eTable 1. Search Strategies
eTable 2. Details of Included Studies
eTable 3. Studies With TAB as Reference Standard
eTable 4. Studies With Clinical Diagnosis as Reference Standard
eTable 5. Study-Specific Definitions of Clinical Diagnosis
eTable 6. Symptoms With Limited Data Available
eTable 7. Physical Findings With Limited Data Available
eTable 8. Sensitivity Analysis: Distinct Reference Standards
eTable 9. Sensitivity Analysis: Prospective vs Retrospective Studies
eTable 10. Sensitivity Analysis: Pretreatment Laboratory Tests
eAppendix. Standard Form for Study Characteristics and QUADAS-2 Items
eFigure 1. PRISMA Flow Diagram
eFigure 2. Overall Summary of QUADAS-2 Items
eFigure 3. Detailed Summary of QUADAS-2 Items
eFigure 4. Forests Plots
eFigure 5. HSROC Curves
eFigure 6. Funnel Plots
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
van der Geest KSM, Sandovici M, Brouwer E, Mackie SL. Diagnostic Accuracy of Symptoms, Physical Signs, and Laboratory Tests for Giant Cell Arteritis: A Systematic Review and Meta-analysis. JAMA Intern Med. Published online August 17, 2020. doi:10.1001/jamainternmed.2020.3050
In patients with suspected giant cell arteritis, which clinical and laboratory findings can help to identify the disease?
This systematic review and meta-analysis of 68 unique diagnostic cohort studies (14 037 unique patients) identified combinations of symptoms, physical signs, and laboratory tests that were informative with regard to the presence or absence of giant cell arteritis, but no single feature taken alone. Headache and scalp tenderness were poorly informative in this population.
These findings suggest that in patients with suspected giant cell arteritis, no single clinical or laboratory feature is sufficient to rule in or rule out the disease; therefore, additional investigations (vascular imaging and/or temporal artery biopsy) are required.
Current clinical guidelines recommend selecting diagnostic tests for giant cell arteritis (GCA) based on pretest probability that the disease is present, but how pretest probability should be estimated remains unclear.
To evaluate the diagnostic accuracy of symptoms, physical signs, and laboratory tests for suspected GCA.
PubMed, EMBASE, and the Cochrane Database of Systematic Reviews were searched from November 1940 through April 5, 2020.
Trials and observational studies describing patients with suspected GCA, using an appropriate reference standard for GCA (temporal artery biopsy, imaging test, or clinical diagnosis), and with available data for at least 1 symptom, physical sign, or laboratory test.
Data Extraction and Synthesis
Screening, full text review, quality assessment, and data extraction by 2 investigators. Diagnostic test meta-analysis used a bivariate model.
Main Outcome(s) and Measures
Diagnostic accuracy parameters, including positive and negative likelihood ratios (LRs).
In 68 unique studies (14 037 unique patients with suspected GCA; of 7798 patients with sex reported, 5193 were women [66.6%]), findings associated with a diagnosis of GCA included limb claudication (positive LR, 6.01; 95% CI, 1.38-26.16), jaw claudication (positive LR, 4.90; 95% CI, 3.74-6.41), temporal artery thickening (positive LR, 4.70; 95% CI, 2.65-8.33), temporal artery loss of pulse (positive LR, 3.25; 95% CI, 2.49-4.23), platelet count of greater than 400 × 103/μL (positive LR, 3.75; 95% CI, 2.12-6.64), temporal tenderness (positive LR, 3.14; 95% CI, 1.14-8.65), and erythrocyte sedimentation rate greater than 100 mm/h (positive LR, 3.11; 95% CI, 1.43-6.78). Findings that were associated with absence of GCA included the absence of erythrocyte sedimentation rate of greater than 40 mm/h (negative LR, 0.18; 95% CI, 0.08-0.44), absence of C-reactive protein level of 2.5 mg/dL or more (negative LR, 0.38; 95% CI, 0.25-0.59), and absence of age over 70 years (negative LR, 0.48; 95% CI, 0.27-0.86).
Conclusions and Relevance
This study identifies the clinical and laboratory features that are most informative for a diagnosis of GCA, although no single feature was strong enough to confirm or refute the diagnosis if taken alone. Combinations of these symptoms might help direct further investigation, such as vascular imaging, temporal artery biopsy, or seeking evaluation for alternative diagnoses.
Giant cell arteritis (GCA) is a “do-not-miss” diagnosis. Prompt diagnosis can avert visual loss.1 Diagnosis can be delayed in those without the classic cranial features, such as headache.2 Treatment for GCA consists of high-dose glucocorticoids tapered during the course of 1 year or more, but this treatment may cause substantial toxic effects, so diagnostic uncertainty must be minimized.3
Making a diagnosis of GCA can be challenging. The American College of Rheumatology 1990 criteria for the classification of GCA in research studies should not be used for clinical diagnosis.4,5 Instead, temporal artery biopsy (TAB; highly specific but with imperfect sensitivity),6 vascular imaging (ultrasonography, computed tomography, magnetic resonance imaging, or positron emission tomography),7 or a combination of these tests are recommended.3,7 These further investigations should be selected based on pretest probability.3,7 The difficulty in practice is how to quantify pretest probability given only symptoms, signs, and, if available, laboratory features. Regression, machine learning models, or clinical scoring systems have been suggested, but these rely on complete information and still require further validation.8,9 Pretest probability might additionally be estimated by using likelihood ratios (LRs) of clinical features to allow sequential bayesian probability revision.10 A previous meta-analysis11 reported pooled estimates of the LRs of clinical and laboratory features for a positive TAB finding. However, this previous meta-analysis included studies comparing TAB-positive vs TAB-negative GCA, which is not appropriate for estimating diagnostic accuracy. The previous meta-analysis also included diagnostic case-control studies, which often overestimate diagnostic accuracy.12,13 Since the earlier meta-analysis,11 many more relevant studies have been published.14-17
We conducted a systematic review and meta-analysis of the diagnostic accuracy of symptoms, physical signs, and laboratory tests for GCA. We provide summary estimates of the sensitivity, specificity, and LRs of these features. We included studies using appropriate reference standards for GCA, including TAB and clinical diagnosis. We excluded case-control studies and studies in which all patients were classified as having GCA.
This study is reported in accordance with the 2009 Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.18 A predefined study protocol was established but not registered. No ethical approval or informed consent was required for the current systematic review and meta-analysis.
We searched PubMed, EMBASE, and the Cochrane Database of Systematic Reviews from December 1940 to April 5, 2020. The search strategy included terms such as giant cell arteritis, temporal arteritis, medical history taking, physical examination, diagnostic imaging, and artery biopsy. The full search strategy was developed together with an experienced medical science librarian (eTable 1 in the Supplement). We included English language records. Case reports and conference abstracts were excluded. The reference lists of included studies were screened for additional records.
We included clinical trials and prospective or retrospective observational studies that met the following criteria: (1) participants were consecutive patients suspected of having GCA; (2) a TAB, imaging test, or clinical diagnosis was used as the reference standard for GCA; (3) a table of the true-positive, false-positive, true-negative, and false-negative counts was either directly available or could be calculated for at least 1 index test (symptom, physical sign, or laboratory test); and (4) at least 5 patients had GCA and at least 5 did not have GCA. The reference standard clinical diagnosis could be based on defined criteria or judgment of 1 or more physicians. We considered healed temporal arteritis (ie, intimal hyperplasia and/or internal elastic lamina disruption in the absence of an arterial inflammatory infiltrate) as a negative TAB result, because it might indicate atherosclerosis rather than GCA.6 We excluded studies in which all patients were diagnosed with GCA and/or the closely related disease polymyalgia rheumatica.19 We excluded case-control studies. Titles and abstracts were screened by 2 independent reviewers (K.S.M.vdG. and M.S.). Full texts were independently assessed in Covidence by 2 reviewers (K.S.M.vdG. and M.S. or S.L.M.). Disagreement between reviewers was resolved by consensus or, if consensus could not be obtained, by consulting a third reviewer (E.B.) who made the final decision.
Study characteristics and data from 2 × 2 tables were extracted by 1 reviewer (K.S.M.vdG.) and checked by a second reviewer (E.B. or S.L.M.). A standardized data sheet was used to collect information on study characteristics (eAppendix in the Supplement). We extracted any clinical or laboratory finding reported, as well as data on age and sex. Composite findings (eg, symptom A plus symptom B) were not recorded. Authors of studies were not contacted. If potential data overlap existed among studies from the same hospital, data were obtained from the largest study. When multiple reference standards were available in 1 study, the clinical diagnosis was used as the reference standard for the main study analysis. A C-reactive protein (CRP) level of less than 0.5 mg/dL (to convert to mg/L, multiply by 10) was considered the reference value unless other laboratory-specific reference values were reported. Disagreement between reviewers was either resolved by consensus or, if consensus could not be obtained, by consulting a third reviewer (E.B. or S.L.M.), who made the final decision.
The risk of bias was evaluated by 2 reviewers (K.S.M.vdG. and E.B.) with the quality assessment of diagnostic accuracy studies (QUADAS-2) tool (eAppendix in the Supplement). The QUADAS-2 tool focuses on the bias and applicability of study results regarding patient selection, the index test, the reference standard, and study flow and timing.20
Study heterogeneity was evaluated by plotting the estimates of sensitivity and specificity in forest plots and receiver operating characteristics (ROC) space. We used hierarchical logistic regression modeling to determine summary estimates of the sensitivity, specificity, diagnostic odds ratio, and LRs by the bivariate model approach, as well as hierarchical summary ROC (HSROC) plots.21 Likelihood ratios of greater than 2.00 or less than 0.50 with 95% CIs not including 1.00 were considered statistically significant.10,22 We performed the following sensitivity analyses: (1) a predefined comparison of LRs in studies using distinct reference standards for GCA; (2) a nonpredefined comparison of LRs in prospective and retrospective studies; and (3) a predefined analysis restricted to pretreatment laboratory tests. Our primary analysis and sensitivity analyses included any index test reported by 4 or more studies. Hierarchical logistic regression modeling analysis and evaluation of funnel plot asymmetry were performed in STATA, version 15.1 (StataCorp LLC) with the metandi, metandiplot, and midas commands.23 Forest plots were created in Review Manager, version 5.3 (Cochrane) and StatsDirect, version 3.2.10 (StatsDirect Ltd).
Of the 1436 reports screened, 68 studies14-17,24-87 fulfilled the selection criteria and were used for the systematic review and meta-analysis (eFigure 1 in the Supplement). These studies included 14 037 patients, of whom 4277 (30.5%) were classified as having GCA (Table 1 and eTable 2 in the Supplement). Most reports were retrospective cohort studies (48 [70.6%])14,15,27,29,31,32,34-41,43-46,48-51,53,54,58,59,62,64-68,70-75,77-81,83-87 and performed at academic centers (56 [82.4%]).16,17,24-35,37,38,40-57,59-62,66-69,72,73,75-84,86,87 TAB was the reference standard in 38 studies (55.9%).14-16,26,30,32,37-45,48,49,52,56,58-60,62,64,65,67,70,72,73,78-81,83-87 The mean or the median length of the TAB specimen was generally greater than 1 cm. A variable proportion of patients underwent bilateral TAB (eTable 3 in the Supplement). In 30 studies (44.1%),17,24,25,27-29,31,33-36,46,47,50,51,53-55,57,61,63,66,68,69,71,74-77,82 clinical diagnosis was the reference standard for GCA; in 8 of these studies,31,46,53,71,75-77,82 all patients underwent TAB, and in 9 studies,17,33,47,51,61,63,68,69,74 patients had a combination of TAB and imaging (eTable 4 in the Supplement). The clinical diagnosis was typically based on clinical and laboratory findings, imaging and/or TAB results, and a good initial response to glucocorticoid treatment (eTable 5 in the Supplement). In 16 of the studies using clinical diagnosis as reference standard,17,29,31,33,34,36,47,51,53,54,57,61,68,69,76,82 patients were all followed up to verify that the clinical diagnosis was not later revised. Only 1 study68 allowed us to evaluate imaging as the reference standard in addition to the clinical diagnosis and TAB.
Patient selection was the principal source of bias (eFigures 2 and 3 in the Supplement). Studies using TAB as the reference standard may have been more prone to selection bias because a sufficient index of clinical suspicion is required to order this invasive test. Conversely, studies using the clinical diagnosis as the reference standard were at high risk of bias because the index test result contributed to the clinical diagnostic decision.
In studies reporting the sex of patients (n = 7798), 2605 (33.4%) of patients were male and 5193 (66.6%) were female. Although headache is considered to be a key symptom for GCA, the positive and negative LRs for headache did not meet our prespecified threshold for statistical significance (Table 2). Double vision provided a positive LR of 1.72 (95% CI, 1.12-2.63). Positive LRs of more than 2.00 were found for limb claudication (6.01; 95% CI, 1.38-26.16), jaw claudication (4.90; 95% CI, 3.74-6.41), and a previous diagnosis of polymyalgia rheumatica (2.07; 95% CI, 0.92-4.65), whereas being older than 70 years had a negative LR of less than 0.50 (0.48; 95% CI, 0.27-0.86). The forest plots and HSROC curves indicated substantial heterogeneity for all statistically significant index tests except for jaw claudication (eFigures 4 and 5 in the Supplement). Overall, we found little evidence of publication bias by evaluation of funnel plot asymmetry (eFigure 6 in the Supplement). Symptoms reported by less than 4 studies are shown in eTable 6 in the Supplement.
A positive LR of more than 2.00 occurred for findings related to temporal artery thickening (LR, 4.70; 95% CI, 2.65-8.33), temporal artery loss of pulse (3.25; 95% CI, 2.49-4.23), temporal tenderness (3.14; 95% CI, 1.14-8.65), an abnormal temporal artery (2.29; 95% CI, 1.61-3.26), anterior ischemic optic neuropathy (2.15; 95% CI, 1.53-3.03), erythrocyte sedimentation rate (ESR) of greater than 60 (2.40; 95% CI, 1.71-3.36), 80 (2.79; 95% CI, 1.78-4.37), and 100 mm/h (3.11; 95% CI, 1.43-6.78), and a platelet count of greater than 400 × 103/μL all (to convert to ×109/L, multiply by 1) (3.75; 95% CI, 2.12-6.64) (Table 3). Negative LRs of less than 0.50 occurred for an ESR of more than 40 mm/h (0.18; 95% CI, 0.08-0.44), more than 50 mm/h (0.48; 95% CI, 0.38-0.62), and more than 60 mm/h (0.42; 95% CI, 0.28-0.61), CRP level of at least 2.5 mg/dL (0.38; 95% CI, 0.25-0.59), or a CRP level of greater than the reference value (0.40; 95% CI, 0.29-0.56). Overall, moderate heterogeneity and little funnel plot asymmetry was observed (eFigures 4, 5, and 6 in the Supplement). Physical findings reported by fewer than 4 studies are shown in eTable 7 in the Supplement.
Results of our sensitivity analyses are provided in eTables 8 to 10 in the Supplement. We found comparable LRs in our comparison of studies with different reference standards (TAB vs clinical diagnosis) or study design (prospective vs retrospective). A pretreatment elevated CRP level showed a sensitivity of 90.1% (95% CI, 76.3%-96.3%) and a negative LR of 0.38 (95% CI, 0.17-0.81) for a diagnosis of GCA. A pretreatment ESR of greater than 50 mm/h had a sensitivity of 87.5% (95% CI, 78.3%-93.1%) and negative LR of 0.27 (95% CI, 0.13-0.57).
This updated meta-analysis provides more precise estimates of LRs associated with symptoms, signs, and laboratory features of GCA. Features that, if present, should upgrade the level of suspicion for GCA are limb claudication; jaw claudication; various temporal artery abnormalities; a platelet count of greater than 400 × 103/μL; ESRs of greater than 60, 80, and 100 mm/h; and anterior ischemic optic neuropathy. Features that should downgrade the level of suspicion for GCA are 70 years or younger; a CRP level in the reference range or less than 2.5 mg/dL; and an ESR of no greater than 40, 50, or 60 mm/h. For most patients with suspected GCA, no single feature is likely to shift pretest probability sufficiently to render further investigation for GCA unnecessary. However, these likelihood ratios may inform clinical decisions, including selection and timing of investigations, and whether to immediately commence high-dose glucocorticoid therapy or await further test results.88,89
Our findings confirm and extend those of the previous meta-analysis,11 which had included 21 studies of 2680 patients. We were able to show that an elevated ESR, especially greater than 60 mm/h, is informative in suggesting a diagnosis of GCA. We improved the precision and clinical utility of the summary estimates. For example, the previous meta-analysis11 estimated the positive LR for double vision as 3.4 (95% CI, 1.3-8.6); with greater patient numbers, we estimate the positive LR as 1.72 (95% CI, 1.12-2.63). We were also able to evaluate the diagnostic accuracy of further features, including transient loss of vision, cerebrovascular accident, limb claudication, central retinal artery occlusion, CRP levels, and platelet counts. Furthermore, we conducted sensitivity analyses to evaluate for bias arising from choice of reference standard, prospective vs retrospective studies, and whether all laboratory tests were explicitly stated as occurring before treatment.
Various tools have been developed that could help to estimate GCA probability. These tools require assessment of a limited set of clinical and laboratory features that were originally selected by expert opinion and then weighted based on expert opinion or statistical methods.8,9 Interestingly, both tools contain features, such as sex, that were not very helpful in changing GCA probability according to our meta-analysis. Some clinical features in these tools, such as symptom duration and alternative diagnosis,8 could not be included in our meta-analysis owing to lack of published data.
Our meta-analysis indicates that some features considered classic for GCA, such as headache, scalp tenderness, and constitutional symptoms, have limited use for upgrading or downgrading the clinical probability of GCA. This does not mean, however, that these symptoms are irrelevant. Our meta-analysis shows that the prevalence of these classic features is high among patients with and without GCA, suggesting that the diagnostic value of these symptoms may have been used up earlier in the care pathway.90 Headache is important in prompting suspicion of GCA and onward referral to a specialist, but once that referral decision has been made, clinicians should be cautious about overvaluing the diagnostic significance of headache and should evaluate patients for the other features identified in our meta-analysis as informative for a final diagnosis of GCA.
Our study was limited by the quality of the studies included. Although we performed a comprehensive search for published studies, we cannot exclude that relevant data was omitted owing to exclusion of non-English articles and conference abstracts. No unpublished data were obtained via contact with authors.
Several sources of bias were present in our meta-analysis. First, studies using TAB may have been at risk of selection bias because the decision for TAB necessarily depends on the presence of clinical and laboratory features to justify this invasive test. Second, clinical diagnosis is subjective and relies on clinical and laboratory features as well as further tests; this circularity could lead to overestimation of the diagnostic accuracy of index tests. Third, many studies were retrospective cohort studies, which could have introduced further selection bias. We mitigated these risks of bias by performing sensitivity analyses, which did not show substantial differences between studies with distinct reference standards or between studies with retrospectively and prospectively gathered data.
The reference standards for GCA may have additional limitations. Although the sensitivity of TAB may be 77% for fulfilment of the American College of Rheumatology 1990 criteria for GCA,91 it is likely lower for the clinical diagnosis in daily clinical practice.6 Some studies in our meta-analysis46,68,71,75-77,82 reported a subgroup of patients with TAB findings that were negative for GCA. Patients with GCA may have had TAB findings negative for GCA in other studies, but these patients were simply classified as not having GCA. Thus, the diagnostic accuracy of clinical and laboratory features might have been underestimated. The clinical diagnosis of GCA might be subjective and strongly related to the experience of the individual physician making the diagnosis. The clinical diagnosis was only ascertained by follow-up in a minority of studies. Nevertheless, we observed comparable LRs of clinical and laboratory features in studies using TAB or the clinical diagnosis as the reference standard.
A clear definition of symptoms was lacking in the studies included in our meta-analysis. This might be relevant for a symptom such as jaw claudication. Jaw claudication typically occurs after 2 to 3 minutes of chewing,92 but temporomandibular joint pain is common in older people and also causes pain with chewing. Lack of a clear definition of jaw claudication might possibly inflate the LR of this clinical feature, because it allows clinicians to classify aching on chewing as either jaw claudication or temporomandibular joint pain based on the clinical judgment that GCA is likely or not. Because jaw claudication is not described in any other disease and might be considered almost pathognomonic of GCA, clinicians may be reluctant to document jaw claudication unless they are fairly sure for other reasons that the patient has GCA.
Glucocorticoid treatment may be commenced immediately when GCA is suspected. This treatment could have affected index test results, particularly the laboratory tests. It was surprising that only few reports explicitly stated that the laboratory test results were obtained before treatment. Our sensitivity analysis for pretreatment laboratory measures could only be performed for an elevated CRP level and an ESR of greater than 50 mm/h. These pretreatment laboratory features tended to show better sensitivity and negative LRs than those obtained in the main study analysis.
The meta-analysis method we used required us to dichotomize continuous variables associated with GCA (age and laboratory values), which is inefficient and likely results in underestimation of diagnostic utility. However, individual patient data meta-analysis would have been needed to overcome this.
Study heterogeneity was observed for various clinical and laboratory features with relevant LRs. Additional prospective studies are needed to confirm the summary estimates of these features. We therefore recommend that complete sets of clinical and pretreatment laboratory data are reported in diagnostic cohort studies, either in summary tables or as raw data. This process would allow investigators to determine summary estimates of diagnostic accuracy parameters with more precision. Prospective studies would ideally consist of all patients who have been evaluated for GCA by every specialty or department in a hospital.90
This systematic review and meta-analysis highlight the clinical and laboratory features that may be informative in making a diagnosis of GCA and that should be assessed when evaluating patients with suspected GCA. They should also be reported in future diagnostic cohort studies. Clinicians should obtain a comprehensive history, physical examination, and laboratory evaluation for each patient suspected of having GCA. No single symptom, physical sign, or laboratory test is sufficient to completely rule in or rule out GCA. An additional imaging test or TAB is typically needed to make a confident diagnosis of GCA. Our study could not determine whether individual LRs can be combined, or whether there is collinearity between particular features (eg, ESRs and CRP levels with constitutional symptoms). Nonetheless, this study provides important data that could inform a future bayesian probability revision approach to investigation, diagnosis, and management of suspected GCA, which would need to be prospectively validated in future studies.
Accepted for Publication: May 25, 2020.
Published Online: August 17, 2020. doi:10.1001/jamainternmed.2020.3050
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 van der Geest KSM et al. JAMA Internal Medicine.
Corresponding Author: Kornelis S. M. van der Geest, MD, PhD, Department of Rheumatology and Clinical Immunology, University of Groningen, University Medical Center Groningen, Hanzeplein 1, Groningen 9700RB, the Netherlands (firstname.lastname@example.org).
Author Contributions: Drs Brouwer and Mackie contributed equally as co–last authors. Dr van der Geest had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: van der Geest, Brouwer, Mackie.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: van der Geest, Brouwer, Mackie.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: van der Geest, Brouwer.
Administrative, technical, or material support: van der Geest, Brouwer.
Conflict of Interest Disclosures: Dr van der Geest reported receiving a speaker fee from Roche paid to the University Medical Center Groningen. Dr Brouwer reported receiving consultancy and speaker fees from Roche paid to the University Medical Center Groningen. Dr Mackie reported receiving support from Roche for attendance of the 2019 European League Against Rheumatism meeting as a coapplicant on a research grant; receiving consultancy fees from Roche and Sanofi SA on behalf of Leeds Institute of Rheumatic and Musculoskeletal Medicine; and serving as a trial investigator for GlaxoSmithKline and Sanofi SA. No other disclosures were reported.
Funding/Support: This study was supported by TARGET partnership grant MR/N011775/1 from the Medical Research Council (Dr Mackie) and the Mandema Stipend from the University Medical Center Groningen (Dr van der Geest).
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Karin Sijtsma, medical science librarian at the Central Medical Library of the University Medical Center Groningen, for her advice on the search strategy. She received no compensation for this work.
Create a personal account or sign in to: