Giant cells in temporal artery biopsies are associated with a reduced risk of large vessel involvement in patients with giant cell arteritis : a retrospective cohort study


 Background: Giant cell arteritis (GCA) is a systemic disease with extensive vascular involvement. There is limited and conflicting information on the relation between patient characteristics at diagnosis and future disease phenotypes. We aimed to investigate predictors of time dependent large vessel involvement (LVI) in a population-based cohort of patients with GCA. Methods: GCA patients with positive temporal artery biopsies (TAB) between 1997and 2010 were identified through a regional pathology register. A structured review of histopathology reports and relevant imaging studies was performed. Cases with LVI through July 2016 were identified. Patients were followed to first LVI, death, migration from the area or July 29, 2016. Event free survival by clinical and histopathologic features was estimated using the Kaplan-Meier method. Potential predictors of LVI were examined using Cox regression models.Results: A total of 274 patients were included. The mean age at GCA diagnosis was 75.7 years. Fifty-one patients (19 %) had documented LVI during the follow-up, corresponding to an incidence rate of 2.4/100 person-years. The median time from GCA diagnosis to the diagnosis of LVI was 4.5 years (interquartile range 0.6-7.4). Thirty-four patients had aortic involvement (67% of those with LVI; 12% of all GCA cases). Survival free of LVI was longer in patients with giant cells in the TAB (75th percentile 14.0 vs 6.7 years; p=0.014). In age-adjusted analysis, the presence of giant cells in the TAB was associated with reduced risk of LVI (hazard ratio 0.48; 95 % confidence interval 0.27-0.86). Conclusions: The negative association with giant cells in the TAB suggests that patients with LVI constitute a subset of GCA with particular disease mechanisms. This underlines the heterogeneity of GCA, which should be further explored in prospective studies.


Introduction
Giant cell arteritis (GCA) is the most frequent systemic vasculitis in the population over the age of 50 in western countries, with incidence estimates of 14-33/100 000 in Scandinavian populations [1,2]. The disease was rst described in 1889 [3] in its most classical form. Temporal artery biopsy (TAB) is the gold standard for diagnosing GCA. The rst biopsy was performed in 1931 [4]. In the biopsy the presence of giant cells, granuloma, disrupted internal elastic lamina, thickened intima and presence of in ammatory cells are indicative of the disease.
Since the rst description other disease subsets have been recognized [5][6][7]. The subsets can be de ned according to clinical phenotypes, imaging ndings, histopathology or laboratory ndings [8]. Although there is no consensus, the most common way of subgrouping patients with GCA is by the clinical phenotypes. classic cranial arteritis (with or without polymyalgia rheumatica (PMR)), GCA with extracranial large vessel involvement (LVI), occult GCA with mainly constitutional symptoms (e.g. in ammation and/or fever of unknown origin), and clinical pure PMR with underlying subclinical vasculitis [9].
Several studies have found that patients with LVI are younger than other GCA patients at the time of disease onset [5,16,17,26]. In a recent retrospective multicenter study in which LVI was identi ed in 45% of patients with GCA, the proportion of patients under 70 years in the LVI group was 60% compared to 27% in patients without LVI. Conversely, the proportion of patients over 80 years were 9% and 26%, respectively, in these groups [16]. Most studies indicate that female sex is a risk factor for LVI in GCA [5,16,17,26].
Arterial bruit, decreased peripheral pulse, discrepancy between arm blood pressures, extremity claudication, Raynaud's phenomenon and aortic valve insu ciency are all signs and symptoms that may be due to LVI, and have been associated with subsequent imaging ndings in several studies [5,10,16,18,23,26].
Different histopathological patterns occurring in GCA and their clinical correlation have been studied [27][28][29][30][31][32]. Giant cells are reported in TAB specimens of patients with GCA with varying frequency, ranging from 37 to 84% [27,[29][30][31][32][33][34][35][36][37][38]. Analyses of the association between the presence of giant cells and different clinical manifestations have generated con icting results. We have previously investigated the predictive value of different histopathologic features of TAB for LVI, and found a negative association with the presence of giant cells in the histopathological reports [22]. To our knowledge, this was the rst study to examine this relationship in patients with biopsy-proven GCA. The present study was undertaken to further investigate predictors for LVI in a larger cohort of patients with biopsy-proven GCA, and, speci cally, to assess the relation between baseline characteristics and time to LVI.

Patients
For identi cation of patients with biopsy-proven GCA, the database of the Department of Clinical Pathology in the county of Skåne (population 1 243 329 on December 2010) was used. Patients with positive TABs performed between 1997 and 2010 who were diagnosed with GCA (N = 840) were identi ed, as previously described [39]. Among these, patients from two de ned areas of the county that had previously been included in two clinical studies [22,40] were pooled for the present analysis (Fig. 1).
Subcohort I consisted of patients from the regional cohort, registered at the pathology unit in Kristianstad (one of four units contributing to the database). These have previously been reported in a retrospective study of large vessel involvement and associated factors [22] Subcohort II consisted of patients from the regional cohort living in the City of Malmö that had been included in a study of associations between clinical characteristics, glucocorticosteroid treatment and TAB ndings [40]. The patients had a registered diagnosis of GCA after participation in two populationbased surveys, the Malmö Diet and Cancer Study (MDCS) [41] and the Malmö Preventive Medicine Project (MPMP) [42], previously identi ed in a study investigating predictors of GCA [43] or had been diagnosed with GCA at Malmö University Hospital within the same time frame.

Data collection
The process of data collection has been described previously in detail [22,43]. The structured review of medical records and histopathology reports of all the identi ed patients included data extraction of items of the 1990 American College of Rheumatology classi cation criteria for GCA [44]. In addition, other clinical data at the time of diagnosis have been collected, e.g. previous PMR diagnosis, PMR at the time of disease onset, laboratory results, initial dose of glucocorticosteroids, visual symptoms and permanent visual impairment related to GCA at diagnosis and during the follow-up were also recorded.
All but 18 of the histopathology reports could be retrieved and reviewed and the presence or absence of giant cells, granuloma, thrombus, fragmented internal elastic lamina (IEL), in ammatory in ltrates, brosis, and luminal stenosis was noted. If they were not mentioned in the report, they were considered absent.
With regard to imaging studies, the patients included were evaluated according to the clinical practice; there was no speci ed protocol for systematic vascular evaluation during the study period. The reports of all relevant radiological and clinical-physiological studies in patients with GCA were reviewed by one of the authors (NN). LVI was de ned as presence of aneurysm, ectasia, or stenosis of the aorta and/or its main branches or positive 18 uoro-2-deoxy-d-glucose positron emission tomography-computed tomography ( 18 FDG PET-CT) as assessed by the radiologist. Aneurysm was de ned as reported aortic dilatation with a loss of wall parallelism and a recorded diameter of ≥ 30 mm. Aortic ectasia was de ned as reported abnormal dilatation without loss of wall parallelism.
Patients with detected LVI before diagnosis of GCA were excluded. For the present study, dates of death and migration from the study area were retrieved from a regional health care register.

Statistics
Patients were followed from date of GCA diagnosis and were censored at the date of: when LVI was rst detected, death, migration from the area or the end of the study (i.e. date of last chart review, July 29, 2016). Event free survival, by baseline clinical characteristics (female sex, PMR symptoms at onset, preexisting PMR, visual symptoms at onset, and TAB ndings -i.e. in ammatory cells, giant cells, granuloma, disrupted IEL, brosis and luminal stenosis), was estimated using the Kaplan-Meier method. Potential predictors of LVI were also examined using Cox regression analysis (covariates investigated: age at disease onset, female sex, PMR at onset, pre-existing PMR, visual symptoms at onset, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), platelet count (Plt), and TAB ndings -i.e. in ammatory cells, giant cells, granuloma, disrupted IEL, brosis and luminal stenosis). For continuous variables, hazard ratios (HRs) were estimated per standard deviation (SD), and, in the case of Plt and CRP, also per quartile, using the lowest as reference. Furthermore, the risk of LVI in those aged < 70 years and those aged 70-80 years was compared to those above 80 at GCA diagnosis (reference category).
Covariates with p < 0.20 in the bivariate analyses were included in multivariate models.

Results
The GCA cohort A total of 286 patients with biopsy-proven GCA were identi ed from the two subcohorts. Twelve patients with LVI before GCA diagnosis were excluded, leaving a study sample of 274. Mean age at diagnosis was 75.7 years [SD 8.09; range 49.5-94.8years]. 76.6% were females. Twenty-four (16%) of those with available information on previous medical history had a pre-existing diagnosis of PMR. At onset of GCA, 23% had current PMR symptoms, and 36% had visual symptoms. Demographics and clinical characteristics, overall and by LVI, are presented in Table 1. Time on GC treatment before TAB was similar in patients with and without LVI (Table 1).

Distribution of the affected large vessels
Aortic ectasia was found in 20 patients (ascendens n = 10, descendens n = 4, abdominal n = 6). In one patient, the whole aorta was ectatic and in four others the aorta had both aneurysmatic and ectatic changes. A total of 19 aortic aneurysms (ascendens n = 13, descendens n = 1, abdominal n = 5) were found in 19 patients. Two patients had dissections, one ascending aorta aneurysm with type-A dissection and one type-B dissection without aneurysm.
The distribution and type of vessel involvement in the fty-one patients with LVI are depicted in Fig. 1.

Predictors of LVI -Survival by baseline features and Cox regression
Demographics: There was no difference in survival time free of LVI between females and males (

Clinical features & laboratory parameters:
Clinical features such as visual symptoms, in ammatory laboratory tests at diagnosis (ESR, CRP, Plt) or symptoms of PMR (pre-existing or current) did not affect the risk of future LVI (Table 3 and Table 4). Among those who developed LVI, fewer suffered from permanent visual impairment or total visual loss, (Table 1), although numbers were too small for meaningful time dependent analyses.
Histopathology features: Survival free of LVI was signi cantly longer in patients with giant cells present in the TAB, 75th percentile 14.0 vs 6.7 years (p = 0.01; Table 2 and Fig. 3). Other histopathology features had no impact on the LVIrisk (Tables 2 and 3).
The presence of giant cells in the TAB was associated with a reduced risk of LVI in bivariate Cox regression (Table 3) as well as in multivariate analysis, adjusted for age (HR: 0.48; 95% CI 0.27, 0.86).

Discussion
In the present study, 19% of patients with biopsy-proven GCA had diagnosed LVI during the follow-up, corresponding to an incidence of 2.4/100 person-years of follow-up. We con rmed a negative association between the presence of giant cells in the TAB and subsequent development of LVI in patients with GCA. We also observed a trend toward reduced risk of LVI with increasing age at GCA-onset, in keeping with previous studies [5,11,16,17,26]. The shift towards onset of GCA below age 70 in the subgroup with LVI, with a corresponding reduction in the number of patients aged > 80 years at diagnosis, is compatible with previous ndings in a study that compared the same age categories [16]. The lower risk in elderly patients in the present study could be due to less extensive investigations with increasing age. On the other hand, in the other study with similar observations, all patients were scanned for LVI [16].
We did not observe an increased risk with elevated laboratory in ammatory markers.
In accordance with some [10,11], but in contrast to other studies [5,16,17,26], there was no major difference between females and males in the occurrence of LVI in this study.
Previous studies indicate that cranial symptoms at the time of GCA diagnosis are negatively associated with LVI [5,7,11,[16][17][18][19][20]. In the present study, severe ocular involvement tended to be less frequent among those with LVI. Other types of cranial symptoms were not evaluated.
In contrast to our previous study of a smaller sample [22], we did not nd any association between PMR symptoms at GCA diagnosis and subsequent LVI. Other previous studies of PMR as a risk factor have yielded divergent results: one reported a negative association with LVI [16], others showed no difference between the two groups [5,10,13,18,19,21], one other study showed a higher proportion with PMR at GCA diagnosis in the LVI group [17] and one study found more patients with PMR prior to GCA diagnosis in the LVI group [11]. Differences in case selection, and in the ascertainment and management of PMR, may explain these discrepancies.   [32,36]. Two other studies found a trend towards a higher occurrence of cranial ischemic complications in the group with giant cells compared to those without [37,38].
Giant cells are fused macrophages, formed through interferon-γ (IFN-γ) stimulation [45]. These macrophages produce among others: reactive oxygen species, matrix metalloproteinases, platelet-derived growth factor and vascular endothelial growth factor [46]. IFN-γ is produced by differentiated Th1 cells [46]. The differentiation is driven by interleukin 12 (IL-12) [46]. High levels of IL-12 [47] and IFN-γ [48] are associated with more ischemic complications. It is possible that the LVI phenotype of GCA is less dominated by Th1 mechanisms considering less frequent presence of giant cells in the TAB, fewer cranial ischemic manifestations and the massive presence of artery tertiary lymphoid organs that lack macrophages [49].
Limitations of this study are related to the retrospective design. Data collection was limited to the recorded information in patient charts, and no standardized physical evaluation was performed. In contrast with other studies [16], we did not limit our analysis to patients that had been investigated with ≥ 1 large vessel imaging scan. Imaging modalities during study period were heterogeneous, nonstructured and often performed due to other indications than suspected vasculitis. Therefore, the incidence of LVI and its extent is likely underestimated due to lack of systematic imaging studies. LVI detected in this study re ect the clinical practice of the time period. Based on the study design, we cannot rule out that some of the changes may be of atherosclerotic origin.
Strengths of this study include the population-based study design, the relatively large patient cohort with a long follow-up period, and the systematic review of pathology reports. Furthermore, we focused on objectively veri ed (biopsy-proven) GCA, which had an incidence in our population that was similar to other studies of Scandinavian populations [2,50].
This study suggests that diverse disease mechanisms may be involved in GCA with different clinical presentation and disease progression. This could affect treatment response and raises the question of whether or not GCA should be treated differently based on phenotype and understanding of the underlying pathophysiology.
Future research should include prospective studies with structured imaging evaluations and standardized quanti cation of the giant cells and other histopathologic features in TABs.
In conclusion, in this study of patients with biopsy-positive GCA, LVI was detected in 1 in 5 patients during follow-up. There was a signi cantly lower risk of LVI in patients with giant cells in the biopsy. This may suggest particular disease mechanisms in LVI of GCA. Availability of data and materials The datasets generated and/or analysed during the current study are not publicly available due to Swedish legislation (the General Data Protection Regulation), but a limited and fully anonymised dataset containing the individual patient data that support the main analyses is available from the corresponding author on reasonable request. Authors' contributions: NN participated in the data collection, statistical analysis and interpretation of the results, and wrote the rst draft of the manuscript.
AM participated in the study design, data collection and interpretation of the results.
KW participated in the data collection and interpretation of the results.
MW assisted in data management and participated in the interpretation of the results.
JÅN assisted in the statistical analysis and participated in the interpretation of the results.
CT participated in the study design, the statistical analysis and the interpretation of the results, and helped draft the manuscript.
All authors read and approved the nal manuscript.