Seminars in Arthritis and Rheumatism 55 (2022) 152017 Contents lists available at ScienceDirect Seminars in Arthritis and Rheumatism journal homepage: www.elsevier.com/locate/semarthrit Subclinical giant cell arteritis in new onset polymyalgia rheumatica A systematic review and meta-analysis of individual patient data Andrea K. Hemmig a, 1, Daniele Gozzoli b, 1, Laura Werlen c, Hannah Ewald d, Markus Aschwanden e, Daniel Blockmans f, Elisabeth Brouwer g, Russell R.C. Buchanan h, i, Dario Camellino j, Corrado Campochiaro k, Marco A. Cimmino l, Hector Corominas m, Viktoria Gloy c, Liesbet Henckaerts f, Diego Kyburz a, n, Patricia Moya-Alvarado m, Claire E. Owen h, i, Mihaela Stegert a, Alessandro Tomelleri k, Yannick van Sleen g, Hiroyuki Yamashita o, Stephan Imfeld e, Christoph T. Berger p, Lars G. Hemkens c, q, r, Thomas Daikeler a, * a Department of Rheumatology, University Hospital Basel, Basel, Switzerland b Department of Internal Medicine, University Hospital Basel, Basel, Switzerland c Department of Clinical Research, University Hospital Basel, University of Basel, Basel, Switzerland d University Medical Library Basel, University of Basel, Basel, Switzerland e Department of Angiology, University Hospital Basel, Basel, Switzerland f Department of General Internal Medicine, Department of Microbiology, Immunology and Transplantation, KU Leuven, University Hospitals Leuven, Leuven, Belgium g Department of Rheumatology and Clinical Immunology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands h Department of Rheumatology, Austin Health, Heidelberg, Victoria, Australia i Department of Medicine, Austin Health, Melbourne Medical School, University of Melbourne, Parkville, Victoria, Australia j Division of Rheumatology, "La Colletta" Hospital, Local Health Trust 3, Arenzano, Italy k Unit of Immunology, Rheumatology, Allergy and Rare Diseases (UnIRAR), San Raffaele Scientific Institute, Milan, Italy l Research Laboratory and Academic Division of Clinical Rheumatology, Department of Internal Medicine, University of Genoa, Genova, Italy m Department of Rheumatology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain n Department of Biomedicine, University of Basel, Basel, Switzerland o Division of Rheumatic Diseases, National Center for Global Health and Medicine, Tokyo, Japan p Departments of Dermatology, Rheumatology and Internal Medicine, University Center for Immunology, University Hospital Basel, Basel, Switzerland q Meta-Research Innovation Center at Stanford (METRICS), Stanford University, Palo Alto, CA, USA r Meta-Research Innovation Center Berlin (METRIC-B), Berlin Institute of Health, Berlin, Germany A R T I C L E I N F O A B S T R A C T Key words: Objectives: To determine the prevalence and predictors of subclinical giant cell arteritis (GCA) in patients with Polymyalgia rheumatica newly diagnosed polymyalgia rheumatica (PMR). Giant cell arteritis Methods: PubMed, Embase, and Web of Science Core Collection were systematically searched (date of last search Subclinical vasculitis July 14, 2021) for any published information on any consecutively recruited cohort reporting the prevalence of Systematic review Meta-analysis GCA in steroid-naïve patients with PMR without cranial or ischemic symptoms. We combined prevalences across populations in a random-effect meta-analysis. Potential predictors of subclinical GCA were identified by mixed- effect logistic regression using individual patient data (IPD) from cohorts screened with PET/(CT). Results: We included 13 cohorts with 566 patients from studies published between 1965 to 2020. Subclinical GCA was diagnosed by temporal artery biopsy in three studies, ultrasound in three studies, and PET/(CT) in seven studies. The pooled prevalence of subclinical GCA across all studies was 23% (95% CI 14%-36%, I2=84%) for any screening method and 29% in the studies using PET/(CT) (95% CI 13%-53%, I2=85%) (n=266 patients). For seven cohorts we obtained IPD for 243 patients screened with PET/(CT). Inflammatory back pain (OR 2.73, 1.32- For re-submission to: Seminars in Arthritis and Rheumatism * Corresponding author: Thomas Daikeler, Department of Rheumatology, University Hospital Basel, Basel, Petersgraben 4, CH-4031 Basel, Switzerland, Phone: +41 61 265 27 09. E-mail address: Thomas.Daikeler@usb.ch (T. Daikeler). 1 Shared first authorship. https://doi.org/10.1016/j.semarthrit.2022.152017 Available online 28 April 2022 0049-0172/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 5.64), absence of lower limb pain (OR 2.35, 1.05-5.26), female sex (OR 2.31, 1.17-4.58), temperature >37◦ (OR 1.83, 0.90-3.71), weight loss (OR 1.83, 0.96-3.51), thrombocyte count (OR 1.51, 1.05-2.18), and haemoglobin level (OR 0.80, 0.64-1.00) were most strongly associated with subclinical GCA in the univariable analysis but not C-reactive protein (OR 1.00, 1.00-1.01) or erythrocyte sedimentation rate (OR 1.01, 1.00-1.02). A prediction model calculated from these variables had an area under the curve of 0.66 (95% CI 0.55-0.75). Conclusion: More than a quarter of patients with PMR may have subclinical GCA. The prediction model from the most extensive IPD set has only modest diagnostic accuracy. Hence, a paradigm shift in the assessment of PMR patients in favour of implementing imaging studies should be discussed. Introduction following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and PRISMA-IPD statement [19,20]. The Giant cell arteritis (GCA) is the most frequent primary vasculitis. The methods were pre-specified in a protocol registered at the Open Science clinical manifestation is heterogeneous and includes cranial symptoms, Framework (OSF) (doi: 10.17605/OSF.IO/749RY) [21]. peripheral vascular claudication, constitutional symptoms, and poly­ myalgia. Polymyalgia rheumatica (PMR) is a related inflammatory dis­ Systematic review order occurring three to ten times more frequently than GCA [1]. Around half of GCA patients report polymyalgia at diagnosis or during Eligibility criteria relapse, while others have a history of PMR before the onset of GCA. Studies were eligible if they reported the prevalence of subclinical Moreover, PMR may be the sole manifestation of GCA. These patients GCA in consecutively recruited, steroid-naïve PMR patients diagnosed present with subclinical vasculitis without specific vasculitic symptoms according to accepted classification criteria at the time of publication. (subclinical GCA), which can be challenging in the clinical routine [2]. We excluded patients with ischemic or vascular symptoms consistent Timely diagnosis and treatment of GCA are essential because, in contrast with GCA (Supplementary Table S1). to PMR, GCA is associated with acute and chronic vascular complica­ tions such as vision loss (up to 19% of cases), arterial stenosis (5% to Search strategy and selection process 29%), stroke (2% to 7%) and the development of aortic aneurysms (up to A medical information specialist (H.E.) systematically searched 27%) [3–10]. In PMR patients with undetected subclinical GCA, stan­ PubMed, Embase (via Elsevier), and the Web of Science Core Collection dard PMR treatment may be inappropriate, with prednisone doses being using keywords and database-specific subject headings for terms rele­ too low to prevent vascular complications [11]. In a retrospective study, vant to PMR, GCA, and diagnostic imaging (date of last search July 14, 18 out of 167 patients initially diagnosed with PMR developed typical 2021) (Supplementary Table S2). cranial vasculitic symptoms or signs of vascular insufficiency in the Three reviewers (D.G., T.D., A.H.) screened titles and abstracts and upper extremities during follow-up. Half of these patients experienced assessed the full texts of all potentially eligible articles. Disagreements severe complications, including permanent vision loss, stroke, and limb were resolved by discussion. If studies contained overlapping patient claudication [11]. This underscores the relevance of subclinical GCA in cohorts, only the largest study was included. We extracted patient PMR patients. Although subclinical GCA in patients with PMR at diag­ characteristics of the total study population and data on the prevalence nosis has been recognized, routine screening of PMR patients for sub­ of subclinical GCA in PMR patients. If the total study population con­ clinical GCA is not considered standard of care in most clinical settings sisted of multiple patient populations (e.g., PMR, GCA, and GCA with [12]. PMR), only data from PMR patients without symptoms suggestive of Subclinical GCA can be detected by vessel wall imaging [13] or GCA were used for analysis. histology obtained by temporal artery biopsy (TAB) [14]. TAB has limited sensitivity for GCA and was the only screening technique for diagnosing GCA before the introduction of temporal artery ultrasound in Individual patient data meta-analysis 1995 [15]. Ultrasound was later extended to the extra-temporal arteries, thus enhancing diagnostic sensitivity for GCA [16]. Today, PET is the Eligibility criteria most widely used technique for GCA screening, especially for extracra­ For the IPD meta-analysis, cohorts were eligible if they included nial large vessel GCA with good diagnostic accuracy (sensitivity of 67% newly diagnosed, steroid-naïve PMR patients who were screened by to 77% and specificity of 66% to 100%) [17]. It is mostly used in com­ PET/(CT) at diagnosis. Case reports, IPD of patients with ischemic bination with computed tomography (PET/CT) [13,18]. symptoms consistent with GCA, and patients with a pre-existing diag­ Amongst publications reporting the prevalence of subclinical GCA at nosis of GCA were excluded. We restricted this analysis to screening by PMR diagnosis, many have considerable biases, such as non-consecutive PET/(CT) to reflect the modern practice in the diagnostic workup for inclusion of selected patients, and may thus be misleading. GCA. The benefits of a screening program would depend on the true prevalence of subclinical GCA at diagnosis of PMR, the costs, and its Identification of PMR cohorts availability. Hence, easily assessable predictors of subclinical GCA in We contacted authors of eligible studies identified during an initial PMR patients would be helpful. Our study, therefore, has two objectives: search on April 16, 2020 and invited them to share the IPD. We pre­ (i) to determine the prevalence of subclinical GCA in patients with new- sented the project at the European League Against Rheumatism (EULAR) onset PMR based on a systematic literature review and (ii) to identify working group for PMR and GCA web conference in 2020 to invite predictors of subclinical GCA in PMR patients by performing a meta- collaborators and identify additional cohorts. analysis of individual patient data (IPD). To this end, we present an example of a prediction model using the identified variables most Data items strongly associated with the outcome. We requested IPD for demographical (age, sex), clinical (cardiovas­ cular risk factors; pain localized at the neck, shoulders, lower back, Methods lower limbs, and hip girdle; morning stiffness and its duration; fever; weight loss), and laboratory data (erythrocyte sedimentation rate [ESR], We report the systematic literature review and IPD meta-analysis C-reactive protein [CRP], IL-6, leukocytes, thrombocytes, haemoglobin level), information on steroid treatment before PET/(CT) (yes/no), GCA 2 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 symptoms (headache, scalp tenderness, jaw claudication, visual prob­ based on the title and abstract. In total, 13 studies were included for lems), and imaging data (joint or vascular region involved) at diagnosis. prevalence estimation (Fig. 1) [31–43]. Definitions of inflammatory back pain and lower limb pain Study characteristics As has been previously described [22], inflammatory lower back The 13 included studies were published between 1965 and 2020 pain was considered present if the patient reported low back pain that [31–43]. Different methods for subclinical GCA screening were used as improved with exercise, but not with rest, and occurred predominantly state-of-the-art diagnostics at the time of publication. The patients were at night [22]. Bilateral diffuse lower limb pain was defined as pain in examined by TAB (n=165) (published from 1965 to 1996) [31–33], both legs, thighs, or anywhere between the knees and ankles [23]. ultrasound (n=135) (from 1996 to 2020) [34–36], and PET/(CT) (n=266) (from 1999 to 2020) [37–43]. We found no studies that used Individual patient data collection magnetic resonance tomography or CT for screening. Collaborating researchers who agreed to share IPD from eligible The aims of the studies were heterogeneous and included the eval­ studies provided data on predefined variables for individual patients. uation of possible subclinical GCA in PMR and assessing patterns of The data were individually reviewed by D.G. and T.D., and checked for articular and periarticular involvement in PMR. Apart from age and sex, discrepancies with published reports, internal consistency, and patients’ characteristics in these studies were mainly not reported completeness. Any ambiguities were resolved by discussion with the (Table 1). collaborating researchers. Prevalence of subclinical GCA in newly diagnosed PMR patients Risk of bias assessment Overall, the 13 included studies reported on 566 PMR patients. Prevalence estimates of vasculitis ranged from 0 to 92%. The combined We assessed the risk of bias using selected items of the Quality pooled prevalence was 23% (95% confidence interval [CI] 14%–36%) Assessment of Diagnostic Studies (QUADAS)-2 tool [24]. We assessed with high heterogeneity between studies (I2=84%) (Fig. 2). patient selection and outcome measurement (i.e., the method used to The pooled prevalence of subclinical GCA among patients screened diagnose subclinical GCA). We tailored the tool to our review question by biopsy was 20% (CI, 7%–46%, I2=86%) [31–33]. Ultrasound signs and added signalling questions regarding sample frame, patient consistent with vasculitis were found in 15% (CI, 3%–50%, I2=86%) recruitment, the total number of eligible patients, and participation rate [34–36]. For the seven PET/(CT) studies, we found a pooled prevalence (Supplementary Table S3). of 29% (CI, 13%–53%, I2=85%) [37–43] (Fig. 3). The method used for subclinical GCA screening did not explain the between-study Statistical analysis heterogeneity. We calculated the pooled prevalence of subclinical GCA using a Predictors of subclinical GCA random-effects model and calculated I2 to quantify the degree of het­ erogeneity across the studies [25]. Study selection, IPD obtained, and patient characteristics For the IPD, we present summary statistics of the included variables We identified ten eligible studies for the IPD meta-analysis: nine by study. We imputed missing data using multilevel joint modelling PET/(CT) studies were identified during the initial literature search multiple imputations with the R package jomo [26]. We performed a [37–41,44–47]. One unpublished Italian study (Milan study; 24 pa­ one-step IPD meta-analysis to compare the odds of having subclinical tients) was identified after the presentation of our study at the EULAR GCA for characteristics of PMR patients. Using the imputed data sets, we working group meeting. We obtained IPD from seven cohorts [40,41, built univariable mixed-effects logistic regression models with vascular 44–47] for 256 patients, of whom 243 were included in the final analysis involvement (no/yes) as the outcome and each patient characteristic (Supplementary Fig. S1 [20]). Three cohorts had no data available (53 under study as a predictor in turn. We then modelled the outcome patients) [37–39]. including all predictors together in a multivariable model. The charac­ Most of the 243 patients were female (n=146; 60.1%), the median teristics were specified as fixed effects, while the study was included as a age was 72.3 years (interquartile range [IQR] 66.4–78.0 years), median random effect. The results for each imputed data set were pooled using CRP was 46.0 mg/L (IQR 19.0–77.7 mg/L), mean ESR was 65.2 mm/h Rubin’s rules [27]. We included the variables that were significantly (standard deviation ± 30.3 mm/h) and 65 patients (27%) were diag­ associated with subclinical GCA or had confidence intervals slightly nosed with subclinical GCA after PET/CT. Cardiovascular risk factors overlapping ‘1’ in the univariable model and kept the direction and such as hypertension was found in 29.6%, dyslipidaemia was seen in magnitude of their association with the outcome after adjusting for the 15.6%, and 4.9% were smokers. Diabetes was found in 9.5%. Detailed other predictors in the multivariable model in a prediction model. We patient characteristics are described in the supplemental material used the R package psfmi [28] to pool the mixed logistic regression (Supplementary Table S4 and Supplementary Table S5). models from the multiply imputed data using the median p-values pooling method and evaluating the model’s performance using Factors associated with subclinical GCA in PMR patients cross-validation. We performed a sensitivity analysis for the univariable In the univariable models, the following variables were associated analyses using the complete cases from the original study data using the with subclinical GCA: inflammatory back pain (odds ratio [OR], 2.73; lme4 package [29]. As an additional sensitivity analysis, we redid the CI, 1.32–5.64), absence of lower limb pain (OR, 2.35; CI, 1.05–5.26), analyses, excluding two studies reporting an extreme prevalence of female sex (OR, 2.31; CI, 1.17–4.58), and the thrombocyte count (OR, vascular involvement. Statistical significance was defined as p < 0.05. 1.51 for an increase in thrombocyte count of one standard deviation; CI, All analyses were performed in R version 4.0.3 (2020-10-10) [30]. 1.05–2.18). An increase in haemoglobin level by 1 g (OR, 0.80 per g/dl; CI, 0.64–1.00) was associated with lower odds for subclinical GCA. Results In the multivariable analysis, inflammatory back pain (OR, 5.71; CI, 1.41–23.06) and absence of lower limb pain (OR, 3.48; CI, 1.16–10.42) Prevalence of subclinical GCA at PMR diagnosis were the only statistically significant predictors (Table 2). Age, pelvic girdle pain, duration of morning stiffness, and elevation Study selection of inflammation markers (ESR, CRP, leukocyte count), were not asso­ A total of 2362 reports were identified after removing duplicates. Of ciated with subclinical GCA, both in the univariable and multivariable those, we reviewed 70 full-text articles deemed potentially relevant models. 3 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 Fig. 1. Flowchart of the study selection process. Prediction model with candidate predictors impact on applicability in most cases. The studies using TAB for We included the variables that had the strongest associations with screening are potentially biased due to the lack of clear-cut definitions of subclinical GCA from the univariable and multivariable models in a PMR and GCA at that time. prediction model and used cross-validation to assess model perfor­ mance. Those variables were inflammatory back pain, lower limb pain, Discussion female sex, temperature > 37◦ , weight loss, thrombocyte count, and haemoglobin level. The prediction model’s ORs and 95% CIs can be In this systematic review of 13 studies reporting on 566 steroid-naive found in the supplementary material (Supplementary Table S6). The patients with PMR about one in four patients were judged to have cross-validation procedure calculated an area under the curve (AUC) of subclinical GCA. The pooled prevalence of subclinical GCA among pa­ 0.66 (95% CI: 0.55-0.75) and a pseudo-R2 of 0.178. tients screened by PET/(CT), the standard nowadays, was even higher, as expected due to the high sensitivity of PET/(CT) in this setting. Sensitivity analyses Evaluation of the temporal artery during routine scanning with The results of the univariable analyses using the complete cases from advanced PET/CT scanners is now possible, which might further in­ the original study data were similar to those obtained from the imputed crease the sensitivity of PET/CT for GCA in the future [48]. data (Supplementary Table S7). In the multivariable analyses excluding The cumulative lifetime risk of patients with PMR developing GCA the two studies that reported extreme prevalence of subclinical GCA may be even higher and has not yet been systematically studied. (Owen et al. [44] with 0% and the Milan study with 71%), inflammatory Altogether, this supports the ‘one disease’ hypothesis that was back pain and lower limb pain were no longer statistically significant already postulated 57 years earlier by Hamrin, i.e. that GCA and isolated (for details see Supplementary Table S8–S10). PMR represent a different spectrum of the same disease [49]. We retrieved IPD from seven cohorts using PET/(CT) for screening. Risk of bias and applicability concerns This enabled us to study characteristics of PMR patients with and Patient selection was a main source of potential bias in most studies without subclinical GCA at diagnosis in the largest existing dataset. In since most patients were recruited from specialized centres or the the univariable analysis, potential predictors of subclinical GCA were description of recruitment was incomplete (overall summary in Sup­ female sex and weight loss and variables indirectly associated with plementary Table S11 and S12), but we deemed this to have limited inflammation. These were thrombocytosis, anaemia, and temperature 4 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 Table 1 Study and patient characteristics. Study Location Patients screened Patients diagnosed Females, n Mean age Diagnostic procedure Vascular area examined for GCA, n with GCA, n (%) (%) (yrs ± SD) Hamrin 1965 Sweden 30 11 (37) NA NA TAB TA [31] Bengtsson 1981 Sweden 67 21 (31) 50 (75) NA TAB TA [32] Myklebust 1996 Norway 68 3 (4) NA NA TAB TA [33] Kraft 1996 [34] Germany 8 0 (0) 7 (88) 70 Ultrasound TA Schmidt 2002 Germany 102 8 (8) 71 (70) 69 Ultrasound TA [35] Burg 2020 [36] Germany 25 10 (40) NA NA Ultrasound TA, Occ, CRA, Carot, Vert, Ax Blockmans 1999 Belgium 5 4 (80) 5 (100) 63 ± 6 PET (4P-VS) Ao, Sub, Carot, Pop, Fem, Tib [37] Moosig 2004 Germany 13 12 (92) 11 (85) 65.5 PET (Vasc-ROI) Ao, Sub, ext Carot, Iliac [38] Blockmans 2007 Belgium 35 11 (31) 20 (57) 68.5 ± 7.2 PET (TVS) Ao, Sub, Carot, Ax, Iliac, Fem [39] Camellino 2012 Italy 64 25 (39) NA NA PET/CT (4P-VS*) Ao, Sub, Carot, Iliac, Fem [40] Corica 2019 Spain 52 6 (12) NA NA PET/CT (qualitative) NA [41] Owen 2020 Australia 33 0 (0) 15 (45) 68.6 ± 7.4 PET/CT (4P-VS* + semi- NA [42] quantitative) Emamifar 2020 Denmark 64 6 (9) NA NA PET/CT (4P-VS*) TA, MA, BA, Ao, Sub, ext [43] Carot, Vert, Iliac, Fem Abbreviations: Ao, aorta; Ax, axillary artery; BA, basilar artery; n, number; Carot, carotid artery; CRA, central retinal artery; CT, computed tomography; ext Carot, external carotid artery; Fem, femoral artery; GCA, giant cell arteritis; Iliac, iliac arteries; NA, not available; MA, maxillary artery; Occ, occipital artery; PET, positron emission tomography; PET/CT, positron emission tomography/computed tomography; Pop, popliteal artery; SD, standard deviation; Sub, subclavian artery; TA, temporal artery; TAB, temporal artery biopsy; Tib, tibial artery; TVS, total vascular score (seven vascular regions were scored as negative (0) or positive (1, 2, or 3), and scores were summed (max score: 21), TVS of ≥1 were regarded as positive for vascular uptake); Vasc-ROI, vascular region of interest divided by individual background value (peripheral region of the lung); Vert, vertebral artery; 4P-VS, four-point visual score ranging from 0 (no visualization of blood vessels) to 3 (intense FDG uptake), scores of ≥2 were regarded as positive for vascular uptake; 4P-VS*, four-point visual score relative to liver uptake (0=no uptake, 1=lower than liver uptake, 2=similar to liver uptake, 3=higher than liver uptake, with scores of ≥2 regarded as positive for vascular uptake). above 37◦ C, but unexpectedly, classical inflammatory markers such as primary care physician. Consequently, it is unclear whether the preva­ ESR and CRP were not amongst them. After multivariable analysis, only lence estimates apply to a broader population in primary care settings. inflammatory low back pain and absence of lower limb pain remained Third, the collected parameters for the IPD data were, in part, significantly associated with subclinical GCA in PMR patients. However, incomplete. While we imputed missing values, this led to more statistical the presence of these two pain symptoms only identifies a minority of uncertainty and weakened the statistical power to detect relevant patients with subclinical GCA. Although lower limb and inflammatory factors. back pain have been reported in the context of newly diagnosed PMR, Fourth, our study only analysed routinely collected laboratory non-inflammatory musculoskeletal pains are frequent in elderly pa­ markers. However, the recent study by van Sleen et al. has shown that tients. Therefore, these two symptoms have to be interpreted with nonroutine biomarkers such as angiopoietin-2 are associated with large caution. The presence of inflammatory back pain showed no statistical vessel vasculitis in PMR [46]. Further studies are needed to identify new correlation with the FDG uptake in the abdominal aorta or in the bursae potential biomarkers of subclinical GCA. of the spinous processes of the lumbar spine (data not shown). In conclusion, this systematic review and meta-analysis indicated To create a tool to predict subclinical GCA, we included all variables that subclinical GCA might affect more than one in four patients with that showed an association with subclinical GCA in the univariable PMR when screened by PET/(CT) at diagnosis. Although we identified model and maintained this relationship in the multivariable model. factors associated with subclinical GCA, the development of a helpful However, the discriminatory ability of this prediction model was prediction model to systematically screen patients with PMR requires modest, with an AUC of 0.66%. more data and the subsequent assessment. Due to the high prevalence of Strengths of the study include the large number of analysed IPD sets subclinical GCA at PMR diagnosis, the role of routine imaging for GCA and the strict inclusion criteria, which led to the exclusion of several warrants discussion. Although PET/CT has been most often used, other large cohorts that had reported subclinical GCA in patients with PMR modalities such as CT, ultrasound, or MRI, each with its specific ad­ but had not included patients consecutively. Thus, although struggling vantages and disadvantages, are available for this purpose [13]. Care­ with potential biases, the prevalence derived from our meta-analysis is fully designed prospective, longitudinal cohort studies, taking the to date the most reliable estimation. potential biases in account would be needed to confirm the presented Limitations to the study include the considerable heterogeneity data. These studies should also assess the long-term damage caused by across the included studies. This was not statistically explained by subclinical GCA, the diagnostic accuracy and the cost effectiveness of diagnostic methodology which evolved over the decades (TAB, ultra­ the employed imaging techniques. sound, PET/(CT)). Possible explanations include different diagnostic criteria, but we would need much more data to investigate this. In Author contributions addition, study focus and population selection were not identical for all studies. DG and AH contributed equally to this work as joint first authors. TD, Second, as GCA is primarily a disease treated by specialists/rheu­ DG, AH, LGH, LW, and HE were responsible for the study conception and matologists, most studies included patients referred to specialist care, design. HE provided the search strategy and execution. Screening of the whereas the majority of PMR patients are usually treated by their studies was conducted by DG, TD, and AH. DG, TD, and AH contributed 5 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 Fig. 2. Forest plot of the prevalence of subclinical GCA in all studies, stratified by the method used for subclinical GCA screening. Abbreviations: CI, confidence interval; CT, computed tomography; PET, positron emission tomography; TAB, temporal artery biopsy; US, ultrasound. Fig. 3. Forest plot of the prevalence of subclinical GCA in PET/(CT) studies. Abbreviations: CI, confidence interval; GCA, giant cell arteritis; PET/(CT), positron emission tomography with or without computed tomography. to data extraction, and evaluation of the included studies. IPD was authorship criteria and that no others meeting the criteria have been collected and provided by AT, CC, CO, DC, DB, EB, HC, HY, LH, MAC, omitted. PMA, RRCB, YvS. LGH and LW contributed to data synthesis. All authors were involved in data analysis and interpretation. DG, TD, AH, LW, LGH Role of funding source drafted the manuscript and all authors revised it critically. All authors had the final responsibility for the decision to submit for publication. AH is supported by a grant from the Swiss Foundation for Research The corresponding author (TD) attests that all listed authors meet on Muscle Diseases (FSRMM). The funding source had no role in the 6 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 Table 2 speaker fees and consulting fees from Roche in 2017, 2018 which were Association of patient characteristics and subclinical GCA in univariable and paid to the UMCG. HC has received honorariums from Abbvie, Gilead, multivariable analysis. Lilly, Novartis, Janssen and Pfizer. LW has received a research grant Univariable analysis Multivariable analysis from the Swiss National Science Foundation. CTB received funding from Predictor OR (95% CI) p- OR (95% CI) p- the FSRMM and the Swiss National Science Foundation. TD received value value funding from the FSRMM. No other disclosures were reported. Age (in years) 0.99 (0.96, 0.700 0.97 (0.93, 0.326 1.03) 1.03) Acknowledgements Sex (female vs. male) 2.31 (1.17, 0.016 1.43 (0.60, 0.428 4.58) 3.45) The authors would like to thank the patient and public involvement a a Hypertension 0.63 (0.26, 0.294 representative Chasper Knapp for his contribution to the study. We 1.49) Dyslipidaemia 0.61 (0.22, 0.349 a a would also like to thank Christian Appenzeller-Herzog, University 1.70) Medical Library, University of Basel, Basel, Switzerland, for peer- Diabetes mellitus 0.36 (0.08, 0.179 0.36 (0.06, 0.260 reviewing the final search strategy and PD Dr. Michael Mayr for sup­ 1.59) 2.13) porting the doctoral thesis of Daniele Gozzoli as well as the Swiss Current smoking status 2.20 (0.40, 0.373 2.22 (0.24, 0.493 Foundation for Research on Muscle Diseases (FSRMM) for supporting 12.23) 20.72) Shoulder girdle pain 0.65 (0.10, 0.662 0.35 (0.04, 0.339 the Ph.D. of Andrea Hemmig. 4.13) 2.92) Pelvic girdle pain 1.01 (0.48, 0.971 0.71 (0.26, 0.520 Supplementary materials 2.16) 1.96) Inflammatory back pain 2.73 (1.32, 0.007 5.71 (1.41, 0.014 5.64) 23.06) Supplementary material associated with this article can be found, in Neck pain 1.45 (0.68, 0.338 0.38 (0.09, 0.169 the online version, at doi:10.1016/j.semarthrit.2022.152017. 3.09) 1.51) Absence of lower limb pain 2.35 (1.05, 0.038 3.48 (1.16, 0.025 References 5.26) 10.42) a a Morning stiffness 1.11 (0.42, 0.843 [1] Buttgereit F, Dejaco C, Matteson EL, Dasgupta B. Polymyalgia Rheumatica and 2.97) Giant Cell Arteritis: A Systematic Review. JAMA 2016 Jun 14;315(22):2442–58. Duration of morning 1.00 (1.00, 0.640 1.00 (0.99, 0.848 https://doi.org/10.1001/jama.2016.5444. stiffness 1.00) 1.00) [2] de Boysson H, Liozon E, Ly KH, Dumont A, Delmas C, Aouba A. The different Temperature > 37◦ 1.83 (0.90, 0.095 1.29 (0.53, 0.590 clinical patterns of giant cell arteritis. Clin Exp Rheumatol 2019 Apr;37(Suppl 3.71) 3.15) 117):57–60 (2). Weight loss 1.83 (0.96, 0.067 1.38 (0.63, 0.432 [3] Salvarani C, Cimino L, Macchioni P, et al. Risk factors for visual loss in an Italian 3.51) 3.04) population-based cohort of patients with giant cell arteritis. Arthritis Care & Weight loss (in kg) 1.10 (0.99, 0.085 a a Research 2005;53(2):293–7. https://doi.org/10.1002/art.21075. 1.22) [4] González-Gay MA, García-Porrúa C, Llorca J, et al. Visual manifestations of giant ESR (in mm/h) 1.01 (1.00, 0.135 a a cell arteritis. Trends and clinical spectrum in 161 patients. Medicine (Baltimore) 2000;79(5):283–92. https://doi.org/10.1097/00005792-200009000-00001. 1.02) [5] Schmidt WA, Seifert A, Gromnica-Ihle E, Krause A, Natusch A. Ultrasound of CRP (in mg/l) 1.00 (1.00, 0.365 1.00 (0.99, 0.886 proximal upper extremity arteries to increase the diagnostic yield in large-vessel 1.01) 1.01) giant cell arteritis. Rheumatology (Oxford). 2008 Jan;47(1):96–101. https://doi.or Haemoglobin (in g/dl) 0.80 (0.64, 0.051 0.81 (0.59, 0.195 g/10.1093/rheumatology/kem322. 1.00) 1.11) [6] Gonzalez-Gay MA, Vazquez-Rodriguez TR, Gomez-Acebo I, et al. Strokes at time of a a Leukocyte count (in G/l) 1.02 (0.87, 0.838 disease diagnosis in a series of 287 patients with biopsy-proven giant cell arteritis. 1.19) Medicine (Baltimore) 2009;88(4):227–35. https://doi.org/10.1097/ Scaled Thrombocyte count 1.51 (1.05, 0.027 1.37 (0.80, 0.249 MD.0b013e3181af4518. (per SD)b 2.18) 2.33) [7] Caselli RJ, Hunder GG, Whisnant JP. Neurologic disease in biopsy-proven giant cell (temporal) arteritis. Neurology 1988 Mar;38(3):352–9. https://doi.org/10.1212/ Abbreviations: CI, confidence interval; CRP, C-reactive protein; ESR, erythrocyte wnl.38.3.352. sedimentation rate; GCA, giant cell arteritis, OR, odds ratio. [8] Nuenninghoff DM, Hunder GG, Christianson TJH, McClelland RL, Matteson EL. a To fit the multivariable model, we excluded variables that were hypothe­ Incidence and predictors of large-artery complication (aortic aneurysm, aortic dissection, and/or large-artery stenosis) in patients with giant cell arteritis: a sized to have less medical relevance for the outcome (hypertension and dysli­ population-based study over 50 years. Arthritis Rheum. 2003 Dec;48(12):3522–31. pidaemia) as well as highly correlated inflammation markers (ESR and leukocyte https://doi.org/10.1002/art.11353. count). [9] García-Martínez A, Hernández-Rodríguez J, Arguis P, et al. Development of aortic b We scaled the variable thrombocyte count to improve model fit. aneurysm/dilatation during the followup of patients with giant cell arteritis: a cross-sectional screening of fifty-four prospectively followed patients. Arthritis Rheum 2008;59(3):422–30. https://doi.org/10.1002/art.23315. study design, in the collection, analysis and interpretation of data, in the [10] Berger CT, Wolbers M, Meyer P, Daikeler T, Hess C. High incidence of severe writing of the report, and in the decision to submit the article for ischaemic complications in patients with giant cell arteritis irrespective of platelet count and size, and platelet inhibition. Rheumatology (Oxford). 2009 Mar;48(3): publication. 258–61. https://doi.org/10.1093/rheumatology/ken480. [11] Narváez J, Estrada P, López-Vives L, et al. Prevalence of ischemic complications in Data availability statement patients with giant cell arteritis presenting with apparently isolated polymyalgia rheumatica. Seminars in Arthritis and Rheumatism.2015;45(3):328–33. https://doi. org/10.1016/j.semarthrit.2015.06.009. The data used and analysed during this study are available from the [12] Dejaco C, Singh YP, Perel P, et al. 2015 Recommendations for the management of corresponding author upon request. polymyalgia rheumatica: a European League Against Rheumatism/American College of Rheumatology collaborative initiative. Ann Rheum Dis 2015;74(10): 1799–807. https://doi.org/10.1136/annrheumdis-2015-207492. Declaration of competing interest [13] Berger CT, Sommer G, Aschwanden M, Staub D, Rottenburger C, Daikeler T. The clinical benefit of imaging in the diagnosis and treatment of giant cell arteritis. AH is funded by a grant from the Swiss Foundation for Research on Swiss Med Wkly 2018 Aug 13;148:w14661. https://doi.org/10.4414/ smw.2018.14661. Muscle Diseases (FSRMM) for her Ph.D. thesis. CC has received hono­ [14] Luqmani R, Lee E, Singh S, et al. The Role of Ultrasound Compared to Biopsy of rariums for lectures and educational events from Roche. CO has received Temporal Arteries in the Diagnosis and Treatment of Giant Cell Arteritis (TABUL): speaking honoraria from Roche and Novartis and meeting sponsorship a diagnostic accuracy and cost-effectiveness study. Health Technol Assess 2016;20 (90):1–238. https://doi.org/10.3310/hta20900. from Roche. DK has received honorariums from Abbvie, Gilead, Lilly, Novartis, Janssen, and Pfizer. EB as an employee of the UMCG, received 7 A.K. Hemmig et al. Seminars in Arthritis and Rheumatism 55 (2022) 152017 [15] Ball EL, Walsh SR, Tang TY, Gohil R, Clarke JMF. Role of ultrasonography in the [34] Kraft HE, Möller DE, Völker L, Schmidt WA. Color Doppler ultrasound of the diagnosis of temporal arteritis. Br J Surg 2010 Dec;97(12):1765–71. https://doi. temporal arteries–a new method for diagnosing temporal arteritis. Klin Monbl org/10.1002/bjs.7252. Augenheilkd 1996;208(2):93–5. https://doi.org/10.1055/s-2008-1035176. [16] Aschwanden M, Kesten F, Stern M, et al. Vascular involvement in patients with [35] Schmidt WA, Gromnica-Ihle E. Incidence of temporal arteritis in patients with giant cell arteritis determined by duplex sonography of 2×11 arterial regions. Ann polymyalgia rheumatica: a prospective study using colour Doppler Rheum Dis 2010 Jul;69(7):1356–9. https://doi.org/10.1136/ard.2009.122135. ultrasonography of the temporal arteries. Rheumatology (Oxford) 2002 Jan;41(1): [17] Duftner C, Dejaco C, Sepriano A, Falzon L, Schmidt WA, Ramiro S. Imaging in 46–52. https://doi.org/10.1093/rheumatology/41.1.46. diagnosis, outcome prediction and monitoring of large vessel vasculitis: a [36] Burg L, Brossart P, Behning C, Schaefer V. Prospective Analysis of the Prevalence of systematic literature review and meta-analysis informing the EULAR Giant Cell Arteritis in Consecutive Newly Diagnosed Patients with Polymyalgia recommendations. RMD Open 2018;4(1):e000612. https://doi.org/10.1136/ Rheumatica [abstract]. Arthritis Rheumatol 2020;72(suppl 10)). https:// rmdopen-2017-000612. acrabstracts.org/abstract/prospective-analysis-of-the-prevalence-of-giant-cell- [18] Imfeld S, Rottenburger C, Schegk E, et al. [18F]FDG positron emission tomography arteritis-in-consecutive-newly-diagnosed-patients-with-polymyalgia-rheumatica/. in patients presenting with suspicion of giant cell arteritis-lessons from a vasculitis [37] Blockmans D, Maes A, Stroobants S, et al. New arguments for a vasculitic nature of clinic. Eur Heart J Cardiovasc Imaging 2018 Aug 1;19(8):933–40. https://doi.org/ polymyalgia rheumatica using positron emission tomography. Rheumatology 10.1093/ehjci/jex259. (Oxford) 1999;38(5):444–7. https://doi.org/10.1093/rheumatology/38.5.444. [19] Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated [38] Moosig F, Czech N, Mehl C, et al. Correlation between 18-fluorodeoxyglucose guideline for reporting systematic reviews. BMJ 2021;372:n71. https://doi.org/ accumulation in large vessels and serological markers of inflammation in 10.1136/bmj.n71. polymyalgia rheumatica: a quantitative PET study. Ann Rheum Dis 2004;63(7): [20] Stewart LA, Clarke M, Rovers M, et al. Preferred Reporting Items for Systematic 870–3. https://doi.org/10.1136/ard.2003.011692. Review and Meta-Analyses of individual participant data: the PRISMA-IPD [39] Blockmans D, De Ceuninck L, Vanderschueren S, Knockaert D, Mortelmans L, Statement. JAMA 2015;313(16):1657–65. https://doi.org/10.1001/ Bobbaers H. Repetitive 18-fluorodeoxyglucose positron emission tomography in jama.2015.3656. isolated polymyalgia rheumatica: a prospective study in 35 patients. Rheumatology [21] Gozzoli, D. S., Hemmig, A., Hemkens, et al. (2021, August 3). Findings consistent (Oxford). 2007 Apr;46(4):672–7. https://doi.org/10.1093/rheumatology/kel376. with subclinical vasculitis in patients with new onset polymyalgia: a systematic [40] Camellino D, Morbelli S, Paparo F, Massollo M, Sambuceti G, Cimmino MA. literature review and a meta-analysis of cohort data. https://doi.org/10.17605/ Similarities exceeds differences in the pattern of joint and vascular positron OSF.IO/749RY. emission/computed tomography uptake in polymyalgia rheumatica and giant cell [22] Sieper J, van der Heijde D, Landewé R, et al. New criteria for inflammatory back arteritis. Arthritis Rheum 2012. https://acrabstracts.org/abstract/similarities-e pain in patients with chronic back pain: a real patient exercise by experts from the xceeds-differences-in-the-pattern-of-joint-and-vascular-positron-emissioncompute Assessment of SpondyloArthritis international Society (ASAS). Ann Rheum Dis d-tomography-uptake-in-polymyalgia-rheumatica-and-giant-cell-arteritis/. 2009;68(6):784–8. https://doi.org/10.1136/ard.2008.101501. [41] Corica ME, Moya P, Fernandez A, et al. THU0603 Suitability of PET-CT in [23] Prieto-Peña D, Martínez-Rodríguez I, Loricera J, et al. Predictors of positive 18F- refractory polymyalgia rheumatica. Ann Rheum Dis 2019;78:593–5. https://doi. FDG PET/CT-scan for large vessel vasculitis in patients with persistent polymyalgia org/10.1136/annrheumdis-2019-eular.6162. rheumatica. Semin Arthritis Rheum 2019;48(4):720–7. https://doi.org/10.1016/j. [42] Owen CE, Poon AMT, Yang V, et al. Abnormalities at three musculoskeletal sites on semarthrit.2018.05.007. whole-body positron emission tomography/computed tomography can diagnose [24] Whiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2: A Revised Tool for the polymyalgia rheumatica with high sensitivity and specificity. Eur J Nucl Med Mol Quality Assessment of Diagnostic Accuracy Studies. Ann Intern Med 2011;155(8): Imaging 2020;47(10):2461–8. https://doi.org/10.1007/s00259-020-04731-z. 529–36. https://doi.org/10.7326/0003-4819-155-8-201110180-00009. [43] Emamifar A, Ellingsen T, Hess S, et al. The Utility of 18F-FDG PET/CT in Patients [25] Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in With Clinical Suspicion of Polymyalgia Rheumatica and Giant Cell Arteritis: A meta-analyses. BMJ 2003 Sep 6;327(7414):557–60. https://doi.org/10.1136/ Prospective, Observational, and Cross-sectional Study. ACR Open Rheumatol 2020; bmj.327.7414.557. 2(8):478–90. https://doi.org/10.1002/acr2.11163. [26] Carpenter JR, Kenward MG. Multiple imputation and its application. Chichester, [44] Owen C, Poon A, Lee S, et al. Fusion of positron emission tomography/computed West Sussex: John Wiley & Sons; 2013. tomography with magnetic resonance imaging reveals hamstring peritendonitis in [27] Little RJA, Rubin DB. Statistical Analysis with Missing Data. 2nd edition. New polymyalgia rheumatica. Rheumatology (Oxford) 2018;57(2):345–53. https://doi. York: John Wiley & Sons; 2002. org/10.1093/rheumatology/kex411. [28] Heymans Martijn W. psfmi: Prediction Model Pooling, Selection and Performance [45] Yamashita H, Kubota K, Takahashi Y, et al. Whole-body fluorodeoxyglucose Evaluation, Across Multiply Imputed Datasets. 2021. R package version 1.0.0. https positron emission tomography/computed tomography in patients with active ://mwheymans.github.io/psfmi/. polymyalgia rheumatica: evidence for distinctive bursitis and large-vessel [29] Bates, D, Maechler, M, Bolker, B, Walker, S. 2015: lme4: Linear mixed-effects vasculitis. Mod Rheumatol 2012;22(5):705–11. https://doi.org/10.1007/s10165- models using Eigen and S4. R package version 1.1-8. 011-0581-x. [30] R Core Team. R: A Language and Environment for Statistical Computing. Vienna, [46] van Sleen Y, Boots AMH, Abdulahad WH, et al. High angiopoietin-2 levels associate Austria: R Foundation for Statistical Computing; 2020. http://www.r-project.org/i with arterial inflammation and long-term glucocorticoid requirement in ndex.html [accessed August 29, 2021]. polymyalgia rheumatica. Rheumatology (Oxford) 2020;59(1):176–84. https://doi. [31] Hamrin B, Jonsson N, Landberg T. Involvement of large vessels in polymyalgia org/10.1093/rheumatology/kez261. arteritica. Lancet 1965;1(7397):1193–6. https://doi.org/10.1016/s0140-6736(65) [47] Henckaerts L, Gheysens O, Vanderschueren S, Goffin K, Blockmans D. Use of 18F- 92723-6. fluorodeoxyglucose positron emission tomography in the diagnosis of polymyalgia [32] Bengtsson BA, Malmvall BE. The epidemiology of giant cell arteritis including rheumatica-A prospective study of 99 patients. Rheumatology (Oxford) 2018 Nov temporal arteritis and polymyalgia rheumatica. Incidences of different clinical 1;57(11):1908–16. https://doi.org/10.1093/rheumatology/kex376. presentations and eye complications Arthritis Rheum 1981 Jul;24(7):899–904. [48] Rottenburger C, Mensch N, Imfeld S, et al. 18F-FDG PET/CT compared with https://doi.org/10.1002/art.1780240706. ultrasound and biopsy for detection of vasculitis of the temporal artery branches. [33] Myklebust G, Gran JT. A prospective study of 287 patients with polymyalgia Swiss Med Wkly 2021;151:w20512. https://doi.org/10.4414/smw.2021.20512. rheumatica and temporal arteritis: clinical and laboratory manifestations at onset [49] Hamrin B, Jonsson N, Landberg T. Arteritis in ‘polymyalgia rheumatica. Lancet of disease and at the time of diagnosis. Br J Rheumatol 1996 Nov;35(11):1161–8. 1964;1(7330):397–401. https://doi.org/10.1016/s0140-6736(64)92784-9. https://doi.org/10.1093/rheumatology/35.11.1161. 8