Participants
The Tasmanian Older Adult Cohort (TASOAC) study is a prospective, population-based study primarily aimed at identifying factors associated with the development and progression of OA and osteoporosis in community-dwelling older adults. Participants aged 50 years and above were selected using a sex-stratified random sampling technique from the electoral roll in Southern Tasmania (population 229,000). A total of 1099 adults (response rate = 57%) consented to participate in the study. Participants were excluded if they had any implants that would prevent them from undergoing a magnetic resonance imaging (MRI) scan or they were living in a nursing home. Participants who consented to participate in the study were invited to attend a clinic at the Menzies Institute for Medical Research, Hobart, Tasmania, between March 2002 and September 2004. They were invited for follow-up clinic assessments at 2.5, 5, and 10 years after the initial clinic assessment. The 10-year follow-up included a sub-study which focused on OA of the hand. Inclusion criteria for a sub-study were that participants were agreeable to have additional radiographic and clinical hand assessment at phase 4. Exclusion criteria for the sub-study included ferromagnetic implants, claustrophobia, and inability to maintain a head first, prone, arm up (superman position) without movement for the duration of the MRI scan. Multiple imaging modalities were performed on a consecutive sample including ultrasound, radiography, MRI, and HRpQCT. Therefore, this study includes 201 participants with hand radiography and HRpQCT measurements. A flow chart of study participants is given in Fig. 1. The study was approved by the Southern Tasmanian Health and Medical Human Research Ethics Committee, and written informed consent was obtained from all participants.
HRpQCT measures at the distal and proximal 2nd DIP, the distal and proximal 2nd PIP, the 1st CMC joint, and the distal radius
Unilateral HRpQCT measurements were performed on the distal and proximal site of 2nd DIP joint (Fig. 2), the distal and proximal site of 2nd PIP joint (Supplementary Figure 1), the 1st CMC joint (Supplementary Figure 2), and the distal radius of the target hand, referred to as the assessed joint. We assessed these hand joints, the 1st CMC joint, the 2nd DIP and PIP joints, as they were in a line which made scanning technically easier and this in plane presentation was used for a parallel study of MRI. The target hand for each participant was the dominant hand unless HRpQCT was contraindicated, in which case the contralateral hand was examined, or the participant was excluded from this study (185 dominant hands and 16 non-dominant hands were scanned and assessed).
HRpQCT (XtremeCT, Scanco Medical AG, Bruttisellen, Switzerland) measurements were performed by immobilizing the target forearm with a flat position in the carbon fiber shell. A dorsal-palmar scout-view X-ray image was obtained to define the scan region of 2nd DIP, 2nd PIP, and 1st CMC [8]. The reference point was defined as the midpoint of joint. Scans were performed in a 9.02mm with 110 slices of region of interest (ROI) at the 2nd DIP, 2nd PIP, and 1st CMC using the manufacturer’s standard patient settings for image acquisition (tube potential of 60 kVp, tube current of 900 μA, 100 ms integration time, isotropic voxel size of 82μm) [8, 9]. Subchondral bone including subchondral bone plate and subchondral trabecular bone has been mostly defined as the bone components lying beneath calcified cartilage, and deeper bone structure [10, 11]. Subchondral regions were evaluated for 20 slices in both distal and proximal sites of the 2nd DIP and 2nd PIP joints, and 110 slices for the 1st CMC using the standard protocol provided by the manufacturer. Due to marked variations in joint shape, the 1st CMC was not separated by the distal and proximal site. The same HRpQCT measurements were performed on the distal radius. The ROI of 9.02 mm (110 CT slices) was at the standardized distance of 9.5 mm from the manually positioned reference line at the end plate of the distal radius.
The semi-automatic contours were created by the device automatically and checked carefully by an operator to segment subchondral plate and trabecular bone. The delineation of the cortical and trabecular compartments was defined automatically using a filter and global fixed threshold to extract the mineralized bone phase using the standard protocol of Scanco software [12, 13]. The reproducibility errors for segmentation and quantification expressed as root mean square coefficients of variation ranged from 0.54 to 3.98% and were <1.5% for volumetric bone mineral density (vBMD) [14]. All scans were graded for motion artifacts using a grading scale from one (no motion) to five (significant blurring and discontinuities). Grading 1–3 could be used. While higher grading of four or higher was rescanned on the day [15].
Bone parameters were derived from the default device using the scanner manufacturer’s software (IPL v5.08b, Scanco Medical AG). Primary bone parameters were measured using the direct model of the device, whereas derived measures were identified with superscript “d” [7, 16]. The following bone parameters were determined for the 1st CMC, the distal and proximal subchondral regions of the 2nd DIP and PIP joints (i.e., the distal and proximal 2nd DIP, the distal and proximal 2nd PIP), and the distal radius: bone areas (total bone area, cortical area, trabecular area), bone density (total, cortical and trabecular vBMD), cortical bone microarchitecture (cortical thickness, and cortical perimeter), and trabecular microarchitecture (trabecular bone volume fraction, trabecular number, trabecular thickness, trabecular separation, and inhomogeneity of trabecular network). Bone parameters and definitions were shown in Supplementary Table 1. The rationale for including all the outcome measures of interest of this study was hypothesis-generating.
Radiographic hand assessment of the 2nd DIP, 2nd PIP, and 1st CMC joints
Radiography is considered the gold standard for the imaging of OA joints. Consequently, a digital posteroanterior radiograph of both hands was acquired. This scan provides a two-dimensional image with high spatial resolution for fine detail of the bones of the hand [17] allowing for the diagnosis of radiographic HOA. For this component of the study, the 2nd DIP joint, 2nd PIP joint, and the 1st CMC joints of the target hand were scored using the OARSI atlas [18] for the presence or absence of osteophytes (0–3) (distal and proximal separately) and joint space narrowing (JSN) (0–3) (the whole joint) by consensus by 2 readers who were a trained rheumatologist and radiographer respectively. The intra-observer agreement of hand OA assessment were published previously. Intra-observer reliability was excellent for both osteophytes and JSN scores (osteophyte score 0.98, JSN 0.94), assessed one week apart (n = 45) [19].
Clinical hand assessment
Bilateral clinical examination of the hands for tenderness, soft tissue swelling, hard tissue enlargement, and deformity was performed and assessed as absent or present by a trained nurse. The assessed joints were the fifteen joints of the hand that include the 1st CMC joint, 1st to 5th metacarpophalangeal joints, 1st to 5th PIP joints, and 2nd to 5th DIP joints. Pain in the joints in the target hand was determined by questioning the participants if they had pain in each individual joint in the preceding seven days. Tenderness was assessed by the examiner exerting pressure onto the joint using their thumb and index finger sufficient to produce whitening of the examiners nail bed [20]. Soft tissue swelling was assessed visually and by palpation by the examiner ascertaining whether the joint appeared swollen. Nodules were assessed by manual examination of each joint and deformity was determined by the appearance of any deviation in the joint from the sagittal plane. Clinical HOA was diagnosed according to the American College of Rheumatology (ACR) classification criteria for hand OA [21] based upon the results of the clinical hand examination. The intra-observer reliability of each of the abnormalities (deformity, nodules, swollen, tenderness) at the joint level was assessed with at least a 1-week interval between the readings, using a kappa statistic in 10 participants with fair to substantial reliability (κ ranging from 0.376 to 0.688) [22].
Anthropometry
Height was measured to the nearest 0.1 cm (with the participant having removed shoes, socks, and headwear) by stadiometer. Weight was measured to the nearest 0.1 kg (with the participant having removed shoes, socks, and bulky clothing) by calibrated electronic scales. Body mass index (BMI, kg/m2) was calculated by dividing weight in kilograms by height in meters squared.
Statistical analyses
T-tests or chi-squared tests were used to assess the differences between continuous and categorical characteristics, as appropriate, of the participants included in this sub-study (n = 201) and the remaining 10-year follow-up sample (n = 367).
Linear-mixed effect models were first run to examine the relationships between osteophyte and JSN scores with HRpQCT measures, including data for all five ROIs in one hand (the proximal and distal 2nd PIP, the proximal and distal 2nd DIP, and the 1st CMC) using fixed effects for age, sex, and BMI, and random intercepts for ROIs. Correlations between observations on the same individual were accounted for using clustered sandwich estimator with robust standard error. In these linear-mixed effect models (both the osteophyte and JSN model), there were significant interactions between joint sites (proximal/distal) and radiographic HOA (osteophyte and JSN scores) on all HRpQCT measures, so all analyses were stratified by proximal (including the proximal 2nd DIP and proximal 2nd PIP) and distal site of joints (including the distal 2nd DIP and distal 2nd PIP), and 1st CMC.
Linear-mixed effect models including fixed effects for age, sex, and BMI, and random intercepts for ROIs were used to examine the relationships between osteophyte and JSN scores with HRpQCT measures at the distal site of joints (i.e., distal 2nd DIP and distal 2nd PIP). Similarly, linear-mixed effect models including fixed effects for age, sex, and BMI, and random intercepts for ROIs were used to examine the relationships between osteophyte and JSN scores with HRpQCT measures at the proximal site of joints (i.e., proximal 2nd DIP and proximal 2nd PIP).
A separate multivariable linear regression model was run for the 1st CMC joint which adjusted for age, sex, and BMI. Lastly, a further linear regression model was run to examine the relationship between osteophyte and JSN scores in the hand (1st CMC, 2nd DIP, 2nd PIP) with HRpQCT measures at the distal radius adjusted for age, sex and BMI.
Additional analyses for distal DIP, proximal DIP, distal PIP, and proximal PIP were run using separate multivariable linear regression models and adjustment for age, sex, and BMI.
All HRpQCT measures were standardized. All statistical analyses were performed using StataSE 16.1 for Windows (StataCorp, College Station, TX, USA). A two-side p-value < 0.05 was considered statistically significant. The family-wise error rate was used to control the overall false-positive rate for multiple testing [23]. In this study, the adjusted alpha level was 0.0167. Using the adjusted alpha level, most results remained statistically significant. Thus, we presented all data.