Positron emission tomography assessment of large vessel inflammation in patients with newly diagnosed, biopsy-proven giant cell arteritis: a prospective, case–control study

Background Positron emission tomography (PET) scan is emerging as a promising imaging technique to detect large-vessel inflammation in giant cell arteritis (GCA). However, the lack of a standardised definition of arteritis based on 18fluorodeoxyglucose (FDG) uptake is an important limitation to the use of PET scan for diagnostic purposes. Objective To prospectively assess the intensity and distribution of FDG uptake at different vascular territories in patients with newly diagnosed GCA compared with controls. Methods 32 consecutive, biopsy-proven, GCA patients treated with glucocorticoids for ≤3 days were included. The control group consisted of 20 individuals, who underwent PET/CT for cancer staging. Maximal standardised uptake value (SUVm) was calculated at four aortic segments, supraaortic branches and iliac-femoral territory. Sensitivity and specificity was calculated by receiver–operator characteristic curves (ROC) analysis. Results Mean SUVm was significantly higher in patients than in controls in all vessels explored and correlated with acute-phase reactants and serum IL-6. Mean of the SUVm at all the vascular territories had an area under the curve (AUC) of 0.830, and a cut-off of 1.89 yielded a sensitivity of 80% and a specificity of 79% for GCA diagnosis. There were no significant differences in AUC among the vascular beds examined. Conclusions FDG uptake by large vessels has a substantial sensitivity and specificity for GCA diagnosis.


INTRODUCTION
Temporal artery biopsy is the gold standard for the diagnosis of giant cell arteritis (GCA) due to the tropism of GCA for the epicranial arteries. 1 2 With a few exceptions, 3 histopathological demonstration of temporal artery inflammation provides the most definitive evidence of GCA. Doppler ultrasonography (DUS) of temporal arteries has emerged as a useful alternative tool in centres where biopsy is not easily available. 1 The diagnosis of GCA may be also supported by demonstrating extracranial artery involvement by imaging. Over the past recent years, positron emission tomography/CT (PET/CT), CT angiography, magnetic resonance imaging (MRI) angiography and DUS have revealed that extracranial involvement in GCA is more frequent than previously anticipated, occurring in 30-74% of patients. [4][5][6][7] PET detection of large-vessel involvement in patients with fever of unknown origin, unexplained constitutional symptoms or apparently isolated polymyalgia rheumatica (PMR) has emphasised its diagnostic potential. 8 9 A limitation of PET as a diagnostic tool is the lack of a standardised definition of vascular inflammation based on the intensity of 18 fluorodeoxyglucose (FDG) uptake. While visual assessment of intensively positive cases may be clear, there is no consensus about the minimal intensity of FDG uptake necessary to define vascular inflammation. Conversely, atherosclerosis and ageing may increase vascular FDG uptake, potentially leading to vasculitis overdiagnosis. 10 In this study, we measured FDG uptake by different vascular territories in a cohort of newly diagnosed patients and controls and performed receiver-operator characteristic curves (ROC) analysis to determine sensitivity and specificity of FDG uptake to detect inflammation at different vascular sites. As a secondary endpoint, we analysed potential correlation between FDG uptake and inflammatory biomarkers.

MATERIALS AND METHODS Patients
Between November 2006 and March 2011, all patients diagnosed with biopsy-proven GCA 2 at our institution were assessed for potential participation in the study. Patients who had received glucocorticoid treatment for >3 days were excluded. Clinical and laboratory data recorded are detailed in the online supplementary methods. The study was approved by the ethics committee (Hospital Clínic, Barcelona).
The control group included 20 patients with no chronic inflammatory diseases, matched for gender, age and cardiovascular risk factors (CVRF), consecutively selected among patients who underwent PET/CT during the same timeframe for early lung cancer staging.
PET/CT protocol PET scans were performed using a hybrid PET/CT (Biograph, Siemens) with an ECAT EXACT HR+BGO PET and a helicoidal CT scanner (Somatom, Emotion). Patients fasted 4 h before injection of 370 MBq of 18F-FDG. Whole-body PET data were acquired 60 min after in three-dimensional mode and for 5 min per bed position. PET images were reconstructed both with and without CT data for attenuation correction. A region of interest (ROI) in 3-D around the vessel was placed manually in transaxial, sagittal and coronal slices. The standardised uptake value (SUV) was calculated based on the measured activity, decay-corrected injected dose and patient body weight. SUVm=maximal activity (ROI) (mBq/mL)/injected dose (mBq)/weight (g).
Four aortic segments (ascending thoracic aorta, aortic arch, descending thoracic aorta and abdominal aorta) and the main tributaries-carotid, subclavian, axillary, iliac and femoral arteries (each bilaterally)-were evaluated. The control group was subjected to the same PET/CT protocol. Assessment of PET data was carried out by two nuclear medicine specialists (FL and MD), who were blinded to clinical and pathological findings. However, unequivocal masked evaluation could not be guaranteed due to the controls' disease.

Statistical analysis
ROC were applied to each vascular territory to calculate sensitivity and specificity. Area under the curve (AUC) comparison was performed by Hanley and McNeil analysis. Cut-offs with best sensitivity and specificity were selected. Mann-Whitney U test or Student t test, when applicable, were used for quantitative data. Correlations were calculated using Pearson's or Spearman's test. Statistical significance was defined as p<0.05. Calculations were performed with the IBM SPSS Statistics (V.20.0, Armonk, New York, USA).

Clinical and laboratory findings of the GCA cohort
Seventy-one GCA patients were diagnosed during the recruitment period. Eight patients refused participation, and 31 had received glucocorticoid treatment for >3 days. The remaining 32 were included. Seventeen of them had been treated for ≤3 days at the time of imaging. Treatment consisted of oral prednisone at 1 mg/kg/day. Two patients received 250 mg intravenous methylprednisolone pulses (1 and 7 pulses, respectively) due to severe cranial ischaemic symptoms.
Online supplementary table S1 shows the clinical and laboratory data of the study group. There were no relevant differences in age, gender or CVRF between patients and controls (see online supplementary table S2).
FDG uptake cut-off for GCA diagnosis SUVm at any vascular territory explored was significantly higher in GCA patients than in controls (table 1). ROC curves and AUCs are displayed in figure 1 and table 1, respectively. Mean of the SUVm observed at all the vascular territories had an AUC of 0.830 (0.715-0.946). A cut-off of 1.89 had a sensitivity of 80% and a specificity of 79%. Mean of the SUVm at supraaortic vessels showed the highest AUC (0.832). In this site, a cut-off of 1.70 achieved a sensitivity and specificity of 81 and 79%, respectively, for the diagnosis of GCA (95% CI 0.720 to 0.946). FDG uptake at the aorta showed lower AUC (0.738), with a sensitivity and specificity of 90 and 42, respectively, using a cut-off of 2.25, and a sensitivity of 58%, specificity of 90% with a cut-off of 2.65 (95% CI 0.598 to 0.881). However, differences in AUCs among territories did not reach statistical significance.
Vascular/liver uptake ratios were also significantly higher in patients than in controls at the right axillary and carotid arteries, Removal of the two patients who had received intravenous methylprednisolone pulses at the time of PET performance did not significantly modify the results (data not shown). Differences in AUCs among different vascular territories did not reach statistical significance. AUC, area under the curve; GCA, giant cell arteritis; PET, positron emission tomography. *Values represent the mean of the SUVm observed at all the vascular beds assessed. **Aorta, Supraaortic branches and Iliofemoral territory represents the mean of the SUVm observed at the different vessels of these areas.
but the overall discriminatory performance was much lower (see  online supplementary table S3).

Relationship between FDG uptake and clinical and laboratory findings
Patients with cranial symptoms presented significantly higher values of maximal and mean SUVm (combined average of all vascular territories) than patients lacking cranial manifestations. No relationship between the intensity of FDG uptake and other clinical findings was observed (table 2). No differences in maximal or mean SUVm were observed between treatment-naïve patients and those who had received glucocorticoids. The maximal and mean SUVm correlated with acute-phase reactants and serum IL-6 concentrations (table 2).

DISCUSSION
The present prospective study, performed in an unselected patient cohort with unequivocal GCA, shows that FDG uptake is significantly stronger in patients than in controls in all vascular territories tested, confirming the diagnostic potential of PET/ CT. 4 8 9 11 PET/CT allows rapid, reproducible and broad vascular evaluation. Nevertheless, there is no standardised definition of vasculitis based on an objective FDG uptake measure, and strategies employed to establish a PET-based diagnosis of GCA has been heterogeneous. Most studies have used qualitative visual assessment or a semiquantitative score using liver uptake as a reference. Visual scoring has a remarkable investigator dependency and interobserver variability. Liver uptake is influenced by individual metabolic activity, glucocorticoid treatment and the time lapse between injection and scanning. 12 We tried to overcome this limitation by objectively quantifying FDG uptake by different vascular beds in patients and controls and performing ROC analysis to determine the optimal cut-off for GCA diagnosis at different vascular territories. FDG uptake by supraaortic branches had the highest AUC, in accordance with a pioneer study showing that supraaortic branches were the most frequently involved when assessed by PET. 4 In this area, an FDG uptake cut-off value of 1.70 had the best sensitivity and specificity. A similar value, in the same territory but with lower performance (AUC=0.72), was reported in a retrospective study of 17 patients with GCA and 3 Takayasu arteritis patients. 13 This observation may be useful to differentiate GCA from other inflammatory aortic diseases that may produce systemic complaints and active aortic FDG uptake, including idiopathic aortitis, periaortitis, IgG4 disease and severe atherosclerosis. 10 14-17 This is crucial since a positive PET/CT may be accepted in the near future as a diagnostic criterion and is currently accepted as such in an ongoing clinical trial with tocilizumab in GCA. 18 The sensitivity and specificity of PET/CT obtained in this study is close to that calculated in a recent systematic review/ meta-analysis of heterogeneous, mostly retrospective studies, and those reported in a retrospective analysis evaluating the impact of PET on the management of patients with suspected large-vessel vasculitis. 19 Interestingly, FDG uptake by the aorta showed a lower AUC, being worse in the abdominal segment where atherosclerosis is more prevalent in the general population. This fact highlights the diagnostic limitation of PET in this territory since aortic FDG uptake may be markedly influenced by ageing or atheroma plaques. Hautzel et al reported a higher sensitivity and specificity of thoracic aorta FDG uptake to detect large-vessel inflammation in a cohort of 18 patients with GCA. 20 A thoracic aorta/ liver ratio of 1.0 had a sensitivity and specificity of 88% and 93%, respectively (AUC = 0.932). However, a substantial proportion of the patients assembled in this cohort were selected on the basis of previously known large-vessel involvement demonstrated by other techniques. In our study, direct, territoryfocused comparison of SUVm between patients and controls discriminated better than vascular/liver ratios.
A retrospective study evaluating how PET/CT results influenced management of patients with suspected GCA suggested that previous glucocorticoid (GC) treatment decreased the diagnostic yield of PET/CT. 19 Sequential assessments have demonstrated, indeed, that FDG uptake decreases after 3 months of treatment. 4 The present study suggests that a short course of therapy (≤3 days) may not substantially reduce the diagnostic accuracy of PET/CT.
In conclusion, this study provides sensitive and specific, territory-focused cut-off values to detect vascular inflammation by PET/CT. A limitation of the study is that while patients were prospectively recruited, controls were retrospectively selected. Another limitation is the relatively small number of patients analysed, although our cohort is among the largest investigated. Further prospective studies using objective cut-offs are necessary to confirm their diagnostic performance in patients with suspected GCA. Table 2 Relationship between clinical and laboratory data and maximal SUV at any vascular territory (SUVm) and mean of the SUVm obtained at every vascular bed assessed (mean SUVm). Values are mean±SD. Removal of the two patients who had received intravenous methylprednisolone pulses at the time of PET performance did not substantially modify the results (data not shown). A, absence; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; F, female; GC, glucocorticoid; IL-6, interleukin 6; M, male; N, no; NS, not significant; P, presence; PET, positron emission tomography; PMR, polymyalgia rheumatica; SUV, standardized uptake value; Y, yes.