.0)	4.0 (3.0, 6.0)	2.5 (2.0, 4.8)	1.5 (1.0, 2.5)	4.0 (3.0, 5.0)	0.4 (0.3, 0.6)	3.0 (2.0, 5.0)	< .001
Duration of opioid infusion (hrs)	132 (77, 240)	134 (69, 254)	115 (65, 198)	161 (80, 287)	60 (34, 99)	75 (40, 158)	100 (45, 204)	< .001
Duration on study (hrs)	169 (82, 275)	148 (75, 260)	136 (83, 218)	178 (81, 298)	123 (77, 174)	83 (53, 167)	136 (60, 233)	.009
	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)
Primary Opioid Used
Fentanyl	57 (95)	54 (98)	42 (52)	24 (30)	46 (88)	5 (8)	30 (100)	< .001
Morphine	3 (5)	1 (2)	39 (48)	55 (70)	6 (12)	57 (92)	0 (0)
Primary outcome at 7 days	12 (20)	6 (11)	20 (25)	18 (23)	0 (0)	2 (3)	8 (27)	<.001
Primary outcome at 14 days	19 (32)	10 (18)	22 (27)	21 (27)	0 (0)	2 (3)	9 (30)	< .001
Benzodiazepines	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)
Any benzodiazepine used	56 (93)	49 (89)	67 (83)	77 (97)	51 (98)	58 (94)	29 (97)	.01
Midazolam used	36 (60)	21 (38)	66 (81)	76 (96)	51 (98)	18 (29)	26 (87)	< .001
Lorazepam used	47 (78)	44 (80)	24 (30)	34 (43)	11 (21)	55 (89)	20 (67)	< .001
	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)
Days receiving benzodiazepines	4 (3, 8)	4 (3, 7)	6 (3, 10)	8 (4, 13)	5 (4, 8)	3 (2,6)	5 (3, 10)	< .001
Average daily dose (mg/kg)*	0.2 (0.1, 0.3)	0.2 (0.1, 0.4)	2.4 (1.1, 4.1)	2.1 (1.4, 3.6)	2.9 (1.7, 4.2)	0.2 (0.1, 0.4)	1.5 (0.8, 3.2)	< .001
Total cumulative dose (mg/kg)	0.6 (0.3, 2.5)	0.9 (0.4, 2.5)	11.9 (3.0, 28.7)	15.7 (6.0, 31.0)	13.7 (5.4, 29.6)	0.6 (0.3, 1.9)	5.4 (2.6, 20.2)	< .001

IQR = interquartile range (25th, 75th percentile); Primary Outcome was defined as doubling of the initially effectiveopioid dose (0–24 hours of therapy). All totals included infusion and bolus doses; morphine doses were converted to fentanyl equivalents using well-defined opioid potency ratios (1:80).

^*Average daily benzodiazepine dose is based only on the days when any benzodiazepines were received.

Figure 4.

Patterns of opioid dosing at the participating sites. Each line represents a single clinical site, showing the median daily opioid dose used at that site each day during the study. Morphine doses were converted to fentanyl equivalents using well-defined opioid potency ratios (1:80).

Discussion

In this study, we found that 16% of all studied patients required doubling of their opioid dose by 7 days, and 20% by 14 days. This outcome was more likely among patients exposed to opioid infusions for 7 days or longer, those receiving midazolam infusions, and among non-postoperative patients. Among postoperative patients, doubling of the opioid dose occurred more frequently in patients receiving opioid infusions for 7 days or longer, and those receiving fentanyl (vs. morphine), midazolam, or lorazepam infusions. We noted a striking variability in the opioid and benzodiazepine use between CPCCRN sites. Finally, we enrolled 500 subjects who met the inclusion and exclusion criteria, but 81 (16%) of these patients required opioid management for less than 24 hours. In the design of future randomized trials, this loss of evaluable subjects must be taken into account when planning accrual and power calculations.

Prolonged opioid exposure may lead to opioid tolerance and withdrawal in children requiring intensive care (19–22); therefore we wanted to identify the patients at risk for these complications. While tolerance may occur in any patient, those exposed to opioids for longer than 3 days and those who require a doubling of their initially effective opioid doses are likely to be at highest risk for these complications (6). This doubling occurred in up to 20% of all ventilated patients receiving opioids for 24 hours, 35% of those receiving opioids for 96 hours or longer, and 43% of those receiving opioids for 14 days or more. These data are consistent with previously reported rates of opioid tolerance in 36–57% of PICU patients (19, 23–28) and were further corroborated in a systematic review (29).

Critically ill children require frequent adjustments of their analgesia/sedation, in response to rapidly changing clinical parameters, worsening illness, necessary invasive procedures, or surgical operations and other indications, but also because of poorly defined goals of sedation, subjective assessments of pain or discomfort, varying expectations of different nurses or physicians, and deeply entrenched practice patterns. These challenges can be addressed in formal analgesia/anxiolysis titration protocols designed for both rapid and effective pain and anxiety relief, but also proactive weaning of these drugs if the patient remains comfortable (30). Research in this area should receive high priority, because iatrogenic injury, with major complications (31, 32) and prolonged hospital stays (27, 33) may occur among children with prolonged opioid exposure (23–25, 34, 35).

In mechanically ventilated children, the effectiveness of opioid analgesia may vary because of worsening pain, or opioid-induced hyperalgesia, or opioid tolerance (6). Currently, there are no clinical, physiological, or biochemical markers to identify patients developing opioid tolerance. In animal models, opioid tolerance is assessed by a need for increased doses and the signs of opioid withdrawal precipitated by giving opioid antagonists (36, 37). This approach is not clinically feasible; therefore, future research must focus on a search for reliable biomarkers that can allow us to investigate the real-time relationships between opioid analgesia, hyperalgesia, tolerance, and withdrawal. We explored a pragmatic definition (doubling of the total initial opioid dose, assuming that effective analgesia was achieved with the initial and subsequent opioid doses) to identify a high-risk group for developing opioid tolerance.

We found that prolonged opioid exposure was associated with doubling of the opioid dose, also noted previously (19, 38). Patients receiving lower initial doses were more likely to develop opioid tolerance compared to those receiving high doses. Inadequate analgesia associated with low initial dosing leads to ongoing pain, hyperalgesia and windup (39), which require much higher opioid doses to finally regain pain control. Benzodiazepines are often added for sedation, but their impact on opioid effectiveness remains unclear. In young animals, midazolam may increase analgesic requirements because it potentiates nociceptive behaviors, sensitizes cutaneous reflexes, and appears to have no sedative effects (40). Other animal studies suggest that midazolam promotes acute opioid tolerance (41). There are no data to suggest similar effects in humans, but our findings support the need to perform similar studies in children.

Univariable analyses showed no differences between morphine and fentanyl, but the logistic regression model showed that primarily using morphine reduced the need for escalating opioid doses, particularly among postoperative patients. This is consistent with the mechanisms of opioid tolerance (6, 20), was noted previously in neonates requiring ECMO (33) but has not been reported in PICU patients. Females receiving postoperative analgesia appeared to have greater odds for doubling of their opioid dose than males, but this effect was barely significant. Gender differences in postoperative pain, opioid analgesia and tolerance are well-known from animal and adult human studies (42–44), but have not been reported in children. Further studies are needed to explore this association.

We confirmed a significant degree of variability in the clinical practices used for opioid analgesia/sedation. Guidelines to reduce practice variability are associated with improved outcomes (45–47). The American College of Critical Care Medicine developed guidelines for analgesia/sedation in critically ill adults in 1995, updated in 2002 (48, 49), but similar guidelines are not available for children. Use of a sedation protocol at one site was associated with reduced variation in median daily opioid doses (Figure 4) and fewer patients at this site required doubling of their initial opioid doses. Practice variation in the use of opioid analgesia may lead to a higher incidence of complications such as oversedation, respiratory depression, hypotension, and opioid withdrawal. Conducting a prospective observational study to define clinical outcomes and identify the populations at risk for complications is important to prepare for clinical trials of opioid analgesia in children (50).

This observational study has several limitations. To achieve 100% accrual of eligible patients, we did not assess the confounding effects of adjuvant drugs on opioid analgesia in this study. Some drugs alter the mechanisms of opioid analgesia and can delay tolerance (e.g. ketamine, clonidine, methadone), although most drug interactions are not well-characterized in children. We did find that children developing tolerance required multiple adjuvant therapies (median 5 vs. 3, p<0.001), perhaps identifying a “difficult to sedate” phenotype among these patients. Another limitation is that we truncated the data collection at 14 days of opioid therapy, thus missing the primary outcome in some patients. However, we assumed that patients likely to require a doubling of their opioid dose would have met this primary outcome by 14 days (27). Another limitation is that these data were obtained from PICUs at tertiary-care children’s hospitals, but the vast majority of PICUs are not located at academic medical centers. We feel that the latter two limitations may underestimate the true incidence of opioid tolerance, which may be greater than 20% in non-academic centers or if data collection was extended beyond 14 days.

Despite these limitations, the present study is timely and relevant (6, 14), as it brings attention to clinical practices associated with unacceptably high risks of iatrogenic injury (1, 2, 5, 6, 13, 19, 23, 29, 31–33), prolonged ICU stays (27, 33) and increased healthcare costs (51–53). Despite the paucity of scientific evidence related to specific drugs or approaches (2, 6, 12–14), developing consensus guidelines for analgesia/sedation in infants and children and defining their daily therapeutic goals based on objective clinical criteria, may help to focus the expectations of bedside clinicians, bring greater consistency to their practice patterns, and thereby prevent some of the complications associated with opioid therapy.