Antibiotic Stewardship in AECOPD Through CRP-Guided Management

Antibiotic Stewardship Through CRP-guided Antibiotic Treatment for Patients With Acute Exacerbation of Chronic Obstructive Pulmonary Disease (AECOPD)

Objectives: To determine whether CRP-guided antibiotic treatment for managing AECOPD in adult patients attending Emergency Departments leads to reduced antibiotic duration, without non-inferior COPD health status with usual care. Hypothesis to be tested: (i) The antibiotic duration in AECOPD patients will be significantly lower for CRP-guided antibiotic discontinuation than usual care; and (ii) COPD health status as measured by the Clinical COPD Questionnaire has no statistically significant difference between two groups. Design and subjects: Multi-center, single-blind, open-label, randomized, combined superiority (antibiotic duration) and non-inferiority (COPD health status) trial in 1,184 adult AECOPD patients presented to A&E. Instruments: Clinical COPD Questionnaire and EuroQol-5D Interventions: Both intervention and control groups follow usual care with GOLD strategy. The intervention group will be recommended to test for serum CRP daily. Antibiotic prescription is considered when CRP >5mg/dL. Once CRP has declined to <5mg/dL and the patient was afebrile for past 48 hours, antibiotic discontinuation will be considered. Communication with Receiving Ward Staffs: Participants in the study may transfer to another departments after treatment/ care in A&E. The following communication would be conducted: A handover note that informs the receiving ward staffs about patients' enrolment to the trial, group assignment, and previous treatments given in A&E. The note would also suggest the investigations for the receiving ward staffs. Telephone handover about intervention group and investigations of the study, and treatments given in A&E to ward. Main outcome measures: The antibiotic duration (total number of antibiotic days) within 28 days and recovery in terms of COPD health status (Clinical COPD Questionnaire total scores) within 14 days from randomisation. Data analysis: Intention-to-treat and cost-effectiveness analyses will be performed. The outcome assessors and data analysts will be blinded to group allocation. Expected results: The intervention group will exhibit reduction in antibiotic duration at 4-weeks, without negatively impacting on COPD health status, compared with the control group.

Introduction Multiple small successes against antimicrobial resistance are urgently needed. The Global Action Plan in 2015 reiterated the importance of tackling antimicrobial abuse and bacterial resistance, which is jeopardizing our ability to manage infectious diseases. Antimicrobial resistance is predicted to cause catastrophic economic impacts in the next 10 years by putting 24 million people into extreme poverty, and will lead to an estimated 10 million deaths annually by 2050. Evidence has shown that an extended course of antibiotics is by no means superior to a shorter one. The over-prescription of broad-spectrum antibiotics in both hospitals and primary care has become a global concern and has driven initiatives to safeguard patient safety through the responsible use of antibiotics as treatments and prophylaxis. In a recent analysis, 55% of pathogenic bacteria identified in sputum culture from AECOPD patients in Hong Kong has demonstrated resistance to at least one antibiotic. Even relatively small changes in prescribing are likely to have beneficial effects on resistance at a population level. There is a need for improved decision-making for antibiotic prescription in AECOPD. International Guideline (GOLD strategy) recommends empirical broad-spectrum antibiotics covering common respiratory pathogens for acute management, based on self-reported subjective AECOPD features, and the risk of bacterial infection. Evidence for the optimal duration of antibiotic treatment is unclear. Antibiotics are reportedly used in 87% of AECOPD patients in US, 74% in UK, and 96% in Hong Kong. AECOPD contributes 21,000 episodes of local hospital admission and readmission in 2019. AECOPD is triggered by infectious and non-infectious precipitants, although 30% are undetermined. Bacterial infection accounts for only 35%-57% among the infected. Inappropriate antibiotic usage may increase the risks of adverse events such as Clostridioides difficile colitis. Moreover, this confers a predisposition to airway colonization by multidrug-resistant bacteria, increasing COPD patients' risk of carrying resistant organisms in their lungs and then exacerbation or progression of pneumonia. The investigators have found that 55% of pathogenic bacteria identified from sputum culture among AECOPD patients were resistant to 1 or more antibiotics. C-Reactive Protein (CRP) may help physicians make more appropriate antibiotic decisions. C-Reactive Protein (CRP) is an acute-phase reactant secreted by the liver in response to bacterial infections. It is synthesized within 4 to 6 hours after tissue injury/inflammation, and levels double every 8 hours, peaking at about 36 hours. CRP is significantly elevated in patients with bacterial infection, representing better treatment effect with antibiotics at elevated CRP values. CRP has been shown to be highly selective for AECOPD and it has excellent diagnostic accuracy when interpreted together with AECOPD symptomatology. Elevated CRPs (>5 mg/dL) in hospitalised AECOPD patients are associated with chest infections, implying a treatment role for antibiotics in this group. Yet antibiotics did not offer higher benefit over placebo among those with normal CRP level. Work done by others CRP-guided antibiotic prescription is effective and safe. In hospitalised AECOPD patients with CRP-guided prescriptions (antibiotic treatment if CRP ⩾50 mg/L, i.e. 5mg/dL) versus a control group with standard GOLD treatment fewer patients in the CRP group were prescribed with antibiotics compared to the GOLD group (31.7% versus 46.2%, p=0.028; adjusted odds ratio (OR) 0.178, 95% CI 0.077-0.411, p=0.029), without a significant difference in the treatment effects between both groups. Yet, the discontinuation of antibiotics in both groups was guided by clinical judgment. The 30-day treatment failure rate was nearly equal in CRP and control groups (44.5% vs. 45.5%, p=0.881), the time to next exacerbation was comparable (32 days vs. 28 days, p=0.713), and the length of hospital stay was also similar (7 days vs. 6 days, p=0.206). On day 30, there was no difference in the symptoms score and quality of life, and no serious adverse events were detected. CRP-guided antibiotic discontinuation is also safe. Whether antibiotic therapy with CRP-guided duration or fixed 7-day duration were clinically non-inferior to fixed 14-day duration in patients with uncomplicated gram-negative bacteraemia was determined in a multicentre trial involved 504 adults. The main outcome was the rate of clinical failure at 30 days, which was 2.4% in CRP-guided patients who experienced recurrent bacteraemia, local suppurative complications, distant complications, re-initiation of antibiotic therapy due to clinical deterioration, or any-cause mortality, compared with a failure rate of 6.6% in the 7-day group and 5.5% in the 14-day group. The CRP-guided treatment met the 10% margin for inferiority. Work done by the research team The team is experienced to conduct trials on antibiotic stewardship with AECOPD. Professor Butler and his research team studied whether point-of-care CRP laboratory testing could be used to safely guide (and perhaps reduce) antibiotic prescribing for AECOPD in primary care practices in the UK. In this multi-centre, open-label, randomised controlled trial, 653 AECOPD patients were randomised to either the CRP-guided treatment group or the usual (GOLD strategy) care group. The results showed that patients in the CRP group reported less antibiotic use within 28 days of randomisation (57.0% vs. 77.4%, adjusted OR 0.31, 95% CI 0.20 to 0.47) and mildly lower (better) clinical COPD questionnaire scores at 2 weeks (adjusted mean difference -0.19 points, 90% CI -0.33 to -0.05). Antibiotics were prescribed less frequently at the initial consultation in the CRP-guided group compared with the usual care group (47.7% vs. 69.7%, adjusted OR 0.31, 95% CI 0.21 to 0.45). There was no evidence of harm to patients, including no differences in diagnosis of pneumonia at the 4-week follow-up and no differences in adverse events related to antibiotics. Both Prins' study and this study provide evidence of the benefits of CRP-guided antibiotic prescription in regard to less antibiotic use, and non-inferiority on the clinical outcomes in AECOPD patients including 30-day treatment failure, length of stay, symptoms score, diagnosis of pneumonia, quality of life, and adverse events. In the previous study antibiotics were found to be prescribed in 33% of patients who had low CRP levels, even though guidance indicated that antibiotics likely would not be of benefit. Patient factors such as expectations for antibiotics, access to antibiotics before consultation with the clinician, and lack of clear guidelines can influence clinicians' antibiotics prescribing behaviour for acute respiratory symptoms. Previous reports also suggested that risk aversion and long turnaround time for CRP resulted in driving them to prescribe antibiotics. Local situation has been well studied. Professor Hui (co-applicant) and Dr. Ko (collaborator) found that an infectious aetiology could be established in only 48.7% of the AECOPD patients. 40.8% of sputum culture was positive. The commonest bacteria identified were Streptococcus pneumoniae, Pseudomonas aeruginosa, and Haemophilus influenza. A&E is an ideal location to start antibiotic stewardship for AECOPD. Prior work has also been conducted by the research team in Hong Kong to ensure the significance of the problem, the feasibility of data collection, and the preliminary results relevant to AECOPD patients with available CRP levels. 21,129 patients were admitted to emergency medicine wards and medical wards through A&E in 2019 (mean age 79 years (IQR 71-86)). Elevated CRP was associated with doubled odds ratio of ICU admission or death (Odds ratio 2.06, 95%CI 1.75 - 2.43). Analysis reveals that 96% of patients with a normal CRP were given antibiotics. Among those with normal CRP, unadjusted statistics suggested those without antibiotics did not have lower odds ratios for deterioration requiring ICU admission and for death (p<0.01). Overall, AECOPD patients were prescribed antibiotics for a mean of 9.4 days (SD 10.3) but the mean duration of an elevated CRP was 6.4 days (SD 6.8). There is an opportunity to shorten antimicrobial regimen for some AECOPD patients. As personalised antibiotic treatments may reduce healthcare resource utilisation, the investigators sought to determine the optimal antibiotics therapy duration for AECOPD patients based on clinical decision-making and CRP results. Research Gaps Antibiotic stewardship involves both appropriate prescription and timely discontinuation of an antibiotic treatment. Limited evidence informs antibiotic prescription practice for AECOPD in A&E. Information on CRP-guided antibiotic discontinuation is absent. Previous studies with hospitalized AECOPD patients with severe symptoms and severe outcomes was insufficient in power. Cost-effectiveness analyses on CRP-guided antibiotic treatment in AECOPD patients are lacking. Clinical Question In adult AECOPD patients in Emergency Departments, does CRP-guided antibiotic treatment, compared with usual care, lead to shorter antibiotic duration without negatively impact on clinical COPD-health status? Aims and Hypotheses to be tested: The present study seeks to establish whether serial CRP level guided clinical judgment can safely and cost-effectively be used to better target antibiotic treatment (include both treatment and discontinuation) for AECOPD in emergency departments to those that are most likely to benefit, so that overall antibiotic exposure duration is decreased without compromising COPD-related health status. The primary objective is to determine whether the addition of a serial CRP levels to usual care for AECOPD leads to shorter antibiotic duration, The co-primary objective is to determine whether the addition of a serial CRP levels to usual care for AECOPD is non-inferior to usual care alone, on COPD health status as measured by CCQ. The secondary objectives are the effectiveness of the addition of serial CRP levels to usual care in AECOPD patients on: General health status as measured by EQ-5D-5L Adverse effects of antibiotics Cost-effectiveness of individualised CRP-guided antibiotic treatments. The investigators hypothesise that: (i) antibiotic duration (defined as 'the number of days of antibiotic treatment') in 28 days among AECOPD patients will be significantly lower for CRP-guided antibiotic treatment (including prescription and discontinuation) than usual care (ii) COPD health status as measured by the Clinical COPD Questionnaire (CCQ) has no statistically significant difference between two groups; (iii) general health status of AECOPD patients as measured by EQ-5D-5L will not be significantly different between the two groups (iv) adverse effects of antibiotic treatment in AECOPD patients will not be significantly different between the two groups (v) CRP-guided antibiotic treatment will be more cost-effective than usual care

Acute Exacerbation of COPD, Lung Diseases, Lung Diseases, Obstructive, Pulmonary Disease, Chronic Obstructive, Respiratory Tract Disease

Standard usual care (following GOLD initiative, including bronchodilator, and systemic steroid). The attending doctor will use the CRP level to inform their decision to continue antibiotics around 3 hours from blood taking. These doctors will be provided pre-study training on CRP interpretation. Every day from randomization, serum CRP testing will be encouraged to take. Once CRP has declined to <5mg/dL and the patient has remained afebrile for past 48 hours, antibiotic treatment will be reviewed for discontinuation. Otherwise, antibiotic treatment will be continued. A switch on the administration route, or a change of antibiotics due to adverse effect, allergy, or suggestion from culture result, is permitted according to in-patient physician's decisions. CRP will continue to be monitored daily upon discharge from that hospital (up to 28 days).

Patients in the control arm will be treated with usual care (GOLD initiative). No CRP would be measured.No CRP would be measured.

The attending doctor will use the CRP level to inform their decision to continue antibiotics around 3 hours from blood taking. These doctors will be provided pre-study training on CRP interpretation. Every day from randomization, serum CRP testing will be encouraged to take. Once CRP has declined to <5mg/dL and the patient has remained afebrile for past 48 hours, antibiotic treatment will be reviewed for discontinuation. Otherwise, antibiotic treatment will be continued. A switch on the administration route, or a change of antibiotics due to adverse effect, allergy, or suggestion from culture result, is permitted according to in-patient physician's decisions. CRP will continue to be monitored daily upon discharge from that hospital (up to 28 days).

Mean antibiotic duration (in days) within 28-days for AECOPD after randomization, measured at the end of each week, using audio/video call until 28 days post randomization. Medication record is also reviewed at 4-weeks post randomization to capture prescription of antibiotics related or unrelated to AECOPD.

The CCQ is a validated and reliable 10-item, self-administered questionnaire. The CCQ consists of three subdomains: symptoms, functional state and mental state. Items are scored on a Likert scale (range 0-60). The final score is the sum of all items divided by 10; separate scores for all three domains can be calculated. Higher scores indicate a worse health status.

Prevalence of potentially pathogenic bacteria cultured from sputum at 4 weeks and the proportion of bacteria that are resistant

Prevalence of commensal organisms cultured from sputum at 4 weeks and the proportion of bacteria that are resistant

Measured by European Quality of Life-5 Dimensions 5-Level questionnaire(EQ-5D-5L). The EQ-5D-5L essentially consists of 2 pages: the EQ-5D descriptive system and the EQ visual analogue scale (EQ VAS). The descriptive system comprises five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each dimension has 5 levels: no problems, slight problems, moderate problems, severe problems and extreme problems. The patient is asked to indicate his/her health state by ticking the box next to the most appropriate statement in each of the five dimensions. This decision results in a 1-digit number that expresses the level selected for that dimension. The digits for the five dimensions can be combined into a 5-digit number that describes the patient's health state. The VAS can be used as a quantitative measure of health outcome that reflect the patient's own judgement.

General health status measured by European Quality of Life-5 Dimensions 5-Level questionnaire (EQ-5D-5L).

EQ-5D is an instrument which evaluates the generic quality of life developed in Europe and widely used. The EQ-5D descriptive system is a preference-based HRQL measure with one question for each of the five dimensions that include mobility, self-care, usual activities, pain/discomfort, and anxiety/depression.

Disease-specific, health-related quality of life over time measured using the Chronic Respiratory Disease Questionnaire Self-Administered Standardized(CRQ-SAS)

The CRQ-SAS is a 20-item self-administered questionnaire covering four dimensions: dyspnoea, fatigue, emotional function and mastery. The response options for each question range from 1 (maximum impairment) to 7 (no impairment). The total score per domain is computed by summing the scores and dividing the total score by the number of items. In the analyses, the CRQ-SAS was divided into two domains: the CRQ-SAS physical domain (the mean of the dyspnoea and fatigue domains) and the CRQ-SAS emotional domain (the mean of the emotional function and mastery domains). Higher scores indicate better HRQoL.

Inclusion Criteria: Being diagnosed with active AECOPD (AECOPD is defined as an event in the natural course of a disease characterized by a change in baseline dyspnoea, cough, and/or sputum that is beyond the normal day-to-day variations with acute onset, which may warrant a change in regular medication in patients with underlying COPD). Known COPD in their medical records. Age 40 years or older. Able to provide informed consent in Cantonese, Mandarin, or English Able to complete the questionnaires during the study period (i.e. 6 months after randomisation) Exclusion Criteria: Patients will be excluded if any ONE of the following are present: Pre-treatment with systemic corticosteroids for the present exacerbation. Pre-treatment with any antibiotics for the present exacerbation, any concurrent infection or prophylaxis. Known clinical stroke in past 6 months Patients with high suspicion of active AECOPD mimics: Pneumonia Congestive heart failure Bronchiectasis Pulmonary embolism Pneumothorax Atrial fibrillation / flutter Lung comorbidities: Cystic fibrosis Tuberculosis Unresolved lung malignancy Progression or new radiographic abnormalities on the chest X-ray. Immunodeficiency disorders such as AIDS, humoral immune defect, ciliary dysfunction etc., and the use of immunosuppressive drugs for more than 28 days. Active inflammatory condition (e.g. flare up of rheumatoid arthritis, gout or polymyalgia rheumatica) or concurrent infection at another site (e.g. UTI, cellulitis) that is likely to produce a systemic response Currently pregnant NEWS2 score of ≥3

Mendelson M, Matsoso MP. The World Health Organization Global Action Plan for antimicrobial resistance. S Afr Med J. 2015 Apr 6;105(5):325. doi: 10.7196/samj.9644. No abstract available.

Strathdee SA, Davies SC, Marcelin JR. Confronting antimicrobial resistance beyond the COVID-19 pandemic and the 2020 US election. Lancet. 2020 Oct 10;396(10257):1050-1053. doi: 10.1016/S0140-6736(20)32063-8. Epub 2020 Sep 29. No abstract available.

Tansarli GS, Mylonakis E. Systematic Review and Meta-analysis of the Efficacy of Short-Course Antibiotic Treatments for Community-Acquired Pneumonia in Adults. Antimicrob Agents Chemother. 2018 Aug 27;62(9):e00635-18. doi: 10.1128/AAC.00635-18. Print 2018 Sep.

Goossens H, Ferech M, Vander Stichele R, Elseviers M; ESAC Project Group. Outpatient antibiotic use in Europe and association with resistance: a cross-national database study. Lancet. 2005 Feb 12-18;365(9459):579-87. doi: 10.1016/S0140-6736(05)17907-0.

Halpin DMG, Criner GJ, Papi A, Singh D, Anzueto A, Martinez FJ, Agusti AA, Vogelmeier CF. Global Initiative for the Diagnosis, Management, and Prevention of Chronic Obstructive Lung Disease. The 2020 GOLD Science Committee Report on COVID-19 and Chronic Obstructive Pulmonary Disease. Am J Respir Crit Care Med. 2021 Jan 1;203(1):24-36. doi: 10.1164/rccm.202009-3533SO.

Stolbrink M, Bonnett LJ, Blakey JD. Antibiotics for COPD exacerbations: does drug or duration matter? A primary care database analysis. BMJ Open Respir Res. 2019 Sep 17;6(1):e000458. doi: 10.1136/bmjresp-2019-000458. eCollection 2019.

Mathioudakis AG, Janssens W, Sivapalan P, Singanayagam A, Dransfield MT, Jensen JS, Vestbo J. Acute exacerbations of chronic obstructive pulmonary disease: in search of diagnostic biomarkers and treatable traits. Thorax. 2020 Jun;75(6):520-527. doi: 10.1136/thoraxjnl-2019-214484. Epub 2020 Mar 26.

Pouwels KB, Dolk FCK, Smith DRM, Robotham JV, Smieszek T. Actual versus 'ideal' antibiotic prescribing for common conditions in English primary care. J Antimicrob Chemother. 2018 Feb 1;73(suppl_2):19-26. doi: 10.1093/jac/dkx502.

Sapey E, Stockley RA. COPD exacerbations . 2: aetiology. Thorax. 2006 Mar;61(3):250-8. doi: 10.1136/thx.2005.041822.

Al-Jaghbeer MJ, Justo JA, Owens W, Kohn J, Bookstaver PB, Hucks J, Al-Hasan MN. Risk factors for pneumonia due to beta-lactam-susceptible and beta-lactam-resistant Pseudomonas aeruginosa: a case-case-control study. Infection. 2018 Aug;46(4):487-494. doi: 10.1007/s15010-018-1147-z. Epub 2018 May 11.

Hu L, Shi Q, Shi M, Liu R, Wang C. Diagnostic Value of PCT and CRP for Detecting Serious Bacterial Infections in Patients With Fever of Unknown Origin: A Systematic Review and Meta-analysis. Appl Immunohistochem Mol Morphol. 2017 Sep;25(8):e61-e69. doi: 10.1097/PAI.0000000000000552.

Ding X, Wu X, Yu C, Hu S. Value of C-reactive protein measurement in hospitalized patients with acute exacerbations of chronic obstructive pulmonary disease. Med J Wuhan University. 2006;27(5):660-3.

Hurst JR, Donaldson GC, Perera WR, Wilkinson TM, Bilello JA, Hagan GW, Vessey RS, Wedzicha JA. Use of plasma biomarkers at exacerbation of chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2006 Oct 15;174(8):867-74. doi: 10.1164/rccm.200604-506OC. Epub 2006 Jun 23.

Llor C, Moragas A, Hernandez S, Bayona C, Miravitlles M. Efficacy of antibiotic therapy for acute exacerbations of mild to moderate chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 2012 Oct 15;186(8):716-23. doi: 10.1164/rccm.201206-0996OC. Epub 2012 Aug 23.

von Dach E, Albrich WC, Brunel AS, Prendki V, Cuvelier C, Flury D, Gayet-Ageron A, Huttner B, Kohler P, Lemmenmeier E, McCallin S, Rossel A, Harbarth S, Kaiser L, Bochud PY, Huttner A. Effect of C-Reactive Protein-Guided Antibiotic Treatment Duration, 7-Day Treatment, or 14-Day Treatment on 30-Day Clinical Failure Rate in Patients With Uncomplicated Gram-Negative Bacteremia: A Randomized Clinical Trial. JAMA. 2020 Jun 2;323(21):2160-2169. doi: 10.1001/jama.2020.6348.

Huddy JR, Ni MZ, Barlow J, Majeed A, Hanna GB. Point-of-care C reactive protein for the diagnosis of lower respiratory tract infection in NHS primary care: a qualitative study of barriers and facilitators to adoption. BMJ Open. 2016 Mar 3;6(3):e009959. doi: 10.1136/bmjopen-2015-009959.

Ko FW, Ip M, Chan PK, Ng SS, Chau SS, Hui DS. A one-year prospective study of infectious etiology in patients hospitalized with acute exacerbations of COPD and concomitant pneumonia. Respir Med. 2008 Aug;102(8):1109-16. doi: 10.1016/j.rmed.2008.03.019. Epub 2008 Jun 24.

Brink A, Alsma J, Verdonschot RJCG, Rood PPM, Zietse R, Lingsma HF, Schuit SCE. Predicting mortality in patients with suspected sepsis at the Emergency Department; A retrospective cohort study comparing qSOFA, SIRS and National Early Warning Score. PLoS One. 2019 Jan 25;14(1):e0211133. doi: 10.1371/journal.pone.0211133. eCollection 2019.

Gillespie D, Francis NA, Carrol ED, Thomas-Jones E, Butler CC, Hood K. Use of co-primary outcomes for trials of antimicrobial stewardship interventions. Lancet Infect Dis. 2018 Jun;18(6):595-597. doi: 10.1016/S1473-3099(18)30289-5. No abstract available.

Kocks JW, Tuinenga MG, Uil SM, van den Berg JW, Stahl E, van der Molen T. Health status measurement in COPD: the minimal clinically important difference of the clinical COPD questionnaire. Respir Res. 2006 Apr 7;7(1):62. doi: 10.1186/1465-9921-7-62.

Offen W, Chuang-Stein C, Dmitrienko A, et al. Multiple co-primary endpoints: medical and statistical solutions: a report from the multiple endpoints expert team of the Pharmaceutical Research and Manufacturers of America. Drug information journal. 2007;41(1):31-46.

Moher D, Chan AW. SPIRIT (standard protocol items: recommendations for interventional trials). Guidelines for Reporting Health Research: a user's manual. 2014:56-67.

Schulz KF, Altman DG, Moher D; CONSORT Group. CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials. Trials. 2010 Mar 24;11:32. doi: 10.1186/1745-6215-11-32.

Butler CC, Gillespie D, White P, Bates J, Lowe R, Thomas-Jones E, Wootton M, Hood K, Phillips R, Melbye H, Llor C, Cals JWL, Naik G, Kirby N, Gal M, Riga E, Francis NA. C-Reactive Protein Testing to Guide Antibiotic Prescribing for COPD Exacerbations. N Engl J Med. 2019 Jul 11;381(2):111-120. doi: 10.1056/NEJMoa1803185.