What is the difference between hydrocodone and tylenol 3

Key Points

Question Does a strong opioid (oxycodone hydrochloride 5 mg or 10 mg 4 times daily) provide better analgesia than a mild opioid (acetaminophen codeine combination 500 mg and 8 mg or 1000 mg and 16 mg 4 times daily) when taken postdischarge among patients who underwent orthopedic surgical treatment?

Findings In this randomized clinical trial of 120 patients with 1 or more acute orthopedic fractures requiring surgical fixation, the mean of daily mean pain scores from day 1 to 7 postdischarge was 4.04 in the strong opioid group and 4.54 in the mild opioid group. The between-group difference was neither statistically nor clinically significant despite a 6-fold higher dose of opioid being delivered in the strong opioid group.

Meaning This study found that oxycodone did not provide superior pain relief compared with combination acetaminophen and codeine during the first 7 days of treatment despite a 6-fold higher dose of opioid being delivered in the strong opioid group.

Importance Patients with a surgically managed fracture are commonly discharged from the hospital with a strong opioid prescription, but limited evidence exists to support this practice.

Objective To test the hypothesis that strong opioids provide greater analgesia than mild opioids over the first week postdischarge from hospital after fracture surgical treatment.

Design, Setting, and Participants This double-blind, superiority, randomized clinical trial was conducted at a single-center, major trauma hospital in Sydney, Australia. Participants were inpatients who had sustained an acute nonpathological facture of a long bone or the pelvis, patella, calcaneus, or talus who were treated with surgical fixation and enrolled from July 27, 2016, to August 22, 2017. Data were analyzed from June through October 2018.

Interventions Initiation at discharge of oxycodone hydrochloride 5 mg of 10 mg (ie, 1 or 2 tablets) or combination acetaminophen and codeine 500 mg and 8 mg or 1000 mg and 16 mg (ie, 1 or 2 tablets) 4 times daily for a maximum duration of 3 weeks.

Main Outcomes and Measures The primary outcome was the mean of daily pain scores collected during week 1 of treatment measured using the Numerical Pain Rating Scale (NRS). Participants were asked to rate their mean pain over the previous 24 hours daily using an NRS score from 0 to 10, with 0 representing no pain and 10 representing the worst pain imaginable. The key secondary outcomes were EuroQol 5-Dimension 5-Level Questionnaire (EQ-5D-5L) responses, worst pain, medication adverse events, global perceived effect, and return to work.

Results A total of 120 patients with 1 or more acute orthopedic fractures requiring surgical fixation were randomized, including 59 patients in the strong-opioid group (43 [72.9%] men; mean [SD] age, 36.0 [14.1] years; mean oral morphine equivalent for days 1-7 of 32.9 mg) and 61 patients in the mild opioid group (47 [77.1%] men; mean [SD] age, 38.2 [13.5] years; mean oral morphine equivalent for days 1-7 of 5.5 mg). From days 1 to 7 postdischarge, the mean daily NRS mean pain score was 4.04 (95% Cl, 3.67 to 4.41) in the strong opioid group and 4.54 (95% Cl, 4.17 to 4.90) in the mild opioid group. The between-group difference of the primary outcome was not statistically significant (−0.50 [95% Cl, −1.11 to 0.12]; P = .11) despite a 6-fold increased dose of opioids being delivered in the strong opioid group.

Conclusions and Relevance This study found that treatment with strong opioid medication subacutely was not superior to treatment with milder medication for treatment of pain among patients with surgically managed orthopedic fractures. These findings suggest that ongoing first-line strong opioid use after discharge from the hospital should not be supported.

Trial Registration Australia New Zealand Clinical Trial Registry No.: ACTRN12616000941460

Opioids have been used for centuries and are the most potent analgesic agents.1,2 Evidence supports the use of opioids to relieve moderate to severe pain, particularly acute and cancer pain.3 In these latter situations, the efficacy of opioids is extensively documented and broadly accepted. However, the evidence does not support using opioid therapy for chronic noncancer pain.4 The prescribing of opioids for chronic pain has increased despite evidence of the association of their long-term use with adverse outcomes.5-7While there is little debate over the short-term use of opioids, their use longer term is controversial, with increasing reluctance among some physicians to prescribe these medications.8 The most powerful opioid analgesics are also the most liable to be associated with abuse and addiction.9 With global attention on increased opioid prescription and the opioid abuse epidemic, reconsideration and reexamination of opioid pain management are being undertaken.

Orthopedic surgeons are among the top prescribers of opioid analgesics.10 Severe postoperative pain is common after orthopedic surgical treatment, and safe and effective management of this pain can be challenging for everyone on the health care team.11 Adequate relief of pain is a measure of patient satisfaction and may be associated with decreased chronic postsurgical pain; however, taking opioids for acute pain is associated with increased likelihood of long-term opioid use.12 Furthermore, increased initial opioid exposure (ie, increased total dose or longer-duration prescription) is associated with increased risks of long-term use, misuse, and overdose,13,14 suggesting that strong opioid prescription following discharge from the hospital should be considered carefully. Many choices are available for postoperative pain, but evidence to inform clinical choice after discharge in the subacute setting is sparse. However, patients with surgically managed fractures are commonly (and in many places, routinely) discharged home from the hospital with a strong opioid prescription.15

In the interest of minimizing the potential problems associated with long-term opioid use, we propose that discharge from the hospital signifies the most suitable time to transition people down the World Health Organization (WHO) analgesic ladder from step 3 (strong opioid) to step 2 (mild opioid) analgesics if strong opioids are not shown to be superior.16 With this in mind, we undertook a superiority trial to determine if a strong opioid (as per current practice) provided better analgesia than a mild opioid (a combination acetaminophen and low dose codeine) when taken postdischarge among patients who received orthopedic surgical treatment.

This study is a double-blind, randomized clinical trial comparing oxycodone hydrochloride (strong opioid) vs combination acetaminophen and codeine (mild opioid) for the treatment of postdischarge pain after orthopedic fracture surgical treatment. This study and protocol (Supplement 1) was approved by the Hunter New England Health Human Research Ethics Committee, New South Wales, Australia. Participants provided written informed consent. This trial has been reported in line with the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline recommendations for randomized clinical trials.

Participants were patients admitted to 1 major trauma hospital in Sydney, Australia, with at least 1 acute fracture requiring surgical treatment from July 2016 to August 2017. The research team obtained relevant medical history to assess the patient against the eligibility criteria. Eligible participants were required to have sustained a nonpathological fracture of a long bone (ie, humerus, radius, ulna, femur, tibia, or fibula) or the pelvis, patella, calcaneus, or talus treated with surgical fixation; be age 18 years or older; and be able to comprehend the study protocol, written in English. Patients were excluded if they had known or suspected multisystem trauma injuries (eg, major head, chest, or abdominal injury); known or suspected major infection after surgical treatment; known or suspected opioid dependency; or contraindications to study treatment or were pregnant or breastfeeding.

Randomization and Masking

A researcher not involved in participant recruitment or data collection generated a randomization schedule a priori using a computer-derived random number sequence. Study medication in the groups was identically overencapsulated and prepared according to the randomization schedule (concealed from study researchers and all participants) and sealed in medication-blinded blister packages. Upon the participant’s discharge from the hospital, the pharmacy dispensed a sealed medication package (in numerical order) to the participant. Once the package was opened, the participant was randomized to 1 of 2 groups (at a 1:1 ratio). The identical overencapsulated study packages ensured concealed allocation. Researchers, participants, and outcome assessors (DEJ and IAH) were blinded. To assess blinding integrity, the participants were asked to guess treatment allocation at the end of the study treatment period. Statistical analysis and preparation of the manuscript results and discussion (consisting of 2 versions, based on 2 possible group allocations) were performed prior to unblinding.

Before study enrollment, participants were administered oral oxycodone during their hospital admission and any other standardized hospital care. Members of the strong opioid group were prescribed oxycodone hydrochloride immediate release at 5 mg or 10 mg (ie, 1 or 2 tablets) 4 times per day. Members of the mild opioid group were prescribed acetaminophen and codeine at 500 mg and 8 mg or 1000 mg and 16 mg (ie, 1 or 2 tablets) 4 times per day. The maximum allowed daily dose (ie, 8 tablets/d) was supplied for 2 weeks (14 days). Study medication was titrated down to cessation during the third and final treatment week (days 15-21). Oxycodone monotherapy was selected to reflect standard therapy, while combination oxycodone and acetaminophen is not locally available. Combination codeine and acetaminophen was chosen as the comparator given that it was the lowest opioid dose available. If the participant considered pain well controlled during days 1 to 14, decreased dosing or cessation was allowed. Participants experiencing uncontrolled pain were discontinued from the study.

Data collection was conducted by researchers in person at baseline and via telephone at days 3, 7, 14, and 21 postdischarge. Participants recorded mean pain, worst pain, and adverse effects daily for 3 weeks or until study completion, with a minimum of days 1 through 7 collected, using a diary (eAppendix in Supplement 2). Study medication was returned at the end of the participants’ study period to calculate tablet use. Baseline data were collected on the hospital ward prior to discharge. Data were collected for the first 7 study days regardless of participants’ compliance with the study treatment.

The primary outcome was pain measured by the Numerical Pain Rating Scale (NRS), which ranges from 0 to 10, with 0 representing no pain and 10 representing the worst pain imaginable. The primary outcome was collected daily, with participants asked to rate their mean pain over the previous 24 hours. The mean of the daily mean pain scores collected from day 1 to 7 postdischarge was calculated. The primary outcome was compared as between-group difference in daily pain during week 1 of treatment (ie, days 1-7).

EuroQol 5-Dimension 5-Level Questionnaire (EQ-5D-5L) answers were collected at day 3 postdischarge and weeks 1, 2, and 3 postdischarge to assess health-related quality of life (HRQL). The EQ-5D-5L descriptive system is a preference-based HRQL measure with 1 question for 5 dimensions, including mobility, self-care, usual activities, pain and discomfort, and anxiety and depression.17 Worst pain was measured daily using the NRS. Participants were asked to rate their worst pain over the previous 24 hours on a scale of 0 to 10, with 0 representing no pain and 10 representing the worst pain imaginable. Medication adverse events were measured daily and recorded in the diary using a list of common adverse effects experienced, with participants asked to answer yes or no for each outcome (eAppendix in Supplement 2). The Global Perceived Effect was measured on days 7, 14, and 21; this questionnaire asks participants to compare their pain to what it was when their injury first occurred. It is measured on a Likert scale, from a score of −5 (vastly worse) to 0 (unchanged) to 5 (completely recovered). Return to work (with responses of yes [full or light duties] or no) was measured on days 7, 14, and 21, indicating a return to preinjury employment after injury.

Descriptive data were collected on study entry during hospital admission. These data included age, sex, employment status, diagnoses, type of surgical treatment, mechanism of injury, insurance status, pain, and quality of life.

The primary comparison was the difference between the means of the daily mean pain scores over the first 7 days, measured on an 11-point NRS. Because group was a patient-level factor, sample size was estimated on this level. A sample of 46 participants per group (with 322 individual NRS pain scores) will provide more than 90% power to show at least a 1.0-unit (on an 11-point pain scale) advantage of oxycodone over acetaminophen and codeine assuming an SD of 1.5 between patients and a 1-sided type 1 error rate of 5%. A sample of 120 participants was randomized to allow for a loss to follow-up of 20%.

The distributions of characteristics of patients included in the study were described for the treatment groups. Significance testing to compare baseline covariates was not used. Data were analyzed by a statistician (JD) who was blinded. All analyses were intention-to-treat analyses (including outcomes for participants who discontinued the study drug). We conducted 3 additional per-protocol analyses of the primary outcome to compare participants who discontinued the study drug during day 1 to 7 for various reasons. Analysis 1 excluded participants who never commenced the allocated medication. Analysis 2 excluded participants who never commenced the allocated medication and those who discontinued the medication owing to adverse effects. Analysis 3 excluded participants who discontinued for any reason. Missing data did not meet the predetermined threshold for imputation. We had no missing data from baseline to primary end point (up to day 7). The primary analysis was a multilevel model using a random slope model with maximum likelihood estimation comparing the mean of daily mean pain scores between treatment groups while accounting for repeated measures over time by patients. The model included days and treatment groups as independent variables. The secondary outcomes were analyzed using similar multilevel models; student t test, and Fisher exact test were used, as appropriate. A generalized estimating equation with cumulative logit link was used to analyze EQ-5D-5L outcomes at days 3 and 7, clustered by patient with days (ie, 3 and 7), treatment group, and interaction days and treatment group as independent variables. The sample size was 1-sided, although there is no mechanism to allow for 1-sided P values in a multilevel model. For all analyses, a P value < .05 was considered statistically significant. The data analysis for this paper was generated using SAS Enterprise Guide statistical software version 7.15 (SAS Institute). Data were analyzed from June through October 2018.

During a 13-month period (July 27, 2016, to August 22, 2017), 899 patients were screened for participation. Of these, 161 individuals were eligible and 134 individuals provided written consent. A total of 120 patients were randomized at a 1:1 ratio, including 59 patients in the strong-opioid group (43 [72.9%] men; mean [SD] age, 36.0 [14.1] years) and 61 patients in the mild opioid group (47 [77.1%] men; mean [SD] age, 38.2 [13.5] years) (Table 1). Patients at baseline had a mean (SD) age of 37.1 (13.9) years, while 105 patients (87.5%) had sustained a single fracture and most fractures were of a lower extremity (67 among 120 patients randomized [55.8%]) (eTable 1 in Supplement 2). The variables appeared evenly distributed between treatment groups and between the randomized and nonrandomized samples (eTable 1 in Supplement 2). Baseline mean pain and worst pain intensity were initially moderate prior to hospital discharge (mean [SD] NRS mean pain score, 4.2 [1.56]; mean [SD] NRS worst pain score, 6.5 [1.9]). Participant flow through to the primary end point (ie, day 7) and beyond (ie, days 8-21) is shown in the Figure. There were 37 patients (30.8%), including 18 patients in the strong opioid group and 19 patients in the mild opioid group, who exited the trial after the primary end point (ie, beyond day 7).

From day 1 to day 7, the intraclass correlation coefficient (ICC) of daily NRS mean pain scores was 0.69. The mean of daily NRS mean pain scores was 4.04 (95% Cl, 3.67 to 4.41) in the strong opioid group and 4.54 (95% Cl, 4.17 to 4.9) in the mild opioid group. The difference between groups in the mean daily mean pain during days 1 to 7 after discharge from the hospital (ie, the primary outcome measure) was not statistically significant (−0.50 [95% Cl, −1.11 to 0.12]; P = .11). The per-protocol analyses supported the intention-to-treat primary analysis (Table 2). The mean (SD) number of adverse effects reported in the strong opioid group was increased vs the mild opioid group, but the difference was not statistically significant (Table 2). The number of patients reporting at least 1 adverse effect daily for the entire study period (days 1 to 21) is shown in eTable 2 in Supplement 2. Furthermore, no statistically significant difference was seen for any secondary outcomes.

Between-group daily mean tablet use was not statistically significantly different in the strong vs the mild opioid group as observed at days 1 to 7 (4.36 tablets [95% CI, 3.84 to 4.89 tablets] vs 4.59 tablets [95% CI, 4.07 to 5.11 tablets]; mean difference, −0.23 tablets [95% CI, −0.91 to 0.45 tablets]; P = .58) or days 1 to 21 (3.33 tablets [95% CI, 2.71 to 3.96 tablets] vs 3.49 tablets [95% CI, 2.88 to 4.11 tablets]; mean difference, −0.16 tablets [95% CI, −0.94 to 0.62 tablets]; P = .69). The morphine equivalent of the mean opioid use during day 1 to day 7 in the strong opioid group was 32.9 mg of oral morphine daily (21.96 mg oxycodone; conversion factor, 1.5), while the morphine equivalent of the mean opioid use in the mild opioid group was 5.5 mg of oral morphine daily (36.48 mg codeine; conversion factor 0.15).18 No secondary outcome measures, including patient-reported outcomes (eTable 3 in Supplement 2), were statistically significantly different between study groups.

At the end of study participation, among patients assigned to oxycodone, 8 patients (13.5%) correctly thought that they were taking oxycodone, 18 patients (30.5%) incorrectly thought that they were taking acetaminophen and codeine, and 33 patients (56.0%) would not guess an assignment. For participants assigned to combination acetaminophen and codeine, 21 patients (34.5%) correctly thought that they were taking combination acetaminophen and codeine, 16 patients (26.2%) incorrectly thought that they were taking oxycodone, and 24 patients (39.3%) could not guess an assignment.

Among patients discharged home following orthopedic surgical fixation for 1 or more acute traumatic bone fractures in this randomized clinical trial, oxycodone did not provide superior pain relief compared with combination acetaminophen and codeine during the first 7 days of treatment, despite a 6-fold higher dose of opioids being delivered in the strong opioid group. This study additionally found that oxycodone did not provide superior pain relief compared with combination acetaminophen and codeine over 3 weeks of treatment. The largest difference in mean NRS pain score between treatments (0.5) was not statistically significant and was less than 0.9 to 1.9, which is a commonly reported range to define the minimal clinically important difference (MCID) in pain.19,20 The 95% CIs of the primary intention-to-treat analysis included the prespecified conservative MCID selected, so the possibility of a clinically important difference (at 1.0 unit) cannot be excluded. However, the findings support the inference that there is no clinically important benefit to strong opioids vs mild opioids. The results further suggest that a mild opioid combination of acetaminophen and codeine may represent a viable alternative to the standard practice of prescribing a strong opioid analgesic for the treatment of surgically managed orthopedic fractures upon discharge.

Strong opioids are commonly prescribed at discharge after orthopedic fracture management.21 There is evidence that strong opioids do not provide superior pain relief to nonopioid or mild opioid alternatives within similar acute settings, but most studies providing this evidence are single-dose trials or of short duration.22-30 There is substantial evidence and acceptance that strong opioid use is not justified longer term in the treatment of chronic noncancer pain.31,32 Furthermore, a greater amount of initial opioid exposure (ie, increased total dose and a longer-duration prescription) is associated with increased risks of long-term use, misuse, and overdose.13,14 Considering the concern pertaining to opioid harms, dose reduction after acute hospitalization and after surgical treatment is a priority.

The concept of nonopioid and opioid dichotomy (ie, mild vs strong pain relief) providing varying levels of effectiveness is evident in the WHO pain ladder that has guided clinicians in the treatment of cancer since 1986 and noncancer pain at a later date.16 Depending on the intensity of pain, nonopioids (eg, ibuprofen and acetaminophen) are prescribed first and then mild opioids (eg, codeine) as necessary, followed by strong opioids (eg, hydrocodone and oxycodone). For individuals sustaining orthopedic fractures, initiation of strong opioids on admission is common. The findings of our study coupled with the existing literature suggest that a reduction down the WHO pain ladder from strong opioids to mild opioids upon discharge, rather than continuing strong opioids, should be supported. This change in prescribing habit could potentially help mitigate the harms from ongoing opioid use by decreasing the number of individuals initially exposed to opioids in the community and the subsequent risk of future addiction. In addition, this study challenges the concept of categorizing pain analgesics in general, at least in terms of their analgesic effects, given that single-dose trials of nonopioids, mild opioids, and strong opioids can provide similar analgesic effects in a variety of clinical settings.33 These findings suggest that current community perceptions and attitudes according to which strong opioids are considered the most effective and most appropriate approach in the management of subacute pain should be rethought.

Strengths and Limitations

This study’s findings should be considered in light of its strengths and weaknesses. The study was adequately powered to detect a conservatively selected MCID (we used the smallest of the reported range within the study18) over the first week of treatment, group allocation was random, the loss to follow-up was minimal, analgesic consumption was closely monitored, and blinding was considered successful, given that most participants were unable to guess allocation. The study followed a pragmatic design, reflecting current prescribing habits, and limited the primary end point to the first week postdischarge assuming that many patients would recover and not require ongoing treatment after this point. This assumption was correct, with almost 31% of patients not requiring study treatment beyond day 7.

This study also has several limitations. Its single-center design and use of a hospital in a region with a high level of socioeconomic disadvantage may limit generalizability. The study excluded fragility fractures among older individuals; thus, the results do not apply to this patient population. The protocol allowed participant completion (dropout) after day 7, adding potential bias for findings from days 8 to 21. Additionally, we acknowledge that while we determined that strong opioids were not superior to mild opioids for pain relief, we could not determine if mild opioids were inferior. However, a larger sample size may have clarified whether adverse effects were increased in the strong opioid group, and if they were increased, establishing noninferiority may not be important.

This study found that treatment with strong opioids was not superior to treatment with a milder opioid medication for postdischarge treatment of subacute pain among patients with surgically managed orthopedic fractures. These findings suggest that ongoing strong opioid use after discharge from the hospital should not be supported.

Accepted for Publication: September 22, 2021.

Published: November 17, 2021. doi:10.1001/jamanetworkopen.2021.34988

Corresponding Author: Deanne E. Jenkin, PhD, Daffodil Centre, University of Sydney, a joint venture with Cancer Council New South Wales, PO Box 572, Kings Cross, NSW 1340, Australia ().

Author Contributions: Dr Jenkin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Jenkin, Descallar, Harris.

Drafting of the manuscript: Jenkin, Descallar.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Jenkin, Descallar.

Obtained funding: Jenkin, Naylor, Harris.

Administrative, technical, or material support: Jenkin, Harris.

Supervision: Jenkin, Naylor, Harris.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded through a competitive government grant from the State Insurance Regulatory Authority, New South Wales government.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Dhalla IA, Mamdani MM , Sivilotti ML , Kopp A , Qureshi O , Juurlink DN . Prescribing of opioid analgesics and related mortality before and after the introduction of long-acting oxycodone. CMAJ. 2009;181(12):891-896. doi:10.1503/cmaj.090784PubMedGoogle ScholarCrossref

Coyle DT , Pratt C-Y , Ocran-Appiah J , Secora A , Kornegay C , Staffa J . Opioid analgesic dose and the risk of misuse, overdose, and death: a narrative review. Pharmacoepidemiol Drug Saf. 2018;27(5):464-472. doi:10.1002/pds.4366PubMedGoogle ScholarCrossref

14.

Brat GA , Agniel D , Beam A , et al. Postsurgical prescriptions for opioid naive patients and association with overdose and misuse: retrospective cohort study. BMJ. 2018;360:j5790. doi:10.1136/bmj.j5790PubMedGoogle Scholar

20.

Myles PS , Myles DB , Galagher W , et al. Measuring acute postoperative pain using the visual analog scale: the minimal clinically important difference and patient acceptable symptom state. Br J Anaesth. 2017;118(3):424-429. doi:10.1093/bja/aew466PubMedGoogle ScholarCrossref

22.

Chang AK , Bijur PE , Lupow JB , Gallagher EJ . Comparative analgesic efficacy of oxycodone/acetaminophen vs codeine/acetaminophen for short-term pain management following ED discharge. Pain Med. 2015;16(12):2397-2404. doi:10.1111/pme.12830PubMedGoogle ScholarCrossref

23.

Chang AK , Bijur PE , Munjal KG , John Gallagher E . Randomized clinical trial of hydrocodone/acetaminophen versus codeine/acetaminophen in the treatment of acute extremity pain after emergency department discharge. Acad Emerg Med. 2014;21(3):227-235. doi:10.1111/acem.12331PubMedGoogle ScholarCrossref

26.

Mitchell A , McCrea P , Inglis K , Porter G . A randomized, controlled trial comparing acetaminophen plus ibuprofen versus acetaminophen plus codeine plus caffeine (Tylenol 3) after outpatient breast surgery. Ann Surg Oncol. 2012;19(12):3792-3800. doi:10.1245/s10434-012-2447-7PubMedGoogle ScholarCrossref

27.

Chang AK , Bijur PE , Esses D , Barnaby DP , Baer J . Effect of a single dose of oral opioid and nonopioid analgesics on acute extremity pain in the emergency department: a randomized clinical trial. JAMA. 2017;318(17):1661-1667. doi:10.1001/jama.2017.16190PubMedGoogle ScholarCrossref

30.

Krebs EE , Gravely A , Nugent S , et al. Effect of opioid vs nonopioid medications on pain-related function in patients with chronic back pain or hip or knee osteoarthritis pain: the SPACE randomized clinical trial. JAMA. 2018;319(9):872-882. doi:10.1001/jama.2018.0899PubMedGoogle ScholarCrossref