Abstract
Objective To compare in images, obtained by high-resolution peripheral quantitative computed tomography (HR-pQCT) and conventional radiography (CR) of the second and third metacarpophalangeal (MCP) joints, the minimal erosive cortical break needed to differentiate between pathological and physiological cortical breaks.
Methods In this single-center cross-sectional study, patients with established rheumatoid arthritis (disease duration ≥ 5 yrs) had their second and third MCP joints of the dominant hand investigated by HR-pQCT and CR. Empirical estimation was used to find the optimal cut-off value for the number of erosions and total erosive volume, which were detectable between patients with and without erosions in the second and third MCP joints according to CR.
Results The total erosive volume in the second and third MCP joints by HR-pQCT for CR-detected erosive disease was estimated to be 56.4 mm3 (95% CI 3.5-109.3). The sensitivity and specificity at this cutpoint were 78% and 83%, respectively, with an area under the receiver-operating characteristic curve (AUC) of 0.81. The optimal cut-off value for the number of erosions by HR-pQCT was 8.5 (95% CI 5.9-11.1) for CR-detected erosive disease in the second and third MCP joints. The sensitivity and specificity at this cutpoint were 74% and 88%, respectively, with an AUC of 0.81.
Conclusion Erosions by HR-pQCT were larger in patients with erosive damage in the second and third MCP joints by CR. We found that CR had poor sensitivity for detecting erosive disease when the erosive volume was < 56.4 mm3 or the number of erosions was < 8.5.
Erosive damage is a cardinal sign of rheumatoid arthritis (RA) and is correlated with disease activity.1 Conventional radiography (CR) of the hands, wrist, and feet is currently the gold standard for evaluating the radiographic disease severity of RA.2,3 The radiographic examinations are most commonly assessed by the Sharp/van der Heijde (SHS) score in observational and clinical trials,2,3 as this method is the most reliable and sensitive for tracking the progression of radiographic damage.4 However, as the treatment of patients with RA has improved, CR is no longer sensitive enough to detect the progression of discrete erosive damage.
Imaging by high-resolution peripheral quantitative computed tomography (HR-pQCT) has been proposed as a modality for assessing erosive cortical breaks in patients with RA because of its high resolution and 3-D imaging.5 A previous study found that HR-pQCT imaging of the second and third metacarpophalangeal (MCP) joints from 1 hand and CR of hands, wrist, and feet had comparable diagnostic accuracy for diagnosing patients with RA having erosive disease.6
The objective of the present study was to evaluate the volume and number of erosive cortical breaks on HR-pQCT of the second and third MCP joints needed to be optimally detectable by CR.
METHODS
Study design. We used data from the RACTX cohort for this cross-sectional single-center study, which consisted of 354 patients with established RA (disease duration ≥ 5 yrs).6 HR-pQCT and CR investigated the second and third MCP joints from 1 hand. The study was designed following the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.7
Patient and public involvement statement. We investigated the presence of erosive disease in patients with RA by 2 different imaging modalities in the current study. Therefore, we did not involve patients regarding the study design, outcome, or recruitment.
Participants. Patients with RA were recruited from the outpatient clinic at the Department of Rheumatology, Aarhus University Hospital. The inclusion period was between March 2018 and October 2020. The patients were examined by HR-pQCT and CR, which were conducted within 3 months of each other. A full medical history was obtained, and a clinical examination was performed for all individuals. Specifically, demographic and clinical data were acquired, including age, sex, disease duration, number of tender and swollen joints, C-reactive protein (CRP), Health Assessment Questionnaire (HAQ), as well as anticitrullinated protein antibody (ACPA), and IgM rheumatoid factor (RF).
Eligibility criteria. Inclusion criteria were diagnosed RA according to the American College of Rheumatology/European Alliance of Associations for Rheumatology 2010 classification criteria,8 disease duration ≥ 5 years, the ability to give informed consent, and age ≥ 18 years. Exclusion criteria were fracture, luxation or prosthesis of the MCP joints in both hands, evidence of active malignant disease, hypocalcemia, impaired renal function (estimated glomerular filtration rate < 35 mL/min), untreated hypothyroidism or hyperthyroidism, or pregnancy.
CR acquisition and scoring. All patients had their hands, wrist, and feet examined radiographically using the standard dorsopalmar projection. The image was generated at a focus distance of 100 to 115 cm, 50 to 55 kV, and 2 to 12 mAs. The radiographs were scored by a single trained reader (RKJ) using the SHS system.3
HR-pQCT acquisition and analysis. An image acquisition protocol endorsed by the Study Group for Xtreme-Computed Tomography in Rheumatoid Arthritis (SPECTRA) was used.9 The second and third MCP joints were imaged using the first generation XtremeCT scanner (Scanco Medical). A 2.7 cm long volume of interest was scanned with a spatial resolution of 82 μm3, a radiograph tube voltage of 59.4 kVp, a current of 900 μA, and an integration time of 100 ms. The scan was performed within a region of 80 slices (6.56 mm) distal and 250 slices (20.5 mm) proximal to the distal end of the third metacarpal head. The dominant hand was scanned except in cases with prior fracture, prosthesis, or luxation in the MCP joints. Digital Imaging and Communications in Medicine images were exported from the HR-pQCT scanner and evaluated by OsiriX medical imaging software (Version 9.0.1, Pixmeo) on a 27-inch cinema screen iMac. Each image was anonymized before analysis. The quality of each scan was evaluated as previously described.10,11 Each joint was divided into the proximal end (metacarpal head) and the distal end (proximal phalanx). Each bone was divided into 4 quadrants (dorsal, radial, palmar, and ulnar).12 Each quadrant was then evaluated for the presence of erosive cortical breaks by 2 experienced readers (RKJ, Jette Barlach [JB]).
Each erosion was measured according to the maximal width, depth, and length. The volume of each erosion was measured by manual segmentation using OsiriX medical imaging software.6 The erosion volume was also calculated as a half-ellipsoid from the measured width, depth, and length.13
The erosions were defined according to the SPECTRA definition.14 In brief, there was a definite cortical break in 2 consecutive slices, in at least 2 perpendicular planes, and with underlying loss of trabecular structure. Last, the cortical break had to be nonlinear to differentiate between erosive cortical breaks and physiological cortical breaks (ie, vascular channels).15 The erosions were measured according to the maximal width, depth, length, and volume.
Ethical approval. The Ethics Committee of Medical Research in Central Denmark Region (1-10-72-437-17) and the Danish Data Protection Agency (1-16-02-33-18) approved the study. The study was registered at ClinicalTrials.gov (NCT03429426). All patients gave informed written consent before inclusion, and the study was performed in agreement with the Declaration of Helsinki.
Sample size. Previously, only a few smaller studies have investigated erosions by HR-pQCT in patients with erosive and nonerosive MCP joints by CR. However, these studies did not assess erosive damage by the SHS score and did not assess the volume of erosions manually.16,17 Therefore, no data are available on the relevant populations that can be used to perform sample size calculations. Still, our study includes the largest patient cohort presently for any study investigating erosive cortical breaks by HR-pQCT in inflammatory rheumatic disease, and the included number of patients is assessed sufficiently to fulfill our aim.5
Statistical methods. Data were analyzed using Stata 13 (StataCorp). The normality of the data distribution was investigated with Q-Q plots and histograms. As the data were nonnormally distributed, data were presented as median (IQR), and statistical significance was tested using the Mann-Whitney U test. The intrareader reliability was investigated by intraclass correlation coefficient (ICC) for 10% of the CR and HR-pQCT scans by a single trained reader (RKJ). The interreader reliability was investigated for 10% of the CR images by 2 trained readers (RKJ, J. Therkildsen). The interreader reliability was investigated for 50% of the HR-pQCT images by 2 trained readers (RKJ, JB). Correlations were calculated by Spearman rank correlation coefficient (ρ). The measured volume in OsiriX and calculated half-ellipsoid volume were compared for each cortical interruption by Bland-Altman plots.18 We assessed optimal cut-off values for the number and total volume of erosions assessed by HR-pQCT to detect erosive disease in the second and third MCP joints by CR using empirical estimations for optimal outcome prediction.19 Sensitivity, specificity, and area under the receiver-operating characteristic curve (AUC) are reported.
RESULTS
A flowchart of patient inclusion is shown in Figure 1. The patient demographics and clinical characteristics are shown in Table 1. Patients were divided into 2 groups with regard to erosive damage in the second and third MCP joints by CR. Two hundred forty-two out of the 353 (68.6%) patients did not present erosions in their second and third MCP joints by CR. The remaining 111 (31.4%) patients had erosive cortical breaks in their second and third MCP joints (Figure 2).
Flowchart showing the identification, inclusion, and exclusion of research subjects. CR: conventional radiography; eGFR: estimated glomerular filtration rate; HR-pQCT: high-resolution peripheral quantitative computed tomography; MCP: metacarpophalangeal; PsA: psoriatic arthritis; RA: rheumatoid arthritis.
Clinical characteristics of the participants.
Images of erosions in the left second MCP joint in a 75-yr-old female patient. CR of the second MCP joint, erosions not visible (A), consecutive slices from HR-pQCT imaging in the axial plane (B1, B2), coronal plane (C1, C2), sagittal plane (D1, D2), and the 3-D segmentation (E). 3-D segmentation was performed using thresholding (3-D Slicer, http://www.slicer.org). CR: conventional radiography; HR-pQCT: high-resolution peripheral quantitative computed tomography; MCP: metacarpophalangeal.
Patients with erosive disease in the second and third MCP joints by CR had larger erosions than patients without erosive disease in the second and third MCP joints by CR; this was seen for all parameters, including maximal width, depth, length, and volume.
Patients with erosive disease in the second and third MCP joint by CR were older, had longer disease duration, had higher SHS and HAQ scores, and a larger proportion of the patients were RF positive. There was not a significant difference in the proportion of ACPA-positive patients in the 2 groups. Likewise, no significant difference regarding demographics and clinical characteristics was observed (Table 1).
Optimal cut-off of HR-pQCT for predicting patients having erosive cortical breaks in the MCP joints detectable by CR. The empirical estimation for the optimal cut-off value for the number of erosions in the second and third MCP joints by HR-pQCT was 8.5 (95% CI 5.9-11.1) erosions for detecting erosive disease in the second and third MCP joints by CR. The sensitivity and specificity at the cutpoint were 74% and 88%, respectively, with an AUC of 0.81 (Table 2; Supplementary Figure S1, available with the online version of this article). Empirical estimation for the optimal cut-off value for the total erosive volume in the second and third MCP joints by HR-pQCT was 56.4 mm3 (95% CI 3.5-109.3) for predicting erosive disease in the second and third MCP joints by HR-pQCT. The sensitivity and specificity at cutpoint were 78% and 83%, respectively, with an AUC of 0.81 (Table 2; Supplementary Figure S2).
The optimal cut-off for the number of erosions, and total erosive volume by HR-pQCT for predicting erosive disease by CR in the second and third MCP joints.
Anatomical distribution of erosions. A total of 2460 erosions were evaluated; 40% of erosions were located in the second metacarpal head. The third metacarpal heads and the second proximal phalanx contained 23% and 24% of the erosions, respectively. The last 12% were in the third proximal phalanx. The radial quadrant showed a strong predilection for erosions; this was seen for both metacarpal heads. The most affected site for erosions in the proximal phalanges was the dorsal quadrant, closely followed by the radial quadrant (Figure 3). The erosions were largest in the second metacarpal head, followed by the third metacarpal head, the second proximal phalanx and lastly, the third proximal phalanx. As for the volume of erosions, the volume was largest in the radial quadrant for the metacarpal heads. In contrast, the volume was largest in the dorsal quadrant for the proximal phalanges (Figure 3).
The total number (A, B) and erosive volume (C, D) of erosions according to the bone in the second and third MH and PB assessed by HR-pQCT in the patients defined as either erosive (A, C) or nonerosive (B, D) in the second and third MCP joints by CR. CR: conventional radiography; HR-pQCT: high-resolution peripheral quantitative computed tomography; MCP: metacarpophalangeal; MH: metacarpal head; PB: phalangeal base.
In the second proximal phalanx, joint space narrowing (JSN) by CR was correlated with the number of erosions (ρ = 0.50, P = 0.01) and total volume of erosions (ρ = 0.47, P = 0.01) by HR-pQCT for the dorsal quadrant only, as illustrated in Figure 4. In the second metacarpal head, JSN by CR was correlated with the number of erosions (ρ = 0.37, P = 0.02) by HR-pQCT for the dorsal quadrant only.
Patients with severe erosive damage. Arrows denote cortical breaks. Note the severe deformity, especially to the dorsal quadrant of the proximal phalanx. CR of the second MCP joint from an 80-yr-old male patient (A1), A2 and A3 are consecutive slices from HR-pQCT imaging, and A4 is the 3-D segmentation. CR of the second MCP joint from a 73-yr-old female patient (B1), B2 and B3 are consecutive slices from HR-pQCT imaging, and B4 is the 3-D segmentation. CR of the second MCP joint from a 72-yr-old male patient (C1), C2 and C3 are consecutive slices from HR-pQCT imaging, and C4 is the 3-D segmentation. 3-D segmentation performed using thresholding (3-D Slicer, http://www.slicer.org). CR: conventional radiography; HR-pQCT: high-resolution peripheral quantitative computed tomography; MCP: metacarpophalangeal.
Volume measures. The volume of erosions measured manually or calculated as a half-ellipsoid analyzed by Bland-Altman plots are shown in Supplementary Figure S3 (available with the online version of this article). The manually measured volume was 5.5 mm3 higher than the half-ellipsoid volume. Still, the interval of 2 SD was wide (−97.4 to 86.3 mm3). When erosions below 20 mm3 were analyzed, the manually measured volume was 0.8 mm3 higher than the half-ellipsoid volume. The interval of 2 SD was (−8.3 to 6.8 mm3). However, it was still evident that the limits of agreement fell with increasing volume.
Reliability. The intrareader reliability (ICC) was 0.998 (95% CI 0.996-0.999) for the SHS score, 0.997 (95% CI 0.995-0.999) for the erosions score, and 0.990 (95% CI 0.979-0.995) for JSN. The interreader reliability was 0.971 (95% CI 0.945-0.985) for the SHS score, 0.966 (95% CI 0.937-0.982) for erosions score, and 0.939 (95% CI 0.887-0.967) for JSN. The intrareader reliability for the number of erosions was 0.963 (95% CI 0.914-0.984), width 0.818 (95% CI 0.622-0.918), depth 0.876 (95% CI 0.728-0.945), length 0.814 (95% CI 0.594-0.918), volume 0.918 (95% CI 0.817-0.964), and ellipsoid volume 0.942 (95% CI 0.863-0.975). The interreader reliability for the number of erosions was 0.826 (95% CI 0.477-0.938), width 0.750 (95% CI 0.486-0.859), depth 0.725 (95% CI 0.108-0.913), length 0.717 (95% CI 0.593-0.799), volume 0.730 (95% CI 0.615-0.808), and ellipsoid volume 0.595 (95% CI 0.482-0.686; Supplementary Figure S3, available with the online version of this article).
DISCUSSION
This cross-sectional study is the first, that we know of, to evaluate the characteristics of erosions not detected by CR of the second and third MCP joints but is detected by HR-pQCT imaging. Direct comparison of erosion detection by HR-pQCT imaging and CR is scarce in the scientific literature.20-22 In a previous study, we investigated the diagnostic value of HR-pQCT imaging of the second and third MCP joints from 1 hand compared to CR of hands, wrist, and feet. Although HR-pQCT imaging has a smaller field of view, we found HR-pQCT had equal diagnostic accuracy.6 In the present study, HR-pQCT imaging detected that if patients presented with fewer than approximately 8 erosions in the second and third MCP joints, erosions were poorly detected by CR. Further, we found that CR did not reliably detect patients with erosive damage beneath approximately 50 mm3.
The majority of erosions have previously been shown to be located in the metacarpal heads.22-25 It is widely accepted that erosions have a predilection for the radial and ulnar quadrants.22,23,25 The current study found that the erosions were most common in the radial quadrant. Still, the quadrant with the second-most erosion was the dorsal quadrant, especially in the proximal phalanges. The reason for this discrepancy could be related to our cohort having longer disease duration and especially older age. Loss of joint space is related to disease duration and, to a larger extent, age.26 The loss of joint space results in the proximal phalanx migrating palmar and proximal toward the metacarpal head. This process erodes more of the bone, primarily at the dorsal quadrant phalangeal base and palmar quadrant of the second metacarpal head. As in the present study, 1 other study found that erosions were more common in the dorsal than the ulnar quadrant. These patients were not as old as in our cohort but had equivalent disease duration.24
Volume measures. Several studies have estimated the volume of erosions from the width, depth, and length using a half-ellipsoid formula.13,27-32 In general, we found that the half-ellipsoid formula, on average, underestimated the volume; this underestimation increased as the volume of erosions grew. The calculated volume based upon a half-ellipsoid formula has not previously been compared with manually measured volume. Yet, the half-ellipsoid formula has been compared with 2 semiautomated algorithms.27,30 The 3 semiautomated algorithms were based on the same principles. Both found that the half-ellipsoid formula tended to underestimate the volume of erosions.27,30 As seen in the current study, Figueiredo et al found that the half-ellipsoid formula performed progressively worse with the increasing size of erosions.30 Because of these results, we recommend that the volume of the cortical breaks should not be calculated by the half-ellipsoid formula.
In general, we observed moderate to high reproducibility with regard to inter- and intrareader reliability for the number and size measures of cortical breaks. The interreader reliability of manually measured volume was higher than the calculated half-ellipsoid volume. Together with a previous study showing the reliability after repositioning is higher for the manually measured volume compared to measures of width, depth, and length,24 our results further illustrate the limitations of this method of estimating erosive volume.
This study has several strengths. First, the study includes the RACTX cohort, which presently is the largest cohort included to investigate erosions by HR-pQCT in inflammatory rheumatic disease.5 Second, none of the patients’ HR-pQCT scans were excluded due to severe deformity, as is seen in many HR-pQCT studies.10,17,21,29,33-36 Third, inter- and intrareader reliability were investigated for HR-pQCT imaging.
In conclusion, erosions by HR-pQCT were larger in patients with erosive damage in the second and third MCP joints by CR. We found that CR had poor sensitivity for erosive disease when the volume was less than approximately 50 mm3. Further, we observed that the patients should have approximately 8 erosions in the second and third MCP joints for optimal detection on CR. Our results demonstrate the potential risk of CR misclassifying patients with RA as having nonerosive disease, despite having considerable erosive cortical breaks to their joints.
ACKNOWLEDGMENT
The authors are grateful for the valuable work in analyzing HR-pQCT scans by Jette Barlach and for the excellent assistance in recruiting and scheduling the patients by Mia Marie Remmer, Lone Thomasen, and Else Sloth Rousing.
Footnotes
The study was financially supported by Aarhus University, The Danish Rheumatism Association, Novo Nordic Foundation, Becket Foundation, and A.P. Møller Foundation. The funding sources did not have any role in the collection, analysis, and interpretation of data. The financial contributors did not influence the study design, collection, analysis, and interpretation of data, the writing of the manuscript, or the decision to submit the manuscript for publication.
EMH reports personal fees from MSD, Pfizer, UCB, and Sobi; and grants from Roche and Novartis, outside the submitted work. BL reports personal fees from Eli Lilly, Amgen, UCB, Gilead, and Gideon-Richter; and grants from Novo Nordisk and Amgen, outside the submitted work. The remaining authors declare no conflicts of interest relevant to this work.
- Accepted for publication October 20, 2022.
- Copyright © 2023 by the Journal of Rheumatology
