Introduction
Depression and the need for novel therapeutics
Depression negatively impacts quality of life and imparts emotional, physical and economic consequences. Depressive symptoms, such as low mood and anhedonia, are associated with increased all-cause mortality and cardiovascular disease.1 The annual economic impact of depression in the USA exceeds US$90 billion.2 Of the 9 million Americans who require medication treatment for depression each year, nearly a third are considered treatment-resistant.2 Treatment-resistant depression (TRD), commonly defined as failure to achieve remission after adequate trials of two antidepressants, is a leading cause of disability,3 suicide4 5 and dementia in older adults.6–8 The prevalence of late-life TRD (LL-TRD) will likely increase as the population ages.9 10 Remission rates with conventional antidepressants are low in LL-TRD,11 and long-term outcomes are dismal, with high recurrence rates in spite of the currently available treatment options.12 13
Current pharmacotherapies for depression target one or more neurotransmitter systems as first-line options,14 with remission in 55%–81% of older adults with depression.15–17 Novel treatments targeting core pathophysiology are desperately needed for the large population of patients who do not benefit from traditional depression treatment strategies. While electroconvulsive therapy18 19 and transcranial magnetic stimulation20 may be viable options for some patients, there remains an unmet need for unique treatment modalities. A potential target for these therapies is sleep disruption,21 which contributes to refractory depression,22 impaired attention and executive dysfunction.23
Sleep, cognition and depression in the ageing brain
Sleep is critical for daily physical and mental restoration. It is composed of rapid eye movement (REM) and non-rapid eye movement (NREM) sleep, each with distinct physiological roles. Slow wave sleep (SWS)/NREM stage 3 (N3)24 is associated alleviation of neurohumoral stress responses.25 Several crucial processes occur during SWS, including synaptic remodelling and memory replay.26 These promote plasticity and memory consolidation.27 Accordingly, SWS enhancement can improve memory.28 29
SWS may be particularly important for maintaining cognitive ability as we age. Slow waves have been shown to regulate the chemical environment within the brain through the glymphatic system30 by stimulating cerebrospinal fluid waves during sleep.31 Proper functioning of this system may be protective against Alzheimer’s disease and related dementias by flushing out amyloid-beta30 32 and tau33 proteins. An inverse relationship exists between the deposition of dementia-associated proteins and measures of SWS in older adults.34 Thus, ensuring sufficient SWS may promote healthier brain ageing. Chronic SWS deficits may damage circuitry underlying the regulation of sleep and circadian rhythms, compounding the effects of poor sleep.
There is a well-established relationship between depression and sleep disturbances. SWS deficiency is a known risk factor for depression recurrence35 and suicidality.36 Numerous antidepressants are known to augment SWS, including lithium,37–39 trazodone,40 mirtazapine,41 sertraline,42 clomipramine,43 ketamine44 and agomelatine.45 SWS is characterised by large-amplitude electroencephalographic (EEG) oscillations known as delta waves in the 0.5–4 Hz frequency band (figure 1A). The squared amplitude, or power, in these slow waves can be expressed as the slow wave activity (SWA).46–48 In healthy individuals, SWA and SWS duration peak during the first N3 cycle of the night then diminish throughout subsequent cycles as REM sleep becomes more frequent. SWA can be viewed as a marker of synaptic plasticity.46–48
Electroencephalographic (EEG) slow waves and burst suppression evoked from a single participant. During natural sleep, the predominance of high amplitude EEG slow waves in the delta band (delta waves) define slow wave sleep (SWS)/stage 3 non-rapid eye movement sleep (N3) (A). EEG slow waves can also be induced during propofol general anaesthesia (B). With high doses of propofol, EEG burst suppression arises (C). Data were acquired from a 71-year-old woman during the pilot study.97
The decline in SWA across successive cycles of SWS is thought to reflect dissipation of sleep pressure and restoration of synaptic homeostasis.49 Reduced nighttime SWA, particularly in the first cycle,43 50–54 appears to be a core feature of depression pathophysiology, regardless of age.55 The shift in peak SWA can be quantified by the delta sleep ratio (DSR), which is computed by dividing the SWA of the first N3 cycle by the SWA of the second N3 cycle.56 Ratios >1.5 are typically observed in patients without depression. Changes in sleep architecture accompany antidepressant response. DSR has been used as a marker of antidepressant responsiveness57 and as a risk factor for recurrence.43 56
Depression-related changes in SWS are often accompanied by abnormal REM sleep. The delay in SWS that is characteristic of depression corresponds to REM intrusion into early sleep cycles. Greater REM durations are also often observed in depressed patients. Restoring SWS to early cycles normalises sleep architecture, allowing REM to dominate later sleep cycles. Later REM onset and reduced REM duration are observed with antidepressant response.
Poor sleep is also associated with cognitive dysfunction in older adults,23 with important implications for depressed patients. LL-TRD is often accompanied by impairments in executive functions,58 59 such as set-shifting (cognitive flexibility) and working memory.60 Poor executive function is associated with inadequate response to oral antidepressant treatment.12 58 61–65 Slowed processing speed, reduced alertness and impaired executive function are also associated with depression and sleep disturbances.66 67 Despite the reciprocal relationships between depression, cognitive dysfunction and sleep disturbances in older adults,68 no studies have specifically targeted SWS to treat TRD and improve cognitive function in these patients.
Propofol modulates sleep and mood
Propofol, a commonly used anaesthetic, induces dose-dependent EEG patterns with varied resemblance to natural sleep.69 70 With doses sufficient to induce unconsciousness, EEG slow waves emerge71 72 (figure 1B) resembling those of N3/SWS70 (figure 1A). Even higher doses elicit burst suppression,73–75 an EEG pattern characterised by episodes of suppressed cortical activity punctuated by bursts of mixed-frequency activity not seen in natural sleep (figure 1C). The electrophysiological similarities between propofol sedation and SWS may underlie its ability to fulfil SWS deficits in rodents.76 Patients often report subjective feelings of restorative sleep after propofol infusions, motivating its use as a treatment for refractory insomnia.77 78 In adults with chronic refractory insomnia, five nightly 2-hour propofol infusions from 22:00 to midnight enhanced SWS at least 6 months after therapy.78 Propofol-related changes in sleep architecture have not yet been linked to antidepressant effects in humans.
While antidepressant effects of propofol have been described, the role of SWS enhancement remains unexplored. After propofol anaesthesia, euphoria,79 80 amorous behaviour81 82 and elation83 may last for several hours.84 An antidepressant effect appears to be dose-dependent.85 Subanaesthetic infusions targeting plasma concentrations of 0.9 µg/mL were associated with no improvements in mood,86 while 10 high-dose propofol infusions, targeting at least 15 min of EEG burst suppression, correlated with antidepressant responses lasting up to 3 months.87 In rodent models, propofol pretreatment augments the antidepressant effects of ketamine88 and moderate-dose electroconvulsive therapy.89 There are numerous potential mechanisms underlying propofol’s antidepressant action. While the primary molecular mechanism of action underlying its hypnotic effects is GABA (γ-aminobutyric acid) agonism,90 91 propofol has additional potentially contributory molecular actions, including NMDA (N-methyl-D-aspartic acid) antagonism92 and AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) agonism.93 94 Recent animal studies demonstrate that propofol blocks dopamine transporters in the nucleus accumbens, leading to increased dopamine within this structure and reduced anhedonia.95 Interestingly, the nucleus accumbens contains a core set of neurons that regulate SWS.96 Thus, further work is needed to investigate relationships between propofol, SWS and depression.
We propose using propofol as a therapeutic probe to enhance SWS in older adults with TRD. Proof-of-concept was provided by a pilot study in which two participants were treated with two propofol infusions.97 One patient demonstrated EEG slow waves and minimal burst suppression during infusions. SWS enhancement was noted on postinfusion nights. The patient subsequently exhibited an antidepressant response that was sustained for at least several months. Another patient did not exhibit slow waves during infusions, postinfusion SWS enhancement, nor antidepressant response. These cases demonstrated that propofol can enhance SWS and improve depression for select individuals, motivating a subsequent clinical trial. Below, we present the protocol for the ongoing open-label phase I study, which will evaluate propofol’s ability to induce EEG slow waves and augment SWS in patients with LL-TRD.
Methods and analysis
Study design overview
Slow Wave Induction by Propofol to Eliminate Depression (SWIPED) (ClinicalTrials.gov NCT04680910) is a phase I clinical trial assessing the safety and feasibility of propofol infusions for SWS enhancement in older adults with depression (figure 2). There is no randomisation in this open-label single-arm investigation. The study tests the hypothesis that targeted intravenous propofol infusions in patients with LL-TRD induce SWA during sedation and augment subsequent sleep SWA. This single-centre trial takes place at Washington University in St. Louis Medical Center. Participants are undergoing two propofol infusions, with each infusion maximising the expression of EEG slow waves. Preinfusion and postinfusion sleep EEG recordings and assessments of mood and cognitive function are used to evaluate the correlations between propofol infusions and changes in sleep architecture, mood and cognition. This protocol is based on the latest approved protocol (version 3.1.5., 30 January 2023).
Study timeline. 15 study participants will undergo two 2-hour propofol infusions, high-density EEG, serial at-home unattended sleep studies, cognitive assessments and evaluations of depression, anhedonia and suicidality. Participants will complete a sleep diary. A blood biobanking study is optional, as marked by asterisks. C-SSRS, Columbia Suicide Severity Rating Scale; EEG, electroencephalography; HD, high density; MADRS, Montgomery-Åsberg Depression Rating Scale; MoCA, Montreal Cognitive Assessment; NIH, National Institutes of Health; PHQ-9, Patient Health Questionnaire-9; PVT, Psychomotor Vigilance Test; SHAPS, Snaith-Hamilton Pleasure Scale.
Study participants
Inclusion and exclusion criteria
This trial enrolled 15 older adults with TRD, defined as persistent depression despite adequate trials of two oral antidepressants (table 1). Exclusion criteria were chosen to maximise safety during propofol infusions (eg, severe left ventricular systolic dysfunction) and to avoid factors that impact SWS (eg, heavy alcohol consumption and certain medications).
Study inclusion and exclusion criteria for enrolment
Recruitment, enrolment and retention
Participants are recruited from the Geriatric Depression Clinic at Washington University in St. Louis, prior clinical trials and social media. Patients are compensated based on the number of study procedures completed, up to US$305 in total. Initial study eligibility is determined via phone screening. The phone screening employs the Patient Health Questionnaire-9 (PHQ-9) to enrol patients who have at least moderate depression (PHQ-9 ≥10).98 Eligible participants undergo a comprehensive review of their medical history, a physical exam, sleep apnoea screening and a urine drug screen. In addition, they undergo the psychiatric and cognitive assessments below. Participants are assigned a study ID with confidentiality of personal health information.
Withdrawal and replacement
Participants may withdraw from the study at any time. The study team may withdraw a participant if there is significant non-compliance with study procedures or if study procedures are poorly tolerated. Patients requiring repeated boluses of ephedrine (total dose >50 mg) for persistent hypotension would be replaced after study session termination.
Participants withdrawn or discontinued before both propofol infusions have been completed would be replaced and would not count towards recruitment target or milestones.
Intervention
Propofol dosing
The Eleveld pharmacokinetic model is used to maximise the stability of propofol effect site concentration in the brain, based on the participant’s sex, age, height and weight.99 To precisely target EEG slow waves while maintaining cardiopulmonary stability, infusion rates range up to 200 µg/kg/min. When possible, real-time target-controlled infusion modelling is implemented. Real-time EEG monitoring during infusions enables dose titration to maximise the expression of EEG slow waves. These visualisations include raw time-series, time-frequency spectrograms and spectral power measures. The presence and amplitude of EEG slow waves and burst suppression guide changes in propofol infusion rates. Dose titration during the first infusion informs subsequent propofol dosing, enabling rapid, stable target engagement of EEG slow waves during the second infusion, administered 2–6 days later.
Infusion visit procedures
Participants fast prior to infusions according to American Society of Anesthesiologists (ASA) guidelines. ACE inhibitors and angiotensin receptor blockers are held on infusion days to mitigate the risk of hypotension. Otherwise, patients generally continue their usual medication regimens throughout the study period.
Infusions take place at the Washington University in St. Louis Clinical Trial Research Unit. After placement of a peripheral intravenous catheter, propofol is infused for 2 hours. Patients are continuously monitored by a board-certified anaesthesiologist. Standard monitors (blood pressure, pulse oximetry, ECG and CO2 capnography) are employed in accordance with ASA guidelines. Supplemental oxygen is administered to all patients. Chin lift/jaw thrust manoeuvres are used to alleviate airway obstruction. Glycopyrrolate, intravenous fluids, phenylephrine and ephedrine are used to treat haemodynamic disturbances. The anaesthetic protocol would be discontinued prematurely if the patient continues to exhibit profound bradycardia, hypotension or hypoxaemia, or sustained cardiac conduction abnormalities refractory to intervention. After each infusion, patients are monitored by a nurse until they meet standard criteria for discharge from the postanaesthesia care unit.100 Participants must be taken home by a responsible adult per institutional policy and are counselled not to work or drive until the next day. There are no provisions for ancillary or post-trial care.
Data collection
Safety outcomes
Adverse events are reviewed at each study visit as primary safety outcomes. The most common adverse events associated with propofol infusions are pain on injection, bradycardia, hypotension, hypoxaemia, hypercapnia, airway obstruction and apnoea. The description of the adverse experience includes the time of onset, duration, intensity, aetiology, relationship to the study drug (none, unlikely, possible, probable, highly probable) and any treatment required.
High-density EEG
Patients undergo 64-channel EEG recordings at enrolment, during propofol infusions and at the second follow-up visit. A MagStim system and 64-channel GSN nets (MagStim, Whitland, UK) are used. Elefix conductive paste (Nihon Kohden, Tokyo, Japan) is injected within the Ag/AgCl electrode sensors to keep impedances below 100 kOhms/electrode. Data are acquired with a Net Amps 400 amplifier and Net Station Software V.5.0. An Axis P3364LV network camera, synchronised to EEG recordings, provides video for post hoc analysis. Eyes open and eyes closed recording periods (10 min each) are obtained.
At-home overnight sleep EEG recordings
Participants conduct at-home sleep EEG recordings using the third-generation Dreem device (Beacon Biosignals, Boston, Massachusetts, USA), a wearable, wireless EEG headband with dry electrodes positioned over frontal (F7, F8 and Fpz) and occipital (O1 and O2) locations.101 To maximise patient compliance and EEG signal quality, device usage and adjustment for proper fit are demonstrated during enrolment. Participants conduct at least two overnight recordings in the week prior to their first propofol infusion, to account for acclimation to device wear and user errors. They provide overnight recordings on the nights of the infusions. After the second propofol infusion, participants will record up to five additional sleep studies over 3 weeks. Participants use a portable wireless hotspot to allow for remote data quality monitoring of at-home sleep recordings.
Psychiatric assessments
Depression severity is evaluated at enrolment using the Montgomery-Åsberg Depression Rating Scale (MADRS).102 In addition, baseline suicidality and anhedonia are assessed on enrolment via the Columbia Suicide Severity Rating Scale (C-SSRS)103 and Snaith-Hamilton Pleasure Scale (SHAPS),104 respectively. The MADRS, C-SSRS and SHAPS questionnaires are repeated approximately 1, 3 and 10 weeks after the second propofol infusion (figure 2). The Feeling Scale105 are used during infusion visits to gauge participants’ affect within an hour before propofol administration and approximately 30 min after return of consciousness.
Cognitive assessments
The Montreal Cognitive Assessment (MoCA)106 is administered at enrolment and approximately 3 weeks after the second propofol infusion (figure 2). An alternate version of the MoCA is used for the post-treatment in-person assessment to prevent confounding due to recall of test items.
To examine cognitive function, feasibility of including the National Institutes of Health (NIH) Toolbox Cognition Battery65 and Psychomotor Vigilance Test (PVT)107 are assessed. Baseline testing occurs at enrolment, with follow-up at approximately 3 weeks after the second propofol infusion (figure 2). The NIH Toolbox Cognition Battery is a computer-based instrument that assesses fluid and crystallised cognition through the following five component tests (and tested cognitive domains): The Flanker Inhibitory Control Test (executive function-sustained attention and inhibitory control), the Dimensional Change Card Sort test (executive function-cognitive flexibility), the List Sorting Working Memory Test (working memory), the Picture Sequence Memory Test (episodic memory) and the Pattern Comparison Processing Test (processing speed). The PVT is used to assess vigilance/alertness in the context of poor sleep and psychiatric illness and can be administered on touchscreen devices.108 It requires the participant to respond as quickly as possible and to maintain sustained vigilance during random intertrial intervals of several seconds. The task generates a right skewed distribution in reaction time (RT), that can be assessed in terms of lapses (outliers) as well as overall metrics of speed (the reciprocal of RT). With sleep deprivation, there is an increased number of lapses (RT >500 ms).109
Assessment of circadian rhythms
Circadian rhythms are assessed over the course of the study. The Morningness-Eveningness Questionnaire110 is administered at enrolment to determine each participant’s baseline chronotype. Participants maintain a daily sleep diary throughout the study period to allow confirmation of bedtimes. Feasibility of 3 weeks of wrist actigraphy is being assessed.
Peripheral blood biomarkers
Optional blood sampling occurs at both infusion visits and at the follow-up visit approximately 3 weeks after the second infusion. Sampling on infusion visits occurs twice, before propofol administration and approximately 1 hour after infusion start. Following centrifugation, plasma samples are banked. Potential future approaches include proteomic analysis with the Olink Neurology 384 Platform, exploration of inflammatory cytokines (eg, interleukin (IL)-6, IL-8) and markers axonal injury (neurofilament light chain), genotyping of sleep/circadian/neuroinflammation loci (eg, adenosine deaminase or adenosine receptor typing) or transmitter receptors (eg, dopamine, P2X7R). Samples are also processed for storage in PAXGene RNA tubes for future differential gene expression and gene set enrichment analyses. Gene expression data may also be used to assess circadian health.111
Data analysis
EEG preprocessing
Deidentified data from overnight sleep recordings are preprocessed by the Dreem manufacturer and imported into MATLAB analysis platform. Through custom-written scripts for MATLAB (Mathworks, Natick, Massachusetts, USA), raw data are down-sampled to 250 Hz and frequency bandpass filtered at 0.5–50 Hz using EEGLAB,112 an open-source analysis toolbox.
High-density EEG data are preprocessed in EEGLAB.112 Data are low-pass filtered at 40 Hz prior to decimation to 125 Hz. A 0.1–0.6 bandstop filter is applied to remove respiratory artefact prior to high pass filtering at 0.5 Hz.34
Sleep scoring
European data format records are imported into Respironics Sleepware G3 software (V.3.7.1, Philips Respironics, Murrysville, Pennsylvania, USA) and evaluated by American Academy of Sleep Medicine (AASM) certified sleep technologists. Recordings are scored manually in 30 s epochs using modified AASM scoring rules previously developed for single frontal EEG analyses113 and described recently.113 114 The sleep stages are aligned with quantitative EEG analytical measures. N2 and N3 (sleep slow waves present in >20% epoch) epochs are used in quantitation of EEG metrics below. Eye movements in frontal EEG allow scoring of REM sleep epochs. Staff are blinded to clinical/cognitive outcomes during sleep staging and EEG quantitative analyses.
SWA and DSR
We quantify the expression of EEG slow waves during propofol infusions and N2 and N3 sleep. Multitaper methods are used for power spectral analysis using the MATLAB Chronux toolbox.115 Spectral estimates are based on 6-second non-overlapping time windows, time-bandwidth product of 3 and five tapers. Distributions of power in the 20–30 Hz and 1–4.5 Hz bands are used in a semiautomated fashion to reject artifactual epochs.34 SWA is calculated at 1 min intervals as the total EEG 0.5–4 Hz power/minute during propofol treatment (sedation SWA) and during overnight SWS (sleep SWA). Average SWA for each night is quantified during the first and second N2/N3 cycles, using custom-written MATLAB scripts.116 Secondary analyses rely on the DSR for each night of sleep, which is computed by dividing the average SWA of the first cycle by the average of the second cycle.97
Burst suppression
Burst suppression has previously been investigated as a potential marker or mediator of the antidepressant effects of propofol87 and electroconvulsive therapy. Total burst suppression duration for each infusion is quantified from infusion EEG data using a previously described voltage-based algorithm.97 117 118
Blinding
Masking of total drug infused and infusion rates is applied to patients and those performing depression, behavioural and cognitive assessments.
Endpoints
Table 2 lists the primary, secondary and tertiary endpoints. This trial primarily assesses the safety and feasibility of administering intravenous propofol infusions for SWS enhancement. Secondary endpoints include changes in suicidality, cognitive ability and sleep architecture. Tertiary endpoints assess the feasibility of acquiring propofol-associated changes in circadian rhythms, psychomotor vigilance, fluid cognition, affect, anhedonia and other depression symptoms.
Primary, secondary and tertiary endpoints of the phase I trial
Data management and statistical analysis
Data will be stored in REDCap and laboratory servers. The sample size was determined using pilot study data (two pretreatment and two post-treatment sleep recordings from one participant).97 We used a mean N3 duration (minutes) of 1.75 (pretreatment) and 10 (post-treatment) and a combined SD of 4.19. For paired measurements within participants, these estimates would only lead to a sample size of 6. This first participant had a >400% augmentation of her baseline N3 duration, which may be greater compared with others. Furthermore, we expect a 20% attrition such that targeting 15 participants seems compatible with a convenience sample size for gauging safety, feasibility and persistence of postintervention sleep SWA promotion. Descriptive statistics and regression analyses will be performed.
Patient and public involvement
Feedback from the National Institute of Mental Health (NIMH) Data and Monitoring Safety Board-D (DSMB-D) members and study participants guide the execution of this investigation and the planned phase II study. After the conclusion of the trial, the study team will provide participants with a summary of study outcomes.
Data sharing
We will share our data with the National Sleep Research Resource within 3 years after completion of the study. Furthermore, we will share our data on the NIMH Data Archive.
Discussion
The SWIPED study aims to enhance SWS with propofol as a novel approach towards depression treatment in older adults. In contrast to current depression treatments, which largely target monoaminergic neurotransmitter systems, we hope to engage endogenous neural circuitry to restore normal sleep architecture, thereby addressing a core pathophysiological feature of the illness. This is similar to efforts with other novel antidepressants such as the neurosteroid brexanolone,119 which is thought to correct pathophysiology associated with post-partum depression.
Phase I of the SWIPED study will assess the safety and feasibility of using propofol to generate slow waves during infusions and enhance subsequent SWS. Pending the results of this investigation, a future phase II clinical trial would assess the association of SWS enhancement with cognitive and clinical outcomes. Phase I will progress to a phase II study if: (1) less than 5% of serious adverse events are directly attributable to propofol infusions, (2) EEG slow waves are present for the majority of the infusion time in at least 60% of participants and (3) total sleep SWA is augmented in at least 40% of participants who complete the study. Phase II will entail a randomised placebo-controlled trial of 70 participants. We anticipate that this approach will further elucidate the relationships between depression, cognitive dysfunction and sleep disturbances and yield new avenues for treating TRD.
Key limitations of this study include its small sample size (n=15) and lack of control groups. We will not be able to compare propofol treatment to (1) no treatment or (2) other methods of SWS enhancement (eg, non-pharmacological enhancement). An intervention that strictly enhances SWS without potential independent mechanisms of antidepressant effect would more incisively address our hypothesis regarding the effect of SWS enhancement on depression. In addition, the findings may be challenging to generalise, given the advanced age of the study participants and the numerous exclusion criteria employed.
Ethics and dissemination
The study design and procedures are guided and approved by the NIMH DSMB-D, which convenes thrice per year. A board-certified anaesthesiologist with no professional conflicts of interest is serving as a clinical site monitor. Their recommendations for improving study conduct have been integrated. The study has been approved by the Washington University Human Research Protection Office (WU HRPO). Any changes to the study protocol are formally approved by both the NIMH DSMB-D and WU HRPO prior to implementation. Full informed consent is obtained from all enrolled participants by trained research staff (online supplemental file 1). This document has been drafted to comply with the Standard Protocol Items: Recommendations for Interventional Trials checklist (online supplemental file 2).
Supplemental material
Supplemental material
The findings of this study will be disseminated via peer-reviewed publications, presentations at scientific conferences and mass media. In addition, data will be stored in the NIMH data archive, which shares deidentified data with qualified researchers and provides summary data to the public. EEG data will be shared via the National Sleep Research Resource within 3 years of study completion.
Acknowledgments
We appreciate the ongoing assistance and feedback provided by study participants and nursing staff in the WU Clinical Trials Research Unit. Contributions from Alec Hester were appreciated. We thank Sherri McKinnon, Sara Fisher, C. Ryan King, Ashley Kennedy and the NIMH DSMB-D for regulatory oversight.
This post was originally published on https://bmjopen.bmj.com