Giant Cell Arteritis: Diagnostic Accuracy of MR Imaging of Superficial Cranial Arteries in Initial Diagnosis—Results from a Multicenter Trial

Introduction

Giant cell arteritis (GCA) is the most common chronic vasculitis of medium- and large-sized arteries in populations with predominantly Northern European ancestry (1,2). The characteristic histopathologic feature of GCA displays a granulomatous inflammation of the vessel wall with multinucleated giant cells (3). Predilection sites of vascular inflammation are the superficial cranial arteries, such as the superficial temporal artery with its branches, and the occipital artery (4). A segmental involvement pattern is typical (5). Classic clinical signs of GCA include new onset of or new type of headache in patients older than 50 years presenting with tenderness of the temporal arteries or decreased pulsation. However, clinical symptoms and their severity may vary depending on the localization of the involved arterial segments (3).

Establishing the diagnosis of GCA may be challenging. The American College of Rheumatology has defined five criteria, which are traditionally in use to date, for classification of GCA (6). Among these criteria, the histopathologic result of a temporal artery biopsy (TAB) is considered the diagnostic standard. Diagnostic certainty is essential in GCA, because systemic corticosteroid (sCS) therapy, with its underlying therapeutic risks, is initiated early for preventing complications (7,8).

Contrast material–enhanced, high-spatial-resolution magnetic resonance (MR) imaging is considered a promising noninvasive diagnostic modality for imaging of vasculitis. In patients with GCA, initial studies have shown that MR imaging reveals inflammatory changes of small superficial cranial arteries (9,10).

The purpose of this prospective multicenter trial was to assess the diagnostic accuracy of contrast-enhanced MR imaging of superficial cranial arteries in the initial diagnosis of GCA.

Materials and Methods

This study was supported by the Deutsche Forschungsgemeinschaft (DFG; BL 1132/1–1). Bracco Imaging (Konstanz, Germany) provided financial support for annual investigator meetings, and Medac (Wedel, Germany) provided financial support for the initiation and maintenance of the multicenter trial online database. The authors had full control of the data and the information submitted for publication. Good clinical practice and the Declaration of Helsinki were strictly followed.

Study Population and Design

This prospective, single-arm, multicenter trial has been registered with the World Health Organization and the German Clinical Trials Register (trial identification no. DRKS00000594). The study was designed and conducted according to Standards for Reporting of Diagnostic Accuracy (11). Following institutional review board approval and after obtaining written informed consent, 185 consecutive patients from three participating university medical centers were included in this study. The Standards for Reporting of Diagnostic Accuracy flowchart is displayed in Figure 1. Patient demographics are presented in Table 1. Neurologists, rheumatologists, and ophthalmologists immediately referred patients, who presented with their first episode of potentially GCA-suspicious symptoms, for MR imaging of the superficial cranial arteries between July 2007 and August 2011 (49 months).

Figure 1:

Download as PowerPointOpen in Image Viewer

Table 1 Patient Demographics

*Numbers in parentheses are percentages. Percentages were rounded.

Inclusion criteria.—The inclusion criteria were as follows: (a) Patients could be either female or male patients; (b) they had to be 50 years old or older; (c) they had to have one or more of the following clinical symptoms and findings, which included new-onset, localized headache, temporal artery tenderness or decreased temporal artery pulse, and erythrocyte sedimentation rate higher than 50 mm/h; (d) they had to have undergone MR imaging of the superficial cranial arteries, indicated for diagnosing or excluding GCA; (e) they should not have undergone sCS therapy or, if they were undergoing sCS therapy, it had to have been initiated 14 days or fewer prior to MR imaging; and (f) they had to have received a definitive diagnosis of GCA or GCA had to have been excluded on the basis of the diagnostic reference criteria described below to be included in this study.

Exclusion criteria.— Patients were excluded as follows: (a) Patients with typical MR contraindications such as implanted pacemakers, metallic foreign bodies, claustrophobia, contrast agent allergy, or renal dysfunction with a glomerular filtration rate lower than 30 mL/h; (b) patients with known GCA; and (c) patients with incomplete or interrupted follow-up after MR imaging were not included in this study.

Diagnostic References

The MR imaging results of this study were analyzed in comparison with two diagnostic references: First, they were analyzed in comparison with our diagnostic reference standard, which included the total study cohort of 185 patients, and second, they were analyzed in comparison with the subcohort of 98 patients, who had undergone the diagnostic standard TAB. Both diagnostic references are described in detail below.

Total study cohort with a definitive and final diagnosis (reference standard).—In all 185 patients, experienced rheumatologists, neurologists, and ophthalmologists established the final diagnosis in an interdisciplinary consensus (total study cohort; patients with GCA-positive results, patients with GCA-negative results) after MR imaging and after all available diagnostic criteria had been worked up. The diagnostic criteria used to establish the diagnosis included the following: (a) histopathologic results of TAB in 98 patients (53.0%), obtained after MR imaging (see “Subcohort of patients with TAB histopathologic results [diagnostic standard]” below); (b) rheumatologic assessment of clinical symptoms according to classification criteria of the American College of Rheumatology; (c) response to sCS therapy; and (d) disease course during a follow-up period of at least 6 months after the first consultation. This clinical reference standard did not include the MR imaging evaluation as diagnostic criterion.

Subcohort of patients with TAB histopathologic results (diagnostic standard).—TAB was considered the diagnostic standard. TAB was performed in 98 of 185 included patients (53.0%) subsequently after the MR study (mean, 2.03 days ± 2.59 [standard deviation]; range, 0–14 days). TAB was performed according to standard surgical techniques: The course of the superficial temporal artery was palpated or visualized by using Doppler ultrasonography (US) distal to the bifurcation and marked for the surgical incision. The vessel was surgically exposed over a length of 3–5 cm of the artery and was segmentally resected. Previous MR images covered the total vessel course to ensure that the TAB specimen had been visualized on MR images, and later, after the MR imaging study, the TAB specimen was resected for evaluation. Because TAB was not performed in all patients, data of this subcohort are presented separately (TAB subcohort). The sole diagnostic criterion was the histopathologic result (TAB positive, TAB negative).

The total study cohort included eight patients with TAB performed prior to MR imaging (mean, 428.9 days ± 394.7; range, 20–1042 days) who were not included in the TAB subcohort, but were in the total study cohort.

MR Image Acquisition

MR imaging was performed in 55 patients with a 1.5-T system (Magnetom Sonata and Avanto, Siemens Medical Solutions, Erlangen, Germany; Achieva, Philips Medical Systems, Cleveland, Ohio) and in 130 patients with a 3.0-T system (Magnetom Trio; Siemens Medical Solutions), with dedicated 12- and 32-channel (for 3.0-T MR) or 16-channel (for 1.5-T MR) phased-array coils. MR imaging sequence parameters were standardized for all hospitals according to results of previous investigations (9,10). Ten minutes after intravenous injection of 0.1 mmol per kilogram of body weight of gadolinium-based contrast agent (MultiHance, Bracco Imaging; Gadovist, Bayer Healthcare, Leverkusen, Germany), fat-saturated T1-weighted spin-echo sequences with a spatial resolution of 195 × 260 µm were performed. Acquisition parameters were adapted to the main magnetic field strength. For the 1.5-T systems, the following parameters resulted in an imaging time of 6 minutes 55 seconds (10 sections): repetition time msec/echo time msec, 535/22; bandwidth, 65 Hz/pixel; field of view, 200 × 200 mm²; matrix size, 1024 × 768; and number of signals acquired, one. At 3 T, repetition time was reduced to 500 msec, and the bandwidth was increased to 76 Hz/pixel; half Fourier encoding (factor, six of eight) was used, resulting in an imaging time of 4 minutes 52 seconds. Two consecutive sets of images were acquired with 10 gapless sections of 3-mm thickness. The two oblique axial acquisitions were caudally aligned and oriented along the skull base on sagittal localizer images covering 60 mm from the skull base in the cranial direction.

MR Image Evaluation

Two board-certified radiologists (J.G. and T.K.) from two different and locally distant university hospitals with 7 and 6 years of experience independently evaluated anonymized MR images. Both readers were blinded to all clinical data, including TAB results. Evaluation of the superficial cranial arteries included six arterial segments: the frontal and parietal branches of the superficial temporal artery and the occipital artery bilaterally (Fig 2a). The evaluation criteria included mural thickening and mural contrast enhancement. Arterial segments were evaluated in cross-sectional views to prevent false-positive results due to partial volume effects, which may occur in oblique vessel sections. Branches of the superficial temporal artery were evaluated from the bifurcation over a segment length of at least 3 cm. The occipital artery was evaluated along its course over at least 3 cm. The vessel course was followed by scrolling through the total stack of 20 × 3-mm oblique axial images.

Figure 2a:

Download as PowerPointOpen in Image Viewer

Figure 2b:

Download as PowerPointOpen in Image Viewer

MR image scoring.—Each arterial segment was assigned a score according to a previously published four-point scale (10): score 0, no mural thickening (vessel wall diameter, < 0.6 mm) and no enhancement; score 1, no thickening but slight mural enhancement; score 2, mural thickening (> 0.6 mm) and significant mural enhancement; and score 3, strong thickening (> 0.7 mm) and strong mural and perivascular enhancement (Fig 2b). Scores 2 and 3 were considered to represent active mural inflammation. Scores 0 and 1 were considered to represent physiologic features.

MR imaging criteria for diagnosing GCA.—The results of MR imaging studies of each patient were considered positive (MR imaging–positive result) or negative (MR imaging–negative result) for GCA by each observer. For an MR imaging–positive result, at least one of six arterial segments had to be assigned a score of 2 or 3. Thus, the maximum MR imaging score among six segments was used to determine the MR imaging test result. MR imaging test results were compared with the diagnostic reference result of the total study cohort and the TAB subcohort.

MR imaging scores for evaluating atypical findings and the effects of sCS therapy.—MR imaging scores were also analyzed in comparison with the diagnostic reference results. Hereby, the mean MR imaging score of all six arterial segments per patient, and not the maximum score, was used as a potential measurement of the degree of vascular inflammation. The mean scores were used to identify potential outliers with atypical test results and to calculate receiver operating characteristic (ROC) curves as a nonbinary variable. MR imaging scores were correlated with the results of the total study cohort and of the TAB subcohort.

Statistical Analysis

Statistical analysis was performed by using software (IBM SPSS Statistics, version 21; IBM, Armonk, NY) and a software package (Prism, version 6.0c; GraphPad Software, La Jolla, Calif). Sensitivity, specificity, positive predictive value, and negative predictive value of MR imaging with 97.5% confidence intervals (CIs) were calculated for each observer in comparison with the total study cohort and with the TAB subcohort (12). The interobserver agreement was calculated by using the Cohen κ test, considering a κ value of greater than 0.80 as outstanding agreement and a κ value of greater than 0.60 as good agreement (13). Continuous data are presented as means with standard deviations. Significant differences were calculated by using the Student t test. A difference with a P value of .05 or less was considered significant. ROC curves were calculated and areas under the ROC curves (AUCs) were compared according to DeLong et al (14).

Results

MR imaging examinations were performed without any complications in all patients. The postoperative course of participants after TAB was not evaluated. The superficial cranial arteries could be evaluated in each patient.

In the total study cohort (n = 185), sensitivity was 78.4% (97.5% CI: 69.2, 86.0) and specificity was 90.4% (97.5% CI: 81.9, 95.8) for observer 1, and sensitivity was 83.3% (97.5% CI: 74.7, 90.0) and specificity was 85.5% (97.5% CI: 76.1, 92.3) for observer 2. In the TAB subcohort (n = 98), sensitivity was 88.7% (97.5% CI: 78.1, 95.3) and specificity was 75.0% (97.5% CI: 57.8, 87.9) for observer 1, and sensitivity was 93.6% (97.5% CI: 84.3, 98.2) and specificity was 75.0% (97.5% CI: 57.8, 87.9) for observer 2.

Table 2 shows detailed results for each reader. Interobserver agreement was good (total study cohort, κ = 0.676; TAB subcohort, κ = 0.718).

Table 2 Accuracy of MR in Initial Diagnosis of GCA in Comparison with the Diagnostic Standard (TAB) and the Reference Standard (Final Clinical Diagnosis)

Note.—FN = false-negative, FP = false-positive, LR = likelihood ratio, NPV = negative predictive value, PPV = positive predictive value, TN = true-negative, TP = true-positive.

*Numbers in parentheses are 97.5% CIs.

Six patients had a GCA-positive result despite a negative TAB result. In four of these patients, MR imaging results were rated positive by at least one observer. In the other two patients, MR imaging results were negative. Of note, sCS was administered for more than 5 days in both patients.

MR Image Scoring

In the total study cohort (Fig 3), the mean MR imaging score was 1.80 ± 0.78 for patients with a GCA-positive result and 0.58 ± 0.43 for patients with a GCA-negative result. In the TAB group (Fig 3), the mean MR imaging score was 1.99 ± 0.67 for patients with a TAB-positive result and 0.79 ± 0.50 for patients with a TAB-negative result. In both groups, patients with a positive diagnostic result had significantly higher MR imaging scores (both groups, P < .001) than patients with a negative result.

Figure 3:

Download as PowerPointOpen in Image Viewer

Superficial Cranial Artery Involvement Pattern

The TAB result was positive in 62 (63.3%) of 98 patients. The arterial involvement pattern of patients with a TAB-positive result for observer 1 is presented in Table 3. Results for observer 2 were comparable with a left-right ratio of 1.03:1. Table 3 shows that involvement of segments was symmetrical in most cases. The frontal branches were marginally more frequently involved than the occipital. In the majority of patients, multiple arteries were simultaneously affected.

Table 3 Segmental Involvement Pattern Detected on MR Images, Including the Left- and Right-sided Frontal and Parietal Branch of the Superficial Temporal Artery and the Occipital Artery, in 62 Patients with a Positive TAB Histopathologic Result

Note.—The results for observer 1 only are included in this table. There were 149 involved left-sided segments and 132 involved right-sided segments, with a 1.03:1 left-right ratio.

Effects of sCS Therapy

In 135 of 185 patients (73.0%) from the total study cohort and in 79 of 98 patients (80.6%) from the TAB subcohort, the exact duration of sCS therapy prior to the MR imaging examination was documented. Thirty-three patients did not receive sCS therapy. In 78 patients of the total study cohort, sCS therapy had been initiated 5 days or fewer prior to the MR imaging examination, and in 24 patients, sCS therapy had been initiated 6–14 days prior to the MR imaging examination. Fifty patients fulfilled the study inclusion criteria but were excluded from this analysis because of incomplete information about sCS therapy.

In the total study cohort, MR imaging sensitivity was 71.4% and specificity was 89.5% for observer 1, and MR imaging sensitivity was 78.6% and specificity was 94.7% for observer 2 in 33 patients without sCS therapy. In 78 patients with sCS therapy for 5 days or fewer, sensitivity was 81.3% and specificity was 84.0% for observer 1, and sensitivity was 85.4% and specificity was 80.0% for observer 2. In 24 patients with sCS therapy for 6–14 days, sensitivity was 72.7% and specificity was 84.6% for both observers.

In the TAB subcohort, MR imaging sensitivity was 89.4% and specificity was 76.2% for observer 1, and MR imaging sensitivity was 95.7% and specificity was 81.0% for observer 2 in 12 patients without sCS therapy. In 56 patients with sCS therapy for 5 days or fewer, sensitivity was 90.0% and specificity was 75.0% for observer 1, and sensitivity was 95.0% and specificity was 75.0% for observer 2. In 11 patients with sCS therapy for 6–14 days, sensitivity was 100.0% and specificity was 87.5% for both observers.

ROC curve analysis for observers 1 and 2 revealed comparable results. Figure 4 shows ROC curves and AUC results for observer 1. ROC curves for observer 2, which are not presented in Figure 4, were comparable, and AUC results were as follows: In the total study cohort, AUC was 0.914 for patients without sCS therapy, 0.898 for patients with sCS therapy for 5 days or fewer, and 0.724 for patients with 6–14 days sCS therapy. AUC was 0.943 for patients without sCS therapy, 0.902 for patients with sCS therapy for 5 days or fewer, and 1.000 for patients with 6–14 days of sCS therapy.

Figure 4:

Download as PowerPointOpen in Image Viewer

In the total study cohort, the ROC curves showed a reduced AUC for patients with sCS therapy for more than 5 days in comparison with the no-sCS group without reaching significance level (P = .08 for observer 1 and P = .06 for observer 2) (Fig 4). In the TAB subcohort, AUCs from ROC curves in patients without sCS therapy, with sCS therapy for 5 days or fewer, or for 6–14 days prior to MR imaging were apparently not significantly different (Fig 4). The ROC curves between the total study cohort and the TAB subcohort without sCS therapy and with sCS therapy for 5 days or fewer also were not significantly different.

Discussion

This prospective multicenter trial provides evidence that contrast-enhanced MR imaging of the superficial cranial arteries has good to excellent diagnostic accuracy in the initial diagnosis of GCA. High diagnostic accuracy is very important for patients who are suspected of having GCA. On the one hand, diagnostic misinterpretation because of false-negative test results or diagnostic delays can lead to aggravation of symptoms in patients with GCA and possibly to severe complications, particularly caused by ischemia. On the other hand, erroneous initiation of sCS therapy because of false-positive test results may expose patients to the whole spectrum of possible side effects of sCS therapy.

Typical MR imaging signs of vascular inflammation in GCA included arterial wall thickening with mural and periadventitial contrast enhancement. Wall thickening and contrast enhancement were assessed by using the previously published four-point MR vasculitis score, which proved to be a reliable semiquantitative measurement tool and which was significantly higher in patients with active GCA than in patients without GCA (10). In patients with GCA, a symmetrical and simultaneous inflammation of all six evaluated superficial cranial artery segments has been observed and is considered a typical involvement pattern (15). In this context, the involvement pattern has to be considered nontypical in at least every fifth patient who, for example, showed signs of inflammation in only one to three segments. The establishment of the diagnosis GCA may be challenging in this group of patients, when the traditional diagnostic standard TAB fails. We assume that MR imaging has the potential to close this gap. MR imaging may also compensate for false-negative histopathologic results due to skip lesions. This other occasional drawback of TAB occurred in two patients with TAB of our cohort who had high MR imaging scores. Histopathologic signs of vasculitis may be missed in TAB specimens because of skip lesions or the involvement of other arterial segments than those sampled (5). These limitations account for the typical disease pattern with segmental involvement and skip lesions, as well as for variable surgical and analytical TAB methods, including different specimen lengths and histopathologic interpretation skills. The length of the biopsy specimen positively correlates with the diagnostic sensitivity of TAB (16). In fact, the true sensitivity of single-sided TAB has been proposed to range between 69% and 98.7% (17). Finally, the clinical assessment including American College of Rheumatology criteria is discussed to be more reliable, allowing a positive diagnostic result despite a negative TAB result. While TAB results have been shown to be negative in about 7.1% of patients with GCA, the clinical evaluation based on American College of Rheumatology criteria may be more sensitive (93.5%) and specific (91.2%) than TAB results alone (18).

The diagnostic accuracy of MR imaging in our study is almost equivalent to previous observations of a consensus reading in a single-center study (n = 32) (9). Previous studies have shown further that MR signs of vasculitis vanish during sCS therapy (19). This accounts for color-coded duplex US, as well (20–22). A recent retrospective study revealed that the sensitivity of MR imaging and US significantly decreases after more than 4 days of sCS therapy (20). From the clinical point of view, early initiation of sCS therapy must not be delayed (8). In our study, MR imaging scores of the total study cohort slowly decreased, when sCS therapy was administered more than 5 days prior to MR imaging. The apparently paradoxical increase in the TAB subgroup is probably caused by the small number of patients with either no sCS therapy or 6–14 days of sCS therapy. The divergence of clinical and pathologic findings during the response to sCS treatment with a quick alleviation of symptoms and persisting histopathologic findings may explain our observations (23).

This study had the following limitations: The diagnostic standard TAB was available in 98 of 185 patients. Patient selection may have been biased because of the expert level of referring physicians in treating vasculitis. This bias may have caused a higher pretest probability for GCA, as reflected in a higher number of patients with MR imaging–positive results in the TAB subcohort. Concomitantly, the higher pretest probability explains why TAB was performed in only 98 patients and why more patients had TAB-positive results. The invasive TAB procedure was particularly requested in patients in whom the clinician had a rather strong suspicion of GCA. The likelihood of GCA was more balanced in the total study cohort of 185 patients. Nevertheless, the total study cohort also included more patients with GCA-positive results (n = 102) than GCA-negative results (n = 83). This led to the assumption that the average disease likelihood in the total study cohort may have been somewhat higher than the expected intermediate level. The final clinical diagnosis may not be accepted as definitive as a positive TAB result, although reassessment during follow-up of at least 6 months was performed. The evaluation of patients with sCS therapy included the duration of therapy prior to MR imaging, but not the cumulative sCS dose. Moreover, GCA is not the only disease that causes arteritis of the superficial cranial arteries. Similarly to GCA, systemic antineutrophilic cytoplasmatic antibody–positive vasculitis, polyarteritis nodosa, and other inflammatory diseases potentially involve the temporal arteries. Although we could show that MR imaging is highly sensitive—maybe more sensitive than TAB—in the detection of small-caliber arteritis, it probably lacks specificity for discrimination of the underlying disease. Finally, the utilization of different MR imaging systems with different magnetic field strengths and coils has caused slight variability in image impression and quality.

The findings of this study may affect future diagnostic workup of patients who are suspected of having GCA. One advantage of MR imaging is its high diagnostic accuracy in the initial diagnosis of GCA. MR imaging is noninvasive and allows for simultaneous and continuous assessment of all superficial cranial artery segments. The simultaneous and continuous assessment of arterial segments is important to prevent false-negative results. Thus, MR imaging may be valuable in patients with atypical involvement pattern. Moreover, MR imaging is repeatable and may therefore be a tool for follow-up of patients and for monitoring therapy response. The good reproducibility of MR imaging may be an advantage over other diagnostic imaging tools.

This prospective, multicenter trial demonstrates that contrast-enhanced MR imaging is accurate and reproducible in the noninvasive, initial diagnosis of GCA. Symmetric and simultaneous inflammation is the typical involvement pattern of superficial cranial artery segments in GCA. Sensitivity of MR imaging probably decreases within days after initiation of sCS therapy.

Author Contributions

Author contributions: Guarantors of integrity of entire study, T.K., T.A.B.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, T.K., J.G., D.D., T.A.B.; clinical studies, all authors; statistical analysis, T.K., J.G., D.D.; and manuscript editing, T.K., J.G., M.B., T.N., S.H., M.R., K.H., T.A.B.