High flow oxygen for vaso-occlusive crisis: a multicentre, prospective, randomised, multi-arm, multi-stage clinical trial (OSONE)

Abstract

Introduction

Sickle cell disease (SCD) is due to the mutation of haemoglobin (Hb), from HbA to HbS and characterised by recurrent vaso-occlusive crises (VOC), which can progress to acute chest syndrome (ACS), a leading cause of death in adults with SCD. Hypoxia is a key modifiable factor in the polymerisation of HbS and the pathogenesis of VOC. High-flow nasal oxygen (HFNO) delivers humidified gas at high oxygen concentrations and flow rates: the former may reverse sickling (metabolic effect) to accelerate VOC resolution and prevent ACS, while the latter may reduce the risk of ACS by mitigating hypercapnia and generating positive airway pressure that limits hypoventilation and atelectasis (pulmonary effect). The study hypothesises that HFNO is a safe and effective strategy for treating VOC and preventing secondary ACS, and will assess this using a multi-arm multi-stage (MAMS) trial design.

Methods and analysis

This is a prospective, multicentre, randomised, open-label controlled trial following an MAMS design with three phases and four arms: one control (low-flow oxygen) and three HFNO intervention arms with varying fraction of inspired oxygen levels (low, intermediate, high). The pilot stage will assess safety and feasibility, using the rate of cardiac and neurological events as the primary endpoint. In the activity stage, arms demonstrating acceptable safety will be compared for efficacy based on the rate of VOC resolution without complications by day 5, allowing selection of the most promising arm. The final efficacy stage will compare the selected HFNO strategy to control, with prevention of secondary ACS by day 14 as the primary endpoint. The study aims to enrol up to 350 VOC episodes in total.

Ethics and dissemination

The study has been granted ethical approval (CPP SUD MEDITERRANEE IV). Following the provision of informed consent, patients will be included in the study. The results will be submitted for publication in peer-reviewed journals.

Trial registration number

NCT03976180.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The trial is being conducted in specialised reference centres for sickle cell disease in France.

• The study employs an innovative adaptive seamless multi-arm multi-stage design.

• Blinding of patients and investigators to the assigned interventions is not feasible in this study.

Introduction

Sickle cell disease (SCD) is one of the most prevalent genetic diseases worldwide, often due to the mutation of haemoglobin (Hb), from HbA to HbS. SCD is characterised by recurrent vaso-occlusive pain crisis (VOC), which may evolve to acute chest syndrome (ACS), the most common cause of intensive care unit admission¹ and death² among adult patients with SCD. Management of VOC mostly involves a symptomatic approach including hydration, analgesics, transfusion and incentive spirometry, which was shown to prevent evolution towards ACS in a small size study (<30).³

The polymerisation of HbS is one major feature in the pathogenesis of vaso-occlusion. Among factors determining the rate and extent of HbS polymer formation, the hypoxic stimulus is one of the most potent and readily alterable. Current guidelines recommend oxygen therapy in patients with VOC in order to maintain a target oxygen saturation of 95%.^{4 5} Low-flow nasal oxygen (LFNO) is routinely used to achieve this normoxia approach, particularly in patients at risk of secondary ACS because they may experience acute desaturation. In contrast, various case series6,8 suggest a potential beneficial role of intensified oxygen therapy targeting hyperoxia for the management of VOC, particularly with the use of hyperbaric oxygen, but the latter is difficult to implement in routine clinical practice.

A recent high-flow nasal oxygen (HFNO) technology allows the delivery of humidified gas at high fraction of inspired oxygen (FiO₂) through nasal cannula.⁹ The FiO₂ can be adjusted up to 100% (allowing hyperoxia that may reverse sickling) and the flow can be increased up to 60 L/min (which generates positive airway pressure and dead space flushing, that may prevent atelectasis and opioid-induced hypercapnia). In patients with acute respiratory failure, HFNO has been shown to improve patient’s comfort, oxygenation and survival as compared with standard oxygen or non-invasive ventilation.⁹

We hypothesise that HFNO may be beneficial during VOC through two primary mechanisms: (1) effects on pulmonary mechanics: by reducing atelectasis and opioid-induced hypercapnia; and (2) metabolic effects: by limiting haemolysis through red cell unsickling. Given the uncertainty surrounding the relative contribution of these mechanisms to the prevention of secondary ACS, as well as the optimal oxygenation strategy (intense hyperoxia with 100% FiO_₂ versus moderate hyperoxia with 50% FiO_₂), we propose an adaptive seamless multi-arm multi-stage (MAMS) trial design. This approach will combine phase II and phase III into a single confirmatory clinical trial.

Objectives

Primary objective

The primary objective is to evaluate the efficacy and safety of HFNO during a VOC at risk of secondary ACS. The two primary hypotheses are: (1) HFNO increases the rate of uncomplicated VOC resolution by day 5; (2) HFNO reduces the incidence of secondary ACS by day 14.

Secondary objectives

The secondary objective is to assess whether HFNO alleviates VOC-associated morbidity, including follow-up of the patient’s clinical status.

Methods and analysis

Trial design

The trial design diagram is provided in figure 1.

Figure 1. Trial design diagram. FiO₂, fraction of inspired oxygen; HFNO, high-flow nasal oxygen; LFNO, low-flow nasal oxygen.

Experimental plan

We will conduct a randomised controlled superiority trial using an MAMS design, composed of three sequential stages and four parallel arms: one control arm and three intervention arms. This adaptive design enables the simultaneous evaluation of several experimental strategies against a common control and allows for the early discontinuation of arms that do not meet predefined safety, feasibility or efficacy criteria. All patient data from every arm and stage will be included in the final analysis.

The three stages of the trial are as follows:

• Pilot stage: this initial stage will assess the safety and feasibility of each intervention arm. An arm will only proceed to the next stage if it is considered both safe and feasible.

• Activity stage: this stage will provide an interim assessment of efficacy, with the primary endpoint being the rate of uncomplicated resolution of VOC at day 5. At the end of this stage, an interim analysis will be performed to identify the most promising experimental intervention. This selected intervention will then be compared with the control in the final stage, following a ‘pick the winner’ strategy.

• Efficacy stage: in the final stage, the selected intervention arm will be compared with the control arm, using the rate of secondary ACS at day 14 as the primary endpoint.

Rationale for the MAMS design

The MAMS trial design offers an efficient and adaptive framework for evaluating multiple interventions within a single study. It allows for the early discontinuation of arms that do not demonstrate sufficient promise based on intermediate endpoints, while continuing recruitment in the control group and the most promising intervention arm. Compared with traditional sequential trials, the MAMS approach can accelerate the generation of high-quality evidence, reduce the required sample size and lower both the cost and duration of the study. These advantages are particularly meaningful in the context of a rare disease such as SCD, where eligible patient populations are limited and the burden of research participation must be carefully balanced.

Eligibility criteria

Inclusion criteria

• Age ≥18 years.

• Diagnosis of major SCD (homozygous sickle-hemoglobin (SS), heterozygous sickle-hemoglobin C (SC), sickle beta zero thalassemia (Sβ⁰) or sickle beta plus thalassemia (Sβ⁺) genotypes).

• VOC is defined as acute pain or tenderness involving at least one body area (limbs, ribs, sternum, skull, spine and/or pelvis), requiring opioid treatment and not attributable to other causes.

• Intermediate-to-high risk of secondary ACS, as defined by the PRESEV (PREdictive SEVerity) score,¹⁰ meeting one of the following:

  • Reticulocyte count >216 G/L.

  • At least two of the following criteria: (1) Categorical Pain Score (CPS)>1 in the spine and/or pelvis, (2) leucocyte count >11 10⁹/L, (3) Hb concentration ≤90 g/L.

For patients on long-term hydroxyurea therapy, only one of the above criteria is required due to the known effects of hydroxyurea on Hb, leucocyte and reticulocyte levels.

• Signed informed consent.

• Affiliation to a social security scheme.

Non-inclusion criteria

• Presence of primary ACS at the time of inclusion, defined as the combination of a clinical sign (chest pain or auscultatory abnormality such as crackles and/or bronchial breath sounds) and a new pulmonary infiltrate identified on chest radiograph, thoracic CT or lung ultrasound.

• Ongoing VOC requiring parenteral opioid therapy for more than 72 hours at the time of inclusion.

• Known pregnancy or current lactation; all women of childbearing potential will undergo a pregnancy test prior to inclusion.

• Known cerebral vasculopathy or history of stroke due to Moya-Moya disease or persistent large-vessel stenosis or occlusion.

• Known ischaemic heart disease or typical anginal chest pain.

• Participation in another investigational drug study at the time of inclusion.

• Previous participation in this study within 28 days of the last randomisation.

• Known legal incapacity.

• Detained individuals or persons under judicial protection or involuntary incarceration.

• Anatomical abnormalities precluding placement of a nasal cannula.

Study interventions

Description and justification of groups

Control group

In the control group, standard LFNO will be administered via nasal prongs until hospital discharge or the onset of secondary ACS, with the goal of maintaining normoxia (target pulse oximetry saturation of 95%). This approach is consistent with current clinical guidelines and usual care practices.^{5 11 12}

Intervention groups

In recent years, HFNO has gained widespread use and is now considered standard of care in several clinical contexts, particularly for respiratory distress in intensive care settings.¹³ The potential clinical benefits of HFNO during VOC may involve two main mechanisms: pulmonary function-related effects and metabolic effects.

Pulmonary effects include enhanced comfort through heated, humidified gas and soft nasal prongs; improved mucus clearance and reduced epithelial injury; decreased work of breathing and atelectasis prevention via positive airway pressure; and improved ventilation efficiency with reduced opioid-induced hypercapnia through dead space washout.¹⁴

Metabolic effects are driven by hyperoxia and include red blood cell unsickling, reduced haemolysis, improved microvascular oxygen delivery, enhanced immune response and decreased mitochondrial oxygen consumption.¹⁵

These mechanisms are expected to lower the risk of secondary ACS during VOC. However, their relative contributions remain uncertain, as does the comparative benefit of intense (100% FiO_₂) versus moderate (50% FiO_₂) hyperoxia. The potential advantages of higher FiO_₂ may be offset by risks such as absorption atelectasis. To address this, three intervention arms will be tested:

• HFNO with low FiO_₂ (21–30%): to isolate the effect of improved pulmonary function under normoxia.

• HFNO with intermediate FiO_₂ (50%) to assess the combined effect of pulmonary support and moderate hyperoxia (FiO_₂ 50% for 24 hours, then 21–30% for 48 hours).

• HFNO with high FiO_₂ (100%) to assess the combined effect of pulmonary support and intense hyperoxia (FiO_₂ 100% for 24 hours, then 21–30% for 48 hours).

Oxygen delivery settings in the four arms of the study are provided in table 1.

Table 1. Oxygen delivery settings in the four arms of the study.

Arm	Control	Intervention	Intervention	Intervention
Intervention	LFNO	HFNO (Airvo2)	HFNO (Airvo2)	HFNO (Airvo2)
Oxygen delivery	0–3 L/min	FiO2: 21–30%	FiO2: 50% for 24 hours, then 21–30% for 48 hours	FiO2: 100% for 24 hours, then 21–30% for 48 hours
Saturation target	Normoxia (target SpO2 95%)	Normoxia (target SpO2 95%)	Moderate hyperoxia for 24 hours	Intense hyperoxia for 24 hours
Gas flow	–	30–60 L/min	30–60 L/min	30–60 L/min

FiO_₂, fraction of inspired oxygen; HFNO, high-flow nasal oxygen; LFNO, low-flow nasal oxygen; SpO_₂, peripheral oxygen saturation.

The AIRVO 2/3 device (Fisher & Paykel Healthcare, New Zealand) was selected for this trial because it includes an integrated turbine, allowing its use outside the intensive care unit (ICU), including in general hospital wards.

In the intervention arms, HFNO will be administered for 72 hours following randomisation. After this period, treatment will switch to standard LFNO, as in the control group. Temporary interruptions of HFNO will be allowed if necessary, particularly during meals, for toileting or in cases of poor tolerance to the nasal cannulas. HFNO will also be discontinued in the event of VOC resolution before day 3. If VOC recurs within the 72-hour treatment window, HFNO may be resumed. In all cases, HFNO will be definitively discontinued at day 3, and all patients will continue with LFNO thereafter.

HFNO will be permanently discontinued if any of the following conditions occur before completion of the 72-hour treatment period:

• The patient withdraws from the study or withdraws consent.

• The attending clinician decides to discontinue the intervention for safety reasons or in the patient’s best interest.

• The patient dies.

Management in all groups

In all groups, oxygen will be weaned before hospital discharge unless the patient is under long-term oxygen therapy. If secondary ACS occurs (primary outcome measure of the final analysis), the management of oxygen therapy will be left at the discretion of the treating physician, in accordance with current recommendations.^{5 11 12}

All included patients will receive standardised care for VOC throughout their hospital stay, in accordance with current recommendations and usual clinical practice.^{5 11 12} All participating sites are referral centres within the MCGRE (Filière de santé des maladies constitutionnelles rares du globule rouge et de l’érythropoïèse) network and follow a shared management protocol for VOC, ensuring consistency with national guidelines. This care includes bed rest, fluid replacement, oral alkaline water, folinic acid, routine use of incentive spirometry for prevention of ACS and multimodal analgesia with intravenous paracetamol and morphine via patient-controlled analgesia. In patients with fever (>38°C), empirical antibiotic therapy may be initiated following microbiological sampling. Transfusion therapy will follow standardised indications based on French national guidelines.^{11 12} For exchange transfusions, the volume of packed red blood cells will be adjusted according to the patient’s Hb concentration.¹¹

Endpoints

Primary endpoints by trial stage

The primary endpoint for Stage 1 (pilot) is the rate of cardiac and neurological events, defined as the occurrence of acute coronary syndrome, acute ischaemic stroke or seizure. This endpoint will be assessed at the end of the pilot stage and monitored throughout the study to provide cumulative safety data. These events were selected based on the known susceptibility of cerebral and coronary circulations to hyperoxia-induced vasoconstriction.¹⁵ Although oxygen toxicity to the central nervous system (Paul-Bert effect) typically requires pure oxygen at supra-atmospheric pressures,¹⁵ the occurrence of seizures will nonetheless be recorded. To minimise risk, patients with known cerebral vasculopathy or ischaemic heart disease will be excluded from the trial.

The primary endpoint for Stage 2 (interim analysis and treatment selection) is the rate of uncomplicated VOC resolution by day 5, defined as resolution without the need for transfusion, ICU admission or resulting in death. This endpoint, specific to the ‘activity’ stage, was chosen for its clinical relevance, sensitivity to treatment and correlation with the final endpoint. An improvement in VOC resolution is expected if secondary ACS is effectively prevented. This intermediate outcome occurs earlier and more frequently than secondary ACS and lies on the causal pathway, making it a suitable screening measure for treatment activity.¹⁶ VOC will be considered resolved when at least three of the following four criteria are met at two consecutive assessments: (1) absence of fever for ≥8 hours; (2) no intravenous opioid requirement in the last 8 hours; (3) the patient can walk or move without pain; (4) spontaneous pain is absent with a CPS of 1 or less.¹⁰

The primary endpoint for Stage 3 (final analysis) is the rate of secondary ACS at day 14, defined as the proportion of episodes with secondary ACS occurring within 14 days of randomisation. Secondary ACS is defined as the new onset of a clinical sign (chest pain or auscultatory abnormality such as crackles and/or bronchial breath sounds) combined with a new pulmonary infiltrate on chest radiograph, CT or lung ultrasound. The prevention of secondary ACS is the primary aim of the trial, given the combined pulmonary and metabolic effects of HFNO. Survival was not selected as the final endpoint due to the low in-hospital mortality associated with VOC and ACS in recent studies (<3%).¹⁷

Secondary endpoints

Each primary outcome described above for a given stage will serve as a secondary outcome in the other stages.

Additional secondary endpoints include the following:

• Volume of transfused red blood cells and volume of exsanguinated blood between randomisation and day 14.

• Pain intensity, assessed using the Visual Analogue Scale¹⁸ and/or the CPS between randomisation and day 14.¹⁹

• Duration of the VOC episode and number of VOC-free days by day 14.

• Reticulocyte count measured at day 2 and day 5.

• Arterial blood gas analysis performed at least once during the first 24 hours of treatment (when available).

• Cumulative dose of intravenous and subcutaneous opioids administered between randomisation and day 14.

• Number of complicated VOC episodes at day 14, defined as the occurrence of at least one of the following events: transfusion, exchange transfusion, mechanical ventilation, shock requiring catecholamine infusion, ICU admission or death.

• Follow-up at day 28, including:

  • Duration of hospital stay (calculated from randomisation to hospital discharge; patients still hospitalised at day 28 will be assigned a duration of 28 days).

  • Number of re-hospitalisations or emergency department visits for VOC or ACS within 28 days post-randomisation.

  • All-cause mortality at day 28.

Sample size calculation

The sample size calculation is based on data from both published and unpublished sources,²⁰ which provide baseline rates and outcome trajectories in patients hospitalised for VOC. These data inform the expected effect sizes in the control and experimental arms.

In the control group, a 50% failure rate of uncomplicated VOC resolution by day 5 (interim endpoint) and a 30% rate of secondary ACS by day 14 (final endpoint) are anticipated under current standard therapy. For the most promising experimental arm (to be selected at interim analysis), these rates are expected to improve to 25% and 14%, respectively. Less active experimental arms are expected to show rates of 35%/17% and 40%/20%.

Using the methodology developed by Friede et al²¹ and implemented by Parsons et al for adaptive seamless trials,²² and assuming a one-sided alpha level of 2.5%, a correlation of 0.5 between the interim and final outcomes, continued follow-up of patients from discontinued arms through the final endpoint, a sample of 40 episodes per arm for the interim analysis (total n=160) and 90 episodes per arm at Stage 3 (total n=180) provides at least 80% power to reject at least one null hypothesis and confirm the efficacy of one experimental arm. Sample size calculations were performed using 1000 simulations with the R package asd.²³

A total of 340 episodes will therefore be included. If the interim analysis has not been completed after at least 45 episodes have been enrolled in one or more study arms, the trial will be temporarily paused until the interim analysis is finalised. Therefore, up to 350 episodes may be enrolled.

Recruitment

Patients will be enrolled for 66 months starting in April 2020. The study timeline is as follows: (1) July 2019: grant award; (2) October 2019: approval by an independent ethics committee; (3) April 2020 to October 2025 inclusion of patients; (5) 2025–2026: the investigators will review the data and check for protocol violations; (6) 2026: the investigators will analyse the data, write the manuscript and submit it for publication.

Allocation of intervention and data collection

A computer-generated randomisation list will be prepared before enrolment of the first patient and updated after each trial stage. Randomisation will be conducted using an independent, web-based centralised system (CleanWeb), available 24 hours a day, 7 days a week. Block randomisation will be stratified by trial site to minimise imbalance between treatment arms.

A unique randomisation number will be assigned to each patient in the order of enrolment and recorded in the screening log. Randomisation will be fixed throughout the trial to ensure contemporaneous allocation of control and intervention patients.

• In Stage 1, patients will be randomised in a 1:1:1:1 ratio to LFNO, HFNO with FiO_₂ 21–30%, HFNO with FiO_₂ 50% or HFNO with FiO_₂ 100%.

• In Stage 2, the same 1:1:1:1 allocation will be used, provided no arm has been discontinued for safety or feasibility reasons.

• In Stage 3, randomisation will follow a 1:1 ratio between LFNO and the most promising HFNO regimen identified at interim analysis.

The Clinical Research Unit of Henri Mondor Hospital (Unité de Recherche Clinique [URC] Henri Mondor) will be responsible for managing the randomisation process.

Use of a re-randomisation design

The OSONE trial will employ a re-randomisation design, in which episodes of VOC, rather than individual patients, serve as the unit of randomisation. This design allows patients to be re-enrolled and re-randomised for each new eligible VOC episode they experience during the trial period.

Re-randomisation designs have been proposed for clinical settings where patients may undergo multiple episodes requiring treatment. Under this approach, the number of times a patient is enrolled is not fixed in advance but depends on the number of qualifying episodes they experience over the course of the study.²⁴ This design is particularly advantageous in trials with short follow-up durations, long recruitment periods and patient populations prone to recurrent events. In such contexts, re-randomisation can improve recruitment rates compared with standard parallel-group designs. This approach is especially relevant in the context of acute sickle cell pain crises, where symptoms recur and each episode typically requires distinct treatment. The re-randomisation design has been found appropriate in this setting.^{25 26}

Blinding methods and provisions to maintain blinding

Blinding of the intervention is not feasible, as HFNO with intermediate (50%) or high (100%) FiO_₂ leads to consistently elevated peripheral oxygen saturation (SpO_₂) values (typically 100%) compared with the control group and the HFNO-FiO_₂ 21–30% group (targeting 95%). Masking SpO_₂ readings from the clinical team would compromise patient safety and is therefore not permitted.

However, to minimise assessment bias, primary endpoints for the interim and final analyses will be evaluated in a blinded manner at the end of the study. A designated physician from the coordinating centre will perform this assessment based solely on the data recorded in the case report forms.

Statistical methods

We will apply the approach described by Friede et al,²¹ which is designed for trials where treatment selection at interim analysis is based on early, phase II-type outcomes, and confirmatory testing at final analysis relies on definitive, phase III-type outcomes. During the ‘activity’ stage, three experimental HFNO regimens (using low, intermediate and high FiO_₂) will be compared with a common control group receiving standard LFNO. Based on interim results, a single experimental treatment will be selected to proceed to the ‘efficacy’ stage, alongside the control arm.

Patients randomised to experimental arms that are discontinued at interim analysis will remain in the trial and be followed through to the final outcome. Subsequent patients will be randomised either to the control group or to the selected experimental arm.

For the analysis of primary and secondary endpoints, all observations will be treated as independent. Under a re-randomisation design, this method has been shown to provide asymptotically unbiased estimates of the treatment effect, maintain correct type I error rates and yield statistical power comparable to or greater than that of an equivalent parallel-group trial, provided the following conditions are met:²⁴ (1) patients are eligible for re-randomisation only after completing the follow-up period from their previous episode and meeting the inclusion criteria again; (2) randomisations for the same patient are performed independently for each eligible episode; (3) the treatment effect remains constant across all randomisation periods.

Primary endpoint

At the interim analysis, the primary endpoint for the ‘activity’ stage (defined as the rate of uncomplicated VOC resolution by day 5) will be assessed across all experimental arms and the control arm. Only one experimental treatment arm, specifically the one with the highest positive test statistic, will be selected to proceed to the efficacy stage alongside the control arm.

For the final analysis, all data on the primary endpoint of the ‘efficacy’ stage (defined as the rate of secondary ACS at day 14) collected across all stages for the selected experimental arm and the control arm will be included. The statistical approach will follow the framework proposed by Friede et al,²¹ in which p values are calculated using Dunnett tests at each stage, based on comparisons of each experimental treatment to the control for the final outcome. This method supports many-to-one comparisons, testing the global null hypothesis that all treatment effects relative to control are zero, against the alternative that at least one treatment has a beneficial effect. P values from all stages will then be combined using the weighted inverse normal method,^{27 28} within a closed testing procedure²⁹ to control the family-wise type I error rate at a prespecified level. In the OSONE study, this methodology will be applied to the specific case in which exactly one experimental arm is selected at the interim analysis.

Descriptive analyses and secondary endpoints

Descriptive statistics will be performed to summarise baseline characteristics, demographics and medical history by randomised group. Quantitative variables will be reported as mean±SD or median with IQR (25th–75th percentiles), depending on the distribution. Qualitative variables will be presented as counts and percentages.

Comparisons between groups will use the χ² or Fisher’s exact test for categorical variables, depending on expected cell counts. For continuous variables, t-tests or Mann-Whitney U tests will be used for pairwise comparisons and analysis of variance or Kruskal-Wallis tests for comparisons across more than two groups. Within-group comparisons across time points will use paired tests: McNemar’s test for categorical variables and paired t-tests or Wilcoxon signed-rank tests for continuous variables, as appropriate. A two-sided alpha level of 5% will be applied for all analyses relating to secondary endpoints. Time-to-event analyses will also be conducted for VOC resolution and secondary ACS occurrence.

Multivariable linear regression models adjusted for baseline characteristics and centre will be used to identify independent determinants of continuous secondary endpoints. Longitudinal changes (eg, in pain scores) will be analysed using mixed-effects linear models to account for repeated measures within patients.

For binary outcomes (including VOC resolution without complication at day 5, complicated VOC, re-hospitalisation and mortality at hospital discharge or day 28), multivariable logistic regression models will be applied, adjusting for baseline characteristics and centre. Model calibration and discrimination will be assessed using the Hosmer-Lemeshow test and area under the receiver operating characteristic curve, respectively.

Safety analyses will be stratified by treatment arm and timing of adverse events, with a particular focus on cardiac and neurological events. Missing or invalid data will be systematically reviewed using source medical records. In addition to complete case analyses, sensitivity analyses will be conducted after missing data imputation using the multiple imputation by chained equations (mice) method.

All statistical analyses will be performed using Stata V.16.1 (StataCorp, College Station, Texas, USA) and R V.4.+ (R Foundation for Statistical Computing, Vienna, Austria), under the supervision of Professor Etienne Audureau (Public Health Department, Henri Mondor Hospital).

Data monitoring

The Trial Steering Committee will oversee the conduct and progression of the study. Study monitors will perform regular on-site visits. Investigators or designated research staff at each centre will be responsible for daily patient screening, enrolment, protocol adherence and electronic case report form (eCRF) completion.

Patient and public involvement

This study was not developed with the input of patients or the public. However, we plan to conduct focused group discussions with patient survivors, hospital and trial staff at the end of the study.

Ethics and dissemination

Ethical approval

The study has been approved by an independent ethics committee (CPP SUD MEDITERRANEE IV) under the registration number: 190805.

Consent to participate

Patients will be included after signing a written informed consent form. If a patient is unable to participate in the consent process at the time eligibility criteria are met, inclusion may proceed with consent from a next of kin. If no substitute decision-maker is available, the patient may be enrolled under a deferred consent procedure. Once the patient has recovered or their next of kin has been contacted (whichever occurs first) consent to continue participation will be sought.

Dissemination policy

The findings will be published in peer-reviewed journals and presented at national and international meetings. The principal investigator and steering committee will be responsible for the communication, reports and publication of the results of the study. All reports will adhere to the Consolidated Standards of Reporting Trials reporting guidelines. Publication rules will follow the international recommendations as set out in The Uniform Requirements for Manuscripts (International Committee of Medical Journal Editors [ICMJE], April 2010).

Confidentiality

All data collected about the study participants and sent to the sponsor by the investigators (or any other specialised collaborators) during and after the clinical study will be anonymised.

Access to data

Investigators will make the documents and individual data required for monitoring, quality control and audit of the study available to designated personnel in accordance with the relevant legislation.

Data statement

The trial steering committee will facilitate the availability of study data in response to legitimate requests, in accordance with applicable regulations.

Discussion

We present the first prospective multicentre study evaluating the use of HFNO in the management of VOC in adult patients with SCD. Only patients at intermediate-to-high risk for secondary ACS, as defined by the PRESEV score,¹⁰ will be eligible. In this population, the likelihood of secondary ACS is high, justifying targeted preventive intervention.

SCD is the most common monogenic disease globally and in France. VOC is the leading cause of hospitalisation in SCD, and ACS remains the primary cause of mortality in adult patients. Currently, no treatment has been proven both safe and effective to abort VOC or prevent its progression to ACS. There is an urgent need for therapies that can both resolve VOC more rapidly and prevent secondary ACS. Moreover, earlier VOC resolution may lead to shorter hospital stays and lower healthcare costs.

Current guidelines recommend LFNO to maintain SpO_₂≥95% during VOC, although its clinical benefit remains unproven.^{5 11} In contrast, case reports and small case series have described clinical improvement with intensified oxygen therapy, including hyperbaric and normobaric^{7 30} hyperoxia. Benefits observed include pain relief during mild^{31 32} or refractory⁶ VOC, improved oxygen delivery, enhanced host defence and reduced mitochondrial oxygen consumption.¹⁵ While hyperbaric oxygen therapy has demonstrated clinical benefit in VOC, its widespread implementation is limited by technical and logistical constraints. In addition, hyperbaric hyperoxia is not without potential theoretical risks. These include increased oxidative stress through reactive oxygen species generation, impaired nitric oxide bioavailability, inhibition of hypoxic pulmonary vasoconstriction and the possible development of absorption atelectasis. Although the Lorrain-Smith effect (hyperoxia-induced pulmonary toxicity) has been demonstrated in animals, it has not been clearly observed in humans with SCD.^{30 33} Additional concerns include suppression of erythropoiesis^{30 34} and a theoretical rebound of sickled cells on oxygen withdrawal,³⁵ though these effects remain controversial³⁶ and inconsistently reported.

HFNO may offer a more practical alternative. It has been shown to improve comfort, oxygenation and survival in patients with acute respiratory failure compared with standard oxygen therapy or non-invasive ventilation.^{9 14 37 38} Its widespread use in paediatric and adult critical care settings has not raised major safety concerns. HFNO may offer a favourable balance of benefit and safety in the management of VOC through two main mechanisms. The first is metabolic: the delivery of adjustable FiO_₂ levels, up to 100%, may reduce haemolysis and promote red blood cell unsickling through hyperoxia, thereby enhancing tissue oxygenation. The second mechanism involves pulmonary function-driven effects: high flow rates of up to 60 L/min generate low-level positive airway pressure,³⁹ which may reduce the work of breathing, prevent atelectasis¹⁷ and flush upper airway dead space. This mechanism helps limit opioid-induced hypercapnia and improves overall gas exchange. However, the relative contribution of these mechanisms to ACS prevention remains uncertain, as does the potential superiority of intense hyperoxia (FiO_₂ 100%) over moderate hyperoxia (FiO_₂ 50%). The study’s innovative design allows for separate evaluation of pulmonary support alone (HFNO with FiO_₂ 21–30%) and in combination with moderate or intense hyperoxia.

We propose an adaptive, seamless, MAMS trial combining phase II and III into a single confirmatory study. This design eliminates the traditional delay between exploratory and confirmatory phases, accelerates the generation of robust data and increases statistical power while reducing the number of participants and overall costs. The inclusion of three experimental arms enhances the ability to identify the HFNO regimen with the most favourable risk-benefit profile.

The trial is ethically justified, as all patients will receive standard VOC care aligned with current guidelines.^{5 11} The intervention poses minimal additional risk, while offering potential for substantial clinical benefit. The benefit-risk assessment supports trial implementation.

We acknowledge that including all major SCD genotypes enhances generalisability but may introduce clinical heterogeneity. Notably, ACS in individuals with haemoglobin SC disease is less well characterised and likely less frequent, and the PRESEV score used to define intermediate-to-high risk of secondary ACS has only been validated in SS and Sβ⁰-thalassaemia, limiting its applicability in this subgroup.