Designing a multi-arm, multi-stage platform trial for venous leg ulceration – Venous leg ulcers: management and eradication, the VEIN platform study.

Patient and Public Involvement

All patients were involved in the development of this study. We conducted online and person-focus groups across the country to involve as many patients as possible. The main research questions, treatments, outcomes, and follow-up methods were developed on the basis of patient feedback. Two of the final study co-applicants were patients with a lived experience of venous ulcers who reviewed the final application to ensure that this was relevant to patients. The final study proposal was developed with methods that allow for easy input and feedback by patients with venous ulcers throughout the planned research.

Overarching aim: To identify the optimal design for research delivery for a large-scale, randomized, multi-arm, multi-stage platform trial assessing the clinical and cost-effectiveness of multiple interventions for venous leg ulcer healing and recurrence.

This research was a grant development exercise and a service evaluation exercise. No ethical approval was required for this research.

The research was conducted in five work streams (WS) with individual aims, as described below:

Work stream 1: Evidence synthesis and identification of research priority areas

A scoping review of the literature was planned to collect evidence on all published treatments for VLU. In addition, specific systematic reviews were planned to provide evidence for individual treatment modalities of particular interest. The aim of this WP was to identify potential treatments for inclusion in a platform trial.

Work stream 2: establishment of the platform development group

The aim of this WS was to identify key individuals to help inform the development of the platform. This was planned across tiers of involvement.

1. The executive group informed the day-to-day development of the platform via weekly meetings, including:

2. Core groups directly informed the development of the platform trial through regular meetings (monthly)

3. Specialist advisory groups (SPAGs) indirectly periodically informed platform development via surveys and questionnaires.

Work stream 3: establishment of key performance indicators (KPIs) and characteristics of the platform trial

The aim of WS3 was to identify the key characteristics of future platform trials, including

Work stream 4: consensus one the optimal design, methodology and delivery of the research

The aim of WS4 was to obtain a consensus on the final design and methodology based on the results of the previous work packages:

Work stream 5: finalising the funding application

The aim of WS5 was to deliver a finalized funding application for the proposed platform trial to the National Institute for Health and Care Research (NIHR) in November 2023.

Design and theoretical framework

The study employed a mixed methods approach, including evidence synthesis (WS1), interviews, questionnaires, focus groups (WS 2–3), statistical simulation to identify the most efficient design and sample size calculation, and model scoping and simulation for cost-effectiveness (WS4). We modelled our platform trial development according to published practical guidelines¹⁶ for platform studies.

The initial work streams were proposed following meetings with lay co-applicants with a lived experience of VLU. The five WS were further developed with weekly meetings by the executive group with regular input from other members of the platform trial governance (core groups, SPAGs). We ensured widespread representation with patients and caregivers from across the UK and international experts involved in the development of the platform. Methodology expertise was sourced from three clinical trial units (CTU: Imperial, Leicester, Edinburgh), with the support of a fourth unit with expertise in designing platform trials (medical research council [MRC] CTU at UCL) and a strong track record of research delivery. In addition, partners from the Vascular Society of Great Britain and Ireland, Royal College of Surgeons of England, Venous Forum of the Royal Society of Medicine, European College of Phlebology, Vascular Research UK, and Vascular and Endovascular Research Network supported this research.

Work stream 1: Evidence synthesis and identification of research priority areas

The first WS started upon the confirmation of funding. A scoping review was performed using the following databases: EMBASE, MEDLINE, Cochrane database, and clinicaltrials.gov using Medical Subject Headings (MeSH) terms pertaining to VLU healing and recurrence, permitting the identification of:

Following the scoping review, further in-depth systematic reviews and meta-analyses were conducted on interventions that were felt to be possible candidates for the trial arms to assess the existing evidence base in greater detail.

NHS library resources were used for all reviews. Team members with expertise in scoping and systematic reviews performed the searches, title, and abstract screening, including searches of reference lists. The analysis was performed, and the results were disseminated to our lay co-applicants.

Work stream 2: Establishment of the platform development group

Concurrent with the evidence synthesis, a platform development group was established. The co-applicants were keen to obtain a broad range of inputs during platform development, while also being aware of the need for robust governance to ensure input could be efficiently and meaningfully incorporated into trial development.

We established three tiers of involvement in the platform development:

Work stream 3: definition of Key Performance Indicators (KPIs) and platform trial characteristics

We performed online focus groups and interviews with the core patient group to inform them of the key priorities from the patient’s perspective. The core patient group, comprising volunteers identified in the participating institutions, participated in two online focus groups, where priorities for the research and possible trial arms were discussed. Although we aimed to maximize diversity in the group, we found that individuals who volunteered for online focus groups were highly educated and well-IT-versed. We developed concerns that the group was not diverse enough from a socioeconomic perspective. Previous research has shown that VLUs are often found in patients experiencing socioeconomic deprivation. Therefore, we decided to develop additional face-to-face patient and public involvement (PPI) events in London and Wakefield to ensure socioeconomic diversity. Similar to the core patient group, these participants were recruited in the outpatient clinic and varicose vein lists of collaborating institutions, but were provided with the option to participate in in-person meetings. These were attended by 20 individuals, improving the diversity of our patient group.

The core clinical groups routinely met on a bi-monthly basis to discuss the results of WS 1 and develop the key platform characteristics with periodic input from the SPAGs via online questionnaires. The following areas were e discussed:

Work stream 4: consensus on the optimal design, methodology and delivery of the research

In WS4, we reached a consensus on the research design. This was performed via online meetings between the executive and core groups with input from SPAGs. During this WS, we finalized the design of the research, including the PICOT framework (Population, Intervention, Comparator, Outcome, Timeframe), establishing the KPIs to assess research delivery and milestones, consensus on how patients are identified and approached efficiently based on unique regional and national characteristics, identification of success and failure criteria for each intervention relating to safety, considerations regarding recruitment challenges, clinical and cost effectiveness, consensus on enablers and barriers for research delivery within the expected timelines, and budgets creating a vehicle for efficient and timely dissemination throughout trial delivery.

From a methodological perspective, we established analysis methods for key outcomes using a simulation-based, frequentist approach to investigate and optimize trial design, considering adaptive elements such as timing and number of interim assessments and corresponding stopping rules, discussing how national registries or cohort studies could be used to streamline trial delivery, and integration of electronic healthcare records for long-term outcome assessment.

As part of this work package, we also finalized patient-facing interfaces for platform trial information for the public. This included the development of a dedicated VEIN website with information on the platform methodology and platform arms, as well as a dedicated infographic to support patient and public engagement, promotion, and dissemination of the research.

Work stream 5: finalising the funding application

In this WP and based on the information from WS 1–4, we prepared an application for the NIHR November 2023 deadline.

Data collection and data analysis

Interviews, focus groups, meetings, and free-text content of questionnaire responses were analyzed by thematic analysis using a mapping approach by team members with qualitative research expertise. The statistical analysis plan was finalized during the platform development process, including the integration of large dataset analytics. Cost-effectiveness analysis was performed by two experienced health economists, including the development of a schematic for a core structure for an economic model that could be adapted for a variety of interventions in VLU, with key health states based on outcomes of interest identified in the WS.

Work stream 1: evidence synthesis

We performed a scoping review of the literature, systematically searched MEDLINE, EMBASE, Web of Science, and CENTRAL databases, and manually searched reference lists (36,365 papers, 2,030 eligible). The scoping review identified six overall treatment approaches: conservative, pharmacological, interventional, preventive, service organization, and miscellaneous (Figure 1). There were no published or ongoing platform/complex RCTs integrating community, primary, or secondary care. We identified a protocol paper for a study exploring the use of pentoxifylline; however, this was in combination with another oral agent and not in the context of MAMS setup.

Figure 1. Results of scoping review.

The results were discussed in workshops and focus groups attended by platform governance members including patients and caregivers. Following consultation, we identified four priority research questions.

Further in-depth systematic reviews were performed or, if present, reviewed to obtain information on potential trial arms. Specifically, the results revealed the following:

Work stream 2: Establishment of the platform development group

We identified all key stakeholders.

Work streams 3 and 4: Establishment of KPIs, priorities for the research and consensus on the research

Study design: The finalized platform trial comprised three multi-arm, multi-stage (MAMS), multicenter, prospective randomized controlled trials with an internal pilot. The trials were conducted across three domains: presence of an active ulcer for < 6 months, presence of an active ulcer for ≥ 6 months, and presence of a healed ulcer. This was believed to provide the greatest opportunity for inclusion in all patients with a history of VLU.

Participants: Adults aged ≥ 18 years with active ulceration or healed ulceration without peripheral arterial disease (palpable pedal pulses or ankle brachial pressure index ≥ 0.8). Domain specific inclusion criteria: Domain 1: presence of active VLU present for < 6 months Domain 2: presence of active VLU ≥ 6 months; Domain 3: healed VLU.

Intervention:

Comparator: standard of care (SoC) NHS

Depending on the domain, intervention and comparators (Table 1):

Outcomes:

Primary outcome: Active ulceration (C6) and percentage healed (index ulcer) at 6 months. In healed ulceration (C5), the time to ipsilateral leg ulcer recurrence.

Secondary outcome: Quality of life (EQ5D-5L), health economic and implementation science evaluation, adverse events (AE), serious adverse events (SAE), and suspected unexpected serious adverse reactions (SUSAR) as per the MHRA guidelines for CTIMPs.

Study design: The proposed VEIN platform trial comprised three MAMS domains running concurrently. Each domain assesses SoC against a drug-based regimen (Pentoxifylline), domain-specific intervention (Prontosan antibacterial wash, endovenous ablation, duplex surveillance), and combination therapy consisting of drug and domain-specific treatments (Figure 2). Within the domains, participants will be randomized in a 1:1:1:1 ratio.

Setting: The study will be delivered in community, primary, and secondary/tertiary care, both in terms of recruitment and delivery of the trial interventions.

Figure 2. Final design of the platform trial.

Inclusion/ exclusion criteria: Participants aged ≥18 years with a history of VLU recruited from community, primary, or secondary/tertiary care. All eligible patients will be recruited without exclusion based on their socioeconomic or ethnic backgrounds. Domain specific criteria include: Domain 1: Presence of active VLU (C6) present for < 6 months; Domain 2: Presence of active VLU (C6) present for ≥ 6 months; Domain 3: Patients with healed VLU (C5). Exclusion criteria: Patients without an active or healed ulcer, and patients with a diagnosis of peripheral arterial disease (impalpable peripheral pulses or ABPI <0.8).

Technologies assessed:

Pentoxifylline: this is a phosphodiesterase inhibitor, which reduces inflammation, blood viscosity, and platelet aggregation. At a dose of 400 mg three times a day, VLU healing rates improved by 21% when used as an adjunct to compression and by 23% with medication alone¹⁷. NICE mentions pentoxifylline as a treatment adjunct for VLU, although its use is unlicensed. Existing evidence is of variable quality, and high-quality studies are required to assess the clinical and cost-effectiveness of this treatment.

Antibacterial wash: Prontosan® is an antibacterial preparation (Betaine and polyhexanide) available in NHS practice indicated for the prevention and removal of biofilms in VLU. The evidence for wound washing is heterogeneous, and there is insufficient evidence to suggest that one approach is superior to the other. The NICE medical technology consultation document found that Prontosan® showed promise in chronic ulceration, but recommended that RCTs be performed to assess the clinical and cost effectiveness of this wound washing approach²³.

Endovenous ablation: This is the standard of care for ulcers for < 6 months, as recommended by national and international guidelines based on the results of the EVRA trial. In EVRA, ulcers > 6 months were excluded, resulting in 1772 (27%) of the screened population being excluded from the trial⁹. There are no data on the clinical and cost-effectiveness of endovenous ablation for ulcers for > 6 months.

Duplex ultrasound surveillance: Duplex ultrasound is the gold standard assessment tool for patients with venous diseases. It provides details on the anatomy and function of the veins in the lower limb. After the intervention, the majority of VLU patients were discharged to their GPs and did not receive follow-up. It is unknown whether intervening in the refluxing veins can prevent ulcer recurrence. The ESCHAR trial found that surgical removal of refluxing veins reduced VLU recurrence rates at 12 months^10,11. Performing duplex ultrasound surveillance and treating any refluxing veins prophylactically may prevent VLU recurrence.

Participants would be randomized to an intervention/comparator according to their domain (clinical presentation, stage C6, or C5).

Domain 1: medication only (pentoxifylline) in addition to SoC; antimicrobial wash only (prontosan) + SoC; combination (pentoxifylline and prontosan) + SoC. Patients will be eligible for superficial endovenous ablation as part of the SoC, and this will not be withheld.

Domain 2: medication only (pentoxifylline) + SoC; superficial endovenous ablation only + SoC; combination (pentoxifylline and superficial endovenous ablation) + SoC.

Domain 3: medication only (pentoxifylline) + SoC; duplex ultrasound surveillance only + SoC; combination (pentoxifylline and surveillance) + SoC.

Identification of additional platform trial arms: In WS3 and WS4, we not only identified the technologies to assess in the first version of the platform trial but also potential technologies to assess in future arms. Statin therapy, deep venous intervention, and skin substitutes were identified as the next tier of interventions to be assessed in the proposed platform trial.

In addition, we developed a process for identifying and selecting additional trial arms. We aimed to include a periodical (six monthly) assessment of the literature to identify the latest technologies in venous ulcer care via regular systematic reviews. The results of the reviews would be presented to the VEIN prioritization committee and undergo the process described in Figure 3 to permit the prioritization of the proposed additional arms.

Figure 3. Process for future adoption of VEIN interventions.

Sample size, recruitment: For all three domains, sample sizes were calculated at 90% power and (one-sided) 2.5% significance rate, with a correction for multiple testing applied to account for each therapy being assessed twice within each domain (once individually and once as part of a combination therapy). The resulting family wise error rate (FWER) for each arm was set at 1.25%.

For Domain 1 we estimated an 86% 6-month healing rate with early ablation and 70% without ablation alongside a 10% absolute rate of improvement at 6 months based on the NICE estimate of wounds healing 1.46 – 1.66 times faster with antimicrobial washes. To investigate a 10% increase in healing rate with a 4-arm, 4-stage MAMS, adjusting for a 10% withdrawal rate and inflation to FWER via interim efficacy testing, a maximum of 2244 patients were estimated.

For Domain 2, we anticipated a 40% healing rate in the SoC arm after 6 months, with an estimated 15% absolute treatment effect, as supported by published data²⁴. Based on a SoC 6-month healing rate of 40%, to investigate a 15% increase within a 4 arm, 4 stage MAMS adjusting for a 10% withdrawal rate and inflation to FWER via interim efficacy testing, a maximum of 1320 patients would be required.

For Domain 3, we estimated recurrence rates of 12% in healthy VLU patients with previous interventions. The recurrence rate with compression alone was approximately 30%. The National Wound Care Strategy Programme (NWCSP) anticipated a 19% absolute recurrence rate based on an estimated 58% of patients with a history of VLU undergoing intervention¹². Using a time-to-event approach and a SoC 1-year recurrence rate of 19%, a 4 arm, 3 stage MAMS trial to detect a hazard ratio of 0.75 (equivalent to a 4.6% absolute reduction in 1-year recurrence) with a 10% withdrawal rate, would require a maximum of 3395 patients.

Sample sizes were initially derived using the NSTAGE and ARTPEP packages in STATA v17.0 and were supplemented by additional simulation work to optimize both the number and timing of interim analyses.

Measurement of costs and outcomes: We planned to develop a platform-wide health cost-effectiveness core model, adapted and reused as required²⁵ for various analyses based on NICE Reference Case specifications²⁶ with flexibility built into the models to allow alternative perspectives and analyses using modular cost and health outcome sets for each CEAP stage. Each domain would contribute data to the wider model, informing the care pathway section they relate to as they report, grow, and refine model outputs over time with unit costs from standard UK sources. A user manual and version control was planned to support staff succession and the creation of a legacy asset. We planned to report model results (QALY based on EQ-5D-5L and costs) alongside within-trial analyses for each domain when it demonstrated superiority or concluded the follow-up. The initial model structure was derived through consultations with clinical experts, an initial broad literature review, and ad hoc focused (non-systematic) reviews to populate specific parameters. Uncertainty was assessed using sensitivity, scenario, and probabilistic sensitivity analyses. To optimize data synthesis and generalizability, a common resource use survey was used in all trials, complemented by Hospital Episode Statistics (HES) and mortality records.

Time / Follow up: Telephone consultations from the core lab were planned at the following time points: Domain 1:6, 12, 18, and 24 weeks. Domain 2: 6, 12, 24 weeks. Domain 3:12, 24, and 52 weeks with further follow-up at 12 weekly intervals up to a maximum of 48 months. Long-term follow-up would be performed via data linkage with HES, ONS, and primary care databases (e.g., CPRD). At follow-up: medication check, eCRF completion, EQ5D completion, review and reporting of AEs, SAEs, SUSARs, drug accountability, and compliance. Throughout the trial, ad hoc ulcer healing and recurrence assessment (photographic evidence from patients remotely or face-to-face assessment where required), implementation work, and PPI (facilitated by the implementation team, PPI leads, and the Equality and Diversity lead) were conducted.

Statistical analysis: This study was planned based on the intention-to-treat principle. For domains 1 and 2, the primary outcome (proportion of patients with index ulcer healing within 6 months of randomization) was assessed using a logistic regression model. For Domain 3, the primary outcome (time to ipsilateral limb ulcer recurrence from randomization) was assessed using a Cox regression (proportional hazards) model.

Secondary (efficacy) outcomes Quality of life (EQ-5D-5L) data were analyzed by comparing SoC against each trial arm individually using ANCOVA adjusted for baseline, stratification variables, and key clinical covariates. Cumulative data on harm were presented to the data monitoring committee at regular intervals.

A detailed statistical analysis plan was planned.