Platelet-rich plasma injection for adults with acute Achilles tendon rupture: the PATH-2 RCT

Achilles tendon

The Achilles tendon is the largest and strongest tendon in the human body. It is the tendon of the gastrocnemius–soleus calf muscle complex. The Achilles tendon inserts to the calcaneum (heel bone), forming the major plantar flexor and stabiliser of the ankle joint. 1 The tendon is rounded and narrow in shape at its midpoint and finally fans out at the insertion (Figure 1). The tendon fibres rotate around 90° as they insert into the calcaneum. The narrowest part of the tendon is approximately 4 cm above the insertion.

FIGURE 1.

Anatomy of the Achilles tendon. The image on the left shows the left Achilles tendon projection at the posterior aspect of the calf and heel (1, calf muscles; 2, Achilles tendon; 3, Achilles tendon insertion to heel); and the image on the right is an illustration of the muscle–tendon unit (1a and 1b, gastrocnemius muscle medial and lateral heads; 1c, soleus muscle; 2, Achilles tendon fibres; 2a, gastrocnemius Achilles fibres; 2b, soleus Achilles fibres; 3, tendon insertion to the calcaneum).

The structure of the muscle–tendon unit allows for effective force to be generated when lifting the heel during locomotive activities such as walking and running. 1,2 The human Achilles tendon also stores and returns energy. 3 The tendon stretches in proportion to the force applied during the downwards motion of the body and then recoils to release most of the energy stored (74%) during the upwards movement.

Blood supply to the tendon is variable but is least in the middle (mid-portion). The mid-portion is the narrowest and a relatively avascular region within the tendon that corresponds to the most frequently injured area. Degenerative changes, common with advancing age and metabolic and chronic diseases, further compromise tendon vascularity and structure. 4 Degenerative changes in the tendon decrease collagen cross-linking and small repetitive tears weaken the tensile strength of the tendon. The Achilles tendon ruptures when it is subjected to a load that exceeds its mechanical capacity.

Achilles tendon rupture epidemiology

The Achilles tendon is the most commonly injured tendon in the human body, accounting for 20% of all tendon ruptures. 5 The incidence of this injury is rising, due to increased engagement in high-impact recreational sport, common particularly among men in the third and fourth decades of life. 6,7 Other factors that might increase the risk of injury include previous injury to the contralateral leg, direct injection of steroids or administration of systemic corticosteroids and fluoroquinolones. 8 In the UK, 11,000 injuries are reported each year. 9 In Denmark, an increase in incidence from 22.1 ATRs per 100,000 person-years in 1991 to 32.6 per 100,000 person-years in 2002 was reported by Gulati et al. 10 The incidence of ATR is 12–18 cases per 100,000 people per year for sedentary professionals. 11 A Cochrane review12 reported rehabilitation and work absence of 63–108 days, highlighting the socioeconomic burden of the injury.

The injury mechanism of ATR is mostly related to loading the tendon during weight-bearing physical activity. Typical acute injuries result from rapid force shifts to the lower leg during sports. This can result in partial or complete tear from direct (acute) trauma. During clinical assessment, patients often describe a history of feeling a direct blow localised to the posterior aspect of the ankle with a sudden onset of pain, swelling, bruising and difficulty with walking, usually during a sport activity. Physical examination of the area commonly shows indentation at the locations of the tendon rupture, soft-tissue swelling and tenderness, but these signs can be masked at the acute stage.

Spontaneous ruptures without a notable trauma, more common in ageing tendons, are typically associated with pre-existing chronic, pathological changes in the tendon. Degenerative changes, such as poor vascular supply, calcification, tendolipomatosis (replacement of tendon tissue with fat), necrosis (cell or tissue death), hypoxia and mucoid degeneration, have been shown to be present in ruptured tendons not related to sports. 13 Once an acute tear to the tendon occurs, the structural, biochemical and functional properties of the tendon are disrupted. Histopathological changes in the injured tissues have been reported. These include high vascularity, collagen disorganisation and hypercellularity relatively close to the ruptured site. There is a reduction in the number and diameter of type I collagen fibres, which are replaced with larger type III fibres. 8

Clinically, a common method of examination to assess the integrity of the Achilles tendon is to lay the patient prone on a plinth with their foot hanging over the edge. In this position, the tendon gap is usually palpable and easy to observe, and on squeezing the calf muscle bulk, ankle plantarflexion can be anticipated if the tendon is intact (Simmonds–Thompson test). 14–16 These methods of assessment may result in misdiagnosis when the tendon gap is filled with oedema or when a plantarflexion response is generated by other intact muscles that also plantarflex the ankle (i.e. tibialis posterior and the peronei). 17 However, significant plantar flexion weakness on directed push-off, abnormal or poor gait pattern and inability to perform a heel rise on the affected leg are often indicative of a complete tendon tear. 18 Diagnosis is usually made clinically from the history and physical assessments, but some clinical centres use ultrasound and magnetic resonance imaging to support diagnosis. There are also treatment protocols that direct management using imaging findings,10,18,19 but this is not currently routine practice in the NHS.

Healing of the tendon (repair and regeneration)

The healing of an injured tendon takes place in three overlapping, successive stages: inflammation, repair and remodelling. These cellular responses are directed by numerous growth factors that are upregulated and activated at various stages of the healing process. 13,20 In the inflammatory stage, haematoma forms and platelets release growth factors and bioactive proteins to attract inflammatory cells, including neutrophils. This haematoma and the inflammatory cells eventually organise into granulation and scar tissue. 13,20 In the repair stage, recruitment, activation and proliferation of tenocytes and fibroblast begins in the injury site. During this phase, neutrophil levels drop while macrophages continue to release growth factors that direct cell activity. In addition, intrinsic and extrinsic tenocytes synthesise and establish a network of extracellular matrix composed of type III collagen fibres. The remodelling stage starts 1–2 months post injury and persists for ≥ 12 months. Tenocytes and collagen mature and become aligned with the direction of stress. Over time, the synthesis of collagen type III fibres decreases, while type I fibres become dominant and fibrous scarring tissue remodel towards the original tendon tissue. 13,20

Although tendons have the ability to heal after injury, the natural process is slow and incomplete. 20 This limited healing capacity of tendons may result from a combination of relatively poor vascularisation and cellular turnover, changes to the matrix structure and the demanding mechanical environment of this connective tissue. 21–24 Poor vascularisation negatively affects the inflammation that is essential for the restoration of the mechanical and biological properties of tendons during the healing process. 25,26 Ageing tenocytes – the dominant cell type in tendons – tend not to be able to facilitate the proper differentiation of progenitor cells, resulting in ineffective healing. This leads to scar formation rather than tendon tissue regeneration, which results in reduced tensile strength. 13,20

The mechanical properties of the healed tendon do not fully recover, leading to weakness of the muscle tendon unit and risk of re-rupture. Even with the best contemporary management, tendon injuries give rise to substantial morbidity that lasts for many months and poses considerable challenges for clinicians and patients during the lengthy healing and recovery period. 27

Management of Achilles tendon rupture

The current treatment strategies for acute ATR are either surgical repair using sutures or non-surgical management by immobilisation of the ankle using a boot or cast. 19 Whether surgical or non-surgical management of ATR is the optimal treatment approach is uncertain. 7,19 The decision for the type of management is often based on patient-related factors such as timing of presentation, comorbidities, activity demands and previous tendon function. 28 Notably, there is a decline in the proportion of tendons treated surgically worldwide. 7

Open, minimally invasive or percutaneous techniques for operative repair seem to show quicker return to work and greater heel-rise height during physical assessments. 29 However, surgical complications including sural nerve injury, keloid formation, wound breakdown or infection are risks to be considered, although they may be reduced when percutaneous surgical repair is used. 29,30

Non-surgical management involves short periods (6–8 weeks) of ankle immobilisation starting in an equinus (plantarflexed) ankle position in a cast, rigid boot or ankle orthoses. 29 Recent developments of early functional rehabilitation and range of motion protocol with early weight-bearing – using controlled ankle motion boots – during non-surgical treatment have demonstrated similar surgical outcomes in terms of re-rupture rates (3–5%) and function. However, non-surgical management does not increase the rate of re-rupture or other complications compared with surgery. 31–33 The inclusion of functional rehabilitation as an adjunct to non-surgical management has not demonstrated improvement in biomechanical tendon properties or tendon elongation, which influence plantarflexion strength. 29,34 Although different surgical techniques and non-surgical approaches can be used, evidence shows similar outcomes. 35 These approaches are not known to alter the existing biological regenerative pathway of the tendon, so the lengthy rehabilitation, reduced function and re-rupture risk (3–5%) all remain. 36 Functional impairments identified by calf atrophy, heel-raise power compared with the opposite side, poor walking pattern (gait) and lower levels of physical activity are reported as outcomes of these management strategies. 37

Platelet-rich plasma

Although significant advancements in the clinical management of ATR have resulted in some improvements in functional outcomes,38 there has been little progress made with regard to the restoration of the structural and biomechanical integrity of an injured tendon to its original state. Thus, it is unsurprising that novel therapies are being explored to diminish the degree and duration of morbidity. Biological adjuncts such as PRP may augment the regeneration of a torn tendon. For example, platelet-derived growth factor (PDGF), a derivative of PRP, seems to upregulate the synthesis of cells in tendons that direct optimal cellular repair response, which results in less scar tissue formation, stimulates healing and improves biomechanics. 20 Evidence to support its clinical application in the management of musculoskeletal conditions remains inconclusive. 39

Platelet-rich plasma is an autologous derivative of whole blood that contains a higher concentration of platelets and other bioactive compounds. PRP is produced by separating the red blood cells from the platelets and other constituents in whole blood. The method of PRP preparation, such as centrifugation or filtration; the concentration and volume of platelets and white blood cells (leucocytes); the activation method; and growth factor levels have all been implicated in the local effects of PRP. 39–41 Unsurprisingly, the nature of the target tissue – acute, subacute or chronic – seems to affect local or systemic effects of PRP. 42

As part of the normal injury response to an acute tendon rupture, platelets immediately contribute to the blood clot that forms within the injured tissue. That response includes multiple growth factors released from the platelets’ α-granules interacting with neutrophils, macrophages and other cells, working synergistically to provide a wound healing mixture. This is part of a cascade within the injured tissue over a period ranging from minutes to an average of several days. 43,44 Active growth factors in PRP in the rupture site promote cell motility and recruit undifferentiated stem cells and surviving tenocytes, which then proliferate to start the healing process. 45 The upregulation of growth factors such as vascular endothelial growth factors (VEGFs) contained within PRP attracts vascular endothelial cells, which stimulate angiogenesis to restore normal tissue conditions. 46 PDGF, insulin-like growth factor 1 (IGF-1) and fibroblast growth factor (FGF) facilitate proliferation of cells and hasten the start of the remodelling phase. 47 Transforming growth factor beta (TGF-β) promotes the formation of fibrous tissue and the expression, organisation and maturation of collagen fibrils, enhancing the biomechanical strength of the tendon. 48

Although there is evidence suggesting that leucocyte-poor PRP may produce better outcomes than leucocyte-rich PRP (L-PRP), leucocytes contained in PRP seem to interact with the tissue of the tendon, promoting healing by stimulating tenocytes within the injured area to produce their own growth factors. 49 Furthermore, PRP assists the initiation of a cascade of steps that induce the synthesis of fibroblasts, collagen fibrils, matrix-degrading enzymes and extracellular matrix. This sequential release of proteins is amplified to recruit a diverse range of cell types to act on the injured tendon and disrupted blood vessels. So far, conclusions from basic laboratory testing have indicated that PRP could be a novel agent for tendon injury. 44 Augmenting the natural biological healing pathway to improve patient recovery in Achilles tendon injury may therefore have significant clinical implications on the health system at large by reducing demands for surgical repair, accelerating rehabilitation, improving physical activity participation and reducing re-injury rates.

Achilles tendon injury and platelet-rich plasma

Although PRP shows some promise as a treatment for promoting faster healing in tendons,20,50,51 the potential enhancement of healing demonstrated in laboratory studies is yet to be evidenced in the patient setting. The mechanisms of action through which PRP exerts its regenerative effects in humans are not completely understood. A systematic review has summarised the evidence from animal experiments that evaluated the use of platelets in the treatment of ATR. 52 The authors reported a moderate beneficial effect of PRP in the treatment of ATR in animal models. The observed effects were attributed to accelerated and enhanced scar tissue maturation. 52 An early systematic review of clinical trials of PRP in humans by Taylor et al. 50 reported faster healing and better function in the management of tendon and ligament injuries. However, a more recent review by Gholami et al. 51 reported no beneficial effects of PRP in the management of sports injuries for pain or function compared with other management options.

A number of studies using different postoperative protocols have investigated the effect of platelets on the treatment of ATR in humans. Two case reports, one of an injured athlete53 and the other of a complete tear in an active 71-year-old male,54 showed positive effects of quick return to play after non-surgical management. In a case–control study by Sánchez et al.,55 12 athletes with Achilles tendon injury were treated with either PRP and surgery, or surgery alone; the PRP and surgery group (six athletes) achieved significantly quicker recovery and better functional outcome measures of greater range of motion and return to play than those treated with surgery alone. However, this effect was seen only at 22 weeks, and a difference was not found at the 1-year follow-up. 56 Similarly, one small clinical trial (15 participants per group) demonstrated no significant difference in structural and functional outcomes between surgery with PRP and surgery alone. 57 However, Zou et al. 58 reported some improvement in isokinetic muscle strength, ankle range of motion and measures of quality of life in a clinical study of 36 people with ATR comparing surgery and PRP with surgery alone.

A randomised controlled trial (RCT) (n = 30) by Schepull et al. 56 showed no difference for functional outcomes or biomechanical tendon properties at 1-year follow-up when PRP with surgery was compared with surgery alone. 59 However, the authors also stated that a limitation in their platelet preparation technique and storage of up to 20 hours resulted in only a 20% release of growth factors from the platelets. The level of activation to release growth factors may have an impact on the regenerative process and could alter the clinical outcome. We observed good activation rates of 69% as a measure of the quality of PRP used in a clinical pilot study,57 in which 20 patients were randomised to receive non-surgical treatment with or without PRP injection or operative management with PRP gel, with standard rehabilitation for both groups. Achilles tendon Total Rupture Score (ATRS) and the Victoria Institute of Sport Activity for Achilles (VISA-A) score showed significantly better PRP group outcomes starting from week 3 to week 24 of follow-up.

All of these small clinical studies used PRP as an adjunct to surgical repair, which may have influenced the effect of PRP on the tendon as there is additional surgical trauma and the mechanical effects of having a suture across the tendon rupture gap. Overall, the results from RCTs with small numbers of participants that have assessed PRP in ATR highlight that the current evidence is inconclusive.

Rationale for the PATH-2 study

Despite significant advances in the management of musculoskeletal injuries (e.g. novel operating techniques),27 slow recovery, morbidity and complication rates persist. PRP as an alternative, regenerative orthobiological agent has gained popularity on the basis that it could augment the management of traumatic musculoskeletal injuries. It is estimated that PRP is used to treat 86,000 tendon, ligament and muscle disorders annually in the USA and Europe. 60 There is also evidence to suggest that PRP injections are being administered in the NHS and private medical practices in the UK. 59

Platelet-rich plasma was initially used only as an adjunct during orthopaedic surgery, but in recent times it has been applied in other areas of sports medicine. Basic experimental research has provided a biological rationale for PRP, with encouraging results from histological and biomechanical assessments; however, its clinical efficacy is yet to be confirmed by findings from studies with robust designs. Although PRP remains an attractive treatment option, owing to its simplicity, patient acceptability, affordability and practicality, the lack of proven clinical effects means that its use is controversial.

Although PRP has been shown to be relatively safe, preparation of PRP has yet to be standardised. 61 For example, some authors have suggested that administration during the early inflammatory stage of healing may disrupt normal underlying physiological processes. This may result in the downregulation of certain growth factors, such as TGF-β, which may promote the formation of fibrous tissue. 61 Others did not find a link with injection stage. Although injecting PRP concentrates may, in theory, increase the risk of infection or exacerbate the underlying microbial activity within spontaneous ruptured tendons,23 no clinical impact on outcome has been reported. 36,50

There is substantial variation in the validity and type of outcomes measured, as well as inconsistency in the effect size of PRP observed from published studies. The underpowered and inadequately designed studies to date suggest that no definite conclusions can be made on PRP as a superior (or inferior) treatment option over standard care in the management of acute ATR. A recent meta-analysis of PRP for orthopaedic conditions62 stated the need for adequately powered studies using disease-specific and patient-important outcomes to investigate the effect of PRP.

We conducted a multicentre, pragmatic, parallel-group, blinded, placebo-controlled RCT to assess the efficacy of PRP. Patients presenting with acute ATR receiving non-surgical management at orthopaedic trauma surgery departments in the NHS in the UK were randomised to receive PRP or placebo (dry-needle injection). PRP and rehabilitation were standardised. Two substudies of blood and tendon biopsy were integrated into the main trial design to assess the components of PRP and to further explore the mechanism of action on injured Achilles tendons.

Participant screening and eligibility assessment

All patients attending orthopaedic or trauma clinics with a suspected acute ATR were screened for inclusion in the trial. Patients were identified for screening for the trial at this clinic visit, normally in the 12 days after the initial visit to hospital, where the attending surgeon (or an extended-scope physiotherapist) confirmed appropriateness for non-surgical treatment and trial eligibility.

Patients with acute ATR were eligible for the trial if they met all of the inclusion criteria and none of the exclusion criteria.

Patients were eligible for inclusion if:

• they were willing and able to give informed consent for participation in the trial

• they were aged ≥ 18 years

• they were ambulatory prior to injury without the use of walking aids or assistance of another person

• they were diagnosed with an acute, complete ATR

• they presented and received trial treatment within 12 days post injury

• the decision had been made for them to receive non-surgical treatment

• they were able (in the investigator’s opinion) and willing to comply with all trial requirements

• they were able to attend a PATH-2 trial hospital site for the 24-week follow-up.

Patients were excluded if they:

• had an Achilles tendon injury at the insertion to the calcaneum or at the musculotendinous junction

• had a previous major tendon or ankle injury or deformity to either lower leg

• had a history of diabetes mellitus

• had a known platelet abnormality or haematological disorder

• were currently using systemic cortisone or a treatment dose of an anticoagulant (i.e. a prophylactic dose for preventing thrombosis was not an exclusion criterion)

• had evidence of lower-limb gangrene/ulcers or peripheral vascular disease

• had a history of hepatic or renal impairment or dialysis

• were female and pregnant or breastfeeding

• were receiving, or had received within the previous 3 months, radiotherapy or chemotherapy

• had inadequate venous access for drawing blood

• had any other significant disease or disorder that, in the opinion of the investigator, might put the participant at risk because of participation in the trial or might influence the result of the trial or the patient’s ability to participate in the trial.

Members of the local research team informed the patient of the trial and carried out the informed consent process, baseline data collection and randomisation.

Standard treatment for this non-surgical population is usually application of a plaster cast, orthopaedic brace or splint during the clinic visit. The PATH-2 trial treatment options required the intervention to be delivered before the definitive immobilisation method was applied. Therefore, the time frame between the informed consent process and treatment was relatively short. To help raise awareness of the trial while patients waited in the clinic, sites displayed trial posters and distributed participant information sheets (PISs) in clinic waiting areas.

The attending clinician conducted a clinical examination and decided with the patient whether management would be surgical or non-surgical. If non-surgical management was appropriate, and the clinician confirmed the patient’s eligibility for the trial, the patient was informed of the trial and given a PIS. The potential participants were allowed as much time as practically possible in this type of acute injury to consider the information, and had the opportunity to ask questions of the attending clinical team and a member of the research team.

Consent was obtained by a member of the local research team who was trained in Good Clinical Practice (GCP) and authorised by the PI to take consent: this could be a research nurse, physiotherapist or surgeon at the local NHS trust, or a nurse assigned from the local National Institute for Health Research (NIHR) clinical research network. The person taking consent presented the PIS and consent form to the participant by means of a verbal discussion. The PIS detailed the exact nature of the trial, the implications and constraints of the protocol, the known side effects and any risks involved in taking part. It was clearly stated that the participant was free to withdraw from the trial at any time for any reason without prejudice to future care and with no obligation to give the reason for withdrawal. The consent form was personally signed and dated by the participant and by the person who obtained consent.

A copy of the signed consent form was given to the participant, and one copy was sent to the trial co-ordinating team in Oxford to facilitate central monitoring. The original signed consent form was retained in the medical notes, and a copy was held in the investigator site file. Consent forms were held in a secure location separately from trial data. Permission was obtained to inform the participant’s general practitioner (GP) of trial participation.

The PIS specified that a blood sample of up to a maximum of 55 ml may be taken. Only a maximum amount was stated, as detailing the exact amount of blood withdrawn per treatment group would have revealed treatment allocation; the amount drawn varied depending on treatment. The PIS also outlined that the sample remaining after treatment for the PRP injection group and the sample for the placebo group would be dispatched to a member of the central research team in the Institute of Inflammation and Ageing, University of Birmingham, for analysis. Samples were anonymised before dispatch, and identified using only the participant’s unique trial number. No laboratory results were reported back to participants or the recruiting centre.

The PIS stated that name and contact details (including mobile phone number, telephone number and e-mail address) would be collected to facilitate follow-up, full data collection and reporting of results. A copy of the contact details would be sent to the trial co-ordinating team in Oxford. These details were used by the trial team to check contact details through NHS Digital and to provide other basic trial-related information that was needed for follow-up.

Permission was sought to allow access to participant data by responsible members of the University of Oxford or the NHS trust for monitoring or audit of the trial to ensure that regulations were complied with.

The 24-week follow-up visit included the heel-rise endurance test (HRET) (primary outcome data collection). The participant was asked to consent to video-recording (without audio) of their ankle and leg movements at the time of the test. Consent for the video-recording was given on a second consent form so that the participant could consent to participation in the trial without consenting to video-recording. The participant personally signed and dated the consent form for video-recording, followed by the person who obtained consent, and the procedures for handling the form were the same as for the intervention consent form, detailed previously. Filming did not include the participant’s face; it focused on the legs, thus reducing the risk of participants being identifiable from the film. The video file (or any still photographs from it) was labelled with the unique trial number, and no identifying details were used. Permission was sought to send the video file to the trial team in Oxford, where it was held and viewed by members of the research team.

At one site (Oxford) where participants were also asked to consent to the biopsy sample collection for substudy 2, a version of the PIS was used that, in addition to the content described previously, invited the patient to attend a separate trial visit during which a small sample of tendon tissue would be removed from the injured tendon. A separate consent form that included consent to giving a tissue sample was provided for these participants. The participant was under no obligation to give a tissue sample.

After injection delivery

The centrifuge produced 8 ml of PRP, of which 4 ml was injected into the participant’s Achilles tendon; 4 ml remained for laboratory analysis (see Chapter 4 for further details of the preparation of these samples). Immediately after the intervention, a member of the local team prepared the remaining PRP (PRP group) and the 5-ml whole-blood sample (PRP and placebo groups) for storage or dispatch. Samples were sent to a member of the central team at the University of Birmingham research laboratory. A single, specialised courier (Davies International, Hampshire, UK) was used for this purpose and a pre-paid account was established with the courier to facilitate ease of sample transport.

The 4 ml of PRP that remained after injection (PRP group only) was prepared for dispatch to the laboratory; 1 ml of PRP was transferred to a microtube and stored at the site in a –70 °C freezer within 2 hours of PRP production. These samples were collected by the dedicated courier at the end of recruitment and transported to the central laboratory for analysis for substudy 1.

In the PRP group, 1 ml of PRP was transferred to a microtube and placed in a trial-specific courier pouch; 2 ml of PRP was transferred into activation vials from a proprietary platelet-activation kit (CB kits, Platelet Solutions Ltd, Nottingham, UK) supplied by the central trial office for platelet activation and fixation. The platelet activation and fixation took around 5 minutes and full instructions were given in the intervention and blood processing manual. When the process of platelet activation and fixation was complete, these two tubes were placed in the courier pouch.

For both groups, the 5-ml venous whole-blood sample was transferred to a tube containing ethylenediaminetetraacetic acid, an anticoagulant, and placed in the courier pouch. Thus, the courier pouch contained four tubes if the participant was randomised to PRP (three tubes containing PRP and one tube containing blood) and one tube if the participant was randomised to placebo (blood only).

The courier pouch was packaged in accordance with instructions in the intervention and blood processing manual and the site contacted the courier for collection and next-day delivery service. Until collection, the courier package was stored at room temperature.

Preparation of PRP and samples and courier dispatch were recorded on a blood tracking sheet. The original version of this form was enclosed with the courier pouch for dispatch to the laboratory. A copy was sent separately to the central trial office and a copy was retained at the site. Blood and PRP analysis results were not reported back to the recruiting centre as they did not have an impact on future treatment.

Injection training

Training in the delivery of both the PRP injection and the imitation injection was provided by the trial team and recorded on a PATH-2 trial training form (treatment related), which was signed by the clinician receiving training and retained at the site. Training consisted of the provision of a training manual for all sites. At most sites, a training video or training session by a PATH-2 trial trainer using the trial kit and trial procedures was provided in addition to this manual. The PI at each site identified surgeons (or extended-scope physiotherapists) to be trained, and recorded those who completed training on the site delegation log. Only those individuals who were trial trained and listed on the delegation log were able to carry out trial treatments. When the delegation log was updated, a copy was sent to the trial co-ordinating office in Oxford. The PI at each site identified local staff who were responsible for preparing the blood and PRP samples for treatment and for later dispatch. Only those individuals who were listed on the delegation log as having this responsibility were able to carry out preparation of the blood and PRP and to manage blood and PRP samples.

Primary outcome

The primary objective of the trial was to evaluate the clinical efficacy of PRP in acute ATR in terms of muscle–tendon function. The primary outcome measure at 24 weeks post randomisation was the Limb Symmetry Index (LSI) of the work performed by each lower limb in joules (J) during the HRET. The HRET is a validated objective measure of Achilles tendon muscle unit function. 65

The HRET involves repetitive concentric–eccentric muscle actions of the plantar flexors in a single-leg stance until exhaustion, with performance quantified as (1) total work in J, (2) number of repetitions and (3) maximum heel-rise height in cm. 65 Total work was selected as the primary outcome measure because this index provides a measure of plantar flexor muscle endurance and Achilles tendon function as it incorporates the maximum height of each repetition. Total work (J) was computed as the product of body mass (kg), total vertical displacement (m) and the constant 9.807 converting kilopond metres to joules. To quantify performance, a linear displacement sensor (MUSCLELAB™, Ergotest Innovation A.S., Porsgrunn, Norway) was attached to the participant’s heel during the test and recorded on software on a laptop computer. A PATH-2-specific user interface version of the MUSCLELAB software was developed with the manufacturer for ease of use in a multicentre trial. Key features of this software and more background on the HRET development are described in further detail in Appendix 2.

Heel-rise endurance test procedure for PATH-2

At the 24-week follow-up visit, the HRET procedure was explained to each participant by the blinded outcome assessor. Standardised participant information and instructions for the HRET involved the participant watching a video demonstration of the HRET and reading standardised written instructions detailing their expected conduct during the test and the test termination criteria. The standardised warm-up involved 5 minutes of usual-pace walking followed by 10 double-leg heel raises. Before testing each leg, the participants were asked to stand on the box being used and were allowed to familiarise themselves with the expected timing of heel raises by lifting both heels together.

The HRET was first carried out on the uninjured limb and then on the injured limb. The test started with the participant standing on one leg on a 10° incline box (so that the ankle was in a dorsiflexed position) with the cord from the linear encoder strapped to the participant’s heel (Figure 3).

FIGURE 3.

A participant performing the HRET. The participant stands on 10° incline box with a cord attached to their heel and connected to a linear encoder, and raises and lowers their heel repeatedly until fatigued.

The following standardisation parameters were adopted:

• Ankle starting position of 10° dorsiflexion, produced by conducting the HRET on a custom-made 10° incline box.

• Knee starting position of full extension.

• Height of each repetition to be as high as possible.

• Pace of 30 raises per minute, guided by a digital metronome.

• Balance support by the fingertips only.

• Strictly defined test termination criteria – participants either stopped (i.e. volitional task failure) or were audibly instructed to stop with both feet flat on the box whenever any of the following test termination criteria were observed: (1) inability to keep pace with the metronome, (2) inability to maintain full knee extension of the standing leg or (3) using more than fingertip support. The desired end point was volitional task failure; however, outcome assessors were encouraged to use verbal prompts whenever the termination criteria were observed and to stop the test if the participant did not respond to two consecutive prompts.

Secure data transfer and confidentiality were assured by copying the HRET data from the encrypted laptop to an encrypted Universal Serial Bus (USB) drive that was sent to the trial office. Sites retained a copy of the entire site HRET data set on the dedicated trial laptop.

Assessor training material consisting of high-quality training videos, which were made by the PATH-2 trial team and produced by Oxford Medical Illustration (Oxford University Hospitals NHS Foundation Trust, Oxford, UK), and a training and reference manual. Face-to-face training was delivered by a member of the PATH-2 trial team to each outcome assessor prior to their first participant’s 24-week follow-up appointment.

The linear encoder, a very sensitive device, recorded minimal movements that might not represent actual heel rises; it could also pick up any additional movements of the heel (e.g. participants tend to step off the box or lift the leg up at the end of the test). To dismiss potential measurement errors, two members of the trial team, who were blinded to treatment allocation, independently reviewed videos of all assessments for which participants consented to recording. The invalid heel-raise repetitions in the HRET data were identified so that could they could be dealt with in the analysis (see Statistical methods).

Secondary outcome measures

The secondary outcome measures were as follows:

• The ATRS,66 a validated patient-reported outcome measure (PROM) incorporating 10 questions relating to muscle strength, fatigue, pain and function at 4, 7, 13 and 24 weeks post treatment. The score range is 0–100.

• The SF-12 version 2 acute, a health-related quality-of-life PROM67 at 4, 7, 13 and 24 weeks post treatment. The scale consists of a 12-item short questionnaire that assesses eight different health domains reflecting both the physical health (physical functioning, role participation with physical health problems, bodily pain and general health) and the mental health (vitality, social functioning, role participation with emotional health problems and mental health) of the evaluated participants.

• A daily pain diary, reported by the participant for 14 days beginning on the day of treatment, using a VAS. The participant was asked to place a vertical mark on a 100-mm horizontal line between the extremes of ‘no pain’ and ‘worst pain imaginable’. Trial office staff measured the distance to the vertical mark and converted the distance to a daily measure (0–100) of pain.

• The pain component of the ATRS at 4, 7, 13 and 24 weeks post treatment. In the ATRS, the participant was asked ‘Are you limited due to pain in the calf/Achilles tendon/foot?’ and invited to circle an integer from 0 to 10 inclusive, 0 indicating major limitations/symptoms and 10 indicating no limitations/symptoms.

• The PSFS,63 a PROM indicating progress on self-selected recovery goals relating to daily activities affected by the injury on an 11-point scale (0–10), at 4, 7, 13 and 24 weeks post treatment.

• The maximum number of repetitions (heel rises) recorded during the HRET68 at 24 weeks post treatment.

• The maximum vertical displacement (cm) recorded during the HRET68 at 24 weeks post treatment.

Adverse events

Complication events reported by participants were explored at two levels: serious adverse events (SAEs) and adverse events (AEs).

Blinding during treatment preparation and intervention

This was a double-blinded (participant and primary outcome assessor) trial. Participants remained blind to allocation throughout the trial. Those staff involved in treatment delivery were aware of treatment allocation because of the nature of the intervention. The primary outcome assessors were not aware of the allocation or treatment received.

Site staff were trained consistently to maintain blinding. Care was taken not to draw the participant’s attention to the amount of blood taken as this could have indicated the intervention group. To facilitate treatment blinding, preparation of the intervention took place in a location away from the treatment room, or the participant waited outside, as there was a difference in the number of consumables handled depending on treatment allocation. When the injection was prepared, it remained out of the participant’s view.

If a participant randomised to the placebo group was within listening distance of the centrifuge, the centrifuge was run on a dummy cycle. The wait time for the participant was approximately 17 minutes; this was the time taken to run the centrifuge to produce PRP. During the interventions, the participant was requested to lie face down with their foot and ankle slightly off the edge of the bed, and there was an option of placing a pillow on the back of the participant’s shoulders below the neck to help visually shield the intervention procedure. The verbal cues used by site staff at the time of treatment delivery did not refer to allocation and were the same for both groups.

Full guidance on ensuring that the correct amount of blood was withdrawn into the appropriate tube or syringe while ensuring that the participant remained blinded to allocation was provided in the PATH-2 trial intervention and blood processing training materials, which were retained in an area where they were not visible to the public.

Facilitating blinding immediately post treatment

The PATH-2 blood sample and treatment CRF collected confirmation that allocated treatment was received. When a treatment other than that allocated was indicated on the CRF, the trial co-ordinating team contacted the site for further details and recorded the details. Additional training or troubleshooting of any issues was instigated when necessary.

In hospital notes and in the letter to the participant’s GP, it was recorded that an injection was delivered according to the random allocation assigned, but the type of injection was not recorded.

To assess the completeness of participant blinding, after the 24-week trial questionnaire and HRET assessments were completed, participants were asked which treatment they believed they received (‘PRP injection’, ‘imitation injection’ or ‘don’t know’) and why they believed this, along with questions around their experience of taking part in the trial. These data were used to compute two blinding indices with 95% confidence intervals (CIs) (James et al. 69 and Bang et al. 70 blinding indices).

Blinded follow-up assessors

Throughout the trial, follow-up was carried out wherever possible by blinded assessors unaware of treatment allocation. A physiotherapist or assessor who was blind to allocation carried out the 24-week trial follow-up assessment, including the HRET.

Sample size

The trial was powered on the work LSI from the HRET at 24 weeks post randomisation and on the ATRS. There were no formal interim analyses of the outcomes planned for the PATH-2 trial. However, the Data and Safety Monitoring Committee (DSMC) agreed to check the sample size assumptions when at least half of the originally planned number of participants had completed the 24-week primary end-point assessment because they were based on the findings of a single previous study conducted in a single centre. 65 This decision was documented in the statistical analysis plan version 1.0.

Original calculation of sample size

It was calculated that 214 participants (107 per treatment group) would provide 90% power to detect a standardised difference of 0.5 in the HRET work measured by the work LSI at 24 weeks post randomisation and with 5% (two-sided) significance allowing for 20% loss to follow-up. This was based on previous data from the non-surgical arm of the 2010 study by Nilsson-Helander et al. ,65 in which a clinically important difference of 10 points with a standard deviation (SD) of 20 points was observed.

This sample size was also to provide 90% power and 5% (two-sided) significance to detect a standardised effect size of 0.5 in the ATRS between the two treatment groups, based on a difference of 11% and a SD of 21.4%.

Sample size review and recalculation

The DSMC agreed at their first meeting that the assumptions for the sample size should be reviewed after approximately half of the participants had been recruited. This review would be a blinded estimate of work LSI variability only. The first variability analysis was undertaken in October 2016 when half of the participants had been recruited (n = 104). It was observed that the SD of the work LSI was smaller than that used in the sample size calculation so no changes to the sample size were recommended (observed SD 17.37, versus sample size SD 20). However, at this time only 27 participants had completed the primary end-point assessment.

As of June 2017, approximately half of the participants had reached their primary end point and recruitment was still ongoing. Therefore, to provide a more robust estimate of the actual variability of the work LSI, a further sample size assumption review was carried out. Using cleaned and validated data from 75 participants, the observed SD was 24, which was higher than that used in the sample size calculation. Based on this observed SD of 24 (being aware that this could go up or down with the addition of further participant data), we recalculated the sample size based on 80% power, which, allowing for 20% loss to follow-up, required a sample size of 226 patients. The DSMC advised the Trial Steering Committee (TSC) and trial management team during this process.

The trial continued to recruit to an agreed minimum of 230 participants (overshooting to account for further minimal fluctuations in sample size assumptions), to ensure that there would be a minimum of 80% power for the primary outcome analysis. These approaches, and the reduction in trial power, were supported by the TSC and by the funder.

Software employed

All analyses outlined here and reported in Chapter 4 were undertaken using Stata^® version 15.0 (StataCorp LP, College Station, TX, USA).

Blinded analysis

Initial exploratory analyses were conducted on a blinded data set (not separated by treatment group) to identify the presence of missing values and to clarify the distribution of continuous variables. These analyses were also used to finalise trial populations and to aid in the identification of key prognostic variables to be included in the adjusted analysis. All subsequent analyses described were conducted on an unblinded data set. Following this blinded analysis, the statistical analysis plan was updated to incorporate any changes.

Data validation

Initial analytical steps assessed the validity of the final, blinded data set. The first of these steps was to manually check, when possible, the use of Stata for importing data and merging data sets for at least 20 participants who were randomly sampled. Once accurate importation was confirmed, data in the data set were validated by checking for duplicate records, checking the values of the range of variables and validating potential outliers against CRFs. Any discrepancies that could not be rectified were referred back to the trial sites. Finally, the production of calculated variables using Stata was manually checked for at least 20 participants who were randomly sampled.

Trial populations

Descriptive analysis

All participants were included in descriptive analyses. Recruitment into the trial was explored, including the numbers of individuals assessed, recruited, randomly assigned to the PRP injection or placebo, receiving treatment, completing the trial protocol and analysed for the primary outcome. Any protocol deviations and violations were investigated and issues with the screening data, including potential incidents of ineligible patients being randomised, were explored. The proportion of unblinded assessors for each assessment was assessed, and the James et al. 69 and Bang et al. 70 indices were calculated to determine whether or not participants were successfully blinded to the intervention they received.

The baseline comparability of participant-level data for each of the treatment groups was summarised. For baseline and follow-up data collected manually through the use of forms or questionnaires, data compliance was explored; all available data collected from these forms were summarised and the proportion of missing items from completed questionnaires was examined. Data-availability patterns for individual variables were assessed both overall and for the two treatment groups separately to explore missing information. Missing values were checked for consistency and the proportion of missing values per variable was assessed. Differentiations were made between partially completed and fully missing outcome data. For measures comprising multiple items, for example SF-12, pro-rata estimation of total and subscale scores was employed for each treatment group, using the appropriate scale-specific scoring guidelines. Imputation of data was utilised for sensitivity checking and is described in further detail in Primary outcome. Imputation was carried out as recommended for each relevant questionnaire, with the imputation value determined by the distribution of the underlying data. Methods utilised for imputation were assessed manually, when possible, for at least 20 participants who were randomly sampled. Sensitivity analyses were conducted to ensure that the missing-at-random assumption for imputation was met.

Comparisons of losses to follow-up were carried out. The proportions of participants defaulting or withdrawing from the trial over the whole period of trial follow-up, and at each analysis time point, were compared between the PRP and placebo groups. The importance of differences identified were assessed using a chi-squared test. For all analyses, a p-value of < 0.05 was considered indicative of a significant difference.

Quality assurance and compliance with the intervention, including any deviations from protocols prior to and during the trial, were assessed. Treatment received was summarised by treatment groups, and time frames for the steps with blood processing were explored to ensure that all participants received their PRP injection within 2 hours of their blood sample being taken. The grade of the health-care professional conducting the procedure was summarised. The importance of apparent differences in compliances was assessed using a chi-squared test.

Finally, data quality and the effect of the treatment received were assessed using CACE analysis in place of the originally planned per-protocol (PP) analysis, as this was deemed the optimal approach. As per the PP analysis, CACE analysis allows for adjustment to account for participants whose received treatment either complied or did not comply with that allocated to them; however, CACE analysis does so without assuming that compliers are the same as non-compliers, as would be the case with a PP analysis. CACE analysis retains information on the original treatment allocation that participants received, which allows us to forgo the core assumption of a PP analysis – that receipt of treatment is random with respect to the outcome predictors – and to enhance our understanding of the effect of PRP on ATR recovery. The decision to use CACE was taken post finalisation of the statistical analysis plan but ahead of the end of follow-up and, therefore, ahead of unblinding.

For the CACE analysis, compliance was defined as participants who received the intervention allocated to them who had sufficient platelets present in their PRP sample to classify it as platelet rich (i.e. higher platelet concentration in PRP than in whole blood). The results from the CACE analysis were reported with standard errors, p-values and 95% CIs, and assessments were made to ensure that the primary outcome estimation was captured within this range.

Standard care for ATR for all trial participants was unaffected by this trial. All participants should therefore have had their injured lower limb immobilised, received a referral for physiotherapy, received delivery of venous thromboembolism prophylaxis and received advice about when to start weight-bearing and ankle-motion exercises. Compliance with these different aspects of standard care was compared between treatment groups using a chi-squared test for categorical data and a Student’s t-test for continuous data.

Analysis of the primary outcome

The primary outcome for this trial was muscle–tendon function, measured by the LSI of maximum work, performed during the HRET (see Primary outcome for more detail). The work LSI was calculated for all trial participants, at 24 weeks (varying by –2 to +8 weeks) post randomisation, as follows:

The proportions of individuals missing data for outcome and explanatory variables were explored, and work LSI data for all participants were assessed for normality. Missing HRET data were defined and handled as follows:

• Participants with true missing data – participants who were lost to follow-up and participants for whom technical errors were experienced during HRETs. The participants who were lost to follow-up were any participants who either did not attend their 24-week follow-up at all or did not remain at their appointment to complete their HRET. The measurements for these participants were kept as missing.

• Participants with true zero measurements – participants who attempted to complete their HRET assessment but their attempts were insufficient for the encoder to record any results. Zero measurements for these participants were included for analysis.

• Participants with potential zero measurements – participants who did not attempt to complete their HRET in at least one leg, despite attending the follow-up appointment. Zero measurements for these participants were included for analysis.

The data followed a near-normal distribution, with a small elevation around zero due to participants defined as having true and potential zero measurements. However, work LSI data did not require transformation prior to inclusion in the final model as post-estimation assessments indicated that the distribution was suitable for the regression technique employed. The impact of zero inflation of the primary outcome measure was assessed using a two-parts model to (1) identify differences between participants with positive work LSI results and participants with zero measures, and (2) identify the impact of the inclusion of participants with zero measures on the primary outcome results. Mean work LSI differences, robust standard errors, p-values and 95% CIs were reported and compared with the core primary outcome results for deviation or interpretation changes.

An unpaired Student’s t-test was used to explore the unadjusted effects of PRP compared with those of placebo.

Multivariate linear regression was used to investigate the effect of PRP on ATR recovery. The base-unadjusted regression model was built using work LSI as the dependent variable and treatment as the key independent variable. The principal analysis model built on this model by adjusting for participant age group and clustering by trial site. Supplementary regression models were built, which adjusted for the effects of sex, body mass index (BMI) and smoking status.

Sensitivity analyses were carried out using imputation of values for missing HRET data to examine the robustness of the conclusions made from the analyses to address the primary aims. Missing HRET data were handled using two approaches. The first was to employ simple imputation of the average concentric displacement (upwards movement); this method was used for missing data associated with specific repetitions of heel rise. The second method to handle missing data was to employ multiple imputation by chained equations; this was used to calculate work LSI for participants with data missing for entire assessments. The multiple imputation using chained equations procedure created a series of complete data sets (observed plus imputed data) in which analyses were carried out individually. The regression parameter estimates plus corresponding standard errors obtained from the analyses of each of the 50 imputed data sets were then combined using the Rubin’s rule approach. When models did not converge during the multiple imputation process, some predictor preselection was carried out based on clinical expertise and subject matter. Sensitivity analyses were employed to assess the assumptions made in preselecting these predictor variables. The method utilised for multiple imputation inclusion was to use the stratification variables already in the principal adjusted model and to further account for any variables with an R² value of ≥ 0.2 following correlation assessments. This resulted in three additional variables being selected: whether or not the participant jogged or ran, undertook weight training or participated in squash before their injury.

For all analyses, a two-sided p-value of 0.05 (5% significance level) was used to indicate evidence of statistical significance.

Analysis of secondary outcomes

The number of repetitions and the maximum heel-rise height recorded during the HRET assessment were explored as secondary outcome measures. Furthermore, four PROMs [pain as measured by the VAS (daily over the 14 days immediately post treatment), ATRS, PSFS and the SF-12 mental and physical health scores] were explored.

The ITT population was used for secondary outcome analyses.

Repeated-measures linear mixed-effects regression models were used to allow the data collected at all follow-up time points to be taken into account. Time elapsed from the intervention to the outcome measurements was included in the models as a random effect factor, considering that not all participants had their follow-up assessments at exactly the same time. Mean differences and 95% CIs were examined. All analyses were adjusted for the stratification factors included in the adjusted analysis of the primary outcome. As pre-injury scores had been collected for the SF-12 data, the models for this PROM were built so that the principal and supplementary analyses adjusted for these scores.

Sensitivity analyses were carried out on the overall ATRS data sets as this was the key secondary outcome. The first of these analyses used the mITT population used in the primary outcome analysis to assess the impact of selecting the mITT population used in the primary outcome analysis. The second applied multiple imputation by chained equations for those participants with data missing with missing ATRSs. The multiple imputation by chained equations technique that was applied to the ATRSs was the same as for the primary outcome, although the additional variables selected for inclusion to inform the imputation were different. The six additional variables used to build the imputation data sets in this situation were BMI as a continuous variable, the date of the ATR, the date when the treatment was received, the date when a participant started carrying out rehabilitation exercises (as recorded at week 13), whether or not a participant was jogging or running by week 24 and whether or not a participant was carrying out ‘do it yourself’ (DIY), heavy housework or gardening by week 24.

As with the primary outcome analysis, data quality was assessed using CACE on the overall ATRS data, to examine the robustness of the conclusions made from the analyses.

Adverse events

The number of participants experiencing one or more AE of any type over the 24 weeks was explored. The number of participants experiencing each specific type of AE over the 24 weeks was also analysed. The number of separate complication events was not explored owing to the opportunity for re-reporting by participants across the different follow-up time points and the nature of the events of interest.

Outcomes for patient and public involvement

In the main trial, participants reported satisfaction at the end of the trial in a post-assessment questionnaire. We asked participants to rate their satisfaction with the trial on a five-point scale of ‘very dissatisfied’ to ‘very satisfied’. Eighty-nine per cent of participants declared themselves ‘satisfied’ or ‘very satisfied’. We invited comments on what could have improved their experience; the vast majority who answered (n = 37) said ‘nothing’ and none asked for fewer forms, although five said that some questions were not relevant and three asked for more information.

An adjustment to the statistical analysis related to how the pain score would be reported is an example of how feedback from our PPI representative improved the conduct of the trial.

The PPI representative reviewed a graphical information slide to be used in future public-facing presentations.

Overall, from participant feedback, we believe that our PPI involvement improved trial conduct and participant experience.

Ethics approval

The trial was given a favourable opinion by the South Central – Oxford A Research Ethics Committee on 11 November 2014 (reference number 14/SC/1333) and each site was granted site-specific approval from its NHS trust research and development department before trial commencement.

Data and Safety Monitoring Committee

The DSMC was established to safeguard the interests of trial participants, monitor the main outcome measures including safety and efficacy, and monitor data quality and completeness. In accordance with the DSMC charter, the DSMC received and reviewed information on the progress and data collection of the trial and advised the TSC on the conduct of the trial.

Trial Steering Committee

The TSC provided expert oversight of the trial on behalf of the sponsor and funder. Through its independent chairperson, the TSC also provided advice to the Trial Management Group (TMG), the funder and OCTRU on all aspects of the trial.

Trial Management Group

The day-to-day management of the trial was overseen by a TMG comprising the chief investigator, the lead research nurse at the lead site (John Radcliffe Hospital), the co-applicants, the trial statistician, research physiotherapists and the trial manager.

Intention-to-treat population

• Primary outcome (work LSI): a total of 201 participants were included, 100 of whom were in the PRP group and 101 of whom were in the placebo group.

• Key secondary outcome (ATRS): a total of 216 participants were included, 107 of whom were in the PRP group and 109 of whom were in the placebo group.

Baseline, randomisation and treatment delivery forms

There were no differences between treatment groups in compliance with baseline, randomisation and treatment delivery forms. Completion rates for all of these forms were 100% for both treatment groups.

Follow-up diaries and questionnaires

There were no differences between treatment groups in compliance with follow-up questionnaires and diaries. A total of 180 out of 229 participants (78.6%) returned their pain diary for the first 2 weeks post intervention. Completion of all follow-up questionnaires at weeks 4, 7, 13 and 24 was ≥ 90% in both intervention groups.

Treatment allocation compliance

Nine per cent (10/113) of PRP group participants did not receive their allocated treatment, whereas all placebo group participants received the intervention assigned to them. Treatment for participants in the placebo group was dependent on the availability of the trial kit for this intervention. However, treatment for participants in the PRP group was further dependent on the ability to withdraw sufficient whole blood to produce the required 8-ml PRP sample. Therefore, it is not unexpected that a higher number of participants in the PRP group did not receive their allocated treatment.

The reasons for participants in the PRP group not receiving their allocated treatment were:

• site staff being unable to take a blood sample from the participant (3/10, 30.0%)

• site staff not being able to take a blood sample of sufficient volume to produce PRP (2/10, 20.0%)

• technical issues encountered with centrifuge (2/10, 20.0%)

• participant not remaining on site to receive treatment (1/10, 10.0%)

• in-date PRP disposable kit not available (1/10, 10.0%)

• site staff unable to obtain PRP disposable kit (1/10, 10.0%).

In all PRP injection participants who were able to have blood samples drawn (103/113, 91.15%), the mean time from taking the blood sample to their PRP injection was 42 minutes (range 12–105 minutes). This breaks down to a mean of 28 minutes from taking the blood sample to producing the PRP sample (range 1–80 minutes) and a mean of 14 minutes from producing the PRP sample to receiving the injection (range 0–45 minutes). All PRP intervention participants who were able to have blood samples drawn received their injection within the expected 2-hour time frame.

Grade of the clinician administering the treatment

A total of 75.5% of participants [total, 173/229; PRP group, 86/113 (76.1%), vs. placebo group, 87/116 (75.0%)] had their treatment administered by a consultant surgeon; 19.2% had their treatment administered by a surgical registrar or research fellow [total, 44/229; PRP group, 19/113 (16.8%), vs. placebo group, 25/116 (21.5%)], 1.3% had it administered by an extended-scope physiotherapist [total, 3/229; PRP group, 1/113 (0.9%), vs. placebo group, 2/116 (1.7%)] and 1.7% had it administered by any other grade of physiotherapist [total, 4/229; PRP group, 2/113 (1.8%), vs. placebo group, 2/116 (1.7%)]. Although we do not have any information for 2.2% of participants (5/229), there were no discernible differences between groups in relation to the grade of clinician administering the treatment (p = 0.706).

Compliance with standard care

Compliance with standard care is reported in Table 2. There were no discernible differences between treatment groups in compliance or in the use of venous thromboembolism prophylaxis [PRP group, 56/113 (49.5%), vs. placebo group, 58/116 (50.0%), difference –0.4%, p = 0.891]; the use of a below-knee cast [PRP group, 47/113 (41.6%), vs. placebo group, 55/116 (47.4%)] or splint/brace [PRP group, 64/113 (56.6%), vs. placebo group, 61/116 (52.6%)] to immobilise the ankle (p = 0.443); or in the receipt of a physiotherapy consultation within 7 weeks [PRP group, 64/113 (56.6%), vs. placebo group, 68/116 (58.6%)] or 13 weeks [PRP group, 100/113 (88.5%), vs. placebo group, 100/116 (86.2%)] (p = 1.000).

The date a participant started weight-bearing was recorded at the 4-week and 7-week follow-ups, and the date they started ankle-motion exercises was recorded at 7 and 13 weeks. It was assumed that participants remembered the date they were weight-bearing or exercising their ankle better if the time point was closer to it. The time to start weight-bearing and the time to start ankle exercises was calculated as the date reported minus the date the ATR occurred. The date reported was determined to be the earliest date reported by the participant that this activity took place. To allow for the fact that participants may have stopped either weight-bearing or exercising their ankle between questionnaires, participant binary responses as to whether or not they were carrying out an activity were taken into account, and if participants answered ‘yes’ in the earlier questionnaire and ‘no’ in the later questionnaire, their reported date was not included.

Following from the creation of the described time to weight-bearing and time to ankle-exercising variables, assessments of their distributions demonstrated that they did not significantly differ from normal. Therefore, differences in time taken to complete these tasks between treatment groups were assessed using Student’s t-test; there were no differences in the time to weight-bearing between the PRP group (mean 27.0 days, 95% CI 24.1 to 30.0 days) and the placebo group (mean 27.7 days, 95% CI 24.4 to 31.0 days) (PRP group vs. placebo group mean difference –0.7 days, 95% CI –3.7 to 5.0 days; p = 0.768) or in the time taken to start ankle-motion exercises between the PRP group (mean 47.4 days, 95% CI 43.6 to 51.1 days) and placebo group (mean 48.5 days, 95% CI 44.7 to 52.4 days) (PRP group vs. placebo group mean difference –1.1 days, 95% CI –4.2 to 6.5 days; p = 0.672).

Missing data assessments

Missing data explorations were carried out to maximise inclusion in the final analysis. The final classification of participants with missing data following these explorations is given in Table 6. (For information on how missing data were classified, see Chapter 2, Analysis of the primary outcome.)

Primary analysis outputs

There were three data sets available for the analysis of the primary outcome:

1. all available HRET assessments

2. all available HRET assessments that were deemed valid on review of video footage (see Chapter 2, Primary outcome, for more information on the validation review)

3. all available HRET assessments that were deemed valid on review and included only participants with assessments available in both legs.

The primary outcome analysis focuses on data set 3 (all valid HRETs available on both legs).

Explorations of the mean work exhibited during the HRET assessment in participants’ injured and uninjured legs on all three data sets were carried out. Basic explorations of the subsequent measure for work LSI in participants with measurements in both legs were undertaken. These explorations demonstrated that the mean work LSI and subsequent HRET measurements for both intervention groups were very similar.

Focusing on the third data set, the mean work LSI in the injured leg was lower in both the PRP group (675.2, SD. 622.3) and the placebo group (748.2, SD 630.3) than the mean work LSI in the uninjured legs in the PRP (1783.8, SD 838.8) and placebo (1825.5, SD 796.2) groups. Comparing the subsequent work LSI measurement in the PRP group (mean 34.9) with the work LSI measurement in the placebo group (mean 38.3) using the Student’s t-test demonstrated no difference between these two groups (unadjusted PRP group vs. placebo group difference –3.467, 95% CI –10.979 to 4.246; p = 0.384). These explorations demonstrate that the work LSI measurements for both groups are very similar.

Analyses using multivariate linear regression adjusting for stratification factors (adjusted analysis and principal analysis) and other confounding factors (fully adjusted analysis) were undertaken to assess the robustness of conclusions based on these results (Table 7). It is clear from these analyses that there is no evidence of any difference between the intervention groups in relation to work LSI, recorded at 24 weeks post recovery. Taking account of stratification factors and the defined prognostic variables has no impact on the results attained. Sensitivity analyses were also undertaken accounting for participants with zero measurements (two-parts model), with individual missing HRET repetitions (simple imputation) and missing entire HRET data sets (multiple imputation) and for compliance (CACE), and these showed that the results were robust and, therefore, they had no impact on the interpretation of the results.

Post-estimation assessments of the fully adjusted primary outcome regression were carried out to ensure that the results reported by the regression analyses were valid. There were no major issues identified during these assessments and, therefore, the model was kept as simple as possible.

No formal subgroup analyses were carried out; however, results were explored visually using forest plots to examine whether or not the effect of PRP compared with placebo was consistent across the stratification factors (age group and trial site) and other prognostic variables (BMI categories, smoking status and sex) (Figure 5). Although the variability was wide for some categories, such as trial site, as was expected owing to the low number of participants randomised in some sites, the treatment effect was consistent across subgroups, showing that the results are robust.

FIGURE 5.

Forest plot of work LSI effect demonstrating the effect of the intervention on subgroups of defined stratification and prognostic factors. a, Number in the placebo arm vs. number in the PRP arm. The Cardiff and Royal Devon and Exeter sites were excluded as neither had patients in the treatment arm completing HRET assessment.

Overall Achilles tendon Total Rupture Score

An overall ATRS of 0 is the lowest (i.e. poorest) score a participant can attribute to their recovery and an overall ATRS of 100 is the highest. At baseline, the unadjusted mean overall ATRS in the PRP group was 14.0, compared with 11.7 in the placebo group. By week 24, this had risen to 64.9 in the PRP group, compared with 65.6 in the placebo group (Table 8).

A repeated-measures mixed-effects regression analysis accounting for the time elapsed since the intervention was received was undertaken (see Table 8 and Figure 6). As with the primary outcome analysis, sensitivity analyses were conducted to assess the robustness of conclusions based on these results. These analyses found no evidence of any difference between the intervention groups in relation to average overall ATRS recorded throughout the recovery period. Taking account of stratification factors had no impact on the results attained. In addition, accounting for participants missing ATRS data (multiple imputation) and for compliance (CACE) had no impact on the interpretation of the results. These analyses have shown that participants’ overall ATRS increased from baseline to 24 weeks post treatment at an almost identical rate in the PRP and the placebo groups, irrespective of adjustment for stratification variables.

FIGURE 6.

The ATRS principal adjusted repeated-measures mixed-effects regression model demonstrating the change in ATRS in PRP and placebo group participants over time. 0 = totally limited because of ATR; 100 = not limited at all because of ATR.

Post-estimation assessments of the fully adjusted primary outcome regression were carried out to ensure that the results reported by the regression analyses were valid. There were no major issues identified during these assessments and, therefore, the model was kept as simple as possible.

As before, no formal subgroup analyses were carried out; however, results for the principal secondary outcome (ATRS) were explored visually using forest plots to examine whether or not the effect of PRP compared with placebo differed based on stratification factors (age group and trial site) and other prognostic variables (BMI categories, smoking status and sex) (Figure 7).

FIGURE 7.

Forest plot of ATRS demonstrating the effect of the intervention on overall ATRS in subgroups of defined stratification and prognostic factors. a, Number in the placebo arm vs. number in the PRP arm. The Cardiff and Royal Devon and Exeter sites were excluded as neither had patients in the treatment arm completing the assessment.

As for the work LSI, this shows that the treatment effect was consistent across subgroups, showing that the results are robust.

Pain-related Achilles tendon Total Rupture Score

Participants’ pain-related ATRSs, a subset of the overall ATRS, increased at a relatively steady rate from baseline to 24 weeks post treatment, with similar rates in the PRP and placebo groups, irrespective of adjusting for stratification variables (Table 9 and Figure 8).

FIGURE 8.

The ATRS pain mixed-effect regression model. Results from the adjusted repeated-measures mixed-effects regression model demonstrating the change in ATRS pain measure in PRP and placebo group participants over time. 0 = totally limited because of pain; 10 = not limited at all because of pain.

Patient-specific functional scale

Participants’ overall PSFS score increased from baseline to 24 weeks post treatment at an almost identical rate in the PRP and placebo groups, irrespective of adjusting for stratification variables (Table 10 and Figure 9).

FIGURE 9.

The PSFS mixed-effects regression model. Results from the adjusted repeated-measures mixed-effects regression model demonstrating the change in PSFS in PRP and placebo group participants over time. 0 = unable to carry out activities; 10 = able to carry out activities at same levels as before ATR.

Maximum concentric displacement and maximum number of repetitions

As with the primary outcome analysis, there was no difference between the intervention groups in the maximum concentric displacement LSI achieved and the maximum number of repetitions achieved (Table 11). After adjusting for all stratification variables, the average maximum concentric displacement LSI [the LSI calculated as (maximum displacement in the injured leg/maximum displacement in the uninjured leg) × 100, to account for potential ability differences] achieved by participants in the PRP group was 55.1 (95% CI 51.4 to 58.8), compared with 55.4 (95% CI 49.6 to 61.3) in the placebo group. This difference was not significant (mean difference –0.3, 95% CI –6.1 to 5.4; p = 0.898). Similarly, after adjusting for stratification variables, the average maximum number of repetitions proportion [calculated as (maximum number of repetitions inthe injured leg/maximum number of repetitions in the uninjured leg) × 100, to account for potential ability differences] achieved by participants in the PRP group was 50.1% (95% CI 43.7% to 56.4%), compared with 60.7% (95% CI 52.8% to 68.7%) in the placebo group. This difference was not statistically significant (mean difference –10.7, 95% CI –21.9 to 0.6; p = 0.061).

Visual analogue scale

Participants’ pain levels as recorded in their pain diaries in the first 2 weeks post intervention demonstrate that the pain experienced by participants decreased at a relatively steady rate from day 1 to day 14 (Figure 10).

FIGURE 10.

Pain VAS mixed-effects regression model. Results from the adjusted repeated-measures mixed-effects regression model demonstrating the change in VAS pain measure in PRP and placebo group participants in the first 2 weeks post intervention. 0 = no pain; 100 = severe pain.

Accounting for the stratification factors, participants in the PRP group reported a mean overall VAS score of 37.3 (95% CI 33.0 to 41.7) at day 1, with the placebo group reporting a similar level of 31.2 (95% CI 26.7 to 35.7). By day 14, participants in the PRP group were reporting a mean VAS score of 9.5 (95% CI 5.2 to 13.9) and participants in the placebo group were reporting a mean VAS score of 13.6 (95% CI 9.0 to 18.1). This difference at day 14 was not significant (mean difference –4.0, 95% CI –10.3 to 2.3; p = 0.210). This demonstrates that the decline in reported pain was at similar rates in the PRP and placebo groups.

Short Form questionnaire-12 items quality of life

Participants’ SF-12 physical component scores (PCSs) increased gradually over time to near-normal levels at 24 weeks post treatment (Figure 11) at an almost identical rate in the PRP and placebo groups, irrespective of stratification variable and accounting for participants’ pre-injury PCSs (Table 12).

FIGURE 11.

The SF-12 PCS mixed-effects regression model. Results from the adjusted repeated-measures mixed-effects regression model demonstrating the change in PCS in PRP and placebo group participants over time, with mean pre-injury score (with 95% CIs) shown.

Participants’ SF-12 mental component scores (MCSs) remained relatively stable and close to their pre-injury scores over the entire trial follow-up period (Figure 12). At 24 weeks post treatment, participants in the PRP group had a MCS that was nearly 3 points lower than that of the MCS in the placebo group after adjusting for stratification variables and accounting for participants’ pre-injury MCSs (p = 0.035) (Table 13).

FIGURE 12.

The SF-12 MCS mixed-effects regression model. Results from the adjusted repeated-measures mixed-effects regression model demonstrating the change in MCS in PRP and placebo group participants over time.

Blood and platelet-rich plasma samples

Venous blood (1–5 ml) was taken from all participants recruited into the trial and anticoagulated within ethylenediaminetetraacetic acid vacutainers (Becton, Dickinson and Company, Plymouth, UK). In the PRP treatment group, an additional 50 ml of venous blood was taken for automated sterile PRP preparation in the centrifuge; 8 ml of sterile PRP was then harvested and 4 ml was used for injection into the Achilles tendon, and the remaining 4 ml was divided into four 1-ml aliquots. One microtube was immediately frozen at –70 °C for storage until the end of the trial when it was used for batch measurement of growth factor levels (see below). The remaining PRP aliquots (each 1 ml) were stimulated at room temperature for 5 minutes by addition to two tubes (Platelet Solutions Ltd, Nottingham, UK) containing either (tube B) saline alone, to provide an unstimulated baseline, or (tube C) adenosine diphosphate and U46619, to fully activate the platelets. Both samples were then fixed using 1 ml of PAMFix (Platelet Solutions Ltd, Nottingham, UK). The whole blood, unfixed PRP and tubes B and C were then transported at room temperature by courier to the Institute of Inflammation and Ageing at the University of Birmingham and processed as soon as possible on arrival.

Cell counting

Whole-blood and unfixed PRP cell counts were conducted using the Sysmex XN-1000 haematology analyser (Sysmex UK, Milton Keynes, UK). The analyser utilises three primary analysis principles: fluorescence flow cytometry, direct current detection with hydrodynamic focusing and sodium lauryl sulfate haemoglobin detection. The WNR (white cell nucleated) channel evaluates white blood cell count, basophils and nucleated red blood cells. An impedance red blood cell count is also reported. The WDF (white blood cell differential) channel also provides a differential count of the neutrophils, lymphocytes, eosinophils, monocytes and immature granulocytes. The platelet fluorescence (PLT-F) channel is a specialised fluorescence optical analysis exclusively for platelets. The PLT-F parameter utilises traditional fluorescence flow cytometry in which platelets are stained with oxazine (a ribonucleic acid binding dye that eliminates any interference mediated by cellular debris). A measurement of platelet production (immature platelet fraction) is also in this channel. The analyser thus reports three platelet counts (platelet impedance, platelet optical and PLT-F), mean platelet volume and the immature platelet fraction. Quality control material (XN Check™, Sysmex UK, Milton Keynes, UK) was tested on a daily basis to ensure instrument performance throughout the trial. The instrument was also enrolled in a national external quality assurance scheme (UK NEQAS, Watford, UK) and maintained on a service contract (Sysmex UK). To determine normal ranges for each cellular parameter, blood samples from 40 healthy volunteers (control cohort) were analysed and normal ranges are represented as the mean value ± 2 SDs.

Measurement of platelet quality

Tubes B and C containing fixed resting and activated platelets, respectively, were analysed for the expression of a platelet-specific activation marker P-selectin (CD62P) by flow cytometry as a measure of platelet quality. After gentle mixing, 5 µl of fixed PRP from tubes B and C was incubated with 5 µl of test antibody [CD62P conjugated to flourescein isothiocyanate (FITC); Beckman Coulter Inc., Brea, CA, USA] or its corresponding isotype control (IgG conjugated to FITC; Beckman Coulter Inc.) with 40 µl of 0.2-µm filtered phosphate buffered saline (pH 7.4) for 20 minutes at room temperature. The samples were then further diluted by the addition of 4 ml of phosphate-buffered saline. The platelets were then analysed by flow cytometry (Accuri™ Flow cytometer, Becton, Dickinson and Company). Platelets were identified according to their characteristic size and granularity (using log-forward scatter and log-side scatter), with a forward-scatter threshold of 25,000 to eliminate noise. An amorphous gate was placed around the platelet population and 10,000 events were collected. Analysis regions were set using the isotype control so that 0.5% of the platelet population was positive. CD62P positivity was then measured and both percentage expression and median fluorescent intensity of all platelet events were recorded.

Measurement of growth factor levels within platelet-rich plasma

At the end of participant recruitment, all frozen aliquots of PRP were shipped to Birmingham for analysis of growth factor levels. Appropriate batches of samples were thawed at 37 °C for 10 minutes. One-fortieth volume of 20% Triton™ X-100 (Roche Diagnostics, Mannhein, Germany) was added to ensure full lysis prior to a further incubation at 37 °C for 10 minutes. Samples were mixed then centrifuged at 1500 g for 10 minutes at room temperature. Supernatants were then removed and simultaneously assayed for five different growth factors (TGF-β1, FGF, VEGF, IGF-1 and PDGF-AB) by commercial enzyme-linked immunosorbent assay (ELISA) kits (Bio-Techne Ltd, Abingdon, UK). Optimal sample dilutions for each growth factor were predetermined by assaying L-PRP samples prepared from control volunteers using the identical method of preparation as in the trial. Diluted samples, blanks and standards were pipetted in duplicate into 96 well plates coated with capture antibodies and incubated as instructed within each growth factor ELISA kit. Plates were then washed four times with the washing buffers provided and the peroxidase-conjugated secondary antibodies provided within each kit were added to each well and incubated as instructed. Plates were then washed four times with washing buffer before addition of substrate solution to each well. The plates were then incubated for 20–30 minutes at room temperature in the dark before the addition of the stop solution. The optical density of each well was then measured immediately using a microplate reader and readings taken at 450 nm. A further reading was taken at 540 nm and the values subtracted from the readings at 450 nm to correct for optical imperfections in the plate. The duplicate readings were averaged for each standard, control and sample, and the zero standard optical density was subtracted. Standard curves were generated by plotting the mean absorbance for each standard on the y-axis against the concentration on the x-axis and the best-fit line generated by regression analysis. Unknown sample values were then read off the regression line and final concentrations were determined by multiplying using the dilution factors used in each assay. The concentration for each growth factor per ml of lysate was reported.

Statistical analysis

The comparability of key blood parameters and bioactive factors for each treatment group was summarised with means and SDs. For whole-blood parameters, a Student’s t-test was used to explore differences between treatment arms.

Correlation between the primary outcome and both key blood parameters and bioactive factors was explored using Pearson’s correlation.

As per analyses in the main study, a p-value of < 0.05 was seen as indicative of a statistically significant difference.

Cell counts within whole blood and platelet-rich plasma

Summary statistics for key blood parameters in whole-blood and PRP samples are given in Table 16. A comparison of the whole-blood counts in the PRP and placebo groups is shown in Figure 13. There was no significant difference in the whole-blood samples between the treatment groups for red blood cell count (mean difference –0.002, 95% CI –0.144 to 0.139; p = 0.974), white blood cell count (mean difference –0.259, 95% CI –0.784 to 0.264; p = 0.330) or platelet count (as PLT-F) (mean difference –19.045, 95% CI –38.310 to 0.040; p = 0.051). Figure 14 illustrates the cell counts within the PRP samples showing the variation in platelet, red blood cell and white blood cell counts, with the whole-blood sample summaries given in the same figure for reference. The mean platelet count (PLT-F) in the PRP sample was 852.5 × 10⁹/l, but with a wide range (6.0–2903.0 × 10⁹/l). Although a small number of PRP platelet counts were lower than original whole-blood counts due to some unforeseen centrifuge problems, the overall mean increase in platelets was 4.1-fold. The mean red blood cell count was 0.9 × 10¹²/l (range 0.1–9.0 × 10¹²/l). As expected, the red blood cell count decreased on average in the PRP by a factor of 5.3 of the original count in whole blood. The mean white blood cell count was 15.1 × 10⁹/l (range 1.7–65.3 × 10⁹/l). As this preparation is a L-PRP, the mean white blood cell count increased by a factor of 2.2, as expected. The white blood cell differential count was 14.6% monocytes (mean 2.2 × 10⁹/l, range 0–21.1 × 10⁹/l), 45.2% lymphocytes (mean 6.8 × 10⁹/l, range 0.1–18.7 × 10⁹/l), 39.2% neutrophils (mean 5.9 × 10⁹/l, range 0.2–30.4 × 10⁹/l), 0.1% basophils (mean 0.02 × 10⁹/l, range 0–0.5 × 10⁹/l) and 0.82% eosinophils (mean 0.1 × 10⁹/l, range 0–1.5 × 10⁹/l).

FIGURE 13.

Cell counts in whole-blood samples, by intervention group. (a) Erythrocytes; (b) leucocytes; and (c) platelets. a, Using the PLT-F measure.

FIGURE 14.

Cell counts in PRP samples, represented against counts in whole-blood samples. (a) Erythrocytes; (b) leucocytes; and (c) platelets. a, Using the PLT-F measure.

Platelet quality

The basal levels of CD62P expression [percentage and mean fluorescence intensity (MFI)] within the platelets of resting basal PRP and activated PRP samples are shown in Table 17 and Figure 15. In resting PRP, the mean CD62P expression was 4.3% with a MFI of 248.6. The majority of PRP samples were therefore of good quality with low levels of activation, with a few exceptions. In the activated PRP samples, the mean CD62P expression was 60.1% with a MFI of 1208.0. The majority of PRP samples were functional, with the majority of platelets in the PRP preparations being shown to be capable of activation and degranulation, with a few exceptions.

FIGURE 15.

Platelet quality in PRP samples, demonstrating both (a) CD62P expression and (b) MFI in activated and resting (unactivated) PRP.

Growth factor levels within platelet-rich plasma

The levels of each growth factor are shown in Table 16 and Figure 16.

FIGURE 16.

Growth factor concentrations in PRP samples.

Blood and platelet-rich plasma variables and clinical outcomes

Baseline blood and the primary outcome measure

Parameters of baseline whole blood taken before the intervention did not correlate with the primary outcome measure at 24 weeks (Figure 17 and Table 18), irrespective of intervention group.

FIGURE 17.

Correlation between whole-blood sample blood cell counts and the primary outcome (work LSI) at 24 weeks. (a) Erythrocytes; (b) leucocytes; and (c) platelets. a, Using the PLT-F measure.

TABLE 18 - Results from primary outcome correlation assessments overall and by intervention group, for whole-blood sample key blood parameters

r, Pearson’s product-moment correlation coefficient.

Proportion of variance in LSI explained by key blood parameter, calculated as (R² × 100).

As PLT-F.

Parameter by group	n	r	% variancea	p-value
Erythrocytes
PRP	97	0.133	1.780	0.193
Placebo	99	0.122	1.496	0.228
Leucocytes
PRP	97	–0.184	3.393	0.071
Placebo	99	–0.172	2.976	0.088
Plateletsb
PRP	94	–0.153	2.329	0.142
Placebo	96	–0.080	0.634	0.440

Platelet-rich plasma and the primary outcome measure

The PRP cell count and growth factor concentration correlation with the primary outcome measure were analysed to assess if the variability in PRP had any effect on the outcome. The results are displayed in Figures 18–20 and Table 19.

FIGURE 18.

Correlation between PRP sample blood cell counts and the primary outcome (work LSI) at 24 weeks. (a) Erythrocytes; (b) leucocytes; and (c) platelets. a, Using the PLT-F measure.

FIGURE 19.

Correlation between platelet quality and the primary outcome (work LSI) at 24 weeks. (a) Activated CD62P expression (%); and (b) activated CD62P MFI expression (%).

FIGURE 20.

Correlation between growth factors and the primary outcome (work LSI) at 24 weeks. (a) IGF-1; (b) TGF-β1; (c) PDGF-AB; (d) VEGF; and (e) bFGF. bFGF, basic fibroblast growth factor.

TABLE 19 - Results from primary outcome correlation assessments overall and by intervention group, for PRP sample key blood parameters

bFGF, basic fibroblast growth factor; r, Pearson’s product-moment correlation coefficient.

Proportion of variance in LSI explained by key blood parameter, calculated as (R² × 100).

As PLT-F.

Parameter	n	r	% variancea	p-value
Blood cell counts
Erythrocytes	91	0.126	1.590	0.234
Leucocytes	91	–0.103	1.053	0.333
Plateletsb	88	0.128	1.649	0.233
Platelet quality
Activated CD62P expression (%)	92	0.038	0.142	0.721
MFI	92	–0.004	0.002	0.966
Growth factors
IGF-1	93	0.124	1.248	0.287
TGF-β1	88	0.005	0.003	0.960
PDGF-AB	90	< 0.001	< 0.001	0.998
VEGF	93	–0.231	5.355	0.026
bFGF	93	–0.102	1.049	0.329

The red blood cell, white blood cell and platelet (as PLT-F) counts and work LSI at 24 weeks linear regression analysis showed no significant correlation (see Figure 18). This was also the case for the platelet-quality measures of interest (see Figure 19) and for all growth factors (see Figure 20) with the exception of VEGF, for which the evidence suggests a negative correlation between PRP VEGF concentration and week 24 LSI performance (p = 0.026), with around 5% of the variance in LSI being explained by this growth factor.

Promotion to patients and the public

Each recruiting centre displayed an ethics-approved patient poster in the fracture clinic, to make patients aware of the trial before they were approached to participate.

The PATH-2 trial has a public-facing web page on the OCTRU trials website (https://path2.octru.ox.ac.uk/; accessed August 2019), where trial information, current recruitment figures and news are publicised. An additional page called Reach Target Event was created in April 2018 with the aim of encouraging site staff to screen and recruit all eligible patients. This was promoted as an alternative to a study day for site staff, and included videos about the trial and the scientific objectives, discussion forums, competitions and cartoons. The Reach Target Event page is now available to the public and was entered into NIHR’s ‘Let’s Get Digital’ competition in June 2017.

Promotion within the trauma and scientific communities

The PATH-2 trial design has been presented at:

• the fifth NIHR Musculoskeletal Trauma Trials Annual Meeting, Warwick, 11 January 2017

• the Trauma Orthopaedic Research Collaboration, Oxford, 6 October 2017

• the sixth NIHR Musculoskeletal Trauma Trials Annual Meeting, Bristol, 10 January 2018

• the fourth Annual Celebrating Trauma Research in the Thames Valley, Reading, 28 February 2018

• Tendon UK, Oxford, 11–12 April 2018.