Which are the most frequently involved peripheral joints in calcium pyrophosphate crystal deposition at imaging? A systematic literature review and meta-analysis by the OMERACT ultrasound – CPPD subgroup

Objectives To identify the prevalence of calcium pyrophosphate crystal deposition (CPPD) using ultrasound and conventional radiology at peripheral joints in patients with suspected or definite CPPD. Methods A systematic literature search was performed in PubMed and Embase using pre-defined search strategies from inception to April 2021 to identify studies that evaluated conventional radiology and ultrasound in detecting CPPD at peripheral joints, including definite or suspected CPPD [Research question 1 (RQ1) and Research Question 2 (RQ2), respectively]. For the meta-analysis, the first, second, and third sub-analysis included studies with the knee, and knee or wrist as the index joint for CPPD (without restrictions on the reference standard) and synovial fluid analysis or histology as a reference standard (without restrictions on the index joint), respectively. Results One-thousand eight hundred and twenty-seven manuscripts were identified, of which 94 articles were finally included. Twenty-two and seventy-two papers were included in RQ1 and RQ2, respectively. The knee had the highest prevalence for RQ1 and RQ2 by both conventional radiology and ultrasound, followed by the wrist with the highest prevalence for RQ1. The hand had the lowest CPPD prevalence. The third sub-analysis showed a higher CPPD prevalence on ultrasound than conventional radiology at the knee (only data available). Conclusion Among all peripheral joints, the knees and wrists could be regarded as the target joints for CPPD detection by imaging. Furthermore, ultrasound seems to detect a higher number of calcium pyrophosphate deposits than conventional radiology, even when using a more restrictive reference standard.

Objectives: To identify the prevalence of calcium pyrophosphate crystal deposition (CPPD) using ultrasound and conventional radiology at peripheral joints in patients with suspected or definite CPPD.
Methods: A systematic literature search was performed in PubMed and Embase using pre-defined search strategies from inception to April 2021 to identify studies that evaluated conventional radiology and ultrasound in detecting CPPD at peripheral joints, including definite or suspected CPPD [Research question 1 (RQ1) and Research Question 2 (RQ2), respectively]. For the meta-analysis, the first, second, and third sub-analysis included studies with the knee, and knee or wrist as the index joint for CPPD (without restrictions on the reference standard) and synovial fluid analysis or histology as a reference standard (without restrictions on the index joint), respectively.
Results: One-thousand eight hundred and twenty-seven manuscripts were identified, of which 94 articles were finally included. Twenty-two and seventytwo papers were included in RQ1 and RQ2, respectively. The knee had the highest prevalence for RQ1 and RQ2 by both conventional radiology and ultrasound, followed by the wrist with the highest prevalence for RQ1. The hand had the lowest CPPD prevalence. The third sub-analysis showed a higher CPPD prevalence on ultrasound than conventional radiology at the knee (only data available).

Introduction
Calcium pyrophosphate deposition (CPPD) is a chronic arthropathy caused by the presence of calcium pyrophosphate (CPP) deposits in articular and periarticular tissues (1). Although the exact incidence and prevalence of CPPD are still unknown, it is considered one of the most common chronic arthropathies (2), characterized by a prevalence that increases with age (3) and can reach up to 13% in the elderly, depending on the assessed joints and the tool used (4). In fact, one of the main issues for epidemiological studies on CPPD is related to the challenges regarding diagnosis.
For a long time, CPPD diagnosis was based on McCarty Criteria, which required both the identification of CPP crystals in synovial fluid analysis (SFA) and the presence of typical calcifications in conventional radiography (CR) for a "definite" diagnosis, while a "probable" diagnosis was defined by SFA or CR positive findings (5). In 2011, a panel of experts from the European League against Rheumatism (EULAR) changed this status. Experts stated that the presence of CPP crystals in the SFA was sufficient for a definite diagnosis. Furthermore, ultrasonography (US) has been endorsed for the first time as a promising tool for CPPD diagnosis (6).
Since then, growing interest in the use of imaging in CPPD has led to an improved definition of the framework for CPPD diagnosis. In particular, US application in CPPD management has been highly improved since its development by the CPPD subgroup of the OMERACT US Working group of a new set of US definitions for CPPD identification (7,8), which demonstrated the reliability and accuracy of CPPD diagnosis (8)(9)(10). A recent systematic literature review (SLR) evaluated the diagnostic performance of CR and US in CPPD diagnosis, showing that both obtained good results with better sensitivity for US and slightly greater specificity for CR (11). Furthermore, an international working group composed of rheumatologists and musculoskeletal radiologist experts in microcrystalline arthritis has recently developed definitions for CPPD identification by CR, which were also assessed for reliability and accuracy, confirming the high specificity of CR for CPPD identification (12, 13).
Accounting for all, the use of imaging is gaining a leading role in CPPD diagnosis and potentially for follow-up in daily practice. However, given the wide range of CPPD joint involvement, it is particularly important to adopt a time-saving approach for US examination by assessing only the most frequently affected peripheral joints, thus increasing the effectiveness and feasibility. The identification of a minimum set of joints could also promote the application of a scoring system, which could be very useful in monitoring the evolution of CPPD.
Thus, the objective of this study was to perform a SLR to estimate the prevalence of CPPD, identified using CR, US or both at the peripheral joints of patients with a suspected or definite CPPD diagnosis, and to establish the most relevant joints for CPPD diagnosis and monitoring.

Methods
The Preferred Reporting Items for Systematic Reviews and Metaanalyzes (PRISMA 2020) guidelines for reporting systematic reviews and meta-analyzes were followed for this review (14).
A protocol defining all phases of this SLR (research questions, search strategy, and inclusion/exclusion criteria for the articles and methods for the analysis) was developed before the beginning of the study and was registered on the PROSPERO platform (Registration Number: CRD42020218155).

Structured search strategy
Two research questions were developed; the first aimed to assess the CPPD prevalence in peripheral joints based on imaging of patients with a definite, crystal proven, CPPD diagnosis [Research Question 1 (RQ1)], and the second one aimed to assess the prevalence of CPP deposits, based on imaging, in patients with suspicion of CPPD diagnosis according to clinical picture [Research Question 2 (RQ2)].
After defining the research questions, the patient, intervention, comparator, outcome (PICO) framework was used to develop the search strategy (15).
PubMed and Embase databases were searched from inception until April 2021. An additional hand search of articles' references was performed to include as many eligible articles as possible. The search strategy was based on both MeSH terms and free text and is illustrated in the Supplementary material S1 (SP1).

Study selection and data extraction
The search included all the studies that evaluated the use of CR and/or US for detecting calcifications at the level of at least one peripheral joint [hand, wrist, elbow, shoulder, acromioclavicular (AC), hip, knee, ankle, foot] in adult patients with suspected or definite CPPD, without any restrictions on the reference test used for diagnosis.
The following study types were eligible for inclusion: crosssectional case-control, cross-sectional cohort, longitudinal casecontrol, longitudinal cohort, retrospective cohort, and retrospective case-control. Case reports, case series, congress abstracts, and studies written in languages other than English were excluded. The titles and abstracts of the retrieved references were screened by six reviewers (AA, EC, EF, GF, FP, and SS) according to pre-defined inclusion and exclusion criteria based on the PICOs. The reviewers worked in pairs to assess the abstracts, and discordant assessments were resolved by consensus.
Relevant full-text articles were evaluated by the same reviewers, and data were extracted using a standardized extraction form. Discordant assessments between the authors were resolved by consensus. Data were extracted using a standardized form, including author, publication year, study type, index test, reference test, inclusion criteria, and number of patients (cases and controls). The data on the frequency of calcifications are summarized in ad hoc tables.
For each article, data on the prevalence of calcifications were collected separately for every peripheral joint, according to the imaging technique applied. If available, data on the involvement of single joint structures (fibrocartilage, hyaline cartilage, tendons) were also retrieved. The frequency of involvement was assessed separately for every joint and according to the imaging techniques used. For each joint and structure, data of monolateral or bilateral involvement of calcifications were collected, divided according to the imaging tool. In case of missing laterality data, they were categorized as unknown.

Assessment of the risk of bias
The risk of bias of the selected studies was assessed using ad hoc instruments applied according to the type of article evaluated. For the diagnostic study, we used the modified version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool (16), while and the Newcastle Ottawa Scale (NOS) was used for the assessment of case-control and cohort studies (17). Data extraction and quality assessment were performed by a single reviewer (EC) and checked by a second reviewer (SS). Any disagreements were resolved by consensus.

Data analysis
Descriptive analyzes and meta-analyzes were performed on the included studies. The descriptive analysis aimed to capture the global prevalence of CPPD in different joints. Except for a relevant degree of variability in the included studies regarding the index joints, reference standards, and imaging techniques, specific meta-analyzes were scheduled to collect as much data as possible from homogenous studies. Thus, the following analyzes were performed: -Descriptive analysis including all studies: Evaluation of CPP deposit prevalence for each joint assessed. All analyzes were divided according to the research question and the imaging method used. If available in the text, the frequency of bilateral involvement at each anatomical site was also provided. -Descriptive analysis including all studies: evaluation of CPP deposit prevalence at the level of the anatomical structures of a single joint. All analyzes were divided according to the research question and the imaging method used. If available, the frequency of bilateral involvement was also provided. -Sub-analysis 1 (SB_1) included only studies that used the knee as the index joint for CPPD diagnosis, independent of the reference standard used. All analyzes were divided according to the research question and the imaging method used. -Sub-analysis 2 (SB_2) included studies that used the knee or wrist as the index joint for CPPD diagnosis, independent of the reference standard used. All analyzes were divided according to the research question and the imaging method used. -Sub-analysis 3 (SB_3) included only studies that used SFA alone (not the McCarty criteria) or histology as a reference standard for the diagnosis of CPPD independently from the index joint and imaging method used. All analyzes were divided according to the research question and the imaging method used.
SB_1 and SB_2 aimed to assess the impact of the index joint, and SB_3 aimed to assess the impact of the reference test on CPP deposit prevalence.

Statistical analysis
The descriptive analyzes were provided as the ratio between the imaging-positive cases and all the cases evaluated (either for the joint or the single articular structure). The results were also presented as percentages. About the bilateral assessment, the results were obtained evaluating all the imaging cases bilaterally positive and the cases evaluated bilaterally (shown in the text as ratios and percentages).
For the sub-analysis, 162 meta-analyzes were performed, one referring to each research question, sub-analysis, and joint analysis. Only meta-analyzes that included at least three studies were considered and presented in this paper. Information on the proportion of participants with CPP deposits in different joints was extrapolated from each study. Pooled estimates [with related 95% confidence intervals (CI)] were calculated using both fixed-and random-effects models. Heterogeneity was calculated using the I 2 index and was high in all analyzes. For this reason, only pooled estimates from random-effects models were reported in the results section. The results were graphically presented using forest plots. All analyzes were performed using the R statistical software (Foundation for Statistical Computing, Vienna, Austria).

Description of the studies
The search strategy identified 1827 records, 1822 from the databases, and five manually searched records (329 duplicates). Of the remaining 1,498 records, 954 were excluded based on their titles and abstracts, and 544 articles entered the full-text evaluation. Considering that 49 full texts were not retrievable (all articles were published before the 1970s), the detailed review included 494 articles. A total of 400 studies were excluded after reviewing the full text, most of which were rejected due to the study type, mainly case reports and case series, followed by outcome. Finally, 94 studies were included in the analysis.
All phases of the selection process are summarized in the Prisma Flow Chart (Figure 1).
All the descriptive results regarding joint involvement are summarized in Table 2.  Frontiers in Medicine 10 frontiersin.org
For RQ2, the values were higher for the TFC, followed by the menisci and hyaline cartilage, but only on US (30,24, and 15%, respectively), while the results were lower at CR (5, 9, and 6%, respectively). Regarding laterality, higher bilateral involvement was recorded for the menisci at CR (28%).
All results regarding CPPD prevalence of joint structures and laterality are shown in SP3.

Meta-analysis
In the SB_1, 73 articles (18, 19, 25-27, 29, 30,   The results of the meta-analysis are summarized in Table 3. Forest Plots of the knees are shown in Figures 2-5. Forest plots of the other joints are shown in SP4.

Assessment of the risk of bias
Most studies (88%, 83/94) showed a low risk of bias in many items, and the overall risk of bias was acceptable. Only a few studies had a high risk of bias (27, 53-55, 58, 65, 71, 90, 100, 104, 105).
Regarding the cohort study, the less fulfilled item was the comparability of the cohorts, which was related to the lack of matching of exposed and non-exposed patients or adjustment for confounders. Usually, the length of follow-up is not evaluable because most of the cohort studies included were mainly crosssectional studies. For case-control studies, the main limitations were related to the representativeness of the cases and controls due to the lack of defined criteria for population selection. Finally, considering the diagnostic studies, the main source of bias was related to the reference standards used. In fact, in three studies, CR was applied as both index and reference standard (27,53,94), while in nine articles, the reference test was the McCarty criteria (20-24, 26, 29, 39, 60). All results of the NOS scale and QUADAS-2 are summarized in SP5.

Discussion
Currently, assessing the prevalence of CPPD remains challenging, mainly because of the heterogeneity of its clinical manifestations (6) and the lack of a non-invasive and accurate diagnostic technique. Furthermore, the natural history of CPPD is still unclear, and the patterns of involvement of the peripheral joints in terms of extent and chronological order have not been defined. In fact, despite being the most evaluated knee joint in the literature, previous studies have shown that radiographic chondrocalcinosis is common in wrists and hips, even in the absence of knee involvement (20,40).
These aspects make imaging a potential cornerstone for CPPD diagnosis and monitoring. In this scenario, US presents several advantages as a noninvasive examination that can be applied to many joints in a short time. Moreover, US has been validated by the OMERACT validation process for diagnosis (10,111). To further improve the application of imaging in CPPD, identifying the joints most frequently affected would improve feasibility and accuracy.
The OMERACT Ultrasound Working group in CPPD performed this SLR to collect the available data on the prevalence of CPPD in peripheral joints, assessed both by US and/or CR, to identify the most relevant joints to scan for CPPD diagnosis and monitoring. In fact, this SLR is the first multi-step approach that will lead to the creation of an US scoring system for CPPD.
Unfortunately, among the included studies, several sources of heterogeneity emerged, as the articles varied in terms of the type and number of joints evaluated, reference standard used, index joint, and CPPD clinical features. These differences made the articles less comparable and introduced biases in the descriptive analysis. Assessing the studies included in RQ1 and RQ2, some differences may be appreciated: for RQ1 (definite diagnosis), imaging was mainly applied on a larger number of joints or on sites different than the knee (only three of 22 articles evaluated the knee alone), using the knee as index joint for the diagnosis, while in the RQ2 (suspected CPPD) the knee was the only joint assessed for diagnosis in almost 50% of the articles, reducing the number of other joints available for analysis. Specific meta-analyzes were performed to address these issues. For each research question, studies were divided according to the index joint for CPPD (knee or wrist) and the reference standard used for diagnosis (selecting only studies that included SFA or histology). This selection led to the identification of the most homogenous study groups, comparable in meta-analyzes that assessed the prevalence of CPP deposits but, on the other hand, reduced the number of patients included in the analysis.
Considering the descriptive analysis, the knee and wrist resulted in the joints being mostly involved in CPPD at both CR and US, independent of the research question. The CPPD prevalence was higher with US at both sites for RQ1, while in RQ2 a higher CPPD prevalence in the knee was detected by US than by CR. In contrast, CR revealed more cases in the wrist than US. Metaanalyzes supported these findings. In fact, according to metaanalyzes, the knee is the joint characterized by the highest CPPD prevalence in RQ2, with values constantly equal to 0.98 in SB_1 and SB_2 (no differences according to the imaging technique used), whereas the prevalence decreased in RQ1 with a higher value in US than in CR [0.93 (0.68-0.99) and 0.79 (0.44-0.95), respectively]. The higher prevalence of CPPD among suspected patients was a surprising result, but is probably explained by the predominant assessment of the knee in RQ2, and by the simultaneous use of CR as an index and reference test in most of the studies included.
The higher CPPD prevalence at the level of the knee when US was applied was also shown in SB_3, which assessed only articles with a reference standard different from imaging [prevalence values: US 0.98 (0-75-1.00), CR 0.63 (0.35-0.84)]. The higher CPPD prevalence by US could be due to the higher sensitivity of this technique compared to CR in detecting CPP deposits at the knee level, as shown in previous studies (11,67).
The results of this SLR also confirmed the common involvement of the wrist in CPPD, even higher than the knee in RQ1 according to . Surprisingly, CPPD prevalence in the wrist was higher in US only in RQ1 and not in RQ2, but this is probably due to the widespread use of CR in RQ2 studies. In fact, the limited data available in the literature regarding a comparative assessment of the wrist showed a higher capability of US in detecting CPP deposits (11). For other joints, the hip showed a lower prevalence than the elbow or shoulder (0.27, 0.41 and 0.37 in the hip, elbow, and shoulder, respectively), but these results were obtained in a small number of patients and should be further addressed. Furthermore, all results were obtained only by CR and could be different if US was applied. Finally, the hand was the joint characterized by the lowest CPPD prevalence, from the 0.10 at the RQ1 to the 0.18 of the RQ2.
In addition, the descriptive results of the articular structures substantially confirmed the prevalence distribution of the whole joint, with higher values at the menisci, TFC, and knee hyaline Frontiers in Medicine 12 frontiersin.org    Forest Plot Knee, Research Question 2 (RQ 2), Sub_analysis 2: patients with suspected CPPD and knee/wrist used as index joint, independently of the reference standard used for diagnosis, analyzed by imaging. Forest Plot Knee, Research Question 2 (RQ 2), Sub_analysis 1: patients with suspected CPPD and the knee joint used as index joint, independently of the reference standard used for diagnosis, analyzed by imaging.
Frontiers in Medicine 15 frontiersin.org cartilage. Again, these results were mainly obtained through CR, which could have some limitations. The results of this SLR, although interesting, should be carefully evaluated. The included studies were heterogeneous in design, population, and reference standards. However, some degree of heterogeneity could be expected, as this is a frequent finding in meta-analysis (112). A common limitation of these studies was the use of CR as an index and reference test simultaneously, leading to a potential misidentification of the CPPD given the low sensitivity of CR, and mainly to an overestimation of the prevalence of deposition. Another issue regarding the joints assessed is that most of the articles evaluated only the knee and/or wrist, and very little data were available for other joints, making the results poorly reliable. Finally, other potentially very sensitive imaging techniques, such as computed tomography (CT) or dual-energy CT, were not included in this systematic review because very little data were available.
On the other hand, this SLR was the first attempt to collect literature data about the distribution of CPP deposits at peripheral joints using imaging techniques mainly applied in clinical practice. Furthermore, this SLR provided results regarding the single joint's structure and bilateral involvement, and these data could be useful in clinical practice. The strengths of this SLR were the identification of sub-groups, ability to reduce the sources of heterogeneity, and the inclusion of meta-analysis aimed to assess the impact of the factors that mainly affected CPPD recognition by imaging: the index joint used to identify the deposits, and the reference standard used to confirm it. Finally, the overall quality of the studies included in the present SLR was acceptable, and the risk of bias was low to moderate.
Considering all the issues that have emerged, the future research agenda should include studies providing polyarticular assessment of CPPD patients, the definition of a tool for monitoring CPPD, and the planning of prospective studies.
In conclusion, the results of this SLR showed that the knee and wrist have the highest CPPD prevalence and should be incorporated into the set of joints for a CPPD follow-up. Furthermore, a higher prevalence of CPP deposits in the US was confirmed. Further, this SLR highlighted the widespread heterogeneity of the studies on CPPD, especially regarding the reference standard applied. This SLR will be the starting point for the development of a US scoring system by the OMERACT US working group for CPPD that could place US as the most validated tool for CPPD assessment both in clinical practice and for research.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmed.2023.1131362/ full#supplementary-material Forest Plot Knee, Research Question 2 (RQ 2), Sub-analysis 3: patients with suspected CPPD diagnosis and reference standard different than imaging, analyzed by imaging.