To investigate the use of a novel surface coil for clinically utilized magnetic resonance imaging (MRI) scanners, in order to describe the microanatomic basis for hand osteoarthritis (OA) at all stages of disease.
To investigate the use of a novel surface coil for clinically utilized magnetic resonance imaging (MRI) scanners, in order to describe the microanatomic basis for hand osteoarthritis (OA) at all stages of disease.
MRI of proximal or distal interphalangeal joints was performed in 58 subjects: 16 patients with early OA (symptom duration ≤12 months), 14 patients with chronic OA, 10 patients with clinically normal asymptomatic joints adjacent to arthritic joints, and 18 normal controls. High-resolution images were obtained with displayed pixel dimensions of 80–100 μm using a 1.5T scanner and a 23-mm–diameter surface coil. All joint structures were evaluated.
The high-resolution images of every joint structure showed comparable abnormalities in both early and chronic OA, including cartilage loss, bone edema, synovial enhancement, osteophytosis, and erosions. Heberden's and Bouchard's node formation occurred at regions where soft tissue bulged through the capsule between the dorsal tendons and collateral ligaments (CLs). Prominent CL thickening or disruption (100% of OA patients) was evident even in joints where cartilage was partially preserved. Clinically normal joints adjacent to OA hand joints showed thickening and enhancement of CLs which was the most common abnormality seen (80% of OA patients). Older normal subjects showed subtle changes within the CLs.
Obtaining high-resolution MR images from clinically utilized scanners represents a novel way for exploring the microanatomic basis of hand arthritis and may have considerable potential in the clinical setting. In the present evaluation in nodal OA, previously unappreciated CL abnormalities were especially common.
The hand is commonly involved in both inflammatory arthritis and osteoarthritis (OA), but the microanatomic basis for hand disease damage and localization is unclear. While synovitis is the primary abnormality in rheumatoid arthritis (RA), and enthesitis/osteitis and synovitis are central to the pathogenesis of spondylarthropathy (SpA), the microanatomic basis for hand OA, especially at the earliest stages of disease, is less apparent.
Osteoarthritis has been viewed primarily in relation to cartilage loss, but the importance of bone abnormalities and joint inflammation in OA has also been well documented (1, 2), as has the need to consider the whole joint organ in disease pathogenesis (3). Conventional magnetic resonance imaging (MRI) has been used to explore the anatomic sites of abnormality in large joint OA, and has enhanced our understanding of disease mechanisms in the knee and hip, where both subchondral bone edema and synovitis are associated with pain, and bone edema itself is associated with progressive disease (4–6).
Knee OA may be difficult to recognize at clinical presentation and patients with knee pain but normal radiographic findings would not normally be referred to a rheumatologist. The clinical pattern of the early stages of distal interphalangeal joint (DIP) and proximal interphalangeal joint (PIP) involvement in OA is often diagnostic, and does not rely on radiographic criteria, which are of limited use in diagnosis of early hand disease and disease at other sites prone to OA, including the knee (7, 8). Furthermore, patients with hand OA are more likely to present to the early arthritis clinic than are those with early knee symptoms; hence, there is the potential to recognize and evaluate OA soon after symptom onset. The spatial resolution limitation of currently available MRI has precluded detailed high-resolution imaging assessment of hand OA. However, previous small pilot studies using custom-built high-resolution MRI equipment showed that MRI was a valid method for the assessment of small joint structures, including the bone trabeculae, cartilage, ligaments, and tendons (9, 10). Dedicated high-resolution MRI scanners have also been developed for animal research, where they show considerable potential in the evaluation of therapy for joint inflammation (11).
An anatomic understanding of the earliest stages of hand OA may have important implications for elucidating its pathogenesis, and high-resolution MRI represents a potentially useful way for such a detailed assessment. This study optimized recently developed high-resolution MRI coils, that can be fitted to existing commercially available scanners routinely used in the clinical setting, to explore the microanatomic basis for hand OA.
Fifty-eight subjects provided informed consent and participated in the study. The study was approved by the local ethics committee at the Leeds Teaching Hospitals National Health Service Trust. In total, images from 40 patients with hand OA and 18 normal control subjects were scanned. All subjects were invited to participate in the study either from rheumatology outpatient clinics or via leaflets and posters about the study distributed in the clinics. Either a DIP or a PIP joint was scanned. Sixteen female patients with a mean age of 56 years (range 49–69 years) had early OA (8 DIP joints, 8 PIP joints) defined by disease duration of ≤12 months (mean 8 months). The disease duration was determined from the onset of symptoms (pain, swelling, or tenderness) in any of the finger joints with clinical OA. Fourteen patients (11 women, 3 men) with a mean age of 60 years (range 51–68 years) and OA symptoms termed chronic (duration >12 months [mean 75 months]) were also imaged (9 DIP joints, 5 PIP joints). Findings on radiography of the hands can appear normal at clinical presentation of OA, and therefore, radiographic depiction of OA was not mandatory for the diagnosis of OA in this study. The patients with OA were diagnosed on clinical grounds with DIP or PIP joint soft tissue swelling or bony thickening and the absence of other arthropathies, including RA, psoriatic arthritis, and gout, or any traumatic injury to the joint selected for imaging.
Ten patients (8 women, 2 men; mean age of 60 years [range 46–72 years]) with hand OA (mean disease duration 120 months; 5 DIP joints, 5 PIP joints) and at least 1 clinically normal joint were randomly selected from the OA cohort, for MRI of the clinically normal joint that had never been symptomatic. This group may represent the earliest stages of OA, since disease tends to progress to affect the uninvolved finger joints over time (7). This group of patients are henceforth referred to as the “latent OA” group.
Eighteen healthy volunteers who had no known OA with asymptomatic finger joints also had 1 finger joint scanned (10 DIP joints, 8 PIP joints). The healthy volunteer group was divided into those ≤40 years old (mean 34 years [range 30–37 years]; 4 women, 4 men) and those >40 years old (mean 55 years [range 42–72 years]; 5 women, 5 men). This group was divided at age 40 years arbitrarily because clinical hand OA usually presents at ∼50 years of age.
MR images were acquired using a 1.5T Gyroscan ACS-NT scanner (Philips, Best, The Netherlands). Subjects were placed in the prone position with the hand extended in front of the body. The surface coil was placed on top of the relevant joint and held in place using sandbags. The DIP and PIP joints are small in relation to most other joints that are routinely studied using MRI. Examination of these joints therefore required methods that measured the MR data at adequate resolution to distinguish the relevant joint features. In previous investigations this was achieved by studying fingers from cadavers, with long imaging times and/or use of specialized imaging coils to obtain the required signal-to-noise levels (9, 10, 12, 13). In the present study, a Philips 23-mm–diameter surface “microscopy” coil was used to acquire adequate signal levels from the small voxels required for high-resolution study of the DIP and PIP joints.
The examination protocol was designed to visualize the different joint structures. This protocol included the acquisition of T1-weighted axial and coronal spin-echo (SE) images for the assessment of the anatomy of most of the structures, T2-weighted fat-suppressed SE images to identify areas of edema and fluid in the joint space and cystic lesions, and proton-density–weighted 2-dimensional SE images to study the collateral ligaments (CLs). In addition, 3-dimensional gradient-echo sagittal images were acquired using a water-selective radiofrequency excitation pulse to assess the cartilage and tendons. After intravenous injection of 10 ml of the contrast agent gadolinium diethylene tetrapentaacetic acid (Gd-DTPA), axial and coronal T1-weighted fat-suppressed SE images were also obtained, highlighting areas of inflammation in the soft tissue and bone. The imaging sequences employed a field of view of either 40 or 45 mm and a data acquisition matrix of 256 × 204 with 512 × 512 Fourier transformed data points. The measured slices were 1 mm thick, and the acquired in-plane pixel dimension was therefore 160–200 μm with a displayed pixel dimension of 80–100 μm. The T1-weighted measurements were made with repetition time (TR) and echo time (TE) values of 475 msec and 18 msec, respectively, and 2 signal averages, while T2-weighted images were acquired using a turbo spin-echo (TSE) sequence and 3 signal averages with a TR value of at least 3,500 msec and a TE of 100 msec. The proton-density–weighted data were obtained using a TSE sequence with centric data acquisition, 2 signal averages, and TR/TE values of 2,000 msec/15 msec. A flip angle of 45° and TR/TE values of 3,000 msec/12 msec were used to obtain the gradient-echo images. The total examination time was ∼50 minutes.
MR images were read at a workstation (Philips Easy Vision) by 2 readers, including an experienced musculoskeletal radiologist. All groups (early OA, chronic OA, latent OA, and normal control) were analyzed in a random order. The observers were blinded to the clinical status and the category of the subjects. The scoring was performed by consensus between the 2 readers. The features of the following anatomic structures were described and evaluated in a dichotomous manner: cartilage, bone cortex, CL, extensor expansion, bone edema, cysts, erosions, osteophytes/nodes (Heberden's or Bouchard's), joint fluid, and capsule/synovium. Where determination of the location of the abnormality (such as bone edema, cysts, erosions, or osteophytes) was possible, this was also recorded.
MRI-determined joint erosions were recorded when a cortical break was evident on T1-weighted images on both coronal and axial images. In contrast, a well-circumscribed area of trabecular absence within the bone was termed a cyst. Osteophytes at either ligament origins or insertions or at tendon insertions were termed enthesophytes.
Regions of synovitis or capsulitis were defined on images acquired after the injection of the contrast agent. The T2-weighted fat-suppressed SE images were used to identify bone marrow edema–like lesions, henceforth referred to as bone edema, and the images were compared with the Gd-DTPA–enhanced images to distinguish between bone edema (showing enhancement) and cysts filled with fluid (nonenhancing).
Fisher's exact test was used to compare the proportion of abnormalities between the groups. Because this was a pilot study, no correction for multiple testing was performed, and the results of significance tests are presented as a guideline only. All analyses were performed with SPSS version 11.0 (Chicago, IL).
Early and chronic OA. All MR studies were satisfactory, with no subject having to be excluded on the basis of technical failure. Virtually every joint structure evaluated showed abnormalities in all patients with either early or chronic OA, reflecting the known fact that OA is a whole organ disease and ultimately a disease of joint failure (Figure 1). The findings in early and chronic OA are compared before presenting the features of latent OA and the normal group.
Articular cartilage. On coronal and sagittal imaging, generalized full-thickness loss of articular cartilage was evident in 4 of the 30 subjects, occurring in both the early (3 of 16) and chronic OA (1 of 14) groups, indicating marked cartilage loss at clinical presentation. A further 5 subjects had generalized full-thickness loss visible only on the coronal plane (2 with early OA and 3 with chronic OA). In the remaining 21 subjects in whom cartilage loss was incomplete and variable, an ulnar predilection for cartilage loss was observed, with 10 of 21 patients (48%) experiencing ulnar loss, as compared with 5 of 21 (24%) experiencing radial loss and 6 of 21 (29%) patients experiencing symmetric but not full-thickness loss. On sagittal imaging, cartilage loss was most pronounced on the volar joint surfaces (Figure 2).
MRI-determined subchondral bone sclerosis, seen as low signal change in the subchondral bone on all sequences, was evident on the distal side of the PIP and DIP joints in 9 of 16 patients with early OA (56%) and in 13 of 14 with chronic OA (93%) (P = 0.04). The trabecular architecture appeared normal and discrete breaks in the subchondral bone, suggestive of microfractures, could not be identified with the imaging resolution obtained in this study.
Collateral ligaments. Ligament abnormalities were universal in both chronic and early disease. Changes seen ranged from thickening and increased signal within the ligament to complete disruption (Figure 3), with Gd-DTPA enhancement evident in the abnormal ligaments (Figure 4). Complete disruption of both CLs was evident in 16 of 30 patients (9 of 16 with early OA and 7 of 14 with chronic OA). While ligament disruption was common in the ligament midsubstance, increased signal was evident throughout the abnormal ligaments, being seen at their origin, midregion, and insertions. This increased signal was seen on both T1- and T2-weighted imaging and would be consistent with myxoid degeneration of the ligament, such as that seen in other ligamentous and tendinous structures. Edema was seen in the adjacent extracapsular tissues.
In 14 of 30 patients in whom complete CL disruption was not a feature, the CLs were abnormal, appearing thickened or with an indistinct appearance, particularly at the proximal attachment. Furthermore, 6 of 14 patients showed Gd-DTPA enhancement of both CLs, with only 1 of the CLs enhanced in 5 of 14. Three patients with chronic OA had no enhancement in the abnormally thickened CLs. In general, all patients with early and chronic OA had some degree of ligament abnormality, and all those with early OA showed Gd-DTPA enhancement (16 of 16), whereas 11 of 14 with chronic OA had signs of enhancement (P = 0.09).
When the pattern of ligament abnormality was compared with the articular cartilage damage in the joint, it was noted that among the 21 of 30 subjects who had cartilage that was partially preserved, 10 of 21 subjects had complete bilateral CL destruction, indicating end-stage ligament disease despite reasonable cartilage preservation. Two of the 4 subjects with complete cartilage loss had complete bilateral ligament disruption. Along with the ligament disruption, joint subluxation was seen in 14 of 30 patients with OA. When present, subluxation was more commonly seen toward the ulnar side (varus subluxation); 11 of 14 patients (79%) had ulnar subluxation, compared with 3 of 14 (21%) with radial subluxation (Figure 3d).
Tendons. Thickening of the extensor tendons was noted close to their enthesis site in almost all of the OA patients (14 of 16 with early OA, 11 of 14 with chronic OA) (Figure 2). All affected tendons had signs of enhancement. In the absence of severe cartilage loss (n = 21), only 16 of 21 subjects had extensor tendon abnormalities, despite the CLs being universally abnormal.
Bone edema. Four major patterns of bone edema were noted. The most common pattern was focal subchondral edema in 8 of 16 patients with early OA and 5 of 14 with chronic OA. Most of these subchondral regions of bone edema (10 lesions in 9 subjects) were on the proximal side of the joints, directly adjacent to sites where the CLs exert pressure on the proximal phalanx.
The second pattern occurred at the CL attachment (14 lesions in 9 patients) (Figure 5a). Bone edema at the CL entheses was seen in 4 of 16 patients with early OA and 5 of 14 with chronic OA patients. In these 14 regions of bone edema at the CL entheses, all except 1 were at the CL insertion on the distal phalanges of either the PIP or DIP joints. Enthesophytes were present in 8 of 14 sites of CL enthesis bone edema. Only 2 of 14 bone edema lesions at the CL entheses (14%) were related to bone erosions, while 7 of 10 of the subchondral bone edema lesions (70%) were associated with bone erosions (P = 0.01).
Diffuse edema on both sides of the joint, where cartilage was lost (representing a “kissing edema” pattern as described for large joints), was seen in 5 of 16 patients with early OA and 4 of 14 with chronic OA. Lastly, small regions of focal edema present elsewhere and not related to subchondral bone or enthesis were seen in 1 of 16 patients with early OA and 1 of 14 with chronic OA.
Erosions and cysts. Erosions, as defined by bone cortex disruption in 2 planes, were evident in 11 of 16 patients with early OA (22 erosions) and 7 of 14 with chronic OA (10 erosions). Two patterns of erosion were evident. The most common type of erosion occurred adjacent to the CLs, and was evident in 17 subjects (24 erosions) (Figure 5b). Less common were central joint erosions, which were evident at sites of cartilage loss in 6 patients with OA (8 erosions). Bone cysts were also visible in a smaller proportion of patients, being present in 3 of 16 patients with early OA and 4 of 14 with chronic OA.
Osteophytes. The most common site for osteophyte development is at the bone–cartilage interface of the more proximal phalanx in both PIP and DIP joints, predominantly on the dorsal proximal side of the joint. These generally appeared hook-like and pointed proximally (Figure 2c). The sites of osteophytes in chronic disease also corresponded with the enhancing soft tissue that was bulging through the joint at areas of least resistance in early disease (Figures 5c and d). Some osteophytes were associated with adjacent bone enhancement following Gd-DTPA administration; these tended to occur in the early OA group.
Enthesophytes. New bone formation at ligament origins or insertions (enthesophytes) was more common in longer standing OA than in early disease, as evidenced by examples in 3 of 16 of patients with early OA (19%) (5 enthesophytes), versus examples in 6 of 14 patients with chronic OA (43%) (9 enthesophytes), and was invariably evident at ligament insertions but not origins. Furthermore, all of the enthesophytes in the patients with early OA (5 of 5) showed Gd-DTPA enhancement in the CLs, while only 4 of 9 enthesophytes in patients with chronic OA (44%) showed enhancement (P = 0.09). Similarly, bone edema adjacent to or within enthesophytes was more common in the early OA group, with 4 of 5 cases (80%) in the early OA group, versus 4 of 9 cases (44%) in the chronic OA group (P = 0.30).
Synovium and joint capsule. Excess joint fluid visible on the T2-weighted fat-suppressed SE sequence was evident in the joint space in a majority of the OA patients (n = 22 [73%]). Fluid and synovial enhancement was especially common in the dorsal recess of the joint in OA. In the axial sequences the central extensor tendon slip and the adjacent lateral bands appear to have been pushed outwards and apart by soft tissue swelling in the PIP joints, and outward in the DIP joints (12 of 16 patients with early OA, 12 of 14 with chronic OA). The bulging of the soft tissue tended to occur through areas of least resistance at sites of anatomic weakness in the joint capsule, between the extensor tendons and CLs, and the flexor tendons and CLs (Figure 5c).
Asymptomatic joint in an OA-affected hand (latent OA). Although the joints scanned in the group of 10 OA patients were apparently unaffected and completely asymptomatic, with no history of pain or swelling, and were clinically normal, these joints showed abnormalities on MRI (Figure 1c). The CLs were the most affected structure (n = 7 [70%]), although abnormalities were often mild with only thickening of the ligament, particularly at the proximal attachment (associated with intrasubstance increased signal and Gd-DTPA enhancement in all cases) (Figures 3b and 4c). Similarly, the extensor tendon was also often affected to some degree, with thickening at the distal attachment (n = 6 [60%]). Three of the 7 patients with ligament abnormalities also had associated bone edema. Four patients with ligament abnormalities had subtle cartilage loss, 1 of whom also had a bone erosion at the site of cartilage loss. Small osteophytes were observed in 5 patients with ligament abnormalities, located mainly on the dorsal phalanx proximal to the joint. Cartilage damage without ligament change was not seen.
Normal subjects. All of the younger control subjects had completely normal joints including ligaments, cartilage, and bone, with a small region of normal synovial fluid evident at the articular margins. However, 5 of 10 older control subjects had subtle abnormality of the CLs, with thickening, and 3 of the 5 had Gd-DTPA enhancement (Figures 1c and 6). Two subjects also had small regions of bone edema, with 1 showing edema adjacent to the abnormal CL. The tendons appeared uniform, and the insertions did not show changes or heterogeneity in signal intensity (Figure 2a). No cysts, erosions, or soft tissue swelling were seen in any of the normal finger joints. Consistent with reports of occasional osteophytes found on radiographs of older asymptomatic subjects, osteophytes were observed in 2 older normal subjects in this study with preservation of joint cartilage.
In this study we used a novel high-resolution MRI coil that can be utilized in clinically available scanners to investigate the microanatomic basis for hand OA. Hand OA was chosen for study because its microanatomic basis is less well defined compared with that of RA and SpA. Consistent with the known features of OA, every joint structure showed variable abnormalities. Severe destructive and inflammatory changes, as suggested by Gd-DTPA enhancement, were evident in every joint structure in early hand OA. Joint collateral ligament abnormalities were especially common, and subtle ligament changes were the most common changes in normal joints adjacent to OA joints.
This study confirmed a number of interesting observations in relation to hand OA and raised a number of new issues. For example, subchondral bone sclerosis was especially common in chronic OA, which may support the recent assertion that subchondral bone sclerosis represents secondary remodeling and is not primarily responsible for cartilage damage as originally suggested (14). The mechanism of bone edema in OA is unknown but is considered to be related to bone microfracture or microcracks. The resolution of the images in this study was not adequate for showing trabecular microcracks or microfractures, but subchondral edema without obvious trabecular fracture or damage to regions of adjacent cartilage was evident. Also, unlike findings in RA, erosions were more commonly seen on MRI in early OA, supporting the concept of the significant healing and remodeling that is known to occur in OA as compared with RA. Joint inflammation as determined by Gd-DTPA uptake was common not only in the synovium in OA but also in most other joint structures, including ligaments, tendons, extracapsular tissues, and bone marrow, and was comparable in degree and location in both patients with early disease and those with painful chronic joint involvement.
The formation of Heberden's and Bouchard's nodes is a characteristic clinical feature of OA. Soft tissue was observed to bulge through the sites of relative weakness of the joint capsule at the 4 locations bounded by the medial and lateral CLs and the flexor and extensor tendons, especially on the dorsal aspect of the joint in early OA. In late OA, Heberden's and Bouchard's node formation was evident at the same sites, suggesting that ossification of tissue at these sites was related to prior inflammation and CL position as previously hypothesized (15), and further suggesting that the phenotype of hand OA is related to the functional anatomy of the joint. Longitudinal studies are needed to confirm this theory.
Perhaps the most interesting result of this study was the common occurrence of ligament abnormalities. To the best of our knowledge, there are no previous studies implicating ligament diseases in hand OA. It is, however, well recognized that trauma or disruption of the ligament of large joints is associated with the subsequent development of OA in experimental and clinical settings. Knee cruciate ligament disruption is associated with subsequent canine OA (16). In humans, chronic knee joint ligament instability is associated with OA development (17–19). Therefore, large joint OA could be seen as an inevitable consequence of ligament failure (14). In the case of small joints it has been shown that the PIP CLs are the primary restraint to varus and valgus joint angulation (20), and that reconstruction of the ligaments as opposed to other joint structures has the greatest benefit in the amelioration of carpometacarpal joint OA (19). While ligament laxity may be important in knee OA, the relationship between joint laxity in the small joints and associated OA remains controversial (21–23). In this study we also noted that other MRI-determined joint abnormalities, including erosion, bone edema, and new bone formation, bore a close anatomic relationship to the abnormal ligaments. Furthermore, older normal subjects had subtle ligament abnormalities, generally showing a lesser degree of enhancement without frank disruption compared with patients who had clinically evident OA. The question arises, therefore, whether the ligament changes are age related or if they are in some way implicated in the pathogenesis of hand OA. Further longitudinal studies specifically examining ligaments in the early stages of hand OA are planned.
Based on the observation that enthesophytes (new bone formation at ligament attachments) were common in OA, it has been suggested that this supports the importance of bone in the pathogenesis of OA (2). However, hand OA is characterized by enthesopathic changes especially in the soft tissue side of ligament enthesis, and also with bone edema at ligament attachments. Therefore, entheseal new bone in OA could simply reflect a reactive bone-forming process at the enthesis, in a manner reminiscent of diseases such as ankylosing spondylitis and psoriatic arthritis, rather than an intrinsic bone disorder (1).
There were a number of limitations to this study. Although we identified the ligament as a prominent site of abnormality in hand OA, we could not obtain histologic proof of ligament disease, since an in vivo study of ligaments precludes the comparison with equivalent sections obtained for histopathologic study. Identification of image features is possible, however, due to good anatomical detail and analogy to findings described in previous reports involving cadaver specimens and animal models (9, 10, 12, 13, 24). Although we studied the early symptomatic phases of OA, it is clear that it is virtually impossible to study the early pathologic phases of hand OA, since patients with symptoms of a few months' duration often had joint abnormalities comparable with those in patients who had chronic disease. This is well described for other sites of OA presentation, including the hip joint (25). However, we referred to these patients as having early OA, because they were assessed following a first symptomatic episode, and we also assessed uninvolved joints that may develop clinical OA at a later stage. Furthermore, we divided the normal subjects into a younger group and an older group, more closely age-matched with the patients.
Because OA is a disease of aging, it is possible that it represents an aberrant age-related process. It was clear that older normal subjects had ligament abnormalities that younger normal subjects did not have. Although the sample sizes of the groups were comparatively small in this pilot study, and multiple tests were performed, the findings consistently indicated prominent ligament damage, particularly in early and chronic OA. Finally, the demonstration of ligament abnormalities in the small joints could reflect the unique anatomy and biomechanics of these joints, and it remains to be proven whether ligament changes are common in other generalized OA-prone sites, including the spine, hip, and knee.
In conclusion, this study used high-resolution MRI in patients with OA to better define the microanatomic basis for early hand joint involvement. The results of the study reaffirm some longstanding concepts relating to OA, in particular the concept of whole joint organ involvement, even at disease onset. The findings also raise important questions about the role of ligaments in hand OA. This MRI technique should become widely available to rheumatologists to investigate the microanatomic basis for OA and other diseases in the clinical setting. In particular, the findings of this study highlight the need for more research into ligament abnormalities in the early stages of OA.
We would like to thank all who volunteered or referred patients for the study. We would also like to thank all the staff at the Leeds General Infirmary MRI Unit for their services with regard to the study, and Dr. Elizabeth Hensor for advice on statistical analysis.