The use of a synthetic shoulder patch for large and massive rotator cuff tears – a feasibility study

The Shoulder Patch for Rotator Cuff (SPARC) study was a pragmatic, two-arm feasibility study. The study protocol was approved by Leeds West Research Ethics Committee (13/YH/0030) and registered on ISRCTN (ISRCTN79844053). Written informed consent was obtained for all participants prior to screening. Participants were recruited from 1st September 2013 to 20th October 2014. For the purposes of this feasibility study participants were followed up for 6 months (follow-up completed 20th October 2015).

Participants were recruited through the Leeds Teaching Hospitals NHS Trust orthopaedic clinics upon referral for rotator cuff surgery. Participants were eligible if they had ultrasound or magnetic resonance imaging (MRI) findings confirming massive rotator cuff full thickness tears (RoCT) (> 5 cm in size), with unacceptable pain and disability following conservative treatment or previous surgery that had failed and would be considered for patch surgery under routine clinical practice. Exclusion criteria were age < 18 years old; history of infection in the shoulder; neurological condition affecting the shoulder girdle; presence of rotator cuff arthropathy (secondary osteoarthritis of the glenohumeral joint as a result of a rotator cuff tear); current treatment for malignancy; pregnancy or lactation; currently participating in other research studies; or inability to give informed consent. Participants with contraindications to MRI were included but did not take part in the MRI component of the study.

The particular intervention was determined in consultation between the surgeon and the participant (Fig. 1 and Additional file 1: Table 1). Management decisions were made at two stages. First, the surgeon and participant jointly made a decision between either conservative management (the anterior deltoid rehabilitation programme) or surgery. Second, an intraoperative decision on surgical management was made by the surgeon. If the torn tendon was mobile and could be pulled back to the tuberosity, arthroscopic repair was performed; if the torn tendon could be mobilised following release to within 1 cm of the tuberosity to allow attachment of a patch, repair was performed using a patch; and if the torn tendon could not be mobilised following release, a debridement was performed. In summary, participants in the ‘Patch’ group had elected surgical management and had massive cuff tears which could not be fully reduced at surgery but were suitable for repair by patch augmentation. The ‘Non-patch’ group had massive tears and elected for either non-surgical rehabilitation, or surgical intervention involving a procedure other than a patch.

The Leeds Kuff Patch (Xiros, Leeds UK) is produced from polyethylene terephthalate (PET, polyester), a non-absorbable biocompatible material that has been in use for the reconstruction of ligaments and tendons for over 25 years. Early generations of artificial tissue contained polytetrafluorethylene and polypropylene, which resulted in inflammatory reactions within the surrounding tissues [34] . However, more modern techniques and materials promote new tissue ingrowth and minimize the occurrence of a foreign body reaction [35].

The design of the patch comprises a base component with an integral reinforcement component. This base component has an “open structure” that acts as a scaffold allowing tissue ingrowth. The polyester is rendered hydrophilic during the manufacturing process while the reinforcement provides enhanced strength for the patch.

Participants were followed up for 6 months with data collected using standardised case report forms at baseline (pre-treatment), 6 weeks and 6 months post-treatment. At all visits, clinical evaluations were performed. Baseline clinical data collection included age, gender, body mass index, duration of symptoms, range of movement, previous treatments for shoulder (including surgery), details of all therapies (including use of intraarticular therapies) in the previous 12 months. Participants completed a series of patient reported questionnaires at each visit to assess pain, function and patient satisfaction (Oxford Shoulder Score (OSS) [36], Shoulder Pain and Disability Index (SPADI) [37], and the Constant Score [38] and quality of life was assessed using the EuroQol five dimensions questionnaire (EQ-5D-5 L) [39]. An increase in the value of the OSS and Constant score at follow-up represents improvement, while for SPADI, a decreased score represents improvement in pain and function. Participants with no MRI contraindications had an MRI scan at baseline and 6 months.

MRI scans of the shoulder were performed using a pre-agreed protocol using a MAGNETOM Verio 3.0 T MRI system (Siemens Healthcare, Erlangen, Germany). Fat and water images were generated from the Dixon images using the scanner vendor’s software. B0 variations due to changes in magnetic susceptibility at tissue interfaces were corrected for using a phase correction method [40]. Imaging analysis was performed using the OSIRIX system (Geneva, Switzerland). Images were contoured on the fat only images from the spinoglenoid notch medially to the first slice showing continuous bone marrow in the scapula laterally. This established a reproducible volume in patients, which did not suffer from signal drop off as previously described [41]. Regions of interest (ROIs) were drawn for the purpose of measuring fat fraction values. The 3 ROIs drawn outlined the supraspinatus fossa, bounded superiorly by the trapezius muscle (SSP), the suprapinatus remnant (SSP_RM) (defined as the muscle remnant within the SSP) and the infraspinatus muscle (Fig. 2). Fat fraction values were calculated for each ROI as the mean voxel value in the fat image ROI divided by the sum of the mean voxels values in the fat and the water images ROIs (fat/(fat+water)). Contouring was undertaken using the methods adopted by Zanetti et al. [41] by a single reader blinded to treatment allocation.

Muscles were also graded with the Goutallier Classification [42] using the 0–4 grading system.

As no hypothesis testing was anticipated no formal power calculations were performed. Published evidence for pilot study sample size suggests that 30 patients per group represent a good balance between accuracy of effect size estimation and feasibility [43‐46]. Therefore we aimed to recruit 60 patients in total. Early in the study the protocol was amended to enable over-recruitment in order to allow for the recruitment of participants who met study criteria but were ineligible for MRI whilst still ensuring a total of approximately 60 participants took part in the MRI component of the study.

Statistical analysis was conducted using STATA software, version 13 (College Station, TX, USA 2013). To understand feasibility, methodological issues relevant to this study were reported according to the framework devised for feasibility studies by Shanyinde et al. [47, 48]. Recruitment and eligibility was assessed using rate of eligibility (%), the conversion of eligible to consent (%) and an assessment of whether required numbers sought were recruited. Adherence to the protocol was assessed qualitatively by broadly reviewing the study objectives against our findings, and the acceptability assessed using the attrition rate (%) [49]. The outcome data was assessed based on the amount of missing data (%) found in each case report form.

The patch and non-patch groups were compared descriptively at baseline. At 6 month follow-up, differences between the medians in each group and associated 95% confidence intervals were assessed using quantile regression, adjusting for baseline scores. The groups were also compared for the proportions in each group that had improvement greater than the previously-reported minimum clinically important difference (MCID) in each shoulder-specific outcome measure [50]. This was expressed as percentages and odds ratios with 95% CI and adjusted for the baseline score for each outcome. MCID for OSS was set at 5 units, SPADI at 15.4 [51] and Constant score at 10.4 [52]. No hypothesis testing was undertaken as the study did not aim to draw inferences from the data.

Improvement was seen at follow-up for all clinical outcomes in both patch and control groups, except for the EQ-VAS which was slightly reduced in the control group (Table 3). Overall, improvements in the patch group were of a greater magnitude than for the non-patch control group. The median score was higher in the patch group compared to controls for the OSS (baseline-adjusted difference 9.76, 95% CI 2.25, 17.29, p = 0.01) and lower for SPADI (difference 22.97, 95% CI 3.0, 42.92, p = 0.03) but no substantive differences were seen for the total Constant score (p = 0.59). Compared to the non-patch group, the patch group had a greater proportion of participants demonstrating improvement (change greater than MCID) in two of the shoulder-specific patient reported outcomes, although not statistically significant: 78% vs 62% for OSS (Odds ratio 2.94, 95% CI 0.78, 11.02,p = 0.11); 63% vs 50% for SPADI (Odds ratio 2.66,95% CI 0.72,9.78, p = 0.15) but had a lower proportion of participants demonstrating improvement for the Constant score (50% vs 62% [Odds ratio 0.57,95% CI 0.18,1.90,p = 0.35)]. Quality of life scores (EQ-5D Index and EQ-VAS) were slightly improved in the patch group compared to the non-patch group.

Results of the sub-analyses showed no substantive differences, all p > 0.05 (Table 4). All comparator groups showed improvement at 6-month follow up with slight variation in the magnitude between them. Compared to the patch-poor group, the patch-better group had better improvement in the Constant score (9.6 score difference between the groups) and less improvement in SPADI (2.4 score difference) having adjusted for their baseline scores; and no difference between the groups for OSS. The non-patch (excluding arthroscopy) group improved by a greater magnitude compared to the non-patch arthroscopy group in all 3 outcome scores although this was also not substantive.

Of the 68 patients eligible for the study, 58 underwent MRI at baseline. Of the patients who did not have a baseline MRI, eight had metallic implants contraindicated to MRI, while one suffered claustrophobia. Another patient attended the scan, but could not tolerate it so it was halted prematurely without sufficient images for analysis. At 6 month follow-up 54/58 underwent MRI, meaning 4 additional patients declining or not attending, alongside the still contraindicated 10 patients from baseline, totalling 14 who did not have the 6 month MRI. On analysis of the MRI scans, for 4 patients, the fat and water images could not be reconstructed from the “in” and “out of” phase images. One patient was imaged in the wrong plane: a technical error in MRI protocol. One patient subsequently withdrew from the study, so their data was not analysed. This resulted in 48 participants with paired MRI scores available: 22 in the Patch group, and 26 in the non-patch group. Complete paired Goutallier data was available for 41 participants.

In the full sample, there was only a slight increase in the fat fraction for supraspinatus (SSP; 53 to 55%) and the infraspinatus (ISP; 26 to 29%) while the supraspinatus remnant (SSP_RM) showed no change (Table 5). These differences were similar and in the same direction when the participants were analysed separately by treatment status (patch vs nonpatch).

A change in Goutallier classification was defined as a difference of at least 1 grade. For the SSP, 11 participants worsened over time (5 in the controls and 6 in the patch), while 30 remained the same and 1 improved. Results were similar for ISP and teres minor.

This feasibility study was designed as a pragmatic study, and thus randomisation and blinding was not assessed. A future larger scale study would likely compare patch-assisted surgery with non-patch assisted surgery, removing the non-surgical component within this feasibility study. As such, we would envisage the use of a standardised intra-operative decision system for surgeons, which would determine whether a participant was eligible for patch-assisted surgery and participants would be randomised at this point into one of two surgical treatment arms. Although the surgeon and potentially the patient (as a result of patch surgery requiring a larger opening, and therefore a larger scar, than non-patch arthroscopic surgery) would be aware of the treatment allocation, an independent assessor could conduct subsequent follow-up visits to reduce potential bias.

The moderate attrition rate at 6 months (11 lost to follow-up) suggests that the protocol was acceptable for participants; however, we did make the decision to remove the 6 week follow-up from the protocol early in the study due to the difficulty of patients in completing functional tests so early after any form of surgery. Missing data was minimal possibly due to the use of a number of routinely used shoulder questionnaires that were easy to complete. There was a noticeably lower completion rate/retention in the non-patch group, with half of these having received non-surgical management. It is possible that having made the decision not to proceed with surgery that these patients became disengaged from involvement in a ‘surgical’ study and there may also have been disengagement from the surgical team since these patients were no longer under their clinical care.

Various scoring systems are quoted in literature for post-operative assessment of rotator cuff surgery. In this current study we selected the OSS and Constant score, since they were the most well validated and frequently used to determine outcomes of RoCTs [9, 57]. In addition, we included the SPADI which has been shown to have reasonable validity and, although devised primarily as a rheumatological outcome score [37, 58], has been used to assess outcomes of RoCTs in the past [59]. We found that the Constant score had the most missing data fields of the three scores, which may reflect the multi-domain nature of the tool with functional tests which are difficult for some RoCT patients to complete and a mixture of patient-reported and clinician-reported outcomes. A recent systematic review of the psychometric properties of patient-reported outcomes in patients with rotator cuff disease found good evidence in support of the measurement properties of SPADI, limited overall evidence for OSS and mixed evidence for the Constant score, with positive evidence for responsiveness and reliability but negative evidence for hypothesis testing [60]. A recent international consensus process involving patients, clinicians and researchers, has defined the core outcome domains for shoulder disorder trials as ‘pain’, ‘physical functioning’, ‘global assessment of treatment success’ and ‘health-related quality of life’ [61]. Work to define a core outcome set based on these domains for clinical trials of people with shoulder pain is ongoing. A future study would therefore incorporate recommendations from this work to ensure alignment with international standards.

The level of MRI acceptability and attendance was high, despite ten patients with contraindications to such scans. In addition, the data available from the scans was of reasonable reliability and accuracy, with only 7 of the 48 paired scans having missing data for analysis of fat fraction and/or Goutallier Classification, which was due to technical errors with the protocols set for those patients’ scans.

This study provides preliminary evidence that any standard shoulder intervention may provide improvement in outcome scores, but that patch repair provides greater improvement in all outcomes at 6 months. The differences between the patch group and controls were substantial (greater than MCID) for the Oxford Shoulder Score. On further subgroup analysis, no substantive differences were found between patch patients with good quality and poor quality tendons, although this finding may have been hampered by small numbers.

MRI outcomes at 6 months demonstrated little difference between the patch group and control group, with both groups demonstrating either similar or slightly worse (i.e. higher) fat fraction within the rotator cuff. It may be that 6 months is too early in the rehabilitation phase following any treatment method to note anatomical changes. To confirm our findings of clinically meaningful improvement using a patch, a definitive trial is needed.