ABSTRACT
Purpose
Opioid use in pediatric oncology is poorly characterized. The primary objective was to describe opioid administration or prescription patterns in pediatric oncology patients.
Methods
In this multi‐center retrospective study conducted across three sites, we included patients aged 0–19 years with a malignant neoplastic disease who received their first chemotherapy between January 2019 and March 2024. Data were extracted from each site's Observational Medical Outcomes Partnership Common Data Model. We described patient characteristics based on: (1) first opioid administration or prescription, overall and for each specific opioid; and (2) all opioid administrations or prescriptions, overall and for each specific opioid. We also evaluated variability in opioid and naloxone use.
Results
Of 2903 pediatric patients with cancer, 2707 (93.2%) were administered or prescribed at least one opioid. Among these patients, fentanyl was the most common (92.4%), followed by morphine (68.7%), hydromorphone (55.3%), and oxycodone (38.6%). Across centers, the proportion of patients receiving fentanyl ranged from 90.9% to 94.9%, morphine from 34.0% to 89.2%, oxycodone from 0.9% to 76.1%, and naloxone from 2.4% to 29.5%. Overall, 10.9% were administered or prescribed naloxone, but among 169 methadone patients, 43.5% were administered or prescribed it. Of the 153,849 total opioid administrations or prescriptions, 22.4% occurred more than 12 months after initiating chemotherapy.
Conclusions
Among pediatric patients with cancer who receive chemotherapy, 93.2% were administered or prescribed at least one opioid; fentanyl was the most common. There was heterogeneity in opioid choice and naloxone use across sites. These findings lay preliminary groundwork for understanding opioid use patterns and identifying potential areas for practice improvement in pediatric oncology.
Keywords: OMOP common data model, oncology, opioid, pediatric
1. Background
Pain is one of the most common and distressing symptoms experienced by pediatric patients with cancer [1, 2, 3]. Pain may arise from procedures required for cancer diagnosis or treatment, such as central venous line insertion, as well as from the cancer itself. It may also occur as a side effect of medications, including dinutuximab and filgrastim, or from treatment‐related complications such as mucositis and typhlitis. Management of pain typically involves both pharmacological and non‐pharmacological approaches, with opioid administration representing an important pharmacological strategy [4].
Little multi‐center information is available regarding opioid administration in pediatric cancer patients. A systematic review of adolescents and young adults identified 11 studies reporting opioid exposure in 12%–97% of patients [5]. While opioid administration remains a cornerstone of effective pain management, there is growing emphasis on opioid‐sparing approaches to minimize adverse effects and reduce the potential for misuse [6]. Understanding current practices and variation across tertiary care centers is essential to guide system‐level interventions related to opioid administration and prescribing in this population.
To better understand the patterns of opioid administration or prescription, leveraging electronic health record (EHR) data provides an efficient means of data collection. Multi‐center evaluation is particularly valuable for improving the generalizability of findings and for identifying variation in practice. However, challenges such as data access, sharing, and standardization can limit multi‐center EHR studies. The Observational Medical Outcomes Partnership Common Data Model (OMOP CDM) facilitates systematic analysis across multiple centers by standardizing both the structure and vocabulary of data [7].
We therefore conducted this study across three tertiary care centers that had implemented the OMOP CDM. Characterizing contemporary opioid prescribing and administration practices in pediatric oncology is an important first step toward understanding practice variation, establishing baseline patterns of care, and informing future studies and quality improvement initiatives aimed at optimizing pain management. The primary objective was to describe opioid administration or prescription patterns in pediatric oncology patients.
2. Methods
The study was approved by the Research Ethics Board or Institutional Review Board at each participating site. Given the retrospective design, use of de‐identified data, and sharing of only summary (not individual level) data, the requirement for informed consent and assent was waived. The OMOP standard concepts used in this study are shown as Appendix 1.
2.1. Sites
The three sites were The Hospital for Sick Children (SickKids) in Toronto, Canada; Lucile Packard Children's Hospital (LPCH) in Palo Alto, United States; and Columbia University in New York, United States.
2.2. Participants
We included pediatric patients with cancer, defined as those with at least one condition within the ancestor concept “malignant neoplastic disease” (concept ID 443392). To distinguish patients receiving active treatment from those in survivorship, we additionally required receipt of at least one chemotherapy medication, defined as a drug within the ancestor concept “antineoplastic agents” (concept ID 21601387). Participants were required to be 19 years or younger at the time of their first chemotherapy administration or prescription, with the first chemotherapy between January 1, 2019, and March 1, 2024. Palliative patients were included. We included all available data from January 1, 2019, through to September 1, 2024 at each participating site.
2.3. Data Source
The data source for this study was the OMOP CDM as locally instantiated at each participating site, with each site using its own implementation approach. As an example, we have previously described the steps used to create the SickKids' implementation, which was based on a curated and validated repository derived from Epic named the SickKids Enterprise‐wide Data in Azure Repository (SEDAR) [8].
The OMOP CDM standardizes the structure of observational data and includes tables such as person, visit_occurrence, procedure_occurrence, drug_exposure, device_exposure, condition_occurrence, measurement, and observation. The visit_occurrence table represents individual healthcare encounters. The procedure_occurrence table captures activities or processes performed for diagnostic or therapeutic purposes. The condition_occurrence table includes diagnoses recorded during encounters, admissions, and discharges, as well as those added to problem lists or during referrals. The drug_exposure table captures both medication administrations, which primarily occur during inpatient encounters, and prescriptions, which are typically filled at external pharmacies.
2.4. Outcomes
The primary outcome was any opioid administration or prescription, defined as a medication within the ancestor concepts “opioids” (concept ID 21604254) and “opioid anesthetics” (concept ID 21604200). We then stratified opioids by the most common types including fentanyl (concept ID 1154029), morphine (concept ID 1110410), hydromorphone (concept ID 1126658), oxycodone (concept ID 1124957), remifentanil (concept ID 19016749), and methadone (concept ID 1103640). Exact counts of inpatient opioid administrations were used. Outpatient opioid prescriptions were recorded as a single prescription event regardless of whether the medication was never taken (e.g., prescribed “as needed”) or was taken multiple times over multiple days, as information on actual outpatient medication use is generally unavailable within the EHR.
2.5. Procedure
Baseline characteristics of the entire cancer cohort, of patients with any opioid administration or prescription, and of those prescribed the most common types of opioids were summarized by age (≥ 10 years at cancer diagnosis and at first chemotherapy), sex, race, ethnicity, and cancer type (leukemia, lymphoma, central nervous system tumor or solid tumor).
We then described patient characteristics based on: (1) first opioid administration or prescription, overall and for each specific opioid; and (2) all opioid administrations or prescriptions, overall and for each specific opioid. These characteristics included order type (administration or prescription), age (< 1, 1–10, and > 10 years), cancer type, and timing relative to first chemotherapy (prior to chemotherapy start (> 6 months, 6 to < 3 and 3 to < 0 months) and after chemotherapy start (0 to < 3, 3 to < 6, 6 to < 12 and > 12 months)). We also described the temporal relationship with surgery, procedure, lumbar puncture, bone marrow aspiration and propofol administration. Additionally, the route of administration and year of exposure were summarized. The number of days opioids were administered or prescribed per patient was then determined.
To evaluate factors associated with opioid administration or prescription, we initially planned to use any exposure to morphine or fentanyl as the outcome. However, nearly all patients received these opioids. Therefore, we defined the outcome as administration or prescription exceeding the median number of days among patients at SickKids. This approach was taken to reduce the number of times each institution had to execute their local script. Potential risk factors examined included age ≥ 10 years at first chemotherapy, female sex, White race, Hispanic ethnicity, and leukemia diagnosis.
2.6. Analysis
Each institution retained control of its patient‐level data. We developed a common analytic script that was distributed to each site and executed locally. Each institution independently ran the same script against its own OMOP database and generated only aggregate summary statistics, which were then shared with the coordinating center for combined analysis. The site‐specific summary statistics that were shared included mean and standard deviation for continuous variables and proportions for categorical variables. These results were synthesized with the meta package in R under a random‐effects model to address variability in patients, practices and OMOP instantiations between sites. To identify factors associated with greater morphine or fentanyl use, we conducted random effects meta‐analysis using inverse variance weighting. Effects were presented as the risk ratio (RR) and 95% confidence intervals (CI). Analysis was conducted in R version 4.4.0.
3. Results
Table 1 presents the baseline characteristics of the study cohort. A total of 2903 pediatric patients with cancer were included across the three sites, of whom 2707 (93.2%) were administered or prescribed at least one opioid. Table 1 and Figure 1a show that among these patients, fentanyl was the most commonly administered or prescribed opioid (92.4%), followed by morphine (68.7%), hydromorphone (55.3%), and oxycodone (38.6%). Appendix 2 illustrates that 90.3% of patients received opioids for two or more days, and that 31.2% were administered or prescribed opioids on 15 or more days. Overall, 10.9% were administered or prescribed naloxone, but it was more common among 1045 oxycodone recipients (24.2%) and 169 methadone recipients (43.5%).
TABLE 1.
Demographic characteristics stratified by any opioid administration or prescription and by opioid type a .
| Characteristic, n (%) | All cancer patients | All opioids | Fentanyl | Morphine | Hydro‐morphone | Oxycodone | Remifentanil | Methadone | Other b |
|---|---|---|---|---|---|---|---|---|---|
| Total number patients | 2903 | 2707 | 2501 | 1861 | 1496 | 1045 | 795 | 169 | 284 |
| Age at cancer diagnosis ≥ 10 years | 1369; 48.1% (95% CI 36.0–60.4; I 2 = 0.9) | 1248; 47.1% (95% CI 34.2–60.5; I 2 = 0.9) | 1133; 46.4% (95% CI 33.0–60.4; I 2 = 0.9) | 797; 44.7% (95% CI 32.6–57.4; I 2 = 0.8) | 858; 60.9% (95% CI 33.6–82.8; I 2 = 0.9) | 585; 55.6% (95% CI 38.5–71.5; I 2 = 0.7) | 358; 47.5% (95% CI 31.0–64.5; I 2 = 0.8) | 106; 63.2% (95% CI 40.0–81.6; I 2 = 0.4) | 156; 54.7% |
| Age at first chemotherapy ≥ 10 years | 1410; 49.4% (95% CI 37.9–61.1; I 2 = 0.9) | 1286; 48.5% (95% CI 35.9–61.2; I 2 = 0.9) | 1170; 47.8% (95% CI 34.8–61.2; I 2 = 0.9) | 825; 46.1% (95% CI 33.8–58.8; I 2 = 0.8) | 884; 62.7% (95% CI 35.2–83.8; I 2 = 0.9) | 601; 57.2% (95% CI 39.5–73.2; I 2 = 0.7) | 378; 49.9% (95% CI 33.0–66.8; I 2 = 0.8) | 111; 67.3% (95% CI 35.3–88.5; I 2 = 0.6) | 163; 57.2% |
| Female sex | 1282; 44.2% (95% CI 42.3–46.1; I 2 = 0) | 1198; 44.3% (95% CI 41.5–47.1; I 2 = 0) | 1084; 43.3% (95% CI 41.3–45.5; I 2 = 0) | 864; 46.4% (95% CI 44.0–48.9; I 2 = 0) | 663; 44.3% (95% CI 38.4–50.3; I 2 = 0.1) | 456; 43.6% (95% CI 40.3–47.0; I 2 = 0) | 352; 44.3% (95% CI 37.7–51.1; I 2 = 0) | 78; 46.2% (95% CI 34.3–58.5; I 2 = 0) | 133; 46.8% |
| White race c | 592; 39.3% (95% CI 24.2–56.8; I 2 = 0.1) | 537; 38.8% (95% CI 24.8–55.0; I 2 = 0) | 496; 38.5% (95% CI 19.1–62.4; I 2 = 0.4) | 271; 39.8% (95% CI 33.4–46.5; I 2 = 0) | 329; 37.6% (95% CI 25.0–52.2; I 2 = 0) | 390; 37.8% (95% CI 24.3–53.4; I 2 = 0) | 136; 44.2% (95% CI 25.6–64.5; I 2 = 0) | 59; 38.9% (95% CI 3.9–90.8; I 2 = 0.4) | 85; 40.3% |
| Hispanic/Latino c | 570; 37.0% (95% CI 11.0–73.7; I 2 = 0.8) | 535; 38.0% (95% CI 13.0–71.6; I 2 = 0.7) | 518; 39.2% (95% CI 12.1–75.0; I 2 = 0.7) | 256; 37.7% (95% CI 14.2–68.8; I 2 = 0.4) | 358; 41.0% (95% CI 31.9–50.7; I 2 = 0) | 402; 38.9% (95% CI 21.0–60.4; I 2 = 0.1) | 92; 29.2% (95% CI 3.9–80.8; I 2 = 0.4) | 60; 39.5% (95% CI 4.9–89.1; I 2 = 0.3) | 85; 40.5% |
| Cancer type | |||||||||
| Leukemia | 1124; 37.8% (95% CI 27.3–49.5; I 2 = 0.8) | 1060; 38.2% (95% CI 27.9–49.7; I 2 = 0.8) | 999; 39.0% (95% CI 28.6–50.5; I 2 = 0.8) | 713; 38.3% (95% CI 33.0–43.9; I 2 = 0.2) | 510; 33.2% (95% CI 22.1–46.7; I 2 = 0.8) | 386; 36.2% (95% CI 26.8–46.7; I 2 = 0.4) | 223; 26.5% (95% CI 16.3–40.2; I 2 = 0.7) | 44; 26.5% (95% CI 14.6–43.1; I 2 = 0) | 107; 36.6% |
| Lymphoma | 419; 14.4% (95% CI 13.9–15.0; I 2 = 0) | 382; 14.1% (95% CI 12.5–16.0; I 2 = 0) | 341; 13.7% (95% CI 10.6–17.5; I 2 = 0.2) | 248; 13.3% (95% CI 12.2–14.5; I 2 = 0) | 208; 14.4% (95% CI 8.7–23.1; I 2 = 0.6) | 152; 14.6% (95% CI 12.6–16.8; I 2 = 0) | 101; 12.6% (95% CI 8.1–18.9; I 2 = 0) | 12; 6.6% (95% CI 0.8–38.2; I 2 = 0.5) | 36; 12.9% |
| CNS | 447; 15.7% (95% CI 10.5–22.8; I 2 = 0.7) | 389; 14.6% (95% CI 9.9–21.0; I 2 = 0.6) | 360; 14.6% (95% CI 10.1–20.7; I 2 = 0.6) | 273; 14.4% (95% CI 8.3–23.9; I 2 = 0.6) | 216; 14.5% (95% CI 10.4–19.8; I 2 = 0.3) | 141; 13.7% (95% CI 9.1–20.1; I 2 = 0.2) | 236; 32.5% (95% CI 16.4–54.3; I 2 = 0.9) | 31; 16.4% (95% CI 1.7–68.9; I 2 = 0.7) | 28; 8.0% |
| Solid tumor | 911; 31.4% (95% CI 28.1–34.9; I 2 = 0) | 874; 32.3% (95% CI 29.4–35.3; I 2 = 0) | 799; 32.0% (95% CI 30.0–33.9; I 2 = 0) | 625; 33.6% (95% CI 30.8–36.5; I 2 = 0) | 560; 37.4% (95% CI 31.0–44.1; I 2 = 0.3) | 365; 34.9% (95% CI 27.7–42.9; I 2 = 0.3) | 234; 29.5% (95% CI 25.4–33.8; I 2 = 0) | 82; 56.2% (95% CI 8.1–94.9; I 2 = 0.9) | 112; 39.5% |
| Unknown | 2; 0.1% | 2; 0.1% | 2; 0.1% | 2; 0.2% | 2; 0.3% | 1; 0.9% | 1; 0.3% | 0 | 1; 0.9% |
Abbreviations: CI, confidence interval; CNS, central nervous system.
Percentages may not add up to 100% because each stratum was estimated independently using a random‐effects model. The values represent the number, percentage, 95% CI and I 2.
The same patient can receive multiple other opioids and thus, individual other opioids will not sum to the total. The most common other opioids were meperidine (n = 110), sufentanil (n = 77), nalbuphine (n = 72), acetaminophen‐opioid combinations (n = 22) and tramadol (n = 17).
Data available from two sites.
FIGURE 1.
Distribution of opioid administrations or prescriptions. (a) The percentage of patients receiving each opioid type (N = 2707 patients); (b) The distribution of each opioid administration or prescription by type (N = 153,849 administrations or prescriptions).
Table 2 summarizes characteristics based on the first time any opioid was administered or prescribed, and the first time each opioid type was administered or prescribed among 2707 patients. The first opioid was more commonly an administration (85.6%) rather than a prescription (14.4%). First opioid exposure frequently occurred on the same day as surgery: 46.5% for any opioid, 57.4% for first fentanyl and 79.7% for first remifentanil.
TABLE 2.
Characteristics of first opioid administration or prescription and first time each opioid type was administered or prescribed (N = 2707 patients) d .
| Characteristic, n (%) | All opioids | Fentanyl | Morphine | Hydro‐morphone | Oxycodone | Remifentanil | Methadone | Other |
|---|---|---|---|---|---|---|---|---|
| Total number patients | 2707 | 2501 | 1861 | 1496 | 1045 | 795 | 169 | 284 |
| Opioid order type | ||||||||
| Administration | 2321 (85.6%) | 2255 (98.8%) | 1428 (76.8%) | 1350 (93.1%) | 737 (67.9%) | 728 (97.8%) | 122 (72.1%) | 213 (79.7%) |
| Prescription | 386 (14.4%) | 246 (1.2%) | 433 (23.2%) | 146 (6.9%) | 308 (32.1%) | 67 (2.2%) | 47 (27.9%) | 71 (20.3%) |
| Age at exposure | ||||||||
| < 1 year | 211 (7.7%) | 190 (7.6%) | 163 (8.8%) | 53 (3.6%) | 47 (4.6%) | 27 (3.7%) | 12 (7.7%) | 13 (4.7%) |
| 1–10 years | 1319 (47.7%) | 1233 (48.1%) | 928 (48.1%) | 596 (36.3%) | 430 (41.4%) | 418 (50.7%) | 47 (28.1%) | 113 (39.8%) |
| > 10 years | 1177 (44.5%) | 1078 (44.2%) | 770 (43.3%) | 847 (60.4%) | 568 (54%) | 350 (45.5%) | 110 (65.6%) | 158 (55.4%) |
| Cancer type a | ||||||||
| Leukemia | 957 (34.0%) | 921 (35.5%) | 672 (35.5%) | 474 (31.0%) | 328 (30.5%) | 215 (24.9%) | 39 (23.4%) | 99 (33.9%) |
| Lymphoma | 280 (10.6%) | 257 (10.6%) | 194 (10.9%) | 182 (12.5%) | 151 (14.1%) | 82 (10.4%) | 16 (10.2%) | 27 (9.1%) |
| CNS | 323 (12.5%) | 304 (12.6%) | 222 (12.6%) | 185 (12.4%) | 137 (13.2%) | 193 (28%) | 26 (15.9%) | 25 (8.2%) |
| Solid tumor | 669 (25.5%) | 632 (26.1%) | 505 (28.7%) | 487 (32.6%) | 322 (31.1%) | 201 (25.2%) | 76 (56.2%) | 102 (35.9%) |
| Unknown | 1 (0.1%) | 3 (0.2%) | 5 (0.3%) | 3 (0.3%) | 2 (0.4%) | 1 (0.3%) | 1 (1.3%) | 4 (1.5%) |
| Prior to cancer | 477 (16.4%) | 384 (14.0%) | 263 (11.0%) | 165 (10.8%) | 105 (10.6%) | 103 (11.9%) | 11 (7.2%) | 27 (9.8%) |
| From chemo start | ||||||||
| 6+ months prior | 281 (11.3%) | 223 (9.6%) | 155 (8.8%) | 107 (8.1%) | 97 (9.7%) | 67 (8.9%) | 11 (6.6%) | 24 (8.7%) |
| 6 to < 3 months prior | 144 (5.3%) | 132 (5.3%) | 68 (3.5%) | 54 (3.6%) | 35 (3.4%) | 50 (6.5%) | 8 (5.0%) | 10 (3.6%) |
| 3 to < 0 months prior | 1199 (44.3%) | 950 (38.0%) | 590 (29.5%) | 423 (26.5%) | 299 (28.6%) | 230 (29.7%) | 15 (9.1%) | 57 (20.2%) |
| 0 to < 3 months after | 924 (31.0%) | 924 (32.5%) | 723 (34.3%) | 511 (32.3%) | 365 (33.7%) | 161 (20.3%) | 39 (23.2%) | 88 (31.0%) |
| 3 to < 6 months after | 53 (2.0%) | 87 (3.6%) | 107 (6.4%) | 131 (8.8%) | 96 (9.3%) | 55 (6.2%) | 23 (13.9%) | 25 (8.9%) |
| 6 to < 12 months after | 44 (1.7%) | 75 (3.2%) | 92 (5.4%) | 106 (7.5%) | 71 (6.9%) | 62 (7.8%) | 26 (15.5%) | 31 (11.0%) |
| 12+ months after | 62 (2.3%) | 110 (4.8%) | 126 (8.2%) | 164 (11.6%) | 82 (8.9%) | 170 (21.4%) | 47 (29.7%) | 49 (16.6%) |
| Relative to surgery b | ||||||||
| Day 0 (same day) | 1142 (46.5%) | 1265 (57.4%) | 528 (31.2%) | 691 (47.4%) | 229 (22.1%) | 549 (79.7%) | 26 (12.2%) | 103 (37.1%) |
| Day 1–4 | 195 (2.4%) | 205 (4.5%) | 179 (11.6%) | 218 (13.2%) | 342 (32.7%) | 63 (8.2%) | 28 (14.2%) | 54 (16.2%) |
| Not surgery | 1520 (51.1%) | 1207 (40.8%) | 1192 (56.6%) | 655 (41.7%) | 503 (48.1%) | 239 (19.4%) | 118 (67.3%) | 136 (47.0%) |
| Same day as procedure | ||||||||
| Yes | 2528 (94.6%) | 2494 (99.7%) | 1579 (93.9%) | 1401 (98.1%) | 959 (79.4%) | 795 (99.8%) | 148 (83.9%) | 258 (90.7%) |
| No | 179 (5.4%) | 7 (0.3%) | 282 (6.1%) | 95 (1.9%) | 86 (20.6%) | 0 | 21 (16.1%) | 26 (9.3%) |
| Same day LP and BMA | ||||||||
| LP only | 109 (3.4%) | 136 (4.0%) | 81 (3.2%) | 53 (3.3%) | 17 (1.7%) | 36 (4.6%) | 0 | 2 (1.2%) |
| BMA only | 206 (4.2%) | 239 (4.6%) | 47 (2.2%) | 59 (4.3%) | 16 (1.4%) | 25 (3.5%) | 0 | 6 (2.8%) |
| LP and BMA | 360 (4.8%) | 402 (5.6%) | 65 (2.0%) | 43 (2.0%) | 22 (1.6%) | 15 (2.0%) | 1 (1.3%) | 0 |
| Same day as propofol | ||||||||
| Yes | 1767 (58.1%) | 2058 (80.1%) | 795 (30.5%) | 837 (49.8%) | 180 (13.5%) | 745 (94.1%) | 26 (9.4%) | 96 (31.9%) |
| No | 940 (41.9%) | 443 (19.9%) | 1066 (69.5%) | 659 (50.2%) | 865 (86.5%) | 50 (5.9%) | 143 (90.6%) | 188 (68.1%) |
| Route of administration e | ||||||||
| Intravenous | 1820 (81.1%) | 1963 (98.1%) | 1128 (80.9%) | 1146 (88.9%) | 0 | 644 (97.7%) | 22 (13.5%) | 155 (84.3%) |
| Oral | 291 (13.2%) | 0 | 350 (16.8%) | 72 (4.3%) | 636 (92.4%) | 0 | 59 (78.7%) | 24 (12.0%) |
| Other c | 127 (2.4%) | 108 (1.9%) | 13 (1.0%) | 71 (1.5%) | 52 (7.6%) | 61 (2.3%) | 13 (14.4%) | 13 (3.8%) |
| Year | ||||||||
| ≤ December 31, 2021 | 1708 (63.1%) | 1551 (61.9%) | 1108 (58.7%) | 855 (55.1%) | 597 (57.0%) | 398 (50.5%) | 60 (35.9%) | 145 (51.1%) |
| > December 31, 2021 | 999 (36.9%) | 950 (38.1%) | 753 (41.3%) | 641 (44.9%) | 448 (43.0%) | 397 (49.5%) | 109 (64.1%) | 139 (48.9%) |
Abbreviations: BMA, bone marrow aspirate; chemo, chemotherapy; CNS, central nervous system; LP, lumbar puncture.
Most recent cancer type prior to first opioid administration or prescription.
May be greater than the total due to multiple surgery windows for the same exposure.
Other includes infiltration, nasal, transdermal, sublingual, epidural, intramuscular, and missing as examples.
Percentages may not add up to 100% because each stratum was estimated independently using a random‐effects model.
Data available from two sites.
Table 3 summarizes characteristics associated with all 153,849 opioid administrations or prescriptions. Figure 1b shows that the most commonly administered or prescribed opioids were morphine (30.9%), hydromorphone (29.6%), fentanyl (21.5%) and oxycodone (12.1%). Patients older than 10 years were more common among those administered or prescribed hydromorphone (70.7%), oxycodone (69.4%) and methadone (82.3%). Regarding cancer type, leukemia was more common among patients administered or prescribed fentanyl (37.1%), solid tumors were more common among those administered or prescribed hydromorphone (42.8%), oxycodone (46.0%) and methadone (51.9%), and central nervous system tumors were more common among those administered or prescribed remifentanil (37.1%).
TABLE 3.
Characteristics of all opioid administrations or prescriptions (N = 153,849 administrations or prescriptions) c .
| Characteristic, n (%) | All opioids | Fentanyl | Morphine | Hydro‐morphone | Oxycodone | Remifentanil | Methadone | Other |
|---|---|---|---|---|---|---|---|---|
| Total number exposures | 153,849 | 33,078 | 47,487 | 45,471 | 18,599 | 3519 | 4243 | 1452 |
| Opioid order type | ||||||||
| Administration | 137,181 (89.3%) | 28,524 (98.6%) | 42,049 (84.4%) | 43,280 (94.5%) | 15,460 (85.1%) | 3293 (99.3%) | 3378 (82.8%) | 1197 (85.8%) |
| Prescription | 16,668 (10.7%) | 4554 (1.4%) | 5438 (15.6%) | 2191 (5.5%) | 3139 (14.9%) | 226 (0.7%) | 865 (17.2%) | 255 (14.2%) |
| Age at exposure | ||||||||
| < 1 years | 8812 (5.8%) | 2369 (7.4%) | 3911 (8.4%) | 1557 (3.1%) | 502 (1.7%) | 160 (5.2%) | 251 (1.1%) | 62 (4.7%) |
| 1–10 years | 65,754 (41.3%) | 15,841 (47.7%) | 25,207 (49.3%) | 15,155 (26.1%) | 6059 (28.8%) | 1824 (45.9%) | 910 (14.2%) | 758 (45.1%) |
| > 10 years | 79,283 (52.8%) | 14,868 (44.8%) | 18,369 (42.1%) | 28,759 (70.7%) | 12,038 (69.4%) | 1535 (46.7%) | 3082 (82.3%) | 632 (49.5%) |
| Latest cancer type a | ||||||||
| Leukemia | 56,449 (35.2%) | 12,727 (37.1%) | 17,541 (33.3%) | 17,764 (33.8%) | 6177 (32.1%) | 690 (17.7%) | 910 (22.5%) | 640 (35.4%) |
| Lymphoma | 15,635 (10.5%) | 4233 (11.3%) | 3554 (11.2%) | 3906 (9.7%) | 2801 (3.8%) | 250 (8.3%) | 749 (3.9%) | 142 (7.0%) |
| CNS | 15,018 (9.9%) | 3774 (11.8%) | 5121 (10.9%) | 2961 (5.6%) | 1711 (3.4%) | 1092 (37.1%) | 260 (1.9%) | 99 (6.2%) |
| Solid tumor | 56,256 (35.9%) | 9492 (29.4%) | 18,248 (35.1%) | 18,153 (42.8%) | 6796 (46.0%) | 1046 (26.2%) | 2064 (51.9%) | 457 (34.2%) |
| Unknown | 545 (0.3%) | 157 (0.4%) | 156 (0.3%) | 183 (0.2%) | 39 (0.2%) | 3 (0.1%) | 2 (0.1%) | 5 (0.4%) |
| Prior to cancer | 9946 (6.8%) | 2695 (8.6%) | 2867 (7.1%) | 2504 (5.4%) | 1075 (5.8%) | 438 (8.3%) | 258 (1.5%) | 109 (7.8%) |
| From chemo start | ||||||||
| 6+ months prior | 6848 (5.0%) | 2140 (6.7%) | 1693 (5.6%) | 1604 (4.0%) | 857 (5.5%) | 263 (8.8%) | 201 (0.7%) | 90 (7.1%) |
| 6 to < 3 months prior | 2572 (1.9%) | 864 (2.6%) | 643 (1.9%) | 445 (1.4%) | 318 (1.8%) | 208 (5.4%) | 64 (1.0%) | 30 (1.2%) |
| 3 to < 0 months prior | 15,705 (10.8%) | 5133 (14.9%) | 4453 (10.7%) | 2759 (7.1%) | 2064 (15.9%) | 874 (27.5%) | 264 (4.2%) | 158 (12.5%) |
| 0 to < 3 months after | 49,659 (30.0%) | 9899 (29.0%) | 17,570 (32.8%) | 14,386 (27.8%) | 5810 (31.2%) | 545 (15.5%) | 901 (22.5%) | 548 (31.9%) |
| 3 to < 6 months after | 24,041 (14.4%) | 3467 (10.5%) | 8747 (13.3%) | 8108 (15.9%) | 2530 (9.0%) | 295 (7.4%) | 684 (14.3%) | 210 (13.3%) |
| 6 to < 12 months after | 22,613 (13.9%) | 4190 (12.1%) | 6634 (13.4%) | 7824 (15.7%) | 2809 (16.3%) | 330 (8.6%) | 658 (15.9%) | 168 (10.5%) |
| 12+ months after | 32,411 (22.4%) | 7385 (22.4%) | 7747 (20.4%) | 10,345 (25.8%) | 4211 (12.8%) | 1004 (27.2%) | 1471 (31%) | 248 (17.6%) |
| Relative to surgery b | ||||||||
| Day 0 (same day) | 29,127 (21.2%) | 16,020 (46.1%) | 4012 (12.9%) | 4366 (11.7%) | 1476 (5.8%) | 2628 (84.1%) | 240 (4.0%) | 385 (27.1%) |
| Day 1–4 | 26,149 (17.4%) | 7098 (14.6%) | 6070 (17.9%) | 6806 (16.5%) | 4938 (22.5%) | 233 (9.0%) | 710 (12.3%) | 294 (16.8%) |
| Not surgery | 103,019 (63.5%) | 12,709 (42.0%) | 37,793 (70.3%) | 35,040 (73.2%) | 12,471 (71.4%) | 861 (14.6%) | 3346 (84.6%) | 799 (48.9%) |
| Relative to procedure | ||||||||
| Yes | 116,913 (91.9%) | 32,790 (99.3%) | 27,522 (90.9%) | 30,967 (92.8%) | 17,203 (82.5%) | 3519 (99.9%) | 3536 (85.8%) | 1376 (95.2%) |
| No | 36,936 (8.1%) | 288 (0.7%) | 19,965 (9.1%) | 14,504 (7.2%) | 1396 (17.5%) | 0 | 707 (14.2%) | 76 (4.8%) |
| LP and BMA | ||||||||
| LP only | 3266 (2.0%) | 1663 (4.7%) | 616 (1.3%) | 439 (0.9%) | 176 (1.2%) | 112 (3.3%) | 6 (0.2%) | 5 (0.5%) |
| BMA only | 2623 (1.0%) | 2396 (3.3%) | 338 (0.6%) | 313 (0.3%) | 113 (0.3%) | 70 (2.2%) | 7 (0.2%) | 20 (1.2%) |
| LP and BMA | 3312 (0.7%) | 2853 (2.8%) | 293 (0.3%) | 261 (0.2%) | 90 (0.1%) | 52 (1.6%) | 5 (0.1%) | 7 (0.5%) |
| Same day as propofol | ||||||||
| Yes | 38,268 (23.7%) | 20,310 (61.8%) | 6464 (11.5%) | 6166 (12.5%) | 1332 (4.1%) | 3334 (94.6%) | 235 (4.9%) | 427 (26.1%) |
| No | 115,581 (76.3%) | 12,768 (38.2%) | 41,023 (88.5%) | 39,305 (87.5%) | 17,267 (95.9%) | 185 (5.4%) | 4008 (95.1%) | 1025 (73.9%) |
| Route of administration d | ||||||||
| Intravenous | 92,515 (69.3%) | 23,681 (97.6%) | 31,218 (76.9%) | 33,043 (83.0%) | 0 | 2970 (97.9%) | 590 (1.8%) | 1013 (92.5%) |
| Oral | 34,171 (26.2%) | 0 | 11,286 (21.1%) | 8527 (13.8%) | 12,185 (98.3%) | 0 | 2039 (87.7%) | 134 (6.7%) |
| Other | 5253 (3.2%) | 1295 (2.4%) | 516 (1.7%) | 856 (2.1%) | 2251 (1.7%) | 163 (2.1%) | 144 (4.5%) | 28 (1.0%) |
| Year | ||||||||
| ≤ December 31, 2021 | 78,978 (50.9%) | 16,533 (50.2%) | 24,914 (50.9%) | 24,503 (51.7%) | 9170 (77.3%) | 1610 (50.6%) | 1620 (41.8%) | 628 (44.9%) |
| > December 31, 2021 | 74,871 (49.1%) | 16,545 (49.8%) | 22,573 (49.1%) | 20,968 (48.3%) | 9429 (22.7%) | 1909 (49.4%) | 2623 (58.2%) | 824 (55.1%) |
Abbreviations: BMA, bone marrow aspirate; CNS, central nervous system; LP, lumbar puncture.
Most recent cancer type prior to opioid administration or prescription.
May be greater than the total due to multiple surgery windows for the same exposure.
Percentages may not add up to 100% because each stratum was estimated independently using a random‐effects model.
Data available from two sites.
Table 3 also shows that fentanyl (46.1%) and remifentanil (84.1%) were frequently associated with surgery. In contrast, opioid administrations or prescriptions were uncommon on the same day as lumbar punctures or bone marrow aspirates. When examining temporal patterns, oxycodone was less commonly administered or prescribed after December 31, 2021 (77.3% vs. 22.7%), whereas methadone was more common after this date (41.8% vs. 58.2%).
Tables 2 and 3, along with Appendix 3, illustrate the timing of opioid administration or prescription relative to the start of chemotherapy. Appendix 3a shows the timing of each patient's first opioid administration or prescription, regardless of opioid type, with each patient represented once (N = 2707). The majority of patients received their first opioid administration or prescription within three months before chemotherapy initiation to three months after. Appendix 3b shows the timing of all administrations or prescriptions, with patients represented multiple times (N = 153,849). Of all opioid administrations, 22.4% occurred more than 12 months after starting chemotherapy.
Table 4 examines variability in opioid choice and naloxone administration or prescription across the three centers. The proportion of opioid recipients administered or prescribed fentanyl was consistent, ranging from 90.9% to 94.9%. In contrast, the use of other opioids and naloxone was more variable, ranging from 34.0% to 89.2% for morphine, 0.9%–76.1% for oxycodone, and 2.4%–29.5% for naloxone. Appendix 4 shows that none of the evaluated factors were significantly associated with greater morphine administration or prescription. In contrast, age ≥ 10 years was associated with lower fentanyl administration, while a leukemia diagnosis was associated with higher fentanyl administration.
TABLE 4.
Percentage of patients administered or prescribed each opioid type or naloxone across the three centers.
| Opioid type or naloxone | Center A (%) | Center B (%) | Center C (%) |
|---|---|---|---|
| Fentanyl | 90.9 | 91.7 | 94.9 |
| Morphine | 89.2 | 78.9 | 34.0 |
| Hydromorphone | 47.0 | 44.1 | 72.9 |
| Oxycodone | 0.9 | 76.1 | 73.9 |
| Remifentanil | 36.8 | 19.2 | 23.8 |
| Methadone | 1.3 | 16.0 | 8.4 |
| Other | 5.7 | 19.6 | 12.8 |
| Naloxone | 2.4 | 14.9 | 29.5 |
4. Discussion
In this multi‐center study leveraging the OMOP CDM, we found that nearly all pediatric cancer patients receiving chemotherapy were administered or prescribed an opioid. Fentanyl was the most common, and it was administered or prescribed in 92.4%. Overall, 10.9% were administered or prescribed naloxone, and it was more frequent among oxycodone and methadone patients. Opioids were uncommonly associated with lumbar punctures and bone marrow aspirations. There was heterogeneity in opioid choice and naloxone use across sites.
The near‐universal administration of opioids in our cohort was surprising. In contrast, a study using the IBM MarketScan Database examined sarcoma patients receiving chemotherapy, radiotherapy, or surgery and found that 64% received an opioid during active treatment [9]. Differences between their cohort and ours include restriction to sarcoma patients, inclusion of patients aged 10–26 years, restriction to outpatient pharmaceutical claims, and inclusion of patients not receiving chemotherapy. In our cohort, opioid exposure occurred predominantly in procedural contexts, with 91.9% of exposures occurring in relation to a procedure and 21.2% occurring on the same day as surgery. However, 63.5% were not temporally associated with surgery. This pattern suggests that the observed opioid use may be driven by peri‐procedural and procedural care. Alternatively, this association may reflect concurrent clinical events, as “procedures” may also include interventions such as transfusions and intravenous line placement.
Recently, opioids have received considerable attention in the press, particularly due to the role of illicitly manufactured fentanyl in the opioid epidemic and overdose deaths [10]. This context is distinct from the use of opioids in pediatric oncology, where they remain essential for pain management. Nevertheless, these broader concerns highlight the importance of opioid stewardship and raise the question of whether opioid use could be optimized, including consideration of non‐opioid medications or non‐pharmacological approaches in appropriate situations [11]. The rare association of opioid administration with lumbar punctures and bone marrow aspirations is reassuring, as these are settings where alternative approaches should be considered.
The substantial inter‐center variation observed in oxycodone (0.9%–76.1%) and naloxone use (2.4%–29.5%) likely reflects a complex interplay of institutional, national, and temporal factors rather than true differences in patient need. Differences in local formularies, prescribing culture, and guideline adherence, as well as variation in policy and regulatory environments, may have contributed to these findings. In addition, evolving opioid stewardship efforts and heightened awareness of opioid‐related harms, particularly during and following the COVID‐19 pandemic, may have influenced prescribing practices over time. While our data do not allow causal attribution, these findings underscore the need for further work to better understand the drivers of prescribing variation and to promote more standardized, evidence‐informed approaches to opioid use in this population.
More specifically, the substantial inter‐institutional variability in naloxone co‐prescribing observed in our cohort may be informative. From a patient safety perspective, naloxone co‐prescribing is widely supported as a harm reduction strategy for patients receiving opioids, with guideline recommendations increasingly encouraging its use in higher‐risk settings. These findings may highlight an opportunity to standardize best practices for naloxone co‐prescribing within opioid stewardship programs [12, 13]. At the same time, our data cannot determine whether higher rates of naloxone prescribing correspond to improved clinical outcomes or more appropriate targeting, nor can they distinguish between proactive prescribing policies [14] and responses to perceived patient risk. As such, the observed variability should be interpreted as reflecting differences in institutional implementation strategies rather than differences in underlying patient need.
In the meta‐regression analysis, we observed an association between age ≥ 10 years and lower fentanyl use. Younger children may be more likely to receive parenteral opioids, including fentanyl, due to challenges with oral administration, whereas older children and adolescents may be more likely to receive oral opioid formulations. We also found that patients with leukemia were more likely to receive fentanyl, which may reflect the higher frequency of procedures such as lumbar punctures and bone marrow aspirations in this population. These results, however, should be interpreted with caution. The need to redefine the outcome for regression, driven by the near‐universal receipt of opioids, represents a limitation; accordingly, this analysis should be interpreted as exploratory.
This study demonstrates both the feasibility and utility of leveraging OMOP to address clinically relevant questions in pediatric oncology. Because only summary‐level data are shared, this approach requires far fewer resources than traditional epidemiologic methods involving chart review, data abstraction, and transfer of individual patient data to a central database. Future studies could use OMOP to assess adherence to guidelines and to compare practices across sites and countries.
A key strength of this study is the relatively large sample size and the likely complete ascertainment of inpatient opioid administration, as medication administrations are typically well documented in the EHR. However, several limitations should be noted. We likely underestimated the duration of outpatient opioid use because prescription data do not reliably reflect the number of days medications were taken. The study was conducted with a small number of institutions, limiting the ability to evaluate between‐site variability. Also, we lacked granular data on the clinical indications for opioid use, limiting the ability to distinguish between procedural, acute treatment‐related, chronic pain, and palliative care contexts. Finally, we were unable to assess the precise timing of opioid administration relative to procedures. Consequently, same‐day associations should not be interpreted as evidence that an opioid was administered specifically for that procedure.
In summary, among pediatric patients with cancer who receive chemotherapy, 93.2% were administered or prescribed at least one opioid; fentanyl was the most common opioid type. There was heterogeneity in opioid choice and naloxone use across sites. These findings lay preliminary groundwork for understanding opioid use patterns and identifying potential areas for practice improvement in pediatric oncology.
Author Contributions
Yuqing Feng: methodology, formal analysis, writing – original draft. Martin Yi: methodology, formal analysis, writing – original draft. Catherine C. Aftandilian: conceptualization, writing – review and editing. Melissa P. Beauchemin: conceptualization, writing – review and editing. Christine Cunningham: writing – review and editing, methodology. David DeStephano: methodology, writing – review and editing. Rhea K. Khurana: writing – review and editing. Emily Larimer: writing – review and editing. Jennifer Oberg: writing – review and editing. Benjamin L. May: writing – review and editing. Keith Morse: writing – review and editing. Karthik Natarajan: writing – review and editing, methodology. David Noyd: conceptualization, writing – review and editing. Victoria Soucek: writing – review and editing, methodology. Yujie Chen: writing – review and editing. Adam Yan: writing – review and editing. Lin Lawrence Guo: methodology, writing – original draft. Lillian Sung: conceptualization, writing – original draft, supervision.
Funding
The authors have nothing to report.
Ethics Statement
This study was approved by the SickKids Research Ethics Board (REB #1000080216) and the Institutional Review Boards of participating sites. The requirement for informed consent was waived.
Conflicts of Interest
The authors declare no conflicts of interest.
Acknowledgments
L.S. is supported by the Canada Research Chair in Data‐enabled Pediatric Precision Medicine. L.S. had full access to the summary data from each center and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Appendix 1. OMOP Standard Concepts Used for Cohort Construction and Outcome Ascertainment
| Purpose | Domain | Ancestor concept ID | Ancestor concept name | Descendant concepts included |
|---|---|---|---|---|
| Pediatric cancer cohort | Condition | 443392 | Malignant neoplastic disease | All standard descendant condition concepts |
| Active cancer treatment | Drug | 21601387 | Antineoplastic agents | All standard descendant drug concepts |
| Primary outcome: opioid exposure | Drug | 21604254 | Opioids | All standard descendant drug concepts |
| Primary outcome: opioid anesthetics | Drug | 21604200 | Opioid anesthetics | All standard descendant drug concepts |
| Opioid subtype | Drug | 1154029 | Fentanyl | Standard descendant drug concepts |
| Opioid subtype | Drug | 1110410 | Morphine | Standard descendant drug concepts |
| Opioid subtype | Drug | 1126658 | Hydromorphone | Standard descendant drug concepts |
| Opioid subtype | Drug | 1124957 | Oxycodone | Standard descendant drug concepts |
| Opioid subtype | Drug | 19016749 | Remifentanil | Standard descendant drug concepts |
| Opioid subtype | Drug | 1103640 | Methadone | Standard descendant drug concepts |
Appendix 2. Number of Days of Opioid Administration or Prescription (N = 2707 Patients) a
| Characteristic | All opioids | Fentanyl | Morphine | Hydro‐morphone | Oxycodone | Remifentanil | Methadone | Other |
|---|---|---|---|---|---|---|---|---|
| Total number patients | 2707 | 2501 | 1861 | 1496 | 1045 | 795 | 169 | 284 |
| Cumulative number of days (%) | ||||||||
| 0–1 | 246 (9.7%) | 428 (16.8%) | 663 (46.1%) | 490 (35.7%) | 345 (34.5%) | 532 (78.3%) | 58 (25.6%) | 214 (78.5%) |
| 2–4 | 564 (20.9%) | 1103 (41.9%) | 472 (24.2%) | 392 (25.4%) | 316 (30.3%) | 235 (19.6%) | 34 (20.9%) | 44 (15.1%) |
| 5–14 | 988 (36.5%) | 816 (33.9%) | 407 (15.2%) | 315 (20.4%) | 260 (24.5%) | 27 (3.5%) | 40 (24.8%) | 22 (5.6%) |
| 15–29 | 430 (15.9%) | 129 (5%) | 174 (4.9%) | 160 (10.2%) | 74 (6.6%) | 0 (0.2%) | 22 (13.3%) | 3 (1.7%) |
| 30+ | 479 (15.3%) | 25 (0.9%) | 146 (3.2%) | 139 (6.5%) | 50 (3.8%) | 1 (0.3%) | 15 (10.4%) | 1 (0.7%) |
| Mean days (95% CI) | 16.5 (6.8, 26.2) | 5.6 (2.2, 9.0) | 5.6 (0, 19.1) | 9.3 (0, 19.2) | 6.6 (0, 13.4) | 1.2 (0, 4.0) | 8.7 (3.2, 14.1) | 1.3 (0, 5.0) |
| Number with naloxone (%) | 372 (10.9%) | 357 (11.1%) | 220 (13.3%) | 324 (17.8%) | 332 (24.2%) | 107 (13.5%) | 88 (43.5%) | 86 (19.1%) |
Abbreviation: CI, confidence interval.
Percentages may not add up to 100% because each stratum was estimated independently using a random‐effects model.
Appendix 3. Distribution of Opioid Timing by First Opioid and All Opioid Administrations or Prescriptions Relative to Chemotherapy Initiation
(a) Illustrates timing of first administration or prescription of opioids regardless of opioid type such that each patient is included only once (N = 2707 patients). In contrast, (b) illustrates timing of all administrations or prescriptions such that patients are represented multiple times (N = 153,849 administrations or prescriptions). Time 0 is first chemotherapy administration.
Appendix 4. Factors Associated With More Morphine and Fentanyl Administrations or Prescriptions
| Characteristics | > Median morphine days a | > Median fentanyl days a | ||
|---|---|---|---|---|
| Risk ratio (95% CI) | p | Risk ratio (95% CI) | p | |
| Age ≥ 10 years at first chemotherapy | 0.94 (0.72, 1.22) | 0.415 | 0.89 (0.87, 0.91) | 0.003 |
| Female sex | 0.96 (0.86, 1.08) | 0.302 | 0.98 (0.91, 1.07) | 0.484 |
| White race | 0.94 (0.76, 1.15) | 0.160 | 0.93 (0.51, 1.70) | 0.366 |
| Hispanic ethnicity | 0.97 (0.52, 1.81) | 0.674 | 1.08 (0.39, 3.00) | 0.532 |
| Leukemia | 1.00 (0.74, 1.34) | 0.969 | 1.45 (1.06, 1.98) | 0.037 |
Abbreviation: CI, confidence interval.
Median days morphine = 5 and fentanyl = 3.
Data Availability Statement
The OMOP datasets used in this study cannot be made publicly available because of the potential risk to patient privacy. However, relevant data may be made available from the corresponding author upon reasonable request and subject to institutional approvals and data‐sharing agreements.
Code Availability: The code is available at GitHub (private repository, access per invite: omop‐multicenter‐pediatric‐cancer/Opioid at main · sungresearch/omop‐multicenter‐pediatric‐cancer).
References
- 1. Dupuis L. L., Johnston D. L., Dix D., et al., “Symptom Screening for Hospitalized Pediatric Patients With Cancer: A Randomized Clinical Trial,” JAMA Pediatrics 179 (2025): 11–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Dupuis L. L., Johnston D. L., Baggott C., et al., “Validation of the Symptom Screening in Pediatrics Tool in Children Receiving Cancer Treatments,” Journal of the National Cancer Institute 110 (2018): 661–668. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Dupuis L. L., Vettese E., Grimes A. C., et al., “Symptom Screening Linked to Care Pathways for Pediatric Patients With Cancer: A Randomized Clinical Trial,” JAMA 332 (2024): 1981–1991. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Loeffen E. A. H., Mulder R. L., Font‐Gonzalez A., et al., “Reducing Pain and Distress Related to Needle Procedures in Children With Cancer: A Clinical Practice Guideline,” European Journal of Cancer 131 (2020): 53–67. [DOI] [PubMed] [Google Scholar]
- 5. Beauchemin M., Dorritie R., and Hershman D. L., “Opioid Use and Misuse in Children, Adolescents, and Young Adults With Cancer: A Systematic Review of the Literature,” Support Care Cancer 29 (2021): 4521–4527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Hall E. A., Shelton C. M., Hagemann T. M., Jasmin H. M., Grey K., and Anghelescu D. L., “A Narrative Review of Pain in Pediatric Oncology: The Opioid Option for Procedural and Surgical Pain,” Paediatric Drugs 27 (2025): 19–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Hripcsak G., Duke J. D., Shah N. H., et al., “Observational Health Data Sciences and Informatics (OHDSI): Opportunities for Observational Researchers,” Studies in Health Technology and Informatics 216 (2015): 574–578. [PMC free article] [PubMed] [Google Scholar]
- 8. Guo L. L., Calligan M., Vettese E., et al., “Development and Validation of the SickKids Enterprise‐Wide Data in Azure Repository (SEDAR),” Heliyon 9 (2023): e21586. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Beauchemin M. P., Raghunathan R. R., Accordino M. K., et al., “New Persistent Opioid Use Among Adolescents and Young Adults With Sarcoma,” Cancer 128 (2022): 2777–2785. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Han Y., Yan W., Zheng Y., Khan M. Z., Yuan K., and Lu L., “The Rising Crisis of Illicit Fentanyl Use, Overdose, and Potential Therapeutic Strategies,” Translational Psychiatry 9 (2019): 282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Paice J. A., Bohlke K., Barton D., et al., “Use of Opioids for Adults With Pain From Cancer or Cancer Treatment: ASCO Guideline,” Journal of Clinical Oncology 41 (2023): 914–930. [DOI] [PubMed] [Google Scholar]
- 12. Hadland S. E., Agarwal R., Raman S. R., et al., “Opioid Prescribing for Acute Pain Management in Children and Adolescents in Outpatient Settings: Clinical Practice Guideline,” Pediatrics 154 (2024): e2024068752. [DOI] [PubMed] [Google Scholar]
- 13. (CRISM) CRIiSM , Naloxone Best Practice Guideline (, 2023), https://crism.ca/naloxone‐best‐practice‐guideline/. [Google Scholar]
- 14. Naloxone Prescription Requirements, https://www.mbc.ca.gov/Resources/Medical‐Resources/Naloxone.aspx. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The OMOP datasets used in this study cannot be made publicly available because of the potential risk to patient privacy. However, relevant data may be made available from the corresponding author upon reasonable request and subject to institutional approvals and data‐sharing agreements.
Code Availability: The code is available at GitHub (private repository, access per invite: omop‐multicenter‐pediatric‐cancer/Opioid at main · sungresearch/omop‐multicenter‐pediatric‐cancer).